SOCIAL PROTECTION DISCUSSION PAPER No. 2519 | FEBRUARY 2025 Who participates in defined contribution pension systems when informality is high? Evidence from 20 years of administrative records from the Dominican Republic Clement Joubert, Thiago Scot, Isaac Rafael Maríñez, Pietter José Regalado, Tatiana Flores and Nicolas Garces Rodriguez © 2025 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington DC 20433 Telephone: +1 (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 of the data included in this work. 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. 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 Group, 1818 H Street NW, Washington, DC 20433, USA; fax: +1 (202) 522 2625; e-mail: pubrights@worldbank.org. Who participates in defined contribution pension systems when informality is high? Evidence from 20 years of administrative records from the Dominican Republic1 February 2025 1. Introduction In a context of accelerated aging and stagnating social insurance coverage, securing adequate incomes in old age without stymieing growth is becoming a primary concern for virtually all countries around the world. One policy option often considered is the adoption of privatized defined-contribution pension systems. A wave of pension privatizations through the 1990s and early 2000s, inspired by the Chilean example, promised to improve participation incentives, reduce informality, and raise pension levels 2. The Dominican Republic --- the context of this note --- privatized its pension system in 2003, moving to a unified defined contribution system vs. a previously fragmented mesh of several defined benefit schemes. Proponents of pension privatization reforms hoped incentives to contribute would improve in privately managed, fully- funded systems: payroll contributions would be perceived as savings rather than taxes (Corsetti 1 This note was prepared by Clement Joubert, Thiago Scot, Isaac Rafael Maríñez, Pietter José Regalado, Tatiana Flores and Nicolas Garces Rodriguez. We extend our gratitude to Antonio María Giraldi, Roscel Savery, and the Superintendency of Pensions (SIPEN) for their invaluable support throughout the project. We also acknowledge the generous funding provided by The World Bank’s Research Support Budget. 2 Examples of full privatization include Chile (1981), Bolivia (1997), Mexico (1997), El Salvador (1998), Kazakhstan (1998), Nicaragua (2000), Dominican Republic (2003), Nigeria (2004); partial privatization cases include Argentina (1994), Uruguay (1996), Hungary (1998), Poland (1999), Costa Rica (2001), Latvia (2001), Bulgaria (2002), Croatia (1999), Estonia (2002), the Russian Federation (2002), Lithuania (2004), Romania (2004), Slovakia (2005), Macedonia (2006), Ghana (2010) Official Use Only 1994); private pension funds would compete and deliver better returns; and public coffers would be relieved from ever-growing pension liabilities. Designing private pension systems is particularly challenging in high-informality countries. The existence of a large informal economy is often associated with limited savings accumulation in private pension systems, as individuals transition into and out of formal employment and, therefore, can have incomplete contribution histories (Berstein, et al. 2006). Pension privatizations did not prevent pension coverage from dropping throughout Latin America during the 90s and 2000s (Rofman, Apella and Vezza 2015) though the exact impact of the reforms is difficult to identify separately from the other macroeconomic forces at play in the period. Stagnating coverage and the strong link between contributions and benefits resulted in low projected replacement rates (Forteza, et al. 2009) which led many countries to reverse course and strengthen the non-contributory features of the systems (Ortiz, et al. 2018). This note provides descriptive evidence on individual contributions histories under the Dominican Republic private pension system, using administrative records. It asks which types of affiliates in the DR pension system participate continuously in a manner that could lead to adequate accumulated savings and pension benefits. Understanding the behavior of contributions in high-informality settings, and how it might vary between individuals and jobs, is a necessary first step to evaluate the performance of these private pension systems and identify necessary reforms. Long-term detailed information on individuals’ work histories is most often absent in representative household or labor force surveys, which makes administrative records an invaluable source of information on life-cycle and cohort dynamics of participation. This note is part of a collaboration between the World Bank’s development research group (DECRG) and development impact group (DIME) and the Dominican Pension Superintendency (SIPEN), the autonomous agency responsible for supervising the private pension system. The main goal of this collaboration is to leverage the rich administrative data managed by SIPEN to provide insight on the performance of the pension system for policymakers and the public and conduct academic research that can guide pension policy in the DR and beyond. In particular, the administrative records of pension contributions constitute a complete dataset of matched employer-employee data, providing a comprehensive view of formal labor market dynamics and complete formal work history for individuals. The World Bank and SIPEN teams collaborated to extract a representative data sample suitable for data analysis and policy research, described in the data section below. While this study centers on the examination of contribution density patterns disaggregated Official Use Only by gender and birth cohorts, Appella and Zunino (2025) concurrently conducted a complementary analysis using the same dataset, with their research focusing specifically on estimating the determinants of contribution transitions. First, we document how often workers contribute to the pension system over their working years and how that varies across workers’ characteristics. We show that the average contribution density (the share of months contributed over potential months) is approximately 40%, meaning that workers contribute on average four out of every ten months when they could have contributed. Focusing on the 2014-2023 period, we document that close to one third of workers almost never work formally, accumulating very few months of contributions, while 20% have almost complete formal work histories. This variation across workers is correlated with characteristics such as geographical location, job industry, and average wages. Second, we take a longer view of the performance of the Dominican system and investigate whether contribution patterns have changed for individuals born in different decades. This is particularly relevant in the Dominican Republic, since economic growth has risen substantially since the privatization of the pension system in 2003, and labor market conditions are therefore likely to also have changed meaningfully. What we document is that younger cohorts, particularly those entering the labor force in the mid-2000s (born after 1980), are contributing at a much higher rate and receiving higher salaries than previous cohorts. This suggests a relative improvement in the perspective of private pension savings for these workers. Last, we dig deeper into heterogeneity across genders. Among formal workers, we find no gender gaps in contribution densities or in average wages in recent years. But the number of men in the system is much higher than that of women, likely reflecting a higher participation in the labor market. To mitigate these gaps, the DR pension system currently relies solely on survivorship pensions, which leave women vulnerable to the risk of divorce. We discuss policy tools employed elsewhere to improve women’s financial autonomy in old age. Official Use Only 2. Context Prior to the establishment of the Dominican Pension System (SDP) under Law 87-01 in 2001, which officially commenced operations in June 2003, the country managed multiple fragmented defined-benefit pension schemes. These schemes lacked the operational and legal capacity to cover all formal-sector workers, and the larger ones, in terms of active worker participation, were financially unsustainable. By 2000, the existing schemes covered only 69% of formal-sector workers, meaning just 30.9% of the workforce was enrolled in a formal pension scheme (Palacios 2003) compared to 44.8% of the workforce employed in the formal sector during that period 3. In 2000, these pension schemes provided benefits to only 18.1% of the population aged 60 and above, with 97% of beneficiaries receiving the minimum pension. At that time, pension liabilities were significantly underfunded, representing around 55% of the national GDP—reflecting the present value projected of expenses extending to 2075 (Palacios 2003). The implementation of the Dominican Pension System and the reforms enacted under Law 87-01 established, for the first time, a universal and mandatory pension system designed to protect all Dominican citizens and residents. This system operates through three financing regimes: contributory 4 , subsidized contributory (not yet implemented), and subsidized. Contributions are mandatory for dependent workers and voluntary contributions are allowed, though they remain rare in practice 5. The minimum pension (Pension Minima Fondo de Solidariedad) requires 300 monthly contributions or 25 years6. Additionally, the reform introduced a privately managed, capitalized defined-contribution model as the core structure of the system. This model is expected to positively influence the system’s financial sustainability and contribute to the growth of internal savings and productive investment 3 Central Bank of the Dominican Republic, available at:https://www.bancentral.gov.do/a/d/2540-mercado- de-trabajo-enft-con-poblacion-ajustada-por-zona-y-regiones. 4 The Dominican Republic's pension system mandates 10% wage contributions (2.9% from employees, 7.1% from employers), capped at 20 times the minimum contributable wage. Retirement is at age 60 with 30 years of contributions. Benefits are offered as programmed withdrawals or life annuities, based on the account balance at retirement. 5 In the sample we use, only 10\% of workers ever make a voluntary contribution and those are in aggregate less than 5\% of mandatory contributions. 6 Since the regime was implemented in 2003, that is not yet attainable. Official Use Only within the country7. Moreover, each financing regime’s minimum pension is now tied to the nation’s minimum wages, as applicable, ensuring adjustments are neither discretionary nor subject to executive decrees. In the contributory regime, pension adjustments are indexed to inflation or recalculated annually based on actuarial analyses, remaining funds, and age of the person. As of December 2023, the number of SDP contributors – formal, salaried employes actively contributing to pension funds – surpassed 2.1 million, approximately 90% of formal-sector workers. Additionally, the system had almost 5 million active affiliates— individuals who have contributed at least once to their pension fund 8. Relative to some of its Latin American peers, the Dominican Pension System faces affiliation-related challenges. By 2023, affiliates represented 62.7% of the Working-Age Population (Winkler and Montenegro 2021) 9, compared to a regional average of 64% and over 70% in countries like Chile and Mexico. However, between 2015 and 2023, the growth rate of affiliates in relation to the WAP in the Dominican Republic surpassed the regional average, suggesting a possible convergence with the Latin American average in the near future 10. In terms of the real return on pension fund investments, the Dominican Republic consistently performed above the regional average from 2015 to 2023, with the exception of 2019 and 2021. During that period, the Dominican Republic pension fund investments had an average real return of 5.24% and the Latin American real return was 3.38% 11. As the Dominican Pension System reaches its 21st year, understanding the contribution patterns of affiliates and their implications for coverage at retirement age and for the 7 As of June 2024, there is a total of 112.669 billion Dominican pesos on pension fund investment in 154 projects in different sectors through "Fondos de inversiones", with construction and solar energy production the most common. More information is available here. 8 Boletín Estadístico Trimestral No. 85 de la Superintendencia de Pensiones (SIPEN) https://www.sipen.gob. do/index.php/publicaciones/boletines-trimestrales 9 World Banks’ Job Diagnostics for Dominican Republic in 2021 document that 58% of workers in the Dominican Republic are informal, when considering informal workers as those not making pension contributions. Report available at: https://documents1.worldbank.org/curated/en/820141619770918898/pdf/Dominican-Republic-Jobs- Diagnostic.pdf 10 Boletín Estadístico Trimestral No. 85 de la Superintendencia de Pensiones (SIPEN) https://www.sipen.gob. do/index.php/publicaciones/boletines-trimestrales 11 Boletín Estadístico Trimestral No. 85 de la Superintendencia de Pensiones (SIPEN) https://www.sipen.gob. do/index.php/publicaciones/boletines-trimestrales Official Use Only adequacy of pension benefits has become essential. This analysis is particularly relevant as the system anticipates a large wave of retirement requests beginning in 2028. 3. Data and Methodology The main data used for this note are administrative records of individual pension contributions provided by pension funds to SIPEN, the system administrator. All results in this note are generated from a dataset including the complete contribution history of 100,000 anonymous contributors between June 2003 and November 2023, randomly selected from the pool of affiliates as of December 2023, from which only 82652 had contributions history12. As it is structured at transaction-level (i.e. we observe each monthly contribution of a worker in a given firm), this sample allows for the construction of a longitudinal dataset following the entire contribution history of these workers. Additionally, we also observe demographic characteristics of workers (gender and age) as well as attributes of the firms they work on (economic sector, firm type and geographical location). We also observe the monetary amount of mandatory contribution for each observation. Given that contributions for individual accounts are a fixed rate of salaries in any period 13, we can indirectly recover wages and generate a history of real wages, adjusted for inflation. It is worth noting that, different from other countries where the ceiling of contribution is low, in the Dominican Republic the ceiling for contributions is twenty contributable minimum wages, meaning that we can observe the actual wage for almost the entire distribution of workers. In our main analysis of contribution densities, we restrict the sample to approximately 60,000 workers that were 32 - 60 years-old in 2023, since we focus on contribution densities in the period 2013-2023 (the last 10 years of the sample). We also provide on the appendix 12 The choice to work with a random sample of workers was driven both by institutional constraints from SIPEN, which can only share subsamples of data, even when anonymized; as well as the computational constraints of working with the universe of affiliates. We have also reproduced these results for a smaller sample of 10,000 random affiliates and obtained very similar results. 13 Mandatory contribution rates have increased from 5% at the inception of the system in 2003 to 8.4% in 2023. The most recent increase was from 8% to 8.4% in February 2020. As previously discussed, the total contribution is 9.97% of wages, since it also includes other contributions for survivorship pensions and covering administrative costs of the system. Official Use Only additional density distribution during the entire period for individuals of working-age (22-60 years-old) in 2023. For the cohort exercises, we make no sample restrictions. Our main measure of contribution density for each individual is the ratio between the total number of months an individual contributed and the number of "potential months contributed", defined as the total number of months during their active years (ages 22-60): 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀ℎ𝑠𝑠 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑑𝑑𝑖𝑖 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑦𝑦𝑖𝑖 =   𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑠𝑠𝑖𝑖 Note that when we consider contribution densities in the entire sample, the potential contribution of some individuals is capped at the number of months since the introduction of the system in June 2003. For the main analysis, which considers contributions in the 2013- 2023 period, the denominator is simply the 120 potential months of contribution. For some exercises, we group individuals in three main groups of contribution density. Those with low contribution have densities below 25%; those with medium between 25 - 75%; and high-contribution workers are those that contributed in over 75% of potential months. When considering heterogeneity of contribution densities, we consider not only fixed worker characteristics’ such as gender, but also employment related traits, such as sector of economic activity and wages. To attribute these characteristics to workers, we consider the last employment observed in the dataset (i.e. workers defined as belonging to construction sector are those for which their last contribution in the panel is from a firm in that sector). To further understand the determinants of contribution density, we also estimate regression models with the following specification: 𝑌𝑌𝑖𝑖  =  𝛼𝛼  + 𝑋𝑋𝑖𝑖 𝛽𝛽 + 𝜀𝜀𝑖𝑖 Where Yi  is the contribution density of worker i, xi are a series of worker-level characteristics, and εi is an error-term. Our main specification is a simple linear regression, and in the Official Use Only appendix, we also report results from logistical regressions using a dummy variable for high- density as the outcome variable. 4. Who participates continuously The population of working age pension affiliates captured in our data comprises 63% of the overall working age population. Therefore, around a third of workers have never made a pension contribution. As seen in Figure A.5, pension affiliates resemble the working population in terms of employment sectors, except that the public sector is overrepresented in the pension records (22% of pension affiliates instead of 5% of workers overall). Figure A.6 shows that Pension affiliates are also more likely to work in firms with more than 500 workers (63% vs. 50% of workers overall) and all but a few work in firms with more than 20 workers (vs. 85% of workers in the overall population). On average, workers contribute less than half of the potential time they could contribute, if they had a continuous formal participation in labor. The main results we present consider workers’ contribution densities over the previous 10 years. Therefore, we look at the 2013 - 2023 period, after the system turned 10, and at workers with at least 10 years of potential experience and younger than the retirement age (i.e. aged 32 - 60 as of November 2023). We plot the histogram of the contribution densities in Figure 1. The average contribution is approximately 40%, meaning that on average workers contributed only in 4 out of 10 potential months in the last decade. That average also hides a substantial dispersion, as shown by the U-shaped histogram. At one end, around 30% of the workers are mostly inactive, contributing less than 10% of potential months. On the other end, approximately 20% contribute close to 100% of their potential, with close to complete formal work contributions in their active years. The remaining 50% of workers have contributions almost uniformly distributed in the range of 10 - 90% of contribution density, showing that formal labor histories can take many forms for workers in the Dominican Republic14. Frequent transitions in and out of formal employment pose a major challenge to pension systems, particularly fully funded individual account systems like The Dominican Republic, in 14 Results are similar if we consider the entire period 2003-2023. On average, the individuals in the sample contributed approximately 40% of the time. In particular, approximately 25% of the individuals contributed nearly nothing (close to 0%), while approximately 10% maintained contributions consistently, reaching close to 100% of their potential. (see Figure A.1). Official Use Only which continuous asset accumulation is paramount to achieve adequate replacement rates. Irregular contributions also jeopardize a worker’s eligibility for minimum pensions which require a long vesting period (25 years). These findings point to the need for reforms aimed at enhancing contribution continuity and ensuring that more workers can qualify for pensions, minimizing the risk of relying on lump-sum payments. Another avenue to consider for reform would be to reduce the minimum pension vesting period, or introduce a progressive schedule where partial vesting results in partial minimum pension benefits. Figure 1. Histogram of contribution density in the past 10 years (Months contributed/120) Note: 42323 observations, from individuals aged 22-60 in 2023. Data comes from a sample of 100.000 randomly selected contributors from the SIPEN affiliates pool as of 2023. Contribution density is calculated as the ratio of the number of months contributed to the feasible contribution of a worker in 10 years. Source: SIPEN administrative records. Contribution densities vary widely depending on workers’ and employment characteristics. The differences in contribution rates we document previously are systematically correlated with several workers’ and employment characteristics. To illustrate these patterns, in Figure A.2 and Figure A.3 we report the results of the coefficients estimated using Equation 2. We first document systematic differences in contribution densities across economic sectors of employers. Using the retail sector (where approximately 15% of workers are employed) as reference, we show that workers in sectors such as Finance, Healthcare, Official Use Only Public Administration, and Education have contribution densities that are on average 7 - 8 p.p. higher (from a baseline of approximately 40% average contribution density). On the other extreme, workers in the construction sector have densities that are on average 10 p.p. lower, and other sectors such as Mining, Water supply and Arts also register substantially lower contribution densities. Unsurprisingly, high-wage workers also register much higher contributions throughout their working years than lower-wage workers. Using as reference workers that earn (in their last jobs) just above the national average minimum wage, we document that those earning less than half that much have densities that are less than 25 p.p. lower and those just below the MW 10 p.p. lower 15 . Conversely, workers earning between one and three minimum wages contribute 5 - 10 p.p. more, while those above three minimum wages have densities that are approximately 25 p.p. higher. This correlation between wages and contribution densities certainly exacerbates differences in the amount workers accumulate in their private retirement accounts: if high-income workers are also more likely to have uninterrupted work histories, that further contributes to their saving behavior and therefore the benefits they will accrue when retire. It is also worth highlighting that, considering their last contributions, approximately half of workers receive less than the national minimum wage; 40% receive between one and three minimum wages; and only 10% receive above three minimum wages this reinforces the challenges faced by the system, since the majority of workers are making contributions based on low salaries and have incomplete contribution histories. We do not observe large differences in contribution densities according to workers’ age or gender. When considering workers of different age groups in 2023, we observe that younger workers (32-35 years-old) have slightly higher average contribution densities than older workers, but those differences are small in magnitude 16. We also do not document any meaningful differences in average contribution densities by gender – a dimension we explore in more detail below. 15 There are several industry and firm-size specific minimum wages in the Dominican Republic, and here we use the national minimum wage of reference, which is not a floor but simply an average of these specific wages. That explains why so many workers are paid below the minimum wage of reference. 16 Figure A.4 shows that in the raw data this younger group is particularly less likely to have very low contribution densities and more likely to have medium contribution densities (25 - 75% of potential months). The share of younger workers with high contribution densities is actually slightly smaller than older individuals. Official Use Only Finally, we also observe large differences in contribution densities in geographical locations. We show that average contributions densities are systematically different for workers employed in firms from different geographical regions, conditional on other characteristics, in Figure A.3. Taking the largest city Santo Domingo as reference, we show that most other regions have higher average contribution densities, ranging from a few percentage points increase to 11-12 p.p. in regions like El Seybo, Espailat, La Romana and Puerto Plata. What many of these "top performer" regions have in common is that relatively few workers are employed there, and a large share of jobs are concentrated in a few specific sectors of firms. In Espaillat and San Juan de la Maguana, for example, most jobs are in the manufacturing sector, while in La Romana employment is concentrated in one large conglomerate that operates ports, tourism and manufacturing activities. 5. Does participation improve over time Our previous analyses focus on the accumulated history of contributions over the decade but are silent on dynamic changes over time. We show that a minority of individuals of working age in 2023 contribute continuously throughout their life, and that those contributions vary systematically across dimensions such as levels of wage and industries. However, these exercises do not focus on changes that might have occurred in the economy over time. For example, we show that the pattern of contribution density is mostly flat by age groups - two individuals, one aged 35 and another 45 in 2023, have on average the same density of contributions in their work history. But these individuals were born a decade apart and might have faced very different labor market conditions in their work trajectory. In this section, we investigate these patterns by following cohorts of workers born in different years throughout their lifetime. This is particularly important in the Dominican Republic, both because the pension system is relatively new and because the country has registered strong economic growth in the last decades. First, the current pension system was implemented in 2003, and we observe a strong increase in affiliates throughout its first decade of existence. This suggests a transition period in which formal workers might not have been registered in the system, despite having formal jobs, and therefore having lower contributions densities. Second, the Dominican economy performed remarkably in the last two decades: GDP grew by almost 6% a year on average and per capita income grew much faster than the rest of Latin America, Official Use Only converging to the average levels of the region (Word Bank 2023). That suggests that younger cohorts might have faced more favorable labor market conditions, including higher salaries and more stable jobs. In this section, we investigate whether this is the case. The probability of contributing at any age has increased substantially for younger cohorts. In Figure 2a, we present how the probability of having any contributions in a year change over the life cycle, for cohorts of individuals born in five-year intervals between 1950 and 1990 17. These are individuals that were between 23 and 53 years-old when the system entered place in mid-2003. It is important to note that, given that the system has only been in place for 20 years, we are unable to see the full work history of any cohort - the older cohorts, born in the 50s, entered the system in the end of their work life, whereas younger cohorts are still of working-age. With that caveat in mind, we observe a significant increase, for a given age, in the share of workers from younger cohorts making contributions. Consider, for example, workers in their early 30s. Among workers born in 1975, approximately 30% are seen making any contributions in their early 30s (that is, in the mid-2000s). That rate increases to around 35% in the 1980s but increases substantially to around 50% for those born in 1990. That last group reaches 30 years-old precisely at the onset of the Covid-19 pandemic, whose effects can be seen across all cohorts who are still of working age in 2020 (the dip observed in most lines at the right tail). Despite that effect, their contributions are still much higher at that age than previous cohorts. Younger cohorts not only contribute more often, but they also contribute larger amounts given rising real wages. In Figure 2b, we plot similar figures across age groups for different cohorts but now show how average real wages (deflated to 2023 prices) changed. Consistent with other evidence on growing wages in the Dominican Republic in the last decades, we document that younger cohorts are in a much steeper wage growth trajectory than older ones. Cohorts born in 1980, 1985 and 1990 have similar trajectories, with clear but small increases in real wages for those born later. The stark difference is when we compare these cohorts with those born in the 1970s or earlier: over their career, these workers started at lower levels of wages and then had slower wage growth. At age 35, for example, workers born in 1980-1985 had average monthly wages of approximately RD$ 30,000, a level of For this figure, we consider the probability of contribution, not among affiliates, but for the entire 17 Dominican population in each cohort-age. We do that by first expanding the numbers in our representative sample to the universe of affiliates, and then considering the total relevant population in the denominator. Official Use Only average incomes that the 1975 cohort only reached in their early-40s and the 1970 cohort in their mid-40s. Jointly, these patterns suggest that younger cohorts will reach the end of their work-life with higher balances in their personal pension accounts. The fact that workers born in more recent years contribute more often and receive higher salaries on average suggests they must be contributing more to their personal accounts. The total accumulation of balance over time depends not only on contributions, but also on the return of these contributions in the AFPs. The overall consistency of average returns in the last decade suggests that it should be the case that younger cohorts are accumulating larger balances towards their retirement in their private accounts. Figure 2.A: % of individuals contributing over total population from 2003 to 2023 Note: Sample of contributors between 22-60 years. Data comes from a sample of 100.000 randomly selected contributors from the SIPEN affiliates pool as of 2023. Cohorts are defined using birth date of each individual, available at contribution level. The share of contribution is estimated yearly as total members of the cohort contributing over cohort members in that year. Population data comes from the United Nations’ population data portal. Source: SIPEN administrative records. Official Use Only Figure 2.B: Average real wage by age and cohort 2003-2023 Note: Sample of contributors between 22-60 years. Data comes from a sample of 100.000 randomly selected contributors from the SIPEN affiliates pool as of 2023. Cohorts are defined using birth date of each individual, available at contribution level. Wages are indirectly observed by contribution levels and are normalized using IPC data to 2023 RD$. Source: SIPEN administrative records. 6. Are there gender gaps in participation Differences between women and men in the frequency of contributions, wages, retirement, and longevity can compound into large gender pension gaps. To start with, women usually exhibit a lower market labor supply: they work fewer years for pay and work part-time more often. In some countries, women who are in the labor force are also more likely to work informally than men. Lastly, they may also retire earlier to provide informal care for elderly household members. This contribution gap is often compounded by lower wages due to occupational segregation, lower accumulated experience and discrimination. As a result, older women - particularly widows - have historically been among the groups most vulnerable to poverty across countries (Even and Macpherson 1994). The gender dimension is thus a key aspect of pension design, and pension rules can play a key role in mitigating or accentuating the differences in pension benefits between men and women. Comparing men and women among pension participants in the DR reveals only minor gaps. Once disaggregated by gender, the contribution density graphs presented in the previous sections exhibit very similar patterns between men and women. The frequencies at Official Use Only each intensity of pension participation are comparable. To the extent that there is a gap in contribution densities, it favors women (Figure 3). They are 0.4 percentage points less likely to have contributed less than 10% of potential months in the past 10 years, and 2.8 percentage points more likely to have contributed over 90% of potential months. Despite a higher proportion of high-density women, gender becomes insignificant in the regression model of contribution density presented in section 4. By implication, the difference must come from different individual or job characteristics between male and female pension affiliates. One of the most salient differences is the gender composition of the workers in public administration, where women take almost 60% of the jobs in this sector. Female and male contributors also exhibit similar contribution levels, and therefore similar wages. Figure 3: Contribution density by gender 2014-2023 Note: 58875 observations, from individuals aged 22-60 in 2023. Data comes from a sample of 100.000 randomly selected contributors from the SIPEN affiliates pool as of 2023. Sex is identified at contribution level for each individual. Contribution density is calculated as the ratio of the number of months contributed to the feasible contribution of a worker in 10 years. Source: SIPEN administrative records. Despite similar contribution densities among pension affiliates, the fraction of men who contribute to the working age population is 6.7 p.p. higher than women’s because they are less likely to be altogether out of the formal labor force (Figure 4a). Women may be less likely to join the pension system because they stay out the labor force more. Over time, the total number of contributions of men and women has followed parallel upward trends until 2020 (figure 4a). The gap appears to have declined post-COVID, with men’s contribution Official Use Only dropping further than women’s in 2020 and recovering more slowly as well. The gap in 2023 was only 2.6pp. Figure 4.A: Share of contributors over population groups by sex Note: 354868 observations, from individuals aged 22-60 in 2023. Data comes from a sample of 100.000 randomly selected contributors from the SIPEN affiliates pool as of 2023. Sex is identified at contribution level for each individual. Population data comes from the United Nations’ population data portal. Source: SIPEN administrative records. Figure 4.B: % of individuals contributing by age and cohort from 2003 to 2023 at population level Official Use Only Note: 14791 Observations, Data comes from a sample of 100.000 randomly selected contributors from the SIPEN affiliates pool as of 2023. Cohorts are defined using the birth date of each individual, available at contribution level. Sex is identified at contribution level for each individual. Population data comes from the United Nations’ population data portal. Source: SIPEN administrative records. Figure 4b reveals that both women and men in younger cohorts have increased their participation in the pension system (as a percentage of the total population) at similar rate until recently. Therefore, the participation gap has remained around 5-10pp across cohorts until recently. When considering the youngest cohorts of entrants, born in the 1990s, it appears that the gender participation gap has shrunk also in absolute terms, relative to older cohorts, particularly after the pandemic. For these individuals, participation rates have become very close across gender over the past few years, which could portend lower pension gaps in the future. Redistributive features of pension benefit rules such as minimum pensions and pension supplements are a powerful tool to correct gender disparities in labor market outcomes. As women are over-represented at low lifetime pension accumulations, minimum pension guarantees usually benefit them more, partially reducing the gender pensions gap. In the DR, the contributory minimum requires a high vesting period by international standard. It is likely that many women with long career interruptions will not qualify. They will most likely rely on the non-contributory features of the system, which were part of the original architecture of the system but have not been implemented to date. Survivorship pensions were historically the main policy tool to provide pension benefits to homemakers, and they can significantly reduce the pension gap by covering married women with no pension rights. However, survivorship pensions are rights derived from the husband’s pension rights, and women do not receive any benefits in case of divorce. As a result, survivorship pensions can make women dependent on their spouse, reduce their autonomy, and fail to cover an important and vulnerable segment of elderly women. Some pension systems have implemented alternatives that insure women against the risk of divorce. In Germany’s public retirement scheme, pension credits are split between spouses upon divorce just like other household assets (Kreyenfeld, Schmauk and Mika 2023). Others have proposed that spouses should share pension credits accumulated during the marriage ( (Klerby, Larsson and Palmer 2019); (Joubert and Todd 2022)). It is also common for pension systems to credit women with children below a certain age limit with pension contributions Official Use Only with pension rights financed from general revenue, to compensate for earnings penalties associated with children in the literature (Kleven, et al. 2019). Some countries constrain pension fund administrators and insurance companies to use gender-neutral mortality tables, so that women’s annuities or programmed withdrawals are not reduced by their higher longevity. In sum, despite small differences among affiliates and progress with new cohorts, significant gender pension gaps can be expected in the coming decades. Currently, the pension system in DR relies solely on survivorship pensions, which leave women vulnerable to the risk of divorce. Many other policy tools can be drawn from international experience to mitigate the impacts of gender pension gaps. 7. Conclusion Administrative data from the Dominican Republic’s pension system shed light on key challenges and reform opportunities to strengthen coverage and retirement security. This note identifies three critical takeaways: Many affiliates exhibit irregular contribution histories, highlighting the need to improve contribution continuity and re-engage those already linked to the formal system. A significant portion of affiliates contribute intermittently, making them prime candidates for behavioral and administrative interventions. Strategies such as personalized reminders, simplified contribution channels, and targeted incentives or enforcement in high-dropout sectors can help. Evidence from (Akbas, et al. 2016) (Azuara, et al. 2021), and (Gars, et al. 2025) shows that these approaches have an effect on existing savers, rather than impacting the contribution base. Effects of individual interventions are typically modest and short-lived, which could reflect compounding barriers and require more comprehensive, multi-faceted interventions. Additionally, since high contribution density is concentrated among high earners, the current vesting requirement for minimum pensions may exclude the majority. Policymakers could consider reducing vesting thresholds or introducing partial minimum pensions to prevent poverty in old age. Gender gaps in pensions stem from lower participation rates, not lower contribution intensity once affiliated. Women are less likely to enter formal employment—likely due to home production roles—but contribute at similar rates and wages once in the system. This makes promoting early labor market entry for women essential. Also, redistributive measures Official Use Only such as minimum pensions may not disproportionately benefit women in the Dominican Republic, unlike in countries where women dominate the low-earning segments (e.g., (Joubert and Todd 2024)). Survivorship pensions remain the main support for homemakers, but they leave women vulnerable in cases of divorce or non-marriage. Expanding protection beyond marital dependency is critical. International models offer alternatives: crediting caregiving periods, using gender-neutral mortality tables, and ensuring pension sharing upon divorce to enhance women’s pension security and autonomy (Klerby, Larsson and Palmer 2019), (Kreyenfeld, Schmauk and Mika 2023). Younger cohorts are entering the formal labor market in greater numbers, suggesting that formalization is most effective at labor market entry. This trend offers a window of opportunity: strategies that embed formalization incentives and reinforce savings behavior early in a worker’s career are likely to yield long-term gains in coverage and adequacy. Ensuring continued contributions over the life course—through early engagement, trust-building, and institutional continuity—will be critical in translating initial formal entry into pension security. The findings in this report open up several avenues for additional research on the Dominican pension system. First, the finding of low average contribution densities, showing that many workers spend a substantial amount of time outside the formal labor markets, calls for additional research on the determinants of entry, permanence, and turnover in formal jobs, as well as on the affiliates' understanding of the incentives in the defined-contribution pension system. Understanding the circumstances of formal labor market exits could be crucial: the willingness to return to formal employment or to make voluntary contributions could be different if low-density affiliates are out-of-work, in informal self-employment, or in informal dependent relationships. Second, the introduction of the system in 2003 included a provision for "late-joiners": any individual aged 45 or above at the moment of their first contribution would be considered a late-joiner and could access the totality of their savings at the moment of retiring. The design of the program, with a sharp discontinuity defined by age, provides a good setting to compare outcomes of similar individuals on each side of the cutoff to understand the impacts of the lump-sum payment on behaviors such as employment during working years, as well as consumption and retirement decisions. Finally, the contribution data (particularly given the high level of contribution ceilings) generates a complete record of employer-employees relationship, which is likely the best source of information on formal labor markets in the Dominican Republic. It could be explored to provide additional insight into topics such as gender inequalities; the role of firms in setting wages; the impacts of labor market policies such as the mininum wage on employment; among many others. This could be Official Use Only particularly fruitful if pension administrative records could be connected to other administrative data, such as tax records. References Aguila, Emma. 2011. "Personal Retirement Accounts and Saving." American Economic Journal: Economic Policy 3, no. 4 (November) 1–2. Akbas, Merve, Dan Ariely, David Robalino, and Michael Weber. 2016. "How to Help the Poor to Save a Bit: Evidence from a Field Experiment in Kenya." SSRN Electronic Journal (10024). https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2803856. Apella, Ignacio Raul, Gonzalo Zunino. 2025. "Labor History and Contribution Density in the Pension System of the Dominican Republic (English)". Policy Research Working Paper;People Washington, D.C. : World Bank Group. http://documents.worldbank.org/curated/en/099223004212511701 Azuara, Oliver, Mariano Bosch, Gustavo Caballero, Fabián Cofré, Stephanie González, Anne Hand, Lukas Keller, Catalina Rodríguez Tapia, and María Teresa Silva-Porto. 2021. "Saving without Barriers: Lessons from the Retirement Savings Laboratory’s Pilot Projects." Interamerican Development Bank. https://publications.iadb.org/es/ahorro-sin-barreras-lecciones-de-las- intervenciones-del-laboratorio-de-ahorro-para-el-retiro. Becerra, Oscar. 2024. "The Effect of Future Pension Benefits on Labor Supply in a Developing Economy." The University of Chicago Press, Economic Development and Cultural Change 72, no. 3 (April) 1527–1566. Berstein, Solange, Guillermo Larraín, Francisco Pino, and Eduardo Morón. 2006. "Chilean Pension Reform: Coverage Facts and Policy Alternatives [with Comments]." Brookings Institution Press, Economía 227–279. https://www.jstor.org/stable/20065500. Cabezon, Francisco. 2023. "The Unequal Protection of Social Security in the Presence of Informality: Theory and Evidence from Chile ." Corsetti, Giancarlo. 1994. "An Endogenous Growth Model of Social Security and the Size of the Informal Sector." Economic Analysis Review 57-76. Even, William, and David Macpherson. 1994. "Why Did Male Pension Coverage Decline in the 1980s?" ILR Review, 47(3) 439-453. Official Use Only Finamor, Lucas. 2023. "Labor market informality, risk, and public insurance." Forteza, Alvaro, Ignacio Apella, Eduardo Fajnzylber, Carlos Grushka, Ianina Rossi, and Graciela Sanroma. 2009. "Work Histories and Pension Entitlements in Argentina, Chile and Uruguay." SSRN Electronic Journal. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1626253. French, Eric, Attila S. Lindner, Cormac O’Dea, and Tom A. Zawisza. 2022. "Labor Supply and the Pension-Contribution Link." NBER. Gars, Jared, Laura Prada, Egon Tripodi, and Santiago Borda. 2025. "Personalized Reminders: Evidence from a Field Experiment with Voluntary Retirement Savings in Colombia." CESifo Working Paper. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5188137. Group, Word Bank. 2023. "Fast Tracking Poverty Reduction and Prosperity for All: Dominican Republic Poverty Assessment 2023." Joubert, Clement. 2015. "Pension Design with a Large Informal Labor Market: Evidence from Chile." International Economic Review 56 (2) 673–694. Joubert, Clement, and Petra E. Todd. 2022. "Gender pension gaps in a private retirement accounts system: A dynamic model of household labor supply and savings." Journal of Econometrics (October). Joubert, Clement, and Petra Todd. 2024. "Gender pension gaps in a private retirement accounts system: A dynamic model of household labor supply and savings." Journal of Econometrics 243, no. 1 . https://www.sciencedirect.com/science/article/pii/S0304407622001622. Klerby, A, B Larsson, and E Palmer. 2019. "Bridging partner lifecycle earnings and pension gaps by sharing NDC accounts." INAPP. Kleven, Henrik, Landais Camille, Posch Johanna, Steinhauer Andreas, and Zweimüller Josef. 2019. "Child Penalties across Countries: Evidence and Explanations." AEA Papers and Proceedings 109 122–26. Kreyenfeld, M, S Schmauk, and T Mika. 2023. "The gender pension gap in Germany: is divorce a gender-equaliser?" Ageing and Society. Lauletta, Maximiliano, and Marcelo Bergolo. 2024. "Pension Privatization, Behavioral Responses,and Income in Old Age: Evidence from a Cohort-Based Reform in Uruguay." Official Use Only McKiernan, Kathleen. 2021. "Social Security reform in the presence of informality." Review of Economic Dynamics 40 (April) 228–251. Moreno, Carla. 2021. "The Impact of Pension System in Labor Markets with Informality." Ortiz, Isabel, Fabio Duran, Stefan Urban, Veronika Wodsak, and Zhiming Yu. 2018. "Reversing Pension Privatization: Rebuilding Public Pension Systems in Eastern European and Latin American Countries (2000-18)." SSRN Electronic Journal. Palacios, Robert J. 2003. "PENSION REFORM IN THE DOMINICAN REPUBLIC." Pardo, Oliver. 2020. "Mandatory Retirement Savings in the Presence of an Informal Labor." SSRN Electronic Journal. Rofman, Rafael, Ignacio Apella, and Evelyn Vezza. 2015. "Beyond Contributory Pensions: Fourteen Experiences with Coverage Expansion in Latin America [in en]." Technical report 92872. World Bank. Winkler, Hernan, and Miriam. Montenegro. 2021. Jobs Diagnostic Dominican Republic. The World Bank. Official Use Only 8. Appendix Figure A.1: Histogram of contribution density (Months contributed/ potential contribution) Note: 81508 observations, from individuals aged 22-60 in 2023. Data comes from a sample of 100.000 randomly selected contributors from the SIPEN affiliates pool as of 2023. Contribution density is calculated as the ratio of the number of months contributed by each individual to the potential contribution each could have, assuming that they start working at age 22. Source: SIPEN administrative records. Figure A.2: OLS model results Official Use Only Figure A.3 OLS model results (Continued) Official Use Only Note: OLS model: Contribution density (10 years) = Age Group + Gender + Industry + Wage Level + Province. Contribution density is calculated as the ratio of the number of months contributed to the feasible contribution of a worker in 10 years. Confidence intervals estimated at 95%. Source: SIPEN administrative records. Figure A.4 Distribution of contribution and densities by age group Official Use Only Note: 176622 observations, from individuals aged 27-60 between 2014-2023. Data comes from a sample of 100.000 randomly selected contributors from the SIPEN affiliates pool as of 2023. Contribution density is calculated as the ratio of the number of months contributed by each individual to the potential contribution each could have, assuming that they start working at age 22. Source: SIPEN administrative records. Figure A.5 Sector of employment among pension affiliates versus the general working age population (2023) Note: The dataset includes 46,209 pension affiliates, drawn from a random sample of 100,000 contributors selected from the SIPEN administrative records as of 2023. The working-age population dataset (4,835,319 individuals) is based on estimates from the ENCFT (Encuesta Continua de Fuerza de Trabajo), a national labor force survey Official Use Only conducted by the Central Bank of the Dominican Republic. Survey data available at: https://www.bancentral.gov.do/a/d/2541-encuesta-continua-encft. Figure A.6 Comparison of employer size distribution among pension affiliates vs. over all population Note: Upper panel shows the share of employment by firm size (left) and the share of firms by size (right) among pension affiliates. The dataset includes 46,209 pension affiliates, drawn from a random sample of 100,000 contributors selected from the SIPEN administrative records as of 2023. Source: Own elaboration. The lower panel reproduces the same two graphs taken from the Jobs Diagnostic Report: Dominican Republic by the World Bank, based on data from the national tax authority (DGII). Image source:https://openknowledge.worldbank.org/server/api/core/bitstreams/9ecfc49d-ef24-55ee-bd21- 55dcd6630868/content. Official Use Only Social Protection & Jobs Discussion Paper Series Titles 2025 No. Title 2519 Who participates in defined contribution pension systems when informality is high? Evidence from 20 years of administrative records from the Dominican Republic 2518 Vertical Skills Mismatch in Türkiye’s Labor Market: Evidence from Refined Measures 2517 Skills Systems for Adults and Out-of-School Youth: 2516 Evaluating the Employment Effects of a Labor Market Integration Program for Syrian Refugees and Turkish Nationals in Türkiye 2515 PES Digitalization Note 2514 Out of Harm’s Way: Adaptive Social Protection for Pandemics in South Asia 2513 State of Social Protection Report 2025: The 2-Billion-Person Challenge. Background Paper # 6: Unlocking the Potential of Household Surveys to Measure Women’s Access to Social Protection 2512 State of Social Protection Report 2025: The 2-Billion-Person Challenge. Background Paper # 5: Riding the Demographic Wave: Pensions and Retirement Income in an Aging World 2511 State of Social Protection Report 2025: The 2-Billion-Person Challenge. Background Paper #4: Optimizing Labor Market Programs and Strengthening Delivery Systems for Impact and Scale 2510 State of Social Protection Report 2025: The 2-Billion-Person Challenge. Background Paper #3: Wake-Up Call for Social Assistance? Unfinished Mission to Reach the Poor and Beyond 2509 State of Social Protection Report 2025: The 2-Billion-Person Challenge. Background Paper #2: Adaptive Social Protection Agenda: Lessons from Responses to COVID-19 Shock 2508 State of Social Protection Report 2025: The 2-Billion-Person Challenge. Background Paper #1: Mind the Gap: Coverage, Adequacy and Financing Gaps in Social Protection for the Extreme Poor and the Poorest Quintile 2507 Service Integration and Case Management for People on the Move: A Review of Selected International Practices 2506 Impact of Climate Change and the Green Transition on Human Capital: A Review of the Evidence from Europe and Central Asia 2505 A slippery slope: the opportunities and risks of digital approaches and technology in Social Protection Systems Official Use Only 2504 De Jure and De Facto Coverage of Parental Benefits in Nepal 2503 Awareness, Access, and Perceptions around Parental benefits among Urban Argentinians 2502 Regulating Markets So More People Find Better Jobs 2501 São Tomé and Príncipe Unpacking Migration Dynamics: Critical Issues and Policy Recommendations To view Social Protection & Jobs Discussion Papers published prior to 2021, please visit www.worldbank.org/sp. Official Use Only ABSTRACT Using rich administrative data from the Dominican Republic’s national pension supervisor (SIPEN), this note analyzes contribution histories to assess the system’s performance. Average contribution density remains low—around 40%—with wide variation by geography, sector, and income, highlighting the importance of encouraging contribution continuity. Younger cohorts exhibit more formalized lifecycle employment profiles, suggesting that formalization interventions may be more effective early in the career. Among formal workers, gender gaps in contribution rates and earnings have largely closed, but women are less likely to enter formal employment. These findings underscore both progress and persistent risks, particularly for workers with fragmented formal careers and for women without spousal protections. The analysis provides a foundation for evaluating and strengthening pension systems in high-informality settings as they face the pressure of accelerated population aging. JEL CODES H55, J46 KEYWORDS Informality, Pension participation, Dominican Republic, Aging, Gender