Gender inequality in Latin America and the Caribbean

Berniell, Inés; Fernández, Raquel; Krutikova, Sonya Working Paper Gender inequality in Latin America and the Caribbean IDB Working Paper Series, No. IDB-WP-01553 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Berniell, Inés; Fernández, Raquel; Krutikova, Sonya (2023) : Gender inequality in Latin America and the Caribbean, IDB Working Paper Series, No. IDB-WP-01553, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0005328 This Version is available at: https://hdl.handle.net/10419/299464 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by-nc-nd/3.0/igo/legalcode Gender Inequality in Latin America and the Caribbean Inés Berniell Raquel Fernández Sonya Krutikova IDB WORKING PAPER SERIES No IDB-WP-01553 Inter-American Development Bank December, 2023 Gender Inequality in Latin America and the Caribbean Inés Berniell Raquel Fernández Sonya Krutikova Inter-American Development Bank December, 2023 Cataloging-in-Publication data provided by the Inter -American Development Bank Felipe Herrera Library Berniell, Inés. Gender inequality in Latin America and the Caribbean / Inés Berniell, Raquel Fernández, Sonya Krutikova. p. cm. — (IDB Working Paper Series ; 1553) Includes bibliographical references. 1. Women -Education-Latin America. 2. Women-Education-Caribbean Area. 3. WomenEmployment -Latin America. 4. Women-Employment-Caribbean Area. 5. Equity-Latin America. 6. Equity -Caribbean Area. I. Fernández, Raquel. II. Krutikova, Sonya. III. Inter -American Development Bank. Vice Presidency for Sectors and Knowledge. IV. Title. V. Serie. IDB -WP-1553 JEL: I24, I25, J16, J24. Keywords: Education and Inequality; Education and Economic Development; Economics of Gender ; Human Capital; Skills; Labor Productivity http://www.iadb.org Copyright © 2023 Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 AttributionNonCommercial -NoDerivatives (CC-IGO BY-NC-ND 3.0 IGO) license (http://creativecommons.org/licenses/by-ncnd/3.0/igo/legalcode ) and may be reproduced with attribution to the IDB and for any non-commercial purpose, as provided below. No derivative work is allowed. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration purs uant to the UNCITRAL rules. The use of the IDB’s name for any purpose other than for attribution, and the use of IDB’s logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this CC -IGO license. Following a peer review process, and with previous written consent by the Inter -American Development Bank (IDB), a revised version of this work may also be reproduced in any academic journal, including those indexed by the American Economic Association’s EconLit, provided that the IDB is credited and that the author(s) receive no income from the publication. There fore, the restriction to receive income from such publication shall only extend to the publication’s author(s). With regard to such restriction, in case of any inconsistency between the Creative Commons IGO 3.0 Attribution -NonCommercialNoDerivatives license and these statements, the latter shall prevail. Note that link provided above includes additional terms and conditions of the license. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter -American Development Bank, its Board of Directors, or the countries they represent. Gender Inequality in Latin America and the Caribbean* In´es Berniell CEDLAS Raquel Fern´andez NYU Sonya Krutikova IFS 2023 1 Introduction This chapter examines gender inequality focusing on two critical spheres in which gender inequality is generated: education and work. Our objective is to provide a current snapshot of gender inequality across key indicators as well as a dynamic perspective that highlights successes and failures. We facilitate a cross-country comparison as well by grouping countries within Latin America by their level of economics development and drawing comparisons with countries outside the region. Finally, we reflect on differences in the ways that gender inequalities play out across different socio-economic groups, particularly those that highlight other sources of inequality. Following a life-cycle approach, we start by examining gender inequalities in both quantity of education and performance at different levels of schooling – from pre-school to higher education. The picture that emerges is mixed. In some important dimensions girls are more advantaged than boys. Girls are more likely to complete secondary and higher education and throughout schooling outperform boys in reading. A key dimension in which they continue to be disadvantaged, however, is performance in math and the likelihood of completing a degree in a STEM subject. Instead, women are over-represented in fields such as Health and Education. This is significant because labor market returns to STEM subjects are much higher than those to Health or Education. Importantly, gender gaps in math are not evident at early stages of primary school and the cross-country variation in gender gaps suggests that environmental factors are at play. This is consistent with evidence we present of gender gaps in confidence in math ability in favor of boys, even controlling for differences in achievement, as well as a large gender gap in aspirations for STEM occupations. Notably, however, the broad consensus around the region among adults is that men and women have the same capacity for science and technology. The second part of the chapter focuses on the worksphere. Here we document significant improvements in female labor force participation over the last 20 years, especially among the least-educated women (those with incomplete secondary education). However, progress has not been equal across all the countries in the region – the pace of improvement in this dimension has been slowest in the least economically developed countries. These are also the countries where a significant proportion of the adult working population, especially among men, continue to hold highly conservative norms about *We are grateful to seminar participants at the LACIR workshop for helpful discussion and to our discussant Florencia Torche for her suggestions. We thank Jessica Bracco, Florencia Pinto, and Juli´an Pedrazzi for excellent research assistance. 1 women’s participation in work. Honing in on the working population, we continue to see persistent significant differences in the quality of jobs men and women do, the remuneration received, as well as the balance between paid market and unpaid non-market work. Even after conditioning on education, employment sector, and occupation, we see a large gender wage gap in favor of men in most countries in the region which has remained constant or even increased over the last 20 years. Furthermore, we see that gender inequalities interact with socio-economic inequalities in the work sphere. For example, gender gaps in indicators of job quality (working in the informal sector and working for a larger firm) are significantly greater among the least-educated group than the rest. Finally, across all of the countries in the region it continues to be the case that only women’s labor market trajectories (not men’s) are hugely impacted by the arrival of children, resulting in declines in labor force participation of up to 40%, with little evidence of recovery in the medium term. This chapter proceeds as follows. Section 2focuses on the education sphere, starting with access to education from the pre-primary to tertiary levels and moving on to consider quality of education and education related attitudes and expectations. The focus of Section 3is on the work sphere. Here we consider participation in work, time spent on work, quality of jobs, remuneration for work, as well as factors that drive the large gender wage gaps that we observe. Section 3.6 concludes. 2 Education In this part of the paper we focus on gender inequality in education at each of the key stages - preprimary, primary, secondary, and tertiary. We consider several dimensions of education including quantity, achievement, and attitudes which may shape educational choices. We utilize GenLAC–CEDLAS harmonized microdata from national household surveys from more than 300 household surveys in the region to compute statistics as well as several sources of internationally comparable assessment data including SERCE, TERCE, PISA, and ECAF.1PISA is the OECD’s Program for International Student Assessment. It evaluates students’ knowledge and skills as they approach the end of their compulsory schooling (at 15 years of age). SERCE (Second Regional Comparative and Explanatory Study, 2006) and TERCE (Third Regional Comparative and Explanatory Study) are regional exams for primary education (third and sixth grades) produced by UNESCO. Nearly every country in Latin America took part in TERCE. The CAF Survey (ECAF) 2015 was carried out by CAF-development bank of Latin America, and has information about adult’s skills in 10 major cities in 10 LAC countries. Table B.1 in Appendix B.2 lists the surveys used in this chapter for 17 Latin American countries. For more details, see Appendix B. We classified countries into three broad income groups based on their GNI per capita (in US$) for the period 2010-2020 and the corresponding World Bank Analytical Classifications: (1) Lower-middle income (LMI), which we defined as countries that were considered LMI at least once in 2010-2020; (2) Upper-middle income (UMI), which we defined as countries that were considered UMI during the entire 2010-2020 period; (3) High income (HI), which we defined as countries that were considered HI at least once in 2010-2020. Note that whenever we show trends for a country group, it consists of the simple (unweighted) average across countries in that group. 1GenLAC is the CEDLAS (Center for Distributive, Labor and Social Studies) initiative to promote gender equity in Latin America and the Caribbean. 2 2.1 The Gender Gap in Education: Quantity This section examines the gender gap in educational attainment as measured by completion of primary, secondary, and tertiary education as well as by attendance of a pre-school program. 2.1.1 Pre-School We start with a snapshot of pre-primary educational attainment, which we measure as the proportion of children age 5 attending pre-school in 2019. There are stark differences in enrollment rates across LA countries: at around 95%, enrollment rates in HI countries are comparable to OECD enrollment rates, but they are significantly lower in LMI countries, at around 70%. Irrespective of levels of enrollment, however, in the great majority of LA countries, as in the OECD, boys and girls are equally likely to be attending pre-school (Figure 1). 2This has been the case for some time, especially in the wealthier countries. Figure 2shows the evolution of pre-school enrollment rates for boys and girls over the last 20 years; with the exception of a small gender gap in favor of girls in LMIs in the early 2000’s, we see gender parity across the period. Figure 1: Pre-school Attendance Among Children Aged 5 96.8 97.5 96.9 95.8 89.4 88.1 99.0 99.8 95.5 95.3 95.6 95.6 0 20 40 60 80 100 Preschool attendance (% 5) ARG CHL PAN URY Average OECD Men Women Men OECD Women OECD (a) High income 94.995.9 84.088.3 98.799.3 79.475.9 88.484.5 97.297.4 98.898.7 91.691.4 0 20 40 60 80 100 Preschool attendance (% 5) BRA COL CRI DOM ECU MEX PER Average Men Women (b) Upper middle income 83.379.1 31.930.8 81.881.7 67.569.7 83.981.9 70.974.0 69.969.5 0 20 40 60 80 100 Preschool attendance (% 5) BOL GTM HND NIC PRY SLV Average Men Women (c) Lower middle income Note: The figure shows, by gender, the share of children age 5 attending pre-school in 2019 or the latest year prior to that (see Table B.1 in Appendix B.2). The average bars show unweighted means. The OECD average is the simple average over the OECD countries’ enrollment rates for individuals aged 5 years old in 2019. Source: authors’ own calculations based on household surveys (GenLAC) and Education at a Glance for the OECD. 2.1.2 Primary School In the majority of LA countries there is also little evidence of gender differences in the completion of primary-school education. Figure 3shows primary school completion rates for individuals who were between 20 and 30 years old in 2019. The absence of gender gaps in primary school enrolment is similar to what we see outside the region across OECD countries. Where there are gender gaps, they tend to be small and in favor of girls - for example, in Brazil, Dominican Republic, Nicaragua, and Honduras. Guatemala stands out as an exception with an 8 percentage point gap in favor of boys. As with pre-primary education, the picture has remained fairly constant over the last 20 years, especially in the wealthier countries (Figure 4). In LMI countries the average showed a small gender gap in favor of boys in the early 2000’s, which closed entirely over the subsequent years. 2Exceptions to this gender parity are the Dominican Republic, Ecuador, and Bolivia. In Colombia, the gap favors girls. 3 Figure 2: Evolution of Pre-school Attendance Among Children Aged 5, 2000-2019 40 60 80 100 Preschool attendance (% 5) 2000-2004 2005-2009 2010-2014 2015-2019 Years Women, LMI Women, UMI Women, HI Men, LMI Men, UMI Men, HI Note: The figure shows, by gender and country income group, the evolution of the share of children age 5 attending pre-school. Each dot represents the (unweighted) cross-country average of their 5-year average. All countries with available data in the corresponding sub-periods are included (the panel is unbalanced for LMI countries). See Table B.1 in Appendix B.2). Source: authors’ own calculations based on household surveys (GenLAC). Figure 3: Primary School Completion 95.8 97.7 97.1 98.0 96.2 97.0 97.7 98.6 96.7 97.8 99.2 99.5 0 20 40 60 80 100 Primary completion (%) ARG CHL PAN URY Average OECD Men Women Men OECD Women OECD (a) High income 81.9 88.5 94.5 96.5 95.1 96.3 81.0 88.0 95.9 96.4 97.3 95.5 91.7 93.9 0 20 40 60 80 100 Primary completion (%) BRA COL CRI DOM MEX PER Average Men Women (b) Upper middle income 95.0 92.2 82.1 87.1 73.8 73.8 73.7 81.8 72.0 63.8 93.1 94.4 81.6 82.2 0 20 40 60 80 100 Primary completion (%) BOL HND SLV NIC GTM PRY Average Men Women (c) Lower middle income Note: This figure shows, by gender, the primary school completion rates. For LAC, the sample is restricted to individuals aged 20-30 years old. The average bars show unweighted means. The OECD average is the simple average over the OECD countries’ primary completion rates for individuals aged 14-16 years old. The year is 2019 or the latest available up to that year (see Table B.1 in Appendix B.2.) Source: authors’ own calculations based on household surveys (GenLAC) and UIS–UNESCO for the OECD. 2.1.3 Secondary School The gender parity that we see in pre and primary school completion rates does not extend to secondary education. Figure 5shows secondary school completion rates for individuals age 20-30 in 2019. Across the large majority of countries we see gaps in favor of women.These are largest in the high income countries; on average women here are 8 percentage points more likely to have completed secondary school than men, significantly greater than the 3.4 percentage points gap of OECD countries. By far the largest gap is in Uruguay at 15 percentage points. Despite being in the HI country group, Uruguay also has among the lowest secondary school completion rates in the region, especially for men, comparable to much poorer countries such as Guatemala, Nicaragua and Honduras. The gaps in favor of women are not a recent phenomenon in the wealthier countries; Figure 6a shows 4 Figure 4: Evolution of Primary School Completion Rates 2000-2019 60 70 80 90 100 Primary completion (%) 2000-2004 2005-2009 2010-2014 2015-2019 Years Women, LMI Women, UMI Women, HI Men, LMI Men, UMI Men, HI Note: The figure shows, by gender and country income group, the evolution of the primary school completion rates. Each dot represents the (unweighted) cross-country average of their 5-year average. Source: see note to Figure 2. Only countries with available data in the corresponding periods are included (unbalanced panel in the case of LMI countries). See Table B.1 in Appendix B.2. that they were already there in HI and UMI countries 20 years earlier. Furthermore, there is no sign of them narrowing over this period - if anything, they have widened slightly. In the poorer countries, the likelihood of completing secondary school was similarly low for women and men in the early 2000’s. Over the last 20 years, however, women have experienced slightly faster-paced improvement than men so that by 2019 these countries are starting to look more similar to the wealthier countries in terms of the gender gap, though not in terms of completion levels. Figure 5: Secondary School Completion (2019) 67.0 75.2 83.9 88.1 61.7 70.8 37.9 49.4 62.6 70.9 84.6 88.0 0 20 40 60 80 100 Secondary completion (%) ARG CHL PAN URY Average OECD Men Women Men OECD Women OECD (a) High income 62.7 72.7 71.1 78.0 54.8 63.8 52.9 67.4 53.7 55.5 84.1 81.3 64.2 69.7 0 20 40 60 80 100 Secondary completion (%) BRA COL CRI DOM MEX PER Average Men Women (b) Upper middle income 75.4 73.7 31.7 39.0 41.6 43.4 36.7 45.7 32.0 29.6 65.7 67.6 47.2 49.9 0 20 40 60 80 100 Secondary completion (%) BOL HND SLV NIC GTM PRY Average Men Women (c) Lower middle income Note: This figure shows, by gender, the secondary school completion rates. The sample is restricted to individuals aged 20-30 years old, except in the case of OECD where it is restricted to individuals aged 25-34 years old. The average bars show unweighted means. Source: authors’ own calculations based on household surveys (GenLAC) and OECD STATS. The year used in the calculations is 2021 in the OECD and 2019 or the latest available up to 2019 in LAC (see Table B.1 in Appendix B.2). 5 and Chile where the math gender gap shrunk by over 10 points (10% of a standard deviation). OECD countries also narrowed their math gender gap over the same time period. Argentina is an exception, with the gender gap in math widening over this time period. While girls are catching up with boys in math, boys are catching up with girls in reading. Figure 13b shows changes in reading gender gaps. These are positive in both periods indicating a persistent girl advantage over that time-span. However, all of the points are on or below the 45 degree line, showing that the female advantage in reading either remained constant or shrunk in all seven countries for which data are available, as in the OECD. Furthermore, the catch up in reading was, on average, of a larger magnitude than the catch-up in math; in four out of the seven countries in the analysis the reading gap shrunk by 12-20% of a standard deviation. Overall the picture over time appears to be one in which gender gaps are shrinking in both math and reading, but boys have decreased their disadvantage in reading more than girls have in math in absolute terms as well as relative to the size of the gap that existed. Figure 13: Change in Gender Gaps in PISA Scores, 2009-2018 ARG BRA CHL COL MEX PER URY OECD -40 -30 -20 -10 0 10 Gap F-M in math, 2018 -40 -30 -20 -10 0 10 Gap F-M in math, 2009 (a) Math ARG BRA CHL COL MEX PER URY OECD -10 0 10 20 30 40 Gap F-M in reading, 2018 -10 0 10 20 30 40 50 Gap F-M in reading, 2009 (b) Reading Note: The figure shows changes in test score gender gaps (F-M) between 2009 and 2018. The test score scale has a standard deviation of 100 points. LAC countries are in orange and the OECD average is in green. Source: authors’ own calculations based on PISA 2009, 2018. 2.2.3 Higher Education We do not have cross-country measures of achievement in higher education in Latin America. We can, however, examine gender differences in the choice of field of study. To do this we use data available from Our World in Data to document the female shares of graduates across main fields of study.3 Figure 14 considers Science, Technology, Engineering, and Mathematics (STEM) subjects only and shows the female proportion of STEM graduates by country. As in many countries around the world, across the region women make up the minority of STEM graduates. Chile stands out as the country with by far the lowest proportion of women graduates in STEM, at around 18%, especially compared to 3See Our World in Data. 12 the other three HI countries in this analysis where this proportion is between 40 and 45%. This figure compares favorably with the OECD average of 31%. Across the majority of UMI and LMI countries, women make up between a third to two-fifths of STEM graduates.4 A more general analysis of female representation across all fields of study (presented in Appendix Figures A.1,A.2, and A.3) shows that, throughout the region. Education and Health stand out as the two fields of study where women are consistently over-represented, making up between 70 and 80% of the graduates across all of the countries included in the analysis. Figure 14: Female Share of STEM Graduates 0 10 20 30 40 50 % of STEM graduates who are female ARG CHL PAN URY Average OECD High income 0 10 20 30 40 50 % of STEM graduates who are female BRA COL CRI DOM ECU MEX Average Upper middle income 0 10 20 30 40 50 % of STEM graduates who are female HND SLV GTM Average Low middle income Note: The figure shows the female share of STEM graduates. The OECD average bar shows unweighted means. Source: Our World In Data. Argentina (2013), Brazil (2014), Chile (2014), Colombia (2014), Dominican Republic (2014), Ecuador (2013), El Salvador (2014), Guatemala (2007), Honduras (2014), Mexico (2012), Panama (2013) and Uruguay (2010). 2.2.4 Adult Skills Lastly, we examine measures of skills in adulthood which capture skills accumulated through the entire education system as well as subsequent work experience. This data was collected by the Development Bank of Latin America (CAF) in 2015 for representative samples of individuals age 15-55 in capital cities of 10 Latin American countries. The survey included assessments of several dimensions of adult skills. Here we examine gender differences in measures of intelligence, verbal skills, and numerical skills. Further details about the data and assessments used in the analysis can be found in Appendix B.5. Figure 15 shows, for each gender, the percent of a standard deviation by which the group on average scores differently than the mean for that variable. Each variable is standardized with a mean of 0 and a standard deviation equal to 1 where the standardization was done using the entire CAF sample of the 10 major cities of the 9 countries depicted (but which also included Venezuela which is not present in our analysis). Thus, the numbers in the figures should be read as the percentage of one standard deviation by which the average score for that gender in the specific country is different from the mean of the entire cross-country sample.5Figure 15a shows standardized scores in the Ravens fluid intelligence assessment. In seven out of nine countries included in this analysis, men outperform women in this assessment but this difference is statistically significantly different from zero only in Bolivia, Brazil and Peru.6The largest gaps are in Peru and Bolivia, where they are 16 and 21% of a standard deviation, respectively.7The exceptions are two countries in the HI group - Argentina and Uruguay - where on 4The under-representation of women in STEM occupations is shown in figure 20. 5The analysis presented here excludes Venezuela and we restrict our analysis to individuals of age 25-55 years old whereas the CAF sample also included individuals age 15-24. 6See Table A.9 in the Appendix for a full analysis. 7The number for the gender gap is obtained via subtraction (F-M), e.g., the gender gap for Peru is -.19 - -.03 = -.16 13 average women score slightly higher than men, although again these differences are not statistically significant at conventional levels. The gender gap in math, already evident in the PISA math scores at age 15, is also present among adults across most of the countries in the region as shown in Figure 15b for the CAF assessment of numerical skills. While the largest gap in the math PISA scores was around a fifth of a standard deviation (Figure 11), among adults the gap in numerical skills are 44% of a standard deviation in Bolivia, 33% for Panama, and 32% for Mexico. These gender gaps are statistically significant in all but one of the countries included in CAF; the exception is Argentina - despite being the country with the largest math gender gap among HI countries in the PISA test. The most pronounced gender gaps in PISA scores (age 15) are in reading and favors girls (Figure 12). Among adults, however, the pattern is less clear-cut (Figure 15c). In most countries women do a bit less well than men in the CAF verbal conceptualization assessment, although the difference is statistically significant only in Bolivia, Colombia and Mexico. Again Bolivia stands out as the country with the largest gap in favor of men (41% of a standard deviation). Overall, a consistent pattern is that women perform significantly worse than men in Bolivia, though the lack of data does not allow us to assess if this would be true for for other LMI countries as well. At the other end of the spectrum, Argentina is distinguished by its low levels of gender inequality in adult skills. Figure 15: Adult’s Skills (a) RAVEN Index (b) Numerical Skills (c) Verbal Conceptualization Note: The figure shows the gender gaps in three measures of skills in adulthood, computed for individuals aged 25 to 55 years old in 2015. Each variable is standardized (mean of 0 and a standard deviation equal to 1). The standardization was done using the entire sample (ages 15 to 55 and the 10 major cities of the 9 countries depicted plus Venezuela). Source: authors’ own calculations based on ECAF 2015 (CAF-development bank of Latin America). 14 2.3 The Gender Gap in Self Confidence and Expectations How does the gender gap in mathematics relate to each gender’s perception of their ability in that field? Several studies, using different country data-sets, find that math self-perception is strongly and positively associated with subsequent achievement in math at all levels of the achievement distribution, even controlling for earlier attainment in math and various child characteristics (Marsh and Martin, 2011;Susperreguy et al.,2018). We use PISA data to examine whether there is evidence of gender differences in math self-perception and ask whether this perception is correct relative to performance. In the 2012 round of PISA, students were asked to indicate how strongly they agreed with five statements relating to their math competence.8The statements included the following: •I am good at mathematics. •I get good grades in mathematics. •I learn mathematics quickly. •I have always believed that mathematics is one of my best subjects. •In my mathematics class, I understand even the most difficult work. The possible answers to these questions are: “very confident”, “confident”, “a little confident”, and “not confident at all”. We code the response of “not confident at all” as zero, “a little confident” as one, “confident” as two, and “very confident” as three. We use an individual’s responses to construct a math self-perception index by summing over the response scores and dividing that sum by 3x5=15. This results in an index with a minimum value of zero and a maximum value of one. Figure 16 shows the distribution of the math self-perception index separately for boys and girls. In both HI and UMI countries it is clear that across the distribution, boys have higher math self-perception than girls. This is consistent with trends we see in the US and is not surprising given that, as discussed previously and can be seen in Figure 17, boys at this age outperform girls in math. Figure 16: Math Self-concept Index Distribution 0 .5 1 1.5 Density 0.2 .4 .6 .8 1 Math self-concept index Boys Girls (a) High income 0 .5 1 1.5 2 Density 0.2 .4 .6 .8 1 Math self-concept index Boys Girls (b) Upper middle income 0 .5 1 1.5 Density 0.2 .4 .6 .8 1 Math self-concept index Boys Girls (c) United States Note: The figure shows, by gender, the distribution of the math self-perception index. The index is constructed based on each student’s responses to five questions related to their math competence, as defined in the text. The density function is obtained by pooling individual responses for all countries in the corresponding income group. Individual weights are reweighted to give equal weight to all nations in the same income group. Source: authors’ own calculations based on PISA 2012. 8For more details about the surveys, see Appendix, Section B.3. 15 Figure 17: Math Score Distribution 0 .001 .002 .003 .004 .005 Density 0200 400 600 800 Math score Boys Girls (a) High income 0 .002 .004 .006 Density 0200 400 600 800 Math score Boys Girls (b) Upper middle income 0 .001 .002 .003 .004 .005 Density 200 400 600 800 Math score Boys Girls (c) United States Note: The figure shows, by gender, the distribution of the PISA math test score for the LAC countries in the sample in (a), the OECD countries in (b), and the US in (c). The standard deviation of the test score distribution is 100 points. The density function is obtained by pooling individual responses for all countries in the corresponding income group. Individual weights are reweighted to give equal weight to all nations in the same income group. Source: authors’ own calculations based on PISA 2012. However, Figure 18a suggests that the math self-perception of girls is lower than that of boys even conditional on actual performance in mathematics. This figure plots the average self-perception index for each percentile of the PISA score separately for boys and girls for each country. Thus, each dot represents a cell defined by country, gender, and percentile of the PISA math score. The orange points (girls) tend to be below the blue points (boys) indicating that for each percentile of the score in PISA mathematics assessment, on average girls have lower math self-perception than boys across the region. This trend is highlighted by the line of best fit for girls being below than for boys (Lowess regression curves). Regression analysis further confirms that these differences are statistically significant. Appendix Table A.10 shows the results of regressing the math self-concept score on the math PISA score, a female dummy, and the interaction between the two for the whole of the region as well as for each of the countries separately. With the exception of Peru, in all LA countries the sum of the female dummy and the interaction effect is negative and statistically significant across the entire math score distribution. The positive coefficient on the interaction term, which is significant in the specification which pools all of the countries (column 1), further indicates that the gender gap in self-perception is narrower at higher levels of attainment. This can be seen in Figure 18a panel a: at the bottom of the PISA math score distribution on average there is roughly a 10 point gender gap (equivalent to 21% of the mean), while among the highest achievers this gap is less than half that number. A notable outlier in this analysis is Peru where the coefficient on the interaction is negative and statistically significant (Table A.10, Col 8), suggesting that the self-perception gender gap widens rather than narrows at higher match achievement levels. 16 Figure 18: Math self-perception by percentile of PISA mathematics score distribution 0 20 40 60 80 100 Average self-concept index 020 40 60 80 100 Test score percentile in Math Women Men (a) Latin America 20 40 60 80 100 Average self-concept index 020 40 60 80 100 Test score percentile in Math Women Men (b) OECD 20 40 60 80 100 Average self-concept index 020 40 60 80 100 Test score percentile in Math Women Men US (c) United States Note: The figure graphs the average math self-perception index (as defined in the text) against each percentile of the PISA score distribution (x-axis). The relationship is shown separately for boys (blue) and girls (orange) by country in LAC (Figure a), by country in the OECD (Figure b), and for the US (Figure c). Each dot in Figure (a) represents a cell defined by country, gender, and percentile of the PISA math score. The blue and orange lines depict the Lowess regression curves illustrating the relationships between the self-perception index and the PISA score percentile for boys and girls, respectively. How do these patterns compare to those of other countries? Figure 18 shows scatter plots for OECD and the US. It suggests that the pattern observed in LA countries is similar to those in the US and OECD where we also see lower math self-concept among girls across most of the math test score distribution and with the gender gap in confidence declining significantly at higher levels of attainment and disappearing at the very top. Self-perceptions of secondary school students may affect outcomes not only through impacts on achievement but also by shaping young people’s expectations. A growing literature shows that expectations play a key role in shaping key educational and career choices (Elsner and Isphording,2017; Wiswall and Zafar,2021). PISA data contains information on pupils’ occupational expectations, which we now turn to. Figure 19 shows gender gaps in the share of 15 year-old students who report that they expect to work in a STEM occupation at age 30 (which includes working in science, engineering, and information and communication technology). Across the LA countries included in PISA, only a small minority of girls (between around 10 and 20% in most countries) report that they expect to work in a STEM17 related occupation. Boys are more than twice as likely as girls to report this expectation. A similar expectations gender gap exist in OECD countries. The gender gap in expectations is especially large in UMI countries; for example in Colombia, Mexico, and Peru boys are around three times more likely to report that they expect to work in a STEM occupation than girls. How correlated might these expectations be with future occupation choices? In Figure 20 we use CEDLAS household survey data from 2018 to document gender gaps in actual proportions of young workers (age 30-40) working in STEM occupations in 2018 in the countries included in PISA. The results indicate that the pattern of proportions of 30-40 year old women working in STEM occupations is broadly similar to that in expectations to work in STEM around the same period. As in the expectations data, the proportion of women working in STEM does not exceed about a fifth of working women in this age group, while the proportion of men is 2-3 times larger in many countries. Furthermore, the countries where the expectations gender gap is largest, such as Chile, Colombia, Peru, and Mexico are also the countries where the gender gap in STEM occupations is largest. Figure 19: Expectations: Work in STEM-related Occupations (a) High income 0 10 20 30 40 50 Students who expect to work in STEM BRA COL CRI DOM MEX PER Boys Girls (b) Upper middle income Note: The figure shows the share of 15 year old students (PISA) who expect to work in a STEM-related occupation at the age of 30. The OECD average shows unweighted means. Source: authors’ own calculations based on PISA 2018. Figure 20: Adults Working in STEM-related Occupations 0 1 2 3 4 5 6 7 % of 30-40 year-olds working in STEM ARG CHL PAN URY Men Women (a) High income 0 1 2 3 4 5 6 7 % of 30-40 year-olds working in STEM BRA COL CRI DOM MEX PER Men Women (b) Upper middle income Note: The figure shows the share of the employed population aged 30-40 years old who work in a STEM-related occupation. The year used in the calculations is 2018 (2017 for Chile). Source: authors’ own calculations based on household surveys (GenLAC). 18 Despite the pronounced gender gaps in math self-concept and expectations to work in STEM occupations, as well as actual patterns of work on STEM occupations among adults, there is widespread agreement in LA countries that women and men have the same capacity for science and technology. Using data from Latinobarometro, Figure 21 shows that across the region in 2018 in the majority of countries, over 90% of respondents agreed or strongly agreed with the statement that “Women have the same capacity for science and technology as men.” The Dominican Republic and Ecuador have the lowest levels of agreement in the region but even here the agreement rate is around 80%. Figure 22 further shows that this high level of agreement extends to both women and men with only small gender differences. There is also little evidence of systematic differences in this view either by education or age cohort - differences across these groups are small and do not follow a distinctive pattern (see Figure A.4 in Appendix). Figure 21: Women have the same capacity for science and technology as men 0 20 40 60 80 100 % agreeing with 'Women have the same capacity for science and tech. as men' ARG CHL PAN URY Average BRA COL CRI DOM ECU MEX PER Average BOL GTM HND NIC PRY SLV Average Note: Individuals age 25-55 years old. This figure shows the percentage of individuals who agree or strongly agree with the statement ‘Women have the same capacity for science and technology as men.’ The bars show unweighted means. Source: authors’ own calculations based on Latinobarometro, 2018. 3 Work This section focuses on gender inequalities in the work sphere, including in labor force participation, employment structure, and wages. As in the prior section, we primarily utilize GenLAC–CEDLAS harmonized microdata from national household surveys and study both the current state as well as the evolution of trends over the last 20 years. We benchmark our analysis against several comparators outside of Latin America, including the US and, for some indicators, France, Indonesia, and Spain.9 We also assess how gender inequalities in the work sphere differ by education and how the presence of children affects women and men differentially. 3.1 Labor Force Participation Across all countries in LAC, over 90% of men between the ages of 25 and 55 were economically active, i.e. either working or actively looking for work in 2019 (Figure 23). As in the rest of the world, the 9These countries were chosen for a variety of reasons. Spain because of its ex-colonial status in LAC, France as an OECD country, and Indonesia to provide comparison as a lower-income country outside LAC. 19 Figure 22: Women have the same capacity for science and technology as men: % of individuals agreeing, by gender 0 20 40 60 80 100 % agreeing with statement High income Upper middle income Lower middle income ARG CHL PAN URY Average BRA COL CRI DOM ECU MEX PER Average BOL GTM HND NIC PRY SLV Average Men Women Note: Individuals aged 25-55 years old. This figure shows the percentage of individuals who agree or strongly agree with the statement ‘Women have the same capacity for science and technology as men.’ The average bars show unweighted means. Source: authors’ own calculations based on Latinobarometro, 2018. proportion of women in this category is significantly smaller (Chioda and Verd´u,2016;Gasparini et al., 2015;World Bank,2011), ranging between an average of 63% in the LMI countries and 74% in the HI countries in the region. Comparing these rates to the U.S. we see that while in the great majority of LA countries, including those in the HI group, female labor force participation rates (FLFP) are lower than that in the US, LFP among men is notably higher. As a result there is a larger LFP gender gap throughout the region than in the US. There has been a large increase in FLFP over the last 20 years, especially in the HI and UMI countries in the region (Figure 24a), although at a lower rate than during the 1990s (Marchionni et al.,2019). Since LFP among men has remained constant over the last 20 years, this has resulted in significant shrinking of the LFP gender gap, particularly in the HI and UMI countries in the region where, on average, the gap has shrunk by 10 percentage points compared to the 5 percentage points decrease in LMI countries. In the US, on the other hand, the change has been much smaller (but women started at a higher level) and the gap has narrowed due both to male LFP falling and female LFP rising (Figure 24). The aggregate patterns mask a significant amount of heterogeneity: across all LAC countries there is a pronounced education gradient in FLFP, with much higher LFP among women with higher education levels. Figure 25 shows how these differ by country and education, comparing LFP across women with incomplete secondary schooling, complete secondary schooling, and complete higher education. This gradient is similar across HI, UMI. and LMI countries in LAC. The average gap of around 25 percentage points between the most and least educated women is comparable in magnitude to the LFP gender gap in HI and UMI countries. It is also comparable to (though slightly lower than) the equivalent gap between less and more educated women in the U.S. The steep education gradient in FLFP is mirrored by a similarly steep education gradient in LFP gender gaps which are above 40 percentage points among the least educated group in the majority of countries, compared to a maximum of 16 percentage points among the most educated group (Figure 26). The similarity in FLFP education gradient between LAC 20 Figure 23: Labor Force Participation in LAC circa 2019 93.3 72.7 92.2 69.2 96.6 73.7 94.7 81.3 94.2 74.2 87.4 77.9 0 20 40 60 80 100 Labor force participation rate (%) ARG CHL PAN URY Average US Men Women Men US Women US (a) High income 90.2 70.3 95.4 72.0 93.3 61.9 95.1 70.8 96.1 67.6 96.1 63.8 95.0 80.6 94.5 69.6 0 20 40 60 80 100 Labor force participation rate (%) BRA COL CRI DOM ECU MEX PER Average Men Women (b) Upper middle income 97.6 73.5 96.5 49.7 94.9 57.7 95.6 62.6 96.2 73.3 92.9 60.3 95.6 62.8 0 20 40 60 80 100 Labor force participation rate (%) BOL GTM HND NIC PRY SLV Average Men Women (c) Lower middle income Note: This figure shows, by gender, the share of the population aged 25-55 years old that is economically active, as defined in the text. The average bars show unweighted means. Source: authors’ own calculations based on LAC household surveys (GenLAC) and the American Community Survey. Survey year is 2019 or the latest year available up to 2019 (see Table B.1 in Appendix B). Figure 24: Evolution of Labor Force Participation in LAC and US 60 70 80 90 100 Labor force participation rate (%) 2000-2004 2005-2009 2010-2014 2015-2019 Years Women, LMI Women, UMI Women, HI Men, LMI Men, UMI Men, HI (a) Evolution of Labor Force Participation in LAC 60 70 80 90 100 Labor Force Participation (%) 2000-2004 2005-2009 2010-2014 2015-2019 Years Women Men (b) Evolution of Labor Force Participation in the US Note: These figures show the evolution of the share of the population aged 25-55 years old that is economically active, as defined in the text. In Panel (a), each dot represents the (unweighted) cross-country average of their 5-year average. In Panel (b), each dot represents the 5-year average for the US. Source: see note to Figure 23. Only countries with available data in the corresponding periods are included (unbalanced panel in the case of LMI countries. See Table B.1 in Appendix B). countries and the US is less evident in the education LFP gender-gap gradient: the difference in the size of the gap between the least and most educated groups is 2-3 times smaller in the US than in most LAC countries. It is interesting to note that at all levels of education, women in HI countries in LAC have higher LFP than women in the US. Thus, the difference in the average LFP stems from fewer women in LAC, on average, obtaining higher levels of education than in the US. Disaggregating trends in FLFP over the last 20 years by women’s education shows that in HI and UMI countries, FLFP increased in all education groups, although slightly faster for women with less than tertiary education, which is as expected given the already high level at the beginning of the period (Figure 27a). For LMI countries, it is notable that LFP has decreased among women with complete secondary schooling: they have gone from having a higher LFP than similarly educated women in LAC 20 years ago to a lower level than these. For the least-educated women (those with incomplete secondary schooling), however, there is a clear positive gradient. In fact, the increase in the rate of 21 3.2.1 Hours worked in the market Among those age 25-55 who were employed in 2019, on average men work longer hours than women in all LAC countries (Figure 32). There has been a decline in the number of hours worked over the last 20 years (since 2000), for both men and women. This can be seen most starkly for men in UMI and HI countries, where the gender gap in work hours has shrunk slightly as a result (Figure 33a). The gender gaps in the average number of hours worked range from a maximum of 12.3 hours in Guatemala to 4.7 hours in El Salvador. While work hours are on average higher in poorer countries, especially among men, gender gaps are similar in magnitude across poorer and richer LAC countries (Figure 32). On average hours worked and the gender gap in the HI LAC countries is similar to that in the US. Figure 32: Hours Worked (2019) 43.3 34.1 45.7 41.0 42.6 36.9 43.2 35.9 43.7 37.0 43.2 38.5 0 10 20 30 40 50 60 Hours worked ARG CHL PAN URY Average US Men Women Men US Women US (a) High income 42.8 37.8 50.7 41.7 47.8 38.7 44.8 38.6 42.7 36.2 52.4 40.3 48.9 40.8 47.2 39.2 0 10 20 30 40 50 60 Hours worked BRA COL CRI DOM ECU MEX PER Average Men Women (b) Upper middle income 49.6 41.1 52.0 39.7 53.9 42.1 49.1 40.1 49.4 41.4 47.2 43.1 50.2 41.3 0 10 20 30 40 50 60 Hours worked BOL GTM HND NIC PRY SLV Average Men Women (c) Lower middle income Note: Individuals aged 25-55 years old. The figure shows, by gender, the average weekly number of hours worked in the market, including all jobs. The average bars show unweighted means. Source: see note to Figure 23. Figure 33: Evolution of the Number of Hours Worked 30 35 40 45 50 55 Hours worked 2000-2004 2005-2009 2010-2014 2015-2019 Years Women, LMI Women, UMI Women, HI Men, LMI Men, UMI Men, HI (a) Hours worked per week in LAC by gender 30 35 40 45 50 55 Hours worked 2000-2004 2005-2009 2010-2014 2015-2019 Years Women Men (b) Hours worked per week in US by gender Note: Individuals aged 25-55 years old. These figures show the evolution of the average weekly number of hours worked in the market, including all jobs. In Panel (a), each dot represents the (unweighted) cross-country average of their 5-year average. In Panel (b), each dot represents the 5-year average for the US. Source: see note to Figure 23. Only countries with available data in the corresponding periods are included (unbalanced panel in the case of LMI countries. See Table B.1 in Appendix B). As in the case of LFP, there is a steep education gradient in the gender gaps in hours worked: across HI, UMI, and LMI countries, the gender gap in working hours among those with tertiary education is around half of that among those who have not completed high school. This education gradient is 28 steeper than in the US, where the gender gap among the most educated is just over a fifth of that among the least educated group (see Figure 34). Figure 34: Gender Gaps in the Number of Hours Worked (F-M) by Education -11.0 -10.0 -7.6 -6.2 -5.8 -3.1 -12.3 -5.7 -3.2 -9.6 -7.6 -5.2 -9.8 -7.3 -4.8 -5.9 -5.3 -4.6 -20 -15 -10 -5 0 Gap in hours worked (F-M) ARG CHL PAN URY Average US Low Middle High Low US Middle US High US (a) High income -6.8 -4.7-4.4 -12.4 -9.6 -3.8 -13.4 -9.5 -3.8 -7.7 -6.0 -4.2 -7.4-8.0 -3.2 -15.4 -12.7 -7.3 -8.6-8.6 -5.0 -10.2 -8.4 -4.5 -20 -15 -10 -5 0 Gap in hours worked (F-M) BRA COL CRI DOM ECU MEX PER Average Low Middle High (b) Upper middle income -8.7 -10.0 -6.6 -13.5 -10.7 -7.0 -13.9 -8.0-7.8 -12.2 -7.1 -3.1 -9.4 -7.0-6.4 -4.1-4.1 -2.7 -10.3 -7.8 -5.6 -20 -15 -10 -5 0 Gap in hours worked (F-M) BOL GTM HND NIC PRY SLV Average Low Middle High (c) Lower middle income Note: Individuals aged 25-55 years old. This figure shows, by education level, the gender gap (F-M) in the weekly number of hours worked in the market, including all jobs. Low refers to less than high-school education; medium denotes high school graduates without higher education; and high indicates completed tertiary education. The average bars show unweighted means. Source: See the note to Figure 23. 3.2.2 Hours worked outside the market The preceding subsection was dedicated to hours worked in the market, excluding time spent on nonmarket work such as care activities and household chores. As is well documented in the literature, around the world the burden of these activities falls disproportionately on women (Charmes,2019). Thus patterns of gender differences in hours worked are likely to differ significantly depending on whether work includes non-market work. We use harmonized data from time-use surveys available for a sub-set of countries processed by GenLAC to document patterns in hours worked once non-market work is included in work hours.11 In this analysis we focus on married or cohabiting individuals age 25-45.12 We refer to this group as married individuals throughout this section for brevity. Figure 35 shows total hours spent on non-market work in housework and care activities by married men and women in each of the LA countries for which data is available and the US. We restrict the sample to individuals between the ages of 25-45.13 Bars with the letter “E” show the total hours spent by employed married women on non-market work. The bars without the “E” are for all married women regardless of their employment status. Men’s average hours spent on non-market work are shown in the last two bars on the right for each country. For men, the “E” indicates that their female partner is employed. As can be seen in the figure, married women in LA spend between a bit over 40 to 70 hrs a week on unpaid non-market work. This is comparable and, for many countries even higher than the time that 11There is some variation in the year for which this data is available across the countries. We use survey years that are closest to each other in timing and conducted before 2020 to align with the rest of our analysis which goes up to 2019. The resulting set includes surveys among which the earliest is 2010 for Peru and latest is 2017 for Costa Rica and El Salvador. 12The data does not allow us to identify who is married/cohabiting with whom so we cannot conduct couple level analysis. 13Although we are utilizing harmonized time use data, cross-country comparability remains limited and should be approached cautiously. This limitation stems from significant heterogeneity in the methodologies and protocols employed in Latin American time use surveys for different countries, as well as variations in the years they were conducted. Refer to the GenLAC Methodology section for more information about the data harmonization process undertaken to achieve the best possible comparability of time use indicators across countries. 29 men spend on market work as shown in Figure 32. Employed women tend to spend slightly less time on non-market work, with hours ranging between 30 in El Salvador and 60 in Mexico. With the exception of Paraguay, at least half of the non-market work time is spent on housework across the region. On the whole, women tend to spend more time on non-market work in HI and UMI countries compared to LMIs countries. The experience of women in couples across LA contrasts with that of US women in couples who, on average, spend much less time on non-market work at just under 30 hours per week (21 hours among employed women). In all countries, men in couples spend less than half the time on housework and care activities than women. In most countries this proportion is closer to a third ranging between 10 hours in Ecuador and Guatemala and around 27 hours in Chile. Across the region, the time spent on non-market work by women exceeds that spent on these activities by men by much more than the excess number of hours that men spend on market work compared to women shown previously (Figure 32), i.e., the total number of hours worked is greater for women. Furthermore, though employed women spend less time on non-market work than women who are not employed, their male partners do not compensate by increasing their non-market work hours as can be seen by comparing the bar without an “E” to that with an “E” for men. The difference in non-market work hours between men in LA versus in the US is much smaller than that for LA women compared to US women. This analysis clearly shows that focusing on time spent on market work only provides a partial view of gender inequality in work time. In order to paint a more complete picture, Figure 36 aggregates time-use survey data on both market and non-market activities to show how women fare relative to men in each of these individually and the two combined. A ratio of less than one indicates that women spend less time on a given activity than men. Darker columns show ratios for all married men and women, while lighter ones include only employed women (in couples) and men whose female partner works. In line with results presented above in Figures 32 and 35, the gender ratio in market work is below one, indicating that women spend less time in market work, whereas the ratios are considerably higher than one for non-market work time. The difference in the two ratios tend to be smaller for employed women and men with employed women as partners. In most countries, the combination of women spending less time on paid work and a lot more time on non-market work than men translates into a ratio above one for total hours worked. Appendix Table A.11 shows that across the region this difference is statistically significant. The ratio is highest in Chile where women spend around 75% more time on work and non-work activities combined than men and lowest in El Salvador and Peru. The ratios increase slightly when conditioning on women being employed. The ratio in market work hours in the US is similar to that of LA countries in the HI group, but household chores and care activities are more evenly distributed, with women spending double the hours that men do, as opposed to triple. Overall, the total hours ratio is below one indicating that, unlike in LA, US women spend less time on the sum of market and non-market work than men. Figure 37 shows how patterns differ by education. Panels (a), (c) and (e) show results for the sample of individuals in couples aged 25-45, as before. Across the region, non-market work times are highest among individuals in the lowest education group. Whereas women in this group tend to spend somewhere between 3 and 6 times as much time on non-market activities as men, the difference is 2-3 times among those in the most educated group. The exception to this is Guatemala where women with higher education spend over 6 times as much time on non-market activities as men. In line with Figure 30 34, the gradient goes in the opposite direction for market work: in all countries the hours that women spend on market work are closer to those done by men for those with higher education relative to those with less education. Combining paid and unpaid hours, there is much less evidence of systematic variation in the F/M ratio in total hours worked (blue column in Figure 37) across education groups. This is consistent with education gradients that go in opposite directions for paid and unpaid work. In the countries where we see some differences, it tends to be the case that the ratio of total hours women spend on work relative to men is highest among the least-educated group. This is the case in Argentina and Chile, for example. Education gradients in F/M ratios of time spent on paid and unpaid work are much less pronounced in the sub-sample of employed women and men whose partners are employed (panels b, d, f in Figure 37). This is especially true for time spent on market work. As before, overall, in most countries we do not see large differences in F/M ratio of total time spent on market and non-market work between education groups. Figure 35: Weekly Non-Market Working Hours (a) High income (b) Upper middle income (c) Lower middle income Note: Individuals in couples aged 25-45 years old. The letter “E” over a bar indicates average values for employed women or, in the case of men, for those men whose partner is employed. Time spent on care activities and household chores are derived from time use surveys. The value of the variable is set equal to zero when the individual does not do participate in an activity. Source: authors’ own calculations based on time use surveys (GenLAC). The year of the time use surveys ranges from 2010 to 2017 (see tables B.4 and B.1 in Appendix B, respectively). Figure 36: Market and Non-market Weekly Working Hours (Ratio F/M) 0 1 2 3 4 5 Ratio of weekly hours spent in market and non-market activities (F/M) ARG CHL URY USA Market Non-market Total (a) High income 0 1 2 3 4 5 Ratio of weekly hours spent in market and non-market activities (F/M) COL CRI ECU MEX PER Market Non-market Total (b) Upper middle income 0 1 2 3 4 5 Ratio of weekly hours spent in market and non-market activities (F/M) GTM PRY SLV Market Non-market Total (c) Lower middle income Note: Individuals in couples aged 25-45 years old. Non-market hours include care activities and household chores. Bars in dark colors show values for all individuals. Bars in lighter colors show the ratio for employed women relative to men whose partners are employed. The value of the variable is set equal to zero when the individual does not do participate in an activity. Source: see note to Figure 35. 31 Figure 37: Market and Non-market Weekly Working Hours, by education (Ratio F/M) 0 1 2 3 4 5 6 Ratio of weekly hours spent in market and non-market activities (F/M) Low Middle ARG High Low Middle CHL High Low Middle URY High Low Middle USA High Market Non-market Total (a) High income, all 0 1 2 3 4 5 6 Ratio of weekly hours spent in market and non-market activities (F/M) Low Middle ARG High Low Middle CHL High Low Middle URY High Low Middle USA High Market Non-market Total (b) High income, employed 0 1 2 3 4 5 6 Ratio of weekly hours spent in market and non-market activities (F/M) Low Middle COL High Low Middle CRI High Low Middle ECU High Low Middle MEX High Low Middle PER High Market Non-market Total (c) Upper middle income, all 0 1 2 3 4 5 6 Ratio of weekly hours spent in market and non-market activities (F/M) Low Middle COL High Low Middle CRI High Low Middle ECU High Low Middle MEX High Low Middle PER High Market Non-market Total (d) Upper middle income, employed 0 1 2 3 4 5 6 Ratio of weekly hours spent in market and non-market activities (F/M) Low Middle GTM High Low Middle PRY High Low Middle SLV High Market Non-market Total (e) Lower middle income, all 0 1 2 3 4 5 6 Ratio of weekly hours spent in market and non-market activities (F/M) Low Middle GTM High Low Middle PRY High Low Middle SLV High Market Non-market Total (f) Lower middle income, employed Note: Individuals in couples aged 25-45 years old. Figures (a), (c) and (e) show values for all individuals. Figures (b), (d) and (f) show the ratio for employed women to men whose partners are employed. Low refers to less than high school; medium denotes high school graduates without higher education; and high indicates completed tertiary education. Non-market hours include care activities and household chores. The value of the variable is set equal to zero when the individual does not do participate in an activity. Source: see note to Figure 35. 3.2.3 Types of Jobs There are some important differences in the types of jobs that men and women do. We categorize employed individuals into four types, depending on what they report their main job to be: wage 32 employees, employers, unpaid workers, or self-employed. Unpaid workers include individuals working on a family farm or business (mostly in retail) without receiving a wage. Figures 38a-38d show that, in all countries, women are less likely to be employers than men and much more likely to be unpaid workers. In some countries, such as Peru, Bolivia, and Ecuador, nearly a fifth of working women have unpaid jobs compared to less than 3% of men whereas in most HI countries there are few unpaid workers of either sex. 33 Figure 38: Share of Employment in Each Type of Job (a) Wage Employees 75.2 78.7 78.0 80.1 63.9 68.3 70.8 76.1 71.9 75.8 89.7 92.6 0 20 40 60 80 Wage-employee (% of employment) ARG CHL PAN URY Average US Men Women Men US Women US High income 65.1 73.9 53.153.8 77.2 82.9 51.1 71.0 59.3 45.6 79.4 70.5 54.4 43.4 62.863.0 0 20 40 60 80 Wage-employee (% of employment) BRA COL CRI DOM ECU MEX PER Average Men Women Upper middle income 42.1 37.9 67.7 51.5 55.1 46.4 58.6 50.4 61.1 56.9 70.2 54.4 59.1 49.6 0 20 40 60 80 Wage-employee (% of employment) BOL GTM HND NIC PRY SLV Average Men Women Lower middle income (b) Employers 3.8 2.0 2.4 1.4 3.6 2.1 4.7 2.4 3.6 2.0 0 5 10 15 Employers (% of employment) ARG CHL PAN URY Average Men Women High income 6.0 3.5 4.7 2.4 4.2 2.0 4.2 2.1 3.6 1.8 9.1 7.5 5.3 3.0 5.3 3.2 0 5 10 15 Employers (% of employment) BRA COL CRI DOM ECU MEX PER Average Men Women Upper middle income 5.9 2.0 4.0 2.1 14.7 6.5 8.1 2.2 8.0 3.7 5.2 4.4 7.6 3.5 0 5 10 15 Employers (% of employment) BOL GTM HND NIC PRY SLV Average Men Women Lower middle income (c) Unpaid Workers 0.1 0.6 0.2 0.5 1.5 5.8 0.2 1.0 0.5 2.0 0 5 10 15 Unpaid workers (% of employment) ARG CHL PAN URY Average Men Women High income 0.6 2.8 0.7 4.5 0.3 1.1 0.1 2.1 2.8 17.2 1.6 4.9 2.9 16.6 1.3 7.1 0 5 10 15 Unpaid workers (% of employment) BRA COL CRI DOM ECU MEX PER Average Men Women Upper middle income 3.1 18.8 3.0 9.9 3.6 7.5 5.5 7.8 3.0 8.6 2.3 5.4 3.4 9.7 0 5 10 15 Unpaid workers (% of employment) BOL GTM HND NIC PRY SLV Average Men Women Lower middle income (d) Self-employed 21.0 18.6 19.4 18.0 31.0 23.9 24.3 20.5 23.9 20.2 10.1 7.2 0 15 30 45 60 Self-employed (% of employment) ARG CHL PAN URY Average US Men Women Men US Women US High income 28.3 19.9 41.439.2 18.3 14.0 44.6 24.7 34.335.3 9.9 17.1 37.436.9 30.6 26.7 0 15 30 45 60 Self-employed (% of employment) BRA COL CRI DOM ECU MEX PER Average Men Women Upper middle income 49.0 41.3 25.3 36.5 26.6 39.6 27.8 39.6 28.0 30.7 22.2 35.8 29.8 37.2 0 15 30 45 60 Self-employed (% of employment) BOL GTM HND NIC PRY SLV Average Men Women Lower middle income Note: Individuals aged 25-55 years old. For each sex, the figures show the share of employment in each of the following four categories, as defined in the text and depending on what the individuals report their main job to be: (a) wage employees, (b) employers, (c) unpaid workers, (d) self-employed. For the US, we only present statistics for wage employees and those in self-employment as the ACS data we use does not have a code for employers and the share of unpaid workers is nearly negligible (less than 0.002 within our sample). In the US, the self-employed category encompasses both incorporated (4.1% for males & 2.4% for females) and not incorporated self-employment (6% for males & 4.9% for females). The average bars show unweighted means. Source: see note to Figure 23. 34 There is a less consistent pattern of gender differences in wage employment and self-employment across countries. In HI countries and about half of the UMI countries, women are more likely than men to be wage employees and men are more likely than women to be self-employed. Across both categories, in the majority of these countries the gender gaps are fairly small. This is also the pattern that we see outside of LAC, for the US. This pattern is reversed, however, in LMI countries, as well as in Ecuador, Mexico and Peru from the UMI country group. On average, in LMI countries, the proportion of women is 10 percentage points lower than that of men in wage employment and about 7 percentage points higher in self-employment. 3.2.4 Job Quality: Informality and Firm Size An important dimension of jobs in LAC is whether they are in the formal or informal sector. Those in the informal sector tend to have less employment protection, fewer formal rights, and less entitlements to in-work benefits. In this analysis we define wage workers without pension rights, non-professional self-employed, and all unpaid workers as working in the informal sector (ILO,2013). Overall, informality is much more widespread in the poorer countries in LAC (see Gasparini and Tornarolli,2009,Perry et al.,2007) for thorough descriptions of labor informality in LAC). Across the LMI countries, around two-thirds of men and women work in the informal sector as their main job, compared to around a third in the HI countries. The largest gender gap in the informal sector is in Peru where 71% of working women report their main job to be in the informal sector, compared to 59% of working men (Figure 39a). There is a great deal of variation in informal sector employment by education. As shown in Figure 39b, less-educated women are much more likely to work in the informal sector than women with a tertiary education. In most LAC countries, the main job of the majority of working women with incomplete secondary school education was in the informal sector in 2019, compared to no more than a quarter of women with complete higher education. We also see much larger gender gaps in informality rates among the less-educated workers compared to those with higher education (see Figure 40a). Indeed, in most countries, for those individuals with a tertiary education the share of men in the informal sector is greater than that of women. On the whole, the gender ratio in the informal sector has been stable for the last 20 years, though there was a period in the early 2000’s when in the UMI countries there was a big decline in this ratio among the least-educated group; that trend has reversed in the more recent years bringing the gap back up to the level of those with complete secondary education (Figure 40b). Another dimension that is likely to capture variation in job quality is firm size. Working for larger may provide several benefits, including more opportunities for progression within the firm and a lower risk of losing a job. Larger firms tend to also be more productive in LAC and thus pay higher wages (Eslava et al.,2021). We define a “large” firm as one with more than 5 employees as this is the measure available for all of the countries in the harmonized data-set. Overall, we see that both men and women are much more likely to be working in a large firm in the richer than the poorer countries in LAC (Figure 41a). With the exception of the Dominican Republic, a greater proportion of employed men report working in a large firm than employed women. On the whole, the size of the gender gap in this dimension is similar across poorer and richer countries. There is a steep education gradient in all of the countries; in most countries over half of working women with tertiary education are employed in 35 Figure 39: Share of employment in informal sector (a) Share of Employment in Informal Sector by Gender 38.7 39.7 24.3 25.6 42.7 41.2 27.6 24.5 33.3 32.7 0 20 40 60 80 Informality rate (%) - expanded def. ARG CHL PAN URY Average Men Women High income 37.9 33.8 53.855.0 28.832.0 51.6 45.4 58.161.9 55.3 61.0 58.8 71.1 49.251.4 0 20 40 60 80 Informality rate (%) - expanded def. BRA COL CRI DOM ECU MEX PER Average Men Women Upper middle income 68.9 75.6 73.776.1 65.368.0 66.2 64.8 61.564.9 57.1 65.4 65.569.1 0 20 40 60 80 Informality rate (%) - expanded def. BOL GTM HND NIC PRY SLV Average Men Women Lower middle income (b) Share of Employed Women in the Informal Sector by Education 71.5 53.1 18.3 49.0 33.6 10.2 81.5 47.6 14.3 48.0 26.7 4.6 62.5 40.2 11.8 0 20 40 60 80 Informality rate (%) - expanded def. ARG CHL PAN URY Average Low Middle High High income 58.5 36.9 7.5 86.8 61.7 12.8 55.9 34.3 7.6 75.1 51.9 13.1 86.2 72.4 16.4 79.3 65.5 35.3 95.0 82.5 25.0 76.7 57.9 16.8 0 20 40 60 80 Informality rate (%) - expanded def. BRA COL CRI DOM ECU MEX PER Average Low Middle High Upper middle income 95.4 86.3 39.1 92.0 52.6 32.0 85.8 64.3 26.0 85.9 63.5 17.6 91.8 76.0 29.9 88.7 62.2 20.1 89.9 67.5 27.5 0 20 40 60 80 Informality rate (%) - expanded def. BOL GTM HND NIC PRY SLV Average Low Middle High Lower middle income Note: Individuals aged 25-55 years old. The figures show the share of employment that is in the informal sector, as defined in the text. In panel (b), low refers to less than high-school education; medium denotes high school graduates without higher education; and high indicates completed tertiary education. The average bars show unweighted means. Source: authors’ own calculations based on LAC household surveys (GenLAC). Survey year is 2019 or the latest year available up to 2019 (see Table B.1 in Appendix B). large firms compared to between a tenth (in LMI countries) and a quarter (in HI countries) of women without a high school degree (see Figure 41b). This is also the group with the largest gender gap as shown in Figure 42a. For example, in high income countries, on average, the proportion of women without a secondary school degree working in large firms is 20 percentage points lower than that of men in this education group, whereas women with higher education are as likely to work in large firms as men. Figure 42b shows that over the last 20 years gender gaps in the likelihood of working for a large firm have, on the whole remained stable or declined slightly. 36 Figure 40: Gender Gap in Informality by Education (F-M) (a) Gender Gap in Informality 13.1 11.0 -0.6 6.3 4.9 0.5 16.8 3.4 -1.1 6.1 1.7 -2.7 10.6 5.3 -1.0 -5 0 5 10 15 20 Gap in informality rate (F-M) - expanded def. ARG CHL PAN URY Average Low Middle High High income 1.0 2.9 0.6 9.7 9.6 1.5 13.7 10.3 -1.3 2.9 4.9 0.1 8.7 14.1 -1.3 5.5 9.9 3.0 8.4 17.3 5.2 7.1 9.9 1.1 -5 0 5 10 15 20 Gap in informality rate (F-M) - expanded def. BRA COL CRI DOM ECU MEX PER Average Low Middle High Upper middle income 5.1 7.5 2.6 6.8 -0.1 -1.4 9.9 12.2 -1.1 8.7 5.6 -9.3 10.0 13.9 -3.1 9.1 12.2 1.7 8.3 8.6 -1.8 -5 0 5 10 15 20 Gap in informality rate (F-M) - expanded def. BOL GTM HND NIC PRY SLV Average Low Middle High Lower middle income (b) Evolution of the Gender Gap in Informality -5 0 5 10 15 20 Informality rate (%) - expanded def. 2000-2004 2005-2009 2010-2014 2015-2019 Years Low Middle High High income -5 0 5 10 15 20 Informality rate (%) - expanded def. 2000-2004 2005-2009 2010-2014 2015-2019 Years Low Middle High Upper middle income -5 0 5 10 15 20 Informality rate (%) - expanded def. 2000-2004 2005-2009 2010-2014 2015-2019 Years Low Middle High Lower middle income Note: Individuals aged 25-55 years old. The figures show, by education level, the gender gap (F-M) in the share of employment that is in the informal sector, as defined in the text. Low refers to less than high-school education; medium denotes high school graduates without higher education; and high indicates completed tertiary education. In Panel (a) the average bars show unweighted means. In Panel (b), each dot represents the (unweighted) cross-country average of their 5-year average. In Panel (a), the survey year is 2019 or the latest year available up to 2019. In Panel (b), only countries with available data in the corresponding periods are included (unbalanced panel in the case of LMI countries. See Table B.1 in Appendix B). Source: see note to Figure 39. 37 Panel a of Figure shows FLFP in 2019 across LAC countries among 25-55 year olds, distinguishing between women without children, those with children age 0-5 and those with older children (age 6-15). Across the region we see the expected pattern of lower LFP among women with children than without, with by far the lowest rates among women with young children. On average, in LMI countries in the region there is a 17 percentage points gap between the LFP of women without children and those with young children. This gap is a bit smaller, but comparable, in the wealthier countries. In half of the LMI countries, a strikingly low proportion (less than 50%) of women with small children were in the labor force in 2019. Figure 47: Female Labor Force Participation, Wages, and Children a) Female Labor Force Participation 86.7 64.0 72.9 77.7 63.2 68.9 80.3 64.5 71.8 87.5 78.0 81.7 83.0 67.4 73.8 0 20 40 60 80 100 Labor force participation rate (%) ARG CHL PAN URY Average No children Children 0-5 Children 6-15 (a) High income 75.3 64.9 71.7 77.1 64.7 72.4 67.6 50.5 61.0 70.9 67.8 76.3 72.9 62.1 67.4 71.5 52.8 62.8 86.6 74.1 83.2 74.6 62.4 70.7 0 20 40 60 80 100 Labor force participation rate (%) BRA COL CRI DOM ECU MEX PER Average No children Children 0-5 Children 6-15 (b) Upper middle income 84.1 62.7 76.8 59.0 40.3 49.8 60.9 45.2 56.0 69.2 51.9 64.1 80.2 66.1 77.7 62.2 44.9 62.4 69.3 51.8 64.5 0 20 40 60 80 100 Labor force participation rate (%) BOL GTM HND NIC PRY SLV Average No children Children 0-5 Children 6-15 (c) Lower middle income b) Wages 5.1 4.6 4.4 7.1 6.6 5.4 6.2 6.0 5.9 5.2 5.1 5.0 5.9 5.6 5.2 0 1 2 3 4 5 6 7 8 9 10 Hourly wage (2005 PPP USD) ARG CHL PAN URY Average No children Children 0-5 Children 6-15 (d) High income 4.6 4.6 3.8 4.2 3.5 3.1 6.3 5.3 5.0 3.1 3.1 3.5 4.6 3.6 3.4 4.1 3.7 3.2 3.2 2.9 2.7 4.3 3.8 3.5 0 1 2 3 4 5 6 7 8 9 10 Hourly wage (2005 PPP USD) BRA COL CRI DOM ECU MEX PER Average No children Children 0-5 Children 6-15 (e) Upper middle income 3.8 3.33.5 3.2 1.82.2 2.6 2.12.3 4.2 2.72.7 3.83.83.7 2.92.92.9 3.4 2.82.9 0 1 2 3 4 5 6 7 8 9 10 Hourly wage (2005 PPP USD) BOL GTM HND NIC PRY SLV Average No children Children 0-5 Children 6-15 (f) Lower middle income Note: Women aged 25-55 years old. Lighter olive bars show the LFP rate (panel a) and average wage (panel b) of women without children; the medium olive bars show LFP (panel a) and wages (panel b) of women with children aged 0 to 5 years old; the darker olive bars show the LFP (panel a) and wages (panel b) of women with children aged 6 to 15 years old. Wages are measured in 2005 PPP USD. The average bars show unweighted means. Source: authors’ own calculations based on household surveys (GenLAC). Survey year is 2019 or the latest year available up to 2019 (see Table B.1 in Appendix B). While there is some, though not complete recovery, in LFP as children get older, a parallel recovery is not observed in hourly wages for most countries. Panel b of Figure 47 shows that on average, across LAC, hourly wages among working women without children are higher than those of working women with children. In many of the countries, especially richer ones, working women with young children have a higher wage than working women with older children. This may reflect various factors: women with younger children are themselves younger, on average, and thus likely to have more education and be in more highly-paid occupations. This is not the pattern that we see in the poorer countries: there, women with older children have the same or a higher wage than women with younger children and in Paraguay and El Salvador there is almost no difference in the hourly wages of the three groups of 44 women. Clearly across most countries in the region there are significant differences in wages of women with and without children. Next we explore how women with children fare in the labor market relative to men with children. We adopt an event-study methodology following the approach taken in several recent studies of the impact of children on the gender wage gap (Figure 47) to quantify the gaps that emerge between men and women after the birth of the first child and how these evolve over the years following that. In order to do this we create pseudo-panels (Kleven et al.,2022) for each of the countries in our analysis using multiple rounds of cross-sectional data from household surveys.17 We build the pseudo panels by matching individuals on age and location: A parent observed at time tis matched to a childless individual in the same region who is pyears younger and was observed in the cross section data pyears before.18 This yields the proxy observations for t=−p.19 Using this pseudo-panel we estimate the following model separately for men and women: yit =X τ6=−1 βτ.I (kit =τ) + X j γj.I (j=ageit) + δy.I (y=t) + εit [1] where yit is the outcome of interest (LFP and earnings) for individual iat time t. The first term on the right-hand side is a set of event time dummies, τ=kit, which indicate the years relative to the birth of the first child of individual iat time t. The events τ≥0 capture the post-child effects relative to the base year which is τ=−1, i.e., one year before the first child was born. The second and third term are a full set of age-in-years dummies and calendar year dummies to control non-parametrically for life-cycle trends and time trends. We scale βτto show our results as a percentage effect compared to the counterfactual outcome without children predicted by the estimated model (see Kleven et al., 2019). Figures 48a-48c plot the results for LFP. In all of the countries we see very similar trends in LFP for men and women in the three years before the birth of the first child, controlling for life-cycle and time-trends, followed by a sharp divergence after the birth of the first child. This divergence is driven entirely by a sharp drop in women’s LFP, which is between 20-30% in HI countries (similar to the 25% estimated for the US in Kleven,2022) and 30-40% in UMI and LMI countries. There is evidence of some recovery over the 5 years following the birth of the child in the majority of UMI and LMI countries but not in the HI countries.20 At the end of the 5 year period, therefore, the decline in the LFP of mothers looks more similar across HI, UMI, and LMI countries. There is no evidence of any decline in 17Results presented here computed by using pseudo-panels in Chile are in line with those shown in Berniell et al. (2021) using Chilean panel data. 18While (Kleven et al.,2022) matches on education as well, we think it is potentially problematic to do so as women may not have finished their education 5 years before giving birth to their first child. Note furthermore that any age restriction (here we start at age 25 for first birth), inevitably introduces a degree of selection. 19In the cross-section we can identify whether the individuals are parents and the year in which their first child was born. Let τ=kit indicate the period relative to the birth of the first child of individual iat time t, and let τ= 0 identify the year that the first child was born. The data allows us to observe individuals once they become parents, but not before, i.e. we do no have information for τ < 0. In order to overcome this problem, we match a parent iof age a observed in year tand τyears relative to the first childbirth to a non-parent individual observed in year t−τ−p, at age a−τ−p, living in the same region. We do this to trace a pseudo-history of 5 years prior to becoming a parent (i.e. with p= 1, ..., 5). If there is more than one observation that could be a match, we collapse the observations using sampling weights. 20Berniell et al. (2021) shows that after the first child is born, the probability of employed mothers having an informal job increases substantially but not for fathers. The availability of informal jobs —characterized by more flexible working hours, but also lower wages and weaker social protection— might serve as a buffer against the drop in female employment. 45 men’s LFP around the time of the birth of their first child or in the 5 years after that. Figure 48: Parenthood and Labor Force Participation (a) High income countries -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth ARG -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth CHL -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth PAN -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth URY (b) Middle income countries -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth BRA -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth COL -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth CRI -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth DOM -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth ECU -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth MEX -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth PER (c) Lower middle income countries -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth BOL -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth HND -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth PRY -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth SLV Note: These figures show, for men and women, the estimated impact of children on labor force participation, from equation 1given the matching methodology described in the text. As the omitted category is τ=−1, the scaled coefficients show the impact of children as a percentage of the counterfactual relative to the year preceding the birth of the first child. Controls include calendar year and age fixed effects. Sample restriction: age for first birth is 25 to 45. The figure displays the 90% confidence intervals. Source: authors’ own calculations based on household surveys (GenLAC). Years 2003-2019. In line with the findings for LFP, Figures 49a-49c show similar evolution in earnings among men and women before the birth of the first child, followed by a sharp drop for women but not for men after this event.21 In fact, in several of the countries, such as Argentina and Bolivia, we see an increase in male earnings in the first few years following the birth of the first child. The drop in women’s earnings varies in magnitude across countries, ranging between around 50% in El Salvador and Bolivia to around 20% in Argentina and Colombia; as with LFP, it tends to be larger in poorer countries. Although there is some recovery in LFP, especially in the poorer countries, this is not reflected in earnings. In several countries including Argentina, Brazil, and Mexico, women’s earnings decline further over this period. There are a few exceptions; some recovery of around 10 percentage points is evident in Ecuador, the Dominican Republic, Peru, and El Salvador. In most countries, the gap between men and women remains constant over the 5 years after the birth of the child as there is also a slight reduction in male 21Note that the earnings include zeroes for women who are not participating in market work. 46 earnings. It is important to interpret what has been called the “motherhood penalty” correctly. The fall in women’s earnings does not reflect discrimination on part of the employer (though this may account for some of it in some countries). A large part is driven by women, and not men, deciding to leave the labor force once they become mothers, or by reducing their work hours, or by switching to a more flexible (and often less well-paid) job. These decisions themselves are driven by factors such as the availability, affordability, and quality of childcare as well as by expectations regarding mothers’ versus fathers’ roles in the care of their child. 47 Figure 49: Parenthood and Earnings (a) High income countries -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth ARG -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth CHL -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth PAN -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth URY (b) Upper middle income countries -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth BRA -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth COL -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth CRI -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth DOM -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth ECU -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth MEX -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth PER (c) Lower middle income countries -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth BOL -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth HND -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth PRY -60 -50 -40 -30 -20 -10 0 10 % change relative to τ = -1 -3 -2 -1 0 1 2 3 4 5 Years from childbirth SLV Note: These figures show, for men and women, the estimated impact of children on labor market earnings, from equation 1. As the omitted category is τ=−1, the scaled coefficients show the impact of children as a percentage of the counterfactual relative to the year preceding the birth of the first child. Controls include calendar year and age fixed effects. Sample restriction: age for first birth is 25 to 45. The figure displays the 90% confidence intervals. Source: see note to Figure 48. 3.5 Gender Roles How individuals believe that men and women should behave in various environments within and outside the home affects how people act, their aspirations, and the opportunities that are open to them, with important consequences for gender equality and the economy (see Fern´andez and Fogli,2009 and Fern´andez et al.,2021). Here we use nationally representative polls from Latinobarometro to examine on how women and men are viewed in several spheres related to work. Latinobarometro asks respondents to indicate their degree of agreement with several statements regarding beliefs about the appropriateness of women working and their relative competence in business and politics. Specifically respondents are asked to indicate whether they strongly agree, agree, disagree or strongly disagree that “A woman should work only if her husband does not earn enough”; “Men make better business executives than women”; and “Men make better political leaders than women.” We start by documenting variation in responses to these statements across countries and over time. We then examine cross-country patterns in gender gaps in the responses, as well as how these gender 48 gaps differ across education groups and cohorts. We end by showing a strong negative correlation between the gender gap in LFP (M-F) and the progressivity of social norms regarding women’s work. Figure 50 plots the proportion of respondents age 25-55 who either disagreed that a woman should work only if her husband does not earn enough for two years: 2008 and 2015.22 Turning first to the more recent year (2015), we see that countries vary significantly in their degree of disagreement, ranging from around 33% in Guatemala to over 80% in Brazil and Chile. There is a clear pattern of more progressive norms in higher-income countries. Whereas in LMI countries between around 35 and 60% of individuals disagree with this statement, in HI countries this range is between just under 70 and around 85%. Panama is a big outlier among the HI countries with significantly more conservative norms as indicated by only 45% disagreeing. Over time (between 2008 and 2015), beliefs in most countries have become more progressive as demonstrated by the fact that most data points lie above the 45 degree line. This trend is evident across poorer and higher-income countries. For example, in Brazil the degree of disagreement increased from just over 60% to nearly 85%, whereas in Nicaragua it went from around 45 to nearly 60%. There are exceptions to this trend in each of the income groups. In Guatemala, Panama, and Mexico the degree of disagreement has decreased over time. Panama stands out again as the biggest outlier with a very substantial of around 20 percentage points or around 30% of those who disagreed with the statement in 2008. Figure 50: A Woman should work only if husband doesn’t earn enough (% disagreeing) BOL GTM HND NIC PRY SLV BRA COL CRI DOM ECU MEX PER ARG CHL PAN URY 30 40 50 60 70 80 % disagreeing in 2015 30 40 50 60 70 80 % disagreeing in 2008 Low-middle Upper-middle High Note: Individuals aged 25-55 years old. This figure shows the percentage of individuals who disagree or strongly disagree with the statement ‘Woman should work only if her husband doesn’t earn enough’ in 2008 and 2015. The different colors refer to the different country groups: HI, UMI, and LMI. Source: authors’ own calculations based on Latinobarometro, 2008 and 2015. Figure 51 disaggregates the aggregate proportion that disagrees in 2015 by gender. On average, men have more conservative views than women (lower disagreement rate). The gender gap is largest in HI countries, where this pattern holds in all countries but Chile and on average the proportion of men disagreeing with the statement is around 10 percentage points lower than among women. The average gender gaps in opinion are smaller in LMI and UMI countries and there is a sizable group of countries, including four out of the seven UMI countries, where the degree of disagreement is either very similar among men and women or even lower for women. 22An individual was said to disagree if they choose either “strongly disagree” or “disagree” as their response. 49 Figure 51: Woman should work only if her husband doesn’t earn enough: % of individuals disagreeing 0 20 40 60 80 % disagreeing with statement High income Upper middle income Lower middle income ARG CHL PAN URY Average BRA COL CRI DOM ECU MEX PER Average BOL GTM HND NIC PRY SLV Average Men Women Note: Individuals aged 25-55 years old. This figure shows the percentage of individuals who disagree or strongly disagree with the statement ‘A Woman should work only if her husband doesn’t earn enough’. The average bars show unweighted means. Source: authors’ own calculations based on Latinobarometro 2015. Throughout the region, without exception, there is a steep education gradient in how conservative beliefs are regarding women working outside the home (Figure 52 ). Those with more education - completed tertiary - tend to have more progressive norms compared to those in the least educated group (incomplete secondary). On average this education gap is largest in LMI countries at over 20 percentage points compared to close to half of that in HI countries. Several countries in the UMI group have gaps closer to the average for LMI countries. These include Colombia, Costa Rica, and Mexico. A number of countries across the income groups stand out for having especially conservative norms (low rates of disagreement) among the less educated group. In Panama, Mexico, Guatemala, Honduras, and El Salvador, less than half of those with incomplete secondary education disagreed with the statement that a woman should only work if her husband does not earn enough. In Guatemala this proportion is around a third. Figure 52 also shows how views on women’s work differs across cohorts, distinguishing between individuals between the ages of 25-34, 35-44, and 45-54, all in 2015. On the whole there is less of a gradient in how progressive these views are by cohort than by education. In some countries, younger cohorts are more progressive (e.g. Honduras and Panama), whereas in others (e.g., Argentina and Chile), older cohorts are more progressive. 50 Figure 52: Woman has to work only if husband doesn’t earn enough: Degree of disagreement with the statement, by education and cohort 0 20 40 60 80 100 % disagreeing with statement ARG CHL PAN URY Average Low edu Medium edu High edu Age 25-34 Age 35-44 Age 45-55 (a) High income 0 20 40 60 80 100 % disagreeing with statement BRA COL CRI DOM ECU MEX PER Average Low edu Medium edu High edu Age 25-34 Age 35-44 Age 45-55 (b) Upper middle income 0 20 40 60 80 100 % disagreeing with statement BOL GTM HND NIC PRY SLV Average Low edu Medium edu High edu Age 25-34 Age 35-44 Age 45-55 (c) Lower middle income Note: Individuals aged 25-55 years old. This figure shows, by education level and cohort, the degree of disagreement (‘disagree’ or ‘strongly disagree’) with the statement ‘A woman should work only if her husband doesn’t earn enough’. ‘Low edu’ refers to individuals with up to a high school education completed, while ‘high’ indicates those who have completed tertiary education. Source: see note to Figure 51. Figure 53 shows the correlation between average opinions regarding women’s work and the size of the gender gap in labor force participation. As can be seen in the figure, in countries in which a higher fraction of the population disagrees with the statement regarding women’s work, the gender gap in LFP is smaller. 51 Figure 53: Gender gap in LFP (M-F) and percent disagreeing with statement ‘A woman should work only if husband doesn’t earn enough’ ARG BOL BRA CHL COL CRI DOM ECU GTM HND MEX NIC PAN PER PRY SLV URY 10 20 30 40 50 Gap in LFP between men and women 30 40 50 60 70 80 90 % disagreeing with statement Note: Individuals aged 25-55 years old. This figure shows the correlation between the degree of disagreement (‘disagree’ or ‘strongly disagree’) with the statement ‘A woman should work only if her husband doesn’t earn enough’ and the gender gap in LFP (M-F). Source: authors’ own calculations based on LAC household and time use surveys (GenLAC, several years) and Latinobarometro 2015. Next we turn to beliefs regarding the roles that women are suited to, focusing on leadership in business and politics. Figure 54 plots the proportion of respondents age 25-55 who disagree (coded as before) with the statement that men make better business executives than women in 2012 and 2019; Figure 55 plots the same for the statement that men make better political leaders than women. On the whole, in both periods, the majority of respondents in all of the countries included in the analysis disagreed with these statements. With respect to attitudes towards leadership in business, there is significantly less variation across countries in 2019 than in 2012. Whereas nearly a third of the respondents agreed with the statement in Brazil in 2012, in Peru this proportion was around 14%. By 2019 the gap between Brazil and Peru had shrunk by 7 percentage points, or 47% of the original gap. Attitudes towards political leadership also became more progressive during this time period. This is especially the case in Uruguay, where by 2019 nearly 100% of respondents disagreed compared to 80% seven years earlier and in Brazil where the level of disagreement went up by over 10 percentage points. from 72% in 2012 (Figure 55). Chile is a clear outlier relative to the general trend of increasingly progressive views over time. Both with respect to business and political leadership, the proportion of individuals who disagreed dropped sharply over this time period. Throughout the region men are more likely to agree than women with the propositions that men make better business and political leaders than women (Figures 56 and 57). The pattern in gender differences is similar for the two statements with the largest gender gaps of close to 20 percentage points observed in Chile. As in the case of beliefs relating to women’s LFP discussed above, there are clear education gradients in views on women’s capacity for business and political leadership roles (Figures 58 and 59); those with less education tend to hold more conservative views. For example, in Colombia, which has one of the steeper education gradients in responses to both statements, there is a gap of around 20 percentage points in the proportion of individuals disagreeing between those in the least and most educated groups. 52 In contrast to education, there is less of a a clear-cut pattern by cohort. In some countries, including Argentina, Colombia, Ecuador, and Nicaragua, there is a somewhat higher level of disagreement with both statements among those in the youngest cohort (age 25-34) than those in the older cohorts as we would expect if more recent cohorts have more progressive norms. However, there are also several countries with the opposite pattern or with essentially no differences across cohorts. Figure 54: Men make better business executives than women (% disagreeing) BRA COL ECU MEX PER ARG CHL URY 65 70 75 80 85 90 % disagreeing in 2019 65 70 75 80 85 90 % disagreeing in 2012 Upper-middle High Note: Individuals aged 25-55 years old. This figure shows the percentage of individuals who disagree or strongly disagree with the statement ‘Men make better business executives than women.’ Source: authors’ own calculations based on Latinobarometro, 2012 and 2019. Figure 55: Men make better political leaders than women (% disagreeing) BRA COL ECU MEX PER ARG CHL URY 65 70 75 80 85 90 % disagreeing in 2019 65 70 75 80 85 90 % disagreeing in 2012 Upper-middle High Note: Individuals aged 25-55 years old. This figure shows the percentage of individuals who disagree or strongly disagree with the statement ‘Men make better political leaders than women.’ The average bars show unweighted means. Source: see note to Figure 54. 53 A Appendix: Tables and Figures Table A.1: Test score regressions using TERCE for 3rd grade students (2013) High income countries Country ARG CHL PAN URY Reading Female 10.92*** 11.81*** 9.606*** 10.36*** 8.139*** 8.902*** 14.19*** 12.60*** (2.838) (2.948) (2.527) (2.717) (3.059) (3.154) (3.334) (3.423) College parents 36.86*** 41.06*** 56.76*** 47.89*** (6.566) (4.621) (6.194) (7.563) Female x College parents 0.0112 -7.494 5.601 15.02 (9.768) (6.570) (9.032) (10.88) 2,589 2,589 2,955 2,955 2,372 2,372 1,937 1,937 Math Female 5.612* 6.323* -1.382 -1.861 6.858** 8.168** 12.46*** 12.94*** (3.121) (3.240) (2.646) (2.833) (3.066) (3.174) (3.906) (3.992) College parents 42.68*** 45.52*** 60.99*** 74.94*** (7.234) (4.761) (6.115) (8.273) Female x College parents 1.136 -0.327 -7.805 -3.955 (10.59) (6.784) (8.934) (12.03) Observations 2,608 2,608 3,110 3,110 2,400 2,400 1,925 1,925 Robust standard errors are reported in parentheses. Calculations based on TERCE 2013. ***p < 0.01, **p < 0.05, *p < 0.1 Table A.2: Test score regressions using TERCE for 3rd grade students (2013) Upper-middle income countries Country BRA COL CRI DOM MEX PER Reading Female 13.08*** 12.69*** 13.53*** 14.55*** 8.147*** 8.077*** 16.71*** 16.82*** 11.04*** 11.74*** 0.491 -0.463 (2.936) (3.045) (2.486) (2.562) (2.546) (2.650) (3.181) (3.545) (2.894) (3.029) (2.483) (2.560) College parents 41.42*** 63.55*** 37.67*** 29.09*** 49.98*** 52.47*** (6.723) (5.064) (5.555) (5.424) (5.815) (5.370) Female x College parents 5.199 -13.33* 1.922 2.387 -3.412 4.522 (9.586) (6.962) (8.066) (7.603) (8.297) (7.698) 2,405 2,405 2,977 2,977 2,552 2,552 2,090 2,090 2,488 2,488 3,458 3,458 Math Female 5.245 4.278 1.928 3.670 -4.127 -5.514** 4.107 2.702 2.802 2.010 -8.136*** -9.307*** (3.466) (3.533) (2.783) (2.919) (2.649) (2.754) (3.070) (3.383) (3.185) (3.338) (2.738) (2.823) College parents 66.30*** 54.62*** 37.24*** 30.92*** 50.08*** 58.82*** (8.196) (5.423) (5.971) (5.223) (6.485) (5.823) Female x College parents 9.433 -7.736 7.921 10.60 4.470 5.555 (11.47) (7.741) (8.388) (7.390) (9.097) (8.327) Observations 2,412 2,412 2,869 2,869 2,551 2,551 2,254 2,254 2,493 2,493 3,451 3,451 Robust standard errors are reported in parentheses. Calculations based on TERCE 2013. ***p < 0.01, **p < 0.05, *p < 0.1 60 Table A.3: Test score regressions using TERCE for 3rd grade students (2013) Lower-middle income countries Country ECU GTM HND NIC PRY Reading Female -0.661 -0.619 -1.540 -2.643 3.165 1.777 0.883 0.899 11.25*** 11.94*** (2.372) (2.498) (2.559) (2.566) (2.588) (2.660) (2.798) (2.951) (3.309) (3.511) College parents 41.05*** 57.56*** 44.65*** 37.13*** 42.81*** (4.555) (8.720) (7.005) (5.707) (6.145) Female x College parents 6.026 21.25* 1.317 5.032 1.688 (6.675) (12.29) (9.212) (8.114) (9.109) 3,499 3,499 3,122 3,122 2,797 2,797 2,555 2,555 2,056 2,056 Math Female -3.157 -0.350 -5.519** -6.336** 1.706 -1.317 -7.883*** -5.927** -4.034 -4.329 (2.696) (2.851) (2.642) (2.632) (2.990) (3.066) (2.716) (2.862) (3.676) (3.897) College parents 51.06*** 73.87*** 41.06*** 45.42*** 40.31*** (5.069) (8.429) (8.247) (5.572) (7.194) Female x College parents -12.21 14.92 20.10* -12.33 7.054 (7.587) (12.02) (10.89) (7.884) (10.55) Observations 3,461 3,461 3,152 3,152 2,717 2,717 2,670 2,670 2,144 2,144 Robust standard errors are reported in parentheses. Calculations based on TERCE 2013. ***p < 0.01, **p < 0.05, *p < 0.1 Table A.4: Test score regressions using TERCE for 6th grade students (2013) High income countries Country ARG CHL PAN URY Reading Female 17.63*** 17.13*** 13.06*** 12.16*** 14.06*** 13.58*** 14.18*** 11.75*** (3.154) (3.250) (2.869) (3.102) (3.224) (3.389) (3.900) (3.959) College parents 41.29*** 34.73*** 58.78*** 62.19*** (8.306) (5.355) (6.296) (9.876) Female x College parents 6.603 2.504 -1.354 22.12 (11.71) (7.621) (8.649) (13.68) 2,766 2,766 3,384 3,384 2,583 2,583 1,979 1,979 Math Female -7.785** -8.928*** -5.727** -6.981** 2.318 1.737 -8.197* -11.50*** (3.153) (3.274) (2.913) (3.129) (3.057) (3.246) (4.287) (4.353) College parents 28.25*** 56.31*** 45.33*** 68.79*** (8.041) (5.313) (5.923) (10.56) Female x College parents 12.11 0.723 2.239 35.08** (11.31) (7.345) (8.286) (14.60) Observations 2,677 2,677 3,261 3,261 2,671 2,671 1,924 1,924 Robust standard errors are reported in parentheses. Calculations based on TERCE 2013. ***p < 0.01, **p < 0.05, *p < 0.1 61 Table A.5: Test score regressions using TERCE for 6th grade students (2013) Upper-middle income countries Country BRA COL CRI DOM MEX PER Reading Female 11.67*** 12.02*** 6.564** 4.468 4.058 2.811 10.79*** 9.075*** 5.789* 6.865** 0.313 -2.265 (3.537) (3.660) (2.876) (2.984) (2.973) (3.097) (2.874) (3.246) (3.286) (3.404) (2.932) (3.038) College parents 43.48*** 49.99*** 38.49*** 27.42*** 69.89*** 63.61*** (8.341) (6.166) (6.561) (4.663) (6.483) (6.157) Female x College parents 10.01 12.30 9.685 8.681 -8.844 4.138 (12.29) (8.854) (9.424) (6.590) (9.260) (8.486) 2,236 2,236 3,287 3,287 2,540 2,540 2,656 2,656 2,566 2,566 3,394 3,394 Math Female -10.20*** -11.24*** -17.02*** -18.77*** -10.43*** -11.42*** -4.738* -6.114** -7.475** -6.027* -24.90*** -27.74*** (3.272) (3.374) (2.598) (2.691) (2.807) (2.920) (2.687) (3.017) (3.371) (3.529) (3.383) (3.502) College parents 39.45*** 60.27*** 43.29*** 26.78*** 65.80*** 76.45*** (7.830) (5.712) (6.300) (4.389) (6.618) (6.964) Female x College parents 21.55* -4.200 6.325 4.973 -6.619 9.883 (11.48) (7.809) (8.893) (6.281) (9.517) (9.633) Observations 2,210 2,210 3,243 3,243 2,509 2,509 2,764 2,764 2,681 2,681 3,711 3,711 Robust standard errors are reported in parentheses. Calculations based on TERCE 2013. ***p < 0.01, **p < 0.05, *p < 0.1 Table A.6: Test score regressions using TERCE for 6th grade students (2013) Lower-middle income countries Country ECU GTM HND NIC PRY Reading Female 3.447 5.111* -2.150 -1.129 7.959*** 8.195*** 12.54*** 12.59*** 12.32*** 17.71*** (2.730) (2.855) (2.729) (2.725) (2.713) (2.756) (2.766) (2.971) (3.472) (3.655) College parents 66.37*** 73.07*** 62.39*** 41.66*** 65.49*** (5.051) (7.575) (6.490) (5.267) (6.337) Female x College parents -11.09 19.58 2.266 -3.940 -24.27** (7.432) (12.13) (9.367) (7.209) (9.542) 3,765 3,765 3,118 3,118 2,990 2,990 2,703 2,703 2,302 2,302 Math Female -10.51*** -10.97*** -20.15*** -18.12*** -10.03*** -9.895*** -7.853*** -9.456*** -2.971 -0.881 (2.693) (2.864) (2.714) (2.705) (2.850) (2.920) (2.524) (2.695) (3.582) (3.820) College parents 47.70*** 89.02*** 60.18*** 34.21*** 51.38*** (4.979) (7.688) (6.747) (4.905) (6.818) Female x College parents 2.623 -9.924 -0.564 6.608 -13.17 (7.311) (11.94) (9.563) (6.769) (9.880) Observations 3,893 3,893 3,118 3,118 2,958 2,958 2,829 2,829 2,394 2,394 Robust standard errors are reported in parentheses. Calculations based on TERCE 2013. ***p < 0.01, **p < 0.05, *p < 0.1 62 Table A.7: Test score regressions using PISA for 15 year-old students (2018) High income countries Country ARG CHL PAN URY Reading Female 16.05*** 12.68*** 19.81*** 23.58*** 14.37*** 16.2*** 23*** 23.41*** (2.57) (3.1) (3.64) (4.1) (2.7) (3.18) (3.24) (3.92) College parents 53.85*** 47.65*** 60.95*** 59.71*** (5.24) (4.24) (6.4) (5.59) Female x College parents 11.92** -5.85 -4.52 2.86 (5.69) (5.28) (5.46) (6.49) Math Female -15.43** -16.33*** -7.47** -3.79 7.66** 5.23 -8.33** -6.73* (2.25) (2.82) (3.65) (3.99) (3.32) (3.83) (3.31) (3.71) College parents 50.55*** 46.6*** 51.84*** 56.7*** 4.84 (3.81) (5.54) (5.43) Female x College parents 5.32 -5.69 -6.35 -2.01 (5.12) (4.42) (5.54) (6.1) Observations 11,975 11,975 7,621 7,621 6,270 6,270 5,263 5,263 Robust standard errors are reported in parentheses. Calculations based on PISA 2018. ***p < 0.01, **p < 0.05, *p < 0.1 Table A.8: Test score regressions using PISA for 15 year-old students (2018) Country Upper-middle income countries BRA COL CRI DOM MEX PER Reading Female 25.68*** 22.76*** 10.33*** 10.37*** 14.38*** 18.28*** 31.08*** 27.63*** 11.12*** 9.79*** 10.52*** 11.59*** (2.11) (2.6) (3.29) (3.15) (3.34) (3.42) (2.41) (2.86) (2.52) (2.92) (3.04) (2.98) College parents 50.83*** 49.25*** 50.4*** 24.35*** 39.09*** 55.12*** (4.22) (5.53) (4.18) (4.83) (4.88) (4.61) Female x College parents 7.54* 6.33 -4.22 11.72** 6.96 -4.46 (4.35) (5.55) (4.95) (4.92) (5.08) (5.02) Math Female -8.6*** -9.65*** -19.52*** -18.85*** -17.67*** -16.2*** 3.15 2.43*** -11.75*** -10.26*** -16.33*** -14.14*** (2.19) (2.43) (3.47) (3.41) (3.93) (3.48) (2.76) (3.05) (2.58) (2.97) (2.85) (3.17) College parents 49.32*** 43.04*** 39.65*** 26.45*** 35.03*** 51.07*** (4.08) (6.11) (4.63) (5.12) (4.78) (4.24) Female x College parents 2.43 3.25 0.53 4.8 -2.36 -6.85 (4.1) (5.55) (4.99) (4.8) (4.52) (4.78) Observations 10,691 10,691 7,522 7,522 7,221 7,221 5,674 5,674 7,299 7,299 6,086 6,086 Robust standard errors are reported in parentheses. Calculations based on PISA 2018. ***p < 0.01, **p < 0.05, *p < 0.1 63 Table A.9: Adult’s skills ARG BOL BRA COL ECU MEX PAN PER URY (1) (2) (3) (4) (5) (6) (7) (8) (9) Panel A Raven PMT Women 0.120 -0.211 -0.131 -0.071 -0.081 -0.091 -0.096 -0.166 0.083 (0.074) (0.077)*** (0.075)* (0.073) (0.069) (0.070) (0.098) (0.075)** (0.075) Observations 721 670 731 730 672 737 437 693 742 (1) (2) (3) (4) (5) (6) (7) (8) (9) Panel B Numerical skills Women -0.036 -0.436 -0.239 -0.234 -0.228 -0.315 -0.326 -0.151 -0.163 (0.074) (0.085)*** (0.084)*** (0.073)*** (0.078)*** (0.074)*** (0.100)*** (0.087)* (0.078)** Observations 702 536 570 709 612 701 386 533 657 (1) (2) (3) (4) (5) (6) (7) (8) (9) Panel C Verbal conceptualization Women -0.030 -0.408 -0.125 0.230 -0.059 -0.243 -0.081 -0.048 0.071 (0.077) (0.080)*** (0.077) (0.075)*** (0.077) (0.075)*** (0.094) (0.077) (0.075) Observations 721 670 731 729 672 709 437 676 739 Note: All these variables are standardized (mean 0, and SD 1). The standardization was done taking into account the whole sample (ages 15 to 55 and the 10 major cities of the 9 countries depicted plus Venezuela). The regression shown in this table was computed for people aged 25 to 55 years old. Robust standard errors are reported in parentheses. Calculations based on PISA 2018. ***p < 0.01, **p < 0.05, *p < 0.1. Source: authors’ own calculations based on ECAF 2015 (CAF-development bank of Latin America). Table A.10: Math self-perception and PISA mathematics score (1) (2) (3) (4) (5) (6) (7) (8) (9) LA ARG BRA CHL COL CRI MEX PER URY Math score 0.0657*** 0.0620*** 0.0451*** 0.1159*** 0.0563*** 0.0972*** 0.0984*** 0.0496*** 0.0948*** (0.0017) (0.0075) (0.0037) (0.0064) (0.0056) (0.0093) (0.0026) (0.0058) (0.0074) Female -8.1301*** -13.3829*** -9.9063*** -9.3920** -16.6369*** -11.8803** -4.8521*** 0.6756 -12.5151*** (0.9713) (4.2537) (2.1325) (3.9487) (3.1385) (5.5784) (1.6205) (3.1447) (4.3967) Math score * female 0.0044* 0.0155 0.0070 0.0029 0.0306*** 0.0109 -0.0004 -0.0165** 0.0110 (0.0024) (0.0106) (0.0053) (0.0092) (0.0080) (0.0135) (0.0038) (0.0082) (0.0103) Constant 23.6258*** 25.6125*** 32.3870*** -1.0126 34.5165*** 16.0082*** 11.2341*** 37.0945*** 12.9666*** (0.7497) (3.0833) (1.5339) (2.8327) (2.2789) (3.9883) (1.1354) (2.2785) (3.2164) Observations 56,528 3,584 11,581 4,395 5,443 2,819 21,881 3,624 3,201 Mean dep. var. 48.25 46.15 46.32 43.96 53.39 51.53 49.53 52.94 48.09 Country FE Yes No No No No No No No No Note: The self-perception index is derived from each student’s responses to five questions about their level of agreement with specific statements related to their math competence, as defined in the text. Robust standard errors are reported in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Source: authors’ own calculations based on PISA 2012. 64 Table A.11: Market and non-market weekly working hours All Employed Market Non Market Total Market Non Market Total Panel A: Low-middle income countries GTM Women −30.52∗∗∗ 39.50∗∗∗ 8.98∗∗∗ −17.59∗∗∗ 27.20∗∗∗ 9.61∗∗∗ Constant 38.06∗∗∗ 9.65∗∗∗ 47.71∗∗∗ 35.65∗∗∗ 11.51∗∗∗ 47.17∗∗∗ PRY Women −25.35∗∗∗ 37.55∗∗∗ 12.20∗∗∗ −13.35∗∗∗ 29.36∗∗∗ 16.02∗∗∗ Constant 43.04∗∗∗ 17.64∗∗∗ 60.68∗∗∗ 43.29∗∗∗ 18.03∗∗∗ 61.32∗∗∗ SLV Women −27.62∗∗∗ 31.02∗∗∗ 3.40∗∗∗ −4.84∗∗ 15.78∗∗∗ 10.94∗∗∗ Constant 48.17∗∗∗ 12.49∗∗∗ 60.66∗∗∗ 47.53∗∗∗ 14.61∗∗∗ 62.13∗∗∗ Panel B: Upper-middle income countries COL Women −26.93∗∗∗ 44.52∗∗∗ 17.59∗∗∗ −2.81∗∗∗ 23.55∗∗∗ 20.74∗∗∗ Constant 47.31∗∗∗ 20.26∗∗∗ 67.57∗∗∗ 46.44∗∗∗ 22.33∗∗∗ 68.77∗∗∗ CRI Women −28.97∗∗∗ 36.72∗∗∗ 7.75∗∗∗ −8.75∗∗∗ 21.35∗∗∗ 12.59∗∗∗ Constant 44.93∗∗∗ 20.13∗∗∗ 65.06∗∗∗ 45.11∗∗∗ 23.03∗∗∗ 68.14∗∗∗ ECU Women −29.48∗∗∗ 43.41∗∗∗ 13.48∗∗∗ −8.47∗∗∗ 34.21∗∗∗ 25.09∗∗∗ Constant 47.20∗∗∗ 10.51∗∗∗ 57.39∗∗∗ 48.56∗∗∗ 12.99∗∗∗ 61.06∗∗∗ MEX Women −33.40∗∗∗ 49.67∗∗∗ 16.27∗∗∗ −16.04∗∗∗ 38.99∗∗∗ 22.95∗∗∗ Constant 51.44∗∗∗ 19.31∗∗∗ 70.75∗∗∗ 52.29∗∗∗ 21.37∗∗∗ 73.66∗∗∗ PER Women −19.86∗∗∗ 34.63∗∗∗ 4.75∗∗∗ −18.04∗∗∗ 26.89∗∗∗ 9.67∗∗∗ Constant 50.66∗∗∗ 15.37∗∗∗ 65.27∗∗∗ 50.98∗∗∗ 15.52∗∗∗ 65.68∗∗∗ Panel C: High income countries ARG Women −26.70∗∗∗ 39.58∗∗∗ 12.90∗∗∗ −13.43∗∗∗ 27.67∗∗∗ 14.24∗∗∗ Constant 44.83∗∗∗ 19.92∗∗∗ 64.74∗∗∗ 44.44∗∗∗ 21.79∗∗∗ 66.22∗∗∗ CHL Women −5.63∗∗∗ 35.03∗∗∗ 29.40∗∗∗ −0.10 28.66∗∗∗ 28.56∗∗∗ Constant 16.51∗∗∗ 27.00∗∗∗ 43.51∗∗∗ 16.58∗∗∗ 27.70∗∗∗ 44.28∗∗∗ URY Women −9.69∗∗∗ 26.14∗∗∗ 6.72∗∗∗ −10.24∗∗∗ 19.43∗∗∗ 10.18∗∗∗ Constant 46.62∗∗∗ 19.67∗∗∗ 65.20∗∗∗ 47.17∗∗∗ 20.20∗∗∗ 66.39∗∗∗ USA Women −16.29∗∗∗ 12.40∗∗∗ −3.89∗∗∗ −7.97∗∗∗ 6.90∗∗∗ −1.07 Constant 39.77∗∗∗ 14.68∗∗∗ 54.45∗∗∗ 39.46∗∗∗ 14.94∗∗∗ 54.40∗∗∗ Note: Married individuals aged 25-45 years old. Columns 1 to 3 show values for the sample of married individuals. Columns 4 to 6 show values for the sample of employed married women or men whose partners are employed. Non-market hours include care activities and household chores. Each variable is equal to zero when the individual does not do such an activity. Source: authors’ own calculations based on time use surveys (GenLAC). The year of the time use surveys ranges from 2010 to 2017 (see tables B.4 and B.1 in Appendix B, respectively). ***p < 0.01, **p < 0.05, *p < 0.1. 65 Table A.12: Oaxaca-Blinder Decomposition Contribution of Explanatory Variables to the Gender Wage Gap F/M log wage log points Percent of gender gap explained Age + Educ. Job Total Total Total Age + Educ. Job Total Total Total Unadjusted Adjusted location charact. expl. unexpl. pay location charact. expl. unexpl. pay gap gap gap gap gap gap (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) Panel A: LMI BOL 0.0098 -0.0044 -0.1042 -0.0988 0.2978 0.1990 4.9% -2.2% -52.4% -49.7% 149.7% 100% 82.0% 74.2% NIC 0.0224 -0.0388 -0.1633 -0.1796 0.2171 0.0375 59.9% -103.5% -435.8% -479.4% 579.4% 100% 96.3% 80.5% SLV 0.0003 -0.0148 -0.1456 -0.1602 0.1592 -0.0010 -27.6% 1477.0% 14487.7% 15937.1% -15837.1% 100% 100.1% 85.3% HND -0.0728 -0.0952 -0.4035 -0.5715 0.3075 -0.2640 27.6% 36.1% 152.8% 216.5% -116.5% 100% 130.2% 73.5% Panel B: UMI PER -0.0140 0.0005 0.0113 -0.0022 0.2673 0.2651 -5.3% 0.2% 4.3% -0.8% 100.8% 100% 76.7% 76.5% DOM -0.0004 -0.0493 -0.0296 -0.0792 0.2770 0.1978 -0.2% -24.9% -14.9% -40.0% 140.0% 100% 82.1% 75.8% MEX -0.0064 -0.0306 -0.1256 -0.1626 0.3104 0.1478 -4.3% -20.7% -85.0% -110.0% 210.0% 100% 86.3% 73.3% ECU 0.0028 -0.0240 -0.0949 -0.1160 0.1996 0.0836 3.4% -28.7% -113.5% -138.8% 238.8% 100% 92.0% 81.9% BRA -0.0093 -0.0699 -0.0500 -0.1292 0.2110 0.0818 -11.4% -85.5% -61.1% -158.1% 258.1% 100% 92.1% 81.0% COL -0.0139 -0.0674 -0.0848 -0.1662 0.2128 0.0465 -29.9% -144.9% -182.3% -357.2% 457.2% 100% 95.5% 80.8% CRI 0.0061 -0.0571 -0.1273 -0.1784 0.1676 -0.0107 -56.6% 531.6% 1185.5% 1660.5% -1560.5% 100% 101.1% 84.6% Panel C: HI ARG -0.0039 -0.0651 -0.0734 -0.1424 0.1897 0.0473 -8.3% -137.8% -155.2% -301.3% 401.3% 100% 95.4% 82.7% CHL 0.0003 -0.0298 -0.0234 -0.0529 0.1827 0.1298 0.2% -22.9% -18.0% -40.7% 140.7% 100% 87.8% 83.3% PAN -0.0124 -0.0854 -0.0859 -0.1837 0.1816 -0.0020 611.0% 4218.6% 4242.4% 9072.0% -8972.0% 100% 100.2% 83.4% URY -0.0030 -0.0478 -0.0675 -0.1184 0.1838 0.0655 -4.6% -73.0% -103.2% -180.8% 280.8% 100% 93.7% 83.2% US 0.0004 -0.0226 0.0400 0.0178 0.1597 0.1775 0.2% -12.7% 22.5% 10.0% 90.0% 100% 83.7% 85.2% Note: Workers aged 25-55 years old, working at least twenty hours a week. The model used for the Oaxaca-Blinder decomposition controls for age, age squared, region of residence, an indicator for living in a rural area, education, sector, occupation (2-digits codes ISCO), an indicator for full-time worker (35+ hours a week), and another for working in the informal sector. Columns 1 to 3 show the result of multiplying the male-female difference in the specified set of variables by the male log wage coefficients associated with those variables. ‘Age + location’ combines the coefficients of age, region, and urban/rural areas. ‘Educ.’ aggregates the education level dummies, while ‘job characteristics’ sums up the explanatory contribution of sector, occupation, full-time, and informal dummies. Column 4 displays the total explained part in the Oaxaca-Blinder decomposition, and column 5 presents the unexplained part. Columns 7 to 12 show the ratio - multiplied by 100 - of the respective values shown in columns 1 to 6 to the total pay gap (i.e., column 6). The unadjusted gender wage gap (column 13) is calculated as the inverse of the exponential of the values shown in column 6, and the adjusted gender gap (column 14) is calculated as the inverse of the exponential of the values shown in column 5. Source: authors’ own calculations based on LAC household surveys (GenLAC) and the American Community Survey. Survey year is 2019 or the latest year available up to 2019 (see Table B.1 in Appendix B). 66 Figure A.1: Female share of graduates in a given field of education: High income countries 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (a) Argentina 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (b) Chile 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (c) Panama 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (d) Uruguay Source: World Bank. Argentina (2011), Chile (2017), Panama (2016) and Uruguay (2017). 67 Figure A.2: Female share of graduates in a given field of education: Upper middle income countries 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (a) Brazil 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (b) Colombia 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (c) Dominican Rep. 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (d) Ecuador 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (e) Mexico 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (f) Peru Source: World Bank. Brazil (2017), Colombia (2018), Dominican Republic (2017), Ecuador (2016), Mexico (2017) and Peru (2017). 68 Figure A.3: Female share of graduates in a given field of education: Lower middle income countries 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (a) Guatemala 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (b) Honduras 0 10 20 30 40 50 60 70 80 90 100 Female share of graduates by field Agriculture Arts Education Services Health Social Sci. Natural sci., math. STEM Engineering, manufacturing Note: Bars shows the number of female graduates expressed as a percentage of the total number of graduates in a given field of education from tertiary education (c) El Salvador Source: World Bank. El Salvador (2018), Guatemala (2015) and Honduras (2018). 69 B Appendix: Data Sources B.1 Definitions of countries income groups and education categories B.1.1 Countries income groups Countries were classified into different income groups based on their GNI per capita (in US$) for the period 2010-2020 and the corresponding World Bank Analytical Classifications. Each country was assigned to one of the following three groups: •Lower-middle income (LMI): Countries that were considered LMI at least once in 2010-2020 •Upper-middle income (UMI): Countries that were considered UMI during the entire 2010-2020 period •High income (HI): Countries that were considered HI at least once in 2010-2020 B.1.2 Education categories •Low: less than high-school education. •Medium: high school graduates without higher education. •High: higher education completed. B.2 GenLAC–CEDLAS household surveys data GenLAC is the CEDLAS (Center for Distributive, Labor and Social Studies) initiative to promote gender equity through the generation, analysis, and dissemination of evidence for Latin America and the Caribbean. For more information about GenLAC, visit www.genlac.econo.unlp.edu.ar. In this chapter we use their processed microdata, from more than 300 household surveys conducted in Latin American countries, to compute statistics. Table B.1 lists the surveys used in this chapter for 17 Latin American countries. Surveys are nationally representative (with the exception of Uruguay before to 2006 and Argentina, where surveys include mainly the urban population, which accounts for around 85 percent of the total population of both countries). GenLAC-CEDLAS made every effort to make statistics comparable across countries and over time by utilizing similar variable definitions in each country/year. The indicators are constructed according to SEDLAC’s processing protocol (CEDLAS and The World Bank). For the specific details about variables’ construction and definitions, refer to GenLAC methodological documents. 76 Table B.1: Household surveys included in the analysis. Country Survey Argentina Encuesta Permanente de Hogares Continua (yearly, from 2003 to 2019) Bolivia Encuesta de Hogares (yearly, from 2001 to 2019) Brazil Pesquisa Nacional por Amostra de Domicilios - Continua (yearly, from 2001 to 2019) Chile Encuesta de Caracterizaci´on Socioecon´omica Nacional (2000 & biyearly in 2003–2017) Colombia Gran Encuesta Integrada de Hogares (yearly, from 2001 to 2019) Costa Rica Encuesta Nacional de Hogares (yearly, from 2000 to 2019) Dom. Republic Encuesta Continua Nacional de la Fuerza de Trabajo (yearly, from 2000 to 2019) Ecuador Encuesta de Empleo, Desempleo y Subempleo (yearly, from 2003 to 2019) Guatemala Encuesta Nacional de Condiciones de Vida (2004, 2006, 2011 & 2014) Honduras Encuesta Permanente de Hogares de Prop´ositos M´ultiples (yearly, from 2001 to 2019) Mexico Encuesta Nacional de Ingresos y Gastos de los hogares (biyearly from 2000 to 2018 Nicaragua Encuesta Nacional de Hogares sobre Medici´on de Nivel de Vida (2001, 2005, 2009, 2014) Panama Encuesta de Hogares (yearly, from 2000 to 2019) Peru Encuesta Nacional de Hogares (yearly, from 2000 to 2019) Paraguay Encuesta Permanente de Hogares (yearly, from 2001 to 2019) El Salvador Encuesta de Hogares de Prop´ositos M´ultiples (yearly, from 2000 to 2019) Uruguay Encuesta Continua de Hogares (yearly, from 2000 to 2019) Note: All samples are restricted to individuals aged 25-55. Source: GenLAC - Evidence for gender equity in Latin America and the Caribbean (CEDLAS, 2022). Table B.2: Definitions of labor market variables. Variable Definition Labor force Economically active population as a percentage of the population aged 25-55. A person is considered economically active if she is either employed or unemployed. participation Hours worked Weekly hours worked in a paid job for workers aged 25-55, including all jobs. Hourly wage Average hourly wage in main occupation (in 2005 PPP USD), for workers aged 25-55 with positive earnings and positive hours worked. Informality Workers in informal jobs as a percentage of the employed population aged 25-55. Informal workers include wage workers without access to social security, self-employed workers who have not completed higher education, and zero-income workers. Employer Employers as a percentage of the employed population aged 25-55. Wage employee Wage employees as a percentage of the employed population aged 25-55. Self-employment Self-employed workers as a percentage of the employed population aged 25-55. Unpaid workers Unpaid workers (mainly family workers or apprentices) as a percentage of the employed population aged 25-55. 77 Definitions of labor market variables (continued). Variable Definition Large-firm Workers employed in large firms (with 5 or more employees) as a percentage of the employed population aged 25-55.workers Occupation Occupation held in a worker’s main job, based on the 2-digit International Standard Classification of Occupations of 2008 (ISCO-08). Not all surveys in the region allow for this classification. Sector Sector of activity of a worker’s main job, based on the 1-digit International Standard Industrial Classification (ISIC-Revision 3). Note: For more details about variables’ construction and definitions refer to GenLAC methodological documents. Definitions of education completion rates •Pre-primary enrollment rate (5 years old): Percentage of five-year-old children enrolled in an educational institution. •Primary completion rate (20-30 years old): Individuals who completed at least primary education as a percentage of individuals aged 20-30. •Secondary completion rate (20-30 years old): Individuals who completed at least secondary education as a percentage of individuals aged 20-30. •Tertiary completion rate (30-40 years old): Individuals who completed at tertiary education as a percentage of individuals aged 30-40. Definition of location It is a variable that captures the geographical macro-regions in which a country is organized. These macro-regions do not usually have a main authority, but instead result from the aggregation of other geographical areas with their own government. For example, each of the macro-regions of Argentina (Greater Buenos Aires, Pampeana, Patagonia, Cuyo, Northwest and Northeast) result from the aggregation of various provinces and each of the macro-regions of Brazil (North, Northeast, Southeast, South, Central -West) result from the aggregation of various federation units. In the case of Peru, the macro-regions (Urban Coast, Urban Sierra, Urban Jungle, Rural Coast, Rural Sierra, Rural Jungle and Metropolitan Lima) result from the combination of regions (Coast, Sierra, Jungle, Lima) and areas (urban and rural). B.3 PISA Evaluations PISA is the OECD’s Program for International Student Assessment. It evaluates students’ knowledge and skills as they approach the end of their compulsory schooling (at 15 years of age). Since their initial 78 administration in 2000, the PISA examinations have been conducted every three years in a number of countries. The tests evaluate students’ abilities in reading, math, and science (and in certain countries, additional topics). In Latin America, the countries that participated in the last edition of PISA (2018) are Argentina, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Mexico, Panama, Peru, and Uruguay. For more details on this survey see OECD (2018). In this chapter, we use PISA data from the years 2009, 2012, and 2018. We analyze tests scores in math and reading, math self-concept, and expectations to work in a STEM occupation at age 30. Tests scores: The standard deviation of the test score distribution is 100 points. Therefore, a 10-point difference on the PISA scale indicates an effect size of 0.10 SD. Math self-concept index: Index constructed based on each student’s responses to five questions about how strongly they agree with a given statement related to their math competence. The five statements considered are: ‘I am good at mathematics’; ‘I get good grades in mathematics’; ‘I learn mathematics quickly’; ‘I have always believed that mathematics is one of my best subjects’; and ‘In my mathematics class, I understand even the most difficult work’. The possible answers to these questions are: “very confident”, “confident”, “little confident”, and “not confident at all”. We code the response of “not confident at all” as zero, “little confident” as one, “confident” as two, and “very confident” as three. We use their responses to construct a math self-perception index by summing over the scores of an individual’s responses and dividing that sum by 3x5=15. This results in an index with a minimum value of zero and a maximum value of one. Available in 2012 only. Expectation to work in a STEM occupation at age 30 : Indicator equal to one if a student expects to work in a STEM-related occupation at the age of 30. STEM-related occupations include science and engineering professionals, and information and communications technology professionals. Available in 2018 and in 2015 only for some countries. B.4 SERCE and TERCE Evaluations These regional exams for primary education are produced by UNESCO. This chapter makes use of the Second Regional Comparative and Explanatory Study (SERCE; 2006) and the Third Regional Comparative and Explanatory Study (TERCE; 2013). These assessments evaluate the learning achievements of thirdand sixth-grade students in reading, math, and science. With the exception of Bolivia, Cuba, Honduras, and Venezuela, nearly every nation in Latin America took part in the TERCE testing. The test score scale has a standard deviation of 100 points. 79 Table B.3: Countries and years included in the SERCE/TERCE analysis Country Years Argentina 2006, 2013 Brazil 2006, 2013 Chile 2006, 2013 Colombia 2006, 2013 Costa Rica 2006, 2013 Dominican Republic 2006, 2013 Ecuador 2006, 2013 El Salvador 2006 Guatemala 2006, 2013 Honduras 2013 Mexico 2006, 2013 Nicaragua 2006, 2013 Panama 2006, 2013 Paraguay 2006, 2013 Peru 2006, 2013 Uruguay 2006, 2013 B.5 Adult’s Skills (ECAF 2015) The ECAF 2015 was carried out by CAF-development bank of Latin America, and has information about adult’s skills in 10 major cities in 10 LAC countries. In this chapter we analyze the data for 9 countries (all except Venezuela). Raven Progressive Matrices Test (Raven PMT; Raven, 1936): It is a nonverbal assessment of fluid intelligence that gauges abstract reasoning using 60 items. The person is asked to find the missing component that completes a certain pattern in each of the 60 items by comparing various forms and using analogies. A brief test with 8 items is utilized in the 2015 CAF Survey. Test of Verbal Conceptualization: This test measures the capacity for inductively producing linguistic concepts. The task entails inferring the relationship or rule that unites two concepts—in this case, “table-chair”—based on the presentation of the stimuli and verbally expressing it (answer: “They are both furniture”). It also requires putting into practice the three fundamental steps of inductive reasoning: codification, inference, and mapping. The exam consists of a sample of questions from the Wechsler Adult Intelligence Scale III’s subtest “Analogies” (WAIS III). The first and last items, which are deemed to be easy, the first two items of medium difficulty, and the first two items of maximum difficulty were used to choose the items. Index of numerical skills: This index is created by summing the results of a test and three questions requiring basic mathematical computations. The exam asks the respondent to count backward from 20 to 0; if they do it properly in the allotted time, they receive 1 point; if not, they receive no points. The 80 respondent is asked to answer real-world mathematics issues. They receive 1 point for each accurate response. There is no credit for wrong responses. The numeral skills index ranges from 0 to 4, with each question taking the value 1 if the respondent answers correctly and 0 otherwise. All these variables are standardized (expressed in standard deviations with respect to the measure). The standardization was done taking into account the whole sample (ages 15 to 55 and 10 cities). The statistics shown in this chapter were computed for people aged 25 to 55 years old. For more details on this survey and the indicators used here, please see Chapter 1’s Appendix in Berniell et al. (2017). B.6 UNESCO Institute for Statistics (UIS) data To compute the OECD indicator shown in Figure 3we use the UIS, UNESCO data, SDG Indicator 4.1.2: Percentage of a cohort of children or young people aged 3-5 years above the intended age for the last grade of each level of education who have completed that grade (i.e. 14-16 years old is the reference age group for calculation of the primary completion rate in the OECD). B.7 IPUMS B.7.1 American Community Survey (ACS) The American Community Survey is a project of the United States Census Bureau that has supplanted the decennial census as the primary source of information about the US population. We utilize the ACS data extracted from IPUMS for the years 2000-2019. B.7.2 IPUMS International In the cohort analysis, we rely on IPUMS International’s harmonized census microdata from Latin American countries, France, Indonesia, Spain and the United States. To facilitate comparative research, IPUMS codes the data consistently across countries and over time. LFP rates in the IPUMS data are not necessarily fully comparable to the CEDLAS (household survey) data for all LAC countries. For example, there is likely to be some variation in the treatment of unpaid work. However, IPUMS invests significant effort into standardizing measures across years and countries so these data are suitable for cross-country and cross-cohort comparisons. There is some variation in the years for which census data are available across LAC countries; for example, while in Chile census data are available for every decade from 2017 to 1960, in Peru there are only data for 2007 and 1993. In particular, Peru is excluded from the cohort analysis in this chapter because IPUMS has data for only two Peruvian censuses, and because there were relevant changes in the definitions of key labor market outcomes between those two censuses. In Bolivia and El Salvador, there were also some changes in the definition of labor market variables across censuses. However, IPUMS argues that the categories are generally comparable over time, so we keep those two countries in our cohort analysis. Table B.4 lists the censuses analyzed in this chapter. For more details about this data, refer to international.ipums.org 81 Table B.4: Census data (Source: IPUMS International) 2010s 2000s 1990s 1980s 1970s 1960s Argentina 2010 2001 1991 1980 1970 Bolivia 2012 2001 1992 1976 Brazil 2010 2000 1991 1980 1970 1960 Chile 2017 2002 1992 1982 1970 1960 Colombia 2005 1993 1985 1973 1964 Costa Rica 2011 2000 1984 1973 1963 Dominican Republic 2010 2002 1981 1970 1960 Ecuador 2010 2001 1990 1982 1974 1962 El Salvador 2007 1992 Guatemala 2002 1994 1981 1973 1964 Honduras 2001 1988 1974 1961 Mexico 2015 & 2010 2005 & 2000 1995 & 1990 1970 1960 Nicaragua 2005 1995 1971 Panama 2010 2000 1990 1980 1970 1960 Paraguay 2002 1992 1982 1972 1962 Peru 2007 1993 B.8 Time Use Surveys Table B.5 provides the list of the time-use surveys used in this chapter, employed to compute market and non-market hours. Non-market hours include care activities and household chores. Household chores are grouped into seven categories, following the guidelines established by the CAUTAL (CEPAL): preparing and serving food, cleaning the house, cleaning and maintenance of clothes and shoes, maintenance and minor reparations in the house, house administration, shopping for the household (including commuting time), caring for plants and pets. In all cases, the commuting and waiting time is included. Indicators related to the use of time are computed for married individuals aged 25-45. For more details about variables’ construction and limitations of the time use data, refer to GenLAC methodological documents. 82 Table B.5: Time use surveys included in the analysis Country Survey Year Argentina Encuesta sobre Trabajo No Remunerado y Uso del Tiempo 2013 Chile Encuesta Nacional sobre Uso del Tiempo (ENUT) 2015 Colombia Encuesta Nacional de Uso del Tiempo (ENUT) 2016 Costa Rica Encuesta Nacional de Uso del Tiempo (ENUT) 2017 Ecuador Encuesta Especifica de Uso del Tiempo 2012 El Salvador Encuesta Nacional de Uso del Tiempo (ENUT) 2017 Guatemala Modulo de Uso del Tiempo de la ENCOVI 2014 Mexico Encuesta Nacional sobre Uso del Tiempo 2014 Paraguay Encuesta sobre Actividades Remuneradas y No Remuneradas (EUT) 2016 Peru Encuesta Nacional de Uso del Tiempo (ENUT) 2010 Uruguay Encuesta de Uso del Tiempo y del Trabajo no Remunerado (EUT) 2013 Note: All samples are restricted to individuals aged 25-45. For more details about variables’ construction and definitions refer to GenLAC methodological documents. 83