Policy Research Working Paper 10605 How Have Gender Gaps in the Colombian Labor Market Changed during the Economic Recovery? María E. Dávalos Diana Londoño Daniel Medina Poverty and Equity Global Practice November 2023 Policy Research Working Paper 10605 Abstract This paper analyzes gender gaps in the Colombian labor techniques, the paper finds that the decline observed in the market, with a particular focus on the impact of the hourly gender wage gap results from two offsetting effects: COVID-19 shock and the economic recovery. Using the explained gap, which favors women due to their endow- household survey and administrative data, the analysis finds ments, primarily education level, and the unexplained gap, significant and persistent gender gaps in favor of men in which favors men and may be associated with discriminatory terms of participation, unemployment, and income. These biases. Moreover, the gender wage gap widens significantly gaps are heterogeneous at the regional level in Colombia, when considering monthly income, showing variations in and are exacerbated among women with children, particu- hours worked between men and women. Given the poorer larly young children, and those with low levels of education. labor market outcomes among women with children, pol- The COVID-19 pandemic widened these gaps, including icies toward the reduction and redistribution of care work from higher female concentration in the most affected sec- within households could contribute to increase women’s tors and potentially associated with the disproportionate opportunities in the labor market. burden of care on women. Moreover, using decomposition This paper is a product of the Poverty and Equity Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at mdavalos@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team How Have Gender Gaps in the Colombian Labor Market Changed during the Economic Recovery? María E. Dávalos, Diana Londoño and Daniel Medina JEL code: J16, J21, J31. Keywords: Gender wage gap, COVID-19, gap decomposition, female labor force participation, care policy. 1. Introduction During 2020, the COVID-19 crisis had a significant and unprecedented negative impact on the global labor market. This was due to an economic and social shock that resulted in a 3.3%1 global economic contraction. The labor market served as the main economic transmission channel to households. The International Labor Organization (ILO) estimated a reduction of 258 million hours worked for full-time jobs worldwide, equivalent to an 8.9% reduction in employment levels, affecting 135 million people (4%). The ILO study also found a sharp decline in the world's labor force participation, with over 100 million people leaving the workforce due to lockdowns and becoming economically inactive. Despite the economic recovery that started in 2021, the ILO forecasted that even in 2023 the employment recovery would still be underway. The impact of the COVID-19 crisis on employment and participation has been uneven globally, affecting certain sociodemographic groups disproportionately. A cross-country study conducted by the International Monetary Fund (IMF) across 38 advanced and developing economies found that 50% and 66% of them experienced significant decreases in women's employment levels compared to men (Bluedorn et al., 2021). This is due to unfavorable changes in sectoral composition, which disproportionately affect women. A more recent study by the World Bank (2023) documents the impact of the shock in the Latin America and Caribbean region, showing that job losses had a significant negative impact on youth, low-educated individuals, informal workers and female workers, and that the recovery has been slower for these groups than for the rest. This paper analyzes gender gaps in unemployment, labor participation, income, and hours worked in Colombia, considering key features of the pandemic and women's endowments. The study uses probabilistic models and standard decomposition methodologies to assess gender gaps and their drivers. It also offers a non-comprehensive view of national policies aimed at closing gender gaps in the labor market. These are the five key conclusions drawn from the research. First, the economic recovery in 2021 did not close the pre-existing gender gaps, which were further exacerbated by the pandemic. This suggests that there are underlying structural factors that need to be addressed. Second, while the gender gap in labor income and hours worked narrowed during the pandemic, this was due to a sharp decline in the levels for men rather than significant progress for women. Third, the formal sector showed more recovery in labor demand for women than the informal sector, but women are still largely concentrated in the informal sector. Fourth, women with children and lower levels of education had worse indicators and wider gaps compared to men and women without children or with higher levels of education, and this situation worsened during the pandemic. Finally, although there are programs in Colombia aimed at reducing gender gaps, such as those promoting economic opportunities for women, there is a need to increase the focus on addressing barriers related to 1 Based on information by the World Bank from 80% of the economies in 2020. care, including the promotion of the care economy, that disproportionately affect women and particularly women with lower levels of education. 2. Literature Review Historically, women have experienced labor gaps in the workforce. These gaps are often attributed to social norms and negative biases that have limited their participation in productive activities (Argys & Averett, 2022). Some of the primary determinants of these gaps include women's role as housewives or caregivers (Blau & Winkler, 2017; Cortés & Pan, 2020), occupational segregation (Blau et al., 2013; Dolado et al., 2003), and differences in education and non-cognitive skills (Grove et al., 2011). Additionally, employers often view mothers as a potential productivity loss due to maternity leave, resulting in hiring biases. While progress has been made in recent years, significant gaps still exist in income (17%) and labor force participation (66% compared to men) in Latin America and the Caribbean. Despite women achieving higher levels of education and reduced birth rates, income and participation gaps persist, especially among low-income households, self-employed individuals, non-formal workers, and rural areas (ILO, 2019). Recent studies from the World Bank (2020) and the ILO (2019) have highlighted the need for continued efforts to reduce these gaps. The COVID-19 economic shock had a significant impact on the labor market, and across many countries the international literature has documented more severe negative effects for women than for men. For example, the Republic of Korea recorded a higher rate of absence among women, particularly due to their role in care work, as well as high rates of part-time employment and an unexplained gap attributed to care-related roles (Ham, 2021). Similarly, women in Japan faced lower demand due to their high participation in fixed-term jobs and concentration in the service sector (Kikuchi et al., 2021). In the United States, the situation was similar, with a more significant contraction in participation and employment levels for married women with children, partly due to restrictions on childcare services (Albanesi & Kim, 2021). The impact of children on labor participation and employment was also found to be significant by other authors in the United States, particularly for women with children between 0-6 years and 6-12 years, in low-income jobs, and among Latinas and Blacks (Lim et al., 2021; Pitts, 2021; Fairlie et al., 2021). Comparable effects on participation were also seen in the United Kingdom2 (Cattan et al., 2020), Australia (Churchill, 2021), and Canada (Quian & Fuller, 2021). Extensive evidence also exists for developing countries. Using High-Frequency Phone Surveys (HFPS) conducted in 13 Latin American and Caribbean (LAC) countries, Cucagna & Romero (2021) analyzed the gendered impact at the onset of the COVID-19 crisis. They found that female workers were 44% more likely than male workers to lose their jobs, with a persistence in the job loss gap as the crisis continued. Furthermore, they found that 56% of job losses were explained by highly female-intensive sectors such as trade, personal services, education, and hospitality, and 2However, Witteveen et al. (2020) find that in the same country, men suffered more layoffs during the pandemic’s early part th an women. that school-age children at home were linked to a rise in job losses among females but not males. Berniell et al. (2021) also used the same surveys to estimate fixed effect models and confirmed that women's labor outcomes were harder hit, with more job losses, income reductions, and fewer job entries. They also discovered that working from home strongly decreased the negative effects, mainly in women with children. Using HFPS in African countries, Contreras-Gonzalez et al. (2022) found a higher probability of job loss for women, urban workers, and youth, with an even greater likelihood for urban workers among women with children. Viollaz et al. (2022) investigated four Latin American countries and found a significant impact on preexisting gaps before the pandemic, with women being less likely to work and more likely to have informal jobs. The authors suggested that a driving factor was the disparity in the share of children, particularly those with school-age children.3 On the flip side, the pandemic also brought about changes in the gender income gap, although the impacts varied depending on women’s occupational choices. Contreras-Gonzalez et al. (2022) found that African non-agricultural companies experienced a greater income loss among women and youth. Doorley et al. (2021) found negative effects on women's wages during the first wave of COVID-19 in Ireland but also interestingly observed a shift in occupational segregation that previously favored men to now favor women's earnings. Additionally, fiscal benefits had a multiplier effect that reduced the gap, which was twice as large as the pre-pandemic reduction. In Canada, Singh et al. (2022) found negative effects on women's income and employment, which increased gender disparities and related to sectoral segregation. Bonacini et al. (2021) used decomposition methods in Italy and discovered that the income gap was greater for women working from home (WFH) 4 - something not expected - with effects on older and married women and those working in the private sector. On the contrary, in the United Kingdom, Kan M-Y (2021) found that women experienced a smaller reduction in their income given their high participation in critical sectors such as the health and care industries. A similar result was found by Aina et al. (2021) in Italy. On the other hand, in Estonia, the pandemic reduced the gaps in employment, income, and hours worked. Again, women with children and those from more affected sectors are penalized with possible long-term effects on inequality (Verdostup, 2021). Hill & Köhler (2021) analyzed gender wage inequality in South Africa during the pandemic and found that the gender wage gap increased significantly by 37%, with a more pronounced effect on the poorest quintile, among which the gap increased by as much as 2.7 times. In Colombia, studies (World Bank, 2019; Garcia-Rojas, 2020 and Hernandez, 2022) show a significant reduction in female participation and employment level, labor formality, and increases 3 Some independent studies also find negative effects on employment outcomes for women: Brazil (Masri et al., 2021), Mexico (Hoehn-Velasco et al., 2021; Juarez & Villaseñor, 2022) and Chile (Egana-del Sol et al., 2020). 4 In the UK, increases in self-perceived productivity for those working at home were also found, except for mothers with children, who also work longer hours (Deole et al., 2021). in the burden of care. The country has smaller wage gaps than the OECD average, but this gap is associated with the unexplained component of the observables and is greater at the extremes of the income and education distributions. 3. Data Sources and definitions We use the Great Integrated Household Survey (GEIH), a probabilistic sampling survey that gathers data on individuals' employment status, including their occupation, income, access to social security, and job search activities, as well as their demographic characteristics such as gender, age, education level, and sources of income. The GEIH is nationally representative and uses a self-weighted sampling method stratified by unequal clusters. The survey's primary aim is to address the factors that affect labor participation and estimate gender-based labor gaps. Additionally, the survey includes information at the regional, departmental, and department capital levels. The GEIH's sample frame was updated in 2022 with information from the 2018 National Population and Housing Census (CNPV), resulting in a structural change in trend series compared to the previous sample frame collected until December 2021. Because our work requires comparing results over time since 2015 and using information from the Mission for the Splicing of the Employment, Poverty, and Inequality Series (MESEP) as a source for labor income variables, we use the 2005 frame GEIH data between 2015 and 2021.5 The labor income variable comprises five categories of income that vary depending on the type of employment relationship, including First Activity Monetary Income (IMPA) for wage earners and self-employed, Income in Kind (IE) for wage earners only, Second Activity Income (ISA) for wage earners, self-employed, and family workers, Monetary Income of the Unemployed and Inactive (IMDI) for the unemployed and inactive, and Income from other sources (IOF) available to everyone. From 2009 onwards, the survey covers 437 municipalities and annually visits an average of 248,000 households concentrated in 22,548 segments. There are limitations to the survey data collected in 2020. Due to government-imposed restrictions in the first few months of the pandemic, data collection in cities was done over the phone. From March to April, the survey was restricted to 39 questions from the usual 200 to reduce respondent rejection. Moreover, data collection restrictions were only relaxed from June onwards, making it unclear how these measures affected data collection and comparability. Therefore, we only make comparisons with 2020 when analyzing primary labor indicators such as unemployment, participation, and employment, as we have information for March and April. When including 2020 5 Nevertheless, the previous section made a descriptive reference to the main indicators for the years 2021-2022 with the information from the new sample framework of 2018. data elsewhere, we do so for informational purposes only, and any comparisons must consider this caveat. Data description Table 2 presents key descriptive statistics of the survey data. Between 2015 and 2019, the survey collected an average of 204,000 observations annually for men and 240,000 for women. However, in 2020, the year most impacted by the pandemic, the number of observations for both groups dropped sharply to 123,000 and 142,000, respectively. By 2021, the data collected had partially recovered, but the observations remained below pre-pandemic levels (188,000 for men and 219,000 for women). Demographic variables such as average age, marital status, and residential area show a slight difference between the groups, while poverty is more prevalent among women. For both groups, the average age is 36 years, and 22% of men live in rural areas, while on average, 19% of women do. Additionally, 56% of men are married or living in a union, and 58% of women are. Finally, 31% of men live in low-income households, while 37% of women do. A further breakdown of these categories reveals that 21% of the population are young people between 18-24 years old, approximately 42.6% are between 25-44 years old, and 36.5% are between 45-59 years old. In terms of children in the household, women live with more children than men. On average, for every man, there are 0.3 children aged 0-3 years in the household, 0.2 children aged +3-6 years, and 0.40 children aged +6-12 years. In comparison, women have, on average, 0.3 children aged 0- 3 years in the household, 0.2 children aged +3-6 years, and 0.45 children aged +6-12 years. Moreover, 21% of men live in households with children aged 0-3 years, while 26% of women do. For +6-12 years, the numbers are 17% for men and 21% for women. Finally, in households with children aged +6-12 years, the numbers are 25% for men and 30% for women. Human capital variables such as years of education and experience show slight differences between the groups. On average, men have 9.6 years of schooling, women have 10.2 years, and both groups have over 20 years of experience. A disaggregation of these categories shows a difference in the higher education component, with approximately 20% of men having higher education while around 25% of women do. However, in non-education or just elementary school, men account for 35%, while women account for 30%. Experience is strongly centered on the group with +15 years of experience, where 57% of both groups are. The survey reveals significant heterogeneity between men and women in the labor component. Between 2015 and 2021, for the population aged 18-59 years, 82% of men were employed, 8.4% were unemployed, and 9% were inactive. In contrast, only 56% of women were employed, 11% were unemployed, and 33% were inactive. Additionally, women have a higher rate of part-time employment (32%) compared to men (12%), indicating that women work almost two times fewer hours per week (23 hours) on average than men (40.3 hours). Salaried workers dominate in both groups, with 49% of men in the private sector and 3.6% in the public sector, while women have higher rates of 51.5% in the private sector and 4.8% in the public sector. Self-employment is more common among men (48%) than women (44%), and both groups have a similar rate of formal employment with social security contributions (around 40%). Regarding the sectoral composition of the occupation, men have a significantly higher relative participation in transportation (13%), construction (10%), and public services (2.7%), while women have a more significant presence in accommodation and food services (12%), education and health services (14%), and other services (14%). Women also have a higher proportion of employees in commerce (21%) and professional services (9%) compared to men (17% and 5%, respectively). In the agricultural sector, only 6% of women work compared to 20% of men. Both groups have a similar representation of around 12% in the industry sector. Finally, women have only 3% representation in the financial and public sectors. 4. Existing gaps and negative impacts of the pandemic Despite progress in narrowing income gaps, gender gaps in employment remain unresolved. The COVID-19 pandemic reinforced pre-existing gender gaps in labor participation and unemployment, resulting in a significant increase in the gender gap from 23.4 to 25.7 percentage points (p.p.) between 2019 and 2020 (Figure 1). The gender gap in unemployment also grew by 2.1 p.p. in the same period (Figure 2). Although both groups experienced a recovery in 2021, the participation level grew faster in men, widening the gender gap by 0.3 p.p. However, the unemployment rate gap remained unchanged. This suggests that the gaps have widened in the pandemic due to worse indicators for women. Across regions, there is significant heterogeneity in the indicators and gaps between the main cities, with some able to reduce the gap in 2021 compared to the pre-pandemic level (Figure 3-5). On the other hand, the wage gap reduced significantly in 2020 due to a greater drop in men's labor income but increased slightly in 2021 as men's income recovered faster.6 In terms of hours worked (Figure 6), men have been reducing their weekly working hours since 2015, and the gap has narrowed following the pandemic. However, there is still a high heterogeneity in the gap across regions (figure 7- 8). In general, all cities achieved a reduction in their labor income gaps during the pandemic, but some cities exceeded pre-pandemic levels in 2021. The sectors that experienced an employment recovery in 2021 had low female participation. However, there are indications of progress in 2022. The sectors with high female participation, such as education, health, accommodation, and food service, experienced a slow-paced recovery of labor demand in 2021, which slowed the adjustment of the pandemic-led gap between men and women. Figure 9 shows a negative relationship between 6 Estimated gap not controlled by sociodemographic characteristics. female sectoral participation (Zone I) in 2019 and the growth of employed persons between 2019 and 2021, indicating that sectors with high female participation were more affected by the contraction in labor demand than sectors with a high concentration of men (Zone II). This pattern is also evident in the job vacancies offered by the public employment services, even though formal demand for the health and financial sectors increased during the same period (Figure 10). In 2022, DANE changed its expansion factors, making it difficult to compare information for years before 2021. However, a review of the trend of participation and unemployment for the 1st semester of 2021 and 2022 (with the 2018 factors) shows that the recovery trend is positive for both men and women (Table 1).7 The participation rate has increased in both groups, especially for women, and the unemployment rate has contracted in both groups,8 with the reduction being more significant in women. Therefore, it can be deduced that there was a gap reduction for both indicators. In 2022, the growth of women's labor demand has been driven by growth in the service sector, such as professional services, accommodation, food and entertainment, commerce, and industry, with more than 10% growth rates pared to 2021. Among these groups, only the industry grew more rapidly among men. Table 1. Labor Market Indicators 2021-2022 1er semester Change (p.p.) 2021 2021 2022 (3)-(1) (3) - (2) Indicator (1) (2) (3) GPR 72.0% 75.7% 76.6% 4.6 0.9 Men UR 12.0% 12.9% 9.6% -2.4 -3.3 GPR 49.1% 48.2% 51.5% 2.4 3.3 Women UR 20.1% 19.4% 15.6% -4.5 -3.8 Expansion factors 2005 2018 2018 Source: GEIH. The lack of recovery can be attributed to persistent structural barriers that impede access for women with children and those with lower educational levels, who face fewer economic opportunities. Women with children under 12 years of age and those with lower levels of education face persistent structural barriers that limit their economic opportunities. Higher education is a protective factor against job losses for women compared to those without it, although there is still a significant gap between women with and without higher education (as shown in Figure 11). In terms of labor participation (Figure 12), there is a wide gap between women with and without higher education, which can be attributed to the latter being more concentrated in the informal sector (Figure 13). 7 Columns (1) and (2) compare the result of the 1st semester of 2021 with both expansion factors, where the gap generated by the change in the projections can be seen. It is particularly strong over the GPR given the change in the working-age population decision (10 to 15 years). 8 In this comparison, all ages are used because the aggregate annex of the GEIH is used. Vulnerability to unemployment increased during the pandemic for people living in urban areas, with ethnic self-identification, and aged between 18-28 years (Figures 14 - 15). Furthermore, there are significant differences in labor incomes and hours worked between men and women. Those without higher education and those with children are particularly vulnerable to larger gaps and worse levels in these labor market indicators (Figure 16). Rural areas and ethnic populations have low labor incomes, below 50%, which is critical. However, higher education can significantly reduce the gap, up to 80 percentage points for women (Figure 18). Finally, the number of women engaged in unpaid family work has declined, while men with children experienced a rise in this type of work in 2020 (Figure 17). The low labor income in rural areas and among the ethnic population is especially critical, with levels below 50%. Higher education improves the income for both groups, with or without children, and for women, it can structurally reduce the gaps up to 80 p.p. (Figure 18). In hours worked per week, the differences are like the previous ones, but proportionally of a smaller magnitude (Figures 19- 20). 5. Methodology The objective of this paper is to assess the influence of the pandemic on preexisting disparities in labor market outcomes by utilizing two methodologies that enable us to approximate the gender- specific impacts of the pandemic on labor participation and income. Estimation of the probability of participating in the labor market Labor participation between men and women presents a high structural gap, as was observed in Section 1. To identify the factors that determine this gap and quantify its importance, we use a probabilistic model to estimate the probability for each group to participate in the labor market based on their characteristics. Given the type of survey, we do not have information that follows individuals over time which limits causal inference. Therefore, we conduct separate estimates (cross-section) for the years 2015, 2017, 2019, and 2021 with the following specification: ( = 1| ∑ + ∑ + > 0) = Φ (∑ + ∑ ) , (1) =1 =1 =1 =1 In this equation, is a dummy equal to 1 if the subject i of the year t within the group g is in the labor force, 0 on the contrary. is a vector containing continuous determinant variables of labor participation such as age, years of schooling, children 0-3 years old, children +3-6 years old, and children +6-12 years old in the household. is a vector of dummies variables containing marital status (1 for married or in a union), rural zone, and dummies for the region of residence. For this exercise, people between 18-59 years of age were selected to limit the results to their working ages and thus avoid self-selection biases related to education and retirement. From the estimation, we obtain the and coefficients (∆( − )⁄∆ ), inputs to predict the marginal effects on the probabilities evaluated over the mean of the independent variables (CME) and the average of the marginal effects (AME). The goal of estimating both effects is to show whether the results are sensitive to the benchmark of the independent variables when evaluating the impact of only changes in one of them. Thus, these estimators allow us to identify the gap between the estimated probabilities for men and women according to specific characteristics and their response to changes in the level of one of the predictor variables. Finally, we estimate the effect of the pandemic as of 2021 on the gaps existing in 2019 in terms of labor participation. The previous estimation also allows us to estimate the Heckman correction equation to correct the self-selection of participation in the labor market over the decomposition of the labor income gaps. Decomposition of the labor income gaps Now, our focus shifts towards estimating the income gaps in labor through two standard decomposition methodologies: the Oaxaca & Blinder (1973) method and the Ñopo (2008) method. These methodologies play a vital role in improving the accuracy of the unadjusted calculation of the gap. They not only address the issue of sampling bias but also break down the gap into components that can be explained by differences in observable characteristics between the two groups, as well as an unexplained component attributed to factors that are not directly observable. To conduct a comprehensive analysis, we incorporated both the Oaxaca-Blinder and Ñopo decomposition methodologies. The Oaxaca-Blinder approach, a seminal parametric method, provides an overview of the average unexplained gender wage gap between male and female workers in the dataset. However, it has limitations in terms of robustness, as it does not account for differences in individual characteristics across the entire distribution, thus hindering comparability between groups. To address this limitation, we employed the non-parametric Ñopo decomposition approach (2008). This approach allows us to estimate gender wage gaps across the entire distribution of individual characteristics and further disentangles the gap into groups that share common characteristics and those that do not. The latter approach is particularly robust in identifying the gap, especially given the significance of gender differences in observable characteristics in Latin American countries, as highlighted by Ñopo (2008), citing the work of Blau and Ferber (1992). By combining both methodologies, we were able to identify gender wage gaps within the common support and outside the common support for Colombia. This provides crucial evidence for making meaningful cross-country comparisons and deepening our understanding of the factors contributing to income disparities between genders. Oaxaca Blinder decomposition This method estimates the labor income equation for both groups to calculate the gap and its components based on observable characteristics, as shown in equation 2. ln = ∑ +∑ + ; = {(), ()}(2) =1 =1 In this equation, W represents the real labor income (per hour or monthly) of the individual i in the year t for the group g, and vectors that carry continuous variables and dummy determinants of labor income, respectively. is the error term. Finally, and are coefficients that capture the impact of each determinant by gender on labor income, these coefficients then allow us to estimate the decomposition components as specified in equation 3. ̂ − ̂ + ( ) ( = (ln ) − (ln ) = (( ) − ( )) ̂ ) (3) ̂ : is related to the differences in endowments that Explained component, (( ) − ( )) men and women may have in observable characteristics ( ). Demographics include age, region of residence, marital status, and metropolitan area; human capital such as years of education and experience; and work-related such as employment relationship (self-employed, employee), economic sector, hours worked, part-time work. ̂, ): this component is associated with unobservables ̂, − Unexplained component, (, )( that cannot be taken into account in the estimation, such as excluded variables like certified experience, type of work specialization, productivity, and skills, among others; non-measurable or observable elements such as bias and discrimination; and errors in the measurement of the variables. Using the specification to estimate the probability of participating in the labor market (equation 1) as a function of the observable variables used in the primary labor income equation and considering the inclusion of having children by age ranges as an exclusion restriction to reduce endogeneity in the estimation of the decomposition, we estimate the Mills ratio (, ) for each group, which is finally included in the income equations to estimate the decomposition (equation 4). ̂ + = (( ) ̂ ) − (( ) ̂ ̂ + ̂ ) ̂ ̂ − ̂ + ( ) ( = (( ) − ( )) ̂ ) + ( ̂ − ̂ ̂ ) (4) ̂ Decomposition of Ñopo (2008) This non-parametric method of decomposition addresses some limitations of the Oaxaca-Blinder method. First, it considers differences throughout the income distribution and not only in the average. Second, it restricts the analysis to comparable individuals, so it does not assume a functional form on the relationship between characteristics and income (Urquidi et al., 2020). Ñopo WGt = (Δ + Δ + Δ ) + Δ0 (5) Unexplained (∆0 ): this component is comparable to the unexplained component of the Oaxaca- Blinder method. Explained (∆ ): the income difference with the observable characteristics (∆ ) in the common support. This component corresponds to the Oaxaca-Blinder explained component. Explained outside of the common support (∆ + ∆ ): The second and third component is the part of the gap that can be explained by the differences between two groups of men (∆ ), ( ) and two groups of women (∆ ), ( ): Those with characteristics that can be matched (common support, group 1) and those that cannot (group 2). The estimations use the following categorical variables in the specification (see the table 2 for category division): Union or married, formal job, rural housing, Part-time job (<40 hours per week), # Children 0-6 years old at household, experience (see table 2), age, area, sector, education and labor relation. 6. Results Probability of participating in the labor market In this section, we present the results of the estimation of the probabilistic model using the transformation of coefficients known as conditional marginal effect at mean (CME) and average marginal effect (AME). Figures 21 to 24 describe these indicators for the number of children in two age categories: 0-3 years old and +3-6 years old. It is evident that in 2019, as the number of children in the household increases, the probability of women's participation in the labor market significantly decreases in households with children aged 0-3 years (Figures 21 and 22). The probability can drop by up to 40 percentage points, regardless of whether the CMEs are evaluated over the mean or the AME. In households with children aged 3-6 years, the estimated probability also decreases with the number of children in 2019 (Figures 23 and 24). However, the confidence intervals do not indicate statistical significance for this decrease. In 2021, the post-pandemic year, the probability of women participating in the labor market dropped sharply in all estimates, including the case of children aged 3-6 years, where the reduction in probability becomes significant in both CME and AME calculations. These results provide important evidence of women's role as caregivers at home and the implications this has on their labor market participation. Another variable considered a determinant in the literature on women's labor participation is household per capita income (see Goldin, 1994). Our estimates reveal that the gender gap is significant in the poorest households, but as income increases, this gap exponentially narrows. Figure 25 shows the CMEs for the mean of this relationship. We observe that in households where per capita income is at the extreme poverty line (161,099 COP), women have a probability of participating in the labor market of 75% in 2019, compared to 95% for men, resulting in a 22- percentage point difference. At the moderate poverty line (350,000 COP), the gap narrows to 17 percentage points, and in households with a per capita income of 1 million COP, the gap is reduced to 13 percentage points. Beyond that, the gap stabilizes at 10 percentage points. This finding is particularly significant as it highlights that in situations of poverty or vulnerability, men tend to assume income-generating roles while women's responsibilities revolve around caring for the home and family. Consequently, women acquire fewer or no skills and experience in productive sectors compared to men, leading to greater economic dependence for partnered women. Again, the pandemic exacerbated this gap, but only in lower-income households. Regarding schooling, the results at the mean in 2019 indicate a probability of approximately 70% for women with no education, implying a gender gap of around 30 percentage points compared to men in the same group (Figures 27-28). As years of education increase, the probability rises and the gap shrinks to around 30 percentage points. However, men consistently have higher participation rates at any level of education. During the pandemic, women with less than 15 years of education were less likely to participate, with activity levels decreasing as the level of education approaches zero. When examining participation gaps by age groups, both the CMEs (Figure 29) and AMEs (Figure 30) reveal higher gaps in labor market participation for older age groups (+20 percentage points), while for individuals in their 20s, the gap is around 10%. Although it is positive that the intergenerational gap has reduced, it is concerning that it still persists. Older women were particularly impacted during the pandemic, likely due to their predominant roles in caregiving activities and household chores. For women residing in rural areas (Figure 31), the AME indicates an 8-percentage point decrease in the likelihood of labor market participation in 2019 compared to urban areas. In 2021, this probability also declined for women in urban areas. Furthermore, being married or in a union (Figure 31) leads to a drop of approximately 13 percentage points in the likelihood of women's labor market participation, while slightly increasing it for men. This dynamic contributes to a structural expansion of the gender gap. These findings highlight the persistence of gender disparities in labor market participation, with certain factors such as age, rural residence, and marital status exacerbating these gaps. Efforts to address these inequalities and promote gender equity in the labor market remain crucial. Gender labor income gaps First, we analyze the Oaxaca-Blinder decomposition results. Figure 33 illustrates the decomposition of real hourly labor income using the Heckman correction from 2015 to 2021. The total income gap for women in 2015 and 2017 was 11% and 9%, respectively. However, the gap sharply decreased to 2% in 2019 (non-comparable result) and returned to 3% in 2021. The unexplained component plays a crucial role in the overall gap, decreasing from 37% in 2015 to 25% in 2019. This reduction can be attributed to differences in discriminatory biases,9 changes in the population group composition, and measurement issues. Segmenting the data by age groups, it is evident that the high gap is primarily driven by the unexplained component in the 29-59 years old group. For this group, the gap decreased from 15% in 2015 to 5% in 2019, while no significant differences were observed in the 18-29 years old group. The coefficients of age, schooling, and experience exhibited notable variations, suggesting changes in the composition of the labor force that led to lower biases against women. The explained component (Table 3) has an average value of -23.5%, with a slight increase of 1.6 percentage points in the pandemic year of 2021. In terms of endowments, women could have a real hourly labor income that is 20% higher than men's. Table 3 highlights the key determinants contributing to the gender gap in 2021, including differences in schooling (-9%), the proportion of people living in rural areas (-3.3%), occupation in formal positions (-2.3%), type of employment relationship (-2.4%), and part-time work (-3.8%, a decrease compared to 2019). For men, only the economic sector shows a favorable explained gap of 1%. Among these factors, schooling has become the most important determinant that contributes to the gender gap in favor of women, increasing in significance over time. The Ñopo methodology yields smaller estimates for the gender income gap but maintains a decreasing trend over time (Table 4). Interestingly, in 2020-2021, the gap turned favorable to women. This difference can be explained by the growth of the men CEO effect, capturing a gap component outside the common support. Furthermore, the unexplained gap experiences a 1- percentage-point reduction between 2019 and 2021, contributing to gap narrowing. The common support remains around 60% in each period, a significant proportion that supports the exercise in calculating gaps between comparable populations. While estimates of hourly labor income are standard in the literature, capturing the returns of the production process unaffected by part-time work, we also examined the dynamics of the gap with monthly earnings. The Oaxaca-Blinder monthly estimates reveal a substantial growth in the gap compared to hourly estimations (29.3% on average, 2019-2021), primarily driven by the unexplained component (Figure 34). This gap consistently decreased since 2015, even during the pandemic. Notably, the reduction in the gap has been mainly due to a decrease in men's labor income rather than favorable reasons. Reviewing differences in coefficients, a significant gap estimated for the number of hours worked closes over time, aligning with the overall gap narrowing. In terms of the explained component (Table 5), it is evident that hours worked is an endowment favoring the gap for men (19% in 2015, 14% in 2021), while for women, education, 9 Changing social norms may not effectively address gender bias. Research by Tribin et al. (2022) using Colombian data shows that 70% believe women are suited for domestic work, despite women spending more time on caregiving. Improved working conditions, education, and income correlate with greater autonomy for women. experience, and type of employment relationship serve as endowments that would explain the gap in their favor in the absence of unobservable factors. Under the Ñopo approach (Table 3), the monthly income gap is reduced by 50% by 2015, consistent with the Oaxaca-Blinder estimate. Once again, the unexplained component is the key factor explaining this gap, tending to align between the methodologies from 2017 onwards (with a difference of 1.4 percentage points in 2019). The explained component plays a smaller role in reducing the overall gap, but it consistently favors women (15% on average). The CEO and MAID effects do not significantly impact the aggregate results as they tend to counteract each other. The common support does not undergo significant changes in the monthly exercise. Following the impact of the COVID-19 pandemic in 2020, the decrease in labor income disparity persisted under the Ñopo decomposition method. Specifically, the reduction in monthly labor income was 10%, while the hourly labor income decreased by -2.4%. The decline in the wage gap can be attributed to a combination of factors, including the reduction of unobserved issues and improvements in women's relative endowments compared to men. 7. Public programs to close gaps Public policy in Colombia, through various programs and strategies, aims at addressing gender gaps. These initiatives aim to improve the conditions of women in the labor market, enhance their demand, and promote their presence in occupations with better prospects and rewards. Table 6 presents a classification of recent measures.10 These policies address areas such as health, education, social protection, and access to better employment opportunities. The strategies include promoting gender equality training, flexible education models, and financing for women's access to higher education, particularly in non-traditional fields like STEM. Other measures focus on removing barriers for women in the job market, such as raising awareness among companies, providing payroll subsidies, and offering tax incentives. Additionally, efforts are made to empower women by improving access to resources, providing technical advice, and facilitating networking opportunities. Measures also aim to enhance women's participation in decision-making processes, political empowerment, and the prevention of violence against women in politics, through initiatives like the Mujer Emprende Fund and the Casas de las Mujeres Empoderadas. Finally, we examine some cross-sectional policies related to the first four axes aimed at improving women's working conditions through targeted resource allocation and incentives to close gender gaps. The "Pact for Equity for Women" is an initiative with the objective of promoting equal access and participation of women in the labor market, enabling them to improve their economic status in an environment free from gender-based violence. One of its expected outcomes is a 4- percentage-point reduction in the income gap below the global level. The "Mecanismo de Protección al Cesante" (Unemployed Protection Mechanism) serves as an instrument to promote 10 This categorization is based on the World Bank's four key axes for gender equality promotion (World Bank Group, 2015). social protection for workers by ensuring access to social security for the unemployed through various funds and solidarity mechanisms. A notable finding from this review of programs is the lack of national measures addressing barriers to accessing care and redistributing care work within households, including the promotion of the care economy, which disproportionately falls on women with lower levels of education. Existing programs, managed by the Colombian Institute of Family Welfare (ICBF), primarily focus on childcare and time release, without providing options for elderly care or complementary measures that facilitate the effective integration of caregivers into the care and labor sectors. 8. Conclusions The gender gap in labor markets presents a significant challenge, particularly impacting vulnerable women, and Colombia is no exception, with the pandemic exacerbating this issue. This study examines recent trends in Colombian labor outcomes within the context of COVID-19. Our focus is primarily on gender gaps in the labor market and the factors influencing them, particularly in sectors where women are more vulnerable, such as home care. Utilizing data from the GEIH, decomposition methodologies, and probabilistic models, we find that the economic recovery observed in 2021 did not close the gender gaps, indicating underlying structural factors. Moreover, the pre-pandemic gender gaps in terms of participation and unemployment were intensified by the crisis. Interestingly, gender gaps in labor income and hours worked decreased during the crisis, but this was primarily driven by a significant decline in men's levels rather than improved conditions for women. Another contributing factor to women's slow recovery in the labor market is the moderate recovery of the informal sector in 2021, in which women have a significant presence. However, there are indications of a more favorable labor market recovery for women in 2022. The analysis reveals that several factors disproportionately affect women and impact their probability of participating in the labor market. Having children in the household, low levels of schooling, and household income are among the key determinants. Women with children can experience a significant decline in the probability of labor market participation, up to 40 percentage points, whereas this remains unchanged for men. Women at the poverty line have a 22-percentage point difference compared to men, and among women with lower levels of education the gender gap widens. These gaps tend to narrow with fewer children, higher income, and higher education levels. Other factors such as age, rurality, and marital status also contribute to significant gaps between men and women, which widened during the pandemic without signs of recovery in 2021. Access to quality childcare appears to be a barrier for these groups. Using the Oaxaca-Blinder approach for gap decomposition, the average hourly income gap decreased from 11% to 3% between 2017 and 2019, a level that remained in 2021. However, further analysis reveals that this result is a net outcome of two key factors: the explained gap, which favors women due to their endowments, primarily education level, and the unexplained gap, which favors men and may be associated with discriminatory biases. Reducing the unexplained component could result in higher hourly labor income for women. The Ñopo (2008) results align with the previous findings but with lower values in the components. When considering monthly labor income, the gap increases significantly for women, averaging 30% between 2019 and 2021. This can be attributed to the importance of hours worked, where women tend to have a higher proportion of part-time jobs. This suggests that the number of hours worked (extensive margin) is a critical determinant of the labor income gap, rather than the hourly income (intensive margin). The review of current government programs aimed at closing the gender gaps highlights a strong focus on promoting economic opportunities for women, particularly in the context of the pandemic. Strategies such as payroll subsidies, hiring incentives, and entrepreneurship funds have been implemented. However, there is a gap in addressing barriers to access to care and redistributing care work within households, which disproportionately affects women with lower levels of education. Existing programs primarily focus on childcare without complementarities to promote the labor market integration of caregivers. Based on the analysis, several key recommendations for improving public policies in this area emerge: 1. Implement a National Care System that integrates benefits and social protection programs to enhance the quality of life for individuals with high dependency levels and their caregivers. This system should also promote women's participation in education and the labor market. 2. Address the correlation between having children and labor gaps by promoting policies that ensure access to care, family planning, advice, and mentoring. Addressing these issues can help break the cycle of poverty and mitigate the impact of gender norms. 3. Advance the mandate of the Intersectoral Commission for Public Care Policy to facilitate coordination and implementation of care policies at the national level. This coordination can lead to more efficient resource utilization and better outcomes. 4. Recognize the importance of local leadership in narrowing gender gaps. Local and regional knowledge and engagement are crucial for effectively challenging culturally entrenched discriminatory norms and driving change11 from within communities. Consequently, local leaders can have a significant role in promoting change. 5. Enhance public information systems with gender breakdowns to enable better diagnosis and design of public policies. The findings also emphasize the vital role of education in improving working conditions for women. Higher levels of education are associated with reduced income gaps, increased labor 11 Lachapelle (2020) provides empirical evidence, based on self-reported data, which demonstrates the significance of training programs for leaders in enhancing their engagement with their community and objectives. Furthermore, Nwaze and Mohammed (2013) present both quantitative and qualitative evidence supporting the crucial role of traditional and religious leaders in promoting initiatives towards the eradication of polio in Nigeria. market participation, and lower unemployment rates. However, occupational segregation remains a challenge, resulting in lower remuneration for jobs predominantly held by women. 9. References Aina, C., Brunetti, I., Mussida, C., & Scicchitano, S. (2021). Who lost the most? Distributive effects of the Covid-19 pandemic (No. 65; INAPP WP). www.inapp.org Argys, L. M., & Averett, S. L. (2022). Women in the Workforce: What Everyone Needs to Know. Oxford University Press. Berniell, I., Gasparini, L., Marchionni, M., & Viollaz, M. (2021). The Role of Work-from-Home in the Gender Asymmetries of COVID-19 (No. 9887; Policy Research Working Paper). http://www.worldbank.org/prwp. Blau, F. D., Brummund, P., & Liu, A. Y. H. (2013). Trends in occupational segregation by gender 1970–2009: Adjusting for the impact of changes in the occupational coding system. Demography, 50(2), 471-492. Blau, F. D., & Winkler, A. E. (2017). Women, work, and family (No. w23644). National Bureau of Economic Research. Bluedorn, J., Caselli, F., Hansen, N.-J., Shibata, I., & Tavares, M. M. (2021). Gender and Employment in the COVID-19 Recession: Evidence on "She-cessions" (WP/21/95; IMF Working Paper). Bonacini, L., Gallo, G., & Scicchitano, S. (2021). Will it be a shecession? The unintended influence ofworking from home on the gender wage gap relatedto the COVID-19 pandemic (No. 771; GLO Discussion Paper). http://hdl.handle.net/10419/229149 Contreras-Gonzalez, I., Oseni, G., Palacios-Lopez, A., Pieters, J., & Weber, M. (2022). Inequalities in Job Loss and Income Loss in Sub-Saharan Africa during the COVID-19 Crisis (No. 10143; Policy Research Working Paper). http://www.worldbank.org/prwp.Theauthorsmaybecontactedatapalacioslopez@worldbank.org. Cortés, P., & Pan, J. (2020). Children and the remaining gender gaps in the labor market (No. w27980). National Bureau of Economic Research. Cucagna, E. & Romero, J. (2021). The Gendered Impacts of COVID-19 on Labor Markets in Latin America and the Caribbean. World Bank, Washington, DC. © World Bank. https://openknowledge.worldbank.org/handle/10986/35191 License: CC BY 3.0 IGO. Deole, S. S., Deter, M., & Huang, Y. (2021). Home Sweet Home: Working from home and employee performance during the COVID-19pandemic in the UK (No. 791; GLO Discussion Paper). Dolado, J. J., Felgueroso, F., & Jimeno, J. F. (2003). Where do women work?: Analyzing patterns in occupational segregation by gender. Annales d'Economie et de Statistique, 293-315. Doorley, K., O'donoghue, C., & Sologon, D. M. (2021). The Gender Gap in Income and the COVID- 19 Pandemic (No. 14360; IZA Discussion Papers). www.iza.org Fairlie, R. W., Couch, K., Xu, H., & Xu Judd, H. (2021). The Evolving Impacts of the COVID-19 Pandemic on Gender Inequality in the U.S. Labor Market: The COVID Motherhood Penalty (No. 29426; NBER WORKING PAPER SERIES). García-Rojas, K., Herrera-Idárraga, P., Morales, L. F., & Ramírez-Bustamante, N. (2020). (She)cession: The Colombian female staircase fall (No. 1140; Borradores de Economía). Goldin, C. (1994). THE U-Shaped Female Labor Force Function in Economic Development and Economic History (No. 4707; NBER Working Paper Series). Goldstein, M., Gonzalez, P., Papineni, S., & Wimpey, J. (2022). Childcare, COVID-19, and Female Firm Exit Impact of COVID-19 School Closure Policies on Global Gender Gaps in Business Outcomes (No. 10012; Policy Research Working Paper). http://www.worldbank.org/prwp. Grove, W. A., Hussey, A., & Jetter, M. (2011). The gender pay gap beyond human capital heterogeneity in noncognitive skills and in labor market tastes. Journal of Human Resources, 46(4), 827-874. Ham, S. (2021). Explaining Gender Gaps in the South Korean Labor Market During the COVID-19 Pandemic. Feminist Economics, 27(1–2), 133–151. https://doi.org/10.1080/13545701.2021.1876902 Hill, R., & Köhler, T. (2021). Mind the gap: The distributional effects of COVID-19 on gender wage inequality in South Africa. ILO (2019) Women in the World of Work. Pending Challenges for Achieving Effective Equality in Latin America and the Caribbean (Thematic Labour Overview 5). Lima: International Labor Organization Regional Office for Latin America and the Caribbean, 188. Retrieved from: https://www.ilo.org/wcmsp5/groups/public/---americas/---ro- lima/documents/publication/wcms_736930.pdf Juárez, L., & Villaseñor, P. (2022). Effects of the COVID-19 Pandemic on the Labor Market Outcomes of Women with Children in Mexico. https://www.inegi.org.mx/programas/eness/2017/ Kikuchi, S., Kitao, S., & Mikoshiba, M. (2021). Who suffers from the COVID-19 shocks? Labor market heterogeneity and welfare consequences in Japan. Journal of the Japanese and International Economies, 59. https://doi.org/10.1016/j.jjie.2020.101117 Lachapelle, P. (2020). Assessing the potential of community foundation leadership through a new conceptual lens. Community Development, 51(2), 93-106. LaSalle, M. (2021). Where Are They Now? Workers with Young Children during COVID-19. Lim, K., & Zabek, M. (2021). Women's Labor Force Exits during COVID-19: Differences by Motherhood, Race, and Ethnicity. Finance and Economics Discussion Series, 2021(066), 1–39. https://doi.org/10.17016/FEDS.2021.067 Mision de Empalme de las Series de Empleo y Pobreza, MESEP, (2010), Metodología de Cálculo de las Cifras de Pobreza, Indigencia y Desigualdad 2009. MESEP, Bogotá, Colombia. Nwaze, E., & Mohammed, A. (2013). An impact evaluation of the engagement of traditional and religious leaders in the Nigerian polio eradication initiative. Sch J Med, 3(5), 53-63. Ñopo, H. (2008). Matching as a tool to decompose wage gaps. The review of economics and statistics, 90(2), 290-299. OECD. (2013). Gender equality, women’s empowerment and development effectiveness: issues brief 7. https://www.oecd.org/dac/gender-development/_%20Issues%20Brief%207.pdf Pitts, M. (2021). Where Are They Now? Workers with Young Children during COVID-19 (No. 10– 2021; Federal Reserve Bank of Atlanta Policy Hub Papers). Qian, Y., & Fuller, S. (2020). COVID-19 and the gender employment gap among parents of young Children. Canadian Public Policy, 46, S89–S101. https://doi.org/10.3138/CPP.2020-077 Singh, V., Shirazi, H., & Turetken, J. (2022). COVID-19 and gender disparities: Labour market outcomes. Research in Economics, 76(3), 206–217. https://doi.org/10.1016/j.rie.2022.07.011 Tribín, A., Gómez-Barrera, A., Pirela-Rios, A. (2022). Distribución del cuidado, roles de género y poder de negociación en Colombia: Un análisis a partir de la ENUT 2020-2021. Informe Quanta - Cuidado y Género. Recuperado de https://cuidadoygenero.org/distribucion-cuidado-enut Tverdostup, M. (2022). COVID-19 and Gender Gaps in Employment, Wages, and Work Hours: Lower Inequalities and Higher Motherhood Penalty. Comparative Economic Studies. https://doi.org/10.1057/s41294-022-00198-z Urquidi, M., Valencia, H., & Durand, G. (2020). Brecha de ingresos laborales por género en Bolivia: un análisis de su evolución en el periodo 1993 a 2018 (No. 1986; Nota Técnica). Viollaz, M., Salazar-Saenz, M., Flabbi, L., Bustelo, M., & Bosch, M. (2022). The COVID-19 Pandemic in Latin American and Caribbean Countries: The Labor Supply Impact by Gender (No. 15091; Discussion Paper Series). www.iza.org World Bank. 2023. From Infection to Inflation. LAC Poverty and Labor Brief. © Washington, DC. Zhou, M., & Kan, M. Y. (2021). The varying impacts of COVID-19 and its related measures in the UK: A year in review. PLoS ONE, 16(September 09). https://doi.org/10.1371/journal.pone.0257286. Appendix. Figure 1. Labor force participation rate, ages 18-59 Figure 2. Unemployment rate, ages 18-59 Figure 3. Gender gap in the labor force Figure 4. Gender gap in unemployment rate, participation rate, ages 18-59 in 23 major cities ages 18-59 in 23 major cities Figure 5. Average real monthly labor income, Figure 6. Average hours worked by weeks, ages 18-59 ages 18-59 Figure 7. Gender gap in average real monthly Figure 8. Gender gap in average hours worked labor income, ages 18-59 in 23 major cities by weeks, ages 18-59 in 23 major cities Figure 9. Women industry participation 2019 Figure 10. Formal women industry vs. employment growth 2021-2019, all ages participation 2019 vs. growth in job offers 2021-2019, all ages Figure 11. Unemployment rate by Figure 12. Labor force participation rate by educationlevel, ages 25-59 education level, ages 25-59 25 100 20 90 15 80 10 70 5 60 0 50 2017 2019 2020 2021 2017 2019 2020 2021 2017 2019 2020 2021 2017 2019 2020 2021 No higher education Higher education No higher education Higher education Men no children Women no children Men children Women children Men no children Women no children Men children Women children Note: children <12 years at home. Note: children <12 years at home. Figure 13. Employed women with children vs. Figure 14. Unemployment rate by informality rate by sector, all ages socioeconomic characteristics, 2019 Ages 18-28 22.9 13.8 Urban Children 14.0 15.7 9.5 4.5 8.1 13.9 9.8 Ethnic self- No higher 22.2 recognition (black education and similar) Men Women Note: mspline regression to estimate the relationship betweeen Note: ethnicity information is for 2020 due to variables. Children <12 years at home. availability of information. children <12 years at home. Figure 15. Unemployment rate by Figure 16. Average real monthly labor income socioeconomic characteristics, 2021 by socioeconomic characteristics 1,600 Ages 18- 28 1,400 28.2 Real labor income (thousands) Men 2019 Women 2019 1,200 16.9 Men 2021 Women 2021 1,000 Urban Children 18.7 20.5 800 12.3 6.5 600 400 10.3 11.2 Ethnic 200 19.1 self- No higher 23.0 recognitio 0 education n (black Ethnic self- No higher Rural Children Ages 41-59 and recognition education similar) (indigena) Men Women Figure 17. unpaid family workers (UFW) indicators, ages 18-59 6.0 50 40 5.0 Growth rate UFW (%) & Inactivity rate (%) 30 20 Occupation UFW (% total) 4.0 10 3.0 0 -10 2.0 -20 -30 1.0 -40 0.0 -50 2017 2019 2020 2021 2017 2019 2020 2021 No children Children Men occupation UFW Women occupation UFW Men GRUFW Women GRUFW Men IR Women IR Note: IR: inactive rate; GRUFW: growth rate unpaid family workers. Chldren <12 years at home. Figure 18. Median real monthly labor income, Figure 19. Hours worked per week by ages 25-59 socioeconomic characteristics 60 Men 2019 Women 2019 Men 2021 Women 2021 50 40 Hours for week 30 20 10 0 Ethnic self- No higher Rural Children Ages 41-59 recognition education (indigena) Figure 20. Average hours worked by week, Figure 21. CME at mean, children in the ages 25-59 household ages 0-3 Figure 22. AME, children in the household Figure 23. CME at mean, children in the ages 0-3, ages 15-59 household ages +3-6, ages 15-59 Figure 24. AME, children in the household Figure 25. CME at mean, Household per capita ages +3-6, ages 15-59 income, ages 15-59 Note: Colombia’s extreme poverty line 161,000 pesos in 2021 and 147,404 pesos in 2019 (2021 prices). Figure 26. AME, Household per capita Figure 27. CME at mean, year of schooling, income, ages 15-59 ages 15-59 Note: Colombia’s extreme poverty line 161,000 pesos in 2021 and 147,404 pesos in 2019 (2021 prices). Figure 28. AME, year of schooling, ages 15-59 Figure 29. CME at mean, age, ages 15-59 Figure 30. AME, age Figure 31. AME, household zone Figure 32. AME, Marital status, ages 15-59 Figure 33. Oaxaca-Blinder decomposition of real labor income per hour, ages 18-59 40% 37% 30% 31% 28% 20% 25% 25% 10% 11% 9% 3% 3% 2% 0% -10% -21% -21% -20% -21% -24% -20% 2015 2017 2019 2020 2021 Explained Unexplained Estimated gap Note: gap is estimated as a proportion of women's income. All estimates are significant at 1% (except 2020, which is 10%). Figure 34. Oaxaca-Blinder decomposition of real monthly labor income, ages 18-59 50% 52% 40% 40% 33% 28% 30% 53% 27% 20% 39% 32% 34% 31% 10% 0% 0% -2% -1% -8% -6% -10% 2015 2017 2019 2020 2021 Explained Unexplained Estimated gap Note: gap is estimated as a proportion of women's income. All estimates are significant at 1% (except 2020 which is not significant). Table 2. Descriptive Statistics by Gender and Year, ages 15 - 59 Men Women Variable 2015 2019 2020 2021 2015 2019 2020 2021 Continuous (mean) Age 36.1 36.2 36.2 36.1 36.3 36.3 36.3 36.4 # Children 0-3 years old at household 0.30 0.30 0.30 0.20 0.30 0.30 0.30 0.30 # Children +3-6 years old at household 0.20 0.20 0.20 0.20 0.20 0.20 0.20 0.20 # Children +6-12 years old at household 0.40 0.40 0.40 0.40 0.50 0.50 0.40 0.40 Years of education 9.2 9.6 9.7 9.9 9.6 10.2 10.2 10.4 Years of experience 20.2 20.0 19.9 19.8 20.1 19.8 19.6 19.7 Real monthly labor income - imputed 1,239,760 1,225,853 1,088,381 1,137,485 986,352 1,029,838 989,716 1,031,265 Real hourly labor income - imputed 6,671 6,615 5,914 6,119 6,313 6,539 6,032 6,272 Hours worked per week 43.0 41.0 36.6 38.7 24.9 23.5 20.0 21.4 Categoricals (%, total) Rural 22.3 21.9 32.8 22.1 19.1 19.0 29.0 19.3 Union or married 56.9 56.5 56.9 55.2 58.5 58.6 59.5 57.2 Poor 20.6 28.4 32.5 32.1 24.5 32.8 37.3 37.4 Age category 18-24 years old 21.6 20.8 20.8 20.4 21.3 21.0 21.5 20.7 +25-44 years old 41.5 42.6 42.6 43.4 41.1 42.1 41.9 42.0 +45-59 years old 36.9 36.6 36.5 36.2 37.6 37.0 36.6 37.3 Children at household (Yes=1) Children 0-3 years 22.1 21.7 22.1 19.8 26.9 26.2 26.5 24.2 Children +3-6 years 17.5 17.3 18.0 16.7 21.2 21.3 22.1 20.9 Children +6-12 years 27.0 25.5 25.3 24.5 31.2 30.4 30.3 29.7 Education category No education 15.5 12.8 13.7 11.1 13.5 10.2 10.6 8.9 Elementary 24.5 22.5 23.2 21.1 22.0 19.5 20.0 18.1 Secondary 6.6 6.4 6.2 6.3 6.6 5.9 6.1 5.7 High School 35.4 38.7 38.7 40.9 35.2 39.5 39.6 41.5 Technical and technological 9.4 9.6 8.8 9.8 12.6 12.7 12.0 13.0 University 6.1 7.1 6.7 7.9 7.2 8.9 8.5 9.6 Postgraduate 2.5 2.9 2.6 2.9 2.9 3.3 3.1 3.2 Experience No experience 0.4 0.4 0.3 0.3 0.7 0.6 0.6 0.5 +0-3 years 7.2 7.1 7.0 7.2 8.0 8.1 8.0 8.1 +3-6 years 10.0 10.0 9.8 10.2 9.8 10.0 10.3 10.4 +7-10 years 11.7 12.3 12.0 12.3 11.3 12.1 12.3 12.1 +10-15 years 12.9 13.1 13.5 13.4 12.7 12.9 13.1 13.1 +15 years 57.8 57.2 57.5 56.6 57.5 56.1 55.6 55.8 Part-time job (<40 hours per week) 11.9 12.6 13.3 12.1 34.1 33.2 33.3 30.4 Formal 39.9 41.3 36.5 38.7 38.7 41.1 40.1 41.4 Labor status Employed 86.0 83.6 80.3 80.4 60.8 58.1 49.7 52.0 Unemployed 6.2 7.6 9.6 9.7 8.5 9.7 12.2 12.1 Inactive 7.8 8.8 10.1 9.9 30.7 32.2 38.1 35.9 Labor relation Private employee 48.6 50.2 47.7 47.8 51.5 52.6 49.6 51.1 Government employee 3.8 3.6 3.3 3.3 5.0 4.4 4.8 4.8 Self-employed 47.7 46.2 49.0 48.9 43.5 42.9 45.6 44.1 Sector Agriculture 19.8 20.1 27.0 20.2 6.3 6.1 9.0 6.1 Electricity, gas, water and mining 2.6 2.6 2.9 2.9 0.8 0.8 1.0 1.0 Industry 12.4 11.3 11.0 11.3 12.2 11.4 10.9 10.4 Construction 12.1 11.6 10.6 11.7 1.1 1.0 1.1 1.1 Trade and vehicle repair 16.9 17.3 15.6 17.4 20.9 21.0 21.2 22.1 Transportation, storage and communications 13.1 12.8 11.8 13.2 3.7 3.1 2.9 3.1 Accommodation and food services 3.8 4.1 3.3 3.8 11.4 12.0 11.9 12.5 Financial and real estate activities 2.7 2.7 2.2 2.7 3.0 3.0 2.8 3.0 Professional and administrative activities 4.3 4.7 4.3 4.9 7.9 8.7 8.3 9.6 Public administration 3.2 3.2 2.8 3.0 3.0 3.1 3.2 3.5 Education and human health 4.6 4.5 3.8 4.3 14.5 14.9 13.6 13.9 Other activities 4.4 4.9 4.6 4.7 15.3 14.9 14.0 13.7 Observations (thousands) 207,102 201,801 123,872 188,251 243,894 235,040 142,129 219,796 Table 3. Explained gap share in real labor income per hour (percentage points), ages 18-59 Variable 2015 2017 2019 2020 2021 Total explained gap -22.6 -23.1 -21.8 -26.7 -23.4 Demographics Age -1.1 -0.6 0.0 -0.4 -0.1 Marital status: married -0.1 -0.1 0.1 0.0 -0.1 Rural area -3.0 -2.4 -2.8 -4.1 -3.3 City -1.9 -1.9 -2.0 -2.0 -1.7 Human capital Schooling -6.8 -7.5 -7.2 -8.4 -9.0 Experience -2.4 -2.4 -3.0 -3.0 -1.7 Work Part-time -5.2 -5.5 -4.5 -3.8 -3.8 Formality -0.2 -0.6 -0.7 -3.2 -2.3 Economic sector -0.2 -0.1 -0.4 0.5 1.0 Employment relationship -1.7 -2.0 -1.3 -2.3 -2.4 Total unexplained gap 36.8 31.4 25.3 27.9 25.4 Total gap 11.1 8.5 3.2 2.2 3.3 Source: GEIH, own calculations. Table 4. Ñopo (2008) decomposition - Real labor income (per hour and month), ages 18-59 Variable 2015 2017 2019 2020 2021 Hour Month Hour Month Hour Month Hour Month Hour Month Total gap Δ 5.7% 25.7% 3.4% 21.2% 1.2% 19.0% -2.0% 10.0% -2.4% 10.3% Not explained Δ0 34.4% 39.2% 27.9% 33.4% 23.7% 31.0% 16.6% 22.7% 22.7% 27.1% CEO effect ΔH 1.8% 0.9% 0.0% -1.1% -1.5% -3.0% -1.3% -1.6% -4.2% -5.5% MAID effect ΔM -5.1% 1.3% -2.3% 3.7% -1.3% 4.7% -1.9% 1.2% -0.8% 4.7% Explained ΔX -25.3% -15.7% -22.2% -14.7% -19.7% -13.6% -15.4% -12.3% -20.1% -15.9% % Men 65% 65% 65% 65% 63% 63% 70% 70% 61% 61% % Women 66% 66% 66% 66% 65% 65% 72% 72% 66% 66% Standard error Δ0 0.008 0.009 0.008 0.008 0.009 0.008 0.009 0.009 0.009 0.009 Table 5. Explained gap share in real monthly labor income (percentage points) , ages 18-59 Variable 2015 2017 2019 2020 2021 Total explained gap 1.6 -0.2 0.5 -6.1 -5.2 Demographics Age -1.1 -0.7 0.0 -0.5 -0.1 Marital status: married -0.2 -0.2 0.0 -0.1 -0.2 Rural area -3.1 -2.5 -3.0 -4.4. -3.5 City -1.9 -1.9 -2.0 -1.9 -1.6 Human capital Schooling -6.6 -7.3 -7.0 -8.1 -8.9 Experience -2.4 -2.4 -3.1 -3.2 -1.8 Work Part-time 19.1 17.7 18.3 17.1 14.8 Formality -0.2 -0.6 -0.7 -3.3 -2.4 Economic sector -0.2 -0.2 -0.5 0.7 1.0 Employment relationship -1.8 -2.1 -1.5 -2.4 -2.5 Total unexplained gap 52.7 39.0 32.4 33.5 30.5 Total gap 52.3 39.7 32.8 28.0 26.6 Source: GEIH, own calculations. Table 6. Public policies that promote the reduction of gender gaps Active programs and policies of Strategy Description the national government 1. Improving human Closing the remaining gender gaps Elementary and secondary school: endowments in health, in health, which includes reducing education, and social maternal mortality and improving 1. Under the framework of the protection. women's access to health care strategy Entornos para la Vida, la services, for instance, in sexual and Convivencia y la Ciudadanía, reproductive health; in education, several programs promote gender which involves getting girls to equality. complete secondary school, 2. Literacy strategies focus on reducing boys' secondary school gender equality and targeting. dropout rates and improving the quality of learning among girls and 3. Modelos Educativos Flexibles. boys, as well as expanding social security networks. Higher education: 1. Promotion strategy to offer professional training in non- traditional programs for women. 2. Financing strategy to promote access and retention in higher education programs that are less traditional for women through the new program Generación E. 3. Por TIC Mujer program. 2. Eliminate restrictions so Focusing on aspects such as safe Signaling: women can access more transportation to and from the and better jobs. workplace; care services for 1. Iniciativa de Paridad de Género children and other family members; Colombia (IPG). training on digital and technological 2. Programa Nacional de Equidad skills needed to access the world of Laboral con Enfoque de Género work and economic opportunities; (PNELG). and the reduction of employment discrimination based on gender. - Raising awareness on labor equality. - Gender equality seals "Equipares". 3. SPE inclusive paths to employment. Payroll subsidies: 4. Programa de Apoyo al Empleo Formal (PAEF). 5. Apoyo a Empresas Afectadas por el Paro Nacional (AEAP). 6. Incentives to create new jobs. Tax incentives: 7. Tax reliefs for creating jobs for vulnerable groups of people. 8. (PNELG) promotion of the Decree 2733 of 2012: tax reliefs for hiring women victims of gender- based violence. 3. Remove the obstacles that Such as land, house, and bank 1. Fondo Mujer Emprende. deprive women of accounts, and facilitate access to ownership and control of financing, technology, and 2. Regalías Mujer. assets. insurance services to make those 3. ALDEA – Fondo Mujer assets profitable. Emprende. 4. Generando Equidad. 5. Autonomous assets. 6. Public purchases with a gender perspective. 4. Fostering the ability and Along with the engagement and Instruction: the women's participation support of men and boys, address in the decision-making challenges related to child marriage, 1. School for instruction in politics process. gender-based violence, social for women. norms, and women's representation, 2. Promotion of non-stereotyped participation and decision-making in activities and equitable the local structures of utilities, relationships. especially water and energy supply in schools, management committees 3. Más mujeres Más democracia. of health care centers and local development committees. Participation and inclusion: 4. Peer lists. 5. Women's empowerment in peace building. 6. Strengthening of territorial gender mechanisms. 7. Pacto por la Equidad de las Mujeres Rurales. Places and protection: 8. Casa de Mujeres Empoderadas program. 9. Protocol for the prevention of violence against women in politics. 5. Cross-sectional to the prior Policies and regulations that outline 1. Pacto XIV “Pacto de Equidad strategies. comprehensive institutional para las Mujeres”. frameworks to promote actions in favor of gender equality. 2. Mecanismo de Protección al Cesante - SPE y FOSFEC. 3. Resolution 758 of March 07, 2016: creates the Gender Subcommittee, which reports to the Comisión Permanente de Concertación de Políticas Salariales y Laborales of the Ministry of Labor. Notes:* Categories from the four key areas of the World Bank's strategies on gender equality (World Bank Group, 2015). Source: Own elaboration based on information from the official websites of the Ministry of National Education, Ministry of Labor, Ministry of Agriculture and Rural Development, National Planning Department, and Presidential Advisor's Office for Women's Equality.