Report No. 37026 The Development Impact of Workers' Remittances in Latin America (In Two Volumes) Volume II: Detailed Findings August 25, 2006 Finance, Private Sector and Infrastructure (LCSFR) and Chief Economist Office (LCRCE), Latin America and the Caribbean Region Document of the World Bank MXP MexicanPesos NBER National Bureauof EconomicResearch ODA Official DevelopmentAssistance OECD Organisationfor EconomicCo-operationand Development OFAC Office of ForeignAssets Control OLS OrdinaryLeast Squares Method PPP PurchasingPower Parity PRAF Programade AsignacionFamiliar PROFECO Procaduria Federaldel Consumidor PRS Political Risk Services PWT PennWorld Tables REER RealEffective ExchangeRate RPS Redde ProteccionSocial RSP Remittances Service Providers SA SouthAsian Region SSA Sub SaharaAfrican Region SWIFT Societyfor Worldwide InterbankFinancialTelecommunication UK UnitedKingdom UNDP / PNUD UnitedNationsDevelopmentProgram US/ USA United States of America USAID United States Agency for InternationalDevelopment USD United States Dollars WAZ Weight-for-Age AnthropometricZ Score WB The World Bank WDI World DevelopmentIndicators WIDER World Institute for DevelopmentEconomics Research woccu World Council of Credit Unions Vice President: PamelaCox Chief Economist: Guillermo Perry Sector Director: Makhtar Diop Sector Manager: Susan Goldmark Task Managers: Pablo Fajnzylber and HumbertoLopez 3 TABLE OF CONTENTS: ACKNOWLEDGMENTS .......................................................................................................... 10 HOW IMPORTANT ARE REMITTANCES INLATINAMERICA? ................................ 11 I.INTRODUCTION ....................................................................................................................... 11 11.THEMAGNITUDEOFREMITTANCESFLOWS: THEGLOBALPICTURE........................................ 12 15 IV.PROFILEOFRECIPIENTS ........................................................................................................ 111.THEMAGNITUDEOFREMITTANCESFLOWS:THEREGIONALPICTURE.................................... 19 V HowRELIABLE AREREMITTANCES FIGURES? ....................................................................... . 32 VI.CONCLUSIONS ...................................................................................................................... 34 BEFORE WORKERS' REMITTANCES, WORKERS' MIGRATION .............................. 36 I.INTRODUCTION ...................................................................................................................... 36 11.LATIN 111AGEPROFILEOFMIGRATION ................................................................................................ . AMERICAN MIGRANTS ABROAD .................................................................................. 37 42 vIV..LABOR EDUCATIONPROFILE ........................................................................................................... 44 MARKET PERFORMANCE ............................................................................................ 49 VI.MIGRATION REMITTANCES .......................................................................................... 51 VI1 CONCLUSIONS.................................................................................................................... . AND 61 REMITTANCES AND POVERTY REDUCTION ................................................................. 62 I.INTRODUCTION 11.REMITTANCES,INEQUALITYAND POVERTY:A MICROECONOMIC ...................................................................................................................... ........................ 62 63 111.INEQUALITYAND POVERTYEFFECTS: APPROACH MACROECONOMIC ................................... EVIDENCE 71 Iv.REMITTANCES,GROWTH AND INVESTMENT ......................................................................... 76 V.REMITTANCESAND OUTPUTVOLATILITY............................................................................. V.CONCLUSIONS ....................................................................................................................... 86 80 ANNEX1:REMITTANCESANDPOVERTYHEADCOUNTS NATIONAL USING POVERTY LINES .......88 REMITTANCES AND HOUSEHOLD BEHAVIOR ............................................................. 90 I.INTRODUCTION 11.REMITTANCESANDHOUSEHOLD ....................................................................................................................... ............................................................................ 91 90 I11.REMITTANCESAND HOUSEHOLD SAVINGS EXPENDITURES ................................................................. Iv.REMITTANCESANDHUMAN 95 CAPITAL ................................................................................ 101 V REMITTANCES AND LABOR . 109 VI.REMITTANCESAND ENTREPRENEURSHIP SUPPLY .................................................................................... ........................................................................... VI1 CONCLUSIONS.................................................................................................................. . 114 119 DO REMITTANCES AFFECT RECIPIENT COUNTRIES FINANCIAL DEVELOPMENT? ................................................................................................................... 120 I.INTRODUCTION ..................................................................................................................... 120 11.MACRO-LEVEL ANALYSIS OF THE ASSOCIATIONBETWEENREMITTANCESAND FINANCIAL DEVELOPMENT ......................................................................................................................... 122 111.MICRO-LEVEL ANALYSIS OF THE ASSOCIATIONBETWEENREMITTANCESAND FINANCIAL DEVELOPMENT IV.CONCLUSIONSAND POLICY RECOMMENDATIONS ............................................................... ......................................................................................................................... 136 153 4 ANNEX1:REMITTANCESAND BANCARIZATIONTHEEXPERIENCESCOLOMBIAAND . OF GUATEMALA ............................................................................................................................ 159 ANNEX2: TAILORINGFINANCIALPRODUCTSTOTHENEEDSTHEREMITTANCE MARKET164 OF .. REMITTANCES, THE REAL EXCHANGE RATE, AND THE DUTCHDISEASE PHENOMENON ....................................................................................................................... 167 I.INTRODUCTION ..................................................................................................................... 167 11 REMITTANCESAND THEREALEXCHANGERATE.................................................................. . 111.EMPIRICALSTRATEGY........................................................................................................ 171 . 177 V. CONCLUSIONSAND POLICY RECOMMENDATIONS ............................................................... IV RESULTS ............................................................................................................................ 180 TECHNICALANNEX.................................................................................................................. 185 187 DO CONDITIONAL CASH TRANSFER PROGRAMS CROWD OUT PRIVATE TRANSFERS? ........................................................................................................................... 192 I.INTRODUCTION ..................................................................................................................... 192 11 CONDITIONAL CASHTRANSFERS REMITTANCES ......................................................... . 195 111.THEREDDEPROTECCIONSOCIAL (Ws)AND THEPUF-I1PROGRAMS........................... AND 197 IV THEDATA . .......................................................................................................................... v.THEEMPIRICAL .................................................................................................. 198 200 VI.ESTIMATION STRATEGY ......................................................................................................... 203 VI1 CONCLUSIONS.................................................................................................................. . RESULTS TECHNICALANNEX.................................................................................................................. 207 209 REMITTANCES, GROWTHAND THE ROLE OF COMPLEMENTARY POLICIES 212 . I.INTRODUCTION .................................................................................................................... 212 I1.POLICY AREASTHAT MAY COMPLEMENT REMITTANCES ...................................................... 111.EMPIRICALEVIDENCE......................................................................................................... 215 Iv.REMITTANCES ..................................................................... 220 232 V CONCLUDING REMARKS ...................................................................................................... 234 . AND DOMESTIC INVESTMENT THE REGULATORY FRAMEWORK FOR REMITTANCES ......................................... 235 I.INTRODUCTION I1 MULTILATERAL 237 I11 ENHANCING .. .................................................................................................................... 235 INITIATIVES ............................................................................................... Iv.TRANSPARENCYTHE COMPETITION................................................................................................. 240 v.ACCESSIBILITYTO............................................................................................................... 256 IN PROVISIONOF REMITTANCE SERVICES .......................................... 251 FORMAL REMITTANCE SERVICES ............................................................ VI.SECURITY ISSUES 260 VI1 POLICYRECOMMENDATIONS . ............................................................................................ 266 REFERENCES .......................................................................................................................... 269 5 LIST OF TABLES: Table 1.2. Recent remittances flows to LatinAmerica and the Caribbean (2002-2005) ..............14 Table 1.1, International flows to low and middle income countries............................................. Table 1.3. Migration and Education ............................................................................................. 26 17 Table 2.1, Education profile o f L A C migrants (% with a given educational level) ..................... 45 Table 2.3. Determinants o f Remittances (per capita) ................................................................... Table 2.2. Determinants o f Remittances (as a share o f GDP) ...................................................... 58 Table 2.4. Determinants o f Remittances (per migrant)................................................................. 59 Table 3.1. Income Gini Coefficient before and after remittances................................................. 60 64 Table 3.3. Poverty Headcounts before and after remittances ....................................................... Table 3.2. Effect on the Gini o f a 10% Increase inRemittances.................................................. 65 Table 3.4. Income Gini Coefficient inCounterfactual Scenario o fNo-Migration ....................... 66 68 Table 3.5. Poverty Headcounts in Counterfactual Scenario o f No-Migration.............................. 69 Table 3.6. Poverty Headcounts among Recipients Households inCounterfactual Scenario o fNo- Migration............................................................................................................................... . . Table 3.7. The impact o f per capita remittances on poverty......................................................... 71 Table 3.8. The impact o f remittances on growth and changes in inequality ................................ 73 75 Table 3.9. Poverty elasticity o f remittances inLatin America 76 Table 3.10. Remittances and Economic Growth .......................................................................... ..................................................... 79 Table 3.11. Remittances and Investment...................................................................................... 80 Table 3.13. Remittances and Growth Volatility ........................................................................... Table 3.12, The Cyclical Behavior o f Remittances inLatin America.......................................... 81 85 Table 3.14. Volatility Effects o f External and Policy Shocks by Remittances Levels (changes in . 86 Table A3.1. Poverty Headcounts before and after remittances (national poverty lines) ..............88 std. dev. of p/c GDP Growth after one std. dev shock) ....................................................... Table A3.2 .Poverty lines) ...................................................................................................................................... Headcounts inCounterfactual Scenario o f No-Migration (national poverty 88 Table A3.3. Poverty Headcounts among Recipients Households inCounterfactual Scenario o f No-Migration (national poverty lines) .................................................................................. 89 Table 4.1, Saving rates by Income Quintile and Remittance Recipient Status............................. 93 Table 4.2. Access to Remittances and Expenditure Shares (OLS, differences with respect to non- recipient households) ............................................................................................................ 98 Table 4.3. Remittances and Expenditure Shares by Counterfactual Household Income Quintiles Table 4.4. Access to Remittances and Children Education - OLS ............................................. ............................................................................................................................................... 99 Table 4.5, Remittances and ChildrenEducation by Mother's Education ................................... 104 105 Table 4.6. Remittances and Health Outcomes............................................................................ 109 Table 4.8. Remittances and Labor Force Participation (with instrumental variables) ................112 Table 4.7. Access to Remittances and Hours Worked................................................................ 111 Table 4.9. Remittances and Labor Force Participation, by Educational Levels (with instrumental 113 Table 4.10. Remittances and Entrepreneurship.......................................................................... variables)............................................................................................................................. 117 Table 5.1. Correlation betweenRemittances and Indicators o f Financial Development.,..........119 Table 4.12. Remittances and Entrepreneurship, by Income Quintile ......................................... Table 4.1 1. Remittances and Self-Employment by Sector ......................................................... 118 127 6 Table 5.2. Panel Estimates o f the Impact o f Remittances on FinancialDevelopment ...............132 Table 5.3. Panel Estimates o f the Impact o f Remittances on FinancialDevelopment with L A C Interaction........................................................................................................................... 133 Table 5.4. Panel Estimates o f the Impact o f Remittances on FinancialDevelopment with LAC16, Table 5.5. Factors that Might Affect the Impact o f Remittances on FinancialDevelopment ....134 LACS, and LAC4 Interactions............................................................................................ 135 Table 5.6. Testing for Differences inthe Use o f Banking Services by Remittance Recipients and Non-Recipients ................................................................................................................... 138 Table 5.7. Remittances and Access to Financial Services inGuatemala, Haiti and the Dominican Republic .............................................................................................................................. 139 Table 5.8. Data Description and Summary Statistics: El Salvador Case Study on the Impact o f Remittances on the Use o f Financial Services .................................................................... 141 Table 5.9. Are Remittance Recipients More Likely to Use Banking Services? Probit estimations for El Salvador .................................................................................................................... 144 Table 5.10. Are Remittance Recipients More Likely to Use Banking Services? FixedEffects 145 Table 5.11. Does the Way inWhich Remittances Are Transferred Affect The Use o f Banking Probit and Instrumental Variables Probit Estimations for El Salvador .............................. Services? Probit estimations for El Salvador...................................................................... 146 Table 5.12. Data Description and Summary Statistics: Mexico Case Study on the Impact o f Table 5.13. The Impact o f Remittances on Bank Deposits Across Mexican Municipalities. OLS Remittances on FinancialDevelopment ............................................................................. 148 Table 5.14. The Impact o f Remittances on Bank Branches Across Mexican Municipalities. OLS Estimations Clustered by State ........................................................................................... 149 Estimations Clustered by State ........................................................................................... 150 Table 5.15. The Impact o f Remittances on Bank Credit Across MexicanMunicipalities. OLS Estimations Clustered by State ........................................................................................... 151 Table 5.16. The Impact o f Remittances on Bank Deposits, Branches and Credit Across Mexican Municipalities .Instrumental Variables (IV) Estimations Clustered by State ..................... 152 AppendixTable: Are Remittance RecipientsMore Likelyto Solicit and Receive Non-bank Credit? Probit estimations for El Salvador ......................................................................... 158 Table 6.1. Remittances and the real exchange rate ..................................................................... 173 Table 6.2. The impact o f remittances on the real exchange rate................................................. 181 Table 6.3, The impact o f remittances on real exchange rate misalignment................................ 184 Table 7.1, Summary statistics o f eligible households surveyed inNicaragua in2000, 2001 and 2002..................................................................................................................................... 201 Table 7.2. Summary statistics o f eligible households surveyed inHonduras in2000 and 2002 202 Table 7.3. Linear ProbabilityModel Estimates o f the Impact o f RPS on the Incidence o f Remittances......................................................................................................................... 204 Table 7.4. OLS Estimates o f the Impact o f RPS on the Incidence o f Remittances .................... 205 Table 7.5. Linear Probability Model Estimates o f the Impact o f PRAF-I1on the Incidence o f Table 7.6. OLS Estimates o f the Impact o f PRAF-I1on the Incidence o f Remittances .............206 Remittances......................................................................................................................... Table 8.1. Remittances, Education and Economic Growth ........................................................ 207 Table 8.2. Remittances, Institutions and Economic Growth ...................................................... 222 225 Table 8.3. Remittances, the FinancialSector and Economic Growth ........................................ 228 Table 8.4. Remittances, the Policy Environment and Economic Growth .................................. 230 7 Table 8.5. Remittances. Complementary Policies and Investment............................................. Table 9.1. Summary o f registration and licensing requirements................................................ 233 242 Table 9.2. Price o f a US$300 remittance from Chicago to Mexico............................................ Table 9.3. Types of remittance service and accessibility considerations ................................... 254 257 Table 9.4. Requirementsto open abank account inthe US .(March 2006) .............................. 259 LIST OF FIGURES: Figure 1.1.Regional distributionof remittances .......................................................................... Figure 1.2.Annual growth rate ofrecorded remittances (1990-2004) ......................................... 13 Figure 1.3.International financial flows: 1990and 2004............................................................. 14 Figure 1.4.Remittances inLatinAmerica: 1980-2003................................................................. 15 Figure 1.6. Shareo f householdreceiving remittances.................................................................. 16 Figure 1.7.Households receivingremittances by quintile ofthe income distribution .................2021 Figure 1.8.Households receiving remittances (%) by quintile of the non-remittances income Figure 1.9. Households receivingremittances by quintileo fthe total income distribution.........23 distribution............................................................................................................................ 22 Figure 1.lo.Educational characteristics of households receivingremittances 25 Figure 1.11 .Average ............................ annual amount reported by recipients ........................................................ 27 Figure 1.12. Incomeshare ofremittances by income quintile (recipients only) ........................... 28 Figure 1.13. Income share o fremittances by income quintile (all households) ........................... Figure 1.14.Incomeand Remittances Distributionby income quintile....................................... 30 31 Figure2.1, Latin American migrants............................................................................................ Figure 1.15. BOP basedversus householdbasedremittances ...................................................... 34 Figure 2.2. Major destination o f LatinAmerican migrants .......................................................... 38 Figure 2.3. Major European destination o f Latin American migrants.......................................... 40 41 Figure2.5. Current Age Profile ofLatinAmerican migrants....................................................... Figure 2.4. Age profile at the time o f arrival for Latin American migrants ................................. 42 Figure 2.6. Education profile o f LAC migrants (age 22+) who arrived during the 1990s ...........43 Figure 2.7. Education profile o f native population versus the Migrants for LAC countries ........44 45 Figure2.9. Share o f college educatedWorkers who receivedtheir degrees at home..................46 Figure 2.8. Share o f college educatedwho have migrated ........................................................... Figure2.10. Shareo fMigrants with College Degrees According to Age ofEntry...................... 47 Figure2.11, Occupational distributionofmigrants...................................................................... 48 Figure2.12. Brainwaste: Home tertiary educatedmigrants 25+ on arrival................................. 49 51 Figure2.13. Migrants as a share o fpopulation............................................................................. Figure2.14. Share ofFemale Migrants ........................................................................................ 52 Figure2.15. Share of College Graduates among Migrants........................................................... 53 ............................................................................ 54 Figure 3.1. Scatter plots o f Remittances, Growth and Investment............................................... Figure2.16. Bank Deposits as a Share ofGDP 54 Figure 3.2. Remittances' Sensitivity to Output Fluctuations inRecipient Countries...................78 82 Figure3.3. Remittances' Sensitivity to Output Fluctuations inSending Countries..................... Figure3.4. Country Estimates o fRemittances' Sensitivity to OwnOutput................................. 82 Figure3.5. The response ofRemittances to Macroeconomic Crises............................................ 83 Figure4.1. Differences inSaving rates by Remittance Recipient Status..................................... 84 92 8 Figure4.2. Estimated Saving Rates out o f Remittance Income ................................................... 95 Figure4.3, Expenditure Patterns by Remittance Recipient Status -Rural Regions .................... Figure4.4. Expenditure Patterns by Remittance Recipient Status -UrbanRegions ...................96 97 Figure4.5. Expenditure inNon-Durables (including food) and Education by Remittance Recipient Status and Counterfactual Income Quintile: Mexico ......................................... 100 Figure4.6. Expenditure inNon-Durables (including food) and Education by Remittance Figure 4.7. Average Years o f Education for Adults (22-65 years old) ....................................... Recipient Status and Counterfactual Income Quintile: Nicaragua ..................................... 101 102 Figure 4.8. Differences in School EnrollmentRates for 12-17 years old by Remittances Recipient Status................................................................................................................................... 102 Figure4.9. Anthropometric Measures for Children aged 1 to 5, by Remittance Recipient Status - Guatemala ........................................................................................................................... 108 Figure4.10. Anthropometric Measures for Childrenaged 1to 5, by Remittance Recipient Status Figure 4.1 1. Labor Force Participation o f Adults (20-59 years old), by Gender and Remittance -Nicaragua......................................................................................................................... 108 Figure4.12. Self-Employment by Remittance Recipient Status................................................ Recipient Status .................................................................................................................. 110 Figure4.13. Business Ownershipby Remittance Recipient Status............................................ 115 115 Figure5.1. Remittances and financial development inLatinAmerican countries..................... 123 174 Figure 6.2. Exports and the real exchange rate........................................................................... Figure6.1. Remittances and the real exchange rate.................................................................... 175 Figure 6.3. Imports and the real exchange rate........................................................................... 176 Figure 8.1. Secondary net enrollment deficit in selected LatinAmerican countries ..................216 Figure 8.2. Institutions and per capita income levels 217 Figure 8.3. Domestic credit to the private sector........................................................................ .................................................................. 219 Figure 8.4. Regional policy index............................................................................................... . . 219 Figure 8.5. Growth, education and remittances: Impact o f a one std. dev. increase inremittances ............................................................................................................................................. 223 Figure 8.6. Growth, institutions and remittances: Impact o f a one std. dev. increase inremittances ............................................................................................................................................. 226 Figure 8.7. Growth, the policy environment and remittances.Growth impact o f a l-sdincrease in remittances.......................................................................................................................... 231 Figure9.1. Range o fprices o f remittance services inthe U.S.-Mexico corridor, 1999-2005 (% of amount sent)........................................................................................................................ 241 Figure9.2. Barriers to entry as reported by surveyed RSPsinthe U.S ...................................... 244 Figure9.3. Minimumamount requiredto set up a new RSP (USD$ Thousand) ....................... 245 Figure9.4. Range inbondingrequirements and level o fremittances by state........................... 245 Figure9.5. Perceptions on why the total cost o f remittance is higherthan the flat commissions 252 Figure9.8. Total price o f a US$300 from Chicago to Mexico (March 2006) ............................ paidby senders .................................................................................................................... 255 Figure 9.9. Total price o f remitting US$300 from various cities inthe US . to Figure 9.10. Channels for Sending Remittances to LAC (2004) ................................................ 2005 (average o f all RSPs monitored)................................................................................ Mexico 1998 - 256 257 Figure9.11 Reasons for not having a bank account.................................................................. . 258 Figure 9.12. Largest expenses reported by RSPs........................................................................ 265 9 ACKNOWLEDGMENTS The Development Impact of Remittances in Latin America is the product o f a collaborative effort o f two units o f the Latin American Region o f the World Bank: the Chief Economist Office and the Finance, Private Sector and Infrastructure Group. The report was prepared by a team led by Pablo Fajnzylber and Humberto Lopez, and comprising Pablo Acosta, Cesar Calderon, Massimo Cirasino, Paola Granata, Mario Guadamillas, Maria Soledad Martinez Peria, Yira Mascaro, Florencia Moizeszowicz, Caglar Ozden, Pedro Olinto, and Emanuel Salinas. Luis Molina, from the Bank o f Spain, also collaborated with the project while he was visiting the World Bank in early 2006. Guillermo Beylis and Namsuk Kim provided excellent research assistance at different times duringthe project. Extensive and excellent advice has been received from Guillermo Perry, Susan Goldmark, Ernesto May, Maurice Schiff, Jose Guilheme Reis, and especially from our peer reviewers, Omar Arias, and Samuel Muzele Maimbo. Comments on some o f the background papers for this report were also received from many participants at the Latin American Region seminar series on Remittances and Development organized in the context o f this project, as well as from Makhtar Diop, Edmundo Murrugarra and Dante Mossi. Ernesto Lopez Chrdova, Jose de Luna Martinez, Manuel Orozco and Anna Paulson provided the data and information used in Chapter 5. To all o f them we are grateful without implication. 10 Chapter 1 How important are remittances inLatin America?* Workers remittances have become a major source of financing in developing countries reaching levels that are comparable to those of Foreign Direct Investment and doubling those of OfJiciaI Development Aid. However, there are a number of questions that are criticalfor determining the development impact of remittances in Latin America: how important are thoseflows in the regional context and how have they evolved over the past years? Khat do we know about remittance recipients at the country and household level? Are they the poor or instead they belong to more accommodated classes? Are the observedpatterns similar across various Latin American countries? Finally, can we rely on oflcial statistics? I.Introduction Workers remittances have become a major source o f financing in developing countries. According to the World Bank's Global Economic Prospects 2006, in 1990 remittances to middle and low income countries amounted to about US$31 billion. Fifteen years later, they are estimated to have reached US$167 billion, o f which about one-fourth to the Latin American and the Caribbean Region (Latin America henceforth). Workers remittances now account for about 30 percent o f total financial flows to development countries, are more than twice as large as official development assistance flows, and represent the equivalent to 2.5 percent o f the gross national income o f the developing world. The raising importance o f remittances flows has been reflected inthe increasing attention beingdevoted to the issue by development practitioners both inacademic and policy circles. At the academic level, a number of recent papers have explored the impact of remittances on poverty (Adams and Page, 2005; Page and Plaza, 2005, Acosta et al. 2006a, and Acosta et al. 2006b), inter-temporal consumption smoothing (Yang, 2005), growth (Ruiz Arranz and Giuliano, 2005; Calderon, Fajnzylber and Lopez, 2006), risk management (Amuedo-Dorantes and Pozo, 2004), education (Cox and Ureta, 2003), labor supply (Rodriguez and Tiognson, 2001), and external competitiveness (Amuedo-Dorantes and Pozo, 2004; Rajan and Subramaian, 2005). 'Thischapterisbasedonthebackgroundpaperforthisreport"RemittancesandDevelopmentinLatinAmerica"by PabloAcosta, Cesar Calderon,Pablo Fajnzylber,andHumbertoLopez. 11 At the policy level, the International Monetary Fund (IMF), the World Bank, and the United Nations Development Program (UNDP) have all addressed the growing importance o f migration and remittances and their impact on development efforts in some o f their flagship publications. For example, the IMF's World Economic Outlook 2005 devoted significant attention to the determinants and implications o f inflows o f workers' remittances, whereas the World Bank's Global Economic Prospects 2006 had as its central topic the economic implications o f remittances and migration. The World Bank has also edited a number o f volumes on migrations and remittances issues (see Maimbo and Ratha, 2005, and Ozden and Schiff, 2006). Similarly, the UNDP's 2005 Human Development Report for El Salvador focused heavily on the development impact o f remittances and the UNDP is now organizing a highlevel meeting on the topic to be held inNew York inthe fall o f 2006. Yet, to a large extent the existing works have focused on the role o f remittances (andor migration) at the global level with relatively little attention paid to regional specific issues. This i s an important limitation because if, as discussed in Chapter 2, the migration patterns o f the Latin American region differ significantly from those o f other regions, then it i s possible that findings based on studies relying on global databases are not hlly applicable to the Latin American context. True, there are also works based on country cases studies that focus on Latin American economies. However, these tend to be based on a limited number o f countries (typically Mexico, El Salvador and Guatemala) and have important methodological differences among them, something that in turn makes it difficult to easily extend and generalize the obtained results to the region. This regional study tries to somewhat fillthe existingknowledge gap for LatinAmerica. To start with, in this chapter we explore the basic facts o f remittances to the region. Among others, we address the following questions: how much and how relevant from an economic perspective are these flows? How have they evolved over the past few years? Are they leveling off or instead continuing to increase? What do we know about those at the receiving end o f the remittances chain? Are they the poor or instead they belong to more accommodated classes? What i s their level o f education? Finally, to what extent can we trust statistics on officially recorded remittances? 11.The magnitudeof remittancesflows: the globalpicture According to Balance o f Payments Statistics' officially recorded remittance flows to developing countries reached US$167 billion in 2005, o f which about US$42 billion to the Latin American region. This would make Latin America together with East Asia and the Pacific (East Asia henceforth) the two top remittances receiving regions o f the world (Figure 1.1, Panel A). ' We would like to note that the BOP based remittance figures used throughout this report do not exactly coincide with the line workers remittance o f the BOP statistics. The data we use, which comes from the World Economic Outlook 2005 database, are (with some exceptions for which we refer the reader to the World Economic Outlook 2005) constructed as the sum o f three BOP items o f the Balance of Payment Statistics Yearbook (IMF): workers' remittances (current transfers made by migrants who are employed and resident in another economy); compensation of employees (wages, salaries and other benefits earned by nonresident workers for work performed for residents o f other countries); and migrant transfers (financial items that arise from the migration or change o f residence o f individuals from one economy to another). 12 n Anicrican an flows (W$8 billion). of thc world, but also cvcry 6 ycars approx ixlgcountrics ovcr thc Panel 3,Per capita ngbascdon total re ststhc LatinArrtcrican, a would bc tlic top rcccivin r (tablc 1.I). anioun This (USS69), Europe an region with the higher growth rate, and now Latin America would be joined by Europe and Central Asia inthe second place o f the ranking (Figure 1.2, Panel B). Figure 1.2. Annual growthrate of recordedremittances(1990-2004) Panel A. Total Flows Panel B. Per capita 20 I 20 15 15 I O 10 5 5 0 0 egend: EAP: East Asia and Pacific; ECA: Europe and Central Asia; LCR: Lain America and the CaribbeanRegion; MENA: MiddleEast andNorthAfrica; SA: SouthAsia; SSA Sub SaharanAsia. Source: GEP andWDI. The relevance o f workers' remittances is also apparent when we compare themwith other international financial flows (both private and official) to low and middle income countries. As table 1.1 indicates, remittances represent about 30 percent o f the US$572 billion total internal financial flows to the developing world, and are estimated to have reached levels equivalent to 80 percent o f foreign direct investment (FDI) flows to developing countries (US$211 billion) and exceed official development assistance (ODA) flows by about US$lOO billion. Remittances are also larger than private non-FDI flows, such as portfolio investment and Bank and trade related lending (US$115 billion). In relation to other financial flows, remittances are particularly important in South Asia, the Middle East and North Africa, and Latin America. Inthe first two regions, they would account for more than half o f the total international flows received whereas inLatinAmerica they would represent about 40 percent Table 1.1.Internationalflows to low and middleincome countries Region - Remittances Remitt. pc FDI Private non-FDI ODA US$million US$ US$million US$million US$million EAP 43 100 23 64563 26660 6916 ECA 19900 42 62211 69089 11869 LCR 42400 80 60843 -4460 6869 MENA 21300 68 5340 2980 10517 SA 32000 22 7151 12670 6758 SSA 8100 11 11276 8400 26004 Total 166800 31 211384 115339 68933 Last availableestimates. Source: GEP andWDI It is worth noting that this situation is radically different from the existing in 1990 when ODA flows accounted for almost half o f the financial flows to developing countries (figure 1.3). In fact, over the past 15 years private flows have increased dramatically at the cost of official flows, and inthis regard workers remittances are not an exception. 14 Figure 1.4. Remittances in Latin America: 1980-2003 E 1 0 .I = 4030 .I a $ 20 D 10 0 Set of 14 ISet of31 ~ Source: Own calculations It would be possible to argue that the previous figures have been artificially inflated by considering them in current rather than constant US$. And in fact, once we take into account the contribution o f US inflation, the over time increase in remittances flows becomes less marked. Yet, in 1980 prices today's remittances flows would still be above US$20 billion (Le. in real terms remittances have increased more ten fold over the past 20 plus years). Or consider the evolution in terms o f the region's GDP. In that case, remittances have also increased dramatically since 1980 when they represented a mere .3 percent o f GDP to about 2.2 percent today. Inother words, no matter how one looks at the evolution o f remittances, the message that emerges i s that they have increased significantly over a sustained period o f time. At the country level, Panel A o f Figure 1.5 reports the remittances to GDP ratio for a number o f Latin American countries in 2004, and table 1.2 reports remittances' annual flows, and growth rates for each country in the region between 2002 and 2005 - with preliminary estimates in the case o f the latter year. These figures indicate that remittances represent more than one half o f GDP (52.7 percent) in Haiti, a ratio that i s the highest in the world - Haiti i s closely followed by Tonga, where in 2004 remittances represented 51.3 percent o f GDP. A second tier o f countries with high remittances to GDP ratios includes Jamaica (17.1 percent in 2004), Honduras (15.6 percent), and El Salvador (15.3 percent). Similarly, in Guatemala, Nicaragua and the Dominican Republic remittances are between 10 and 12 percent o f GDP. The importance o f those flows can also be illustrated by comparison with other private capital flows. Thus, in Guatemala, Honduras, El Salvador and the Dominican Republic, remittances are equivalent to respectively 14, 4, 3 and 2 times FDI flows. Even in Colombia and Ecuador, where in relative terms remittances are lower than in several Central American and Caribbean countries, remittances represent respectively 197 and 112percent o f FDI. 16 Table 1.2. Recent remittancesflows to Latin America and the Caribbean(2002-2005) 21,773 25.4 1.7 2.3 2.7 Central America Guatemala 1,600 2,147 2,623 2,825 20.9 6.9 8.7 10.1 El Salvador 1,936 2,106 2,422 2,616 10.6 13.5 14.1 15.3 Honduras 718 867 1,153 1,699 33.3 11.0 12.6 15.6 Nicaragua 377 439 531 645 19.7 9.4 10.6 11.7 Costa Rica 251 321 355 442 20.8 1.5 1.8 1.9 Panama 85 85 82 80 -1.8 0.7 0.7 0.6 Belize 15 19 19 19 7.1 1.7 1.9 1.8 Caribbean DominicanRepublic 1,960 2,061 2,228 2,370 6.5 9.1 12.5 12.1 Haiti 1,325 1,622 1,862 2,086 16.3 38.2 54.9 52.7 Jamaica 1,260 1,398 1,487 1,520 6.4 17.0 17.5 17.1 TrinidadandTobago 79 62 193 317 58.8 0.8 0.5 1.5 Barbados 109 113 114 125 4.6 4.3 4.1 4.0 Antigua andBarbuda 11.1 26.7 16.9 17.7 16.7 1.5 3.5 2.1 Dominica 3.5 3.5 5.1 4.7 10.4 1.4 1.3 1.9 St, Kitts andNevis 3.7 3.6 3.9 4.0 3.4 1.0 1.0 1.0 St. Vincent 3.5 3.7 3.7 3.9 4.3 1.0 1.0 0.9 St. Lucia 4.0 4.3 3.1 2.2 -17.8 0.6 0.6 0.4 South America Colombia 2,480 3,076 3,521 3,802 15.3 3.1 3.9 3.6 Brazil 2,449 2,821 3,216 3,501 12.7 0.5 0.6 0.5 Ecuador 1,432 1,539 1,645 1,830 8.5 5.9 5.7 5.4 Peru 705 860 1,029 1,269 21.6 1.2 1.4 1.5 Argentina 586 700 823 864 13.8 0.6 0.5 0.5 Bolivia 158 171 182 193 7.0 2.0 2.1 2.1 Paraguay 202 222 233 172 -5.1 3.9 4.0 3.4 Chile 12.4 12.5 21.9 30.8 35.4 0.0 0.0 0.0 Uruguay 36.1 31.8 31.7 28.6 -7.5 0.3 0.3 0.2 Suriname 15.1 23.5 21.3 21.4 12.4 1.6 2.3 1.9 Source: Own calculations using BOP and WE0 data. Preliminary estimates. 17 Figure 1.5. Remittancesto LatinAmerica in 2004 PanelA. (YOof GDP) 60 1 I PanelB. (US$ millions) PanelC. Per capita 600 I Source: Own calculations using BOP data 18 The classification o f recipients changes significantly on a US$ basis. Overall, the country with the highest absolute remittances flows (Panel B o f Figure 1.5) is Mexico, which received $21.8 billion in 2005. This would represent 45 percent o f total flows to Latin America in that year ($48.3 billion) and would make Mexico the largest world recipient in 2005, followed by India, Philippines, China and Pakistan, with respectively $12.1, $9.8, $6.9 and $5.8 billion. Colombia and Brazil are ranked gthand 1lth among the top remittance receiving countries in the world, with flows o f respectively $3.8 and $3.5 billion in 2005. Other countries with more than $2 billion o f remittances inflows in 2005 ($2.5 billion on average) include Guatemala, El Salvador, the Dominican Republic and Haiti. Also worthy o f note are Ecuador, Honduras, Jamaica and Peru, where remittances averaged $1.6 billion in2005. On a per capita basis (panel C o f Figure 1.5) the country with the highest level o f remittances would be Jamaica with approximately $550 per capita, followed by Barbados with about $400 per capita and El Salvador with flows o f approximately $350. The average for the 28 countries being considered here would be $128 per capita per year, but that amount increases to $270 among the ten countries with highest per capita remittances - a group that also includes the Dominican Republic, Haiti, Antigua and Barbuda, Guatemala, Mexico, Honduras, and Trinidad and Tobago. I t i s also worth noting that between2002 and 2005 remittances to Latin America are estimated to have grown at a rate o f 18.7 percent per year. While this rapid growth has been largely drivenby Mexico's 25 percent annual growth in remittances flows, annual rates close to or above 20 percent have also been observed in Guatemala, Honduras, Nicaragua, Costa Rica, Trinidad and Tobago and Peru (Table 2.5). IV. Profile of recipients The BOP data reviewed in the previous section allows for a cross country comparison o f remittances. Yet, it gives no information about those at the receiving end. For example, are those receiving remittances the poorest groups in society so that remittances can be expected to have a large impact on poverty reduction? Or instead are they from more accommodated classes so that remittances, while certainly increasing welfare in the recipient country may not have that much o f an impact on poverty? In order to address these issues, it is necessary to have household specific information and therefore a natural way forward i s looking at household surveys. Unfortunately, nationally representative household surveys with specific questions on remittances are only available for eleven Latin American countries: Haiti, El Salvador, Honduras, the Dominican Republic, Nicaragua, Guatemala, Ecuador, Paraguay, Mexico, Bolivia and Perum4On a more positive tone, it is worth noting that in terms o f BOP data these countries would represent more than two-thirds o f remittances to the Region. We next review the profile o f recipients that emerge from these household surveys. A twelfth country with data on remittance receipts is Jamaica. Unfortunately as the household survey o f reference does not include information on other income sources (Le,, labor income), it is not included in the analysis that follows. Nevertheless, the Jamaican case is explored in Chapter 4 (remittances and household behavior). 19 How many households receive remittances? The number o f households receiving remittances in the Latin American region varies significantly across countries (figure 1.6). For example, in Haiti more than 25 percent o f the households reported having received remittances in 2001. At the other extreme, only 3 percent o f the Peruvian households would benefit from these flows. In between, remittances reach between 10 and 25 percent o f the households in the Dominican Republic, El Salvador, Nicaragua, and Honduras; between 5 and 10 percent in Mexico and Guatemala; and finally between 3 and 5 percent in Bolivia, Ecuador and Paraguay. Thus remittances are quite a popular phenomenon inthese countries. Figure1.6. Share of householdreceivingremittances 25 20 Y Ee 15 $t 10 5 0 Source: Own calculations using household data Who receives remittances in Latin America? A natural question that arises from the previous discussion regards the situation along the income distribution o f those households receiving remittances. Figure 1.7 shows the relative position o f Latin American recipient families according to income quintiles both for non- remittances income and for total income (Le. including remittances) for the eleven Latin American and Caribbean countries for which remittances micro-data are available. The figure i s computed by averaging the results for the individual countries (see also below for country details). Inspection o f this figure reveals two important elements. First, the income data that excludes remittances indicates that, on average for the 11 countries under analysis, 30 percent o f the households receiving remittances are in the lowest quintile o f the income distribution; the remaining 70 percent would be distributed more or less homogenously over the 4 richest quintiles. In other words, remittance flows seem to be directed to all income groups but the weight o f the poorest families appears to be larger than the weight o f everybody else. Second, once we take into account remittances income, recipient households climb up significantly inthe income ladder. In fact, after we take into account the role o f remittances only 10 percent of the 20 contrast, on thc 1 2 3 4 s Incornc Qunitile Panel B. Total iacome 1 2 3 3 5 Q1 Q3 Q4 El Safvador Guatemala 40 Jfl Haiti Honduras 3uI Q1 02 QS Nicaragua xo 30 20 I0 If Paraguay Dominican Republic 22 Figure 1.9. usctiolds receiving rcmittariccs by yuintile of the total income distr Bolivia Ecuador 91 Q2 Q3 Q4 ys yl Q2 Q3 Q4 QS ElSalvador Guatemala Haiti Honduras Paraguay Peru Dominican Republic 40 23 The data underlying figure 1.7, however, show significant country heterogeneity (figure 1.8), and while it i s clear that in many cases migrants and recipient families tend to come from definite socioeconomic segments o f the society, those segments considerably vary across Latin American countries. For instance, inthe case o f Mexico, remittance recipients are predominantly poor: 61 percent o f the households that report receive remittances fall inthe first quintile o f non- remittances income whereas only 4 percent o f them would be in the top quintile. Similarly, in Paraguay 42 percent o f recipients are in the first quintile o f the distribution and only 8 percent are in the top quintile. Other countries where at least 30 percent o f remittances recipients are in the lowest quintile (i.e. where there flows tend to be directed towards the lower quintile) are Ecuador, El Salvador, and Guatemala. In contrast, in Peru and Nicaragua the distribution of remittances across households is completely different. For example, in Peru less than 6 percent o f the households that receive remittances belong to the lowest quintile while 40 percent belong to the top quintile. Or take the case o f Nicaragua, where only 12 percent o f the recipients are in the first quintile while 33 percent belong to the fifth quintile. Thus in these two countries remittances seem to be flowing towards the richest. In between the group of Mexico, Paraguay, Ecuador, El Salvador, and Guatemala, and the group o f Peru and Nicaragua, there are four countries (Bolivia, Honduras, the Dominican Republic and Haiti) where remittances appear to be homogeneously distributed across the distribution o f income, or exhibit a U-shaped distribution (i.e. remittances flow towards the poorest and the richest inthe same proportion and more than towards the three middle deciles). This situation changes dramatically when we analyze the economic status o frecipients on the basis o f total income (including the value of remittances). Infact, figure 1.9 suggests that (i) the share of recipients that belong to the lowest quintile falls dramatically in all the countries; and (ii)with the exception o f Mexico and to a lesser extent o f Paraguay and o f El Salvador where 50, 40, and 34 percent of recipients respectively continue to be in the first and second quintiles, in the rest o f the countries more than half o f recipients are now in the two highest quintiles. Not surprisingly, this concentration is particularly marked in those countries where migrants seem to come from richer classes. Figure 1.9 indicates that in Peru more than 75 percent (50 percent) o f recipients are now in the highest two (top) quintiles o f the income distribution. The situation i s similar in Nicaragua, a country where more than 60 percent o f recipients belong to the top two quintiles. Thus, on the basis o f this analysis one would expect that remittances may have quite different inequality and poverty impacts indifferent countries. Remittances and education Understanding the selectivity process o f migration i s important in order to evaluate the overall impact o f migration on poverty and inequality (see Acosta, Calderon, Fajnzylber, and Lopez, 2006b for a technical discussion o f the relevance o f selectivity issues in this context; see also Chapters 3 and 4 for a more basic discussion). One important dimension o f the selection into migration process i s likely to be related to the patterns o f educational attainment patterns in the households with migrants, and ideally we would like to compare the educational attainment 24 !1.10. Educationalcharacteristics of households receiving remit Bolivia Ecuador Haiti Honduras 15 I O 5 ct 0 L-4 x x _ Nicaragua Paraguay iij 0 1-4 5-7 x - I 1 1'I u I1 5-7 8-11 i 2 t DominicanRepublic 25 According to that survey, the average education o f adult migrants i s 6.83 years, while non-migrant adults have on average 5.61 years o f education, which reinforces the idea emerging from figures 1.8 and 1.9 that migrants are positively selected inNicaragua (see also Chapter 2 for a discussion o f the education profile o f L A C migrants in the US). For the rest o f the countries, the only relevant information available relates to the schooling o f migrants' family members. However, the evidence for Nicaragua suggests that the level o f education o f migrants closely resembles that o f non-migrant adult members in the same household^.^ In other words, it seems that one can get a good approximation o f the patterns o f educational selection into migration usingdata on adult family members left behindby migrants. With this motivation, figure 1.10 presents the incidence o f remittance recipient households throughout the distribution o f the average years o f schooling o f adults aged 16 to 65. For example, in the case o f Nicaragua only 10 percent o f those without education receive remittances whereas about 30 percent o f those with 12+ years do. Similarly, in Peru only 2 percent o f those without education are recipients against 6 percent o f those with 12+ years. That is, remittances are relatively more important among those with higher education than among those with lower educational levels. On the contrary, in Mexico, Paraguay or Ecuador recipients tend to be more representative among the less educated groups. Take the case o f Mexico, where between 10 and 15 percent o f those with less than 4 years o f education receive remittances against less than 2.5 percent among those with 12+ years. Or take the case o f Paraguay where the corresponding shares for least and most educated groups would be 8 and 3 percent. Inthese two cases, remittances are relatively more relevant among less educated groups. Table 1.3 shows the proportion o f households with migrants across the educational distribution o f the four countries (El Salvador, Haiti, Honduras and Nicaragua) for which information i s available not only on households with remittances receipts but also on household with members who have migrated abroad. InEl Salvador, 14.5 percent of the surveyed families reported having international migrant members, while 19.3 percent o f them receive remittances - including money received from distant family members and friends abroad. The equivalent figures for Haiti, Honduras and Nicaragua show that respectively 29.9 percent, 10 percent and 12 percent o f the families have migrants, but respectively 27.2 percent, 10.9 percent and 15.6 percent receive remittances Table 1.3. Migration and Education Years o f Education ElSalvador Honduras Haiti Nicaragua 0 18.37 6.85 21.08 11.37 1-4 14.09 7.28 29.74 7.49 5-7 14.24 11.42 38.94 12.34 8-11 14.27 13.89 43.52 17.89 12 + 12.87 11.81 62.54 12.90 Avg. educ non-migrants HHS 6. Avg. educ migrants HHS 5.93 6.09 3.91 6.36 The table reports the % o f migrants by average years o f education inthe household. Source: 0 4 calculations usinghousehold s&vky data 'InNicaragua, the average years of schooling of adults inhouseholds withmigrants is 6.66, compared to 6.83 for migrants, and 5.61 for adults innon-migrants households. 26 Overall the results indicate the presence o f negative educational selection into migration inMexico, Guatemala, El Salvador and Paraguay,6 and ofpositive educational selectioninHaiti, Peru, Honduras, the Dominican Republic and Nicaragua. Our finding for Mexico i s consistent with Ibarraran and Lubotsky (2005), who show that Mexican migrant families tend to be less educated than their non-migrant counterparts. The evidence presented here is less conclusive for Bolivia and Ecuador, where migrants are likely to be drawn from several segments o f the educational distribution. Figure 1.11. Average annualamount reportedby recipients PanelA. US$ 1000 1 900 800 700 600 500 400 300 200 100 0 PanelB.Percent of income LAC refers to the un-weighted average of the 11 countries inthe figure. Source: Own calculations using the last available household survey. It i s also worth noting that Mexico, Guatemala and El Salvador are the three Latin American countries whose migrants in the US have the lowest educational levels (see chapter 2). 27 Fig e 1.12.Income.shareof remittancesby incomequintile(recipientsa Bolivia Ecuador 40 1 30 20 10 0 1 2 3 4 5 I 1 2 3 4 5 1 El Salvador Guatemala "" I I 40 60 40 20 20 0 0 1 2 3 4 5 I 1 2 3 4 5 1 Haiti Honduras 48 1 1 2 3 4 5 I 1 2 3 4 5 1 Mexico Nicaragua 40 20 0 1 2 3 4 5 1 2 3 4 5 Paraguay 40 40 30 20 20 10 0 0 1 2 3 4 5 1 2 3 4 5 50 7 I 1 2 3 4 5 1 Source: Own calculations using the last available household survey. 28 How important are remittancesfor recipients? An issue o f particular interest to infer the relevance o f remittances in this context is the amount received by recipient households. Figure 1.11 addresses this issue and suggests that for the 11 countries under analysis, on average, the typical household receiving remittances reports about US$500 per year (panel A). Here, there is also significant variation and whereas in the Dominican Republic, Mexico and Honduras remittances may be close to US$900 per receiving household per year, in Nicaragua and Haiti they would be much smaller. In fact, in Nicaragua household remittances would be estimated at about US$120 per receiving household whereas in Haiti they are estimated at US$191. Among the remaining countries, Peruvian and Ecuadorian households would report receiving between US$600 and US$700, Salvadoran US$430, and Guatemalan, and Paraguayan between US$200 and US$400. The country variability is much smaller when we look at remittances as a share o f the income o f the receiving families. In fact, apart from Nicaragua and Peru, remittances tend to represent between 30 and 50 percent o f household income (Panel B). However, much more variation across countries i s observed in the manner in which the share o f remittances in recipient households' income varies across income quintiles. Indeed, as seen in figure 1.12, while the income share o f remittances is generally higher among recipient households located in lower income quintiles-the rich are less dependent on remittances than the poor -the decline as ones moves up inthe income distribution i s quite slow in some countries and very fast in others. As an example o f the former case, in Mexico, Haiti, Paraguay and Bolivia the decline in the share o f remittances inincome from the bottom to the top quintile is o f about 10 percent. Inother words, in these countries the share o f remittances in income i s very similar in poor and rich households. In the other extreme, in Guatemala and El Salvador remittances represent respectively 63 and 55 percent o f income in households located in the first income quintile, but they fall to less than 20 percent of income among households in the fifth quintile. Similarly, in Peru and Nicaragua remittances are 50 percent lower as a percentage o f recipients' income in the bottom compared to the top quintile.Honduras and Ecuador are inan intermediate position, with declines inthe income share o f remittances among recipients o f about 30 percent from the first to the fifth quintile. How regressive is the distribution of RemittancesIncome? Given that a larger share o f recipients tend to be located in the upper quintiles o f the income distribution (figure 1.9) - the only exception being Mexico - the finding that remittances are higher as percentage o f poor recipients' income compared to their richer counterparts (figure 1.12) does not imply that remittances have a progressive effect on the distribution o f income. In fact, as seen in figure 1.13, with the only exceptions o f Mexico, El Salvador and to a lesser extent Paraguay and Guatemala, remittances are a bigger share o f total income in the upper quintiles o f the income distribution - considering all households in each quintile, regardless of their remittances recipient status. Thus, for instance, in Haiti, almost 25 percent o f the income of households in the top quintile comes from remittances, compared to respectively 5 and 10 percent for the first and second quintile. Large differences o f this sort are also found in the Dominican Republic and Honduras, where among the richest 20 percent remittances represent 29 respectively 11 and 8 percent o f total income, compared to an average o f 5 and 3 percent respectively for households inthe two lowest quintiles. Fi; e 1.13. Incomeshare of remittancesby incomequintile(all househ Bolivia Ecuador 3 1 4 3 2 / I i o 1 2 3 4 5 I 1 2 3 4 5 El Salvador Guatemala l5 1 1 2 3 4 5 1 2 3 4 5 Haiti Honduras 30 - I I 1 0 1 1 2 3 4 5 1 2 3 4 5 Mexico Nicaragua 4 1 1 2 3 4 5 1 2 3 4 5 Paraguay Peru 2 1 I2 1 1 2 3 4 5 1 2 3 4 5 Dominican Republic l5 1 I O 5 0 1 2 3 4 5 1 Source: Own calculations using the last available household survey. 30 Figr 1.14. Incomeand RemittancesDistributionby income quintile Bolivia Ecuador 80 1 6._ 0 , I 60 40 40 20 20 0 0 1 2 3 4 5 I 1 2 3 4 5 1 El Salvador Guatemala 50 1nn -" _I I 40 30 40 20 20 I O 0 0 1 2 ' 3 4 5 1 2 3 4 5 Haiti Honduras 100 1 80 60 40 20 0 1 2 3 4 5 1 2 3 4 5 Mexico Nicaragua I 50 80 -, 40 60 4 30 20 10 0 1 2 3 4 5 1 2 3 4 5 Paraguay Peru 40 60 40 20 20 0 0 1 2 3 4 5 , I 1 2 3 4 5 1 Dominican Republic 80 i 60 40 Total Income (incl.Rem.) 20 0 1 2 3 4 5 I S :e: Own calculations using the last available household survey. 31 More direct evidence on the extent to which remittances have a regressive effect on the distribution o f income i s provided in figure 1-14,which reports each quintile's share o f both total income and total remittances. The results suggest not only that remittances are distributed in a quite unequal fashion, but also that that are generally distributed more unequally than total income. Thus, inthe 11 countries for which we have data the first three quintiles - the poorest 60 percent o f the population - receive only a quarter o f total remittances, while the top quintile receives on average 54 percent o f those flows. For comparisons purposes, on average the richest 20 percent respond for 51 percent o f total household income, which suggest that the distribution o f remittances i s only slightly more unequal than that o f income. Figure 1.14, however, reveals that inthe cases o f Mexico, El Salvador, Guatemala, and Paraguay remittances are less unequally distributed than total income - e.g. the poorest 60 percent receive 41 percent o f remittances compared to 29 percent o f income. In contrast, in the other 7 countries the first three quintiles respond for only 16 percent o f total remittances, compared to 26 percent o f total income, thus suggesting that remittances have a regressive effect on the income distribution. While these calculations are subject to a number o f caveats which we address in detail in chapter 3 - e.g. in the absence o f remittances households would probably have generated incomes that are likely to be higher than the observed non-remittances income - the evidence so far does not suggest that remittances could play an important role in reducing the very high levels o f income inequality observed inLatin America. V. How reliable are remittances figures? So far we have reviewedthe magnitude o f remittance flows to the LatinAmerican region, their evolution over time, and the profile of receiving households. Yet we have also mentioned that remittances data can be subjected to significant mismeasurement issues to the point that it could be possible that part o f the observed upwardtrend inobserved remittances is simply driven by improvements in recording procedures. This clearly raises many questions regarding the quality and coverage o f the data with which we have been working, which could in turn affect the robustness of the analyses basedon these data. Inother words, casual evidence indicates that there may be important deficiencies with the officially recorded statistics on remittances. Inthis section we perform an exercise that we hope will give some indications o f the extent to which the data described inthis chapter can be trusted. Remittances data are typically computed from Balance of Payments statistics for aggregate analysis and from household surveys for country specific studies. As argued in the World Bank's Global Economic Prospects 2006, there are a number o f reasons o f why these two sources may not coincidea7To start with, there i s not a single definition o f remittances. Should they just include current transfers by migrants that are employed and resident inanother country, or should they also include compensation o f employees (i.e. current transfers by nonresident workers)? Similarly, what should be the treatment o f migrants' transfers (Le. financial items that arise from migration from one economy to another)? 'Indeedremittances from these two sourcesof datararelycoincidefor individualcountries. 32 Moreover, even when one agrees on a common definition there are still a number o f important issues that may affect the quality o f the data. Inaddition to the implicit difficulties o f capturing flows through informal channels, weaknesses in data collection may imply that many formal flows, especially if small, go unrecorded or are misclassified as tourism receipts or nonresident deposits. On the other hand, household survey based remittances have also a number o f associated problems. First, the corresponding household surveymay not be as "national representative" as it should. Moreover, even if the household i s representative for the purposes o f analyzing income levels, it may not be representative when the focus i s the analysis o f migration and remittances. A second aspect would be related to possible recall bias estimates o f remittances' receipts: families usually decide to pool their income regardless o f their source, and therefore are presumably more likely to remember whether they received financial aid from abroad or not, than to remember the exact amount. Finally, the timing o f the survey data collection may also have an impact. Anecdotic references suggest that remittances' flows tend to be highly volatile within a year, with peaks at particular dates (Le., Christmas, Mothers' Day, and birthdays). If survey data do not ask for the amount perceived as remittances during the whole year period (as in the case o f the Dominican Republic, Ecuador, Paraguay and Peru, for instance), depending on the data collection period, remittances could be underestimated. Similarly, in countries with volatile exchange rates, it is difficult to value (in dollars) the amounts remitted when they are reported in local currency values (as in the majority o f household surveys). Yet, if official figures are reliable one could also expect a strong positive correlation between the Balance of Paymentsbased and household based estimates. Figure 1.15 compares the estimates (as a ratio to GDP) obtained using the two available sources o f data that we have for the eleven Latin American countries for which the household surveys report data on remittances. Inspection o f this figure suggests a number o f interesting issues. First, BOP based estimates tend to be larger than household survey data estimates. The only case where the latter i s higher i s Mexico and only by .1 percentage points. Second, the discrepancies can be very important. For example, in the cases o f Haiti and El Salvador BOP based data produce estimates that are 10 percentage points o f GDP higherthan Household survey data. In Nicaragua, the differences are smaller but still o f a substantial magnitude (about 6 percent o f GDP). The medianremittances to GDP ratio for the 10 countries under analysis would be 4.7 percent on the basis o f the Balance of Payments data and 2.7 on the basis o f household survey data. It is worth noting, however, that these discrepancies may be more predictable than they appear. Or in other words that the observed discrepancies reflect systematic differences more than unexplained gaps. In fact, Figure 1.15 indicates that after a logarithmic transformation the Balance of Payments and household based series o f remittances are strongly correlated: the estimated R2 o f a simple OLS regression is almost -80 indicating that most o f the variance in observed Balance of Payments based figures can be explained with the findings o f household survey data. 33 The slope o f the regression line, estimated at .94, is also quite close to 1. Even though if by working with a sample size o fjust 11 observations we cannot pretend to do rigorous statistical inference it i s worth noting that the null hypothesis that this slope i s equal to zero i s rejected by the data whereas the nullofbeingequal to 1cannot be rejected. Inother words, the differences in the estimated remittance flows emerging from the two different sources o f data seem to be more a question o f scale. These results, which are virtually unchanged when Haiti is eliminated from the sample to control for extreme values, would suggest that a good rule o f thumb relating remittances from these two sources would be givenby: BOP Remittances=l.73 x (HHRemittances) Inother words, Balance of Payments based remittances statistics tend to be on average 70 to 75 percent larger than household baseddata. Figure 1.15. BOPbasedversus householdbased remittances 4 1 3.5 j 0.94 I n 0+0.55 Haiti 3 IIn(B0P)=R' =0.79 2 2.5 =d 2 L 1.5 1 0.5 0 0 0.5 1 1.5 2 2.5 3 3.5 I HHsurvey based Source: Own calculations VI. Conclusions. This chapter has reviewedthe evolution o fremittances flows to Latin America usingboth balance o f payments statistics and household survey data from eleven Latin American countries. I t has also explored the varying economic and educational profile o f the households that receive remittances using household survey data from those eleven countries. It has been argued that remittances flows to Latin American are very important regardless o f whether one looks at them as total flows (about US$40 billion), per capita flows (US$SO per person per year), their contribution to total financial flows (about 40 percent), or their share o f regional GDP (2.2 percent). The chapter has also noted that remittances have dramatically increased over the past 25 years (twenty-fold innominal terms and ten-fold inreal terms), and that there i s no indication 34 o f them leveling off. At the country level, they are relatively most important among Central American and Caribbean countries. As for the characteristics o f remittance receiving households, the analysis o f household survey data indicates that their distribution by income and adult educational attainment varies considerably across Latin American countries, with households with remittances income coming mainly from the bottom o f the distribution in some countries - e.g. Mexico and Paraguay - but an opposite pattern being found in others - e.g. Haiti, Peru and Nicaragua. This would indicate that the impact o f remittances on poverty and inequality cannot be expected to be the same across the different countries o f the region. Finally, it has also been noted that Balance of Payments statistics and household survey data do indeed produce dramatically different estimates o f the importance o f remittances for receiving countries. However, the differences between those two data sources appear to be somewhat predictable, something that at least in our sample should ease concerns about data quality. 35 Chapter 2 Before workers' remittances, workers' migration * While workers' remittances would not occur if those sending them Migration and remittances are twofaces of the same coin. had not migrated in thefirst place, migration is often motivated by the desire to improve the welfare of those left behind, mainly by sending them money. Thus, migration profiles in terms of destination and characteristics are likely to influence remittance patterns. The present chapter aims to describe Latin American migration in an international perspective, and relate it to the remittances patterns that are sent home by the region's migrants. What do we know about the destinations chosen by Latin American migrants? What are the age and education profiles of the region's migrants and what is their labor market performance in destination countries? How do these characteristics affect remittancesjlows? I.Introduction There are many countries in the world for which migration is a key economic and social issue. For many small countries, like Tonga and Moldova, remittances account for the largest portion o f their capital inflows. Larger countries, such as Sri Lanka, Bangladesh and the Philippines, have millions o f citizens for whom working abroad for several years at a time i s an important economic activity. Countries like India, Korea and Taiwan, have seen some o f their most educated workers migrate abroad over the last three decades, but they also have benefited tremendously through the closer economic links these migrants established between their native and adopted countries. For others with large diasporas in Europe, such as Turkey, ex-Yugoslav Republics and the North African countries, migration poses complex cultural challenges in addition to economic ones. The experiences o f Latin American countries have certain similarities with some o f the examples mentioned above. As discussed in Chapter 1, remittances are extremely important for many o f the countries inthe region and there are countries like Haiti where remittances are close * This chapter is based on the background papers for this report "Migration Patterns in Latin America and the Caribbean: An Overview of Destination, Age, Educationand OccupationProfiles" and "Remittance Flows in Latin America andthe Caribbean:Patternsand Determinants"bothby Caglar Ozden. 36 to 40 percent o f GDP, Mexico where official flows are estimated at more than US$l0 billion, or Jamaica where annual remittances are about US$500 per person. Similarly, as discussed below, many countries in Central America and especially in the Caribbean also suffer from significant emigration o f highly educated workers, what i s usually referred as the bruin drain. On the other hand, there are elements that are specific to Latin America. For example, unlike Asian countries where migrants tend to spend a few years abroad and then return to their home countries, most migration from Latin America i s permanent and a significant portion i s undocumented. Also, regarding final destinations, most Latin American migrants (and inthe case o f Mexico, virtually all migrants) choose the U S due to the geographic proximity, existence of social networks and relatively more accepting culture o f migrants. This chapter first reviews the patterns o f migration o f Latin American countries and to that end it relies on the censuses o f the destination countries (mainly the US) which offer a detailed picture o f the existing stock o f migrants. It then explores the profile o f migrants interms o f age, education, and performance in the labor market o f the destination country. Finally, the chapter relates the observed characteristics o f the stock o f migrants to the observed official flows o f remittances as discussed in Chapter 1. We have to acknowledge that the picture presented in this chapter i s undoubtedly biased because due to data limitations we do not take into account South to South migration flows, which in some cases can be important. For example, it i s well known that a large number o f Nicaraguans migrate towards Costa Rica. Household survey data for 2001 show that 59 % o f Nicaraguan international migrants go to Costa Rica (29% to the US). According to Bank's Poverty Assessment for Costa Rica roughly 20,000 Nicaraguans migrated to Costa Rica each year in the late 1990s (although that number declined to about 9,000 per year in the 2000s). By 2005, it i s estimated that the number o f Nicaraguans in Costa Rica was around 300,000 persons, a significant amount taking into account a population o f about 5 million for Nicaragua. Similarly, according to the 2001 Haitian household survey, 20% o f Haitian international migrants go to the Dominican Republic. However, to the extent that one is willing to assume that most o f Latin American migrants take as destination a developed OECD country, then the biases will be relatively small. 11.LatinAmerican migrantsabroad How many Latin Americans live abroad? Inprinciple, countries with a larger population living abroad - both in absolute and relative terms - would experience more economically significant remittance flows. Indeed, the World Bank Global Economic Prospects 2006 states that the stock o f migrant workers i s associated with higher worker remittances. Panels A and B o f Figure 2.1 present the overall number o f migrants from the region using two different definitions and destinations. Panel A presents migration to the U S whereas Panel B focuses on migration to all OECD countries from the region. Finally, Panel C presents the ratio o f the labor force abroad to the total native labor force o f country o f origin. In these panels, the countries have been grouped according to geographic region and economic similarity. First, we list Mexico, followed by the Central America countries, the Caribbean countries, Andean countries and finally the rest o f South America. 37 Figure2.1. Latin American migrants. PanelA. Total Latin American migrantsin the US (thousands) PanelB. Latin American migrantsin the OECD labor force (thousands) IO w 0 1ow I W 10 L PanelC. Share of migrantsabroadrelativeto homepopulation,2000 1 Io allmigrants tertiaryeducatedI 80% 70% 60% 50% 40% 30% 20% IO% Source: Own calculations 38 Before proceeding, the reader should note that these graphs use a logarithmic scale so that Mexican migration does not visually overwhelm the other smaller countries. In fact, the differences between Mexico and the rest o f the Latin American countries are very marked. Panel A indicates that, in 2000, there were close to 10 million Mexicans living in the US. That same year, according to the US Census figures, the number o f Cubans (870 thousand) or Salvadorans (820 thousand) in the US (i.e. the countries with the second and third highest number of migrants, respectively), represented less than 10 percent o f the Mexican figure. One common critique o f the Census i s that it undercounts undocumented migrants for a variety o f reasons. Among these, the desire o f the migrants to minimize their contact with the US government out o f fear o f deportation i s considered the most serious one. However, the Census claims to have measures to correct for the problem and many studies confirm that the problem is no greater than 10 percent o f the population inmost cases. In addition to Cuba and El Salvador, there are several other countries with a stock of migrants betweenhalf a million and one million inthe US: Dominican Republic (680 thousand), Jamaica (550 thousand), and Colombia (510 thousand). In absolute terms, and apart from the small Caribbean Island o f St. Kitts and Nevis, the country with lowest number o f migrants inthe US is Paraguay with less than 13 thousand. On the whole, the total number of Latin American migrants inthe U S increased from 8.6 million in 1990 to about 16 million in 2000 (an 86 percent increase). Evenexcluding Mexico, which clearly dominates the absolute figures, migration to the US from the other countries in the region increasedby 63 percent from 4.2 million migrants in 1990 to 6.8 million in2000. Among the countries inthe region, Honduras experienced the most rapid growth interms o f the number o f migrants - from 112 thousand in 1990 to 281 thousand (i.e. a 150 percent increase) ten years later. It is followed by Brazil with a 120 percent increase from around 95 thousand migrants to about 210 thousand and Mexico which more than doubled its migrant population in the US. Duringthis period, there were only two Latin American countries whose number o f migrants in the US declined: Panama (from 121 thousand to 104 thousand) and Dominica (from 18 thousand to 17 thousand). Inabsolute terms, the largest increases inmigrants in the U.S.between 1990 and 2000 were from Mexico (4.7 million) followed by El Salvador (343 thousand). Panel B o f Figure 2.1 presents the number o f migrants from the region inthe labor forces o f all OECD countries. These data, which are based on the Docquier and Marfouk (2005) dataset on brain drain, excludes children and other migrants who are not in the labor force. It i s worth noting that there is significant correlation with Panel A since the U S i s the main destination for the migrants from many o f the countries. Yet, for many South American countries, Europe continues to be an important destination. Insome cases, migrants to the U S from South America tend to represent less than 50 o f those countries' migrants. Panels A and B o f Figure 2.1 hide an important issue. Eventhough there might be many migrants from a given country, say Mexico, migration can be relatively more important for countries with much smaller populations. In fact, the number o f Mexican migrants in the US is 39 larger than the total population o f several o f the countries in the region. To address this bias, Panel C presents the ratio of migrants to the overall labor force inthe home country I Figure2.2. Major destinationof LatinAmerican migrants PanelA. 1990 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% UnitedStates HEurope 0 Australia+Canada 0 Japan Mexico 1 PanelB. 2000 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% U HUnitedStates Europe 0 AustraliaXanada 0 Japan W Mexico Source: Own calculations The most striking feature o fthis panel i s that there i s significant level o fmigration from a number of small countries, especially inthe Caribbean. For example, as of 2000 on average 30 percent of the labor force of many Caribbean Islands had migrated. An extreme case is that o f Grenada, where close to 50 percent o f the population had migrated. For the non Caribbean countries, migrants as a share o f the origin country population would represent on average about 40 10 percent (6 percent for South America). Thus even if Mexico and the Central American countries tend to top the ranking o f migrants in absolute terms, the small Caribbean Islands clearly dominate the migration charts when we look at migration flows in relation to each country's population. To explore the destination choices o f Latin American migrants in more detail, Figure 2.2 presents the share o f migrants in the US, Europe, Australia + Canada, Japan and Mexico for 1990 (Panel A) and 2000 (Panel B) respectively. The U S i s clearly the main destination for the migrants from Mexico and Central America. In fact, Mexico i s the second most important destination for migrants from Central America since it serves as the first stop on their way to the US. For the Caribbean countries, Europe (especially the United Kingdom) and other OECD countries (mainly Canada) are the main destinations after the US, accounting for 50 percent o f the migrants for several countries. Europe i s a major destination for Andean and other South American countries, mainly due to colonial and ethnic links. An interesting example i s the migration from Brazil to Japan where grandchildren o f the ethnic Japanese who migrated at the turn o f the century are now going back to Japan for employment. In short, the U S is the destination for less than 50 percent o f the migrants in South America and slightly above 50 percent for the Caribbean countries. Nevertheless, since the vast majority o f migrants from LatinAmerica and the Caribbean are from Mexico and Central America, over 90 percent end up inthe U.S. Among the European destinations o f Latin American migrants, Figure 2.3 indicates that language seems to play an important role. This figure i s restricted to the Caribbean and South American countries and it indicates that the Caribbean migrants' preferred destination is the United Kingdom whereas that o f the South Americans i s Spain. Figure 2.3. Major Europeandestinationof LatinAmerican migrants iource: Own calculations 41 Y Scttrrcc:csCensusf2UQO) 42 w to bc older, Thc cxt 43 for Central Am arrived during the 1990s - ~x x x Source: from TahIc 2.1, which ucation for thrcl: ca. In fact, a majority of migrants ation, Fa contrast, a majority of 44 would have secondary education. Finally, the most educated migrants are those from South America where more than three-quarters have secondary or tertiary education. Table 2.1. Educationprofile of LAC migrants(YOwith a given educationallevel) Primary Secondary Tertiary Mexico and Central America 51 37 12 Caribbean 34 53 13 South America 24 48 28 Source: Own calculations basedon the 2000 US Census data One o f the key questions is what factors explain this divergence. One possibility is the overall education level in the underlying population in the home country. In Figure 2.7, the horizontal axis presents the tertiary school enrollment in the home country. This ratio ranges from 1.5 percent in Haiti to the 20 percent in many countries such as Costa Rica, Panama Venezuela and Argentina. The vertical axis i s the ratio o f tertiary educated among the migrants who are currently inthe US. Figure2.7. Educationprofileof nativepopulationversus the Migrants for LAC countries I + 0 m Venezuela * Argentina v) 3 0 + + c Brazil Chile + St KiMNevis c0 * Costa Rica Dominica + Haiti Ecuador Daminloan Republic onduras /// I Mexi00 0 - # - r I I . I ' I I I Source: US Census (2000) and World Development Indicators (2005) Figure 2.7 indicates that Mexico and most o f the Central American countries are below the 45 degree line; that is, migrants are drawn from the lower end o f the education spectrum in the home country. Incontrast, the Caribbean and the South American countries are above the 45 degree line which means the migrants are proportionally more educated than those who remain behind. For example, even though education levels in Brazil and Mexico are similar, their migrants are starkly different intheir education profiles. One possible explanation o f this finding i s that it i s relatively easy for Mexican and Central Americans to migrate to the US whether 45 QSCSO pcrccnt of' a Europe 46 gcr portion of a vcry s work by Oscquicr and mcrica, thc ~ i ~ lcvcl ofr thca ~ ~ ~ ~ Figure 2,9. Share of CUI! d Workers who re d their degrees at home. m ." w 47 2.10. Share of 48 VeLabor market performance ci. In othcr war c w ent age 22-1.) who were below 17 at time of arrha1 of social and ccononxk and Garibbcanc o f Mcxiccl and Ccnfral Which migrant ~ h a ~ ~ ~ ~ e ~ i faini cremitfancc patterns? I t is not hard to argue that s t s migrant p~pula~ionp r o b ~ one yf i s ~ ~ o most important ~ ~ ~ c0 ~ i n a ~ ~ ~ differcnt ways of m e ~ s ~thci ~~ 0~ ~ # ~of tmigrants will affect r a ~ o ~ t ~ ~ ~ ~ toaBSSUIXICothat~thcatots1 ~ ~ ~ e Figure 2.13 prcscrrts thc avcrage fcvcl of as a share of the ho n Latin American sub-rcgions to the US ~ ~ ~ ris~the~sub-rcgionn b ~ a at I 5 pcrccnt o f t t in 1980. Thc samc num c r a m in 1980) arid around 1E3 74 l-2 10 8 6 4 2 0 I 52 A sccond c h a r a c ~ ~ ~tostake into account is thc inco i t i ~ ti tcndcncy for ~~~~~n~~ from one country to cam signific lo.yxrer and it is Figure 2,lJ. Share of FemaleMigrants 58 56 5.4 52 50 48 4.6 44 42 40 important variable is f thc highly cducatcd cspforcd in dctait ca chapter and it is an 53 14 j 12 t 10 t 8 6 4 2 0 1991 1992 1993 7994 1995 7996 1997 1998 1999 2000 45% ; 40% 35% 30?# 25% 20% 15% -I OOh 5Yo 0Yo 1 1988 7990 1992 1994 3996 1 2oc Source: Own calculations Chapter 5 below argues that remittances may have positive effects on financial development o f the recipient countries and provides new evidence on the subject. However, the reverse i s likely to be true as well: migrants would be more likely to send money back home if the financial sector was more developed and efficient. Furthermore, financial development i s likely to increase the portion of remittances flowing through formal channels which are, in return, more likely to be accurately captured in official remittance statistics. Figure2.16 presents the Bank Deposit/GDP ratio for the three sub-regions o f LAC and indicates that the Caribbean i s the more financially developed sub-region, which can partially explain the higher level o f remittances. Another frequently used measure o f financial development i s the ratio o f Credit to GDP, which we as a proxy use inthe empirical section. The final set o f variables employed in our analysis influences remittance flows through the macroeconomic conditions o f recipient countries. The first one is per capita GDP. The motivation for including this variable is that if income levels are low in a country, migrants might be more likely to remit money to their families. The second variable is economic growth. Ifthe investment climate is favorable and a country is growing rapidly, migrants mightbe more likely to remit for investment purposes. The corresponding flows may be destined to real estate investments, which are highly favored by migrants in many countries, physical capital investments, such as businesses and capital goods, or human capital investments, such as the education o f their children. Third is size o f the country. There i s much evidence that argues that smaller and isolated small economies can not sustain many economic activities which induces their citizens to migrate. Thus, smaller countries - controlling for the above two factors - are likely to attract lower levels o f investment. Empirical strategy The goal o f this section is to identify which o f the above listed variables are significant determinants o f remittance flows for the Latin American countries. The first step i s to estimate the following equations for three differentmeasures o f remittances. (1) RemittancedGDPi, = a + +++ psLog Migrants/Populationit+ p 2BankDeposit/GDPi, p3 FemaleMigrantRatioit + p 4CollegeEducatedRatioi, MigrantIncomeit + p 6GDPGrowthit p7LOg GDP/Capitait+ p8 Log GDPii + Eit (2) Log Remittanceskapitai, = a + + p3 Log MigrantsAbroadit+ p2 BankDeposit/GDPi, + psFemaleMigrantRatioit+ + p CollegeEducatedRatioit 4 Log Migrantlncomei, p 6GDPGrowthi, + p7L0gGDP/Capitait+ p8Log GDPit + it (3) Log Remittances/migrantit = a ++ psLog + Log MigrantsAbroadit + p 2BankDepoMGDPi, p3 FemaleMigrantRatioit + p 4CollegeEducatedRatioit Migrantlncomeit + p 6GDPGrowthit 55 where i and t are a country and time subscript respectively. Each equation is first estimated using OLS with robust standard errors and the log o f the native populations as weights in order to avoid attaching too much weight to smaller or larger countries in the sample. Then we estimate the same equations usingboth country and year fixed effects to isolate time invariant country specific conditions - such as distance to the U.S.- and period-specific effects that have the same effect in all countries - such as for example the business cycles inthe US. One o f the crucial issues in the migration literature is that migrants are not a random sample o f the underlying native population o f the source country. In some cases, they are more educated than their fellow citizens, which is the case with the Asian and Middle Eastern migrants inthe US as well South American migrants as identifiedabove. Inother cases, they come from rural areas or are from a certain region. Inthe Latin American migration context, a large portion o f the migrants, especially from Central America and Mexico, are relatively unskilled migrants who plan to send home significant amount o f remittances. In other words, while the level o f migration affects remittance levels, the desire to remit also influences migration levels. This endogeneity bias needs to be controlled in our estimation via instrumental variables. Thus, we add a first stage estimation to each equation listed above where we use several instrumentsfor the migration variables. These are the cost o f obtaining a passport, the ratio o f urban population and the population density in each country. Empirical Results Tables 2.2 to 2.4 report the results o f estimating equations (1) to (3), respectively. The first two columns correspond to OLS estimation, the third and fourth to fixed effects, and the last two to instrumental variables estimation. The mainmessages are as follows: 0 Migration levels increase remittance/GDP levels and decreases remittances sent per migrant. The impact on remittances received per capita is ambiguous. As expected migration levels are positively correlated with remittance levels when remittances are measured as the remittance/GDP ratio. On the other hand, an increase in the number o f migrants decreases remittances sent by each migrants. This i s likely to be due family reunification effect - as more people migrate the aggregate amount sent i s divided among more senders. 0 Increases in the overall education levels of migrants tend to reduce remittances sent. This finding confirms the predictionthat once we control for migrants' incomes, higher education levels are associated with lower remittances. Note however, that the effect loses some degree o f significance in the IV estimation even though the coefficients are still negative. This i s likely to be due to the interaction between this variable and the instruments. 56 0 Economic growth in the recipient country tends to increase remittance levels. Higher economic growth i s a sign o f stability and more abundant investment opportunities. This i s likely to encourage migrants to send remittances for investment purposes Financial development variables have a significant and positive effect on remittances. In the OLS and fixed effects estimations, financial development variables, DepositdGDP and CredidGDP, have ambiguous signs. However, in the IV estimation, they have positive and generally significant coefficients indicating that financially more developed countries attract higher levels o f remittances. This is complementary to the findings on the positive effects of remittances on financial development that are explored indetail inlater chapters. 0 The share of female migrants does not have a significant effect on remittanceflows. Inthe cases o f remittances per capita and per migrant, the coefficients on this variable are negative as predicted but they are non-significant inmost cases. Inthe case o f remittances as a ratio o f GDP, the share o f female migrants has a positive coefficient but again i s not significant in most specifications. Income per capita and size of the economy influence remittance flows. Overall economic prosperity is likely to be among the most important determinants o f remittance flows. These variables, log GDPhpitu and log GDP have highly significant coefficients, especially inthe IV estimates. GDP/capita and total GDP respectively reduce and increase remittances flows, both as a share of GDP and on a per migrant basis. On the other hand, the impact on remittances received per capita i s not statistically different than zero. 57 Table 2.2. Determin nts of Remittan es (as a share 01 ZDP) Variable OLS Fixed effects IV (3) (4) (5) (6) Migrants abroadlpopulation 0.139 0.129 1.124 0.941 t-stat 5.26 5.29 1.78 1.63 1.42 1.95 Bank deposit/GDP 0.499 0.97 6.356 t-stat 0.38 0.48 1.69 Bank credit/GDP 1.541 -1.596 0.716 t-stat 1.20 -0.82 0.29 Share of female migrants 8.058 7.579 9.598 12.46 21.936 11.036 t-stat 1.83 1.63 0.77 1.00 0.68 0.40 Share of college graduate migrants -3 1.522 -3 1.794 -29.587 -31.619 -20.783 -25.453 t-stat -5.31 -5.53 -3.33 -3.66 -0.33 -0.48 L g of average income of migrants 0.359 0.326 -0.737 -0.808 -1.788 -2.108 t-stat 1.72 1.56 -1.07 -1.15 -0.72 -0.96 G D P growth 0.04 0.039 0.066 0.057 0.004 0.005 t-stat 0.93 0.91 2.54 2.04 0.12 0.15 L g of GDP per capita -1.174 -1.284 -20.819 -20.6 -50.145 -46.142 t-stat -2.59 -2.81 -4.46 -4.31 -2.52 -2.45 L g of GDP 0.072 0.122 14.212 14.47 45.449 41.383 t-stat 0.45 0.78 2.97 2.94 2.14 2.04 Constant 3.305 3.357 -206.527 -215.062 t-stat 0.74 0.76 -2.03 -2.07 Observations 270 270 270 270 226 226 R2 0.459 0.462 0.891 0.891 0.149 0.427 F-statistic (o-value) u 0 0 0 0 0.6 0.76 smittances to GDP on the variables in the first column. IV estimation uses as instrument for remittances passport costs, the ration of urban population and the population density ineach country. Source: Own calculations 58 Table 2.3. Dete inces (per capita Variable Fixed effects (3) (4) migrants abroad/population 0.516 0.484 0.05 -0.052 1.734 2.27 t-stat 6.04 5.90 0.10 -0.10 4.20 2.88 Bankdeposit/GDP -1.411 1.278 2.293 t-stat -2.21 1.37 2.29 Bankcredit/GDP -0.894 -0.047 1.398 t-stat -1.36 -0.05 1.33 Share of female migrants -1.214 -0.872 -7.747 -6.607 -3.57 -0.991 t-stat -0.54 -0.39 -1.87 -1.57 -0.64 -0.16 Share of college graduate migrants -9.922 -10.999 -9.247 -10.226 -7.765 -6.23 t-stat -4.10 -4.60 -1.53 -1.65 -1.14 -0.72 Lg of averageincome of migrants 0.516 0.484 -0.111 -0.142 -0.635 -0.718 t-stat 3.58 3.34 -0.37 -0.47 -1.42 -1.42 GDP growth -0.008 -0.009 0.0167 0.013 0.013 0.015 t-stat -0.57 -0.60 1.86 1.44 1.08 1.36 Lgof GDP per capita 1.298 1.265 -3.651 -3.886 -2.386 -2.392 t-stat 11.39 10.91 -2.46 -2.57 -1.10 -0.95 Lg of GDP -0.743 -0.679 3.259 3.694 2.52 2.795 t-stat -8.54 -8.28 2.12 2.35 1.32 1.28 Constant 5.363 4.638 -43.573 -51.385 t-stat 2.47 2.10 -1.57 -1.82 Observations 270 270 270 270 226 226 R2 0.579 0.575 0.913 0.911 0.366 0.318 F-statistic @-value) 0 0 0 0 0.17 0.12 mces per capita on !variables in the first column. IV estimation uses as instrument for remittances passport costs, the ration of urban population and the population density in each country. Source: Own calculations 59 Table 2.4. Deter nces (per migra Variable Fixed effects (3) (4) migrants abroadlpopulation -0.483 -0.516 -0.949 -1.052 -1.483 -2.248 t-stat -5.68 -6.29 -1.87 -2.00 -1.85 -2.75 Bank deposit/GDP -1.411 1.278 3.742 t-stat -2.21 1.37 3.03 Bank credit/GDP -0.894 -0.046 2.814 t-stat -1.36 -0.05 2.89 Share of female migrants -1.214 -0.872 -7.747 -6.607 -12.51 1 -13.149 t-stat -0.54 -0.39 -1.87 -1.57 -1.29 -1.24 Share of college graduate migrants -9.922 -10.999 -9.247 -10.226 -9.1 -4.255 t-stat -4.I O -4.60 -1.53 -1.65 -1.82 -0.92 L g of average income of migrants 0.516 0.484 -0.111 -0.142 0.265 0.442 t-stat 3.04 3.36 -0.37 -0.47 0.39 0.57 GDPgrowth -0.008 -0.009 0.016 0.013 0.032 0.044 t-stat -0.57 -0.60 1.78 1.44 2.46 3.67 L g of GDP per capita 0.298 0.265 -4.65 1 -4.886 -14.245 -12.214 t-stat 2.61 2.30 -3.I 3 -3.23 -4.20 -3.34 L g of GDP 0.257 0.32 4.259 4.694 12.226 9.709 t-stat 2.95 3.95 2.78 2.98 3.26 2.42 Constant 5.363 4.638 -43.573 -51.385 t-stat 2.47 2.I O -1.57 -1.82 Observations 270 270 270 270 225 225 R2 0.326 0.319 0.859 0.858 0.436 0.417 F-statistic (o-value) 0 0 0 0 0.03 0 .1 Note. The table reports the results of regressing the log o f re ttances migrant on the variables in the ~~ ~ first column. IV estimation uses as instrument for remittances passport costs, the ration of urban population and the population density ineach country. Source: Own calculations 60 VII. Conclusions This chapter has reviewed the overall migration patterns from Latin American and the Caribbean countries, especially interms o f the destinations chosen, age and education profiles o f migrants and their labor market performance in the US. Even though most analysts categorize migrants from the region as a homogenous group, we find that there are important differences across countries and sub-regions. Mexican and Central American migrants dominate overall migration flows in terms o f numbers. They are also much younger, less educated and tend to migrate primarily to the US. South American migrants are older, more educated and a large portion favors Europe due to ethnic and cultural linkages. While South American migrants are much more educated than the rest, brain drain i s a more serious problem for other countries, especially smaller ones that do not provide many labor market opportunities for skilled workers. On the other hand, a large portion o f the highly educated migrants from the region complete their education in the US. Given that they might not have received the same education if they have stayed at home, migration provides them and their home countries with higher economic potential. We also observe significant "brain waste" - many college educated migrants from the region endup inunskilledjobs that are not commensurate with their education levels. The patterns outlined in the chapter were aimed at providing an overview o f general migration trends, not definite answers, in order to contextualize the following chapters, which are preoccupied mainly with remittances and their impact on home countries. The questions that were posed need to be explored in fkrther detail if the goal i s to design policies and institutions that lead to win-win-win outcomes for the migrants themselves, and for destination and home countries. As for the results o f relating the characteristics o fmigrants to observed remittance flows, the main findings o f this chapter are the following: (i) Migration levels increase remittance/GDP levels and decrease remittances sent per migrant. The impact on remittances received per capita i s ambiguous. (ii)Increases in the overall education levels o f migrants reduce the level o f remittances sent. (iii)Economic growth in the recipient country increases remittance levels. (iv) Financial development variables have a significant effect on remittances: other things equal, countries with more developed financial sectors receive more remittances. (v) The share o f female migrants does not have a significant effect on remittance flows. And (vi) income per capita and size o f the economy influence remittance flows, with larger and poorer economies receiving more remittances. 61 Chapter 3 Remittances and Poverty Reduction* To the extent that remittances increase per capita income, ease credit constraints and compensate for negative shocks to recipient countries, they are likely to help reduce poverty and inequality, as well as boost domestic investment and economic growth. However, remittances also tend to be associated with a reduction in the income generating capacity of households with migrants. Moreover, they may reduce labor supply and lead to real exchange rate appreciations, all of which could potentially increase income volatility. Thequestion thus remains as to whether remittances do in fact improve social outcomes and bolster economic growth and investment. Are all the above potential impacts of remittances economically relevant? Do thepositive and negative effects of remittances compensate each other? What is the net effect of remittances onpoverty andgrowth? I.Introduction Have the large remittances flows received by Latin America contributed to reduce poverty in the region? In principle, given that in many cases remittances go to poor households and directly increase their level o f income, an unequivocal positive answer could be expected. Moreover, to the extent that remittances ease credit constraints and reduce risk and volatility, they could also promote higher levels o f investment in physical and human capital, and have dynamic effects on growth and poverty reduction. There are, however, several reasons that warrant circumspection in the assessment o f the development impact o f remittances. First, as shown in the previous chapter the position of migrants inthe distribution of income and education varies considerably across sender countries. As a result, the impact of remittances on poverty reduction should also be expected to vary by country and region. Second, according to the New Open Economy Macroeconomics, if remittances lead to increases in a country's domestic wealth, they could reduce labor supply, increase the demand for non-tradable goods and generate real exchange rate appreciations which, * This chapter i s based on the background papers for this report "Remittances and Development in Latin America" and "What is the Impact of International Remittances on Poverty and Inequality in Latin America?" by Pablo Acosta, Cesar Calderon, Pablo Fajnzylber, and Humberto Lopez, and the papers "Remittances, Growth, and Policy Complementarities," and "Remittances and Growth Volatility: an Investigation into the Business Cycle Properties of Workers' Remittances," by Cesar Calderon, Pablo Fajnzylber, and Humberto Lopez. 62 inturn, could hurt competitiveness and growth. Third, to the extent that remittances arejust one among other consequences o f migration, they cannot be treated as exogenous transfers. Rather, it i s reasonable to believe that inmany cases remittances and migration also entail potential losses o f income, associated with the migrants' absence from their families and communities. Moreover, depending on the demographic characteristics o f migrants, "brain drain" effects could have negative effects on productivity and welfare. In this chapter we explore both the static and dynamic effects of remittances on development using micro and macroeconomic data and techniques. In particular, we investigate the links between remittances, poverty and inequality usinghousehold surveys and cross-country regression analyses. We then analyze the macroeconomic effects o f remittances in more detail, considering not only their impact on economic growth, but also that on volatility and investment. 11.Remittances,InequalityandPoverty: a Microeconomic approach As argued in the seminal paper by Stark et al. (1986), the impact o f migration and international remittances on income inequality depends on the magnitude o f remittances in relation to income from other sources and upon the position o f remittance-receiving households in the distribution of income. In their perspective, when migration is incipient, its costs and related uncertainty are likely to be high, so that migrants can be expected to come from among the better off. As a result, at this point remittances would tend to have an un-equalizing effect. Over time, however, the uncertainty and costs involving migration are likely to diminishwith the spread o f information and contacts across a wider range o f households, which may lead to increases in migration among the worse off, with potentially favorable effects on poverty and income inequality. The simplest way o f analyzing whether remittances increase or decrease income inequality i s by comparing estimates o f the Gini coefficient obtained respectively with observed non-remittances household income and total income. As shown in table 3.1, with the exception o f Nicaragua and Peru, the other 9 Latin American countries for which data on remittances i s available exhibit higher Gini coefficients for non-remittances income, suggesting that if remittances were exogenously eliminated inequality would increase. Quantitatively, however, the estimated potential changes inthe Gini coefficient are small. One o f the factors that could explain the small inequality reducing effect o f remittances i s the fact that their own distribution is generally very unequal and that in several countries remittances tend to go to relatively well off households. This is illustrated in table 3.2, which reports the results o f a complementary simulation involving the effect on the Gini coefficient o f marginal increases in remittances income - as opposed to assuming that remittances are eliminated altogether. The calculations are based on Stark et al.'s (1986) decomposition o f the Gini coefficient for total income, as a function o f inequality inremittances and non-remittances income, the share o f each income source in total income, and the extent to which remittances and other sources o f 63 income are correlated with total income.' More specifically, the Gini coefficient o f total income GTcan be expressedas: K GT = RkGkSk k=l where k i s an index for the different components o f household income (Le., non- remittances and remittances income), Rk is the Gini correlation o f component k with total income, Gk i s the Gini coefficient o f component k, and sk i s the share o f income source k intotal income. In this context, Stark et al. (1986) show that the effect o f a small percentage change in one source o f income - e.g. remittances - on the Gini coefficient o f total income can be calculated as follows: Table 3.1. Income GiniCoefficientbefore and after remittances Difference in Gini Difference in Country Coefficient Gini beforelafler Diff' in Country CoefficientGini beforelafler DiK in % remitt. % remitt. Bolivia (2002) Honduras (2002) Non-RemittancesIncome 0.557 Non-RemittancesIncome 0.572 Total Income 0.555 -0,002 -0.4% Total Income 0.559 -0.013 -2.3% Ecuador (2004) Mexico(2002) Non-RemittancesIncome 0.505 Non-Remittances Income 0.491 Total Income 0.499 -0.006 -1.2% Total Income 0.481 -0.010 -2.1% El Salvador (2000) Nicaragua (2001) Non-RemittancesIncome 0.514 Non-RemittancesIncome 0.517 Total Income 0.486 -0.028 -5.4% Total Income 0.518 0.001 0.3% Guatemala(2000) Paraguay(2003) Non-RemittancesIncome 0.596 Non-RemittancesIncome 0.520 Total Income 0.586 -0.010 -1.8% Total Income 0.516 -0.004 4.7% Haiti (2001) Peru(2002) Non-RemittancesIncome 0.670 Non-RemittancesIncome 0.476 Total Income 0.669 -0.001 -0.1% Total Income 0.476 0.000 0.I% DominicanRepublic(2004) Non-RemittancesIncome 0.531 Total Income 0.520 -0.011 -2.0% Source: Acosta et al. (2006). As reported in table 3.2, the Gini coefficient for remittances i s above 0.9 in all 11 countries. In some o f them, however, the high inequality in the distribution o f remittances i s partially compensated by a low Gini correlation between remittances and total income, which implies that in those cases remittances are often received by low income households. This i s the case o f Mexico, where the Gini correlation o f remittances with total income i s 0.289, El Salvador (0.303), Paraguay (0.437) and Guatemala (0.541). Not surprisingly, we find that in these four countries a marginal increase in remittances has an equalizing albeit small effect on income This is the approach adopted by Taylor (1992), Barham and Boucher (1998), Adams (2004), Taylor et al. (2005) and Acosta et al. (2006). 64 inequality. Inparticular, it is estimated that a 10 percent increase inremittances would lead to an average reduction o f 0.1 percent inthe Gini o f total income inthose four countries. Incontrast, in the other seven countries, where the households that receive remittances are also located in the upper deciles o f the income distribution - e.g. Haiti, Nicaragua, the Dominican Republic and Peru - a marginal increase in remittances is associated with an increase in total income inequality: the Gini increases by 0.2 percent on average for each 10 percent increase in remittances income. Table 3.2. Effect on the Giniof a 10% Increase in Remittances Effect of Share in Gini Effect of Share in Gini 10% Country Total Gini correlation 10% increase Total Gini correlation Country , Income Coefficient with Total in remit.on Coefficient with Total Increase Income Gini Income remit.onin Gini Bolivia (2002) Honduras(2002) Non-Remittances Income 0.985 0.557 0.991 Non-Remittances 0.933 0.572 0.971 Remittances 0.015 0,990 0.728 0.04 Remittances 0.067 0.954 0.634 0.06 Total Income 1.000 0.555 1.000 Total Income 1.000 0.559 1.000 Ecuador (2004) Mexico (2002) Non-Remittances Income 0.964 0.505 0.980 Non-Remittances 0.980 0.491 0.987 Remittances 0.036 0.969 0.620 0.07 Remittances 0.020 0.976 0.289 -0.08 Total Income 1.000 0.499 1.000 Total Income 1.000 0.481 1.000 El Salvador (2000) Nicaragua(200I) Non-Remittances Income 0.941 0.514 0.972 Non-Remittances 0.976 0.517 0.992 Remittances 0.059 0.910 0.303 -0.26 Remittances 0.024 0.959 0.773 0.10 Total Income 1.000 0.486 1.000 Total Income 1.000 0.518 1.000 Guatemala(2000) Paraguay (2003) Non-Remittances Income 0.969 0.596 0.986 0.986 0.520 0.995 Remittances 0.031 0.958 0.541 -0.04 .Non-Remittances Remittances 0.014 0.985 0.437 -0.02 Total Income 1.000 0.586 1.000 Total Income 1.000 0.516 1.000 Haiti (2001) Peru (2002) Non-RemittancesIncome 0.794 0.670 0.926 Non-Remittances 0.989 0.476 0.992 Remittances 0.206 0.916 0.937 0.58 Remittances 0.01I 0.989 0.865 0.09 Total Income 1.000 0.669 I.000 Total Income 1.000 0.476 1.000 Dominican Republic (2004) Non-Remittances Income 0.914 0.531 0.950 Remittances 0.086 0.933 0.740 0.28 Total Income 1.000 0.520 I.000 * Effects of a I% increase in per capita incomefrom different sources on the Gini coefficient of Total Income. Source: Acosta et al. (2006). To complement the previous simulations on the effects o f remittances on inequality, we report in Table 3.3 poverty headcounts calculated before and after excluding remittances from the total income o f recipients. We employ two commonly used headcount poverty indicators, based on poverty lines o f US$1 and US$2 per person per da measured at PPP values, corresponding respectively to "extreme" and "moderate" poverty. `"As expected, with this very simple approach remittances lead to large reductions in poverty levels, especially in those countries where migrants tend to come from the lower deciles o f the income distribution - notably Mexico, El Salvador and the Dominican Republic, where extreme poverty i s estimatedto fall by more than 35 percent, and moderate poverty by an average o f 19 percent. The reductions inpoverty headcounts that result from taking remittances income into consideration are smaller loThe US$ 1 and US$2 a day are values measured in 1989 international prices and adjusted to local currency using purchasing power parities (PPP) to take into account local prices. Partly because of their simplicity, these are the standard indicators used for international poverty comparisons. For instance, they are periodically presented in the World Development Indicators (World Bank). 65 whenusinglocally defined country-specific national poverty lines (appendix table A3.l)." Thus, while Mexico, El Salvador and the Dominican Republic are still the countries where largest reductions are obtained, the corresponding average changes are o f 15 percent for extreme poverty and 8 percent for moderate poverty. One concern with the above inequality decompositions and poverty simulations is that they implicitly make the unrealistic assumption that remittances can be treated as exogenous transfers by migrants. However, it i s reasonable to believe that in many cases migration also entails potential losses o f income, associated with the migrants' absence from their families and communities. In other words, remittances are not exogenous transfers but rather they substitute for the home earnings that migrants would have had if they had not decided to leave their countries to work abroad. Table 3.3. PovertyHeadcountsbeforeand after remittances Country Less than US$ Diff in Less than US$ Diff. in 1 aday(PPP) % 2 aday(PPP) % Country Less than US$ Diff. in Less than US$ Diff in 1 aday(PPP) % 2 aday(PPP) % Bolivia (2002) Honduras (2002) Non-Remittances Income 18.400 35.261 Non-Remittances Income 18.681 34.772 Total Income 17.764 34.674 Total Income 16.155 31.731 Diff. beforeiafler remitt. -0.6 -3.5% -0.6 -1.7% Dif.befordafler remitt. -2.5 -13.5% -3.0 -8.7% Ecuador (2004) Mexico (2002) Non-Remittances Income 12.531 28.849 Non-Reminances Income 5.268 15.029 Total Income 11.198 27.147 Total Income 3.165 12.695 Dif. beforeiafler remitt. -1.3 -10.6% -1.7 -5.9% Dif.befordafler remitt. -2.1 -39.9% -2.3 -15.5% El Salvador (2000) Nicaragua (2001) Non-Remittances Income 12.116 23.743 Non-Remittances Income 8.763 23.323 Total Income 7.700 18.607 Total Income 8.260 22.552 Dif. beforeiafler remitt. -4.4 -36.4% -5.I -21.6% Dif.befordafler remitt. -0.5 -5.7% -0.8 -3.3% Guatemala (2000) Paraguay (2003) Non-Remittances Income 24.413 41.715 Non-Remittances Income 6.839 16.188 Total Income 21.578 39.087 Total Income 6.057 15.333 Dif. beforeiafler remitt. -2.8 -11.6% -2.6 -6.3% DIE befordafler remitt. -0.8 -11.4% -0.9 -5.3% Haiti (2001) Peru (2002) Non-Remittances Income 60.229 76.343 Non-Remittances Income 4.193 15.644 Total Income 53.425 71.414 Total Income 4.185 15.539 Dif. beforeiafler remitt. -6.6 -11.3% -4.9 -6.5% Dif.befordafler remitt. 0.0 -0.2% -0.1 -0.7% Dominican Republic (2004) Non-Remittances Income 7.219 16.072 Total Income 4.688 12.836 Dif.beforeiafler remitt. -2.5 -35.1% -3.2 -20.1% Source: Acosta et al. (2006). . Inorder to consider these effects one needs to estimate the value that household income would have taken had migrants stayed in their households. To impute per capita household income for migrant families in that counterfactual scenario we predict per capita income levels for households with remittances on the basis o f a reduced-form specification for the determinants of income among household without remittances. The estimated model is the following: I'We emphasize the calculations based on common international poverty lines (US$1 and US$2 PPP) because they are better suited for international comparisons. However, the appendix to this chapter reports similar calculations based on locally defined poverty lines, which may be of interest for country-specific analysis. It is worth noting that poverty headcounts are in all cases much higher when using local poverty lines - which are higher than their international counterparts. 66 where Yj represents per capita non-remittances income, xi i s a vector o f household characteristics (demographic and location covariates), Hj is a set o f characteristics o f the household head, and pj i s unobserved heterogeneity in income generation. The procedure consists on estimating equation (3) for the sub sample o f households that do not receive remittances, making a correction for possible selection biases. The estimated coefficients allow predicting the counterfactual non- migration income for remittances-recipient households. As explained below, in order to avoid the reduction in income variability for migrant families due to the used o f predicted values, we follow Barham and Boucher (1998) inadding to the predicted values o f income a simulated error component. Several issues need to be discussed. First, in the absence o f information on migrant characteristics - which i s the case in nine out o f our eleven countries - it i s necessary to make some basic assumptions about the number and the demographics o f migrants. Inthis respect, we follow Rodriguez (1998) - and most o f the evidence on international migration - in assuming that on average remittances are sent by a single adult male family member. In addition, we assume that the migrant has the average years o f education o f other adults in the household. To test the robustness o f our results to the above assumptions we compare the corresponding results with those obtained when detailed data on migrants' demographics is employed in lieu o f the above assumptions-this is possible only for Nicaragua and Haiti. Second, OLS estimates o f equation (3) would be inconsistent if pi i s not independently identically distributed (i.i.d.). Inother words, ifmigrants are not randomly selected from the pool o f households, estimates o f equation (3) based on the sample o f households without migrants could suffer from selection bias. To control for that possibility, we add variables that approximate the "propensity to migrate", in the context o f the two-step estimation framework proposed by Heckman (1979). l2 With the assumption stated before concerning migrant characteristics, the estimated coefficients for (3) are used to impute the counterfactual non- remittances per capita household income for recipient families. With this variable one can proceed to calculate the levels o f poverty and inequality that would have prevailed had migration and remittances not taken place. It must be noted, however, that as mentioned by Rodriguez (1998) the variance of the counterfactual income predicted on the basis o f observable household characteristics i s artificially small, because it ignores unobserved determinants o f income. To deal with this problem, we follow Barham and Boucher (1998) and add to the predicted household income a random error component drawn from a distribution with the same properties (mean, variance) o f the actual estimated errors. We pursue this approach and obtain 1,000 different estimates o f the imputedcounterfactual non-remittances income for families with migrants, and the same number o f estimates for the poverty and inequality levels that would have prevailed in the above l2 A similar approach is followed inBarham and Boucher (1998) and Adams (2006). The exclusion restrictions that we employ for the non-remittances selection equation are an index o f household assets, the percentage o f households that receive remittances in the respective county or province o f residence (a proxy for the presence o f migrant networks), and their interaction. When information on household assets is missing (Bolivia, Ecuador and Honduras), the network variable is interacted with the number o f adult males, which ensures variability at the household level. Additional methodological details can be found inAcosta et al. (2006). 67 counterfactual scenario. This allows us to report not only point estimates for those variables but also 95% confident intervals, based on the 25th and 975th estimates o f the variables - after sorting them ascending order. The resultingestimates for the impact o f remittances on inequality and poverty are reported intables 3.4 and 3.5 respectively. Table3.4. Income GiniCoefficientinCounterfactualScenario of No-Migration Gini Differencein Gini Differencein Country "E Coefficient Gini beforeiafler remitt, in Country Coefficient Gini beforeiafler Diff. in % remitt. Bolivia (2002) Honduras(2002) Non-RemittancesIncome 0.556 Non-RemittancesIncome 0.565 95% Confidence lnterval(0.553; 0.561) 95% Confidence lnferval(0.564; 0.567) Total Income 0.555 -0.001 -0.3% Total Income 0.559 -0.006 -1.1% Ecuador(2004) Mexico (2002) Non-RemittancesIncome 0.501 Non-RemittancesIncome 0.477 95% Confidence Interval(0.500;0.503) 95% Confidence lnterval(0.477; 0.478) Total Income 0.499 -0.002 -0.5% Total Income 0.481 0.004 0.7% El Salvador (2000) Nicaragua(2001) Non-RemittancesIncome 0.497 Non-RemittancesIncome 0.528 95% Confidence lnferval(0.494; 0,501) 95% Confidencelnferval (0.519; 0.539) Total Income 0.486 -0.011 -2.1% Total Income 0.518 -0.010 -1.8% Guatemala(2000) Paraguay(2003) Non-RemittancesIncome 0.603 Non-RemittancesIncome 0.515 95% Confidence lnferval(0.596; 0.615) 95% Confidencelnferval (0.514; 0.517) Total Income 0.586 -0.017 -2.9% Total Income 0.516 0.001 0.2% Haiti (2001) Peru(2002) Non-RemittancesIncome 0.725 Non-RemittancesIncome 0.478 95% Confidence Interval(0.703; 0.756) 95% Confidence lnterval(0.476; 0.481) Total Income 0.669 -0.056 -7.7% Total Income 0.476 -0.002 -0.3% DominicanRepublic (2004) Non-RemittancesIncome 0.519 95% Confidence lnterval(0.514; 0.525) Total Income 0.520 0.001 0.3% Source: Acosta et al. (2006). Our results suggest that the Gini coefficients that would have prevailed inthe absence o f migration would have been generally higher, with the largest difference obtained for Haiti (7.7%), followed by Guatemala (2.9%), El Salvador (2. I%), Nicaragua (1.8%) and Honduras (1.1%). The negative effects o f remittances on inequality are much smaller inthe other countries and even a positive effect i s obtained for the cases o f Mexico and the Dominican Republic. The results for the countries for which remittances' impact i s minor or even favoring inequality are consistent with the findings o f previous studies that have made attempts to calculate counterfactual pre-remittances income for families with migrants.l3Overall, the estimated inequality reducing effects o f remittances are found to be relatively small - 2.7% on average, when significant - albeit they tend to be comparatively larger in countries where remittances represent a higher share o f income. l3Rodriguez (1998), for instance, finds that remittances increase inequality in the Philippines, and the effect rises from 1.27 percent to 7.90 percent when using imputed income instead o f reported non-remittances income. Similarly Barham and Boucher (1998), for the case o f Bluefields (Nicaragua), find that the Gini for household income falls from 0.47 to 0.43 when using reported figures, but inequality actually rises from 0.38 to 0.43 after correcting the pre-remittances distribution using imputed income for migrant families. 68 Table 3.5. Poverty Headcounts in Counterfactual Scenario of No-Migration Country US$ 1 aday Diff. US$2 aday Diff. US$ 1 aday Diff. US$2 aday Diff. (PPP) i n % (PPP) in% country (PPP) i n % (PPP) in % Bolivia (2002) Honduras (2002) Non-RemittancesIncome 17.999 35.052 Non-RemittancesIncome 16.715 33.155 95% ConfidenceInterval (17.842; 18.184) (34.824; 35.279) 95% Confidence Interval (16.608; 16.820) (32.993; 33.307) Total Income 17.764 34.674 Total Income 16.155 31.731 Diff. beforeiafter remitt. -0.2 -1.3% -0.4 -1.1% Dit beforeiafler remitt. -0.6 -3.4% -1.4 4.3% Ecuador(2004) Mexico (2002) Non-RemittancesIncome I1.665 28.082 Non-RemittancesIncome 3.079 12.603 95%ConfidenceInterval (11.594; 11.741) (27.960; 28.221) 95% ConfidenceInterval (3.019; 3,145) (12.480; 12.731) Total Income 11.198 27.147 Total Income 3.165 12.695 Dif. beforeiafter remitt. -0.5 -4.0% -0.9 -3.3% Dif.beforeiafler remitt. 0.1 2.8% 0.1 0.7% El Salvador (2000) Nicaragua(2001) Non-RemittancesIncome 8.215 20.055 Non-RemittancesIncome 8.226 22.848 95% ConfidenceInterval (8.077; 8.375) (19.824; 20.311) 95% ConfidenceInterval (8.012; 8.528) (22.427; 23.345) Total Income 7.700 18.607 Total Income 8.260 22.552 Dif. beforeiafter remitt. -0.5 -6.3% -1.4 -7.2% Dif. beforeiafler remitt. 0.0 0.4% -0.3 -1.3% Guatemala (2000) Paraguay(2003) Non-RemittancesIncome 23.630 41.379 Non-RemittancesIncome 6.066 15.373 95% Confidence Interval (23.335; 23.931) (41.055; 41.710) 95% Confidence Interval (5.999; 6.145) (15.256; 15.521) Total Income 21.578 39.087 Total Income 6.057 15.333 Dif. beforeiafter remitt -2.1 -8.7% -2.3 -5.5% Dif. beforeiafter remitt. 0.0 -0.1% 0.0 -0.3% Haiti (2001) Peru(2002) Non-RemittancesIncome 57.541 74.376 Non-RemittancesIncome 4.186 15.555 95% ConfidenceInterval (56.929; 58.138) (73.793; 74.992) 95% ConfidenceInterval (4.186; 4.192) (15.533; 15.888) Total Income 53.425 71.414 Total Income 4.185 15.539 Dif. beforeiafler remitt. -4.1 -7.2% -3.0 -4.0% Dit beforeiafler remitt. 0.0 0.0% 0.0 -0.1% Dominican Republic (2004) Non-RemittancesIncome 4.364 13.008 95% ConfidenceInterval (4.247; 4.488) (12.777; 13.270) Total Income 4.688 12.836 Dif.beforeiafter remitt. 0.3 7.4% -0.2 -1.3% Source: Acosta et al. (2006). As for the impact o f migration and remittances on poverty, the results intable 3.5 suggest that the failure to correct for the reduction in income associated with the absence o f migrants from their households may lead to grossly overestimating the poverty reducing effect o f remittances. In particular, we find that remittances reduce poverty headcounts in only 6 out o f the 11 countries for which data is available -the exceptions being the Dominican Republic, Mexico, Nicaragua, Paraguay and Peru - and they reduce poverty gaps in only 3 cases - Ecuador, Guatemala and Haiti.I4In two cases, the Dominican Republic and Nicaragua, we even find that remittances are linked to small increases in extreme poverty - respectively o f 7.4 and 0.4 percent. Thus, for very poor households inthose countries the income lost due to the absence o f migrants from their households i s less than compensated by the money they send home, possibly because they also under-perform inthejob market indestination countries. Considering all eleven countries and assuming that the remittances share in GDP i s as given by BOP statistics, the average estimated impact o f remittances on poverty headcounts i s such that a 1 percentage point increase in the remittances to GDP ratio reduces moderate and extreme poverty by respectively .37 and 0.29 percent. Interestingly, the countries for which we find the largest inequality and poverty reducing effects are not necessarily those where remittances recipients tend to come from lower income groups. Consider, for instance, the cases o f El Salvador, Guatemala, Mexico and Paraguay, l4The results on poverty gaps are available inAcosta et al. (2006). 69 where remittances recipients tend to be relatively less educated than the general population - there i s "negative" selection into migration - and remittances are more progressively distributed than total income. Only in two o f them (El Salvador and Guatemala) do we find that remittances are associated with significant reductions in both inequality and poverty. Moreover, in the other two countries where remittances do appear to reduce poverty and inequality - Haiti and Honduras - remittances recipients are more likely to be found among highly educated individuals, and remittances income i s distributed more unequally than total income. Ifanything, what the four countries where remittances have the largest effects on poverty and inequality have incommon is that they are amongthose inwhich remittances are highestwith respect to GDP. One complementary approach proposed by Schiff (2006) i s to estimate the impact o f remittances on poverty focusing exclusively on the population o f households which receive remittances. The point is that while the impact o f remittances on national poverty levels may be limited, the effect on the poverty status o f the families with migrants couldbe much larger. Table 3.6 reports the results o f such analysis, using as a reference group only remittance recipient households rather than all the survey sample - even though the latter continues to be used, as described above, for estimating the counterfactual income o f recipients. The first interesting finding from this analysis is that the initial poverty rates -before migration- among households with migrants are lower than those found in the general population. In Peru and Nicaragua, for instance, moderate poverty levels in the no-migration counterfactual had been estimated to be close to 15 and 22 percent respectively at the national level, but they are close to 1 and 12 percent respectively among recipients. Other countries where migrants are considerably less likely to be poor than the average household in the country include Haiti, the Dominican Republic and Honduras. Second, as expected, we find that the effects o f remittances on poverty are much larger than in the case when non-recipient households were included in the calculations. The largest absolute reductions in extreme poverty are found in Haiti and Guatemala, with rates that are respectively 15 and 10.7 percentage points below those in the scenario o f no migration. Similarly, both in the latter countries and in Bolivia, Honduras and Ecuador, the absolute reductions in moderate poverty are between 10 and 17 percentage points. Incontrast, remittances are found to have the effect o f slightly increasing poverty levels - even among migrants - inboth Mexico and the Dominican Republic. 70 Table 3.6. PovertyHeadcountsamongRecipientsHouseholdsin CounterfactualScenarioof No-Migration Country US$1a day Diff. in USS2 a day Diff. in US$Ia day Diff, in % U S I 2 a day Diff. in (PPP) % (PPP) % Country (PPP) (PPP) % Bolivia (2002) Honduras (2002) Non-Remittances Income 15.103 32.610 Non-Remittances Income 7.694 22.785 95% Confidence Interval (10.297; 20.118) (26.619; 39.209) 95% Confidence Interval (6.703; 8.699) (21.260; 24.318) Total Income 8.570 22.181 Total Income 2.769 10.124 Diff. beforelafter remitt. -6.5 -43.3% -10.4 -32.0% Dif.beforeiafter remitt. -4.9 -64.0% -12.7 -55.6% Ecuador(2004) Mexico (2002) Non-Remittances Income 7.357 22.977 Non-Remittances Income 1.719 12.967 95% Confidence Interval (6.060; 8.826) (20.666; 25.539) 95% Confidence Interval (0.835; 2.810) (10.979; 15.185) Total Income 2.953 10.934 Total Income 3.632 15.942 Dif.beforeiafter remitt. -4.4 -59.9% -12.0 -52.4% Dif.beforeiafter remitt. 1.9 111.3% 3.0 22.9% El Salvador(2000) Nicaragua (2001) Non-Remittances Income 5.255 17.870 Non-Remittances Income 3.348 12.309 95% Confidence Interval (4.555; 6.074) (16.658; 19.327) 95% Confidence Interval (2.049; 5.103) (9.751; 15.363) Total Income 2.774 10.757 Total Income 3.730 10.815 D i t beforeiafter remitt. -2.5 -47.2% -7.1 -39.8% Dif. beforeiafter remitt. 0.4 11.4% -1.5 -12.1% Guatemala(2000) Paraguay(2003) Non-Remittances Income 20.995 38.085 Non-Remittances Income 4.052 14.799 95% Confidence Interval (17.362; 24.527) (34.098; 42.246) 95% Confidence Interval (2.098; 6.018) (11.594; 18.158) Total Income 10.307 21.341 Total Income 3.979 14.127 Dif. beforeiafter remitt. -10.7 -50.9% -16.7 -44.0% Dit beforeiafter remilt. -0.1 -1.8% -0.7 -4.5% Haiti (2001) Peru (2002) Non-Remittances Income 47.225 63.877 Non-Remittances Income 0.051 1.269 95% Confidence Interval (45.053; 49.452) (61.686; 66.060) 95% Confidence Interval (0.018; 0.692) (0.244; 2.530) Total Income 32.238 53.030 Total Income 0.000 0.804 D i t beforeiafter remin. -15.0 -31.7% -10.8 -17.0% Dif.before/after remitt. -0.1 -100.0% -0.5 -36.6% Dominican Republic (2004) Non-Remittances Income 1.715 7.746 95% Confidence Interval (1.195; 2.265) (6.738; 8.864) Total Income 3.266 7.349 D i t beforeiafter remitt. 1.6 90.4% -0.4 -5.1% Source: Acosta et al. (2006). 111.Inequality and PovertyEffects: Macroeconomic Evidence An alternative way o f estimating the impact o f remittances on poverty is by means o f a cross-country regression analysis. This i s the approach taken by Adams and Page (2005), and the IMF's World Economic Outlook (2005). Both o f them find that countries that receive remittances have lower poverty levels. In particular, Adams and Page find that a 10 percent increase inper capita remittances would lead to a 3.5 percent decline inthe share o f people living in poverty inthe corresponding country, and the World Economic Outlook concludes that a 2.5 percentage point increase in the remittances to GDP ratio i s associated with a 0.5 percentage point decrease in poverty. However, neither study allows the effect o f remittances to vary by country or region. Moreover, as noted inWorld Economic Outlook (2005), because both studies control separately for the effect o f per capita income and income inequality, they miss the effects o f remittances that operate through changes in those variables. Thus, they are both likely to under-estimatethe poverty effects o f remittances. Table 3.7 reports the results o f estimating the impact o f remittances on poverty using Adams and Page's (2005) approach, but allowing the effect to be different in the L A C region than inthe rest o f the world. The estimated equation is o f the following form. 71 where P i s a poverty measure o f the FGT family (i.e. headcount, poverty gap, or squared poverty gap), Y is a measure o f per capita income (either GDP or household mean income), G i s the gini coefficient, R stands for remittances measured on a per capita basis, lac i s a dummy variable that takes a value o f 1 if country ii s in the Latin American region and 0 otherwise u i s an error term, and i and t identify countries and time periods, respectively. With this set up, a priori one could expect the coefficient on income to be negative and that on the Gini to be positive The coefficients on remittances and its interactive with the L A C dummy would be our main parameters o f interest. A negative value for the former would indicate that remittances lead to lower poverty at the global level, while the sign o f (p3+p4)would determine the direction o f the poverty impact o f remittances inLatinAmerica. As seen in table 3.7, the signs and magnitudes o f the coefficients for inequality and per capita income are in line with those reported by other studies: averaging across all specifications -withdifferentsetsofinstrumentalvariables-a 1percentincrease inpercapitaincomelevels would lead to a poverty decline o f about 1.1 percent, whereas a 1 percent increase in the Gini coefficient would lead to an increase in poverty o f about 1.7 percent. As.for the impact o f remittances on poverty, table 3.7 suggests that the effect tends to be negative at the global level - although significant inonly two o f the six specifications -but close to zero inthe specific case o f the L A C region. There are a number o f potential reasons that may explain these somewhat disappointing results. One i s the choice o f variable used to proxy for remittances flows. For example, the World Economic Outlook (2005) relies on a similar framework but rather than relating poverty to remittances on a per capita basis it uses remittances as a percentage o f GDP. However, replicating the above exercise with this alternative remittances variable does not yield additional support to the idea that remittances significantly lower poverty, either in Latin America or elsewhere. A second explanation could be related to the specification o f the model per se. As noted above, in principle one could expect remittances to affect poverty by changing either average income or inequality. Thus, it is difficult to interpret the parameter o f remittances in a regression model that controls for the two mainpotential drivers o f poverty reduction. To deal with this problem, we next assume that per capita income levels follow a lognormal distribution, so that the effect on poverty o f a change in remittances will depend on: (i)the impact that remittances have on growth; (ii)how growth is translated into poverty reduction; (iii)the simultaneous impact that remittances have on inequality (as measured by the Gini coefficient), and finally (iv) how inequality changes are translated into poverty reduction. This can be expressed as follows. d P R - d Y R d P Y +---- d G R d P G d R P d R Y d Y P dR G d G P Under fairly general conditions it is possible to derive theoretical values for the parameters in channels (ii)and (iv), which correspond to the growth elasticity o f poverty y and the inequality elasticity o fpoverty 4, expectedto be respectively negative andpositive: 72 y =--o. aY P i=,,p With these elements in mind, a change inpoverty due to a change inremittances (which we will denote by R) can be expressed as: --_-- d P R - d Y R x y+--xb. dG R aR P aR Y dR G (7) Table 3.7. The impact of per capita remittanceson poverty (1) (2) (3) (4) (5) (6) OLS I V IV I V IV IV Y -1.22 -0.78 -1.06 -1.21 -0.59 -1.62 t-stat -8.43 -5.26 -1.01 -4.51 -0.62 -4.98 Gini 2.51 0.56 1.87 2.10 0.69 2.73 t-stat 4.57 2.17 0.91 2.77 0.79 2.70 Remitt.per capita -0.20 -1.11 -1.01 -0.75 -2.05 0.50 t-stat -3.08 -4.77 -0.48 -1.53 -0.97 0.90 Remitt.per capitax lac 0.23 1.69 0.54 0.46 0.85 0.25 t-stat 2.68 3.37 1.02 3.16 1.30 1.12 Instrument Instrument Instrument Instrument Instrument Dist. Educ. Gov. Stab. Growth-Dist Growth-Mig. Ho: No impactin LAC (p-val) 0.76 0.21 0.78 0.52 0.45 0.09 # Observations 100 100 90 90 100 100 The dependent variable is the headcount poverty rate (using a $1 PPP poverty line, in logs). Explanatory variables: per capita income (in logs), the Gini (in logs), lagged remittances per capita, and lagged remittancesper capita interactedwith a dummy for LatinAmerica." Source: Acosta et al. (2006). Measuring the effect o f inequality on poverty is slightly more complex than estimating the impact o f growth. Indeed, inequality can change in infinite manners. Although intuitively progressive distributional change i s likely to reduce poverty, this result cannot be generalized without additional assumptions. To make the problem o f the impact o f inequality changes on poverty tractable, we follow one possibility i s to assume that income follows a particular parametric distribution. For example, Lopez and Serven (2006) compare the theoretical quintile shares according to a log-normal distribution with their empirical counterparts using data from 794 household surveys and argue that the log-normal approximation fits the empirical data extremely well. Hence, the empirical problem becomes one o f estimating the impact o f remittances on income growth and inequality. Inorder to estimate the links between remittances and growth inthe data, our empirical strategy i s based on the addition o f a measure o f remittances to an otherwise standard empirical growth regression: "IVheadingsindicatethattheequationhasbeenestimatedusinginstrumentalvariables.Instrumentsconsideredin the differentspecifications are Distance to the main sendingarea (Dist.), Secondary Education(Educ), Government Stability (Gov. Stab), Growth of sender countries weighted by distance (Growth-Dist) and Growth of sender countriesweightedby the stock ofmigrantsinreceivingcountry. 73 where y i s the log o f per capita income, Y i s a measure o f remittances, x represents a set o f control variables other than lagged income, which we shall discuss shortly, i s a country- specific effect, and'" i s an i.i.d. error term. According to (8), growth depends on initial income, remittances and current andor lagged values o f the control variables, and the impact o f remittances on growth for LatinAmerica would be givenby Pl+P2. As for the impact o f remittances on inequality, we follow a similar strategy and estimate the following specification: where g i s the (logged) Gini coefficient, and pi andEif are a country-specific effect, and an i.i.d. error term. Analogous to equation (8), the impact o f remittances on the changes ininequality for LatinAmerica would be givenby alfa2. Table 3.8 reports the results o f this exercise.I6 Overall, it appears that remittances tend to be good for growth (Le. a higher remittances to GDP ratio tends to be associated with higher growth), both at the global level and in Latin America. In fact, we cannot reject the hypothesis that the impact o f remittances on growth i s the same in Latin America and in the rest o f the world. On the other hand, remittances seem to lead to higher income inequality at the global level but to either reduce or leave inequality unchanged in Latin American (something to be welcomed given the high inequality levels of the region). These basic messages are robust to the use o f different instrumental variables to correct for the possible endogeneity o f remittances - e.g. the possibility o f remittances also being affected by growth or inequality. These results suggest that remittances tend to reduce poverty in Latin America. In terms o f the magnitude o f the estimated effects, table 3.9 reports the elasticity o f poverty with respect to the remittances to GDP ratio obtained for the average Latin American country, for each o f the specifications used. l6With regard to the control variables, rather than adding to the already huge variety of growth models, in this chapter we opt for considering a set o f controls that has already been used in a number of empirical growth studies. This is the one used by Perotti (1996), Forbes (2000), Banerjee and Duflo (2003), and Knowles (2005), and it includes the average years o f secondary education of the male population, the average years o f secondary education of the female population, and a measure of market distortions: the price of investment goods relative to the one in the US.All these variables are measuredinlevels at the beginning of the period. 74 Table 3.8. The impactof remittanceson growth and changesin inequality (1) (2) (3) (4) (5) (6) (7) Distance Distance Migration Migration Distance Migration Level Growth Level Growth Level, Growth Level, Growth Impactof Remittanceson Growth Remittances(t-1) 0.005 ** 0.012 0.012 0,008 * 0.013 * 0.011 * 0.009 * (1.75) (3.48) (4.57) (3.45) (3.16) (4.46) (4.64) Remittances(t-I) x LAC 0,000 0,002 -0.003 0.002 -0.008 0.000 0.003 (-0.05) (0.43) (466) (0.49) (-1.39) (-0.22) (1.10) Remittancesin LAC 0.005 * 0.014 * 0.009 * 0,010 * 0.006 ** 0.011 * 0.012 * p-val 0 0 ,005 ,001 .07 0 0 Impactof Remittanceson Inequality Remittances(t-I) 0.007 * 0.010 * 0.011 * 0.008 * 0,009 * 0,005 * 0.008 * (5.43) (5.61) (9.92) (6,59) (6.51) (5.99) (6.IO) Remittances(t-1) x LAC -0.014 * -0.012 * -0,011 * -0,011 * -0,012 * -0.011 * -0.013 * (-10.80) (-3.03) (-9.41) (-5.61) (-11.01) (-13.49) (-7.73) RemittancesinLAC -0.007 -0.001 0.000 -0,003 -0.003 .0.006 -0.006 p-val 0.000 0.600 0.950 0.030 0.140 0.000 0.000 Notes: The dependent variables are either logged changes in per capita income (upper panel) or the Gini (lower panel). Other explanatory variables: either lagged p/c income or the lagged Gini (in logs), average years of secondary education of the female and male population, a measure of market distortion (givenby the price of investment goods), remittances (as a percentageof GDP in logs), andremittances interactedwith a regionaldummy for LatinAmerica. All regressions include a constant. Remittances in LAC presents the value correspondingto the impact of Remittances on LatinAmericangrowth or inequalitychanges.Source: Acosta et al. (2006)." As seen in table 3.9, there is substantial heterogeneity in the effects o f remittances on poverty depending upon the country's initial conditions, as given by the ratio o f per capita income to the poverty line, and the Gini coefficient. Inthe Latin American context, however, this range can be substantially narrowed as most countries in the region have Gini coefficients o f around 0.5.'* Thus, considering that in the average Latin American country remittances are 4.9 percent to GDP, a one percentage point increase in the remittances to GDP ratio i s estimated to lead to reductions in poverty that vary between 0.08 percent for poorer countries to 1.12 percent for richer countries, with an average estimated reduction o f 0.37 percent, which i s hlly in line with our micro-econometric results for moderate poverty.Ig On the whole, contrary to the analysis based on Adams and Page's (2005) approach, when we empirically model the transmission channels from remittances to poverty we find that (i) remittances lead to lower poverty; and (ii) this effect is more marked inLatinAmerica than elsewhere. The regressions are performedusingthe systemGMh4 estimator. The first columnuses only internalinstruments.The second to seventh columns use lagged internal instruments for the Gini, average years o f secondary education of the female and male population,andthe measure of market distortion plus an externalinstrumentwhich equals the per capita income levelinthe main migrant host countries weighted distance (Distance) or number of migrants (Migration). Level indicates that the external instrument is the value of the host countries per capita GDP series. Growth indicates that the external instrument is the growth rate of the host countries per capita GDP series.Level, growth combines the levelsand growthinstruments. ''See for example Figure 4.7 inPerry et al. (2006). l 9This is the estimatedpoverty effect o f one percentage point increase in the remittances to GDP ratio in a country where that ratio is 4.9 percent and the Gini Coefficient is 0.5, using the average poverty elasticity from all alternative choices o f instrumental variables reported in table 8. 75 Table 3.9. Povertyelasticityof remittancesinLatinAmerica GMM GMM,Distance, Level Gini Coefficient Gini Coefficient V / Z 0.30 0.40 0.50 0.60 v/z 0.30 0.40 0.50 0.60 6 -1.17 -0.68 -0.46 -0.33 6 -0.94 -0.51 -0.32 -0.20 3 -0.56 -0.34 -0.24 -0.18 3 -0.59 -0.33 -0.20 -0.14 2 -0.32 -0.20 -0.15 -0.12 2 -0.40 -0.23 -0.15 -0.10 1.5 -0.19 -0.13 -0.10 -0.08 1.5 -0.29 -0.18 -0.12 -0.08 1 -0.07 -0.06 -0.05 -0.04 1 -0.16 -0.1 1 -0.08 -0.06 GMM,Distance, Growth GMM,Migration, Level Gini Coefficient Gini Coefficient v/z 0.30 0.40 0.50 0.60 v/z 0.30 0.40 0.50 0.60 6 -0.53 -0.29 -0.17 -0.11 6 -0.92 -0.52 -0.33 -0.23 3 -0.35 -0.19 -0.12 -0.08 3 -0.52 -0.30 -0.19 -0.14 2 -0.25 -0.14 -0.09 -0.06 2 -0.34 -0.20 -0.13 -0.10 1.5 -0.18 -0.11 -0.07 -0.05 1.5 -0.23 -0.14 -0.10 -0.07 1 -0.10 -0.07 -0.05 -0.03 1 -0.12 -0.08 -0.06 -0.05 GMM,Migration, Growth GMM,Distance, Level & Growth Gini Coefficient Gini Coefficient VI2 0.30 0.40 0.50 0.60 v/z 0.30 0.40 0.50 0.60 6 -0.7 1 -0.40 -0.26 -0.18 6 -1.41 -0.80 -0.52 -0.37 3 -0.37 -0.22 -0.15 -0.11 3 -0.74 -0.44 -0.29 -0.21 2 -0.23 -0.14 -0.10 -0.07 2 -0.46 -0.28 -0.19 -0.14 1.5 -0.15 -0.10 -0.07 -0.05 1.5 -0.30 -0.19 -0.14 -0.1 1 1 -0.07 -0.05 -0.04 -0.03 1 -0.14 -0.10 -0.08 -0.06 GMM, Migration, Level & Growth Gini Coefficient V / Z 0.30 0.40 0.50 0.60 6 -1.49 -0.84 -0.55 -0.38 3 -0.80 -0.47 -0.3 1 -0.22 2 -0.49 -0.30 -0.2 1 -0.15 1.5 -0.33 -0.2 1 -0.15 -0.11 1 -0.15 -0.11 -0.09 -0.07 Notes: The table reports the estimated poverty elasticities o f remittances/GDP for different values o f the Gini coefficient and o f per capita income to the poverty line v/z. I t assumes an initial level of remittances of 10 percent of GDP. Source: Acosta et al. (2006). IV. Remittances,GrowthandInvestment While the above reported results suggest that remittances are positively and significantly related to growth, they are based on a very parsimonious specification. Thus, the question remains as to the extent to which the link between remittances and growth i s robust to the inclusion o f additional control variables. This question i s particularly relevant given that previous cross national empirical evidence on the impact o f remittances on growth has not been conclusive. Faini (2002), for instance, has found that remittances have a positive impact on economic growth, thus confirming the predictions o f the new economic migration literature (Stark and Lucas, 1988; Taylor, ,1994) according to which remittances promote investment and entrepreneurial activities. However, a broader study conducted for the IMF's World Economic Outlook (IMF, 2005) finds that there is no significant relationship between remittances and growth in GDP per capita.20In addition, Chami, Fullenkamp and Jahjah (2005) find a negative association between the growth rate o f immigrant remittances and growth in GDP per capita for a sample o f 113 countries over the period 1970-98. Similarly, Giuliano and Ruiz-Arranz (2005) find that the impact o f remittances on growth is not significant when remittances are simply added as an additional explanatory variable in a growth regression. They claim that it i s likely that the impact o f remittances on growth may depend on some structural features o f the 2oThis study uses a sample of 101developingcountries over the period 1970-2003. 76 economy, and find evidence that remittances enhance growth in countries with shallower financial markets. There have also been a few studies dealing with the specific case o f the Latin America and Caribbean region. Solimano (2003) evaluated the impact o f remittances on per capita growth for two Andean countries, Colombia and Ecuador, during the period 1987-2002. He found that lagged remittances have a positive and significant impact on growth in Colombia, but the relationship i s not significant for Ecuador. Mundaca (2005) evaluated the effects o f workers' remittances on growth for countries in Central America, Mexico and the Dominican Republic duringthe period 1970-2003. She found that remittances promote growth inthe region and that the impact is even stronger inthe presence o f anactive banking sector inthe credit market. Inthe same vein, using a sample o f Caribbean countries, Mishra (2005) found that private investment rises by 0.6 percentage point o f GDP in response to a 1 percentage point o f GDP increase in remittance inflows. The fact that the results on the growth-remittances link have been somewhat ambiguous i s not surprising given the counter-cyclicality o f those flows, which suggests that remittances tend to respond negatively to economic growth. Thus, failure to correct for reverse causality and other sources o f endogeneity in remittances flows may lead to misleading conclusions regarding the causal relationship from remittances to economic growth. From a conceptual point o f view, this relationship could be motivatedby the possibility that workers' remittances may help ease credit constraints thus allowing individuals to not only increase their consumption and reduce poverty, but also to augment to lead to higher investments in physical capital, education, health care, and the creation or expansion o f micro-enterprises, all o f which could eventually be reflected in higher aggregate investment and economic growth. As seen in figure 3.1, even simple scatter plots between remittances, growth and investment using a large cross section o f countries tend to suggest the presence of positive correlations between remittances, investment and growth. Panel data estimates on the impact o f remittances and growth are presented inTable 3.10, using a sample o f 67 countries, o f which 21 are from LatinAmerica and the Caribbean. We find that remittances have a positive and significant impact on growth, and that this effect i s robust to the use o f external and time varying instrumental variables to control for the potential endogeneity o f rernittancesa2lAll control variables are found to be significant and with the sign that would be a priori expected. That is, growth i s found to be higher for countries with lower levels o f income, higher levels o f education, deeper financial markets, more trade openness, and better institutions, and to be discouraged by excessive government burden, higher inflation and real exchange rate overvaluation. These results improve upon previous estimates, which have either overlooked the issue o f the possible endogeneity o f remittances, or have addressed it using time-invariant instrumental variables (e.g. IMF, 2005) or internal instruments only (e.g. Giuliano and Ruiz-Arranz, 2005). 22 Moreover, following Loayza, Fajnzylber and Calder6n (2005), we 2' The only exception is given by column [4] where we instrument remittances with their own lagged levels and differences. This may be inappropriate if remittancesare influenced in an inter-temporaloptimizingframework by future shocksto economic growth. 22 We use two external instruments constructedby Aggarwal, Demirguc-Kunt and Martinez Peria (2005), based on the real output per capita of the countrieswhere remittancesoriginate. The first instrument is the averageoutput per 77 now use a wide set o f control variables as potential growth determinants, thus reducing possible omitted variable biases. Figure3.1. Scatter plots of Remittances, Growth and Investment 1.I Workers' Remittances vs. Economic Growth CHN " 1 i! BWA k tQ MUS KNA ION MYS? ATGLKA EOg:; CPV LSO .: GAB I NIC y=0.3x+1.1137 R2= 0.0641 4 2 Work."' Rlmltt.nsi*I. p=rs.nta$p 01 QDP (In loa.) 1.2 Workers' Remittances vs. Domestic investment KNA LSO GAB MDG y = 0.0423~ 3.0747 + RZ= 0.0595 _ _ -4 .3 .2 1 0 1 2 3 4 5 W0rk.r.' Rimlttms.8 88 p.lOlntaO. 01 QDP (In loo.) Source: Own calculations The magnitude o f the estimated effect o f remittances on growth is, however, relatively small in economic terms. For the average LatinAmerican country inthe sample, for instance, the increase in remittances from 0.7 percent o f GDP in 1991-1995 to 2.3 percent o f GDP in 2001- 2005 is estimated to have led to an increase of only 0.27 percent per year (27 basis points). Moreover, as seen in the final four columns o f Table 3.10, when domestic investment is included as an additional explanatory variable, the effect o f remittances on growth ceases to be significant. This may imply that one o f the main channels through which remittances affect growth is by increasing domestic investment, capita of the top country destinations for migrants across the world weighted by the (inverse o f the) distance between the remittance-sender and the remittances-recipient country. The second instrument is the average output per capita o f the top five country destinations for migrants in the OECD weighted by the share of migrants o f the recipient country in each o f these five destinations. We refer to these two variables as the "Distance" and the "Migration" instruments, respectively. 78 Table 3.10. Remittancesand EconomicGrowth Growth Resre.stons wkhout Invaslmml CrDWthRelmsllO~lnsiudlngInvestment I11 PI 131 PI bog& LIY + om 141 [SI [61 171 Exop.nous Oin1.nce Mipnilon Exog*nou 0i.U"O. ~ i p ~ t i ~ n b w d LIV + om Vartabie R.mMnC.. h l N I M n l I~SI~YIM~I of Rmmances L m n u n n i bitmm.nt in.ir~mnt of Rmmnnsem initial GDP per capita 4.354 ** -0.281 - 4.296 ** -0.349" -0.438' -0.841 - -0 e48 - -0.524 (1" bV) (0.08) (0.06) (0.09) (0.08) (0.08)* (0.07) (0 08) (0.09) Lflmwlm lnveslrnenlRate 4.110' 0.645 '* 5.988 *. 4.325 (A.a peroentw11 GOP,Inlops) (0.85)* (0.32) (0.20) (0.19) p . . s Education 0.257 *' 0.258 0.217 0 340 ** 0.220 ** 0.303 ** 0.203 *' 0.219' lSeCmdar#EnmlimnLh1-1 (0.09) (0.11)- . (0.10) (0.08) (0.10) (0.13) (0.13) (0.11) FinanclatDepth 0.620 ** 0.304 0.499 0,523- 4.108 4.448** -0.228 * 0.007 (Prhle Domeslio CndHlo GOP, inI-) (0.19) (0.16) (0.19) (0.17) (0.17) (0.13) (0.14) (0.18) Inslilulioni 3.888 ** 4.236 ** 4,105.' 3.676.. 2.918'. 2.657 2.338 : 2.934" (ICRGPolnkalRbkIndex.Inlop.) (0.31) (0.27) (0.31) (0.29) (0.41) (0.35) (0.39) (0.41) Trade Openness(TO) 0.328 '* 0.431 0.422 ** 0.503 ** -0.095 -0.283 * -0.305 -0.155 (RealEqrarti md Import. bGOP,In 1-8.) (0.11) (0.11) (0.10) (0.12) (0.12) (0.15) (0.13) (0.11) Lack of PricaStability - -0.007 -0.006 ** 4.007 ** .0.006 *' 4.008 ' -0.005" -0.007 -0.008" (InLllon nle,Inl~liQO+lnl.rsle)) (0.00) (0.00) (0.00) (0.00) (0.00)* (0.00) (0.00) (0.00) RER Overvalueton 4 011** -0.012 -0.012 .0.010 ** -0.005 * -0.001 0.000 -0.003* (0.00) - (PmponionaiIndex InI-, overvaluatbnn>O) (0.00) (0.00) (0.00) (0.00) * (0.00) (0.00) (0.00) GovernmentBurden -0882 ** -0.882 ** -0.828 *. -0.942** 4.019' -1.061 -0.929 *' -1.026 ** - - (GerwralOovl Comumplbnh I-) (0.18) (0.19) (0.18) (0.10) (0.16)' (0.13) (0.11) (0.13) RemilUnces 0.167 0.226 0.239 ** 0.025 0.063 * 0.042 0.048 0.039 (won.n Remrnmmlo COP, Inlop.) (0.04) (Q.W (0.04) (0.07) IOW (0.05) (0.04) (0.05) No. Counlriea 87 07 67 07 07 87 67 87 No. Ob6BNBliOn0 273 273 273 273 273 273 273 273 Specification Tests (p-vduns) .Sargan Test (0.28) (0.31) (0.37) (0.84) (0.56) (0.55) (0.53) .2nd. Order Correlation (0.19) (0.19) (0.19) (0.20) (0.29) (0.29) (0.21) Direct estimates o f the effect o f remittances on the ratio o f investment to GDP confirm this hypothesis (table 3.11). Inparticular, the results suggest that from 1991-1995 to 2001-2005, the increase in remittances to LAC was responsible for a 2 percent increase in the share o f domestic investment to GDP. Given the estimated effect o f investment on growth from Table 3.10, this implies that between 1991 and 2005 the impact o f remittances on economic growth that took place through increased rates o f investment was equivalent to 13 basis points per year, or about one half of the total impact o f remittances on growth estimated during that period. The remaining effects o f remittances on growth could take place through other channels, including the reductionof aggregate volatility, the accumulation ofhumancapital andpossible increases in entrepreneurship. However, the finding o f a relatively small overall impact o f remittances on growth suggests that either the economic importance o f those other channels is limited, or there are considerable negative compensating effects o f remittances on labor supply and the real exchange rate. We then examine these issues in detail, starting in the next section with a direct analysis o f the business cycle properties o f remittances and their impact on growth volatility. 79 Table 3.11. RemittancesandInvestment 111 ' 121 131 141 Exogenous Distance Migration Lagged Lev + Dlff Variable Ramittancea Inatrument Instrument of Remittances Persistence Lagged Investment Ratio 0.520 ** 0.527 ** 0.547 ** 0.541 ** (in logs) (0.02) (0.02) (0.02) (0.01) GU2kxth EconomicGrowth 0.037 ** 0.035 ** 0.034 ** 0.037 ** (In percentages) (0.00) (0.00) (0.00) (0.00) Education 0.035 ** 0.021 -0.032 ** 0.020 (Secondary Enrollment,In logs) (0.02) (0.01) (0.01) (0.01) FinancialDepth 0.047 ** 0.051 ** 0.057 ** 0.041 ** (PrivateDomestlc Credit to GDP. in logs) (0.01) (0.01) (0.01) (0.01) institutions 0.319 ** 0.347 ** 0.415 ** 0.235 ** (ICRG PoiitlcalRlsk Index. in logs) (0.06) (0.06) (0.07) (0.04) Price of Investment -0.055 ** -0.066 ** -0.058 ** -0.067 ** (in logs) (0.02) (0.02) (0.02) (0.02) Lack of PriceStability -5.81E-04 ** -3.55E-04 "* -2.99E-04 -7.28E-04 ** (lnflatlonrate, in log[lOO+inf.rate]) (0.00) (0.00) (0.00) (0.00) Workers' Remittanew Remittances 0.007 ** 0.017 ** 0.015 * 0.003 (Workers Remittancas to ODP, in ioga) (0.00) (0.01) (0.01) (0.00) No. Countries 66 66 66 66 No. Observations 268 268 268 268 Specification Tests (p-values) - Sargan Test - 2nd. (0.36) (0.30) (0.20) (0.27) Order Correlation (0.95) (0.98) (0.89) (0.99) All regressionsincludea constant and time dummies. * ('7 denotes statlstlcalsignificance at the 10 (5)percent level Source: Own calculations V. Remittancesand Output Volatility Output fluctuations in developing countries are substantially more volatile than those in industrial economies, with median welfare cost o f business cycles ranging from 10 to 30 times the costs for industrial economies (Pallage and Robe, 2003). To the extent that remittances exhibit a counter-cyclical behavior, they could play a crucial role in smoothing out developing countries' output fluctuations, and helping maintain macroeconomic stability. This could be the case if remittances were dominated by compensatory transfers, sent by migrants to their families in order to offset or prevent income shortfalls due to negative external shocks (e.g. natural disasters and financial crisis, among others). Such transfers would raise household income at times when recipient economies are indownturnphases of their economic cycles. However, it is also possible that remittances respond to profitable investment opportunities in recipient economies' ,thus operating as standardprivate capital flows and behaving pro-cyclically. Previous evidence on the business cycle properties o f remittances flows has so far been inconclusive. Chami, Fullenkamp and Jahjah (2005) find that in general remittances are negatively associated to the income gap o f the recipient country with respect to the UnitedStates. Similarly, Mishra (2005) and Sayan (2006) show that remittances tend to be countercyclical, respectively among Caribbean countries, and for low and lower-middle income countries. However, when analyzing the behavior o f remittances sent by Turkish workers from Germany, 80 Sayan (2004) finds that they are positively related to real output in Turkey, hence, appearing to behave pro-cyclically. Moreover, the evidence presented by Giuliano and Ruiz-Arranz (2005) suggests that in most cases remittances co-move directly with output fluctuations in recipient countries, with a higher degree o f pro-cyclicality incountries with shallower financial systems. A common shortcoming o f previous attempts at determining the cyclical properties o f remittances flows is that they have not controlled for the potential endogeneity o f output fluctuations. Table 3.12 presents the results o f panel regression-based correlations between the cyclical components o f remittances and real output in recipient countries, for a sample of 26 Latin American countries, controlling for the possible presence o f reverse causality from remittances to GDP, and for the fact that both variables could be affected by a common deterministic time trend, as well as by changes inthe output o f remittances-sending countries.23 Table 3.12. The CyclicalBehavior of RemittancesinLatinAmerica Pooled Fixed and Time Effects Domestic Fweign Domestic Foreign SampleI Filter GDP growth GDP growth GDP growth GDP growth 1. Ordinary Least Squares First-Differencecyclicalcomponents -1.1425 '* 1.7358 -1.2963 ** -8.9426 (0.491) (1281) (0.533) (28.801) Band-PassFilter cyclical components -1.3736 ** 2.2620 ** -1.5691 ** -3.9082 (0.581) (1.029) (0.529) (25.843) Hcdrick-PrescottFilter cyclicalcomponents -1.1917 ** 1.9613 ** -1.3809 ** 3.5604 (0.597) (0.990) (0.529) (24.636) II. Instrumental Variables (IV) First-Differencecyclical cwnponents -1.1575 ** 1.7269 -1.1601 ** -9.0473 (0.475) (1.277) (0.544) (28.866) Band-PassFilter cyclical components -1.2913 ** 2.1792 ** -1.3976 ** -3.7560 (0.573) (1.028) (0.540) (25.927) Hcdrick-PrescottFilter cyclicalcomponents -1.1293 * 1.9001 * -1.2213 *` 3.7368 (0.586) (0.988) (0.541) (24.708) Source: Own calculations We find that there i s a negative and significant relationship between remittances and real output in the remittance-recipient country regardless o f the estimation technique used. Moreover, our estimates imply that reductions in real output below the trend are associated with more than proportional increases in the ratio o f remittances to GDP above the trend. Thus, for instance, a one standard deviation reduction in real output o f 2.7 percent below the trend would be associated with an increase in the ratio o f remittances to GDP above the trend o f approximately 3.8 percent (using band-pass filter and instrumental variable estimates). On the other hand, we 23 For the sake o f robustness, we use different filters to compute the cyclical components o f remittances and GDP: first differences, Hodrick-Prescott filter, and band-pass filter. We also include country-specific and time-specific fixed effects. To instrument real output o f remittance-recipient countries we follow Fatas and Mihov (2003) inusing lagged levels o f this variable and current and lagged fluctuations in international crude oil prices. The real output o f the remittance-sending country is approximated by the average foreign GDP growth o f the top country destinations for migrants weighted by the (inverse o f the) distance between sender and recipient countries ("distance foreign GDP growth") and the average foreign GDP growth o f the Top 5 OECD country destinations for migrants weighted by their share o fmigrants ineach country ("migration foreign GDP growth"). 81 aces and thc real output of' rcmi ut of both recipient and sending c loping world ffigurcs 3.2 and 3.3) Figure 3.3. Rcmittanccs' Sensitivit) to Output Fluctuations in Sending Countries 82 One concern with the above estimates i s that they reflect average responses o f remittances to output in the whole sample o f L A C countries. Figure 3.4 reports time series estimates obtained country-by-country. Although these results are statistically significant in only 8 o f the 26 countries, they suggest considerable heterogeneity, across Latin American countries. Indeed, we find that in 16 out o f 26 countries remittances behaved counter-cyclically, with an average coefficient for the output o f the recipient country o f -2.88. The countries where these estimates are significant include Ecuador, Argentina, Costa Rica and Mexico. On the other hand, in the remaining group of 10 countries, remittances appear to behave pro-cyclically, with significant positive output coefficients for El Salvador, Paraguay and Venezuela. It thus appears that the extent to which remittances operate as compensatory transfers or profit driven capital flows differs considerably across countries. Figure3.4. Country Estimates of Remittances'Sensitivityto Own Output 4 - 2 - 0 - -2 - 4 - -8- -8- Source: Own calculations Notwithstanding the somewhat conflicting results on the business cycles properties o f remittances flows, several papers have shown that workers' remittances play a critical role in reducing the vulnerability o f home-recipient households to large negative shocks. Using aggregate data for 87 countries from 1970 to 2002, Yang (2006) finds that among poorer developing countries remittances increase substantially in the aftermath o f natural disasters. Kapur (2004) and Hysenbegasi and Pozo (2002) conclude that remittances increase sharply after large adverse macroeconomic shocks and currency crisis, respectively. Using microeconomic data on Jamaican households, Clarke and Wallsten (2003) show that remittances have operated as a partial insurance mechanism, covering 25 percent o f hurricane damages. Similar conclusions have been obtained by Yang and Choi (2005) for the Philippines, where workers' remittances replace approximately 60 percent o f exogenous declines in income, and the World Bank (2005), which shows that remittances have increased after floods in Bangladesh, and hurricanes in the Dominican Republic, Haiti and Honduras. 83 Figure 3.5 shows the evolution o f remittances in the wake o f sudden stops in capital flows and currency crises.25 We find that in Latin America, as well as in the rest o f the developing world remittances (as a percentage o f GDP) have clearly declined in the periods preceding both types o f severe negative shocks, but they have increased considerably thereafter. Inparticular, inthe year after a sudden stop remittances to LatinAmerica have increased by 0.35 percent o f GDP (0.43 percent for currency crises). By the 4th year the average increase with respect the level o f remittances preceding the sudden stop is 0.75 percent o f GDP (2.74 percent for currency crises). The result that on average remittances behave counter-cyclically, as well as the evidence that they tend to increase after severe crises suggest that they should reduce the volatility o f output per capita o f recipient countries. This has been previously highlightedby the International Monetary Fund (2005), which finds that a 2.5 percentage point increase in the ratio o f remittances to GDP is correlated, on average, with a one-sixth decline in aggregate output volatility. Moreover, remittances are found to improve the creditworthiness o f recipient countries, as illustrated by a positive and robust association between remittances and credit ratings for sovereign debt. For instance, model-based calculations show that including remittances in creditworthiness assessments would improve the credit ratings o f Haiti by two notches (from CCC to B-) and o f Nicaragua by one notch (from CCC+ to B-).As shown by the World Bank (2005), these improvements would imply a reductioninthe sovereign spread o f 334 and 209 basis points, respectively. Figure3.5. The response of Remittancesto MacroeconomicCrises 0 0 0.0 0 4 0 2 0.0 -0.a -04 -All D-uloping (mxr I.lmnd.> -LAC EAPl -0- -0- Changer In Rernlttanser after Sudden Stops 1 -2.0J Changes In Rernlttanser after Currency Crlses Source: Own calculations 25We define and time those events following respectively Frankel and Cavallo (2004) and, for currency crises, the episodes defined in Frankel and Rose (1996), Kaminsky and Reinhart (1999), and Goldstein, Kaminsky and Reinhart(2000). 84 Additional evidence on the impact o f remittances on the volatility o f output growth i s provided in Table 3.13, which shows the impact o f workers' remittances on the standard deviation o f GDP per capita growth, controlling for standards determinants o f growth volatility. The later include macroeconomic policy variables, external shocks and country- and time- specific effects. We find that there is a robust negative relationship betweenworkers' remittances and growth volatility. This implies that countries with larger remittance flows (as percentage o f GDP) tend to have less volatile (or more stable) real output fluctuations. Economically speaking, a one standard deviation increase in remittances (1-72) would reduce the standard deviation o f growth inreal output per capita by more than 10 percent, from a sample average o f 3.01 to 2.67. The volatility reducing effects o f remittances increase with per capita income, as evidenced by the estimated negative sign o f an interactive between remittances and GDP per capita. Quantitatively, the reduction in volatility that follows a one standard deviation increase in remittances is almost twice as large for countries close to the 80thpercentile o f the distribution o f per capita income (about $3000), compared to countries with the sample's medianincome (about $1,000). In contrast, for those countries with per capita incomes below $300 (the 15'h percentile o f the distribution), remittances are estimated to heightenmacroeconomic volatility. Table 3.13. Remittances and Growth Volatility S.D. Growth in Real GDP per capita [I1 [21 131 Variable Exogenous Distance Migration MacroeconomicPolicv Variables Inflation Volatility 0.006 ** 0.007 ** 0.009 ** (S.D. annual logdifferencesof CPI) (0.00) (0.00) (0.00) Monetary Policy Volatility 0.133 ** 0.133 ** 0.135 ** (S.D. annuallogdifterencesof Money) (0.02) (0.01) (0.01) Fiscal Poiicy Volatility 0.100 ** 0.100 **' 0.102 ** (S.D. annual logdifferencesof Govl. Consumption) (0.01) (0.01) (0.01) RER Overvaluation 0.010 ** 0.009 ** 0.007 ** (Proportionalindexinlogs, OMKyaluaIlonif XI) (0.00) (0.00) (0.00) Systemic Banking Crises 1,121 ** 1.133 ** 1.187 ** (Frequencyofyaars undercrises:0.1) (0.07) (0.06) (0.06) Trade Openness(TO) -0.541 ** -0.493 ** -0.266 ** (RealExports and ImpOltslo GDP, inlogs) (0.12) (0.12) (0.11) Remittances 4.183 ** 4.202 " -0.219 " (Workers Remittancesto GDP, logs) (0.02) (0.02) (0.02) Volatiltv of Foreicm Shocks Volatility of Terms of Trade Changes 0.009 ** 0.010 ** 0.011 ** (S.D. annual logdifferencesof TOT) (0.00) (0.00) (0.00) Volatility of Foreign Growth Volatility 0.105 ** 0.113 ** 0.044 (S.D. annuallogdbrences d ForeignGrwvth) (0.05) (0.05) (0.05) Countriesi Observations 89 I359 89 / 359 89 i 359 - Sargan Test - 2nd. (0.48) (0.54) (0.49) Order Correlation (0.44) (0.46) (0.55) Source: Own calculations We also find evidence that external shocks, fiscal and monetary policy shocks, real exchange overvaluations and banking crises have a smaller volatility increasing effect in countries with higher levels o f remittances. Indeed, interactives between measures o f those various shocks and the share o f remittances in GDP are found to have negative and significant coefficients in our volatility regressions. To illustrate this finding, table 3.14 reports the impact on volatility o f various types o f shocks for countries located in different deciles o f the remittances distribution. In the case o f external shocks we find that for countries in the first 85 decile o f the remittances distribution -with a share o f remittances in GDP below 0.1 percent - a one standard deviation change ina country's terms o f trade or inthe volatility o f its main trading partners increases the standard deviation o f per capita GDP growth by respectively 0.27 and 1.06. In contrast, the effect o f those shocks for countries in the sixth decile o f the remittances distribution - with an average remittances share o f 2.3 percent o f GDP - i s only a 0.08 increase in volatility (for both types of shocks). Similarly, more volatile fiscal and monetary policies, higher real exchange rate overvaluations and more frequent banking crises increase growth volatility to a much lower extent incountries with higher remittances shares. Table 3.14. Volatility Effects of Externaland PolicyShocks by RemittancesLevels (changes in std. dev. of p/c GDP Growth after one std. dev. shock) Shocks to Growth Volatility Distribution of Volatile External Shocks Volatile Policy Shocks Crises Remittances Terms of Foreign Monetary Fiscal RER Banking (deciles) Trade Growth Policy POIICY Overvaluation Crises I 0.272 ** 1.062 '* 1.286 ** 1.333 ** 0.819 ** 0.724 ** (0.04) (0.19) (0.16) (0.13) (0.16) (0.07) II 0.202 ** 0.708 ** 1.001 ** 1.097 ** 0.522 *' 0.599 *' (0.02) (0.08) (0.07) (0.05) (0.07) (0.03) 111 0.162 ** 0.507 ** 0.838 ** 0.963 ** 0.354 ** 0.528 ** (0.01) (0.05) (0.04) (0.03) (0.04) (0.02) IV 0.130 ** 0.342 *' 0.705 *' 0.852 ** 0.216 '* 0.469 ** (0.01) (0.05) (0.04) (0.03) (0.04) (0.02) V 0.102 *' 0.204 *' 0.593 '* 0.760 ** 0.100 *' 0.420 ** (0.01) (0.03) (0.03) (0.02) (0.03) (0.01) VI 0.077 *' 0.076 * 0.490 ** 0.674 ** -0.007 0.375 ** (0.01) (0.04) (0.03) (0.03) (0.03) (0.01) VI1 0.052 *' -0.048 0.390 ** 0.592 '* -0.111 ** 0.331 ** (0.01) (0.03) (0.03) (0.02) (0.03) (0.01) Vlll 0.030 '* -0.161 ** 0.299 ** 0.516 ** -0.205 ** 0.291 ** (0.01) (0.03) (0.03) (0.02) (0.03) (0.01) IX 0.010 -0.264 '* 0.215 ** 0.448 ** -0.292 ** 0.255 ** (0.01) (0.04) (0.03) (0.03) (0.03) (0.01) X -0.033 -0.479 ** 0.042 0.304 ** -0.472 '* 0.179 ** (0.02) (0.12) (0.10) (0.08) (0.10) (0.04) Source: Own calculations V. Conclusions This chapter has shown that migration and remittances have a significant poverty reducing effect that appears to operate mainly through increases in per capita income o f remittances-receiving countries. Our cross-country and micro-based estimates yield very similar conclusions, suggesting that for each percentage point increase in the share o f remittances to GDP, the fraction o f the population living inpoverty is reduced by about 0.4 percent. However, the poverty and inequality impact o f remittances varies considerably across countries, depending on their general level o f development, initial inequality and on the segments o f the income distribution where remittances-receiving households are concentrated. The chapter has also shown that remittances have a positive and significant impact on growth. This effect i s robust to corrections for the potential endogeneity o f remittances, and the use a wide set o f control variables as potential growth determinants. However, the magnitude o f 86 the estimated effect of remittances on growth is found to be relatively small in economic terms. For the average Latin American country in the sample, for instance, the increase in remittances from 0.7 percent o f GDP in 1991-1995 to 2.3 percent o f GDP in 2001-2005 i s estimated to have led to an increase o f only 0.27 percent per year inper capita GDP growth. With regard to the channels through which remittances affect economic growth, direct estimates o f the effect o f remittances on the ratio o f investment to GDP suggest that about one half o f the impact o f remittances on growth takes place through increased rates o f domestic investment. Another important channel i s the reduction o f aggregate volatility. Indeed, the evidence indicates that remittances behave counter cyclically inmost countries o f the region and they increase sharply after macroeconomic crises. Moreover, after controlling for various sources o f external and policy shocks, we find that remittances significantly reduce growth volatility, both directly and by diminishing the impact on the economy o f external and macroeconomic policy shocks. 87 Annex 1:Remittances and PovertyHeadcountsusingNationalPovertyLines Table A3.1. Poverty Headcounts before and after remittances (national poverty lines) Extreme Moderate Extreme Moderate Country Poverty Diff. in Poverty DiK in Poverty Diff in Poverty Diff.in (National % (National % Country (National % (National % Poverty Line) Povertv Line) Poverty Line) Poverty Line) Bolivia (2002) Honduras (2002) Non-Remittances Income 38.086 61.780 Non-Remittances Income 39,136 60.000 Total Income 37.480 61.254 Total Income 36.039 57.027 Diff.beforeiafter remitt. -0.6 -1.6% -0.5 -0.9% Dif. befordafter remitt. -3.1 -7.9% -3.0 -5.0% Ecuador (2004) Mexico (2002) Non-Remittances Income 27.484 54.449 Non-Remittances Income 24.291 29.970 Total Income 25.793 52.628 Total Income 22.171 28.009 Dit befordafter remitt. -1.7 -6.2% -1.8 -3.3% Dit befordafter remitt. -2.1 -8.7% -2.0 -6.5% El Salvador (2000) Nicaragua(2001) Non-Remittances Income 22.715 43.063 Non-Remittances Income 29.470 52.351 Total Income 17.684 38.737 Total Income 28.522 51.441 Dif.beforelafter remitt. -5.0 -22.1% -4.3 -10.0% Dif.beforelafter remitt. -0.9 -3.2% -0.9 -1.7% Guatemala (2000) ' Paraguay(2003) Non-Remittances Income 34.077 60.536 Non-Remittances Income 24.025 47.727 Total Income 32.060 58.780 Total Income 23.097 47.039 Dit beforeiafter remitt. -2.0 -5.9% -1.8 -2.9% Dif.befordafter remitt. -0.9 -3.9% -0.7 -1.4% Haiti (2001) Peru(2002) Non-Remittances Income 89.613 96.526 Non-Remittances Income 23.107 44.047 Total Income 85.391 94.521 Total Income 22.923 43.676 Dif.beforelafter remitt. -4.2 -4.7% -2.0 -2.1% Dif.befordafter remitt. -0.2 -0.8% -0.4 -0.8% Dominican Republic (2004) Non-Remittances Income 25.539 53.934 Total Income 21.741 49.952 Dif.beforelafter remitt. -3.8 -14.9% -4.0 -7.4% Table A3.2. Poverty Headcounts in Counterfactual Scenario of No-Migration (national poverty lines) Extreme Moderate Extreme Moderate Poverty DifX Poverty Diff. Poverty Diff. Poverty Diff. Country (National in% (National in % Country (National in % (National in % $e) P o v e r t v w ) Bolivia (2002) Honduras (2002) Non-Remittances lncome 37.801 61.660 Non-Remittances Income 37.566 58.973 95% Confidence Interval (37.583; 38.021) (61.433; 61.904) 95% Confidence Interval (37.403; 37.735) (58.797; 59.148) Total Income 37.480 61.254 Total Income 36.039 57.027 Diff. beforeiafter remitt. -0.3 -0.8% -0.4 -0.7% Dif. beforeiafterremitt. -1.5 -4.1% -1.9 -3.3% Ecuador (2004) Mexico (2002) Non-Remittances Income 26.507 53.629 Non-Remittances Income 22.025 27.923 95% Confidence Interval (26.353; 26.669) (53.430; 53.850) 95% Confidence Interval (21.868; 22.189) (27.750; 28.089) Total Income 25.793 52.628 Total Income 22.171 28.009 Dif. beforeiafter remitt. -0.7 -2.7% -1.0 -1.9% Dif. befordafter remitt. 0.1 0.7% 0.1 0.3% El Salvador (2000) Nicaragua (2001) Non-Remittances Income 18.933 40.034 Non-Remittances Income 28.778 52.265 95% Confidence Interval (18.693; 19.178) (39.700; 40.348) 95% Confidence Interval (28.348; 29.253) (51.659; 52.871) Total Income ~. 17.684 38.737 Total Income 28.522 51.441 ~ Dif.beforelafter remitt. -1.2 -6.6% -1.3 -3.2% Dit befordafterremitt. -0.3 -0.9% -0.8 -1.6% Guatemala (2000) Paraguay(2003) Non-Remittances Income 33.300 59.675 Non-Remittances Income 23.216 47.203 95% Confidence Interval (32.911; 33.682) (59.323; 60.049) 95% Confidence Interval (23.059; 23.397) (47.008; 47.415) Total Income 32.060 58.780 Total Income 23.097 47.039 Dif.beforeiafter remitt. -1.2 -3.7% -0.9 -1.5% Dif.beforelafter remitt. -0.1 -0.5% -0.2 -0.3% Haiti (2001) Peru (2002) Non-Remittances Income 87.959 94.894 Non-Remittances Income 22.951 43.759 95% Confidence Interval (87.406; 88.494) (94.447; 95.347) 95% Confidence Interval (22.910; 22.996) (43.678; 43.845) Total Income 85.391 94.521 Total Income 22.923 43.676 Dit beforelafter remitt. -2.6 -2.9% -0.4 -0.4% Dif.befordafter remitt. 0.0 -0.1% -0.1 -0.2% Dominican Republic (2004) Non-Remittances Income 22.270 50.680 95% Confidence Interval (21.957; 22.624) (50.198; 51,156) Total Income 21.741 49.952 Dif.beforeiafterremitt. -0.5 -2.4% -0.7 -1.4% 88 Table A3.3. Poverty Headcounts among Recipients Households in Counterfactual Scenario of No-Migration (national poverty lines) bxtreme Moderate Extreme Moderate Country Poverty DifT Poverty Diff. Poverty DifT Poverty Diff. (National in % (National in % Country (National in% (National in % Povertv Line) Povertv Line) Povertv Line) Povertv Line) Bolivia (2002) Honduras (2002) Non-Remittances Income 35.750 56.598 Non-Remittances Income 27.453 52.000 95% Confidence Interval (29.310; 42.239) (49.960; 63.206) 95% Confidence Interval (26.048; 28.888) (50.372; 53.785) Total Income 26.346 44.918 Total Income 13.476 34.206 Diff, beforelafter remin. -9.4 -26.3% -I1.7 -20.6% Dif. before/after remitt. -14.0 -50.9% -17.8 -34.2% Ecuador (2004) Mexico (2002) Non-Remittances Income 21.375 47.294 Non-Remittances Income 26.176 34.229 95% Confidence Interval (19.069; 23.791) (44.219; 50.534) 95% Confidence Interval (23.506; 28.899) (31.501; 37.31 I) Total Income 10,380 31.987 Total Income 28.680 36.680 Dif. before/after remitt. -11.0 -51.4% -15.3 -32.4% Dif.before/after remitt. 2.5 9.6% 2.5 7.2% El Salvador (2000) Nicaragua (2001) Non-Remittances Income 16.761 39.532 Non-Remittances Income 16.210 35.937 95% Confidence Interval (I5.505; 18.002) (37.864; 41.224) 95% Confidence Interval (13.110; 19.387) (31.965; 39.995) Total Income 10.293 32.826 Total Income 14.623 30.792 Dif. beforeiafter remitt. -6.5 -38.6% -6.7 -17.0% Dif. beforelafter remin. -1.6 -9.8% -5.1 -14.3% Guatemala (2000) Paraguay (2003) Non-Remittances Income 31.918 53.975 Non-Remittances Income 24.590 52.221 95% Confidence Interval (27.961; 35.852) (49.812; 58.064) 95% Confidence Interval (20.702; 28.991) (46.745; 57.502) Total Income 18.830 44.621 Total Income 21.641 48.199 Dif. before/after remin. -13.1 -41.0% -9.4 -17.3% Dif. before/after remitt. -2.9 -12.0% 4 0 -7.7% Haiti (2001) Peru (2002) Non-Remittances Income 79.825 89.526 Non-Remittances Income 2.732 12.239 95% Confidence Interval (77.725; 81.825) (87.852; 91.114) 95% Confidence Interval (1.313; 4.376) (9.330; 15.223) Total Income 70.322 88.126 Total Income 1.769 9.184 Dif. beforelafter remitt. -9.5 -1 1.9% -1.4 -1.6% Dif.beforeiafter remin. -1.0 -35.2% -3.1 -25.0% Dominican Republic (2004) Non-Remittances Income 15.136 40.880 95% Confidence Interval (13,811; 16.412) (38.855; 42.947) Total Income 12.826 37.685 Dif. beforelafter remin. -2.3 -15.3% -3.2 -7.8% 89 Chapter 4 RemittancesandHouseholdBehavior* By increasing the income of recipient households, remittances can lead to changes in savings, expenditure patterns and other household behaviors. For instance, remittances may allow previously poor families to meet their basicfood needs and subsequently increase their expenditures in housing, education or health. Expenditure patterns can also change if migrants tie remittances 'jlows to specijk expenditures, or ifmigration changes the preferences or incentives of those that are left behind. For instance, to the extent that migrants tend to work in occupations requiring limited schooling, the returns from investments in education may be lowerfor those that are envisaging international migration. Health outcomes can also change possibly through a combination of increases in income and information transferred by remittances senders. Finally, by affecting local labor market conditions and household budget constraints, remittances may alter labor force participation decisions, and tilt individuals' occupational choices towards home production and/or entrepreneurship. I.Introduction Do recipient households save a fraction o f remittances income? Are remittances spent mostly on "conspicuous" consumption goods? Do households that receive remittances destine a larger fraction o f their income to investments inhousing? How do expenditures in education and health vary with remittances? Are they reflected in better educational and health outcomes? Are remittances associated with changes in labor market participation? These questions have usually been at the center o f the public debate on the impact o f remittances on recipient communities. However, there i s little rigorous empirical evidence on a wide set o f countries to inform that debate, which is thus often based on country case studies and anecdotal evidence. This chapter aims at reducing this gap, using the household survey data employed in chapters 1 and 3 to describe recipients' profiles and estimate remittances poverty effects. In particular, to allow for the impact o f remittances to vary throughout the income distribution, we use the household income estimated in chapter 3 for the counterfactual scenario o f no migration, to compare the ~~ * This chapter i s based on the backgroundpaper "Remittances and HouseholdBehavior," preparedby Pablo Acosta and Pablo Fajnzylber. 90 behavior o f recipient households with other households with similar characteristics (including per capita income prior to migration). Since some o f the behaviors under analysis exhibit different patterns by gender and also across urban and rural areas, when appropriate we also allow the impact o f remittances to vary along those dimensions. 11.RemittancesandHouseholdSavings One o f the main channels through which migration and remittances can affect household welfare i s by providing mechanisms to smooth consumption in the context o f negative external shocks. Indeed, in the absence o f efficient credit and insurance mechanisms, migration and remittances can play an important role by allowing households to diversify their income sources and thus operate as ex-ante risk coping mechanisms. In addition, households also react to negative shocks - ex-post - by sending some its members to work abroad, or by asking existing migrants for additional monetary assistance during bad times. That this effectively happens in practice i s confirmed by the evidence presented in the previous chapter regarding the counter- cyclicality o f remittances flows. A third channel through which remittances could help households smooth out the effects o f negative shocks and increase their welfare is by allowing increases in savings and the accumulation o f assets. Once again, the macro level evidence presented in the previous chapter on the positive effect o f remittances on investment rates would suggest that aggregate savings are likely to be affected as well. Moreover, as shown in the next chapter, both macro and micro data indicate that remittances tend to increase bank deposits, which suggests that recipient households are able to save some o f the income from that source. Despite these indications, there i s no direct evidence regarding the saving behavior o f remittances recipients. This section attempts to fillthat gap on the basis o f household surveys from six LAC countries, which contain information on income (including remittances) and expenditures.26 We calculate saving rates as the difference between total income and expenditures as a fraction o f the former. The darker bars in Figure4.1 below show the differences between saving rates o f recipients and non-recipients: in four out o f the six countries, recipients save more than non-recipients, the only exceptions being Mexico and El Salvador. However, since saving rates are known to increase with incorney2'the saving behavior o f remittance recipients in the latter countries could be driven by the fact that, as shown in chapter 1, migrants in Mexico and El Salvador tend to come from lower income quintiles. To avoid potential spurious saving differentials between recipients and non-recipients, we estimate a simple model o f savings as a function o f income quintiles - using the counterfactual pre-migration income variable calculated in the previous chapter - as well as other demographic characteristics o f the household. The estimated equation is o f the following form. 26These countries are Mexico, El Salvador, Guatemala, Peru, Nicaragua and the Dominican Republic. 27See Butelmann and Gallego (2001) and the references therein. 91 ad urhcr householdcharit as rhc brighter bars in can and Salvadoran rcc In ordcr 10 inc how r ~ ~ ~ ~ hold behavior ~ n~ r~ othe~incomt ~ ~ t ~ i ~ ~ ~ h tion, wc haw also ~ ~ I ~ usaving~ c dfor ~ h various inconic q ~ i n ~ani ~ ~ s l ~ c us. This is rcportcdi .Ibelowwhich,forthesix C O U ~ 92 the expected pattern of savings increasing with household income (column l).30 same The pattern i s observed for non-recipients (column 3). Inthe case o f remittances recipients, however, saving rates increase with income only in the case o f the Dominican Republic, with different patterns emerging in the other five countries (column 2). Thus, in Mexico, Peru and Guatemala saving rates do increase from the first to the 2ndor even 3rd income quintiles, but they fall again for recipients located in the upper quintiles, thus creating an inverted-Upattern between savings and income. Possible explanations include the absence from the household, inthe case o f better- off recipients, o f a larger number o f income generating migrants - see chapter 2 on the evidence that richer migrants tend to bring along most o f their direct relatives. However, in El Salvador and Nicaragua a U-shaped pattern is suggested by the data, with lower income recipient households exhibiting saving rates that are above those o f the middle class, which in turn saves less than households inthe top income quintiles. Table 4.1. Savingrates by IncomeQuintileand RemittanceRecipientStatus Country Counterfactual Savings Rates Savings Rates Savings Rates Savings Rates Savings Rates ncome Quintilet All Rem. Recip Non-Recip Iifference R-NR3LS Coefficientsfor Rem Mexico QI -0.051 0.013 -0.054 0.067" 0.073" 4 2 0.029 0.064 0.026 0.037' 0.001 4 3 0.090 0.112 0.088 0.025 -0.016 4 4 0.158 0.088 0.163 -0.075*** -0.112*** Q5 0.244 0.017 0.249 -0.231*** -0.277'" El Salvador Ql 0.285 0.514 0.252 0.262"' 0.244". 4 2 0.480 0.484 0.479 0.004 -0.007 4 3 0.534 0.501 0.544 -0.043'" -0.037"' 4 4 0.592 0.503 0.621 -0.118*** -0.093"' Q5 0.653 0.518 0.687 -0.170'*' -0.167"' Guatemala Ql -0.418 -0.042 -0.479 0.437*** 0.413"' 4 2 -0.184 0.057 -0.220 0.277"' 0.308*** 4 3 -0.089 -0.070 -0,090 0.020 0.060 4 4 0.020 -0.012 0.023 -0.035 -0.221*** 0.021 45 0.180 -0.025 0.201 -0.226"' Peru Ql 0.035 0.105 0.034 0.071 0.061 4 2 0.130 0.104 0.131 -0.027 -0.050 4 3 0.199 0.278 0.196 0.081 0.088; 4 4 0.268 0.181 0.273 -0.092 -0.081" Q5 0.388 0.192 0.398 -0.206*** -0,183"' Nicaragua QI 0.009 0.378 -0.005 0.383"' 0.562'" 4 2 0.182 0.314 0.17 0.144** 0.141 '* 4 3 0.286 0.258 0.291 -0.033 -0.012 4 4 0.358 0.338 0.363 -0.025 0.020 Q5 0.461 0.275 0.521 -0.246'" -0.149"' Dom. Rep. Ql -0.363 -0.194 -0.388 0.194"' 0.182*** Q2 -0.32 -0.087 -0.373 0.286"' 0.282"' 4 3 -0.191 -0.088 -0.226 0.138"' 0.154'" 4 4 -0.08 -0.024 -0.103 0.078"' 0.120*** Q5 0.167 0.038 0.204 -0.166**' -0.118*** Notes:*** Signi: mt at 1% level. Simificant at 5' eve!. Sianific * at 10% level. Source:Author's calculationsbasedon householdsurveys 30It must be noted that the calculation of saving rates is plagued by problems o f under-reporting of both expenditures and income. The fact that the gravity of these problemsvaries across the household surveys o f different countries is reflectedinthe large variance of savingrates across countries-e.g. inthe case of El Salvadorthe under- reporting o f expenditures is likely to be more serious than that of income, which is reflected in excessively high saving rates. For this reason , we focus on within-country comparisons, both across income quintiles and between recipientsandnon-recipients. 93 As for the differences between recipients and nom-recipients, in the six countries recipients in lower income quintiles exhibit higher saving rates than non-recipients (column 4). However, these differences tend to diminish and become negative for households located higher up inthe income distribution. The same pattern is obtained when these differences are calculated in a regression framework, after controlling for other household characteristics, by adding to equation (1) a set o f interactives between Ri and the income quintile "dummy" variables. Estimates for these interactives are reported in the last column o f table 4.1. Once again, the results suggest that saving rates tend to increase for poorer remittances receivinghouseholds, but they tend to decrease for better-off households. One possible interpretation o f this finding i s that for those in the lower quintiles remittances operate more as an ex-ante risk coping mechanism, while for those located in the upper quintiles remittances become significant after the corresponding households have been hit by negative income shocks which have reduced their saving capacity. Moreover, as mentioned above, it is also possible that the number o f absent income generating migrants i s larger among richer households. It must be noted, however, that the fact that some remittance recipients save less than other households with similar characteristics does not necessarily imply that they consume the income from remittances entirely (Le. that their saving rate i s not positive). By itself, that result suggests only that the propensity to save out o f the remittances income is lower than the corresponding saving rate from non-remittances income. Indeed, as described in equation (2) below, the observed saving rate o f recipient households can be expressed as a weighted average o f the saving rates for those two sources o f income. where Si and Wi are respectively the saving rate out o f income source i and the share in total income o f that source, and the indexes r and nr denote remittances and non-remittances income respectively. Assuming that the saving rate out o f non-remittances income i s equal to the saving rate o f non-recipients (SI = Snr), equation (2) implies that the difference between the savings o f recipients and non-recipients (S2 - SI) will be positive (negative) if and only if Sr i s larger (smaller) than Snr.Inthis framework, inorder to determine the value o f Sr after having estimated (S2 -S,), oneneedstoapplythefollowingtransformation: Figure 4.2 reports saving rates out o f remittances income calculated on the basis o f equation (3), using estimates o f (S2 - SI) obtained either from the differences in means between the samples o f recipients and non-recipients (darker bars), or those derived from regressions that control for income quintile and other household characteristics (brighter bars). The figure also reports, for comparison purposes, the observed total saving rates for those receiving remittances (S2). Inalmost all cases, our estimates o f S, are positive, suggestingthat recipients do not spend remittances entirely (the only exception is Peru, when household characteristics are controlled for). Overall, the result that recipient households tend to save a positive fraction o f their remittances income - even if the saving rate is lower than that for non-remittances income - is consistent with the findings o f the previous chapter regarding the positive impact o f remittances on domestic investmentand growth. 94 durable goads, htsusiitg, 95 tility of that sorf o f in urban and in rural arcas, spend less than tlz cation andhealth. Figur Thc cotitrots arc the ~amcas in equation ( I f 96 Rclativcly similar rcsults - reductions in food sha dijrablcs, durxblcs, hcsllth cation - arc obtained wh into accourtt, and o f remittances i s controlled for.35However, to the extent that savings and expenditure patterns vary with income - see previous section - the above estimates o f averages effects on all remittances recipients may be misleading, in the sense that they could mask possible differences across households located in different parts o f the income distribution. To investigate these possible differences, we allow the impact o f remittances on household behavior to vary across quintiles o f the distribution o f per capita household income. Moreover, in order to also capture the possible effect o f remittances on expenditures that operates through increases in household income, we classify remittances recipients according to the income that they were estimated to have earned prior to migration (see chapter 3). The results, reported in table 4.3, indicate that remittances effects are quite different across segments o f the income distribution, and they also vary considerably across countries. Table 4.2. Access to Remittances and ExpenditureShares (OLS, differenceswith respect to non-recip :nt housek Ids) Dependent Variable I1 Food I Non-Durables Durables Housing Education Health IMexico' I -0.031*** I NIA 0.014*** -0.006** 0.003 0.021*** (0.004) (0.003) (0.004) (0.003) 0.002* 0.003** 0.019*** 0.006*** (0.001) (0.001) (0.003) (0.001) IGuatemala I -0.034*** I 0.010* 0.006** 0.000 0.009*** 0.009* (0.006) (0.005) (0.002) (0.002) (0.003) (0.005) PeN2 -0.043*** 0.024*** -0.006 -0.001 0.009** 0.016*** (0,008) (0.007) (0.004) (0.003) (0.004) (0.005) Nicaragua -0.014* -0.002 0.000 0.003 0.008 0.005 (0.007) (0.006) (0.001) (0.003) (0.005) (0.006) Jamaica' -0.002 -0.007 0.004*** 0.003** -0.005* 0.009*** (0.005) (0.005) (0.001) (0.001) (0.003) (0.002) DominicanRepublic' -0.013*** -0.002 0.000 0.000 0.003* 0.012*** (0.004) (0.003) (0.001) (0.001) (0.002) (0.003) INotes: *** Significant at 1% level. ' Foodand ** Significantat 5% levi*Significant a1 0% level. other non-durable goods together. *Only urban areas. 'Coefficients for housing inJamaica and Dominican Republic are multiplied by 10. Source:Author's calculations basedon householdsurveys Inthe case of Mexico, for instance, remittances recipients inthe lower quintiles exhibit a pattern that i s similar to the one observed for the population as a whole. Indeed, they increase expenditures in durable goods, housing and human capital, mainly at the expense o f non- durables. In contrast, their richer counterparts exhibit higher expenses in non-durable goods and lower expenditures in housing improvements and education. These results would suggest that, at 35 The following instrumental variables are used for remittances: the percentage o f households that receive remittances in the respective county o f residence (a proxy for the presence o f migrant networks), and its interaction with household level characteristics that affects the decision to migrate, including among them an indicator for the presence of 0-5 years old children in the household, the number o f adult males in the household, and the average educational level among adults. 98 least inMexico, remittances are used in a more productive way by poorer households. Arguably, remittances have the effect o f relaxing budget constraints that limit the housing and human capital investments o f poorer families. For richer households, on the other hand, the results suggest that the above budget constraints are not binding, so that remittances have the effect of increasing consumption o f food and non-durable goods. This is illustrated in figure 4.5, which shows that in Mexico recipients in the first quintile experience a reduction in food and non- durable goods expenses and an increase in educational expenditures in comparison with non- recipients o f similar characteristics, but both changes tend to become smaller - and even have their signs inverted-as one moves up inthe income distribution. I r I I Remittances -0.063**' NIA 0.016' 0.013" 0.0201 O.O15** Remittances'Q2 0.028 NIA 0.000 -0.027"' -0.005 0.003 Remittances'Q3 0.036' NIA 0.008 -0.023''' -0.025' 0.004 Remittances'Q4 0.045" NIA -0.012 -0.020'' -0.024' 0.011 Remittances'QS 0.074*** NIA -0.022 -0.020' -0.051*** 0.019 Remittances -0.004 -0.01I 0.012'*' 0.004 -0.003 0.002 Remittances'Q2 -0.040*** 0.018 -0.007' 0.001 0.026.' 0.003 Remittances'Q3 -0.046*** 0.036" -0.013'*' 0.000 0.020* 0.004 Remittances'Q4 -0.043'** 0.030" -0.010** -0.004 0.023" 0,004 Remittances'QS -0.022' 0.001 -0.011** -0.003 0.030*'* 0.006' Remittances -O.OSO*** 0.040' 0.006* 0.016 -0.013" 0.001 Remittances'Q2 0.003 -0.007 0.003 -0.018 0.014'' 0.006 Remittances'Q3 -0.003 -0.031 0.006 -0.015 0.025"' 0.019 Remittances'Q4 0.016 -0.046" -0.005 -0.015 0.031'" 0.019 Remittances'QS 0.049" -0.047' -0.002 -0.019 0.026"" -0.007 Remittances 0.003 0.012 -0.004 -0.013'" -0.003 -0.008 Remittances'Q2 -0.078 0.044 0.002 0.018' 0.008 0.024 Remittances'Q3 -0.057 0.006 0.003 0.013'' 0.020 0.027' Reminances*Q4 -0.055 0.015 0.005 0.016** 0.004 0.032" Reminances*QS -0.018 -0.001 -0.014** 0.004 0.015 0.018 Remittances 0.056' -0.013 0.007 -0.002 -0.041 ** -0.008 Remittances*Q2 -0.057 0.008 0.000 0.000 0.039' 0,009 Remittances'Q3 -0.094''' 0.015 0.003 0.003 0.050" 0.023 Remittances'Q4 -0.074" 0.018 -0.009 0.005 0.042" 0.017 Remittances*Q5 -0.067'* 0.006 -0.015 0.007 0.064'" 0.005 Remittances -0.026** 0.019' 0.007*'* -0.001 -0.010 0.009. Reminances*Q2 0.02 1 -0.024' -0.001 0.003 0.004 0.000 Reminances'Q3 0.025' -0.028" -0.001 0.004' 0.005 -0.002 Remittances'Q4 0.039'. -0.035'' -0.010'' 0.006 0.006 -0.001 Remittances'QS 0.033 -0.052"' 0.003 0.006 0.011 0.004 Remittances -0.016 0.024" 0.001 0.005 -0.010 0.000 Remittances*Q2 -0.004 -0.013 0.004 -0.001 0.013* 0.001 Remittances'Q3 -0.018 -0.023. 0.002 -0.008 0.010 0.030' Remittances'Q4 0.005 -0.027*' -0.002 -0.006 0.015' 0.010 Remittances*QS 0.028 -0.042**' -0.008*** -0.004 0.017" 0.005 I Notes: *** Significant at II% level.** Significant atI5% level.* I I I I I ' Food Significant at 10% level. and other non-durable goods together. 'Only urban areas. 'Coeficients for housing inJamaica and Dominican Republic are multiplied by IO. Source: Author's calculations basedon householdsurveys 99 drop out of school, which butc to icnt h ~ ~ ~ ~ h ~ l d ~ . Figure 4.8, renrrs in School ~ ~ ~ r oRates for 12-17 yea I ~ ~ ~ n t ittan ReeiDient Status Sourcc: Author's calcuiacion bascd on h o ~ s ~sitrvcys ~ ~ l ~ # 102 Some o f these differences, however, could be attributed to the fact that, as shown in chapter 1, households that report remittances exhibit considerably different demographic and income characteristics than non-recipients. We thus estimate regression models that attempt to deal with this problem, using a specification that closely follows Hanson and Woodruff`s (2003), and focus on the accumulated schooling o f children aged 10 to 15. The models, which follow equation (4) below, are estimated separately by gender, and urban and rural areas whenever possible (provided sufficient variability inremittances' receipts i s available). =a+PXi +yMi+Xi+mi+E, (4) where Ej represents the number o f school grades completed by child i, Xi is a vector o f child and household characteristics (age o f the child, a dummy for the child being the oldest in the household, indicators for the number o f children o f different ages inthe household, presence o f a 0-5 year old child, family home ownership, counterfactual income quintile), Mi i s a set o f characteristics o f the child's mother (indicators for mother's education, mother's marital status, mother's head o f household status, and a quartic in mother's age), Ci represent community characteristics (the proportion o f households with sanitary services in the county o f residence, the proportion o f household heads working in agricultural activities in the county o f residence, and state/province indicators), Ri is a "dummy" variable for households that receive remittances, and ci i s a random error. Regression estimates for the S coefficient are presented intable 4.4. Our results suggest that access to remittances is positively and significantly associated with higher educational attainment in 6 out o f 11 countries - the exceptions being Mexico, Paraguay, Peru, Jamaica and the Dominican Republic. The estimated positive impact o f remittances varies by gender and across rural and urban areas - e.g. in Ecuador an impact is found only for urban areas. Since the actual amount remitted i s likely to have a differential impact depending on the magnitude o f the transfer, we have also replicated this analysis looking at the impact o f the per capita value o f remittances received by the household. The results suggest that higher remittances also increase schooling in the Dominican Republic, and among boys inurban Mexico.36 Previous evidence on Mexico has suggested that the positive effects o f remittances on schooling vary with the educational attainment o f the children's' parents, being generally larger when the latter are low. Differential effects o f this sort could be due to the fact among poorer families - with lower levels o f adult schooling - remittances could have a more sizable effect in terms o f relaxing budget constraints that keep children out o f school. However, one could also expect an opposite effect - remittances having a smaller impact on education when the schooling 36A potentialmethodological concern associated with the above results is that unobserved household characteristics that affect their propensity to have migrants and receive remittances could also be driving the decisions to keep children in school. We have attempted the use o f instrumental variables to address this simultaneity bias, but have failed to find appropriate instruments that are sufficiently correlated with the migration status o f households but do not otherwise affect the educational attainment o f children. While the instruments used in other sections of this chapter -the fraction o f households that receive remittances in the county and its interaction with household characteristics - have passed standard specification tests, the fact that the size of the sample o f children aged 10 to 15 i s much smaller than in other sections, has led coefficients and standard errors to increase considerably in comparison with ordinary least squares, leadingus to focus on the latter. 103 o f parents is low - if less educated parents exhibit lower preferences for educational over other alternative expenditures. To find out which effect dominates, we re-estimate equation (4) adding an interaction term between remittances (Ri) and a variable that indicates whether the mother has four o f more years o f education. The corresponding results are reported intable 4.5. Table 4.4. Access to Remittances and ChildrenEducation OLS - Age Group 10-15 1 ars Old DependentVariable Accumulatc Schooling Sample Rural Urban Boys Girls Boys Girls Mexico -0.149 0.113 -0.023 -0.235 (0,129) (0.095) (0.177) (0.192) El Salvador 0.482*** 0.312*** 0.245** -0.010 (0.110) (0.097) (0.098) (0.097) Guatemala 0.448** 0.313* 0.291 0.454** (0.177) (0.164) (0.230) (0.177) Honduras 0.427*** 0.495*** 0.298*** 0.360*** (0.100) (0.098) (0.078) (0.080) Ecuador 0.187 0.112 0.289** 0.314** (0.149) (0.151) (0.125) (0.134) Paraguay -0.012 0.208 (0.195) (0.182) 0.098 0.244** (0.107) (0.098) -0.070 0.157 (0.146) (0.120) 0.437*** 0.375*** (0.153) (0.133) Jamaica I -0.114 -0.005 (0.150) (0.1 13) DominicanRepublic 0.077 0.118 (0.107) (0.094) I Notes: *** Significantat YOlevel. **Significant i5% level. * Significantat 10% level. RuralandUrbanareas together. ' Source: Author's calculationsbasedon householdsurveys Our findings confirm those o f previous papers on Mexico, suggesting that the positive effect o f remittances on education tends to be larger when the schooling o f parents i s low. For instance, among rural girls in Mexico, Paraguay and Peru,37our previous results suggested no effect o f remittances on educational attainment but we now find a positive and significant effect for those whose mother have at most three years o f educational attainment, whereas for the remaining children the effect i s estimated to be close to zero - although slightly negative in Mexico and Paraguay. Similarly, in Guatemala and Honduras, some o f the previously estimated positive effects o f remittances are now found to be larger in magnitude for children with 37Inthe case o fPeru, the sample includesbothurbanandrural children. 104 uneducated mothers and considerably smaller for those whose mothers have at least four years o f schooling. 4ge Group 10-15 Years Old Dependent Variable Accumulated Schooling Zountry Variable Rural Urban Boys Girls Boys Girls Mexico Receive Remittances -0.082 0.329" -0.041 -0.573 (0 192) (0.141) (0.329) (0.553) Receive Remittances * -0.144 -0.417** 0.024 0.461 Mother Educ 4 Years or More (0.240) (0.186) El Salvador Receive Remittances 0.511*** 0.251** (0.129) (0.115) (0.170) (0.186) Receive Remittances * -0.1 16 0.229 -0.203 0.297 Mother Educ 4 Years or More (0.212) (0.176) (0.197) (0.206) Guatemala Receive Remittances 0.482** 0.223 (0.200) (0.186) (0.337) (0.23 1) Receive Remittances * -0.179 0.450 -0.323 -1.336*** Mother Educ 4 Years or More (0.389) (0.318) (0,408) (0.3 IS) Honduras Receive Remittances 0.581*** 0.662*** 0.731*** 0.554*** (0.142) (0.155) (0.1 78) (0.209) Receive Remittances * -0.317" -0.328* -0.564*** -0.247 Mother Educ 4 Years or More (0.193) (0,184) (0.193) (0.220) Ecuador Receive Remittances 0.278 -0.106 0.502 0.805' (0.233) (0.237) (0.331) (0.463) Receive Remittances * -0.138 0.386 -0.239 -0.547 Mother Educ 4 Years or More (0.277) (0.287) (0,344) (0.475) Paraguay Receive Remittances 0.056 0.433' (0.271) (0.235) Receive Remittances * -0.133 -0.476 Mother Educ 4 Years or More (0.374) (0.345) Haiti Receive Remittances 0.043 0.273** (0.120) (0.111) Receive Remittances * 0.229 -0.1 11 Mother Educ 4 Years or More (0.237) (0.220) Peru Receive Remittances 0.187 0.393*** (0.296) (0.144) Receive Remittances * -0.362 -0.343* Mother Educ 4 Years or More (0.3 38) (0.207) Nicaragua Receive Remittances 0.577** 0.554** (0,260) (0.221) Receive Remittances * -0.208 -0.296 Mother Educ 4 Years or More (0.310) (0.258) Jamaica ' Receive Remittances 0.510 -0.236 (0.465) (0.435) Receive Remittances * -0.668 0.253 Mother Educ 4 Years or More (0.484) (0.443) Dom. Rep. ' Receive Remittances -0.148 0.301 (0.242) (0.208) Receive Remittances * 0.282 -0.242 IMother Educ 4 Years or More I (0.263) (0.230) I Notes: *** Significant at 1% level.**Significant at 5% level.*Significant at 10% level. I Rural and Urban areas together. Source: Author's calculations based on household surveys Overall, with the addition o f Paraguay and Peru to the list o f countries where remittances affect educational attainment, we are left with only two countries (Jamaica and the Dominican Republic) inwhich those effects are always non-significant. As for the differences by gender and urban status, our estimation results do not allow easy generalizations and suggest that the various 105 potential effects o f migration and remittances - relaxation o f budget constraints, social disruption, changes in the returns o f schooling - carry different relative weights depending on the country and socio-economic group involved. In any case, it appears that remittances tend to relax budget constraints that otherwise would force children to leave school and reduce their educational attainment, but this effect i s sometimes restricted to those with less educated mothers. Health Outcomes Few papers have addressed the impact o f migration and remittances on child health. The exceptions are mostly focused on infant mortality. Brockerhoff (1990) and Ssengonzi, De Jong and Stokes (2002) investigate the effects o f female migration on the survival chances o f their children in Senegal and Uganda, respectively. They find that rural to urban migration significantly increases child survival chances. Kanaiaupuni and Donato (1999) analyze the effects o f village migration and remittances on infant survival outcomes inMexico, and conclude that remittances reduce infant mortality. However, the authors reach an opposite conclusion for the effect o f migration: higher rates o f infant mortality in communities experiencing intense migration. Finally, with data on Mexican municipalities Lopez-Cordova (2006) concludes that larger proportions o f remittances and migrant households at the community level are associated with lower infant mortality rates. Further evidence on the impact o f migration on child health has been provided by Hildebrandt and McKenzie (2006). The authors investigate the impact o f international migration on several children health outcomes inMexico. Their results show that migrant households have lower rates o f infant mortality and higher birth weights. Moreover, they find evidence that migration also raises maternal health knowledge and the likelihood that the child was delivered by a doctor. On the other hand, preventative health care (such as breastfeeding, visits to doctors, and vaccinations) seem to be less likely for children from migrant households. No previous study has investigated the impact o f remittances and migration on anthropometric indicators for young children (weight-for-age, height-for age). Hoddinot and Kinsey (2001) present results on how external shocks drastically affect child growth in developing countries. Both weight and heightmeasures are good indicators o f health status, with different consequences in the short run and long run. For instance, lower weight i s associated with malnutrition and higher mortality risk. Similarly, lower stature in childhood is strongly correlated with lower body size in adulthood, with negative consequences on earnings and productivity (Thomas and Strauss, 1997), and increased risk o f cardiovascular and lungdiseases. With this motivation, the present section aims at uncovering evidence on the effect o f international migrant remittances on anthropometric health indicators typically used in the health literature, and with well known links with child growth. We also estimate the impact o f remittances on the probability that the delivery o f children born in the year preceding the survey was assisted by a doctor, and on the probability that children aged 2 to 5 received the complete set o f required vaccinations. The anthropometric measures on which focus are the Weight-for- Age (WAZ) and Height-for-Age (HAZ) Z-Scores for children aged 1 to 5 years old. These are 106 standardized measures o f performance inweight and height, and consist on comparing each child o f a given age to a reference group. The reference tables for WAZ and HAZ are taken from the CDC Growth Charts for US (Kuczmarski et al., 2000).38 Inparticular, the following formula i s used for calculating the Z-Scores. where X i s the measure o f interest (weight, height) for the child, M i s the median o f the corresponding variable, S i s the generalized coefficient o f variation, and L i s the power in the Box-Cox transformation taken from the CDC reference tables for a given age group. In order to avoid extreme values and outliers due to misreporting, we follow Hoddinott and Kinsey's (2001) recommendation to drop children with z-scores greater than 6 or lower than -6. As the original reference tables account for children with 1.5, 2.5, 3.5, etc. months, in order to match our data the reference months are rounded up (i.e., 1.5 month in the reference table are equal to 2 months).39 Figures 4.9 and 4.10 show the distribution o f weight-for-age and height-for-age anthropometric z-scores, using kernel density estimation, for the cases o f Guatemala and Nicaragua, the only two L A C countries in which the household surveys used in this report provide the information needed for calculating the health indicators employed in this section. Plot densities o f the above described anthropometric indicators for children aged 1-5 years old from remittance recipient and non-recipient households are estimated using kernel densities. The figures show that children from recipient households have both higher weight-for-age and height-for age z-scores. Kolmogorov-Smirnov tests for equality o f distributions reject the equality o f distributions for recipient and non-recipient households, and suggest that remittances are in fact associated with better children anthropometric scores. In order to test whether these results are driven by the differential characteristics of households with and without migrants, we estimate a regression model similar to the one used for educational attainment (equation 4), changing only the dependent variable, from years o f schooling to the four health indicators measuredabove. Inaddition to this basic specification, we estimate a modified version o f equation (4), where the indicator for remittances recipients i s interacted with a dummy variable for the second quintile o f the income distribution - using the counterfactual income prior to migration - and a dummy for households located in the third to fifthq~intiles.~' 38ReferenceTables can be foundat www.cdc,gov/nchs/about/major/nhanes/growthcharts/datafiles.htm 39For WAZ, the original tables used are the "Weight-for-age charts, birth to 36 months", for children from 0 up to 36 months.For older children,values where extrapolated usingthe "Weight-for-agecharts, 2 to 20 years". Similarly, for HAZ the reference tables used are "Lenght-for-age charts, birth to 36 months", for children from 0 up to 36 months, and "Stature-for-agecharts, 2 to 20 years", for older children. 40We group the 31dthrough5'hquintiles due to the relatively small sample size for some ofthe estimations. 107 Figure4.9. Anthropometric Measures for Children aged 1to 5, by RemittanceRecipient Status-Guatemala. Children 1 5 Years Old - Weight-for-Age Z Scores Height-for-AgeZ Scores rnJ Y - r_ 0 - Source: Author's calculations based on household surveys Figure4.10. AnthropometricMeasures for Children aged 1to 5, by RemittanceRecipient Status -Nicaragua. Children 1 5 Years Old - Weight-for-Age Z Scores Height-for-AgeZ Scores Source: Author's calculations basedon household surveys Table 4.6 reports the corresponding results, including the coefficients on free-standing dummy variables for the second through fifth income quintiles.As confirmed by our estimates, both weight- and height-for-age indexes tend to increase monotonically and significantly with household income, and so does the likelihood o f doctor-assisted deliveries in the case of Nicaragua. Moreover, controlling for pre-migration income, children from households that report receiving remittances tend to exhibit higher health outcomes than those from non-recipients households with similar demographic and socio-economic characteristics. While the relatively small sample sizes make most of the estimated interactives between remittances and income quintilesnon-significant from a statistical point o f view, inmost cases the results clearly indicate 108 that the impact o f remittances on children health i s concentrated on low income households located inthe first quintile o f the income distribution. Country Guatemala Nicaragua Child Dependent Variable Weight-for- Height-for- ReceivedAll Weight-for- Height-for- ReceivedAll Age Z-Score Age Z-Score Vaccines Deliveredby Doctor by Age Z-Score Age Z-Score Vaccines Doctor 2nd Income Quintile 0.117** 0.141** 0.011 0.006 0.154* 0.230** -0.028 0.104* (0.058) (0.060) (0.011) (0.015) (0.085) (0.091) (0.047) (0.060) 3rd Income Quintile 0.233*** 0.385** 0.016 0.054*** 0.077 0.327*** -0.01 1 0.085 (0.060) (0.067) (0.013) (0.025) (0.099) (0.109) (0.054) (0.070) 4th Income Quintile 0.325*** 0.479** 0.010 0.023 0.263** 0.594"' -0.126 0.168* (0.073) (0.076) (0.016) (0.023) (0.117) (0.113) (0.062) (0.079) 5th Income Quintile 0.594*** 0.686** 0.026 0.013 0.352** 0.594*** -0.102 0.263** (0.091) (0.098) (0,018) (0.025) (0.138) (0.136) (0.078) (0.082) Remittances 0.211** 0.213 0.065** 0.255*** 0.306 0.289 0.119 0.297** (0.089) (0.228) (0.021) (0,160) (0.394) (0.347) (0.225) (0.090) Remittances*Q2 -0.327 0.084 -0.082 -0.034' -0.370 -0.079 0.034 -0.463* (0.283) (0.264) (0.079) (0.007) (0.457) (0.421) (0.275) (0.225) Remittances*Q3-Q4-Q5 -0.423 0.004 -0.041 -0.036*** -0.252 -0.148 0.071 -0.623*** (0.272) (0.253) (0.071) (0.006) (0.418) (0.385) (0.252) (0.114) V. Remittancesand Labor Supply The impact o f remittances on labor supply i s in principle ambiguous. For individuals from households with migrants, the net additional income derived from remittances could have the "income effect" of increasing the demand for leisure and reservation wages, with a consequent reduction in labor force participation. However, out-migration also has the direct effect o f reducing the size of the labor force, and the ensuing upward pressure on local wages could inturn create a "substitution effect" away from leisure, with a consequent increase inlabor supply for those living in areas with high migration rates. In the case of Mexico, for instance, Mishra (2004) estimates that emigration raised average wages by 8% between 1970 and 2000. In addition to the above factors, in households with recent migrants the need to replace the income lost due to the migration o f wage earners could reinforce the effect o f higher market wages, resultinginan increase inthe labor participation o f stayers. Previous papers on the subject suggest that that remittances tend to reduce labor force participation in rural Mexico (Hanson, 2005) and El Salvador (Acosta, 2006), but they do not have a significant effect in Nicaragua (Funkhouser, 1992). In particular, Hanson finds that in Mexico, receiving remittances from abroad reduces both the likelihood o f working outside the home and the hours worked by both males and females. While Hanson interprets the latter result as evidence o f increasing intra-household specialization - e.g. remittances income allowing families to "buy back" some o f the labor time o f the women that stayed in Mexico - he implies that caution should be used when interpreting the result o f a lower male labor supply in households with migrants. Indeed, despite Hanson's efforts to control for self-selection into 109 If LAC countrics. Inour males in that agc graup and thus rcport ZGT males and females, inurban and rural areas - the only exception being Peru, where recipients are almost absent inrural areas.4*The estimated models are o f the following form. where Lj represents either the number o f hours worked by individual i or a dummy variable that takes the value one if he/she i s active in the labor market (is either working or looking for a job), Xi i s a vector o f personal characteristics (a quartic in age, indicators for educational attainment, marital status), Hi i s a set o f household characteristics (household size and composition, home ownership, and state/province indicators), Ri is a "dummy" variable for households that receive remittances, and Ej is a random error. Results are presented in tables 4.7 and 4.8, respectively. Table 4.7. Access to Remittances and HoursWorked Age Group 20-59 ' Brs Old Dependent Variable Hours wor: ilast week Sample Rural Urban Males Females Males Females Mexico -15.473*** -13.187**4 12.686*** -7.561 *** (1.022) (1.979) (1.473) (2.175) El Salvador -4.498*** -12.257**` -5.546*** -8.580*** (0.986) (2.089) (0.847) (1.115) Guatemala -0.969 -15.253*** -6.180*** -7.228*** (1.426) (3.418) (1.601) (2.427) Honduras -3.052*** -12.813*** -6.307*** -10.067*** (0.704) (2.046) (0.816) (1.152) Ecuador -2.186* -4.399** -3.391*** -5.622*** (1.264) (1.723) (1.1 17) (1.621) Paraguay -15.395** -5.583 10.865 7.506 (7.247) (13.027) (11,680) (10.789) Haiti -8.410*** -2.925 -1.221 2.308 (2.831) (3.069) (1.170) (1.412) lperu 12.711 *** -7.455*** (1.870) (2.306) Nicaragua -3.096 -0.701 -7.216*** -7.776*** (1.889) (5.208) (1.643) (2.134) Dominican Republic -5.278*** -8.844*** -7.240*** -10,245*** 1.301 2.828 1.031 1.514 INotes: *** Significant a YOlevel. ** Significant 5% level. * Significant i t I0% level Source: Author's calculations based on household surveys 4'Control variables include a quartic in age, indicators for years o f education, marital status, indicators for the number o f children o f different ages in the household, presence o f a 0-5 year old child, the number o f adult males and females in the household, family home ownership, and state/province indicators. The sample is restricted to individuals aged 20 to 59. The instruments for remittances include the incidence o f households with remittances in the county and its interaction with household level characteristics that affect the decision to migrate. The validity o f the instruments is confirmed both in the first stage regressions and by means o f Sargan's over-identification tests. 111 Confirming previous evidence, our results suggest that in all the 10 countries for which data i s available, remittances have the effect o f reducing the number o f hours worked per week. This negative effect is present both inurban and rural areas, the only exceptions beingParaguay and Haiti, where hours worked are reduced in rural areas only, and Nicaragua, where the estimated effect is significant in urban areas only. As for differences by gender, no clear generalizations are possible, with five out o f ten countries showing larger effects among females and four exhibiting the opposite pattern. Table 4.8. Remittances and Labor Force Participation(with instrumentalvariables) Age Group 20-59 ' u s Old Dependent Variable Labor Forcr articipation Sample Rural Urban Males Females Males Females Mexico -0.329** * -0.245*** -0.097 0.023 (0.048) (0.049) (0.135) (0,146) El Salvador -0.087 -0.598*** -0.032 -0.309*** (0.070) (0.083) (0,101) (0,100) Guatemala 0.007 -0.095 -0.228 -0.261 (0.170) (0.221) (0.140) (0.206) Honduras -0.006 -0.135** -0.096 -0.095 (0.068) (0.067) (0.094) (0.095) Ecuador -0.228** -0.154* -0.310** 0.211 (0,090) (0.085) (0.134) (0.203) Paraguay -0.009 0.052 -0.530* 0.908** (0.092) (0.174) (0.321) (0.380) Haiti 0.254 0.338** 0.114** 0.263*** (0.164) (0.165) (0.051) (0,088) Peru -0.334*** -0.284** (0,099) (0.123) Nicaragua -0.181* 0.337 -0.211 * -0.008 (0.099) (0.208) (0.127) (0,140) Jamaica -0.047 -0.027 -0.128** -0.056 (0.05 1) (0.05 1) (0.064) (0.061) DominicanRepublic -0.222** -0.010 -0.108 -0.131 (0.092) (0.126) (0.071) (0.082) Notes: * *** Significant a1 )/o level. **Significant 5% level. Significant at 10% level. Source: Author's calculations based on household surveys Similar results are obtained for the decision to participate in the labor market using a "probit" model and assuming remittances to be exogenous. Indeed, such a model yields negative effects in 10 out o f 11 countries - data i s now available also for Jamaica - and the weakest results are obtained for Paraguay and Nicaragua (where significant results are restricted to urban areas), as well as for Haiti (where the effects are non-significant). However, when remittances are allowed to be endogenous and are appropriately instrumented, their negative impact on labor force participation ceases to be significant in a number o f cases (table 4.8) - e.g. their impact on 112 labor supply becomes non-significant in Guatemala as well as in urban areas in Mexico, Honduras, and the Dominican Republic. Moreover, inboth Paraguay and Haiti, females living in urban areas are found to be more likely to participate in the labor force when receiving remittances, suggesting that the latter may be having a social disruption type o f effect that in those cases dominates possible reductions inreservation wages. 4ge Group 20-59 Years Old 3ependent Variable Labor Force Participation Zountry Variable Rural Urban Males Females Males Females Mexico Receive Remittances -0.434*** -0.283*** -0.250 0.188 (0,055) (0.066) (0.1s)) (0,204) Receive Remittances Educ 4+ * 0.214*** 0.059 0.233 -0.223 (0.073) (0.069) (0.186) (0.203) El Salvador Receive Remittances -0,108 -0.623"' -0.045 -0.376"' (0,076) (0.087) (0.120) (0.108) Receive Remittances * Educ 4+ 0.050 -0.072 0.018 0.118 (O.OS9) (0,067) (0.074) (0.072) 3uatemala Receive Remittances -0,037 -0.119 -0,164 -0.407' (0,183) (0.223) (0.191) (0.224) IReceive Remittances * Educ 4+ 0.214 0.417' -0.091 0.328 (0.168) (0.248) (0,175) (0.223) Honduras Receive Remittances 0.032 -0.148** -0,257** -0.326'. (0.084) (0.097) (0.130) (0.129) Receive Remittances Educ 4+ * -0.067 0.022 0.212" 0.315*** (0.076) (0.086) (0,093) (0,102) Ecuador Receive Remittances -0.256 -0.158 -1.161" 0.241 (0.158) (0.127) (0.454) (0.368) Receive Remittances Educ 4+ * 0.035 0.006 0.870** -0.033 (0.160) (0.120) (0.432) (0.3W) Paraguay Receive Remittances -0.125 -0.231 -0.296 0.461 (0,138) (0.203) (0,587) (0.498) Receive Remittances Educ 4+ * 0.167 0,469' -0.269 0.537 (0,148) (0.260) (0.579) (0.460) Haiti Receive Remittances 0.029 0.463'. 0.029 0.241** (0.215) (0,191) (0.074) (0.099) IReceive Remittances * Educ 4+ 0.288* -0,204 0.109 0.056 (0.175) (0,148) (0.081) (0.082) Peru Receive Remittances -3.458' -0.748' (1.780) (0.410) IReceive Remittances * Educ 4+ 3.190' 0.522 Nicaragua Receive Remittances -0.366" 0.380 -0.215 0.038 (0.161) (0.232) (0.163) (0.212) Receive Remittances Educ 4+ * 0.298** -0.093 0.007 -0.054 _ _ _ _ _ _ _ _ _ ~ (0.152) (0.227) (0.141) (0.171) Jamaica Receive Remittances -0.133 -0.075 0.125 -0.100 (0,129) (0,115) (0.169) (0.144) Receive Remittances * Educ 4+ 0.092 0.053 -0.289* 0.050 (0.128) (0.118) (0.176) (0.IS I ) Dom. Rep. Receive Remittances -0.220' -0.075 -0,194' -0.222' (0,122) (0.158) (0,103) (0,121) Receive Remittances a Educ 4+ 0,001 0.104 0.096 0.108 (0.093) (0.120) I Notes: ***Significant at I% level.**Significant at 5% level.* (0.083) (0.102) Significant at 10% level. Source: Author's calculations basedon householdsurveys 113 One possible concern with the above results is that remittances may be expected to have very different effects on the labor supply decisions o f individuals with different levels o f skills. Indeed, for those with higher levels o f schooling remittances income is likely to represent a smaller fraction o f total income, so that income effects derived from remittances may be relatively less important than substitution effects associated with changes in labor market conditions - e.g. increases in wages as a result o f out-migration. To test for this possibility, we have re-estimated the model underlyingtable 4.8 introducing an interactive between remittances and a "dummy" variable for individuals that have at least 4 years o f schooling (table 4.9). The findings reported in table 4.9 suggest that the reductions in labor supply caused by remittances tend to be much smaller among individuals with higher levels o f schooling. Evidence in this respect is found in 8 out o f 11 countries, the only exceptions being El Salvador, the Dominican Republic and Jamaica. Moreover, in the cases o f rural females in Guatemala and Paraguay and rural males in Haiti, for those with at least 4 years o f schooling the effect o f remittances appears to be that o f increasing labor supply, which i s consistent with either social disruption effects or with changing labor market conditions that affect remittances recipients to a larger extent. However, although depending on the country the effects can be quite different across genders, between rural and urban areas and by levels o f schooling, the overall conclusion from this section i s that remittances are more likely to reduce and not increase labor supply. VI. Remittances and Entrepreneurship Despitethe fact that remittances most often tends to reduce labor force participation, for those individuals who remain active in the labor market, remittances could potentially increase the range of available productive activities. Indeed, assuming that many would-be entrepreneurs are subject to credit constrains, one could expect remittances to provide the means for financing the opening or expansion o f small businesses. There are, however, other factors which would lead to expect fewer business owners among remittances recipients. First, ifhouseholds with migrants tend to be smaller, the pool o f non-paid family workers can end up being too small, thus making them less apt to undertake entrepreneurial activities. Second, since migration i s an inherently risky activity - especially when illegal as it i s frequently the case inLatinAmerica - households may not want to combine it with (other) entrepreneurial activities, which are also intrinsically risky. Third, the creation o f a new business can make less economic sense in the contexts in which migration often tends to develop in the first place. As pointed out by Massey and Parrado (1998), migrants often lack training and entrepreneurial experience, their families are young and have growing consumption needs, and their communities are typically distant from markets and lack basic infrastructure. Finally, to the extent that individuals who pertain to families with migrants have a higher probability o f frequent than elsewhere in the world - they could have lower incentives to engage in eventually migrating themselves - particularly in LAC, where temporary migration is less entrepreneurial activities, whose returns may take a long period o f time to materialize. Previous evidence for L A C i s limited to Mexico and Nicaragua. In the latter case, Funkhouser (1992) finds that remittances increase self-employment among males but not 114 t f S While the above tabulations are a useful first approximation, as in the case o f other types o f household and individual behaviors examined in this chapter, this type o f analysis i s likely to be biased by the fact that recipient and non-recipients are quite different populations. Thus, we proceed to compare the incidence o f self-employment and business ownership in a regression framework that controls for the influence o f personal and household characteristics on the probability o f being engaged in each o f those entrepreneurial activities. The estimated model i s as follows. Bi =a+pXi+fli +mi+Ei (7) where Bi represents either a dummy variable for being self-employed or an indicator for beinga business owner, while Xi and Hi are vectors o f respectively personal and household characteristics (include income quintiles, and dummies rural areas and male individuals). As before, Ri is a "dummy'yvariable for households that receive remittances, and Ei i s a random error term. Results are presentedintable 4.10.42 Our results indicate that only in6 out o f 11countries (El Salvador, Guatemala, Honduras, Ecuador, Paraguay and the Dominican Republic) remittances are significantly and positively associated with self-employment. The results are non-significant in three other countries and in the cases o f Jamaica and Haiti a negative relationship i s suggested by the data. In terms of the absolute magnitudes o f the estimated effects, the largest impact o f remittances i s found for Mexico and Guatemala, where the probability o f being in self-employment increases by more than 10 percentage points. As for the effects on business ownership, our estimates suggest positive and significant effects inonly four cases: Mexico, El Salvador, Honduras and Peru. Inorder to investigate the economic activities in which recipients are more likely to be engaged, we also estimate the impact o f remittances on the probability o f being a self-employed inthe agricultural, manufacturing and commerce sectors -the estimatedequation is similar to (6) changing only the dependent variable.43 The results, reported in table 4.11, suggest that remittances are most likely to increase the likelihood o f an individual being engaged in agricultural self-employment, with positive and significant effects in five countries, compared to three cases o f positive effects for manufacturing, and two cases o f positive links for the commerce sector (Guatemala and Haiti). Negative links between remittances and self- ' employment are found only for the agricultural sector, in the cases o f Haiti, Jamaica and the Dominican Republic. 42The complete set o f control variables invectors Xi andHiincludes dummies for income quintiles, rural areas, male individuals, a quartic inage, indicators for educational attainment and marital status, measures o f household size and composition, an indicator for home ownership, and state/province dummy variables. The sample is restricted to individuals aged 20 to 59. As in the analysis o f household expenditures, we have also estimated models in which remittances are instrumented using a proxy for the presence o f migrant networks interacted with household characteristics that affect the decision to migrate. The results are qualitatively similar so we concentrate on the above described "probit" model. 43Because the fraction o f individuals who report business ownership i s relatively small in all countries, we do not further disaggregate it by economic sector. 116 Table 4.10. Remittancesand Entrepreneurship Sample 20-59 Y ars Old DependentVariable Self-Employment Own Business Mexico 0.108** * 0.013** (0.018) (0.007) El Salvador 0.012 0.036*** (0.009) (0.006) Suatemala 0.101*** -0.011 (0.021) (0.008) Honduras 0.031*** 0.031*** (0.011) (0.007) Ecuador 0.032** 0.011 (0.014) (0.007) Haiti -0.029* ** -0,001 (0.011) (0.004) Peru 0.001 0.022* (0.027) (0.013) Paraguay 0.068** -0.002 (0.030) (0.009) Nicaragua 0.018 0.002 (0.021) (0,008) Jamaica -0.038* * -0.001 (0.015) (0.002) Dominican Republic 0.035*** 0.001 0.013 0.003 *Notes:*** Significant at /o level.** Sign at 5% level. Significant at 10% level. Source: Author's calculationsbasedonhousehold surveys Several o f the factors that were mentioned at the beginning o f this section as potentially related to the probability that remittance recipient households engage in entrepreneurial activities are likely to vary with income. For instance, if remittances relax credit constraints that limit the opening o f new businesses this i s likely to happen to a greater extent among poorer households. However, complementary assets and skills which may be needed to leverage remittances income are likely to be less abundant among poorer households. Finally, assuming that risk aversion i s decreasing in wealth, poorer households may also be more reluctant to engage in other risky activities besides migration. To uncover which factor prevails in practice, we re-estimate equation (6) adding interactives between remittances and income quintile "dummy" variables. The results, reported in table 4.12, suggest different patterns for self-employment and business ownership. Thus, in the case o f the latter, remittances tend to have a positive effect on the probability o f individuals from poor households - located in the first income quintile - being business owners. However, this effect becomes increasingly smaller as one moves up in the income distribution, a pattern that i s most clear in the cases o f Mexico, El Salvador, Honduras, and Ecuador. In contrast, for the case o f self-employment, the probability that a household from 117 the first quintile engages in that activity i s actually smaller in6 out o f 11 countries - significantly so in 3. However, in most cases the net effect o f remittances eventually becomes positive for upper income q ~ i n t i l e s .Overall, our results tend to confirm that access to remittances may ~ ~ positively affect the incentives for entrepreneurship inmost income quintiles, with a larger effect on business ownership among poorer households, and a larger effect on self-employment for middleto upper class ones. Table 4.11. Remittancesand Self-Employment by Sector Sample 20-59 Ye rs Old in Self-E iployment Dependent Variable Agriculture Manufacturing Commerce Mexico 0.017*** 0.005 0.004 (0.004) (0.005) (0.009) El Salvador 0.005 *** 0.001 0.001 (0.002) (0.003) (0.005) 3uatemala 0.047*** 0.006 0.022** (0.014) (0,008) (0.012) Honduras 0.003 0.003* 0.005 (0.004) (0.004) (0.006) Ecuador 0.012* 0.009** -0.003 (0.007) (0.005) (0.008) Haiti -0.133*** 0.004 0.029** (0.015) (0.005) (0.014) Peru -0.001 -0.001 -0.005 (0.017) (0.007) (0.014) Paraguay 0.028* -0.007 0.012 (0.018) (0.006) (0.016) Nicaragua 0.003 -0.006 -0.001 (0.008) (0.004) (0.010) Jamaica -0.028** * 0.001 0.000 (0.008) (0.004) (0.003) Dominican Republic -0.007* * 0.008* -0.008 (0.003) (0.005) (0.008) Notes: *** Significant at /o level. ** Sigi E at 5% level. * gnif. at 10% lev( L. Source: Author's calculations based on household surveys 44An interesting and once again sui generis pattern is found for Mexico, where the likelihood o f self-employment increases even for the poorest households but increases even more for those inthe top income quintile. 118 Table 4.12. Remittancesand Entrepreneurship,by IncomeQuintile Sample Sample 20-59 Years Old %pendent Variable Dependent Variable Self-Employment1 Own Business I I I vlexico Remittances 0.088'' 0.048" PeN Remittances -0.268 NIA Remittances'Q2 0.088' -0.01I Remittances'Q2 0.291 Remittances'Q3 -0.038 -0.010 Remittances'Q3 0.274 Remittances'Q4 -0.026 -0.019'' Remittances'Q4 0.306 Remittances'Q5 0.148" -0.020' Remittances'Q5 0.368 31 Salvador Remittances -0.056"' 0.092"' Paraguay Remittances -0.005 0.071 Remittances'Q2 0.095"' -0.012 Remittances'Q2 -0.012 -0.018 Remittances'Q3 0.071'' -0.020 Remittances'Q3 0.063 -0,029' * Remittances'Q4 0.081"' -0.041"' Remittances'Q4 0.121 -0.030'' Remittances'QS 0.115*** -0.050"' Remittances'Q5 0.I63 -0.020 katemala Remittances 0.091 0.034 Nicaragua Remittances 0.121 0.048 Remittances'Q2 0.050 -0.031 Remittances'Q2 -0.092 -0.00; Remittances'Q3 0.007 -0.013 Remittances'Q3 -0.061 -0.01; Remittances'Q4 -0.026 -0.028 Remittances'Q4 -0.102 -0.oot Remittances'QS 0.012 -0.039 Remittances'Q5 -0.095 -0.06t Honduras Remittances -0.033 0.119"' Jamaica Remittances -0.081** NIP Remittances'Q2 0.077 -0.020 Remittances'Q2 0.063 Remittances*Q3 0.058 -0.037" Remittances'Q3 0.028 Remittances'Q4 0.090' -0.051*** Remittances'Q4 0.057 Remittances'QS 0.054 -0.062"' Remittances'QS 0.113 Ecuador Remittances -0.064 0.113*** DominicanRep. Remittances 0.082 0.00: Remittances'Q2 0.142" -0.046*** Remittances'Q2 -0,018 -0.01: Remittances'Q3 0.139'' -0.042'" Remittances'Q3 0.009 0.00; Remittances'Q4 0.079 -0.034'' Remittances'Q4 -0.065 -0.001 Remittances'QS 0.110' -0.048'" Remittances'Q5 -0.084 -0.008 Haiti Remittances -0.065' -0.002 Remittances'Q2 0.018 0.023 Remittances'Q3 0.053 -0.007 Remittances'Q4 0.007 -0.006 Remittances'QS 0.047 I I -0.009I Source: Author's calcu' ions based on household surveys VII. Conclusions The econometric results presented in this chapter suggest that the effects o f remittances on household behavior vary considerably across countries and between different socio-economic groups. In particular, saving rates tend to increase for poorer recipient households but the opposite effect i s obtained for richer ones. The composition o f household expenditures, on the other hand, i s altered in the direction o f increasing human capital investments, but except for Mexico this effect i s restricted to households located in the middle to upper segments o f the income distribution. Moreover, while there is evidence that for some specific groups - defined by country, gender, and urban status -remittances increase children's educational attainment, the impact i s often restricted to children with low levels o f parental schooling. In the case o f health outcomes the results are restricted to two countries - Nicaragua and Guatemala - but in both cases they suggest that remittances improve children health, particularly among low income households. Similarly, a negative link i s found between remittances and labor supply, but the effects are often restricted to individuals with low levels o f schooling. Finally, entrepreneurship tends to be stimulated by remittances, but the effects once again vary considerably by income quintile. 119 Chapter 5 D o remittances affect recipient countries financial development*? As researchers and policy-makers have come to notice the increasing volume and stable nature of remittances to developing countries, a growing interest has emerged regarding their development impact along various dimensions. Surprisingly, little attention has been given to the question of whether remittances promote financial development in recipient countries. Yet, this issue is important becausefinancial systems perform a number of key economicfunctions and their development has been shown to foster growth and reduce poverty. Furthermore, this question is relevant since many argue that banking remittance recipients will help multiply the development impact of remittances. I.Introduction Whether and how remittances might affect financial development is a priori unclear. The notion that remittances can lead to financial development in developing countries i s based on the concept that money transferred through financial institutions paves the way for recipients to demand and gain access to other financial products and services, which they might not have otherwise (Orozco, 2005). At the same time, providing remittance transfer services allows banks to "get to know" and reach out to unbanked recipients or recipients with limited financial intermediation. For example, remittances might have a positive impact on credit market development if, as individuals receive sizeable transfers from abroad that are shown to be stable, banks become more willing to extend credit to remittance recipients. Moreover, even if higher bank lendingto remittance recipients does not materialize, overall credit in the economy might increase ifbanks' loanable funds surge as a result o f deposits linked to remittance flows. Similarly, because remittances are typically lumpy, recipients might have a need for financial products that allow for the safe storage o f these funds. In the case o f households that receive their remittances through banks, the potential to learn about and to demand other financial products i s even larger. On the other hand, because remittances can also help relax This chapter was co-authored by Maria Soledad MartinezPeria, Yira Mascaro, and Florencia Moizeszowicz, with extensive support from Paola Granata for annex 1 and 2. The chapter is based on research conducted with Asli Demirguc-Kunt, Chris Woodruff, and Ernesto Lopez Cordova. We are grateful to Emesto Lopez Cordova, Jose de LunaMartinez,ManuelOrozco, andAnna Paulsonfor providingus data andinformationusedinthis`chapter. 120 individuals' financing constraints, they might lead to a lower demand for credit and have a dampening effect on credit market development. Also, a rise in remittances might not translate itself into an increase incredit to the private sector ifthese flows are instead channeled to finance the government. Finally, remittances might not increase bank deposits if they are immediately consumed or if remittance recipients distrust financial institutions and prefer other ways to save these flows. Recent accounts o f financial institutions' attempts to "bank" remittance recipients - by lowering remittance fees and by offering specially designed products - suggest that financial institutions perceive the likely impact o f remittances on financial development to be positive.45 However, empirical research on the impact o f remittances on financial development is largely lacking. One exception is a recently completed study by Aggarwal, Demirgug-Kunt, and Martinez Peria, (2005). Using balance o f payment (BOP) statistics for over 90 countries during the period 1975-2003, the study uncovers a positive relationship between remittances and financial development. However, this study looks at all developing countries combined and does not test whether this relationship holds across regions and, in particular, for Latin America and the Caribbean (LAC). This chapter investigates the association between remittances and financial development for Latin American countries both at the macro and micro level. At the macro level, using the data and empirical approach pursued by Aggarwal et al. (2005), we compare the impact o f remittances on financial development for countries within Latin America and outside o f the region. At the micro level, the chapter presents research on the association between remittances and the use o f banking services in Latin America. With data from 19 household surveys for 11 Latin American countries, we test whether the proportion o f households that use financial services i s different between remittance recipients and non-recipients in Latin America. Furthermore, we present results from detailed studies on El Salvador and Mexico, two o f the largest remittance recipients in the region. These case studies allow us to investigate the association between remittances and financial development more rigorously, controlling for other household characteristics and addressing concerns about reverse causality. We also complement the more rigorous econometric analysis with findings derived from ad-hoc surveys and follow up interviews o f selected banks in Colombia and Guatemala (also among the largest remittance recipients inthe region). This i s intended to illustrate the increasing interest o f banks inthe remittances business, reported key contributing or limiting factors for the "bancarization" o f remitters and senders, and incipient efforts to develop specialized products for cross-selling o f services to remittances recipients. The findings from this chapter can be summarized as follows. The macro level analysis suggests that remittances have a positive impact on the financial development o f developing countries overall, but this effect is smaller for LatinAmerican countries. The micro level analysis reveals that while there i s evidence that the likelihood o f usingdeposit accounts i s higher among remittance recipients, no such effects are present thus far when it comes to bank credit. 45See Orozco and Fedewa (2005) for a summary o f recent efforts by banks in Latin America to convert remittance recipients into bank clients. 121 Though more research i s required to understand what i s driving these results, the chapter endorses a number o f measures to enhance the impact o f remittances on financial development. First, the chapter highlightsthe importance o f policies that facilitate migrant's access to banking services, such as initiatives that provide illegal migrants valid forms o f identification in migrant recipient countries, promote financial literacy, reduce the costs o f sending remittances through banks and promote the overall use o f formal financial institutions. Second, the chapter supports policies that allow financial institutions from Latin America to market their services directly to their diasporas. Third, the chapter underlines the need for governments in Latin America to promote greater outreach by reducing the regulatory burdeno f opening branches and by allowing banks to provide services through alternative means o f delivery like mobile banking and by entering into partnership with non-financial firms that offer greater geographical coverage (e.g., post offices, retail stores, and cooperatives). Fourth, the chapter advocates policies that stimulate competition in the financial sector as a way to guarantee greater outreach for both remittance senders and recipients, at lower costs. Finally, the chapter argues that in order to foster a link from remittance flows to loan use and credit market development issues such as weak creditor rights, inefficient contract enforcement mechanisms, lack o f collateral, and government crowding will have to be considered and tackled by governments as potential areas in need of reform. The rest o f the chapter is organized as follows. Section I1explores the macro-level association between remittances and financial development for Latin America. Section I11presents results from a micro-level analysis o f the relationship between remittances and financial development. Usingdata from 19 household surveys conducted in 11 countries, Section 1II.a. presents tests for differences in the percentage o f households that use financial services among remittance recipients and non-recipients. Sections 1II.band 1II.c look deeper into whether remittances foster financial development and the use o f banking services inLatin America, by focusing on the case o f El Salvador and Mexico - two o f the region's largest remittance recipients. This i s complemented by Annex 1, based on ad-hoc surveys and follow up interviews o f banks in Colombia and Guatemala (complemented by other descriptive documents), which discusses the increasing interest o f banks in the remittances business and reports key contributing or limiting factors for the "bancarizationl' o f remitters and senders. Annex 11, also based on findings from these ad-hoc surveys and complementary documents, describes incipient efforts for banks in Latin America to develop specialized products for cross-selling o f services to remittances recipients. Finally, Section IV concludes and offers some policy implications. 11.Macro-levelanalysis of the associationbetweenremittances and financial development II,1Basiccorrelations We begin our analysis o f the association between remittances and financial development inLatin America, by plotting country-level data on bank deposits and bank credit to the private 122 sector,46along with balance o f payment statistics on remittances over the period 1975-2002 (see Figure5.1). All three series are expressed as a share of GDP. Figure5.1. Remittancesand financialdevelopmentin LatinAmericancountries ." of GDP Argenlina % of GDI %ofGDI 0.8 6 of GDP Barbados 90 ..i 0.7 80 -Deposits ...... ~ 0.6 70 - PrivateCredit -Remittances (len axis) 0.5 60 70 - .... 0 4 50 - 40 - 0.3 30 ........ 30 - 0.2 20 - 5 - 0.1 10 - . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1976 1978 I980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 - ,ofGDP Belize % of OD 6 of GDP BolMs 3DI 6 0 , 90 7 1.8 - 1.6 10 1.4 8 I.2 1.O -Deposita ......PrivateCredit 6 0.8 -Remittances (Icfl axis) 4 0.6 0.4 I Oi 2 0.2 i , , , , , , , , , , , , , , < , , I . 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2 35of 6 of GDP Colombln % of GDI -Deposits 30 25 -Remittances (len axis) 20 I S 10 5 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2004 2002 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 19992001 2003 46Bank deposits include all demand, savings, and time deposits held at deposit money banks as reported in the IMF's International Financial Statistics. Bank credit refers to claims on the private sector held by deposit money banks. These numbers also come from the IMF's InternationalFinancial Statistics. 123 Figure5.1: Remittances and financial development in Latin American countries (cont.) I #; & ::] of GDP Domlnlea %OfGD Deposits Coats Rlcs % of GDI of GDP 2.0 I 90 1 %i!/ r 8 -Private Credit 1.6 1.4 .......... ......... 25 1.2 50 ..... 1i ........... I.o 40 I5 *. * . . I * . . ......... .... 't. 0.8 -Deposits ......PrivateCredit , , to., 20 30 10 -Remittances (left axis) I , , , , 1977 1979 1981 1983 I985 1987 1989 1991 1993 1995 1997 1999 2001 2003 1986 1988 1990 1992 1994 1996 I998 2000 2002 I.r l I of GDP DomlnlcsnRepubllr % of GI % ofGDP Ecuador . % o f GD r 9 -Deposits I ......Private Credit -Remittances (left axis) L 6 1 5 - 4 I 5 - ....--. ..... ..... 3 10 - * , e i4 2 5 l2 I - I . , , . . . . . . . . . . . . . . . . . . . . . . . 1 . -1 I , , , I , , , , , , 1 . I975 1977 1979 1981 1983 1985 I987 1989 1991 1993 1995 1997 19992001 2003 1990 1991 1992 I993 1994 1995 1996 1997 1998 1999 2000 2001 I50 OfGDP ElSdvsdor % of GDI 6 of GDP Grenada %ofOD Deposits 18 90 1 r 8 45 ...... PrivateCredit PrivateCredit 4O -Remittances (left axis) 14 - I 2 /B I 60 - .......... 5 - 10 1 50 - 25 ...... ............ 4 20 4 8 40 - : 3 / / , , ......Deposits , , , , i; L 6 1 - 4 ,, , , , , , & R e TPrivate:Credit ( y a ) , , , , , , , 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 1986 I987 I988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 124 Figure5.1: Remittances and financial development in Latin American countries (cont.) iofGDP Gustemsls % of GDP HaIH %ofGD 25 1, 45 1 ..q:: 40 -Deposits /dJ ......Private Credit 35 1 -Remittances (let?axis) 30 I5 - 25 j I O - ' - 20 I 5 .................. ............ I O 5 - I O 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 t . 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2W3 L45 GDP of Honduras % 01"GDP % of GDP Jamalcs % of GDI 1 r 14 45 1 12 40 35 30 - I O 30 25 - 8 15 6 20 I 5 4 I O - I O 5 - 2 1 I 5 i . . . . . . . . . . . . . . . ,,m',l- . . . . . . ' I975 1977 1979 19811983 1985 1987 1989 1991 1993 1995 1997 19992001 2003 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 20W 2002 b of GDP Mexico % ofGDPI :::i /% of GDP Nicaragua % of GD 25 1 i 60 i 20 - 50 - I 1.2 j 40 - 15 - 1.0 I ' 30 - ~ I O - 0.8 0.6 ! 20 - ..... 5 1 0.4 ! I O - -Remittances (let? axis) -Remittances (let?axis) 7 1979 1981 1983 I985 1987 1989 1991 1993 1995 1997 1999 2001 1 1992 1993 1994 1995 1996 1997 1998 1999 2WO 2W1 2002 2W3 125 Figure5.1: Remittancesand financialdevelopmentin Latin American countries (cont.) - - ?Df 4.5 8 % of GDP PS"Nll* /. jDI of GDP Paraguay h -Deposits 0.8 ...... 0.7 ......De~osils 4.0 ::i 25 - Private Credit 0.6 -Remittances (left axis) 3.5 60 : 20 - 3.0 ..... 0.5 2.5 0.4 2.0 0.3 10 ...... ........ . . * e . e.-. I.5 20 0.2 I.o 0 1 0.5 5 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 19992001 2003 1 % 30 ofGDP . Peru % of GDP 120 , SI. Kith and Nevls % of GDI /g-- r I O r I O 100:;!; ..... ..... t 0 8 60 - 80 4 L 0 6 40 ........ ..... 3 ..... .* -Deposits ......Private k 0 4 -Deposits Credit 20 - ...... Private Credit 2 -Remittances (lei3 axis) * 0 2 -Remittances (left axis) I . '1990 1991'1992'9993'1994'1995'1996'19971998'1999'2000'2001'2002'2003 . I %90 GDP of - St. Lurla % of GDP/ 6 of GDP *..*. 80 . St. Vincent and the Grenadines % o-f GD 9 70 60 ; ......... 60 - 6 5 50 I 50 - 5 4 !40 ........ 4 3 3o 1 ................. 3 ~30 - I20 -DeDosits 2 ......Private 20 . -Deposits 2 tI Credit 10 ......Pnvate J Credit ~ I y i I -Remittances (lei3 axis) -Remittances (lei3 axis) , , , , , , , , , , , ' , , , , , , 1986 1988 1990 1992 1994 1996 1998 20W 2002 % of GDP Trlnldid mdTobqo 50 7 % ofGDP r 1.0 I ;d, 1 ::; 35 - 30 .... 7 0.5 1 - 0.4 i I5 , , , , , , , , I O 5 . 0 1 1 . 19751977I979I9811983I985I987I9891991I993I995199719992041 . ~ 126 Le 5.1, St. Kitti andNevis -03610 -0.5200f' St. Lucis -0.3952 -0.5752" St. Vincent and the GrcnadiIiss -0.62 63*+* -0.8606*** uf A na and Mcxicu, it is to disccm a clear crtt an ittsincos. For these two tries, periods of pos Citl 1ca, 127 between these variables have alternated with periods where remittances increase, as financial development collapses. This last type o f episode - when remittances and financial development appear to be negatively correlated - seems to correspond to periods o f financial banking crises in these countries. It i s foreseeable that during such periods both countries underwent recessions, and remittances increasedinresponse to crises, producing the negative association. Table 5.1 shows country by country correlations between remittances and each o f the measures o f financial development. These results largely confirm the patterns highlightedabove. For Belize, Dominica, St. Kitts and Nevis, St. Lucia, and St. Vincent and the Grenadines we observe a negative correlation between remittances and both measures o f financial development. Consistent with what Figure 5.1 shows, no clear macro-level relationship appears to exist between financial development and remittances for Argentina and Mexico. Similar results are found for Haiti and Panama. On the other hand, for the remaining 16 countries, the correlations confirm the positive associationbetween financial development and remittances. 11.2Empirical estimations While the graphs and correlations above are helpfulin describing the association between remittances and financial development, a more rigorous empirical approach is required to obtain a more definitive answer regarding the link between these variables. Inparticular, it i s important to control for other factors that might affect both remittances and financial development and to correct for biases that might arise as a result o f measurement error, reverse causation, and omitted country characteristics. We empirically examine the relationship between financial development and remittances by estimating equations (1) through (3): where irefers to the country and t refers to the time period from 1975 to 2003. The dependent variable, FD, is either the share o f bank deposits to GDP or the ratio o f bank credit to the private sector to GDP, depending on the regression. These are common indicators o f financial development (see King and Levine, 1993). Rem refers to the ratio o f remittances - as recorded in balance o f payment statistics - to GDP.X i s a matrix made up o f variables that the literature has found to affect financial development: including country size, the level o f economic development (as measuredby GDP per capita), the inflation rate, financial liberalization, capital and current account openness48. 48Data on financial development, country size, GDP per capita, and inflation come from the International Financial Statistics (IMF) and the WorldDevelopment Indicators (World Bank). Information on dual exchange rates systems comes from the Annual Report on Exchange Arrangements and Restrictions (IMF). 128 Equation (1) models the relationship between remittances and financial development for all developing countries combined, allowing for individual country effects. By including an interaction term between a dummy for Latin American countries and the measures o f financial development, equation (2) allows the relationship between remittances and financial development to be different for countries in the region vis-&vis other developing countries. In particular, LAC i s a dummy that equals 1 for the 25 countries in the Latin American region included in our sample. The net impact o f remittances on financial development for countries in LatinAmerica is given by the sum o f 61and 82. Because the graphs and correlations presented above suggest that there are differences across countries within the region when it comes to the link between remittances and financial development, equation (3) separates Latin American countries into three groups. Specifically, we distinguish between the 16 countries for which we found a positive correlation between remittances and financial development (hence forth referred to as LACZ6), the 4 countries (LAC4) for which the correlation between remittances and financial development were insignificant, and the remaining 5 countries (LACS), which exhibited a negative correlation between remittances and financial development. The impact o f remittances on LAC16 countries i s given by the sum o f q l plus q2, which for LAC5 countries can be measured by the sum o f q l plusq3, while the net impact o f remittances on LAC4 countries is givenby q1plusq4. An important complicationinempirically studyingthe impact o f remittances on financial development is the potential for endogeneity biases as a result o f measurement error, reverse causation, and omitted variables, Officially recorded remittances are known to be measured with error. Ina recent paper based on a survey o f central banks, De Luna Martinez (2005) reports that balance o f payment statistics produced by developing countries often neglect remittances received via money transfer operators4' and almost always exclude those transferred via informal means such as hawala operators, friends, and family members. Furthermore, estimates o f the size o f informal remittances range from 20 to 200 percent o f official balance o f payment statistics, suggesting that the scope for measurement error is very large (Freund and Spatafora, 2005). The possibility that remittances affect financial development, giving rise to estimation biases due to reverse causality i s also justified. Better financial development might lead to larger measured remittances either because financial development enables remittance flows or because a larger percentage o f remittances are measured when those remittances are channeled through formal financial institutions. In addition, financial development might lower the cost o f transmitting remittances, leading to an increase in such flows. Finally, omitted factors can explain both the evolution o f remittances and o f financial development, also leading to biases in the estimated impact o f remittances on financial development. Thus, obtaining valid estimates for equations (1)-(3) requires an estimation technique that can deal with measurement error, omitted variables and reverse causality. 49 There is anecdotal evidence that official estimates o f remittance flows in LAC have increased sharply during the last few years partly due to better reporting. At least for the case o f Guatemala, this was reported to be the case due to recently introducedregulatory changes related to reporting agencies. For more details see The World Bank, (forthcoming, 2006). 129 Following the approach pursued by Aggarwal et al. (2006) we present Instrumental Variables (IV) estimations where we use economic conditions - GDP per capita, real GDP growth, and the unemployment rate - in the top remittance-source countries (ie., the countries from which migrants send money) as instruments for the remittances flows received by the countries in our sample. Economic conditions in the remittance-source countries are likely to affect the volume o f remittance flows that migrants are able to send, but are not expected to affect financial development in the remittance receiving countries inways other than through its impact on remittances or through the effect on other variables we already control for like exports or capital flows. Because bilateral remittance data are largely unavailable, we identify the top remittance-source countries for each country in our sample, using bilateral migration data from the OECD's Database on Immigrants andExpatriates. This dataset identifies the top five OECD countries that receive the most migrants from each remittance-recipient c~untry.'~ Here we assume that these OECD countries receive the bulk o f the migrants from the countries in our sample and account for the majority o f the remittance flows sent to the countries in our sample. We construct three instruments by multiplying, respectively, the GDP per capita, the real GDP growth, and the unemployment rate, in each o f the top five remittance-source countries by the share o f migration to each o f these five OECD c~untries.'~To check the validity o f our set o f instruments, in all estimations, we report the Sargan test o f overidenditfying restrictions. The joint null hypothesis for this test is that the instruments are uncorrelated with the error term and that excluded instruments are correctly excluded from the estimated equation. Table 5.2 shows results from the instrumental variables estimations on the impact o f remittances on financial development including all developing countries together but allowing for country fixed effects and including time dummies to mitigate concerns about omitted variables. Sargan's tests o f overidentifying restrictions reported at the bottom o f the table confirm that the instrumentsused are valid in that they are uncorrelated with the error term. We present regressions including and excluding a dummy to control for periods o f domestic financial liberalization, since this variable i s available for a smaller set o f countries. We find that remittances have a positive and significant impact on both bank deposits and bank credit. A 1 percentage point increase in the share o f remittances to GDP results in approximately a 5 percentage point rise inbank deposits and credit to GDP. Other variables such as log o f GDP, GDP per capita, inflation and financial liberalization are also significant inthese estimations and have the expected sign. InTable 5.3 we report results for equation (2), where we allow for remittances to have a different impact among L A C and non-LAC countries by including an interaction between remittances and a dummy for Latin American countries. We find that remittances continue to have a positive impact on financial development, but this effect i s smaller among L A C countries. For this group o f countries, a one percentage point rise in remittances leads to at most a 4 percentage point increase in deposits and credit to GDP. However, in LAC, contrary to the case o f non-LAC countries,the impact o f remittances over credit i s somewhat larger than for deposits. 50http://www.oecd.ore/document/S110,2340,en 2825 494553 34063091 I 1 1 I,OO.html. Note that the bilateral migration data i s only available for 2000, so the weights we use are constant. The time variation arises from the series on the GDP per capita, real growth rate, and unemployment rate inremittance-source countries. 130 Table 5.4 shows results allowing remittances to have a different impact across LAC16, LACS, and LAC4 countries. For LAC16 and LAC4 countries we find that remittances have a positive impact on credit and deposits. Among LAC16 countries, a 1percentagepoint increase in remittances results in approximately a 3 percentage point rise in bank depositskredit, while for LAC4 countries this effect i s closer to 2 percentage points. On the other hand, among LAC5 countries remittances appear to have no statistically significant impact on financial development. In sum, all the estimations presented so far indicate that remittances have a positive impact on financial development, but the effect i s smaller for Latin American countries relative to other developing economies. What explains this finding? This i s a hard question to tackle and one for which a definite answer is not possible given the data available. Nevertheless, below we discuss some candidate explanations. First, the impact o f remittances on financial development might be smaller for Latin American countries if remittance recipients in these countries relative to recipients in other countries are less likely to use financial institutions to receive remittances. This could be because o f greater distrust o f these institutions by recipients in LAC. Though there i s no formal way o f testingthis hypothesis, the fact that crises have beenmore recurrent and severe inLatin America, as shown inTable 5.5, i s consistent with this possibility. 52 '*Another possible reason for mistrust, but which cannot be formally tested either i s the public perceptionof an excessive focus of banks on profits and lack of commitmentto support the needs of poorer clients. This is cited in Annex I1as reportedby banks interviewedin Colombia. However, it is unclear if this is a generalized perceptionin Colombia, if it is the case in other countries in LAC, or even if both o f these are true, if this is different from the case in other regions. 131 Table 5.2. PanelEstimatesof the Impact of Remittanceson FinancialDevelopment. InstrumentalVariablesFixedEffectsEstimates. Eco IBankDepositsto GDP Bank Credit to GDP Remittances to GDP 15.244 4.166 5.055 5.34 [5.29]*** [5.13]*** [5.331*** [5.39]*** Log o f GDP 30.293 30.229 35.086 45.328 [4.41]*** [4.25]*** [5.301*** [5.221** * GDP per capita 5.457 4.065 6.652 4.234 [2.61]*** [1.71]* [3.33]*** [1.46] [nflation -0.002 -0.002 -0.002 -0.002 [2.54]** [3.01]*** [1.69]* [1.631 Dual exchange rate 0.562 0.09 -1.767 -2.301 [0.35] [0.06] [1.16] [1.36] Other flows to GDP -0.022 0.02 -0.014 0.004 [0.60] [0.62] [0.39] [0.10] Exports to GDP 0.055 0.068 -0.069 -0.094 [0.84] [1,151 [1.11] [1.30] Financial liberalization 3.606 4.078 [2.72] *** [2,52] * * Observations 1150 910 1143 910 R-squared 0.63 0.68 0.52 0.51 Country dummies Yes Yes Yes Yes Time dummies Yes Yes Yes Yes Sargan test o f overidentifying restrictions 0.18 0.3 2.54 3.87 P-value for Sargan test 0.91 0.86 0.28 0.14 , iere FD refers to financi development measured as the ratio o fbank deposits and, separately, bank credit to GDP. Remittances to GDP is the share o f remittances to GDP. X is matrix o f controls including: GDPper capita, measured in constant dollars; Log of GDP, stated inconstant dollars; Inflation, defined as the percentage change in the GDP deflator; Dual exchange rates, which is a dummy capturingperiods whenmultipleexchange rates were ineffect; Financial liberalization, which is a dummy identifying periods o f liberalization indomestic interest rates; Otherflows to GDP, defined as foreign direct investment +Non-FDIprivate inflows + aid to GDP and Exports to GDP is the ratio o f total exports to GDP. GDP per capita, real GDP growth, and unemploymentrates inremittance-source countries weighted by migration are used as instruments. Time dummies are included but not shown. Absolute value o f t statistics in brackets, *, **, and *** denote significance at 10,5, and 1percent. 132 ttanccs ;I.Ob]* 1 1 .is] Rciiuttanccs in LAC 3.894 2.085 4.217 3-22? HO: Remittances for LAC=O 5.51 2.69 8.23 3.9 P-Value 0.02 0.10 0.00 0.05 t,h>i.r\:l1lh 1150 9 10 11-43 9ltJ Drefers 133 I 5 31 13.074 f 8, t 53 [1.873" /2.42]** [1.%Sf* LACf 6 x Rctnittnnccs to GDP -13 656 -9.618 -15 428 [1.721" -7.659 if,84]* 12 17]** [I731" 6.93 [a 06j*' -0,001 [ 1 441 [QEJ [0.77j -u924 [0.JJ] [1.O3] [1,Oit.j Othcr flows tu ODP -0.021 -0.054 -0 042 [0.30j [U.SZj lO.SO] -0 126 -0 192 -0 312 l0.671 [ 1.441 11 393 6.05 6,927 134 Table 5.5. Factors that Might Affi t the Iml ict of Rei ittances c Financi Development I' Latin Eastern EastAsia South East Middle East Sub- America Europeand and the and North Saharan Zentral Asir Pacific Asia Africa Africa Use of banking services by migrants in the US % migrants with a savings account in the U.S. ' 30.00 38.00 47.90 48.40 33.10 62.60 Crisesfrequency and severity Episodes pernumber of countries * 0.8 0.9 0.4 0.4 0.4 0.8 Average fiscal costs of crises (% of GDP) ' 20.1 8.1 16.7 n.a. n.a. 12.5 Average durationof crises episodes (years) 3.5 1.9 4.2 n.a. n.a. 1.5 Average output loss (% of GDP) 20.7 6.5 22.4 n.a. n.a. 3.3 Physical bankpresence Branchesper 1,000 square km 4.78 7.46 15.51 19.98 33.78 6.28 Number of branches per 100,000people 9.31 8.57 6.82 4.45 9.46 2.89 ATMs per 1,000 squarekm 10.29 16.08 71.59 3.17 50.81 14.28 Number of ATMs per 100,000 people 18.96 19.89 21.42 1.08 12.49 6.68 Cost of usingbanking services Fees associated with checking accounts (as a percentageof GDP per capita) 0.90 0.20 0.00 5.5 0.46 10.40 Fees associated with savings accounts (as a percentageof GDP per capita) 0.57 0.05 0.05 3.04 0 1.77 Average fees on loans (as apercentageof GDP per capita) 1.62 1.23 1.30 4.54 0.80 1.30 Legal rights and contract enforcement Legal RightsIndex 3.8 5.6 5.3 3.8 4.1 4.4 Time to enforce contracts (days) 461.3 393 406.8 385.5 432.1 438.5 ata sources: 'Osili, U.O.and A. Paulson, (2005); Caprio,G. and D. Klingebi (2003); Martinez Peria, M.S.,(2005); Doing Business:LegalRights Index reflectsthe legal rights of borrowersand lendersand measuresthe degree to which collateral ' 'HI P. and D. ingebiel(200 a Beck, T., Demirguc-Kunt,1 and bankruptcy laws facilitate lending. The index ranges from 0 to 10, with higher scores indicating that collateral and bankruptcy laws are better designedto expand access to credit. Time to enforce contracts measuresthe efficiency of the judicial (or administrative) system in the collectionof overdue debt. The time requiredfor disputeresolutionis recordedin calendar days, countedfromthe moment the plaintifffiles the lawsuit incourt until settlementor payment. Second, the impact o f remittances on financial development in Latin America might be smaller if remittance recipients inthese countries are less likely to receive remittances via banks, either because Latin American migrants do not use banks as much to send the remittances or because the beneficiaries do not use banks as much at the receiver end. Presumably, in countries where remittances are more frequently received via banks, recipients are more likely to open bank accounts and use other financial services. Also, in recipient countries where banks play a large role in the remittances business, banks are in a better position to learn about remittance recipients and are more likely to seek them out as clients for other products. As suggested by surveys o f remittances senders performed by the IDB, Banks respond for only 7 percent o f all remittances sent to Latin America, compared to 78 percent for money transfer operators and 11 percent o f transfers made through people traveling to home countries (see chapter 9). Furthermore, this hypothesis i s also supported by statistics on the percentage o f migrants in the U.S.that have bank accounts obtained from Osili and Paulson (2004). Their numbers, shown on Table 5.5, indicate that Latin American migrants residinginthe U.S. are less likely to have bank accounts when compared to migrants from other regions. One possible explanation for this result 135 might be related to the fact that migrants from LatinAmerica tend to be less educated, as shown in Chapter 2. Overall, the lower usage of bank accounts by migrants in the U.S.is noteworthy because it might suggest that Latin American migrants are less likely to send money back home through direct deposits in banks. Regarding the receiver end o f the transaction, IDB 2006 estimates percentages o f the distribution market for 13 countries inLatin America and calculates a regional average based on these data, adjusted by the volume o f remittances in each country. The findings suggest that banks are an important means o f remittance payment inhome countries (about 49 percent), only slightly smaller than all other alternatives combined. However, this report also indicates that a very small percentage o f the remittances paid by banks actually enter the financial system through existing or new accounts, acting these banks more as payment agents much like retail payers in small busines~es.~~ Third, remittances might not spur as much deposits or credit growth in Latin America if access to physical banking outlets i s more limited inthis region than inother countries. Table 5.5 shows data reported in Beck, Demirguc-Kunt, and Martinez Peria (2005) on the average number o f branches and ATMs per square kilometer and per 100,000 inhabitants across regions. While the presence o f physical banking outlets per capita is not consistently lower for Latin American countries relative to other regions, there is some evidence that geographic penetration i s low in Latin America. In particular, the number o f branches per area i s the lowest among all regions, suggesting that perhaps distance to the nearest branch is an obstacle for remittance recipients to demand and use financial services inLatinAmerica. Fourth, the impact o f remittances on financial development in Latin America might be smaller than that observed for other countries if the costs o f banking are larger in the region. Recently, collected data from Beck, Demirguc-Kunt, and Martinez Peria (2006) suggest that the costs o f maintaining a bank account and the fees associated with loans are higher in Latin America than inmost countries with the exception o f some in South East Asia and inAfrica. Finally, even if the supply o f loanable funds increases with remittances, credit might not rise in Latin America due to weaker creditor protection and poorer contract enforcement. Legal rights rankings and statistics on the number o f days to enforce a contract from the World Bank's Doing Business (2005) dataset shown on Table 5.5 reveal that Latin America ranks below all other regions along these dimensions. The data presented in Table 5.5 offer some support for the validity o f the factors discussed above in explaining why remittances might have a smaller impact on financial development in Latin America vis-a-vis other regions. However, further research is requiredto validate this tentative evidence and to establish the degree to which each o f these factors i s important. 111.Micro-level analysisof the associationbetweenremittancesand financial development While it i s useful to investigate the relation between remittances and financial development using macro-level data, as discussed in chapter 3 cross country analysis have 53At least for the cases analyzed through ad-hoc surveys and field interviews, exclusivity agreements o f banks with money transfer operators have hindered the banks' capacity to convert these recipients into clients (see Annex I). 136 important limitations. In what follows we use household-level survey data to investigate the association between remittances and financial development. Here we equate financial development with greater use o f financial services. First, we compare some simple statistics on the proportion of households with deposit accounts and loans among remittance recipients and non-recipients in Latin America. Second, we look deeper into the link between remittances and the use o f financial services inthe region by conducting case studies on El Salvador and Mexico. Both countries are among the largest recipients o f remittances in Latin America and the data available inboth cases is detailed enough to allow us to correct for biases resulting from omitted variables and endogeneity. III.1.Evidenceporn household-level tests Using household surveys for 11 Latin American countries - Bolivia, Dominican Republic, Ecuador, El Salvador, Guatemala, Haiti, Honduras, Jamaica, Nicaragua, Peru, and Suriname - we now'investigatewhether the use o f financial services differs between households that receive remittances and those that do not. In particular, we examine the proportion o f households that have a deposit account and the share o f those that have outstanding bank credit. For comparison, we also report on the share o f households with non-bank credit among remittance recipients and non-recipients. Table 5.6 shows summary statistics for a total o f 19 surveys conducted inthe 11 countries mentioned above. Data on bank accounts i s available in 13 o f the 19 surveys, information on bank credit outstanding i s contained in 16 o f the surveys, and there are 12 surveys that include data on non-bank credit. The proportion o f households with bank accounts for remittance recipients exceeds that for non-recipients in 11 out the 13 surveys for which deposit information i s available. However, these differences are statistically significant in only 5 surveys and 4 countries.54Inthe case o f the proportion o f households that receive credit, we find that ratios are higher among remittance recipients in9 out o f 16 surveys. Differences are statistically significant in the right direction in only four cases. Finally, the proportion of remittance recipient households with credit outstanding from non-bank sources exceeds that for non-recipients in 7 out o f the 12 surveys. However, these differences are only significant inone case. All in all, household level data provides some evidence consistent with the hypothesis that the use o f financial services i s more prevalent among remittance recipient households. This i s particularly the case for deposit holdings and less so for credit. However, this evidence needs to be taken with a grain o f salt for at least two important reasons. First, the tests conducted do not control for other households characteristics that might account for differences in the use o f financial services. Second, these simple statistics suggest a correlation between remittances and the use o f financial services, but are inno way proof o f causality. 54 In the case of Honduras, the ratios come from Orozco (2005) and he does not report a test for differences across recipients and non-recipients. 137 To mitigate at least the first problem, we use household survey data from three countries which have been analyzed elsewhere in the report - Guatemala, Haiti and the Dominican Republic - to estimate simple "probit" models o f the probability o f having deposit accounts, outstanding bank credit and non-bank credit, controlling for household income and other household characteristics - i.e. age o f the household head, educational attainment o f adults, household size and composition, urban status o f the locality and province "dummy" variables. The estimates for the impact o f remittances are shown in table 5.7, which also reports the coefficients o f variables representing the 2ndthrough 5'h quintiles o f the income distribution - using counterfactual pre-migration incomes estimated in chapter 3. In the three countries the results show that, as expected, the probability o f having deposit accounts increases monotonically with income, and so does the likelihood o f having outstanding bank credit - except in the case o f Guatemala. Moreover, within given income quintiles, households with access to remittances are significantly more likely to report deposit accounts. However, at least in Haiti and the Dominican Republic, recipients are also less likely to have bank and non-bank outstanding credit suggesting a role for remittances to relax financing constraints. Country Guatemala DominicanRepublic Haiti Dependent Bank Bank Non-Bank Bank Bank Non-Bank Bank Bank Non-Bank Variable Deposits Credit Credit Deposits Credit Credit Deposits Credit Credit Remittances 0.114*** -0.011 0.013 0.078';" -0.022*** -0.012 -0.001 ** -0.022*** (0.020) (0,009) (0.018) , (0.01I) (0.007) (0,008) (0.000) (0.007) Countefactual Income Quintiles Q2 -0.016 -0.022"' 0.016 0.084"' 0.017 0.029*** 0.036*** -0.001 0.018* (0.016) (0.007) (0.016) (0.019) (0.012) (0.01 I) (0.015) (0.001) (0.012) Q3 0.033* -0.012 0.022 0.160*** 0.074*** 0.032*** 0.040*** 0.001 0.024** (0.018) (0.008) (0.017) (0,020) (0.015) (0.012) (0.015) (0.001) (0.012) 4 4 0.116*** -0.014 0.020 0.235*** 0.090*** 0.033*** 0.075*** 0.001 0.047"' (0.021) (0.008) (0,018) (0,022) (0,016) (0.013) (0.019) (0.001) (0.014) Q5 0.238*** 0.018* 0.008 0.378*** 0.142*** 0.008 0.194*** 0.005*** 0.040*** I (0.028). . (0.012) (0.020) (0.024) . . (0.019) . . (0.013) . . (0.030) (0.003) (0.016) level. The above results are complemented inwhat follows with a more detailed analysis o f the relationship between remittances and the use o f financial services for two o f the largest remittance recipients inLatinAmerica: El Salvador and Mexico. The datasets we use allow us to control for a variety o f factors that might also influence the use o f financial services. Finally, the analysis we conduct tackles the concern that remittances might be endogenous by conducting instrumental variables estimations. 111.2Evidenceporn two casestudies: El Salvador and Mexico 111.2.1El Salvador" El Salvador is Latin America's fifth largest recipient o f remittances in dollar terms and the third largest as a share of GDP. Duringthe last decade, remittances have averaged almost 1.4 ''The ~ analysis on El Salvador is based on research by Demirguc-Kunt and Martinez Peria (2006) 139 billion dollars or 14 percent o f GDP. El Salvador i s also among the most financially developed countries in the region. In 2003, deposits and private bank credit in this country reached 42 percent and 39 percent o f GDP, respectively. To investigate the relationship between remittances and financial development for El Salvador, we use data from a nationally representative rural panel survey conducted by the Fundacion Salvadoreiia para el Desarrollo Economico y Social (FUSADES), and the Rural Finance Program at the Ohio State University. The panel data set i s composed o f four biennial observations for 1995, 1997, 1999, and 2001. The surveys include 738 households in 1995, 623 in 1997, 697 in 1999, and 690 in 2001. There are 719 households that were surveyed twice or more times and 451householdsthat appear inall four surveys. The surveys identify households that receive remittances and contain information about whether members o f the household use banking services. Inparticular, the surveys ask whether households have a bank account and whether they solicited bank loans and have outstanding bank credit. The surveys also contain information about the remitters. Inparticular, they identify the length o f time that they have been outside o f El Salvador. Finally, the survey contains other socio-economic data about the households including information on households' income, housing, education, and occupation. Table 5.8 provides detailed definitions for the variables used from the FUSADES surveys and reports descriptive statistics. Our empirical analysis focuses on estimating different variants o f equations (3), (4), and (5) below: Credit outstandingi,t poi+p1 Remittancesi,t+P~'X,t+~i,t (4) Credit applicationit = 6oi+ 61Remittanceqt+ 62t4,t+~i,t (5) iisthehouseholdidentifierandtrepresentsthetimeperiod. Deposit i s a dummy that equals 1 if the household has a bank deposit account and 0 otherwise. Credit outstanding is a dummy that takes the value 1 if the household has any outstanding loans and 0 otherwise. Credit application refers to a dummy variable that equals one ifthe household applied for a loan inperiod t. Inall three equations, Remittances is a dummy variable that identifies those households that receive international remittances. X is a matrix o f household level controls, including household income, educational level o f the head o f household, the number of rooms in the house, and a dummy for whether at least one member o f the household is employed in agricultural activities. Because all three dependent variables are discrete, equations (3)-(5) are estimated as probit models. Panel A in Table 5.9 shows probit results for the likelihood that a household has a bank account. We report marginal effects as opposed to coefficients. We include estimations for households that appear at least twice and, separately, four times in the four year database. Our estimations indicate consistently that households that receive remittances have a higher 140 likelihood (between 0.12 and 0.16 percentage points higher) of owning a deposit account regardless of which householdcharacteristics we control for and independentlyof whether we focus on householdsthat appear inat least two or inall four ofthe surveys. To investigate whether remittance recipients are more likely to use bank credit we distinguish between the likelihood of applying for a loan during the survey period and the probability of havingoutstanding credit. Panel B presents the results for the likelihood of having outstanding debt. Once again, we report results for households that were surveyed two or more times and for those includedinall four surveys. Consistentlyacross all estimations, we find that the likelihood of having outstanding bank debt does not seem to be affected by whether households receive remittances. The dummy for whether the householdreceives remittances is always insignificant.As shown inthe appendix, this is also true if we focus on non-bank credit instead. Table 5.8. Data Description and Summary Statistics: El Salvador Case Study on the Impact of Remittances on the Use of Financialt :rvicc - - VariableName Variable Definitions Source Obs. Mean Std. Dev. Min Max International Dummy equals to 1 indicates Encuestaa familias -- - 2,715 0.20 0.40 0 1 Remittances that the household receives salvadorefias,Fundaci6n international remittances Salvadorefiaparael DesarrolloEconomicoy Social (FUSADES) Bank Deposits Dummy equals to 1 indicates Idem 2,747 0.17 0.38 0 1 that the household has bank deposits Bank Credit Dummyequals to 1 indicates Idem 2,748 0.10 0.30 0 1 that the household has outstanding bank credit Application for Dummy equals to 1 indicates Idem 2,748 0.11 0.32 0 1 Bank Credit that the household has applied for bank credit HouseholdIncome Total Income for Household Idem 2,696 14.88 17.44 0 225 (inthousands ofcolones) Agricultural Dummyequals to 1 indicates Idem 2,748 0.54 0.50 0 1 Employment that there is at least one member of the household working inagricultural activities Rooms in Total rooms inhousehold Idem 2,7 19 1.74 0.91 1 7 Household Household Educational level o f the head Idem 2,720 2.78 3.13 0 16 Schooling- o fthe household udy analysis o f the linkbetween rei ttances and the use of financial services in El Salvador. Data comes from a conducted by Fundacion Salvadorefia para el Desarrollo Economico y Social (FUSADES) in 1995, ,1997, 1999, and 2001. Approximately 700 households were included each year. Resultso f the analisis are show on tables 8 and 9. 141 Results for the likelihood o f applying for a bank loan on the year o f the survey (shown on Panel C o f Table 5.9) are similar to those discussed above for the probability o f having outstanding credit. Once again the dummy for remittance recipients is never significant. That is, independently o f whether we look at households that are present two or four years inthe sample, and regardless o f which type o f household characteristics we control for, we find that remittance recipients are as likely to have outstanding debt as those that do not receive remittances. The estimates reported so far on whether households that receive remittances are more likely to use financial services might be biased if there are omitted relevant household characteristics or if the likelihood that a household receives remittances i s endogenous. To address these concerns, Table 5.10 presents estimations including household fixed effects and, separately, instrumenting for remittances. Including household fixed effects should help mitigate the concern that the association between remittances and the use o f banking services is driven by omitted household characteristics. However, because for many households there i s no variation intheir status as remittancerecipients or as users ofbanking services, the number ofobservations when we include fixed effects is smaller since those households for which there i s no change are dropped from the estimations. Fixedeffects estimations yield very similar results to those reported above. We find that remittance recipient households are more likely than non-recipients to have a bank account but are no different when it comes to the likelihood o f having outstanding debt or applying for a loan. The fact the FUSADES surveys contain information about the migrants who sends the remittances allows us to instrument for the likelihood that a household receives remittances. In particular, we speculate that the length o f time since the migrants left El Salvador i s likely to influence whether households receive remittances, but it i s not likely to independently affect the households' use o f banking services. Thus, we use the number o f years migrants have resided abroad and the number o f years squared (to allow for a non-linear relationship between length o f receives remittance^.'^ stay overseas and sending remittances) as an instrument for the likelihood that a household The instrumental variable estimations confirm the results discussed above. Namely, we find that remittance recipient households are more likely than non-recipients to have a bank account, but the probability that recipient households apply for and have outstanding credit i s not different for remittance recipient and non-recipient households. What role does the way in which remittances are received play in the results reported so far? The surveys for El Salvador contain information as to whether remittances are received via banks, money transfer operators, relatives, etc for two years: 1999, 2001. Table 5.12 shows 56Basedon a survey of Latin American migrants in the U.S. conducted in November 2001 by Bendixen and Associates, Orozco (2003) reports that there is a non-linearrelationshipbetweenremittancessent and the lengthof stay in the U.S. Inthe first years of arrival, migrants are less likely to remit partly becauseof their limited income andbecauseof obligationsincurredwhen arrivinginthe country.As immigrants settleover time, they sendmore, in part because they earn more. However, after an extended period of time, remittances decline in part because of greater obligations faced by the migrants in the host country and because ties to the country of origin weaken in some cases. 142 results for the likelihood o f having bank accounts, having outstanding credit, and soliciting loans where we interact the remittance recipient dummy with a dummy indicating whether remittances were received through the banking sector. The results show that the likelihood that remittance recipients have bank accounts i s twice as large if remittances are channeled through the banking sector. Onthe other hand, no such effect i s present on the credit side. that receive funds via banks - are more likely to open bank accounts relative to other individuals. Overall, the findings for El Salvador indicate that remittance recipients - especially those However, when it comes to credit, we do not observe any differences in credit application and usage by remittance recipients and non-recipients. Since in El Salvador the financial system (banks and non-banks) i s relatively deeper than inother L A C countries, more research i s needed to explore if there are other limitations to ease the typical constraints to broaden access to credit for segments o f the population that have limited credit payment history and insufficient collateral, o f ifthese constraints are different for remittance recipients. III.2.2Mexico '' Mexico i s Latin America's largest recipient o f remittances in dollar terms so the potential for remittances to impact financial development i s perhaps largest in this country. Though remittances are 2 percent o f GDP, they amount to approximately 10 percent o f banking sector deposits and credit. Also, in recent years both Mexican and foreign banks have taken steps to enter the remittance business and to cross-sell products to remittance senders and recipients, a move that if successful should be reflected ingreater financial de~elopment.~~ To analyze the relationship between remittances and financial development for the case o f Mexico, we combined information from the 2000 Mexican Census on the share o f households across Mexican municipalities that receive remittances with municipal level information from the Comisi6n Nacional Bancaria y de Valores (CNBV) on the size o f the commercial banking sector, and use o f their services across municipalities. Inparticular, we collected information on the number of commercial bank branches, the value and number of commercial bank deposits, and the amount o f commercial bank loans outstanding. Finally, to control for other municipal level characteristics that might affect financial development, we also collected information for each municipality on GDP per capita, the average years o f schooling among household heads, the percentage o f households with access to piped water, the percentage o f households with a telephone line, the share households with at least one member employed inagricultural activities, and the percentage o f the population within the municipality residing in rural localities (those with less than 2500 people). Table 5.12 provides definitions, data sources, and descriptive statistics for all the variables included inthe Mexico case study. 57This section relies heavily on the analysis conducted by Demirguc-Kunt, Lopez Cordova, Martinez Peria, and Woodruff (2006). See Hernandez-Coss (2004) and Orozco (2004) for a description o f efforts by US. banks to penetrate the remittance business to Latin America. 143 Table 5.9. Are Remittance Recipients More Likely to Use Banking Services? Probit - estimations for ElSalvador PanelA: Likelihoodofhavingabank account Including households surveyed two or more times Includine households surveyedfour times InternationalRemittancesRecipient 133 0.149 0.133 0.122 0.146 0.147 ,143 0.145 0.141 0.130 0.160 0.142 60'"' 7.11". 6.60*** 6.09*** 7.16*** 6.91*** ,73*** 5.45"' 5.66*** 5.17*** 6.17*** 4.98*** HouseholdIncome 0.004 0.003 0.004 0.002 7.35*** 4.93'1' 5.60*** 3.80*** Agricultural Employment -0.002 0.018 0.0I 7 0.027 0.13 1.17 0.90 1.33 Rooms in Household 0.063 0.041 0.061 0.040 7.93"' 5.02*** 6.19*** 3.79*** Household Schooling 0.027 0.022 0.030 0.023 12.05*** 9.30*** 10.44*** 7.40*** Observations 197 2446 2497 2474 2472 2396 700 1576 1700 1632 1632 1452 PanelB: Likelihood of ha1 g.outstandingbank debt Including households surveyed two or moretimes Includinghouseholds surveyedfour times 1 InternationalRemittances Recipient 1.004 0.001 -0.003 -0.003 -0.003 0.002 0.006 0.008 0.008 0.007 0.006 0.011 25 0.07 0.22 0.21 0.20 0.16 0.30 0.38 0.39 0.36 0.31 0.53 HouseholdIncome 0.001 0.000 0.001 0.000 2.20" 1.25 1.35 0.58 Agricultural Employment -0.014 -0.009 -0.025 -0.006 1.20 0.71 1 1.63 0.36 Rooms in Household A An? 2.08'. 1.10 0.94 0.14 HouseholdSchooling 0.006 0.005 0.008 0.007 3.34*** 2.73*** 3.36*** 2.71"' Observations 199 2448 2499 2476 2474 2398 1700 1576 1700 1632 1632 1452 PanelC: Likelihood of plyingfor a bank loan Including households surveyedtwo or more times Includinghouseholdssurveyedfour times InternationalRemittances Recipient 1.015 -0.01 -0.014 -0.015 -0.014 -0,010 0.002 -0.003 0.000 -0.003 -0,002 -0.003 94 0.63 0.90 0.98 0.89 0.63 1.10 0.15 0.00 0.15 0.09 0.13 HouseholdIncome 0.001 0.000 0.000 0.000 1.90' 0.99 0.98 0.21 Agricultural Employment -0.017 -0.012 -0.029 -0,008 1.36 0.95 1.80' 0.46 Rooms in Household 0.009 0.003 0.006 0.000 1.41 0.50 0.69 0.02 HouseholdSchooling 0.007 0.006 0.009 0.008 3.42*** 3.00*** 3.44*** 2.90*** Observations 199 2448 2499 2476 2474 2398 ,,700 1576 1700 1632 1632 1452 The table shows PI data is describedin Table 5.7. Marginal effects are shown instead oicoefficients. *, **, *** denotesignificance at 10, 5, and 1, percent. 144 Probi and InstrumentalVariablesProbitEstimations for ElSalvador PanelA: Likelihood of havingabank account Fixedeffects estimations Instrumentalvariablesestimations InternationalRemittances Recipient ,285 0.287 0.284 0.272 0.285 0.274 231 0.243 0.231 0.217 0.236 0.234 .06*** 4.99*** 5.03"' 4.78*** 4.97*** 4.62*** 14". 6.22*** 6.12*** 5.72*** 6.29*** 5.91*** HouseholdIncome 0.003 0.003 0.004 0.003 2.28" 2.25** 7.33*** 4.95*** Agricultural Employment 0.052 0.056 -0.002 0.020 1.28 1.32 0.11 1.26 Roomsin Household 0.054 0.042 0.064 0.041 2.07** 1.59 7.77"' 4.87*** HouseholdSchooling 0.01 0.009 0.027 0.022 0.90 0.74 11.92*** 9.13"' Observations 05 882 905 892 887 852 460 2411 2460 2437 2435 2361 PanelB: Likelihoodof hr ngoutstandingbank debt Fixedeffects estimations Instrumentalvariablesestimations InternationalRemittances Recipient 3.056 -0.045 -0.058 -0.029 -0.048 -0.008 ).010 -0.005 -0.009 -0.Q15 -0.016 -0.097 .83 0.65 0.85 0.41 0.69 0.11 .39 0.18 0.36 0.57 0.65 0.59 Household Income 0.002 0.002 0.001 0.003 1.33 1.28 2.19** 1.20 Agricultural Employment 0.035 0.033 -0.013 -0.041 0.73 0.66 1.08 0.58 Rooms in Household 0.016 0.017 0.012 0.042 0.48 0.50 2.04" 1.07 Household Schooling -0,029 -0.03 0.006 0.030 1.99** 1.96** 3.32*** 2.74*** Observations 53 640 653 627 642 603 462 2413 2462 2439 2437 2363 PanelC:Likelihood of iplying for a bank loan Fixedeffects estimations Instrumentalvariablesestimations InternationalRemittances Recipient 3.071 -0.056 -0.073 -0.054 -0,064 -0.031 1.039 -0.035 -0,038 -0.042 -0.045 -0.043 .08 0.84 1.1 I 0.80 0.97 0.44 .41 1.23 1.38 1.53 1.65* 1.54 Household Income 0.001 0.001 0.001 0.000 1.13 1.07 1.79* 0.87 Agricultural Employment 0.031 0.035 -0.016 -0.011 0.67 0.74 1.22 0.82 Roamsin Household -0.007 -0.008 0,009 0.004 0.21 0.26 1.43 0.54 HouseholdSchooling -0.011 -0.009 0.007 0.006 0.80 0.61 3.31"' 2.94". Observations 19 703 719 689 708 662 ,&462 2413 2462 2439 2437 2363 The table shows fixed effects probit and instrumental variables probit results for the likelihood that remittance recipients in El Salvador use banking services. The data is described in Table 5.7. Marginal effects are shown : 145 Table 5.11. Does the Way inWhich Remittances Are TransferredAffect The Use of BankingServices? - Probitestimationsfor ElSalvador nternationalRemittances Impact .115 0.146 0.115 0.109 0.133 0.149 3.95]*** [4.79]*** [3.93]*** [3.73]*** [4.48]*** [4.86]*** nt Remx Remsent throughbanks .192 0.179 0.195 0.161 0.168 0.166 2.61]*** [2.22]** [2.65]*** [2.171 ** [2.191** [1.961** ncome 0.005 0.004 [7.801*** [5.22]*** \gricultural Employment -0.029 -0.005 [1.231 [0.20] iooms in Household 0.062 0.039 [5.17]*** [3.15]*** -Iousehold Schooling 0.034 0.027 [9.301*** [7.13]*** Ibservations 372 1335 1372 1372 1359 1322 PanelB:Likelihooc )fhaving an outstandingbank debt when money is sent through banks 'nternationalRemittancesImpact 0.013 -0.005 -0.013 -0.014 -0.0 12 -0.007 0.731 [0.30] [0.74] [0.77] [0.67] 10.411 :nt Rem x Rem sent throughbanks 1.055 0.05 1 0.056 0.052 0.040 0.042 1.141 [1.021 [1.15] [1.071 [0.86] [0.85] .ncome 0.001 0.001 [2.75]*** [1.541 4gricultural Employment -0,007 0.002 [0.46] [O. 171 Zooms inHousehold 0.007 0.002 [1.091 [0.23] HouseholdSchooling 0.009 0.007 [4.301*** [3.521 *** ~~ 3bservations 373 1336 1373 1373 1360 1323 Panel C: Likeli lod of applying for abank loan when money is sent through banks [nternationalRemittancesImpact 0.018 -0.011 -0.018 -0.019 -0.0 17 -0.013 0.981 [OS61 [0.99] [1.021 [0.92] [0.67] [ntRemx Remsent throughbanks 1.053 0.049 0.053 0.050 0.037 0.040 1.041 [0.94] [1.061 [0.98] [0.76] [0.77] [ncome 0.001 0.001 [2.65]*** [1.441 Agricultural Employment -0.006 0.003 [0.40] [0.21] Rooms in Household 0.007 0.001 [1.01] [0.15] Household Schooling 0.009 0.008 [4.46]*** [3.72]*** Observations ,.373 1336 1373 1373 1360 1323 g for an interaction effect with a dummy for whether remittances are received via banks. The data is described in Table 5.7. Marginal effects are shown instead o f coefficients.*, **, *** denote significance at 10,5, and 1, percent. 146 We examine the impact o f remittances on financial development across Mexican municipalities by estimating equations (6)-(9) below: #DepPerCapitai=Po + PlRemittancesi+ Pz'Xi+ ui (7) BranchesPerCapitai=$o+ 41Remittancesi + $;Xi + Wi (9) DepGDP refers to the share o f the value o f commercial bank deposits expressed as a percentage o f municipal GDP. #DepPerCapita stands for the number o f commercial bank deposit accounts per capita. CreditGDP i s the share o f the value o f commercial bank loan outstanding to GDP. Finally, BranchesPerCapita i s the ratio o f the number o f commeicial bank branches to the population ineach municipality. Inall four equations, Remittances is the percentage ofhouseholds inthe municipality that receive remittances from abroad and the matrix X stands for the municipal level controls discussed above, including municipal GDP per capita, the rurahrban nature of the municipality , the level o f education, and the degree o f access to physical infrastructure (water and telephones). Table 5.13 presents results for the share o f deposits to GDP and the number o f deposit accounts per capita. We find that municipalities where a larger percentage o f the population receives remittances also tend to have higher ratios o f deposit accounts per capita and deposit amounts to GDP. A one standard deviation increase in the share o f households that receive remittances leads to, approximately, a 0.14 standard deviation change in ratio o f deposits per capita and a 0.2 standard deviation change inthe share o f deposits to GDP. Table 5.14 shows the results for the number o f branches per capita. Regardless o f which municipal level variables we control for, we find that municipalities where a larger share o f the population receive remittances tend to have a higher number o f branches per capita. A one standard deviation change in the share o f households that receive remittances brings about approximately a 0.15 standard deviation change in the ratio o f branches per capita. Contrary to the findings on deposits and branches, results on Table 5.15 indicate that there does not appear to be a significant association between remittances and credit to GDP across municipalities. The estimates discussed so far for Mexico would be biased and inconsistent in the event that remittances are endogenous or there are unobserved municipal characteristics. To correct for this possibility, Table 5.16 reports results where we instrument for remittances. Inparticular, we use two different variables as instruments: namely, (i) short-run deviations in rainfall (measured as rainfall for 1999 minus the average historical rainfall for most o f the 20thcentury relative to the standard deviation o f rainfall during the last century) and (ii)the sum o f the distance o f each municipality to the railroad network in existence during the 1920s plus the distance from that point to the US-Mexico border. The rational for including short-run deviations inrainfall resides on the fact that variations in rainfall can have a negative impact on households' income and 147 increase the need for remittances. On the other hand, it i s less likely that short-run shocks to municipal income will affect financial development directly. Table 5.12. DataDescriptionand Summary Statistics: Mexico Case Study on the Impact of Remittan es on FinancialD velopment Variable Name Variable Definitions Source Obs. Mean Std. Dev. Min Max Remittance- Percentof households INEGI, XI1Censo 2,427 6.56 7.72 0.00 53.71 receiving receivingremittances Generalde Poblaci6n households y Vivienda 2000, Cuestionario Ampliado Rural Localities Percentof individuals Idem 2,427 62.19 35.62 0.00 100.00 residinginlocalities<2500 Household Average years of school in Idem 2,427 4.48 1.63 0.00 13.57 Schooling the population 15 years and older Indigenous Percentof householdheads Idem 2,427 23.99 35.45 0.00 100.00 Language speaking an indigenous language Householdwith Percento f householdswith Idem 2,427 72.92 23.86 0.00 100.00 Piped Water pipedwater Householdwith Percentage of households Idem 2,427 12.17 13.39 0.00 80.3 1 Telephone with telephone Householdwith Percentage of households Idem 2,427 36.91 22.85 0.00 97.38 Agriculture with at leastone member in agriculture Bank Deposits Deposits to GDP (%) Comisi6nNacional 2,262 0.07 0.66 0.00 28.88 Bancariay de Valores Bank Credit Credits to GDP (%) Idem 2,383 0.02 0.22 0.00 6.86 Bank Accounts Bank Accounts to Idem 2,415 54.35 174.49 0.00 3,717.58 Population(1,000 inhab.) Bank Branches Bank Branchesto Idem 2,254 0.24 0.76 0.00 17.76 Population(10,000 inhab.) GDP per capita Municipal income per CONAPO 2001 2,426 3.43 2.62 0.15 32.88 capita as estimatedby CONAPO Distance to US Distance to USBorder or LopezCordova, 2,427 884.14 364.23 0.50 2,162.73 Border Distance to USBorder + Emesto, (2004) Distance to Railroad Rainfall standard Rainfall standarddeviation Idem 2,427 0.02 0.25 (0.83) 0.81 deviation The table descri IS the data used in the em rical analysis o f the ipact of remittances on financial development across Mexican municipalities. See tables 11 2, and 13for empirical iesults. 148 r * L * * * * * * * * * * 3 & m * 8 s O L O b so9 O C ? C ** ** * * Table 5.14. The Impactof Remittanceson BankBranchesAcross MexicanMunicipalities. OLS Estimations Clusteredby State BranchesPer Capita HouseholdsReceivingRemittances 0.0061 0.0081 0.0060 0.0033 0.0083 0.0052 0.0075 [2.45]** [2.77]*** [2.47]** [1.79]* [2.82]*** [1.90]* [2.741 ** GDP per capita 0.0586 [9.35]*** IndigenousLanguage -0,0015 [3.62]*** Householdswith Telephone 0.0167 [10.55]*** HouseholdSchooling 0.0796 [5.17]*** Householdswith Agriculture Labor -0.0037 [6.131** * Householdswith PipedWater 0.0012 [4.11]*** Rural Localities -0.003 -0.0049 -0.0013 -0.0031 -0.0041 -0.005 [3.23]*** [5.11]*** [2.1I]** [3.80]*** [4.81]*** [5.501 *** Constant 0.1543 0.1346 0.5011 0.0574 -0.0138 0.5565 0.3757 [3.93]*** [1.81]* [6.97]*** [1.30] [0.14] [7.36]*** [5.591*** Observations 2231 2230 2231 2231 2231 2231 2231 R-squared 0.01 0.33 0.25 0.42 0.30 0.27 0.24 OLS estimates of the impact oj i e percentage of households receiving remittances on bank branches in Mexico. Robust t-statistics are shown from estimations where standard errors are clustered by Mexican state. *, **, - - ~ *** denotes significanceat 10,5, and 1percent. 150 Table 5.15. The ImpactofRemittanceson BankCreditAcross MexicanMunicipalities. OLS Estimations Clusteredby State Creditto GDP Households ReceivingRemittances -0,0001 0.0000 -0.0001 -0.0003 0.0000 -0.0001 0.000 [1.25] [0.02] [1.47] [3.13]*** [0.37] [2.23]** [0.47] GDP Per Capita 0.0029 [7.08]*** IndigenousLanguage -0.0001 [3.12]*** Householdswith Telephone 0.001 [8,801*** HouseholdSchooling 0.0051 [4.41]*** Householdswith Agriculture Labor -0.0002 [4.63]* ** Householdswith PipedWater 0 01 [4.33]*** Rural Localities -0.0001 -0.0002 0.0000 -0.0001 -0.0002 -0.0002 [2.63]** [4.68]*** [0.26] [3.15]*** [4.58]*** [4.98]*** Constant 0.0091 0.0066 0.0242 -0.0009 -0.0083 0.0276 0.0165 [3.93]*** [1.59] [5.95]*** [0.41] [1.49] [6.14]*** [4.94]*** Observations 2359 2358 2359 2359 2359 2359 2359 R-squared 0.00 0.18 0.12 0.27 0.19 0.14 0.12 OLS estimates of the impact o the percentage of households receiving remittances on bank credit in Mexico. Robust t-statistics are shown from estimations where standard errors are clustered by Mexican state. *,**, *** denotes significanceat 10,5, and 1percent. 151 * * * * * * * I C - * * * * N 2 * * * * * The sum of the distance to the border plus distance to the 1920s railroad is used as an instrument for remittances because previous studies have found that in large part the migration patterns we observe today that give raise to remittance flows are driven by historical migration patterns (see Woodruff and Zenteno, 2001, Hanson and Woodruff, 2003, McKenzie and Rapoport, 2004, and Lopez-Cordova 2004 among others). Historical migration flows in Mexico were heavily affected by municipalities' proximity to the railroad and to the U.S.border. Thus, we assume that proximity to the railroad and border fostered historical and, hence, current migrations from Mexico and through this effect influence the share o f households that receive remittances today. By using distance as an instrument we are implicitly assuming that this variable does not affect financial development other than through. its effects on remittances. Since it i s possible that the establishment o f the railroad and/or proximity to the U.S.may have fostered economic growth in some municipalities, causing them to be more financially developed today, we also present estimations where we eliminate municipalities in close proximity to the railroad or the border -those within 40 or 50 kilometers. We present instrumental variables (IV) estimations inTable 5.16. Inparticular, we report results controlling both for GDP per capita and the urbadrural character o f the m~nicipality.~' The first-stage F-statistics for the significance of the instruments in explaining remittances are reported at the bottom o f the table along with Hansen's tests o f overidentifying restrictions. Both the high F-statistics and the fact that we cannot reject the null o f overidentifying restrictions suggests that our instrumentsare valid. The IV results are generally consistent with what we reported above. In particular, we find that the share o f households that receives remittances has a positive impact on bank deposits (both interms o f numbers and amounts) and bank branches across municipalities, but in2 out o f 3 specifications does not appear to influence the extent o f credit market development - the exception being the specification with all municipalities, for which the quality o f the instruments based on distance to the railroadmay be more questionable. IV. Conclusions andpolicyrecommendations Using both macro and micro data, this chapter investigated the impact o f remittances on the development o f the banking sector and the use o f banking services. At the macro level, the results indicate that remittances have a positive impact on financial development (considering both deposit and credit services) inLAC, but the effect i s lower than inother developing regions. At the micro level, we find some evidence consistent with the hypothesis that the use o f financial services i s more prevalent among remittance recipient households, particularly for deposits. However, the impact for L A C countries at the macro level is larger for credit than for deposits, contrary to the case for non-LAC countries. To the extent that this relatively larger increase in credit as a result o f remittances flows has not resulted in broader access to credit for remittance beneficiaries, as summarized below, this impact also has potentially important distributional effects that would be larger for L A C countries. 59However,results do not change ifwe includeother municipal level controlsinsteadof these ones. 153 At the micro level, based on case studies for El Salvador and Mexico for which relevant data were available we found similar results on the deposit side, even when controlling for household characteristics) but not for credit. For the case o f El Salvador, overall findings indicate that remittance recipients are more likely to open bank accounts relative to other individuals, specially for households that receive funds via banks (the impact i s twice as large if remittances are channeled through the banking sector), but no evidence is found in credit application and usage o f credit by remittance recipients and non-recipients. For the case o f Mexico, results indicate that municipalities where a larger percentage o f the populationreceives remittances also tend to have higher ratios o f deposit accounts per capita and deposit amounts to GDP. Also, regardless o f which municipal level variables we control for, we find that Mexican municipalities where a larger share o f the population receives remittances tend to have a higher number o f branches per capita. Results based on IV estimations to account for the possibility o f endogeneity o f remittances or unobserved municipal characteristics are generally consistent with these findings. That is, the share of households that receives remittances has a positive impact on bank deposits (both in terms o f numbers and amounts) and bank branches across municipalities. On the credit side, in2 out o f 3 specifications the share o f households that receives remittances does not appear to influence the extent o f credit market development (with the exception having the lowest quality o f instruments employed for the estimation). Similarly, the Colombian and Guatemalan descriptive case studies reported in annex 1based on ad-hoc surveys and follow up interviews o f banks show that these banks have been thus far much more effective at selling savings products than credit to remittance recipients. In terms of income characteristics of households and financial sector development aspects, analysis o f household data for three countries (Guatemala, Haiti, and the Dominican Republic), which have been analyzed elsewhere in the report suggests that the probability o f having deposit accounts increases monotonically with income (using the counterfactual pre- migration incomes estimated in chapter 3) and so does the likelihood o f having outstanding bank credit, except inthe case o f Guatemala. However, at least inthe case o f Haiti and the Dominican Republic, recipients are also less likely to have bank and non-bank outstanding credit, suggesting a role for remittances to relax financing constraints. Though more research i s required to better understand what drives the findings o f our analysis, the results suggest some tentative conclusions and policy implications.6o Overall, evidence suggests that remittances have a positive impact on financial development but there still i s room for improvement, especially regarding remitters and credit. The descriptive analysis o f the Colombian and Guatemalan experiences hint that there i s potential growth for banks not only on money transfer services but also on savings and credit as: (i) they have been growing rapidly inthe remittance distribution business; (ii) growth has been supported by the eliminationo f their obstacles and exploitation or development o f competitive advantages; (iii) some o f the remaining limitations for growth could be solved relatively easily and banks are planning on taking action; 6o Arguably, giventhe wide variety o fremittancespatterns that appear to exist among LAC countries, which ledto the classificationof countries into two sets basedon plottedflows, more specific recommendationsmay haveto vary dependingon the set of countries.However, the following focuses onbroadrecommendationssince further research is neededto undertakesuch endeavor, which requires more datathan availablefor this study (e.g., not enough observationsto carry out detailedempiricalanalysis for each group of countries identifiedbased onplotted flows). 154 and (iv) a proportion o f remitters and beneficiaries that have not been "bancarized" seem to be interested inusingbanks' financial services. Both banks and governments in remittance recipient and sender countries have an important role to play to facilitate the leveraging o f remittance flows and increase financial deepening, which affects both the supply and demand o f financial services. First, Banks could step up efforts to "bank" migrants and provide credit by using remittances as collateral. Ongoing efforts to continue to "bank" migrants are important, since they will increase the likelihood that migrants send their remittances through bank accounts., In the last couple o f years some Latin American banks have started to look into ways to enter the remittance origination market including placing agencies in migrant recipient countries (see annex 1). In addition, U.S.banks and credit unions have made significant inroads in providing remittances and other services to migrants.61For this trendto continue, banks need to continue to facilitate access to their services by lowering their cost and by tailoring their products to meet migrants' needs.62 In turn, efforts to bank remitters will facilitate turning beneficiaries into clients. Second, although banks in LAC have undertaken steps to "bank" remittance recipients, these efforts need to be steppedup. Orozco (2005) documents the strategies currently pursued by nine financial institutions in the region to attract remittance recipients - such as offering special promotions for opening accounts and training on how to use banking services. H e finds that these efforts have resulted in one fifth o f recipients becoming bank clients. Clearly, more needs to be done. H e points out that banks are most successful when they have a strong presence in both remittance sending and recipient communities and when they invest in community outreach and socially oriented product offerings. An important step for banks to increase credit to remittance market participants i s to systematically collect and analyze information on remitters and beneficiaries transactions to develop products tailored to their needs. The descriptive case studies presented in Annex 1, and ongoing or planed initiatives to develop specialized products presented inAnnex 2 hrther illustrate the potential limitations, avenues for improving them, and ongoing efforts at tailoring o f products. Third, regarding the facilitation o f access to financial services to remitters, it is important for governments inboth migrant recipient countries and sending countries to work inpartnership to implement programs that foster financial access for migrants. This includes financial literacy efforts supported by governments in remittance recipients countries, for instance through consulates and other offices placed in migrant recipient countries. In this context, the ongoing collaboration between Mexico and the U.S. through the "Matricula Consular", the Automated Clearing House (ACH) system, and the initiatives o f the Institute for Mexicans living Abroad are good examples to pursue. "Forexample, asreportedbyOrozco(2004), by2003Hispanicsrepresented20percentofBankofAmerica's new accounts and, as o f 2004 Wells Fargo attracted 250,000 new clients since offering Intercuenta Express, an account- to-account service allowing customers at Wells Fargo branches to pay $10 to transfer up to $1,000 directly into their beneficiaries' BBVA-Bancomer account in Mexico. `*Suro (2005) finds that the cost and flexibility o f banking services are one o f the main reasons why Latin American migrants in the U.S.are not using banks. 155 The issuance o f the Mexican government o f the Matricula ConsuIar, an alternative form o f id for undocumented migrants, which has been endorsed by the U.S.Department o f Treasury now allows illegal Mexican migrants in the U.S. to apply for banking services. On the other hand, the ACH. established by the Federal Reserve Bank and the Mexican Central bank, has sought to reduce transactions costs for remittances sent via banks is also another form o f fruithl collaboration (although its impact has not been as large as anticipated, partly due to the relevance o f exchange rate differentials for intermediaries, which would be kept by the Mexican Central bank for transactions under the A C H (see The World Bank, 2006 forthcoming, and Chapter 9). 63Finally, to promote financial literacy, the Mexican Foreign Ministry's Institute for Mexicans Living Abroad invites banks like Wells Fargo, Bank o f America, and Citibank to discuss remittances and other services with home town association leaders in the U.S.In order to foster the use o f banking services by migrants, efforts such as those undertaken by Mexico need to continue and be embraced by more countries inthe Latin American region. Efforts that directly link migrants with financial institutions from their home countries might also help increase the impact o f remittances on local financial development. Inthis sense, it is important that governments in receipient countries allow financial institutions to operate abroad and to offer financial services to the migrant communities, since such measures have been shown to increase the amount o f remittances flowing into the domestic financial sector in other regions.64 Fourth, Governments could also adopt policies that support efforts to convert remittance recipients into bank clients. These could include: (i) minimizing the regulatory costs to banks o f opening branches and other outlets to serve these communities; (ii)fostering greater outreach by allowing non-traditional methods o f delivering banking services such as mobile banking or correspondent arrangements with post offices, retailers, and cooperatives that are better positioned to serve recipient communities; (iii) promoting competition inthe financial sector are expected to lead to greater outreach; and (iv) continue improving the collection and analysis o f information on remittance flows and patterns. Inparticular, efforts to promote competition could include: (a) provision o f more accurate information about the size and structure o f the market to increase banks' interest in engaging in a larger role and develop specialized products, as they perceive opportunities for economies o f scale that could reduce the cost and increase the benefits o f banking beneficiaries; (b) promoting more clarity inthe pricing o f remittances transfers (more than half o f the remittersreport they don't know why their beneficiaries receive less money than they had estimated-see Chapter 9 for more details on the opacity o f fee structures), which could increase demand pressures to offer better products and reduce costs; and (c) promoting partnerships between formal financial institutions and other financial intermediaries (e.g., microfinance institutions and NGOs) that may have broader outreach to poorer (but bankable) segments o f remittance recipients. 63The ACH allows banks to originate credit payment by sending them to the Federal Reserve for processing via the same methodused for domestic operations, hence potentially reducing the cost o f those transactions. 64 The 2006 Global Economic Prospects on migration and remittance mentions examples for countries such as China, Eritrea, India, Israel, Lebanon, Pakistan, and the Philippines. 156 Channeling o f remittances flows to increase credit inthe economy and among remittance recipients is likely to be harder than broadening deposits with those flows. Evendiagnosing why this is not happening at the moment is not easy.. Limitations presented inAnnex 1, as reported by selected banks in Guatemala and Colombia, could be further explore to improve diagnosis and analyze potential to ease these limitations. Notwithstandingthe difficulties, the potential for increasing the impact o f remittances i s perceived to be large, at least based on the market test o f increased interest o f banks in furthering their role in the market and recent ease o f limitations banks identified, as reported in Annex 1; this coupled with the incipient development o f specialized products, as reported inAnnex 2. One potential problem can be lack o f demand. Remittance recipients may not demand credit because the flows they receive from their relatives overseas relax their credit constraints and substitute for credit (the evidence for Haiti and the Dominican Republic appears to support this hypothesis, at least partly). Also, they mightnot have viable productive projects to pursue or they may be supported by other third parties. However, the results presentedinthis chapter for El Salvador do not support the hypothesis that the demand for credit is lower among remittance recipients. Nor does it suggest that remittance recipients are any different when it comes to obtaining credit. Data permitting, future research to assess if existingprofitable clients o f banks have similar household or income characteristics as remittance beneficiaries that are not client could increase the banks' interest inbroadening access to this group. Supply side factors including lack o f collateral o f beneficiaries, and insufficient understanding o f credit requirements for beneficiaries by banks might all be factors affecting the leveraging o f the impact o f remittances on financial development. Also, there may be prudential and regulatory obstacles hinderingthe use o f remittances flows to mitigate the typical constraints to credit for those beneficiaries (or remitters) that have limited credit history and no collateral. To the extent that remittances can increasingly be used to compensate for insufficient credit payment history or even as substitute collateral, as it is increasingly the case o f Banco Solidario inEcuador discussed inAnnex 2, the more likely it is that banks will increasingly broadenaccess to credit to remitters. Furthermore, given some evidence that many beneficiaries may have received credit from the informal sectors, governments and banks could support the collection o f information on remittances payments and enhance capacity o f credit bureaus to consider such information in assessing credit history and analyze behavior patterns o f potential clients. More detailed surveys o f remittance receipients should be undertaken to better understand why credit does not seem to be increasing as a result o f remittances and provide more specific policy recommendations. Finally, institutional and macroeconomic factor might also help explain why credit to the private sector in Latin American i s not growing as a result o f remittances. Issues such as weak creditor rights, inefficient contract enforcement mechanisms, crowding out as a result o f the government demand for credit due to fiscal pressures, and lingering effects o f repeated crisis in LAC may also needto be considered and tackle by governments seeking to leverage the impact o f remittances on financial development. 157 r c 3 Annex 1: RemittancesandBancarization The Experiencesof Colombia and - Guatemala This descriptive annex presents more details on banking practices related to remittances in two cases, Colombia and G~aternala,~~based on ad-hoc surveys and follow-up interviews carried out during April and May, 2006 at selected banks that have an important participation in the remittances market in each country.66 Other descriptive studies have also been incorporated. The objective is to complement the analytical findings o f the chapter to: (i)illustrate the increasing interest o f banks on the remittances market both on the deposit and, more recently, on the asset side, (ii)identify potential limitations reported by banks for the development o f the banking sector and banking services based on remittances, and (iii)help better understand possible alternatives for overcoming these limitations inthe near future. The ColombianExperience Survey responses and interviews with bank representatives reflect the increased interest and ongoing efforts to further engage in the remittances market, largely based on the size o f the market and perceived potential for bankingprofitable clients. They report that as a result o f these recent efforts, remittance recipients are increasingly opening bank deposit accounts, but they have been less effective thus far at converting remitters into clients, or at increasing credit (for both remitters and recipients). Ingeneral, a first set of obstacles to deepen the banking system through remittances are related to the non-predominant share o f banks in the remittance distribution market.67There are several factors that help explain banks' still secondary role in this market, including: (i)their relatively recent entrance in the market;68Qi) image disadvantages as compared with their main competitors, foreign exchange houses;6 (iii)exclusivity agreements with international counterparts (mostly money transfer operators) that have a given (sometimes small) percentage o f the remittances origination market and commit banks not to pay remittances o f other businesses; (iv) slower customer service for remittance recipients compared to foreign exchange 65Inthe LAC region, Colombia andGuatemalaare, respectively,the third andfourth largestremittancerecipients in absolute terms. In2005, ColombiareceivedUS$4,126 millions(equivalentto 4.1 percent of GDP) while GuatemalareceivedUS$2,993 millions(or 9.3 percent of GDP inrelative terms) (IADB 2006 b). 66Although interviewedbanksrepresent an importantpercentageofthe remittancemarketin each country, due to diversity amongbankingfinancial institutions andthe informalnatureof the surveys they maynot necessarilybe representativeof the country or the region. 67Notwithstandingdata scarcity, Orozco(2006) estimates that Colombianbanks have about 40 percentofthis market,while foreign exchange housesrepresent about 47 percent.Inthis sense, Colombiadiffersfrom the weighted regional average (accordingto country volume), where banksare estimatedto havegreatest percentageof the remittancedistributionmarket (closeto 50 percent),followedby retailstores with approximately38 percent(IADB 2006 a). It is worth notingthat these market participation estimates are basedon data collectedby Orozco from 13 LAC countries and a total of 50 enterprises that offer money transfer services (Orozco 2006). 68Banks started growing inthe remittanceretaildistributionmarketafter 2001. 69Banks reportthe needto improvetheir own imageto gainmarket share, as they believethe public perceives them to be excessivelyfocusedonprofitsandhas some difficulty intrustingthemwith their funds. 159 houses; and (v) to a lesser extent, inadequate location o f branches for targeting remittances recipients inthe case o f smaller banks.70 More recently, the ability o f banks to ease some o f these limitations and to develop new competitive advantages and exploit existent ones have allowed them to gain market share in the last few years.71 Banks have improved their image in the remittance market as their customers have had a positive experience with their services; signed agreements with non-financial businesses with a wide geographic penetration; and, in some cases, provided their clients with ATM cards to minimize payment waiting time. Additionally, banks report to have gained a competitive edge by offering financial services specially tailored to the market72and decreasing exchange rate spreads; as well as exploiting existent advantages such as offering greater security73and, for larger banks, a wide branch distribution network. A second set o f reported obstacles is directly related to the actual broadening of access to credit for recipients and remitters. These include limitations that are more specific to the remittance market and others that are typical o f credit markets but have not yet been resolved by exploiting the characteristics o f remittances flows. Among the former i s the competition from non-supervisedretail businesses as the reported preference for durable consumption goods seems to have provided an opportunity for retailers to provide direct credit. Banks argued that the interest rate ceilings they face by regulation has given an advantage to retailers, as the latter are not subject to these and can compensate higherrisksby charging higher interest rates.74 Typical limitations in the credit market that are also relevant in the remittance market include: lack o f information (insufficient credit history and payment behavior), inadequate risk and scoring models, low demand, and little collateral. Banks report constraints from low demand for credit, because they perceive remittances are mainly used for consumption purposes, to cover basic needs and purchase durable consumption goods.75Although the predominance o f consumption can limit credit demand, residual remittances available for saving and investment 'OThis refers to the actual withdrawal o f remittances at banks' offices. As banks not always had separatecustomer lines to pay remittances, foreign exchange houses have been able to offer a faster service in some cases. 71 Although foreign exchange houses traditionally held most o f the market, they have been loosing grounds to other businesses. In2001 foreign exchange houses paid 95 percent o f remittances, but their share declined to 81 percent in2004 (Banco de la Republicade Colombia 2004). Moreover, Orozco (2006) estimates that their share declinedto 41 percent by 2006 72 The development o fproducts adaptedto the needs o f the remittance market was at different stages in interviewed banks, but all report a large focus on this strategy. Some banks were already offering new products and getting positive results from this strategy, while others were still on the process o f designing them. 73 This refers to the confidence that can be generated when a renownedbank is involved inthe transaction and the greater personal security that can be enjoyed receiving the payment. As banks offer several services aside from money transfers, a client leaving from one o f its branches might not be carrying cash and, therefore, is less likely to be victimized. 74Nevertheless, the influence o f retailers is limitedby the fact that they can only provide credit for the products they sell and banks have the ability to further adapt their credit products. 75This expenditure pattern couldbe explained, at least inpart, by the fact that LAC migrationtends to be longer- term as compared with other regions (see Chapter 2). Therefore, migrants and their families back home might not feel the urge to save inorder to invest and lift their living standards upon the sender's return to the home country. 160 recipient household^.^^ (and paying-off credit) may still be significant as compared with the saving capacity o f non- The problem o f insufficient collateral in the case o f remittance recipients emerges because many o f them are unemployed and can only offer these transfers as collateral. In principle, remittances can serve as collateral since they can occur periodically through a sustained period o f time and be sufficient to cover credit payments. Although interviewed banks were not yet offering credit by taking remittances as guarantee o f payment, there are incipient efforts for change as some are tracking remittances and planning on using this information to offer credit to recipients. Finally, a third group o f identified obstacles for developing the financial sector are those related with remitters. Banks have found several difficulties for converting immigrants into clients and, more specifically, for being able to offer credit to them. First, banks found that these activities require buildinga new infrastructure inexternal markets with business partners that can absorb a relevant percentage o f profits.77 Also, increased competition for the distribution o f remittances and the bancarization o f recipients have led banks to focus their energy and resources inthe domestic market. Notwithstanding the limitations to convert remitters into clients, banks continue to perceive this group as an attractive business opportunity as many o f them have not yet been "bancarized" by foreign competitors and are reported to be interested in saving, investing intheir home country, and making sure that their transfers serve to lift the living standards o f their families by granting them access to education, housing or supporting entrepreneurial activities. As a result, banks are increasing their efforts to take advantage o f the perceived business opportunity. The Guatemalan Experience As in Colombia, in Guatemala the share o f commercial banks in the retail remittance distribution market seems to have increased markedly in recent years (The World Bank 2006. forthcoming) and recipients have increasingly been converted into clients, although they participate mostly through banks' liabilities rather than assets.78 On the other hand, the "bancarization" o f remitters has also been limited by the small participation o f banks in the remittance origination market. Although data on the market structure o f the retail distribution o f remittances in Guatemala are particularly scarce and incomplete, estimates indicate that once small shops and other retailers are taken into account, Guatemalan banks do not have the largest percentage o f the 76This is the case in Guatemala, where remittanceshave a considerable impacton householdincome. Evenwhen consumptiongoods constitutethe main expenditureofrecipienthouseholds, they still manageto save and invest morethannon-recipienthouseholds. 77Infact one of the interviewedbanks hadlauncheda mortgage creditproduct for migrantsbut laterwithdrew it sinceprofit marginswere insufficient. 78As Table 6 shows, while 30.7% o fbeneficiaries holdsavingaccounts, only 0.57% havebank credit. 161 market.79 However, the relevance o f banks in the market has been increasing as limiting factors have eased and banks have exploited competitive advantages. Among obstacles, Guatemalan banks report that exclusivity agreements have constrained their expansion in remittance distribution. These agreements have also hindered the bancarization o f recipients by prohibiting banks to transfer remittances to recipients' bank accounts (both existing and new). This clause limits the benefits o f holding saving accounts as it forces beneficiaries to go to the bank to receive their remittance payments, even if they already have an account. The prohibition also increases the costs o f converting beneficiaries into customers as banks. Ithis process, banks need to assign an ID to remittance beneficiaries, track remittances history and link this information with the transactions that beneficiaries do as bank clients. This implies developing or purchasing compatible information systems to track information (one for remittances and another one for other transactions) and dedicating qualified personnel to analyze trends and explore opportunities for cross selling o f other products. Nevertheless, the relevance o f exclusivity features i s decreasing as banks are broadening their agreements with other agents. On the other hand, there are several elements contributing to the growth o f banks inthe remittance distribution market and deepened bancarization through remittances. A good example i s the tailoring o f products to the needs o f remittance sector, such as the program Los Chapines Estumos Unidos o f Banco Industrial, the largest commercial bank in Guatemala (see Annex 2). All banks interviewed are also making efforts to collect and analyze information on remittances transactions and beneficiaries, which should put them in a better position to take advantage o f some existent but underutilized strengths. The reported shift from informal towards formal transfers recently motivated by AML-CFT regulations that improve recording o f official flows; banks broader range o f financial services; and, in some cases, their relatively widespread branch network are also in favor o f their growth in remittance distribution (The World Bank, 2006 forthcoming). The latter is particularly the case o f two large banks in Guatemala with large rural networks in geographic regions that are large remittance recipients, which have only started to explore the potential for banking beneficiaries. Regarding the potential growth o f bank liabilities in the remittance sector, a 2004 survey conducted by the World Council o f Credit Unions 81shows some encouraging evidence. Even when it is estimated that close to 70 percent o f remittance recipients do not have bank accounts inGuatemala (see Table 5.6), 65 percent of the respondents expressed interest indepositing part o f their remittances in financial institutions and saving 22 percent o f their remittances (Mesbah- Khavari, Evans and 2005). 79 Banks haveabout 23.5 percent of the remittance distribution market. Most of the market is heldby retailers, who account for 72.8 percent of the business.Source; IADB (2006 a). Largerbankswith wide distribution networksthat already target lower incomeindividualswill probablybe ina particularly good position to increasetheir participationby taking advantageof these features once they adapt their roducts andmarketingto the remittancesectorbasedon this information. !I The World Council of Credit Unions, Inc.(WOCCU) conductedasurvey ofusers o f its InternationalRemittance Network (IRnet) services at the five Guatemalancredit unionsduringthe spring of 2004, covering502 individuals (Mesbah-Khavari, Evans and 2005). I62 When interviewees o f the WOCCU survey were asked about the reasons for not having a savings account ina financial institution, they mentioned they did not have enough income, were not used to save, had other more profitable investments, did not know how to do it, or believed they did not meet the requirements (Mesbah-Khavari, Evans and 2005). Among others, these limitations could be minimized by further adapting deposit services to remittances beneficiaries needs (e.g. eliminating management costs, offering attractive interest rates, and lowering minimumrequiredbalance) and explaining the requirements for opening an account. Competition from informal credit providers and lack o f demand and collateral have been obstacles for granting credit to Guatemalan remittance recipients. The WOCCU survey found that: (i)at the overall sample level, 35% o f remittance recipients' households were using credit in 2003; (ii) loans were providedby informal sources (informal moneylenders, supplier credit, most friends or relatives); and (iii)most remittances beneficiaries do not apply for credit as they lack collateral and fear their applications will be denied (Mesbah-Khavari, Evans and 2005). Credit demand i s also influenced by recipients' preferred distribution o f income as Guatemalans primarily use their remittances to cover basic household expenditures rather than to invest in business, agriculture or residential property.82 There are, conversely, some elements that can lead to credit growth in the remittance sector. The fact that recipients are already usinginformal credit raises the opportunity to increase formal credit for beneficiaries by offering better terms, if collateral and lack o f sufficient credit history can be addressed through remittances. Guatemalan banks are starting to take advantage o f remittances to solve the typical limitations o f access to credit for low income individuals as they are now using proven patterns o f stable remittances flows during a given period o f time as a positive factor when assessing credit risk o f potential clients. Finally, even when Guatemalans recipients assign the biggest portion o f their remittances to consumption, their ability to save and invest i s bigger than non-recipient households as the effect o f remittances on income is considerable. 82 Accordingto a recent USAID survey, remittancerecipients spend56 percent o fremittanceson coveringbasic needs, while the remaining44 percentis distributedamong healthandhomeinvestmentsanddebt repayment. 163 Annex 2: Tailoring FinancialProductsto the Needsof the RemittanceMarket The following paragraphs provide examples o f financial products that have been designed by Latin American financial institutions to satisfy the particular needs o f remittances senders and recipients. These examples seek to highlightthe distinctive and innovative features o f these products rather than to provide a detailed description. 0 Banco Solidario: gaining market share through product adaptation Banco Solidario i s a key player in the Ecuadorian remittance market. Even when it entered the market fairly recently (2002), it has managed to rapidly gain participation in the remittance retail distribution business by providing remittance senders and beneficiaries with savings and credit products tailored to their needs (Orozco and Fedewa 2005) and establishing strategic alliances with international banks inseveral countries (most importantly saving banks in Spain). g3 Remitters who hold saving accounts at Banco Solidario are waived transfer fees (since December 2003) and have the option o f designating beneficiaries who can withdraw funds in their home country. Saving accounts have low costs and allow the customer to have funds available upon re uest and a saving plan that lets clients decide how much to save each month and for how long.74 Banco Solidario also offers mortgages and loans to migrants and their families with several innovative features. Mortgages are sold to migrants through a network o f sale offices in Spain and Italy, have low saving requirementsg5, and a minimum down payment o f 30% (Dorham 2005). Relatives o f the remitter can also apply for a similar mortgage by showing proof o f the last three remittances received from Spain (Dorham 2005). Beneficiaries who receive their remittances from Spain can also apply for loans o f up to US$ 5,000 by demonstrating they are receiving remittances, even when they are client o f other financial institutions. Banco Solidario also offers loans to remitters o f up to US$10,000, that can be used for several purposes (e.g. migration travel costs, improvement o f micro-enterprises and housing). 0 Banco Salvadorefio: reaching senders Banco Salvadoreiio has opened several branches in the US to reach senders. So far, the institution has seven agencies in the U S cities with the higher concentration o f Salvadorian 83These alliances with saving banks (cajas de ahorro) have formed a network, called "Red de Servicios Financieros a Emigrantes", that includes more than 9,000 outlets in Spain (see http:/lwww.banco- solidario.con7/noticia.uh~'?iiotID=5).Through these network, migrants in Spain transferred $120.4 million inthe three years through July 2005, and they deposited $7.8 million, or 6% o f that figure, into savings accounts (Derham 2005). 84Administrative costs are $1,50 for accounts ifthe average balance is smaller than $300. 85Initial savings for mortgages can start at $40. 164 migrantsSg6This has allowed the bank to charge competitive fees for transfers and migrants to transfer funds directly to beneficiaries' savings account, capturing 12 per cent o f El Salvador's remittance market (Orozco and Fedewa 2005). Financial products have also been adapted to the needs o f Salvadorian remitters and recipients, who are given preferential treatment to facilitate the establishment o f a long-term relationship (Orozco and Fedewa 2005). In particular, the bank has developed a mortgage product line for remitters,which offers three products according to the legal status o f the migrant (i.e. legal residents, work-permit holders and undocumented). ''Mortgages to acquire homes (used or new) take the property as collateral and can finance up to 90% o f the value o f the house for a maximum term o f 15 years. Mortgages to acquire lots also take the property as collateral but can finance up to 70% o f the value for a maximum term o f 10years. Banco Industrial: tailoring credit to recipients Banco Industrial (BI), the largest Guatemalan bank, has developed the "Los Chapinos Estamos Unidos" (Guatemalans Are United) program, which includes money transfer, saving, and credit services tailored to the needs o f the remittance market. The program's saving and credit services are mainly sold to remittances recipients. Credit products consider the recipient's remittance history to assess risk and determine the maximum loan amount (70 percent o f remittance average) and are generally short-term (maximum o f 2 years). BI offers two credit products within its "Los Chapines Estamos Unidos" program: Los Chapines Estamos Unidos Loan and Remittances in Advance. While the former offers a 2 percent monthly interest rate and a longer term (6 to 24 months), the latter carries higher interest rates (4 to 6 percent) and a maximum term o f 3 months. Bansej?:networking to reach economies of scale and undersewed areas Bansefi (Banco de Ahorro Nacional y Servicios Financieros - National Savings and Financial Services Bank) is a public development bank created as part o f a governmental strategy designed to upgrade the Popular Savings and Credit Sector, which i s formed by semi-informal and informal financial intermediaries that serve unbanked areas (De la Torre, Gozzi and Schmukler 2006).88 The strategy seeks to support these institutions to increase financial sector deepness and promote access to financial services by lower-income individuals through a networking model that links financial intermediaries to support common activities and reach economies o f scale (Coutinho 2006). Bansefi administers "L@ Red de la Gente" (The People's Network), a strategic association among the bank and financial institutions belonging to the Popular Savings and Credit Sector, which i s facilitating the "bancarization" o f underserved sectors (Le. lower and 86 These are Los Angeles, Santa Ana, San Francisco and Van Nuysin California, Houston (Texas) and Las Vegas (Nevada). "Undocumentedmigrantsareofferedshortertermmortgagesandarenotrequiredtopresentdocumentationto prove their legal status inthe US, remittances history and tax declarations. Bansefi was established through an organic law, sanctioned by congress inApril, 2001, which mandatedthe transformation o f the existent Patronato del Ahorro Nacional (PAHNAL). 165 middle income individuals). Aside from exploiting operational and commercial economies o f scale8', L@Red de la Gente performs activities directly related to remittances, including: (i) negotiating fees with foreign banks and transfer companies to lower transfer costs; (ii) channeling the transferred funds; and (iii)attracting senders and beneficiaries to open saving accounts that generate financial records (Coutinho 2006). e Peru: mortgagesfor recipients. Fondo MIVIVIENDA, a Peruvian public fund that provides housing finance through private institutions, modified one o f its products (CrCdito MIVIVIENDA) to facilitate access for remittances recipients. Currently, close relatives o f Peruvian immigrants that meet specified requirements can request a mortgage to the private banks that participate inthe program and their remittances are considered as part o f their income to lower their risk profile. In order to qualify, all recipients must have received and deposited in their accounts a monthly remittance that should not be smaller than the monthly payments o f the desired mortgage for a period o f six months. In turn, private financial institutions that belong to the program define other requirements for remitters and recipients and can tailor their products to the needs o f their clients. For instance, Banco de CrCdito has developed the product "MIVIVIENDA con Remesas" that offers three different mortgages, two for recipients that have an additional source o f income (Plans 1and 2) and one for direct relatives o f remitters that have remittances as their sole source o f income (Plan 3). While Plan 1requires saving for 12 months and a minimum down payment o f lo%, Plan 2 requires saving for 6 months and a minimumdown payment o f 20 %. Finally, Plan 3 offers two options: saving for 9 months (with a minimum down payment of 30%) or saving for 12 months (with a minimumdown payment o f 20%). 89Economies ofscale are sought by providingan internetbasedfinancial services networkanda trademark for the productsofthe Netmembers. 166 Chapter 6 Remittances, the real exchange rate, and the Dutch Disease phenomenon* So far, this report has been arguing that the more remittances a country receives, the higher its werfare level. Yet, flows that are too large relative to the size of the receiving economies, as those observed in a number of Latin American countries, may also bring a number of undesiredproblems. Among thoseprobably the mostfeared in this context is the Dutch Disease. This chapter explores the empirical evidence regarding the impact of remittances on the real exchange rate. Do countries with high remittanceflows experience losses of competitiveness due to a real exchange rate appreciation? Should policy makers react to this potentialphenomenon? Ifso, what are the options that they have to address this concern? I,Introduction In the previous chapters we have argued that remittances may have a number of beneficial effects for the welfare o f the receiving countries. The evidence presented so far inthis report (see Chapter 3) suggests that at the country level higher remittances inflows tend to be associated with lower poverty indicators and higher growth rates. Beyond these typical income dimensions o f welfare, remittances seem to reduce output volatility (a measure o f risk faced by countriesg0), and at least in some countries and for some socio-economic groups lead to improvements in social indicators (chapter 4). Yet, the magnitude o f these flows relative to the size o f the receiving economiesg' impliesthat remittances may also pose an important number o f challenges. For while these inflows may ease external financing constraints and therefore hold the potential for higher investment by developing countries, in many circumstances remittances * This chapter is based on the background paper for this report "Remittances and the real exchange rate: Are there reasonsto be concerned about Dutch Disease in Latin America?" by Humberto Lopez and L.Molina. 90 See Perry, Arias, Lopez, Maloney and Serven (2006) for a discussion o f non-monetary dimensions (including risks considerations) ofwelfare. 9'Recall that as discussed in Chapter 1, there are a number of Latin American countries where remittances are above 10 percent o f GDP and in the case o f Haiti they reach almost 40 percent. 167 are so large that they can impact macroeconomic stability and more specifically carry a potential for Dutch Disease type o f phenomena (see the International Monetary Fund's World Economic Outlook 2005, and the WorkBank's Global Economic Prospects 2006). Workers remittances can be viewed as a capital inflow, and therefore the theory on the Dutch Disease phenomenon associated to surge in inflows (perhaps because o f a discovery of new natural resources) can also be applied inthis context. Very briefly, the rationale for a Dutch Disease type o f effect would be as follows. Remittances have a positive impact on the incomes o f receiving households, and hence they will also tend to positively affect c o n s ~ m p t i o nTo . ~ ~the extent that some o f this consumption i s directed towards the non tradable sector where competition i s likely to be somewhat limited, remittances will tend to drive up the price o f non tradable goods relative to that o f tradables and therefore contribute to a real exchange rate appreciation. Obviously, the pressure on the real exchange rate will be somewhat mitigated if (i) there are productivity gains, particularly in the non tradable sector, that offset the effects o f the increasing demand; and/or (ii)a large share o f the remittances is channeled to the external sector via additional imports so that the price effect on non tradable goods is limited. Yet inprinciple it seems difficult to justify that these effects are enough to mitigate appreciating pressures. In turn, there are a number of additional macroeconomic effects that can result from a real exchange rate appreciation associated to remittances flows. They include the following: 0 Adverse effects on the tradable sector o f the economv. Although remittance flows are likely to lead to an expansion o f the non tradable sector (as a result o f the increase experienced in the domestic demand), both export and import-competing industries (Le. the tradable sector o f the economy) would be adversely affected by the real exchange rate appreciation and the associated loss o f international competitiveness. The negative impact of remittances on the tradable sector may be reinforced if they also fuel inflation and higher prices result in higher economy-wide wages.93 This effect would further be magnified if as argued in Chapter 4, remittances also reduce the labor supply. In these circumstances, the non tradable sector may be inthe position o f passing onto prices some o f the wage pressures, but this i s likely to be much more difficult for a tradable sector facing international competition which as a result will loose competitiveness. Widening of the current account deficit. In principle, it is difficult to justify that an increase in the domestic demand will be passed in full to the non tradable sector. So, to the extent that some o f the remittances-induced-consumptioni s directed towards tradable goods, there will be an increase in the demand for imports. This coupled with the losses o f international competitiveness o f the domestic firms mentioned in the previous paragraph would likely result in deteriorations of the external position. For example, according to World Bank (2003) the surge inremittances observed in El Salvador during the 1990s was the most likely factor behindthe worsening o f the country's trade deficit which over the 1990s deteriorated from less than 7 percent o f GDP to almost 14 percent o f GDP. 92For example, FUSADES,a Salvadoranthink-tankputsthe share of remittances consumedinthat country at about 70 percent. 93This is a typical resultof economic modelswith labor mobility. 168 Weaker monetary control, inflationary pressures. and the sectoral allocation o f investment. If remittances flows do not leave the country through a widening o f the current account, large flows will push up monetary aggregates, potentially derailing inflation targets. Experience also indicates that prices o f financial assets and particularly o f real estate can rise rapidly following a surge inremittances, something that inturn may introduce significant distortions in the economy and affect the sectoral allocation o f investment and lead to overinvestment insome sectors (i.e. real estate). On the whole, the previous discussion highlights a number o f problems that policy makers may have to face in the context o f a surge in remittances. True, to the extent that these Dutch Disease phenomena are part o f the natural adjustment process towards a new equilibrium, they should not be a matter o f particular concern for policy makers. Indeed, if we view remittances as a positive shock to the economy, then the real appreciation and related effects experimented by the receiving country would simply be part o f the inevitable relative price adjustment process that goes with favorable shocks. Yet ifthis real appreciation i s very dramatic, or the adjustment process towards the new equilibrium i s uneven (Le. not fully consistent with the change in economic fundamentals at each point intime) policy makers may wish to mitigate, to the extentpossible, its adverse effects on export industries. Inprinciple, one could also mention two additional reasons of concern for policy makers that are usually mentioned in the context o f surges o f capital inflows. One i s the potential for a flow reversal over the medium run. This i s important because if there i s hysteresis in the real sector, a real exchange rate appreciation may wipe out important sectors o f the economy that would not reappear even if the currency subsequently depreciates. The other potential concern is a very sudden appreciation that cannot be accommodated and therefore brings a very painful adjustment. However, it must be noted that the documented stabilityg4and counter-cyclicality o f remittances would lead one to assume that the probability o f short-run reversals or sudden adjustments i s quite low, leaving as main reason for concern the magnitude o f the real appreciation associated to the remittance inflows. Against this background, what does the economics literature has to say about the evolution o f the exchange rate in countries that have experienced important increases o f remittances? The truth i s that the existing empirical literature is very limited and less than unanimous. For example, Amuedo-Dorantes and Pozo (2004) rely on cross country econometrics techniques and find that in a sample o f 13 Latin American countriesg5a doubling o f workers remittances would lead on average to a real exchange rate overvaluation o f about 22 percent. This estimate would be robust to the presence o f fixed effects in the data, and to the use o f IV estimation techniques to account for reverse causality from the exchange rate to remittances. However, Rajan and Subramanian (2005), who rely on a cross national dataset o f 3-digit industry value added growth data to explore whether remittances have a differential impact 94The stability of remittances over time is one o f the most important differences between remittances and other types of capital inflows. In fact, as noted by the Word Economic Outlook 2005, non-FDI private capital flows, exports, FDI, and even official aid all show greater volatility than remittances. 9sThe countries in their sample are: Argentina, Belize, Bolivia, Colombia, Dominican Republic, El Salvador, Guatemala, Honduras, Jamaica, Mexico, Nicaragua, Peru, and Trinidad & Tobago. 169 depending on the labor intensity of the different industries, find that unlike other types o f capital flows (particularly aid flows) remittances do not seem to have a negative impact on external competitiveness. Rajan and Subramanian (2005) argue that this could be the result o f remittances being directed to a large extent towards unskilled-labor intensive activities - e.g. goods and services providedby micro-enterprises - andor tradable sectors such as manufacturing, and thus having limited effects on the prices o f skilledlabor and other relatively scarce resources. This chapter addresses these issues and contributes to the existing limited literature along several dimensions. First, it discusses the different channels through which remittances can affect the real exchange rate using a framework where the equilibrium exchange rate is characterized by an external equilibrium similar to those analyzed in asset market models (Mussa, 1984 and frenkel and Mussa, 1985) and an internal equilibriumbased on a productivity differential model as those inBalassa (1964) and Samuelson (1964). Second, it provides estimates o f the impact o f remittance flows on the real exchange rate using a large cross national data set rather than information for a limited number o f countries. Our approach allows testing whether there are regional differences andmore specifically whether Latin America i s different in this context. Note that one o f the main differences between the work o f Amuedo-Dorantes and Pozo (2004) and Rajan and Subramain (2005) i s the coverage o f the data. Thus if the impact o f remittances on the real exchange rate i s different in Latin American than inthe rest o f the world, then the different findings o f Amuedo-Dorantes and Pozo (2004) and Rajan and Subramain (2005) should not be surprising. In fact, to anticipate some o f the results in Section IVYthis chapter argues that although remittances flows do not appear to affect the real exchange rate in countries outside o f Latin America, in this region they are likely to contribute to a significant real exchange rate appreciation. These results are robust to the presence o f fixed effects in the data, potential reverse causality from the exchange rate to remittances, and variations inthe set o f control variables. Third, the chapter also explores the extent to which the estimated appreciation in Latin America i s consistent with the change in economic fundamentals implied by the increase in remittances or instead whether it can be attributed to changes inthe misalignment component o f the real exchange rate (i.e. changes in the underlyingreal overhndervaluationo f the currency). To also anticipate our results on this, we find that the evolution o f the Latin American real exchange seems to be driven by a combination o f changes in the equilibriumreal exchange rate and changes inthe degree o f misalignment. Finally, on the basis o f its empirical results the chapter discusses a number o f options for policy makers concerned with the impact that a surge in remittances may have on the external competitiveness o f the country. The rest o f the chapter is organized as follows. In Section I1we consider an additional number o f theoretical considerations that may explain why remittances may lead to a real exchange rate appreciation and review the evolution o f remittances, the real exchange rate, exports and imports for the 8 largest receivers o f remittances (as a percentage o f GDP) in Latin America. Section I11reviews the empirical strategy used to assess the impact o f remittances on the real exchange rate. In Section IV, we present the results o f estimating two econometric models. One relates changes in the real effective exchange rate to the ratio o f remittances to GDP. The second uses as explanatory variable the changes in a measure o f real 170 exchange rate overvaluation. The basic idea here is trying to disentangle how much of the observed changes in the real exchange rate are due to changes in the equilibrium exchange rate and hence consistent with the evolution o f economic fundamentals. Finally Section V closes with some conclusions and a review o f policy options. 11.Remittances and the realexchangerate. Theoretical considerations Remittances can potentially affect the real exchange rate through three main channels (see the technical annex for a formal discussion). First, remittances may affect the external equilibriumo f the economy by raising the net foreign asset position o f the country. For example, the theoretical models o f Mussa (1984), Frenkel and Mussa (1985), Alberola and Lopez (2001) and Aberola et al. (2002) imply that the external equilibrium o f the economy will be reached when any current account imbalance is compensated by a sustainable flow o f international capital. In turn, the rate o f sustainable capital flows will be a function o f the stock o f foreign assets and liabilities o f the economy, so that changes to the net foreign asset position o f the country will lead to changes inthe real equilibrium exchange rate. Given that international remittances are transfers o f foreign currency that unlike other types o f international flows have no obligation associated, remittances will have a direct impact on the net financial position o f the country vis a vis the rest o f the world. Note inthis regard that the impact of remittances on the stock o f net foreign assets differs from the impact o f other flows such as loans or foreign direct investment flows. In the case o f a loan, there is an associated liability (the repayment) and therefore the contribution to the net foreign asset position o f the country is given by the difference between the proceeds and the net present value o f the repayment obligations. In this regard, loans will positively affect net foreign assets to the extent that they have a positive grant component. On the other hand, foreign direct investment flows coming into the home country will increase the foreign liabilities and therefore, will lead to a decline o f the net foreign asset position. Second, remittances can also affect the internal equilibrium o f the economy understood as the situation where domestic capital and labor are efficiently utilized. If as discussed above, remittances lead to an acceleration in the demand for services, inflation will tend to be higher in these sectors which typically are not tradable (and hence somewhat protected from competition) leading to a real exchange rate appreciation (the traditional Balassa-Samuelson effect). Similarly, market rigidities may result in productivity differentials between sectors. For example, if remittances raise the reservation wage, then excessive wage pressures inthe tradable sector may lead to employment adjustments to maintain competitiveness, whereas inthe non-tradable sector employers may admit these pressures because they can pass them onto prices. As a result, remittances can also lead to higher productivity growth and lower inflation inthe tradable sector through their potential impact on the reservation wage, One implication o f this discussion i s that whether remittances are primarily used for household consumption or investment purposes will have a direct impact on the way they affect the real exchange rate, with remittances that are 171 predominantly consumption oriented having more o f an appreciating impact on the real exchange rate. A third possibility for remittances to affect the real exchange rate is through their impact on growth (see Chapter 3 for evidence in this regard), although in this case the impact on the exchange rate i s likely to be uncertain. On the one hand, an acceleration inthe growth rate would lower the stock o f net foreign assets as a percentage o f GDP and hence this would lower the real exchange rate (Le. growth would have the same impact as an increase in the liabilities o f the country). If on the other hand, the net foreign asset position o f the country is negative vis a vis the rest o f the world, the increase in the rate o f growth would lower the liabilities to GDP ratio and hence lead to an appreciation. On the internal front, faster growth would be associated with a real exchange appreciation. Higher growth would lead to higher internal demand and the mechanism described above in the description o f the internal adjustments will apply (Le. the Balassa-Samuelson argument). Thus on the whole, the impact o f a growth acceleration could go ineither direction or evencancel each other and have no impact. How do the data look like? We now move beyondthe theoretical considerationsjust discussed and focus on the data. Table 6.1 reports the (annualized) change inthe (log) real exchange between 1990 and 2003 for a number o f Latin American countries, together with the initial and final levels o f remittances (in percent o f GDP). The remittances data are the same we have been using throughout the report. The real effective exchange rate is from the IMF's International Financial Statistics and it is defined as the relative price o f domestic to foreign goods, so that increases imply a real exchange rate appreciation. Table 6.1 indicates that over 1990 and 2003 the real effective exchange rate appreciated in 11 o f the 20 countries under consideration. The countries where it appreciated the most are Belize, Ecuador, El Salvador, Guatemala, and Haiti. Interestingly, Ecuador, El Salvador, Guatemala, and Haiti are among the top receiver countries o f remittances o f the region (as percentage o f GDP). Other countries where the real effective exchange rate appreciated over the period under consideration by at least 1 percent per year on an annual basis are Honduras, Jamaica and Mexico. The first 2 o f those are also among the countries o f the region with a high remittance to GDP ratio, whereas Mexico i s the country with the larger remittances flows o f the world. Figure 6.1 presents similar information but now restricting the sample to the 8 Latin American countries with the largest remittances to GDP ratio in 2002. Even though we should stress that this figure can only provide evidence o f unconditional correlations, it i s suggestive o f a positive relationship between the evolution o f remittances and the real effective exchange rate. In fact in most of the 8 countries in the figure it is possible to observe a real exchange rate appreciation in parallel to an increase in the remittances to GDP ratio. The first apparent exception to this rule would be Nicaragua, where the evolution o f the real exchange rate over the early 1990s and 2000s appears to move in the opposite direction to what one could expect. The second exceptions are given by Ecuador and the Dominican Republic, where marked real depreciations followed the crisis o f 1999 and 2002, respectively, at a time when remittances were 172 increasing substantially. However, in the case o f the Dominican Republic the real exchange rate and the remittances to GDP ratio also moved inparallelbefore 2002. Table 6.1. Remittances and the real exchange rate. REER 1/ Remittances 2/ Growth Initial Final Argentina -1.6 0.7 0.5 Belize * 32.6 3.4 1.7 Bolivia -0.3 0.1 1.5 Brazil -5.3 0.1 0.6 Chile 0.1 0.0 0.0 Colombia 0.5 1.1 4.0 Costa Rica 0.3 0.2 1.8 Dominican Republic -0.4 5.0 12.7 Ecuador 3.0 0.5 5.7 El Salvador 3.4 7.4 16.1 Guatemala 3.3 1.6 8.7 Haiti 3.0 2.7 41.2 Honduras 1.1 2.1 12.5 Jamaica 1.o 4.4 17.7 Mexico 1.8 1.2 2.4 Nicaragua ** -1.3 0.5 10.6 Panama -1.2 0.3 0.7 Paraguay -1.6 0.6 3.7 Peru -0.6 0.3 1.4 Venezuela *** -1.5 0.1 0.2 1I percent 21percent og GDP Initial Period is 1990 and Final Period is 2003 ***BelizeInitial Period=l990, Final Period=2002 ***Nicaragua, Initial Period=l992, Final Period=2003 Venezuela, Initial Period=l997, Final Period=2003 Source: Author's calculations. Note also that in some o f the other countries the observed real exchange rate appreciations have been quite dramatic. For example in Ecuador, El Salvador, Guatemala, and Haiti, over the 1990-2003 period the real exchange rate appreciated by about 40 percent. In the cases o f Honduras and Jamaica the recorded appreciation is more modest but it would still be in the 20 percent range. In Figure 6.2 we now compare the evolution of the real exchange and the volume o f exports o f good and services. The figure indicates that the only countries where export volumes have significantly increased over the 1990-2003 period are El Salvador, a country where exports increased from about 19 percent o f GDP in 1990 to close to 30 percent in 2003, and Ecuador where exports increased by almost 20 percentage points to close to 55 percent o f GDP. In Honduras, and Nicaragua export volumes would have been more or less stable over this period oscillating between 30 and 40 percent o f GDP in the case o f Honduras and hovering around 25 percent in the case o f Nicaragua (although with a large variance). In the rest o f the countries 173 under analysis, we observe declines in export volumes which in some cases have been quite dramatic. For example, in Guatemala and Jamaica export volumes have fallen over the period under analysis by about 10percentage points o fGDP. Figure 6.1. Remittances and the real exchange rate Dominican Republic Ecuador l A . 10 4.8 4.3 4.23 4.1 0 4 ElSalvador Guatemala 17 . . A 9 1 4.8 4.7 4.6 I 4.5 b 4.4 3 4.3 Haiti 7 Honduras Jamaica Nicaragua 20 5 I 4.8 4.8 4.8 10 4.7 % 4.6 3 4.5 4.6 4 4 0 4.6 I-CRemittancedGDP 6REER ' --CRemittances/GDP --CREER 1 1 Note: increases inthe REER index indicate a real appreciation. Source: Author's calculations. I74 Figure 6.2. Exports and the real exchange rate Dominican Republic 38 4.80 55 . Ecuador 4.80 36 4.75 460 4.55 24 4.50 4.45 4.40 EL Salvador Guatemala I 4.90 22 T 4.90 +Export/GDP -Log of REER ~ Haiti Honduras 30 1 5.20 , 5.10 I-CEXDOWGDP -Log of REER I 1; Jamaica Nicaraaua I 5.10 ::::[ 60 5.00 4.70 35 E:b4 30 4 40 25 4.30 J+EXPO~~/GDP -Log of REER 1 (+Export/GDP -Log of REER1 I Note: increases in the REER index indicate a real appreciation. Source: Author's calculations. I75 Figure 6.3. Imports and the real exchange rate Dominican Republic Ecuador 7 4.80 4.80 El Salvador ::;:1 1:: 44 7 - i 4.90 28 4.80 4.70 1 4.60 '6 4.60 '6 4 - - 4.50 32 4.50 4.40 4 22 4 40 30 4.30 20 4.30 Haiti Honduras 50 I 7 r 5.20 1 6 0 1 ,5.10 45 5.00 4.90 4.80 4.80 '6 - 20 4.70 4.60 4 15 4 40 10 4.20 35 4.50 +Imports -CREER- -- I Jamaica Nicaragua % 5 00 f :: 1 6o 4.90 4.80 4.70 '6 460 4 4 50 45 4.40 1 1 I Note: increases in the REER index indicate a real appreciation. Source: Author's calculations. As for the evolution o f imports, figure 6.3 indicates that over the 1990-2003 there i s only 1 country where the imports to GDP ratio fell (Ecuador). Inthe other 7 countries under analysis imports increased. True in some countries like Guatemala, Jamaica and Nicaragua only slightly 176 (by less than 10 percentage points o f GDP), but in the cases o f Honduras and Haiti the increase has been quite marked: 15 percentage points o f GDP in Honduras and close to 30 percentage points o f GDP inHaiti. Overall, the previous figures would indicate that in general increases inremittances have been accompanied by real exchange rates appreciations, and these in turn by declines in exports and increases in imports, elements that could be taken as a lose o f competitiveness. The next section explores whether the empirical evidence i s also supportive o f some causality from development on the remittances front to the evolution o f the real exchange rate. 111. Empiricalstrategy. Empirical model To explore the existence o f causal links betweenthe real exchange rate and remittances in the data, we rely on the following regressionmodel: where q i s the log o f the real effective exchange rate,96 A i s the first difference operator (such that Aq, = q, -q,-, ), R i s a measure o f remittance flows (in the case o f this paper the remittances to GDP ratiog7), x represents a set o f control variables, which we shall discuss shortly, vi is a country-specific effect, andu, i s an i.i.d. error term. In (1) i and t are a country and a time index respectively. exchange appreciation we should find that p > 0. If,however, remittances have no impact on the Our primary focus is the estimate o f p in equation (1). If remittances lead to a real real exchange rate then we should findp =O. Although it i s theoretically possible here we do not consider the possibility o f < 0, since this would imply that remittances contribute to real exchange rate depreciation."(See the technical annex for a discussion) The previous model can easily be extended to allow remittances to have a different impact inLatin America and inthe rest o f the world by simply adding and additional regressor to the specification inequation (1):" 96Recall that as argued above increases in q correspond to a real exchange rate appreciation. 97In this context, it seems more appropriate to work with the remittances to GDP ratio than with for example remittances per capita. The reason i s that in principle one would expect that the real exchange rate is more affected when remittances are large relative to the size o f the economy than when they are large relative to the population. 98This however would require that the main effect o f remittances on the exchange rate is through the growth channel described above and that the country has a positive net foreign asset position. 99This can occur if regional migration flows patterns are different and as result migrants remit for different purposes. Once again remittances that result in higher consumption will likely lead to an appreciation o f the real exchange rate whereas remittances that lead to higher investment may have less o f an effect. 177 AqiI = w'xi, p,R, + +p2R,, x +vi+vi,, lac where lac i s a dummy variable that takes a value o f 1 ifthe country inquestion i s inLatin America and 0 otherwise. One additional issue that needs attention i s the specific set o f control variables that are included inx. Here we follow to a large extent the strategy inAmuedo-Dorantes and Pozo (2004) and consider the terms o f trade, government expenditure, the world real interest rate, and GDP per capita income. These variables would capture respectively the effects o f potential external shocks, fiscal policy differences, changes in external financial conditions, and productivity gains.loo Despite the broad similarities between equations (1) and (2) and those estimated by Dorantes and Pozo (2004) there are a number o f important differences. First, our results are based on a large cross national dataset rather than on a limited number o f countries (in the case o f Dorantes and Pozo, 2004, 13 Latin American countries). As argued previously if Latin American migration patterns are different from those o f other regions (see Chapter 2 for a discussion on this issue), this could also be reflected in a differentiated effect o f remittances on the real exchange rate. Second, we focus on changes inthe exchange rate rather than on its levels. The reason for this is that since the real effective exchange rate variable is an index, it is no possible to make cross country comparisons on the basis o f the levels o f the variable."' Fixed effects estimation can somewhat mitigate this problem but it i s not likely to fully address it in a satisfactory way. Note that this lack o f cross country comparability also prevents us from estimating dynamic models for the real exchange rate where for example the changes in the variable o f interest are related in addition to a control set, to the lagged endogenous variable so that one can capture mean revertingforces. The need to work with the first difference o f the real exchange rate inturn implies that it is probably more appropriate to work with transformations o f the original explanatory variable set. For example, rather than working with the levels o f the terms o f trade we work with their changes. Similarly, rather than working with GDP per capita we work with GDP growth rates. A third difference with respect to Amuedo-Dorantes and Pozo (2004) is that we explore whether the results are robust to excluding GDP growth from the specification. The reason for this is that if remittances affect per capita growth and growth in turn affects the real exchange rate either h la Balassa-Samuelson or as in portfolio models, then econometric models that control for the evolution o f income levels will not capture the full impact o f remittances on the exchange rate (i.e. these models will produce results that are biased). looThe remittancesdata are as in Aggarwal, Demiguc-Kunt, and Martinez Peria (2005) and the data come from the IMF's World Economic Outlook 2005 database. As for the rest o f the variable, the terms of trade and per capita growth variables come from the World Development Indicators database, government consumption is from the World Economic Outlook 2005 database, and the world real interest is computed using the US interest rate (6- month) andUSproducerprice index from the International Financial Statistics. lo'For example, the baseyear all the countriesinthe sample will havethe same value, say 100; clearly, this does not implythat the realexchangerate levelis comparable amongthem. 178 Fourth, our variable o f interest is the remittances to GDP ratio whereas Amuedo- Dorantes and Pozo (2004) use remittances per capita. We believe that our choice is more appropriate in this context since it i s likely to better capture the importance o f remittances flows relative to the size o f the economy. Finally, we also explore the extent to which the results for the real exchange rate are driven by changes in fundamentals and changes in the disequilibrium level. In this regard, we also present the results o f a regression model that has as dependent variable an index o f overvaluation based on Dollar (1992). To better understand the idea here, consider the following decomposition o f the real exchange rate: where the bar on top o f the variable indicates that an equilibrium value and the hat indicates a disequilibriumvalue. Then our strategy i s basedon the following regression: Aii, = K ' x ~ , +mi,+vi +Uir. (4) Note that same simple manipulations o f (l), and (4) yield (3) Aqi, = Aqi, -Aii, =( - w)'x, (p-6)R, +error K + so that from those regressions it is possible to recover the impact o f remittances on the real equilibrium exchange rate. Econometric issues One problem that has to be faced before proceeding with the results o f the estimation i s the potential reverse causality from the real exchange rate to the remittances. Rajan and Subramanian (2005) note that countries that had overvalued exchange rates in the early 1990s received significantly lower remittances during the rest o f the decade, and argue that it i s plausible that if emigrants perceive an overvalued exchange rate, they may switch to sending goods directly. Similarly, emigrants can find more attractive sending remittances home following devaluation. In this regard, the expected positive causal relationship from remittances to the real exchange rate can be contaminated by a possible negative causal relation from changes inthe real exchange rate to remittances. To somewhat address the potential reverse causality one can resort to IV estimation techniques and follow Agganval, Demirguc-Kunt, and Martinez Peria (2005) who propose two instruments for remittances: the level o f output per capita of the host countries o f migrant workers weighted by (i)distance between sending and receiving country and (ii)share o f migrants o f the receiving country inthe sending country. Inthe empirical section below we use as instrumentsthe (logged) levels and first differences o f these two variables. 179 IV. Results Remittances and the real exchange rate We start this section by presenting results corresponding to equation (1) under the restriction that the GDP growth rate does not enter in the equation, so that if remittances affect the exchange rate though their impact on growth this specification should capture their full impact. Below we also explore the results when we add per capita growth to the regressor set. The estimates in Column (I)f Table 6.2 would suggest that indeed remittances would lead to a o real exchange rate appreciation. Judging from point estimates, this basic model indicates that a 1 percentage point increase inthe remittances to GDP ratio would lead to a real effective exchange rate appreciation o f 1.5 percent. Thus a doubling o f the remittances to GDP ratio would lead to a real exchange rate appreciation o f 3 percent. This estimate is much lower than the estimate o f Amuedo-Dorantes and Pozo (2004), which as noted above would be slightly above 22 percent. The estimates for the rest o f the variables in this specification carry the expected signs and all are statistically different from zero. As discussed above, higher interest rates and improvements in the terms o f trade would inprinciple be associated with a higher real exchange rate. Regarding higher government consumption, it would not be unreasonable to assume that it will fall disproportionately on non-tradables and therefore, ceteris paribus, lead to an appreciation o f the exchange rate on the basis o f Balassa-Samuelson considerations. Column (2) reports the results o f extending the previous basic model as in equation (2); that is, it reports the results of a model that allows remittances to have a different impact on the real effective exchange rate inLatin America and inthe rest o f the world. With this specification the estimates o f the impact of remittances on the real exchange rate are much less accurate than inthe previous model. For example, neither the parameter ofremittances nor that ofremittances interactedwith the Latin American dummy is now significantly different from 0. Yet, the p-value (.02) o f the null that in Latin America remittances do not affect the real exchange rate (i.e. the nullthat the parameter ofremittances plusthe parameter ofremittances interactedwith the Latin American dummy equals 0) indicates that this hypothesis should be rejected. Inother words, the result that remittances lead to a real exchange rate appreciation seems to be driven by the Latin American countries in the sample and it could be possible that the effect observed in the global sample i s specific to the region. 180 Could these results be biased by reverse causality considerations? As discussed above, in principle one would expect that a potential causal relationship from the real exchange rate to remittances would introduce a negative bias in the previous results. However, missing variables considerations could make the sign o f the bias go in any direction. Columns (3) and (4) explore whether the previous findings are robust to the use o f IV estimation techniques using as instrumentsthe variables described in Section 111.2 above. These specifications indicate that, if anything, the impact o f remittances on the real exchange rate would be much larger than in the models that treat remittances as an exogenous variable (i.e. the OLS estimates seem to be affected by a negative bias). In fact, while the estimates in column (1) suggested an impact parameter o f about 1.5 with the IV specification the estimated parameter is almost five-fold (7.4). Similarly, for the specifications where Latin America is allowed to have a region specific impact: in this case, OLS estimates suggest an impact parameter o f 2 for Latin America whereas IV estimates suggest 12.2. That is, depending on the estimation method, an increase in remittances by 1 percentage point o f GDP would lead to a real exchange rate appreciation o f between 2 and 12 percent. Thus doubling the remittances to GDP ratio would result in an appreciation o f the real exchange rate o f between 4 and 24 percent. We have to acknowledge here that these results are far from accurate and magnitude o f the point estimates o f the parameter p. For example, in the case o f the therefore that we have to be extremely careful especially if interest centers inthe order o f model in column (3) the 95 percent interval for the parameter o f the remittances to GDP ratio i s givenby 1.5-13. More dramatically, the specification in column (4) impliesthat in the Latin American context, the 95 percent interval for the parameter measuring the impact o f remittances on the real exchange rate i s given by 1.6-22. The basic findings described above are maintained when per capita growth i s included in the real exchange rate equation. Columns (V) to (VIII) o f table 6.2 are a replica o f columns (I) to (IV) but now also including among the regressors per capita growth, which in these exercises i s assumed to be exogenous (i.e. we do not instrument for it even when using IV estimation techniques). Columns (IX) and (X) would instrument for per capita growth as well as for remittances. On the whole, the basic message that remittances will contribute to a real exchange rate appreciation i s maintained regardless o f the specification considered (including or not growth, instrumentingor not for remittances and growth). However, as inthe case o f equations (11) and (IV) the driver behind the estimated impact o f remittances on the real exchange rate seems to be the Latin American region. In fact, outside Latin America, we cannot reject the nullthat remittances have no impact on the real exchange rate. As for the estimates ofthe rest o fthe variables inthe control set, it is worth noting that they do not change much with respect to the specifications that did not include per model in column (V) - and the point estimates are negative in all but one specification. capita growth. Interestingly the per capita growth rate i s only significant in one case - That is, judging from point estimates countries that grow faster would find that the growth effect tends to compensate for the remittances effect. On the whole, this could 182 indicate that the portfolio balance effect inequation (9) somewhat dominates the Balassa- Samuelson effect. Remittances and real exchange rate misalignment What's the driver behindthe real exchange rate appreciation that one could expect to see following a surge in workers' remittances? I s it a reaction o f the real equilibrium exchange rate to a positive shock and hence a natural adjustment to a new equilibrium? Or instead, is it a temporary deviation from the equilibrium?We now empirically assess this issue and proceed to regress David Dollar's measure o f real exchange rate misalignment (see Dollar, 1992, for details regarding the calculation o f this variable) on the same set of regressors discussed above. That is, we now estimate equation (3), using David Dollar's index to proxy for ii,.The results o f this exercise are presented in table 6.3. lo2 Inspection o f this table indicates that the parameter o f remittances tends to be significant inthe Latin American countries but insignificant inthe rest o f the world. That is, inthe Latin American context it would be difficult to defend that all the changes inthe real exchange rate would be driven by an adjustment towards the new equilibrium. On the other hand the point estimates o f the impact o f remittances on the degree o f real exchange rate misalignment are always smaller than the point estimates o f the impact o f remittances on the observed real exchange rate, an indicationthat the observed changes in the real exchange rate are a combination o f adjustment towards the new equilibrium and some apparent overshooting. For example, the basic models in columns (I) and (V) would suggest that only one-fourth o f the observed change in the real exchange rate would be due to adjustments inthe real equilibriumexchange rate (Le. an estimated impact o f remittances on the real exchange rate o f 1.4-1.5 against an estimated impact o f about 1.1 on the measure o f misalignment change). This share falls to about one-tenth in the models in columns (11) and (VI) (estimated impacts o f 2.0-2.1 against 1.9 for the exchange rate and misalignment measure respectively). On the other hand, the IV based estimates would suggest that between one-third and one-half o f the fluctuations would be due to equilibrium adjustments. On the whole, we take these results as an indicationthat the observed fluctuations inLatinAmerica are to a large extent drivenby the misalignment component. However, for the rest o f the world the results in tables 6.2 and 6.3 combined would suggest that remittances have no effect on either the real exchange rate or any o f its components (equilibriumor misalignment). lo* We would like to acknowledge here that the results in Table 6.3 are likely to be influenced by the particular measure o f misalignment used (i.e. the one by Dollar, 1992). Unfortunately, data availability limits even the possibilities to measure the degree to which the results are influenced by that particular choice. 183 3 V. Conclusionsand policyrecommendations Inthis chapter we reviewedthe impact of workers remittances on the real exchange rate and concluded that although remittances do not seem to have an impact on the real exchange rate outside the Latin American region, inLatinAmerica they are likely to lead to a real appreciation. Moreover, we have also explored whether the estimated appreciation in the regional context would be consistent with the natural appreciation that one would expect in the real equilibrium exchange rate following a positive shock, or instead whether the observed changes are more likely to be drivenby changes inthe misalignment component. The findings o fthe chapter would indicate that not all o f the observed changes are consistent with the equilibrium changing according to the new fundamentals. Against this background a natural question that arises i s what Latin American policy makers can do about that real appreciation and therefore about the potential loses in international competitiveness that may come with large remittance flows to the region. We now discuss these issues. 0 Rein in fiscal policv. Fiscal restraint is probably one o f the only tools that governments have to prevent overheating, and avoid a real exchange rate appreciation inthe context o f a surge in international workers remittances. Beyond the theoretical reasoning in support o f this tool,'03 the estimates presented in this paper indicate that increases in the government consumption to GDP ratio would be associated with real appreciations. Yet, it must also be noted that our estimates indicate that the impact o f this variable tends to be much lower than the impact o f remittances. In other words, the adjustment needed to stabilize the real exchange rate may be quite large and therefore constrained by political economy considerations. 0 Avoid sterilization. One natural question in this context i s the extent to which countries should try to sterilize the remittances inflows. Sterilization can be defined as the exchange o f government paper for foreign exchange so that the monetary base is insulated from the remittances flows (other sterilization-type policies would include increases in reserve requirements on all or selected parts o f bank deposits). Sterilizing operations could be effective if used over the short run, but may prove infeasible if needed on a sustained basis for two main reasons. First, the magnitude o f the remittances would make the quasi-fiscal costs o f sterilizing these flows untenable. Large remittances inflows coupled with Latin American spreads that for the ten top receiving countries range from 141 bp in Mexico to almost 300 bp in Jamaica would in fact make this alternative extremely expensive (to the point that even assuming no pressure on the domestic interest rate, in a number o f countries the cost o f sterilizing the inflows in full would be measured in tenths o f percentage points). Second, sterilization would possibly put pressure on the domestic interest rates something that may attract other type o f inflows in search o f high returns and this in turn would put more pressure on the IO3That fiscal expansions (contractions) lead in the presence o f perfect capital mobility to a real exchange rate appreciation (depreciation) is a typical result o f the basic IS-LMmodel. 185 exchange rate. In this regard, if sterilization i s implemented without fiscal adjustment, (Le. tight money plus loose fiscal) it would not be unlikely to observe a further appreciation. 0 Microeconomic interventions. Although the thrust o f responses to surges on capital inflows o f any type (including remittances) may be expected to be in the macro policy arena, there are a number o f microeconomic interventions that governments can implement. The discussion in section I1 above suggested that rigidities in labor and product markets could contribute to a real appreciation inthis context because o f Balassa- Samuelson type o f arguments. Thus efforts aimed at making domestic markets more efficient could also ease exchange rate pressures. More generally, microeconomic interventions that make the economy more competitive could somewhat offset the real exchange rate pressures. 0 Finally. accept some amreciation. Taking together all the elements inthe paper and to the extent that fiscal adjustment and microeconomic interventions may not be enough to correct the upward pressures in the real exchange rate, it is possible that Latin American policy makers will have to accept some real appreciation, especially in those countries with substantial flows. This loss o f competitiveness, however, should not be viewed as a cost associated to remittances but rather as a reflection o f the changing conditions brought by the significant remittances flows. 186 TechnicalAnnex The argument that a real exchange rate appreciation is a natural outcome in the presence o f remittances can be illustrated with a simple model o f exchange rate determination. Here, we follow Alberola et al. (1999) and assume that there are two countries in the world, each producing two goods: one tradable (subscript T, in what follows) and one non-tradable (N). The real exchange rate (q) i s defined as the relative price o f domestic to foreign goods in the consumption basket, p andp*,re~pectively,'~~expressed indomestic currency: q =p-(s+p.) I where s is the (log) nominal exchange rate, defined as the price of foreign currency in terms o f domestic currency. Thus, an increase inq represents an appreciation o f the real exchange rate. The consumer price index (CPI) for each country is a weighted-average o f the tradable, non-tradable, and imported (tradable) prices, all expressed intheir own currency: where the as are the weights of the respective goods inthe consumer basket. Substitutingthese expressions in (l), assuming that aN= a;, and rearranging terms we obtain where: .. qx = bT-(s +pi)] i s the relative price of domestic to foreign tradables and q, = [(pN-p,) -(pi -pf )] i s the price o f non-tradables relative to tradables across countries. The first component o f (3) ((1-aT-a;)qx) captures the competitiveness o f the economy and determines the evolution o f the foreign asset position, while the second (aNq,) plays a central role inadjusting excess demand across sectors inthe economy. Each relative price adjusts to achieve equilibrium in one of the markets, and hence we will denote qx and qr as the internal and the external relative prices, respectively. The equilibrium exchange rate ( q,where the bar denotes equilibrium values) will require simultaneous equilibrium in both markets, and thus it will be a combination o f the equilibrium internal and external relative prices. We next characterize the external and internal equilibrium o fthe economy. IO4An asterisk denotes foreign variables. 187 External equilibrium Portfolio models o f real exchange rate determination (Mussa 1984) focus on asset equilibrium, as defined by the attainment o f agents' desired foreign asset stock. Over time, the accumulation o f net foreign assets (F)i s givenby the current account balance (CA),which equals the trade balance (AN),plusthe net income that residentsreceive (or pay) onF: AF= CA=Jw+i *F (4) where i* is the international interest rate, which is assumed given. It will be more convenient to focus on the trajectory o f the foreign asset stock relative to GDP, which can be written Af=ca=xn+(i *-@f (5) wheref andxn denote the ratios to GDP o f the respective uppercase variables, andg is the rate o f GDP growth. Ifthe Marshall-Lerner condition holds an increase in the relative price o f domestic tradables qx shifts consumption toward foreign tradables and worsens the trade balance. Consistent with this interpretation, it i s plausible to assume that the trade balance as a percentage o f GNP (xn) is given by: The capital account deficit reflects the desired rate o f accumulation o f net foreign assets by the home country, which is assumed to dependon the divergence between the current level o f assets as.a percentage o f GNP (n and the desired equilibrium level ( f ) , itself determined by exogenous factors such as saving preferences and demographics which will not be modelled here: Af = ca = a(? -f ) a>O. (7) Equation (7) indicates that if the actual net foreign asset position is below its desired level, agents will accumulate assets to reach the target; conversely, iff is greater than f agents will be reduce their asset holdings untilthey reach 7 . Equating (7) and (5) after using(6), and solving for qxwe get: Equation (8) shows that the external real exchange rate depends on (i)the divergence between current and equilibrium asset holdings; and (ii) the current stock o f net foreign assets$ Defining the equilibrium external real exchange rate qx as that consistent withf =f (Le. the exchange rate consistent with asset holdings at their equilibriumlevel) it follows that 188 From equation (9) it follows that improvements in the equilibrium net foreign asset position $would lead to a real exchange rate appreciation. Similarly, increases in the international interest rate i* would also lead to a real exchange rate appreciation. On the contrary a higher growth rate would be associatedwith a lower equilibriumreal exchange rate. Internal equilibrium The differential behaviour o f sectoral relative prices between countries determines the evolution o f the internal real exchange rate. Sectoral prices are inturn related to the evolution o f sectoral productivity. These notions can be illustrated usinga simple model with two production factors, labor (L) and capital (K). Output in each sector is determined by a Cobb-Douglas productiontechnology: where 0<8,6<1 represent the intensityof labor in each sector. Labor is perfectly mobile between sectors (but not across countries), implyingnominal wage equalization: Labor i s paid the value o f its marginal product dY@Lj=W/Pj. Under Cobb-Douglas technology the ratio o f marginal productivities i s proportional to the ratio of average productivities: From (12) it follows that the (log) sectoral price differential i s equal to the labor productivity differentials plus a drift capturing the relative intensity o f labor. Expressing with lower case the natural logarithms o f sectoral labor productivities, (12) reduces to Neglecting constant terms and denoting n =[(yT-y N )-(yi -y,)], the internal equilibriumexchangerate isjust: Thus in line with the argument put forward by Balassa (1964) and Samuelson (1964), productivity differentials between the tradable and non tradable sectors relative to the foreign country will also affect the evolution o f the real exchange rate. In particular, productivity gains 189 in the domestic tradable sector relative to the domestic non tradable, would result in a real exchange rate appreciation. Remittances and the real exchange rate First, remittances may affect qx. Remittance flows will raise the net foreign asset position o f the countryfand therefore will affect the real exchange rate. When remittances have a positive impact on f (Le. when remittances affect the country's willingness to hold foreign assets increases) the evolution qx and therefore o f q would be driven by a change in fundamentals. In this particular case, remittances would lead to an appreciation o f both the real equilibriumand the observed exchange rate. When instead, remittances affectf but not f ,one would expect to observe temporary changes in qx driven by (f -f ) as indicated by equation 8 above. Remittances would also lead to an appreciation of the real exchange rate, but inthis case would not be consistent with the fundamentals o f the external equilibrium. In practice, it i s possible that changes inthe real exchange rate are driven by both changes in the equilibriumand adjustments towards the new equilibrium.This would be the case if f adjusts to the remittances but less than 1 for 1. Second, remittances can also affect 4,. Ifremittances lead to acceleration inthe demand for services, inflation will tend to be higher inthese sectors, which typically are not tradable and hence lead to a real exchange rate appreciation. Productivity differentials can also be the result of market rigidities. For example, if remittances raise the reservation wage, then excessive wage pressures in the tradable sector may lead to employment adjustments to maintain competitiveness, whereas in the non-tradable sector employers may admit these pressures because they can pass them onto prices. As a result, remittances can also lead to higher productivity growth and lower inflation in the tradable sector through their potential impact on the reservationwage. A third possibility for remittances to affect the real exchange rate is through their impact on growth (see Chapter 3 for evidence in this regard), although in this case the impact on the exchange rate i s likely to be uncertain. On the one hand, looking at the expression for the external equilibrium, an acceleration in the growth rate would lower the stock o f net foreign assets as a percentage o f GDP and hence this would lower the real exchange rate. Note here, that faster growth would also lower the stock o f foreign liabilities as a percentage o f GDP and therefore this would act as a counterbalancing effect (Le. in countries with an outstanding liability position, a growth acceleration leads to an appreciation). In other words, a growth acceleration would reduce the weigh given to the portfolio model in the determination o f the equilibriumexchange rate. On the other hand, looking at the expression for q, ifremittances have a positive impact on growth, then one should expect a real exchange appreciation. In fact, to a large extent the Balassa-Samuelson argument is based on the observation that economic growth tends to be associated with an increase in the relative price o f non-tradables. The reason is that productivity gains tend to concentrate in the tradable sector, because it i s more capital intensive and 190 technological progress i s usually embedded in new capital. With equal nominal wage increases intradable andnontradable sectors, due to labor mobility, equilibrium wouldonly be attainedby an upward adjustment in non-tradable prices. Thus on the whole, the impact of a growth acceleration could go ineither direction or even cancel each other and have no impact. 191 Chapter 7 D o Conditional Cash Transfer Programs Crowd Out Private Transfers?* There is an extensive economics literature, both theoretical and empirical, thatfocuses on whether public spending crowds out private spending and on the implications for policy design of this potential crowding, Could then it be also possible that public transfers afSect remittances, which in essence areprivate transfers, and that increases in theformer result in at least apartial decline in the latter? I.Introduction There is an extensive economics literature that has focused on whether public spending crowds out private spending. Martin Bailey (1971) first proposed that a unit of public consumption would likely be valued as much as q (O0. We next move to explore what the empirical evidence has to say on this issue. Remittances, Human Capital and Economic Growth In our first experiment we test whether there are complementarities between education and remittances for the growth process. Empirically, we replace Complement in (1) with our indicators o f human capital: enrollment rates in primary and secondary schooling. Table 8.1 presents the estimation results for enrollment rates in each level o f education as well as for three different sets o f estimates depending upon the specific instrument used to control for potential reverse causality from growth to remittances. More specifically and as done previously in this report we rely on the Distance and Migration instruments first introduced by Aggarwal, Demiguc-Kunt, and Martinez Peria (2005). In all the cases, the models are estimated using a system GMM-IV estimator. Inspection o f this table indicates that inmost cases the coefficient o f remittances i s negative and significant, whereas that corresponding to the interaction between remittances and education i s positive and also ~ignificant."~This implies that the growth benefits o f remittances rise as the level o f human capital increases. Inother words, education and remittances seem to be complementary factors inthe growth process. The estimates in table 8.1 can also be used to compute the educational threshold for which remittances have no impact on growth. For example, using the estimates corresponding to secondary schooling and the Distance instrument we find this threshold at a low 27 percent (located in the 40* percentile o f the distribution of secondary enrollment in our developing country sample). Beyond this critical level, remittance-recipient countries would start reaping growth benefits. In fact, the marginal growth impact o f remittances for the median countries in the distribution o f secondary enrollment rates is 0.04, with a one standard deviation increase in remittances generating an increase in growth rate o f 7 basis points. In contrast, countries with enrollment rates in the 75th percentile o f the distribution have a marginal impact o f remittances on growth o f 0.19 percent per year. 'I5To save space and lower the technicality o f the discussion we do not discuss the results o f the specification tests (Sargan test of overidentification and test for second order serial correlation). Yet, it is worth noting that the tests do not reveal any particular problem with the specification o fthe model. 22I ++ " d 2 + + + + + + + + + + + + +w + + i t ++ * ++ ++ ++ ++ ++ ++ ++ ++ N N N ++ ++ ++ ++ + + + * s A Panel A o f Figure 8.5 plots the growth response o f higher remittances as a function o f the level o f secondary schooling across developing regions and across regions. It shows that, ceteris paribus, regions with higher rates of enrollment in secondary education display the largest potential growth benefits associated to surges in remittances. A one standard deviation increase in the ratio of remittances to GDP would raise the growth rate of Eastern Europe by 33 basis points per year while, on average, annual growth in East Asia and Latin America would be higher by -26and .25 percent per year, respectively. Inpanel B o f figure 8.5, we repeat the same exercise but now focusing on Latin American countries. This panel indicates that there i s substantial variation in the growth benefits o f remittances across countries, a reflection o f the significant differences in human capital within the region. For instance, 9 out of 19 countries with data on secondary enrollment for 2000 have a growth response to higher remittances below the average response for Latin America. Guatemala i s the country with the smallest potential response -an increase in the growth rate of 0.08 percent per year. On the other hand, Brazil and Argentina have the largest potential growth benefits due to higher remittances (between0.34 and 0.37 percent per year), with growth effects slightly larger than the ones observed by the average inEasternEurope. Figure 8.5. Growth, education and remittances:Impact of a one std. dev. increase in remittances PanelA. Regions of the world I 0.35 7 0.33 0.25 0.26 PanelB. Latin American Countries I 0.40 0.35 1 0.30 -I Source: Author's calculations. 223 A complementary interpretation o f the above results is that the impact o f progress on the educational front i s amplified when remittances are higher. Thus, the growth effect o f increases in secondary enrollment rates in countries which start from low levels would be even larger if remittances flows are considerable. To get an idea, assume that Guatemala, El Salvador and Nicaragua (lowest levels of secondary enrollment in 2000) reform their educational system and raise their enrollment rates to the median level in the region (Colombia, 70 percent enrollment rate in secondary schooling in 2000). The growth benefits o f this change, in excess to those attributed to their high remittances (Le. the extra bonus), would range between .07 percent per year (El Salvador and Nicaragua) and $17percent (Guatemala). Remittances, Institutions, and Economic Growth Inour second experiment we explore the potential complementarity betweenremittances and institutions and to this end we rely on the ICRG index o f political risk publishedby the PRS Group to proxy for our Complement variable in (1). Inthis subsection we also look at two o f the components o f this overall index o f institutions in order to test the forces behind the impact o f institutions and remittances on growth. Specifically, we use the indices o f political institutions and the socio-economic environment. The former sub-index, captures the responsiveness o f the government to its people. The latter sub-index, on the other hand, comprises information on the government's ability to carry out its programs and stay inoffice, the socio-economic pressures at work in society that could restrain government actions, and factors affecting risk to investment (such as contract expropriation, profits repatriation, and payment delays). Table 8.2 reports the estimates for this specification for each o f the indices under analysis usingonce again the Distance and Migration variables to instrumentfor remittances.Il6 The first three columns of Table 8.2 correspond to the Distance instrument, whereas the latter three columns use the Migration instrument. Inspection o f this table indicates that both groups o f estimations show analogous results. The estimates inthe table also indicate that the parameter of remittances enters the specification with a negative (and statistically significant) sign. Similarly, the parameter corresponding to the interaction term i s always positive and significant suggesting that the marginal impact o f remittances i s higher for countries with higher.levels o f institutional quality. Figure 8.6 shows the effects o f a surge o f remittances by region taking into account their average institutional levels. Ineach case we assume a positive shock equivalent to a one-standard deviation to remittances. The simulations indicate that in Latin America an increase in remittances would lead to a higher growth rate o f .5 percent per year, only surpassed by Eastern Europe (about .63 percent) and only slightly by the Middle East (0.52 percent). The poor institutional records o f Sub-Saharan Africa and South Asia would be reflected on their lower potential growth benefits across regions --only .36 and .25 percent respectively. 'I6Although the results are not reported, we should note that when treating remittances as exogenous the results were qualitatively similar. 224 * ** ** * ** * * 3 m 2 ** ** ** * ** * * ** ** ** ** ** ** ** ** ** ** * ** ** ** * * ** * * * * * * * ** * * * * * * * * s A n b Figure8.6. Growth,institutionsand remittances: Impactof a one std. dev. increasein remittances PanelA. Regions of the world 0.7 1 0.63 0.6 0.5 0.4 0.36 0.3 0.2 0.1 0.0 Panel B. Latin American Countries I 0.9 0.8 1 . . iource: Author's calculations Across Latin American countries, the growth response to higher remittances also varies significantly. Countries with poorer institutions, such as Venezuela, only observe their growth rates increase by approximately 0.2 percent per year following an increase inremittances. On the other hand, countries with better institutions, like Costa Rica, would experience surges ingrowth rates higher than 0.7 percent per year. Our findings here indicate that if countries improve their overall institutional framework, remittances can be hrther allocated into productive activities and, hence, foster growth. For instance, if Latin American countries were to improve their institutional framework to the levels displayed by the regional leaders, growth gains would range from 0.1percentage points (Panama) to more than 0.5 percentage points per year (Venezuela). 226 Remittances, Financial Depth, and Economic Growth As discussed above, from a theoretical point of view it is not clear whether remittances and financial sector development should complement or substitute each other. To empirically explore this issue, Table 8.3 reports the results o f a growth regression that includes an interaction term between workers' remittances and measures o f financial development such as domestic credit to the private sector and liquid liabilities and money as a percentage o f GDP. The first three columns o f this table present the results for "Private Credit." As in previous tables each column corresponds to a different instrument set for our remittance variables. The results indicate that remittances promote growth, but that the effect declines as the financial system becomes deeper -that i s p,>O and p2<0, This finding i s robust to the use o f private credit or liquidliabilitiesas measures o f financial development. For instance, for Argentina the growth benefits o f an increase in remittances"' i s 0.46 percentage points for Argentina; for Peru it would be o f .38 percent and for Brazil o f 0.32 percent. These three countries roughly correspond to the 25, 50 and 75 percentile of the distribution o f our financial development proxy. It i s worth noting, however, that our estimates in Table 8.3 suggest that the impact o f remittances on growth is always positive (the marginal impact o f remittances on growth would turn negative for levels o f financial development that are above the maximum value o f that variable inour sample). Which reasons can be behind these results? Although it i s difficult to say from a single regression like this, one can hypothesize that the substitutability between remittances and financial depth i s likely to reveal problems o f access to credit by poorer households (especially those located in rural areas and devoted to agricultural activities), who may use remittances to finance high-returnprojects in their portfolio (e.g. human capital accumulation). Inother words, remittances would make the budget constraint faced by the poor less binding, and this would be more important in countries with less developed financial sectors where the poor will have even less access to credit. In fact these results are in line with the findings o f Giuliano and Ruiz- Arranz (2005) where remittance-driven growth i s higher in countries with less developed financial systems. Clearly, this should not be interpreted as an indication that countries should move towards shallower financial sectors. As it i s clear from table 8.3 (and more generally from the growth literature), a well developed financial sector i s a critical element in any growth strategy. If anything this result would indicate that countries experiencing a decline in remittances can partially make up for it by further developing the financial sector. '" As inprevious exercises we assume that remittances increase by one standard deviation. 227 Remittances, the Macroeconomic Policy Environment and Growth We have already discussed above the arguments put forward by Burnside and Dollar (2000, 2004) arguing that the effectiveness o f international aid on raising the growth rate o f the economy may depend on a sound economic policy environment that does not generate macroeconomic uncertainty. A similar argument i s also made by Faini (2002) for remittances. To hrther explore this possibility we follow Burnside and Dollar (2000) and use the coefficient estimates o f the baseline growth regression reported inthe first column o f Table 3.10 in Chapter 3 to construct a policy environment index that comprises trade openness, inflation, and government burden: Pol = t-0.329 * Trade Openness- 0.0073 * Inflation-0.862*GovernmentBurden. In this setup higher (lower) values of Pol indicate a good (bad) policy environment, characterized by less (more) policy distortions and to the extent that a good policy environment i s a complement o f remittances we would expect a positive and significant coefficient for the parameter o f the interaction term. Clearly, in this index trade openness i s associated with a good policy environment whereas inflation and the government burdenwith a bad one. Table 8.4 reports the results o f the regression analysis where we replace the indicators o f trade openness, inflation and government burden with our policy environment index. The first three columns present the base results including the Pol indicator, whereas in the fourth to sixth columns we include its interaction between remittances as an additional explanatory variable. Inspection o f this table indicates that the coefficient estimates for the growth determinants have the expected sign: (a) there is conditional convergence, (b) growth is fostered by higher education, deeper financial markets, better institutions and a sound economic policy environment, and (c) growth is hinderedby higher levels o f real exchange rate overvaluation. As seen intable 8.4, the parameter o f remittances continues to be positive and significant with this specification. Our point estimates indicate that a one percent increase in remittances would lead to an increase in per capita growth o f about .2 percent. Moreover, when the interaction term i s included in the regression not only do we find that remittances still exert a positive impact on growth, but also that remittances are more effective in enhancing growth prospects in countries with better economic policies (as captured by our policy index). Our estimates suggest that a one-standard-deviation increase in workers' remittances will raise the rate o f growth in the recipient country by 0.32 percentage points for countries in the 25'h percentile o f the distribution o f policy environment. Growth benefits are only slightly larger for countries in the median and 75th percentile (0.35 and 0.38 percentage points per year, respectively). 229 ** ** ** ** ** * * ** ** ** ** * * ** ** ** ** * * * ** * * * * * * * *C *Y * * 0 ccr N * * * * I * I * * * Figure 8.7 reports the growth response to higher remittances conditional on the economic policy environment o f the recipient country and region. We observe that, on average, the growth response in Latin America -an increase in growth rate by .36 percent- i s in line with the average response o f the developing countries as well as the MiddleEast and North Africa region. According to figure 8.7, the regionthat would benefit the most from a surge inremittances given its policy environment would be East Asia, where an increase of (one standard deviation in) remittances would be associated with a growth acceleration of .42 percent. This finding conforms to the discussion in Section I1where we argued that the East Asian and Pacific region has the best policy indices. Within the Latin America region, growth benefits would be higher for those countries with better economic policy environments - e.g. Mexico or El Salvador (Panel B). Figure8.7. Growth,the policy environment and remittances. Growthimpactof a 1-sd increaseinremittances PanelA. Regionsof the world n45 -. .- 0.420 0.383 0.391 0.40 - 0.358 0.357 0.357 0.35 0.338 - 0.30 - 0.25 - 0.20 0.15 0.10 1- 0.05 - 0.00 I PanelB. LatinAmerican Countries I 0.40 - 0.35 - 0.30 - 0.20 0.15 - 0.10 - 0.05 - i r r r Source: Author's calculations 231 IV. Remittancesand Domestic Investment. The argument used in this chapter to justify the complementarity in the growth process between remittances and other policy interventions relies on the idea that when some conditions are present in home countries those at the receiving end o f remittances flows will have more incentives to invest the corresponding funds. It i s thus natural to test whether the complementarities between remittances and other policies also extend to the determinants of domestic investment."' In order to investigate this possibility we now proceed to estimate an empirical model that relates the investment rate to a set o f controls, remittances, and an interaction between remittances and the variables used inthe previous section to capture progress inthe four differentpolicy areasunder analysis. That is, our modelnow is: Investment rate,,,=/3,,`Z,,, +/3,R,,,+@,,, *Complement,,,+p, +77, +E,,, (2) where now Z is a set o f control variables (see Box 3) including the lagged investmentrate and the rest ofthe notation is as inequation (1). Box 3. Controlset in the empiricalmodelfor the investment rate. The empirical equation for the investment rate is based on the simple accelerator model relating the investment rate to per capita GDP growth and the cost o f capital, which we proxy with the price o f investment goods from the PWT6.1. This basic model is augmented with a number o f variables aimed at capturing elements related to the investment climate. More specifically we also include among the explanatory variables the ratio o f total secondary enrollment (regardless o f age) to the population o f the age group corresponding to that level is the proxy for human capital, the stock o f claims on the private sector by deposit money banks and other financial institutions, expressed as a ratio to GDP as a measure o f financial sector development. The empirical model also includes the ICRG index o f political risk to capture country institutional aspects, and the average annual CPI inflation rate as a proxy for the macroeconomic environment. Finally, the model also takes into account potential dynamics inthe investment rate by includingthe lagged dependent variable inthe control set. Table 8.5 reports the results o f this new exercise. Each column inthe table corresponds to one o f the areas identified above as a potential complement to remittances (Le. education, institutions, financial sector development, and the policy environment) and the reported results as based on the system GMM estimator using the Distance instrument.'" To a large extent the results echo those found inthe previous section for the determinants o f growth. 'leNote inany case, that failure to support this hypothesis would invalidatethe results ofthe previous section. `I9A full set of results can be found in "Remittances, growth, and policy complementarities" by C. Calderon, P. Fajnzylber,and H.Lopez. 232 * * * * * * * * 3 I * * ** * * * * ** ** ** ** * * * m m N 3 I n -.- GI n 0 GI 0 .-c 3 c a $ A First,remittances and human capital appear to be complementary factors inthe process o f physical capital accumulation. For instance, the investment effects o f remittances are higher as the enrollment rate in schooling increases. Similarly, the marginal impact o f remittances on investment is larger for countries with higher levels o f institutional quality and better policy environment as captured by our policy index. On the contrary, and consistently with the findings in Section 111, we also found that remittances and the development o f the domestic financial market appear to be substitutes inthe capital accumulation process. In fact, according to the estimates in table 8.5, the impact o f remittances i s positive and decreasing as domestic financial markets become deeper. The marginal impact o f remittances on investment becomes negative and (in most cases) not statistically significant for higher levels o f financial development (beyond 85th and 90th percentiles). Thus, the largest increases in domestic investment driven by remittances take place inless financially developed countries. For instance, for countries with levels ofprivate credit to the domestic sector (as a percentage o f GDP) in the 25th percentile o f the distribution (2.65 in logs) a one standard deviation increase in remittances would raise the investment coefficient by almost 1percentage point o f GDP - from an average o f 20.3 to 21.3 percent o f GDP. Inturn, this higher investment would be transformed into an increase in the growth rate of the economy o f 0.3 percentage points per year. On the other hand, for those countries with an average level o f financial development the marginal impact o f remittances on investment i s 0.015. Inthis case, an increase in the ratio o f remittances to GDP would lead to higher investments by approximately 0.5 percentage points o f GDP and this would be transformed, in turn, in a growth increase o f 0.17 percent peryear. V. Concluding remarks Inchapter 3 of the report we argued that remittances are likely to have a positive effect on growth and poverty reduction. This chapter has extended the analysis there and has explored the extent to which policy makers can enhance the development impact o f remittances by implementing a number o f policies that, in addition to being pro growth per se, can also result in an extra bonus inthe presence of significant remittance flows. The results discussed above suggest that the incentives to invest remittances inproductive activities may be sensitive to some structural features and economic policies in the remittance- recipient country. More specifically, it has been found that remittances are more effective in raising investment and enhancing growth incountries with higher levels o f human capital, strong institutions and good policy environments. It then follows that inorder to maximize the development impact o f remittances countries need to (i)maintain sound macroeconomic economic policies; (ii)promote human capital development; and (iii)strengthen the institutional framework. The chapter has also explored whether a more developed financial sector may also complement remittances, but our results are more suggestive o f substitution than o f complementary effects. 234 Chapter 9 The Regulatory Framework for Remittances*.. Remittance transactions are inherently private operations and, as such, regulation does not address in any way the allocation of remittance funds, which receivers clearly have thefreedom to spend or invest as they choose. Within this scope, regulatory concerns are aimed at facilitating the provision of formal remittance services at the lowest costpossible to as many users as possible, while maintaining high levels of security in the system. In most cases, regulatory issues are linked to the expansion of formal remittance channels, rather than with the regulation of informal ones, which is inherently dijjcult. I.Introduction For migrants sending money home, remittance services have been traditionally expensive, with fees o f up to 20% o f the principal sent, depending o f the size and type o f the transfer and destination."' Fees structures themselves have been opaque (with hidden charges and poor exchange rates), and have penalized transfers o f small amounts o f the type commonly made by migrants. Moreover, cost savings achieved through technological advances in payment systems have not necessarily translated into lower prices for remittance services.'*' Accordingly, reduction o f the price o f remittances has been considered a major target for many multilateral initiatives and regulatory efforts (see next section). However authorities have shied away from imposing direct price controls in remittance services, favoring mechanisms aimed at increasing transparency, enhancing competition in the system, and reducing barriers for users to access a wider range o f service providers. Prices for remittance services have indeed decreased in those corridors where competition is highest, but have remained largely unchanged in corridors with low volume and competition. Another important dimension o f the regulation o f remittances is related to the risk o f remittance channels being used for illegal purposes, including money laundering (ML) and financing o f terrorism (FT). While the same risk i s faced by other financial sector activities, anonymous transactions, weak record keeping, non-transparent settlement systems, and the * This chapter i s based on the background papers prepared for this report "Regulatory Issues related to Facilitating Remittances Flows and Security in the System" by Emanuel Salinas Mufioz, and "The Payment System Infrastructure and Remittances inLAC: Main Issues and Policy Recommendations" by Massimo Cirasino and Mario Guadamillas "'Frias, I "' Michael. 2005 Global Economic Prospects 2006, InternationalRemittances and Migration, World Bank 235 absence o f regulatory oversight make remittance systems attractive vehicles for illicit activities.'22This i s especially the case in informal systems, which usually have much weaker security mechanisms. Regulation aimed at avoiding misuse o f the system commonly requires service providers to positively identify their clients and assess the legality o f their transactions. Compliance with these regulations i s costly, increasing the price o f the service for users, and posing barriers to entry or formalization o f new Remittance Service Providers (See Box 1). Similarly, these regulations have made many financial institutions reluctant to service sectors o f the population that cannot demonstrate lawful residency in the host country which, by some accounts, could be as much as 12 millionpeople inthe case o f the UnitedStates.123 Inthis respect, the main challenge for authorities is to ensure the integrity o f the system by restricting the opportunities for misuse while minimizing the disruption and cost o f the service for bona fide participants. Box 1. Formality inthe provisionof RemittanceServices It is difficult to distinguish between the "informal" and "formal" remittance systems. Neither its cash basis nor the occasional character o f its customer base provide a basis for saying that a transfer inwhich a person sends money through an unregisteredtravel agent (informal) i s fundamentally different from a similar transaction conducted through a credit union (formal). Definition problems are compounded by the lack o f consensus about terminology, ranging from "informal funds transfer systems," "alternative remittance systems," "underground banking," "ethnic banking," to "informal value transfer system". 124 Although it is difficult to distinguishbetween the "formal" and "informal" sector, a clear distinction is whether or not a regulatory framework is applied to the remittance provider. In this paper, "informal" is defined as a provider functioning without regulation or oversight o f financial supervisors for its remittance services. Therefore, a "formal" provider i s one regulated and overseen by competent government agencies for its remittance services. In this paper, it i s important to keep in mind that remittance providers who are registered or licensed under a regulatory regime become part o f the formal sector. In countries that have put inplace a regime to regulate the remittance sector, providers that over time fail to be registered or licensed, would be illegal providers. In countries that have not yet issued regulations for remittance providers (Le., no registration or 122IMF(2005) According to Passel (2006), the Current Population Survey by the Pew Hispanic Center shows that there were 11.1 million unauthorized migrants inthe United States inMarch 2006. The Center developed an estimate o f 11.5 to 12 million for the unauthorized population as o f March 2006. `24A more detailed description o f the background, workings, and settlement methods o finformal remittance systems is provided inEl Qorchi et a1 (2003) and will not be further elaborated here. 236 licensing requirements or a supervisory structure), it seems appropriate to refer to a regulated formal sector and an unregulated informal sector, where the informal sector would not necessarily be deemed illegal. Informal providers may interact with the banking (formal) sector for some aspects o f their activities. For example, informal providers may utilize the payments mechanism of banks for settlements, but they also settle balances outside the banking system, for instance through other money transfer operators, trade transactions, and/or cash couriers. Because formality usually entails observance o f specific operational requirements, formal Remittances Service Providers (RSP) face compliance costs and rigidity in their processes. As opposed to this, informal remittance systems can be attractive to users because o f their speed, low cost, convenience, versatility, and potential for anonymity. Accordingly, a major challenge for authorities seeking to strengthen the integrity o f remittance systems i s to have a suitable regulatory framework to bring the informal remittance providers into the formal arena. Effective regulations should not impede the flows o f remittances nor drive remittance systems underground through excessive regulatory requirements. This chapter is divided in five sections. Section 2 summarizes the multilateral initiatives to enhance remittance flows through regulatory mechanisms. Section 3 addresses competition in remittance services in light of regulatory requirements and access to payment systems. Section 4 focuses on accessibility considerations, reviewing the obstacles for remittances senders and receivers to use all formal channels. Section 5 addresses regulatory measures to ensure security inremittance services and Section 6 proposes policy recommendations and lists areas for further action. 11.Multilateral Initiatives In January 2004, the Presidents of the Americas in the Declaration of Nuevo Leon, MCxico, stated: "We recognize that remittances are an important source o f capital in many countries o f the Hemisphere. We commit to take concrete actions to promote the establishment, as soon as possible, o f necessary conditions, inorder to achieve the goal o f reducing by at least half the regional average cost o f these transfers no later than 2008 and report on progress achieved at the next Summit o f the Americas inArgentina in2005. We will adopt, as needed or appropriate, measures such as: the promotion o f competition between the providers o f these services, the elimination o f regulatory obstacles and other restrictive measures that affect the cost o f these transfers, as well as the use o f new technologies, while maintaining effective financial oversight." Furthermore, at their meetings in Boca Raton, Florida inFebruary 2004, the G-7 Finance Ministers and Central Bank Governors included in their general statement the following commitment: "We aim to reduce the impediments that raise the cost o f sending remittances". Duringthe G8 Summit inthe US, inJune 2004, it was stated that "G8 countries will work with 237 the World Bank, IMF, and other bodies to improve data on remittance flows and to develop standards for data collection in both sending and receiving countries. G8 countries will also lead an international effort to help reduce the cost o f sending remittances. The developmental impact o f these flows may be fostered by increasing financial options for the recipients o f these flows. The objectives o f the G8 programs are: 0 Make it easier for people in sending and receiving countries to engage in financial transactions through formal financial systems, including by providing access to financial literacy programs, where appropriate, and by working with the private sector to extend the range and reach o f these services. 0 Reduce the cost o f remittance services through the promotion o f competition, the use o f innovative payment instruments, and by enhancing access to formal financial systems in sending and receiving countries. In some cases, remittance costs between sending and receiving countries have been reducedby up to 50 percent or more. G8 countries believe that similar reductions o f highcosts could be realized inthe case o f other countries. 0 Promote better coherence and coordination o f international organizations that are working to enhance remittance services and heighten the developmental impact o f remittance receipts indeveloping countries. 0 Encourage cooperation between remittance service providers and local financial institutions, including micro-finance entities and credit unions, in ways that strengthen local financial markets and improve access by recipients to financial services. 0 Encourage the creation, where appropriate, o f market-oriented local development funds and credit unions that give remittance-receiving families more options and incentives for productively investing remittance flows. 0 Support dialogue with governments, civil society, and the private sector to address specific infrastructure and regulatory impediments. For example, governments should ensure non-discriminatory access to payment systems for the private sector, consistent with strong supervisory standards, and work together to modernize overall financial infrastructure." Against this background, the World Bank and the Committee on Payments and Settlement Systems (CPSS) convened, in November 2004, a Task Force to address the needs o f international policy coordination for remittance systems. The output from this Task Force forms a basis for the development, regulation and oversight o f remittance systems inthe future. Following its mandate, in March 2006 the Task Force produced a report on "General Principles for International Remittances service^"'^^, describing key features and functions that should be satisfied by remittance systems, providers and financial intermediaries. These principles are intended to be clear and universally applicable international standards, its main '" Henceforth referred to as "the General Principles." 238 focus being to identify the main characteristics o f sending and receiving remittances and the related infrastructures with a view to improving them. In this regard, the General Principles cover areas such as transparency and consumer protection, payment system infrastructure, legal and regulatory environment, market structure and competition, and governance and risk management. The Report also identifies what the role o f the remittance service providers and authorities should be to achieve the public policy objective o f a safe and efficient market for remittance services. The Principles and related roles o f the authorities and remittance service providers are summarized inBox 2 below and are discussed indetail throughout the chapter. In parallel with the finalization of the General Principles, the Bank is developing, together with other international financial institutions (IFIs), a Guidance Note with detailed guidelines and actions for the implementation o f the General Principles, as well as practical suggestions on how to organize a stocktaking exercise. Remittance systems in sending and receiving countries will then be assessed against this framework. The methodology will be available in May 2006. The Bank plans to be involved in these assessments together with other development banks, the IMF and authorities o f sending and receiving countries. The Bank, in cooperation with other International FinancialInstitutions, will also support the implementation o f policy recommendations and action points that stem from these assessments. Inthis regard, any necessary actions will be integrated inthe context o f the reform o f national payment systems, a process in which the Bank has been involved in more than 70 countries over the past 12 years. Through these reforms, many o f the preconditions for increasing efficiency in the provision o f remittance services have been already created, both through better retail payment systems and by making payment system oversight more effective. Box 2: the GeneralPrinciplesand relatedroles The general principles are aimed at the public policy objectives o f achieving safe and efficient international remittance services. To this end, the markets for the services should be contestable, transparent, accessible and sound. Transparency and consumerprotection General Principle 1. The market for remittance services should be transparent and have adequate consumer protection. Payment system infrastructure General Principle 2. Improvements to payment system infrastructure that have the potential to increase the efficiency o f remittance services should be encouraged. Legal and regulatory environment General Principle 3. Remittance services should be supported by a sound, predictable, non- discriminatory and proportionate legal and regulatory framework inrelevantjurisdictions. 239 Market structure and competition General Principle 4. Competitive market conditions, including appropriate access to domestic payments infrastiuctures, should be fostered inthe remittance industry. Governance and risk management General Principle 5. Remittance services should be supported by appropriate governance and risk management practices. Roles of remittance serviceproviders andpublic authorities A. The role of remittance service providers. Remittance service providers should participate actively inthe implementation o f the General Principles. B. The role of public authorities. Public authorities should evaluate what action to take to achieve the public policy objectives through implementation o f the General Principles. 111.EnhancingCompetition As mentioned before the cost o f sending remittances has been traditionally high, with fees o f up to 20% depending o f the size, type and destination o f the transfer.126By some accounts, reducing the cost o f remittances to 5% o f the amount remitted would free up more than US$ 1 billion per year for migrants and their In turn, the positive effect in the disposable income o f poor migrants i s deemed to enhance their incentives to sendmoney home. The high fees charged by RSPs is a major challenge for policy makers and multilateral initiatives aimed at facilitating international migrant remittance flows to developing countries. However, rather than calling for direct regulation o f prices, most-o f these initiatives focus on enhancing competition inthe market as a way to create incentives for efficiency and make prices more representative o f the costs to the RSPs. Prices for remittances have decreased in the last few years in some Latin American corridors served by multiple RSPs. A well known example i s that o f the U.S.-Mexico corridor (see Figure 9.1). In contrast, prices have remained stable or even increased in less competitive corridors. Examples include the corridors between the US and Colombia, Honduras and Guatemala.12' This evidence suggests that pricing o f remittance services i s largely determinedby the level of competition faced by operators, and does not necessarily reflect the actual costs o f the services. '26Frias(2005) 127Suro and Bendixen(2002) 12'Orozco (2004) 240 Infact, according to the 2006 Global Economic Prospects, transaction costs for RSPsare much lower and dissociated from the fees charged to users, and there is room for decrease in both costs and prices.'29However, the incentives for incumbent RSPs to realize cost savings and reduce prices are limitedinsmaller corridors where the limited volume o f operations represents a natural barrier to entry to new operators. In these cases many o f the benefits o f competition can be realizedprovidedthat there is contestability inthe market (see box 3). Figure 9.1. Range of prices of remittanceservices in the US.-Mexico corridor, 1999-2005 I (%of amount sent).13" I 200, 18 0 I6 0 14 0 I?0 _ _ % 100 B O 6 0 4 0 2 0 0 0 Pink dots and upper line represent the maximum price, red dots and middle line represent average price and blue dots and lower line represent minimum price at eachperiod. Source: ProcuraduriaFederaldel Consumidor (PROFECO) Box 3. Enhancingefficiency in low-volumecorridorsthrough contestability Even in absence o f regulatory or infrastructure barriers to entry, competition can be limited by the size o f the corridor, as the volume of operations in small remittance corridors may not afford the existence o f multiple RSPs. In these cases, many o f the benefits o f competition can be achieved if there i s contestability in the market. A contestable market i s one where transparency and absence o f barriers to entry create incentives for incumbents to provide high quality services at fair prices, as new entrants could easily start up and reap the market. Fostering contestability i s arguably one o f the most important responsibilities of authorities in low-volume corridors. The role o f authorities i s twofold: 0 Eliminateunnecessary regulatory entry requirements to new operators, and 0 Strive to ensure appropriate access to domestic payments infrastructures in fair conditions World Bank (2006a) Basedon a remittance of US$300. 241 242 encourage entities currently in the informal sector to come under the regulatory regime. However, in practice formalization requirements are far from being consistent, even within the same country (see Box l), which demonstrates that they are still not based on objective security considerations, and do not create adequate incentives for the formalization o f operators. According to the FATF, countries should define the adequacy o f licensing or registration requirements based on their specific circumstances. Licensing provides the authorities with the power to perform a pre-qualification due diligence on the RSPs' operational and basic security systems, including those related to AML/CFT considerations. Moreover, a license for a limited period o f time could allow the authorities to enforce periodic reviews o f RSPs at the time o f license renewals. Registration, on the other hand, poses significantly lower entry barriers to new RSPs and its main objective i s to encourage all RSPs to identify themselves and commit to comply with AML / CFT requirements inthe course of their operations. However, since registered RSPs are not required upfront to have systems or procedures in place for basic security, authorities are required to set up monitoring mechanisms to ensure compliance throughout RSPs' on-going operations. 133 In determining whether a licensing or a registration requirement is imposed for new RSPs, authorities must consider that regulatory activities such as surveillance and enforcement o f regulation pose significant resource requirements given the number o f RSPs that can operate at any given time. Regulatory requirements should thus be set up in a realistic way considering not only the benefits o f increasedsecurity but also the consequent costs for the authorities. Since remittance services, by definition, imply operations in multiplejurisdictions, RSPs are subject to different regulatory frameworks at the same time, which increases the complexity and costs o f regulatory compliance. The homologation o f regulatory requirements for formalization o f RSPs has been considered a way to enhance competition by providing certainty and a level playing field to new entrants. While international homologation may be difficult to achieve, authorities should strive to ensure standardization o f entry requirements at a national level. Moreover, in accordance with the General Principles, regulatory requirementsto set up a new RSP should be clear, non-discriminatory and commensurate with the type and size o f the operations o f the RSP. The issue o f proportionality o f regulation is essential to ensure a fair and competitive market. Over-regulation or unduly high requirements to entry in the remittance market increases costs which are passed over to users through price increases. Similarly, while regulatory hurdles can deter RSPs from formalizing their operations, they do not necessarily impede the operations o f informal RSPs, which can operate without proper authorization. In turn, this can create an environment o f unfair competition (where 133It is important to consider that regulatory activities such as surveillance and enforcement of regulation pose significant resource requirements for authorities, given the number o f RSPs that can operate at any given time. Regulatory requirements should then be set up in a realistic way considering the costs and benefits for the authorities. 243 informal RSPs can offer lower prices by not being subject to regulatory and where the systemis more open to misuse. Box 4. Regulatoryrequirements inthe UnitedStates as barriersto entry A recent survey'35o f RSPs operating in corridors between the U.S. and different regions showed that the single most important regulatory barrier that they faced to enter the remittance market was obtaining a license to operate, a constraint mentioned by 40 percent o f respondents. By the same token, surety bonds required to operate were seen as a constraint by 1in4 o f these firms (see figure 9.2). Figure9.2. Barriersto entry as reportedby surveyed RSPsinthe U.S. 04 I4 10% 15% 00% 15% 30% >$% 40% 1 0 udicatli ah i e s matedat LeastmpuibyrlwU5 n;lJlt.aw mmonmud Source: Andreassen (2006) There i s significant lack o f consistency in regulatory requirements among different jurisdictions within the United States. Indeed, the minimum amount required by regulation to set up a new RSP (including net worth and bonding) varies drastically across states, from a low US$ 15,000 inAlabama to US$1.5 million inPennsylvania. Surety bonds requirements also differ significantly. Moreover, the fact that bonding requirements in many states are fixed irrespectively o f the volume o f operations o f RSPs, creates a cost bias against smaller RSPs and diminishes the prudential merit o f this requirement. There i s no evident rationale for these divergences in regulatory requirements across states, as they do not appear to be proportionate to the overall volume ofremittances ineach state (figure 134According to Andreassen (2005), around 55% o f RSPs surveyed in the U.S.considered that informal competitors are a considerable or major obstacle, versus less than 30% that had the same view regarding formal competitors. 244 9.4) or each operator. In turn, this fragmented and inconsistent regulation within the United States has acted as a significant constraint to new operators, limiting competition and thus helping to maintain the preponderant situation o f incumbent RSPs. Figure9.3. Minimum amount requiredto Figure9.4. Range in bondingrequirements set up a new RSP (USD$Thousand) and level of remittancesby state I 1,600 IlsllBond (Min) -Bond (Max) +,-Remittances 7 Minimum Net Worth EMinimum Bond 1,400 100 13 f - 1,200 8 0;J 1,000 7 0 P* 800 6 0 I$ ts -- 5 0 600 3 2 . z m 4 0 8: 400 3 02P : LL 200 200 20 5, 100 1 0 Source: Ratha and Riedberg (2005) Source: Ratha and Riedberg (2005) and IADB(2004) Therole ofpayment systems infacilitating competition RSPs require constant access to payment and settlement systems in the course of their operations. The degree o f development o f those systems and the extent to which new RSPs can access them largely determines the potential for competition in the market. Indeed, "remittance services, except perhaps those that are entirely cash based, dependat some stage on the domestic payment infrastructure for settlement (and sometimes also for transfer of information)". 136Even though from a technological perspective access of new RSPs to existing payment systems should be relatively easy to achieve either directly or indirectly (through a bank), many RSPs face formal restrictions to accessingthese systems. Direct access to national payment systems i s normally granted to well capitalized and established banking institutions. Indeed, in a recent survey, De Luna and Martinez (2005) found that 35 out of 40 authorities in different countries were not favorable to openingthe clearing and settlement systems to smaller institutions, which are perceived as lacking the required technology platform, and hence would arguably pose a risk to the system. 135This case study draws on data and conclusions from Andreassen (2006). All figures are taken from the aforementioned document unless otherwise stated. 136CPSS-WB Consultative Report: General Principles for InternationalRemittance Services, March 2006, pages 16- 17. 245 RSPs can have indirect access to payment and settlement systems through banks. Intheory, this should not create unfair competitive disadvantages to RSPs as regulation o f payment and settlement systems should ensure that indirect access to RSPs i s provided in fair conditions. However, there are concerns that indirect access could be unduly constrained due to competition (when both the stand-alone RSP and the bank provide remittance services), or regulatory concerns. For non-bank RSPs operating in the US., accounts with commercial banks are their only way to access the official payment and settlement systems and hence an essential component in their operations. However, more than 40 percent o f RSPs recently considered that limitedaccess to the bank settlement system was an obstacle for their on-going operations. More importantly, 65 percent o f respondent RSPs in the same survey reported problems opening and maintaining accounts with US. banks. This may be linked to the fact that many banks are concerned that participation in remittance activities could make them subject to heightened regulatory oversight and costs, even though authorities have expressed that this i s not the case.13* According to the General Principles, the problem o f accessibility to payment systems, as well as the challenges for cross-border settlement can be addressed through bilateral or multilateralefforts linking existingnetworks in sending and receiving countries (see box 5). Box 5. Buildingcross-border payment The safety and efficiency o f remittance services can be affected by payment systems in the relevant markets and the way that these systems are accessedand usedby RSPs or by banks acting for RSPs. In addition to improvements in the domestic payment infrastructure, the safety and efficiency of cross-border remittances may be hrther improved by the coordination andor adoption across the relevant payment systems of, for example, communications standards and payment message formats that facilitate greater interoperability as well as rules, procedures and operating hours that support straight- through processing. It may also be possible to link the relevant domestic retail payment systems o f sending and receiving countries, particularly where the domestic payment systems inboth countries are well developed and have wide geographical coverage and where remittance volume betweenthe countries is high. Sometimes such initiatives may be undertaken by the market itself. However, given, first, the diverse nature o f the institutions involved and thus the potential for conflicting interests and, second, the uncertainty about the scale of h t u r e flows and thus whether investment in the initiative is justified, in many cases the authorities, and in particular central banks, may want to facilitate the consideration o f these possibilities. In general, cross-border or cross-system initiatives require a high level o f bilateral (or possibly 137Andreassen (2005). 13*Bair (2005). 139The text o f this box is quoted from the CPSS-WB Consultative Report: General Principles for International Remittance Services, March 2006, pages 16-17 246 multilateral) cooperation on technical, regulatory and oversight matters and, accordingly, the extensive involvement of central banks, regulators, payment system operators, banks and bankers' associations and other industry representatives from both jurisdictions. In some cases central banks themselves have established bilateral cross-border links between the payment systems they operate. There are several other initiatives ongoing to evaluate ways to expand the use o f existing international networks and platforms (e.g. the major international card networks, SWIFT) to provide new or improved remittance services. Also particularly important could be international initiatives to standardize the message formats used by individual payment systems and the international banking community generally, since, even without direct links betweendomestic payment systems, standardized formats could do much to enable banks and other RSPs to process payment instructions without the need for expensive manual intervention. InLatin America, efforts to develop cross-border payment and settlement systems have been conducted through private or public initiatives with different degrees of success. For example official efforts to link US payments and settlement systems to Mexican banks (FEDAchI4') to eliminate costly wire-transfers have failed to raise significant interest from commercial banks largely due to revenue expectation^.'^' Conversely, private efforts to link networks of Credit Unions appear to be successful at attracting new entrants and providing remittances services at a significantly lower cost (see Box 6). However, these arrangements can create barriers to new competitors if they require exclusivity. This i s especially the case when the disbursing RSP has a large network such as those o f the post office, telecom companies or large store chains. While exclusivity agreements betweenprivate businesses are difficult to prevent, local governments should ensure that public networks are open to different RSPs rather than beinglimitedby exclusivity agreements. Box 6. Linkingcreditunionsthroughthe InternationalRemittanceNetwork(IRnet) The IRnet is a platform created by the World Council of Credit Unions (WOCCU), a multinational trade organization, creating a direct link for remittances between credit unions inorigin and destination countries worldwide. '42 The stated objective of the IRnet is offering lower-price remittance services. Savings to I4OThe Federal Reserve Bank's Automated Clearing House system for Mexico. 14'Exchange rate differentials in operations through the FEDAch facility are kept by the Central Bank. At the same time, many large commercial banks in Mexico can achieve higher profitability inremittances transactions when they use their own internal cross-border systems (as many are subsidiaries of foreign banks) or closed agreements with capturing agents inthe U.S. 14'The coverage o f IRnet includes Latin America, Asia, Africa, Europe andAustralia. 247 users appear significant with a US$l,OOO remittance from the U S to Mexico costing US$10'43versus US$30'44for a similar service with a leading Money Transfer Operator (MTO). Access to IRnet for new participants appears fairly simple and exceptionally affordable, with signup fees between US$300 and US$750 (depending on the size o f the credit union) and maximum quarterly fees o f US$150. But most importantly, transaction fees are between US$0.25 and US$0.75. Even though the IRnet is primarily a network for credit unions, it is open to affiliated MTOs (MoneyGram, Travelex and Vigo Remittance Corp~ration'~~),which can act as capturing or disbursingagents. An essential element for the development o f this service i s its accessibility. To this respect, after legal consultation, WOCCU has issued clear guidelines to member Credit Unions in the U.S. aimed at facilitating the access of undocumentedmigrants to financial services providedby these en ti tie^.'^^ ISource: WOCCU website www.woccu.org unless otherwise stated Payment Systems Development and Oversight A sound and appropriate legal framework is generally considered the basis for a sound and efficient system o f payments, including remittance services. The legal environment should include the following: 1) laws and regulations o f broad applicability that address issues such as insolvency and contractual relations between parties; 2) laws and regulations that have specific applicability to payment systems (such as legislation on electronic signature, validation o f netting, settlement finality); and 3) the rules, standards and procedures agreed to by the participants o f a payment system. The legal infrastructure should also cover other activities carried out by both public and private sector entities. For example, the legislative framework may establish clear responsibilities for the central bank or other regulatory bodies such as oversight o f the payments system or the provision o f liquidity to participants in these systems. Finally, relevant pieces o f legislation that have impact on the soundness o f the legal framework on the payments system include: law on transparency and security o f payment instruments,terms and conditions; antitrust legislation for the supply o f payment services; and legislation on privacy. While laws are normally the appropriate means to enforce a general objective in the payments field, in some cases regulation by the overseers might be an efficient way to react to a rapidly changing environment. In other cases, specific agreements among participants might be adequate; in this case an appropriate professional assessment o f the enforceability o f these 143Source: WOCCU (2006) 144Source: Author's calculation based on `Money in minutes' service from Western Union. This figure does not include the exchange rate implicit fee. 145Vigo Remittance Corporation has been recently acquired by Western Union. The scope o f Vigo's participation on IRnet going forward is not clear. 146 The guidelines also clarify that identification requirements to open an account can be met with a foreign government document such as passport or consular identitycard (commonly known as matricula consular). 248 arrangements is usually required. Finally, since the payments system typically includes participants incorporated in foreign jurisdictions or, the payments system might operate with multiple currencies or across borders, in some cases it may be necessary to address issues associated with foreignjurisdictions. The oversight role o f the central bank i s currently at the heart o f the international debate and the function i s emerging as key in central bank activity.I4' Direct involvement o f the central bank in managing clearing and settlement systems has been, in all countries, the first step to governing the overall structure and operation o f a country's payments system and ensuring that the desire to limit systemic risk especially in the area of large-value payment systems is adequately taken into account. In many cases, this role stems from the need to ensure a widespread adoption o f the more advanced technology in the fbnd transfer mechanisms and to avoid possible discriminations in the access to payment services. In all cases, in order to pursue the public interest in the payments system, central banks should ensure that the systems they operate comply with the principles and guidelines they establish, and as overseers, ensure the (financial and operational) reliability and efficiency o f the clearing and settlement systems they do not operate. The oversight role o f the central bank is more likely to emerge in its relevance when the payments reform is complete and the central bank will be called to ensure a proper monitoring o f the reliability and efficiency o f the domestic system on an on-going basis. In recent years, in an increasing number of countries, payments system oversight has been explicitly entrusted to central banks by law (see Box 7). Specifying the objectives in relevant legislation may be the most direct way for providing a well-founded legal basis for the central bank to implement its policies and make it accountable in pursuing its goal and mandate in the payments system. For countries facing the implementation o f reforms in the payments system, it is o f utmost importance for the central bank to have a well-founded legal framework that clearly defines its payments system role and objectives. InMay 2005 the CPSS publisheda report totally devotedto payment system oversight in G10 countries. Among its main conclusions, this report stated that effective cooperation among market participants, between regulators and market participants and among regulators i s essential for the development o f a sound and efficient payments system. On one hand, the use o f payment instruments generates significant externalities on the demand side, since the usefulness o f an instrument i s strictly linked to the degree o f its acceptance and use for transaction purposes ("network economies"). Consequently, widespread use o f new payment instrumentsand services relies heavily on public confidence in them. On the other hand, within the payments system, the supply o f services can be affected by coordination failures due to the existence o f conflicts o f interests (and information costs) as well as the intermediaries' unwillingness to cooperate. This can lead to "sub-optimal" equilibria in the organizational arrangements as to the system's reliability and efficiency. One o f the main roles o f the payments system overseer i s therefore that o f making up for coordination failures inthe market for payment services. 147 Recent examples are the focus on central bank's responsibilities in the CPSS Core Principles report and the BIS/IOSCO recommendations for securities settlement systems and the paper on Payments system oversight of the Bank o f England. See also Bossone B. and Cirasino M., (2001), "The oversight of the payments system - A framework for the development and governance of payment systems in emerging economies", Western Hemisphere Payments and Securities Clearanceand Settlement Initiative ResearchSeries, CEMLA-WorldBank, Mexico City. 249 Box 7. Scope of the Oversight Function As for the scope o f the oversight function, at the international level there is consensus on the fact that systems posing systemic risks should fall under the direct control o f the overseer. Typical examples o f these systems are those that handle transactions o f a high value at both the individual and aggregate level. For example, the CPSS Task Force on Core Principles identified the main responsibilities o f the central bank in applying the core principles for systemically important payment systems. However, a broader scope including retail payment systems i s common inmany countries, especially developing ones. "Central banks have different roles in retail payments, including remittance services, depending on their responsibilities, policies and powers. It may be desirable that central banks monitor developments in the market for remittances to assess their significance for safety and efficiency. In some cases, central banks' responsibilities may also make it appropriate for them to oversee certain remittance services. To the extent that central banks provide payment services, they may be able, where appropriate, to enhance these services to support the smooth functioning o f international remittance services. Examples might include the development o f new services that support cross-border payments or enhancing existing services to make them more useful for supporting cross-border payments. The central bank should cooperate with other public authorities to address significant policy issues arising from remittance market structures and performance. Central banks may wish to enter into discussions with the private sector and other central banks to facilitate the achievement o f public policy objectives regarding remittance services and to foster international cooperation." 14' Cooperation problems may be especially relevant within inter-bank clearing and settlement systems. Infact, inthese systems the risk profiles- both at the system level and at the level o f the individual intermediary - may not be fully assessedby participants. In addition, the concern with having to support less reliable intermediaries may lead larger participants to discriminate against smaller ones, even when these are technically eligible to participate in the system. Finally the payment system industry also depends on agreements between producers to ensure that different components o f the system are compatible. Most recently, the emergence o f new types o f non-bank intermediaries and payment instruments has strengthened the need for a comprehensive level o f cooperation inthe payments system. With regard to the cooperation among regulators, the safety and efficiency objectives o f payment and securities settlement systems may be pursued by a variety o f public sector authorities, in addition to the central bank and the securities commission. Examples o f these regulators include: legislative authorities, ministries o f finance and competition authorities. There are also complementary relationships between oversight, bank supervision and market surveillance. Appropriate cooperation among supervisors can be achieved in a variety o f ways, for example, exchanges o f views and information betweenrelevant authorities may be conducted 14'CPSS-WB Consultative Report: General Principles for InternationalRemittance Services, March 2006, page 22. 250 by holding regular or ad hoc meetings. Agreements on the sharing o f information may be useful for such exchanges. IV. Transparency inthe provisionof remittanceservices Inorder to ensure that users can make informed decisions and have the ability to choose their best service option, the elimination o f unnecessary barriers to entry for RSPs, and the development o f an efficient and fair payment system, must be complemented by measures directed at guaranteeing transparency in the remittances market, together with accessibility to formal RSPs (see section 5). As discussed below, this i s a responsibility for both the authorities and service providers. The role of serviceproviders Lack o f transparency in the remittance markets has been constantly highlighted as a significant problem. Remittance senders are oRen unaware o f the different direct and indirect costs and fees charged by RSPs and therefore ignore the total price o f their remittance transaction until the money is delivered to their relatives. Moreover, many senders are unaware that there are costs in addition to the initial fee they are charged. Ina recent survey o f remittance more than half o f the respondents mentioned that the sums delivered to their relatives were lower than they expected, but they did not know why (figure 9.5). Only a small proportion o f respondents identifiedexchange rate differentials as an additional cost intheir transaction. The calculation o f the total price of a remittance services is complex and requires the collection o f various explicit and implicit components (see box 8). The components o f the price o f remittances include fees, exchange rate differentials and other less transparent implicit costs such as float and fees charged by disbursing agents. Even inpresence o f many different options for remittance services, users may find it very difficult and costly to calculate the total cost for various competing services. Accordingly, the choice o f provider may not necessarily be the best or the least costly for the users' needs, and i s likely drivenby the marketing tactics o f RSPs. The problem of transparency is exacerbated when a given RSP offers various remittance services with similar features but significantly different prices (see box 9). According to the General Principles, RSPs should disclose the total price o f their remittance services, the conditions and the characteristics o f the service in a way that is clear and easy to understand by their common users. Arguably, the way in which this information i s presented should be standardized, in order to facilitate the comparison across different RSPs. However, the General Principles do not call for direct regulation o f providers regarding disclosure o f price information. Rather they suggest that other mechanisms such as self- regulatory efforts or definition o f best practices at an industry level may prove more efficient as ways to enhance transparency. '49Suro andBendixen( 2002). 25 1 Figure9.5. Perceptions onwhy the total cost of remittanceis higherthan the flat commissionspaidby senders I 60% 1 53% 50% - II -0 5 40% - g e 30% - 26% 's 20% s - 16% 10% - 2Yo 2% 0% , , Tax Other Exchange Fee Don1know rate MY I Source: Suro and Bedixen (2002). Box 8. The calculationof the totalprice of remittances Direct fees are the most explicit component o f the price and are charged on a transaction basis. Frequently these are flat fees with little or no variation regardless o f the amount o f the transaction, which works in detriment o f smaller remittances as it can represent a large percentage o f the money sent (figure 9.6). Figure9.6. Fees as percentageof the mone sent Illinoisto ElSalvador (March - 2006). - 40% 35% 20% 14% I - 50 100 200 300 400 500 600 700 800 900 1000 Amount of the remittance (US$) Source: Author's calculationbased on information from RSP Exchange rate differential represent a second important component o f remittances costs. RSPs usually obtain additional revenues from the differential between the exchange rate in the market and the rate set by the RSP for disbursement of the remittance in local currency. It is often the case that the RSP with the lowest fees charges a high exchange I 5 OBased on the service `Money in Minutes' from Illinois to El Salvador. March 3Ist2006. 252 rate differential (see figure 9.7). Given the fact that many users are not aware o f this sdditional cost, advertisement o f fees by RSPs provides incomplete information and can be misleading. Figure9.7. Fees and exchange ratecosts."' 1 Western Union Delgado Citibank MoneyGmrn OrderExpress Ria EnLia Fees Exchangerate rewnue ource: Author's calculations based on data from PROFECO and Oanda Currency Exchange. Float. RSPs can also make revenues by holding the remittance funds for a period longer than needed and investing them in overnight transactions. Accordingly, the speed o f the service is also a factor determining the overall cost o f the remittance. Additional fees may be charged by at disbursement, especially when this i s done by agents rather than through branches o fthe capturing RSP. Box 9. Moneyin minutes or next day Western Union i s the largest RSP in the US - Mexico corridor, with the largest market share. In this corridor Western Union offers six different remittance services, which are mainly differentiated by the speed o f the service, and the disbursing agent locations. The information on total pricing (including fees and the exchange rate differential) o f only one o f the six types o f services is publicly available (through the internet). Information on total pricing can be disclosed on an ad-hoc basis to users by capturing agents. However, comparative price o f the same transaction under different types o f services are not directly provided. The total price o f the service i s usually disclosed to the users inthe receipt o f the transaction, issued after the operation has been processed. As shown in table 9.2 below, immediate availability o f the remittance at destination i s priced almost 60% higher than next-day service. Considering that the same capturing and disbursing agents (and Dresumablv same network and infrastructure) are used in both Is'Based on a US$300 remittance from N e w York to Mexico calculated with data as o f March 6 2006. The exchange rate cost was calculated using the differential between the exchange rate applied by each RSP and the average interbank exchange rate of the same date (MXP 10.5958 per USD$I). 253 cases, it i s not clear that the incremental costs incurred by the RSP for immediate delivery o f remittance canjustify the highprice differential. It is frequently the case that users cannot easily differentiate among the different services offered by a given RSP or do not know that cheaper options are also provided by the same operator. Source: Author's calculations based on information from Western Unionweb site and capturing agent. Therole of authorities in enhancing transparency in the market Even inamong the most competitive corridors, there is still a wide divergence inpricing of services. Figure 9.8 below shows the range o f prices for remittances within the same corridor and similar conditions.154 The wide divergence observed (with the most expensive service costing 235% more than the cheapest) can largely be attributed to lack o f transparency.'55 Senders are often unaware o f the prices that other providers would charge for the same type o f ~ervice"~,and this i s due to several reasons. First, it i s difficult and costly for users to collect the relevant information to make a complete assessment o f costs. Relevant information is difficult to obtain and is frequently provided only in English. Second, frequently there are costs such as exchange rate differentials and fees at disbursing agent that cannot be quantified until after the transaction has been made. Lastly, total prices for remittance services change frequently. An informed choice o f RSP would require a continuous monitoring o f the different providers. Is'Includesfees, commissionsand exchangerates differential. 153Telecommis the state-ownedtelegraphservice, still widely usedinMexico. 154Although the services vary in terms of the speed of the service and additional accessory features the basic services are similar. Given the fact that all the RSPs listed in this figure are MTOs, accessibility restrictions are equally low in all cases. A survey of remittancesenders conductedby PROFECO, the consumer protection agency in Mexico, found that 61 percent of respondents did not know how much they would be charged by a different provider for the same remittancetransaction.PROFECO 1998 254 Figure 9.8. Total price of a US$300from Chicago to Mexico (March 2006) , Fba 6lvm Western lhwn / Telecom Westernlhbn/ Next day h'ajaparaI Paga dohr MneyGram Order Express h'akpara h'axipaga Delgado Western Unwn/ mnsdiate 0 2 4 6 8 10 12 14 16 Source:Author's calculationswith data from PROFECO. The role o f the authorities is thus to actively facilitate transparency through collection and publication o f comparative prices and conditions o f service among different RSPs. These efforts can be complementedby the provision o f basic financial literacy to users inorder to foster awareness o f the different service options available to them. At the present time, there are many official efforts in this direction in recipient countries (see Box 10). However, transparency in origin countries is likely to have a higher impact as it i s senders who ultimately decide on the choice o f service provider. As per regulatory considerations, it is important to note that the General Principles suggested that a `name-and-shame' approach through the abovementioned publication o f comparative prices would create public awareness and consequently put pressure on RSPs to lower their prices. This would be an alternative mechanism to direct intervention inthe market to enforce transparency. I Box 10. Official efforts for transparency in the US -Mexico remittance market In 1998 PROFECO, the consumer protection agency in Mexico, created a unilateral program to increase transparency in the remittance corridor from the US. Although the program initially relied on Mexican Consulates in the U.S. to collect the information on transaction costs, many RSPs eventually requestedparticipating actively by providing the information themselves. At the present time, PROFECO monitors and reports on a weekly basis comparative statistics o f the prices charged by 24 different RSPs. Even though the program has expanded geographically over time, it currently monitors prices in only the nine cities with the largest concentration of Mexican migrants. Since RSPs' prices vary by state in the U.S.,there may be significant differences between the states monitored and those not coveredby the program. 255 Both commissions and exchange rate differentials have indeed decreased significantly in the states where this program operates. At the same time, price differential across different origin states have also decreased (see figure 9.9 below). However, several other factors may have also supported this improvement. A significant constraint to the program i s the limiteddissemination o f the information collected. At the present time, the main dissemination channels are postings in Mexican Consulates, a dedicated information hotline (within Mexico), and PROFECO's internet website. In this respect, diffbsion o f this information, and overall transparency in the market could be hrther enhanced by the participation o f a counterpart in the U.S. such as the Bureau o f Consumer Protection. Figure 9.9. Total price of remitting US300 from various cities in the U.S. to Mexico 1998-2005 (average of all RSPs monitored). V. Accessibilityto formal remittanceservices As mentioned inthe previous sections, the existence o fnumerous providers o fremittance services ina given corridor does not ensure the efficiency o f the market. The wide divergences in prices still observed among RSPs can be explained in part by lack o f transparency. However, these divergences can also be caused by differentiated access to specific service providers, at both the countries o f origin and destination o f the transaction. Over the past two decades, the market has evolved from one dominated by labor- intensivephysical transmission and courier services to a market dominated by cash-to-cash wire transfers through MTOs (see figure 9.10). The entrance o f financial institutions into the remittances market has made available more cost-efficient transactions such as account-to-cash and account-to-account remittances, which are slowly gaining market share. 256 14).157 MTOs People Bank Mail Credit Union tramling Source: Multilateral Investment Fund, IDB However, constraints to accessibility appear to be inversely related to the level o f efficiency o f new channels. For example, while remittance services can be cheapest through bank services (see table 9.3), a large percentage o f migrants do not have access to banking facilities. 15* Physical Informal Difficult to monitor and Lowest: no identification or delivery providers and quantify due to informal reporting requirements and arguably courier services nature of the service. few constraints to amounts. Cash-to-cash MTOs Usually highest among formal Low: identification usually required RSPs only for transactions above US$300. Foreign IDSaccepted Account-to- Financial Usually cheaper than many High: Requires that sender has a cash Institution with MTOs bank account disbursingagent. Account-to- Financial Cheapest. Can be nil due to Highest: Requires that both sender account Institutions only cross-selling of other financial and recipient have bank accounts Historically, migrants' access to bank accounts in the U.S. has been constrained by founded and unfounded considerations, o f which the most important one related to the legal status o f migrants (figure 9.11). Other constraining factors appear to be related to expectations o f costs, level o f income and, to a lesser extent, distrust infinancial institutions. Inthese cases, lack o f information, rather than regulation may be the sources o f constraints. These factors are analyzedbelow. '*'Based on a survey of 3,802 remittance senders in 37 states o f the US.and the District o f Columbia in 2004. A survey of remittance senders in 2002 showed that only in 26% o f the cases both sender and recipient o f the remittance had a bank account. 257 I Figure 9.11. Reasons for not having a bank account $ u) 45% 40% '0 35% 30% 25% b e! 20% c 15% ;10% 5% p. 0% Legal status Minimum Distrust banks Guarantees deposits I Source: Suro and Bendixen (2002) Regulatory constraints to accessibility at the country of origin'59 The regulation inthe U.S.requires that, for security purposes, financial institutions verify the identityofapplicants when opening up newbank accounts.While federal regulation does not expressly forbid the provision o f financial services to undocumented applicants, this issue i s far from being clear, creating concerns among both financial institutions and a large segment o f prospective clients. I 6 O To date, many financial institutions provide financial services only to individuals that can prove their legal residence in the U.S. (see table 9.4 below). This appears to be largely due to seemingly unfounded concerns among many banks that provision o f financial services to undocumentedmigrants or specific ethnic groups could subject them to more stringentregulatory oversight, with the consequent increasedcosts and loss of efficiency (Bair, 2005). There have beensignificant efforts both from the private and the public sector to reduce accessibility constraints to migrants. According to Orozco (2005), the Treasury Department opened access to bank accounts to undocumented individuals of Mexican nationality. In2002, the Treasury Department advised Congress that, under the terms of the regulation to ensure security in the system (USA PATRIOT Act), an official identification issued by the Mexican government (consular identity card) can be used as a valid form o f identification to open an account with a financial institution. Authorities from other recipient countries should seek to implement similar programs to provide their nationals with a form o f identification that facilitates access to banks inthe U.S. 159 This section focuses on the situation in the US., as the country o f origin o f most o f the remittances to Latin America. According to some estimates (Lowell and Suro, 2002), at least two-fifths o f the adult Latino immigrant opulation is made up o f individuals not authorized to be in the country. p61 Itis not clear whether identifications of migrants o f other nationalities would have similar treatment. 258 Table 9.4. Reauirementsto oDena bank account inthe U.S. (March 2006) 1 Citibank Bea U.S. citizen or Resident Alien Have a U.S.address Have a Social Security Number or Tax ID Bank o f America Social Security numbers Home addresses for the last twelve months Driverslicense, US Passport or State IDnumbers Credit card or bank account number for funding your new account Current email address USBank Social Security Number Driver's License, State ID or Military ID number and expiration date. Email address Harris Bank Social Security number Employment information (company name, address and phone number) Driver's license or other IDnumbers mce: Informationprovided inthe wet te of each institution. Compiledby author. Financial institutions are increasingly providing services to undocumented migrants in a simplifiedway, for example: 0 The World Council o f Credit Unions (WOCCU) has issued a recommendations documentI6' to member Credit Unions in the U.S. clarifying that a) official documents issued by foreign governments are acceptable forms o f identification for opening up an account, and b) federal regulation does not prohibit providing services to undocumented individuals. However, it highlightsthat state-specific regulations may do so. 0 The Coalicion Internacional de Mexicanos en el Extranjero (CIME) has reached agreements with various banks in Illinois in order to reduce requirementsfor opening up new bank accounts and to provide these services to undocumented migrants. However, this initiative is limited to banks in Chicago and has recently been expanded to Indianapolis. Non-regulatory determinants of accessibility From the demand side, beside the constraints posed by regulation and banks' policies, there appears to be a significant problem o f information as many undocumented migrants have the perception that banks might share their information with immigration authorities, which would expose them to deportation. By the same token, many migrants do not trust financial institutions in general, and many consider that financial services are inadequate for them. Many of these reservations appear to be based more on word-of-mouth than on first-hand negative experiencewith banks. WOCCU, "HOW serve undocumented individuals", World Council o f Credit Unions, Inc., Washington D.C. to 259 The cost o f the financial services can be significant, and is frequently considered too burdensome both by and by some market practitioners.164 Banks typically charge transaction fees and commissions for account management to users that have an average balance below certain threshold (frequently set above US$l,OOO), which i s above the means o f many migrants given their low income. Ina recent survey among migrants that use remittance services (MIF 2004) only 38% o f respondents that have a yearly income o f below US$30,000 have a bank account. This is particularly noteworthy considering that the majority o f migrants seem to be concentratedinthat income bracket. Accessibility issues in receiving countries The quality o f the financial services infrastructure in recipient countries is critical to ensure security and efficiency o f the service, and hence create incentives for senders to use more efficient channels. Even though migrants are wary o f high fees, their choice o f channel for remittances is significantly influenced by the convenience, security and reliability at the receiving end. Accessibility can be improved by enabling more financial institutions to participate in remittances. In particular, savings and loans, credit unions and microfinance companies may be well positioned to act as disbursing agents, as their networks may be closer to the usual recipients o f remittances than that o f large commercial banks. In this sense, authorities at recipient countries should ensure that there are no unduly regulatory constraints to the participation o f these entities. VI. Security Issues Remittance channels can and have been used for illicit purposes, which include money launderingI6', as well as fraud and financing o f terrorism activities.166Any funds transfer system can be misusedfor money laundering or terrorist financing. As put in an IMF report, "like their counterparts in the traditional financial sector, operators o f remittance systems could be misled 163 Sur0andBendixen(2002). 164 According to recent remarksby Citibank officials, the process of sending remittancesthrough commercialbanks i s "expensive, challenging and time consuming" both for the customer and the bank, [largely due to regulatory requirements].FromBalancingCompetition andRegulationin the RemittanceMarket, Citibank, BobAnnibale, IDB InternationalForumon Remittances2005 165 Remittance providers have been associated with facilitating tax evasion, capital flight, circumventing excise requirementsand smuggling. See ElQorchi et a1(2003) 166 Examples of remittances being used to fund terrorism, civil wars and liberationstruggles include remittances being used to acquire arms for guemlla in Somalia, remittances originated in Sweden to fund the Free Aceh Movement, in Canada for the LiberationTigers of Tamil Eelamand from the U.K.for the Kshmiricause. Basedon Kapur, Davesh, 2004, "Remittances: The New Development Mantra?", G-24 Discussion Paper No.29, UN Conference on Trade and Development, Geneva, Switzerland. And Chee Sung Lee and Maud Boekkerink "Two current issues facing developingcountries". 260 by persons in terms o f both the nature and intended purpose o f the transaction as well as the identity o fthe persons sendingandreceiving the funds inquestion."'67 The risk of misuse o f remittance channels i s highest among informal remittance providers that are completely unknown to the regulatory or supervisory bodies. As argued by the IMF, informal providers may not fully understandthe consequences o f not undertaking adequate due diligences regarding their customers' identities, the nature o f their businesses, and their possible engagement inillicit activities. Thus, "while some customers may value anonymity, this i s a key risk for ML/FT. It is believed that the risk o f misuse of remittance systems would be reduced if transfers were channeled through remittance systems that are subject to regulations and monitoring by national authorities."'68 Accordingly, the FATF considers that security in the system can be enhanced by a combined approach including: 0 Efforts by national authorities to encourage and enable the use o f formal systems (such as banks) by lowering the costs and increasing access to these systems to all users as discussed inthe previous section, and Putting in place a regulatory framework which includes licensing or registration together with AMLEFT requirements for money transfer providers. Moreover, the FATF recommends that those requirements should be the same as those followed by banks and other financial institutions. The main AML / CFT regulations are the Bank Secrecy the Patriot Act"' and directives from the Office o f ForeignAssets Control (OFAC) inthe U.S(see box 11). Box 11.AML and CFT regulations FATF recommendations. '71 Extracts from The forty recommendations, Financial Action Task Force on Money Laundering (FATF), revised 22 October 2004. Recommendation 5: Financial institutions should undertake customer due diligence measures, including identifying and verifying the identity o f their customers, when carrying out occasional transactions that are wire transfers. Recommendation 7: Financial institutions should, in relation to cross-border `67IMF (2005) ' IMF (2005) 16'Regulationincludedin31USC Sections 5311-5330 and 12USC Sections 1818, 1829 and 1951 1959 - I7ORegulationincludedinUniting and StrengtheningAmericaby Providing AppropriateTools Requiredto Intercept and Obstruct Terrorism Act of 2001. 17'Extract from World Bank (2006a) 261 correspondent banking and other similar relationships, in addition to performing normal due diligence measures: (a) Gather sufficient information about a respondent institution to understand fully the nature o f the respondent's business and to determine from publicly available information the reputation o f the institution and the quality o f supervision, including whether it has been subject to a money laundering or terrorist financing investigation or regulatory action. (b) Assess the respondent institution's anti- money laundering and terrorist financing controls. (c) Obtain approval from senior management before establishing new correspondent relationships. (d) Document the respective responsibilities o f each institution. (e) With respect to "payable-through accounts", be satisfied that the respondentbank has verified the identity o f and performed on-going due diligence on the customers having direct access to accounts o f the correspondent and that it i s able to provide relevant customer identification data upon request to the correspondent bank. Recommendation 8: Financial institutions should pay special attention to any money laundering threats that may arise from new or developing technologies that might favor anonymity and take measures, if needed, to prevent their use in money laundering schemes. Inparticular, financial institutions should have policies and procedures inplace to address any specific risks associated with non-face to face business relationships or transactions. Recommendation 23: . Other financial institutions should be licensed or registeredand , , appropriately regulated, and subject to supervision or oversight for anti-money laundering purposes, having regard to the risk o f money laundering or terrorist financing inthat sector. At a minimum, businesses providing a service of money or value transfer, or o f money or currency changing should be licensed or registered, and subject to effective systems for monitoring and ensuring compliance with national requirementsto combat money laundering and terrorist financing. Extracts from Special recommendations on terrorist financing, FATF, revised22 October 2004. VI. Alternative remittance: Each country should take measures to ensure that persons or legal entities, including agents, that provide a service for the transmission o f money or value, including transmission through an informal money or value transfer system or network, should be licensed or registeredand subject to all the FATF Recommendations that apply to banks and non-bank financial institutions. Each country should ensure that persons or legal entities that carry out this service illegally are subject to administrative, civil or criminal sanctions. VII. Wire transfers: Countries should take measures to require financial institutions, including money remitters, to include accurate and meaningful originator information (name, address and account number) on funds transfers and related messages that are sent, and the information should remain with the transfer or related message through the payment chain. This section draws from MoneyGram (2006a) 262 Countries should take measures to ensure that financial institutions, including money remitters, conduct enhanced scrutiny o f and monitor for suspicious activity funds transfers which do not contain complete originator information (name, address and account number). ThePatriot Act "' The Patriot Act required that all RSPs adopt a written anti-money laundering program that is designed to ensure monitoring and reporting o f suspicious transactions. This program must include: 0 Internal policies, procedures and controls to verify customers identification, file reports, keepingrecords and respondto law enforcement requests 0 Designation o f a compliance officer responsible for ensuring that procedures are followed 0 Provisiono f on-going training to employees on compliance to procedures 0 Independentreview ofthe anti-money laundering program Record keepingrequirements: 0 Transactions below US$1,000 require presentation o f identification o f sender that includes name and address (identification cards issued by a foreign government can be accepted). Identificationo f the receiver i s not required. Transactions (both send and receive) above US$3,000 require collecting information on identification, address and occupation o f sender andreceiver. 0 Transactions above US$lO,OOO require filing o f a standard Currency Transaction Report (CTR). CTRs mustbe filed with the InternalRevenue Service. 0 Records o f transactions above US$3,000 must be retained for five years. Similarly, RSPs must report `suspicious activity' such as structuring (splitting remittances in small transactions to avoid reporting) through a Suspicious Activity Report. However there are no clear guidelinesas per the definition of `suspicious', and this i s left to the individual agent or operator to identify and report. Reporting to the OfJice of Foreign Assets Control (OFAC) The OFAC is the unit of the U.S. Department of the Treasury in charge of enforcing economic and trade sanctions against targeted foreign countries, terrorists and drug cartels. OFAC issues a list of governments, terrorists, drug traffickers and those engaged in proliferation of weapons of mass destruction. This list is known as Specially Designated Individuals (DSNList). Any transaction with individuals or entities included in the DSN List is prohibited by OFAC regulation, and businesses are required to identify and freeze assets o f organizations or individuals includedinthe list and report to the Department o f Treasury. 263 Criminalpenalties Non-compliance with the anti-money laundering regulations derived from the Bank Secrecy Act, USA Patriot Act and OFAC can be punishedby highfines and prisonterms. Businesses can be heldcriminally liable for the acts o f the employees. Enforcing Regulations Even if regulations for ensuring adequate security inpayments and remittances systems are rightly established, their application at the local level can be challenging, requiringa balance between strict enforcement, proportionality to the risk o f misuse and avoidance o f disruption causedby unduly burdensomerequirements(see Box 12). According to the General Principles, AML and CFT regulations should be equally applicable to all RSPs irrespectively o f their legal form (i.e. financial institutions or commercial companies). The regulatory framework for remittances should thus be defined on a functional, rather than institutional basis so that it can be extended to any RSPs that are not currently under regulatory oversight. As opposed to this, the uneven application o f regulatory requirements across RSPs creates both loopholes that can be used for illicit purposes, as well as competitive disadvantages for regulated entities.17) However, this principle is difficult to apply in many countries, where authorities and applicable regulations depend on whether the RSP i s a financial institution or just a commercial company. Moreover, regulation is not harmonized throughout Latin America and in many countries appears to be minimal and enforcement limited(Orozco, 2005). Box 12. The FATF's Best Practices on regulatoryrequirements According to the FATF, the implementation of adequate regulatory requirementsshould take into consideration the following tasks: Supervisors need to have a better understanding o f the nature o f the remittance business to design a regulatory framework; 0 Supervisors should show sufficient flexibility in dealing with remittance providers, but ensure that a regulatory system works effectively; 0 Supervisors should have adequate resources and capacity to regulate remittance '73Non regulated RSPs can offer services at an artificially low price as they are not subject to the considerable regulatory costs incurred by formal competitors. 264 systems; 0 Supervisors need to analyze the flow o f remittance funds to detect irregularities and possible misuse o f the system; 0 Supervisors need to issue clear policies on how to deal with remittance providers that choose not to participate inthe regulatory system. Source: IMF(2005) The cost of compliance The cost o f compliance with AML / CFT regulations can be very high. It includes the setting up o f monitoring and reporting systems, the establishment and enforcement o f internal procedures, the constant training o f personnel and frequent external audits to the various systems. Moreover, RSPs are subject to reporting requirements with multiple authorities and, in the case of the U.S.also to additional regulation at the state level. Not surprisingly, in a recent survey of RSPs the costs of regulatory compliance have been identified as the third most important expense, after rent and salaries, mentioned by 45 percent o f RSPs (figure 9.12). Figure 9.12. Largest expenses reportedby RSPs 0% !nu m% 3ou 40% SOY m 73% s o y Flrminmdionhse#- Source: Andreassen (2006) based on a survey of 73 remittance firms in 6 U.S.states. The FATF advises that countries should ensure that oversight is commensurate with the risk of misuse to avoid unnecessary costs and inefficiency. At the same time, countries should seek to 'formalize' RSPs through licensing or registration, as a mechanism to help enforce AML/CFT requirements. In turn, the decision to either license or register RSPs should take into account that the former creates relatively high costs o f pre-qualification (due diligence) which could deter new entrants, while registration poses low barriers to entry, but requires sufficient resources for ex-post monitoringby authorities. 265 Similarly, the General Principles suggest that any regulation o f remittances should balance the benefits o f increased safety of the system with the potential costs and inefficiencies created. Regulatory costs can translate into higher remittance prices for formal RSPs, they can increase barriers to entry for new formal competitors, and create incentives for the proliferation o f informal RSPS."~All in all, regulators need to take into account the trade off between increased security and the creation o f unnecessary constraints to the flow o f bona fide remittances. VII. PolicyRecommendations Competition Regulatory requirements for new RSPs should balance the need to maintain security in the system with eliminating unnecessary hurdles to bona fide entrants. Entry requirementsmust be clear, transparent and uniformly applicable to all entrants o f similar characteristics. These requirements should also be proportional to the risk and volume of operations. While homogeneity across countries i s a desirable but hard to achieve goal, requirements should at leastbe homogenized within the same country. Unduly regulatory barriers to the use of payment and settlement systems (either directly or indirectly) should be removed. While direct access may not be strictly necessary, indirect access through financial institutions should not translate into competitive disadvantages for RSPs. Exclusivity agreements between capturing and disbursing RSPs can limit competition in the market, especially when they involve wide networks. While these may be difficult to prevent when both RSPs are private entities, local authorities at destination countries should maintain public networks - e.g. the postal systems-open to different service providers. Authorities in low volume corridors should ensure contestability in their local markets as a way to create incentives for efficiency of incumbent RSPs. Transparency 0 Authorities should take an active role in the collection and dissemination o f information on comparative prices o f remittance services by different providers. This information should be disseminated through adequate channels ina form that is clear and easily understandable. Bi- lateral efforts should be made so that this information i s disseminated both at countries o f origin and destination o fremittances. 0 Many authorities have avoided direct regulation and enforcement o f transparency by RSPs. However, the experience in some countries shows that RSPs have found in their best interest to cooperate with official efforts for transparency on a voluntary basis. Similarly, the See Orozco (2005). 266 establishment o f best practices at the industry level can be encouraged as a form o f self- regulation. The price information disclosed by RSPs should be clear, easily understandable, and providedinthe languages o f the corresponding corridor. This information should also be presentedunder an industry-wideformat to facilitate users' comparisons across RSPs. Accessibility Accessibility to financial services at the country o f origin o f remittances can be limitedby unnecessarily highdocumentation requirementsthat affect particularly illegal immigrants. o Inthe cases where this is due to actual regulation, authorities should review the merit and need for such constraints. o In other cases this is not a direct regulatory requirement, but appears to be derived from misinterpretation o f regulation or unfounded banks' expectations o f stronger oversight by regulators if they provide services to undocumented migrants. Authorities should ensure clarity inregulatory requirements and transparency intheir oversight processes in order to avoid misinterpretations that can translate into artificial constraints to accessibility. Financial services may also be under-utilized by immigrants based on perceptions o f high costs or distrust o f financial institutions. Financial services such as bank accounts can indeed pose significant transaction costs and administration fees, especially to users that cannot maintain a minimumbalance. While it i s unlikely that this i s an issue that can be addressed by authorities, there is room for the provision o f basic financial literacy that can lower mistrust on banks while helping migrants to understand the costs and benefits of using financial institutions for remittances and other services. Joint efforts by authorities and financial institutions should be considered. Accessibility to financial services at the country o f destination o f remittances can be improved by permitting access to remittance services to smaller financial institutions such as credit unions, savings and loans and microfinance companies. Because of their focus, these institutions may be closer to the recipients o f remittances than mainstream commercial banks. Securiv 0 The definition and implementation o f regulatory measures to ensure security in the system must be based on authorities' efforts to collect information and hlly understand the characteristics of the remittances markets. 0 Basic security regulations to avoid criminal misuse o f remittance channels should be applicable to all RSPs. Uneven application o f these regulations can hamper the security o f the systemby creating loopholes than can be exploited for illegal purposes. Similarly uneven regulation can create market disruptions through unfair competitive advantages in favor o f less-regulated entities. However, the even application o f security regulation may pose significant challenges to authorities in countries where the regulatory framework is different 267 across RSPs that operate as financial institutions and those that are established as commercial companies. 0 Given the fact that compliance with regulatory requirements is costly and can create operational inefficiencies, regulation to ensure security o f the system should be proportionate to the risk o f misuse and the volume o f operations o f the RSPs. 268 REFERENCES Acosta, P. (2006): "Labor Supply, School Attendance, and Remittances from International Migration: The Case o f El Salvador" World Bank Policy Research Working Paper 3903, The World Bank, Washington, DC. Acosta, P., C. Calderon, P. Fajnzylber, and H. Lopez (2006 a): "Remittances and Development inLatin America", forthcoming, WorldEconomy. Acosta, P., C. Calderon, P. Fajnzylber, and H. Lopez (2006 b): "What is the Impact of International Remittances on Poverty and Inequality in Latin America?", mimeo, The World Bank, Washington, DC. Adams, R. (2004): "Remittances and Poverty in Guatemala." World Bank Policy Research Working Paper 3418, The World Bank, Washington, DC. Adams, R. (2005): "Remittances, Household Expenditure and Investmentin Guatemala" World Bank Policy ResearchWorking Paper No. 3532, The World Bank, Washington, DC. Adams, R. (2006): "Remittances and Poverty in Ghana." World Bank Policy ResearchWorking Paper 3838, The World Bank, Washington, DC. Adams, R., and J. Page (2005): "Do International Migration and Remittances Reduce Poverty in DevelopingCountries?", WorldDevelopment 33: 1645-1669. Aggarwal, R., A. Demirguc-Kunt, and M. S. Martinez Peria (2005): "Do Workers' Remittances Promote Financial Development?" mimeo, The World Bank, Washington, DC. Ahmed, S. (1986): "Temporary and Permanent Government Spending in an Open Economy: Some Evidence for the UnitedKingdom", Journal of Monetary Economics 17, 197-224. Alberola, E., S. G. Cervero, H.Lopez, and A. Ubide (1999): "Global equilibriumexchange rates: euro, dollar, `ins,' `outs' and other major currencies in a panel cointegration framework", IMF Working PaperNo. 175, International Monetary Fund, Washington, DC. Alberola E., S. G. Cervero, H. L6pez and A. Ubide (2002): "QUO vadis euro?", European Journal of Finance 8 (4), 352-70. Alberola, E. and H. Lopez (2001): "Internal and external exchange rate equilibrium in a cointegration framework. An application to the Spanish peseta", Spanish Economic Review, 3(l), 23-40. Amuedo-Dorantes, C. and S. Pozo (2004): "Workers' Remittances and the Real Exchange Rate: A Paradox of Gifts", WorldDevelopment 32: 1407-1417. 269 Andreassen, 0. (2006): "Remittance Service Providers in the United States: How remittance firms operate and how they perceive their business environment", The World Bank, Washington, DC. Anibale, B. (2005): "Balancing Competition and Regulation in the Remittance Market", Citibank, remarks at the IDB International Forumon Remittances 2005, Washington, DC. Arellano, M. and 0. Bover (1995): "Another Look at the Instrumental Variable Estimation o f Error Component Models". Journal of Econometrics 68: 29-51. Aschauer, D. (1985): "Fiscal Policy and Aggregate Demand" American Economic Review, 75, 117-27. Attanasio, 0. and M. Szekely (2000): "Saving in Developing Countries: Inequality, Demographics and all that", mimeo, UniversityCollege London, London. Attanasio, 0. and V. Rios Rull (2000): "Consumption smoothing in island economies: Can public insurance reduce welfare?", European Economic Review, 44, 1225-58. Bailey, M.J. (1972): "The Optimal FullEmployment Surplus", Journal of Political Economy, 80, 649-61. Bair, S. C. (2005): "Improving access to the U.S.Banking System among recent Latin American immigrants", Center for Public Policy and Administration, University o f Massachusettsand The MultilateralInvestmentFund, Washington, DC. Balassa, B. (1964): "The Purchasing Power Parity Doctrine: A Reappraisal," Journal of Political Economy, 72,584-96. Banco de la Republica de Colombia (2004): "Principales Resultados de la Encuesta de Costos de Transaccih de Remesasde Trabajadores en Colombia". Banerjee, A. and E. Duflo (2003): "Inequality and Growth: What Can the Data Say?", Journal of Economic Growth 8,267-299. Banerjee, A.V. and A.F. Newman. (1993): "Occupational Choice and the Process o f Development." Journal of Political Economy 101,274-298. Barham, B., and S. Boucher (1998): "Migration, Remittances, and Inequality: Estimating the Net Effects o f Migration on Income Distribution", Journal of Development Economics 55: 307-331, Barro, R.J. and J.W. Lee, (2001): "International Data on Educational Attainment: Updates and Implications." Oxford Economic Papers 53(3), 541-63. Beck T., A. Demirguc-Kunt, and R. Levine, (2000): "A New Database on Financial Development and Structure", WorldBank Economic Review 14 (3), 597-605. 270 Beck, T., A. Demirgiiq-Kunt and M.S. Martinez Peria (2005): "Reaching out: Access to and Use o f Banking Services Across Countries". World Bank Policy Research Working Paper No. 3754, Washington, DC. Beck, T., A. Demirgiiq-Kunt and M. S. Martinez Peria (2006): "Banking Services for Everyone? Barriers to Bank Access Around the World", mimeo, The World Bank, Washington, DC. Bekaert, G., H. Campbell and C. Lundblad, (2005): "Growth Volatility and Financial Liberalization." Journal of International Money and Finance, forthcoming. Bendixen & Associates (2001): "Survey of Remittance Senders: U.S to Latin America", Bendixen & Associates, Coral Gables, Florida. Blejer, M.I., M.S. Khan (1984): "Government Policy and Private InvestmentinDeveloping and Countries", IW StafSPapers, 31(2), 379-403. Borensztein, E., J. De Gregorio, and J.-W. Lee, (1998): "How Does Foreign Direct Investment Affect Economic Growth?", Journal of International Economics 45(l), 115-35. Burnside,C.A., and D.Dollar, (2000): "Aid, Policies, and Growth." American Economic Review 90(4), 847-868. Burnside, C. and D.Dollar (2004): "Aid, Policies, and Growth: Revisiting the Evidence", Policy ResearchWorking Paper Series 3251, The World Bank, Washington, DC. Butelmann, A. and F. Gallego (2001): "Estimaciones de 10s Determinantes del Ahorro de 10s Hogares en Chile (1988-1997)", in Analisis Empirico del Ahorro en Chile, ed. by F. MorandC and R. Vergara, Banco Central de Chile and Centro de Estudios Publicos, forthcoming. Calderon, C., and R. Fuentes, (2005): "LCuanto explican las reformas y la calidad de las instituciones el crecimiento chileno?" Central Bank o f Chile Working Paper 314, Santiago de Chile. Calderon C., P. Fajnzylber and H. Lopez (2006): "Remittances, growth, and policy complementarities'y, mimeo, The World Bank, Washington, DC. Calderon C., P. Fajnzylber and H. Lopez (2006 a): "Remittances and growth volatility: an investigation into the business cycle properties o f workers' remittances", mimeo, The World Bank, Washington, DC. Caprio, G. and D. Klingebiel (2003): "Episodes o f Systemic and Borderline Financial Crises 1970s-2002", The World Bank, Washington, DC. 271 Cavallo, E. and J. Frank,el(2004): "Does Opennessto Trade Make Countries Less Vulnerable to Sudden Stops? Using gravity to Establish Causality", NBER Working Paper 10957, National Bureau o f Economic Research, Inc. Chami, R., C. Fullenkamp and S. Jahjah (2003): "Are immigrant remittance flows a source of capital for development", IMF Working. N o 03/89, International Monetary Fund, Washington, DC. Cheikhrouhou, H., R. Jarque, R. El-Swaify, and A. Villanueva-Villareal (forthcoming): "The U.S-Guatemala Remittances Corridor Analysis", The World Bank, Washington, DC. Clarke, G. and S. J. Wallsten. (2003): "Do Remittances Act Like Insurance? Evidence from a Natural Disaster inJamaica", The World Bank ,Washington, D.C. Cox, D., Eser, Z., and E. Jimenez (1998): "Motives for Private Transfers Over the Life Cycle: An Analytical Framework and Evidence for Peru", Journal of Development Economics, 55(l), 57-80. Cox-Edwards, A,, and M. Ureta (1998): "International Migration, Remittances, and Schooling: Evidence from El Salvador", Journal of Development Economics 72: 429-461. Coutinho, L (2006): "Networking: The Case o f Bansefi", mimeo, The World Bank, Washington, DC . Cutler, D. M. and J. Gruber (1996): "Does Public Insurance Crowd our Private Insurance," Quarterly Journal of Economics, 112(2), 391-430. De Ferranti, D. G. Perry, I.Gill, J.L. Guasch, N.Schady, W. Maloney and C. Sanchez Paramo (1992): "Closing the Gap inEducation and Technology", The World Bank, Washington, DC. De la Torre, A,, J. Gozzi, and S. Schmukler (2006): " Innovative Experiences in Access to Finance: Market FriendlyRoles for the Visible Hand?", The World Bank, Washington, DC. Derham, M.(2005): "Every Dollar Counts", LatinFinance 173 (December). De Luna Martinez, J. (2005): "Workers' Remittances to Developing Countries: A survey with Central Bank on selectedpolicy issues", The World Bank, Washington, DC. Demirgriiq-Kunt, A. and M. S. Martinez Peria (2006): "Remittances and the Use o f Banking Services. Evidence from El Salvador" mimeo, The World Bank, Washington, DC. Demirgriiq-Kunt, A., E. Lopez-Cordova, M. S. Martinez Peria, and C.Woodruff (2006): "Remittances and Financial Development. Evidence from Mexican Municipalities" mimeo, The World Bank, Washington, DC. 272 Docquier, F.and A. Marfouk (2005): "International Migration by Educational Attainment (1990 -2000)," Chapter5 inC.OzdenandM.Schiff(eds.) InternationalMigration,Remittances and the Brain Drain. Palgrave McMillanPress (for the World Bank). New York Dollar, D. (1992): "Outward-Oriented Developing Economies Really Do Grow More Rapidly: Evidence from 95 LDC's, 1976-1985," Economic Development and Cultural Change40(3), 523- 44. Easterly, W., C.A. Rodriguez and K. Schmidt-Hebbel (1994): "Public Sector Deficits and Macroeconomic Performance", Oxford University Press, New York. El Qorchi, M:, S. Maimbo and J. Wilson (2003): "Informal Funds Transfer Systems: An Analysis of the Informal Hawala System", IMF OccasionalPaper No. 222, Washington, DC. Evans, P. and G. Karras (1996): "Convergence Revisited", Journal of Monetary Economics, 37, 249-65. Faini, R. (2002) : "Development, Trade, and Migration", Revue d'Economie et du De`veloppement, Proceedings from the ABCDE Europe Conference, 1-2: 85-116. Faini, R. (2002): "Migration, Remittances and Growth." manuscript, University o f Brescia, Brescia, Italy. Fatas, A. and I.Mihov (2003): "The Case for Restricting Fiscal Policy Discretion", Quarterly Journal of Economics 118(4), 1419-47. Forbes, K. (2000): "A Reassessment o f the Relationship between Inequality and Growth", American Economic Review 90(4), 869-87. Frankel, J. and A. Rose (1998): "The Endogeneity o f the Optimum Currency Criteria," Economic Journal 108, 1009-25. Frenkel, J. and M. Mussa (1985): "Asset Markets, Exchange Rates, and the Balance of Payments", in R.W. Jones, P.B. Kenen (eds.) Handbook o f International Economics Vol. 2, Ch. 14. Frias, M. (2005): Supervisory Insights, "Linking International Remittance Flows to Financial Services: Tapping the Latino Immigrant Market", FederalDeposit Insurance Corporation. Frumkin, S. (2005): "Remittances: A gateway to banking for unbanked immigrants", Comptroller o fthe Currency Administrator o fNational Banks. Funkhouser, E. (1992): "Migration from Nicaragua: Some Recent Evidence", World Development 20: 1209-18. 273 FUSADES (1995): "Uso productivo de las remesas familiares", Fundaci6n Salvadoreiia para el Desarrollo Econ6mico y Social, La Libertad, El Salvador. Gallego, F., and N. Loayza. (2002): "The Golden Period for Growth in Chile: Explanations and Forecasts." In Economic Growth: Sources, Trends, and Cycles, edited by Norman Loayza and Raimundo Soto, Central Bank o f Chile, Santiago de Chile. Galor, 0.and J. Zeira (1993): "Income Distribution and Macroeconomics." Review of Economic Studies 60, 35-52. Giuliano, P. and M. Ruiz-Arranz (2005): "Remittances, Financial Development and Growth". IMFWorking Paper 05/234, International MonetaryFund, Washington, DC. Glewwe, P. and P. Olinto (2004): "Evaluating o f the Impact o f Conditional Cash Transfers o f Schooling: An Experimental Analysis of Honduras' PRAF Program" USAID. Goldstein, M. G., L.Kaminsky, and C. M.Reinhart. (1999): "Assessing FinancialVulnerability: An Early Warning System for Emerging Markets". Institute for International Economics, Washington, DC. Hanson, G. (2005): "Emigration, Remittances, and Labor Force Participation in Mexico", mimeo, University o f California, San Diego, California. Hanson, G. H. and C. Woodruff (2003): "Emigration and Educational Attainment in Mexico" mimeo, Universityo f California, San Diego, California. Heckman, J. J. (1979): "Sample selection bias as a specification error", Econometrica ,47, 153- 161. Hernandez-Coss, R. (2004): "Lessons from the US-Mexico Remittances Corridor on Shifting from Informalto FormalTransfer Systems". The World Bank, Washington, D.C. Honohan, P. and D. Klingebiel (2003): "The Fiscal Cost Implications o f an Accommodating Approach to Banking Crises" Journal of Banking and Finance 27(8), 1539-60. Hysenbegasi, A., and S. Pozo. (2002): "What Prompts Workers to Remit? Evidence using a Panel o f Latin American and Caribbean Nations." Citigroup, Irvin Texas and Department o f Economics, Western MichiganUniversity. Ibarraran, P. and D. Lubotsky (2005): "Mexican Immigration and Self-selection: New Evidence from the 2000 Mexican Census", NBER Working Papers No.11456, National Bureau o f Economic Research, Inc. Inter-American Development Bank (2002): "Survey of Remittance Senders: U.S. to Latin America", Inter-AmericanDevelopment Bank, Washington, DC. 274 (2003): "Remittance Senders and Receivers: Tracking the Transnational Channels", Inter-AmericanDevelopment Bank, Washington, DC. (2004): "Sending money home: Remittance recipients in the Dominican Republic and Remittance Senders from the U.S.", Inter-American Development Bank and Columbia University,New York. (2004): "State by state surveyo fremittance senders: U.S.to Latin America", Inter- American Development Bank, Washington, DC. (2004a): "Remittance flows to Latin America and the Caribbean, 2004", Inter- American Development Bank, Washington, DC. (2006b): "Remittances 2005: Promoting Financial Democracy", Inter-American Development Bank, Washington, DC. (2006a): "Enviando Dinero a Casa: Marcador de la Industria de Remesas", Inter- American Development Bank, Washington, DC. International Monetary Fund (2005): "Balance o f Payment Statistics Yearbook", International Monetary Fund, Washington, D.C. (2005): "World Economic Outlook" International Monetary Fund, Washington, DC. (various issues): "Annual Report on Exchange Arrangements and Exchange Restriction", International Monetary Fund, Washington, DC. (various issues): "Balance o f Payments Statistics Yearbook", International Monetary Fund,Washington, D.C. (various issues): "International Financial Statistics", International Monetary Fund, Washington, DC. (2005): "Approaches to a regulatory framework for formal and informal remittance systems: experiences and lessons", IMF, Washington, DC. Kaminsky, G. L. and C. M. Reinhart (1999): "The Twin Crises: The Causes o f Banking and Balance-of-Payments Problems", American Economic Review 89(3), 473-500. Kapur, D. (2004): "Remittances: The New Development Mantra?", G-24 Discussion Paper No.29, UNConference on Trade and Development, Geneva, Switzerland. Kaufman, D., A.Kraay and M-Mastruzzi (2005): "Governance Matters IV: Governance Indicators for 1996-2004", Policy Research Working Paper 3630, The World Bank, Washington, DC. 275 Khan, M. S. and C. M. Reinhart (1990): "Private Investment and Economic Growth in DevelopingCountries", WorldDevelopment, 18,19-27. King, R. and R. Levine (1993): "Finance and Growth: Schumpeter Mightbe Right", Quarterly Journal ofEconomics, 108,717-37. Knowles, S. (2005): "Inequality and Economic Growth: The Empirical Relationship Reconsidered in the Light o f Comparable Data", The Journal of Development Studies, 41, 135- 159. Kormendi, R. (1983): "Government Debt, Government Spending, and Private Sector Behavior", American Economic Review, 73,994-1010. Levin, A. and L. K. Raut (1997): "Complementarities between Exports and Human Capital in Economic Growth: Evidence from the Semi-industrialised Countries", Economic Development and Cultural Change, 46 (l), 155-74. Lindert, K., E. Skoufias and J. Shapiro (2005): "Redistributing Income to the Poor and the Rich: Public Transfers in Latin America and the Caribbean", Regional Study Working Paper, The World Bank, Washington, DC. Lindert, K., E. Skoufias and J. Shapiro (2006): "How effectively do public transfers in Latin America redistribute income?" Regional Study Working Paper, The World Bank, Washington, DC Loayza, N.,P. Fajnzylber, and C. Calderon (2005): "Economic Growth inLatin America and the Caribbean: Stylized Facts, Explanations and Forecasts", The World Bank, Washington, DC. L6pez Cbrdova, E. (2005): "Globalization, Migration and Development: The Role o f Mexican Migrant Remittances" Economia, Journal of the Latin American and Caribbean Economic Association, forthcoming. L6pez, H., and L.Molina (2006): "Remittances and the real exchange rate: Are there reasons to be concerned about Dutch Disease in Latin America?", mimeo, The World Bank, Washington, DC. Lopez, H., and L. ServCn (2006): "A Noma1 Relationship? Poverty, Growth and Inequality". World Bank Policy ResearchWorking Paper 3814, The World Bank, Washington, DC. L6pez, H.,K. Schmidt-Hebbel, and L.Serven (2000): "How effective i s Fiscal Policy inraising National Saving?" TheReview of Economics and Statistics LXXXII (2), 226-38. Lowell, B. L.and R. Suro (2002): "How Many Undocumented: The numbersbehindthe US.- Mexico", MigrationTalks, Pew Hispanic Center, Washington, DC. Maimbo S., and D. Ratha (2005): "Remittances: Development Impact and Future Prospects", The World Bank, Washington, DC. 276 Maluccio, J. A. (2004): "Education and Child Labor: Experimental Evidence from a Nicaraguan Conditional Cash Transfer Program", forthcoming, inChild Labor inLatin America, P. Orazem, G. Sedlaceck, and Z. Tzannatos (editors), International Food Policy Research Institute, Washington, D.C. Maluccio, J. A,, and R. Flores (2004): "Impact evaluation o f a conditional cashtransfer program: The Nicaraguan Red de Proteccibn Social", Food Consumption and Nutrition Division Discussion Paper (forthcoming), International FoodPolicy ResearchInstitute, Washington, DC. Massey, D., and E. Parrado (1998): "International Migration and Business Formation in Mexico", Social Science Quarterly 79: 1-20. Mattoo ,A,, I.C. Neagu and C. Ozden (2005): "Brain Waste? Educated Inmigrants in the U.S. Labor Market", World Bank Policy Research Working Paper No. 2581, The World Bank, Washington, DC. McKenzie, D. and H. Rapoport (2005): "Migration and Education Inequality in Mexico", mimeo, Inter American Development Bank, Washington, DC. Mesbah-Khavari, D., A. Evans, and JKlaehn (2005): "Credit Union Remittance Services in Guatemala: Expanding the Access o f Low-Income Remittance Recipients to Financial Institutions", United StatesAgency for International Development (USAID). Mishra, P. (2003): "Emigration and Wages in Source Countries: Evidence from Mexico" manuscript, Department of Economics, ColumbiaUniversity, New York. Mishra, P. (2005): "Macroeconomic Impact o f Remittances inthe Caribbean", forthcoming, IMF WorkingPaper, International Monetary Fund,Washington, DC. MoneyGram (2006): "Anti-Money Laundering Compliance Guide", MoneyGram, Minnesota, U.S. MoneyGram (2006a): "Agent Compliance Manual", MoneyGram, Minnesota, U.S. Morris, P., G. Duncan and C. Rodrigues (2004): "Does money really matter? Estimating impacts o f family income on children's achievement with data from random assignment experiments", Paper prepared for the "Next Generation" project, MDRC, New York. Mundaca, G. (2005): "Can Remittances Enhance Economic Growth? The Role o f Financial Markets Development", mimeo, Department o f Economics, Universityo f Oslo, Oslo. Mussa, M. (1984): "The Theory of Exchange rate Determination", in J. Bilson, and R. Marston (eds), Exchange Rate Theory and Practice, NBER Conference Report, University o f Chicago Press, Chicago. 277 North, D. (1990): "Institutions, institutional change, and economic performance", Cambridge U. Press, Cambridge. Olinto, P. (2006): "Do Conditional Cash Transfer Programs Crowd Out Private Transfers? Evidence from Randomized Trials inRural Honduras and Nicaragua", mimeo, The World Bank, Washington, DC. Organization for Economic Co-operation and Development (2005): "Database on Immigrants and Expatriates" OECD, Paris, France. http:llwww.oecd.orddocunient/5 1/0,2340,en 2825 494553 34063091 1 1 1 1.OO.html Orozco, M. (2003a): "Changes in the Atmosphere? Increase in Remittances, Price Decline and New Challenges", Inter-American Dialogue, Washington, DC. Orozco, M. (2004): "The Remittance Marketplace: prices, policy and Financial Institutions", Pew Hispanic Center Report, Georgetown University and Institute for the Study o f International Migration, Washington, DC. Orozco, M. (2005): "Markets and Financial Democracy: The Case for Remittance Transfers", Journal of Payment Systems Law, 2, 166-215. Orozco, M. and Fedewa, R. (2005): "Leveraging Efforts on Remittances and Financial Intermediation" Report Comissioned by the Inter-American Development Bank, Washington, DC. Osili, U.O. and A. Paulson (2004a): "Prospects for Inmigrant-Native Wealth Assimilation: Evidence from Financial Market Participation" Working Paper No. 18, Federal Reserve Bank o f Chicago. Ozden, C. (2006): "Migration Patterns in Latin America and the Caribbean: An Overview o f Destination, Age, Education and Occupation Profiles", mimeo, The World Bank, Washington, DC. Ozden, C., and M. Schiff (2006): "International Migration, Remittances, and the Brain Drain", The World Bank, Washington, DC. Page, J., and S. Plaza (2005): "Migration, Remittances, and Development: A Review o f Global Evidence", mimeo, The World Bank, Washington, DC. Pallage, S., and M.A. Robe. (2003): "On the Welfare Cost o f Economic Fluctuations in DevelopingCountries", forthcoming, International Economic Review. Passel, J. (2006): "The Size and Characteristics o f the Unauthorized Migrant Population in the U.S.", Pew Hispanic Center, Washington, DC. 278 Perotti, R. (1996): "Growth, Income Distribution and Democracy", Journal of Economic Growth, 1, 149-87. Perry, G., 0. Arias, H. Lopez, W. Maloney, and L. Servtn (2006): "Poverty Reduction and Growth: Virtuous andVicious Circles", The World Bank, Washington, DC. PROFECO (1998): "El Mercado de envio de dinero de Estados Unidos a Mexico", Procuraduria Federaldel Consumidor, Mexico City. PROFECO (2004): "Programa quien es quien en el envio de dinero de Estados Unidos a Mexico, origen y evolution", Procuraduria Federal del Consumidor, Mexico City. Rajan, R., and A. Subramanian (2005): "What Undermines Aid's Impact on Growth?", mimeo, International Monetary Fund, Washington, DC. Ratha, D. (2003): "Workers' Remittances: An Important and Stable Source of External Development Finance," in Global Development Finance 2003, The World Bank, Washington, DC. Ratha, D. and J. Riedberg (2005): "On reducing remittance costs", The World Bank, Washington, DC. Rodriguez, E. R. (1998): "International Migration and Income Distribution in the Philippines", Economic Development and Cultural Change, 46(2), 329-50. Rodriguez, E. and E. Tiongson (2001): "Temporary Migration Overseas and Household Labor Supply:Evidence from UrbanPhilippines", International Migration Review, 35, 1185-204. Ruiz-Arranz, M., and P. Giuliano (2005): "Remittances, Financial Development, and Growth". IMFWorking PaperNo. 05/234, Washington, DC. Samuelson,P.A. (1964): "Theoretical Notes on Trade Problems," Review of Economics and Statistics, 46, 145-54. Sander, C. and S. Mainbo (2003): "Migrant Labor remittances in Africa: Reducing obstacles to developmental contributions". Working paper No. 64, Africa Region Working Paper Series, The World Bank, Washington, DC. Sayan, S. (2004): "Guest Workers' Remittances and Output Fluctuations in Host and Home Countries", Emerging Markets Finance and Trade, 40(6), 68-81, Sayan, S. (2006): "Business Cycles and Workers' Remittances: How do Migrant Workers Respond to Cyclical Movements o f GDP at Home?", IMF Working Paper No. 06/52, Washington, DC. 279 Schoeni, R. (2002): "Does Unemployment Insurance Displace Familial Assistance." Public Choice. 110,99- 119. Schoeni, R. F. (1996): "Does Aid to Families with Dependent Children Displace Familial Assistance?", RandWorking Paper, DRU-1453-RC. Skoufias, E. (2005): "PROGRESA and Its Impact on the Welfare o f Rural Households in Mexico", ResearchReport 139, International Food ResearchInstitute, Washington, DC. Solimano, A. (2003): "Remittances by Emigrants: Issues and Evidence", WIDER Discussion PaperNo. 2003/89, World Institute for Development Economics Research, UnitedNations University,Helsinki. Spatfora, N. and C. Freud (2005): "Remittances: Transaction Costs, Determinants and Informal Flows", World Bank Policy ResearchWorking Paper 3704, The World Bank, Washington, DC. Stark, O., J. Taylor, J., and S. Yitzhaki (1986), "Remittances and Inequality", Economic Journal 96,722-740. Suro, R. (2005): "Survey o f Mexican Migrants: Attitudes about Immigration and Major Demographic Characteristics", Research& Publications, Pew Hispanic Center, Washington, DC. Suro, R. and S. Bendixen (2002): "The Remittance Process and the Unbanked", Pew Hispanic Center and Multilateral InvestmentFund, Washington, DC. Suro, R., S. Bendixen and D. Benavides (2002): "Billions in motion: Latino immigrants, remittances and banking". Pew Hispanic Center Report, Georgetown University and Institute for the Study ofInternational Migration, Washington, DC. Taylor, J. (1992): "Remittances and Inequality Reconsidered: Direct, Indirect, and Intertemporal Effects", Journal of Policy Modeling 14, 187-208. Taylor, J. E., J. Mora, R. Adams, and A. Lopez-Feldman (2005): "Remittances, Inequality, and Poverty: Evidence from RuralMexico", mimeo, University o f California. Teruel, G., and B. Davis (2000): "Final Report: An Evaluation o f the Impact o f PROGRESA Cash Payments on Private Inter-Household Transfers", International Food Policy Research Institute, Washington, DC. U.S.Bureauofthe Census(2000): "International DataBase", Washington, DC. United Nations (2005): "Human Development Report" United Nations Development Programme, New York, NY. WOCCU (2004): "A technical guide to remittances", World Council o f Credit Unions, Inc., Washington, DC. 280 WOCCU (2006): "How to serve undocumented individuals", World Council of Credit Unions, Inc., Washington, DC. Woodruf, C. and R. Zenteno (2004): "Remittances and Microenterprises in Mexico", mimeo, University of California, SanDiego, California. World Bank (2002): "World DevelopmentIndicators database", The World Bank, Washington, DC . (2003): "World DevelopmentReport 2003: SustainableDevelopment inaDynamic World", The World Bank, Washington, DC. (2005): "World DevelopmentIndicators database", The World Bank, Washington, DC. (2005): "Doing Business", The World Bank, Washington, DC. (2005): "Economic Growth inthe 1990s, Learning Froma Decadeof Reform",The World Bank, Washington, DC. (2005): "The U.S.-Mexico Remittances Corridor Analysis", The World Bank, Washington, DC. (2006): "Global Economic Prospects:Economic Implications of Remittancesand Migration", The World Bank, Washington D.C. (2006a): "The U.S.-Guatemala Remittances Corridor Analysis", forthcoming, The World Bank, Washington, DC. and Committee on Payment and Settlement Systems (2006): "General principles for international remittance services, consultative report", Bank for International Settlements and The World Bank, Washington, DC. Yang, D. (2006): "Coping with Disaster: The Impact of Hurricanes on International Financial Flows, 1970-200l", mimeo, University of Michigan, Ann-Arbor. Yang, D.and H.Choi. (2005): "Are RemittancesInsurance?Evidencefrom RainfallShocks in the Philippines", Ford Schoolof Public Policy Working Paper Series No. 2005/005, University ofMichigan, Ann Arbor. wb164661 L:\rernit\deliveryUII\Vol2 062906.doc 08/25/2006 3:52:00 PM 281