The World Bank Poverty and Equity Kristina Vaughan and Monica Robayo-Abril Inequality of opportunity in Bulgaria Policy Note Abstract: Using data from a recent EU-SILC module on intergenerational mobility, this policy note explores to what extent the high levels of inequality prevalent in Bulgaria are due to inequality of opportunity, that is, inequality due to circumstances beyond an individual’s control . The results show that over half of the income inequality in Bulgaria is due to inequality of opportunity, the highest share in the European Union. Using Shapley's decomposition techniques, we find that disparities in parents’ education explain most of the inequality of opportunity. Related to this, Bulgaria is one of the countries in Europe where children’s education is strongly linked to that of their parents, resulting in little educational intergenerational mobility. Policies to tackle high levels of inequality of opportunity mainly center around reforming the educational system as gaps in educational attainment are evident as early as the early childhood level, and schools have typically been inequality reinforcing. Greater access to quality early childhood education, delayed tracking of students until they have acquired foundational skills, and improved access to quality tertiary institutions are key to reducing inequality of opportunity. Complementary analysis suggests there are sizable growth gains to be made in Bulgaria if human capital attainment is improved and reducing inequality of opportunity presents one such pathway. Inequality of Opportunity in Bulgaria1 1. Introduction Income inequality in a society is considered acceptable from an equity standpoint insofar as differences in income result from differing levels of effort, more commonly referred to as inequality of effort (Roemer 1993; Van de Gaer 1993). However, when circumstances at birth either directly or indirectly determine income, more commonly referred to as inequality of opportunity (Roemer 1993; Van de Gaer 1993), this is viewed as a violation of core equity principles. Several studies have analyzed the degree to which inequality of opportunity is prevalent across countries, primarily focusing on education, the labor market, and income. Most recently, Palmisano, Biagi, and Peragine (2021) analyzed inequality of opportunity in tertiary education in 31 countries in Europe and found big geographical differences, with Nordic countries having lower rates of inequality of opportunity than Mediterranean and Eastern countries. Across all countries, parental education and occupation were the strongest determinants of inequality of opportunity. Abras et al. (2012) examined inequality of opportunity in the labor market across countries in Europe and Central Asia. They found substantial heterogeneity in inequality of opportunity and its determinants, though they cite a strong role for parents' education and gender. Finally, Marrero and Rodríguez (2011) examined the impact of inequality of opportunity on disposable income using data from 23 European countries. They found that Nordic, Western European, and some of the more prosperous Eastern European Union (EU) countries report low inequality of opportunity. In contrast, the Mediterranean, Atlantic, and poorer Eastern EU countries report high inequality of opportunity. While the strand of the literature that quantifies inequality of opportunity and its determinants is relatively well developed, complementary literature empirically identifying the macroeconomic impacts of inequality of opportunity is substantially less explored. This is mostly due to demanding data requirements, econometric complexities, and the difficulty in disentangling inequality of opportunity from inequality of effort. In one of the few papers that attempt this, Marrero and Rodríguez (2013) evaluate the impact of inequality of opportunity on growth, disentangling inequality of opportunity from inequality of effort. The authors use data from the Panel Study of Income Dynamics in the United States and find a robust, strong negative relationship between inequality of opportunity and economic growth. Specifically, a one standard deviation reduction in inequality of opportunity increases decade growth by about 1.24 percentage points and steady-state income by about US$274 per person. Disparities in health outcomes, years of schooling, and quality of education by parental background in Bulgaria provide suggestive evidence of inequality of opportunity along these 1 This analysis was prepared as part of the Poverty and Equity Program for Bulgaria in the Poverty and Equity Global Practice of the World Bank.. The analysis was conducted by a team including Kristina Vaughan (Economist), Santiago Garriga (Consultant), and Monica Robayo- Abril (Senior Economist, World Bank). The authors are grateful for comments received from Tom Bundervoet, Marc Schiffbauer and Desislava Nikolova. 1 dimensions. Given the importance of these outcomes to human capital formation and, subsequently, of human capital formation to growth, it stands to reason that there is potential growth to be unlocked if Bulgaria manages to reduce inequality of opportunity by addressing the key factors behind it. While the impact of reducing inequality of opportunity on health outcomes, years of schooling, and educational quality is challenging to quantify, the Country Economic Memorandum Long-Term Growth Model projections provide estimates of improving these outcomes on growth. However, it is unlikely that reducing or eliminating inequality of opportunity would lead to as large a magnitude of the effects presented here. The results suggest that improving expected years of schooling from 12.3 years to 13.6 years, or that of the 75th percentile in Europe and Central Asia or that of the median of Baltic countries, would yield yearly average human capital growth as high as 0.16 percentage points during 2022–2050 (0.23 percentage points during 2035–2050), translating into an increase in average yearly growth of 0.10 percentage points during 2022–2050 (0.15 percentage points during 2035–2050). Similarly, improving educational quality (proxied by harmonized test scores), which is assumed to affect learning immediately but have less muted effects on older children, from a baseline of 0.71 to 0.87 (the level of Estonia), is expected to increase the yearly average human capital growth by 0.36 percentage points during 2022–2050 (0.52 percentage points during 2035–2050), translating into an increase of 0.23 percentage points in yearly average growth (0.34 percentage points during 2035–2050). Though not explored directly in this paper, improving health outcomes through increasing the adult survival rate and the 'not stunted' rate from a baseline of 0.87 and 0.93 to 0.93 and 1.0, respectively, or that of the 75th percentile of Europe and Central Asia, is estimated to improve average human capital growth by 0.02 percentage points during 2022 –2050 (0.03 percentage points during 2035–2050). The combined impact of improving years of schooling, education quality, and health under the reforms mentioned above suggests that the impact on annual human capital growth could be as high as 0.63 percentage points during 2022–2050 (0.91 percentage points during 2035–2050). The combined impact translates into improvements in gross domestic product (GDP) growth of 0.4 percentage points annually during 2022–2050 (0.6 percentage points during 2035–2050). The analysis undertaken in this policy note complements the existing literature by studying inequality of opportunity in education, the labor market, and income in Bulgaria. Bulgaria has consistently recorded the highest level of income inequality within the EU-27 and has seen a substantial widening in inequality over time. One of the primary purposes of this note is to explore the degree to which high levels of income inequality in Bulgaria can be explained by high levels of inequality of opportunity and assess how this compares to that of other EU-27 countries. Using decomposition techniques, we also identify the most salient determinants of inequality of opportunity in Bulgaria and contrast the relative importance of these factors across other EU-27 countries. Finally, we discuss potential policies that can help alleviate inequality of opportunity and identify potential channels through which this can unlock unrealized growth. 2 2. Inequality of opportunity in Bulgaria Over the past decade, Bulgaria has consistently recorded one of the highest rates of income inequality in the EU, standing at 37.2 in 2022 compared with 29.6 for the EU-27 average (Figure 1). Furthermore, income inequality, as measured by the per adult equivalent Gini coefficient, has generally been widening over time—increasing from 33.2 in 2011 to 37.2 in 2022 (Figure 1). Related to this, Bulgaria consistently reports one of the highest poverty rates in the EU-27, standing at 20.6 percent in 2022 compared with 16.2 percent for the EU-27. Figure 1: Per adult equivalent Gini Index 2011–2021 for Bulgaria and the EU-27 43.0 41.0 39.0 37.0 37.2 Gini Index 35.0 33.0 31.0 29.0 29.6 27.0 25.0 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 Bulgaria EU-27 Source: [ilc_di12], Eurostat. Estimates refer to income reference year, not survey year. A methodological change (in sampling) in 2016 may affect the comparability of the temporal series before and after The persistently high and widening inequality in Bulgaria can be largely attributed to disparities in labor income. A comparison of the concentration coefficient of various sources of income with the Gini coefficient suggests that labor income is the main factor contributing to a high Gini coefficient (Figure 2), with labor income more unevenly distributed than total income among Bulgarians.2 Similarly, a Shapley decomposition of factors affecting inequality over time also points to labor income as the primary source of widening inequality with labor income responsible for the majority of the 4.7 percentage point increase in the Gini coefficient over the period 2010-2020. (Figure 3). 2 Concentration coefficients are compared to the Gini coefficient. Negative values of the concentration coefficient suggest that a higher share of that component goes to those in the lower part of the distribution and thus may be reducing inequality. Values of the concentration coefficient which are positive, but below that of the Gini coefficient, suggest that the majority of this component goes to richer portions of the distribution but to a lesser degree than total income. Concentration coefficients higher than the Gini coefficient are an indication that the component is more unequally distributed than total income. 3 Figure 2: Concentration coefficients by Figure 3: Contribution to per adult equivalent Gini changes sources of income and Gini coefficient, between 2010 and 2020 as a share of the total change 2020 Pension Income Gini (total income) 41.8 Other Income Other Income Soc. Assist. Income Labor income Pension Income Share Employed Soc. Assist. Income Share of Adults Ad.Eq. Ratio Labor income 51.5 Tax Ratio -20 0 20 40 60 -2 -1 0 1 2 Source: Author's computations based on the EU-Statistics of Income and Living Conditions (SILC). Note: Labor income includes wages, noncash wages, and self-employment income; Social assistance income includes sickness benefits, disability benefits, education allowance, housing allowance, child allowance, social allowance; Pension income includes private pensions, old age benefits, survivor benefits; Other income includes unemployment benefit, income from capital, household transfers, income received by people ages under 16. Given the role labor income plays in persistent and widening inequality, we consider individual labor income as our main outcome of interest but consider disparities in access to the labor market and tertiary education as other outcomes of interest. We limit our sample to individuals ages 25–55 to lessen the impact of deviations from lifetime income common at the early stages of an individual's career and preretirement. We also exclude self-employment income from the sample due to the difficulties with accurately recording this type of income, most notably underreporting, an inability to separate personal finances from business finances in reporting income, and high items of nonresponse. We do not anticipate this will change the results significantly, given that the share of self-employment income in total (monetary) income of households was just 6 percent on average for 2012–2021. The main dataset for this analysis is the 2019 EU-SILC fielded across the EU-27 countries capturing the income year 2018. In addition to information on income, demographics, labor market, and poverty and social exclusion indicators, an ad hoc module on intergenerational transmission of disadvantages was fielded in 2019.3 The next module with this information was collected in 2023, but microdata was not available yet at the time of this study. Respondents were asked about their circumstances growing up, including the household's 3 https://ec.europa.eu/eurostat/web/income-and-living-conditions/database/modules Though ideally we would have liked to be able to speak to changes in inequality of opportunity of time, the intergenerational module is only fielded periodically and suffers from issues of comparability across modules over time. The next module of intergenerational transmission of advantages and disadvantages was fielded in 2023 and microdata is still not available. 4 financial situation; parent's presence, education, and employment; and degree of urbanization, among other indicators. We supplement the information from the 2019 EU-SILC with data on ethnicity from the Bulgarian National Statistics Institute. We follow a similar approach used in the European Bank for Reconstruction and Development (EBRD) transition report 2016–2017 to compute inequality of opportunity. 4 As a first step, we compute the Gini coefficient of the existing labor income distribution to get a baseline measure of labor income inequality. We then regress labor income on a series of variables aimed at capturing circumstances either determined at birth or when the individual was 14 years. We include the following variables as our measure of circumstances: respondents' gender; self-reported ethnicity (ethnic majority); whether they grew up in a dual-parent household; the education status of their most educated parent (tertiary educated relative to secondary educated and primary educated relative to secondary educated); the self-reported financial situation of the household (moderately bad/bad/very bad, moderately good/good/very good); and the degree of urbanization of the place of residence (urban versus rural). We then predict labor income based solely on these circumstance variables and recompute the Gini coefficient based on this new labor income distribution. This new Gini is our measure of inequality of opportunity, that is, the income inequality that would prevail if income was solely determined by circumstances at birth or when the respondent was 14 years. The remaining inequality is assumed to be attributed to effort, preferences, and choices. Relative inequality of opportunity (RIOO)—the proportion of income inequality that is due to individual circumstances—is then computed as the ratio of inequality of opportunity to total income inequality. It should be noted that the reliability of these estimates is largely subject to the availability of data capturing these circumstances and assumptions surrounding the correlation between included variables and the error term in the regression. It is also likely that some factors that capture circumstances, such as the quality of parents' social networks or the quality of neighborhoods children grow up in, have been excluded from the regression as these are not observed in our dataset. The estimates presented in this paper should thus be interpreted as a lower bound. 4 The full report can be found at https://www.ebrd.com/documents/oce/pdf-transition-report-201617-inequality-of-opportunity.pdf. 5 Table 1: Regression of labor income on circumstance variables Regressor Coefficient Male 1,640*** (254) Ethnic majority 697 (439) Dual-parent household 611 (544) Highest educated parent - primary −1,535*** (367) Highest educated parent - tertiary 4,373*** (342) Urban 1,568*** (280) Financial situation - moderately good 610 (440) Financial situation - good/very good 1,999*** (462) Constant 2,573*** (729) N 4,076 R2 0.114 Note: *** p<0.01, ** p<0.05, * p<0.1. Excluded categories are financial situation - moderately bad/bad/very bad and highest educated parent - secondary. Standard errors are in parentheses. Results from the first-stage regression (Table 1) suggest that being male, having the highest educated parent with tertiary education relative to secondary education, growing up in an urban as opposed to rural area, and growing up in a household with the financial situation described as good/very good relative to moderately bad/bad/very bad, all have a positive and statistically significant impact on labor income. Conversely, having a parent with primary education as opposed to secondary education has a negative and statistically significant impact on labor income. We repeat the analysis by urban/rural and younger (25 – 40) and older (41–55) cohorts to allow for geographic differences and structural changes that have occurred over time. For the sake of brevity, we do not display the first-stage regression results here. Computation of the RIOO suggests that Bulgaria has an RIOO of 52 percent, or put differently, 52 percent of the inequality that prevails in Bulgaria is due to inequality due to circumstances beyond an individual's control. RIOO is higher in urban areas, potentially due to more diversity in the circumstances, but similar across older and younger cohorts. 6 To determine the relative impact of each of these factors on overall inequality of opportunity, we perform a Shapley decomposition on the R-squared of the first-stage regression, which is assumed to capture the portion of the variation in income that can be predicted by the circumstance variables. The relative importance of each of the factors is displayed in Figure 4. Figure 4: RIOO Shapley decomposition of R-squared of income regression on circumstances by national, urban-rural, young-old cohort 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 National Rural Urban Young (25-40) Old (41-55) Gender Ethnicity Presence of parents Parent's education Urban/ Rural Financial situation of the household RIOO Source: Author's computations based on the EU-SILC. At the national level, results suggest that parental education is the most important determinant of inequality of opportunity accounting for 55 percent of the total. The impact of parental education is more pronounced in urban areas at 61 percent, almost double that of rural areas at 31 percent, potentially due to more variation in parental education in urban areas. Parental education is more critical for older cohorts at 59 percent than younger cohorts at 51 percent. The degree of urbanization growing up is the second most important factor at the national level, accounting for 16 percent of the inequality of opportunity. It is relatively more important in rural areas (25 percent) than in urban areas (7 percent) and for older cohorts (20 percent) than younger cohorts (11 percent) in determining income, potentially due to an improvement in internal labor mobility across cohorts. The household's financial situation growing up accounts for 15 percent of inequality of opportunity at the national level, more significant for urban areas and younger cohorts. Being male is a strong determinant of inequality of opportunity in rural areas, accounting for just over a fifth compared with 10 percent in urban areas. By contrast, ethnicity and the presence of both parents growing up have relatively low importance in explaining inequality of opportunity except in rural areas where ethnicity accounts for 13 percent of inequality of opportunity, potentially due to more variation in ethnicity in rural areas. We repeat the above analysis for the EU-27 countries, excluding the ethnicity variable. Comparisons across other countries in the EU-27 suggest that in 2018, Bulgaria had the highest level of 7 RIOO in income in the EU-27 at 52.8 percent, more than double that of Sweden, the country with the lowest RIOO at 21.1 percent. From these results, Bulgaria emerges as one of the countries where parental education matters strongly for inequality of opportunity (58.8 percent), along with Malta (67.8 percent), Luxembourg (56.4 percent), and Romania (49.8 percent). This starkly contrasts to Sweden, where parental education contributes to only 13.6 percent of total inequality of opportunity. Figure 5: RIOO across EU-27 countries, 2018 60 52.8 50 40 30 21.1 20 10 0 CYP DNK LUX AUT IRL HUN MLT BGR CZE NLD SVK FRA GRC LVA ESP EST LTU ROU SWE BEL FIN HRV ITA POL PRT Source: Author's computations based on the EU-SILC. Figure 6: Shapley Decomposition of RIOO across EU-27 countries, 2018 100 90 80 70 60 50 40 30 20 10 0 CYP DNK AUT HUN IRL LUX MLT CZE NLD BGR SVK GRC LTU LVA ESP EST FRA BEL HRV ITA POL ROU SWE FIN PRT Gender Presence of parents Parent's education Urban/Rural Financial situation Source: Author's computations based on the EU-SILC. 8 Even before earnings, there may be inequalities of opportunity in access to the labor market. To account for this possibility, we examine the impact of inequality of opportunity оn employment. Our approach here is closely related to that of Abras et al. (2012), who examined the impact of inequality of opportunity in getting a job with subjectively desirable characteristics and having suffered no event of economic distress or shock related to employment in the past year.5 Given that the outcomes are binary, we require a different framework to assess inequality of opportunity. We use the Dissimilarity Index (D- index) within the Human Opportunity Index (HOI) framework to determine inequality in employment, abstracting away from issues with selection into labor force participation. The D-index measures whether opportunities, in this case, labor force outcomes, are allocated equitably, comparing different circumstance groups' probabilities of access to the various labor market variables of interest (World Bank 2012). The D-index ranges from 0 to 1 (or 0 to 100). A D-index equal to 0 implies that access to an opportunity is the same among the general population irrespective of one's circumstances. In contrast, a D-index of 1 (or 100) indicates that a group is completely excluded from access. The D-index is also commonly interpreted as the share of opportunities that would have to be 'reallocated' across different groups so that all groups would have equal access. We first estimate a logit model of the probability where we regress access to an opportunity, here labor market outcome variables, on the individual circumstance variables defined earlier. We then obtain the predicted probabilities of the logit model for each individual based solely on his or her circumstances and use these to estimate the D-index. As before, we perform a Shapley decomposition to determine the marginal contribution of each circumstance variable to inequalities in access.6 At the national level, Bulgaria recorded a relatively low D-index of employment of 3.6. The D- index is higher in urban areas and for younger cohorts, indicating greater employment disparities along these dimensions depending on one's circumstances. Much of this disparity is explained by parents' education (36 percent) followed by ethnicity (30 percent). In rural areas, ethnicity accounts for the largest share of the D-index at 35.3 percent, followed by parents' education. Interestingly, parents' education matters less for younger cohorts (34.8 percent) than for older cohorts (40.1 percent) though the share is still high. Conversely, urbanization matters more for younger cohorts (17.6 percent) than for older cohorts (11.7 percent). 5 The full paper can be found at https://izajold.springeropen.com/articles/10.1186/2193-9020-2-7. 6Further details on the approach can be found at http://www1.worldbank.org/poverty/visualizeinequality/Files/Documentation/HOI- Stata.pdf. 9 Figure 7: D-index and Shapley decomposition of D-index of logit of employment on circumstances by national, urban- rural, young-old cohort 100 5 80 4 Share of D-index D-index 60 3 40 2 20 1 0 0 National Rural Urban Young (25-40) Old (41-55) Gender Ethnicity Presence of parents Parent's education Urban/ Rural Financial situation of the household D-index Source: Author's computations based on the EU-SILC We again repeat the analysis across all the EU-27 countries for which data are available and exclude the ethnicity variable. Bulgaria has the highest D-index in the EU-27 countries, indicating that it records the most significant disparity in access to employment due to differences in individual circumstances, though still small. Bulgaria is one of the countries reporting the highest contribution of parental education to the D-index at 51.9 percent, surpassed only by Czechia (53.8 percent) and Sweden (51.9 percent), indicating the relatively high importance of parent's education compared with other EU- 27 countries. The geographic location of growing up explains a further 22.2 percent of the D-index in employment, surpassing the EU-27 average of 12.4 percent. In contrast to other countries, gender explains relatively less about the differences in employment across circumstance groups, accounting for only 0.5 percent compared with 18.1 percent for the EU-27 average. 10 Figure 8: D-index of employment across EU-27 countries, 2018 4.0 3.5 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.3 0.0 CYP DNK AUT HUN IRL LUX MLT BGR CZE NLD SVK GRC LVA ESP EST FRA LTU BEL HRV POL ROU ITA SWE FIN PRT Source: Author's computations based on the EU-SILC. Figure 9: Shapley decomposition of D-index of logit of employment on circumstances 100 90 80 70 60 50 40 30 20 10 0 CYP DNK AUT HUN LUX IRL MLT BGR NLD GRC LVA CZE ESP LTU ROU SVK EST FRA BEL FIN ITA SWE HRV POL PRT Gender Presence of parents Parent's education Urban/Rural Financial situation Source: Author's computations based on the EU-SILC. Given the strong link between tertiary education and the probability of employment and earnings, we examine the inequality of opportunity in tertiary education attainment and discuss intergenerational persistence in educational attainment. As before, we use the D-index within the HOI framework to determine inequality of opportunity in tertiary education attainment and use a Shapley decomposition to determine the relative contribution of each circumstance to differences in tertiary education attainment within Bulgaria. 11 Bulgaria recorded a relatively high D-index of 31.8, indicating that there are large disparities in tertiary education attainment based on individual circumstances. The D-index is higher in urban (40.9) than rural areas (24.3), indicating that tertiary educational attainment is much more unevenly distributed in urban than rural areas. Older cohorts reported a slightly higher D-index compared with younger cohorts. Figure 10: D-index and Shapley decomposition of D-index of logit of tertiary education on circumstances by national, urban-rural, cohort 100 45 40 80 35 Share of D-Index 30 D-Index 60 25 20 40 15 20 10 5 0 0 National Rural Urban Young (25-40) Old (41-55) Gender Ethnicity Presence of parents Parent's education Urban/ Rural Financial situation of the household D-index Source: Author's computations based on the EU-SILC. Bulgaria has the third highest D-index of tertiary education within the EU-27, indicating that tertiary educational attainment is relatively highly unevenly distributed within the population compared with other countries. Interestingly, while parental education is important in determining the differences in educational attainment within Bulgaria, accounting for 45 percent of the differences in tertiary attainment, this is relatively low within the EU-27, where the average is 54 percent. The geographic location of growing up is also of relatively high importance within Bulgaria accounting for 23 percent of the inequality in tertiary education attainment, exceeding the EU-27 average of 14 percent. Last, growing up in relatively well-off household accounts for 18.6 percent of the variation, above the EU-27 average of 16.3 percent. 12 Figure 11: D-index of tertiary education across EU-27 countries, 2018 40 35 31.7 30 25 20 15 10 5 0 CYP DNK AUT LUX HUN IRL MLT BGR CZE NLD SVK GRC LVA ESP EST FRA LTU BEL HRV ITA POL ROU SWE FIN PRT Source: Author's computations based on the EU-SILC. Figure 12: Shapley decomposition of D-index of logit of tertiary education on circumstances 100 90 80 70 60 50 40 30 20 10 0 DNK LUX CYP AUT HUN IRL MLT BGR CZE EST LTU SVK FRA GRC LVA NLD BEL ESP SWE FIN HRV PRT ROU ITA POL Gender Presence of parents Parent's education Urban/Rural Financial situation Source: Author's computations based on the EU-SILC. Since educational attainment matters so strongly for later life outcomes in Bulgaria, it is useful to examine the extent to which an individual's educational attainment—and not just tertiary education—depends on that of their parents. Relative intergenerational mobility of education (RIME) is defined as the extent to which an individual's position in the education distribution is independent of 13 (or dependent on) the position of their parents and commonly measured as (one minus) the correlation between an individual's years of schooling and the years of schooling of his or her most educated parent. Here we rely on data from the 9th round of the European Social Survey (ESS) data conducted in 2019 by the European Research Infrastructure. The ESS contains more detailed educational classifications compared with the EU-SILC for both the respondent and the self-reported respondent's parental education status, allowing for more accurate estimates of RIME. To preserve comparability, we restrict our ESS sample to those ages 25–55. RIME is closely related to the concept of RIOO—generally, societies with high levels of RIME tend to have lower RIOO as children's mobility is less dependent on their parental characteristics, including education. This pattern is evident when considering the correlation between RIOO and RIME for the EU-27 countries for which data are presented in Figure 13. Figure 13: Scatterplot of RIOO and RIME 90 80 DNK LVA 70 FIN NLD SWE EST LTU 60 ESP IRL CZE FRA POL AUT HRV 50 HUN BEL ITA RIGME SVK CYP PRT BGR 40 30 20 10 0 20 25 30 35 40 45 50 55 60 RIOO Source: Author's computation based on the 2019 ESS and the 2019 EU-SILC. Note: Individuals ages 25–55. Data on Greece, Romania, Germany, Luxembourg, and Slovenia were not available. Looking across the EU-27 countries, we see that Bulgaria has the highest level of intergenerational persistence of education, or put differently, the lowest level of RIME, with the education of children highly correlated with the education of their most educated parent. Alternate measures of RIME, such as transition probabilities across different points in the education distribution, suggest a similarly low RIME—a child whose most educated parent has educational attainment in the bottom quartile of the education distribution has only a 9 percent chance of having educational attainment in the top quartile of the education distribution compared with a 63 percent chance of having educational attainment in the lowest quartile of the education distribution. 14 Figure 14: Correlation between children and parent's years of schooling 0.6 0.55 0.5 0.4 0.3 0.2 0.1 0.0 DNK AUT CYP DEU HUN IRL NLD LVA ESP BGR LTU CZE SVN SVK EST SWE FRA FIN BEL ITA POL HRV PRT Source: Author's computation based on the 2019 ESS. Figure 15: Probability of a child with a parent's educational attainment in the bottom quartile of the educational distribution would transition to the lowest and highest quartile 80 70 63.0 60 50 40 30 20 9.4 10 0 DNK CYP HUN AUT DEU IRL NLD LTU BGR SVN SVK FRA LVA ESP EST CZE FIN SWE PRT ITA POL HRV BEL Bottom Quartile-->Bottom Quartile Bottom Quartile-->Top Quartile Source: Author's computation based on the 2019 ESS. Throughout this analysis, a common theme has emerged—individuals' later life outcomes are strongly dependent on their parents' educational attainment and, to a lesser extent, the degree of urbanization of the place they grew up. The strong dependence of later life outcomes on parental education is partly due to the high correlation between the educational attainment of parents and that of their children. The strong dependence on childhood geographic location likely reflects geographic differences in the accessibility and quality of opportunities and institutions. Such dependence on individuals' outcomes on childhood geographic location and the outcomes of their parents results in an intergenerational persistence of poverty and inequality that is challenging to reverse. Indeed, Bulgaria shows strong signs of intergenerational transmission of disadvantages—poverty rates among individuals 15 in our sample whose most educated parent has primary education stand at 36.9 percent compared with 9.2 percent for secondary education and 3.5 percent for tertiary education. Similarly, for individuals who grew up in an urban area, poverty rates are 14.8 percent compared with 16.7 percent for those who grew up in rural areas. Until access to and quality of institutions, especially the educational system and the labor market, is such that they can compensate for the intergenerational transmission of disadvantages, Bulgaria will continue to struggle with high rates of poverty and inequality. Prior complementary analysis done by the World Bank suggests that gaps in educational attainment by parental background and region emerge early in life and that inequalities in access to early childhood and care institutions help perpetuate these gaps (World Bank 2021). This is reflected in substantially lower rates of early childhood and care enrollment among Roma children and children in rural areas (World Bank 2021). The vast literature linking early childhood cognitive and socioemotional skills to later life outcomes (Cunha and Heckman 2007) and complementary literature indicating that it is difficult to acquire skills later in life (Carneiro and Heckman 2003; Cunha and Heckman 2008) suggest that these early disadvantages present in Bulgaria are hard to reverse and perpetuate later in life. Furthermore, instead of secondary schools contributing to reducing the preexisting inequalities and often perpetuated from the primary and early childhood level, the structure of the Bulgarian secondary school system exacerbates these inequalities. At the secondary level, gaps in learning among ethnic minority students, students from rural areas, and students with parents who have lower educational attainment are reflected in lower reading, math, and science Programme for International Student Assessment (PISA) test scores (Figure 16). 7 School value added is significantly higher for Bulgarian-speaking students (compared to Romani or Turkish-speaking students), for students whose parents have higher levels of educational attainment (bachelor's or post-graduate degrees), or students with employed parents (as opposed to unemployed or inactive parents) reinforcing the premise that schools are inequality-compounding institutions (World Bank 2012). 7PISA is an international assessment that tests the skills and knowledge of 15-year-old students in reading, mathematics, and science. A total of 79 countries and economies took part in the 2018 assessment. 16 Figure 16: Reading, math, and science PISA test scores by ethnicity, parent's education, and location 500 460 420 380 340 Other Post secondary, Large City Bulgarian Short-cycle tertiary Town Primary City Bachelors, Masters Village Small Town Lower Secondary Upper Secondary non-tertiary or Doctorate Ethnicity Parent's education Urban/Rural Reading Math Science Source: Organisation for Economic Co-operation and Development (OECD) PISA database, https://pisadataexplorer.oecd.org/ide/idepisa/report.aspx. Note: Scores are from 2018. Ethnic majority/minority is proxied by the main language spoken at home. The poor performance among disadvantaged students is partly a function of a system that groups low-performing students into vocational schools at an early age, which hinders the development of crucial math and reading skills necessary to secure higher-paying jobs in the workforce.8 This is reflected in lower PISA scores in reading, math, and science for vocational schools compared with general schools (Figure 17). Such early segregation has resulted in Bulgaria being the country in the EU with one of the largest indexes of school social segregation (Gortazar, Kutner, and Herrera-Sosa 2014; World Bank 2018), meaning that students with similar backgrounds tend to go together to secondary education more than in other countries, indicated by the abovementioned study (Figure 18). In fact, in Bulgaria, the degree of segregation is so strong that the type of peers matters more for students' performances than their own background characteristics (World Bank 2012). Many Bulgarians do not complete secondary education, as evidenced by Bulgaria's high early school leave rates of 12.8 percent in 2020, surpassing the EU-27 average of 9.9 percent.9 Early school leaving tends to be higher among students in rural areas (25.5 percent) compared with those in cities (5.6 percent) and towns (12.7 percent), damaging employment prospects and rendering participation in higher education unattainable.10 Disparities in access to tertiary education are then perpetuated in the labor market through employment, unemployment, and wages disparities. In 2020, the employment rate among those with 8In Bulgaria, academic tracking takes place at the end of grade 7 based on the selection of students after a national high-stakes test, generating divergent educational paths from primary education that have a lifetime impact on students from a socioeconomic perspective (World Bank 2012). 9 Eurostat [edat_lfse_30]. 10 Eurostat [edat_lfse_14]. 17 primary education was only 48.1 percent compared with 74.0 percent for those with secondary education and 87.6 percent for those with tertiary education. Unemployment rates show similar disparities—in 2020, the unemployment rate among those with primary education was 14 percent compared with 4.6 percent for those with secondary education and 2.5 percent among those with tertiary education.11 Large wage disparities are also evident—in 2018, employed tertiary-educated individuals reported labor income on average 3.7 times higher than those with primary education. Individuals in rural areas, as well as those on average less educated, face the additional complexity of less favorable employment opportunities in rural areas, representing an additional source of inequality of opportunity. In 2019, the employment rate among individuals ages 20–64 was 78.2 percent among individuals in urban areas compared with 66.1 percent among those in rural areas. Regional disparities in unemployment rates were similarly present—the unemployment rate among individuals living in urban areas was 3.0 percent compared with 7.6 percent among those living in rural areas.12 Figure 17: Reading, Math and Science test scores by Figure 18: School Segregation Index, 2012 school type 500 459 465 458 Bulgaria 0.63 450 Hungary 406 379 389 Romania 400 Germany 350 Czech Republic Slovakia 300 Turkey Poland 250 Slovenia 200 Greece USA 150 Lithuania 100 Croatia Latvia 50 Russia Serbia 0 Estonia Reading Math Science Great Britain Denmark General Vocational Canada Montenegro Sweden Norway Finland 0.30 0.40 0.50 0.60 Source: OECD PISA database, Source: World Bank 2012. https://pisadataexplorer.oecd.org/ide/idepisa/report.aspx . Note: Scores are from 2018. 11 [lfs_urgaed] Eurostat. 12 Author’s computations based on 2019 EU-LFS. 18 3. Concluding Remarks To address systemic inequality of opportunity in Bulgaria, policies should be put in place to tackle inequalities at each point they occur. Unfortunately, at each step of the education system, culminating in the labor market, opportunities are missed to reduce preexisting inequalities, and the structure of institutions often exacerbates these inequalities. High user fees and other access costs remain a significant burden to participation in early childhood and preschool education among children from disadvantaged backgrounds (World Bank 2021). The recent policy decision to remove user fees for kindergartens effective April 2022 is likely to assist in increasing participation, particularly among disadvantaged students, though user fees are only part of the cost parents face. Greater financial support to parents of children from disadvantaged backgrounds and rural areas so that they can ensure clothing and appropriate shoes for their kids, especially during the cold season, remains a necessary precursor to improving access. Additionally, further steps to improve the quality of early childhood education need to be partnered with increased access to make meaningful gains in reducing inequality. At the secondary level, delaying tracking until children are older can allow students the time to develop crucial math and reading skills to boost future long-term employability. Early warning systems supported by regular measuring and monitoring of educational outcomes can be used to identify children at risk of being left behind or dropping out. Students identified as at risk through these systems can then be placed in remedial and re-engagement programs. Such programs could include free extracurricular and summer activities to help boost engagement and minimize summer learning losses.13 In addition, financial incentives to attract and retain high-quality teachers in rural and disadvantaged areas can help improve the educational outcomes of these students. Last, modernizing the curricula to reflect the evolving nature of education and the demands of the labor force can help boost the employment prospects of future graduates. Furthermore, continuing the thrust to reform the vocational education and training (VET) curriculum to ensure students graduate with skills relevant to and in demand by the labor market and facilitating better partnerships with potential employers through work-based learning schemes can go some way toward improving the wage prospects of those who graduate from the vocational track. Remedial classes to address deficiencies in foundational skills acquisition brought forward from the secondary level can also help boost labor market outcomes. Increasing the quality of the secondary and upper secondary institutions and delaying early tracking will go some way toward improving the ability of individuals from disadvantaged backgrounds to participate in tertiary education. Increasing access to forms of financial support, introducing pathways from vocational schools to universities, introducing flexible options for partaking in 13Summer learning losses refer to students from disadvantaged backgrounds losing ground academically to their more well-off peers as they are less likely to be engaged in stimulating activities over the summer. 19 tertiary education, and expanding the number of occupationally oriented tertiary colleges are other steps that can be taken to improve future job opportunities and labor market outcomes among students from disadvantaged backgrounds. Encouraging continued adult learning can help reengage parents and young adults who may have become disenfranchised by the formal education system and left with low levels of education. Bulgaria does not have a strong culture of lifelong learning, with one of the lowest participation rates (1.6 percent versus 9.2 percent in 2020), and efforts to improve this would likely require strong financial incentives and targeted campaigns.14 Additionally, since some of these adults may be engaged in the labor market, programs should be sufficiently flexible to allow for their continued participation in the labor market. Boosting connectivity in rural areas can go some way toward reducing inequality of opportunity due to geography. Improving the availability and quality of transport between urban and rural areas by rehabilitating and expanding the rail and road networks can accelerate the development of rural areas, promote greater labor mobility, and widen the range of employment prospects for rural workers. For digital infrastructure, there is a need to accelerate the deployment of broadband access by building backbone infrastructure and last-mile connectivity in underserved areas with currently only 1 percent of rural households benefiting from high-capacity network technology, well below the EU average of 24 percent (World Bank 2021). Improvements in digital infrastructure can facilitate greater access to digital jobs, thereby overcoming the lack of access to labor market opportunities in rural areas. Finally, within the labor market itself, encouraging and facilitating the upskilling and reskilling of workers through more significant expenditure and more relevant active labor market policies will help address some of the employment and wage disparities by educational attainment. Future job projections on job openings suggest that individuals with low levels of educational attainment are at risk of getting left further behind in the labor market, with only 4 percent of new job openings suitable for them compared with 48 percent each for those with medium and high education qualifications (Cedefop 2020). Now, with the COVID-19 pandemic and the accelerated push toward digitization along with necessary structural adjustments as part of the European Green Deal, individuals with low educational attainment should take appropriate steps to reskill and upskill. Bulgaria's commitment to improve the digital skills capacity of the workforce as part of its Digital Bulgaria 2025 comes as welcome news amidst this challenge;15 however, Bulgaria still faces substantial constraints to improving its digital skills. It has one of the lowest levels of spending on active labor market policies16 relative to the EU-27 average (0.6 percent versus 1.7 percent of GDP in 2019) and one of the lowest level of digital skills acquisitions (29 percent versus 56 percent for the EU-27).nica Robayo (mrobayo@worldbank.org) and Stefanie Brodma 14 Eurostat [sdg_04_60]. 15 https://digital-skills-jobs.europa.eu/en/actions/national-initiatives/national-strategies/bulgaria-digital-bulgaria-2025-national- programme. 16 [lmp_expsumm] European Commission - Directorate-General for Employment, Social Affairs and Inclusion (DG EMPL). 20 Annex 1: The Human Opportunity Index (HOI) The Human Opportunity Index (HOI) is a measure of the coverage rate of an opportunity, discounted by inequality in its distribution across circumstances groups. An opportunity is defined as access to a good or service, which is accepted as universal by society. A circumstance is defined as individual, household, geographic characteristics outside individual's control, for example, gender, ethnicity, parental education, wealth, geographic location. A circumstance group is a set of individuals with the same set of circumstances, for example, all ethnic Bulgarian males with tertiary highest parental education growing up in rural areas in a dual-parent household with a moderately good financial situation.17 The formal computation of the HOI is as follows: = ( − ) × where D is the Dissimilarity Index C is the coverage rate or the percent of individuals that have access to an opportunity. The Dissimilarity Index (D-index) measures whether existing opportunities (access to services) are allocated equitably, comparing different circumstance groups' probabilities of accessing a given opportunity. The D-index ranges from 0 to 1. A D-index equal to 0 implies that access to an opportunity is the same among the general population irrespective of one's particular situation (for example, whether urban or rural dweller or whether male or female). In contrast, a D-index of 1 indicates that a group is completely excluded from access. The D-index is interpreted as the share of opportunities that would have to be 'reallocated' across different groups of children so that all groups would have equal access.18 The Dissimilarity Index (D) is computed as follows: = ∑ | − | = where C is the coverage rate m is the total number of circumstance groups αk is the share of group k in the total population Ck is the coverage rate of circumstance group k. 17 http://www1.worldbank.org/poverty/visualizeinequality/Files/Documentation/HOI-Methodology.pdf. 18 https://www.worldbank.org/en/topic/poverty/lac-equity-lab1/equality-of-opportunities/hoi-d-index. 21 References Abras, Ana, Alejandro Hoyos, Ambar Narayan, and Sailesh Tiwari. 2012. "Inequality of Opportunities in the Labor Market: Evidence from Life in Transition Surveys in Europe and Central Asia." Background Paper for the World Development Report 2013, World Bank, Washington, DC. https://openknowledge.worldbank.org/handle/10986/1213. Carneiro, P., and J. J. Heckman. 2003. “Human Capital Policy.” In Inequality in America: What Role for Human Capital Policies? edited by James J. Heckman, Alan B. Krueger, and Benjamin M. Friedman. Cambridge: MIT Press. Cedefop. 2020. "Skills Forecast 2020: Bulgaria." Cunha, F., and J. Heckman. 2007. "The Technology of Skill Formation." American Economic Review 97 (2): 31–47. Cunha, F., and J. J. Heckman. 2008. "Formulating, Identifying and Estimating the Technology of Cognitive and Noncognitive Skill Formation." Journal of Human Resources 43 (4): 738–782. Gortazar, L., D. Kutner, and K. Herrera-Sosa. 2014. "How can Bulgaria Improve its Education System?" Washington, DC: World Bank. Marrero, G., and J. G. Rodríguez. 2011. "Inequality of Opportunity in Europe." Review of Income and Wealth 58 (4): 597–621. Marrero, G. A., and J. G. Rodríguez. 2013. "Inequality of Opportunity and Growth." Journal of Development Economics 104: 107–122. Palmisano, F., F. Biagi, and V. Peragine. 2021. “Inequality of Opportunity in Tertiary Education: Evidence from Europe.” Res High Educ 63: 514–565. Roemer, J. E. 1993. "Pragmatic Approach to Responsibility for the Egalitarian Planner." Philosophy & Public Affairs 20: 146–166. Van de Gaer, V. 1993. Equality of Opportunity and Investment in Human Capital. Catholic University of Leuven, Faculty of Economics. World Bank. 2009. Bulgaria: Improving the Quality and Relevance of Education for All. Washington, DC: World Bank. https://openknowledge.worldbank.org/handle/10986/12542. World Bank. 2012. How Can Bulgaria Improve Its Education System? An Analysis of PISA 2012 and Past Results . Washington, DC: World Bank. https://openknowledge.worldbank.org/handle/10986/20776. World Bank. 2018. Improving the Quality of Bulgaria's Education System: A Review of Bulgaria's Performance on the 2015 Programme for International Student Assessment (PISA). World Bank, Washington, D.C World Bank. 2021. Bulgaria Systematic Country Diagnostic. Washington, DC: World Bank. https://openknowledge.worldbank.org/handle/10986/36842. 22