Inequality of Opportunity in Income and Education: Evidence from Central and Eastern Europe

Abstract

Inequality of opportunity is a critical issue that significantly impacts the socioeconomic landscape. Understanding the variations in income and educational attainment has become increasingly important, as these disparities are often shaped by social determinants such as individual effort and circumstances, which can affect educational outcomes and income potential throughout life. The aim of this paper is to quantify the inequality of opportunity and to examine how circumstances beyond an individual’s control influence income and level of education in the Central and Eastern European countries. We draw on recent data from the fourth wave of the Life in Transition Survey and employ inequality of opportunity indices, Shapley and Oaxaca decompositions, and econometric models to capture the structure and magnitude of the effects. Our findings reveal that inequality of opportunity in income is mainly due to gender and, to a smaller extent, parental education, while educational attainment is mainly influenced by parental education, books at home and the mother’s occupational sector. The results provide a robust foundation for supporting targeted policies in education, employment, and pay equity, and they indicate the need to tailor strategies to the specific contexts of each Central and Eastern European country.

1. Introduction

Inequalities are a relatively simple concept to grasp, yet particularly complex in terms of the many forms they take. Social inequality refers to unequal access to and use of resources across various domains—such as health, education, and employment—that lead to disparities based on gender, race or ethnicity, age, and social class. It is closely linked to the concept of social stratification.

The persistence and deepening of inequalities may jeopardise achieving internationally agreed-upon objectives (United Nations—Department of Economic and Social Affairs, 2020). Over the past 10 to 15 years, inequalities have become increasingly sensitive, important, and challenging. “Leave no one behind” has become the rallying cry in the global fight against inequality, promoted through the 2030 Agenda. As early as 2012, Stiglitz warned that the levels of inequality reached in recent decades are morally unacceptable, undermining both economic stability and the political climate (Stiglitz, 2012).

Inequality has consequences that extend beyond economic development; it is detrimental to society as a whole and poses threats to peace and security (Brinkman et al., 2013). It is important to recognise that resentment over injustice, unequal access to public goods or social services, and political or social exclusion can lead to unrest, hostility and violence. Discrimination based on race, ethnicity, language, or religion and restrictions on legal, political, or social rights and freedoms are all root causes of inequality.

In recent years, the concept of inequality of opportunity has received growing attention in the literature. A distinction is thus made between inequalities that stem from individual effort and those resulting from circumstances beyond an individual’s control (Brunori et al., 2013). Inequality of opportunity in income distribution is a critical area of research that investigates how individual income disparities arise not only from personal effort but also from circumstances outside individual control, such as family background, education, and geographical location (Checchi & Peragine, 2010). This concept underscores the ethical dimensions linked to economic disparities and emphasises the need for effective policy interventions to reduce such inequities.

One effective strategy for combating social and economic inequality is education (Hofmarcher, 2021). However, to achieve this, educational systems must give all students an equal opportunity regardless of their circumstances. That means an individual’s academic performance should be determined by their abilities and effort rather than by outside influences like their social or economic background (Marrero et al., 2024). Thus, inequality of opportunities in education can be both an impediment to success at a personal level, as education influences labour market outcomes, and to economic development at a macro-level, as studies show (Marrero & Rodríguez, 2013; Marconi, 2018).

Our analysis focuses on Central and Eastern European (CEE) countries that are members of the European Union. This region experienced relatively equal income distributions during the socialist period but experienced a growth in inequality after transitioning to market economies. Initially, this increase was generated by rapid privatisation and deregulation, while in recent years it has been accentuated by technological progress, digitalisation, and globalisation, as well as relatively weak income redistribution policies (Kuzmar & Piatek, 2024). These structural changes have influenced income inequality and access to education, changing the patterns of disparity arising from the inherited circumstances and individual effort (Brzezinski & Salach, 2022).

This paper explores inequality of opportunity (IOp) using recent data from the fourth wave of the Life in Transition Survey (LiTS). We investigate how income and educational attainment are influenced by circumstances beyond the individual’s control, in a transgenerational context. We aim to measure the contribution of circumstances to observed inequalities and to identify the key influencing factors of opportunity in the CEE countries. We use a mix of methods: IOp indices, Shapley, Oaxaca decomposition, and econometric models in order to capture both the structure and magnitude of these effects.

Our analysis is structured around a set of research hypotheses that reflect both theoretical expectations and empirical patterns observed in previous studies:

H1. 

A considerable part of income inequality can be traced back to circumstances such as family background, gender, or place of birth.

H2. 

Even when individual effort is considered, circumstances continue to play a strong role in shaping income.

H3. 

Educational outcomes are strongly tied to early-life conditions, with individuals from less educated families being more likely to attain lower levels of education.

The originality of this study derives from its focus on the countries of Central and Eastern Europe, a region underexplored in the literature on inequality of opportunities but marked by deep socio-economic transformations. We use recent data from the 2022–2023 wave of the Life in Transition Survey. Our approach integrates two key dimensions of inequality—income and education—providing a broad view of how circumstances shape socio-economic outcomes. Building on this regional and dimensional focus, our methodological approach strengthens the contribution of the study. We use Shapley and Oaxaca decompositions to estimate the contributions of different circumstances to inequality, with the latter allowing meaningful comparisons across countries. In addition, we apply regression techniques suited to each outcome: cross-sectional OLS for income inequality and multinomial logit for education. By combining these complementary methods with recent survey data, our approach offers a more complete and nuanced understanding of inequality of opportunity than focusing on a single dimension, country, or method.

The remainder of the paper is structured as follows: Section 2 reviews the relevant literature on inequality of opportunity in income and education. Section 3 describes the data and methodology, while Section 4 presents the results, covering both income and education inequality of opportunity and robustness and sensitivity analyses. Finally, Section 5 concludes with a discussion of the main findings and their policy implications.

2. Literature Review

According to Roemer (1998), the determinants of a person’s advantage (i.e., desired outcomes such as income or occupational status) should be separated into circumstances and efforts. Circumstances are factors exogenous to the individual, such as gender, race, family environment or place of birth, education or occupational status of the parents (Marrero & Rodríguez, 2012; Niehues & Peichl, 2014). These factors can affect a person’s outcomes but are not themselves influenced by those outcomes (Brunori et al., 2013).

Efforts, on the other hand, include factors that contribute to individual outcomes and that can be influenced by personal decisions and actions, such as the level of education attained or participation in the labour market (Checchi & Peragine, 2010). However, Roemer and Trannoy (2016) argue that it is necessary to distinguish between effort influenced by circumstances (accountable effort) and the effort itself (raw effort).

Inequality can therefore be broken down into two components: inequality of opportunity and inequality of effort. While the latter is considered acceptable at the societal level, the former is regarded as unfair and should be addressed (Ferreira & Gignoux, 2011; Brunori et al., 2013; Roemer & Trannoy, 2016; Villar, 2017).

Other authors have similarly defined exogenous factors that influence individuals’ opportunities: circumstances are the initial conditions that can shape an individual’s trajectory yet lie beyond their control (Bussolo et al., 2023). Furthermore, Lefranc et al. (2006) argue that if an individual could choose their circumstances before knowing how much effort would later be required, they would naturally opt for the circumstances offering the best set of opportunities.

Inequality of opportunity occurs when individuals’ initial circumstances influence their outcomes. Thus, a fair society is defined not by equal outcomes but by equality of initial opportunities (Hassine, 2012).

Another way to analyse inequalities of opportunity is by examining intergenerational income mobility (Piraino, 2015), based on the idea that high intergenerational persistence of income signals inequality of opportunities in the labour market. In cross-country studies, the degree of intergenerational mobility is often estimated using the elasticity between a son’s income (or educational level) and that of his father, the percentage difference in income between the child’s generation and the parents’. Many studies focus on the father-son relationship to avoid complications arising from the historical view of women’s participation in the labour market (Corak, 2016).

Regarding measurement methods, Roemer (1993) argues that equality of opportunity exists when all individuals who exert the same degree of effort receive the same reward. He suggests calculating, for each group of individuals defined by the same set of circumstances, the minimum outcome achieved by those exerting the same level of effort—that is, those belonging to the same group. Building on this perspective, two main approaches to measuring inequalities of opportunity have emerged: the ex post and the ex ante approaches. In the ex post approach, equality of opportunity is achieved when all individuals exerting the same effort obtain the same result. In the ex ante approach, equality of opportunity is achieved when all individuals have access to the same set of opportunities, regardless of their personal circumstances (see Checchi & Peragine, 2010; Marrero & Rodríguez, 2012).

Van de Gaer and Ramos (2020) further distinguish between two measurement methods: the direct and the indirect approach. In the direct method, inequality of opportunity is measured as inequality in a counterfactual distribution where all differences are attributed to circumstances, entirely removing the effect of effort. In the indirect method, observed inequality is compared with the inequality that would exist if all individuals had identical circumstances. Counterfactual distributions can be constructed using either parametric (linear or log-linear regressions) or non-parametric (group averages) methods.

Across Europe, numerous studies have addressed the issue of inequalities of opportunity. For example, Marrero and Rodríguez (2012) apply a parametric approach based on the method proposed by Ferreira and Gignoux (2011), which involves estimating a linear regression between individuals’ disposable income and a broad set of circumstance variables, and then quantifying the inequality component between groups—representing the share of total inequality explained by circumstances (an ex ante approach and parametric estimation). They use data from the 2005 EU-SILC for 23 European countries and adopt the Theil index as a measure of inequality.

In the circumstances, the authors include parental education and occupation, nationality and perceived economic conditions in childhood. As expected, the results are mixed. The Nordic countries (Denmark, Finland, Norway and Sweden), the continental Western Europe (Germany, the Netherlands, Austria, Belgium and France) and some of the richer Eastern European countries (Slovakia, the Czech Republic, Slovenia and Hungary) form the group with low levels of inequality of opportunity. The group with high inequality of opportunity consists of the Mediterranean countries (Italy, Greece and Spain), the Atlantic countries (Portugal, Ireland, and the United Kingdom) and the poorer Eastern European countries (Estonia, Latvia, Poland and Lithuania). Similar findings are reported by Roemer and Trannoy (2016): in Denmark, the income distributions by parental education are much closer than in Hungary, indicating a higher degree of equality of opportunity.

A similar approach was taken by Andreoli and Fusco (2017), who analysed the evolution of inequality of opportunity in several European countries between 2005 and 2011. Using EU-SILC data from 2005 and 2011 for 19 European countries, they measure both ex post and ex ante inequality of opportunity. The main circumstance variable is the father’s education, and individuals’ gross annual income is expressed in purchasing power parity (PPS). The results are consistent with those mentioned above: the Nordic countries and Germany have the lowest levels, while the CEE countries display higher levels.

Andreoli and Fusco (2017) found that between 2005 and 2011, changes in inequality of opportunity were small and statistically insignificant, with a few exceptions—such as Austria, where inequality increased, and Finland, where it decreased. Similar results, based on an ex ante analysis, were obtained by Brzezinski (2019): inequality of opportunity increased significantly between 2004 and 2010 in Austria, Greece and Spain, but decreased in Portugal and Poland.

Another multi-country analysis was conducted by Bussolo et al. (2023), who use harmonised microdata from four national sources—SHIW (for Italy), SOEP (for Germany), HBS (for France) and BHPS/UKHLS (for the United Kingdom)—all integrated through the Luxembourg Income Study (LIS). The dataset covers the period 1978–2014, including more than 600,000 observations on individuals aged between 25 and 80. Circumstances include gender, age, region of birth, immigration status, and parental education (converted into years of study). The results indicate a general trend of decreasing inequality of opportunity in Germany, France, and the United Kingdom, but stagnation—or even a slight increase—in Italy, in line with the aforementioned studies. An interesting observation is that in all the countries analysed, gender differences have become less and less relevant in explaining income. The authors also suggest that the observed decline in inequality of opportunity may reflect a growing influence of unobservable factors, such as individual abilities.

At the European level, several studies have analysed the inequality of opportunity within individual countries. One example is the work of Checchi and Peragine (2010) on Italy. They used data from the Survey on Income and Wealth (SHIW) conducted by the Bank of Italy for 1993, 1995, 1998 and 2000, focusing only on individuals earning income from salaried work, to avoid distortions caused by the self-reported earnings of the self-employed. The main circumstance considered is the parents’ education level, with the analysis broken down by geographical regions (North vs. Centre-South) and gender (men vs. women), to capture the variations in opportunities to access the labour market.

Their findings indicate that inequality of opportunity accounts for around 19.5% of total income inequality in Italy according to the ex post approach, and 14.8% according to the ex ante method. In addition, they found that southern regions of Italy face a higher incidence of inequality of opportunity, especially among women, than the northern regions. Discrimination based on family origin is persistent and more pronounced in less developed regions, with negative effects on labour force participation.

For Spain, Pérez-Mayo (2019) uses EU-SILC microdata from 2011 to analyse inequality of opportunity from a territorial perspective. Inequality of opportunity is measured through a combination of the regression method proposed by Bourguignon et al. (2007) and the parametric approach of Ferreira and Gignoux (2011): equivalent income is estimated as a function of circumstances, and from this, a counterfactual distribution of income was constructed in which the influence of these circumstances is eliminated. Based on this distribution, the author calculates two indicators: IOL (absolute level of inequality of opportunity) and IOR (the share of inequality explained by circumstances). A Shapley decomposition is also applied to estimate the contribution of each circumstance individually. The results indicate that 10% of total income inequality in Spain is due to uncontrollable circumstances, with parental education and occupation being the main drivers, each contributing more than 20% to inequality of opportunity. Regional differences are significant: in Cantabria, the inequality of opportunity is relatively low (3%), whereas in Extremadura it reaches as high as 33%.

A substantial body of research has also been conducted in non-European countries. For example, Piraino (2015) analyses South Africa using data from the National Income Dynamics Study (NIDS) and the Project for Statistics on Living Standards and Development (PSLSD). His estimates suggest that between 17% and 24% of income inequality can be attributed to inherited circumstances (father’s education and the individual’s race). Race remains a significant factor influencing mobility and economic opportunities in South Africa, even twenty years after the end of apartheid.

Niehues and Peichl (2014) compare Germany and the United States using data from the German Socio-Economic Panel (SOEP) and Panel Study of Income Dynamics (PSID), respectively. They use fixed-effect models for panel data in order to extract an individual effect that remains constant over time, which they interpret as the maximum sum of circumstances beyond the individual’s control. In addition, this fixed-effect estimate is included in a regression model to assess the upper bound of the impact of circumstances on income. For Germany, estimates show that between 47% and 62% of total income inequality can be attributed to circumstances, while for the United States, this proportion ranges between 33% and 36% for annual income.

For Brazil, Bourguignon et al. (2007) use data from the Brazilian National Survey (PNAD) for 1996. The authors compare the actual income distribution with a counterfactual distribution in which all individuals would have the same opportunities. The gap between observed and counterfactual distributions is interpreted as the share of income inequality of opportunity. The results show that, depending on the cohort, between 10% and 37% of income inequality is explained by observable circumstances (i.e., father’s and mother’s education, father’s occupation, race, and region of birth). Once again, parental education appears as the most important determinant of inequality of opportunity, outweighing both race and region of origin.

Hassine (2012) takes a similar approach to the case of Egypt. The study uses data from three national labour force surveys conducted in 1988, 1998 and 2006. These surveys provide information on individual income, parents’ educational and occupational characteristics, gender, region of birth, as well as effort variables such as the individual’s own educational level and age at entry into the labour market. The findings show that around 30% of the total income inequality in Egypt can be attributed to inequalities of opportunity when the Theil index is used to measure inequality and about 17% when using the Gini index. The analysis also shows that 70% of the impact of circumstances on income is direct, while the remining 30% is transmitted through effort-related variables.

A more extensive study, covering inequality of opportunity in six Latin American countries, was carried out by Ferreira and Gignoux (2011). Their work has since become a key reference for many subsequent analyses. Using nationwide survey data, they examine per capita household income as the outcome variable. The authors construct an inequality of opportunity index in two versions: an absolute measure, which captures the total level of inequality of opportunity (IOL—Inequality of Opportunity Level), and a relative measure, which shows its share in the total inequality of economic outcomes (IOR—Inequality of Opportunity Ratio). Their non-parametric estimates of the IOL for household income vary between 0.14 in Colombia and 0.23 in Brazil, while the IOR ranges from 25% in Colombia to 36% in Guatemala. The parametric estimates are only slightly lower in both cases: the IOL ranges from 0.13 in Colombia to 0.22 in Brazil, and the IOR from 23% in Colombia to 34% in Guatemala. The ranking of countries remains stable regardless of whether IOL or IOR is used, and regardless of the estimation method applied (parametric or non-parametric).

The literature on income inequality shows that circumstances beyond individual control play a decisive role, forming the basis for H1 and H2. Marrero and Rodríguez (2012) find that parental education and occupation explain a significant share of income disparities across Europe, with clear contrasts between the Nordic and Mediterranean countries. Andreoli and Fusco (2017) confirm the persistence of such inequalities linked to family background, while Brzezinski (2019) documents their increase in several European states. Evidence from outside Europe, such as Piraino (2015) for South Africa and Bourguignon et al. (2007) for Brazil, further demonstrates the importance of inherited circumstances. Moreover, Hassine (2012) shows that circumstances continue to shape income outcomes even when effort variables are accounted for. Collectively, these findings substantiate the formulation of H1 and H2.

While the studies discussed so far focus on inequality of opportunity in relation to income, other research has examined the concept in different spheres, such as health, living conditions or education.

Education is widely recognised as a key means of breaking the cycle of poverty and reducing social and economic disparities (Hofmarcher, 2021). At the European Union level, one of the five headline targets of the Europe 2020 strategy was to target poverty and social exclusion. As previously mentioned, the Agenda 2030, which is the ongoing plan for the European Union, affirms that no one should be left behind (Rolfe et al., 2021). Educational systems should become more inclusive, more equitable, and more focused on the pupils rather than becoming a political instrument. The academic results should only depend on students’ abilities and effort, not any other factors, such as factors related to social, economic or cultural origin. However, Poder et al. (2016) show that policies targeting education can modify—either by strengthening or diminishing—the influence of an individual’s family background.

Marrero et al. (2024) investigate inequality of opportunity in education by looking at the PISA 2018 database. The influence of school characteristics on inequality of opportunity in education is especially pronounced in countries like those in Central Europe, Italy, Greece, or the United Kingdom, where such inequality is more pronounced. Among the circumstances that impact this type of inequality, the importance of the cultural environment in the initial family and that of the parental occupation came to the foreground.

Camarero Garcia (2022) analyses the implications of increased learning intensity upon inequality of opportunity, which occurred during the 2000s in several German federal states (the length of secondary education by one year was decided, without altering the overall curriculum). Their results indicate that a more intensive learning rhythm increased educational inequality, since students from families with greater resources were better positioned to adapt through private tutoring or other forms of academic support. The impact was more pronounced in mathematics and science than in reading, suggesting that some subjects offer more flexibility in curriculum adaptation than others.

Villar (2017) analyses inequality of opportunity in education using data from the Programme for International Student Assessment (PISA 2012). The measurement is based on a method derived from the Theil index. The data include information on student performance correlated with family and socioeconomic characteristics, such as parental education and available educational resources. Villar’s results show that educational inequality of opportunity is generally high, although levels vary considerably between countries. In the Nordic countries (Finland and Norway), inequality of opportunities is relatively low, reflecting a fairer education system. By contrast, in countries like Mexico, Chile, and Hungary, a much larger share of variation in student performance is explained by the family’s socioeconomic background, which suggests that the environment of origin plays a decisive role in educational success.

In the field of education, the literature similarly highlights the persistent influence of family background on outcomes, which guided the formulation of H3. Villar (2017) shows that the extent to which student performance depends on parental education and socioeconomic conditions varies considerably across countries, while Marrero et al. (2024) demonstrate that parental occupation and cultural environment play a decisive role in explaining educational disparities in Europe. These contributions suggest that early-life conditions are a consistent determinant of educational trajectories, providing the foundation for H3.

3. Data and Methodology

Estimating the inequality of opportunities and its share in total income inequality is difficult, mainly due to the limited availability of data. Datasets containing sufficient information about both circumstances and individual effort are scarce, making it challenging to identify the factors that determine income differences between individuals accurately. Moreover, the ability to make comparisons between countries or over time is restricted for the same reason.

For our empirical analysis, we used data from the Life in Transition Survey (LiTS), a large-scale, internationally comparable household survey conducted by the European Bank for Reconstruction and Development (EBRD) in collaboration with the World Bank. Using a standardised questionnaire, LiTS is nationally representative and covers both urban and rural areas, with more than 1000 households per country. The most recent wave, LiTS IV, was carried out between 2022 and 2023 and covered 37 economies. For our study, the Central and Eastern European countries were selected: Bulgaria, Croatia, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, the Slovak Republic, and Slovenia. The survey was conducted using face-to-face interviews carried out in respondents’ homes, targeting adults aged 18 and over. A random sampling method was used in all countries, except Czechia, where quota sampling was applied.

This dataset includes key information that can support the identification of factors influencing individual income levels and the assessment of inequality of opportunity, both in terms of income and education. Specifically, the survey collects data on respondents’ parental background—such as the education level and employment sector of both mother and father—as well as on childhood conditions, including place of birth and cultural context, proxied by the number of books in the household. Together, these variables constitute an essential set of circumstances relevant for measuring inequality of opportunity. In addition, the data are comparable across countries, which adds value by enabling both the ranking and comparative analysis of national contexts. The focus on Central and Eastern European (CEE) countries is motivated by their relative homogeneity due to their comparable socio-economic trajectories, institutional reforms, and similar inequality and social mobility challenges.

The two outcome variables targeted in our study were the level of education and the respondent’s net monthly income. In our estimates, education was introduced as a categorical variable with three levels (low, medium, and high), classified according to the International Standard Classification of Education (ISCED), which ensures comparability across countries. The low level of education (ISCED 0–2) includes no formal education, primary school, or lower secondary education. The medium level (ISCED 3–4) covers upper secondary or post-secondary non-tertiary education. The high level (ISCED 5–8) refers to tertiary education, which includes short-cycle tertiary programs, bachelor’s, master’s, and doctoral studies. Regarding the level of net income, since the data were initially collected in national currencies, the amounts were converted into euros to ensure cross-country comparability. The conversion was based on the average exchange rate during the data collection period for six non-euro countries: Bulgaria, Croatia, Czechia, Hungary, Poland, and Romania. In the analysis, we used the natural logarithm of net income to account for the skewed income distribution. The dataset used was supplemented with a series of circumstances and effort variables, described in

Table 1. Description of variables.

Variable	Type	Description	Descriptive Statistics
Income	Continuous	Net monthly income (euro)	Mean: 1076.8
			Std. dev: 1265.06
Circumstances
Gender	Binary	1—male	43.2%
		2—female	56.8%
Place of birth	Binary	1—urban	63.8%
		2—rural	36.2%
Father’s education	Categorical	1—low education ISCED 0–2	40.2%
		2—medium education ISCED 3–4	46.8%
		3—higher education ISCED 5–8	13.0%
Mother’s education	Categorical	1—low education ISCED 0–2	44.0%
		2—medium education ISCED 3–4	42.8%
		3—higher education ISCED 5–8	13.2%
Father’s sector of employment	Categorical	1—Not working	1.4%
		2—Agriculture, Manufacturing or Mining	45.5%
		3—Construction, Transportation, Public utilities, Wholesale trade, or Retail trade	34.0%
		4—Finance, Insurance, Real Estate, Services or Public Sector	19.1%
Mother’s sector of employment	Categorical	1—Not working	11.0%
		2—Agriculture, Manufacturing or Mining	36.9%
		3—Construction, Transportation, Public utilities, Wholesale trade or Retail trade	19.00%
		4—Finance, Insurance, Real Estate, Services or Public Sector	33.1%
Books at home in childhood	Categorical	1—none or fewer than 10 books	20.4%
		2—10 to 100 books	58.4%
		3—more than 100 books	21.2%
Country	Categorical	11 CEE countries: Bulgaria, Croatia, Czechia, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, the Slovak Republic, and Slovenia	Around 9% each country
Effort
Current place of residence	Binary	1—urban	63.9%
		2—rural	36.1%
Respondent’s education	Categorical	1—low education ISCED 0–2	17.8%
		2—medium education ISCED 3–4	55.5%
		3—higher education ISCED 5–8	26.7%
Respondent’s sector of employment	Categorical	1—Agriculture, Manufacturing, Mining or Construction	27.2%
		2—Transportation, Public utilities, Wholesale trade or Retail trade	21.4%
		3—Finance, Insurance, Real Estate, or Services	28.5%
		4—Public Sector (education, administration, health care)	22.9%
Work experience	Categorical	1—less than 3 years	19.4%
		2—3 to 6 years	11.9%
		3—more than 6 years	68.8%
Training or formal education last year	Binary	1—yes	16.7%
		2—no	83.3%
Technical skills	Binary	1—yes	35.5%
		2—no	64.5%
Foreign language skills	Binary	1—yes	49.8%
		2—no	50.2%

.

To meet the research objectives and given the structure of the available data, this study considered a set of methodological approaches. The first methodological aspect addressed in this study is the estimation of inequality of opportunity (IOp) for the two outcomes under analysis, income and education. We used the iop command implemented in Stata software (version 9.2) (Juárez & Soloaga, 2014), as it provides options that can be used for both types of variables used in our study and is grounded in a robust methodological framework. Moreover, it allows for detailed decompositions that facilitate a deeper understanding of inequality of opportunity by identifying the relative importance of circumstances and enabling a comparative ranking of the countries analysed based on their IOp levels. As a further step, the analysis was extended by estimating regression models aimed at identifying the key determinants of income and education, as well as examining how these factors influence the respective outcomes. For the income analysis, we employed an Ordinary Least Squares (OLS) regression model. At the same time, for the education outcome, we used a multinomial logistic regression model, which is appropriate for modelling categorical variables. This methodological choice reflects the distinct nature of the two dependent variables and ensures that the estimation approach aligns with their specific distributional properties.

To estimate the inequality of opportunity, we consider y as the outcome variable of interest and a finite set of distinct individuals $i ϵ \left\{1, 2,\dots , N\right\}$. An individual’s outcome (income or level of education), denoted by y_i, is modelled as a function of their level of effort (e_i) and a vector of circumstances (C_i) that characterise their background, such that y_i = f(C_i, e_i). Circumstances are considered beyond the individual’s control, as they cannot be altered through personal decisions. In contrast, effort is viewed as a matter of individual control, though it is still partly shaped by those circumstances (Marrero & Rodríguez, 2012).

The population is partitioned into P mutually exclusive and collectively exhaustive groups, denoted by G = {G₁, …, Gp}, so all individuals within a group share identical circumstances. In other words, each individual belonging to group p has identical values of C_i. The distribution of effort within group p is noted with F_p, and e_p(r) represents the effort level corresponding to the r-th quantile of that distribution. Equality of opportunity is attained when a person’s income is independent of their socio-economic background or parental origins (Bourguignon et al., 2007; Lefranc et al., 2006). This can be formally stated as follows:

$$F_p\left(y\right)=F_q\left(y\right), \forall p,q|G_pϵG,|G_qϵG$$

(1)

When circumstances are extensive, the number of observations per type becomes too small to estimate distribution functions reliably. A practical alternative is to use a specific moment of these distributions, typically the mean vector, denoted by u. Equality of opportunity requires, at a minimum, that all elements of the vector u be identical, that is:

$$u_p\left(y\right) = u_q\left(y\right), \forall p, q | G_{p }ϵ G, | G_{q }ϵ G$$

(2)

The level of inequality of opportunity can be calculated by applying a standard inequality index, denoted I(u). Among the various inequality measures that satisfy the fundamental principles identified in the inequality literature, Ferreira and Gignoux (2011) opt for the Mean Logarithmic Deviation (also known as Theil 0). This index belongs to the generalised entropy class, allowing additive decomposability. Thus, the Theil index is decomposed into between-group and within-group inequality components, as follows:

$$T_0(y) = T_0(u) +\sum _{p=1}^P{\displaystyle \frac{n_p}{N}} T_0(y_p)$$

(3)

where n_p denotes the population size of type p. The overall Theil index reflects total outcome inequality, while the between-group component measures inequality of opportunity.

In our study, the inequality of opportunity in net income is assessed following an ex ante framework, using a parametric approach inspired by Ferreira and Gignoux (2011), which relies on a log-linear OLS regression model—specifically, a reduced-form linear specification that estimates advantages as a function of observable circumstances. This approach is particularly suitable for datasets like LiTS that include rich information on circumstance variables while also being robust to limited sample sizes. The method provides a lower-bound estimate of true inequality of opportunity by treating the proportion of total outcome inequality explained by observable circumstances as a minimum estimate of the overall influence of circumstances.

To analyse inequality of opportunity in education, we adopted the approach proposed by Paes de Barros et al. (2009). This approach consists of estimating a logistic regression model and using the dissimilarity index as the inequality measure, given that the choice of metric depends on the nature of the outcome variable.

The inequality of opportunity was estimated using the iop algorithm implemented in Stata, which applies OLS for continuous outcomes and logit models for categorical or binary ones, as described above. Inequality of opportunity is first computed in absolute terms using predicted values from these regressions, then expressed in relative terms as the ratio to the total inequality calculated for the original outcome.

To gain deeper insight, we extended the analysis by decomposing the inequality of opportunity in income and education. We began by looking at how much each circumstance contributes to inequality of opportunity for each outcome, using a Shapley decomposition. In this approach, we calculate the inequality index for every possible combination of circumstance variables. From these calculations, we can determine each variable’s average marginal contribution.

Despite its computational intensity, it ensures order independence and an exact breakdown of total inequality. Second, we examined cross-country differences in inequality of opportunity through an Oaxaca-type decomposition. This approach focuses on differences between groups, in our case, countries. Inequality of opportunity can differ either due to variations in the distribution of circumstances, namely a composition effect, or due to differences in how these circumstances influence outcomes, which is the association effect. To separate these effects, inequality of opportunity is computed for each country, and counterfactuals are constructed by applying the coefficients from one country to the distribution of circumstances in another (Juárez & Soloaga, 2014).

The income estimation relied on a multivariate regression model with a log-linear specification, which means that net income is transformed using the natural logarithm, and the explanatory variables included in the model are in their original values:

$$log y_{i }=\alpha +{\beta }_1·x_{i1}+ {\beta }_2·x_{i2}+\cdots + {\beta }_n·x_{ip}+ {\epsilon }_i$$

(4)

where y is the net income, x₁ … x_p are the independent variables, $\alpha$ denotes the intercept, ${\beta }_i$ are the parameters associated with the explanatory variables, $\epsilon$ represents the error term, and i indicates the analysed individuals (Wooldridge, 2012). The classical assumptions of the multivariate linear regression model are a continuous dependent variable, the absence of multicollinearity among the explanatory variables, and, for the error term, a normal distribution, homoscedasticity, and no autocorrelation. If the assumption of homoscedasticity is violated, which is common in cross-sectional data, Stata allows for robust estimation of standard errors.

The multinomial logistic regression is an extension of binary logistic regression, which is used when the dependent variable has three or more nominal categories (Agresti, 2010). In our case, the outcome variable is the education level, with three categories (low, medium, and high). The model estimates the likelihood of each outcome category based on a set of predictors, comparing each with a baseline category.

The multinomial model relies on logistic functions. A multinomial logit model with n predictors has the following form:

$$log \left({\displaystyle \frac{y_{j }}{y_J}}\right)={\alpha }_j+{\beta }_1·x_{j1}+ {\beta }_2·x_{j2}+\cdots + {\beta }_n·x_{jn}+ {\epsilon }_j$$

(5)

with j = 1, …, J − 1, n = 1, …, N.

Independent variables can be continuous, categorical, or dummy. If a predictor is categorical, dummy variables are created using a reference level for each, and the coefficients are interpreted accordingly. Since the dependent variable in our model has three categories, two separate equations are estimated—each contrasting one category at a time with the baseline category. As a result, the impact of the predictors will vary across comparisons, providing different effects of each factor on different levels of the outcome.

For the multinomial logit model, we verified the Independence of Irrelevant Alternatives (IIA) assumption, evaluated goodness-of-fit using Hosmer–Lemeshow tests in a one-versus-rest framework, and measured the model’s discriminative power through the area under the ROC curve. Detailed outcomes of these checks are presented in Section 4.

4. Results and Discussion

4.1. Inequality of Opportunity in Income

The first objective of this study was to estimate the inequality of opportunity in the 11 CEE countries analysed, both in absolute and relative terms. The results are presented in

Table 2. Inequality of opportunity in net income.

	Absolute	Relative
Inequality of opportunity	0.000702 *	11.96% *
Standard error a	0.000092	0.000092
Number of observations	1263

Note: ^a Bootstrap-based estimation with 500 repetitions; * Statistically significant at the 5% level. Source: Authors’ calculations using Stata.

. The estimates indicate an absolute value of the inequality of opportunity in net income (in logarithms) of 0.000702. The relative value of 11.96% is of greater practical relevance, indicating that approximately one-eighth of the total income inequality can be attributed to circumstances rather than individual effort. The reliability of these results is supported by the very small standard errors calculated by a bootstrap procedure with 500 repetitions.

Applying the Shapley decomposition allowed a more in-depth analysis of the contribution of each analysed circumstance to the previously identified inequality of opportunity in income. The results indicate that the most important factor outside the individual’s control that contributes to inequality of opportunities is gender, reconfirming once again the existence of gender disparities that exert a strong influence on income (

Table 3. Shapley decomposition of inequality of opportunity in net income by circumstances.

Variable	Absolute Value	% of IOp
Gender	0.000340	48.40%
Place of birth	0.000023	3.33%
Father’s education	0.000099	14.13%
Mother’s education	0.000108	15.44%
Books at home in childhood	0.000052	7.42%
Father’s sector of employment	0.000006	0.83%
Mother’s sector of employment	0.000055	7.79%
Country	0.000019	2.66%

Source: Authors’ calculations using Stata.

). Almost half (48.40%) of the inequality of opportunity in income is determined by the gender of the respondent, suggesting that gender income inequality remains a crucial issue in Central and Eastern European countries.

Furthermore, the results showed that parental education is an important contributor to inequality of opportunity. It should be noted that maternal education is slightly more important than paternal education (15.44% versus 14.13%), results that are in line with similar studies in the field, which state that the mother’s level of education has a greater impact on children’s life outcomes than the father’s (Erola et al., 2016).

A greater influence of the mother is also observed for the sector of activity. Circumstances related to the sector of activity of the parents explain around 10% of the inequality of opportunity in income, but this time the difference between parents is important (7.79% for the mother versus only 0.83% for the father). A possible explanation for this difference is that gender roles are still pronounced in the CEE countries , making the mother’s sector of activity more influential for the household’s financial resources and, consequently, for inequality of opportunity (Metcalfe & Afanassieva, 2005).

The number of books the respondent had in their household during childhood was included in the study as a proxy for the socio-economic status of the household of origin (as in Poder et al., 2016; Rolfe et al., 2021). The Shapley decomposition results indicate a 7.42% contribution of this variable to inequality of opportunity in income, exceeding the contributions associated with place of birth (3.33%) or country (2.66%).

The results obtained from the Shapley decomposition are consistent with those reported by Mogila et al. (2022), in the sense that their study also identifies parental education as an essential factor in determining inequality of opportunity, and gender as a substantial component in countries such as Estonia and Slovakia. However, the set of countries analysed is not identical, which allows only a partial comparison of the results.

The decomposition of inequality of opportunity by country, based on the Oaxaca method, allowed the ranking of the analysed states and highlighted major differences between Central and Eastern European countries regarding inequality of opportunity in income, as shown in Figure 1.

The highest value is recorded in Latvia, followed by Hungary and Estonia, suggesting that in these countries, a larger share of income variation is explained by circumstances beyond individual control included in the analysis. In this context, public policies aimed at reducing these disadvantageous circumstances could have a more substantial impact in attenuating inequalities. On the other hand, countries such as Romania, Poland, and Lithuania registered much lower levels of inequality of opportunity, which may indicate either that income differences are less determined by circumstantial factors or that there are other mechanisms (e.g., labour market structure, fiscal redistribution) that mitigate their effect. Countries with moderate levels of inequality are Bulgaria, the Czech Republic, Croatia, Slovakia, and Slovenia. The relatively large differences observed between the CEE countries indicate considerable regional heterogeneity. These findings are consistent with those of Marrero and Rodríguez (2012), who also identified Hungary and Estonia among the CEE countries with high inequalities of opportunities in terms of income. Czechia and Slovakia ranked among those with the lowest values.

The next step of the analysis involved examining how the set of circumstances (C_j) influences income, specifically by investigating the results of the OLS regression model estimation $log {income}_{i }=\alpha +{\beta }_i·C_{i\mathrm{j}}+ {\epsilon }_i$, presented in

Table 4. OLS regression results—income explained by circumstances and effort variables.

Variable	Model 1—Circumstances		Model 2—Circumstances and Effort
	Coefficient	Std Error	Coefficient	Std Error
Gender (ref: male)	−0.3343 *	0.0306	−0.2880 *	0.0333
Place of birth (ref: rural)	−0.0436	0.03204	−0.0295	0.0313
Father’s education (ref: low ISCED 0–2)
medium: ISCED 3–4	0.1065 *	0.0352	0.0741 *	0.0358
high: ISCED 5–8	0.1002 **	0.0564	0.0426	0.0535
Mother’s education (ref: low ISCED 0–2)
medium: ISCED 3–4	0.1241 *	0.0366	0.0801 *	0.0376
high: ISCED 5–8	0.2763 *	0.0576	0.1915 *	0.0556
Father’s sector of employment (ref: Not working)
Agriculture, Manufacturing or Mining	−0.1393	0.0930	−0.1230	0.0998
Construction, Transportation, Public utilities, Wholesale trade or Retail trade	−0.1525	0.0946	−0.1153	0.1003
Finance, Insurance, Real Estate, Services or Public Sector	−0.2386 *	0.0983	−0.2118 *	0.1028
Mother’s sector of employment (ref: Not working)
Agriculture, Manufacturing or Mining	0.0219	0.0563	0.0366	0.0544
Construction, Transportation, Public utilities, Wholesale trade or Retail trade	0.0985	0.0639	0.0819	0.0628
Finance, Insurance, Real Estate, Services or Public Sector	0.1270 *	0.0608	0.0846	0.0592
Books at home in childhood (ref: none or fewer than 10 books)
10 to 100 books	0.1312 *	0.0502	0.0489	0.0471
more than 100 books	0.1900 *	0.0595	0.0540	0.0572
Country (ref: Romania)
Bulgaria	−0.1940 *	0.0631	−0.1352 *	0.0570
Croatia	0.2968 *	0.0680	0.2860 *	0.0614
The Czech Republic	0.3849 *	0.0730	0.4651 *	0.0736
Estonia	0.4427 *	0.0635	0.4642 *	0.0608
Hungary	−0.1525 *	0.0756	0.0171	0.0741
Latvia	0.0608	0.0662	0.1184 **	0.0604
Lithuania	0.2733 *	0.0853	0.2793 *	0.0849
Poland	0.2362 *	0.0778	0.2741 *	0.0772
Slovakia	0.1915 *	0.0656	0.2627 *	0.0634
Slovenia	0.5638 *	0.0628	0.5650 *	0.0568
Current place of residence (ref: urban)			−0.0312	0.0265
Respondent’s education (ref: low ISCED 0–2)
medium: ISCED 3–4			0.1039 *	0.0521
high: ISCED 5–8			0.2796 *	0.0621
Respondent’s sector of employment (ref: Public Sector—education, administration, health care)
Agriculture, Manufacturing, Mining or Construction			0.0743	0.0609
Transportation, Public utilities, Wholesale trade or Retail trade			0.0855 **	0.0455
Finance, Insurance, Real Estate or Services			0.1348 *	0.0393
Experience (ref: less than 3 years)
3 to 6 years			0.2537 *	0.0632
more than 6 years			0.2654 *	0.0529
Training recently (ref: yes)			−0.0951 *	0.0323
Technical skills (ref: yes)			0.0638 **	0.0345
Foreign language skills (ref: yes)			0.1271 *	0.0301
Constant	6.9243 *	0.1295	6.2832 *	0.1732
R-squared	0.324		0.393
Number of observations	1263		1234

Note: * Statistically significant at the 5% level; ** Statistically significant at the 10% level. Source: Authors’ calculations using Stata.

—Model 1. This brings added value through the possibility of observing the effects of circumstances in more detail, especially categorical ones.

The results indicated that the respondent’s gender strongly influences income. The negative sign, as expected, indicates that women earn considerably less than men, which underlines the persistence of gender wage disparities with potential implications for labour market equality and household welfare. Regarding the place of birth, no statistically significant differences are observed between individuals born in urban areas and those born in rural areas.

Both medium and higher levels of father’s education are associated with higher incomes compared to low education, although the difference between medium and higher education is small, suggesting limited economic returns from the father’s additional schooling. In contrast, in the case of mother’s education, the results show an income advantage for respondents whose mothers have a medium level of education, compared to the reference category, and an even more pronounced advantage for those whose mothers hold a higher education degree. This pattern highlights the stronger intergenerational transmission of economic benefits from maternal education, which may reflect both the mother’s role in shaping children’s skills and aspirations, as well as the better employment prospects, income stability, and social networks that mothers with higher education are able to transfer to their families.

Regarding the parents’ sector of activity, the results obtained show a negative impact on the respondent’s income when the father worked in Finance, Insurance, Real Estate, Services or Public Sector, compared to the reference category of those not employed. This counter-intuitive result may reflect specific features of labour markets in CEE countries, where wages in public services or certain parts of the financial and real estate sectors have historically been relatively modest, while parents not formally employed may have relied on alternative or entrepreneurial sources of income. In the case of mothers, those employed in Finance, Insurance, Real Estate, Services or Public Sector offered an advantage to their children compared to the reference category, the estimated coefficients being statistically significant, associated with a positive impact on the respondent’s income.

The impact of the number of books in childhood is positive and significant: the more books present in the household, the higher the income, confirming the relationship between early educational resources and later earnings.

To assess the effects of the country on income, we chose Romania as the reference country, the one with the lowest inequality of opportunities. The most considerable differences are recorded in Slovenia, Estonia, and Czechia, where incomes are significantly higher than in Romania. Moderate positive differences are observed in Croatia, Lithuania, Poland, and Slovakia. Only Bulgaria and Hungary recorded, on average, lower incomes than in Romania. Latvia does not show statistically significant differences in income compared to Romania.

The estimate of relative inequality of opportunity in income of 11.96%, together with the Shapley decomposition showing gender and parental education as the circumstances with the largest share in IOp, as well as the first regression model explaining 32.4% of income variation solely through circumstances, all support Hypothesis 1 that a considerable part of income inequality is attributable to background characteristics.

The next step was to estimate an extended model by adding variables that quantify individual effort. The results are presented in Table 4—Model 2. The results show that gender remains a significant determinant, with the coefficient associated with women being negative and robust in both models, but with a lower magnitude in Model 2, suggesting that effort variables mediate part of the gender pay gap. Regarding parental educational capital, the father’s education loses its relevance when individual effort is accounted for, the higher level becoming statistically insignificant, while the mother’s education retains its significant influence, even if the coefficients decrease slightly. This asymmetry suggests that the mother’s education is more directly and persistently linked to the children’s income. The occupational structure of parents has a limited effect: only employment in the Finance, Insurance, Real Estate, Services or Public Sector remains significantly associated with income, but the magnitude of the effects is modest. Finally, cross-national differences remain strong: countries such as the Czech Republic, Estonia, Lithuania, Poland, Slovakia, and Slovenia have significant and stable income advantages over Romania, while Bulgaria’s disadvantage remains. Overall, the results suggest that, although some of the variation in income is explained by individual effort, certain circumstances of origin—in particular gender, maternal education, and national context—continue to exert significant and robust effects on income.

Regarding the set of variables that quantify individual effort, the signs obtained are consistent with economic expectations. The respondent’s education has a positive and statistically significant impact: a medium level of education is associated with a higher income compared to the reference category of low education, and the effect of higher education is even more substantial. These results reflect the well-documented economic returns to schooling, whereby each additional level of education enhances human capital, increases productivity, and improves access to better-paid jobs in the labour market.

The sector in which the respondent works matters in terms of income. The results indicated that, on average, incomes are higher for respondents employed in Finance, Insurance, Real Estate or Services compared to the reference category, and the positive impact is even more substantial for those working in the Public Sector (education, administration, health care).

Work experience, as expected, plays an important role: the greater the experience, the higher the earnings. However, no substantial differences are observed between mid-career and senior levels of experience, which may indicate diminishing returns to work experience, as the additional income gains become more limited at later stages of the career.

Respondents who reported participating in a training program or formal education in the past 12 months had, on average, lower income. This negative association may be explained by the fact that most often these programs are attended by people at the beginning of their careers or who need to improve their skills, and the results of these trainings have not yet been translated into income.

A high self-rated level of skills, both technical and foreign language, is associated with higher average incomes, with the results indicating stronger effects in the case of foreign language skills. This finding underscores the increasing value of language proficiency in globalised labour markets, where foreign language skills can expand access to higher-paying positions and complement technical abilities.

The increase in the coefficient of determination from 0.324 (the model with circumstances only) to 0.393 (circumstances plus effort) indicates that, even after accounting for individual effort, circumstances continue to play a substantial role in explaining income, thus supporting Hypothesis 2.

We checked both models against the classical OLS assumptions, focusing on multicollinearity, normality of residuals, and heteroscedasticity. The diagnostic tests indicated deviations from normality and confirmed the presence of heteroscedasticity, while multicollinearity did not appear to be a concern. To address these issues, the final estimations presented above were conducted using robust standard errors, which improve the validity of statistical inference under such conditions. Even so, the results should be interpreted with caution, as potential endogeneity implies that the identified relationships capture associations rather than clear causal effects. Nevertheless, the findings provide valuable insights into the persistence of family background influences on income, even after controlling for effort.

4.2. Inequality of Opportunity in Education

The results presented in

Table 5. Inequality of opportunity in education.

	Absolute—Dissimilarity Index	Standard Error a
Inequality of opportunity—high education versus low & medium education	0.2835 *	0.0137
Number of observations	2933

Note: ^a Bootstrap-based estimation with 500 repetitions; * Statistically significant at the 5% level. Source: Authors’ calculations using Stata.

show the absolute level of inequality of opportunity in education estimated using the dissimilarity index with the iop command in Stata with the logit option. This index is applied to a binary variable that measures the respondent’s level of education, distinguishing between those with higher education and those with a low or medium level of education. The circumstances variables included in the estimation are gender, place of birth, parents’ education, number of books in the household, the parents’ sector of activity and the country of residence.

The results show that the inequality of opportunity in education is 0.2835. The dissimilarity index takes values between 0 and 1; therefore, the value obtained suggests that a considerable share of individuals’ educational attainment variation can be explained by circumstances beyond their control (family background or socio-economic conditions), rather than by their own efforts or choices. The bootstrap standard error of 0.0137 is relatively accurate, suggesting that the calculated value of the inequality of opportunity is statistically reliable and unlikely due to sampling variation.

The Shapley decomposition breaks down the inequality of opportunity in education, showing how much each circumstance contributes (

Table 6. Shapley decomposition of inequality of opportunity in education by circumstances.

Variable	Absolute Value	% of IOp
Gender	0.0066	2.35%
Place of birth	0.0201	7.18%
Father’s education	0.0652	23.28%
Mother’s education	0.0736	26.27%
Books at home in childhood	0.0467	16.66%
Father’s sector of employment	0.0201	7.16%
Mother’s sector of employment	0.0467	16.66%
Country	0.0012	0.44%

Source: Authors’ calculations using Stata.

). The results indicate that parental education is the most important factor, accounting for practically half of the total inequality. The mother’s education plays a somewhat larger role than the father’s, suggesting a substantial influence on educational outcomes. Hu and Qian (2023) also found that mothers’ level of education plays an equally or even more important role in intergenerational mobility than fathers’ level of education. Kong et al. (2015) also highlight the importance of mothers’ educational background on children’s intelligence.

The presence of books at home during childhood explains about 17% of the inequality, suggesting the importance of the home learning environment. Having access to books at home echoes the family’s cultural and educational interests and encourages reading habits and cognitive development from an early age. This consequently shapes a child’s academic skills and motivation, transforming them into a key factor in creating unequal educational opportunities. Other studies have also shown that the cultural and educational background of the family is important when analysing inequality of opportunity. For example, Heppt et al. (2022) argued that the quantity of books available at home strongly predicts students’ academic performance, while Evans et al. (2010) found that children raised in homes with books tend to complete, on average, three more years of education compared to those from homes without books, regardless of their parents’ education level, occupation, or social class.

The mother’s occupation sector contributes approximately 17% to the overall inequality of opportunity. This suggests that the mother’s work environment has an important effect on a child’s educational outcomes (it could influence household resources, time availability, or social networks). In contrast, the father’s employment sector is less influential, possibly reflecting changing family dynamics or differences in parental involvement in children’s education.

Gender and place of birth appear to contribute relatively little in this case. Similarly, the country of residence does not seem to play a significant role, accounting for less than 1% of the inequality of opportunity in education. Understanding these contributions is important because it helps identify which areas—especially parental education and home environment—need the most attention to reduce educational inequality of opportunity.

The Oaxaca decomposition allowed for the analysis of the inequalities of opportunity in education by country. The results presented in Figure 2 highlight considerable variations in inequality of opportunity across Central and Eastern European countries. Romania stands out with the highest level of inequality of opportunity, indicating a strong dependence of educational achievements on factors beyond individuals’ control and reflecting persistent structural barriers in equal access to education. Hungary and Slovakia are also in the upper part of the ranking, with high levels of inequality of opportunity in education.

At the opposite end, Czechia has the lowest level of inequality of opportunity in education in the region, which indicates a relatively lower influence of circumstances on the educational level achieved. Estonia, Lithuania, and Latvia are also at the bottom of the ranking, suggesting fairer education systems in terms of the distribution of opportunities. Poland, Bulgaria, Slovenia, and Croatia belong to the middle zone, which indicates moderate levels of inequality of opportunity in education in these countries.

The results reveal challenges in reducing the impact of circumstances on educational outcomes in Slovakia, Hungary, and especially Romania, as well as the need for public policies aimed at enhancing the equity of education systems.

The results of the multinomial logistic regression model, presented in

Table 7. Multinomial logistic regression results—education level explained by circumstances.

Variable	Medium Versus Low Education		High Versus Low Education
	Coefficient	Std Error	Coefficient	Std Error
Gender (ref: male)	−0.3987 *	0.1343	0.0454	0.1574
Place of birth (ref: rural)	−0.4346 *	0.1309	−0.4473 *	0.1568
Father’s education (ref: low ISCED 0–2)
medium: ISCED 3–4	1.0919 *	0.2069	1.5119 *	0.2222
high: ISCED 5–8	1.3690 *	0.4958	2.4306 *	0.4908
Mother’s education (ref: low ISCED 0–2)
medium: ISCED 3–4	0.8308 *	0.2447	1.2916 *	0.2558
high: ISCED 5–8	−0.2158	0.4196	0.9028 *	0.4127
Father’s sector of employment (ref: Not working)
Agriculture, Manufacturing or Mining	0.5793 **	0.3051	0.9417 **	0.5164
Construction, Transportation, Public utilities, Wholesale trade or Retail trade	0.5304 **	0.3212	0.4416	0.5288
Finance, Insurance, Real Estate, Services or Public Sector	0.4766	0.3557	0.7836	0.5506
Mother’s sector of employment (ref: Not working)
Agriculture, Manufacturing or Mining	0.1394	0.1815	0.0809	0.2325
Construction, Transportation, Public utilities, Wholesale trade or Retail trade	0.4157	0.2830	0.6080 **	0.3229
Finance, Insurance, Real Estate, Services or Public Sector	0.3297	0.2397	0.8103 *	0.2830
Books at home in childhood (ref: none or fewer than 10 books)
10 to 100 books	1.0493 *	0.1468	1.7912 *	0.2122
more than 100 books	1.4340 *	0.2477	2.7467 *	0.2936
Country (ref: Romania)
Bulgaria	−0.0520	0.3177	−0.0644	0.3571
Croatia	0.5100	0.3110	0.5418	0.3601
The Czech Republic	0.4555	0.4748	−0.2991	0.5006
Estonia	−0.2151	0.3265	−0.3222	0.3612
Hungary	−1.5019 *	0.3261	−1.9577 *	0.4019
Latvia	−0.0895	0.3214	−0.0013	0.3560
Lithuania	14.3893 *	0.4506	15.5448 *	0.4312
Poland	−1.3549 *	0.3365	−0.7002 **	0.3686
Slovakia	0.0714	0.3778	−0.5980	0.4220
Slovenia	0.0217	0.2986	0.2686	0.3352
Constant	0.7627	0.5139	−2.4245 *	0.7365
Pseudo R-squared	0.2085
Number of observations	2933

Note: * Statistically significant at the 5% level; ** Statistically significant at the 10% level. Source: Authors’ calculations using Stata.

, detail the impact of each analysed circumstance on the chances of achieving a medium or higher level of education compared to a low level of education.

In the transition from a low level of education to a medium level, gender plays a significant role: women are less likely than men to reach this level. This effect is no longer significant for higher education, suggesting that gender disparities are mostly manifested in the intermediate stages of the educational pathway, not in tertiary education. The place of birth has a negative effect on both medium and higher education, but the association is slightly stronger for the latter. This indicates that people born in rural areas face persistent disadvantages, which may translate into lower access to skilled jobs and reduced long-term income prospects, thereby perpetuating intergenerational inequality.

Regarding the level of education achieved by parents, the results indicated significant and strong effects, highlighting that parents’ education is a central determinant of the educational level achieved by the respondent. In the case of both the father and the mother, a higher level of education increases the child’s probability of obtaining medium or higher education, with the effects being more pronounced for higher education. More precisely, the mother’s medium level of education has a positive and significant impact on both the respondent’s chances of achieving secondary and higher education. However, the mother’s tertiary level of education does not significantly influence the probability of having secondary versus low education, but it becomes an important factor for higher education. This difference suggests that the effects of the mother’s education are more visible and stronger when she is a tertiary education graduate. The father’s education proves to be the most important predictor of the respondent’s education, both medium and high, with accentuated effects when the father has a university degree. In such cases, respondents are 2.4 times more likely to achieve a high educational level, compared to the reference category, ISCED 0–2. These findings underscore the central role of parental education in shaping intergenerational mobility: higher parental education not only improves children’s educational attainment but also enhances their future earning capacity and labour market opportunities.

Our results suggest that the father’s employment in Agriculture, Manufacturing or Mining significantly increases the likelihood of children achieving medium or high education levels, compared to the reference category of not working. In contrast, for higher education, there is a positive effect in cases where mothers worked in Construction, Transportation, Public utilities, Wholesale trade, or Retail trade and in Finance, Insurance, Real Estate, Services or Public Sector.

The presence of books in the household during childhood is positively and significantly associated with achieving both medium and higher levels of education. The effects are considerably stronger for higher education. Access to a larger number of books reflects a learning-oriented family environment, which facilitates not only the completion of secondary education but also the continuation of the academic path to the university level. A respondent who had more than 100 books in the house in childhood is 2.7 times more likely to be a graduate of higher education (compared to having a low level of education) than one who had no more than 10 books. This highlights the long-term economic significance of early access to cultural resources, as it fosters higher educational attainment, which in turn translates into greater human capital accumulation and improved labour market outcomes.

Country-level differences are observed, with Hungary and Poland showing significantly lower chances of both medium and high education, indicating structural disadvantages relative to Romania (the baseline in this case). By contrast, Lithuania displays very large and significant positive coefficients, suggesting a much stronger likelihood of educational attainment. For the remaining countries, the coefficients appear to be statistically insignificant. The results highlight a heterogeneous pattern across Central and Eastern European countries.

We performed several robustness checks for our multinomial logit model to ensure that the model is generally well specified. The likelihood ratio test confirmed that the explanatory variables jointly contribute to the model. Also, the Hausman tests provided no evidence against the IIA assumption, suggesting that the multinomial logit specification is appropriate in this context. The Hosmer–Lemeshow goodness-of-fit test and ROC analysis indicated that the model provides an overall satisfactory fit and discriminative ability. While the specification tests support the adequacy of the multinomial logit model, the results should be interpreted with caution. The cross-sectional nature of the data and potential unobserved heterogeneity limit the scope for strong causal claims, meaning that the estimated effects reflect robust associations rather than definitive causal relationships.

Our third research hypothesis is supported by the estimate of inequality of opportunity in education of 0.28% and the Shapley decomposition. The latter shows that parental education and the presence of books are the circumstances that contribute the most to IOp. In addition, the multinomial regression model results emphasised the strong influence of the father’s level of education and the books in the origin family, hence substantiating this hypothesis.

4.3. Robustness and Sensitivity Analyses

To reinforce the validity of our findings, the analysis is complemented by a series of robustness and sensitivity checks. Robustness is examined through re-estimations on selected subsamples, more precisely, random subsamples of 70%, 75%, and 80% of the data and on two age-based subsamples (18–54 years and 54+ years), which test the stability of the results across different population groups. Sensitivity is assessed using the Oster (2017) test of coefficient stability, which evaluates the influence of potential omitted variable bias. The variables of interest were gender, mother’s education, and father’s education, as these were found to explain a large share of income inequality of opportunity in the Shapley decomposition analysis. Together, these complementary analyses strengthen the credibility of the results and the reliability of the conclusions drawn.

We started the investigation of robustness with Model 1, which explains income by circumstances. The results largely confirm the coefficients’ robustness in magnitude and sign (see Table S1). The gender variable remains negative and strongly statistically significant across all specifications, suggesting a persistent disadvantage for women in terms of income. Place of birth (urban/rural) is not statistically significant, a result that holds across all subsamples. Parental education is positively associated with respondents’ income, although the effects differ between mothers and fathers. For mothers, both medium and high levels of education show positive and significant coefficients in most specifications, indicating a robust and consistent influence on income. For fathers, the effects are more modest: medium education is positive and significant in the baseline model and in most subsamples, while high education, although positive, does not consistently reach statistical significance. Thus, both parents’ education matters, but the impact of the mother’s education proves to be more consistent. The effects of the parents’ occupations are less stable: the coefficients for the father’s sector are significant only for some subsamples. For the mother, the Finance, Insurance, Real Estate, Services or Public Sector is positively and significantly associated with income in the basic model and some sub-samples, but not among respondents aged over 54. The number of books in the household during childhood serves as a robust indicator of cultural capital, with both the 10–100 and over-100 categories displaying positive and significant coefficients across all subsamples, though effect sizes vary slightly. Country-level effects are also stable, with consistent coefficients in sign and order of magnitude. Countries such as Slovenia, Estonia or the Czech Republic show strong and significant positive differences compared to Romania in all specifications, confirming the robustness of the results. These robustness checks show that the main explanatory variables—gender, parental education, cultural capital, and differences between countries—remain significant and consistent regardless of the subsample or age segment considered.

The analysis for Model 2, which extends the previous specification by including effort-related variables, also indicates that the results are robust (see Table S2). It is worth noting that the set of circumstance variables shows stable patterns: gender remains negative and highly significant across all subsamples, place of birth is consistently insignificant, while parental education displays mixed effects, with the mother’s education more robust than the father’s education. Country effects are likewise stable across specifications. Turning to effort-related variables, the respondent’s own education is a stable determinant. Medium levels of education are positive and significant in most specifications, while higher education is strongly positive and consistently significant, particularly for the younger group. Employment sector also matters: the public sector shows stable positive effects across subsamples. Experience contributes robustly as well, with longer tenure consistently associated with higher income, though effects weaken in the 54+ group. Foreign language proficiency exerts a strong and stable positive influence across subsamples, while technical skills appear less robust. Overall, the robustness analysis confirms that key effort variables—education, employment sector, experience, and language skills—exert consistent and significant effects on income, strengthening the credibility of the findings.

Sensitivity analysis using the Oster test for the two OLS models supports our results (see Table S3). For gender, the high δ values indicate the stability of the estimates. In practical terms, this means that it would take very strong omitted variables to overturn the negative relationship. The effect, therefore, remains stable even after adding further controls. On the other hand, for parental education, we found a moderate robustness. For example, for father’s education (medium or high), the test indicates that the results of the model may be more vulnerable to the omission of variables. In the case of maternal education, δ is higher in Model 1, but decreases in Model 2, suggesting a higher sensitivity of the coefficients to the model specification. Overall, these results indicate that the relationship between gender and income is robust, while the effects associated with parental education should be interpreted with more caution as they are potentially affected by omitted variables.

The robustness analysis of the multinomial logit model for medium versus low education shows an overall stability of the main coefficients in different random subsamples and age group subsamples (see Table S4). The coefficient for gender remains significant in most subsamples and for the 18–54 age group, indicating the effect is robust to sample variation. Place of birth and the number of books at home show consistent and significant coefficients in all subsamples, confirming the stability of these key variables. Father’s education coefficients remain significant and consistent across subsamples and age groups, demonstrating high robustness. Maternal education shows more variability: coefficients for the medium level remain generally significant, while those for the high level often become insignificant in certain subsamples and age groups. This pattern is consistent with the baseline model, suggesting the effect of maternal education is more fragile. Coefficients for the parents’ occupation sectors remain insignificant across subsamples or age groups.

For high versus low education, the coefficients for gender remain insignificant in all subsamples and age groups, with no changes in sign (see Table S5). Place of birth retains its significance and direction in most subsamples and age groups, also suggesting robustness. Parents’ education coefficients as well as those of the number of books at home remain significant and consistent across subsamples and age groups, demonstrating high stability. Country coefficients show some variations in significance, but the sign remains generally stable, indicating that the results are robust, even if the statistical significance differs between subsamples. Therefore, the model for high versus low education also demonstrates strong stability and robustness of the main coefficients.

While potential endogeneity cannot be fully ruled out due to the lack of suitable instruments or panel data, the robustness and sensitivity analyses indicate that the main results are stable. Key determinants, including gender, parental education, cultural capital, and country differences, retain their direction and, in most cases, their statistical significance across subsamples, age groups, and model specifications, providing confidence in the reliability of the findings.

5. Conclusions

Inequality of opportunity is a critical issue with significant implications for socio-economic development. Understanding income and educational level variations is becoming increasingly important for substantiating public policies and reducing existing discrepancies. Our study assessed the extent to which initial circumstances and individual effort contribute to inequality of opportunities in income and education, providing an integrated picture of the factors that generate these differences in Central and Eastern European countries.

The estimation of inequality of opportunity in income indicates that circumstances beyond the individual’s control drive 11.96% of income inequality. The dominant factor is gender, responsible for 48.4% of the inequality of opportunities, which confirms the persistence of significant wage differences between women and men. Parents’ education is another major determinant, with a greater influence of maternal education (15.44%) than paternal education (14.13%), and the mother’s sector of activity has a net higher impact than that of the father. The socio-economic status of the family of origin, approximated by the number of books in childhood, contributes 7.42%, while place of birth and country have low shares. The analysis by country shows high levels of inequality of opportunity in Latvia, Hungary, and Estonia; moderate levels in Bulgaria, Croatia, Czechia, Slovakia, and Slovenia; and low levels in Lithuania, Poland, and Romania.

The results of the OLS model estimation indicate that initial circumstances alone explain 32.4% of the income variation, confirming the hypothesis that background factors drive a considerable part of inequality in income. Gender remains the strongest determinant, with a significant negative effect on women, while parental education, especially the mother’s education, and the number of books in childhood, clearly influence income. The analysis by country highlighted important differences: Slovenia, Estonia, and the Czech Republic have significantly higher average incomes than Romania, while Bulgaria and Hungary are the only ones with values lower than Romania.

The extension of the model by including variables that measure individual effort confirms the robustness of the circumstances’ effects but also highlights the decisive role of personal effort. The respondent’s education has a strong positive impact, enhanced by working in high-value-added sectors like Finance, Insurance, Real Estate or Services. Work experience and self-evaluated skills, especially language skills, bring clear income benefits. Recent participation in training programs is associated with lower income, probably reflecting the characteristics of the participants: early career or professional transitions.

These results point out that an essential part of the income inequality in CEE countries is driven by initial circumstances, particularly the gender and education of parents, which calls for targeted policies. The priority is to reduce gender pay gaps and expand access to quality education for children from disadvantaged backgrounds. In the specific context of CEE labour markets, this implies a stronger focus on pay equity legislation, improved childcare support to facilitate female labour force participation, and measures to reduce occupational segregation by encouraging women’s entry into technical and high-growth sectors, which can in turn help reduce structural imbalances and foster more equitable labour market outcomes across the region. Supporting early cultural capital and stimulating high-value-added sectors can generate positive intergenerational effects. The differences between countries indicate the need to adapt policies to the national context, with structural interventions where the IOp is high and the strengthening of existing mechanisms in countries with low IOp. The results of the regression models also emphasise the role of individual effort—education, experience, skills—suggesting that training programs correlated with labour market demand may ensure a more rapid conversion into income benefits. For the CEE region, this requires better alignment between vocational training systems and the skill demands of emerging industries, as well as targeted support for workers in regions facing economic restructuring or higher unemployment risks.

The inequality of opportunity in education arises at 0.2835, suggesting that circumstances like family background or socio-economic condition in childhood explain a considerable proportion of the variation in an individual’s educational attainment. The Shapley decomposition pointed to parental education as the most important factor, followed by the existence of books at home and by the mother’s occupational sector. Gender, place of birth, the country and the father’s occupational sector contribute less than 10% each. Background educational factors (measured by paternal education and the availability of books in childhood) contribute more than 65% to the inequality of opportunities in children’s educational attainment. The Oaxaca decomposition indicated that in Romania, the inequality of opportunity in education is the highest, while in the Czech Republic, it is the smallest. However, in Hungary, Slovakia, and Poland, the inequality of opportunity in education was also high (values between 0.35 and 0.40). On the other hand, in Latvia, Lithuania and Estonia, the inequality of opportunity is almost the same, around 0.25.

The results of the multinomial logit regression confirm previous findings, indicating that parental education has the strongest impact on inequality of opportunity: a higher level of education increases the child’s probability of obtaining medium or higher education, the effects being more pronounced for higher education. The findings suggest that paternal employment in Agriculture, Manufacturing, or Mining is associated with a higher likelihood of children attaining medium or high education, whereas maternal employment in Trade, Services, or the Public sector is positively linked to children achieving higher education levels. An interesting aspect that arose from the regression is the gender influence: gender disparities mostly manifest in the medium education level, not tertiary education, with women being less likely than men to reach the medium level. Regarding the place of birth, it seems that people born in rural areas face persistent disadvantages. The presence of books in the household during childhood is positively associated with achieving both medium and higher levels of education, indicating that an early exposure to reading can influence a child’s academic skills and motivation. The analysis indicates that patterns vary considerably across Central and Eastern European countries.

The strong influence of parental education and early cultural resources highlights the need to reduce background-related disparities from an early age. Educational reforms in CEE countries should focus on expanding access to quality preschool in disadvantaged and rural areas, reducing funding gaps and rural–urban divides, and ensuring more equitable resource allocation. In addition, vocational training and dual-education systems aligned with labour market demand can support students from less advantaged backgrounds. Initiatives promoting reading and access to learning resources through libraries and book distribution can further strengthen opportunities and help reduce inequalities in educational attainment.

Our findings enrich the existing literature on inequality of opportunity, indicating that approximately 12% of income inequality in the CEE countries is attributable to circumstances beyond individuals’ control, with notable regional heterogeneity, similar to Marrero and Rodríguez (2012), who also found high inequalities of opportunity in Hungary and Estonia, and lower values in Czechia and Slovakia. Compared to other contexts, the share of inequality explained by circumstances is relatively low: estimates range from 17–24% in South Africa (Piraino, 2015), 10–37% in Brazil (Bourguignon et al., 2007), around 30% in Egypt (Hassine, 2012), and between 33–36% in the US (Niehues & Peichl, 2014). These comparisons suggest that, while inherited circumstances matter in CEE, their impact is smaller than in many non-European countries. The study also makes a theoretical contribution by combining the estimation of IOp indices, Shapley and Oaxaca decomposition, and econometric methods into a unified framework. This approach makes it possible to quantify both the share of initial circumstances in income and education inequality and the distinct influence of each factor. At the practical level, the results provide a robust foundation for supporting targeted policies in education, employment, and pay equity while underscoring the need to tailor them to the specific contexts of each CEE country.

As a limitation of the study, it should be noted that the methods used to estimate the inequality of opportunity provide a lower-bound estimate of the real level of inequality of opportunity, an aspect also mentioned in the Methodology Section. In addition, due to the small sample size for each individual country, estimations were carried out at an aggregate level for all 11 CEE countries. At the country level, only the results from the Oaxaca decomposition were presented, not country-specific models. Furthermore, the use of cross-sectional LiTS data restricts the analysis to a single point in time, preventing us from directly assessing the extent to which inequalities in income and education are transmitted across generations, and it may also involve a degree of selection bias due to non-random survey participation. Another limitation relates to potential endogeneity, which could not be fully addressed because the dataset does not provide suitable instrumental variables or panel data. As a result, causal interpretations should be treated with caution, since temporal relationships and unobserved heterogeneity cannot be adequately controlled.

Despite these limitations, the multi-method approach employed in this study provides robust and valuable insights into inequality of opportunity in Central and Eastern Europe and can serve as a basis for subsequent research. Future studies may focus on including more variables to better explain the determinants of income and educational attainment, replicating the analysis using the new wave of LiTS data to capture changes over time, or expanding the analysis at the EU level to identify more cross-country patterns.