Policy Research Working Paper 11126 Where You Are Born Matters Inequality of Opportunities and Intergenerational Mobility across Colombia’s Territory María Eugenia Dávalos Juan Manuel Monroy Poverty and Equity Global Department May 2025 Policy Research Working Paper 11126 Abstract The circumstances into which individuals are born are incorporate place of birth and a more granular geographic beyond their control, yet they play a significant role in lens to capture the extent of regional disparities. The find- shaping people’s economic opportunities and are thus ings show that 49 percent of the Gini income inequality key drivers of inequality and its persistence over time. is explained by circumstances at birth, and place of birth Understanding the role of place of birth is essential to under- accounts for up to half of these inequalities. Intergenera- standing inequality of opportunities and social mobility, tional mobility measures at the department (province) level both of which directly affect overall inequality. This paper also reveal striking disparities in opportunities across the uses machine learning techniques and data from Colombia, country. These findings underscore the critical role that one of the most unequal countries in Latin America and place of birth plays in perpetuating inequality, providing the Caribbean, to estimate inequality of opportunity and important insights for policies aimed at promoting social intergenerational education mobility indexes. The analysis mobility and reducing territorial disparities. This paper is a product of the Poverty and Equity Global Department. 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 Where You Are Born Matters: Inequality of Opportunities and Intergenerational Mobility across Colombia’s Territory María Eugenia Dávalos1 and Juan Manuel Monroy2 Keywords: Intergenerational Mobility, Education, Inequality of Opportunities, Latin America, Colombia JEL Classification: I2, D3, D63, J6, J62, O15 1 Senior Economist at the World Bank’s Poverty and Equity Global Practice, Latin America and Caribbean region. mdavalos@worldbank.org 2 Consultant at the World Bank’s Poverty and Equity Global Practice. jmonroybarragan@worldbank.org. I. Introduction Income inequality can be driven by factors within an individual's control, such as effort (Roemer, 1993). However, the stark reality contradicts a uniform landscape of opportunity, even under equal effort, as not all individuals are afforded equal prospects to generate income not live in poverty (World Bank, 2005). For instance, less than 15 percent of people born into the bottom half of the education ladder in the median developing country in the 1980s made it to the top quarter (World Bank, 2018), revealing the extent to which a person's socioeconomic status, such as educational attainment or income level, is partly dependent on circumstances at birth, such as their parents' socioeconomic status (Van der Weide et al, 2021). The literature argues that inequality of opportunity—inequality that stems from circumstances beyond one's control, such as place of birth—can hinder human capital accumulation and growth (Marreno and Rodriguez, 2013; Carranza, 2020; Van der Weide et al., 2024). The philosophical stance, supported by scholars like Arneson (1989), Cohen (1989), Roemer (1993, 1998a), and Ramos and Van de Gaer (2012), posits that individuals should only be accountable for factors within their control, and Mahler and Ramos (2019) argue that inequality arising from circumstances is inherently unfair and should be minimized. Increased socio-economic mobility and equality of opportunities can help address long-term inequality and promote social stability (World Bank, 2018). Numerous studies have found that greater relative mobility and higher equality of opportunity are associated with lower inequality (Corak 2013; Brunori et al., 2013; World Bank, 2018). The well-known Great Gatsby curve, for example, finds that higher economic mobility is associated with lower income inequality (Krueger, 2012; Jantti et al, 2006; Corak 2013; World Bank, 2018; DiPrete, 2020; Durlauf et al., 2022). With place of birth as one circumstance that people are born into, the extent to which there is an uneven spatial distribution of resources and services can influence mobility and opportunities (Narayan et al, 2013; Connolly, Corak and Haeck, 2019; Brunori et al, 2023a). High spatial inequalities are reflected, for example, in lower access to quality education and healthcare in certain municipalities within the same country (World Bank, 2024), shaping opportunities for the next generation and resulting in lower relative mobility 3 (Connolly, Corak and Haeck, 2019; Chetty and Hendren, 2018; World Bank, 2018). Conversely, evidence for the United States and Canada shows that higher mobility at the subnational level is associated with less residential segregation, lower inequality, higher quality public schools, and stronger social networks and 3Relative mobility is understood as the extent to which a person's socioeconomic status is independent of circumstances at birth, such as their parents' socioeconomic status, place of birth, or sex. 2 family structures (Chetty et al, 2014; Connolly, Corak and Haeck, 2019; Corak, 2013). Recent studies integrate a subnational focus when studying intergenerational mobility. For instance, within-country heterogeneity is analyzed in Buscha et al (2021) for England and Wales, in Alesina et al (2021) for Africa, in Munoz (2021) for Chile, or in Hong and Gruijters, (2024) for China; and by population groups in Asher et al., (2024) for India. These authors find substantial heterogeneity at the subnational level in mobility. For instance, in the case of Chile the relative mobility in education at the commune level varies from 0.54 in Quemchi to 0.97 in San Pedri de Atacama. The Latin America and the Caribbean region has long been characterized as one of the most unequal regions (Clavijo et al, 2021; Garparini and Cruces, 2021; Brunori, Ferreira and Neidhöfer, 2023). The nature of that inequality matters: more than 44 percent of current income inequality in LAC measured by the Gini coefficient is attributed to inherited factors, ranging from 44 percent in Argentina to 63 percent in Guatemala (Brunori, Ferreira and Neidhöfer, 2023). 4 Within the region, Colombia stands out as the country with highest income inequality (0.546 Gini in 2023), also one of the highest around the world. Four out of five Colombians consider the distribution of income to be unfair or very unfair (data from Latinobarometer 2023). Inequality of opportunity has been estimated at nearly 47 percent of the total inequality measured by the Gini using the Encuesta de Calidad de Vida (ECV) in 2010 (Brunori et al., 2014). 5 Social mobility has also been very low over time, always in the lowest ranges of international comparisons (Angulo et al, 2012) and tends to be higher in regions with higher income levels (Galvis and Meisel, 2014). International comparisons place Colombia as a country with high absolute mobility but very low relative mobility (Van der Weide et al., 2023). One dimension of inequalities in Colombia is that of territorial disparities. Individuals living in certain areas of the country have lower access to services and to economic opportunities, perpetuating poverty. For example, in Colombia, certain municipalities have over 90 percent of avoidable infant deaths and of 10- year-old children who cannot read and comprehend a simple text (World Bank, 2024). Yet, while there is national-level data on socio-economic mobility, there is little information on the heterogeneity of socio- economic mobility within the territory, and no recent estimates on inequality of opportunity and how place of birth matters. 4 Results in Brunori, Ferreira and Neidhöfer (2023) found that when considering the MLD coefficient, the inherited inequality for Latin America ranged from 16 percent in Argentina to 32 percent in Guatemala. In Colombia, inequality of opportunities explains between than 44 and 47 percent of total inequality when using the Gini coefficient and between 16.9 to 18.75 when reporting the MLD. Nevertheless, in this approach estimates for Colombia include a limited list of circumstances where the municipality at birth is not considered. 5Ferreira and Melendez, (2012) have found 20 percent of the total inequality measured by the Mean log Deviation (MLD) using the Encuesta de Calidad de Vida (ECV) in 2010 using Ferreira and Gignoux’s (2011) approach. 3 In this paper we explore two main issues: (i) the extent to which place of birth determines inequality of opportunities and (ii) inter-generational socio-economic mobility in Colombia at the subnational level. First, it proposes new estimates of inequality of opportunity, using more up-to-date data and alternative methods, and considering an additional circumstance relevant to the country context, namely exposure to internal armed conflict. Second, we go beyond national-level measures of intergenerational mobility in education, and develop subnational (department-level) estimates, comparable to those available for a large range of countries, thus allowing for international benchmarking of Colombia’s departments. The analysis uses data from the ECV (Encuesta de Calidad de Vida) 2019 to 2022. Our results reveal an important role of place of birth in determining opportunities and mobility. A high share of income inequality in Colombia is predetermined by circumstances at birth, with place of birth impacting people’s prospects for inter-generational mobility. For example, around 49 percent of per capita household income inequality as measured by the Gini coefficient is explained by parents’ educational background, place of birth, sex, and ethnicity. We also find a significantly higher dependance between parents and children’s incomes in places like Nariño, La Guajira and Vichada (with relative mobility estimated at 0.35, 0.36 and 0.37, respectively), compared to 0.61 in places like Bogotá. This paper contributes to the literature on inequality of opportunities and inter-generational mobility. First, it expands the research that has focused on these themes. For the inequality of opportunity index, we follow the approach used in Brunori, Hufe and Mahler (2023) and applied in Brunori et al (2024), and Atamanov et al (2024). We provide more recent estimates for Colombia, testing for various methodological approaches, and test the inclusion of top income adjustments methodologies to account for the entire distribution. Second, we go beyond national-level estimates on social mobility, in a country that is highly unequal, to zoom in to the subnational level and show the different prospects that people have depending on where they were born. Third, this paper advances the literature by including the additional circumstance of presence of internal armed conflict in the municipality of birth. The evidence for Colombia shows that presence of internal armed conflict is associated with lower access to services, human capital accumulation and perceptions of liberties and freedom (World Bank, 2024; Arjona et al, 2024), which can affect opportunities. The paper is structured as follows. Section II outlines the data and methods used to construct measures of inequality of opportunity and subnational inter-generational mobility for Colombia, and Section III presents the main findings around both. Section IV includes several robustness checks for variables with missing information, sensitivity of the results to changing certain circumstances, and an initial testing of innovative methods of top incomes. Section V finalizes with key conclusions and insights. 4 II. Data and Methods Inequality of Opportunities Following Roemer (1998) and Van der Gaer (1993), the main model of Inequality of Opportunities (IOp) distinguishes between circumstances and effort when measuring welfare, such as income. The most common structure for measuring inequality of opportunity follows the form of inequality in predicted � ), to inequality in observed income, (), where the first refers to predicted income by inherited income, ( � can be seen as the circumstances or absolute IO (Brunori, Ferreira and Salas-Rojo, 2023). In other words, contrafactual inequality in income due solely to circumstances at birth or as a smoothed per capita income that measures the average level within each "type" or subgroup with similar observable conditions (Brunori, Hufe and Mahler, 2023). Then, the Relative Inequality of Opportunity (IOr) follows: �) ( = () There is a growing literature on measuring inequality of opportunities, and applications for several countries (Bourguignon, 2018; Ferreira and Peragine, 2016; Ramos and Van de Gaer, 2016; Roemer and Trannoy, 2013; Brunori, Ferreira and Neidhöfer, 2023; Brunori, Hufe and Mahler, 2023; Atamanov et al., 2024). In this paper we follow the Ferreira and Gignoux (2011) accounting for direct ex-ante IO (the so-called lower bound) and building upon Ferreira and Melendez’s (2012) application for Colombia. 6 The latter used 1997, 2003, 2008, and 2010 surveys and focused on household and labor income to assess the role of municipality size as well as particular regions in driving inequality. They found that circumstances such as parent’s education, municipality size, and rural areas determine disadvantages in terms of inequality of opportunities. Additionally, and understanding the limitations of the latest approach, which potentially leads to create a downward bias, we follow Brunori, Hufe and Mahler’s (2023) approach by estimating conditional inference regression trees and random forest. Following Brunori, Hufe and Mahler (2023), and Atamanov et al.’s (2024) application, the conditional inference regression trees (Conditional Trees) approach consists of partitions of the sample into circumstance types, where a regression tree algorithm divides the data into different nodes (each node being a circumstance). Here, the goal is to predict income outside the sample. At each tree branch, the algorithm 6 It is known as the parametric approach. It uses OLS to regress income at the individual level on the circumstances, determining a set of coefficients to calculate a predicted income on average due to circumstances, where differences in income that are orthogonal to circumstances at birth are attributed to effort (Ferreira and Gignoux, 2011). 5 identifies the circumstance split that results in the most significant income difference. 7 On the other hand, the Random Forest approach consists of hundreds of trees, each utilizing a subset of observations and a subset of circumstances at each node, with the ultimate predictions being the average of all the trees (more methodological details in Brunori, Hufe and Mahler, 2023).The regression tree approach adds more variability to the model, dealing with downward biases more efficiently given that it considers nodes by partitioning the data instead of averages for the entire distribution. Additionally, the forest tree methodology tackles difficulties due to potential small changes in the data that might alter the splitting points, it diminishes arbitrary selection of specifications without any functional form selection needed (Brunori, Hufe, and Mahler, 2023; Atamanov et al., 2024). 8 Refer to the version of the regression tree for Colombia in Annex figure A1. Using more recent data from the ECV (Encuesta de Calidad de Vida, pre- and post-COVID-19 2019 and 2022 surveys), we explore how much individuals’ incomes depend on circumstances at birth. The selection of these two years allows assessing how structural inequality of opportunity is in the presence of such a strong economic shock. As measures of living standards, we focus on per capita income inequality and labor income inequality for the population between 25 and 85 years old with reported income. We include the following circumstances in the measurement of an inequality of opportunities index: (i) Education: the maximum educational level of parents reported in the ECV; (ii) Place of birth: the area of birth (rural or urban); the equivalent region of birth (following Ferreira and Melendez, 2014); the size of the population in the municipality of birth divided into four groups; and a variable that measures the presence of internal armed conflict in the municipality of birth; (iii) Ethnicity: indigenous, NARP (Negro, Afro Colombian, Raizal or Palenquero), or neither; (iv) Sex: man or woman, excluded from the household income specification given that it is individual and not at the household level; when included, it is not statistically significant as in Ferreira and Melendez (2014). 7 This process uses statistical tests to avoid over-fitting. First, the algorithm circumstances that are significantly related to income, with a p-value less than .01 after a Bonferroni adjusting for multiple tests. It then picks the circumstance with the smallest p-value. For this circumstance, it tests all possible ways to split the data to find the one that shows the biggest difference in income between groups. If the circumstance is continuous or ordered, it chooses a splitting point. If it is categorical, it divides the values into two groups. This process continues until no more significant splits are found or the maximum depth is reached (Atamanov et al., 2024). 8 We follow Brunori, Hufe and Mahaler’s (2018) algorithm. They discuss fine-tuning "tree" and "forest" methods to estimate inequality of opportunity. The goal of fine-tuning is to balance downward (underestimation) and upward (overestimation) biases. For trees, fine-tuning involves selecting an α value that determines when tree nodes are split. A less strict α allows more splits, risking overfitting. To prevent this, cross-validation chooses the α value that best predicts out-of-sample results. For forests, fine- tuning selects values for α, the number of predictors considered at each split, and the number of subsamples used to build the trees. Again, a cross-validation process, based on the "out-of-bag" error, finds the combination of values that minimizes out-of-sample prediction error. In both cases, the goal is to find a model that generalizes well to new data, providing more reliable estimates of inequality of opportunity. 6 We define a municipality under internal armed conflict based on historical data from Fernandez (2010), LeGrand (1988) included in the Municipal Panel dataset of CEDE (see Acevedo, and Olivella, 2014), and the Violent Presence of Armed Actors in Colombia report (ViPAA, see Osorio et al., 2019). We consider two periods taken from those data sources, 9 where we assume that people registered in the Household Survey (the ECV) who were born during that period in their respective municipality, were exposed to the presence of internal armed conflict, and subsequently we treat that as a circumstance. 10 This definition has limitations given that individuals have the value of 1 if the time of birth overlaps with periods of internal armed conflict in a municipality. Robustness checks such as widening the age period of child exposure to internal armed conflict could be explored in future versions. We report two measures of inequality: the Mean Log Deviation (MLD) to facilitate comparison with previous studies for Colombia, and the Gini coefficient, which is less sensitive to extreme values and helps address underestimation issues (Aaberge et al., 2011; Palmisano et al. 2022; Brunori et al., 2019; Atamanov et al., 2024). Intergenerational Mobility in Education We build upon the International IGM database constructed by Van der Weide et al. (2024). To ensure methodological consistency, we initially use the Colombia ECV 2013 survey to replicate results for the reference year in the International IGM database. However, this paper draws from more recent data. Specifically, we pool the ECV 2019, 2021, and 2022 to strengthen sample power to be able to disaggregate results at the department level and include department of birth in the analysis. 11 Colombia has 32 departments and 1,112 municipalities. We concentrate on two common concepts in the inter-generational mobility literature: absolute and relative mobility. Absolute mobility refers to the proportion of individuals who achieve a higher level of education than their parents. Relative mobility, on the other hand, indicates the extent to which an individual's socioeconomic success is not influenced by the socioeconomic status of their parents. There are several ways to measure inter-generational mobility. Selecting similar indicators to those in the international 9The first period encompasses the years 1901-1931 and 1948-1953, during which the presence of partisan insurgency by municipality is documented. The second period begins in 1973, with annual data available, and considers attacks by various violent actors, including guerrillas, paramilitaries, and military forces. We estimate the presence of violence in a municipality within a five- year span for any actor. Consequently, we imputed a dummy variable (internal armed conflict presence) to the ECV by municipality and year of birth, assigning a value of 1 when the individual was born in that particular municipality during the specified period. 10 Due to the absence of data for the periods 1932-1947 and 1954-1973, we assume the presence of internal armed conflict only in municipalities where conflict was persistent. This means we consider municipalities that reported conflict presence both before and after these periods. We then imputed this information to household members in the ECV based on their place of birth and year of birth. 11 See annex Figure A9 comparing sample sizes. 7 database allows to draw comparisons across countries, to assess how Colombia’s departments fare in educational mobility compared to countries of different income levels. Among a set of indices estimated in Van der Weide et al (2024) we focus on two. First, the share of respondents with more years of schooling (ℎ ) than both of their parent’s ( ) conditional on parents not reaching the maximum years of schooling as a measurement of absolute mobility: = �ℎ > � < max ) Second, 1-beta, or 1 minus the coefficient from regressing respondent’s years of schooling on parent’s years of schooling as a measurement of relative mobility: = 1 − , ℎ = + ∗ + There is a debate regarding potential bias when using the regression coefficient relative to a correlation coefficient. For instance, Emran et al. (2019) use data from Bangladesh and India to estimate IMG from coresident samples (a sample which is only available for individuals living in the same household with their parents) showing significant downward bias in the regression coefficient, while the correlation coefficient is estimated to have a much smaller bias. Nevertheless, Munoz and Siravegna (2023) show that the correlation coefficient is not always less biased by co-residency compared to the regression coefficient. Moreover, in terms of relative mobility, the Pearson correlation coefficient and rank-based indicators derived from education data appear less reliable for ranking economies compared to the intergenerational regression coefficient, despite having a smaller co-residence bias. Given the findings in the latest study, we report the 1-beta coefficient as a measure of relative IMG. Nevertheless, when computing the correlation coefficient, we found a correlation between both indicators of about 0.81. However, it is important to raise a challenge when dealing with education variables, which is commonly observed in coarse bins. The literature identifies challenges in measuring IMG due to unobservable parent- child correlations within broad education categories, making conventional rank mobility measures potentially biased (Asher et al., 2017). In other words, parent-child correlations within bins may be unobservable. Parents with the same educational level may have different true outcomes due to education quality or broad survey categories, leading to different expected performances for their children (Van de Weide et al., 2023). Certain authors (Asher et al. 2017, 2024) have tried to tackle this issue by developing a nonparametric method to bound mobility measures and propose a new measure of upward mobility, 8 considering the expected education rank of a child born to parents in the bottom half of the education distribution. III. Results Inequality of Opportunity Table 1 shows that approximately one-fifth of income inequality using MLD can be attributed to circumstances at birth, but about a half when considering the Gini coefficient. 12 For labor income, this rises, and ranges between 22.4 to close to 30 percent in MLD, and between 48.6 and 50.5 percent using the Gini coefficient. 13 Table 1. Results on absolute and relative IO, 2022 Panel A: MLD Outcome: Per capita Income Outcome: Labor Income Conditional Random Conditional Random Parametric Parametric Trees Forest Trees Forest IO 0.131 0.121 0.124 0.122 0.104 0.112 IOr 23.6 21.1 21.7 29.6 22.4 23.9 Obs 151,882 151,882 151,882 90,996 90,997 90,997 Panel B: Gini Outcome: Per capita Income Outcome: Labor Income Conditional Random Conditional Random Parametric Parametric Trees Forest Trees Forest IO NA 0.274 0.278 NA 0.255 0.265 IOr NA 48.5 49.2 NA 48.6 50.5 Obs NA 151,882 151,882 NA 90,997 90,997 Next, we aim at determining to what extent each circumstance explains the inequality of opportunity both in per capita income and labor income. Table 2 presents the Shapley value of circumstances under the parametric approach and the contribution of each circumstance estimated under the Random Forest 12 The specification reported in Table 1 includes variables such as ethnicity, parent’s education with imputations for unreported registers using one variable for both parents (specification 1 in table 4, see robustness checks section), as well as a pool of place at birth variables (born rural/urban area, born municipality size, presence of internal armed conflict in born municipality). 13 This is similar to the results we found for 2019 (presented in Table A1), although the Inequality of Opportunity index (IOr) is slightly higher for total household per capita income in 2022 compared to 2019. 9 methodology. 14 Our findings suggest the place of birth, all together, accounts for around a third of the total inequality of opportunity (circumstances) but could explain nearly 53.6 percent when using the Random Forest methodology. The latter suggests that under the parametric approach some crucial variables in the Colombian context such as the department of birth could be underestimated. 15 Table 2. Shapley value and Contribution circumstances for the main specification Shapley (under parametric Contribution (under random approach using MLD) % forest) % Per capita Labor Per capita Labor Income Income Income Income Parents Education 54.7 53.6 41.2 41.6 Ethnicity 6.5 9.2 5.2 6.4 Place of birth 38.8 35.7 53.6 50.1 Department of birth 2.7 2 22.5 23.6 Municipality category of birth 17.4 17.1 12.1 8.8 Rural/Urban Municipality 10.3 11 9.2 9.2 Presence of Internal Armed Conflict in 8.4 5.6 9.8 8.5 Municipality when Born Sex na 1.6 na 1.9 Intergenerational education mobility With around half of the inequality of opportunity determined by individuals’ socio-economic background (namely parents education), we then explore inter-generational mobility in education, bringing in the spatial lens to opportunities within Colombia. First, the International IGM database reveals that, as in LAC countries, Colombia faces high absolute levels of mobility. Seventy-eight percent of Colombian adults have more years of schooling that both of their parents. Nevertheless, relative mobility is among the lowest worldwide, well below that of OECD countries and of some countries in the region such as Brazil or Argentina. That is, there is high persistence in the ranking of the education distribution across generations. In fact, Colombia is one of the few countries in 14In annex table A4 we report the Shapley value for different specification considering the parametric approach. 15 One potential reason for these differences is that the parametric approach enforces a fixed functional form on the relationship between circumstances and income. For instance, the impact of parental education might differ by birth region, but the method assumes it is uniform for all, leading to a downward bias in IOp, and specifically in variables with higher variation and more categories such as the department of birth with 33 categories. In contrast and as explained in Brunori, Ferreira and Neidhöfer (2023), the Random Forest approach addresses both downward and upward biases efficiently by interacting all circumstance variables in the regression, assigning individuals to the average income within their specific type and under collections of hundreds of trees using a subsample of observations and circumstances at each node and tree. 10 the world (among those with available data) where high absolute mobility coexists with extremely low levels of relative mobility (Figure A1). One of the main contributions of this paper, as mentioned, is to explore mobility outcomes at the subnational level. We find that there is high heterogeneity of social mobility in education across Colombian departments. Figures 1 and 2 and Table 4 present mobility indices, absolute and relative, for Colombia’s departments and separately for sons and daughters. In terms of absolute mobility, Colombian departments display high mobility, showing the country’s overall progress in educational attainment (Figure 1). However, there is high dispersion in relative mobility, revealing wide spatial disparities in the opportunities to move up the socio-economic ladder (Figure 2). In general, peripheric departments in the Pacific, Caribbean and Orinoquía Region show the lowest rates of absolute mobility (Figure A2). Some of them also experience low levels of relative mobility (Figure A3). While Cundinamarca and Casanare have the highest levels of both absolute and relative mobility, La Guajira, Guainía and Vichada are among the departments with the lowest mobility in the country. Women are more likely to surpass the education of their parents, but in some departments such as Guainía or Amazonas the gender gap in relative mobility is considerably large (Table A3). Compared to OECD countries, with an average relative mobility in education of 0.67, certain departments display even lower levels than the already-low mobility exhibited by the Colombia national figure (Table 3). See Annex Table A2 for more granular comparisons with OECD countries. Figure 1. Absolute Mobility: Years of education Figure 2. Relative Mobility: 1-Beta Coefficient, (share of adults, YOS), cohort 1980 cohort 1980 11 Table 3. Intergenerational mobility, all children, and maximum years of education for both parents, cohort 1980 Parents Children Observations in education education Absolute Relative Department HS (mean) (mean) Mobility Mobility OECD 56665 12.3 14 0.57 0.67 San Andres 840 9.7 13 0.72 0.67 Bogotá 3918 9.2 13.5 0.77 0.61 Middle East & North Africa 60687 5.5 9.5 0.66 0.61 Atlántico 3913 8 11.7 0.73 0.6 East Asia & Pacific 38925 7.3 9.6 0.6 0.6 Cundinamarca 2809 5.8 10.9 0.81 0.58 Arauca 2077 4.9 9.1 0.76 0.57 Casanare 1744 4.8 9.9 0.82 0.57 Amazonas 1814 4.7 9.4 0.79 0.57 Chocó 2478 5.3 9.5 0.71 0.57 Europe & Central Asia 10746 11.2 12.2 0.48 0.57 Quindío 1859 7.1 11.7 0.77 0.57 Latin America & Caribbean 53459 6.3 9.5 0.68 0.56 Valle del Cauca 3539 7.5 11.6 0.75 0.56 Cesar 2899 5.5 10.4 0.78 0.54 Risaralda 2380 6.4 11 0.78 0.53 Guaviare 892 4.5 9.5 0.82 0.53 Meta 2637 6.3 11 0.79 0.52 Sucre 3158 4.7 9.9 0.81 0.52 Magdalena 3044 5.7 9.9 0.76 0.51 Bolivar 4101 6.1 10.1 0.75 0.51 Córdoba 3431 4.8 9.8 0.81 0.51 Antioquia 4380 6.7 11.1 0.77 0.5 Tolima 3031 6 10.7 0.79 0.5 Norte de Santander 3338 5.8 10.3 0.76 0.5 Caldas 2666 6.4 11.3 0.8 0.5 Colombia 89282 6.5 11 0.78 0.49 Santander 3464 6.6 11.4 0.8 0.49 Caquetá 2537 4.8 9 0.78 0.49 South Asia 43300 4.3 7.4 0.55 0.48 Putumayo 1762 4.7 9.4 0.82 0.46 Boyacá 3399 5.6 10.8 0.83 0.45 Vaupes 1534 3.2 8.1 0.87 0.45 Huila 2966 5.3 10.2 0.83 0.43 Guainía 1622 3.5 7.4 0.76 0.41 Cauca 3317 4.4 9 0.81 0.4 Vichada 1542 4.1 7.2 0.69 0.37 La Guajira 2818 4.7 8.8 0.71 0.36 Nariño 3373 4.8 9.2 0.81 0.35 12 In terms of trends, absolute mobility had been increasing until the 1980s cohort, with a higher pace in places that are considerably poor such as Amazonas, Oriental and even the Caribbean region (Figure 3). This is likely explained by a ceiling effect: as economies grow and the average level of education rises, it becomes increasingly challenging for individuals to surpass the achievements of their parents (Narayan et al., 2018). In terms of relative mobility, it has been increasing over time and the gap between regions has shrunk along cohorts (Figure 4). Nevertheless, there are still differences in opportunities between regions (Figures 3 and 4), with poorer areas experiencing lower mobility (Figures A4 to A7). Figure 3. Trends in Absolute Mobility by Figure 4. Trends in Relative Mobility by cohorts cohorts grouped by regions grouped by regions Lower educational mobility is associated with poverty and overall inequality. Departments with high poverty (monetary and multidimensional) and inequality levels also have low absolute mobility (Figure A8) and low relative mobility (Figure A9 and Figure A10). For instance, La Guajira, Vaupés, and Guainía have relatively lower levels of both absolute and relative mobility as well as high multidimensional poverty levels, similar to that of the south-west region (Cauca, Narino, Putumayo, and Caquetá). IV. Robustness Checks Several robustness checks are carried out and included in Tables 4 and A1 to A4 to assess the sensibility of the index to one of the main variables included, namely that of education of the parents. The main challenge with these variables stems from choices of including mother or father’s education separately or maximum education (one variable for both parents), and for filling missing values in the dataset (more so for mothers’ education). Retrospective variable (education of parents) tends to suffer from non-response. In particular, 13 given instances of missing data on parents’ education (mother and/or father), we test imputation techniques for missing parent education. We run multiple specifications for consistency checks under different assumptions treating the variable education of parents. First, we impute the education level for missing observations in the retrospective variable, given that the Household Survey separately asks for individual education within the household. We compute estimates by considering the maximum education level of both parents, rather than using two separate variables (one for the mother and one for the father). In cases in which some children do not report their parents' education, we identify the actual level self-reported by the individual by considering their relationship. 16 In the second specification type, we do not perform any imputation but consider the maximum educational level of both parents. In contrast, in the third and fourth specifications, we treat the parents' education variables separately for each parent. Specifically, we impute missing values in the third specification but do not perform imputation in the fourth specification. The IOr decreases in specifications that include imputed values and larger sample sizes. For example, the IOr in specifications 1 and 3, when considering imputed values, are approximately 1 percentage point lower than when no imputation is applied to parental education. This is particularly notable with the IOr standing at 45 percent when a single variable for both parents is used (specification 1), and 46.3 percent when the education levels of the mother and father are computed separately (specification 3). On the other hand, when the sample size is reduced (specifications 3 and 4, see Table 4 below) due to the separate consideration of mother’s and father’s education, the Inequality of Opportunities (IOr) increases slightly. Specifically, the IOr rises from 48.5 percent in specification 1 to 49.1 percent in specification 3, and from 48.8 percent in specification 2 to 49.5 percent in specification 4. 17 Second, we conduct multiple specifications for consistency checks by incorporating additional variables that more accurately capture the place of birth. We consider specifications both with and without the population size of the birth municipality, defined as follows: i) fewer than 50,000 inhabitants; ii) more than 50,000 but less than 100,000 inhabitants; iii) more than 100,000 but less than 500,000 inhabitants; and iv) more than 500,000 inhabitants. Excluding municipality variables, as well as not accounting for critical circumstances such as the presence of conflict in Colombia in the municipality of birth, could lead to an underestimation of the role of birth circumstances in explaining inequality of opportunities. For instance, in Specification 1, the IOr increases by 1.7 percentage points (pp) when including variables for the 16 For example, a grandmother to a mother living in a household where the household head is her child, but the mother has not reported information in the self-reported variable. 17 Annex tables A1 and A2 present results using the parametric approach. 14 municipality of birth, and by an additional 1.8 pp when considering the dummy variable for the presence of internal armed conflict in the birth municipality. Table 4. Results on absolute and relative IO for different specifications using Conditional Trees, (Gini reported for 2022) Per capita income Labor income Municipality Municipality Without With Without With + internal + internal municipality municipality municipality municipality Specification type/key variable variables variables armed variables variables armed conflict conflict IO 0.255 0.264 0.274 0.247 0.255 0.255 S1: Education imputed, one variable for both parents IOr 45.0% 46.7% 48.5% 47.2% 48.7% 48.6% N 151,883 90,997 IO 0.264 0.271 0.278 0.264 0.268 0.269 S2: Education no imputed, one variable for both parents IOr 46.2% 47.5% 48.8% 49.9% 50.7% 51.0% N 138,262 82,917 IO 0.265 0.274 0.281 0.253 0.270 0.270 S3: Education imputed, father and mother separated IOr 46.3% 47.8% 49.1% 47.7% 50.9% 51.0% N 124,676 75,098 IO 0.270 0.281 0.287 0.271 0.281 0.126 S4: Education no imputed, father and mother separated IOr 46.6% 48.5% 49.5% 50.3% 52.2% 52.3% N 111,812 66,796 Finally, given that the literature has identified that household surveys often overlook high-income households, (Alvaredo et al. 2013, Atkinson et al. 2011, Bourguignon, 2018, Burkhauser et al. 2012, Cowell and Flachaire, 2007, Jenkins 2017, Lustig, 2020; Cowell and Flachaire, 2007; Morgan 2018), and it is especially true for Colombia (Alvaredo and Londono, 2013; Diaz, 2015), we implement an additional exploratory analysis to measure the IO in Colombia accounting for top incomes. High-income earners represent a small population share and tend to underreport or even not respond about their income or consumption in household surveys (Alvaredo and Londono, 2013; Diaz, 2015; Anand and Segal, 2016). For example, numerous studies have shown that in low- and middle-income countries, per capita income or consumption measured by household surveys frequently lags behind welfare outcomes from National Accounts (Altimir, 1987; Fesseau and Mantonetti, 2013; Alvaredo et al, 2018). In that regard, several methodologies are developed to correct inequality indicators by taking into account the missing top income households (Bourguignon, 2018; Lustig, 2020; Cowell and Flachaire, 2015). As part of the robustness checks, we applied the methodology proposed by Blanchet et al. (2022), which uses 15 together replacing and reweighting approaches to adjust the top of the income distribution with an endogenous selection of the merging point. One of the advantages of the methodology relies on the potential scope of replicating socioeconomic characteristics within the household, which allows us to get data on education for the corrected population by assuming the socioeconomic characteristics of the replaced population within the survey. Table 5. Results on absolute and relative IO for different specifications adjusting for the missing rich, 2022. Per capita income Without municipality With municipality Municipality + internal Specification type/key variable variables variables armed conflict (1) (2) (3) IO 0.382 0.430 0.420 S1: Education imputed, one variable for both parents IOr 64.2% 72.2% 70.5% N 152,609 IO 0.417 0.459 0.466 S2: Education no imputed, one variable for both parents IOr 69.2% 76.3% 77.3% N 138,963 IO 0.416 0.431 0.434 S4: Education imputed, father and mother IOr 68.6% 71.0% 71.6% separate N 125,345 IO 0.420 0.446 0.448 S3: Education no imputed, father and mother separate IOr 68.6% 72.9% 73.2% N 112,389 We followed a similar approach and application in Baquero et al (2023) for Colombia using tax records, but this time adjusting the ECV for 2022 to better capture the entire income distribution, and then estimate IO indicators as in the table 5. Our estimates suggest an increase in the Gini coefficient from 0.565 without adjustment to 0.595 by correcting for the missing rich. The adjustment supposes a significant increase in the inequality of opportunities in absolute terms for specifications 2 and 3, as well as an increase in the relative inequality of opportunities index driven by the increment in the absolute IO, 18 given the adjustment at the top of the income distribution. IOr is estimated around 70.5 percent, which is close to the inherited inequality found in South Africa, ranging from 67.4 to 73.6 percent (see Brunori, Ferreira and Salas-Rojo, 2023). Nevertheless, the literature has identified the need to validate methodologies that adjust for the top earners by comparing multiple techniques (Lustig, 2020). This is an area that warrants further exploration. Additionally, future research in this domain can focus on two main areas: i) improving the identification of socioeconomic characteristics of top earners through imputation techniques, and ii) using linked data with 18 Results using the parametric approach are shown in Annex table A3, where despite a slight increase in the absolute IOp, the relative IO decreases. 16 administrative records to better capture the actual socioeconomic characteristics of households along the entire distribution. In this context, Colombia, as a data-rich country, possesses a set of administrative records, such as the Registro Social de Hogares, which can potentially be explored as studies by Leites et al (2022) in Uruguay. V. Conclusions This paper presents new estimates of inequality of opportunity in Colombia and subnational (department- level) estimates of intergenerational mobility in education. On the latter, absolute mobility is measured as the share of respondents with more years of schooling than both of their parent’s, and relative mobility as 1 minus the coefficient from regressing the respondent’s years of schooling on parent’s years of schooling. We find a large role for place of birth in shaping inequality of opportunity, and significant heterogeneity in inter-generational education mobility at the subnational level. The results show that around half of inequality of total and labor income is explained by individuals’ circumstances at birth and that, among those, 53.6 and 50.1 percent, respectively, are attributed to place of birth. We also find that educational mobility varies significantly at the subnational level, with people who are born in poorer places facing lower relative mobility. Bogotá, for one, has the highest relative mobility in education and closer to that of the OECD average, while poorer departments such as Nariño, La Guajira and Vichada have very low relative mobility. The estimates presented in this paper shed light on the vast spatial disparities that can take place in one country and that are masked by national estimates. They offer an opportunity to bring these inequalities to the forefront of the debate and can have implications for the country’s poverty and inequality reduction agendas, including on policies to ensure equity in access to opportunities and promote territorial development. 17 References Aaberge, R., Mogstad, M., & Peragine, V. (2011). Measuring Long-Term Inequality of Opportunity. Journal of Public Economics, 95(3-4), 193-204. Acevedo, K y Bornacelly Olivella, I. (2014). Panel municipal del CEDE. Universidad de los Andes, Facultad de Economía, CEDE. Disponible en: http://hdl.handle.net/1992/8510 Ayala-García, Jhorland. 2017. Capítulo 4. Movilidad social. Pág.: 103-137. En Estudios sociales del Pacífico colombiano. Banco de la República de Colombia. ISBN 9789586643740 Alesina, Alberto, Sebastian Hohmann, Elias Papaioannou, and Stelios Michalopoulos. (2021). “Intergenerational Mobility in Africa.” Econometrica 89 (1): 1–35. Alesina, A., Hohmann, S., Michalopoulos, S. and Papaioannou, E. (2021), Intergenerational Mobility in Africa. Econometrica, 89: 1-35. https://doi.org/10.3982/ECTA17018 Alvaredo, F.A., Londoño Vélez, J.: Altos ingresos e impuesto de renta en Colombia, 1993-2010, Revista de economía Institucional, Universidad Externado de Colombia, 16(31), pp. 157–194 (2014) Anand, S., & Segal, P. (2015), The global distribution of income. In A. B. Atkinson, & F. Bourguignon (Eds.). Handbook of income distribution (2A, pp. 937–979). Amsterdam: North- Holland. Arjona, Ana M., SarahMoore, and SilviaOtero-Bahamón. 2024. “Subnational Inequality after Armed Conflict in Colombia: Mapping a Research Agenda.” Background paper, World Bank, Washington, DC. Unpublished. Atamanov,Aziz; Cuevas,Pablo Facundo; Lebow,Jeremy Aaron; Mahler,Daniel Gerszon. (2024) New Evidence on Inequality of Opportunity in Sub-Saharan Africa : More Unequal Than We Thought (English). Policy Research working paper ; no. WPS 10723; PEOPLE Washington, D.C. : World Bank Group. http://documents.worldbank.org/curated/en/099558203182421649/IDU1a3c568111b02514f9d19 e221936be7486403 Atkinson, A.B.: Measuring Top Incomes: Methodological Issues. In: Atkinson, A., Piketty, T. (eds.) Top Incomes over the Twentieth Century: A Contrast between Continental European and English-Speaking Countries. Oxford University Press, Oxford (2007) Arti Grover and William F. Maloney (2022). Proximity without Productivity: Measuring Agglomeration Effects with Plant-level Output and Price Data. 18 Baquero, Juan Pablo, Davalos, Maria Eugenia and Monroy Barragan, Juan Manuel, (2023), Revisiting the Distributive Impacts of Fiscal Policy in Colombia, No 10520, Policy Research Working Paper Series, The World Bank, https://EconPapers.repec.org/RePEc:wbk:wbrwps:10520 Blanchet, T., Flores, I. y Morgan, M. (2022). The weight of the rich: improving surveys using tax data. The Journal of Economic Inequality, 20, 119-150. https://doi.org/10.1007/s10888-021- 09509-3 Bourguignon, F. (2018), "Inequality of opportunity", in Stiglitz, J., J. Fitoussi and M. Durand (eds.), For Good Measure: Advancing Research on Well-being Metrics Beyond GDP, OECD Publishing, Paris, https://doi.org/10.1787/9789264307278-7-en. Bourguignon, F. (2018) Simple adjustments of observed distributions for missing income and missing people. J Econ Inequal 16, 171–188 https://doi.org/10.1007/s10888-018-9388-8 Brunori, P., Hufe, P., and Mahler, D. (2023). The roots of inequality: Estimating inequality of opportunity from regression trees and forests*. The Scandinavian Journal of Economics, 125(4), 900-932. https://doi.org/10.1111/sjoe.12530 Brunori, P., Ferreira, F., and Neidhöfer, G. (2023). Inequality of opportunity and intergenerational persistence in Latin America, WIDER Working Paper Series wp-2023-39, World Institute for Development Economic Research (UNU-WIDER). Brunori, P., Ferreira, F. H. G., & Peragine, V. (2023). Inherited inequality: A general framework and an application to South Africa (III Working Paper No. 107). International Inequalities Institute, London School of Economics and Political Science. http://eprints.lse.ac.uk/120308/1/III_working_paper_107.pdf Brunori, P., Palmisano, F., and Peragine, V. (2019). Inequality of Opportunity in Sub-Saharan Africa. Applied Economics, 51(60), 6428-6458. Brunori, P., Ferreira, F., and Peragine, V. (2013). Inequality of opportunity, income inequality and economic mobility : some international comparisons. Washington, DC: World Bank. https://documentsinternal.worldbank.org/search/17150436 Buscha, F., Gorman, E., & Sturgis, P. (2021). Spatial and social mobility in England and Wales: A sub‐national analysis of differences and trends over time. The British Journal of Sociology, 72(5), 1378–1393. https://doi.org/10.1111/1468-4446.12885 Burkhauser, R. V., S. Feng, S. P. Jenkins and J. Larrimore (2012), “Recent trends in top income shares in the USA: reconciling estimates from March CPS and IRS tax return data,” Review of Economics and Statistics 94, pp. 371–88. CAF (2022) Inherited inequalities: The role of skills, employment, and wealth in the opportunities of new generations. Buenos Aires, Argentina. ISBN: 978-980-422-277-1. Carranza, R. (2020). Inequality of outcomes, inequality of opportunity, and economic growth (No. 534). ECINEQ, Society for the Study of Economic Inequality. 19 Carranza, R. (2023) Upper and lower bound estimates of inequality of opportunity: A cross- national comparison for Europe. Review of Income and Wealth 69(4). https://doi.org/10.1111/roiw.12622 Chetty, R., Hendren, N., Kline, P., Saez, E., Turner, N. (2014). Is the United States still a land of Opportunity? Recent trends in intergenerational mobility. Am. Econ. Rev.Pap. Proceed. 104, 141– 147. Chetty, R., and Hendren, N. (2018). The impacts of neighborhoods on intergenerational mobility I: Childhood exposure effects. The Quarterly Journal of Economics, 133(3), 1107–1162. https://doi.org/ 10.1093/qje/qjy007 Clavijo, I., Mejía-Mantilla, C., Olivieri S., Lara, G., Romero, J. (2021). Mind the Gap : How COVID-19 is Increasing Inequality in Latin America and the Caribbean. Washington, D.C. : World Bank Group. https://documentsinternal.worldbank.org/search/33277635 Connolly, Marie, Corak, Miles, Haeck, Catherine, 2019. Intergenerational Mobility between and within Canada and the United States. The Journal of Labor Economics, S243-S778. Corak, M., 2013. Income inequality, equality of opportunity, and intergenerational mobility. J. Econ. Perspect. 27, 79–102. Emran, M. Shahe, Greene, William, Shilpi, Forhad, April Corak, Miles, 2021. The Canadian Geography of Intergenerational Mobility. The Economic Journal, 130 (631), 2134-2174 Cowell, F., Flachaire, E. (2015) Statistical Methods for Distribution Analysis. In: Atkinson, A.B., Bourguignon, F. (eds.) Handbook of Income Distribution, vol. 2. Elsevier, North-Holland, Amsterdam Christoph Lakner, Silvia Redaelli, Daniel Gerszon Mahler, Rakesh Gupta N. Ramasubbaiah, and Stefan Thewissen. 2018. Fair Progress? Economic Mobility across Generations around the World. Washington, DC: World Bank DiPrete, Thomas A. 2020. ‘The Impact of Inequality on Intergenerational Mobility’. Annual Review of Sociology 46(1):379–98. doi: 10.1146/annurev-soc-121919-054814. Durlauf, Steven N., Andros Kourtellos, and Chih Ming Tan. 2022. ‘The Great Gatsby Curve’. Annual Review of Economics 14(1):571–605. doi: 10.1146/annurev-economics-082321-122703. Emran, M. S., Greene, W., & Shilpi, F. (2018). When Measure Matters: Coresidency, Truncation Bias, and Intergenerational Mobility in Developing Countries. Journal of Human Resources, 53 (3), 579{607. Galvis-Aponte, L. A., & Meisel-Roca, A. (2014). Aspectos regionales de la movilidad social y la igualdad de oportunidades en Colombia. Revista De Economía Del Rosario, 17(02), 257–297. https://doi.org/10.12804/rev.econ.rosario.17.02.2014.03 20 Garparini L., and Cruces G. (2021) The changing picture of inequality in Latin America. United Nations Development Programme. Retrieved from https://www.undp.org/latin- america/publications/changing-picture-inequality-latin-america Fernández, Manuel. (2012). Violencia y derechos de propiedad: El caso de La Violencia en Colombia. Ensayos sobre Política Económica, 30(69), 111-147. Ferreira, F. H., & Gignoux, J. (2011). The Measurement of Inequality of Opportunity: Theory and an Application to Latin America. Review of Income and Wealth, 57(4), 622-657. https://doi.org/10.1111/j.1475-4991.2011.00467.x Ferreira, F. H. G. and V. Peragine. (2016) Individual Responsibility and Equality of Opportunity, in M. D. Adler and M. Fleurbaey (eds), The Oxford Handbook of Well-Being and Public Policy, Oxford University Press, New York, 746–84 Ferreira, F., & Meléndez, M. (2012). Desigualdad de resultados y oportunidades en Colombia 1997-2010. Documentos CEDE # 40., Universidad de los Andes Hong, Q., Gruijter, R. (2024). A lost land of opportunity? The geography of intergenerational educational mobility in China. Population, Space and Place, 30, e2784. https://doi.org/10.1002/psp.2784 Jantti, M., Bratsberg, B., Roed, K., Raaum, O., Naylor, R., Osterbacka, E., Bjorklund, A., Eriksson, T. (2006). American exceptionalism in a new light: a comparison of intergenerational earnings mobility in the Nordic Countries, the United Kingdom and the United States. Mimeo Jenkins, S. 2017. "Pareto Models, Top Incomes and Recent Trends in UK Income Inequality," Economica, London School of Economics and Political Science, vol. 84(334), pages 261-289, April. Krueger, A.B.(2012). The Rise and Consequences of Inequality in the United States. Washington D.C, Speech at the Center for American Progress. Latinobarometro (2023). Latinobarómetro 1995-2023. 25 años monitoreando las sociedades latinoamericanas. Presentation in IDB, November 1st, 2023, by Marta Lagos. https://dds.cepal.org/redesoc/video?id=2217 LeGrand, C. (1988). Colonización y protesta campesina en Colombia (1850-1950). Bogotá: Centro editorial Universidad Nacional de Colombia. Leites, M., Ramos, X., Rodríguez, C., and Joan, V. (2022). Intergenerational mobility along the income distribution: estimates using administrative data for a developing country, Documentos de Trabajo (working papers) 22-05, Instituto de Economia - IECON. Lustig, N. (2020) The missing rich in household surveys: causes and correction approaches, Working Paper 75, CEQ Institute, Tulane University Maloney, William F., and Felipe Valencia-Caicedo. (2016). “The Persistence of (Subnational) Fortune.” The Economic Journal 126 (598): 2363–401. Marrero, G. A., & Rodríguez, J. G. (2013). Inequality of Opportunity and Growth. Journal of Development Economics, 104, 107-122. 21 Morgan, M. (2018), “Essays on Income Distribution. Methodological, Historical and Institutional Perspectives,” Ph.D. dissertation, Ecole Doctorale n°465, Ecole des Hautes Études en Sciences Sociales, Paris, France Muñoz S., Ercio; Siravegna, Mariel (2023) : When measure matters: Coresidence bias and intergenerational mobility revisited, IDB Working Paper Series, No. IDB-WP-01469, InterAmerican Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0004881 Munoz, E. (2021). Intergenerational Educational Mobility within Chile. Available at SSRN: https://ssrn.com/abstract=3837299 or http://dx.doi.org/10.2139/ssrn.3837299 Narayan, A.,; Van der Weide, R., Cojocaru, A., Lakner, C.,; Redaelli, S., Mahler, D.,; Ramasubbaiah, R., Thewissen, S.. (2018). Fair Progress?: Economic Mobility Across Generations Around the World. Equity and Development. © Washington, DC: World Bank. http://hdl.handle.net/10986/28428 Narayan, A., Saavedra, J., ;Sailesh, T. (2013). Shared prosperity : links to growth, inequality and inequality of opportunity. Washington, DC: World Bank. Neidhöfer, G., Serrano, J., & Gasparini, L. (2018). Educational inequality and intergenerational mobility in Latin America: A new database. Journal of Development Economics, 134, 329-349. https://doi.org/10.1016/j.jdeveco.2018.05.016 Ianchovichina, Elena. 2024. The Evolving Geography of Productivity and Employment: Ideas for Inclusive Growth through a Territorial Lens in Latin America and the Caribbean. World Bank Latin American and Caribbean Studies. Washington, DC: World Bank Osorio, Javier, Mohamed Mohamed, Viveca Pavon, and Susan Brewer-Osorio. 2019. "Mapping Violent Presence of Armed Actors in Colombia," Advances of Cartography and GIScience of the International Cartographic Association, 1(16):1-9. Palmisano, F., & Peragine, V. (2022). Inequality of Opportunity: Theoretical Considerations and Recent Empirical Evidence. In Advances in Economic Measurement: A Volume in Honour of DS Prasada Rao (pp. 349-386). Singapore: Springer Nature Singapore. Ramos, X. and Van de gaer, D. (2016), Approaches To Inequality Of Opportunity: Principles, Measures And Evidence. Journal of Economic Surveys, 30: 855-883. https://doi.org/10.1111/joes.12121 Roemer, J. E. (1993). A Pragmatic Theory of Responsibility for the Egalitarian Planner. Philosophy & Public Affairs, 146-166. Roemer, J.E. (2012) On several approaches to equality of opportunity. Economics and Philosophy 28: 165–200. Roemer, J.E. and Trannoy, A. (2013) Equality of Opportunity, forthcoming in A.B. Atkinson and F. Bourguignon (eds.), Handbook of Income Distribution (pp. 217–300), North-Holland. Van der Weide, R., Lakner, C., Mahler, D. G., Narayan, A., & Gupta, R. (2023). Intergenerational mobility around the world: A new database. Journal of Development Economics, 166, 103167. https://doi.org/10.1016/j.jdeveco.2023.103167 22 World Bank. (2005). World Development Report 2006: Equity and Development. © Washington, DC. http://hdl.handle.net/10986/5988 World Bank. (2018). Fair progress?: Economic mobility across generations around the world. Washington, DC: World Bank. Retrieved from https://www.worldbank.org/en/topic/poverty/publication/fair-progress-economic-mobility- across-generations-around-the-world. World Bank. (2024). Colombia - Poverty and Equity Assessment: Trajectories - Prosperity and Poverty Reduction in the Colombian Territory. © Washington, DC: World Bank. http://hdl.handle.net/10986/42557 License: CC BY-NC 3.0 IGO 23 Annex Table A1. Results on absolute and relative IO for different specifications using a parametric approach (MLD reported), 2019 Capita income Labor income Municipality Municipality Without With + internal Without With + internal Specification type/key municipality municipality armed municipality municipality armed variable conflict conflict S1: Education imputed, one IO 0.103 0.118 0.133 0.101 0.121 0.136 variable for both parents IOR 16.5% 19.0% 21.4% 21.9% 26.2% 29.5% 167,033 N 101,361 S2: Education no imputed, IO 0.108 0.124 0.137 0.116 0.137 0.153 one variable for both parents IOR 17.1% 19.6% 21.7% 24.2% 28.5% 31.9% 149,986 N 91,167 S4: Education imputed, father IO 0.115 0.131 0.151 0.122 0.141 0.160 and mother separately IOR 18.0% 20.4% 23.4% 24.9% 28.8% 32.6% 130,753 N 79,880 S3: Education no imputed, IO 0.124 0.142 0.156 0.140 0.162 0.179 father and mother separately IOR 18.8% 21.6% 23.6% 27.7% 32.2% 35.4% 117,801 N 71,484 Table A2. Results on absolute and relative IO for different specifications using a parametric approach (MLD reported), 2022 Per capita income Labor income Municipality Municipality Without With Without With + internal + internal municipality municipality municipality municipality armed armed variables variables variables variables conflict conflict (1) (2) (4) (5) Specification type/key variable (3) (6) S1: Education imputed, one IO 0.095 0.111 0.131 0.091 0.107 0.122 variable for both parents IOr 17.1% 20.0% 23.6% 22.3% 26.1% 29.6% N 151,882 90,996 S2: Education no imputed, IO 0.106 0.122 0.135 0.107 0.123 0.133 one variable for both parents IOr 18.6% 21.5% 23.8% 25.2% 29.1% 31.3% N 138,262 82,916 S3: Education imputed, father IO 0.113 0.128 0.144 0.112 0.127 0.139 and mother separated IOr 19.7% 22.2% 25.1% 26.0% 29.6% 32.3% N 124,675 75,097 S4: Education no imputed, IO 0.123 0.140 0.157 0.131 0.162 0.157 father and mother separated IOr 20.9% 23.7% 26.6% 29.2% 33.3% 35.0% N 111,812 66,795 24 Table A3. Results on absolute and relative IO for different specifications adjusting for the missing rich using a parametric approach (MLD reported), 2022 Per capita income Without With Municipality + municipality municipality internal armed variables variables conflict Specification type/key variable (1) (2) (3) S1: Education imputed, one variable for IO 0.081 0.108 0.128 both parents IOr 13.0% 17.2% 20.4% N 152,608 S2: Education no imputed, one variable for IO 0.092 0.120 0.138 both parents IOr 14.4% 18.7% 21.5% N 138,963 S4: Education imputed, father and mother IO 0.101 0.128 0.149 separate IOr 15.5% 19.6% 22.7% N 125,345 S3: Education no imputed, father and IO 0.111 0.141 0.155 mother separate IOr 16.6% 21.0% 23.1% N 112,389 Table A4. Shapley value of circumstances: Including variables regarding the place of birth at municipality level Per capita income Labor Income Municipalit Municipality + Without With y + internal Without With internal armed municipality municipality armed municipality municipality conflict (1) (2) conflict (4) (5) (6) (3) Parents Education 83.6 57.2 54.7 81.5 55.8 53.6 Ethnicity 12.6 10.3 6.5 13.1 10.5 9.2 Born Department 3.8 1.8 2.7 3.9 1.9 2.0 Born Municipality Group na 19.3 17.4 na 17.7 17.1 Rural/Urban Municipality na 11.3 10.3 na 12.2 11.0 Presence of Internal Armed Conflict in Municipality of na na 8.4 na na 5.6 birth Sex na na na 1.5 2.00 1.6 25 Table A5. Results on absolute and relative intergenerational mobility, OECD and Colombian departments Parents education Children Absolute Country/Department Observations in HS Relative Mobility (mean) education (mean) Mobility United Kingdom 761 12.79 15.15 0.65 0.82 Korea, Rep. 2042 11.69 14.98 0.77 0.81 New Zealand 116 14.33 14.97 0.52 0.81 Israel 1486 12.47 13.65 0.47 0.80 Denmark 519 13.83 14.39 0.56 0.80 Australia 2843 13.38 13.62 0.38 0.77 Iceland 248 14.85 16.40 0.60 0.76 Finland 1025 13.86 15.24 0.58 0.76 Netherlands 727 12.65 15.51 0.71 0.75 Canada 3305 14.41 14.71 0.41 0.74 France 804 11.84 14.88 0.73 0.73 Germany 1346 14.71 15.06 0.46 0.73 Japan 380 13.76 14.31 0.37 0.71 Lithuania 796 13.82 14.56 0.47 0.70 Belgium 870 12.39 14.75 0.66 0.69 Greece 218 10.64 13.37 0.66 0.68 Spain 1028 9.03 15.49 0.87 0.68 Norway 809 14.25 14.87 0.51 0.67 United States 3660 14.12 14.41 0.36 0.67 San Andres 840 9.65 12.98 0.72 0.67 Italy 332 11.80 13.66 0.50 0.67 OECD 145947 12.15 13.94 0.58 0.67 Switzerland 774 13.66 12.34 0.30 0.66 Poland 1143 11.46 14.43 0.75 0.65 Ireland 1459 11.05 15.82 0.92 0.65 Sweden 769 14.05 14.55 0.47 0.64 Costa Rica 1681 6.82 9.00 0.57 0.64 Slovenia 689 12.53 14.18 0.61 0.64 Chile 12708 10.47 13.01 0.65 0.62 Bogotá 3918 9.19 13.49 0.77 0.61 Estonia 1062 14.58 14.47 0.38 0.61 Latvia 284 13.80 13.96 0.46 0.61 Slovak Republic 461 12.79 13.89 0.56 0.61 Czechia 1291 13.53 13.97 0.47 0.61 Atlántico 3913 8.00 11.71 0.73 0.60 Mexico 7928 7.85 10.98 0.68 0.60 Cundinamarca 2809 5.84 10.85 0.81 0.58 Arauca 2077 4.88 9.08 0.76 0.57 Casanare 1744 4.78 9.92 0.82 0.57 Amazonas 1814 4.74 9.38 0.79 0.57 Chocó 2478 5.29 9.52 0.71 0.57 Quindío 1859 7.14 11.69 0.77 0.57 Valle del Cauca 3539 7.46 11.57 0.75 0.56 Cesar 2899 5.49 10.37 0.78 0.54 Risaralda 2380 6.35 11.01 0.78 0.53 Guaviare 892 4.49 9.46 0.82 0.53 Meta 2637 6.27 11.02 0.79 0.52 Sucre 3158 4.74 9.89 0.81 0.52 Austria 955 12.94 14.37 0.57 0.52 26 Portugal 726 8.08 12.39 0.81 0.52 Magdalena 3044 5.66 9.91 0.76 0.51 Bolivar 4101 6.10 10.10 0.75 0.51 Córdoba 3431 4.81 9.84 0.81 0.51 Antioquia 4380 6.68 11.10 0.77 0.50 Tolima 3031 6.04 10.67 0.79 0.50 Norte de Santander 3338 5.85 10.25 0.76 0.50 Caldas 2666 6.37 11.31 0.80 0.50 Colombia 89282 6.53 11.01 0.78 0.49 Santander 3464 6.58 11.39 0.80 0.49 Caquetá 2537 4.79 9.00 0.78 0.49 Putumayo 1762 4.66 9.35 0.82 0.46 Boyacá 3399 5.62 10.80 0.83 0.45 Vaupes 1534 3.18 8.13 0.87 0.45 Huila 2966 5.28 10.22 0.83 0.43 Türkiye 735 6.44 10.31 0.70 0.42 Guainía 1622 3.48 7.40 0.76 0.41 Cauca 3317 4.42 9.04 0.81 0.40 Hungary 685 12.22 13.15 0.46 0.40 Vichada 1542 4.06 7.20 0.69 0.37 La Guajira 2818 4.74 8.76 0.71 0.36 Nariño 3373 4.78 9.24 0.81 0.35 27 Table A6. Results on absolute and relative intergenerational mobility by gender in Colombian departments Absolute mobility (years of Relative mobility (1-beta) education) Department Girls Boys Girls Boys Bogotá 0.79 0.74 0.64 0.59 Arauca 0.77 0.75 0.63 0.52 Cundinamarca 0.88 0.73 0.62 0.52 Atlántico 0.76 0.7 0.61 0.61 Casanare 0.84 0.8 0.6 0.54 San Andres 0.82 0.62 0.6 0.7 Valle del Cauca 0.78 0.71 0.59 0.51 Quindío 0.81 0.73 0.57 0.55 Cesar 0.82 0.74 0.56 0.52 Chocó 0.73 0.69 0.56 0.57 Risaralda 0.81 0.75 0.56 0.49 Sucre 0.83 0.79 0.56 0.49 Bolivar 0.78 0.72 0.53 0.49 Caldas 0.82 0.78 0.53 0.46 Magdalena 0.79 0.72 0.53 0.49 Norte de Santander 0.79 0.73 0.53 0.46 Tolima 0.82 0.74 0.53 0.47 Santander 0.83 0.78 0.52 0.46 Córdoba 0.82 0.8 0.51 0.51 Meta 0.83 0.75 0.51 0.54 Antioquia 0.82 0.71 0.5 0.5 Putumayo 0.82 0.81 0.5 0.41 Caquetá 0.82 0.73 0.47 0.54 Guaviare 0.8 0.85 0.47 0.59 Amazonas 0.75 0.83 0.46 0.68 Boyacá 0.85 0.8 0.45 0.45 Cauca 0.82 0.81 0.43 0.37 Huila 0.86 0.79 0.43 0.42 Vichada 0.7 0.69 0.39 0.35 Nariño 0.82 0.8 0.37 0.32 La Guajira 0.72 0.71 0.36 0.37 Vaupes 0.87 0.87 0.32 0.58 Guainía 0.71 0.81 0.27 0.56 28 Figure A1. Regression Tree for Colombia 29 Figure A2. Absolute and Relative mobility around the world, cohort 1980. 30 Figure A3. Absolute Mobility: Years of Figure A3. Relative Mobility: 1-Beta education (share of adults, YOS), cohort 1980 Coefficient, cohort 1980 31 Figure A4. Absolute Mobility by cohorts Figure A5. Relative Mobility by cohorts grouped grouped by department’s poverty levels. by department’s poverty levels. Figure A5. Absolute Mobility by cohorts Figure A7. Relative Mobility by cohorts grouped grouped by gender and poverty status by gender and poverty status 0.70 0.80 0.60 0.70 Boys-Poor 0.50 Boys-Poor Boys-nonpoor Boys-nonpoor Girls-Poor Girls-Poor Girls-nonpoor 0.40 Girls-nonpoor 0.60 0.30 0.50 0.20 1940 1950 1960 1970 1980 1990 Cohort (which decade individuals are born in) 1940 1950 1960 1970 1980 1990 Cohort (which decade individuals are born in) 32 Figure A6. Relationship between absolute mobility and welfare indicators Panel A: Absolute Mobility vs Multidimensional Poverty Panel B: Absolute Mobility vs Income Inequality (Gini coefficient) 33 Panel C: Absolute Mobility vs Monetary Poverty (%) Figure A7. Relationship between relative mobility and welfare indicators Panel A: Relative Mobility vs Multidimensional Poverty 34 Panel B: Relative Mobility vs Income Inequality Panel C: Relative Mobility vs Monetary Poverty 35 Figure A8. Multidimensional vs Relative Mobility, cohort 1980 Figure A9. Replicating overall absolute and relative indicators for ECV 2013, Panel A: Absolute Mobility Panel B: Relative Mobility Absolute: Continuous Relative: BETA 1 .8 .8 .6 .6 .4 .4 .2 .2 0 dad max mom dad max mom dad max mom dad max mom dad max mom 0 dad max mom dad max mom dad max mom dad max mom dad max mom 1940 1950 1960 1970 1980 1940 1950 1960 1970 1980 International DB Replicated International DB Replicated 36