Policy Research Working Paper 11166 Intergenerational Income Mobility around the World A New Database Ercio Munoz Roy van der Weide Development Economics Development Research Group July 2025 Policy Research Working Paper 11166 Abstract This paper introduces a new global database with estimates association between income mobility and inequality (known of intergenerational income mobility for 87 countries, cov- as the Great Gatsby Curve) continues to hold across this ering 84 percent of the world’s population. This marks a wider range of countries. The database also reveals a positive notable expansion of the cross-country evidence base on association between income mobility and national income income mobility, particularly among low- and middle-in- per capita, suggesting that countries achieve higher levels of come countries. The estimates indicate that the negative intergenerational mobility as they grow richer. This paper is a product of the Development Research Group, Development Economics. 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 atrvanderweide@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 Intergenerational Income Mobility around the World: A New Database∗ noz† Ercio Mu˜ Roy van der Weide‡ JEL Codes: D31, D63, J62, O15. Keywords: Intergenerational mobility, Inequality, Income, Poverty. ∗ The authors would like to thank Garance Genicot, Stephen Jenkins, Dean Jolliffe, Gianluca Violante, as well as the participants at the 2025 MBS Sustainable Development Conference in Prato for providing comments and suggestions. The authors are also grateful to Garance Genicot and Debraj Ray for providing replication materials for their measure of upward mobility. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. † Social Sector, Inter-American Development Bank (erciom@iadb.org). ‡ Corresponding author: Development Economics Research Group, World Bank (rvander- weide@worldbank.org). 1 Introduction Intergenerational income mobility measures the extent to which a child’s income depends on his or her parents’ income. Higher mobility is often associated with greater “equality of opportunity” (see, e.g., Corak, 2013; Chetty et al., 2014; Cholli and Durlauf, 2022). The primary contribution of this study is the compilation of a global database with estimates of income mobility for 87 countries, collectively covering 84 percent of the world’s popula- tion. This is achieved by processing individual data from at least 156 surveys and employing the Two-Sample Two-Stage Least Squares (TSTSLS) approach proposed by Bjorklund and Jantti (1997). TSTSLS makes the estimation of the intergenerational income elasticity possi- ble when long-running panel surveys are not available. It uses parental education and, where available, occupation as predictors for parental income. As retrospective data on parental ed- ucation and occupation are more widely available than long-running panels are, this greatly expands the country coverage that can be achieved (without it, our global database of income mobility would include only a handful of countries). We next describe key features of the cross-country variation in intergenerational income mobility and verify whether empirical regularities that have been established using data for high-income and emerging economies carry over to developing countries (and the world at large). Theory predicts that the intergenerational income elasticity (IGE) varies with structural parameters, including the efficacy of private and public investments in generating human capital, returns to education, and the progressivity of public investments (see, e.g., Becker and Tomes, 1979, 1986; Becker et al., 2018; Solon, 2004, 2014). In these models, parents with more human capital and higher income have a higher capacity and more resources to invest in the education of their children; moreover, they have greater incentives to do so (see the excellent discussion in Corak, 2013). In turn, this predicts a negative relationship between intergenerational income mobility and income inequality. Naturally, this relationship is endogenously determined (see, e.g., Corak, 2013; Durlauf et al., 2022); the factors described above simultaneously drive both 2 income mobility and income inequality in the direction that produces a negative correla- tion between the two.1 All else being equal, higher returns to schooling increase the inter- generational persistence in income (lower income mobility) and increase income inequality. Similarly, more progressivity of public investments reduces both intergenerational income persistence and income inequality. Theory also offers predictions regarding the relationship between intergenerational income mobility and national income. Naturally, this relationship is also endogenously determined, and the sign of this relationship may vary. Under certain conditions, public redistributive policies may be growth enhancing (see, e.g., Benabou, 2000), while growth and higher na- tional income create more fiscal space for redistributive public policies (which in turn can increase mobility). Examples of such public policies include progressive taxes and trans- fers, public education, and policies that encourage residential integration. In the model put forward by Benabou (2000), redistributive policies achieve increased growth through a more efficient allocation of investment resources (particularly for education) to more credit- constrained households. However, redistribution also introduces tax distortions to effort, such that the net effect of redistribution can go either way (and, by extension, the correla- tion between income mobility and national income can, in theory, go either way). There is extensive empirical evidence confirming the negative relationship between mo- bility and inequality, which has become known as the Great Gatsby Curve (GGC).2 In one of the earliest and most cited studies, it was established using data for 13 high-income countries and emerging economies (see Corak, 2013).3 The GGC has arguably become the most well- known stylized fact in the empirical literature on intergenerational mobility (for a synthesis of theoretical and empirical work on it, see Durlauf et al., 2022). There are fewer empirical investigations of the relationship between intergenerational income mobility and national 1 See Durlauf and Seshadri (2017) for a discussion about the intertemporal correlation between mobility and inequality derived from a theoretical model that has socioeconomic segregation as the main mechanism. 2 This relationship was given this name after a speech by Alan Krueger (Krueger, 2012). 3 Other early studies recognizing this empirical relationship include Hassler (2007) and Andrews and Leight (2009). 3 income, and they mostly exploit within-country variability (e.g., Guell et al., 2018).4 Empirical studies of the largest cross-sectional number of income mobility observations are arguably within-country studies that estimate mobility at the district level.5 Promi- nent examples are empirical applications to the United States and Canada by Chetty et al. (2014), Corak (2020), and Connolly et al. (2019).6 Districts with comparatively high rates of mobility are found to have lower inequality, better public-school systems and are less residentially segregated.7 While the number of empirical studies conducting cross-country comparisons of intergen- erational income mobility is large, the number of countries in these studies tends to be small.8 For reviews of this body of work, see e.g., Solon (2002), Corak (2013), Blanden (2013), Jantti and Jenkins (2015), Mazumder (2015), and Mogstad and Torsvik (2023).9 Since the seminal study by Bjorklund and Jantti (1997), which compares estimates of income mobility be- tween Sweden and the United States, the empirical question of how the United States (and the United Kingdom) compares to the Nordic countries in terms of intergenertional income mobility has received much attention (e.g., Bratsberg et al., 2007; Jantti et al., 2006; Corak et al., 2014; Landerso and Heckman, 2017; Helsø, 2021).10 The majority of the empirical literature, however, provides estimates of income mobility for a single country and borrows estimates for other countries from the existing literature to put the country in question in context (e.g., Rohenkohl, 2023; Kenedi and Sirugue, 2023) or uses earlier findings for the same country to reconcile divergences (e.g., Mazumder, 2005; Aaronson and Mazumder, 4 Other studies examine educational mobility and national income or growth, e.g., Neidh¨ ofer (2019), Neidh¨ ofer et al. (2024) and van der Weide et al. (2024). Similarly, Marrero and Rodriguez (2013) and Ferreira et al. (2018) study the relationship between inequality of opportunity and growth. 5 Recent work on intergenerational educational mobility also includes district level estimates (e.g., Alesina et al., 2021; Card et al., 2022; Dodin et al., 2024 Asher et al., 2024; Munoz, 2024). 6 Similar research for countries outside North America include Acciari et al. (2022), Britto et al. (2022), Cortes et al. (2022), and Deutscher (2020), among others. 7 Recent work has emphasized the role of social capital in explaining the relationship between upward mobility and other correlates (Chetty et al., 2022). 8 Some attempts to conduct cross-country analysis at a larger scale include OECD (2018) and Narayan et al. (2018). 9 Torche (2014) provides a review including the literature in sociology with a focus on Latin America. 10 Other examples of two-country comparisons include Blanden et al. (2014) and Stockhausen (2021). 4 2008). To the best of our knowledge, the largest cross-country income mobility database to date was recently compiled by Ray and Genicot (RG; 2024), using an alternative measure of mobility. Their indicator of upward mobility does not require panel data (or retrospective data on parental education and occupation) for estimation, which expands the number of countries for which this measure of mobility can be obtained. Their database contains estimates for 122 countries. When the data are used to reproduce the GGC, the negative relationship between income mobility and inequality does not appear to hold across the large cross-section of countries (and for their measure of mobility). The authors attribute this result to the inclusion of countries from the developing world (which were excluded in earlier empirical studies that established the GGC). This suggests that stylized facts established for the developed world might not carry over to the developing world (or the world at large). “It is entirely possible that the Gatsby curve as measured by IGE will look different in the expanded set of countries. (There are no data to examine this possibility.),” (Ray and Genicot, 2023). Now that we have compiled these data, we can revisit these stylized facts using the measure of income mobility with which they were first established. We find the following. First, income mobility is seen to vary significantly between countries; the IGE can be as low as 0.14 (for Sweden) and as high as 0.96 (for Madagascar). On average, income mobility is found to be lower in the developing world. Second, the GGC holds across the 87 countries in our database (including 44 developing countries) when evaluated using the IGE. Third, we observe a positive relationship between intergenerational income mobility and national income; the relationship is somewhat weaker for the developing world. Fourth, consistent with predictions from theory, we observe a positive correlation between income mobility and the progressivity of public spending and a negative correlation between income mobility and the returns to education. It should be noted that these results do not establish causal mechanisms. However, our global database on intergenerational income mobility can assist 5 with further investigations into such mechanisms. RG’s upward mobility shows a negative correlation with national income. In summary, upward mobility and mobility measured through the IGE have contrasting relationships with both inequality and national income. The correlations observed for RG’s upward mobility measure sit well with the interpretation of upward mobility as a measure of growth (at lower percentiles of the income distribution). There is a large empirical literature on the relationship between inequality and growth, and it finds that the results on this relation- ship are inconclusive (consistent with the flat relationship between upward mobility and inequality) (see, e.g., Banerjee and Duflo, 2003). Similarly, the convergence effect implies a negative relationship between growth and national income (as low-income countries have greater scope for catch-up growth), which is consistent with the negative correlation between upward mobility and national income. The discussion above highlights the fact that RG’s upward mobility and the IGE de- scribe different features of the intergenerational transmission process—and that the two corresponding global databases complement each other. The remainder of the paper is organized as follows. Section 2 describes our methodology, data sources, and global coverage. Section 3 describes the database of intergenerational mobility and a set of stylized facts. Section 4 compares this database with RG’s global mobility database. Section 5 concludes. 2 Data and methodology 2.1 Estimation of intergenerational mobility Consider the following standard intergenerational regression specification: yi,t = c + βyi,t−1 + εi,t , (1) 6 where yi,t measures the log income earnings of individual i from generation t (where t refers to the child’s generation and t − 1 refers to the parent’s generation) and εi,t is a random 2 error term with mean zero and variance σε . The coefficient of interest is β , which measures the IGE. Higher values of β indicate greater intergenerational persistence and are thereby associated with lower levels of intergenerational mobility. For reviews of the literature, see, e.g., Solon (1999), Bjorklund and Jantti (1997), Black and Devereux (2011), and Jantti and Jenkins (2015). Ideally, yi,t and yi,t−1 measure (log) permanent incomes (i.e., lifetime earnings). Unfor- tunately, data on lifetime earnings are rarely available (Jenkins, 1987; Grawe, 2006).11 A commonly adopted alternative is to evaluate wage earnings around the age of 40, which is found to provide a reasonable approximation of permanent income (see, e.g., Haider and Solon, 2006; Lee and Solon, 2009).12 Age-adjusted (log) incomes are obtained by regressing individual (log) income on a polynomial of age minus 40: yi,t = γ0 + γ1 (Ai,t − 40) + γ2 (Ai,t − 40)2 + ui,t , (2) where Ai,t equals the age of individual i from generation t at the time the individual’s income is measured. This regression is estimated for each generation separately by means of Ordinary Least Squares (OLS). The residual from this regression offers an estimate of the individual’s income at the age of 40 (ignoring the intercept), and will be used as age-adjusted ¯i,t ). The quadratic specification, when evaluated over income data (henceforth denoted as y a suitable age range (the sample used typically includes individuals between the age of 30 and 55), is found to fit the data well (see, e.g., Chadwick and Solon, 2002; Osterberg, 2000; Peters, 1992).13 The next challenge is that surveys generally do not collect data on parental income. 11 Surveys typically collect data on current income earned over a given reference period. 12 A small number of papers have also used consumption (e.g., Bruze, 2018; Waldkirch et al., 2004). 13 The use of third- or fourth-order polynomials instead does not qualitatively alter our results. Nybom and Stuhler (2016) provide an examination of the biases introduced by the use of proxies for permanent income using high-quality administrative data. 7 While in some cases data on parental earnings can be retrieved from long-running panel surveys, these are rare exceptions. If we were to restrict ourselves to countries for which long-running panels are available, the database with estimates of intergenerational income mobility would include only a handful of countries. Retrospective data on parental education, age, and occupation are more common. These parental characteristics can be used to predict parental income earnings. The resulting predicted incomes can then be used as an instrument in the intergenerational income re- gression (see the seminal paper by Bjorklund and Jantti, 1997). This approach, referred to as TSTSLS, requires two independent survey samples. The first survey corresponds to the generation of children and contains data on their income earnings as well as data on their parents’ socioeconomic background (e.g., highest educational attainment, main occupation). The second survey corresponds to the parents’ generation (typically predating the offspring survey by about 15 to 20 years) and contains data on the income earnings, education, and main occupation of “pseudo-parents”. The estimation of β proceeds in three steps. First, estimate a Mincer-type income equa- tion using the older survey that is deemed representative of the current population of par- ents. Second, use the estimated model coefficients (i.e., return to education and experience) to predict parental income earnings given the retrospective data on parental education and occupation in the new offspring survey. Third, regress child (log) income (age adjusted) on predicted parental (log) income (also age adjusted). Our estimates of the IGE are potentially subject to several sources of bias.14 Let us highlight two sources that are arguably most relevant. The first is sample selection bias. We measure the intergenerational transmission of labor earnings, but not all working-age individuals earn labor income. And selection into waged employment is not random. As the share of waged employment varies between countries, the magnitude of this selection bias is also subject to variation. 14 See Nybom and Stuhler (2017) for biases on other dependence measures. 8 The second bias is characteristic of the TSTSLS estimator. This bias is introduced when parental education has a direct positive effect on child income beyond the indirect effect through parental income. In this case, TSTSLS estimates are subject to upward bias. The stronger the instrument, i.e., the stronger the correlation between parental education and parental income (leaving less variation in income that is not captured by paternal education), the smaller will be the upward bias. The magnitude of the bias is estimated to be in the range of 10–15 percent, which is small relative to between-countries differences (Bjorklund and Jantti, 1997). On the upside, TSTSLS helps protect against another source of bias. Parental income based on incomplete earnings histories may introduce measurement error. As a result, OLS estimates will be subject to attenuation bias (Solon, 1992). TSTSLS addresses this bias by using an instrument for parental income in the form of parental education and occupation. Applying the same method of estimation across all countries hopefully helps mitigate the impact of bias on cross-country comparisons. orklund and J¨ Since the original empirical application by Bj¨ antti (1997), which obtains (and compares) estimates of income mobility for the United States and Sweden, TSTSLS has been replicated across a wide range of contexts and has been established as the standard method of estimation in cases where direct information on parental income is not available. Importantly, it allows us to significantly expand the coverage of the developing world, as the existing evidence base is heavily biased toward the high-income world. 2.2 Data sources To construct a global database of intergenerational income mobility, we identify two rounds of nationally representative survey data, approximately 15 years apart, covering 78 economies. On average, the newer survey collected data around 2015 (most between 2010 and 2019), while the older survey collected data around 2001 (most between 1993 and 2007). These surveys collected the following information: (1) in the newer survey round, in- 9 come, labor force status, type of employment, and (usually through retrospective questions) parental characteristics such as educational attainment and, where available, occupation; (2) in the older survey round, income, labor force status, type of employment, educational attainment, and occupation. When multiple options for a given country are available, we select the survey based on the quality of information, particularly with respect to income variables and labor market information. Furthermore, for countries for which there is an empirical literature on intergen- erational income mobility, we use the same surveys when possible (e.g., Adli and Sahadewo, 2023; Daza, 2021; Piraino, 2015; Stockhausen, 2021; Yan Deng, 2022). Table A1 in the Appendix provides details on the surveys identified and the years of the survey rounds being used. We use several types of surveys. For most developing countries, we use cross-sectional household income or expenditure surveys. For most countries in Europe, we use EU Statis- tics on Income and Living Conditions (EU-SILC) and the intergenerational transmission of disadvantages module (collected in 2011, 2019, and 2023), which asks about parental background when respondents were 14 years old. For an additional number of countries in Europe, we use the Life in Transition Survey (LITS). We also use longitudinal surveys (treating two rounds of the survey as cross-sectional surveys) in a small number of countries such as Australia (Household, Income and Labour Dynamics in Australia; HILDA), Ger- many (Socio-Economic Panel; SOEP), Indonesia (Indonesia Family Life Survey; IFLS), the Republic of Korea (Korean Labor and Income Panel Study; KLIPS), the Russian Federation (Russian Longitudinal Monitoring Survey; RLMS), and the United States (Panel Study of Income Dynamics; PSID). We include estimates from the literature for nine additional countries for which we were unable to obtain access to the necessary survey data.15 For Argentina, the Netherlands, and Uruguay, the estimates employ the TSTSLS method with two rounds of nationally 15 The sources and methods are detailed in Table A2 of the Appendix. 10 representative data (see Jimenez, 2011; EqualChances, 2018; Araya, 2019). Estimates for Malaysia, Nepal, New Zealand, Pakistan, the Philippines, and Singapore use a variety of methods that, in general, employ an instrumental variable strategy or OLS on high-quality linked data (see Grawe, 2001; Greenaway-Mcgrevy, 2024; Javed and Irfan, 2014; Dacuycuy, 2018; Ng, 2013).16 Our estimates of income mobility are merged with cross-country data on selected cor- relates. We use GDP data from the World Development Indicators (WDI) database of the World Bank, which is completed with estimates from the Maddison Project Database (Bolt and van Zanden 2025) in cases where GDP data from the WDI are missing. Country and regional classifications are also obtained from the World Bank.17 Data on income inequal- ity, as measured by the Gini index, come from the World Bank’s Poverty and Inequality Platform (PIP). In the select few cases where PIP data are missing, we obtain estimates of the Gini index from the “All the Ginis” database.18 For each country, linear interpolation is used to impute Gini values for between-survey years (where survey-based estimates of the Gini index are not available). An indicator of tax-and-transfer progressivity constructed using income inequality pre- and post-fiscal system is obtained from Fisher-Post and Gethin (2023).19 Finally, returns to education are obtained from Montenegro and Patrinos (2023), and missing years are completed by means of interpolation. In cases where the time coverage of the correlates does not line up, we use the closest year available. For IGE estimation, it is not required that the income variables are expressed in real terms. This is a requirement, however, when estimating the upward mobility measures proposed by RG (2023). Accordingly, we use consumer price indices obtained from the World Bank’s WDI and the October 2024 World Economic Outlook database of the International 16 Two rely on coresident samples, which may generate bias (see, Emran et al., 2018; Francesconi and Nicoletti, 2006). A number of articles also study mobility by caste or for the rural sector (e.g., Asadullah, 2012; Bevis and Barrett, 2015; Hnatkovska et al., 2013). 17 See the latest classification available at https://datahelpdesk.worldbank.org/knowledgebase/articles/906519- world-bank-country-and-lending-groups. 18 Available at https://stonecenter.gc.cuny.edu/research/all-the-ginis-alg-dataset-version-february-2019/. 19 Available at https://amory-gethin.fr/data.html. 11 Monetary Fund to express income in currency from 2011. In some cases, countries also implemented currency changes between survey rounds; thus, we use the exchange rates at the time of those changes. 2.3 Sample definitions Our samples of interest are defined as follows. Sample of sons. This sample consists of male individuals between 30 and 54 years of age who report positive earnings and are paid employees. In a number of countries where the sample is relatively small, we expand this sample to include individuals between 25 and 60 years of age (or relax the constraint of using only paid employees). We exclude indi- viduals with missing information on the education and/or occupation (when the respective survey collects such information) of their fathers. Outliers, defined as observations with a studentized residual in absolute value equal to or greater than 5, are excluded when the intergenerational income regression is estimated. For each sample, information about the exact age range, restriction based on type of employment, and the availability of occupation can be found in Table A3 of the Appendix. Sample of synthetic parents. This sample consists of male individuals between 30 and 54 years of age who report positive earnings and are paid employees. In countries where the sample of sons includes individuals between 25 and 60 years of age, we similarly expand the sample age range of synthetic fathers. We exclude individuals with missing information about their educational attainment and/or occupation (when the respective survey collects such information). Outliers, defined as observations with a studentized residual in absolute value equal to or greater than 5, are excluded when the Mincer-type regression used to predict parental income is estimated. For each sample, information about the exact age range, restriction based on type of employment, and the availability of occupation can be found in the Appendix (Table A3). 12 2.4 Variable construction The variables that we use are constructed as follows. Child income. We construct daily labor earnings for all individuals active in the labor market. We use information about the reference or recall period (e.g., per day, week, month) to express earnings for a given period in daily terms. The precise definition of earnings in terms of the different components included (e.g., cash, in-kind, bonuses) and whether the earnings are gross or net of paid taxes varies by survey wave. In most cases, we use local currency units.20 We express income in real terms when replicating RG’s indicators of upward mobility with our micro data. Paternal income. We construct the daily labor earnings of the father as the predicted age-adjusted daily labor earnings. These earnings are predicted according to the method- ology described in section 2.1 using retrospective information about parents’ education and occupation, when available. For the synthetic parents, we construct daily labor earnings in the same way as we do with child income. Education. We use educational attainment as an instrumental variable for paternal income in our TSTSLS application. For each individual in the newer survey, we construct the educational attainment of the father using retrospective information or the panel struc- ture, when feasible. In the older survey, we work with the educational attainment of the respondents (i.e., “synthetic parents”). Different surveys measure education outcomes differ- ently, ranging from years of schooling to a set of categories that are not homogeneous across countries or, sometimes, within countries (i.e., to describe paternal educational attainment, the newer survey contains a set of educational categories that differ from those available for synthetic parents in the older survey). We harmonize the data on a case-by-case basis to make the information in the newer survey comparable with that in the older survey. In some cases, doing so requires constructing years of schooling from categorical variables. For this purpose, a repository of country-level International Standard Classification of Educa- 20 The EU-SILC data are an important exception, as they come with earnings expressed in euros. 13 tion (ISCED) mappings over time (starting in 1970) is available to facilitate conversion from categories to years by the year of schooling to which the individual’s schooling tenure corre- sponds. If information is missing in the ISCED sources for the mapping exercise, additional economy-specific information is used, or the following rules of thumb for converting ISCED categories to completed years of schooling are applied: ISCED 1: 0 years; ISCED 1: 6 years; ISCED 2: 9 years; ISCED 3: 12 years; ISCED 4: 13 years; ISCED 5: 15 years; ISCED 6: 16 years; ISCED 7: 18 years; and ISCED 8: 21 years.21 In other cases, aggregating categories is required to obtain a common minimum. Occupation. When possible, we use occupation as an additional instrumental variable for paternal income. For each individual in the newer survey, data on paternal occupation is obtained from the retrospective module or by means of the panel structure when feasible. In the older survey, we work with the occupation data for the respondents (that is, “synthetic parents”). The way in which occupation data are collected may vary between and within countries (i.e., to describe paternal occupation, the newer survey contains a set of occupa- tional categories that differs from those available for synthetic parents in the older survey). We harmonize the data on a case-by-case basis to make the information in the newer survey comparable with that in the older survey. 2.5 Country coverage Our database includes estimates for 87 countries that account for 84 percent of the world population (Table 1). These countries account for 94 percent of the population in high- income economies and 81 percent of the population in developing countries. Within the latter, in all regions except the Middle East and North Africa region, the population coverage is more than 60 percent. For the Middle East and North Africa, the coverage is 40 percent. 21 To construct educational variables, this approach is adopted by Van der Weide (2024). 14 Table 1: Coverage Income group/region Number of economies Population (%) High-income economies 43 94.48 Developing economies 44 81.20 East Asia and the Pacific 6 91.22 Europe and Central Asia 9 63.40 Latin America and the Caribbean 8 83.02 Middle East and North Africa 4 40.35 South Asia 4 89.14 Sub-Saharan Africa 13 65.59 World 87 83.63 Notes: The table shows the number of countries our database covers, and the population share (using UN population estimates for 2015) that these countries account for. 3 Income mobility around the world 3.1 A first look at the database Income mobility shows large cross-country variation, with lower levels of mobility concen- trated in the developing world (similar to the patterns for intergenerational mobility in education observed by Van der Weide et al., 2024) (see Figure 1). Our estimates of income mobility indicate that 11 of the 15 least mobile countries are in Africa, Latin America, and South Asia, with income persistence coefficients in excess of 0.64. African countries with low levels of income mobility are not limited to the continent’s poorest countries, such as the Democratic Republic of Congo, Niger, and Madagascar. They also include large emerg- ing economies such as Nigeria and South Africa. In Latin America, low-mobility countries include Colombia, Guatemala, and Mexico. In South Asia, India records the lowest level of income mobility in our database. In contrast, among the 15 countries with the highest estimates of income mobility (with income persistence coefficients below 0.28), 12 are high-income countries, and 11 are in Europe and Central Asia. The three most mobile countries in our database are all Nordic countries (Sweden, Norway, and Finland, with Denmark not far behind). 15 To illustrate the magnitude of cross-country variation in income mobility levels across countries, consider two fathers, one earning twice the income of the other father. All else being equal, the son of the higher-income father can expect to earn 50–70 percent more than the son of the lower-income father in Brazil, Argentina, Peru, Indonesia, China, and the United States. This income divergence increases to more than 80 percent in the Arab Republic of Egypt, Tunisia, and Guatemala. By comparison, the son of the higher-income father can expect to earn less than 25 percent more in Sweden, Finland, Norway, Denmark, and Canada. Figure 1: Intergenerational income elasticity around the world Notes: The map shows country-level estimates of the intergenerational income elasticity. Coun- tries are classified into quintiles based on the mobility measure. A higher value indicates lower intergenerational income mobility. Unlike educational mobility, income mobility is also a function of the functioning of the labor market. Countries may achieve a high level of educational mobility (i.e., provide a level playing field in regard to the accumulation of human capital) but fail to achieve a high level of income mobility (i.e., are unable to provide a level playing field in regard to the labor market). Think of countries where good jobs are in short supply and tend to be reserved for individuals from privileged and well-connected backgrounds. Figure 2 plots our estimates of intergenerational persistence in income against estimates 16 of persistence in education from Van der Weide et al. (2024). As expected, the two measures are highly correlated. However, there are several countries for which a notable divergence is observed. Egypt and Tunisia stand out with levels of income mobility that are low given their levels of educational mobility.22 The reverse can also be observed. Countries where income mobility is higher than what might be expected based on their levels of educational mobility include Serbia and Turkey. Note that the Nordic countries rank at the top in terms of income mobility but are on par with other high-income countries in terms of educational mobility. Figure 2: Intergenerational mobility in income and education 1 MDG GTM EGY Correlation coef.: 0.479 TUN Intergenerational income elasticity .8 IND XKX MEX COD NGA COL ZAF PER NER ARG BGR ETH USA ECU .6 CYP CHL TZAPAN ROU MYSBIH BRA GHA MNE CHN MAR TGO IDN ESP UGA GBR LVA MNG BOLMWI CZE HUN SVK ITA PHL IRL URY PAK NPL FRA VNM .4 LTU DEU KEN TWN PRT BEL POL HRV EST LBR UZB GRC RUS AUT JOR KAZ NLD JPN ISL CHE AUS TUR LKA SVN KOR TJK DNK MDA .2 CAN SRB NOR FIN SWE 0 0 .2 .4 .6 .8 Educational persistence (beta coefficient) Notes: Estimates of educational persistence for father and sons are obtained from Van der Weide et al. (2024). Data for Luxembourg, Malta, New Zealand, and Singapore are missing. We match our database with the educational mobility of the cohort born in the 1980s. 3.2 Correlates predicted by theory Theory predicts that the IGE varies with structural parameters that include the efficacy of private and public investments in generating human capital, the returns to education, and the progressivity of public investments (Becker and Tomes, 1979, 1986; Becker et al., 2018; Solon, 2004, 2014). In these stylized models, the resulting IGE increases with the 22 Other countries with significantly higher persistence in income than in education include South Africa, Madagascar, and Guatemala. 17 returns to education (and with efficacy of investments into human capital) and decreases with the progressivity of public investments. Higher returns to education in the parents’ generation mean that parents with more human capital earn higher incomes and thus have more resources to invest in the education of their children. Higher returns to education in the child’s generation mean that parents with the necessary resources furthermore have greater incentives to invest in the human capital of their offspring. More progressive public investments partially compensate for the inequalities in private investments, which helps level the playing field and lower the intergenerational persistence in income. There is good cross-country data on returns to education (by Montenegro and Patrinos, 2023) and the progressivity of public spending (by Fisher-Post and Gethin, 2023), both compiled in recent years.23 Progressivity is measured by the change in inequality (evaluated as the ratio of the average income of the top 10% to that of the bottom 50%) between pre- and post-tax and transfers (Fisher-Post and Gethin, 2023). Merging these data with our database on income mobility allows us to verify whether the IGEs respect these theoretical predictions.24 The empirically observed correlations between the IGE and the selected correlates are seen to conform with theory, see Figure 3. Income mobility (evaluated by 1 − IGE ) is observed to be positively correlated with progressivity (left panel), and negatively correlated with returns to schooling (right panel). 3.3 The Great Gatsby Curve revisited Arguably the most famous stylized fact established in the empirical literature on intergener- ational mobility is the Great Gatsby Curve (GGC), which captures the empirical observation that low (high) levels of intergenerational mobility often coincide with high (low) levels of income inequality (measured in the parents’ generation). The GGC was first established us- 23 Data on the parameters of the human capital production function is unfortunately not available at this point. 24 Holter (2015) uses data from 11 countries to document a strong negative cross-country correlation between IGE and measures of tax progressivity. 18 Figure 3: Intergenerational mobility vs. progressivity and returns to education 1 1 Correlation coef.: 0.341 Correlation coef.: -0.367 SWE SWE Intergenerational Mobility (1-IGE) Intergenerational Mobility (1-IGE) NOR FIN NOR FIN SRB SRB .8 KOR TJK MDA CAN DNK .8 DNK CAN KOR TJK MDA SGP SVN SGP SVN LKA TURCHE AUS AUS CHE LKA TUR ISL NZL ISL JPN NLD JPN NLD RUS JORKAZ JOR RUSGRC LBR EST GRCPOL AUT UZB HRV BEL EST BEL AUT HRV POL LBR TWN .6 PHL VNM URY KEN LTU DEU PRT FRA .6 VNM DEU LTU FRA URY PRT PHL KEN PAK IRL NPL PAK LUX SVK MWI ITA ITA SVK IRL LUX MWI MNG BOL CZE HUN LVA GBR MNG LVA CZE HUN BOL UGA IDN MAR ESP ESP IDN MAR UGA CHN TGO CHN BIHGHA MNE BRA BIH GHA BRA MYS ROU ROU CYP CHL TZACHL .4 TZA ECU PAN USA .4 USA ECU PAN ETH BGR MLT ARG BGR MLT ARG ETH NER PER ZAF PER NER ZAF NGA COL NGA COL COD XKX MEX MEX COD .2 IND .2 IND TUN TUN EGY EGY GTM GTM MDG MDG 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 4 8 12 16 20 24 28 Progressivity Returns to tertiary education Confidence interval Local polynomial fit Linear fit Confidence interval Local polynomial fit Linear fit (a) Progressivity of tax and transfers (b) Returns to education Notes: Progressivity of tax and transfers is measured as the change in inequality (average income ratio of the top 10% to the bottom 50%) between pre- and post-tax-and-transfers national income using data from Fisher-Post and Gethin (2023). Data for New Zealand, Singapore, Taiwan, and Malaysia are missing. Returns to education are obtained from Montenegro and Patrinos (2023) and are estimated using the “Mincerian” method with data from household surveys. Data for Cyprus, Kazakhstan, Kosovo, Montenegro, New Zealand, Taiwan, and Uzbekistan are missing. We match our mobility estimates with data from 20 years earlier, using the year in which income was collected in the newer survey as our reference point. For example, for an estimate that uses income for children collected in 2015, we match with data on progressivity and returns to education from 1995. ing data for 13 countries (mostly high-income and emerging economies) compiled by Corak (2013). Reproducing the GGC using our global database with estimates of income mobility for 87 countries shows that this stylized fact continues to hold across this wider range of coun- tries covering both the developed and developing world. The positive correlation coefficient equals 0.435 and is statistically significant at the 1 percent level. The Nordic countries (i.e., Denmark, Norway, Sweden, and Finland) stand out as countries with some of the highest levels of income mobility and low levels of inequality. Examples of countries occupying the other corner of the GGC, with low levels of mobility and high levels of income inequality, include South Africa and Guatemala, among others. 19 Figure 4: Intergenerational mobility and income inequality 1 Correlation coef.: 0.435 SWE Intergenerational Mobility (1-IGE) NORFIN SRB CAN .8 MDA KOR TJK DNK SGP SVN TUR LKAAUS CHE NZL ISL JPN NLD RUS JOR KAZ GRC AUT UZB EST LBR POL BEL HRV TWN .6 KEN URY PHL PRT VNM LTU FRA DEU PAK IRL NPL MWI ITA LUX SVK BOL GBRLVA MNG HUN UGA ESP IDN CZE TGO MAR CHN BRA MYS GHA BIH MNE ROU CHL CYP .4 PAN ECU USA TZA ARG ETH BGR MLT ZAF PER NER COL NGA MEX COD XKX .2 IND TUN EGY GTM MDG 0 .4 .5 .6 .7 .8 Equality (1-Gini coefficient) Confidence interval Local polynomial fit Linear fit Notes: Income inequality is measured with the Gini index obtained from World Bank’s Poverty and Inequality Platform (PIP) supplemented with data from “All the Ginis” database. We match our mobility estimates with inequality data from 20 years earlier, using the year in which income was collected in the newer survey (circa 2015) as our reference point. For example, for an estimate that uses income for children collected in 2015, we match with data on inequality from 1995. 3.4 Intergenerational mobility versus national income A lesser-known stylized fact describes the relationship between intergenerational mobility and national income. Van der Weide et al. (2024) observe that intergenerational mobility in education increases with GDP per capita. Figure 5 revisits this stylized fact with our estimates of income mobility. National income per capita is measured 20 years prior to the year in which income data for children are collected (i.e., the same lag that is used in the GGC). This confirms that the empirical observation that countries achieve higher rates of intergenerational mobility as they grow richer holds for both education and income mobility. Should intergenerational mobility be expected to increase with national income? While we can only speculate at this point, theory offers some guidance. Different channels can, in theory, pull mobility in different directions as countries develop. In the models developed by, e.g., Becker and Tomes (1979, 1986) and Solon (2004), increases in returns to education will lead to higher levels of intergenerational income persistence. All else being equal, this would predict a negative relationship between income mobility and national income. 20 Candidate channels that could rationalize a positive relationship are progressive public transfers and investments (including public education) and improvements in credit markets. As countries develop, the effect of credit constraints arguably declines as credit markets become more efficient and poorer households are able to secure the necessary resources to invest in their children, which would reduce intergenerational persistence (see, e.g., Maoz and Moav, 1999; Owen and Weil, 1998). Public interventions that partially compensate for inequalities in private investments constitute another plausible channel through which higher levels of income mobility could be realized as countries develop (see, e.g., Benabou, 2000). Richer countries will have more fiscal space available to finance and implement such public policies.25 It is well documented that public spending (on, e.g., public education, healthcare, childcare) increases with national income, not only in absolute value but also as a share of national income, which is known as Wagner’s law (Wagner, 1890). Figure 5: Intergenerational mobility and GDP per capita 1 Correlation coef.: 0.382 SWE Intergenerational Mobility (1-IGE) FIN NOR SRB .8 TJK MDA KOR CAN DNK SVN LKA TUR AUSSGP ISL CHE NZL JPN NLD LBR JOR KAZ RUS GRC AUT UZB EST HRV POL BEL TWN PRT .6 VNM KEN PHL LTU URY DEU FRA NPL PAK IRL MWI SVK ITA LUX MNG BOL LVA HUN GBR UGA CZEESP CHN TGO MAR IDN GHA BIH MYS MNE BRA ROU CHL CYP .4 TZA ECU PAN USA ETH BGR MLT ARG NER PER ZAF NGA COL COD XKX MEX .2 IND TUN EGY GTM MDG 0 500 1000 2000 4000 8000 16000 32000 64000 GDP per capita (2021 USD PPP, logarithmic scale) Confidence interval Local polynomial fit Linear fit Notes: GDP data is from the World Development Indicators supplemented with data from the Maddison project where necessary. We match our mobility estimates with GDP data from 20 years earlier, using the year in which income was collected in the newer survey (circa 2015) as our reference point. For example, for an estimate that uses income for children collected in 2015, we match with GDP data from 1995. 25 Mayer and Lopoo (2008) use state-level data in the US to show that government spending increases mobility. 21 4 Comparison with RG’s global mobility database 4.1 RG’s measures of upward mobility Ray and Genicot (2023) put forward measures of upward mobility that have attractive axiomatic properties. A practical advantage of their mobility measures is that they can be estimated using repeated cross-sectional surveys (i.e., no panel data are required). This feature greatly expands the number of countries for which upward mobility can be evaluated. Utilizing this property, RG compiled a companion global database of upward mobility that includes 122 countries covering a large share of the world’s population.26 As upward mobility and the IGE capture distinct aspects of the intergenerational transmission process, the two global databases complement each other. RG distinguish between absolute and relative upward mobility, which we will denote by M and R, respectively: −α 1 E [yit ] M = − log −α (3) α E [yit−1 ] −α −α 1 E [yit ] 1 E [yit −1 ] R = − log − + log , (4) α E [yit ] α α E [yit−1 ]−α for α > 0, where yit measures income for individual i from generation t. The parameter α governs the degree of growth progressivity; with higher values of α, the measures of upward mobility will attach greater weight to growth of the poor. The estimates of M and R compiled by RG are based on percentiles of the income distributions coming from the World Inequality Database (WID) database. We obtain a companion set of estimates of RG’s upward mobility measures using the same micro data used to estimate the IGE coefficients in our database to verify whether the stylized facts established with the measures of upward mobility are robust to the source of data used.27 26 A smaller database with 71 countries was used in Ray and Genicot (2023). 27 Figure A1 of the Appendix plots estimates of RG’s mobility indicators using our microdata versus those estimated in Ray and Genicot (2023) using WID. Estimates with these different data sources show important differences, especially for relative mobility. 22 To understand how RG’s measures of intergenerational upward mobility relate to the IGE and the levels of income inequality across generations, it may be helpful to evaluate analytical expressions for the upward mobility measures under the assumed linear intergenerational transmission regression with normally distributed errors:28 log yit = c + β log yit−1 + σu uit , (5) 2 where uit is normally distributed with mean zero and unit variance. Let µt and σt measure the unconditional mean and variance of log income for generation t. Under these assumptions, it follows that: −α E [yit 2 ] = exp (−αµt + α2 σt /2) (6) E [yit ]−α = exp (−αµt − ασt 2 /2). (7) Similarly substituting these moments into the expression for M yields: 1 M = − log exp −αµt + α2 σt 2 2 /2 − (−αµt−1 + α2 σt −1 /2) (8) α 1 1 = − −α(µt − µt−1 ) + α2 (σt 2 2 − σt−1 ) (9) α 2 1 2 2 = (µt − µt−1 ) − α(σt − σt −1 ). (10) 2 Similarly substituting these moments into the expression for R, we obtain: 1 R = − 2 log exp −αµt + α2 σt /2 − (−αµt − ασt 2 /2) α 1 + log exp −αµt−1 + α2 σt2 2 −1 /2 − (−αµt−1 − ασt−1 /2) α 1 1 1 1 = − log exp (α2 + α)σt 2 + log exp (α2 + α)σt2 −1 α 2 α 2 1 2 2 = − (1 + α) σt − σt−1 . 2 28 This makes the relationship between upward mobility, the IGE, and inequality explicit. 23 Equation (10) confirms that the upward mobility measure M aligns well with the income 2 growth rates of the poor. Using a first-order Taylor expansion for σt 2 ¯ 2 , i.e., σt around σ ≈ ¯ 2 + 2¯ σ σ (σt − σ ¯ ), eq. (10) is seen to approximate: ¯ (σt − σt−1 ). M ≈ (µt − µt−1 ) − ασ (11) Under the assumption of log-normally distributed income, it can be verified that this approx- imation to M equals the income growth rate at percentile pα = Φ(−ασ ¯ ) between generations t and t − 1: M ≈ log yt (pα ) − log yt−1 (pα ), (12) where Φ() is the quantile function of the standard normal distribution, and where we used the fact that log income at percentile p in generation t satisfies: log yt (p) = µt + σt Φ−1 (p). α > 0 implies that pα < 0.5, i.e., M measures growth below the median. By increasing the value of α, one lowers the income percentile pα along which growth is measured. To see how M and R relate to the IGE coefficient β , consider a re-parameterization in terms of implied steady-state values. The steady-state mean and variance of log income are seen to satisfy: c ¯ = µ (13) 1−β 2 σu ¯2 σ = . (14) 1 − β2 2 2 Accordingly, (µt − µt−1 ) and (σt − σt−1 ) may be written as: µ − µt−1 ) µt − µt−1 = (1 − β )(¯ (15) 2 2 2 σt − σt 2 σ 2 − σt −1 = (1 − β )(¯ −1 ). (16) 24 By substituting these into the expressions for M and R, we obtain: α M = (1 − β ) (¯ µ − µt−1 ) − 2 σ 2 − σt (1 + β )(¯ −1 ) (17) 2 1 ¯ 2 − σt R = − (1 + α)(1 − β 2 ) σ 2 −1 . (18) 2 It follows that the upward mobility measures M and R are reduced to zero when the moments µ and σ 2 are in steady state. By extension, M and R exhibit a stronger (weaker) relationship with the IGE β when the income distribution is further away from (closer to) steady state. Figure 6 empirically illustrates how RG’s absolute upward mobility is closely related to annualized growth at the bottom of the income distribution (first graph) and that the strength of this association decreases when considering average growth (second graph) or growth at the top of the distribution (third graph).29 Figure 6: Growth rates versus Ray and Genicot’s indicators 8 8 8 Correlation coef.: 0.967 Correlation coef.: 0.775 Correlation coef.: 0.501 6 6 6 Annualized growth rate (9th decile) Annualized growth rate (1st decile) Annualized growth rate (Average) 4 4 4 2 2 2 0 0 0 -2 -2 -2 -4 -4 -4 -4 -2 0 2 4 6 8 -4 -2 0 2 4 6 8 -4 -2 0 2 4 6 8 Absolute mobility (RG) Absolute mobility (RG) Absolute mobility (RG) Notes: This figure plots the indicator of absolute mobility proposed by Ray and Genicot (2023) against growth rates at different percentiles of the income distribution. The graph includes 75 countries. The indicators of mobility are computed in Ray and Genicot (2023) while the growth rates are estimated by the authors using the same underlying data. 29 Figure A2 in the Appendix replicates this figure using relative mobility showing that the same pattern does not hold in this case. Qualitatively similar patterns are found using our data (see Figures A3 and A4). 25 4.2 The IGE versus RG Given that RG’s measure of upward mobility does not require panel data, the authors are able to obtain estimates for 122 countries. Our estimates of the IGE show a weak correlation with our estimates of RG’s absolute mobility measure (Figure 7).30 This observation suggests that these two indicators capture different aspects of the intergenerational transmission process (and that their lack of a strong association is not due to differences in data sources). Figure 7: Intergenerational elasticity of income vs. Ray and Genicot’s indicators 1 Correlation coef.: 0.067 SWE FIN NOR SRB Intergenerational Mobility (1-IGE) .8 CAN DNK KOR MDA SVN LKA AUS CHE ISL TUR JPN GRC AUT KAZ JOR BEL HRV EST POL TWN DEU .6 KEN PRT FRA LTU VNM IRL ITA LUX MWI SVK GBR MNG HUN BOL LVA ESP UGA CZE MAR IDN CHN BRAMNE BIH GHA ROU CYP CHL .4 PAN USA TZA ECU MLT ETH BGR NER ZAF PER COL NGA MEX XKX COD .2 IND TUN EGY GTM MDG 0 -5 0 5 10 15 Absolute mobility (RG) Notes: This figure plots the intergenerational income elasticity against the indicator of absolute mobility proposed by Ray and Genicot (2023). The graph includes 73 countries. RG indicators of mobility are estimated with the same sample used by the authors to estimate the IGE. 4.3 RG’s upward mobility versus inequality and national income When the estimates are plotted against lagged income inequality to verify whether the GGC applies when intergenerational mobility is measured using RG’s upward mobility measure, the estimated relationship is found to be flat (or even slope in the opposite direction), see Ray and Genicot (2023) and Genicot et al. (2024). RG attribute the weaker GGC to the inclusion of countries from the developing world 30 Relative mobility is also weakly correlated, see Figure A7 in the Appendix. Our IGE estimates are also weakly correlated with those in Ray and Genicot (2023), which use data from WID (Figure A5). 26 (which were not included when the original version of the GGC was established; see, e.g., Corak, 2013). Plotting estimates of income mobility for the 84 countries included in our database, which covers 81 percent of the population in the developing world and 84 percent of the world population, shows that the GGC continues to hold when mobility is measured using the IGE (compare Figures 4 and 8a). That suggests that there are other reasons why the GGC does not hold for the upward mobility measure. One candidate reason is that RG’s upward mobility measure captures a different feature of the intergenerational transmission process. As highlighted in Section 4.1 and noted by Ray and Genicot (2023), the upward mobility measure is closely aligned with growth (at lower percentiles of the income distribution). This may help explain the weaker relationship with income inequality. There is a large empirical literature on the relationship between inequality and growth which by and large finds that this relationship is weak (see, e.g., Brueckner and Lederman, 2018; Van der Weide and Milanovic, 2018; Banerjee and Duflo, 2003; Baselgia and Foellmi, 2023). Plotting the estimates of upward mobility against national income similarly shows that the empirical relationship changes sign when compared to the relationship observed for the IGE (compare Figure 5 to Figure 8b).31 The negative correlation between upward mobility and national income sits well with the interpretation of upward mobility as a measure of growth. The negative association in this case describes a convergence effect (or regression to the mean), i.e., lower-income countries have greater scope for catch-up growth. 5 Concluding remarks This study builds the largest cross-country database on intergenerational income mobility to date, with estimates of the IGE for 87 countries covering 84 percent of the world’s population. This marks a significant expansion of the existing cross-country evidence base on income mobility, particularly among low- and middle-income countries. The most comprehensive 31 Figure A6 in the Appendix replicates the figure using RG’s relative mobility showing a different pattern. 27 Figure 8: Absolute mobility vs. national income and inequality 15 15 Correlation coef.: -0.004 VNM Correlation coef.: -0.333 VNM 10 JOR MDA 10 MDA JOR COD KAZ COD KAZ Absolute Mobility (RG) Absolute Mobility (RG) GHA GHA NGA CHE NGA CHE ECU ECU LVA BIH BIH LVA ROU SVK SVK ROU 5 TZA POL LTU 5 TZA LTUPOL TUR EST CHN CZE CHN TUR EST CZE CHL SRB XKX BGR XKX SRB CHL BGR BOL IDN BOL IDN MWI MNE MWI MNE BRA PER KOR IND IND PER BRA KOR COL UGA EGY UGA EGY COL AUS AUS ZAF ETHUSA LKA FRA IRL HUN HRV FIN SWE SVN ETH LKA ZAF HRV HUN SVN USA FIN SWE IRL 0 TUN PRTMAR LUX DEU DNK 0 MAR TUN PRT FRADEU DNK LUX CYP NOR CYP ITA NOR ITA CAN AUT MLT TWN MLT TWN CAN AUT KEN MDG NER JPN BEL NER MDG KEN JPN BEL MEX ISL MEX ISL PAN MNG MNG PAN GRC GRC GBR GBR GTM GTM -5 -5 ESP ESP .4 .5 .6 .7 .8 500 1000 2000 4000 8000 16000 32000 64000 Equality (1-Gini coefficient) GDP per capita (2021 USD PPP, logarithmic scale) (a) Income inequality (b) Gross domestic product Notes: This figure plots the indicator of absolute mobility proposed by Ray and Genicot (2023) against the Gini index and GDP. The graph includes 73 countries. RG indicators of mobility are estimated using the same sample as the authors to estimate the IGE. Income inequality is measured with the Gini index obtained from World Bank’s Poverty and Inequality Platform (PIP) supplemented with data from “All the Ginis” database. GDP data is from the World Development Indicators supplemented with data from the Maddison project where necessary. We match our mobility estimates with GDP and inequality data from 20 years earlier, using the year in which income was collected in the newer survey as our reference point. For example, for an estimate that uses income for children collected in 2015, we match with data on inequality from 1995. cross-country studies prior to ours include estimates for 21 to 29 countries (Durlauf, 2022; Mogstad and Torsvik, 2023). An inspection of the variation in income mobility across the 87 countries yields several empirical observations. First, the IGE is seen to vary significantly, ranging from levels near 0.1 for the Nordic countries to values near 1 in parts of the developing world. Second, in line with predictions from theory, intergenerational mobility is found to be negatively correlated with returns to education and positively correlated with the progressivity of the fiscal system. Third, the negative association between income mobility and inequality that was first established for 13 mostly high-income countries (known as the Great Gatsby Curve) also holds across this wider range of countries that includes 44 countries from the developing world. Fourth, income mobility is found to have a positive association with national income per capita. Future research could expand this work in several directions. First, given the differences 28 in labor markets between developed and developing countries and the fact that we estimate the intergenerational transmission in labor earnings, the extent to which sample selection shapes our findings is an open question. Second, it would be insightful to establish empirically through what channels countries achieve higher levels of intergenerational mobility as they grow richer (and evaluate the relative importance of these channels). Third, our study focuses on father-son pairings. While this is a standard choice in the literature, further research could expand it to include mothers and daughters. 29 References Aaronson, Daniel and Mazumder, Bhashkar (2008). Intergenerational Economic Mobility in the United States, 1940 to 2000. Journal of Human Resources, 43, number 1. Acciari, Paolo, Polo, Alberto and Violante, Giovanni L. (2022). And Yet It Moves: Intergen- erational Mobility in Italy. American Economic Journal: Applied Economics, 14, number 3, 118–163. Adli, Muhammad Imam and Sahadewo, Gumilang Aryo (2023). Estimating Intergenerational Income Mobility in Indonesia. Jurnal Ekonomi dan Pembangunan, 31, number 1, 43–58. Alesina, Alberto, Hohmann, Sebastian, Michalopoulos, Stelios and Papaioannou, Elias (2021). Intergenerational Mobility in Africa. Econometrica, 89, number 1, 1–35. Andrews, Dan and Leigh, Andrew (2009). More Inequality, Less Social Mobility. Applied Economics Letters, 16, number 15, 1489–1492. Araya, Federico (2019). Evidence about Intergenerational Earnings Mobility for a Developing Country: The Uruguayan Case. El Trimestre Economico L, XXXVI (2), number 342, 265–305. Asadullah, M. Niaz (2012). Intergenerational Wealth Mobility in Rural Bangladesh. Journal of Development Studies, 48, number 9, 1193–1208. Asher, Sam, Novosad, Paul and Rafkin, Charlie (2024). Intergenerational Mobility in In- dia: New Measures and Estimates Across Time and Social Groups. American Economic Journal: Applied Economics, 16, number 2, 66–98. Banerjee, Abhijit V and Duflo, Esther (2003). Inequality and Growth: What Can the Data Say? Journal of Economic Growth, 8, 267–299. 30 Baselgia, Enea and Foellmi, Reto (2023). Inequality and Growth. In Handbook of Labor, Human Resources and Population Economics (ed. K.F. Zimmermann), WIDER Working Paper, volume 2023, pp. 1–41. UNU-WIDER. Springer, Cham. org L. (2018). Becker, Gary S, Kominers, Scott Duke, Murphy, Kevin M and Spenkuch, J¨ A Theory of Intergenerational Mobility. Journal of Political Economy, 126, number S1, S7–S25. Becker, Gary S and Tomes, Nigel (1979). An Equilibrium Theory of the Distribution of Income and Intergenerational Mobility. Journal of Political Economy, 87, number 6, 1153–1189. Becker, Gary S and Tomes, Nigel (1986). Human Capital and the Rise and Fall of Families. Journal of Labor Economics, 4, number 3, S1–S39. enabou, Roland (2000). Unequal Societies: Income Distribution and the Social Contract. B´ The American Economic Review, 90, number 1, 96–129. Bevis, Leah E.M. and Barrett, Christopher B. (2015). Decomposing Intergenerational In- come Elasticity: The Gender-differentiated Contribution of Capital Transmission in Rural Philippines. World Development, 74, 233–252. orklund, Anders and J¨ Bj¨ antti, Markus (1997). Intergenerational Income Mobility in Sweden Compared to the United States. The American Economic Review, 87, number 5, 1009– 1018. Black, Sandra E. and Devereux, Paul J. (2011). Recent Developments in Intergenerational Mobility. In Handbook of Labor Economics (eds David Card and Orley Ashenfelter), volume 4b, chapter 16, pp. 1487–1541. Elsevier. Blanden, Jo (2013). Cross-country Rankings in Intergenerational Mobility: A Comparison 31 of Approaches from Economics and Sociology. Journal of Economic Surveys, 27, number 1, 38–73. Blanden, Jo, Haveman, Robert, Smeeding, Timothy and Wilson, Kathryn (2014). Inter- generational Mobility in the United states and Great Britain: A Comparative Study of Parent-child Pathways. Review of Income and Wealth, 60, number 3, 425–449. Bolt, Jutta and van Zanden, Jan Luiten (2025). Maddison-style estimates of the evolution of the world economy: A new 2023 update. Journal of Economic Surveys, 39, number 2. antti, Markus, Eriksson, Bratsberg, Bernt, Røed, Knut, Raaum, Oddbjørn, Naylor, Robin, J¨ ¨ Tor and Osterbacka, Eva (2007). Nonlinearities in Intergenerational Earnings Mobility: Consequences for Cross-country Comparisons. Economic Journal, 117, number 519, 72– 92. Britto, Diogo, Fonseca, Alexandre, Pinotti, Paolo, Sampaio, Breno and Warwar, Lucas (2022). Intergenerational Mobility in the Land of Inequality. IZA Discussion Papers, number 15611. Brueckner, Markus and Lederman, Daniel (2018). Inequality and Economic Growth: The Role of Initial Income. Journal of Economic Growth, 23, 341–366. Bruze, Gustaf (2018). Intergenerational Mobility: New Evidence from Consumption Data. Journal of Applied Econometrics, 33, number 4, 580–593. Card, David, Domnisoru, Ciprian and Taylor, Lowell (2022). The Intergenerational Trans- mission of Human Capital: Evidence from the Golden Age of Upward Mobility. Journal of Labor Economics, 40, number S1, S39–S95. Chadwick, Laura and Solon, Gary (2002). Intergenerational Income Mobility Among Daugh- ters. The American Economic Review, 92, number 1, 335–344. 32 Chetty, Raj, Hendren, Nathaniel, Kline, Patrick and Saez, Emmanuel (2014). Where is the land of opportunity? The geography of intergenerational mobility in the United States. Quarterly Journal of Economics, 129, number 4, 1553–1623. Chetty, Raj, Jackson, Matthew O., Kuchler, Theresa, Stroebel, Johannes, Hendren, Nathaniel, Fluegge, Robert B., Gong, Sara, Gonzalez, Federico, Grondin, Armelle, Jacob, Matthew, Johnston, Drew, Koenen, Martin, Laguna-Muggenburg, Eduardo, Mudekereza, Florian, Rutter, Tom, Thor, Nicolaj, Townsend, Wilbur, Zhang, Ruby, Bailey, Mike, Bar- a, Pablo, Bhole, Monica and Wernerfelt, Nils (2022). Social Capital I: Measurement ber´ and Associations with Economic Mobility. Nature, 608, 108–121. Cholli, Neil A and Durlauf, Steven N (2022). Intergenerational Mobility. In The Inequality Reader (eds D Grusky, N Dahir and C Daviss), NBER Working Paper Series, 5 edn. NBER Working Paper Series. Connolly, Marie and Haeck, Catherine (2019). Intergenerational Mobility Between and Within Canada and the United States. Journal of Labor Economics, 37, number S2, 595–641. Corak, Miles (2013). Income Inequality, Equality of Opportunity, and Intergenerational Mobility. Journal of Economic Perspectives, 27, number 3, 79–102. Corak, Miles (2020). The Canadian Geography of Intergenerational Income Mobility Cana- dian Intergenerational Income Mobility. The Economic Journal, 130, number 631, 2134– 2174. Corak, Miles, Lindquist, Matthew J. and Mazumder, Bhashkar (2014). A Comparison of Upward and Downward Intergenerational Mobility in Canada, Sweden and the United States. Labour Economics, 30, 185–200. es, Javier, D´ Cort´ errez, Pablo, Montecinos, Alexis, Troncoso, Pablo A. and ıaz, Juan D., Guti´ 33 Villarroel, Gabriel I. (2022). Decentralizing the Chilean miracle: regional intergenerational mobility in a developing country. Regional Studies, 57, number 5, 785–799. Dacuycuy, Lawrence B. (2018). Social Mobility in the Philippine Labor Market. The Philippine Review of Economics, 55, number 1 and 2, 1–40. Daza, Nancy A. (2021). Intergenerational Earnings Mobility in Mexico. Quantitative Social Science Working Paper, 21-10, 1–31. Deutscher, Nathan and Mazumder, Bhashkar (2020). Intergenerational mobility across Australia and the stability of regional estimates. Labour Economics, 66. ule, Paul (2024). Dodin, Majed, Findeisen, Sebastian, Henkel, Lukas, Sachs, Dominik and Sch¨ Social Mobility in Germany. Journal of Public Economics, 232, 105074. Durlauf, Steven and Seshadri, Ananth (2017). Understanding the Great Gatsby Curve. NBER Macroeconomics Annual, 32, number 1. Durlauf, Steven N., Kourtellos, Andros and Tan, Chih Ming (2022). The Great Gatsby Curve. Annual Review of Economics, 14, 571–605. Emran, M. Shahe, Greene, William and Shilpi, Forhad (2018). When Measure Matters: Coresidency, Truncation Bias, and Intergenerational Mobility in Developing Countries. Journal of Human Resources, 53, number 3, 579–607. Equalchances (2018). Equalchances: The World Database on Equality of Opportunity and Social Mobility (beta version May 2018). ¨ Ferreira, Francisco H.G., Lakner, Christoph, Lugo, Maria Ana and Ozler, Berk (2018). Inequality of Opportunity and Economic Growth: How Much Can Cross-Country Regres- sions Really Tell Us? Review of Income and Wealth, 64, number 4, 800–827. Fisher-Post, Matthew and Gethin, Amory (2023). Government Redistribution and Devel- opment Global Estimates of Tax-and-Transfer Progressivity, 1980-2019. mimeo. 34 Francesconi, Marco and Nicoletti, Cheti (2006). Intergenerational Mobility and Sample Election in Short Panels. Journal of Applied Econometrics, 21, number 8, 1265–1293. Genicot, Garance, Ray, Debraj and Concha-Arriagada, Carolina (2024). Upward Mobility in Developing Countries. NBER Working Paper, 32391. Grawe, Nathan (2001). Intergenerational Mobility in the U.S. and Abroad: Quantile and Mean Regression Measures. Ph.D. Thesis. University of Chicago. Grawe, Nathan D. (2006). Lifecycle Bias in Estimates of Intergenerational Earnings Persis- tence. Labour Economics, 13, number 5, 551–570. Greenaway-Mcgrevy, Ryan and So, Yun (2024). Temporal and Regional Variation in Inter- generational Income Mobility in New Zealand. New Zealand Economic Papers, 1–25. uell, Maia, Pellizzari, Michele, Pica, Giovanni and Rodr´ G¨ e V. (2018). Cor- ıguez Mora, Jos´ relating Social Mobility and Economic Outcomes. Economic Journal, 128, number 612, F353–F403. Haider, Steven and Solon, Gary (2006). Life-Cycle Variation in the Association between Current and Lifetime Earnings. American Economic Review, 96, number 4, 1308–1320. ıguez Mora, Jos´ Hassler, John, Rodr´ e V. and Zeira, Joseph (2007). Inequality and Mobility. Journal of Economic Growth, 12, number 3, 235–259. Helsø, Anne Line (2021). Intergenerational Income Mobility in Denmark and the United States*. The Scandinavian Journal of Economics, 123, number 2, 508–531. Hnatkovska, Viktoria, Lahiri, Amartya and Paul, Sourabh B. (2013). Breaking the Caste Barrier: Intergenerational Mobility in India. Journal of Human Resources, 48, number 2, 435–473. 35 Holter, Hans A. (2015). Accounting for Cross-country Differences in Intergenerational Earn- ings Persistence: The Impact of Taxation and Public Education Expenditure. Quantitative Economics, 6, 385–428. J¨ ¨ antti, Markus, Bratsberg, Bernt, Røed, Knut, Raaum, Oddbjørn, Osterbacka, Eva, orklund, Anders and Eriksson, Tor (2006). American Exceptionalism in a New Light: Bj¨ A Comparison of Intergenerational Earnings Mobility in the Nordic Countries, the United Kingdom and the United States. IZA Discussion Paper, number 1938, 1–40. antti, Markus and Jenkins, Stephen P (2015). Income Mobility. In Handbook of Income J¨ cois Bourguignon), volume 2, Distribution, Amsterdam (eds Anthony B Atkinson and Fran¸ pp. 807–935. North Holland. Javed, Sajid A. and Irfan, Mohammad (2014). Intergenerational Mobility: Evidence from Pakistan Panel Household Survey. The Pakistan Development Review, 53, number 2, 175–203. Jenkins, Stephen (1987). Snapshots versus Movies: ’Lifecycle Biases’ and the Estimation of lntergenerational Earnings Inheritance. European Economic Review, 31, 1149–1158. enez, Maribel (2011). ¿Movilidad o Persistencia Intergeneracional del Ingreso en la Jim´ on Emp´ Argentina? Una Aproximaci´ ıa Pol´ ırica. Revista de Econom´ ıtica de Buenos Aires, 9 y 10, number 5, 91–143. Kenedi, Gustave and Sirugue, Louis (2023). Intergenerational income mobility in France: A comparative and geographic analysis. Journal of Public Economics, 226. Krueger, Alan (2012). The rise and consequences of inequality in the United States. Speech, Cent. Am. Prog., Washington, DC. https://obamawhitehouse.archives.gov/sites/default/files/krueger cap speech fi- nal remarks.pdf. 36 Landersø, Rasmus and Heckman, James J. (2017). The Scandinavian Fantasy: The Sources of Intergenerational Mobility in Denmark and the US. The Scandinavian Journal of Eco- nomics, 119, number 1, 178–230. Lee, Chul-In and Solon, Gary (2009). Trends in Intergenerational Income Mobility. The Review of Economics and Statistics, 91, number 4, 766–772. Maoz, Yishay D. and Moav, Orner (1999). Intergenerational Mobility and the Process of Development. Economic Journal, 109, number 458, 677–697. ıguez, Juan G. (2013). Inequality of Opportunity and Growth. Marrero, Gustavo a. and Rodr´ Journal of Development Economics, 104, 107–122. Mayer, Susan E. and Lopoo, Leonard M. (2008). Government Spending and Intergenerational Mobility. Journal of Public Economics, 92, number 1-2, 139–158. Mazumder, Bhashkar (2005). Fortunate Sons: New Estimates of Intergenerational Mobility in the United States using Social Security Earnings Data. The Review of Economics and Statistics, 87, 235–255. Mazumder, Bhaskar (2015). Intergenerational Mobility: A Cross-National Comparison. In Emerging trends in the social and behavioral sciences: An interdisciplinary, searchable, and linkable resource (eds Robert Scott and Stephen Kosslyn). John Wiley & Sons. Mogstad, Magne and Torsvik, Gaute (2023). Family Background, Neighborhoods and Inter- generational Mobility. Handbook of the Economics of the Family, 1, number 1, 327–387. Montenegro, Claudio E. and Patrinos, Harry Anthony (2023). A data Set of Comparable Es- timates of the Private Rate of Return to Schooling in the World, 1970–2014. International Journal of Manpower, 44, number 6, 1248–1268. Munoz, Ercio (2024). The Geography of Intergenerational Mobility in Latin America and ıa LACEA Journal, 23, 333–354. the Caribbean. Econom´ 37 Narayan, Ambar, Van der Weide, Roy, Cojocaru, Alexandru, Lakner, Cristoph, Redaelli, Silvia, Gerszon Mahler, Daniel, Ramasubbaiah, Rakesh and Thewissen, Stefan (2018). Fair Progress?: Economic Mobility Across Generations Around the World. The World Bank. ofer, Guido (2019). Intergenerational Mobility and the Rise and Fall of Inequality: Neidh¨ Lessons from Latin America. Journal of Economic Inequality, 17, number 4, 499–520. ofer, Guido, Ciaschi, Mat´ Neidh¨ ın (2024). Social ıas, Gasparini, Leonardo and Serrano, Joaqu´ Mobility and Economic Development. Journal of Economic Growth, 29, 327–359. Ng, Irene Y.H. (2013). The Political Economy of Intergenerational Income Mobility in Singapore. International Journal of Social Welfare, 22, number 2, 207–218. Nybom, Martin and Stuhler, Jan (2016). Heterogeneous Income Profiles and Lifecycle Bias in Intergenerational Mobility Estimation. Journal of Human Resources, 51, number 1, 239–268. Nybom, Martin and Stuhler, Jan (2017). Biases in Standard Measures of Intergenerational Income Dependence. Journal of Human Resources, 52, number 3, 800–825. OECD (2018). A Broken Social Elevator? How to Promote Social Mobility. Osterberg, Torun (2000). Intergenerational Income Mobility in Sweden: What do Tax-data Show? Review of Income and Wealth, 46, number 4. Owen, Ann L and Weil, David N (1998). Intergenerational Earnings Mobility, Inequality and Growth. Journal of Monetary Economics, 41, number 1, 71–104. Peters, Elizabeth H. (1992). Patterns of Intergenerational Mobility in Income and Earnings. The Review of Economics and Statistics, 74, number 3, 456–466. Piraino, Patrizio (2015). Intergenerational Earnings Mobility and Equality of Opportunity in South Africa. World Development, 67, 396–405. 38 Ray, Debraj and Genicot, Garance (2023). Measuring Upward Mobility. American Economic Review, 113, number 11, 3044–3089. Rohenkohl, Bertha (2023). Intergenerational income mobility: New evidence from the UK. Journal of Economic Inequality, 21, number 4, 789–814. Solon, Gary (1992). Intergenerational Income Mobility in the United States. The American Economic Review, 82, number 3, 393–408. Solon, Gary (1999). Intergenerational mobility in the labor market. In Handbook of Labor Economics (eds Orley Ashenfelter and David Card), volume 3 PART, chapter 29, pp. 1761–1800. Elsevier B.V. Solon, Gary (2002). Cross-Country Differences in Intergenerational Earnings Mobility. Journal of Economic Perspectives, 16, number 3, 59–66. Solon, Gary (2004). A Model of Intergenerational Mobility Variation over Time and Place. In Generational Income Mobility in North America and Europe, chapter 2. Cambridge University Press. Solon, Gary (2014). Theoretical Models of Inequality Transmission across Multiple Gener- ations. Research in Social Stratification and Mobility, 35, 13–18. Stockhausen, Maximilian (2021). Like father, like son? A comparison of absolute and relative intergenerational labour income mobility in Germany and the US. Journal of Economic Inequality, 19, 667–683. Torche, Florencia (2014). Intergenerational Mobility and Inequality: The Latin American Case*. Annual Review of Sociology, 40, number 1, 619–642. van der Weide, Roy, Lakner, Christoph, Mahler, Daniel Gerszon, Narayan, Ambar and Gupta, Rakesh (2024). Intergenerational mobility around the world: A new database. Journal of Development Economics, 166, 103167. 39 van der Weide, Roy and Milanovic, Branko (2018). Inequality is Bad for Growth of the Poor (But Not for that of the Rich). The World Bank Economic Review, 32, number 3, 507–530. Wagner, A (1890). Finanzwissenchaft. Winter, C. F., Leipzig. Waldkirch, Andreas, Ng, Serena and Cox, Donald (2004). Intergenerational Linkages in Consumption Behavior. Journal of Human Resources, 39, 355–381. Yan, Weibo and Deng, Xiaolan (2022). Intergenerational Income Mobility and Transmission Channels in a Transition Economy: Evidence from China. Economics of Transition and Institutional Change, 30, number 1, 183–207. 40 Appendices Table A1 details the list of countries and the surveys used for our estimates. Table A2 details the sources and method of estimation for IGE estimates drawn from previ- ous literature. Table A3 provide details about specification choices used in our TSTSLS estimation. Figure A1 plots indicators of mobility proposed by Ray and Genicot (2023) estimated using WID data to estimates using the same sample that the authors used to estimate IGE. Figure A2 plots the indicator of relative mobility proposed by Ray and Genicot (2023) against growth rates at different percentiles of the income distribution. Mobility indicator is computed in Ray and Genicot (2023) while the growth rates are estimated by the authors using the same underlying data. Figure A3 plots the indicator of absolute mobility proposed by Ray and Genicot (2023) against growth rates at different percentiles of the income distribution, both estimated using the same sample that the authors use to estimate the IGE. Figure A4 plots the indicator of relative mobility proposed by Ray and Genicot (2023) against growth rates at different percentiles of the income distribution, both estimated using the same sample that the authors use to estimate the IGE. Figure A5 plots IGE against the indicators of mobility proposed by Ray and Genicot (2023) computed using data from WID. Figure A6 plots the indicator of relative mobility proposed by Ray and Genicot (2023) against the Gini index and GDP with RG indicators of mobility estimated using the same sample as the authors to estimate the IGE. Figure A7 plots the intergenerational income elasticity against the indicator of relative mo- bility proposed by Ray and Genicot (2023), both estimated using the same sample that the authors used to estimate the IGE. 41 Table A1: List of surveys used Economy Newer survey Year Older survey Year AUS HILDA 2015 HILDA 2001 AUT EU-SILC 2019 EU-SILC 2004 BEL EU-SILC 2019 EU-SILC 2004 BGR EU-SILC 2011 EU-SILC 2007 BIH LITS 2016 LSMS 2001 BOL EH 2008 EIH 1992 BRA PNAD 2014 PNAD 1996 CAN CGSS 2014 CGSS 1999 CHE EU-SILC 2011 EU-SILC 2007 CHL CASEN 2013 CASEN 1998 CHN CFPS 2012 CHIP 2002 COD E123 2012 E123 2004 COL ENCV 2013 ECH 2001 CYP EU-SILC 2023 EU-SILC 2005 CZE EU-SILC 2019 EU-SILC 2005 DEU SOEP 2017 SOEP 2002 DNK EU-SILC 2019 EU-SILC 2004 ECU ECV 2013 ECV 1999 EGY ELMPS 2012 ELMPS 1998 ESP EU-SILC 2011 EU-SILC 2004 EST EU-SILC 2019 EU-SILC 2004 ETH LSMS-ISA 2013 HICES 2000 FIN EU-SILC 2019 EU-SILC 2004 FRA EU-SILC 2019 EU-SILC 2004 GBR EU-SILC 2011 EU-SILC 2005 GHA GLSS 2012 GLSS-IV 1998 GRC EU-SILC 2019 EU-SILC 2004 GTM ENCOVI 2014 ENCOVI 2000 HRV LITS 2016 HBS 2004 HUN EU-SILC 2011 EU-SILC 2005 IDN IFLS 2014 IFLS 2000 IND IHDS 2011 NSS50-SCH10 1993 42 Table A1 continued from previous page Economy Newer survey Year Older survey Year IRL EU-SILC 2019 EU-SILC 2004 ISL EU-SILC 2019 EU-SILC 2004 ITA EU-SILC 2019 EU-SILC 2004 JOR JLMPS 2010 HIES 2002 JPN JGSS 2012 JGSS 2000 KAZ LITS 2016 HBS 2001 KEN STEP 2013 ILFS 1998 KOR KLIPS 2014 KLIPS 1998 LBR HIES 2014 CWIQ 2007 LKA STEP 2012 HIES 2002 LTU EU-SILC 2019 EU-SILC 2005 LUX EU-SILC 2019 EU-SILC 2004 LVA EU-SILC 2019 EU-SILC 2005 MAR ENNVM 2006 ENNVM 1991 MDA LITS 2016 HBS 2001 MDG ENEMPSI 2012 EPM 1999 MEX EMOVI 2011 ENEU 1991 MLT EU-SILC 2011 EU-SILC 2007 MNE LITS 2016 HHS 2004 MNG LITS 2016 HIES 2002 MWI LSMS-ISA 2013 IHS-I 1997 NER LSMS-ISA 2014 LFS 2002 NGA LSMS-ISA 2012 GHS 1997 NOR EU-SILC 2023 EU-SILC 2004 PAN ENV 2008 EH 1995 PER ENAHO 2014 ENAHO 1999 POL EU-SILC 2019 EU-SILC 2005 PRT EU-SILC 2019 EU-SILC 2004 ROU EU-SILC 2023 EU-SILC 2007 RUS RLMS 2011 RLMS 1996 SRB LITS 2016 HBS 2003 SVK EU-SILC 2019 EU-SILC 2005 SVN EU-SILC 2019 EU-SILC 2005 43 Table A1 continued from previous page Economy Newer survey Year Older survey Year SWE EU-SILC 2019 EU-SILC 2004 TGO QUIBB 2015 QUIBB 2006 TJK LITS 2016 LSMS 1999 TUN TLMPS 2014 ENE 2000 TUR LITS 2016 HICES 2002 TWN TSCS 2015 TSCS 2000 TZA LSMS-ISA 2012 HBS 2000 UGA LSMS-ISA 2014 UNHS 1999 USA PSID 2015 PSID 1999 UZB LITS 2016 HBS 2000 VNM STEP 2012 VHLSS 2002 XKX LITS 2016 HBS 2003 ZAF NIDS 2012 PSLSD 1993 44 Table A2: List of sources from the literature Economy IGE Year Source Method ARG 0.642 2006 Jimenez (2011) TSTSLS MYS 0.537 1989 Grawe (2001) TSTSLS NLD 0.304 2010 Equalchances (2018) TSTSLS NZL 0.296 2015 Greenaway-Mcgrevy and So (2024) OLS NPL 0.436 1995 Grawe (2001) TSTSLS (one cross-section split by age) PAK 0.438 2010 Javed and Irfan (2014) Coresident children using IV PHL 0.432 2006 Dacuycuy (2018) Coresident children using OLS SGP 0.260 2010 Ng (2013) Interval regression with IV URY 0.429 2013 Araya (2019) TSTSLS 45 Table A3: Specification and data details Code O W Age P Income Notes AUS N Y 30-54 Y gross earnings AUT Y Y 30-54 Y gross earnings BEL Y Y 30-54 Y gross earnings BGR Y Y 30-54 Y gross earnings BIH N Y 30-54 Y earnings BOL N Y 30-54 Y earnings BRA N Y 30-54 Y earnings CAN N Y 30-54 Y earnings CHE Y Y 30-54 N gross earnings CHL N Y 30-54 Y earnings 46 CHN N Y 30-54 N average earnings in 5 years; earnings Income per capita in old survey for rural sector COD N Y 30-54 Y earnings Excluded values lower than 2.5 in older survey COL N Y 30-54 Y earnings; gross earnings CYP Y Y 30-54 Y gross earnings CZE Y Y 30-54 Y gross earnings DEU Y Y 25-60 N net earnings DNK Y Y 25-60 N gross earnings ECU N Y 30-54 Y earnings; gross earnings EGY N Y 30-54 Y earnings ESP Y Y 25-60 N net earnings EST Y Y 30-54 Y gross earnings ETH N Y 25-60 Y household income per worker; earnings Income per worker in old survey. Table A3 continued from previous page Code O W Age P Income Notes FIN Y Y 30-54 N gross earnings FRA Y Y 30-54 Y gross earnings GBR Y Y 30-54 Y gross earnings GHA N Y 30-54 Y earnings GRC Y Y 30-54 Y net earnings GTM N Y 30-54 Y earnings HRV N Y 25-60 Y earnings HUN Y Y 30-54 Y gross earnings IDN N Y 30-54 Y net earnings IND N Y 30-54 Y earnings 47 IRL Y Y 25-60 Y gross earnings ISL Y Y 30-54 Y gross earnings ITA Y Y 30-54 Y net earnings JOR N Y 30-54 Y earnings JPN N N 30-54 N earnings KAZ N N 30-54 Y earnings KEN N Y 25-60 Y gross earnings; earnings KOR N Y 30-54 Y earnings LBR N N 30-54 N household income per worker; earnings Income per worker in old survey LKA N N 30-54 N earnings; net earnings LTU Y Y 30-54 Y gross earnings LUX Y Y 30-54 Y gross earnings LVA Y Y 30-54 Y net earnings Table A3 continued from previous page Code O W Age P Income Notes MAR N Y 30-54 Y earnings MDA N N 30-54 N earnings MDG N Y 30-54 Y earnings MEX Y Y 25-60 N earnings MLT Y Y 30-54 Y gross earnings MNE N Y 25-60 Y earnings MNG N Y 30-54 Y earnings MWI N Y 30-54 Y earnings NER N Y 30-54 Y earnings NGA N Y 30-54 N earnings 48 NOR Y Y 30-54 Y gross earnings PAN N Y 30-54 Y earnings; gross earnings PER N Y 30-54 Y earnings POL Y Y 30-54 Y gross earnings PRT Y Y 30-54 Y net earnings ROU Y Y 30-54 Y gross earnings RUS N Y 30-54 N earnings SRB N N 30-54 N earnings SVK Y Y 30-54 Y gross earnings SVN Y Y 30-54 Y gross earnings SWE Y Y 30-54 Y gross earnings TGO N Y 25-60 Y earnings TJK N N 25-60 Y earnings Table A3 continued from previous page Code O W Age P Income Notes TUN N Y 25-60 Y earnings TUR N N 30-54 Y net earnings; earnings TWN N Y 30-54 Y earnings TZA N Y 30-54 Y earnings UGA N Y 30-54 Y earnings USA Y Y 30-54 Y earnings UZB N Y 30-54 N earnings VNM N Y 30-54 Y earnings XKX N N 25-60 Y net earnings; earnings ZAF Y Y 30-54 N gross earnings 49 Notes: This table provides information about data and specification choices that vary between countries. Y refers to Yes and N refers to No. Columns “O” details whether the estimates use occupation to predict parental income. Columns “W” details whether the estimates use survey weights. Columns “P” details whether the estimates only use paid employees in the estimation sample. Figure A1: RG’s indicators using our data versus WID 15 4 Absolute mobility (Ray Genicot, 2024) - Authors data 3 HUN VNM Relative mobility (RG) - Authors data ECU 10 Correlation coef.: 0.398 JOR 2 Correlation coef.: 0.057 VNM UGA GHA 1 BRA PER MAR POL CHE NGA ECU DEU AUS GRC KEN COL USA MEX 0 SWECHE ITA CHL NGA TUN FIN JPN 5 TZA POL IDN CAN FRA TUR CHN EGY MDG DNK LUX GBR CHL BGR IDN -1 ZAF IND AUT IRL BEL TUR BRA IND PER EGY UGA COL MNG AUS USA -2 ZAF HUN FIN IRL SWE ETH LKA FRA GHA 0 LUX ITA DNK DEU AUT MAR TUN JOR BGR CAN MDG BEL JPN KEN NER -3 MEX MNG GRC LKA GBR -4 TZA NER CHN ETH -5 ESP -5 ESP -5 0 5 10 -5 -4 -3 -2 -1 0 1 2 3 4 Absolute mobility (Ray Genicot, 2024) Relative mobility (RG) - Ray and Genicot (2023)'s data (a) Absolute mobility (b) Relative mobility Notes: This figure plots indicators of mobility proposed by Ray and Genicot (2023) estimated using WID data to estimates using the same sample that the authors used to estimate IGE. The graph includes 47 countries. 50 Figure A2: Growth rates versus RG’s indicators 8 8 8 Correlation coef.: 0.626 Correlation coef.: -0.167 Correlation coef.: -0.482 6 6 6 Annualized growth rate (9th decile) Annualized growth rate (1st decile) Annualized growth rate (Average) 4 4 4 2 2 2 0 0 0 -2 -2 -2 -4 -4 -4 -4 -2 0 2 4 6 8 -4 -2 0 2 4 6 8 -4 -2 0 2 4 6 8 Relative mobility (RG) Relative mobility (RG) Relative mobility (RG) Notes: This figure plots the indicator of relative mobility proposed by Ray and Genicot (2023) against growth rates at different percentiles of the income distribution. The graph includes 75 countries. The indicator of mobility is computed in Ray and Genicot (2023) while the growth rates are estimated by the authors using the same underlying data. 51 Figure A3: Growth rates versus RG’s (absolute) indicators using authors’ data 15 15 15 10 Correlation coef.: 0.958 10 Correlation coef.: 0.897 10 Correlation coef.: 0.842 Annualized growth rate (9th decile) Annualized growth rate (1st decile) Annualized growth rate (Average) 5 5 5 0 0 0 -5 -5 -5 -10 -10 -10 -15 -15 -15 -15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15 Absolute mobility (RG) Absolute mobility (RG) Absolute mobility (RG) Notes: This figure plots the indicator of absolute mobility proposed by Ray and Genicot (2023) against growth rates at different percentiles of the income distribution. The graph includes 73 countries. RG indicators and the growth rates are estimated using the same sample as the authors to estimate the IGE. 52 Figure A4: Growth rates versus RG’s (relative) indicators using authors’ data 15 15 15 Correlation coef.: 0.264 Correlation coef.: -0.318 Correlation coef.: -0.317 10 10 10 Annualized growth rate (9th decile) Annualized growth rate (1st decile) Annualized growth rate (Average) 5 5 5 0 0 0 -5 -5 -5 -10 -10 -10 -15 -15 -15 -15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15 Relative mobility (RG) Relative mobility (RG) Relative mobility (RG) Notes: This figure plots the indicator of relative mobility proposed by Ray and Genicot (2023) against growth rates at different percentiles of the income distribution. The graph includes 73 countries. RG indicators and the growth rates are estimated using the same sample as the authors to estimate the IGE. 53 Figure A5: Intergenerational income elasticity versus RG’s indicators 1 1 Correlation coef.: -0.091 Correlation coef.: 0.040 FIN SWE SWE FIN Intergenerational Mobility (1-IGE) Intergenerational Mobility (1-IGE) .8 CAN DNK .8 DNK CAN AUS CHE LKA AUS CHE LKA TUR TUR JPN JPN GRC AUT JOR JOR AUT GRC BEL POL BEL POL DEU DEU .6 FRA KEN VNM .6 FRA KEN VNM ITA IRL ITA IRL LUX HUN GBR MNG HUN LUX MNG GBR ESP UGA ESP MAR IDN CHN CHN IDN MAR UGA BRA GHA BRA GHA CHL CHL .4 USA ECU TZA .4 USA TZA ECU BGR ETH ETH BGR ZAF PER NER ZAF PER NER NGA COL NGA COL MEX MEX .2 IND .2 IND TUN TUN EGY EGY MDG MDG 0 0 -5 0 5 10 -5 -4 -3 -2 -1 0 1 2 3 4 Absolute mobility (RG) Relative mobility (RG) (a) Absolute mobility (b) Relative mobility Notes: This figure plots the intergenerational income elasticity against the indicators of mobility proposed by Ray and Genicot (2023). The graph includes 47 countries. The indicators of mobility are computed in Ray and Genicot (2023) using data from WID. 54 Figure A6: Relative mobility vs national income and inequality 4 4 Correlation coef.: -0.137 COD HUN COD HUN ECU ECU 2 2 Correlation coef.: 0.190 VNM VNM GTM UGA UGA GTM Relative Mobility (RG) Relative Mobility (RG) BRA PER BIH BIH PER BRA BOL PRTMAR POL SRB MARBOL SRB POL PRT ROU ROU SVK SVK COLMEX KEN GRC HRV ISLDEU COL HRV GRC AUS ISL 0 NGA TUN USA CHE AUS KORTWN XKX NORSWE CZE 0 KEN NGA TUNXKX MEX KOR CZE TWN DEU USA SWE CHE NOR PAN MWI CHL ITA EST MWI PAN CHL ITA LVA JPN LTU MNE FIN LVALTU EST MNE FIN IDN FRA CANCYP SVN IDN SVN JPN CYPFRA EGY EGY CAN MDG GBR IRL LUX DNK MDG GBR LUX ZAF IND IND ZAF IRLDNK KAZ AUT KAZ AUT TUR BEL TUR BEL MNG MLT MNG MLT -2 -2 GHA MDA BGR GHA MDA BGR JOR JOR LKA LKA -4 NER TZA CHN -4 NER TZACHN ETH ETH ESP ESP -6 -6 .4 .5 .6 .7 .8 500 1000 2000 4000 8000 16000 32000 64000 Equality (1-Gini coefficient) GDP per capita (2021 USD PPP, logarithmic scale) (a) Income inequality (b) Gross domestic product Notes: This figure plots the indicator of relative mobility proposed by Ray and Genicot (2023) against the Gini index and GDP. The graph includes 73 countries. RG indicators of mobility are estimated using the same sample as the authors to estimate the IGE. Income inequality is measured with the Gini index obtained from World Bank’s Poverty and Inequality Platform (PIP) supplemented with data from “All the Ginis” database. GDP data is from the World Development Indicators supplemented with data from the Maddison project where necessary. We match our mobility estimates with GDP and inequality data from 20 years earlier, using the year in which income was collected in the newer survey as our reference point. For example, for an estimate that uses income for children collected in 2015, we match with data on inequality from 1995. 55 Figure A7: Intergenerational elasticity of income vs. RG’s indicators 1 Correlation coef.: -0.021 FIN SWE NOR SRB Intergenerational Mobility (1-IGE) .8 DNK CAN KOR MDA SVN LKA AUS CHE TUR ISL JPN JOR KAZ AUT GRC BEL EST HRV POL TWN .6 LTU DEU KEN PRT VNM FRA IRL ITA LUX MWI SVK MNG GBR LVA CZE BOL HUN ESP CHN IDN MAR UGA MNE BIH BRA GHA ROU CYP CHL .4 TZA PAN USA ECU ETH BGR MLT NER ZAF PER COL NGA XKX MEX COD .2 IND TUN EGY GTM MDG 0 -5 -4 -3 -2 -1 0 1 2 3 4 Relative mobility (RG) Notes: This figure plots the intergenerational income elasticity against the indicator of relative mobility proposed by Ray and Genicot (2023). The graph includes 73 countries. RG indicators of mobility are estimated using the same sample that the authors used to estimate the IGE. 56