DEPARTAMENTO DE ECONOMÍA Pontificia Universidad Católica del Perú DEPARTAMENTO DE ECONOMÍA Pontificia Universidad Católica del Perú DEPARTAMENTO DE ECONOMÍA Pontificia Universidad Católica del Perú DEPARTAMENTO DE ECONOMÍA Pontificia Universidad Católica del Perú DEPARTAMENTO DE ECONOMÍA Pontificia Universidad Católica del Perú DEPARTAMENTO DE ECONOMÍA Pontificia Universidad Católica del Perú DEPARTAMENTO DE ECONOMÍA Pontificia Universidad Católica del Perú DEPARTAMENTO DE ECONOMÍA Pontificia Universidad Católica del Perú DT DECON DOCUMENTO DE TRABAJO GENDER GAP IN PENSION SAVINGS: EVIDENCE FROM PERU’S INDIVIDUAL CAPITALIZATION SYSTEM Nº 513 Javier Olivera y Yadiraah Iparraguirre DOCUMENTO DE TRABAJO N° 513 Gender gap in pension savings: Evidence from Peru’s individual capitalization system Javier Olivera y Yadiraah Iparraguirre Junio, 2022 DOCUMENTO DE TRABAJO 513 http://doi.org/10.18800/2079-8474.0513 http://doi.org/10.18800/2079-8474.0513 Gender gap in pension savings: Evidence from Peru’s individual capitalization system Documento de Trabajo 513 © Javier Olivera y Yadiraah Iparraguirre Editado e Impreso: © Departamento de Economía – Pontificia Universidad Católica del Perú Av. Universitaria 1801, Lima 32 – Perú. Teléfono: (51-1) 626-2000 anexos 4950 - 4951 econo@pucp.edu.pe http://departamento.pucp.edu.pe/economia/publicaciones/documentos-de-trabajo/ Encargada de la Serie: Janina V. León Castillo Departamento de Economía – Pontificia Universidad Católica del Perú jaleon@pucp.edu.pe Primera edición – Junio, 2022 ISSN 2079-8474 (En línea) mailto:econo@pucp.edu.pe file:///d:/Users/mirtha.cornejo/Dropbox/cisepas%20(1)/Procesando-Documentos%20de%20Trabajo/jaleon@pucp.edu.pe Gender gap in pension savings: Evidence from Peru’s individual capitalization system∗ Javier Olivera† Yadiraah Iparraguirre‡ June 30, 2022 Abstract We study the gender gap in pension funds in Peru, a country where the main pension system is based on individual retirement accounts. We exploit randomly selected sam- ples of administrative pension fund individual registers collected between 2005 and 2019 and find a gender gap in favour of men at each percentile of the distribution of pension funds. The unconditional gender gap decreases along the percentiles until it reaches a sort of “glass ceiling” around the 85th percentile, and then it increases substantially. We also detect heterogeneity by birth cohorts, indicating that older cohorts show higher gender gaps in pension saving because of the capitalization process. Moreover, we find that awareness about pension fund risk management –a proxy for financial literacy– increases the disper- sion of pension savings over the distribution and, therefore, increases inequality and the gender gap. This situation is aggravated by the fact that Peru has very low levels of finan- cial literacy. Key words: Gender gap, Pension savings, Financial literacy, Unconditional quantile, Peru JEL-classification: D31, G23, J16, J32. ∗We are grateful to the research assistance provided by Marcelo Lozada. We also thank the helpful comments provided by the participants at the 29th Colloquium of Pensions and Retirement Research (IPRA), Patricia Vera, and Fernando Rios-Avila. We are also grateful to the Department of Economics of PUCP for financial support. The authors declare that they have no financial or material interests in the results of this study. All the results and interpretations provided in this study are of exclusive responsibility of the authors. †Luxembourg Institute of Socio-Economic Research (LISER), and Department of Economics, Pontificia Uni- versidad Catolica del Peru; e-mail: javier.olivera@liser.lu; e-mail: olivera.j@pucp.edu.pe; ORCID ID: https: //orcid.org/0000-0002-8769-7845 ‡Innovations for Poverty Action, and Department of Economics, Pontificia Universidad Catolica del Peru; e- mail: yadiraah.iparraguirre@pucp.edu.pe; e-mail: yiparraguirre@poverty-action.org. 1 http://javier.olivera@liser.lu http://olivera.j@pucp.edu.pe https://orcid.org/ 0000-0002-8769-7845 https://orcid.org/ 0000-0002-8769-7845 mailto:yadiraah.iparraguirre@pucp.edu.pe mailto:yiparraguirre@poverty-action.org Brecha de género en el ahorro pensionario: Evidencia del sistema de capitalización individual de Perú* Javier Olivera** Yadiraah Iparraguirre*** 29 de junio de 2022 Resumen Estudiamos la brecha de género en los fondos de pensiones en Perú, país donde el prin- cipal sistema de pensiones se basa en cuentas individuales de retiro. Analizamos muestras seleccionadas aleatoriamente de registros administrativos individuales de fondos de pen- siones recopilados entre 2005 y 2019. Encontramos una brecha de género a favor de los hombres en cada percentil de la distribución de los fondos de pensiones. La brecha de gé- nero, no condicionada, disminuye a lo largo de los percentiles hasta alcanzar una especie de “techo de cristal” alrededor del percentil 85, y luego aumenta sustancialmente. También detectamos heterogeneidad por cohortes de nacimiento, indicando que las cohortes de ma- yor edad muestran mayores brechas de género en el ahorro pensionario debido al proceso de capitalización. Además, encontramos que el conocimiento sobre la gestión del riesgo de los fondos de pensiones –una variable que captura alfabetización financiera– aumenta la dispersión de los ahorros pensionarios en la distribución y, por lo tanto, aumenta la de- sigualdad y la brecha de genero en los fondos de pensiones. Esta situación se ve agravada por el hecho de que Perú tiene niveles muy bajos de alfabetización financiera. Palabras claves: Brecha de género, Fondos de pensión, Alfabetización financiera, Regre- siones por cuantiles, Perú Códigos JEL: D31, G23, J16, J32. *Agradecemos la asistencia en investigación que brindó Marcelo Lozada. También agradecemos los valiosos comentarios proporcionados por los participantes en el 29° Coloquio de Investigación sobre Pensiones y Jubila- ciones (IPRA), Patricia Vera y Fernando Ríos-Ávila. También agradecemos al Departamento de Economía de la PUCP por el apoyo financiero. Los autores declaran que no tienen intereses económicos ni materiales sobre los resultados de este estudio. Todos los resultados e interpretaciones proporcionados en este estudio son de exclusiva responsabilidad de los autores. **Luxembourg Institute of Socio-Economic Research (LISER) y Departamento de Economía, Pontificia Uni- versidad Católica del Perú; e-mail: javier.olivera@liser.lu; e-mail: olivera.j@pucp.edu.pe; ORCID ID: https: //orcid.org/0000-0002-8769-7845. ***Innovations for Poverty Action y Departamento de Economía, Pontificia Universidad Católica del Perú; e- mail: yadiraah.iparraguirre@pucp.edu.pe; e-mail: yiparraguirre@poverty-action.org http://javier.olivera@liser.lu http://olivera.j@pucp.edu.pe https://orcid.org/ 0000-0002-8769-7845 https://orcid.org/ 0000-0002-8769-7845 mailto:yadiraah.iparraguirre@pucp.edu.pe mailto:yiparraguirre@poverty-action.org 1 Introduction Various Latin American countries reformed their pension systems by implementing schemes based on individual retirement accounts (IRA) during the wave of structural reforms of the 1990’s. The previous public pension schemes – mostly based on Pay-as-you-go (PAYG) fi- nancing – were fully or partially replaced with the new IRA systems. Among the frequently mentioned goals of these reforms were solving public financial imbalances and facilitating in- dividuals to access to better pensions by means of individual capitalization. There are studies assessing the benefits and problems created by these reforms in different dimensions such as pension adequacy, saving behaviour and capital market development (e.g. Bosch et al. (2013); Arenas de Mesa (2019); Altamirano-Montoya et al. (2018)), but less on gender gaps in pension savings accumulation. It has been found that the key factors to understand the gender gap and its trends in pensions are the effects of the life course of women, including their participation in the labour market, differential mortality, and the institutional characteristics of the pension system (Bando, 2019; Madero-Cabib et al., 2019). Despite advances in educational attainment of women, the family roles and the labour market characteristics are still strong determinants of gender gaps. Women often participate less in the labour market, spend disproportionately more time in household tasks and have lower wages (Cordova et al., 2021; Arza, 2015; Madero-Cabib et al., 2019). It has been illustrated that these conditions may hamper the ability of women in the long-term to accumulate pension savings and generate adequate levels of pensions (Altamirano-Montoya et al., 2018). The design of the pension system and its rules have an important role on determining pen- sion outcomes by gender. For example, minimum contributions spells, minimum pensions, retirement age, mortality tables used to compute pension benefits and the link between bene- fits and earnings have a key role on the gender differences observed in pensions (Arza, 2015; Bertranou, 2001). This means that IRA systems, which strengthen the link between lifetime wages and pensions, prompt a new set of gender equality issues. The gender gap could fur- ther expand if we consider that financial knowledge or the ability to make better investment choices or annuity management is more prevalent among men than among women (Lusardi and Mitchell, 2010; Hastings et al., 2010; Fonseca et al., 2012). We seek to contribute to the literature on gender pension gap by studying the case of Peru, a country where the main compulsory pension system is based on IRA. This is the Private Pension System (known as SPP, due to its Spanish name). Although there is an alternative public pension system (known as SNP, due to its Spanish name), most of the new workers enrol into the SPP (in 2019, the ratio between new affiliates of the SPP to the SNP was about 5 to 1). We use representative samples of the non-retired population affiliated to the SPP, randomly selected from administrative registers in 2005, 2006, 2013, 2015, 2016, and 2019. Our data allows us to analyse gender gaps in pension balances along birth cohorts and across the years of our period 2 of analysis. There are not many empirical studies assessing gender gaps in pension wealth generated in IRA systems. Note, however, that a recent study by Cordova et al. (2021) finds in Germany that the low participation and capitalization of women in private pension plans may explain their lower pensions. Thus, low pensions are mainly explained by the type of occupation, income level, hours of work and the presence of children at home. Also in Germany, Flory (2012) finds that the pension wealth gap is heterogeneous by birth cohort groups. The gender gap tends to be lower among younger cohorts than among older cohorts. In the United Kingdom, Foster and Smetherham (2013) find that the factors strongly associated to individual contribution rates are the type of occupation, income, economic position and having young children, with these predictors being more salient for women than for men. Studies on net worth show a gender gap favouring men over women. Schneebaum et al. (2018) exploit the Household, Finance and Consumption Survey (HFCS) to study the gender gap throughout the distribution of net wealth across single-person households in eight European countries. They find that the gap increases with the percentiles of the distribution of net wealth, showing evidence of a “glass ceiling” problem. When looking at the gap by type of wealth, this gap does not come from asset wealth in the household, but from the disparity in occupational pensions. Furthermore, they find that the gender gap increases when both the cohort is older and the percentile is higher, with the exception of Germany and Spain. Likewise, the study by Meriküll et al. (2021) on Estonia finds an increasing gender gap in favour of men across the quantiles of household net wealth. The factors explaining this result are labour market status (self-employment, retirement) education, occupation type, and marital status. Moreover, studies like the ones by Frémeaux and Leturcq (2020) and Sierminska et al. (2019) exploit individualized wealth portfolios to understand several drivers of gender wealth gaps. Like some of the previous studies mentioned, we also use unconditional quantile regressions to estimate gender gaps along the distribution of pension wealth. We find a gender gap in favour of men at each percentile of the distribution pension funds, yet this decreases constantly along the percentiles until reaching a sort of “glass ceiling” around the percentile 85. From that point, the gender gap rapidly increases in the top section of the distribution of pension wealth. Overall, our results point out that on the one hand, we observe a reduction of the gender gap in pension wealth across birth cohorts, which is line with other findings about the increase of female labour participation and wage improvements. However, on the other hand, we observe that this not enough to reduce consistently the pension wealth gap when we observe longer spells in the capitalization system. The capitalization of individual pension contributions –operating via the capital market returns of pension funds– amplifies any early and small gender gap observed at the beginning of the working life. Moreover, in a country where only 28% of adult population scores correctly in financial lit- eracy questions about interest rate, inflation and risk diversification (see Klapper et al. (2015)) is important to know whether this could have a role on the gender pension wealth gap. We are able 3 to capture financial literacy by observing how individuals move from default options in the risk composition of their portfolio investments, which we call Awareness of portfolio management. As has been found in other studies (e.g. Lusardi et al. (2017) find an effect of financial literacy on household wealth inequality), we document that financial literacy may increase inequality in pension savings. Furthermore, the importance of Awareness of portfolio management in ex- plaining pension savings gap increases along the distribution of pension wealth, meaning that differences in financial literacy may exacerbate gender gaps as well as pension savings inequal- ity. The remainder of the document is organized as follows. Section 2 describes the institutional framework of the pension system in Peru. Section 3 describes the data and empirical strategy. Section 4 reports and discusses the main results, and Section 5 exploits available information on portfolio risk choices to study the role of financial literacy on the distribution of pension savings and gender gaps. Section 6 presents additional results regarding a measure of extended pension wealth and the distribution of income. Finally, Section 7 presents the conclusions. 2 Institutional background The Peruvian pension system is composed of two schemes which represent two mutually ex- clusive options to the individuals. First, the SPP is a defined contribution (DC) system, which is based on individual retirement accounts (IRA) and started in June 1993. The introduction of this type of system was part of a wave of pension reforms, inspired in the Chilean case, well-spread across Latin America during the 1990’s. The pension fund managers (the so-called AFP) are firms receiving the pension contributions and investing the individualized savings on investments tightly regulated by the Superintendent of Banking, Insurance and Pension Funds (known as SBS due to its Spanish name). Second, the National Pension System (SNP) is a defined benefit (DB) system operating as a PAYG system. The individuals must choose one of these pension systems at the beginning of their working lives. If the SPP is chosen, the individ- ual must remain there, but a shift from SNP to SPP is possible at any time. In order to recognize part of the contributions made to the SNP, the government has issued “Recognition Bonds” which values are monthly updated by the official prices index. So far, there are three types of bonds, issued as of 1992, 1996 and 2001. There are currently four AFP in Peru: Prima, Integra, Profuturo and Habitat. There were other AFP in the past, but they gradually left the market or merged with other firms over time. After several changes in the regulation, currently the worker contributes 10% (plus administrative fees and insurance premium) of the gross salary to the AFP, or 13% to the SNP.1. IRA systems have been strongly criticized for their distributive impacts. These systems tend to favour people with higher incomes and do not guarantee minimum pensions, leaving many people with small pensions during old age and limiting their eligibility for social assistance 1The contribution rate was 11% in 1993-1995, 8% in 1996-2005 and 10% since 2005 4 programs. At least in Peru, an individual is eligible to social pensions only if she is older than 65, extreme poor and has no private or public pensions. The dis-equalizing effects of IRA systems are aggravated in contexts with high levels of informality and high turnover between formal and informal employment, which reduces the frequency of contributions. Likewise, these systems are criticized for their high fees and administrative costs. Yet, these systems have also attracted support when the assessment of the IRA system is focused on the positive contributions made on national savings, economic growth and the development of new annuity markets. The SPP could also exacerbate the inequalities observed in labor income through the capi- talization process and the disparity in the frequency of contributions among individuals. This means that the process of capitalization and the fact that richer individuals contribute more fre- quently could generate larger differences in pension savings than in incomes. In addition, these inequalities may further increase once we take into account the absence of minimum guaran- teed benefits in the SPP and the fact that the computation of pensions use gender-differentiated mortality tables, which favour men over women.2 This differs from DB systems, which tend to reduce inequality through minimum guaranteed pensions and the use of unisex mortality tables to determine the amount of pensions. Consequently, gender inequality in pension savings could be significant in IRA type systems. We can see some of the main factors driving gender gaps in pension savings with the following stylized equations: Bi = a 65− j ∑ j=25 (wi jdi j)(1+ r)(65− j) i = m, f (1) Gender Gap = Bm−B f = a 65− j ∑ j=25 (wm jdm j−w f jd f j)(1+ r)(65− j) (2) The equation 1 indicates the level of pension balance (Bi) accumulated at retirement age by men (m) or women ( f ). The value depends on income (w), frequency of contributions (d ∈ [0,1]), contribution rate (a), return rate (r), and the period of capitalization (65− j) (as- sumed between age 25 and 65). The gender gap is reported in equation 2, showing the difference of pension balances accrued by men and women at retirement. We observe that two main com- ponents affect the level of the gender gap, which are the wages and the capitalization process. On the one hand, we have the differences in labor income (weighted by frequency of contri- butions, which may reflect occupation status and degree of formality), and on the other hand we have the capitalization process driven by the return rate and the length of the capitalization period. A gender gap could potentially be observed at any period of the labour span of the 2Women have larger life expectancy than men, which must be reflected in the annuity price formula used to compute pensions in the SPP. For example, the difference in pensions attributed to differential sex mortality would be about 10% (in favour men over women) when we compute an annuity at age 65 using the SPP’s official life tables for a single a man and a single woman with the same level of pension balance, and a discount interest rate of 3%. 5 individual, but it is key to recognize that only at retirement age we could fully account for all the capitalization process affecting the full value of pension savings. 3 Data and empirical strategy 3.1 The data We use cross-sectional samples of the total non-retired population from the SPP administrative registers as of 2005, 2006, 2013, 2015, 2016, and 2019. The samples are random, stratified and representative of the following strata for each sampling year: 5-year age group, sex and year of enrolment in the SPP. These are the only available data sets including information about each individual’s pension balance, management fees, income and some demographic variables. For each year, the sample is equivalent to 2% of the total non-retired population in the SPP3. The initial sample size is composed of 600,360 observations, which corresponds to individ- uals aged between 21 and 64 in each sample year. We do not consider individuals older than 65 as this is the legal retirement age. After dropping individuals with no information on pension balance (165), affiliated for less than one month (1,307), being in pension fund type 0 (1,536), and with zero pension balance (64,152), we obtain our final sample of 533,200 observations. The balances of value zero may reveal that the individual is not able or does not want to accu- mulate pension funds. Given that our interest rests on assessing pension fund gender gaps of people who do save for pensions, we removed the individuals who have not pension savings.4 Thus, our analysis is representative of the non-retired population who have at least contributed once to their retirement savings in the SPP. The micro-data include information on age, gender, employment condition and income at the individual level. The data also include information on the pension account, such as the enrollment date, AFP, last contribution date, pension balance, type of fee (load factor or balance fee), type of pension fund, information about recognition bonds, and contribution density. This last variable indicates the share of contributions made by the individual with respect to the theoretical total number of contributions that the individual should made, so that the values ranges between 0 and 1. However, as this variable is only available for the samples extracted in 2015, 2016, and 2019, we are still able to use the date of the last registered contribution, available for all the sample years, to compute a proxy variable. The variable regular contributor takes value one if the last contribution registered for the individual was made in the sampling year, and zero otherwise. 3The sample size is 1.8% of the SPP non-retired population for 2005 and 2006, and is 2% for each of the other years. 4Among the individuals with zero pension balance, 28% have been enrolled in the SPP for 15 years or more, 45% between 5 and 14 years, and 23% between 1 and 4 years. Possible explanations for this behaviour are that the individual was enrolled while she was working in an activity with no obligation to contribute (informal sector or as self-employed) or was inactive (e.g. students). There are also the so-called “ghost affiliates”, who are individuals who never realized they were enrolled at some point by an AFP salesman. 6 There are four main types of pension funds. Fund type 0 is designed to maintain capital, offers both very low return and volatility and is intended for individuals who are in the process of acquiring a pension. Fund type 1 includes investments with relatively low returns and volatil- ity and is mandatory for individuals aged 60-65, unless the individual has expressly chosen to be assigned to fund type 0 or 2. Fund type 2 includes investments with moderate growth and volatility and combines both fixed-income instruments and equities. Fund type 3 is generally composed of investments with higher returns and volatility such as equities. 5 When an individ- ual enrolls for the first time into an AFP, the default pension fund is type 2. Choosing another type of pension fund requires a special administrative procedure. Following Bernal and Olivera (2020) we use these pension fund risk defaults to compute a measure about how active individ- uals are regarding their portfolio management. The variable Active portfolio management takes the value of one if an individual under 60 has a pension fund type 1 or 3 or whether an indi- vidual older than 60 has a pension fund other than type 1; and takes value zero otherwise. This variable indicates that the individual has taken action to move away from the default pension fund risk portfolio. We argue that this variable captures awareness about risk diversification and may therefore be a proxy for financial literacy. We expect that more sophisticated individual investors will be more likely to deviate from the defaults. 3.2 Empirical strategy First, we use OLS regressions to explore and estimate the gender gap in pension balance ac- cording to the following equation: Bi = α +β1malei +β jC ji +πyeari + γ jC ji×malei +X ′ i θ + εi (3) Where Bi is the pension balance of individual i, C ji is an indicator variable for the birth year cohort j of the individual, yeari is the year of the sample draw, malei is an indicator variable for men, X ′ i is a vector of covariates, and εi is the error term. As we are interested in estimating the gender gap magnitudes along the distribution of pension wealth, we will perform unconditional quantile regressions (UQR). These regressions are based on an extension of the Recentered In- fluence Function (RIF), which provides a linear approximation of the unconditional quantiles of the dependent variable of analysis (Firpo et al., 2009). This function is defined as the following: RIF(B;Qτ ,F) = Qτ + IF(B;Qτ ,F) (4) Where Qτ is the value of pension balance B at quantile τ in the unconditional distribution F of pension balances, and IF(B;Qτ ,F) is the quantile influence function. The influence function 5Fund type 0 invest 100% on fixed-income instruments. Fund type 1 invests up to 90% in short-term fixed- income instruments and up to 10% in equities; fund type 2 invests up to 55% in short-term fixed-income instru- ments and up to 45% in equities; and fund type 3 is composed of investments up to 80% in equities and up to 20% in short-term fixed-income instruments. 7 of a statistic (in our case, the quantile) indicates how sensible is this statistic to different areas of the distribution (Choe and Van Kerm, 2018). This function is represented as follows: IF(B;Qτ ,F) = (τ− I[B≤ Qτ ]) fB(Qτ) (5) Where fB(Qτ) is the density function up to the percentile τ and I[B≤Qτ ] is a binary variable that takes the value of one when the value of B is lower than the corresponding percentile, and zero otherwise. Replacing equation 5 in 4 we obtain equation 6 that represents the RIF. Using the RIF assures that the change in its average value over time is equal to the change in the statistic of interest (Davies et al., 2017). RIF(B,Qτ) = Qτ + (τ− I[B≤ Qτ ]) fB(Qτ) (6) Once the RIF estimators of B are computed, the following equation can be estimated by OLS: RIF(B,Qτ) = α +β1malei +β jC ji +X ′ i θ + εi (7) The RIF regression allows us to evaluate the impact of any covariate on the statistic of in- terest and determine which variable is associated with the greatest influence on the distribution. The coefficients obtained from the regression can be interpreted as how much an infinitesimal change in the distribution of the covariate influences a given quantile, maintaining everything else constant, which is known as the unconditional quantile partial effect (UQPE). This method allows us to focus on a certain point of the distribution of the pension balance (high or low percentile), regardless of whether the values of the covariates are the same (Firpo et al., 2009; Choe and Van Kerm, 2018). This represents an advantage for studying gender gaps compared to the conditional quantile regression (CQR) method. The reason is that the CQR estimates can only be interpreted on a set of individuals sharing covariates with the same values and cannot be used to estimate the impact of a variable of interest on the corresponding unconditional per- centile (Firpo et al., 2009). Thus, we could evaluate the effect of increasing the participation of men or increasing a year of affiliation in a certain upper percentile of the distribution, in which there are probably more men than women. In addition, we use the Oaxaca-Blinder decomposition based on the RIF regressions for men and women at given percentiles. Following the standard decomposition equation, we have the following: B̄m,τ − B̄ f ,τ = ( Xm−X f ) βm,τ +X f ( βm,τ −β f ,τ ) (8) Where B̄m,τ − B̄ f ,τ represents the gender gap in pension wealth at percentile τ , Xm and X f represent the average values of the explanatory variables, and βm,τ and β f ,τ are the coefficients that come from running RIF regressions for men and women. The decomposition allows the gender gap to be separated into two components: a component explained by differences in the 8 characteristics between men and women, and an unexplained component that has its origin in the differences in the returns of the variables that are typically linked to gender discrimination in the labor market. 4 Results 4.1 Descriptive analysis Table 1 reports the percentiles of the distribution of pension savings for men an women and the raw gender gap observed at each of these percentiles. In average, men owns 37% more pension funds than women do, but this gap is different across the distribution of pension funds.6 The gender gaps show a sort of U shape, being about 60% higher for men in the three first deciles, and then reducing until the 89th percentile. From that point, the gap increases quickly towards the top section of the distribution of pension savings. For example, the gender gap grows from 22% at the 90th percentile to 43% at the 99.5th percentile. The findings are similar to other studies assessing gender wealth gaps. For instance, Anglade et al. (2017) also find a U- shaped distribution of gender gaps in the distribution of wealth of single individuals in Ecuador. Given that pension fund accumulation in IRA systems can mimic savings from earnings – which is an important component of financial wealth – it is not surprising that our distribution of gender gaps follow established patterns in gender wealth gaps. The results of Table 1 also confirm the high levels of inequality in the distribution of pension wealth. For example, women (men) in the 99th percentile owns 41 (49) times more savings than women (men) in the 50th percentile. Overall, the Gini index of pension savings is 0.75, while the Gini index of income in 2019 is 0.42 (according to World Bank Open Data). These differences in the distribution of pension savings and income are well aligned with general patterns showing that wealth tend to be more unequally distributed than income. When we replicate the results of Table 1 for each sampling year (see Figure A.1 in the Appendix), we observe that the raw gender gaps (measure in men-to-women ratios of pension savings) have been increasing in the lower percentiles of the distribution of pension funds between 2005 and 2019, while the gaps observed in the top percentiles have shown a rather stable pattern for the same period.7 6The raw gender gap is 7,942 Soles, which represents about 8 minimum wages in Peru. 7We only plot the percentile 50th for the bottom group, but other percentiles in the bottom section of the distribution also show an increase in the raw gender gaps. We also notice that in 2005 and 2006, women had larger levels of pension wealth than men until about the percentile 80th. The percentile 99th shows a decrease in gender gaps (47% in 2005 and 35% in 2019), but the other top shares show do not change much in the period. 9 Table 1: Raw gender gaps in pension savings in 2019 (Soles) Variables Mean Percentile 10 20 30 40 50 60 70 80 90 95 99 99.5 Male 29,332 439 1,204 2,469 4,579 7,875 13,117 21,430 36,224 67,340 114,827 323,676 477,979 Female 21,390 274 750 1,534 2,803 4,943 8,383 14,505 27,016 55,359 87,017 239,763 335,168 Gap (M-F) 7,942 164 454 934 1,776 2,932 4,734 6,925 9,207 11,982 27,810 83,913 142,811 P-value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Men-to-women ratio 1.37 1.60 1.61 1.61 1.63 1.59 1.56 1.48 1.34 1.22 1.32 1.35 1.43 Note: The table uses the sample of registers drawn in December 2019 (N=124,942, N male=76,029, N female=48,913). Table 2 reports the means of pension balances by gender and various characteristics of the individuals for the sampling year 2019. As expected, older cohorts accumulate more pension savings than younger cohorts; and we also observe a reduction in the absolute amounts of gender gaps among younger cohorts. Yet, the gender gaps in percentage terms do not show a clear decreasing pattern along birth cohorts as with the case of the absolute values, but at least we see that the percentage gender gap is higher in the the oldest cohorts than in the youngest cohorts. Table 2: Unconditional means by gender in pension savings in 2019 (Soles) Variables Total Male Female Diff (M-F) Gap in % Mean S. D. Mean S. D. Mean S. D. Difference S. E. All 26,223 (71,536) 29,332 (80,145) 21,390 (55,222) 7,942*** (414) 37.1 Birth cohorts 1996-1998 1,441 (1,698) 1,549 (1,810) 1,302 (1,533) 247*** (38) 19.0 1989-1991 7,488 (10,216) 7,857 (10,529) 7,002 (9,768) 855*** (191) 12.2 1979-1981 23,471 (37,944) 25,427 (38,005) 20,399 (37,647) 5,029*** (731) 24.6 1969-1971 48,808 (98,800) 51,970 (108,708) 43,178 (77,806) 8,792*** (2,258) 20.4 1959-1961 69,919 (166,585) 74,857 (178,212) 59,501 (138,400) 15,356** (6,334) 25.9 Years enrolled in SPP 1-3 2,142 (5,958) 2,450 (7,656) 1,784 (2,906) 666*** (86) 37.3 9-11 14,155 (24,872) 14,912 (27,581) 12,957 (19,790) 1,956*** (453) 15.1 19-21 40,335 (65,332) 42,263 (70,628) 36,810 (54,159) 5,453*** (1,213) 14.8 25-27 89,839 (168,542) 93,014 (180,244) 82,536 (137,661) 10,478*** (3,835) 12.7 Regular contributor No 10,828 (31,089) 12,103 (34,833) 8,863 (24,076) 3,240*** (278) 36.6 Yes 37,319 (88,397) 41,683 (98,950) 30,495 (68,164) 11,188*** (671) 36.7 AFP Habitat 12,305 (70,062) 15,374 (87,196) 8,569 (40,035) 6,805*** (1,025) 79.4 Integra 32,697 (73,963) 35,632 (81,267) 28,064 (60,380) 7,568*** (783) 27.0 Prima 25,513 (77,090) 29,615 (87,993) 19,686 (57,698) 9,929*** (783) 50.4 Profuturo 27,945 (59,036) 29,051 (62,107) 25,674 (52,096) 3,377*** (741) 13.2 Recognition Bond No 23,766 (61,627) 26,516 (68,949) 19,498 (47,788) 7,018*** (359) 36.0 Yes 186,133 (254,390) 206,694 (281,452) 151,320 (195,731) 55,374*** (12,003) 36.6 Note: The table uses the sample of registers drawn in December 2019 (N=124,942, N male=76,029, N female=48,913). The mean differences are computed using two-sample equal variance t-tests by gender. *p<0.10, **p<0.05, ***p<0.01. . The years of affiliation play an important role in the accumulation of savings, particularly if the contributions are more frequent. The descriptive statistics show that the gender gap amounts 10 can increase substantially with the number of years affiliated in the SPP. Figure 1 exploits the entire pooled sample of 2005-2019 and shows the gender gap for each period of affiliation, regardless the calendar year. This also shows a substantial increase in the gender gap value for each year of affiliation. The relatively large intervals of confidence reveal that there is still substantial heterogeneity within each period of savings accumulation. Note, however, that the gender gaps in percentage terms tend to decrease with the number of years of affiliation, which could indicate that for each period of accumulation there are women who are not completely “left behind” in terms of pension balance accumulation. This is somewhat supported by the fact that the percentage gender gap is similar for those individuals with a regular contribution behaviour as for those showing an irregular contribution pattern, meaning that men and women are similarly distributed in terms of contribution frequency. This last point is confirmed by the results shown in Figure A.2 in the Appendix, which reports a similar distribution of pension contributions between men and women. Figure 1: Unconditional gender gaps by number of years enrolled in SPP (pooled sample) 0 1 0 0 0 2 0 0 0 3 0 0 0 4 0 0 0 5 0 0 0 6 0 0 0 7 0 0 0 G e n d e r g a p 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Years affiliated in SPP Note: The figure uses the pooled sample of 2005-2019. The vertical lines indicate confidence inter- vals at 95%. The mean is one of the moments of the whole distribution and it could mask different values of the gender gaps along the distribution of pension funds. In Table 1, we have already observed that the gender gap is very high at the 99th percentile of the pension fund distribution, reaching 83,913 Soles, while this is 2,932 Soles at the median of the distribution. Figures A.3. and A.4 in the Appendix report the gender gaps at different percentiles of the distribution of pension funds across cohorts and selected sampling years. In all cases, the gender gap increases substantially since around the 90th percentile. Regarding the gender gap by AFP, there are some notable differences steaming from the 11 composition of affiliates in each firm. AFP Habitat reports the largest percentage gender gap of all AFP, that is men owns in average 79% more pension savings than women do. Yet, this gap is only 13% in AFP Profuturo. One of the reasons behind this difference is that Habitat is the youngest firm in the market and hence their affiliates have been participating fewer years in the SPP (people have contributed in average 59 months in Habitat, and 198 months in Profuturo). We noticed before that the percentage gender gap falls with the years of enrolment in the SPP. Lastly, the gender gap by Recognition Bond (RB) status shows a much larger gap for those affiliates who have this bond than for those who have not the bond. The reason is that the individuals with RB are mostly older individuals who have already capitalized sizable savings. Nevertheless, there are practically no differences in the percentage gender gap by RB. Figure 2 shows an additional way of exploring gender differences in pension savings. This figure plots the share of women within the percentiles of the distribution of pension funds ob- served in four different years, and also reports the average share of women in the SPP (in dotted lines). First, we observe that the average participation of women in the SPP masks important differences across the distribution of pension funds. Second, we observe how the plots moves towards a more clear negative-slope curve from 2005 to 2019, meaning that the participation of women decreases within the richest percentiles and increases within the poorest percentiles. While in 2005, the share of women in the percentiles (mostly under the 70th percentile) was similar and around the average share, in 2019 we observe a strong negative relation between women participation and the percentiles of pension funds. These results could point to a deteri- orating position of women in the distribution of pension savings. 12 Figure 2: Share of women across the unconditional distribution of pension savings .2 .2 5 .3 .3 5 .4 .4 5 .5 .5 5 .6 S h a re o f w o m e n 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile 2005 .2 .2 5 .3 .3 5 .4 .4 5 .5 .5 5 .6 S h a re o f w o m e n 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile 2013 .2 .2 5 .3 .3 5 .4 .4 5 .5 .5 5 .6 S h a re o f w o m e n 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile 2016 .2 .2 5 .3 .3 5 .4 .4 5 .5 .5 5 .6 S h a re o f w o m e n 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile 2019 Note: The figures show the share of women across the unconditional distribution of pension balance for each year. The adjusted curves show the lowess-smoothed shares of women, and the dotted lines indicate the average share of women in the SPP. 4.2 Pooled OLS Table 3 shows the estimates of pension balance on a pooled sample including all the year data sets. With no covariates, apart from year fixed effect, the gender gap in pension savings is on average 5,513 Soles in favour of men. Once we control for AFP and birth cohort, the gap is 3,405 Soles (Model 3), which represents about about 16% and 63% of the mean and median pension balances in the sample, respectively. Model 4 shows the results from the estimation of equation 3, which includes interactions between cohorts and gender. These estimates are useful to retrieve the expected gender gap by cohort (β1 + γ jC ji) plotted in Figure 3. Model 5 adds interactions between sample year and gender, and between sample year and cohorts. This allows us to retrieve the gap by cohort for each year showed in Figure 4. 13 Table 3: OLS estimates of pension savings (2005-2019) Variables (1) (2) (3) (4) (5) Male 5,513.0*** 3,461.8*** 3,405.2*** 18,342.5** 18,582.2** (160.7) (142.7) (143.2) (7,980.9) (8,092.6) Regular contributor 20,144.5*** 20,249.8*** 20,370.5*** 20,616.2*** (128.3) (128.9) (130.3) (132.3) Recognition Bond 59,334.1*** 59,022.7*** 59,640.7*** 60,591.9*** (1,172.9) (1,202.6) (1,214.5) (1,227.8) Years enrolled in SPP -123.8** -347.1*** -331.1*** 124.1** (56.3) (56.3) (56.1) (53.3) Years enrolled in SPP^2 /100 12,364.0*** 12,213.5*** 12,144.2*** 10,186.8*** (275.5) (280.5) (279.5) (267.7) Constant 5,624*** -12,821.5*** -22,487.6*** -35,817.2*** -38,167.1*** (146.4) (346.9) (4,421.0) (6,268.6) (6,389.1) Year Yes Yes Yes Yes Yes AFP Yes Yes Yes Yes Cohort Yes Yes Yes Cohort*Male Yes Yes Year*Male Yes Cohort*Year Yes Observations 533,200 533,200 533,200 533,200 533,200 R-squared 0.008 0.155 0.161 0.162 0.162 Notes: The sample corresponds to the pooled samples drawn in 2005, 2006, 2013, 2015, 2016 and 2019. The dependent variable for all regressions is the pension balance. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Figure 3 reveals that the gender gap reduces among younger cohorts. Approximately, up to the cohort 1983, there is a statistically significant positive gender gap, but younger cohorts tend to exhibit a gender gap that is not statistically different from zero. It is interesting to observe a decline in the gender gap among younger cohorts, but Figure 3 could be masking some important heterogeneity across sample years. The capitalization of pension balances could potentially hinder individuals who do not contribute frequently and/or have lower incomes. Thus, if women were more likely to have lower incomes, then a longer period enrolled in the pension system could exacerbate the differences in the pension pots between women and men. Figure 4 could help to observe this. 14 Figure 3: Conditional gender gaps by cohorts (pooled sample) − 2 5 0 0 0 2 5 0 0 7 5 0 0 1 2 5 0 0 1 7 5 0 0 2 2 5 0 0 E s ti m a te d c o e ff ic ie n ts 1956 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 Birth year Notes: The figure plots the sum of estimated coefficients from Model 4 of Table 3, which rep- resent the expected gender gap by birth-year cohort. The dotted lines indicate 95% confidence intervals. Figure 4 shows the expected gender gap by cohorts observed in 2005 and 2019, the two most distanced sample years of our data. It is clear from the figure that for a given cohort, the pension gap increases with the length of the period affiliated in the SPP.8 On the one hand, we observe a decreasing gender gap among younger cohorts, which is in line with other findings in the labour market, but on the other hand, the length of time participating in the pension system increases the gap by means of the capitalization process. 8Starting in 2005, the gender gap increases on average by S/. 1,550, S/. 2,356, S/. 3,409 and S/.4,285 in 2013, 2015, 2016 and 2019, respectively. 15 Figure 4: Conditional gender gap by cohorts in 2005 and 2019 − 2 5 0 0 0 2 5 0 0 7 5 0 0 1 2 5 0 0 1 7 5 0 0 E s ti m a te d c o e ff ic ie n ts 1956 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 Birth year 2005 2019 Notes: The figure plots the sum of estimated coefficients from Model 5 of Table 3, which represent the expected gender gap by birth year cohort and sample year. 4.3 Unconditional quantile regression analysis We obtain two important findings in the OLS regressions: (1) there is a positive gender gap in pension balances favouring men over women; (2) the gap is heterogeneous at the cohort level, being higher for older cohorts and lower for younger groups. Given that we are interested in studying the gender gap along the distribution of pension savings, we estimate unconditional quantile regressions using RIF regressions. As explained before, the RIF regressions allow us to measure how a marginal increase in the participation of men affects the dispersion of pension savings, which could indicate an increase in the gender pension savings gap. We exploit the sample of year 2019 to avoid contaminating our results with different distributions of previous years. Unlike the pooled sample, we introduce some additional variables that are available in that year and are interesting to explore. These covariates are the contribution density (a continuous variable between 0 and 1), type of pension fund risk and region of residence. Table 4 reports the RIF regression coefficients for the 25th, 50th, 75th, 90th, 95th and 99th quantiles, along with bootstrapped (1,000 iterations) standard errors. 16 Table 4: Unconditional quantile regression coefficients on logs of pension savings (2019) Variables OLS Q25 Q50 Q75 Q90 Q95 Q99 Male 0.194*** 0.253*** 0.196*** 0.102*** 0.0540*** 0.152*** 0.207*** (0.00589) (0.0139) (0.00996) (0.00966) (0.0110) (0.0154) (0.0273) Contribution density 3.537*** 4.158*** 4.082*** 3.337*** 2.027*** 1.816*** 1.309*** (0.00923) (0.0196) (0.0137) (0.0146) (0.0188) (0.0266) (0.0455) Years enrolled in SPP 0.375*** 0.598*** 0.472*** 0.202*** 0.0409*** 0.0296*** -0.0142* (0.00203) (0.00460) (0.00297) (0.00290) (0.00319) (0.00444) (0.00729) Years enrolled in SPP^2/100 -0.823*** -1.572*** -1.100*** -0.175*** 0.245*** 0.249*** 0.355*** (0.00733) (0.0159) (0.0110) (0.0111) (0.0132) (0.0188) (0.0324) Recognition Bond 0.716*** -0.110*** 0.219*** 0.946*** 2.276*** 3.753*** 6.893*** (0.0293) (0.0417) (0.0366) (0.0382) (0.0820) (0.161) (0.457) Fund type 2 -0.271*** 0.298*** -0.0773 -0.890*** -1.255*** -1.471*** -2.013*** (0.0477) (0.0607) (0.0552) (0.0718) (0.130) (0.233) (0.709) Fund type 3 0.0981** 0.0839 0.192*** 0.104 -0.0149 0.257 -0.749 (0.0491) (0.0636) (0.0584) (0.0766) (0.138) (0.245) (0.724) Constant 4.558*** 1.078*** 3.254*** 7.443*** 10.63*** 11.58*** 14.25*** (0.0769) (0.164) (0.118) (0.121) (0.172) (0.283) (0.805) Observations 124,829 124,829 124,829 124,829 124,829 124,829 124,829 R-squared 0.751 0.425 0.591 0.532 0.320 0.203 0.081 Notes: The regressions use the sample drawn on December 2019. The dependent variable for all regressions is the pension balance in logarithm. All regressions control for birth cohort, AFP, and region. Bootstrapped standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The regression results show that a marginal increase in the proportion of men will pull the overall pension balance distribution upwards. In other words, the replacement of women by more men leads to a more unequal distribution, which could imply a growing gender gap in detriment of women. Figure 5 illustrates the non-monotonic effects of gender across the different quantiles since the effect of gender is different at each point of the distribution. 17 Figure 5: Male coefficients of unconditional quantile regressions, 2019 Notes: The graph plots the coefficients for male of unconditional quantile regressions whose specification is the same as in Table 4. The shadowed area indicates 95% confidence intervals. The first result observed in Figure 5 is that a larger share of men generates a positive gender gap and increases the dispersion of pension savings for all the percentiles. A second result is that the gender gap decreases consistently along the quantiles until approximately the 85th quantile, from where the gap starts to grow rapidly towards the top quantiles. Thus, greater participation of men in the top quantiles accelerates the grow of the gender pension savings gap, and generates more inequality. This result may reflect a sort of “glass ceiling effect”. Explanations for this effect are related to the fact that women are more likely to be in jobs with lower salaries and in lower ranked positions in the firm than men do. Thus, women have fewer opportunities and less resources to capitalize in the pension system. Given the direct link between pension savings, labour income and occupation status in an IRA system, it is not surprising finding a gendered ceiling in pension savings. 4.4 Oaxaca–Blinder decomposition of gender differentials in pension funds Our full regression results of the Oaxaca-Blinder decomposition can be consulted in Table A.3 in the Appendix, while Figure 6 below plots the estimated raw gender gap and the unexplained component of such gap along distinctive quantiles of the pension fund distribution. The gender gap is positive for any quantile and follows the previous reported negative trend until about the 90th quantile from where it increases towards the upper end of the distribution. The unexplained share of the gap decreases consistently along the quantiles until about the 90th quantile (56%, 47%, 42%, 40%, and 28% for the 10th, 25th, 50th, q75th, and 90th quantiles, respectively), and then it increases to 54% at the 95th quantile and 72% at the 99th quantile. This behaviour is 18 consistent with the previously described “glass ceiling effect”. The most important variables increasing the unexplained gender gap in pension savings in the top quantiles (95th and 99th) of the pension funds distribution are density of contributions and years of affiliation. These vari- ables substantially determine the pension balance accrued across years and are directly linked to labour market outcomes (incomes, occupation, skills, etc.) in an IRA system. Thus, the factors behind the unexplained or “discriminated” part of gender income gaps in the labour market will not only be translated to the gender pension savings gap, but will also be exacerbated because of the capitalization process over time embedded in the IRA system. Figure 6: The gender gap in quantiles of pension savings 0 .1 .2 .3 .4 .5 D if fe re n c e i n l n (p e n s io n b a la n c e ), m e n − w o m e n 10 25 50 75 90 95 99 Quantile Raw gap Unexplained gap Notes: The graph plots the raw and unexplained gender gaps in pension balance across quantiles of the distribution of pension funds. The estimates are based on Oaxaca-Blinder RIF decomposition, which are reported in Table A.3 in the Appendix. The vertical axis shows the estimated values of the raw gap and the unexplained gap (men minus women). The bars indicate 95% confidence intervals. 5 The role of financial literacy We explore the role of financial literacy on gender gaps in pension savings and on the distri- bution of pension savings. We capture financial literacy by exploiting the individual choices (or not) of pension fund portfolios with different risk attributes. As the individual has always the option of opting out of the default pension fund risk allocation, we consider that this action implies awareness with portfolio management and therefore may involve better levels of finan- cial literacy. This strategy has also been employed by Bernal and Olivera (2020) when they analyse a reform of pension fund management fees in Peru in which the affiliates could opted out of the default set by the pension policy. It has been noted that knowledge or awareness of risk diversification enable individuals to make better choices of annuities (Lusardi and Mitchell, 19 2010; Hastings et al., 2010; Banks et al., 2015) and correct decisions in retirement (Clark et al., 2011; Agnew and Szykman, 2011). As defined before in section 3, we use the variable Active portfolio management to capture the role of financial literacy on gender gaps. Table 5 exploits the samples obtained between 2013 and 2019 and provides a first insight about how financial literacy, gender, and pension wealth distribution could be linked.9 Overall, we observe that only 6.2% and 4.8% of men and women, respectively, could be considered as financially savvy, which implies that a large majority of people are not opting out from the default pension fund risk choices set up by the regulation. In the bottom half of the distribution of pension funds, only 0.9% and 1.3% of women and men are actively managing their portfolios, yet these shares increase along richer groups in the distribution of pension savings. Thus, the richer the individual, the higher the like- lihood to be financially savvy. For example, 28.2% and 34.7% of women and men belonging to the top 1% share are actively managing their portfolios; these figures are 20.3% and 44.7% for women and men belonging to the top 0.1% share. Furthermore, we observe an increasing difference between the percentage of women and men with financial literacy along the distribu- tion of pension funds. This difference is 1.4 percentage points across all affiliates, but it is 24.3 percentage points for the affiliates located in the top 0.1% of the distribution of pension funds. All in all, women tend to be less financially savvy than men, even among the richer group of affiliates, which could contribute to expand the gender gap in pension wealth. Table 5: Percentage of people with Active portfolio management by pension savings shares (2013-2019) Pension savings share Female Male Diff (M-F) P-value N Overall 4.8 6.2 1.4 0.00 430,698 Bottom 50% 0.9 1.3 0.3 0.00 215,368 P50-90th 7.0 7.4 0.4 0.00 172,280 Top 10% 20.1 22.8 2.7 0.00 43,050 Top 5% 24.7 28.0 3.2 0.00 21,516 Top 1% 28.2 34.7 6.5 0.00 4,299 Top 0.5% 27.4 37.2 9.8 0.00 2,148 Top 0.1% 20.3 44.7 24.3 0.00 429 Notes: The sample is composed of individuals from sampling years 2013, 2015, 2016, and 2019. For each sample year, we compute 1,000 quantiles for the distribution of pension savings, and then we pool all the samples. We use the pooled sample to compute the percentage of people with Active portfolio management within distinctive groups of quantiles, irrespective of the sample year. The pooled sample size is 430,698 observations: N=94,315 in 2013, N=103,399 in 2015, N=108,091 in 2016, and N=124,942 in 2019. Following on our previous models of unconditional quantile regressions, we assess the role of financial literacy on the distribution of wealth and gender gaps. The only difference with 9The pension funds with different risk compositions were implemented in 2006. However, our sample of December 2006 has not available information about the pension fund type chosen by the individual. This is why we use the other available samples drawn between 2013 and 2019. 20 respect to the models of Table 4 is that this time we include two additional covariates in the regressions: the dummy variable Active portfolio management and its interaction with Male. The full results of these regressions are reported in Table A.4 in the Appendix, but we plot our coefficients of interest in Figure 7. We plot the combined coefficients, and their 95% confidence intervals, indicating three distinctive groups: savvy financial males (i.e. males who actively manage their portfolios), savvy financial females, and no savvy financial males. We observe that, in general, financial literacy has stronger effects at higher quantiles, pulling the distribution of pension wealth upwards. In other words, a greater participation of individuals (regardless of gender) with financial knowledge contributes to a greater dispersion of pension wealth, i.e. more inequality. A second observation is that financial savvy males can contribute more than financial savvy females to the dispersion of pension savings (the coefficients curve of financial savvy males is always above the coefficients curve of females in Figure 7). Therefore, the gender gaps in pension savings could expand along the distribution of pension funds. Figure 7: Unconditional quantile coefficients of Active portfolio management 0 .5 1 1 .5 2 2 .5 3 U n c o n d it io n a l q u a n ti le r e g re s s io n c o e ff ic ie n ts 25 50 75 90 95 99 Quantile Financial savvy males Financial savvy females No financial savvy males Notes: The graph plots the UQR coefficients for financial savvy males, financial savvy females and no financial savvy males retrieved from Table A.4 in the Appendix. The vertical bars indicate 95% confidence intervals. 6 Additional results 6.1 Extended pension wealth Some SPP affiliates who were before in the public pension system have Recognition Bonds (RB), which represent past pension contributions made to the public system. These bonds are paid at retirement and introduced into the pension balance of the affiliates in order to compute pension amounts. The total pension wealth of these affiliates should include the BR value. Thus, 21 we use a concept of “extended pension wealth” by adding the updated RB value to the pension balance. We run the same quantile regressions we used before on this new outcome and report the results in Table A.5 in the Appendix. The results are practically the same we obtained in Table 4. 6.2 Imputed income One of the limitations of the registers data is the limited availability of updated wages. For the 2019 sample, about 58% of the individuals have contributed in the year of the sampling draw and therefore they have updated information for their wages, 35% have outdated wage information but have the date of last contribution, and 7% have no wage information nor the date of last contribution. We do not use wage information in our main analysis due to its limited availability, but at least we could attempt to impute and update them to explore the relationship between our results on gender pension wealth gaps and gender income gaps. The procedure to impute monthly earnings for the 2019 sample is as follows: the initial value of the wage is the last value recorded in the sample. In case the value corresponds to any year before 2019, we update the recorded value by inflation and wage premiums per cohort (5-year groups), sex, and contribution behavior.10 The imputation of incomes for the affiliates who have not this information uses the predicted values from a regression of wage (in logs) against sex, recognition bond, decile of contribution density, type of administrative fee, AFP, type of pension risk fund, affiliation duration in the SPP, percentile of pension balance, age, age squared, and region. In the SPP, the contributions are calculated over wages which value must be at least equal to the minimum wage (equal to 930 Soles). Thus, we set up that earnings cannot be lower than the official minimum wage. Table 6 reports the means and percentiles of the distribution of monthly earnings in our 2019 sample. The raw gender gap in earnings is 19%, which is about half the gender gap in pension savings (37%). We do not observe gaps along the first three deciles because of our assumption that affiliates must earn at least the minimum wage. The gender income gap is about 11% at the 40th percentile and increases smoothly until 18% at the 95th percentile. Then, the income gap grows quickly to 31% and 40% at the 99th and 99.5th percentiles, respectively. The values of the gaps in income and pension wealth can greatly differ along the first percentiles and even beyond the media of both distributions, but their values tend to converge towards the top 1% and beyond (see Table 1). At the median of each distribution, the pension wealth gap is 59%, but this is only 15% in incomes; while that at the 90th percentile, the pension wealth gap is 22% and the income gap is 16%. As reported in other studies assessing gender wealth gaps, we also observe a larger gender gap along the distribution of pension wealth than along the distribution of incomes. 10The wage premiums are estimated using the 2015, 2016 and 2019 samples. We estimate the variations in the median wages by sex, birth cohorts, and whether the individual contributed in the sampling year or not. 22 Table 6: Raw gender gaps in earnings and pension savings-to-earnings ratio in 2019 (Soles) Variables Mean Percentile 10 20 30 40 50 60 70 80 90 95 99 99.5 Distribution of earnings: Male 2,445 930 930 930 1,119 1,378 1,649 2,034 2,724 4,421 7,000 17,141 24,643 Female 2,053 930 930 930 1,009 1,200 1,430 1,760 2,335 3,800 5,945 13,106 17,574 Gap (M-F) 393 0 0 0 110 178 219 274 389 621 1,055 4,035 7,069 P-value 0.00 1.00 1.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Men-to-women ratio 1.19 1.00 1.00 1.00 1.11 1.15 1.15 1.16 1.17 1.16 1.18 1.31 1.40 Note: The table uses the sample of registers drawn in December 2019. The table shows the means and percentiles of the distribution of monthly earnings, which include updated and imputed salaries for individuals who did not contribute in 2019 or had missing income information. . We also compute the ratio between pension savings and monthly earnings for each indi- vidual, which indicates the number of income months accumulated in the retirement account. Some studies on income and wealth inequality (e.g. Piketty and Zucman (2014) and Cowell et al. (2017)) use the wealth to income ratio to explore the evolution and country differences of wealth inequality due to changes in asset prices and income returns. The left-hand side panel of Table 7 reports the average of individual pension savings-to-earning ratios by distinctive in- come groups of the distribution of earnings. Overall, men have 2.2 more months of incomes in their pension accounts, yet this gap is 3 months for the individuals who belong to the bottom 50% of the earnings distribution. For the individuals belonging to the top 10%, top 5% and top 1% of the earnings distribution, the gaps are equal to 1.2, 1.8, and 2.2. Note however that these values use the earnings of the corresponding income group, which are higher in the top groups, and hence the level of pension wealth is larger in the top income groups. For example, Table 6 shows that a woman and man in the 99th income percentile earn about 11 and 12 times more than a woman and man in the 50th percentile. In order to capture these differences in income, the right-hand side panel of Table 7 shows the pension savings-to-earning ratios expressed as the mean pension saving of a particular income group over the mean income across all affiliates rather than the mean income for that group. Overall, the gender gap is 3.5 months of average income, but the gap is larger in the top income groups: 10.9, 17,5 and 36.5 months for top 10%, top 5%, and top 1% income shares, respectively. 23 Table 7: Raw gender gaps in pension savings-to-earnings ratio in 2019 (Soles) Using mean earnings of each income group Using overall mean earnings Mean Bottom 50% P50- 90th Top 10% Top 5% Top 1% Mean Bottom 50% P50- 90th Top 10% Top 5% Top 1% Male 10.9 9.6 11.5 14.3 14.6 12.7 12.8 4.2 11.0 57.3 81.4 150.1 Female 8.8 6.6 10.9 13.1 12.8 10.5 9.3 2.9 10.1 46.4 63.9 113.6 Gap (M-F) 2.2 3.0 0.6 1.2 1.8 2.2 3.5 1.3 1.0 10.9 17.5 36.5 P-value 0.00 0.00 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.00 Note: The table uses the sample of registers drawn in December 2019. The left-hand side panel shows the average of individual pension savings-to-earning ratios by distinctive income groups of the distribution of earnings. The right-hand side panel is similar to the other panel, but the pension savings-to-earning ratios is computed as the mean pension saving of a particular income group over the mean income across all affiliates rather than the mean income for that group. . 7 Conclusions Our study uncovers a large gender gap in favour of men in pension savings. Although this gap falls for younger cohorts (because the salary gap is also lower in these groups), the IRA system’s capitalization process may be reversing this improvement, expanding therefore the gap in pension savings across the ilfe-cycle. We also explore gender gaps along the distribution of pension savings and find that the gap favouring men is always positive at each percentile, but it decreases until it reaches a kind of “glass ceiling” around the 85th percentile, where the gap increases substantially. The low levels of financial education captured by the individual’s risk management of pension fund portfolios (i.e. the ability to opt out from default choices of risk-specific pension funds) contribute to the increase of inequality in the distribution of pension funds and on widening the gender pension savings gap. Indeed, pension-savings-rich individuals have higher levels of awareness with risk portfolio management, and among them, financial savvy males contribute more than financially savvy females to increasing inequality in pension savings. Thus, we observe that on one hand, financial literacy is an important determinant of overall pension wealth inequality, and on the other hand, it is also key to explain increasing gender gaps in pension savings. This situation is not helped by the fact that Peru has very low levels of financial literacy, and therefore, policy-makers should rethink about the design of existing default choices in the risk composition of pension fund portfolios. Requiring greater financial knowledge about the returns and risks of pension funds may mostly affect groups with low financial literacy such as individuals employed in low-skilled occupations and women. We also notice that some instruments that can attenuate gender pension savings inequality, such as minimum pension guarantees and unisex life tabes, are absent in Peru’s IRA system. Thus, extending social assistance pension programs and/or setting pension guarantees could help in the short run. Nevertheless, in the long-run there are pending problems for the gender gap in pension wealth and pensions. For example, there is still room to improve the adequacy 24 of benefits, the distribution of household chores between men and women, and the social pro- tection culture, in particular among people who do not work in the formal labour market. Overall, our results could be useful to other countries with IRA systems or countries that are considering increasing the relative importance of these systems in their pension models. These systems can improve the incentive alignment between individual contributions and retirement savings, but one danger is exacerbating gender pension gaps. Furthermore, more need to be done regarding the periods when women are less able to contribute (e.g. due to child birth and childbearing) or in the events of divorce as the IRA are fully individualized and are not part of the shared household wealth. We also consider our paper is useful to investigate a component of wealth distribution (pension savings) that is closely related to earnings savings. Very few developing countries have wealth surveys and are not able to study wealth distribution patterns, but some of them have IRA systems. Thus, a way forward to study wealth distributions with more detail is exploiting data of individual pension accounts as we have done in our paper. We wish our study opens new avenues for future research on gender gaps in wealth and pensions, in particular in developing countries, which still show low levels and high dispersion on financial literacy. 25 References Agnew, J. and Szykman, L. (2011). Annuities, financial literacy and information overload. In Mitchell, O. S. and Lusardi, A., editors, Financial Literacy: Implications for Retirement Security and the Financial Marketplace, volume 101, chapter 9, pages 158–178. Oxford University Press New York. Altamirano-Montoya, Á., Berstein, S., Bosch, M., Huitrón, M. G., and Oliveri, M. L. (2018). Presente y futuro de las pensiones en América Latina y el Caribe. Inter-American Develop- ment Bank. Anglade, B., Useche, P., and Deere, C. D. (2017). Decomposing the gender wealth gap in ecuador. World Development, 96:19–31. Arenas de Mesa, A. (2019). Los sistemas de pensiones en la encrucijada: desafíos para la sostenibilidad en América Latina. CEPAL. Arza, C. (2015). The gender dimensions of pension systems: policies and constraints for the protection of older women. United Nations Women. Bando, R. (2019). Evidence-based gender equality policy and pay in Latin America and the Caribbean: progress and challenges. Latin American Economic Review, 28(1):1–23. Banks, J., Crawford, R., and Tetlow, G. (2015). Annuity choices and income drawdown: evi- dence from the decumulation phase of defined contribution pensions in england. Journal of pension economics & finance, 14(4):412–438. Bernal, N. and Olivera, J. (2020). Choice of pension management fees and effects on pension wealth. Journal of Economic Behavior and Organization, 176:539–568. Bertranou, F. M. (2001). Pension reform and gender gaps in Latin America: What are the policy options? World Development, 29(5):911–923. Bosch, M., Melguizo, A., and Pagés, C. (2013). Mejores pensiones, mejores trabajos: Hacia la cobertura universal en América Latina y el Caribe. Banco Interamericano de Desarrollo, Washingnton, D.C. Choe, C. and Van Kerm, P. (2018). Foreign workers and the wage distribution: what does the influence function reveal? Econometrics, 6(13):41. Clark, R. L., Morrill, M. S., and Allen, S. G. (2011). Pension plan distributions: The impor- tance of financial literacy. In Mitchell, O. S. and Lusardi, A., editors, Financial Literacy: Implications for Retirement Security and the Financial Marketplace, volume 101, chapter 3, pages 40–58. Oxford University Press New York. Cordova, K., Grabka, M. M., and Sierminska, E. (2021). Pension wealth and the gender wealth gap. Technical Report 1141, SOEPpapers on Multidisciplinary Panel Data Research. Cowell, F. A., Nolan, B., Olivera, J., and Van Kerm, P. (2017). Wealth, top incomes, and inequality. In Hamilton, K. and Hepburn, C., editors, National Wealth: What is Missing, Why it Matters, chapter 8, pages 175–203. Oxford University Press, Oxford, UK. 26 Davies, J. B., Fortin, N. M., and Lemieux, T. (2017). Wealth inequality: Theory, measurement and decomposition. Canadian Journal of Economics, 50(5):1224–1261. Firpo, S., Fortin, N. M., and Lemieux, T. (2009). Unconditional quantile regressions. Econo- metrica, 77(3):953–973. Flory, J. (2012). The Gender Pension Gap: Developing an Indicator Measuring Fair Income Op- portunities for Women and Men. Study, Federal Ministry for Family Affairs, Senior Citizens, Women and Youth, Berlin. Fonseca, R., Mullen, K. J., Zamarro, G., and Zissimopoulos, J. (2012). What Explains the Gender Gap in Financial Literacy? The Role of Household Decision Making. Journal of Consumer Affairs, 46(1):90–106. Foster, L. and Smetherham, J. (2013). Gender and Pensions: An Analysis of Factors Affecting Women’s Private Pension Scheme Membership in the United Kingdom. Journal of Aging & Social Policy, 25(3):197–217. Frémeaux, N. and Leturcq, M. (2020). Inequalities and the individualization of wealth. Journal of Public Economics, 184:104145. Hastings, J. S., Mitchell, O. S., and Chyn, E. T. (2010). Fees, Framing, and Financial Literacy in the Choice of Pension Manager. Wharton Pension Research Council Working Papers No. 215, pages 1–26. Klapper, L., Lusardi, A., and Van Oudheusden, P. (2015). Financial literacy around the world: Insights from the standard & poor’s ratings services global financial literacy survey. World Bank. Washington DC: World Bank. Lusardi, A., Michaud, P.-C., and Mitchell, O. S. (2017). Optimal financial knowledge and wealth inequality. Journal of Political Economy, 125(2):431–477. Lusardi, A. and Mitchell, O. (2010). Financial Literacy and Planning: Implications for Retire- ment Wellbeing. Wharton Pension Research Council Working Papers No. 211, pages 1–38. Madero-Cabib, I., Biehl, A., Sehnbruch, K., Calvo, E., and Bertranou, F. (2019). Private pension systems built on precarious foundations: A cohort study of labor-force trajectories in Chile. Research on Aging, 41(10):961–987. Meriküll, J., Kukk, M., and Rõõm, T. (2021). What explains the gender gap in wealth? Evidence from administrative data. Review of Economics of the Household, 19(2):501–547. Piketty, T. and Zucman, G. (2014). Capital is back: Wealth-income ratios in rich countries 1700–2010. The Quarterly Journal of Economics, 129(3):1255–1310. Schneebaum, A., Rehm, M., Mader, K., and Hollan, K. (2018). The gender wealth gap across European countries. Review of Income and Wealth, 64(2):295–331. Sierminska, E., Piazzalunga, D., and Grabka, M. M. (2019). Transitioning towards more equal- ity? Wealth gender differences and the changing role of explanatory factors over time. IZA Discussion Papers Series No. 12404. 27 Appendix Figure A.1: Raw gender gaps in pension savings at specific percentiles of the pension savings distribution (2005-2019) 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 M en -to -w om en ra tio 50th 90th 95th 99th 99.5th Note: The figure plots the raw ratio of men’s pension savings to women’s pension savings at distinctive percentiles of the distribution of pension funds of each available sample year. A1 Figure A.2: Histograms of contribution density by gender (2015-2019) 0 1 2 3 4 5 6 7 0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1 female male F re q u e n c y ( p e rc e n ta g e ) Contribution density (0 to 1) A2 Figure A.3: Cohort-specific gender gap across the unconditional distribution of pension balance (2019) 0 2 0 4 0 6 0 8 0 1 0 0 G e n d e r g a p ( 0 0 0 ’s S o le s ) 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile Cohort 1959−1961 0 1 0 2 0 3 0 4 0 5 0 G e n d e r g a p ( 0 0 0 ’s S o le s ) 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile Cohort 1969−1971 − .5 0 .5 1 1 .5 G e n d e r g a p ( 0 0 0 ’s S o le s ) 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile Cohort 1979−1981 − .2 0 .2 .4 .6 G e n d e r g a p ( 0 0 0 ’s S o le s ) 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile Cohort 1989−1991 Note: The figures show the lowess-smoothed gender gaps across the unconditional distribution of pension balance for each cohort in 2019. A3 Figure A.4: Cohort-specific gender gap across the unconditional pension fund distribution in various years 0 5 1 0 1 5 2 0 G e n d e r g a p ( 0 0 0 ’s S o le s ) 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile 2005 0 1 0 2 0 3 0 4 0 G e n d e r g a p ( 0 0 0 ’s S o le s ) 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile 2013 0 1 0 2 0 3 0 4 0 5 0 G e n d e r g a p ( 0 0 0 ’s S o le s ) 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile 2016 0 1 0 2 0 3 0 G e n d e r g a p ( 0 0 0 ’s S o le s ) 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Pension fund percentile 2019 Note: The figures show the lowess-smoothed gender gaps across the unconditional distribution of pension balance for each year. A4 Table A.1: Additional descriptives (1) (2) (3) (4) (5) (6) 2005 2006 2013 2015 2016 2019 Pension balance 9,261.02 14,580.27 20,105.94 22,376.41 24,292.47 26,239.68 (105.14) (181.62) (184.16) (193.65) (229.78) (202.43) Pension balance + RB 12,560.84 17,689.31 21,497.46 23,488.38 25,262.47 26,676.40 (153.45) (222.42) (202.49) (208.09) (242.91) (208.56) Age 37.33 37.69 38.07 38.47 38.50 38.33 (0.04) (0.04) (0.03) (0.03) (0.03) (0.03) Years enrolled in SPP 7.22 8.22 10.83 11.64 11.92 12.30 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Regular contributor 0.57 0.61 0.62 0.59 0.58 0.58 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Contribution density - - - 0.48 0.49 0.48 (0.00) (0.00) (0.00) Recognition Bond (RB) 0.11 0.10 0.04 0.03 0.03 0.02 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) RB value 3,300 3,109 1,392 1,112 970 437 (68.09) (65.84) (34.24) (28.79) (26.83) (15.69) Active portfolio management - - 0.06 0.06 0.05 0.06 (0.00) (0.00) (0.00) (0.00) Notes: The table shows the main descriptive statistics of our sample. A5 Table A.2: Unconditional quantile regressions of pension balance (2005, 2013 and 2019) Variables OLS Q25 Q50 Q75 Q90 Q95 Q99 2005: Male -0.00079 -0.0278 -0.0984*** -0.0539*** 0.204*** 0.256*** 0.361*** (0.0129) (0.0253) (0.0185) (0.0139) (0.0238) (0.0291) (0.0445) Regular contributor 1.391*** 2.010*** 1.811*** 0.792*** 0.618*** 0.509*** 0.281*** (0.0130) (0.0255) (0.0191) (0.0135) (0.0219) (0.0260) (0.0381) Years enrolled in SPP 0.640*** 1.198*** 0.792*** 0.0724*** -0.0384*** -0.124*** -0.184*** (0.00861) (0.0172) (0.0104) (0.00752) (0.0124) (0.0149) (0.0224) Years enrolled in SPP^2 -3.050*** -6.674*** -3.810*** 0.682*** 1.298*** 1.770*** 1.942*** (0.0621) (0.117) (0.0813) (0.0627) (0.106) (0.129) (0.197) Recognition Bond 1.158*** 0.666*** 1.103*** 1.388*** 2.440*** 2.751*** 2.687*** (0.0218) (0.0290) (0.0271) (0.0301) (0.0703) (0.0983) (0.176) Constant 3.974*** 0.690*** 3.410*** 7.505*** 8.310*** 9.150*** 11.11*** (0.163) (0.246) (0.182) (0.154) (0.336) (0.461) (0.949) 2013: Male 0.0950*** 0.133*** 0.0944*** -0.0531*** 0.0592*** 0.119*** 0.227*** (0.00915) (0.0169) (0.0129) (0.0127) (0.0134) (0.0216) (0.0314) Regular contributor 1.779*** 2.321*** 1.934*** 1.502*** 0.836*** 0.926*** 0.527*** (0.00984) (0.0176) (0.0133) (0.0131) (0.0134) (0.0209) (0.0294) Years enrolled in SPP 0.253*** 0.418*** 0.396*** 0.105*** -0.00887** -0.0398*** -0.0967*** (0.00355) (0.00720) (0.00473) (0.00438) (0.00435) (0.00670) (0.00965) Years enrolled in SPP^2 -0.440*** -1.129*** -0.940*** 0.299*** 0.499*** 0.661*** 0.746*** (0.0162) (0.0302) (0.0218) (0.0220) (0.0235) (0.0372) (0.0558) Recognition Bond 1.222*** 0.423*** 0.831*** 1.633*** 2.545*** 4.532*** 5.013*** (0.0234) (0.0278) (0.0250) (0.0348) (0.0646) (0.135) (0.275) Constant 3.617*** -0.156 4.083*** 6.863*** 9.346*** 10.17*** 12.80*** (0.0827) (0.121) (0.0920) (0.104) (0.138) (0.279) (0.633) 2019: Male 0.174*** 0.225*** 0.169*** 0.0933*** 0.0472*** 0.138*** 0.187*** (0.00797) (0.0152) (0.0117) (0.0110) (0.0115) (0.0156) (0.0272) Regular contributor 1.815*** 2.192*** 2.101*** 1.606*** 0.962*** 0.884*** 0.648*** (0.00841) (0.0152) (0.0122) (0.0117) (0.0119) (0.0160) (0.0262) Years enrolled in SPP 0.335*** 0.548*** 0.425*** 0.162*** 0.0219*** 0.0195*** -0.0180** (0.00249) (0.00489) (0.00312) (0.00303) (0.00324) (0.00446) (0.00719) Years enrolled in SPP^2 -0.663*** -1.386*** -0.918*** -0.0141 0.340*** 0.328*** 0.408*** (0.00909) (0.0168) (0.0120) (0.0123) (0.0139) (0.0194) (0.0329) Recognition Bond 1.250*** 0.502*** 0.831*** 1.466*** 2.608*** 4.077*** 7.155*** (0.0295) (0.0334) (0.0302) (0.0407) (0.0852) (0.164) (0.460) Constant 4.199*** 1.234*** 3.427*** 6.598*** 9.262*** 9.945*** 12.72*** (0.0672) (0.113) (0.0833) (0.0862) (0.122) (0.200) (0.580) Notes: Bootstrapped standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. N = 49,448 for 2005, N = 94,315 for 2013, and N = 124,829 for 2019. All regressions control for AFP, and birth cohort. A6 Table A.3: Results of the detailed decomposition of the gender gap in pension balance over quantiles (2019) Q10 Q25 Q50 Q75 Q90 Q95 Q99 Male 6.086*** 7.479*** 8.973*** 10.23*** 11.12*** 11.65*** 12.69*** (0.0139) (0.0112) (0.00956) (0.00825) (0.00858) (0.0115) (0.0190) Female 5.616*** 6.998*** 8.511*** 9.881*** 10.92*** 11.37*** 12.39*** (0.0172) (0.0140) (0.0124) (0.0121) (0.0106) (0.0121) (0.0229) Gap 0.470*** 0.481*** 0.463*** 0.353*** 0.196*** 0.277*** 0.300*** (0.0221) (0.0179) (0.0156) (0.0146) (0.0137) (0.0167) (0.0298) Explained gap Total 0.209*** 0.256*** 0.268*** 0.211*** 0.141*** 0.127*** 0.0834*** (0.0113) (0.0123) (0.0121) (0.00967) (0.00776) (0.00866) (0.0103) Contribution density -0.0145* -0.0159* -0.0159* -0.0121* -0.00804* -0.00747* -0.00520* (0.00786) (0.00866) (0.00862) (0.00658) (0.00437) (0.00406) (0.00284) Years enrolled in SPP 0.819*** 0.927*** 0.683*** 0.267*** 0.0745*** 0.0592*** 0.00503 (0.0271) (0.0284) (0.0209) (0.00978) (0.00775) (0.0108) (0.0189) Years enrolled in SPP^2/100 -0.534*** -0.592*** -0.382*** -0.0508*** 0.0686*** 0.0742*** 0.102*** (0.0201) (0.0203) (0.0133) (0.00526) (0.00665) (0.00931) (0.0163) Recognition Bond -0.000398 -0.000140 0.000275* 0.00127* 0.00303* 0.00521* 0.00925* (0.000257) (0.000123) (0.000165) (0.000695) (0.00165) (0.00284) (0.00504) Fund type 2 -0.0137*** -0.00782*** 0.000430 0.0173*** 0.0285*** 0.0346*** 0.0620*** (0.00388) (0.00266) (0.00188) (0.00206) (0.00283) (0.00383) (0.00681) Fund type 3 0.00262 0.00203 0.00292*** 0.00200** 0.000789 0.00252 -0.0181*** (0.00203) (0.00141) (0.00104) (0.000956) (0.00118) (0.00170) (0.00344) Unexplained gap Total 0.261*** 0.225*** 0.195*** 0.142*** 0.0552*** 0.151*** 0.216*** (0.0201) (0.0140) (0.0102) (0.0102) (0.0114) (0.0153) (0.0291) Contribution density -0.0113 -0.0293 -1.35e-05 -0.237*** 0.0172 0.194*** 0.0636 (0.0273) (0.0190) (0.0139) (0.0138) (0.0155) (0.0208) (0.0397) Years enrolled in SPP 0.602*** 0.0348 -1.053*** -1.034*** 0.199** 0.287*** 0.639*** (0.145) (0.100) (0.0735) (0.0728) (0.0821) (0.110) (0.210) Years enrolled in SPP^2/100 -0.263*** 0.0329 0.545*** 0.272*** -0.387*** -0.294*** -0.586*** (0.0868) (0.0603) (0.0442) (0.0439) (0.0493) (0.0657) (0.126) Recognition Bond 0.000246 -0.00106 0.000235 0.00131 0.00243* 0.0126*** 0.00118 (0.00258) (0.00179) (0.00131) (0.00131) (0.00146) (0.00200) (0.00374) Fund type 2 0.214 0.0635 0.199 0.426*** -0.161 0.0855 -2.199*** (0.261) (0.181) (0.133) (0.132) (0.148) (0.196) (0.378) Fund type 3 0.0103 0.00707 0.00748 0.00944* -0.000118 0.0124* -0.0861*** (0.00984) (0.00683) (0.00501) (0.00499) (0.00557) (0.00740) (0.0144) Constant -2.384* -2.294** -0.104 0.240 0.484 -0.703 3.952** (1.349) (0.939) (0.694) (0.709) (0.754) (0.960) (1.934) Notes: The regressions use the Blinder-Oaxaca decomposition method, based on RIF. Bootstrapped standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. N = 75,961 for group 1 (male), N = 48,868 for group 2 (female). All regressions control for AFP, birth cohort, and region. A7 Table A.4: Unconditional quantile regression coefficients on logs of pension savings, including financial literacy variables (2019) Variables OLS Q25 Q50 Q75 Q90 Q95 Q99 Male 0.194*** 0.259*** 0.202*** 0.102*** 0.0415*** 0.122*** 0.158*** (0.00804) (0.0157) (0.0119) (0.0110) (0.0111) (0.0147) (0.0251) Active portfolio management 0.901*** 0.470*** 0.942*** 1.465*** 1.384*** 1.637*** 1.317*** (0.0266) (0.0345) (0.0369) (0.0533) (0.0782) (0.121) (0.230) Male*Active portfolio mana. -0.0615* -0.228*** -0.211*** -0.00303 0.302*** 0.683*** 1.055*** (0.0324) (0.0414) (0.0440) (0.0636) (0.0956) (0.152) (0.304) Regular contributor 1.782*** 2.178*** 2.068*** 1.554*** 0.906*** 0.809*** 0.578*** (0.00824) (0.0151) (0.0121) (0.0116) (0.0117) (0.0156) (0.0252) Years enrolled exact 0.323*** 0.540*** 0.412*** 0.147*** 0.00556* -0.00236 -0.0400*** (0.00246) (0.00491) (0.00312) (0.00300) (0.00319) (0.00440) (0.00732) Years enrolled exact^2/100 -0.656*** -1.389*** -0.909*** 0.000870 0.357*** 0.350*** 0.432*** (0.00893) (0.0167) (0.0119) (0.0121) (0.0137) (0.0191) (0.0330) Recognition Bond 1.192*** 0.452*** 0.773*** 1.407*** 2.542*** 3.974*** 7.028*** (0.0290) (0.0338) (0.0307) (0.0416) (0.0847) (0.162) (0.457) Constant 4.922*** 2.178*** 4.083*** 6.995*** 8.684*** 9.592*** 14.14*** (0.496) (0.455) (0.639) (0.637) (0.836) (1.267) (4.283) Observations 124,829 124,829 124,829 124,829 124,829 124,829 124,829 R-squared 0.566 0.325 0.451 0.403 0.266 0.180 0.079 Notes: The regressions use the sample drawn on December 2019. The dependent variable for all regressions is the pension balance in logarithm. All regressions control for birth cohort, AFP, and region. Bootstrapped standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. A8 Table A.5: Unconditional quantile regression coefficients on log of extended pension wealth (2019) Variables OLS Q25 Q50 Q75 Q90 Q95 Q99 Male 0.194*** 0.253*** 0.196*** 0.101*** 0.0546*** 0.152*** 0.202*** (0.00590) (0.0139) (0.00997) (0.00969) (0.0110) (0.0153) (0.0275) Contribution density 3.529*** 4.155*** 4.079*** 3.329*** 2.010*** 1.777*** 1.294*** (0.00923) (0.0196) (0.0137) (0.0146) (0.0189) (0.0266) (0.0456) Years enrolled in SPP 0.375*** 0.598*** 0.473*** 0.202*** 0.0411*** 0.0265*** -0.0120 (0.00203) (0.00460) (0.00297) (0.00290) (0.00321) (0.00444) (0.00729) Years enrolled in SPP^2/100 -0.823*** -1.571*** -1.101*** -0.174*** 0.241*** 0.257*** 0.339*** (0.00733) (0.0159) (0.0110) (0.0111) (0.0133) (0.0188) (0.0324) Recognition Bond 0.959*** -0.0511 0.307*** 1.197*** 3.062*** 4.701*** 8.415*** (0.0296) (0.0417) (0.0373) (0.0365) (0.0762) (0.163) (0.487) Fund type 2 -0.279*** 0.293*** -0.0832 -0.917*** -1.285*** -1.601*** -1.859*** (0.0474) (0.0608) (0.0554) (0.0720) (0.130) (0.231) (0.718) Fund type 3 0.0930* 0.0817 0.188*** 0.0893 -0.0479 0.111 -0.635 (0.0489) (0.0637) (0.0585) (0.0768) (0.138) (0.243) (0.733) Constant 4.356*** 1.343*** 3.180*** 6.565*** 9.494*** 10.57*** 13.90*** (0.0704) (0.169) (0.116) (0.109) (0.135) (0.212) (0.611) Observations 124,829 124,829 124,829 124,829 124,829 124,829 124,829 R-squared 0.752 0.425 0.590 0.533 0.331 0.216 0.093 Notes: The dependent variable for all regressions is the log of extended pension wealth, that is the sum of the pension balance and the updated Recognition Bond. All regressions control for birth cohort, AFP, and region. Bootstrapped standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. A9 ÚLTIMAS PUBLICACIONES DE LOS PROFESORES DEL DEPARTAMENTO DE ECONOMÍA  Libros Elena Álvarez (Editor) 2022 Agricultura y desarrollo rural en el Perú: homenaje a José María Caballero. Lima, Departamento de Economía PUCP. Aleida Azamar Alonso, José Carlos Silva Macher y Federico Zuberman (Editores) 2022 Economía ecológica latinoamericana. Buenos Aires, México. CLACSO, Siglo XXI Editores. Efraín Gonzales de Olarte 2021 Economía regional y urbana. El espacio importa. Lima, Fondo Editorial PUCP. Alfredo Dammert Lira 2021 Economía minera. Lima, Fondo Editorial PUCP. Adolfo Figueroa 2021 The Quality of Society, Volume II – Essays on the Unified Theory of Capitalism. New York, Palgrave Macmillan. Carlos Contreras Carranza (Editor) 2021 La Economía como Ciencia Social en el Perú. Cincuenta años de estudios económicos en la Pontificia Universidad Católica del Perú. Lima, Departamento de Economía PUCP. José Carlos Orihuela y César Contreras 2021 Amazonía en cifras: Recursos naturales, cambio climático y desigualdades. Lima, OXFAM. Alan Fairlie 2021 Hacia una estrategia de desarrollo sostenible para el Perú del Bicentenario. Arequipa, Editorial UNSA. Waldo Mendoza e Yuliño Anastacio 2021 La historia fiscal del Perú: 1980-2020. Colapso, estabilización, consolidación y el golpe de la COVID-19. Lima, Fondo Editorial PUCP. Cecilia Garavito 2020 Microeconomía: Consumidores, productores y estructuras de mercado. Segunda edición. Lima, Fondo Editorial de la Pontificia Universidad Católica del Perú. Adolfo Figueroa 2019 The Quality of Society Essays on the Unified Theory of Capitalism. New York. Palgrave MacMillan. https://departamento.pucp.edu.pe/economia/libro/agricultura-desarrollo-rural-peru-homenaje-jose-maria-caballero/ https://departamento.pucp.edu.pe/economia/libro/10743/ https://departamento.pucp.edu.pe/economia/libro/la-economia-ciencia-social-peru-cincuenta-anos-estudios-economicas-la-pontificia-universidad-catolica-del-peru/ https://departamento.pucp.edu.pe/economia/libro/la-economia-ciencia-social-peru-cincuenta-anos-estudios-economicas-la-pontificia-universidad-catolica-del-peru/ https://departamento.pucp.edu.pe/economia/libro/la-historia-fiscal-del-peru-1980-2020-colapso-estabilizacion-consolidacion-golpe-la-covid-19/ https://departamento.pucp.edu.pe/economia/libro/la-historia-fiscal-del-peru-1980-2020-colapso-estabilizacion-consolidacion-golpe-la-covid-19/ http://departamento.pucp.edu.pe/economia/libro/the-quality-of-society-essays-on-the-unified-theory-of-capitalism/ Carlos Contreras y Stephan Gruber (Eds.) 2019 Historia del Pensamiento Económico en el Perú. Antología y selección de textos. Lima, Facultad de Ciencias Sociales PUCP. Barreix, Alberto Daniel; Corrales, Luis Fernando; Benitez, Juan Carlos; Garcimartín, Carlos; Ardanaz, Martín; Díaz, Santiago; Cerda, Rodrigo; Larraín B., Felipe; Revilla, Ernesto; Acevedo, Carlos; Peña, Santiago; Agüero, Emmanuel; Mendoza Bellido, Waldo; Escobar Arango y Andrés. 2019 Reglas fiscales resilientes en América Latina. Washington, BID. José D. Gallardo Ku 2019 Notas de teoría para para la incertidumbre. Lima, Fondo Editorial de la Pontificia Universidad Católica del Perú. Úrsula Aldana, Jhonatan Clausen, Angelo Cozzubo, Carolina Trivelli, Carlos Urrutia y Johanna Yancari 2018 Desigualdad y pobreza en un contexto de crecimiento económico. Lima, Instituto de Estudios Peruanos. Séverine Deneulin, Jhonatan Clausen y Arelí Valencia (Eds.) 2018 Introducción al enfoque de las capacidades: Aportes para el Desarrollo Humano en América Latina. Flacso Argentina y Editorial Manantial. Fondo Editorial de la Pontificia Universidad Católica del Perú. Mario Dammil, Oscar Dancourt y Roberto Frenkel (Eds.) 2018 Dilemas de las políticas cambiarias y monetarias en América Latina. Lima, Fondo Editorial de la Pontificia Universidad Católica del Perú. http://departamento.pucp.edu.pe/economia/libro/las-alianzas-publico-privadas-app-en-el-peru-beneficios-y-riesgos/ http://departamento.pucp.edu.pe/economia/libro/las-alianzas-publico-privadas-app-en-el-peru-beneficios-y-riesgos/ http://departamento.pucp.edu.pe/economia/libro/las-alianzas-publico-privadas-app-en-el-peru-beneficios-y-riesgos/  Documentos de trabajo No. 512 Poder de mercado, bienestar social y eficiencia en la industria microfinanciera regulada en el Perú. Giovanna Aguilar y Jhonatan Portilla. Junio 2022. No. 511 Perú 1990-2020: Heterogeneidad estructural y regímenes económicos regionales ¿Persiste la desconexión entre la economía, la demografía y la geografía? Félix Jiménez y Marco Arroyo. Junio 2022. No. 510 Evolution of the Exchange Rate Pass-Throught into Prices in Peru: An Empirical Application Using TVP-VAR-SV Models. Roberto Calero, Gabriel Rodríguez y Rodrigo Salcedo Cisneros. Mayo 2022. No. 509 Time Changing Effects of External Shocks on Macroeconomic Fluctuations in Peru: Empirical Application Using Regime-Switching VAR Models with Stochastic Volatility. Paulo Chávez y Gabriel Rodríguez. Marzo 2022. No. 508 Time Evolution of External Shocks on Macroeconomic Fluctuations in Pacific Alliance Countries: Empirical Application using TVP-VAR-SV Models. Gabriel Rodríguez y Renato Vassallo. Marzo 2022. No. 507 Time-Varying Effects of External Shocks on Macroeconomic Fluctuations in Peru: An Empirical Application using TVP-VARSV Models. Junior A. Ojeda Cunya y Gabriel Rodríguez. Marzo 2022. No. 506 La Macroeconomía de la cuarentena: Un modelo de dos sectores. Waldo Mendoza, Luis Mancilla y Rafael Velarde. Febrero 2022. No. 505 ¿Coexistencia o canibalismo? Un análisis del desplazamiento de medios de comunicación tradicionales y modernos en los adultos mayores para el caso latinoamericano: Argentina, Colombia, Ecuador, Guatemala, Paraguay y Perú. Roxana Barrantes Cáceres y Silvana Manrique Romero. Enero 2022. No. 504 “Does the Central Bank of Peru Respond to Exchange Rate Movements? A Bayesian Estimation of a New Keynesian DSGE Model with FX Interventions”. Gabriel Rodríguez, Paul Castillo B. y Harumi Hasegawa. Diciembre, 2021 No. 503 “La no linealidad en la relación entre la competencia y la sostenibilidad financiera y alcance social de las instituciones microfinancieras reguladas en el Perú”. Giovanna Aguilar y Jhonatan Portilla. Noviembre, 2021. No. 502 “Approximate Bayesian Estimation of Stochastic Volatility in Mean Models using Hidden Markov Models: Empirical Evidence from Stock Latin American Markets”. Carlos A. Abanto-Valle, Gabriel Rodríguez, Luis M. Castro Cepero y Hernán B. Garrafa-Aragón. Noviembre, 2021. No. 501 “El impacto de políticas diferenciadas de cuarentena sobre la mortalidad por COVID-19: el caso de Brasil y Perú”. Angelo Cozzubo, Javier Herrera, Mireille Razafindrakoto y François Roubaud. Octubre, 2021. No. 500 “Determinantes del gasto de bolsillo en salud en el Perú”. Luis García y Crissy Rojas. Julio, 2021. No. 499 “Cadenas Globales de Valor de Exportación de los Países de la Comunidad Andina 2000-2015”. Mario Tello. Junio, 2021. No. 498 “¿Cómo afecta el desempleo regional a los salarios en el área urbana? Una curva de salarios para Perú (2012-2019)”. Sergio Quispe. Mayo, 2021. No. 497 “¿Qué tan rígidos son los precios en línea? Evidencia para Perú usando Big Data”. Hilary Coronado, Erick Lahura y Marco Vega. Mayo, 2021. No. 496 “Reformando el sistema de pensiones en Perú: costo fiscal, nivel de pensiones, brecha de género y desigualdad”. Javier Olivera. Diciembre, 2020. No. 495 “Crónica de la economía peruana en tiempos de pandemia”. Jorge Vega Castro. Diciembre, 2020. No. 494 “Epidemia y nivel de actividad económica: un modelo”. Waldo Mendoza e Isaías Chalco. Setiembre, 2020. No. 493 “Competencia, alcance social y sostenibilidad financiera en las microfinanzas reguladas peruanas”. Giovanna Aguilar Andía y Jhonatan Portilla Goicochea. Setiembre, 2020. No. 492 “Empoderamiento de la mujer y demanda por servicios de salud preventivos y de salud reproductiva en el Perú 2015-2018”. Pedro Francke y Diego Quispe O. Julio, 2020. No. 491 “Inversión en infraestructura y demanda turística: una aplicación del enfoque de control sintético para el caso Kuéalp, Perú”. Erick Lahura y Rosario Sabrera. Julio, 2020. No. 490 “La dinámica de inversión privada. El modelo del acelerador flexible en una economía abierta”. Waldo Mendoza Bellido. Mayo, 2020. No. 489 “Time-Varying Impact of Fiscal Shocks over GDP Growth in Peru: An Empirical Application using Hybrid TVP-VAR-SV Models”. Álvaro Jiménez y Gabriel Rodríguez. Abril, 2020. No. 488 “Experimentos clásicos de economía. Evidencia de laboratorio de Perú”. Kristian López Vargas y Alejandro Lugon. Marzo, 2020. No. 487 “Investigación y desarrollo, tecnologías de información y comunicación e impactos sobre el proceso de innovación y la productividad”. Mario D. Tello. Marzo, 2020. No. 486 “The Political Economy Approach of Trade Barriers: The Case of Peruvian’s Trade Liberalization”. Mario D. Tello. Marzo, 2020. No. 485 “Evolution of Monetary Policy in Peru. An Empirical Application Using a Mixture Innovation TVP-VAR-SV Model”. Jhonatan Portilla Goicochea y Gabriel Rodríguez. Febrero, 2020. No. 484 “Modeling the Volatility of Returns on Commodities: An Application and Empirical Comparison of GARCH and SV Models”. Jean Pierre Fernández Prada Saucedo y Gabriel Rodríguez. Febrero, 2020. No. 483 “Macroeconomic Effects of Loan Supply Shocks: Empirical Evidence”. Jefferson Martínez y Gabriel Rodríguez. Febrero, 2020. No. 482 “Acerca de la relación entre el gasto público por alumno y los retornos a la educación en el Perú: un análisis por cohortes”. Luis García y Sara Sánchez. Febrero, 2020. No. 481 “Stochastic Volatility in Mean. Empirical Evidence from Stock Latin American Markets”. Carlos A. Abanto-Valle, Gabriel Rodríguez y Hernán B. Garrafa-Aragón. Febrero, 2020. No. 480 “Presidential Approval in Peru: An Empirical Analysis Using a Fractionally Cointegrated VAR2”. Alexander Boca Saravia y Gabriel Rodríguez. Diciembre, 2019. No. 479 “La Ley de Okun en el Perú: Lima Metropolitana 1971 – 2016.” Cecilia Garavito. Agosto, 2019. No. 478 “Peru´s Regional Growth and Convergence in 1979-2017: An Empirical Spatial Panel Data Analysis”. Juan Palomino y Gabriel Rodríguez. Marzo, 2019.  Materiales de Enseñanza No. 5 “Matemáticas para Economistas 1”. Tessy Váquez Baos. Abril, 2019. No. 4 “Teoría de la Regulación”. Roxana Barrantes. Marzo, 2019. No. 3 “Economía Pública”. Roxana Barrantes, Silvana Manrique y Carla Glave. Marzo, 2018. No. 2 “Macroeconomía: Enfoques y modelos. Ejercicios resueltos”. Felix Jiménez. Marzo, 2016. No. 1 “Introducción a la teoría del Equilibrio General”. Alejandro Lugon. Octubre, 2015. Departamento de Economía - Pontificia Universidad Católica del Perú Av. Universitaria 1801, San Miguel, 15008 – Perú Telf. 626-2000 anexos 4950 - 4951 http://departamento.pucp.edu.pe/economia/ DDD513-Carátula DDD513-Segunda hoja DDD513-Contratapa DDD513-Abstract y texto GenderGapPensionsPeru_v3 GenderGapPensionsPeru_SpanishAbstract.pdf GenderGapPensionsPeru_v3.pdf DDD513-ultimas publicaciones