Page 1 SOUTH ASIA REGION A WORLD BANK DOCUMENT PREM WORKING PAPER SERIES The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the Poverty Trends in Bangladesh During the Nineties Rinku Murgai and Salman Zaidi May 2004 Report No. SASPR-2B World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. 30863 Page 2 P OVERTY T RENDS IN B ANGLADESH DURING THE N INETIES Rinku Murgai and Salman Zaidi 1 1 Poverty Reduction and Economic Management Unit, South Asia Region, The World Bank. Contact information: rmurgai@worldbank.org , szaidi5@worldbank.org . This paper was prepared as a background paper for the Bangladesh Poverty Assessment. We would like to thank Faizuddin Ahmed, Syed Nizamuddin, Zahid Hussain, Kapil Kapoor, Martin Ravallion, Zaidi Sattar, Shekhar Shah, Binayak Sen, and participants of the Bangladesh Poverty Assessment workshops for useful comments and suggestions. . Page 3 Page 4 About the SASPR Working Paper The purpose of the SASPR Working Paper Series is to provide a quick outlet for sharing more broadly research/analysis of issues related to development in South Asia. Although the primary source of such research/analysis in SASPR staff, other contributors are most welcome to use this outlet for rapid publication of their research that is relevant to South Asia’s development. The papers are informal in nature and basically represent views/analysis of the concerned author(s). All papers submitted for publication are sent for an outside review to assure quality. I provide only a very light editorial touch. For enquiries about submission of papers for publication in the series or for copies of published papers, please contact Naomi Dass (telephone number 202-458-0335). Sadiq Ahmed Sector Director South Asia Poverty Reduction and Economic Management World Bank, Washington D.C. Page 5 Page 6 T ABLE OF C ONTENTS 1. INTRODUCTION........................................................................................................................1 2. TRENDS IN POVERTY INCIDENCE DURING THE NINETIES ........................................1 2.1 M ETHODOLOGY .........................................................................................................................................................1 2.2 A DJUSTING P OVERTY L INES FOR C HANGES IN C OST - OF -L IVING ...................................................................2 2.3 P OVERTY AND I NEQUALITY T RENDS ....................................................................................................................4 2.4 R EGIONAL T RENDS ....................................................................................................................................................7 3. GREATER PROGRESS IN FIRST OR SECOND HALF OF THE DECADE?.........................8 3.1 P ROGRESS OVER THE D ECADE ...............................................................................................................................8 3.2 P ROGRESS OVER THE F IRST AND S ECOND H ALF OF THE N INETIES ...............................................................9 4. SENSITIVITY ANALYSIS AND ROBUSTNESS CHECKS .......................................................9 4.1 C OMPARABILITY OF HES D ATA S ETS ..................................................................................................................10 4.2 A LTERNATE A PPROACHES TO D ERIVING P OVERTY E STIMATES ...................................................................11 4.3 O THER E VIDENCE OF C HANGES IN L IVING S TANDARDS ...............................................................................14 5. COMPARING BANGLADESH TO SOUTH ASIA AND EAST ASIA...................................... 17 6. SUMMARY OF POVERTY TRENDS ANALYSIS .................................................................... 19 APPENDIX TABLES........................................................................................................................... 21 BIBLIOGRAPHY .................................................................................................................................27 L IST OF T ABLES T ABLE 1. T RENDS IN CBN P OVERTY M EASURES .............................................................................................................4 T ABLE 2. T RENDS IN I NEQUALITY : G INI C OEFFICIENTS ................................................................................................7 T ABLE 3. R EGIONAL T RENDS IN P OVERTY ........................................................................................................................7 T ABLE 4. T RENDS IN N OMINAL AND R EAL PCE: N ATIONAL AND S ECTORAL ..........................................................8 T ABLE 5. H EADCOUNT R ATES : CBN-M ETHODOLOGY E STIMATES ...........................................................................11 T ABLE 6. H EADCOUNT R ATES : CPI AND TP- BASED E STIMATES ................................................................................12 T ABLE 7. H EADCOUNT R ATES : DCI- BASED E STIMATES ...............................................................................................13 T ABLE 8. H EADCOUNT R ATES : CBN I NCOME - BASED E STIMATES .............................................................................14 T ABLE 9. B ANGLADESH AND S OUTH A SIA : C OMPARISON OF S ELECTED I NDICATORS OF C HILD N UTRITION .18 T ABLE 10. I NTERNATIONAL C OMPARISONS OF S ELECTED D EVELOPMENT I NDICATORS .....................................18 T ABLE A1 B UDGET SHARES OF ITEMS WITH UNIT - VALUE INFORMATION IN THE HES..........................................21 T ABLE A2 R ELATIVE WEIGHTS OF ITEMS COVERED IN THE P RICE I NDEX ...............................................................22 T ABLE A3 S ELECTED UNIT - VALUES (T K ./ UNIT ) FROM THE S URVEYS .......................................................................23 T ABLE A4 C OMPOSITE P RICE I NDICES : 1991/92 – 1995/96 AND 1995/96 – 2000.................................................24 T ABLE A5 CBN P OVERTY L INES : U PDATING 1991-92 L INES WITH THE C OMPOSITE P RICE I NDEX ..................24 T ABLE A6 P OVERTY L INES : R EAPPLYING THE CBN METHODOLOGY TO EACH DATA SET ..................................25 T ABLE A7 S HARE OF HOUSEHOLD BUDGET ALLOCATED TO FOOD ITEMS ................................................................25 Page 7 T ABLE A8 P OVERTY L INES : U PDATING 1991-92 L INES WITH THE CPI....................................................................25 T ABLE A9 P OVERTY L INES : U PDATING 1991-92 L INES WITH THE HES-TP............................................................26 L I ST OF FIGURES F IGURE 1: C UMULATIVE D ISTRIBUTIONS OF M ONTHLY R EAL PCE: N ATIONAL , U RBAN AND R URAL ................6 F IGURE 2: C ONTRASTING P ROGRESS OVER THE F IRST AND S ECOND H ALF OF THE N INETIES .............................9 F IGURE 3: A VERAGE Q UANTITIES C ONSUMED ( GRAMS PER CAPITA PER MONTH )..................................................16 L IST OF B OXES B OX 1. A RE P OVERTY E STIMATES A CROSS C OUNTRIES IN S OUTH A SIAN C OMPARABLE ?...................................17 Page 8 1 Poverty Trends in Bangladesh during the Nineties Abstract: Analysis of data from various Bangladesh Household Expenditure Surveys (HES) suggests considerable progress at poverty reduction during the 1990s. About 50% of the country’s population lived below the poverty line in 2000 compared to 59% in 1991-92. Poverty in rural areas continues to be higher than in urban areas, but has declined at a fairly rapid rate in both sectors during the nineties. While the survey data and National Accounts show similar amounts of progress in Bangladesh over the decade as a whole, they present conflicting pictures of the pattern of growth over the decade: the National Accounts series indicate progress to have taken place at roughly equal rates over the first and second halves of the nineties, while the HES series show most of the progress at poverty reduction to have taken place during the first half. Introduction 1. The performance of the Bangladeshi economy in the past decade has been relatively strong, with annual growth in gross domestic product (GDP) averaging about 5% during the 1990s. Between 1991 and 2000, real GDP increased by 52 percent in real terms, with gross output in agriculture, services, and the industrial sector increasing by about 33 percent, 50 percent, and 86 percent respectively. Given the widespread interest in linkages between growth, equity, and poverty reduction, investigating the extent to which this impressive growth performance translated into reduced incidence of poverty in the country is an important one. 2. The Household Expenditure Surveys (HES) series conducted by the Bangladesh Bureau of Statistics (BBS) are the main data source for estimation of poverty in Bangladesh. These surveys are designed by BBS to be comparable over time (i.e. in terms of methodology, questionnaire content, interviewing procedures, etc.), and have been carried out in Bangladesh at regular intervals. This paper presents the main findings of the analysis of the 2000 Household Income and Expenditure Survey (HIES), as well as of earlier rounds of the HES series (i.e. the 1991-92 and 1995-96 surveys) to assess changes in poverty incidence in Bangladesh during the past decade. The analysis presented was carried out in close collaboration with BBS. 3. Trends in poverty and inequality in Bangladesh during the 1990s are presented in Section 2, which also outlines the various steps followed to derive these estimates. Section 3 compares selected findings from the various HES data sets with other data sources such as the National Accounts. This section includes a discussion of the extent to which the main HES findings are corroborated by these data sources, as well as highlights areas where the two present conflicting trends. Section 4 presents a brief discussion of the extent to which the three HES data sets are comparable with one another. In addition, estimates of poverty obtained by following alternate methodologies are also presented in this section, along with trends in other selected indicators of living standards. Section 5 contrasts the pace of poverty reduction in Bangladesh with its neighboring countries in South Asia and East Asia. Finally, Section 6 concludes by summarizing some of the main findings of the paper, as well as outlining areas where further work and research might prove fruitful. Trends in Poverty Incidence during the Nineties 2.1 Methodology Page 9 2 4 . BBS and the World Bank used the Cost-of-Basic-Needs (CBN) method to derive poverty lines and poverty measures from the 1991-92 and 1995-96 HES (BBS, 1997; World Bank, 1999). To summarize briefly, the CBN approach entailed three main steps: First, a food bundle yielding 2,122 kcal per day per person was chosen comprising rice, wheat, pulses, milk, mustard oil, beef, fresh water fish, potato, other vegetables, sugar, and bananas. Purging reported unit values in the survey data of possible variation due to differences in the quality of items consumed, the prices of the various food items in this bundle were estimated for 14 different geographic regions to ascertain the total cost of consuming this bundle in different parts of the country. 2 The second step was then to estimate the cost of basic non-food needs. Following the approach proposed by Ravallion (1994), two non-food allowance components were calculated: the first obtained by taking the amount spent on non-food items by those households whose total consumption was equal to their regional food poverty line (corresponding to the lower poverty line), while the second was obtained by taking the amount spent on non-food items by those households whose food consumption was equal to the regional food poverty line (corresponding to the upper poverty line). The third step in calculating the lower and upper poverty lines for each region entailed simply adding up the cost of purchasing the food bundle in each region to the respective non-food allowance components. The lower poverty lines thus incorporated a minimal allowance for non-food goods (the typical non-food spending of those who could just afford the food requirement) while the upper poverty lines made a more generous allowance (the typical non-food spending of those who just attained the food requirement). 5. In assessing trends in poverty over the decade, we hold fixed in real terms the poverty lines estimated by the CBN method at the beginning of the period – i.e. 1991-92 – and update in subsequent years each region’s base year poverty line for changes in the cost-of-living using a region- specific price index. The methodology used to derive these regional price indices is described briefly in the next section, while the poverty estimates obtained through following this approach are presented in Section 2.3. Alternative estimates of poverty trends obtained using poverty lines derived through other methodologies are discussed in Section 4. 2.2 Adjusting Poverty Lines for Changes in Cost-of-Living 6. There are several data sources that could potentially be used for estimates of cost-of-living increases needed to update the 1991-92 poverty lines. For instance, we could (i) rely on estimates of inflation from official sources such as the consumer price index (CPI) or the GDP deflator series, or (ii) use price indices derived from the HES datasets themselves using information on unit values of various consumption items collected in the surveys. 3 7. In some sense, the CPI is the natural choice for updating the poverty lines, as it is the standard yardstick used by most to assess changes in the cost-of-living over time. However, in the case of Bangladesh, the official CPI suffers from two main shortcomings: (i) it is based on a set of weights that have not changed since 1985-86, and may therefore be quite out-of-date in relation to current consumption patterns, and (ii) the national index, which is derived by aggregating urban and rural price indices, may be a poor proxy for changes in price levels in different regions. 2 The 14 regions used comprised: 1. Dhaka SMA, 2. Other urban areas of Dhaka division, 3. Rural areas of Dhaka and Mymensingh, 4. Rural areas of Faridpur, Tangail, and Jamalpur, 5. Chittagong SMA, 6. Other urban areas of Chittagong division, 7. Rural areas of Sylhet and Comilla, 8. Rural areas of Noakhali and Chittagong, 9. Urban areas of Khulna division, 10. Rural areas of Barishal and Pathuakali, 11. Rural areas of Khulna, Jessore, and Kushtia, 12. Urban areas of Rajshahi, 13. Rural areas of Rajshahi and Pabna, and 14. Rural areas of Bogra, Rangpur, and Dinajpur greater districts. 3 This, for instance, was the approach used by Deaton and Tarozzi (2000), who used similar data from the Indian National Sample Survey Organization (NSSO) data sets to derive inflation rates over time as well as across regions for their analysis of poverty trends in India. Page 10 3 8 . By contrast, an important advantage of using the HES data to derive price indices is that not only do the surveys report unit value information relating to actual transactions – i.e. rather than prices listed or reported by shops – but also that these data permit one to calculate region-specific indices to take into account differential rates of inflation across various parts of the country. However, one drawback of using data from surveys is that they rarely have information on prices of non-food items, and thus provide only a partial picture of the change in the aggregate price level. Food and non-food items (mainly fuels) for which unit values can be calculated from the HES surveys account for approximately two-thirds of total household expenditures. 4 The budget shares not covered in urban areas tend to be higher, which is a reflection of the relatively greater importance for urban consumers of goods such as housing and transportation for which we have no information on unit values. If the prices of these non-covered items change at a rate different from those items included in the index, then the price indices derived from the HES data may not fully capture changes in the cost-of-living over time. 9. As both the above alternatives – the CPI or HES-based price indexes – each have their advantages as well as shortcomings, we combined the two into a composite index so as to capitalize on the relative strengths of both approaches. 10. The HES-based price indices were derived in four steps. First, expenditures on various items in the HES were divided into 14 groups. These groups were chosen so as to retain as much disaggregation as possible (to minimize heterogeneity within categories) as well as to be comparable across the three survey years. 5 Second, unit values (by dividing expenditures by quantity) of the most commonly consumed item within each of the expenditure groups were calculated for each household. For each group, the median of the unit values within each geographic region was calculated. 6 Using the price of the most commonly consumed item within each group and medians (which are more robust to outliers as compared to means) for the summary region-specific unit values helped minimize the problem that the calculated unit values are contaminated by choice of quality rather than providing information on market price alone. Third, average budget shares of the 14 main expenditure groups were calculated for each survey year. Finally, region-specific Törnqvist price indexes were then calculated using budget shares of the expenditure groups along with median prices of the selected items. 7 The Törnqvist price indices for each region k were calculated as follows: p p w w P n j k j k j k j k j Tk å = ÷ ÷ ø ö ç ç è æ + = 1 0 1 0 1 10 ln 2 ln where P Tk denotes the Törnqvist price index for region k , 1 and 0 denote the two years of comparison, w k 1j and w k 0j are the respective budget shares, and p k 1j and p k 0j are the respective prices for good j in the two years of comparison. 11. Once the HES-based price indexes for each region had been derived from the survey data, we took a weighted average of these and the non-food component of the official CPI (disaggregated by urban and rural sectors) to derive region-specific cost-of-living indices for 1995-96 and 2000, the relative weights being the budget shares of covered goods in each region for the HES price index, and balance (i.e. one minus these budget shares) for the non-food CPI. The composite price indices were then used to update the 1991-92 CBN poverty lines to 1995-96 and 2000. 8 4 Budget shares are presented in Appendix Table A1. 5 Appendix Table A2 lists the relative budget share weights of each group in the overall HES price index for each year. 6 The median values of the unit values for the three surveys are reported in Appendix Table A3. 7 We used the chained Törnqvist price index in preference to the Laspeyres or Paasche indexes because it uses budget shares averaged between consecutive years, and therefore allows for changes in consumption patterns over time. 8 The composite price indices, as well as the CBN poverty lines for each region derived using these, are presented in Appendix Tables A4 and A5 respectively. Page 11 4 12. The derived composite price indices show cost-of-living in Bangladesh to have increased by, on average, about 16% between 1991-92 and 1995-96, and by about 12% between 1995-96 and 2000. Note that the overall 30% increase in price level between 1991-92 and 2000 implied by these indices is somewhat lower than the 35% increase in the GDP deflator over the same period, and much lower than the 52% increase in the overall CPI. We will return to the implications for poverty trends of this difference between the composite price index and the CPI in Section 4. 2.3 Poverty and Inequality Trends 13. Headcount rates based on both the upper as well as lower poverty lines show poverty in Bangladesh to have declined considerably during the nineties (Table 1). In 2000, 50% of Bangladesh’s population was poor (as measured by the upper poverty line) as compared to 59% in 1991-92. Similarly, 34% of the country’s population was very poor (i.e. below the lower poverty line) in 2000 as compared to 43% in 1991-92. Thus, according to both the upper and lower poverty estimates, the incidence of poverty in Bangladesh declined by about 9 percentage points over the course of the decade. Throughout the decade, poverty in rural areas remained higher than in urban areas; however, the overall decline in poverty incidence over time was roughly equal across the two sectors. 9 Table 1. Trends in CBN Poverty Measures Upper Poverty Line Change (Upper Line) Lower Poverty Line 1991-92 1995-96 2000 1991-92 to 1995-96 1995-96 to 2000 During the Decade 1991-92 1995-96 2000 HEADCOUNT RATE (P 0 ): National 58.8 51.0 49.8 -7.8 -1.2 -9.0 42.7 34.4 33.7 Urban 44.9 29.4 36.6 -15.5 +7.2 -8.3 23.3 13.7 19.1 Rural 61.2 55.2 53.0 -6.0 -2.2 -8.2 46.0 38.5 37.4 POVERTY GAP (P 1 ): National 17.2 13.3 12.9 -3.9 -0.4 -4.3 10.7 7.6 7.3 Urban 12.0 7.2 9.5 -4.8 +2.3 -2.5 4.9 2.6 3.8 Rural 18.1 14.5 13.8 -3.6 -0.7 -4.3 11.7 8.6 8.2 SQUARED POVERTY GAP (P 2 ): National 6.8 4.8 4.6 -2.0 -0.2 -2.2 3.9 2.5 2.3 Urban 4.4 2.5 3.4 -1.9 +0.9 -1.0 1.5 0.7 1.2 Rural 7.2 5.3 4.9 -1.9 -0.4 -2.3 4.3 2.8 2.6 14. The poverty gap (P1) estimates how far below the poverty line the poor are on average as a proportion of that line. The squared poverty gap (P2) takes into account not only the distance separating the poor from the poverty line, but also inequality among the poor. Trends in these measures broadly mirror those observed with the headcount rates. Both measures confirm that urban poverty remained lower than rural poverty throughout the decade. In addition, however, these measures also suggest that rural areas experienced greater reductions than urban areas in the depth and severity of poverty. 9 During the 1990s, the overall decline in poverty in Bangladesh as a whole (9.0%) was greater than in either urban (8.3%) or rural (8.2%) areas because (i) the share of population living in urban areas increased significantly during the period, and (ii) the incidence of poverty in urban areas was considerably lower than in rural areas. Page 12 5 1 5. Plotting the cumulative distributions for monthly real per capita expenditures (PCE) in Bangladesh (national, urban, and rural, respectively) for the three years confirms that trends in poverty between 1991-92 and 2000 (as well as between 1991-92 and 1995-96) are robust to the choice of the poverty line over the range of virtually all possible poverty lines (Figure 1). This is true for both the urban and rural sectors – the cumulative distributions for real PCE in 2000 are everywhere below and to the right of the cumulative distributions for 1991-92, indicating first-order stochastic dominance. Page 13 6 F i g u r e 1 : C u m u l a t i v e D i s t r i b u t i o n s o f M o n t h l y R e a l P C E : N a t i o n a l , U r b a n a n d R u r a l Cumulative fraction of population N a t i o n a l , b y y e a r P C E : t a k a p e r c a p i t a p e r m o n t h 1 9 9 1 - 9 2 1 9 9 5 - 9 6 2 0 0 0 9 0 4 0 0 8 0 0 1 2 0 0 1 6 0 0 2 0 0 0 0 . 2 . 4 . 6 . 8 1 Cumulative fraction of population U r b a n , b y y e a r P C E : T a k a p e r c a p i t a p e r m o n t h 1 9 9 1 - 9 2 1 9 9 5 - 9 6 2 0 0 0 9 0 4 0 0 8 0 0 1 2 0 0 1 6 0 0 2 0 0 0 0 . 2 . 4 . 6 . 8 1 Cumulative fraction of population R u r a l , b y y e a r P C E : T a k a p e r c a p i t a p e r m o n t h 1 9 9 1 - 9 2 1 9 9 5 - 9 6 2 0 0 0 9 0 4 0 0 8 0 0 1 2 0 0 1 6 0 0 2 0 0 0 0 . 2 . 4 . 6 . 8 1 1 9 9 5 - 9 6 1 9 9 1 - 9 2 2 0 0 0 1 9 9 5 - 9 6 1 9 9 1 - 9 2 2 0 0 0 1 9 9 1 - 9 2 1 9 9 5 - 9 6 2 0 0 0 Page 14 7 1 6. The HES data sets show much greater progress during the first half of the decade compared to the second half (Table 1): poverty in Bangladesh as measured by the HES fell by almost 8 percentage points between 1991-92 and 1995-96, but then by less than 1 point between 1995-96 and 2000. The data show rural poverty to have declined throughout the nineties, though at a considerably more rapid rate during the first half as compared to the second half (6.0 points drop vs. 2.2 points respectively). In urban areas, the HES series show poverty to have fallen a spectacular 15.5 percentage points during the first half, but then to have increased by about 7.2 percentage points in the latter half. As the pattern of decline in poverty over the two halves of the decade indicated by the HES series is quite different from that suggested by other data sources, we return to examine this puzzle in more detail in the next section. Table 2. Trends in Inequality: Gini Coefficients Upper Poverty Line Lower Poverty Line 1991-92 1995-96 2000 1991-92 1995-96 2000 National 0.259 0.302 0.306 0.272 0.315 0.318 Urban 0.307 0.363 0.368 0.311 0.369 0.370 Rural 0.243 0.265 0.271 0.251 0.267 0.275 17. Trends in inequality measured by Gini coefficients are reported in Table 2. The HES data sets suggest that inequality in Bangladesh has increased over time. Almost all of the increase occurred between 1991-92 and 1995-96, while, by contrast, inequality did not change much during the second period. While the Gini coefficients in 2000 are higher than for 1995-96, the Lorenz curves for the two years lie close to each other and cross at around the 80th percentile, suggesting that inequality changes between 1995-96 and 2000 cannot be ranked unambiguously. Over the decade, the rise in inequality in urban areas was considerably higher than that in rural areas. 2.4 Regional Trends 18. The HIES data reveal that the incidence of poverty varies quite considerably across different parts of the country, from a low of 39.7% in Barishal to a high of 61.0% in Rajshahi division (Table 3). Between 1991-92 and 2000, the decline in poverty was highest in Dhaka, followed by Barishal and Rajshahi divisions. By contrast, poverty in Chittagong appears to have stagnated during the nineties. Table 3. Regional Trends in Poverty Upper Poverty Line Change (Upper Line) Lower Poverty Line 1991-92 1995-96 2000 1991-92 to 1995-96 1995-96 to 2000 During the Decade 1991-92 1995-96 2000 All Divisions 58.8 51.0 49.8 -7.8 -1.2 -9.0 42.7 34.4 33.7 Barishal . 49.9 39.7 - -10.2 - . 39.1 28.8 Chittagong 46.5 52.3 47.7 5.8 -4.6 1.2 24.6 28.6 25.0 Dhaka 58.7 40.1 44.8 -18.6 4.7 -13.9 42.3 27.8 32.0 Khulna 59.9 55.0 51.4 -4.9 -3.6 -8.5 47.2 36.4 35.4 Rajshahi 71.8 61.8 61.0 -10.0 -0.8 -10.8 59.7 46.9 46.7 Page 15 8 Greater Progress in First or Second Half of the Decade? 19. How consistent are the HES findings – i.e. with regard to differential progress in poverty reduction over the two halves of the nineties – with other data covering the same period? We start first by presenting some summary statistics from the HES data sets, and then compare these with the National Accounts as well as wage data compiled by BBS. Our aim is to assess the extent to which these data report similar trends over the period under study. Table 4 reports trends in nominal and real per capita expenditures (PCE) by sector. Table 4. Trends in Nominal and Real PCE: National and Sectoral Mean per-capita expenditures (Tk. per month) Change (Percent) 1991-92 1995-96 2000 1991-92 to 1995-96 1995-96 to 2000 During the Decade NOMINAL PCE: National 550 764 876 39% 15% 59% Urban 829 1,344 1,390 62% 3% 68% Rural 503 649 747 29% 15% 49% REAL PCE: National 550 657 677 20% 3% 23% Urban 829 1137 1049 37% -8% 27% Rural 503 562 583 12% 4% 16% 3.1 Progress over the Decade 20. The HES data show nominal PCE in Bangladesh to have increased by about 59% during the nineties (Table 4). This is consistent with (if somewhat lower than) the 67% increase in nominal per capita private consumption or 76% increase in nominal GDP per capita reported by the National Accounts over the same period. 10 The 23% growth in real PCE is also consistent with the 24% increase in real per capita private consumption observed in the National Accounts. 11 21. Trends in PCE estimated from the HES data within the urban and rural sectors are also broadly consistent with changes in manufacturing and agricultural wages, which are typically considered to be very good proxies for living conditions. In particular, taking into consideration that inequality increased over the nineties, the 68% and 49% increase in nominal PCE in urban and rural areas conforms closely with the 64% and 44% increase in manufacturing and agricultural nominal wage indices respectively over the same period. 12 10 National Accounts Statistics of Bangladesh: Revised estimates, 1989-90 to 1998-99, BBS, Dhaka, December 2000, as well as latest GDP estimates for FY99—FY01. Note that the HES figures are based on a 19% increase in population between 1991-92 and 2000 while the NA assume a 15% increase over the same period; recalibrating the HES-estimates assuming the same population increase as in the NA raises nominal growth in PCE to 61%. 11 Even though the increase in nominal PCE from the HES is lower than that reported in the NA, real PCE growth rates are similar because the price index used to deflate nominal PCE in the former is lower than the GDP deflator. 12 Nominal wage index series presented in the 1999 Statistical Yearbook of Bangladesh (BBS, 2001). The respective wage series have been extrapolated past 1998-99 using the growth rates for the 97-98 to 98-99 period. Page 16 9 2 2. Finally, the relatively higher increase in nominal PCE in urban vs. rural areas (68% vs. 49%) is also in line with the differential rate of growth of agriculture and other sectors over the same period. The National Accounts show per capita output of the agriculture sector to have increased by 52%, the services sector to have increased by 80%, and per capita output of the industrial sector to have increased by 100% in nominal terms over this period. 3.2 Progress over the First and Second Half of the Nineties 23. While the HES and NA estimates are fairly consistent with regard to growth in PCE over the nineties, the two series present differing snapshots of the pattern of growth over time. The HES series show most of the increase in PCE to have taken place over the first half of the decade, while the NA series indicate a very similar magnitude of change over the two periods. According to the HES series, mean PCE increased by 39% between 1991-92 and 1995-96, but by only 15% between 1995-96 and 2000; the National Accounts, by contrast, show per capita private consumption to have increased by 28% and 31% over the two periods (Figure 2). Figure 2: Contrasting Progress over the First and Second Half of the Nineties 24. Which of the two – the HES or the National Accounts – gives the correct picture of rate of progress over the two halves of the decade? In the absence of clear evidence in support of either standpoint, it is difficult to make a definitive assessment in this regard. On the one hand, it is difficult to reconcile the 62% and 3% increase respectively in urban PCE during the first and second periods reported by the HES to trends in the manufacturing wage index as well as sectoral GDP growth rates. This would suggest that the HES series may have overestimated growth in urban PCE between 1991/92 and 1995/96, while underestimating the increase that took place between 1995/96 and 2000. However, on the other hand, both levels as well as trends in rural PCE reported by the HES are consistent with trends in the agricultural wage index. By contrast, it is difficult to explain why the acceleration in per capita GDP growth in the latter half of the nineties indicated by the National Accounts was not accompanied by a commensurate rise in agricultural wages over the same period. Sensitivity Analysis and Robustness Checks 39% 15% 59% 28% 31% 67% 29% 37% 76% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 1991/92 to 1995/96 1995/96 to 2000 1991/92 to 2000 Survey PCE NA Priv. Cons. NA GDP per capita 62% 3% 68% 26% 30% 64% 29% 15% 49% 22% 18% 44% 0% 10% 20% 30% 40% 50% 60% 70% 80% 1991/92 to 1995/96 1995/96 to 2000 1991/92 to 2000 Urban PCE Manufac. Wage Rural PCE Agri. Wage Page 17 10 4.1 Comparability of HES Data Sets 25. The HES/HIES series are the main source of data used for estimation of poverty in Bangladesh. While the same organization (Bangladesh Bureau of Statistics) has conducted the surveys, and similar survey methodology, questionnaires, interviewing procedures, etc. have been used over time, it is nonetheless worth investigating if any of the changes adopted over time have compromised the degree of comparability of data sets. We examine in turn three important aspects of the surveys and their implementation: 26. Questionnaire design: Consumption data in the HES series is collected using three different recall periods. Data on food consumption is collected on a daily basis (i.e. separate record of consumption on each day) through multiple visits, while data on non-food items is collected using a monthly and annual recall period. The lists of individual food and non-food line items covered in each survey have remained fairly similar over the years. However, in 1995-96, in addition to the usual income and consumption modules, an education module was added for the first time to the questionnaire. In 2000, the coverage of the questionnaire was expanded further, with additional modules added on housing, health, fertility, and economic activities. BBS was careful to plan field operations and distribution of workload amongst interviewers in a manner that took into account the additional time needed to canvass the more detailed questionnaires, so it is unlikely that this expansion in coverage adversely affected comparability of the consumption data collected. However, this possibility cannot be completely ruled out, and is therefore worth noting. 27. Field implementation procedures: Over the years, there have been a few changes in the way the food consumption module has been administered in the HES. In 1991-92, these data were collected over a 14-day period through daily visits (i.e. 14 daily records through 14 visits). However, in 1995-96, BBS switched to collecting food consumption data for 7 days only (i.e. 7 daily records through 7 visits). Finally, in 2000, BBS reverted back to the 14-day period, but with data collection taking place every alternate day (i.e. 14 daily records through 7 visits). Could it be that the sharp rise in PCE observed during the first period was due to less respondent fatigue during the 1995-96 round on account of the fewer number of visits undertaken? 13 Once again, as was the case with changes in questionnaire coverage, this possibility cannot be ruled out entirely, but we do not think it played a major role in accounting for the large difference between the two rounds. For one, we observe a significant decline in the share of PCE attributable to food items over this period, the exact opposite of what would have happened if higher food consumption had been reported by households in the 1995-96 round. Moreover, closer scrutiny of food consumption data from the 2000 survey provides no clear evidence of any statistically significant difference in food consumption estimates derived using data from the 1st week or 2nd week of the interview (i.e. which could be attributed to respondent fatigue). 14 28. Sampling: In all three surveys, household interviews were conducted in the same set of Primary Sampling Units (PSUs), a feature that enhances considerably the degree of comparability between the three data sets. For instance, all 371 PSUs covered in the 1995-96 HES were also covered in the 2000 HIES sample. However, in addition to the original 70 PSUs from the statistical metropolitan areas (SMA), an extra 70 PSUs were added in 2000 from the SMA. These extra PSUs were randomly selected from all SMA PSUs in the sample frame, and hence provide a very powerful means to assess 13 The reasoning being that respondent fatigue arising from multiple visits over an extended period results, after a certain number of visits, in progressively less food consumption being reported for each additional day. Since the 1995-96 survey entailed half the number of visits as the 1991-92 survey, the data collected in this survey would have been less susceptible to this problem. 14 Specifically, we tested if there was any significant difference in food consumption reported by households for the first 7 days compared to food consumption reported for the second 7 days of the interviewing cycle. The difference between the two estimates (18 Tk. per capita per month, with the former being higher than the latter) was not statistically significant. Page 18 11 t he extent to which the HES sample is representative of the SMA. PCE estimates across the two sub- samples (i.e. the 70 old and 70 new PSUs) were found to be extremely close to one another (1,545 and 1,557 Tk. per month respectively), thus confirming that the HES sample in urban SMA is in fact a good representation of all households that reside in these areas. This also suggests that the modest increase in urban PCE during the second period is unlikely to be attributable to any particular sampling-related quirks. Had this been the case, it would be hard to explain why 700 additional randomly selected households (10 in each PSU) yielded estimates of PCE that were in such close conformance to the rest of the sample. 4.2 Alternate Approaches to Deriving Poverty Estimates 29. Clearly the poverty estimates presented earlier (henceforth referred to as the CBN estimates) depend critically on how the underlying poverty lines were derived. In this section, we explore five alternative approaches to estimating these lines. First, rather than updating the 1991-92 CBN poverty lines to subsequent years, an alternative is to simply apply the same methodology to re-estimate poverty lines; i.e. the 1995-96 and 2000 poverty lines can be derived by applying the CBN approach to each data set separately. 15 Poverty estimates derived using this approach are referred to as CBN methodology estimates. Second, rather than using the composite price index to update the 1991-92 poverty lines, one could instead use the CPI for this purpose: the poverty estimates this gives rise to are henceforth referred to as CPI-based estimates. Third, rather than using the composite price index combining the HES-based index and the CPI to update the poverty lines, the HES-based Törnqvist index alone could be used for this purpose, in turn yielding the TP estimates. Fourth, one could also estimate poverty incidence in the country using the Direct Calorie Intake (DCI) method. Finally, poverty estimates could also be derived using the same CBN poverty lines derived in Section 2.2, but instead using the per capita income (PCI) HES aggregates rather than per capita expenditure (PCE) as the welfare yardstick. Poverty estimates obtained through each of these approaches are presented and discussed in this section. 30. CBN methodology estimates: Poverty trends using this approach applied to all three years are surprising (Table 5): they show poverty to have declined somewhat between 1991-92 and 1995-96, but then to have risen back to 1991-92 levels by 2000. A major drawback of this approach is that if living standards in a country improve over time, and even poor households spend a larger share of their income on non-food items, the allowance made for these items in the poverty line increases over time as well. In Bangladesh, the share of spending on non-food items has increased not just for the overall population, but also for households in the bottom two quintiles of the income distribution from 27.2% in 1991-92 to 32.8% in 1995-96, to 35.4% in 2000. 16 As a result, reapplying the CBN methodology to each year makes progressively larger non-food allowances in 1995-96 and 2000, and the poverty lines therefore no longer reflect basic-needs bundles of constant value in real terms. Assessing trends in absolute poverty over time presupposes that the same yardstick was used at all points in time, a condition that is violated by the CBN-methodology estimates. Table 5. Headcount Rates: CBN-Methodology Estimates Upper Poverty Line Change (Upper Line) Lower Poverty Line 1991-92 1995-96 2000 1991-92 to 1995-96 1995-96 to 2000 During the Decade 1991-92 1995-96 2000 National 58.8 53.1 59.7 -5.7 6.6 0.9 42.7 35.6 38.0 Urban 44.9 35.0 47.1 -9.9 12.1 2.2 23.3 14.3 23.4 15 This was the approach applied to the 1995-96 HES data by BBS and in the World Bank’s Poverty Assessment (World Bank, 1999). Poverty lines estimated by this method are presented in Appendix Table A6. 16 Budget shares are reported in Appendix Table A7. Page 19 12 Rural 61.2 56.7 62.9 -4.5 6.2 1.7 46.0 39.8 41.7 31. CPI or HES TP-based estimates: 17 An alternative to using the composite price index to update the 1991-92 poverty lines is to instead use the official CPI for this purpose. This approach is also a CBN-based method in that it also attempts to keep poverty lines constant in real terms, but assumes that the true change in cost-of-living during the nineties was the higher rate reflected by the CPI rather than that indicated by the composite price index. The CPI-based poverty estimates show the percentage of population that was very poor increased by 5% and the percentage that was poor (below the upper poverty line) to have increased by approximately 4% over the decade (Table 6). In urban areas, these estimates show negligible decline over the decade overall, since a sharp decline in poverty between 1991-92 and 1995-96 was followed by a period of rising poverty in the latter half of the nineties. In rural areas, by contrast, these estimates suggest that poverty increased quite considerably over the decade. However, one reason why the official CPI may overestimate the increase in price level is that the weights used to construct this index are by now fairly out-of-date (the expenditure weights used have not been updated since 1985-86). Second, the CPI is a Laspeyres price index; as is well known, such indices tend to over-estimate the increase in the price level over extended periods of time, as they do not take into account substitution in consumption towards goods whose relative price has fallen. Table 6. Headcount Rates: CPI and TP-based Estimates Upper Poverty Line Change (Upper Line) Lower Poverty Line 1991- 92 1995-96 2000 1991-92 to 1995-96 1995-96 to 2000 During the Decade 1991-92 1995-96 2000 CPI-based estimates: National 58.8 56.6 63.0 -2.3 6.4 4.1 42.7 40.5 47.6 Urban 44.9 32.8 44.9 -12.1 12.1 0.0 23.3 15.7 27.3 Rural 61.2 61.3 67.5 0.1 6.2 6.3 46.0 45.4 52.8 HES TP-based estimates: National 58.8 48.5 43.6 -10.3 -4.9 -15.2 42.7 32.1 27.7 Urban 44.9 29.1 34.3 -15.8 5.2 -10.6 23.3 13.3 16.8 Rural 61.2 52.4 45.9 -8.8 -6.5 -15.3 46.0 35.9 30.5 32. Similarly, one could instead use the lower HES-based Törnqvist price indices to update the 1991-92 poverty lines to 1995-96 and 2000. On account of the lower assumed rate of inflation, the TP-based poverty estimates show poverty to have declined at a more rapid rate during the nineties (Table 6). However, as discussed earlier, if the price of goods not covered in the HES-based Törnqvist price indices increased at a more rapid rate than the goods on which these indices were based, these indices would underestimate the increase in price level, and hence overestimate the decline in poverty. 33. DCI estimates: The DCI method entails choosing a minimum threshold of caloric consumption as a measure of welfare, and then defining as poor any household with a caloric intake less than this threshold. One advantage of this method is that as long as the calorie threshold is kept fixed, poverty measures estimated at different points in time represent the same living standard, and are therefore readily comparable over time. In its summary reports on the 1991-92 and 1995-96 Household Expenditure Surveys, BBS adopted a nationwide upper threshold of 2,122 kcal and a lower threshold of 1,805 kcal per person per day. We used the same thresholds to estimate poverty 17 CPI and HES-TP based poverty lines are reported in Appendix Tables A8 and A9, respectively. Page 20 13 i ncidence with the 2000 HIES. The headcount rates at the national and sectoral level using this method are summarized in Table 7. 18 Table 7. Headcount Rates: DCI-based Estimates Upper Caloric Threshold (2,122 kcal per day) Lower Caloric Threshold (1,805 kcal per day) 1991-92 1995-96 2000 1991-92 1995-96 2000 National 47.5 47.5 46.5 28.0 25.1 24.4 Urban 46.7 49.7 53.0 26.3 27.3 27.3 Rural 42.6 47.1 44.8 28.3 24.6 23.7 34. As the table shows, the DCI-based poverty estimates provide mixed support for our preferred CBN poverty estimates. On the one hand, the decline in poverty incidence during the 1990s is much lower using the DCI method as compared to the CBN approach. However, consistent with the CBN estimates, the DCI estimates also show urban poverty, as measured at the upper caloric threshold, to have increased from 50% to 53% during the latter half of the decade. The cumulative distributions of caloric intake for the latter two surveys suggest that the finding of rising urban poverty in this period is robust to choice of a caloric threshold between 1,800 and 2,500 kcal per person per day: any cutoff within this interval would yield an increase in urban DCI poverty estimates. Also consistent with the CBN estimates, the DCI estimates of rural poverty incidence (and nationwide) at the lower caloric threshold show that poverty declined during the nineties, with much of the gains accruing in the first half. However, this pattern is not borne out at the upper caloric threshold. 35. The main differences between the DCI and CBN estimates relate to progress in poverty reduction in urban areas during the nineties, and to rural-urban comparisons. First, unlike the CBN method, which suggested substantial progress in urban poverty reduction, the percentage of population below a minimum caloric threshold increased between 1991-92 and 2000 in urban areas. Second, the CBN estimates consistently show rural poverty to be higher than urban poverty, whereas the DCI measures suggest that during the nineties, poverty rates in urban areas have become higher than in rural areas. However, this is not surprising since the DCI method applies the same caloric threshold to both urban and rural areas, which translates to a corresponding poverty line (in terms of real per capita expenditures) that is considerably higher in urban than rural areas. 19 36. CBN income-based estimates: Compared to earlier rounds, the 2000 HIES questionnaire included much more comprehensive coverage of different income sources of households. This enabled the construction of per capita income measures in addition to the more traditional per-capita consumption measures, which in turn can be used to estimate the incidence of poverty (Table 8). For 2000, CBN poverty estimates based on income are over 5 percentage points lower than those based on per capita consumption (44.2% vs. 49.8%). As was the case with consumption-based poverty estimates, the CBN-income estimates show rural poverty to be considerably higher than urban poverty (47.5% vs. 31.2%). The table also includes CBN income-based poverty estimates for earlier 18 1991-92 and 1995-96 headcount rates are BBS estimates. In order to ensure that the caloric conversion factors we applied to the 2000 data were comparable to those used earlier by BBS, we re-estimated poverty rates for 1995-96. Our estimates for 1995-96 were found to be very similar to those computed by BBS, suggesting that all three estimates presented above are comparable. 19 Because of higher food prices and lower caloric requirements (for instance, because of less physically demanding labor), the urban calorie Engel curve tends to lie lower than the rural calorie Engel curve. This implies that if one were to use a common minimum caloric threshold for urban and rural areas, caloric requirements would be achieved only at much higher PCE in the urban areas (Bidani and Ravallion, 1994). Page 21 14 y ears. However, it is important to note that, on account of the considerably improved coverage of different income sources in the most recent HIES questionnaire, the poverty measures for 2000 are not comparable to those for 1991-92 and 1995-96. It is nevertheless interesting to see that – as was the case with the DCI as well as CBN-poverty estimates – the CBN income-based estimates also show urban poverty to have increased between 1995-96 and 2000. Table 8. Headcount Rates: CBN Income-based Estimates Upper Poverty Line Change (Upper Line) Lower Poverty Line 1991-92 1995-96 2000 1991-92 to 1995-96 1995-96 to 2000 During the Decade 1991-92 1995-96 2000 National 56.8 49.3 44.2 -7.5 -5.1 -12.6 45.5 34.8 30.6 Urban 56.5 28.3 31.2 -28.2 3.1 -25.3 41.6 14.0 16.8 Rural 56.9 53.5 47.5 -3.4 -6.0 -9.4 46.1 38.9 34.1 37. In sum, while the DCI poverty estimates provide mixed support for the CBN estimates, the CBN methodology as well as CPI-based estimates shows virtually no progress in poverty reduction in Bangladesh over the nineties. Not only does this run counter to the expected drop in poverty suggested by the National Accounts which show a 3% rise in per capita incomes per year, but these alternate poverty estimates are also sharply at odds with other indications within the same data sets showing considerable improvement in living conditions in Bangladesh during this period. These findings are presented and discussed in the following section. 4.3 Other Evidence of Changes in Living Standards 38. One of the main drawbacks of the DCI method is that it makes no allowance for improvements in composition of the food bundle consumed. Analysis of the average quantities of different food items consumed reveals that per capita consumption of virtually all major food groups (with the notable exception of rice, wheat, and pulses) increased substantially during this period (Figure 3). For instance, between 1991-92 and 2000, per capita consumption of fish increased by 9%, meat increased by 48%, poultry increased by 120%, milk increased by 55%, cooking oil by 26%, while sugar consumption increased by 11%. While per capita wheat consumption dropped considerably in 2000, 20 and rice consumption declined marginally (by about 3%), consumption of potatoes increased by 25% during this period. Similarly, the 13% decline in consumption of pulses is probably also due to substitution away from pulses towards higher-value sources of protein. 39. Figure 3 also reports per capita food consumption for households in the bottom two quintiles in urban and rural areas, and shows that the trend reported above was not confined to upper income groups only; consumption of fish, meat, poultry, and milk by the poor also increased considerably over the nineties. 40. In addition to confirming improvements in living conditions in Bangladesh during the nineties, the graphs also support the earlier finding of rapid progress during the first half of the decade, but then modest decline in poverty between 1995-96 and 2000. For instance, while there are indications 20 Part of the decline in wheat consumption is probably due to two important factors: (i) sharp rise in the relative price of wheat in relation to rice, and (ii) lower distribution of wheat through the various food assistance programs. During a period of bumper rice production and large rice stocks in the Public Food Distribution System (PFDS), rather than curtail procurement, in some instances the Government resorted to distributing rice instead of wheat. Page 22 15 o f considerable improvement in consumption patterns between 1991-92 and 1995-96, the pattern over the latter half of the nineties is much murkier. 41. Finally, the graphs also help vividly illustrate the difference in living conditions between urban and rural areas. As indicated by the CBN poverty estimates presented earlier, virtually all households in the bottom two quintiles in urban and rural areas fall below the poverty line. However, across these two groups, there are striking differences in consumption patterns: levels of consumption of rice, wheat, vegetables, and pulses for the urban and rural poor are fairly similar; however, on average the urban poor consume considerably more high-value food items like meat, fish, poultry, milk, oils, and sugar. In contrast to what the DCI poverty estimates suggest, the poor in urban areas, at least in terms of the range and quantities of different food items they consume, appear to be considerably better off than their counterparts in rural areas. Page 23 1 6 F i g u r e 3 : A v e r a g e Q u a n t i t i e s C o n s u m e d ( g r a m s p e r c a p i t a p e r m o n t h ) R i c e W h e a t V e g e t a b l e s P u l s e s F i s h M e a t P o u l t r y P o t a t o e s M i l k O i l S u g a r O : O v e r a l l U : U r b a n U 2 : U r b a n : b o t t o m 2 q u i n t i l e s R : R u r a l R 2 : R u r a l : b o t t o m 2 q u i n t i l e s L E G E N D 0 5 0 0 1 0 0 0 1 5 0 0 2 0 0 0 2 5 0 0 3 0 0 0 3 5 0 0 4 0 0 0 4 5 0 0 5 0 0 0 O U R U 2 R 2 0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 O U R U 2 R 2 0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 O U R U 2 R 2 0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 O U R U 2 R 2 0 5 0 0 1 0 0 0 1 5 0 0 2 0 0 0 2 5 0 0 O U R U 2 R 2 0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 O U R U 2 R 2 0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 O U R U 2 R 2 0 2 0 0 0 4 0 0 0 6 0 0 0 8 0 0 0 1 0 0 0 0 1 2 0 0 0 1 4 0 0 0 1 6 0 0 0 O U R U 2 R 2 0 2 0 0 4 0 0 6 0 0 8 0 0 1 0 0 0 1 2 0 0 1 4 0 0 1 6 0 0 1 8 0 0 O U R U 2 R 2 0 2 0 0 4 0 0 6 0 0 8 0 0 1 0 0 0 1 2 0 0 O U R U 2 R 2 0 2 0 0 4 0 0 6 0 0 8 0 0 1 0 0 0 1 2 0 0 1 4 0 0 1 6 0 0 O U R U 2 R 2 1 9 9 0 - 9 1 1 9 9 5 - 9 6 2 0 0 0 Page 24 17 Comparing Bangladesh To South Asia And East Asia 42. Bangladesh’s pace of poverty reduction compares favorably with its South Asian neighbors. The decline in income poverty of about one percentage point per year during the nineties is in sharp contrast to the virtual stagnation Bangladesh experienced during the eighties. 2 1 The reduction in poverty in Bangladesh during the nineties also compares favorably with other countries in the region. Box 1 . Are Poverty Estimates Across Countries in South Asian Comparable? While trends in poverty reduction are comparable across countries in South Asia, estimates of poverty incidence levels are not. The national statistical offices in India, Pakistan, and Bangladesh all prepare poverty estimates using data from fairly similar national household surveys conducted on a regular basis in their respective countries. In India, the NSSO Consumer Expenditure Survey Series is used to estimate the incidence of poverty in the country. The Government of India Planning Commission’s latest estimates using the 1999-00 survey show poverty in India to be 26.1 percent. In Pakistan, the Household Income and Expenditure surveys (HIES) conducted by the Federal Bureau of Statistics are used to estimate the incidence of poverty. Using data from the 1998-99 HIES, the incidence of poverty in the country was estimated to be 32.6 percent. In Bangladesh, two measures of poverty are estimated by BBS, corresponding to the upper and lower poverty lines. Using the upper poverty line, poverty was estimated to be 49.8 percent in 2000, while the lower poverty line yielded estimates of extreme poverty of 33.7 percent. Is poverty in Bangladesh so much higher than in India or Pakistan, as indicated by these estimates? Cross-country comparisons of poverty are fraught with complex measurement and comparability issues and, amongst other factors, depend on the yardstick used to assess poverty levels in the respective countries. However, comparing the poverty lines used across these three countries suggest that part of the reason why poverty estimates in Bangladesh are so much higher than in either India or Pakistan is that a considerably higher poverty line is used to assess poverty (see table below). In US dollar terms, the upper poverty lines in use in Bangladesh are considerably higher than those in India and Pakistan (at prevailing exchange rates, not PPP- adjusted like the often used $1 per person per day line) Poverty Lines in India, Pakistan, and Bangladesh Country Year Poverty line per capita per month Poverty Line US$ (prevailing Ratio to upper line (in US$) in (local currency) exchange rate) Bangladesh India Urban: 1999/00 Rs. 454 9.88 0.62 Rural: Rs. 328 7.14 0.57 Pakistan Urban: 1998/99 Rs. 665 13.27 0.84 Rural: Rs. 589 11.76 0.95 Bangladesh Upper line: Urban: 2000 Tk. 832 15.85 1.00 Rural: Tk. 652 12.42 1.00 Lower line: Urban: Tk. 628 11.96 0.75 Rural: Tk. 549 10.46 0.84 43. In India, where the economy grew at about 6 percent per annum during the nineties, consensus is emerging that poverty declined by roughly 5-10 percentage points over a 6 year period between 1993-94 and 1999-00, a magnitude similar to that observed in Bangladesh. However, in Pakistan where the rate of GDP growth has slowed down considerably in the latter part of the nineties, recent evidence suggests that poverty has more or less stagnated over the nineties. And in Sri Lanka, poverty declined at a considerably slower pace, by 6 percentage points between 1985 and 1995. 21 Earlier World Bank estimates show poverty in Bangladesh to have been stagnant at 59 percent between 1983-84 and 1991-92 (World Bank, 1999). Similarly, Ravallion and Sen (1996) estimate that rural poverty in Bangladesh declined only marginally from 54 percent in 1983-84 to 53 percent in 1991-92. Page 25 18 44. How do non-income indicators of living standards in Bangladesh compare to other countries? Using measures of stunting, wasting, and children underweight from Demographic and Health Surveys carried out in India and Bangladesh in 1998-99 and 1999-00 respectively, Bangladesh compares favorably with India (Table 9). The comparison with Pakistan and Sri Lanka is more mixed. While Bangladesh has lower rates of stunting and wasting than Pakistan, the percentage of underweight children is far greater. Table 9. Bangladesh and South Asia: Comparison of Selected Indicators of Child Nutrition Nutrition Status Bangladesh India Pakistan Sri Lanka Indicator 1999-00 1998-99 1990-91 1987 Stunting (height-for-age) % below 2 std. deviations 50 57 57 34 % below 3 std. deviations 20 32 36 - Wasting (weight-for-height) % below 2 std. deviations 9 13 10 13 % below 3 std. deviations 1 2 1 - Underweight (weight-for-age) % below 2 std. deviations 56 58 46 48 % below 3 std. deviations 17 24 19 - Source Various DHS Reports. For comparability, comparison limited to children 24-35 months (24-36 for Sri Lanka). 45. Comparisons of other development indicators show that Bangladesh, with a lower GNP per capita, has done reasonably well on some dimensions but lags with respect to others when compared with other South Asian countries (Table 10). It has lower population growth and mortality rates than both India and Pakistan. Access to improved water supply is better in Bangladesh, although this success is being threatened by the problem of arsenic contamination of groundwater. Adult literacy remains a problematic area relative to other countries, although Bangladesh has made significant strides in improving gender parity in enrollments. 46. While cross-country comparisons always require some care, the recent experiences of Vietnam, a country with the same GNP per capita as Bangladesh, may point to what is possible. Between 1993 and 1998, Vietnam experienced a 21 percentage point drop in poverty, spurred in large part by an ambitious reform program that included land reform, liberalization of agricultural input and output markets, freeing up the informal sector, and equitable investments in human capital. 22 Between 1992 and 1998, the average annual GDP growth rate in Vietnam was a spectacular 8.4 percent, with agricultural growth averaging 4.5 percent, industrial growth 13 percent, and the services sector growing by 8 percent per annum. In addition to progress in reducing consumption-based poverty, Vietnam has also achieved substantial progress in educational and health outcomes, which are now comparable to those of other East Asian countries that have much higher income levels. Vietnam’s experience suggests that the poverty reduction payoffs to further reforms and institutional development in Bangladesh could be substantial. Table 10. International Comparisons of Selected Development Indicators Indicator Bangladesh China India Pakistan Thailand Vietnam GNP per capita: US$ 370 780 450 470 1,960 370 Population growth: % 1.6 1.1 1.8 2.5 1.2 1.8 Urban population: % of total 24 32 28 36 21 20 Health Male life expectancy at birth: years 58 68 62 61 70 66 Infant mortality: per 1,000 live births 73 31 70 91 29 34 22 See World Bank (2000) for more details on progress in Vietnam during this period. Page 26 19 Under-5 mortality rate: per 1,000 96 36 83 120 33 42 Access to water and sanitation (% of population with access) Access to improved water source 84 90 81 60 89 36 Access to sanitation 35 21 16 30 96 21 Literacy and Education Male illiteracy: % of age 15 & older 49 9 33 42 3 5 Female illiteracy: % of age 15 & older 71 25 57 71 7 9 Net primary school enrollment 75 100 77 .. 88 100 Source : World Development Indicators. Notes : Estimates are from 1999, or most recent estimates reported in the Database. Summary of Poverty Trends Analysis 47. Both survey-based CBN poverty estimates as well as those based on the National Accounts show that the nineties were a period of declining poverty in Bangladesh. The proportion of the very poor declined from 43% in 1991-92 to 34% in 2000, while the proportion of the poor fell from 59% to 50%. Poverty in rural areas continues to be higher than in urban areas, but has declined at a fairly rapid rate in both sectors during the nineties. The improvement in living conditions is evidenced not only by increases in PCE, but also by a shift in composition of the food bundle consumed by the poor towards more high value items. 48. While the incidence of poverty has fallen considerably during the decade, examination of the total number of individuals living below the poverty line reveals a more sobering picture: the total population living below the upper poverty line in 2000 remained virtually unchanged (at about 63 million) compared to 1991-92, while the population living below the lower poverty line declined somewhat from 45.2 million in 1991-92 to 42.5 million in 2000. The analysis carried out has also brought to light a number of puzzles that warrant further attention: 49. Discrepancy with National Accounts: The discrepancy between the HES and National Accounts series relates to the pattern of growth over the decade, with the HES surveys indicating much more modest progress at poverty reduction over the latter half of the decade as compared to the NA statistics. Assessing which of the two gives the correct picture of poverty trends is problematic as there exists supporting evidence for both standpoints. 23 On the one hand, since 1995- 96, the Bangladeshi economy as a whole (agriculture in particular) has performed quite well. Inflation has remained within single digits, the price of rice is virtually unchanged in real terms, and per capita availability of essential food items has improved considerably. Secondary data on wages and agricultural incomes also point toward improvements in living standards through the nineties. On the other hand, HES data suggest that much of the increase in PCE as well as improvement in composition of the average food bundle consumed has taken place over the first half of the nineties. Similarly, while the share of total expenditures allocated to food has gone down considerably over the decade, most of this decline took place over the first half. Finally, enrollment rates derived from the two data sets suggest that the proportion of primary school-age children attending school has declined between 1995-96 and 2000. 50. One important contra-indication to the otherwise bleak picture painted by the HES series for the latter half of the nineties is the 34% increase in wages (in nominal terms) and 23% increase in median crop revenues per capita between 1995-96 and 2000, in contrast to the much smaller 15% increase in mean consumption. Amongst the possible reasons the latest HES data set may 23 In addition, there are several reasons (e.g. differences in items included in NA consumption vs. survey consumption measures, and various sources of measurement errors in both the NA and survey data) why it is not surprising to find a discrepancy between National Accounts and household survey based consumption estimates (Deaton, 2000). Page 27 20 u nderestimate improvement in living standards could be that the constructed welfare measure excludes important items for which expenditure has increased considerably in recent years (purchase of livestock and other assets damaged or destroyed by the 1998 floods), that it does not fully capture improved access to publicly provided goods and services, or that it does not fully account for increase in household savings rates, etc. 24 Further analysis of shifts in employment patterns indicated by the most recent Labor Force Survey, analyzed in conjunction with the poverty profile yielded by the 2000 HIES survey (in particular, the relationship between poverty and sources of household income) may be helpful in better understanding the relationship between aggregate growth and poverty during the latter half of the decade. 51. Discrepancy with the CPI: A second important issue worth noting concerns the discrepancy in inflation estimates from the survey data and the official CPI series. As pointed out earlier, the Törnqvist indices derived from the survey data suggest that the price level in Bangladesh increased by about 20% between 1991-92 and 2000 while the CPI series show a rise of about 52% over the same period. Part of the discrepancy relates to the fact that the CPI is a Laspeyres index, which tends to over-estimate inflation over long time periods since budget-share weights are fixed at the base year level. In Bangladesh, these weights have not been revised since 1985-86, and may be quite out-of- date in relation to current consumption pattern. Given the widespread use of the CPI, updating these weights merits serious consideration by BBS. 52. Impact of Rural-Urban Migration on Poverty Estimates: Finally, the last issue we’d like to draw attention to concerns differentials in living standards between the urban and rural sectors. Notwithstanding the observed stagnation in urban poverty rates in recent years, living standards appear to be considerably higher in urban as compared to rural areas (as suggested by greater consumption of higher-value food items by the urban poor compared to the rural poor). The influx of migrants from rural to urban areas of Bangladesh appears to have continued unabated through the nineties. Results from the “quick-count” carried out for the recent Population Census suggest that the share of the country’s urban population has risen from around 14% in 1991-92 to over 20% in 2000. 53. A final question we would like to pose is whether the rural-urban cost of living differential implicit in our choice of poverty lines may have led to a slight underestimation of the decline in poverty. Recall that our choice of poverty lines for the two sectors was tied to the rural-urban differential embedded in the 1991-92 poverty lines which we updated to 1995-96 and 2000 using region-specific cost-of-living indices. These lines imply that the cost-of-living is anywhere up to 41% higher in urban as compared to rural areas. However, what if this overestimates the difference in cost-of-living across the two sectors? Consider the case of a person who is just above the poverty line in the rural sector, and who moves to the urban sector where he obtains a job generating a real income gain less than the difference in poverty lines across the two sectors. 25 Though that person may be better off in his new residence, the poverty measures used will show an increase in both urban as well as rural sectors (there is one less non-poor person in rural areas, and one more poor person in urban areas). Further investigation into the extent and nature of migration trends in Bangladesh (what types of individuals moved? what jobs were they engaged in before moving to urban areas? what types of jobs did they take up in their new residence? etc.) will doubtless be an important topic for future research. 24 Unfortunately, since comparable data on wages and incomes are not available in the 1991-92 HES, it is not possible to investigate the same trends over the 1991-95 period. 25 The argument outlined follows the one outlined in Ravallion (1994). Page 28 21 Appendix Tables Table A1 Budget shares of items with Unit-value Information in the HES BUDGET SHARES (%) REGION 1991-92 1995-96 2000 SMA Dhaka 65.3 53.4 52.2 OU Dhaka 74.8 61.0 58.7 R. Dhaka 75.9 72.3 71.5 R. Faridpur Tangail Jamalpur 82.4 74.8 72.2 SMA Chittagong 60.6 63.2 55.8 OU Chittagong 71.8 62.2 58.4 R. Sylhet Comilla 78.0 76.0 66.6 R. Noakhali Chittagong 74.2 71.5 63.1 U. Khulna 68.6 64.9 58.3 R. Barishal Pathuakali 80.3 73.1 66.3 R. Khulna Jessore Kushtia 75.3 70.4 68.3 U. Rajshahi 71.8 61.5 60.8 R. Rajshahi Pabna 76.7 70.2 71.4 R. Bogra Rangpur Dinajpur 78.7 71.4 68.5 Page 29 22 T able A2 Relative Weights of Items Covered in the Price Index FOOD GRAINS VEGETABLES PULSES FISH EGGS REGION 91 95 00 91 95 00 91 95 00 91 95 00 91 95 00 SMA Dhaka 35.2 29.7 29.4 9.4 8.4 8.5 4.3 3.5 3.6 12.3 13.9 13.5 1.4 2.1 2.2 OU Dhaka 44.9 37.8 33.3 8.0 7.8 7.7 3.6 3.4 3.4 8.1 10.6 11.0 1.3 1.7 1.6 R. Dhaka 52.8 44.9 39.6 6.7 8.4 8.0 2.9 2.3 3.2 7.0 10.0 11.4 0.7 0.8 1.8 R. Faridpur Tangail Jamalpur 55.5 49.9 43.0 6.7 8.0 8.2 3.0 2.2 2.6 6.9 8.3 9.5 0.7 0.9 1.1 SMA Chittagong 41.7 35.0 32.8 7.8 8.4 9.0 2.6 2.4 2.9 10.2 11.1 11.0 0.6 1.1 1.3 OU Chittagong 41.0 38.2 32.1 6.6 7.7 9.7 3.5 2.7 3.1 11.3 11.3 13.0 1.0 2.0 0.8 R. Sylhet Comilla 44.4 45.2 36.3 8.0 7.7 9.3 3.0 2.5 2.7 9.9 10.7 13.0 1.1 0.7 0.7 R. Noakhali Chittagong 45.4 41.3 37.9 7.3 8.9 9.3 2.6 2.1 3.2 9.7 12.7 13.4 0.7 0.8 0.7 U. Khulna 42.8 39.5 31.9 8.1 8.6 8.8 3.2 3.0 3.0 9.8 11.6 13.0 1.1 1.7 1.9 R. Barishal Pathuakali 49.2 45.6 38.8 6.2 7.5 9.8 3.9 3.4 3.1 8.3 11.7 13.5 0.9 1.0 1.2 R. Khulna Jessore Kushtia 53.1 47.8 42.8 8.3 9.0 9.2 2.2 2.0 2.1 7.7 9.5 10.0 0.7 0.9 1.1 U. Rajshahi 42.4 39.0 36.1 9.1 8.6 9.4 3.0 2.8 2.5 7.8 8.4 9.0 1.0 1.3 1.7 R. Rajshahi Pabna 54.2 48.9 45.0 8.2 9.3 8.9 1.9 2.3 2.6 4.8 6.7 8.3 0.6 0.9 1.0 R. Bogra Rangpur Dinajpur 55.7 50.7 47.6 8.6 8.8 9.8 2.3 1.7 1.5 5.8 6.5 7.1 0.7 0.9 1.1 MEAT SALT & SPICES MILK SUGAR COOKING OILS 91 95 00 91 95 00 91 95 00 91 95 00 91 95 00 SMA Dhaka 4.9 6.9 8.4 6.5 5.0 6.4 2.1 3.6 3.7 2.2 2.4 2.1 4.4 4.4 3.9 OU Dhaka 2.1 4.9 5.9 6.8 4.9 5.8 3.2 3.0 4.4 2.2 1.4 1.9 3.9 3.7 3.5 R. Dhaka 2.0 2.9 4.2 6.8 5.4 6.9 1.9 3.4 3.3 1.3 1.7 1.3 3.2 3.2 3.6 R. Faridpur Tangail Jamalpur 1.6 2.1 3.5 6.3 5.6 7.0 1.8 2.5 2.7 1.3 1.2 1.5 3.2 3.3 3.2 SMA Chittagong 3.7 4.8 7.0 4.6 4.1 6.9 3.9 3.0 2.7 2.0 1.9 2.6 2.8 2.8 3.3 OU Chittagong 3.5 4.6 6.1 4.6 4.4 7.0 2.6 2.7 2.5 1.9 1.2 2.0 3.4 2.7 2.9 R. Sylhet Comilla 3.0 2.5 3.9 7.7 5.1 6.9 1.8 2.2 2.7 1.6 1.5 2.0 3.8 2.9 3.2 R. Noakhali Chittagong 3.4 3.9 6.1 6.0 5.1 6.4 1.3 1.4 1.7 1.9 1.5 1.2 2.6 2.7 2.9 U. Khulna 3.7 5.1 6.2 6.4 4.3 6.4 2.2 2.6 2.6 1.6 2.1 2.0 4.6 3.9 4.1 R. Barishal Pathuakali 1.9 2.4 3.1 6.5 6.2 5.8 1.3 1.8 2.2 1.6 2.4 1.2 3.5 3.1 3.7 R. Khulna Jessore Kushtia 2.2 3.4 4.2 5.7 4.4 6.3 0.7 1.7 2.2 1.5 2.2 1.7 4.3 4.2 3.7 U. Rajshahi 4.8 5.8 7.0 5.8 5.1 7.8 2.3 3.0 2.8 1.5 2.0 1.6 4.2 4.5 3.9 R. Rajshahi Pabna 2.8 3.6 4.1 5.8 4.7 8.4 1.4 2.3 1.9 1.3 2.0 1.4 3.0 3.7 3.8 R. Bogra Rangpur Dinajpur 3.0 4.1 4.2 5.5 4.9 6.1 1.3 2.1 2.1 1.8 1.5 1.4 2.9 2.7 2.9 FRUITS SOFT DRINKS BETEL&TOBACCO FUELS 91 95 00 91 95 00 91 95 00 91 95 00 SMA Dhaka 1.6 3.5 4.0 2.2 3.1 0.6 5.1 4.8 1.9 8.6 8.9 11.8 OU Dhaka 0.7 2.6 5.0 1.5 1.2 0.4 4.4 4.8 3.2 9.2 12.2 12.8 R. Dhaka 1.0 2.0 2.8 0.3 0.8 0.1 4.8 4.7 3.1 8.7 9.7 10.7 R. Faridpur Tangail Jamalpur 1.0 1.4 2.0 0.3 0.4 0.1 5.3 5.1 3.2 6.3 9.0 12.2 SMA Chittagong 0.9 2.6 2.7 3.1 5.7 1.2 5.0 7.6 4.0 11.1 9.5 12.5 OU Chittagong 1.7 2.5 2.7 2.5 4.1 0.9 7.2 6.8 4.6 9.3 9.1 12.7 R. Sylhet Comilla 1.5 1.7 2.4 1.2 2.2 0.5 6.5 6.6 5.2 6.5 8.6 11.0 R. Noakhali Chittagong 1.1 1.5 1.8 3.3 3.3 0.4 8.2 6.8 4.1 6.5 8.1 10.8 U. Khulna 1.2 2.1 2.6 1.6 1.4 0.4 4.7 3.6 2.0 9.0 10.5 15.0 R. Barishal Pathuakali 0.9 0.9 2.6 1.4 1.1 0.2 5.6 5.7 3.7 8.7 7.3 11.0 R. Khulna Jessore Kushtia 1.0 1.4 1.6 0.3 0.3 0.2 4.2 3.6 1.9 8.0 9.5 13.1 U. Rajshahi 1.0 1.2 2.1 1.4 2.4 0.3 4.0 3.8 2.1 11.9 12.1 13.7 R. Rajshahi Pabna 0.3 1.8 1.2 0.4 0.5 0.2 5.1 3.2 1.6 10.1 10.2 11.7 R. Bogra Rangpur Dinajpur 0.5 1.0 1.9 0.5 0.8 0.1 3.8 4.1 2.8 7.5 10.3 11.4 Page 30 23 T able A3 Selected Unit-values (Tk./unit) from the Surveys Coarse Rice (kg) Masur (kg) Puti (kg) Hen’s egg (per 10) Beef (kg) REGION 91 95 00 91 95 00 91 95 00 91 95 00 91 95 00 SMA Dhaka 12 15 14 29 40 40 30 50 67 23 35 35 60 60 80 OU Dhaka 12 14 12 28 40 40 20 40 50 19 30 30 50 60 70 R. Dhaka 12 13 12 28 40 40 20 25 40 13 30 30 48 60 70 R. Faridpur Tangail Jamalpur 12 12 12 28 40 40 20 32 40 13 25 30 48 50 70 SMA Chittagong 13 14 13 30 40 40 24 48 50 23 35 40 60 70 80 OU Chittagong 12 14 13 28 40 40 23 40 50 23 30 40 60 70 80 R. Sylhet Comilla 12 13 12 28 40 40 20 27 40 13 25 30 60 60 80 R. Noakhali Chittagong 12 13 12 29 40 40 20 40 40 17 30 30 54 60 80 U. Khulna 12 13 12 28 40 40 22 30 44 19 30 30 48 60 70 R. Barishal Pathuakali 12 13 12 28 40 40 20 30 40 13 25 30 40 60 70 R. Khulna Jessore Kushtia 11 12 11 27 30 40 20 30 40 13 20 25 40 50 67 U. Rajshahi 12 12 12 28 40 40 22 30 40 18 30 30 40 50 63 R. Rajshahi Pabna 10 12 12 26 40 40 20 24 40 13 20 25 40 40 60 R. Bogra Rangpur Dinajpur 11 11 11 26 35 40 20 24 40 13 20 25 39 40 60 Potato (kg) Milk (kg) Sugar (kg) Mustard Oil (ltr) Salt (kg) 91 95 00 91 95 00 91 95 00 91 95 00 91 95 00 SMA Dhaka 8 8 10 20 22 26 31 32 40 50 60 60 10 10 10 OU Dhaka 7 8 10 13 16 20 31 36 32 50 60 60 10 10 10 R. Dhaka 7 8 8 12 16 16 30 32 32 55 60 60 8 10 10 R. Faridpur Tangail Jamalpur 7 8 8 12 13 16 31 40 32 53 60 60 7 8 7 SMA Chittagong 7 8 10 14 16 20 30 32 33 55 75 80 7 10 10 OU Chittagong 7 9 10 14 16 24 32 30 36 52 67 50 8 10 10 R. Sylhet Comilla 8 8 10 12 16 20 30 32 36 59 60 60 8 10 10 R. Noakhali Chittagong 8 8 10 12 15 20 30 32 34 51 60 60 8 8 8 U. Khulna 7 8 8 14 16 18 30 30 32 58 60 60 8 10 10 R. Barishal Pathuakali 6 8 8 10 12 16 30 32 30 54 60 60 8 10 10 R. Khulna Jessore Kushtia 6 8 8 10 12 12 30 30 30 53 60 60 7 10 9 U. Rajshahi 7 7 8 12 14 16 30 30 32 50 60 60 9 10 8 R. Rajshahi Pabna 6 8 8 10 12 14 30 28 32 50 60 50 8 8 7 R. Bogra Rangpur Dinajpur 7 8 6 10 10 12 30 30 31 51 60 60 8 8 6 Bananas (kg) Coke/Fanta (no.) Prepared Betel Kerosene (ltr) 91 95 00 91 95 00 91 95 00 91 95 00 SMA Dhaka 15 20 20 7 8 15 0.5 0.5 1.0 18 18 16 OU Dhaka 10 20 20 6 7 11 0.5 0.5 1.0 18 16 16 R. Dhaka 10 14 16 7 7 15 0.5 0.5 1.0 18 16 18 R. Faridpur Tangail Jamalpur 10 16 20 6 4 17 0.5 0.5 1.0 18 15 16 SMA Chittagong 9 20 27 7 8 15 0.5 0.5 1.0 18 16 16 OU Chittagong 10 20 20 7 8 12 0.5 0.5 1.0 18 16 16 R. Sylhet Comilla 8 20 20 7 7 17 0.5 0.5 1.0 16 16 16 R. Noakhali Chittagong 10 20 20 6 7 15 0.5 0.5 1.0 16 16 16 U. Khulna 9 16 15 6 8 15 0.5 0.5 1.0 20 16 16 R. Barishal Pathuakali 8 12 20 5 8 15 0.5 0.5 1.0 16 16 15 R. Khulna Jessore Kushtia 10 12 12 6 8 19 0.5 0.5 0.5 20 20 16 U. Rajshahi 10 16 13 7 6 15 0.5 0.5 1.0 18 16 16 R. Rajshahi Pabna 8 16 10 7 7 15 0.5 0.5 0.5 20 17 18 R. Bogra Rangpur Dinajpur 10 12 13 6 7 14 0.5 0.5 0.5 20 16 16 Page 31 24 T able A4 Composite Price Indices: 1991/92 – 1995/96 and 1995/96 – 2000 1991-92 to 1995-96 1995-96 to 2000 REGION Food HES Index Covered budget sh. Non- Food CPI Composite Price Index Food HES Index Covered budget sh. Non-Food CPI Composite Price Index SMA Dhaka 1.20 59% 1.20 1.20 1.10 53% 1.16 1.13 Other Urban Dhaka 1.20 68% 1.20 1.20 1.03 60% 1.16 1.08 Rural Dhaka 1.12 74% 1.26 1.16 1.07 72% 1.20 1.11 Rural Faridpur Tangail Jamalpur 1.08 79% 1.26 1.12 1.07 74% 1.20 1.12 SMA Chittagong 1.20 62% 1.20 1.20 1.09 59% 1.16 1.12 Other Urban Chittagong 1.20 67% 1.20 1.20 1.09 60% 1.16 1.12 Rural Sylhet Comilla 1.12 77% 1.26 1.15 1.11 71% 1.20 1.15 Rural Noakhali Chittagong 1.17 73% 1.26 1.19 1.06 67% 1.20 1.11 Urban Khulna 1.12 67% 1.20 1.14 1.06 62% 1.16 1.10 Rural Barishal Pathuakali 1.17 77% 1.26 1.19 1.05 70% 1.20 1.10 Rural Khulna Jessore Kushtia 1.16 73% 1.26 1.19 0.98 69% 1.20 1.05 Urban Rajshahi 1.07 67% 1.20 1.11 1.08 61% 1.16 1.12 Rural Rajshahi Pabna 1.13 73% 1.26 1.17 1.04 71% 1.20 1.10 Rural Bogra Rangpur Dinajpur 1.04 75% 1.25 1.10 1.01 70% 1.20 1.09 Table A5 CBN Poverty Lines: Updating 1991-92 Lines with the Composite Price Index 1991-92 1995-96 2000 REGION ZL ZU ZL ZU ZL ZU SMA Dhaka 480 660 574 791 649 893 Other urban Dhaka 399 482 480 580 521 629 Rural Dhaka 425 512 492 593 548 659 Rural Faridpur Tangail Jamalpur 432 472 484 529 540 591 SMA Chittagong 523 722 627 867 702 971 Other urban Chittagong 517 609 619 730 694 818 Rural Sylhet Comilla 432 558 499 644 572 738 Rural Noakhali Chittagong 438 541 522 645 582 719 Urban Khulna 482 635 552 727 609 803 Rural Barishal Pathuakali 413 467 494 558 546 616 Rural Khulna Jessore Kushtia 420 497 499 592 527 624 Urban Rajshahi 446 582 496 647 557 726 Rural Rajshahi Pabna 459 540 535 630 586 690 Rural Bogra Rangpur Dinajpur 426 487 468 535 510 582 Note: ZL is the lower poverty line; ZU is the upper poverty line. Amounts are in Tk. per person per month. Page 32 25 Table A6 Poverty Lines: Reapplying the CBN Methodology to each data set Region 1991-92 1995-96 2000 Ratio: 1995-96 to 1991-92 Ratio: 2000 to 1995-96 ZL ZU ZL ZU ZL ZU ZL ZU ZL ZU SMA Dhaka 480 660 613 950 746 1,236 1.28 1.44 1.22 1.30 OU Dhaka 399 482 584 931 724 979 1.46 1.93 1.24 1.05 R. Dhaka 425 512 523 661 575 712 1.23 1.29 1.10 1.08 R. Faridpur Tangail Jamalpur 432 472 521 604 581 703 1.21 1.28 1.12 1.16 SMA Chittagong 523 722 561 749 646 937 1.07 1.04 1.15 1.25 OU Chittagong 517 609 564 704 659 842 1.09 1.16 1.17 1.20 R. Sylhet Comilla 432 558 515 584 614 848 1.19 1.05 1.19 1.45 R. Noakhali Chittagong 438 541 548 638 628 858 1.25 1.18 1.15 1.34 U. Khulna 482 635 541 779 646 872 1.12 1.23 1.19 1.12 R. Barishal Pathuakali 413 467 522 639 600 748 1.26 1.37 1.15 1.17 R. Khulna Jessore Kushtia 420 497 481 563 556 694 1.15 1.13 1.16 1.23 U. Rajshahi 446 582 499 628 576 752 1.12 1.08 1.15 1.20 R. Rajshahi Pabna 459 540 480 582 522 624 1.05 1.08 1.09 1.07 R. Bogra Rangpur Dinajpur 426 487 457 570 529 671 1.07 1.17 1.16 1.18 Note: ZL is the lower poverty line; ZU is the upper poverty line. Amounts are in Tk. per person per month. Table A7 Share of Household Budget Allocated to Food Items Overall Population Bottom 40% Nominal Tk per person per month 1991-92 1995-96 2000 1991-92 1995-96 2000 PCE 550 764 876 326 427 473 PCE Food 353 432 463 236 287 305 PCE Non-Food 197 332 413 89 140 168 Share of PCE on food 64.2 56.5 52.8 72.6 67.2 64.6 Table A8 Poverty Lines: Updating 1991-92 Lines with the CPI Region 1991-92 1995-96 2000 ZL ZU ZL ZU ZL ZU SMA Dhaka 480 660 590 812 729 1004 OU Dhaka 399 482 491 593 607 733 R. Dhaka 425 512 523 630 647 779 R. Faridpur Tangail Jamalpur 432 472 531 580 656 717 SMA Chittagong 523 722 643 889 794 1098 OU Chittagong 517 609 635 750 785 926 R. Sylhet Comilla 432 558 532 686 657 848 R. Noakhali Chittagong 438 541 539 665 666 822 U. Khulna 482 635 593 782 732 966 R. Barishal Pathuakali 413 467 509 574 628 709 R. Khulna Jessore Kushtia 420 497 516 612 638 756 U. Rajshahi 446 582 549 715 679 884 R. Rajshahi Pabna 459 540 564 665 697 821 R. Bogra Rangpur Dinajpur 426 487 524 599 648 740 Note: ZL is the lower poverty line; ZU is the upper poverty line. Amounts are in Tk. per person per month. Page 33 26 T able A9 Poverty Lines: Updating 1991-92 Lines with the HES-TP Region 1991-92 1995-96 2000 ZL ZU ZL ZU ZL ZU SMA Dhaka 480 660 577 795 641 883 OU Dhaka 399 482 485 585 501 605 R. Dhaka 425 512 479 576 512 617 R. Faridpur Tangail Jamalpur 432 472 469 513 501 548 SMA Chittagong 523 722 634 876 691 956 OU Chittagong 517 609 624 736 682 805 R. Sylhet Comilla 432 558 486 627 540 697 R. Noakhali Chittagong 438 541 512 632 544 671 U. Khulna 482 635 543 717 574 757 R. Barishal Pathuakali 413 467 487 550 512 578 R. Khulna Jessore Kushtia 420 497 490 580 479 568 U. Rajshahi 446 582 480 626 517 674 R. Rajshahi Pabna 459 540 523 616 544 641 R. Bogra Rangpur Dinajpur 426 487 447 510 453 518 Note: ZL is the lower poverty line; ZU is the upper poverty line. Amounts are in Tk. per person per month. Page 34 27 Bibliography Bangladesh Bureau of Statistics (BBS). 1997. Summary Report of the Household Expenditure Survey 1995- 96. Dhaka. Bangladesh Bureau of Statistics (BBS). 2001. 1999 Statistical Yearbook of Bangladesh. Dhaka. Bidani, B. and M. Ravallion. 1994. “How Robust is a Poverty Profile?” The World Bank Economic Review , 8, pp.75-102. Deaton, A. 2000. Counting the World’s Poor: Problems and Possible Solutions, Princeton University. Deaton, Angus. and A. Tarozzi. 2000. Prices and Poverty in India. Research Program in Development Studies, Princeton University. Ravallion, M. 1994. Poverty Comparisons. Harwood Academic Press, Switzerland. Ravallion, M. and B. Sen. 1996. “When Method Matters: Monitoring Poverty in Bangladesh.” Economic Development and Cultural Change, 44: 761-792 . World Bank. 1999. From Counting the Poor to Making the Poor Count. South Asia Poverty Reduction and Economic Management Unit, Washington D.C. World Bank. 2000. Vietnam Development Report 2000: Attacking Poverty. Washington DC.