Fast Tracking poverty reduction and prosperity for all Dominican Republic Poverty Assessment 2023 Background Note on Poverty and Social Protection Central America and the Dominican Republic Country Management Unit Latin America and the Caribbean Region April 2023 © [2023] International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy, completeness, or currency of the data included in this work and does not assume responsibility for any errors, omissions, or discrepancies in the information, or liability with respect to the use of or failure to use the information, methods, processes, or conclusions set forth. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org Contents 3 Poverty and Social Protection ............................................................................................................... 4 3.1 UNCOVERING GAPS IN EDUCATION AND LABOR MARKET OUTCOMES .......................................... 5 3.2 ROLE OF SOCIAL PROGRAMS ........................................................................................................... 10 3.2.1 Background ................................................................................................................................ 10 3.2.2 Effects of CCT Programs on Education Outcomes ..................................................................... 13 3.2.3 Imputed Effects of CCT Programs on Employment and Earnings .............................................. 18 3.2.4 Effects of Job Training Programs on Employment and Wages .................................................. 20 3.3 CONCLUSIONS AND POLICY IMPLICATIONS .................................................................................... 22 4 REFERENCES ........................................................................................................................................ 23 5 ANNEXES ............................................................................................................................................. 24 Figures Figure 1. Education access, enrollment and completion rates for DR and LAC, 2000-2019 ........................ 5 Figure 2. Primary and lower secondary completion rates for DR 2004 - 2020............................................. 5 Figure 3. Enrollment rates by age and ICV Level, Q4 2021 ........................................................................... 6 Figure 4. Population with complete secondary education or more (18 – 24 years old) by sex (2021) ........ 6 Figure 5. Labor force participation rate by ICV and sex (2006 – 2021) ........................................................ 7 Figure 6. Informality and average labor income (25-54 years old) by ICV (2021) ........................................ 7 Figure 7. Employment rate and monthly labor income for 18-64 years old by sex and ICV Level (2021) ... 8 Figure 8. Young Population (18-24) classified as NEETs by ICV level and sex, (2021) .................................. 8 Figure 10. Returns to educational achievement in the DR (2002 – 2021) .................................................... 9 Figure 11. Education level of the economically active population (2002 -2021) ......................................... 9 Figure . Evolution of coverage of the main 3 CT programs (total households, 2004-2022) ....................... 12 Figure 12. Coverage, adequacy, targeting and efficiency of social protection programs for selected group of countries and the Dominican Republic (2019) ....................................................................................... 13 Figure 13. Cumulative density function distribution of the ICV for Avanza+, Aprende+ groups and for nonbeneficiaries.......................................................................................................................................... 14 Figure 14. Poverty changes decomposition (2013-2021) ........................................................................... 15 Figure 15. Figure Extreme poverty changes decomposition (2013-2021).................................................. 15 Figure 15. Figure ATT Effects of AVANZA+ on High School Graduation...................................................... 16 Figure 16. ATT Effect of AVANZA+ on High School Graduation .................................................................. 16 Figure 18. ATT Effects of AVANZA+ on Years of Schooling ......................................................................... 16 Figure 19 . ATT Effects on Years of Education ............................................................................................ 16 Figure 20. Increase in the probability of being employed and formality due to CCT ................................. 19 Figure 21. Growth of expected earnings due to CCT .................................................................................. 19 Boxes Box 1. Description of key features of operational processes of the PROSOLI Youth Program................... 10 Box 2. Further Evidence of the Effects of Supérate School Transfers on High School Completion ............ 17 Box 3. Estimating the expected earnings after the exposure to the Avanza program ............................... 19 Box 4. The Effects of Prosoli (now Superate) Capacity Building Program on Labor ................................... 21 3 Poverty and Social Protection The last two chapters described how welfare of Dominicans has improved in many respects during the last two decades, albeit unevenly across regions and income groups. Many poor households need basic support to start growing their incomes and to elevate themselves out of poverty. Nutritional interventions and cash transfers can help the poor to meet their most basic needs and improve the income opportunities of future generations through investments in health and education. To sustain income over their lifecycles, the poor and vulnerable also need access to better jobs. This chapter aims to assess whether key social programs and interventions in the Dominican Republic have built human capital and raised incomes of the poor. The chapter is organized as follows. Section 1 describes gaps in access to education and labor markets, which, despite secular improvements, still argue for social protection programs. Section 2 reviews the coverage and performance at reaching the poor of social protection in the DR. This is followed by an assessment of whether two of the main social protection programs in the country -namely, conditional cash transfers and vocational training- improve welfare. The final part of this section considers the implications of these findings for the current social protection system. The work draws primarily from three data sources. First, we have access to two SIUBEN data waves. The first wave corresponds to the period 2004-2008 and contains information on recipient and non-recipient families before the former were enrolled in the conditional cash transfer programs, thus serving as our baseline data. Wave 1 includes the ICV score used to target families during most of our analysis period, all household-level variables underlying the ICV index, as well as other sociodemographic characteristics of households and their members. We also have access to SIUBEN’s third census, conducted in 2018, containing the same set of variables as the first one. Together, waves 1 and 3 allow us to analyze changes in households’ conditions across time. The second data source is the cash transfer programs payroll data provided by the Social Subsidies Administrator (ADESS). As noted in chapter 2, the historical payroll of beneficiaries for the main cash transfers provided in the Dominican Republic is administered by the ADESS. It consists of a historic panel starting in October 2004, with all transfers provided to each beneficiary household head, including date and amount of each transfer. This information can be directly matched to the SIUBEN household data through the household’s head national ID. In other words, these data allow us to classify families according to their beneficiary status, identify the amount of each type of transfers they have received and for how long. Aside from these administrative datasets related to the conditional cash transfer programs, we make use of several waves of the National Labor Force Survey, Encuesta Nacional Continua de Fuerza de Trabajo or ENCFT. The nationally representative ENCFT, administered by the Central Bank of the Dominican Republic for more than two decades, is the source for the country’s official poverty and labor market statistics. As such, the main goal of the survey is to collect information on income generation activities and on public and private transfers receipts of family members. We use the ENCFT a to assess the labor market returns of social protection programs in the country. 3.1 UNCOVERING GAPS IN EDUCATION AND LABOR MARKET OUTCOMES Chapter 2 showed that non-monetary welfare has improved in the past two decades. The proportion of people who were multidimensional poor (proxied through a quality-of-life Index)1 declined from 60 percent in 2005 to 25 percent in 2021. Improvements in educational attainment, discussed below, account for an important part of the reduction in multidimensional poverty. The country has seen noticeable progress in educational access and attainment. The country has achieved almost universal access to primary education, especially in urban areas, and increased secondary enrollment by over 80 percent. The primary completion rate (PCR) increased from 79 percent in 2000 to 93 percent in 2019. Secondary enrollment has also increased remarkably. However, the enrollment and completion rates are low relative to the regional average (figure 1) as is the quality of education, as DR ranks 146th out of 148 countries with low quality education (OCDE, 2019). Figure 1. Education access, enrollment and Figure 2. Primary and lower secondary completion rates for DR and LAC, 2000-2019 completion rates for DR 2004 - 2020 100 98.3 97.9 120 97.7 93.3 85.3 81.6 78.8 100 90 60.1 80 Percentage 80 Percentage 60 70 40 12.2 60 4.0 3.6 2.8 20 0 50 DR LAC DR LAC 40 2007 2012 2004 2005 2006 2008 2009 2010 2011 2013 2014 2015 2016 2017 2018 2019 2020 2000 2019 Children out of school (% of primary school age) Lower secondary completion rate, total (% of relevant age group) Primary completion rate, total (% of relevant age group) Primary completion rate, total (% of relevant age group) School enrollment, secondary (% gross) Source: WDI Indicators. Note: Lower secondary in the Dominican context alludes to the first cycle of secondary education (1st to 3rd grade). The average schooling gains conceal large disadvantages faced by poor children. Analysis based on the ENCFT reveals that school enrollment gaps between poor and better off children widen after primary education. The phenomenon is particularly salient among 15 and 18 year-olds, ages corresponding to the final year of the first and second cycles of secondary education, with school attendance rates of just 50 and 21 percent, respectively, for children in households classified as ICV-1 (figure 3). The Dominican Republic has the highest prevalence of child marriage in LAC (UNICEF 2019), especially common in rural areas and among poorer and less educated girls. According to the ENHOGAR-MICS household survey conducted in 2019, 32 percent of women aged between 20-24 at the time of the survey were married or in a union before the age of 18 (9 percent before their 15th birthday). These girls are more 1 The quality-of-life index (Indice de Calidad de Vida or ICV as abbreviated in Spanish) measured by the Single Beneficiary Selection System (Sistema Unico de Beneficiarios or SIUBEN as abbreviated in Spanish) reflects household deprivations in access to basic services (electricity, sanitation and water), housing materials (roof, walls and floor), durable goods and education (level of education of household head, average education and literacy of household members, and school attendance of children aged 6-14), as well as the level of overcrowding and the share of under-fives in the household; all measured through 15 variables. Those variables get combined through statistical techniques to produce a score which serves to classify households according to four categories: ICV-1: Extreme Poor; ICV-2: Moderate Poor; ICV-3: non-poor but vulnerable; and ICV-4: non-poor. likely to drop out of school than their peers who marry later and they tend to complete fewer years of education (UNICEF 2019). Child marriage has also been associated with intra-family violence, and violence against girls. Poor teenage girls are at a particular disadvantage. As girls enter adolescence, existing gender roles and social norms continue to affect their ability to stay in school and to transition into work, especially among the poorest. Adolescent pregnancy, at 93 births per 1,000 women aged between 15-19 years in 2018, remains exceptionally high in the DR by global standards and well above the LAC average of 62 (World Bank DR Gender Assessment, 2022). And teenage pregnancy is strongly associated with not working or studying. According to the most recent Labor Force Surveys in the country (ENCFT), boys cite lack of interest in school and joining the labor market as the main reasons for leaving their studies. Girls drop out of secondary school predominantly due to financial constraints and for family reasons, including becoming pregnant. As a result, secondary completion rates are much lower among poor girls than for the rest of the population (figure 4). Keeping teenagers in school, especially girls is a clear priority given the multiple negative consequences associated with dropping out. Figure 3. Enrollment rates by age and ICV Figure 4. Population with complete secondary Level, Q4 2021 education or more (18 – 24 years old) by sex (2021) 100% 100 90% 80% 80 70% Percentage 60% 60 50% 40% 40 30% 20% 20 10% 11.3 20.6 27.6 43.5 50.1 68.7 80.2 89.0 0% 0 7 - 12 13 14 15 16 17 18 1 2 3 4 (avg) Age ICV categories ICV1 ICV2 ICV3 ICV4 Male Female Source: Own calculations based on ENCFT. The fact that children and teenagers growing up in poverty have lower rates of school enrollment and completion raises concerns about their employment prospects and social mobility more broadly. Fifteen years ago, the participation of men in labor markets was twice as high as that of women. Although women have been gradually catching up, even in 2021 only around 56 percent were working or actively looking for a job compared to around 77 percent of men (figure 5). Much higher shares of working-age adults living in poverty (income quintile 1) were informally employed and earning lower salaries compared to their peers in the ranks of the non-poor (figure 6). While the male labor force participation rate is about average for Upper-middle-Income countries and for LAC, the female rate has, until only recently, remained much lower than the LAC average. Figure 5. Employment rate by poverty status Figure 6. Informality and average wage income and sex (2004-2021) (25-54 years old) by income quintile (2021) 80 40 73 80 65 33 70 35 61 58 Dominican pesos 30 60 60 43 Percentage 25 50 18 Percent 20 15 40 40 13 15 11 30 10 20 20 5 10 0 0 I II III IV V 0 Quintile 2004 2006 2008 2010 2012 2014 2016 2017 2019 2021 Male Non poor Male Poor Average monthly labor income (thousands) Female Non poor Female Poor Informality rate (%) Source: Own calculations based on ENFT y ENCFT. Note: The official national labor market survey (ENFT) was methodologically updated from 017 giving way to the renewed (ECNFT) , as a result, the indicators calculated with the ECNFT are not strictly comparable with previous years. Women lag relative to men in many other labor market outcomes There are large gender gaps in employment and labor income. As of 2021, women are more likely to experience unemployment (12 percent compared to 4 percent), work less (5 hours per day on average across economic sectors) and earn less (on average, 85 percent of men’s earnings, falling to 56 percent in the informal sector) than men. These gaps are even wider for the poor, with the average employment rate for men twice that for women living in households classified as ICV-1 level (figure 7). Looking across education levels, the unemployment rate for women with basic education2 was more than 3 times higher in 2021 than the rate for comparable men (14.8 vs 4.6 percent), and around twice that observed among men at the intermediate and advanced levels. These gaps are related to the disproportionate shouldering of housework and care activities by women; in 2016, women allocated more than 31 hours to unpaid work per week, three times the amount allocated by men. This difference is larger in rural and poorer regions (World Bank DR Gender Assessment, 2022). In 2021, the incidence of young (18-24 year-old) women not engaged in education, employment, or training (NEETs) is 23 percent compared to 12 percent of young men, a gap particularly acute among members of the poorest (ICV- 1) households (figure 8). 2 Basic education is defined as incomplete secondary or less. Figure 7. Employment rate and monthly labor Figure 8. Young Population (18-24) classified as income for 18-64 years old by sex and ICV Level NEETs by ICV level and sex, (2021) (2021) 41.3 25 100 45 40 Dominican pesos (thousands) 20 80 31.1 35 Percentage (%) 30 23.7 15 60 22.0 21.5 Percent 25 17.9 15.5 10 40 20 13.7 12.2 12.0 11.5 15 5 20 10 4.9 4 13 5 13 6 15 15 22 5 0 0 ICV 1 ICV 2 ICV 3 ICV 4 0 ICV 1 ICV2 ICV 3 ICV4 Labor income Employment rate Female Labor income Male Labor income Female Male Total Female Employment rate Male Employment rate Source: Own calculation based on ENCFT. Source: Own calculation based on ENCFT. *Defined for non-attending school population. The economic crises triggered by the COVID-19 pandemic reinforced pre-existing inequalities in education and employment. The bottom 40 percent of the income distribution saw a 9 percent drop in employment, while school attendance for those aged 3 to 18 also declined by 28 percent. The top 60 percent of the income distribution experienced drops of 7 percent and 25 percent, respectively. Women made up 42 percent of the labor force in 2019 but accounted for 54 percent of those who stopped working or seeking employment in 2020, with urban women being particularly impacted. The informal sector, which constitutes around half of female employment, was responsible for 60 percent of job losses. By 2020, 37 percent of young women were classified as NEET (Not in Employment, Education or Training) compared to 21 percent of young men, with a total estimated NEET population of 373 thousand people. This represents a 60 percent increase, or 141 thousand additional NEETs compared to the same period in 2019. The gaps in access to education and jobs impede progress on poverty reduction. Labor earnings are the main source of livelihood for most Dominicans, representing 70 percent of total household income on average in 2019. Real wages dropped dramatically after the banking crisis of 2004 and were fairly stagnant up until 2013. Only then did wage income start growing in real terms and labor incomes started to catch up with productivity growth. Nevertheless, this was not enough to make up for the lost ground since 2004.3 There are sizeable returns to secondary education in the Dominican Republic. Education and other forms of human capital are widely considered essential for economic growth and poverty reduction. High school completion among the economically active population has grown from 23 to 37 percent in the last decade (see Figure 11). On average, completing high secondary schooling leads to about 25 percent increase in hourly wage-earning relative to not having an education degree. Estimates from the 2021 National Labor Force Survey (Encuesta Nacional Continua de Fuerza de Trabajo – ENCFT as abbreviated in Spanish) indicate that employment rates for individuals in urban and rural areas who complete upper secondary is 3 Winkler, H. and M. Montenegro. 2021. “Dominican Republic Jobs Diagnostic.� World Bank, Washington, DC. 69 percent, 23 p. p. higher than individuals with incomplete secondary (46 percent). In addition, individuals who complete high school in urban (rural) areas earn US$ 88 (57) more a month than individuals with incomplete secondary education. Figure 9. Returns to educational achievement in Figure 10. Education level of the economically the DR (2002 – 2021) active population (2002 -2021) 0.5 100% 10.4 13.9 0.4 80% 20.1 37.1 0.3 8.7 60% Return Percentage 10.7 0.2 9.6 40% 0.1 10.6 45.2 20% 24.0 0.0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 0% 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 Completed primary None or incomplete primary Complete primary Complete lower secondary Incomplete secondary Complete lower secondary Complete upper secondary Complete upper secondary Complete tertiary Source: Authors calculation based on a mincer equation. The estimation controls for age, age squared, urban, region, sex, and informality. Despite improvement in enrollment and the seemingly high economic returns to education, approximately one in two male students and one in three female students do not complete upper secondary school, concentrated among the poor. Closing the gender and poor/non-poor jobs gaps would help reduce poverty in the country. In 2017- 2019, poverty dropped 4.7 percentage points in the Dominican Republic. A recent Gender Assessment performed a poverty change decomposition analysis by income and employment sources disaggregated by sex. Such analysis revealed that 1.2 percentage points of the reduction in poverty were due to the increase in the employment rates and wages of women. In other words, despite facing disadvantages in terms of labor market inclusion and remuneration, women contributed with one quarter (25 percent) of the total reduction in poverty during this period. The Jobs Diagnostic for the Dominican Republic identifies five factors that may have contributed to anemic real wage growth during the last two decades. On the labor demand side, the contributing factors include the lagging adjustment of minimum wages, market concentration and anti-competitive behavior in various economic sectors, and factor-biased technological change. On the supply side, the factors include a protracted increase in labor force participation rates and the existence of gender discrimination in connection with a growing labor market participation of women. These issues are discussed at length in the Jobs and Gender Diagnostics. The next section of this chapter provides an overview of the main social programs that the Dominican government has put in place to improve the education and employment opportunities for the poor and then delves into their results. 3.2 ROLE OF SOCIAL PROGRAMS 3.2.1 Background The DR has made significant progress in building a Social Protection (SP) System over the past 15 years. Following the economic crisis in 2003, poverty in the Dominican Republic spiked in 2004 reaching 50 percent (compared to 32 percent in 2002). One of the measures that the government took to address the situation was the launch of the SOLIDARITY PROGRAM in October 2004 which granted conditional cash transfers (Comer es Primero – CEP for its acronym in Spanish) for food to socioeconomically vulnerable families. Between 2005 and 2011, the food transfer coverage almost tripled reaching close to 0.6 million families by the end of 2011, at which point additional non-conditional transfer schemes were established. Together, these transfers now constitute the main component of the Dominican non-contributory social protection system reaching close to 40 percent of the country’s total population by mid-2022. In December 2005, the program instituted the first schooling-related transfer, the Incentivo a La Asistencia Escolar or ILAE. It was targeted to all eligible households with children in school age (6-21) to promote school attendance at primary and secondary levels. The overall aim of CEP/ILAE is to support family consumption and contribute to reducing the intergenerational transmission of poverty by encouraging investment in health and education through conditioning on pre and post-natal exams, immunization programs, routine checkups for children until the age of five, and on school enrollment and attendance.4 By 2012, the Solidarity Program had given way to the Progresando con Solidaridad Program (PROSOLI) with interventions aimed at increasing citizen awareness, income generation (see Box 1), access to ICT, and at reducing the digital divide. Additionally, in June of 2013, a new conditional cash transfer was established specifically targeting high school age children to further encourage attendance at this level (Bono Estudiantil Estudiando Progreso – BEEP). Box 1. Description of key features of operational processes of the PROSOLI Youth Program A World-Bank-financed project, “Integrated Social Protection and Promotion Project�, P147213, was implemented by the Social Cabinet to support the PROSOLI Youth Program during the period 2015-2022. Under Sub-Component 1.3. Increase access of eligible Conditional Cash Transfer users to productive opportunities, the Project aimed at increasing the employability of young women and men (ages 18-29 years) in extremely poor and moderately poor PROSOLI households through two main series of activities: (i) the carrying out of technical, vocational and life skills training courses and provision of apprenticeships; and (ii) the carrying out of periodic diagnostics of employers’ training needs and training for the Eligible Training Providers. PROSOLI was the technically responsible unit of carrying out the activities related to this sub-component, Implementation benefitted from existing good practice and lessons learned from the former Bank-supported Youth Development Project (7371-DO), including, among others, competitive selection of training providers (COS) and focus on employers’ needs. These youth employability activities were implemented through a partnership between the National Training Institution (INFOTEP) and the Social Cabinet/PROSOLI. Finally, this Sub-Component financed technical assistance and capacity-building to enhance the knowledge and capacity of PROSOLI to link extremely and moderately poor households to income-generating opportunities through self-employment. 4 See PROSOLI Operations Manual (2017). Key features and operating processes of this Sub-Component. The Project comprised in-classroom training and on- the-job learning through internships at private firms. The training course includes a Life Skills Training Course (LSTC) and a Vocational Technical Training Course (VTTC). Both components are taught by the Private Training Providers (COS), which were regulated and by the INFOTEP (National Training Institution). The COS were competitively selected through a bidding process launched by the Social Cabinet. The operating costs incurred by INFOTEP were reimbursed by Social Cabinet for the evaluation of proposals and supervision of training course delivery. The COS were responsible for delivering the training courses and identifying vacancies at private firms for beneficiaries to participate in two- month internships. The competitive selection process was defined in the operations manual of the Project, including the payment schedule and accountability scheme. The beneficiaries were selected by PROSOLI, using a first selection of the eligible households, grouping by the smallest geographical unit available, and performing a random selection and classification in as many segments as necessary for the bidding process. Selected households were notified by the Family Liaison, who communicated the course options available for that specific round of the training program. Households were responsible for communicating which qualifying member (18-29 years) would be attending the course as well as the chosen VTTC. The household must also inform the Social Cabinet about the beneficiary's bank account information, which was used for the stipend and payment of remuneration during the internship. The Family Liaison, jointly with the Social Cabinet’s UTP, opened bank accounts for those beneficiaries that did not have one at enrollment time. Once the information was collected, PROSOLI matched beneficiaries to training courses. The Social Cabinet/PROSOLI were responsible for the implementation of the training program and entered into an interinstitutional agreement with INFOTEP. The inter-institutional agreement established payment schemes and amount between Social Cabinet and INFOTEP for services monitoring and evaluation, as well as the supporting documentation that had to be approved by PROSOLI. PROSOLI communicated to INFOTEP once a year, the course content to be included in the bidding documents as well as information on the geographic distribution of the courses to be competitively selected. The Social Cabinet launched the bidding process to competitively select COS. The COS submitted to INFOTEP proposals to enter the bidding process and be selected, sign a contract with Social Cabinet’s UTP for the provision of training courses during the in-class phase, as well as assigning beneficiaries to firms during the internship phase. INFOTEP received COS’ proposals, evaluated and recommended to the Social Cabinet/PROSOLI; and during course delivery, INFOTEP monitored course delivery and internships. The Social Cabinet made payments to COS against course delivery according to the arrangements set forth in the contract between these parties. Finally, INFOTEP was responsible for supervising implementation during in class and on-the- job-phases. The Social Cabinet/ PROSOLI validated and paid youth the stipends during the training and the internships. Overall, PROSOLI was in charge of all technical aspects for the implementation of this sub-component and was the unit approving the product and services of INFOTEP and the COS to process payments. A total of 30,000 youth were supported through this scheme over the course of four years from 2015 to 2019. In 2021, the SUPERATE Program was launched, which emphasizes building skills to raise incomes and employability building on past experiences with active labor market policies. SUPERATE continues to make use of conditional cash transfers to support consumption and encourage the accumulation of human capital through the Aliméntate, Aprende and Avanza transfers, which replace the Comer es Primero, ILAE and BEEP transfers. More importantly, SUPERATE provides access to active labor market policies (ALMPs), such as technical vocational training and apprenticeships (see box 1).5 5 SUPERATE is in fact more than a CT scheme, as it expands the scope of its predecessor, PROSOLI. SUPERATE includes a larger set of CTs, comprising: (i) conditional cash transfers paid based on the compliance of co- responsibilities in education and health; (ii) newly created components such as unconditional cash transfers (UCT) for emergency response; (iii) economic inclusion services; and (iv) support services for households and individuals (providing identity cards to individuals, energy and fuel subsidies, and housing improvements). Ministerio de Economia, Planificacion y Desarrollo de la Republica Dominicana (2019). The DR has an SP system that is generally efficient, gender-sensitive and well-targeted to the poor, but there are some gaps in coverage. According to the 2021 ENCFT,6 around 43 percent of Dominican households receive social assistance benefits, with the share rising to 57 percent of households in the bottom 20 percent of the income distribution. The share of beneficiaries among low-income households is higher for female headed households (62 percent) than for male headed households (53 percent). Moreover, 60.1 percent of the households registered in SIUBEN are headed by a woman with 39.5 percent of them categorized as extreme poor or poor (ICV1 or ICV2, respectively).7 SUPERATE programs have recently expanded to have coverages comparable to those of other LAC and Upper middle-income countries (UMC), yet coverage gaps persist. Between 2005 and 2011, the food transfer coverage of SUPERATE almost tripled, reaching close to 0.6 million families by the end of 2011. By 2019, the coverage of social assistance programs for the bottom 20 percent of the income distribution in the DR was comparable to the regional LAC average, as well as to upper-middle countries and world- wide. Another expansion occurred during the pandemic to reach 1.6 million households, which was partially rolled back to reach roughly 1.37 million conditional transfer recipient households by 2022, equivalent to almost 40 percent of the country’s population. The CT Aliméntate has by far the largest coverage followed by Aprende and Avanza (figure 12). 94,148 poor and extremely poor households are eligible for SUPERATE but are not enrolled. Another 107,632 eligible households lack an identity document to obtain coverage under the program. Figure 11. Evolution of coverage of the main 3 CT programs (total households, 2004-2022) 1,600,000 1,400,000 1,200,000 Beneficiary HHs 1,000,000 800,000 600,000 400,000 200,000 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 Alimentate Aprende Avanza Source: ADESS. Avanza was not transferred during 2017-2018 Public transfers have contributed to poverty alleviation in the DR, but there is room for improving adequacy, targeting and efficiency. Recent studies find that most direct transfers in the Dominican Republic are pro-poor.8,9 . Although real wages were stagnant between 2004-13, cash transfers together 6 The ENCFT is the official national continuous survey for labor market and poverty statistics in the Dominican Republic. 7 SIUBEN establishes 4 groups of beneficiaries according to a proxy-means test named �ndice de Calidad de Vida for its Spanish acronym – ICV1, ICV2, ICV3 and ICV 4 groups–, with ICV1 and ICV2 being the groups corresponding to extreme poor and poor beneficiaries respectively. 8 World Bank. 2021a. “Dominican Republic Public Expenditure Review. � World Bank, Washington, DC. 9 World Bank. 2021b. “Poverty and Distributional Impacts of Fiscal Policy in Dominican Republic. �World Bank, Washington, DC. with remittances helped reduce poverty (chapter 1). Cash transfers also played an important role in mitigating the impact of the COVID-pandemic: in 2020, official poverty increased by 2.4 percentage points to 23.4 percent but would have reached 29 percent in the absence of emergency cash transfers.10 However, the DR lags its LAC comparators in terms of adequacy and incidence of the benefits of social assistance among the poorest households. As a result, the social protection system reduces the poverty headcount by only 10 percentage points, 24 percent of the poverty reduction for LAC countries and 35 percent for UMC countries (Figure 12). Figure 12. Coverage, adequacy, targeting and efficiency of social protection programs for selected group of countries and the Dominican Republic (2019)11 80 60 Percent 40 20 0 Adequacy of benefits of all Benefits incidence of all Coverage of all social Poverty Headcount reduction social assistance (bottom 20) social assistance (bottom 20) assistance (bottom 20) (%) - All Social Protection and Labor (bottom 20) DR Latin America and the Caribean Upper Middle Income World Source: World Bank – ASPIRE database. Note: The indicators for LAC, UMC and the World are calculated for the average (2010-2019) 3.2.2 Effects of CCT Programs on Education Outcomes By reviewing the existing evidence and conducting new analytical work, the next section asks whether the conditional cash transfer programs from PROSOLI on food and education (Comer es Primero, Incentivo a la Asistencia Escolar and Bono Escolar Estudiando Progreso) improved schooling outcomes in the Dominican Republic. Available data allow us to analyze the effects of SUPERATE’s high school conditional cash transfer on educational attainment. SIUBEN datasets provide information on beneficiary households before inclusion in the program and several years after they started receiving transfers. We exploit variation in Avanza coverage due to financial and administrative constraints of the program to estimate its impact on high school graduation and years of completed education. Our main estimation focuses on children in the 18- 10 COVID bajo la lupa – Efectos de la COVID-19 en la pobreza monetaria, desigualdad y mercado de trabajo. MEPYD 2020. 11 Benefit incidence: Percentage of benefits going to each group/quintile of welfare distribution relative to the total benefits going to the population. The indicator includes both direct and indirect beneficiaries. Adequacy of benefits: The total transfer amount received by all beneficiaries in a quintile as a share of the total welfare of beneficiaries in that quintile. Coverage: Percentage of population participating in social protection and labor programs (includes direct and indirect beneficiaries). Poverty headcount reduction: Simulated change (%) on poverty headcount due to SPL programs. Indicators for all SPL programs provide the totals summing up the social assistance, social insurance, and labor market figures. 23 years range in SIUBEN 2018, as these children were directly exposed to Avanza while of high school age. We estimate the impact of Avanza by comparing beneficiary children exposed to this transfer, the Avanza+ group, versus beneficiary children whose households did not receive it but were otherwise eligible, the Aprende+ group.12 These groups received the food transfer and were eligible to both school transfers, Aprende and Avanza, however, the second group was not exposed to Avanza due to the restrictions cited above. To address the potential bias induced by nonrandom assignment of this transfer among eligible households, we employ the propensity score matching (PSM) estimator. PSM computes the average treatment effect by matching individuals in the treatment group with their closest counterpart in the control group in terms of observable characteristics at baseline. The advantage of this method in constructing balanced treatment and control samples hinges on reducing the dimensionality of the comparison to just one variable or indicator, the propensity score, which is a function of baseline characteristics. A limitation of PSM is that it is not robust to selection into treatment based on unobservable characteristics. Treatment and controls were comparable at baseline. Indeed, the mean ICV difference before matching between the two groups is 0.07 and 0.16 standard deviations in urban and rural areas respectively. A similar pattern is observed when comparing ICV components separately (see Annex 3), including years as beneficiaries of the food transfer. Therefore, our groups for estimation are relatively well balanced on average to start with. Figure 13 plots the cumulative density function (CDF) of the ICV for these two groups and for nonbeneficiaries, separately for urban and rural areas. While the CDFs for the Avanza+ and Aprende+ groups almost overlap, the CDF for nonbeneficiaries is shifted to the right, indicating that the two beneficiary groups had lower living standards at the beginning of the program, in line with the program targeting scheme. However, Avanza households have received US$ 364 more transfers on average than Aprende households, precisely, the difference we exploit in this exercise (see tables in Annex 3). Figure 13. Cumulative density function distribution of the ICV for Avanza+, Aprende+ groups and for nonbeneficiaries Source: Author’s calculations based on SIUBEN. 12 As for other transfers, the lack of coverage of AVANZA is mainly explained by SUPERATE budget constraints. However, in the case of school transfers, it is also explained by administrative constraints given the incompleteness of national enrollment records, SUPERATE tries to close this gap through schools' visits, however, the program is not able to confirm enrollment status of all eligible students. To estimate the propensity score we use all variables underlying the index for the proxy means test for program targeting (the ICV).13 Given differences in living conditions across urban and rural areas, the program used different ICV models to target each zone. Therefore, in our estimations we first split the sample by area of residence. In a second exercise, we split the sample by gender to analyze whether there is evidence of differential impacts between boys and girls. To further control for potential bias induced by differences in high school enrollment decisions between households in our two main groups, we only include children in SIUBEN 2018 that had completed at least one year of high school or were enrolled in high school in this dataset. 1415 A first step assessment indicates a suitable PSM design. The first condition that must be met in a PSM design is that for nearly every observation in the treated group there should be a counterpart in the control group with a similar level of the propensity score. In other words, there should be a significant overlap between the propensity score distributions of treatment and control groups. Figures 14 and 15 plot the propensity score distribution for households that received Avanza (and Aprende), the Avanza+ group, and for households that receive APRENDE but did not receive Avanza, the Aprende+ group. In both urban and rural areas, there is a high degree of overlap. Figure 14. Poverty changes decomposition (2013- Figure 15. Figure Extreme poverty changes 2021) decomposition (2013-2021) Source: Author’s calculations based on ECNFT 13 In our estimations we also include province and household headship and year of birth fixed effects. 14 Taking advantage of PROSOLI’s targeting scheme and the ICV eligibility thresholds we also explored the possibility of a Regression Discontinuity Design to estimate school transfers effects around the eligibility cutoff. However, the exploration analysis suggests the identifying assumptions required to apply this method are not met. Specifically, statistical evidence indicates the continuity assumption of the running variable around the cutoff, the ICV in this case, does not hold. 15 The creation of Avanza could have influenced households’ high school enrollm ent decision, specially, for those whose children were not still enrolled in this level by the start of Avanza. By focusing on children with at least one year of completed high school in SIUBEN 2018, we address the potential bias from the self-selection of families and students into secondary education. The PSM estimates indicate that receiving additional transfers during high school has economically important effects on secondary school graduation rates and years of completed education. Recall that we are comparing children in households receiving Avanza transfers against children in Avanza eligible households with similar sociodemographic characteristics at the start of the program that did not receive Avanza transfers. All results are statistically significant at standard significance levels. Receiving Avanza transfers increases the probability of completing high school by 13.6 percentage points. Comparing this treatment effect size against the high school graduation rate in the control group (59.1 percent) implies a relative increase of 23 percent. These results are important given the low secondary school completion rate among the most vulnerable groups and given the high labor market returns to completing this level of education as discussed in detail in section 1. Estimates of Avanza’s effect on high school completion are similar for males and females and are slightly higher in rural areas (14.7 percent) than in urban areas (12.6 percent; figure 15). Gender-specific treatment effects are also similar between rural/urban areas (figure 16). Figure 16. Figure ATT Effects of AVANZA+ on Figure 17. ATT Effect of AVANZA+ on High School High School Graduation Graduation 11.6 Urban 12.6 Male 14.4 11.5 Rural 14.7 Female 14.8 Urban Rural Source: Authors calculations Exposure to Avanza also translates into 0.52 additional years of completed education. Moreover, and in line with previous results, the impact is higher for rural areas versus urban areas: 0.58 years (about 7 months) versus 0.47 years (about 5 and a half months). After splitting the estimation sample by gender, we can see the different treatment effects across areas are mainly driven by the difference in female students (figure 19). Figure 18. ATT Effects of AVANZA+ on Years of Figure 19 . ATT Effects on Years of Education Schooling Urban 0.46 0.47 Male 0.53 0.39 Rural 0.58 Female 0.55 Urban Rural Source: Authors calculations Younger siblings in households that earlier received this transfer but subsequently lost access to it experience much more modest improvement in schooling outcomes. We further investigate whether children in the Avanza group who enrolled in high school during the interruption period of the transfer, mid-2016 through early 2019, completed more years of education compared to their closest high school peers in terms of baseline characteristics in the Aprende group. In this case, the estimated treatment effect in rural areas is only 0.047 years and is not statistically significant, whereas in urban areas the treatment effect is statistically significant with a magnitude of 0.097, much lower than the nearly half a year of education from our previous comparison. In other words, transfers conditional on high school attendance that were anticipated by households but never paid ex-post have a much smaller impact on educational outcomes than the same conditional transfers that were ultimately paid. Evidently, then, the Avanza transfer has its intended effect of encouraging schooling, but only if it is paid and not just promised. The analysis above builds upon the earlier work of Hernandez et al (2022) described in Box 2. The key advantage of the present set of results lies in the use of SIUBEN 2018, for which households can be traced back directly to baseline data from SIUBEN 2004.16 Despite this substantial improvement and somewhat reassuringly, both analyses yield similar quantitative findings. Box 2. Further Evidence of the Effects of Supérate School Transfers on High School Completion In 2013, Supérate incorporated Avanza (then called BEEP), a second school transfer financed by the Ministry of Education and targeted at high school students in beneficiary households. Avanza implied an increase in school transfers of up to 3.3 times the amount received relative to Aprende (See Annex 1) (then called ILAE), the basic school transfer (which started in 2005 and was conceived for all eligible households with school-aged children and then targeted exclusively at primary school students after the creation of BEEP). In a recent work (Hernandez et al, 2022) the authors take advantage of the partial coverage of Avanza (due to restrictions in the Ministry of Education school enrollment verification process and Supérate budget constraints) to study the effects on high school completion of receiving additional school transfers at this level. The database gathered for this study combines three administrative datasets: the Ministry of Education’s National Exams database, the historic payroll data of Supérate provided by the Social Subsidies Administrator (ADESS), and 16 Aside from avoiding the need for a household matching algorithm between baseline and outcomes database, the use of SIUBEN 2018 provides schooling outcomes measured 13-14 years after baseline instead of only 10-11 years after; this allows us to estimate impacts on years of completed education directly. We can also assess robustness by limiting our analysis to students with at least some high school, thereby precluding potential selection bias from high school enrollment decisions of households. two waves of SIUBEN’s census data. SIUBEN’s first wave was collected mainly between 2004 and 2008 and contains information on both beneficiary and non-beneficiary households before the former were included in the program. The estimation sample includes beneficiary students who took the mandatory tests to graduate from primary education between 2010 and 2013, as they could have been exposed to the additional transfers implied by Avanza for at least one year of their high school education.17 Furthermore, the study compares the incidence of high school completion between students whose households receive Avanza transfers versus students whose households receive Aprende transfers only, but who otherwise share similar program participation probabilities at baseline.18 Results suggest that receiving Avanza is, on average, associated with an 11.7 to 13.2 percentage-point higher probability of completing high school relative to not receiving these additional transfers. Considering average graduation rates among the matched control groups, the estimated effects represent an increase in the high school graduation rate of 23 to 25.3 percent compared to the counterfactual (i.e., with no Avanza transfers). The authors do not find evidence of differential impacts across gender nor area of residence. In addition, the transfers seem to play an especially key role during the last high school year of targeted students. Effect of 10 additional US$ in high school transfers on delaying age for having first child (in months) 0.35 Relative to incomplete secondary education 0.4 0.6 Relative to completed primary education 0.6 Rural Areas Urban Areas Source: authors' calculations. Back-of-the-envelope cost-benefit calculations indicate that impacts on high school completion from school transfers translate into non-trivial effects later monthly salaries, employment prospects and time delay for having a first child. This analysis is relevant in a context where the returns to education are not small and considering the DR has one of the highest teen pregnancy rates in the LAC region. For example, the authors combine the estimated effects of additional cash transfers on high school completion with estimates of delay in having a first child by education level obtained from the National Demographic and Health Survey, ENDESA (2013). Doing so indicates that 10 additional US dollars per year transferred to the group of girls that received both school transfers translate into a 0.6-month delay in having their first child, both in urban and rural areas, relative to those who only completed primary education, and around 0.4 months relative to those with an incomplete secondary education (see figure above). Similarly, the authors calculate 10 additional US dollars in high school transfers per year increases employment probability by 1.3-1.5% in urban areas and 1.2-1.3% in rural areas, and monthly wages by US$ 1.2-1.3 in urban areas and US$ 1.3 in rural areas, compared to completed primary education only. Source: Hernandez et al (2022) 3.2.3 Imputed Effects of CCT Programs on Employment and Earnings 17 Avanza (previously BEEP) was suddenly discontinued in July 2016 by the Ministry of Education and restarted in early 2019, this time financed directly by Supérate with modifications in the transferred amounts. 18 Households of students in the first group also received Aprende (previously ILAE) during their childrens’ primary education years. In addition, the Aprende-only group were still eligible to continue receiving this transfer during high school if they were not covered by Avanza. To translate the educational impacts of Avanza transfers into labor market outcomes, we turn to data from a recent cross-sectional labor force survey (ECNFT 2018). In Box 3, we describe estimation of the full economic returns to high school completion. High school completion increases the probability of being employed for young women, but not for young men. According to a multinomial model (see Box 2, equation (ii) and Figure A2.2) the completion of secondary education is associated with a 17 percent increase in the likelihood of employment for women, but not for men. A completed high school education is strongly correlated with an increase (32 percent) in the probability of being employed in the formal sector when compared to individuals without such education (figure 10). The provision of conditional cash transfers would increase the labor income of those exposed to the program. The previous analysis indicates that the 13.6 percentage point increase in high school completion associated with the programs translates into an increase in expected monthly total labor income by 2.5 percent (see figure 21) and in hourly labor income by 1.5 percent. These results are driven by higher employment in the formal sector (see annex 2) coupled with the wage premium associated with completed secondary education. Figure 20. Increase in the probability of being Figure 21. Growth of expected earnings due to employed and formality due to CCT CCT 8% 9.35 9.01 0.35 0.321*** 0.322*** 0.328*** 8.58 Dominican Pesos (Thousands) 0.299*** 9 0.30 9.23 7.83 6% 8.78 0.25 8.37 Growth (%) 7 Marginal effect 0.20 0.168*** 7.53 4% 0.15 4.0% 0.10 5 2% 2.5% 2.6% 0.05 0.021 0.018 0.00 1.4% 0% 3 -0.05 -0.008 Total Male Female Urban Total Male Female Urban Growth Baseline Simulated Being employed in the formal sector Being employed Source: Authors calculation based on the ECNFT. Note: See the methodologic details in Box 1 and Annex 1. Note: The results show the marginal effects after a multinomial logit (see Box 1 (ii)). The estimation used specific samples to desegregate the results by the shown characteristics. Note: Results are estimated for the population exposed to the program. Box 3. Estimating the expected earnings after the exposure to the Avanza program Using a sample from the ECNFT 2018 of individuals aged 22 to 34 years old (exposed to the program) living with their parents,19 the analysis followed 5 steps: (i) Estimation of Mincerian equations for formal and informal workers to predict the wage premium of finishing secondary or more education on wages (see results in Annex 1). (1) 𝐿𝑛𝑌𝑖𝑓,𝑖𝑛 = 𝛽0 + 𝛽1 𝜕𝑖 + 𝛽2 𝑋𝑖 + 𝑢𝑖 19 The selection of the sample was determined to match the characteristics of the selected group for the ATT estimation. The analysis also required to control for retrospective household characteristics of the individuals exposed to the Avanza program, in that sense we restricted the ECNFT sample only for those individuals in the interest age group still living with their parents to control for the education of the latest. where, 𝐿𝑛𝑌𝑖𝑓,𝑖𝑛 = Logarithm of monthly earnings of individual i for both formal (f) and informal workers (in) 𝜕𝑖 = Dummy variable of completed education level, including HS (high school completion). 𝑋𝑖 = age, age squared, sex, urban areas, region, and education level of the parents. 𝑢𝑖 = error term (ii) A multinomial logit was used to predict the probability of formal employment, informal employment, or unemployment (base outcome). As follow: 1 Pr(𝑦 = 𝑢𝑛𝑒𝑚�𝑙𝑜𝑦𝑒𝑑) = 𝑓𝑜𝑟𝑚𝑎𝑙 𝑖𝑛𝑓𝑜𝑟𝑚𝑎𝑙 1+ 𝑒 𝑋𝛽 + 𝑒 𝑋𝛽 𝑓𝑜𝑟𝑚𝑎𝑙 𝑒 𝑋𝛽 Pr(𝑦 = 𝑓𝑜𝑟𝑚𝑎𝑙) = 𝑓𝑜𝑟𝑚𝑎𝑙 𝑖𝑛𝑓𝑜𝑟𝑚𝑎𝑙 1 + 𝑒 𝑋𝛽 + 𝑒 𝑋𝛽 𝑖𝑛𝑓𝑜𝑟𝑚𝑎𝑙 𝑒 𝑋𝛽 Pr(𝑦 = 𝑖𝑛𝑓𝑜𝑟𝑚𝑎𝑙) = 𝑓𝑜𝑟𝑚𝑎𝑙 𝑖𝑛𝑓𝑜𝑟𝑚𝑎𝑙 , 1+𝑒 𝑋𝛽 +𝑒 𝑋𝛽 where 𝑋 𝑎𝑛𝑑 𝛽𝑘 are vectors of covariates and coefficients respectively. The covariates include sex, age, age squared, area of residence, region dummy, education level and parents’ education level. (iii) Prediction of the expected earnings (EE): Using the predicted earnings for formal 𝐸𝑓 and informal workers 𝐸𝑖 and the probability of being formal 𝑃𝑓 or informal 𝑃𝑖 workers given their observable characteristics to compute: (3) 𝐸𝐸 = 𝑃𝑓 ∗ 𝐸𝑓 + 𝑃𝑖 ∗ 𝐸𝑖 (iv) Using the treatment effects (TE) of AVANZA for both urban and rural areas estimated in the previous section we construct counterfactual high school completion rates (HS’). HS’ = HS + TE (v) Replace HS by counterfactual HS’ in all the prediction equations and compute the corresponding expected earnings: (4) 𝐸𝐸′ = 𝑃𝑓 ′ ∗ 𝐸𝑓 ′ + 𝑃𝑖 ′ ∗ 𝐸𝑖 ′ (vi) The difference 𝐸𝐸 ′ − 𝐸𝐸 represents the expected earnings gain attributable to AVANZA. 3.2.4 Effects of Job Training Programs on Employment and Wages The last section showed how conditional cash transfers may have reduced the intergenerational transmission of poverty by encouraging investments on education through the fulfillment of the co- responsibilities associated with the transfers. But high school completion alone may not be sufficient to escape poverty. As section 2 noted, the current government has emphasized capacity building for employability as part of its renewed strategy against poverty. The strategy continues to make use of conditional cash transfers but augments these incentives with offers of technical vocational training and apprenticeships as a route to better-paid jobs and entrepreneurship. Yet, a large empirical literature shows that the returns to active labor market interventions are often low. Despite this disappointing evidence in other contexts, a recent randomized-control trial study shows that a scaled-up program of vocational training for youths of SUPERATE beneficiary households had substantial positive effects on earnings and transitions into formal employment (box 4). Notably, this intervention directly targeted poor youth. Box 4. The Effects of Prosoli (now Superate) Capacity Building Program on Labor General Objective In 2015 the federal government of the Dominican Republic launched the initiative Progresando Unidos (ProUni) to strengthen the federal social security framework and provide additional benefits to the families of the beneficiaries of the national conditional cash transfer (CCT), Progresando con Solidaridad (ProSoli). One of the programs that are part of ProUni provides vocational training to children of beneficiaries aged 18-29. The training consists of a 40-day vocational course in one of 11 different trades (such as hairdressing, or commercial vehicle driving) plus a two-month internship at a local firm in the sector. The program aims to increase labor opportunities for youth given the high unemployment rate for this age group. A total of 30,000 youth received these interventions over the course of four years. Evaluation Design An impact evaluation of the program was conducted at the end of 2019 using an oversubscription design. In particular, field teams recruited up to 30 students per course with an enrollment capacity of only 20 to 25 students, selected at random. The remaining individuals were used as controls. In total, the study tracks 6,224 youth in more than 200 courses divided into 2 groups: the control group, that did not receive any intervention, and the treatment group, who would attend the course and obtain an internship in a local firm for two months. Data The study uses 4 rounds of student survey data collection: skills tests (during and after the intervention), a first follow-up (around 6 months after the intervention), a second follow-up (around a year and a half after the intervention) and an endline (more than two years after the intervention). Using youth with at least one round of follow up data, yields an estimation sample of 2,311 students. Methodology The analysis uses a panel dataset composed on all responses from the skills tests, the two follow up surveys and endline data. The sample of treated students is not balanced with respect to its peers in the control group mostly because only treated students who actually signed up in the course were followed, while the ones who rejected the course were not part of the sample with data. Thus, we perform propensity score matching regressions on baseline observable variables (education, age, gender, having or not children, province, baseline salary level and type of vocational course level). This matching yields a comparable sample between the two groups and also provides regression results with adjusted SE for estimated propensity score matching. Main Results As expected because of our selected data, most of the treated students in the sample (92 percent) confirm having done the vocational course. However, using all of our available data, we observe no effect of the program in increasing labor opportunities for those in the treatment arms. Indeed, both control and treatment have an employment rate of around 47 percent, with only around 15 percent contributing to social security. Even so, treatments do display higher quality jobs: they earn a higher income from working (on average, USD 43 more every month, an increase of 11 percent with respect to their peers in the control group) and in the long term, this increase becomes even more noticeable and more of them pay TSS taxes, which indicates a more formal job. Additionally, youth who have done the training display higher job searching skill. The evidence thus indicates that the program led to better jobs and preparation for the labor market, if not a higher probability of employment. 3.3 CONCLUSIONS AND POLICY IMPLICATIONS We have documented persistent human capital deficits in the DR, both relative to its LAC counterparts as well as between groups within the country, especially by poverty status and gender. Only half of youth in the DR complete high school, despite a high estimated return to secondary education in the labor market. Taken together, these facts suggest that individuals face financial barriers to human capital investment. At the macro-level, the Jobs Diagnostic (World Bank, 2021a) reports that real wages in the DR have stagnated, attributing this lack of trend to, among other factors, skill-biased technical change. In other words, Dominican workers may be losing the “race between education and technology� (Goldin and Katz, 2008). Two supply-side policies embedded in the DR’s principal Social Protection program could help future workers win this race. These policies are designed to alleviate financial constraints on human capital investment for adolescents and young adults from low-income households. The first policy is a conditional cash transfer for high school enrollment. Our empirical analysis of its causal impact reveals a substantial gain in completed years of education and a higher rate of high school completion for participants. The second policy provides vocational training, again targeted to poorer households, and there is strong evidence that this too leads to wage gains and better jobs for its beneficiaries. One caveat to these promising findings is that, in estimating the returns from these interventions, there is no accounting for general equilibrium effects that would accompany scale up of such supply-side policies. In short, the income gains presented here may be overly optimistic were these interventions to be implemented writ large. That said, interventions on the supply side of the labor market should not be considered in isolation, but rather as complements to demand-side policies such as encouraging FDI and liberalizing markets. Finally, efforts to encourage high school completion in and of themselves will not be effective in the DR without serious reforms to the educational system that lead to improvements in the quality of education.20 4 REFERENCES COVID bajo la lupa – Efectos de la COVID-19 en la pobreza monetaria, desigualdad y mercado de trabajo. MEPYD 2020. Goldin, C., & Katz, L. F. (2010). The race between education and technology. Harvard University Press. OECD (2019). PISA 2018 Results (Volume I): What Students Know and Can Do. OECD publishing. Winkler, H. and M. Montenegro. 2021a. “Dominican Republic Jobs Diagnostic.� World Bank, Washington, DC. World Bank. 2021b. Poverty and Distributional Impacts of Fiscal Policy in Dominican Republic.� World Bank, Washington, DC. 20 https://documents.worldbank.org/en/publication/documents- reports/documentdetail/933991624785121641/dominican-republic-learning-poverty-brief-2021 5 ANNEXES Table A1. Cash Transfer Components and Amounts for SUPERATE (former PROSOLI), 2022 Actual Actual Transfer Level transfer transfer Frequency Conditionalities RD$ US$ Health check-ups for Aliméntate (former Comer es 1,650 30.0 Monthly pregnant women and Primero CEP) kids 0-5 ILAE (Incentivo a la Asistencia Elementary 300 5.45 Bimonthly Escolar) HS (1st gra) 400 7.27 Bimonthly HS (2nd gra) 500 9.09 Bimonthly HS (3rd gra) 600 10.91 Bimonthly School enrolled and HS (4th gra) 800 14.55 Bimonthly assistance >= 80 BEEP (Bono Escolar Estudiando percent Progreso) HS (5th gra) 1,000 18.18 Bimonthly HS (6th gra) 1,200 21.82 Bimonthly HS - Technical 1,400 25.45 Bimonthly (all grades) IES (Incentivo Educacion Superior) 500 9.09 Monthly Enrolled in college Notes: 56.7 Exchange rate (2022 avg) = 55 RD/USD ANNEX 2. RESULTS IN LABOR OUTCOMES AND SALARIES OF THE AVANZA PROGRAM Results of a Mincerian equation shows that the completion of high school among population aged 22 to 34 years old (exposed to the Avanza program) are associated with a monthly and hourly wage premium of around 20 percent. The monthly wage premium could be higher for those in the formal sector (24 percent) compared to those in the informal sector (8 percent). See figure A2.1. Figure A2.1. Wage premium for education for youth aged (22-34) years old. Monthly labor income Hourly labor income Informal Formal Informal Total Formal workers Total workers workers workers (1) (2) (3) (4) (5) (6) Ln labor Ln labor Ln labor Ln hourly labor Ln hourly labor Ln hourly VARIABLES income income income income income labor income Complete primary 0.142** -0.0753 0.0633 0.159*** -0.0826 0.0799 (0.0622) (0.0768) (0.0490) (0.0613) (0.0739) (0.0498) Incomplete secondary 0.0652 0.0732 0.0614* 0.0193 0.0901* 0.0267 (0.0484) (0.0495) (0.0372) (0.0520) (0.0479) (0.0392) Complete secondary or more 0.0767* 0.242*** 0.196*** 0.160*** 0.282*** 0.209*** (0.0446) (0.0460) (0.0339) (0.0432) (0.0424) (0.0324) Constant 6.446*** 8.176*** 7.313*** 1.030 2.604*** 2.006*** (1.069) (0.930) (0.771) (1.088) (0.986) (0.766) Observations 1,681 1,805 3,486 1,671 1,753 3,424 R-squared 0.135 0.251 0.173 0.119 0.256 0.170 Controls YES YES YES YES YES YES Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: Controls include experience, experience squared, gender, urban area, region, and education level of parents. Source: ECNFT 2018. The specification can be found in Box 1 equation (i). The selected sample refers to population aged 18-24 years old living with their parents to control for parents’ education. The result of a multinomial probit shows also shows the completion of high school among population aged 22 to 34 years old (exposed to the Avanza program) are associated with an increase in around 30 percent of being employed in the formal sector. No significant results are shown in increases on employability for males, but women are 16 percent more likely to be employed when the high school degree is completed. Table A2.2. Increases in the probability on employment and formalization associated with the completion of high school decree - Marginal effects after multinomial probit for selected samples. (Population aged 22-34 years old) Marginal effect Std Err. Total Formal sector 0.3207*** 0.024864 Employed 0.0205098 0.022093 Male Formal sector 0.3219*** 0.025944 Employed 0.0077898 0.017091 Female Formal sector 0.299*** 0.085446 Employed 0.168*** 0.086049 Urban Formal sector 0.3275*** 0.030187 Employed 0.0181 0.025749 Note: The specification of the multinomial model could be found in box 1. The expected earnings of workers are predicted using the results of the multinomial probit and the mincerian wage equation, which consider the workers' observable characteristics (see box 2 for methodological details). These predictions are then compared to a scenario where the effects of the AVANZA program on high school graduation are used to estimate the impact on expected earnings. As shown in Figure A2.3, the increase in employability and formalization, resulting from improved human capital formation, is estimated to increase expected earnings by an average of 2.5 percent per month. Table A2.3. Expected earnings at baseline and simulated (completion of high school) for population exposed to the ILAE and BEEP programs (aged 22-34 years old) Mean of expected earnings Baseline Simulated diff percent diff Hourly income (Dominican pesos) 48.5 49.2 0.7 1.5 percent Monthly income (Dominican pesos) 8365 8574 209.0 2.5 percent *Mean expressed in DOM pesos after estimated unbiased predictions. ANNEX 3. SUMMARY STATISTICS AT BASELINE OF GROUPS FOR ESTIMATION OF SCHOOL TRANSFERS EFFECTS ON SCHOOL OUTCOMES BY AREA OF RESIDENCE URBAN AREAS Aprende+ Avanza+ Variable Mean Mean Baseline ICV 53.28 53.93 Prop. members < 6 years 1.50 1.13 Prop. attending school 2.23 2.17 HH schooling level 0.94 0.98 HH head schooling level 0.82 0.93 Prop. of working members 1.28 1.19 Dwelling type 1.85 1.85 Public water source 1.44 1.48 Toilet access 1.91 1.95 Cooking fuel type 0.91 0.93 Garbage disposal method 1.69 1.71 HH head gender 0.25 0.21 Floor type 0.94 0.95 Wall type 2.41 2.46 Roof type 1.12 1.14 Lightning source 0.99 0.99 Appliances ownership 1.27 1.33 Overcrowding 2.07 1.87 Years as beneficiaries 11.34 11.22 3,890.27 4,264.23 Total amount transferred (US$, up to 2018) Observations 120,546 51,521 Source: SIUBEN 2004. Notes: For ICV calculation, original variables were transformed to indexes. Variables are measured at the household level. RURAL AREAS Aprende+ Avanza+ Variable Mean Mean Baseline ICV 44.33 46.07 Prop. members < 6 years 1.60 1.24 Prop. attending school 2.24 2.17 HH schooling level 0.88 0.92 HH head schooling level 0.68 0.78 Prop. of working members 1.35 1.22 Dwelling type 1.95 1.94 Public water source 0.84 0.87 Toilet access 1.62 1.61 Cooking fuel type 0.64 0.70 Garbage disposal method 1.03 1.07 HH head gender 0.39 0.34 Floor type 0.83 0.86 Wall type 1.97 2.01 Roof type 1.01 1.02 Lightning source 0.82 0.86 Appliances ownership 0.93 1.00 Overcrowding 2.20 1.99 Years as beneficiaries 11.38 11.51 3,925.55 4,279.90 Total amount transferred (US$, up to 2018) Observations 74,416 38,407 Source: SIUBEN 2004. Notes: For ICV calculation, original variables were transformed to indexes. Variables are measured at the household level. ANNEX 4. ATT ESTIMATES OF SCHOOL TRANSFERS ON HIGH SCHOOL GRADUATION AND YEARS OF EDCUATION ON MAIN ESTIMATION SAMPLE (AGE GROUP 18-23) High School Completion Years of Education ATT Avanza Explosare Urban Rural Urban Rural Coef. 0.1258 0.1468 0.4635 0.5804 s.e. 0.0090 0.0120 0.0312 0.0398 Obs. T=0 6,204 3,555 6,204 3,555 Obs. T=1 43,896 24,420 43,896 24,420 Avg T = 0 0.606 0.577 11.329 11.154 Note: Coefficients correspond to the pscore matching average treatment on the treated (ATT) estimates of exposure to additional schools transfers during high school. The dependent variable in columns 2 and 3 is a dummy variable that indicates if the student graduated from high school and the number of completed years of education in columns 4 and 5. T=1 identifies the Avanza+ (treatment) group and T=0 the Aprende+ (control) group. The estimations include all variables used to construct the quality-of-life index (ICV) plus province and year of birth fixed effects. We implement the pscore matching estimator the teffects command in Stata (caliper of 0.01 and one match only). We use the linearized pscore and we dropped observations below the first percentile of the pscore distribution for our treatment group and above the 99th percentile of the distribution for our comparison group. We use a logit specification for pscore estimation. The standard errors (s.e.) reported are heteroskedastic robust. Obs. = number of observations. Avg. = average value of dependent variable. All coefficients are significant at the 1% level. High School Completion Years of Education ATT Avanza Exposure Urban Rural Urban Rural Coef. 0.1161 0.1442 0.4635 0.5305 s.e. 0.0377 0.0165 0.0312 0.0505 A. Males Obs. T=0 3,555 2,159 6,204 2,159 Obs. T=1 24,458 14,151 43,896 14,151 Mean T = 0 10.937 0.493 11.329 10.757 Coef. 0.1145 0.1481 0.3854 0.5530 s.e. 0.0128 0.0175 0.0492 0.0609 B. Females Obs. T=2 2,623 1,379 2,623 1,379 Obs. T=3 19,400 10,315 19,400 10,315 Mean T = 0 0.712 0.700 11.857 11.739 Note: Coefficients correspond to the pscore matching average treatment on the treated (ATT) estimates of exposure to additional schools transfers during high school. The dependent variable in columns 2 and 3 is a dummy variable that indicates if the student graduated from high school and the number of completed years of education in columns 4 and 5. T=1 identifies the Avanza+ (treatment) group and T=0 the Aprende+ (control) group. The estimations include all variables used to construct the quality-of-life index (ICV) plus province and year of birth fixed effects. We implement the pscore matching estimator the teffects command in Stata (caliper of 0.01 and one match only). We use the linearized pscore and we dropped observations below the first percentile of the pscore distribution for our treatment group and above the 99th percentile of the distribution for our comparison group. We use a logit specification for pscore estimation. The standard errors (s.e.) reported are heteroskedastic robust. Obs. = number of observations. Avg. = average value of dependent variable. All coefficients ars significant at the 1% level. ANNEX 5. ATT ESTIMATES OF SCHOOL TRANSFERS ON YEARS OF EDCUATION ON YOUNGER COHORT (AGE GROUP 15-17) Years of Education ATT Avanza Exposure Urban Rural Coef. 0.0974*** 0.0470^ s.e. 0.0251 0.0366 Obs. T=0 2,076 1,113 Obs. T=1 27,130 15,040 Mean T = 0 9.890 9.889 Note: Coefficients correspond to the pscore matching average treatment on the treated (ATT) estimates of exposure to additional schools transfers during high school. The dependent variable is the number of completed years of schooling. T=1 identifies the Avanza+ (treatment) group and T=0 the Aprende+ (control) group. The estimations include all variables used to construct the quality-of-life index (ICV) plus province and year of birth fixed effects. We implement the pscore matching estimator the teffects command in Stata (caliper of 0.01 and one match only). We use the linearized pscore and we dropped observations below the first percentile of the pscore distribution for our treatment group and above the 99th percentile of the distribution for our comparison group. We use a logit specification for pscore estimation. The standard errors (s.e.) reported are heteroskedastic robust. Obs. = number of observations. Avg. = average value of dependent variable. ^ indicates not statistically significant. *** indicates statistically significant at 1% level.