Results from international large-scale assessments of adults’ skills show that individuals with certain socio-demographic characteristics are more likely to have high levels of 21st-century skills than their peers. Young adults, adults with tertiary-educated parents, adults with parents who worked in high-status occupations and adults raised in cities consistently outperform their peers in literacy, numeracy and adaptive problem-solving. Gender disparities are domain-specific: women have higher proficiency in literacy whereas men have higher proficiency in numeracy and adaptive problem solving. Skills differences are widest at the tails of the distribution, suggesting “sticky floors” for disadvantaged groups and “glass ceilings” for high-achieving women. Breaking the analysis of skills disparities down by socio-demographic characteristics reveals that the nature of disadvantage is multi-layered. International comparisons show that inequalities related to gender and income are linked to skills disparities within countries. These disparities are already present among school-aged children, suggesting that future generations may face similar challenges to today’s adults.
OECD Skills Outlook 2025
Abstract
This chapter reviews evidence on skills disparities in OECD countries across five dimensions: gender, socio-economic background (through indicators of parental education and occupation), age, immigrant background and childhood residential context. It draws on international large-scale educational assessments to identify disparities among adults and young people across achievement levels, countries and time. Disparities exist not only in the core 21st-century skills of literacy, numeracy and adaptive problem solving, but also in ancillary social and emotional skills and willingness to delay gratification. Socio-economic background is the most consistent driver of disparities, with advantages compounding over time. Context matters, with societal level inequality, place of residence and family resources shaping outcomes. Immigrant background gaps often narrow once language and family factors are considered, while gender gaps are domain-specific, favouring women in literacy and men in numeracy and adaptive problem solving. Patterns observed among adults are already present among school-aged children, suggesting that today’s disparities are likely to persist without intervention. Many forms of disadvantage compound, with certain forms of disadvantage amplifying the effects of other disadvantages. Understanding these patterns is essential for designing targeted policies to narrow skills gaps. The key findings explored in this chapter include:
Disparities by socio-economic background (parental education and occupation):
Family socio-economic background is the strongest and most pervasive predictor of adult skills disparities. For example, in every country, adults with tertiary-educated parents have higher skills than those whose parents are not tertiary educated: differences correspond to 0.55 standard deviation (SD) difference in literacy, 0.53 SD in numeracy and 0.55 SD in problem solving.
Socio-economic disparities in core 21st-century skills are wider at the bottom of the skills distribution. The advantage in literacy for adults with at least one tertiary-educated parent corresponds to 0.44 SD among the highest achievers but 0.73 SD among the lowest achievers.
Socio-economic disparities in core 21st-century skills are wider in countries with greater income inequality. For example, in Chile, total income disparities are relatively large (0.45 for Gini1), and skills differences related to parental occupation are also wide (0.7 SD for literacy). In contrast, in Poland, incomes are more equally distributed (Gini 0.27), and skills differences related to parental occupation are relatively contained (0.4 SD for literacy).
Among 15-year-olds, socio-economic disparities remain substantial, although narrowed in the first quarter of the 21st century. For example, the difference in mathematics between students with and without tertiary-educated parents shrank from 0.47 SD to 0.37 SD between 2003 and 2022.
Disparities between men and women:
Gender gaps in 21st-century skills are domain specific. On average, adult women outperform men in literacy (0.04 SD), but men outperform women in numeracy (0.17 SD) and adaptive problem solving (0.06 SD). In 30 out of 31 countries, men outperform women in numeracy.
Gender gaps vary by achievement level. Men’s advantage in numeracy is especially strong among top performers.
Between-country differences in gender gaps in numeracy reflect the level of gender inequality present in a society. More inequality is associated with smaller numeracy gender gaps in favour of men.
The magnitude of gender gaps differs depending on other characteristics. For example, the gender gap in numeracy favouring men is larger among adults with tertiary-educated parents (0.20 SD) than among adults with non-tertiary educated parents (0.16 SD).
Among teenagers, boys and girls have different strengths. Boys have higher achievement than girls in creative problem solving (0.08 SD), financial literacy (0.05 SD), computational thinking (0.04 SD) and mathematics (0.10 SD). Girls have higher achievement than boys in collaborative problem solving (0.29 SD), global competence (0.26 SD), creative thinking (0.25 SD), civic knowledge (0.22 SD), computer and information literacy (0.20 SD), and reading (0.24 SD).
Age:
Information-processing skills peak in young adulthood and generally decline with age. This pattern is widespread among OECD countries – with the exception of Sweden and New Zealand. On average, 16-29 year-olds outperform 50-65 year-olds by about 0.48 SD in literacy, 0.36 SD in numeracy and 0.57 SD in adaptive problem solving.
Age disparities are widest at the bottom tail of the achievement distribution. For example, the difference in literacy between 16-29 year-olds and 50-65 year-olds corresponds to 0.56 SD at the 10th percentile but 0.41 SD at the 90th percentile.
Immigrant background:
On average, the children of immigrants have lower skills than individuals with similar background characteristics but without an immigrant background. Differences correspond to 0.14 SD in literacy, 0.14 SD in numeracy and 0.11 SD in adaptive problem solving. However, these differences vary greatly across countries.
Childhood residential context:
Disparities in core 21st-century skills by childhood residential context are significantly in favour of city dwellers. Whether individuals grew up in a village, town or city shapes the opportunities they have to develop and maintain their skills over the life course. On average, adults who grew up in cities have higher literacy (0.20 SD), numeracy (0.18 SD) and adaptive problem solving (0.17 SD) than those who grew up in villages.
The urban–rural gap in 21st-century skills is largely driven by differences in the socio-economic background of individuals across different residential contexts. When comparing individuals with similar parental circumstances, the city–village gaps shrink to 0.08 SD in literacy, 0.06 SD in numeracy and 0.05 SD in adaptive problem solving.
Among 15-year-olds, disparities related to school location have slightly increased over time. The difference in mathematics achievement between young people who attended school in a city and those who attended school in a village grew from 0.22 SD in 2003 to 0.26 SD in 2022.
1. The Gini coefficient measures the extent to which distribution (for example of income) among individuals or households within an economy deviates from a perfectly equal distribution. A Gini index of 0 represents perfect equality, while an index of 100 implies perfect inequality.
Crises and megatrends have intensified the importance of individuals possessing the necessary skills to adapt, innovate and thrive in rapidly changing work and social environments, with core 21st-century skills – literacy, numeracy and adaptive problem solving – becoming increasingly important for current and emerging job roles (see Chapter 1, Section 1.1). However, evidence suggests that these skills are unevenly distributed across populations, reflecting broader inequality in which constraints and outcomes reinforce each other over time, allowing gaps to become entrenched across the life course. This unequal distribution limits the ability of certain groups to fully participate in and benefit from modern economic and civic life.
This chapter reviews existing evidence on how 21st-century skills vary by a range of socio-demographic characteristics: gender, parental education, parental occupation, immigrant background, childhood residential context and age. Drawing on data from the 2023 Survey of Adult Skills, the Programme for International Student Assessment (PISA), the International Computer and Information Literacy Study (ICILS), and the International Civic and Citizenship Education Study (ICCS), it offers a detailed assessment of the magnitude and nature of these skill gaps among adults and youth. Together, the findings shed light on both persistent and emerging patterns in the distribution of 21st-century skills, and explores how these gaps can be narrowed through targeted efforts within and beyond the education system.
Individual chances of skills formation are strongly influenced by socio-demographic characteristics (Hanushek and Woessmann, 2011[1]; Van de Werfhorst and Mijs, 2010[2]); however, the strength of these associations varies across countries, indicating that national characteristics and policy choices can shape the extent to which individuals’ inherited circumstances shape their opportunities to develop their skills.
2.1.1. Disparities by parental education and occupation
Research suggests that parents influence their children’s cognitive and social and emotional development (Cunha and Heckman, 2007[3]; Demange et al., 2022[4]; England-Mason and Gonzalez, 2020[5]; Grusec and Davidov, 2019[6]; Guhin, Calarco and Miller-Idriss, 2021[7]; Leerkes, Bailes and Augustine, 2020[8]). Social and emotional skills can be transmitted across generations (Attanasio, de Paula and Toppeta, 2024[9]), reinforcing social stratification, i.e. how societies categorise individuals into groups based on socio-demographic characteristics (Farkas, 2003[10]; Gruijters, Raabe and Hübner, 2023[11]). Children of parents with higher educational qualifications or who work in high-status occupations (defined in this chapter as managers, professionals, technicians and associated professionals) typically exhibit higher levels of information-processing skills and different behavioural tendencies. These disparities emerge early and persist through adolescence into adulthood, although the evolution varies by country context (Borgonovi and Pokropek, 2021[12]; Dickson, Gregg and Robinson, 2016[13]). At the same time, countries differ with respect to whether initial differences grow or shrink as individuals leave school and enter highly differentiated learning opportunities in further education, training or the labour market (OECD, 2021[14]).
Parental socio-economic status shapes both objective resources and young people’s educational and career expectations (Breen and Goldthorpe, 1997[15]; Bodovski, 2013[16]; Lareau, 2011[17]). Stimulating home environments, characterised by the presence of educational resources and activities, promote early advantages in cognitive development, creating cumulative advantages that grow over time (Lareau, 2011[17]). High parental expectations, strategic residential choices and social networks further reinforce these advantages among families with socio-economically advantaged backgrounds (Chetty et al., 2022[18]; Owens, Reardon and Jencks, 2016[19]; OECD, 2024[20]).
Adolescents have been shown to evaluate educational decisions in light of perceived risks and benefits. Those from disadvantaged backgrounds perceive higher risks and often lower their expectations despite good academic performance, whereas advantaged youth maintain high aspirations regardless of performance (Bernardi and Valdés, 2021[21]). As a result, disadvantaged students respond more strongly to signals about their likelihood of academic success, frequently opting for lower-risk educational tracks (Holm, Hjorth-Trolle and Jæger, 2019[22]). All these factors contribute to the widening of initial disadvantage over time, as small differences in circumstances in childhood, in the absence of compensatory factors, tend to compound over time.
Adults from families with socio-economically advantaged backgrounds display higher levels of information-processing skills in all countries (Figure 2.1, Panels C and D). Differences are medium-sized: the difference between adults with and without tertiary-educated parents corresponds to 0.55 SD in literacy, 0.53 SD in numeracy and 0.55 SD in adaptive problem solving. Differences related to parental occupation are similar: the difference between adults with and without parents in high-status occupations corresponds to 0.51 SD in both literacy and numeracy, and 0.49 SD in adaptive problem solving. Differences related to parental education and parental occupation are widespread: adults with a more advantaged background have higher levels of information-processing skills than their peers with a less advantaged background in all countries. Differences by parental education are widest in Portugal for literacy (0.86 SD), numeracy (0.80 SD) and adaptive problem solving (0.84 SD), and differences by parental occupation are widest in Portugal for literacy (0.7 SD), and in Chile for numeracy (0.68 SD) and adaptive problem solving (0.67 SD).
Over the past century, the make-up of families has changed significantly. Box 2.1 looks at how parental education and occupation have changed across generations and the impact this has had on the distribution of family resources in OECD countries.
Average, country min., country max., and number of countries with positive score-point difference, OECD average
Note: Adults aged 16-65. All average score-point differences in Panels A–D are statistically significant at the 5% level. The triangle marks the country with the largest score-point difference between the respective groups. The diamond marks the country with the smallest score-point differences between the respective groups. Numbers in square brackets next to skills indicate the number of countries with positive score-point differences. Parental education (at respondents’ age 14) is based on the International Standard Classification of Education (ISCED) 2011 and distinguishes between adults with at least one tertiary-educated parent (ISCED 2011 5, 6, 7 and 8) and those with no tertiary-educated parent. Parental occupation (at respondents’ age 14) is based on the International Classification of Occupations (ISCO) and grouped into high-status: managers, professionals, and technicians and associate professionals (ISCO 1-3); and low-status: clerical support workers; service and sales workers; skilled agricultural, forestry and fishery workers; craft and related trades workers; plant and machine operators, and assemblers; and elementary occupations (ISCO 4-9). Country-specific results are provided in Tables 2.A.2.1, 2.A.2.2, 2.A.2.3, and 2.A.2.4 in Annex 2.A.
Source: Calculations based on OECD (2024[23]), Survey of Adult Skills (PIAAC) 2nd cycle database, www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html.
In this report, a family’s socio-economic background is measured through indicators of the highest educational attainment and occupational status of parents, comparing adults with and without at least one parent with a tertiary-education degree and adults with and without at least one parent working as a professional or manager. However, considering only the educational qualification and occupational status of the most advantaged parent fails to distinguish between individuals with and without the added advantage of having both parents with tertiary qualifications or working in high-status occupations.
Over the past century, the make-up of families has changed significantly. Only 7% of adults aged between 56 and 65 in 2023 (those born between 1958 and 1967) have both parents educated at the tertiary level, whereas this share is 29% among those aged between 16 and 25 (those born between 1998 and 2007) (Table 2.A.1.1 in Annex 2.A). Conversely, the share of adults with both parents who achieved, at most, a lower-secondary qualification is 40% among 56-65 year-olds but only 9% among 16-25 year-olds. Among the oldest birth cohort, only 31% of individuals with at least one tertiary-educated parent also have a second parent educated at the tertiary level, compared to 52% in the youngest birth cohort.
Changes in educational attainment over time are mirrored by changes in occupational status: women’s increased participation in education has been accompanied by increased participation in the labour market, with many employed in higher-status occupations such as managers and professionals (International Classification of Occupations [ISCO 1-2]). The share of adults with both parents in ISCO 1-2 occupations increases across younger age cohorts, with 7% of 56-65 year-olds having both parents in this occupational category compared to 16% of 16-25 year-olds (Table 2.A.1.2 in Annex 2.A). Among the oldest birth cohort, only 20% of individuals with at least one parent working in high-status occupations also have a second parent working in a high-status occupation compared to 34% in the youngest birth cohort.
The significant changes in educational attainment and occupation detailed in Tables 2.A.1.1 and 2.A.1.2 in Annex 2.A are largely driven by increases in the participation of women in tertiary education and their increased employment prospects, rather than by changing preferences in terms of the characteristics men and women look for in a spouse. The share of individuals with parents who obtained similar levels of educational qualifications and the share of individuals whose parents worked in a similar type of occupation remain relatively stable across birth cohorts. The biggest changes are visible for adults who have tertiary-educated mothers among those whose fathers are also tertiary educated. This share corresponds to 38% of 56‑65 year-olds and increases to 71% of 16-25 year-olds (Panel A in Figure 2.2). Similarly, the share of adults with mothers who worked as managers and professionals among those whose fathers also worked as managers and professionals increased from 38% among 56-65 year-olds to 53% among 16-25 year-olds (Panel B in Figure 2.2). Meanwhile, few changes are visible among adults with tertiary-educated fathers whose mothers are also tertiary educated, and among adults with fathers working as managers and professionals whose mothers are also managers and professionals.
Total students in OECD countries, with countries weighted by population, OECD average
Note: Panel A: The category “Total share of parents with matching educational qualifications across all levels” denotes the share of respondents with cumulative share of parents with the same educational qualification across all educational levels. The category “Tertiary-educated mothers among those with tertiary-educated fathers” denotes the share of respondents with tertiary-educated mothers and fathers among respondents with tertiary-educated fathers. The category “Tertiary-educated fathers among those with tertiary-educated mothers” denotes the share of respondents with tertiary-educated fathers and mothers among respondents with tertiary-educated mothers. The category “Both parents tertiary-educated among those with at least one tertiary-educated parent” denotes the share of respondents with tertiary-educated mothers and fathers among those with tertiary-educated mothers or tertiary-educated fathers.
Panel B: The category “Total share of parents with matching classification of occupation” denotes the share of respondents with cumulative share of parents with the same classification of occupation. The category “Mothers as managers and professionals among those with fathers as managers and professionals” denotes the share of respondents with mothers and fathers as managers and professionals among respondents with fathers as managers and professionals. The category “Fathers as managers and professionals among those with mothers as managers and professionals” denotes the share of respondents with fathers and mothers as managers and professionals among respondents with mothers as managers and professionals. The category “Both parents managers and professionals among those with at least one parent as a manager or professional” denotes the share of respondents with mothers and fathers as managers and professionals among those with mothers as managers and professionals or fathers as managers and professionals.
Source: Calculations based on OECD (2024[23]), PIAAC 2nd cycle database, www.oecd.org/en/about/programmes/piaac/piaac-data.html.
These changes, due to assortative mating and increases in female educational attainment, have profound implications for education systems and skills acquisition. Resources available in households where both parents are tertiary educated and earn incomes from higher-status occupations are very different from the resources available in families where neither is the case. There are also challenges to evaluating trends over time, as whereas historically, many individuals with at least one tertiary-education parent had only one such parent, now most have two. This creates difficulties for direct comparisons of skills levels by parental education. However, it also suggests that the depth of advantage potentially experienced by children with socio-economically disadvantaged backgrounds relative to that of more advantaged households has deepened over time, and resource differences are likely to have grown. Therefore, compensatory measures that might have been sufficient and effective in the past may no longer be equally effective today or in the future.
Findings also indicate that the composition of households according to parental education between adults who grew up in cities and those who grew up in villages has become more pronounced over time. For example, among 56-65 year-olds in 2023 (born between 1958 and 1967), 11% of those who grew up in cities had both parents educated at the tertiary level compared to only 4% of those who grew up in villages, a difference of 7 percentage points (Table 2.A.1.3 in Annex 2.A). This difference was as large as 15 percentage points among adults born between 1998 and 2007. Hence, findings suggest that long-term trends in the concentration of individuals with tertiary-level qualifications in urban centres have exacerbated disadvantage in rural communities.
2.1.2. Disparities by gender
Skills disparities between men and women are generally modest (Hyde, 2014[24]) but can still be consequential to the extent that they shape educational paths and career choices. Girls tend to outperform boys in verbal/language-related tasks, leading to better reading achievement, while boys often have a small edge in spatial and mathematical tasks. Gender differences evolve with age, reflecting educational and occupational trajectories (Rebollo-Sanz and de la Rica, 2020[25]; Borgonovi, 2022[26]; OECD, 2019[27]). The advantage of girls and women in literacy peaks in adolescence and diminishes by adulthood, while the advantage of boys and men in numeracy steadily increases (Borgonovi, Choi and Paccagnella, 2021[28]; Solheim and Lundetræ, 2016[29]; Lundetræ et al., 2014[30]; Rosdahl, 2014[31]; OECD/Statistics Canada, 2005[32]). Men and boys are more likely to report being willing to choose larger delayed rewards over smaller immediate ones than women and girls (Falk and Hermle, 2018[33]), but in practice they are less likely to actually choose the larger delayed rewards over the smaller immediate rewards (Silverman, 2003[34]).
Starting from an early age, boys and girls are often socialised into different skill-building experiences, reinforcing gender stereotypes and influencing skill development trajectories: girls are often encouraged to read and express emotions, whereas boys may receive toys that promote the development of the prerequisites of mathematics skills (OECD, 2015[35]). Gender stereotypes can also undermine young people’s confidence in counter-stereotypical domains. Adolescent girls, for instance, often underestimate their mathematical ability despite high performance, affecting their further skill development (Huang, 2012[36]). School environments and broader social structures can magnify or reduce gaps (OECD, 2015[35]).
Differences between men and women in core 21st-century skills vary depending on the skill considered (Figure 2.1, Panel B). For example, while on average women outperform men in literacy by 0.04 SD, men score higher than woman in numeracy and adaptive problem solving (0.17 SD and 0.06 SD, respectively). Across countries, women outperform men in 9 out of 31 countries in literacy. In numeracy, men outperform women in all countries except Croatia, and the gender gap in numeracy is largest in Switzerland at 0.29 SD. In adaptive problem solving, men outperform women in 12 out of 31 countries.
2.1.3. Disparities by age
Skills disparities evolve over the life course, influenced by formal, non-formal and informal learning, life experiences, and cohort-specific socio-economic contexts. Age also moderates the effects of gender, family background and migration (Borgonovi et al., 2017[37]; Rebollo-Sanz and De la Rica, 2020[38]). Information-processing skills peak in early adulthood and typically decline thereafter, and more sharply for lower-educated individuals. Longitudinal studies indicate that skills use at work and home mitigates age-related decline, particularly among highly educated individuals (Hanushek et al., 2025[39]).
On average, 16-29 year-olds have higher levels of core 21st-century skills than 50‑65 year‑olds. Differences are small- to medium-sized when using “Cohen’s d benchmarks”1 but large when considering the standardised difference in achievement over the course of a school year among PISA participants (around 0.20 SD) (Avvisati and Givord, 2023[40]), or the range of effects on achievement of educational interventions (Kraft, 2020[41]): the difference corresponds to 0.48 SD for literacy, 0.36 SD for numeracy and 0.57 SD for adaptive problem solving (Cohen, 2013[42]). Younger groups also report being more willing to give up something beneficial today in order to benefit more from it in the future (0.35 SD difference) (Figure 2.1, Panel A). Differences in information-processing skills related to age favour young adults in virtually all countries. The only exceptions are New Zealand, where older adults score higher in literacy, and New Zealand and Sweden, where they score higher in numeracy, although these differences were not statistically significant.
2.1.4. Disparities by immigrant background
Evidence on adult populations suggests that in most countries, immigrants or the children of immigrants, i.e. those born in a country to migrant parents or who migrated when under 18, have lower levels of information-processing skills than individuals without an immigrant background (OECD, 2016[43]). The unadjusted difference related to immigrant background is 0.14 SD for literacy, 0.13 SD for numeracy and 0.10 SD for adaptive problem solving (Figure 2.3, Panel A). After accounting for parental education and parental occupation, the differences are 0.14 SD for literacy and numeracy and 0.11 SD for adaptive problem solving.
In OECD countries, the educational attainment of migrants has significantly increased over recent decades, reflecting global increases in educational attainment (Barro and Lee, 2013[44]) and selective immigration policies (OECD, 2024[45]). Attracting, selecting and retaining migrants with skills adapted to the host-country labour market has become a key policy objective in many OECD countries, and differences in migrant composition explain a large part of between-country differences in the set of information-processing skills of migrant populations (OECD, 2018[46]). Language barriers, particularly for those migrating as adults, significantly impact their ability to express their information-processing skills’ potential, with proficiency declining notably for those who migrated after the age of 12 (Cathles et al., 2021[47]; OECD, 2018[46]). Box 2.2 details the role of linguistic distance in people’s skills.
Adjusted and unadjusted difference in literacy, numeracy and adaptive problem solving, by immigrant background and childhood residential context, OECD average
Note: Adults aged 16-65. Unadjusted differences are the differences between the two averages for each contrast category. Adjusted differences are based on a regression model that takes into account differences associated with parental education and parental occupation. All adjusted and unadjusted differences are statistically significant at the 5% level. Childhood residential context (at respondents’ age 14) refers to whether the respondent grew up in villages, towns or cities. Groups by immigrant background distinguish between children of immigrants and non-immigrants. Children of immigrants were born in the country in which they currently reside, but their parents were not, or they were born in a different country and moved to their current country of residence before the age of 18. Non-immigrants were born in their current country of residence, as were their parents. Country-specific results are provided in Tables 2.A.2.5 and 2.A.2.6 in Annex 2.A.
Source: Calculations based on OECD (2024[23]), Survey of Adult Skills (PIAAC) 2nd cycle database, www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html.
Chapter 1 details how in the 21st century, individuals must possess a broad skillset that includes social and emotional skills and the capacity to delay gratification, alongside information-processing skills. By examining a range of skills, it is possible to consider how a certain personal characteristic may act as a barrier to skills proficiency in one domain (e.g. literacy) but a potential advantage in another (e.g. numeracy or adaptive problem solving). Recognising these interconnections highlights the multifaceted nature of human potential and reveals pathways for more inclusive learning and employment practices.
A holistic perspective on skills development also encourages the recognition of the strengths of populations that are usually disadvantaged in labour markets and societies. For instance, factors such as linguistic distance – the degree of dissimilarity between an individual’s native language (L1) and second language (L2) – might hinder proficiency in acquiring a new language that is very different from the language an individual already knows, but it could also spur cognitive adaptations that foster greater flexibility in other areas (Isphording, 2014[48]; Paradis, 2011[49]). Linguistic distance influences not only the direct acquisition of linguistic skills but also amplifies cognitive challenges during L2 usage, often referred to as the “foreign language effect” (Takano and Noda, 1995[50]). Appreciating such compensatory mechanisms can help policymakers, educators and employers better tap into underutilised talents across society. Ultimately, understanding a broader range of skills enables the design of interventions and opportunities that nurture and showcase the full spectrum of human capabilities, benefiting not only the individuals involved but also the communities and organisations they belong to.
Language proficiency in a consciously acquired second language can be significantly affected by how dissimilar the new language is compared to the language an individual already knows in terms of pronunciation, grammar, script and vocabularies. Greater dissimilarity between L1 and L2 generally leads to increased difficulty in linguistic processing (understanding L2 as it is being heard/read). This leads to cognitive fatigue, because greater cognitive resources are needed for linguistic processing. Therefore, cognitive resources are less likely to be mobilised for non-linguistic information processing (Isphording, 2014[48]; Takano and Noda, 1995[50]). Consequently, individuals facing greater L1 to L2 dissimilarity may experience compounded difficulties in both using L2 and performing cognitively demanding tasks in L2 (Norman and Bobrow, 1975[51]). As such, linguistic distance emerges as a significant factor influencing both linguistic and broader information-processing skills (Borgonovi and Ferrara, 2020[52]; Isphording, 2014[48]; Kuperman, 2022[53]).
Although linguistic distance is associated with poorer literacy outcomes (Borgonovi and Ferrara, 2020[52]; Isphording, 2014[48]; Kuperman, 2022[53]), its effects on other competencies can be more nuanced. For instance, Kuperman (2022[53]) examined the impact of linguistic distance on the outcomes of Canadian immigrants. While greater linguistic distance was associated with lower literacy scores, numeracy skills were unaffected. One explanation for this is that numeracy tends to be more “language independent” than literacy, with existing numeracy skills more readily transferred to a new linguistic environment, whereas proficiency in literacy is more closely tied to the specific language in which they were originally developed (Kuperman, 2022[53]).
Learning a second language has been linked to structural anatomical changes in the brain (Li, Legault and Litcofsky, 2014[54]). For example, bilingual experiences can produce cross-domain effects, enhancing both linguistic abilities and general cognitive functions (Lehtonen et al., 2018[55]; Li, Legault and Litcofsky, 2014[54]). Even among older adults, learning a second language has been linked to improvements in attentional switching, inhibition, working memory and increased functional connectivity (Ware et al., 2021[56]). A meta-analysis by Nucette et al. (2024[57]) explored the correlation between foreign language instruction and mathematical skills in young adolescents, highlighting that language learning could positively influence proficiency in numeracy. Therefore, although individuals with larger linguistic distance may lag in literacy, they may cultivate compensatory skills that benefit them in numeracy and other domains.
The results presented in Figure 2.4 support prior evidence on linguistic distance having compensatory effects on skills other than literacy. Linguistic distance, as measured by Levenshtein Distance Normalised by Divergence,1 shows a significant negative correlation with literacy both before and after accounting for socio-demographic characteristics (-0.26 SD and -0.12 SD, respectively). However, when literacy is held constant, linguistic distance is positively correlated with numeracy after controlling for socio-demographic characteristics (0.02 SD). Moreover, the initially negative relationship between linguistic distance and adaptive problem solving (-0.23 SD) disappears almost entirely (0.01 SD, not significant) once accounting for literacy and background factors. These patterns suggest that while linguistic distance presents initial barriers to literacy acquisition, it may foster other skills, particularly in numeracy, through compensatory mechanisms.
Effect of a one-standard-deviation increase in linguistic distance on core 21st-century skills, OECD average
Note: All coefficients before accounting for socio-demographic characteristics are statistically significant at the 5% level. Filled diamonds indicate that coefficients after adjusting for background characteristics are statistically significant at the 5% level. Socio-demographic characteristics include gender, age, respondents’ educational attainment, educational attainment of parents, place of living at the age of 14, years in country of assessment, whether two languages were learned at home in childhood and still understood, type of immigration pattern, and literacy score (in the models for numeracy and adaptive problem solving). All models include country fixed effects.
Source: Calculations based on OECD (2024[23]), Survey of Adult Skills (PIAAC) 2nd cycle database, www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html.
1. The Levenshtein Distance Normalised by Divergence (LDND) is a measure of linguistic distance developed by the Cross-Linguistic Linked Data project hosted by the Max Planck Institute for the Science of Human History. It provides a systematic approach to quantifying the similarity between languages (Wichmann et al., 2010[127]). The LDND metric is based on the Automated Similarity Judgment Program, which compares the phonetic and lexical features of different languages using the Swadesh list – a standardised set of words with shared meanings among languages. Phonetic transcriptions of the words are analysed using a composite Levenshtein distance, calculated as the minimum number of single character edits needed to transform one language’s transcription into another’s (Bakker et al., 2009[126]). The average distance across all words is then adjusted to produce the final LDND score, which is a continuous measure ranging from 0 to just above 100, where lower scores indicate greater similarity. This approach provides a standardised and empirically validated framework for cross-language comparison that is applicable in various research contexts.
2.1.5. Disparities by childhood residential context
Where individuals live plays an important role in shaping their labour market and social opportunities (Chetty et al., 2014[58]; 2018[59]; OECD, 2025[60]). While cities have historically promoted economic equality, recent decades have seen increasing inequality and diminished opportunities for social mobility in cities (Connor et al., 2025[61]). Place of birth impacts skill development opportunities due to financial, occupational, social and caregiving constraints on mobility, with many adults remaining close to their hometowns despite limited opportunities (Artamonova and Syse, 2021[62]; Ferreira, Gyourko and Tracy, 2011[63]; Ganong and Shoag, 2017[64]; Gobillon and Seold, 2021[65]; Spring, Gillespie and Mulder, 2023[66]).
Regional contexts affect social and emotional skills and academic achievement, independent of parental background. Urban areas typically offer better educational infrastructure and resources, whereas rural areas experience constraints in educational choices and quality (Atherton et al., 2023[67]; OECD, 2017[68]; Rentfrow, Gosling and Potter, 2008[69]; Echazarra and Radinger, 2019[70]). Rural areas also face challenges in early childhood education and care (ECEC), including higher child-to-staff ratios, fewer pro-social teaching practices and limited availability. Urban centres attract more funding, better-qualified staff and more comprehensive resources (OECD, 2019[71]; Echazarra and Radinger, 2019[70]).
In around three out of four countries, adults who grew up in cities have higher proficiency in literacy, numeracy and adaptive problem solving than adults who grew up in villages or rural areas, as indicated in Panel B in Figure 2.3. On average, the unadjusted differences correspond to 0.20 SD in literacy, 0.18 SD in numeracy and 0.17 SD in adaptive problem solving. Differences are smaller when comparing the two groups after controlling for parental education and occupation: 0.08 SD for literacy, 0.06 SD for numeracy and 0.05 SD for adaptive problem solving.
Disparities in social and emotional skills and willingness to delay gratification can be observed across different socio-demographic groups (Box 2.3).
In a world where success increasingly hinges on collaboration, adaptability and self-regulation, social and emotional skills and a willingness to delay gratification complement the core 21st-century skills of literacy, numeracy and adaptive problem solving. Mapping how these skills vary by age, gender, socio-economic background, immigrant background and childhood residential context can highlight disparities that a reliance on traditional achievement metrics miss and equip policymakers with more complete information for the design of interventions. These measures are descriptive and context-dependent and therefore should not be interpreted as fixed traits or value judgements about individuals or groups (see the Reader’s Guide).
Disparities by age: Older adults are more likely to report being emotionally stable and not easily upset (i.e. scoring high on emotional stability) than younger adults (0.12 SD) on average across OECD countries. They are also more likely to be compassionate and assume the best in people (with a difference in agreeableness of 0.24 SD) and are more likely to be reliable (with a difference in conscientiousness of 0.50 SD) (Figure 2.5, Panel A). Younger adults are more likely to be full of energy and to come up with new ideas (with a difference in extraversion of 0.11 SD and a difference in open-mindedness of 0.18 SD).
Disparities by gender: Men score higher than women in emotional stability in all countries (0.40 SD), but women score higher than men in agreeableness (0.34 SD) and conscientiousness (0.15 SD). Gender differences are less pronounced for extraversion and open-mindedness (Figure 2.5, Panel B). Men also report being more willing to delay gratification than women (0.06 SD), with significant differences observed in 15 out of 31 countries (see Table 2.A.2.2 in Annex 2.A).
Disparities by socio-economic background: In all countries, adults with socio-economically advantaged backgrounds score higher on open-mindedness than adults with a socio-economically disadvantaged background (0.32 SD difference on average between individuals with and without tertiary-educated parents, and 0.29 SD difference on average between individuals with parents who worked in high-status rather than low-status occupations). Adults with a socio-economically advantaged background are also more likely to score higher than individuals with a disadvantaged background on extraversion and emotional stability, although differences are less pronounced (Figure 2.5, Panels C and D). Adults with socio-economically disadvantaged backgrounds score higher on conscientiousness than individuals with more advantaged backgrounds. For example, in 13 out of 29 countries, adults with parents who worked in low-status occupations score higher on conscientiousness than adults whose parents worked in high-status occupations (0.07 SD difference on average), and in 20 out of 29 countries, individuals with no tertiary-educated parents score higher on conscientiousness than individuals with tertiary-educated parents (0.13 SD difference on average). Socio-economic differences in agreeableness are small on average (slightly favouring adults with non-tertiary-educated parents by 0.03 SD and favouring adults whose parents worked in high-status occupations by 0.02 SD) and vary across countries. Differences in willingness to delay gratification related to socio-economic background are quantitatively small but pervasive in all participating countries. Adults whose parents obtained tertiary education report a higher willingness to delay gratification than adults whose parents did not (0.21 SD difference). Similarly, adults whose parents worked in a high-status occupation when they were children report a higher willingness to delay gratification than adults whose parents did not (0.18 SD difference).
Average, country min., country max., and number of countries with negative effect sizes, OECD average
Note: All average score-point differences in Panels A–D are statistically significant at the 5% level. The triangle marks the country with the largest effect size between the respective groups and the diamond marks the country with the smallest effect size between the respective groups. The share next to each element considered indicates the percentage of countries with negative effect sizes. See the note for Figure 2.1 for the definitions of groups based on parental education and parental occupation. Country-specific results are provided in Tables 2.A.2.1, 2.A.2.2, 2.A.2.3 and 2.A.2.4 in Annex 2.A.
Source: Calculations based on OECD (2024[23]), Survey of Adult Skills (PIAAC) 2nd cycle database, www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html.
Disparities by immigrant background: Differences in social and emotional skills between individuals with and without an immigrant background are small, but largest when considering open-mindedness. On average, children of immigrants score higher on open-mindedness than individuals without an immigrant background (0.09 SD on average, Figure 2.6, Panel A) – even after controlling for socio-economic background (i.e. parental education and occupation). Empirical evidence suggests that higher levels of open-mindedness correspond with a heightening of the intention to migrate (Canache et al., 2013[72]), with studies on intergenerational transmission of skills, attitudes and behaviours suggesting that children behave similarly to their parents (Dohmen et al., 2011[73]; Liefbroer and Elzinga, 2012[74]; Grönqvist, Öckert and Vlachos, 2016[75]). Therefore, children of immigrants may develop behavioural tendencies associated with higher levels of open-mindedness, which may involve the ability to discern whether challenges could be beneficial (Tucker, 2023[76]). Moreover, growing up in a household of immigrants means that children are likely to be exposed to a range of languages, norms, values, literary texts and worldviews, which may broaden their ability to understand and reflect different perspectives.
Adjusted and unadjusted difference in delayed gratification and social and emotional skills, by immigrant background and childhood residential context, OECD average
Note: Unadjusted differences are the differences between the two averages for each contrast category. Adjusted differences are based on a regression model that takes into account differences associated with parental education and parental occupation. Darker colours denote differences that are statistically significant at the 5% level. See the note for Figure 2.3 for the definitions of groups based on immigrant background and childhood residential context. Country-specific results are provided in Tables 2.A.2.5 and 2.A.2.6 in Annex 2.A.
Source: Calculations based on OECD (2024[23]), Survey of Adult Skills (PIAAC) 2nd cycle database, www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html.
Disparities by childhood residential context: Differences in social and emotional skills related to childhood residential context are small, except for open-mindedness, extraversion and delayed gratification (Figure 2.6, Panel B). Adults who grew up in cities score higher in open-mindedness than adults who grew up in villages (0.17 SD difference, after controlling for parental education and occupation), which is in line with findings from the literature (Militaru et al., 2023[77]). They also score higher on extraversion and delayed gratification (0.06 SD difference each, after controlling for parental education and occupation). These differences do not reflect different socio-economic background but may be due to the fact that cities are characterised by greater population density and higher diversity than villages, meaning that children living in cities are more likely to be exposed to individuals differing from them and from each other culturally, socially and economically (Lee et al., 2025[78]). Furthermore, the higher concentration of institutions such as museums, libraries and availability of cultural events in cities may encourage curiosity and engagement with new ideas among children who grew up in cities rather than villages. Finally, urban environments typically offer more varied educational and extracurricular opportunities (Echazarra and Radinger, 2019[70]), which may foster adaptability and open-mindedness. In 9 out of 10 countries, adults who grew up in cities also report a greater willingness to delay gratification than adults who grew up in villages or rural areas, but this difference is small (0.06 SD after controlling for parental education and occupation).
2.2.1. Disparities in core 21st-century skills: From basic to advanced skills
Average gaps in skills disparities between different groups mask large differences at the extremes of the proficiency distribution in literacy, numeracy and adaptive problem solving. Examining skills disparities among the highest achievers helps to identify the extent to which individuals from different backgrounds who share the talent needed to steer innovations such as artificial intelligence (AI) development of green technologies have been adequately supported in their skills development trajectory. Similarly, examining between-group differences in skills disparities among those with the lowest levels of achievement helps to identify the need for investments in lifelong learning to ensure that all adults, irrespective of their backgrounds, have the basic levels of proficiency needed to perform the core economic and social activities needed for informed participation in society and labour markets. Disparities among the highest achievers point to the success or failure of, for example, tertiary-education institutions and countries’ innovation hubs to ensure equal opportunities for men and women and for individuals from advantaged and disadvantaged backgrounds. By contrast, disparities among the lowest achievers point to the success or failure of, for example, compulsory schooling in motivating and engaging young people with a lower interest and aptitude for processing information. Looking beyond the mean therefore provides a clearer, more detailed basis for policy design and resource allocation than relying on averages. In concrete terms, the analysis compares gaps in skills not only for the “average” adult but also for those at the very top (90th percentile) and very bottom (10th percentile) of the skills distribution. This allows us to see whether disparities are larger/smaller among high and low achievers, and which groups are most affected. This section complements the analyses above by detailing differences for average adults in OECD populations and disparities at the 10th (bottom 10% scoring adults) and 90th (top 10% scoring adults) percentiles of the distribution of literacy, numeracy and adaptive problems solving skills.
Among the highest achievers, gender disparities favour men in all domains, as indicated in Figure 2.7, Panel B. In numeracy, the average advantage of men is 0.17 SD; however, at the 90th percentile it is 0.28 SD and at the 10th percentile it is only 0.05 SD. In literacy, men outperform women by 0.05 SD among the highest achievers, whereas women outperform men by 0.15 SD among the lowest achievers. Similarly, the advantage of men in adaptive problem solving corresponds to 0.15 SD among the highest achievers, but women outperform men by 0.04 SD among the lowest achievers. These differences are discussed in Section 2.2.2, which considers the intersectional nature of disadvantage and reflects on the interplay between gender and socio-economic background.
Age disparities favour young adults and are widest among the lowest achievers in literacy, numeracy and adaptive problem solving. Panel A in Figure 2.7 reveals that the difference in numeracy in favour of 16-29 year-olds relative to 50-65 year-olds corresponds to 0.28 SD at the 90th percentile but 0.46 SD at the 10th percentile. In literacy, 16-29 year-olds outperform 50-65 year-olds by 0.41 SD among the highest achievers and by 0.56 SD among the lowest achievers. Similarly, the advantage of young adults in adaptive problem solving corresponds to 0.53 SD among the highest achievers and 0.62 SD among the lowest achievers. These findings could reflect age-related skills depreciation (Paccagnella, 2016[79]) and restricted opportunities for adult learning (Hanushek et al., 2025[39]) disproportionately affecting older adults, who already possess limited human capital. By contrast, younger cohorts who have just completed a more broadly uniform schooling may benefit from a de facto minimum proficiency floor that leads to a relatively compressed lower tail of the achievement distributions. Over the life course, older workers with low information-processing skills are less likely to access further training, have technology-rich jobs or engage in cognitively stimulating activities; however, older adults with high levels of information-processing skills can offset natural decline through continuous professional practice and by mobilising experience. The result is a more polarised distribution among older individuals due to a sizeable share of this group sliding further behind, while the lower tail of the youth distribution remains comparatively compact, accentuating age-related disparities precisely where achievement is weakest.
Disparities related to socio-economic background favour adults with advantaged backgrounds and are widest among the lowest achievers. For example, Panels C and D in Figure 2.7 indicate that the advantage in literacy skills of adults with at least one tertiary-educated parent corresponds to 0.44 SD among the highest achievers and 0.73 SD among the lowest achievers. The advantage in literacy skills of adults with at least one parent who worked in a high-status occupation corresponds to 0.41 SD among the highest achievers and 0.65 SD among the lowest achievers. Similarly, the advantage in numeracy of adults with at least one tertiary-educated parent corresponds to 0.43 SD among the highest achievers and 0.68 SD among the lowest achievers. For adults with at least one parent who worked in a high-status occupation, the advantage in numeracy skills corresponds to 0.42 SD among the highest achievers and 0.62 SD among the lowest achievers.
Disparities related to immigrant background favour non-immigrants and are widest among the lowest achievers. Differences in information-processing skills between individuals with and without an immigrant background are especially wide among the lowest achievers, whereas disparities are lowest among the highest achievers (Figure 2.7, Panel E). For instance, while the mean literacy difference between non-immigrants and immigrants is 0.14 SD, the difference among the lowest achievers is 0.26 SD and among the highest achievers 0.06 SD.
Disparities related to childhood residential context favour those who grew up in cities and are widest among the highest achievers in adaptive problem solving but do not differ significantly between the highest and lowest achievers in literacy and numeracy. The mean difference between adults who grew up in cities and those who grew up in villages is 0.2 SD in literacy, 0.18 in numeracy and 0.17 in adaptive problem solving (Figure 2.7, Panel F). In adaptive problem solving, the difference among lowest achievers is 0.15 SD, while it is 0.19 SD among the highest achievers.
Average, 90th and 10th percentile difference in skills by socio-demographic characteristic, OECD average
Note: Adults aged 16-65. All mean score-point differences in Panels A–F are statistically significant at the 5% level. Filled diamonds indicate that differences at the 10th or 90th percentile are statistically significant at the 5% level. The pale colour diamond marks the score-point difference at the 10th percentile between the respective groups. The darker colour diamond marks the score-point differences at the 90th percentile between the respective groups. See the note for Figure 2.1 for the definitions of groups based on parental education and parental occupation. See the note for Figure 2.3 for the definitions of groups based on immigrant background and childhood residential context. Country-specific results are provided in Tables 2.A.2.7, 2.A.2.8, 2.A.2.9, 2.A.2.10, 2.A.2.11 and 2.A.2.12 in Annex 2.A.
Source: Calculations based on OECD (2024[23]), Survey of Adult Skills (PIAAC) 2nd cycle database, www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html.
The findings shown in the figures above suggest that “who you are” matters differently at different levels of achievement, reflecting within-group differences between individuals with a higher and lower propensity to excel in core 21st-century skills. The wide gap associated with socio-economic background among low achievers could reflect compensatory advantage processes, i.e. the fact that families with a socio-economically privileged background can mobilise financial, cultural and social capital to ensure that if their children have low levels of academic potential they nonetheless gain valuable skills (Bernardi, 2014[80]; Bernardi and Valdés, 2021[21]). It could also reflect the fact that such parents are more able to engage in intensive and strategically structured parenting to support their children (Lareau, 2011[17]).
2.2.2. The intersectional nature of skills disparities
Disparities in skills arise from the interplay of multiple characteristics such as gender, socio-economic background and geographical context. Intersectionality underscores the importance of examining these factors simultaneously, as each dimension can compound or mitigate inequalities in complex ways (Crenshaw, 1991[81]). Individuals sharing one characteristic, like parental education, may experience different outcomes due to their gender or childhood location. An intersectional approach allows for the more accurate identification of subgroups facing compounded disadvantages, providing insights that could be used to provide targeted support and promote skills development over the life course. This perspective allows policymakers to create flexible, targeted interventions that address specific vulnerabilities (Christoffersen, 2021[82]). This section adopts an intersectional approach to illustrate the added vulnerability that arises from the combination of sources of childhood disadvantage.
Differences between groups in core 21st-century skills vary systematically by gender, parental education and childhood residential context. Detailed differences for specific core 21st-century skills and population groups are available in Annex Table 2.A.3. This section presents illustrative examples to highlight the importance of policy making taking into account the multiplicative nature of disadvantage to provide adequate and tailored support.
Gender disparities
Among school-aged populations, evidence suggests that young people’s socio-economic background is especially impactful for boys, and that boys from disadvantaged backgrounds and/or attending schools with disadvantaged peers are especially likely to suffer from low levels of academic achievement (Autor et al., 2023[83]; Legewie and DiPrete, 2012[84]). These findings reflect the fact that gender gaps in numeracy are especially pronounced among individuals with high levels of numeracy skills (Ellison and Swanson, 2010[85]; Fryer and Levitt, 2010[86]), who tend to have a socio-economically advantaged background and live in urban rather than rural settings.
The disadvantage of women with socio-economically advantaged backgrounds could reflect the specific barriers they may face among high achievers, such as women’s attitudes towards risk taking and competition (Eckel and Füllbrunn, 2015[87]; Niederle and Vesterlund, 2007[88]), as well as stereotypes and discrimination regarding women’s lower potential to be “brilliant” (Leslie et al., 2015[89]; Storage et al., 2020[90]). The wider gender gaps among high achievers could also underscore how context-specific social norms and stereotypes may influence boys and girls differentially depending on their family background and where they grew up. Social norms in more advantaged urban environments may reinforce or elevate boys’ engagement with academic pursuits, amplifying their lead in domains such as numeracy, which are prized in the labour market.
Gender disparities in numeracy are wider – favouring men – among those with at least one tertiary-educated parent (0.20 SD), and slightly narrower among those without (0.16 SD) (Figure 2.8). The figure also illustrates how the gender gap in numeracy proficiency between men and women differs depending on parental educational attainment in different countries. Among adults without any tertiary-educated parent, the gender gap in favour of men is widest in England (United Kingdom) (0.34 SD) and Switzerland (0.33 SD), whereas in Croatia, Hungary, Israel, Latvia, Lithuania, New Zealand, Poland, the Slovak Republic and the United States it is less than 0.1 SD. Among those with at least one tertiary-educated parent, the gender gap in favour of men in numeracy is widest in the Netherlands (0.31 SD) and is smaller than 0.1 SD in Croatia, Italy, Lithuania and Poland. The difference in the gender gap in numeracy in favour of men between individuals with and without a tertiary-educated parent is largest in favour of those without a tertiary-educated parent in Italy – where it is as large as 0.24 SD; it is largest in favour of those with a tertiary-educated parent in New Zealand, where it is as large as 0.18 SD.
Difference in numeracy proficiency between men and women, by parental education
Note: Adults aged 16-65. The figure shows the gender gap in numeracy among adults where no parent is tertiary-educated and among parents where at least one parent is tertiary-educated. See the note for Figure 2.1 for the definitions of groups based on parental education.
Countries are ranked in descending order of the score-point difference between men and women among those without tertiary-educated parents. Source: Calculations based on OECD (2024[23]), Survey of Adult Skills (PIAAC) 2nd cycle database, www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html.
Childhood residential context
Disparities in information-processing skills between adults with and without tertiary-educated parents are especially wide among those who grew up in villages rather than cities. For example, the difference in literacy between individuals with and without a tertiary-educated parent is 0.61 SD among individuals who grew up in villages but 0.52 SD among those who grew up in cities Annex Table 2.A.3. Similarly, the differences in numeracy and adaptive problem solving skills between individuals with and without a tertiary-educated parent are 0.57 SD for numeracy and 0.6 SD for adaptive problem solving among individuals who grew up in villages compared to 0.51 SD, and 0.52 SD among those who grew up in cities.
These findings suggest that a family’s cultural, economic and social resources exert a stronger influence on long-term outcomes in settings with limited opportunities for skills development. For example, in rural areas, lower population density often reduces choice in terms of school educational programmes or high-quality training programmes (Echazarra and Radinger, 2019[70]). Similarly, amid generalised teacher shortages in many OECD countries (OECD, 2024[91]), rural areas face especially strong barriers to attracting and retaining teachers (Echazarra and Radinger, 2019[70]), with highly educated individuals typically preferring to live and work in urban centres with rich cultural amenities and greater labour market opportunities for their partner (Bacolod, Blum and Strange, 2009[92]; Berry and Glaeser, 2005[93]). Moreover, economic activities in rural settings – such as energy extraction, agriculture, forestry and fishing, food production and processing, and logistics – often do not support the cultivation of advanced information-processing skills to the same extent as activities in urban settings. Consequently, individuals with socio-economically disadvantaged backgrounds from rural areas may be particularly constrained by limited resources within the home and in their environment, hindering their ability to develop and maintain robust skills over time.
Figure 2.9 shows that the “city advantage” in numeracy is not the same for all adults but depends on the education level of a person’s parents. For example, in Israel, adults without tertiary-educated parents gain the most from having grown up in a city rather than a village: those raised in cities score around half a standard deviation higher in numeracy than similar adults who grew up in villages. By contrast, adults with tertiary-educated parents gain the most from having grown up in a village rather than a city: those raised in cities score one-tenth of a standard deviation lower in numeracy than similar adults who grew up in villages. In Norway the pattern is reversed: whereas there is no difference in numeracy between adults who grew up in urban and rural settings for adults with tertiary-educated parents, among adults without tertiary-educated parents, numeracy proficiency is lower among those who grew up in cities rather than a village. These differences reflect between-country differences in how economic opportunities and the availability of high-quality skills development possibilities differ across urban and rural settings.
Average difference in numeracy proficiency between adults who grew up in cities and villages, by parental education
Note: Adults aged 16-65. The figure shows the gender gap in numeracy among adults who grew up in cities and among adults who grew up in villages. See the note of Figure 2.3 for the definition of groups by childhood residential context.
Countries are ranked in descending order of the score-point difference between those who grew up in cities and villages among those without tertiary-educated parents.
Source: Calculations based on OECD (2024[23]), Survey of Adult Skills (PIAAC) 2nd cycle database, www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html.
Disparities in information-processing skills, alongside social and emotional skills, can influence economic productivity, social mobility and individuals’ well-being. Examining differences in the size of disparities across countries can shed light on whether contextual factors may strengthen or weaken skills differentials across specific population groups. This section considers whether differences between countries in the size of skills disparities between men and women and between individuals with socio-economically advantaged and disadvantaged backgrounds reflect the degree of inequality present in different countries. Figure 2.10 summarises results regarding the associations between country-level characteristics and the size of disparities in skills.
2.3.1. How differences across countries in skills disparities between men and women relate to country-level characteristics
Across OECD countries, reducing gender disparities in education and the labour market is a common policy goal, and empirical evidence based on data from PISA has revealed that in countries with greater gender equality – as measured by the Gender Inequality Index (GII) – the gender gap in mathematics among adolescents in favour of boys is smaller, whereas the gender gap in reading in favour of girls is wider (Guiso et al., 2008[94]). Although this pattern was expected to hold among adults, further empirical evidence has since indicated that in countries with higher levels of gender equality (as measured by the GII), differences between men and women in numeracy proficiency are wider, in favour of men, as is the gender gap in participation in science, technology, engineering and mathematics (STEM) fields of study (Balducci et al., 2024[95]; Borgonovi, Choi and Paccagnella, 2018[96]; Herlitz et al., 2024[97]; Stoet and Geary, 2018[98]). This is sometimes referred to as the “gender equality paradox” (Stoet and Geary, 2018[98]). Understanding why these patterns emerge is crucial for designing policies that effectively address the root causes of persistent differences in outcomes between men and women, and can help to refine theories on how social environments interact with individual choices and potential.
Most empirical analyses of the association between the size of gender skills differentials in a country and that country’s norms towards gender equality rely on measures of societal-level gender inequality that reflect women’s under-representation in the economic, political and social life of a country. Such metrics quantify women’s access to labour markets and political representation but fail to portray qualitative differences in the opportunities men and women have available. For example, in a particular society, men and women could have similar levels of employment but work in very different occupations, with women being considerably more likely to work as teachers and nurses, and men being more likely to work as engineers and welders.
This section characterises societal-level gender disparities through the newly constructed Gender Occupational Integration Index (GOII). The GOII measures how equally women are represented in men-majority occupations and how equally men are represented in women-majority occupations (see Box 2.4 for a detailed description of the two indices). Associations derived from the GOII are similar to those obtained when considering GII, which is the widely used measure of societal inequality (Annex Table 2.A.4). Figure 2.10 illustrates country-level associations between the indicator of societal-level gender equality, using the GOII, and country-level indicators of gender disparities in skills, providing evidence that can be used to assess how social norms shape the relative performance of men and women across various skills domains. Societal-level gender equality is not systematically associated with the size of gender disparities in core 21st-century skills. Most country-level associations are quantitatively small or medium size, except for numeracy (Figure 2.10, Panel B). This means, for example, that gender disparities in literacy (a skills domain with a relatively large variation in the size of gender gaps across countries) tend to be just as large in countries in which occupational segregation between men and women is low and in countries in which it is high.
Figure 2.10 shows that gender gaps in numeracy in favour of men tend to be wider in countries with higher levels of gender equality. For example, in Canada the gender gap in numeracy is large (0.27 SD), and on average women and men are less likely than in other countries to work in occupations with a majority of workers of the same sex (GOII=37).
These findings indicate that among adults, the gender equality paradox applies only to proficiency in numeracy and not to other domains. Furthermore, the results suggest that the paradox is not due to measurement: estimates are aligned regardless of whether measures of gender equality that characterise qualitative differences in the participation of men and women in the labour market (likelihood of employment in counter-stereotypical occupations, i.e. the GOII) or measures that characterise quantitative differences (employment rates) are used (i.e. the GII) (see Annex Table 2.A.4).
Note: To maintain readability, a selection of countries and economies is labelled. For more information on the Gender Occupational Integration Index (GOII), refer to Box 2.4 and Chapter 4.
Source: Calculations based on OECD (2024[23]), Survey of Adult Skills (PIAAC) 2nd cycle database, www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html.
These findings suggest that societal norms and stereotypes may persist even in highly egalitarian contexts. Although structural barriers may be lower, gendered expectations about suitable roles and interests may still influence individuals’ decisions. The gender equality paradox may reflect differences in the relative achievement of boys and girls in different academic domains and, in turn, the extent to which relative achievement shapes educational and career choices. In more equal societies, girls achieve at a level that is closer to that of boys than in less equal societies, but they generally continue to be especially proficient in domains not typically associated with enrolling in STEM subjects or pursuing STEM careers (OECD, 2015[35]). As a result, they tend to choose pathways of further education and training that do not strengthen their numeracy skills (Jansen, Becker and Neumann, 2021[99]). While these observed relationships can inform policy discussions, a more definitive understanding will require longitudinal or multilevel designs that control for shared cultural, linguistic and historical factors (Richardson et al., 2020[100]).
Indicators of societal-level gender inequality each have advantages and disadvantages, depending on the intended use (Permanyer, 2009[101]). One of the most commonly used indicators is the Gender Inequality Index (GII), which is a composite measure of inequality between men and women that captures three dimensions: 1) reproductive health; 2) political empowerment; and 3) labour market participation (UNDP, 2025[102]). A low GII value indicates small differences in outcomes on the three dimensions between men and women, and a high GII value indicates large differences in outcomes between men and women (Permanyer, 2009[101]).
Although the GII offers broad country coverage and covers labour market disparities, its approach is relatively limited in this area as it only considers labour force participation rates. Its composite structure also makes it difficult to conduct more focused analyses of specific labour market outcomes. In particular, the GII cannot determine gender differences in labour market participation by sector and whether observed differences arise from women’s under-representation in some sectors, men’s under-representation in other sectors or both, which is an important distinction given that these patterns can vary considerably by country and sector (see Chapter 4).
To address these limitations, the Gender Occupational Integration Index (GOII) was developed to examine labour market occupational integration for men and women separately, with results combined into a single indicator to reflect how societal-level social norms shape the concentration of men and women in different economic activities. Put simply, whereas the GII captures how many women work, the GOII indicates the extent to which men work as, for example, teachers or nurses and women as truck drivers or engineers.
The GOII is calculated as the sum of the percentage of women working in men-majority occupations and of men working in women-majority occupations. Men-majority occupations are those where the weighted proportion of men’s employment across all countries and economies in pooled 2023 Survey of Adult Skills first and second cycles exceed 75%. Women-majority occupations are those where the weighted proportion of women’s employment across all countries and economies in pooled 2023 Survey of Adult Skills first and second cycles exceed 75%. This percentage is broadly aligned with other literature that defines gender-dominated occupations as those with female (or male) percentages between 60% and 80% (Bächmann, 2022[103]; Keane, Russell and Smyth, 2017[104]; McCaughey, 2023[105]; Torre, 2018[106]). The proportions were calculated at the three-digit ISCO level. Ten occupation categories (less than 0.5% of the sample) were excluded due to an insufficient number of observations. Armed forces occupations were also excluded from the analysis. This classification thus categorises occupations as men- and women-majority among all 2023 Survey of Adult Skills participating countries and economies together, disregarding local contexts. This ensures that, in the next step, the same men- and women-majority occupations are compared among countries.
Once each occupation has been classified as men-majority, women-majority or neither, shares of women in men-majority and men in women-majority workers were calculated in each country and economy participating in the 2023 Survey of Adult Skills. These shares have been summed to produce the GOII value, expressed in percentages. Its theoretical minimum is 0% (no equality), indicating that no women work in men-majority occupations and no men work in women-majority occupations. If 50% of women in a country were to work in men-majority occupations and 50% of men work in women-majority occupations (perfect equality), its value is 100%.
The top three men-majority occupations are building frame and related trades workers (men-employment share of 97.8%), electrical equipment installers and repairers (97.6%), and building finishers and related trades workers (97.5%). The top three women-majority occupations are secretaries (women-employment share of 93.6%), nursing and midwifery professionals (91.7%), and childcare workers and teachers' aides (90.7%).
On average across the OECD, only 29% of workers are employed in jobs where their own sex accounts for less than one-quarter of the workforce; however, there is considerable cross-national variation in occupational gender integration (Figure 2.11). At one end of the spectrum, Poland and Lithuania have the lowest integration, with only about 15% of workers employed in jobs where their own sex accounts for less than one-quarter of the workforce (determined as all countries participating in the pooled first and second cycles of the Survey of Adult Skills). Similarly, in Croatia, Portugal, the Slovak Republic and Estonia, less than 25% of workers are employed in jobs where their own sex accounts for less than one-quarter of the workforce. The United States and Spain are the only countries in which over 40% of workers are employed in jobs in which their own sex accounts for less than a quarter of the workforce.
Sum of the percentage of women working in men-majority occupations and of men working in women-majority occupations
Note: Gender Occupational Integration Index measures the sum of the percentage of women working in men-majority occupations and men in women-majority occupations. Its theoretical minimum is 0% (no equality), indicating that no women work in men-majority occupations and no men work in women-majority occupations. Men-majority occupations are those where the weighted proportion of men’s employment across all countries and economies in the pooled first and second cycles of the Survey of Adult Skills exceed 75%. Women-majority occupations are those where the weighted proportion of women’s employment across all countries and economies in pooled 2023 Survey of Adult Skills first and second cycles exceed 75%. The proportions were calculated using occupations at the three-digit ISCO level.
Countries are ranked in descending order of the Gender Occupational Integration Index.
Source: Calculations based on OECD (2017[107]), PIAAC 1st cycle database, www.oecd.org/en/data/datasets/PIAAC-1st-Cycle-Database.html and OECD (2024[23]), PIAAC data and methodology, www.oecd.org/en/about/programmes/piaac/piaac-data.html.
Table 2.1 indicates, in detail, differences in the shares of workers in women-majority and men-majority occupations, of men and of women in occupations where their sex is a minority, alongside gender-specific rates of labour force participation or employment.
Note: Includes 16-65 year-olds. Men-majority occupations are those where the weighted proportion of men’s employment across all countries and economies in pooled 2023 Survey of Adult Skills first and second cycles exceed 75%. Women-majority occupations are those where the weighted proportion of women’s employment across all countries and economies in pooled 2023 Survey of Adult Skills first and second cycles exceed 75%. The proportions were calculated at three-digit ISCO level. Potentially employed individuals are adults who were not students or retired, and were either unemployed; actively looked for paid work; or did not look for work due to looking after the family or home, due to being temporarily sick or injured, due to not believing any jobs were available, or due to not getting around to looking yet; and those who stopped working involuntarily in their last job due to a temporary job coming to an end, the job not matching skills, reorganisation, mass lay-offs or plant closure, or family reasons. It is expressed as a percentage of all employed workers.
Source: Calculations based on OECD (2017[107]), PIAAC 1st cycle database, www.oecd.org/en/data/datasets/PIAAC-1st-Cycle-Database.html and OECD (2024[23]), PIAAC data and methodology, www.oecd.org/en/about/programmes/piaac/piaac-data.html.
2.3.2. How differences across countries in skills disparities linked to socio-economic background relate to country-level characteristics
Evidence indicates that countries with greater income inequality also have lower social mobility, which Alan Kruger referred to as the Great Gatsby Curve (Corak, 2013[108]). Historically, worsening inequality has been shown to go hand in hand with declining social mobility. In contexts where incomes are spread out very unevenly, the children of disadvantaged families may find it more difficult to improve their position in adulthood by developing their skills. If socio-economic disparities shape opportunities for skills development among different groups, whether through access to family resources, access to high-quality education or broader community support, differential skills development could be one of the key channels through which inequality reduces social mobility.
A critique of the Great Gatsby Curve is that, in its original formulation, it fails to distinguish different forms of inequality (Bukodi and Goldthorpe, 2018[109]). Inequality arising from the willingness and ability of a society to reward talent and achievement is, in fact, qualitatively different from inequality that reflects innate privilege (OECD, 2025[60]). Therefore, this section considers two indicators of inequality: total income inequality, as measured using the Gini coefficient, and inequality of opportunity, as measured by the OECD’s absolute inequality of opportunity indicator, which reflects how much income varies between people with different backgrounds (OECD, 2025[60]). The OECD absolute inequality of opportunity indicator reflects how much the household market income of two people who put similar effort to succeed but came from different family circumstances can be expected to differ.
Investigating whether cross-country differences in skills gaps between individuals with socio-economically advantaged and disadvantaged backgrounds reflect country differences in inequality can reveal the extent to which broader inequalities filter through into skills formation. Societies that exhibit both high income inequality and large disparities in key skills may be at particular risk of sustaining low levels of social mobility, as predicted by the Great Gatsby Curve. Conversely, those that manage to cultivate more inclusive skills development may mitigate some of the long-term harms associated with high inequality.
This section explores these issues by examining the country-level associations between total income inequality and inequality of opportunity on the one hand, and disparities in information-processing skills across adults with different socio-economic backgrounds on the other. The findings presented in Table 2.2 show that associations between socio-economic disparities in information-processing skills and total income inequality are strong or medium size, with skills disparities more pronounced in countries with greater inequality. By contrast, associations with inequality of opportunity are weaker2. For example, around 18% of the between-country variation in differences in numeracy between individuals with and without tertiary-educated parents can be explained by differences across countries in income inequality. Similarly, around 21% of the between-country variation in differences in numeracy between individuals who grew up in cities and those who grew up in villages can be explained by differences across countries in income inequality.
Variation explained and sign of relationship
Note: For the definition of the absolute inequality of opportunity index see OECD (2025[60]). The strength of the association is assessed by R‑squared (R²) values, which are categorised based on Cohen’s (2013[42]) benchmarks. Specifically, an R² above 0.13 reflects a medium or substantial effect. The Gini coefficient is based on household disposable income, i.e. income after taxes and transfers adjusted for household size. The Gini coefficient takes values between 0 (where every person has the same income), and 1 (where all income goes to one person). See the note for Figure 2.1 for the definitions of groups based on parental education and parental occupation. See the note for Figure 2.3 for the definitions of groups based on childhood residential context.
Source: Calculations based on OECD (2025[60]), To Have and Have Not – How to Bridge the Gap in Opportunities, https://doi.org/10.1787/dec143ad-en; OECD (2024[23]), Survey of Adult Skills (PIAAC) 2nd cycle database, www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html; and OECD (2024[110]), Income Distribution Database, (database), Data Explorer, http://data-explorer.oecd.org/s/17a.
Analyses suggest that within-country disparities in information-processing skills reflect how economic resources are distributed: in societies with more unequal distributions of economic resources, the skills gap between individuals with different socio-economic backgrounds is wider. The results also indicate that, as observed for gender disparities, numeracy is more closely linked to a society’s level of inequality than either literacy or adaptive problem solving skills.
For example, Figure 2.12 shows how in Chile and Israel, total income inequalities are relatively large (Gini = 0.45 and 0.34, respectively) and skills disparities between individuals related to parental occupation are also wide (0.7 SD difference in numeracy and 0.6 SD difference in adaptive problem solving, and 0.7 SD difference for both countries in literacy). In contrast, in Poland and the Slovak Republic, incomes are more equally distributed (Gini = 0.27 and 0.23, respectively, Figure 2.12), and skills disparities between individuals based on parental occupation are relatively narrow (0.4 SD and 0.3 SD difference in numeracy, 0.3 SD and 0.2 SD difference in adaptive problem solving and 0.4 SD and 0.3 SD difference in literacy, Figure 2.12).
Note: To maintain readability, a selection of countries and economies is labelled. Total income inequality is based on the Gini coefficient. The Gini coefficient is based on household disposable income, i.e. income after taxes and transfers adjusted for household size. The Gini coefficient takes values between 0 (where every person has the same income), and 1 (where all income goes to one person). Data for total income inequality for the Flemish Region (Belgium) refer to Belgium and for England (United Kingdom) to the United Kingdom. See the note for Figure 2.1 for the definitions of groups based on parental occupation.
Source: Calculations based on OECD (2024[110]); Income Distribution Database, (database), Data Explorer, http://data-explorer.oecd.org/s/17a and OECD (2024[23]), Survey of Adult Skills (PIAAC) 2nd cycle database, www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html.
Adolescence marks a critical juncture in educational trajectories and future labour market choices. Research consistently shows that adolescents’ proficiency in reading and mathematics is a strong predictor of later academic success, influencing decisions about upper secondary specialisations, university enrolment and eventual career pathways (Hakkarainen, Holopainen and Savolainen, 2013[111]; Wang, 2013[112]). Changes over the first quarter of the 21st century in adolescent reading and mathematics skills can shed light on whether gaps across different population groups, such as gender, immigrant background, childhood residential context and socio-economic background, are narrowing or widening. Such information can help policymakers evaluate the evolving quality of initial education and target interventions to reduce disparities that emerge at an early age.
By analysing recent trends in the achievement of adolescents, policymakers can better anticipate whether new cohorts have key prerequisites for successful participation in further education and training. Information on the skills levels of young people is also needed to design interventions that mitigate longer-term disparities by reshaping the provision of further education and training to align with what young people know and can do. Conversely, evidence of narrowing divides might signal that existing policies are helping to equalise opportunities, although sustained commitment may still be required to maintain progress.
Ultimately, tracking shifts in skills disparities during adolescence offers a more forward-looking perspective than relying solely on cross-sectional data from adult populations. It allows policymakers to identify emerging trends before they fully manifest in the workforce, thereby guiding the design of proactive strategies to ensure that future adults, regardless of gender, parental background, immigration status or childhood residential context, can develop the skills demanded by rapidly evolving labour markets. This section compares trends in achievement in reading and mathematics between 2003 and 2022, the period for which comparable evidence on both domains can be tracked among adolescents in a large number of OECD countries.
2.4.1. Trends in achievement disparities among 15-year-old students
Between 2003 and 2022, the gender gap in reading and mathematics proficiency fluctuated. Figure 2.13 presents the average proficiency by gender in reading and mathematics among 15‑year-old students (Panel A) and the evolution of the gender gap over the 2003 to 2022 period (Panel B). In reading, the gender gap in favour of girls increased by six PISA score points between 2003 and 2009 (from 34 to 40 points, corresponding to an increase from 0.34 SD to 0.40 SD difference3). However, starting in 2012, the gender gap steadily narrowed, and by 2022, it was 25 PISA score points (corresponding to 0.25 SD difference). In mathematics, the gap favouring boys decreased between 2003 and 2018, from 11 PISA score points to 5 score points (corresponding to a decline of the gap from 0.11 SD to 0.05 SD), but widened again from 5 score points to 9 score points between 2018 and 2022 (corresponding to an increase in the gap from 0.05 SD to 0.09 SD).
Figure 2.13 suggests that changes in gender gaps over the period were due to diverging trends in achievement, which declined among both boys and girls. Between 2012 and 2018, the reduction in the gender reading gap in reading, which favoured girls, was driven by improvements in boys’ achievement, whereas girls’ achievement remained stable. In contrast, between 2018 and 2022, the period corresponding to the COVID-19 pandemic, reading achievement decline was more pronounced among girls than boys. The increase in the gender gap in mathematics between 2018 and 2022 was due to the fact that both boys’ and girls’ mathematics achievement declined over the period, but the decline was steeper for girls than for boys.
Panel A: Mathematics and reading scores for boys and girls, OECD average. Panel B: Differences in average mathematics and reading scores between boys and girls (boys minus girls)
Note: Panel B: Lines in lighter colours represent gender disparities in specific countries between 2003 and 2022. Lines in dark colours represent the OECD average.
Source: Calculations based on OECD (2003[113]), PISA 2003 Database, www.oecd.org/en/data/datasets/pisa-2003-database.html; OECD (2006[114]), PISA 2006 Database, www.oecd.org/en/data/datasets/pisa-2006-database.html; OECD (2012[115]), PISA 2012 Database, www.oecd.org/en/data/datasets/pisa-2012-database.html; OECD (2015[116]), PISA 2015 Database, www.oecd.org/en/data/datasets/pisa-2015-database.html; OECD (2018[117]), PISA 2018 Database, www.oecd.org/en/data/datasets/pisa-2018-database.html; and OECD (2022[118]), PISA 2022 Database, www.oecd.org/en/data/datasets/pisa-2022-database.html.
Whereas gender disparities differ markedly depending on the skill considered, disparities in reading and mathematics between 15-year-old students with and without tertiary-educated parents are closely aligned. Similarly, results remain consistent when parental occupational status is used as an indicator of socio-economic background. Figure 2.14 therefore reports only the evolution of disparities in mathematics achievement between 15-year-old students with and without tertiary-educated parents from 2003 to 2022. Disparities by parental educational attainment narrowed during this period but remained large: in 2022, the score-point difference in mathematics achievement between young people with and without a tertiary-educated parent was 37 score points (corresponding to 0.37 SD difference), whereas in 2003 it was 47 score points (corresponding to 0.47 SD difference). The narrowing of disparities between 2003 and 2012 was primarily due to the fact that mathematics skills declined over the period among all young people, but the decline was steeper among young people with tertiary-educated parents.
Panel A: Mathematics scores for students with tertiary and non-tertiary educated parents, OECD average. Panel B: Differences in average mathematics scores between students with tertiary and non-tertiary educated parents
Note: Panel B: Lines in lighter colours represent disparities between students with and without tertiary educated parents in specific countries between 2003 and 2022, and the darker line represents the OECD average. Parental education is based on students’ responses. Information on their mothers’ and fathers’ education. were used to derive the index of highest education level of parents. The index is equal to the highest ISCED level of either parent.
Source: Calculations based on OECD (2003[113]), PISA 2003 Database, www.oecd.org/en/data/datasets/pisa-2003-database.html; OECD (2006[114]), PISA 2006 Database, www.oecd.org/en/data/datasets/pisa-2006-database.html; OECD (2012[115]), PISA 2012 Database, www.oecd.org/en/data/datasets/pisa-2012-database.html; OECD (2015[116]), PISA 2015 Database, www.oecd.org/en/data/datasets/pisa-2015-database.html; OECD (2018[117]), PISA 2018 Database, www.oecd.org/en/data/datasets/pisa-2018-database.html; and OECD (2022[118]), PISA 2022 Database, www.oecd.org/en/data/datasets/pisa-2022-database.html.
There are notable mathematics skills disparities among students who attended schools in cities compared to villages, with the skills gap fluctuating between 2003 and 2022. Figure 2.15 illustrates trends from 2003 to 2012 in the average mathematics achievement of 15‑year‑old students who attended schools in cities compared to those in villages. Disparities in reading are similar. On average, students who attend school in cities have higher mathematics skills than those who attend schools in villages (Figure 2.15, Panel A), and the disparity between the two groups widened from 22 score points in 2003 (corresponding to 0.22 SD difference) to 26 score points in 2022 (corresponding to 0.26 SD difference).
Panel A: Mathematics scores for students who attended school in villages and cities, OECD average. Panel B: Differences in average mathematics scores between students who attended school in villages and cities (city minus villages)
Note: Panel B: Lines in lighter colours represent disparities between students who attended in schools in cities and villages in specific countries between 2003 and 2022, and the darker line represents the OECD average. The figure reports the mathematics achievement and the mathematics achievement gap depending on whether students attended school in a village or a city. Villages are defined as communities (villages or towns) with up to 15 000 people, cities are defined as communities (cities or megacities) with 1 million and more people.
Source: Calculations based on OECD (2003[113]), PISA 2003 Database, www.oecd.org/en/data/datasets/pisa-2003-database.html; OECD (2006[114]), PISA 2006 Database, www.oecd.org/en/data/datasets/pisa-2006-database.html; OECD (2012[115]), PISA 2012 Database, www.oecd.org/en/data/datasets/pisa-2012-database.html; OECD (2015[116]), PISA 2015 Database, www.oecd.org/en/data/datasets/pisa-2015-database.html; OECD (2018[117]), PISA 2018 Database, www.oecd.org/en/data/datasets/pisa-2018-database.html; and OECD (2022[118]), PISA 2022 Database, www.oecd.org/en/data/datasets/pisa-2022-database.html.
2.4.2. Extending the analysis: Skills disparities in ancillary 21st-century skills related to gender, childhood residential context and parental education
The range of skills analysed in this report is limited by the availability of cross-country comparable data on adult populations in OECD countries. However, analyses can be considerably extended when assessing disparities in ancillary 21st-century skills among young people. This is because a wider range of assessment instruments has been developed and administered to schooled populations. These include measures for creative problem solving (OECD, 2014[119]), collaborative problem solving (OECD, 2017[120]), global competence (OECD, 2020[121]), financial literacy (OECD, 2023[122]), creative thinking tasks (OECD, 2024[123]), civic knowledge (IEA, 2022[124]), computational thinking (IEA, 2023[125]), and computer and information literacy (IEA, 2023[125]).
Results from this further analysis show that there are variations in these ancillary skills related to gender, childhood residential context and parental education. Figure 2.16, Figure 2.17 and Figure 2.18 summarise disparities between boys and girls, between young people who attended schools in rural and urban settings, and between those with and without tertiary-educated parents in each of these ancillary skills, while additionally showing average disparities in reading, mathematics and science for the latest available data from 2022. Figure 2.16 displays standardised scores in the assessments as well as differences at the 90th percentile and 10th percentile. On average across OECD countries, boys outperform girls in creative problem solving (0.08 SD), financial literacy (0.05 SD), computational thinking (0.04 SD) and mathematics (0.10 SD). Girls outperform boys in collaborative problem solving (0.29 SD), global competence (0.26 SD), creative thinking (0.25 SD), civic knowledge (0.22 SD), computer and information literacy (0.20 SD), and reading (0.24 SD). In domains where boys outperform girls on average, differences are larger among top-scoring students. In contrast, in domains where girls outperform boys, differences are smaller among top-scoring students, with differences larger among the 10% of bottom-scoring students.
Note: Estimates are standardised to a total OECD mean of zero and a standard deviation of one in the respective databases. Darker bars denote mean differences that are statistically significant at the 5% level. All differences at the 90th and 10th percentile are statistically significant at the 5% level. OECD averages exclude the following OECD countries due to unavailable data: In PISA 2012: Costa Rica, Greece, Iceland, Latvia, Lithuania, Luxembourg, Mexico, New Zealand and Switzerland; in PISA 2015: Ireland, Poland and Switzerland; in PISA 2018: Australia, Austria, Belgium, Czechia, Denmark, Estonia, Finland, France, Germany, Hungary, Iceland, Ireland, Italy, Japan, Luxembourg, Mexico, the Netherlands, New Zealand, Norway, Poland, Portugal, Slovenia, Sweden, Switzerland, Türkiye and the United States; in PISA 2022 for “creative thinking”: Austria, Ireland, Japan, Norway, Sweden, Switzerland, Türkiye, the United Kingdom and the United States. OECD averages include the following OECD countries: PISA 2022 “financial literacy”: Austria, the Flemish Community of Belgium, Canadian provinces (Alberta, British Columbia, Manitoba, New Brunswick, Newfoundland and Labrador, Nova Scotia, Ontario and Prince Edward Island), Costa Rica, Czechia, Denmark, Hungary, Italy, the Netherlands, Norway, Poland, Portugal, Spain and the United States; ICCS 2022: Colombia, Denmark, Estonia, France, Italy, Latvia, Lithuania, Netherlands, North Rhine-Westphalia (Germany), Norway, Poland, Schleswig-Holstein (Germany), the Slovak Republic, Slovenia, Spain and Sweden; ICILS 2023: Austria, Chile, Czechia, Denmark, Finland, the Flemish Community of Belgium, France, Germany, Greece, Hungary, Italy, Korea, Latvia, Luxembourg, the Netherlands, Norway, Portugal, the Slovak Republic, Slovenia, Spain, Sweden and the United States.
Source: Calculations based on IEA (2022[124]), ICCS 2022 Database, https://doi.org/10.58150/ICCS_2022_data_edition_2_including_process_data; IEA (2023[125]), ICILS 2023 Database, https://doi.org/10.58150/ICILS_2023_data; OECD (2012[115]), PISA 2012 Database, www.oecd.org/en/data/datasets/pisa-2012-database.html; OECD (2015[116]), PISA 2015 Database, www.oecd.org/en/data/datasets/pisa-2015-database.html; OECD (2018[117]), PISA 2018 Database, www.oecd.org/en/data/datasets/pisa-2018-database.html; and OECD (2022[118]), PISA 2022 Database, www.oecd.org/en/data/datasets/pisa-2022-database.html.
Students who attended schools in cities (with over 100 000 inhabitants) outperform their peers who attended schools in villages in all domains, with differences ranging between 0.1 SD for computational thinking to 0.4 SD for global competence, as indicated in Figure 2.17. In creative problem solving, global competence, financial literacy, civic knowledge, computational thinking, and computer and information literacy, differences related to school location are larger among top-scoring students. In contrast, in creative thinking, differences related to school location are larger among bottom-scoring students.
Students with tertiary-educated parent(s) outperform their peers without tertiary-educated parents across all assessments and domains (Figure 2.18). The socio-economic gap is larger among top-scoring students in all assessments except for creative thinking and computer and information literacy.
Note: Estimates are standardised to a total OECD mean of zero and a standard deviation of one in the respective databases. Darker bars denote mean differences that are statistically significant at the 5% level. Filled markers denote differences at the 90th and 10th percentile that are statistically significant at the 5% level. OECD averages exclude the following OECD countries due to unavailable data. In PISA 2012: Costa Rica, Greece, Iceland, Latvia, Lithuania, Luxembourg, Mexico, New Zealand and Switzerland; in PISA 2015: Ireland, Poland and Switzerland; in PISA 2018: Australia, Austria, Belgium, Czechia, Denmark, Estonia, Finland, France, Germany, Hungary, Iceland, Ireland, Italy, Japan, Luxembourg, Mexico, the Netherlands, New Zealand, Norway, Poland, Portugal, Slovenia, Sweden, Switzerland, Türkiye and the United States; in PISA 2022 for “creative thinking”: Austria, Ireland, Japan, Norway, Sweden, Switzerland, Türkiye, the United Kingdom and the United States. OECD averages include the following OECD countries: PISA 2022 “financial literacy”: Austria, the Flemish Community of Belgium, Canadian provinces (Alberta, British Columbia, Manitoba, New Brunswick, Newfoundland and Labrador, Nova Scotia, Ontario and Prince Edward Island), Costa Rica, Czechia, Denmark, Hungary, Italy, the Netherlands, Norway, Poland, Portugal, Spain and the United States; ICCS 2022: Colombia, Denmark, Estonia, France, Italy, Latvia, Lithuania, Netherlands, North Rhine-Westphalia (Germany), Norway, Poland, Schleswig-Holstein (Germany), the Slovak Republic, Slovenia, Spain and Sweden; ICILS 2023: Austria, Chile, Czechia, Denmark, Finland, the Flemish Community of Belgium, France, Germany, Greece, Hungary, Italy, Korea, Latvia, Luxembourg, the Netherlands, Norway, Portugal, the Slovak Republic, Slovenia, Spain, Sweden and the United States.
Source: Calculations based on IEA (2022[124]), ICCS 2022 Database, https://doi.org/10.58150/ICCS_2022_data_edition_2_including_process_data; IEA (2023[125]), ICILS 2023 Database, https://doi.org/10.58150/ICILS_2023_data; OECD (2012[115]), PISA 2012 Database, www.oecd.org/en/data/datasets/pisa-2012-database.html; OECD (2015[116]), PISA 2015 Database, www.oecd.org/en/data/datasets/pisa-2015-database.html; OECD (2018[117]), PISA 2018 Database, www.oecd.org/en/data/datasets/pisa-2018-database.html; and OECD (2022[118]), PISA 2022 Database, www.oecd.org/en/data/datasets/pisa-2022-database.html.
Note: Estimates are standardised to a total OECD mean of zero and a standard deviation of one in the respective databases. All mean differences and differences at the 90th and 10th percentile are statistically significant at the 5% level. Parental education is based on students’ responses. Information on their mothers’ and fathers’ education was used to derive the index of highest education level of parents. The index is equal to the highest ISCED level of either parent. OECD averages exclude the following OECD countries due to unavailable data. In PISA 2012: Costa Rica, Greece, Iceland, Latvia, Lithuania, Luxembourg, Mexico, New Zealand and Switzerland; in PISA 2015: Ireland, Poland and Switzerland; in PISA 2018: Australia, Austria, Belgium, Czechia, Denmark, Estonia, Finland, France, Germany, Hungary, Iceland, Ireland, Italy, Japan, Luxembourg, Mexico, the Netherlands, New Zealand, Norway, Poland, Portugal, Slovenia, Sweden, Switzerland, Türkiye and the United States; in PISA 2022 for “creative thinking”: Austria, Ireland, Japan, Norway, Sweden, Switzerland, Türkiye, the United Kingdom and the United States. OECD averages include the following OECD countries: PISA 2022 “financial literacy”: Austria, the Flemish Community of Belgium, Canadian provinces (Alberta, British Columbia, Manitoba, New Brunswick, Newfoundland and Labrador, Nova Scotia, Ontario and Prince Edward Island), Costa Rica, Czechia, Denmark, Hungary, Italy, the Netherlands, Norway, Poland, Portugal, Spain and the United States; ICCS 2022: Colombia, Denmark, Estonia, France, Italy, Latvia, Lithuania, Netherlands, North Rhine-Westphalia (Germany), Norway, Poland, Schleswig-Holstein (Germany), the Slovak Republic, Slovenia, Spain and Sweden; ICILS 2023: Austria, Chile, Czechia, Denmark, Finland, the Flemish Community of Belgium, France, Germany, Greece, Hungary, Italy, Korea, Latvia, Luxembourg, the Netherlands, Norway, Portugal, the Slovak Republic, Slovenia, Spain, Sweden and the United States.
Source: Calculations based on IEA (2022[124]), ICCS 2022 Database, https://doi.org/10.58150/ICCS_2022_data_edition_2_including_process_data; IEA (2023[125]), ICILS 2023 Database, https://doi.org/10.58150/ICILS_2023_data; OECD (2012[115]), PISA 2012 Database, www.oecd.org/en/data/datasets/pisa-2012-database.html; OECD (2015[116]), PISA 2015 Database, www.oecd.org/en/data/datasets/pisa-2015-database.html; OECD (2018[117]), PISA 2018 Database, www.oecd.org/en/data/datasets/pisa-2018-database.html; and OECD (2022[118]), PISA 2022 Database, www.oecd.org/en/data/datasets/pisa-2022-database.html.
References
[62] Artamonova, A. and A. Syse (2021), “Do older parents’ assistance needs deter parent-child geographic divergence in Norway?”, Health & Place, Vol. 70, p. 102599, https://doi.org/10.1016/j.healthplace.2021.102599.
[67] Atherton, O. et al. (2023), “Rural–urban differences in personality traits and well‐being in adulthood”, Journal of Personality, Vol. 92/1, pp. 73-87, https://doi.org/10.1111/jopy.12818.
[9] Attanasio, O., Á. de Paula and A. Toppeta (2024), Intergenerational mobility in socio-emotional skills, The IFS, https://doi.org/10.1920/wp.ifs.2024.2624.
[83] Autor, D. et al. (2023), “Males at the tails: How socioeconomic status shapes the gender gap”, The Economic Journal, Vol. 133/656, pp. 3136-3152, https://doi.org/10.1093/ej/uead069.
[40] Avvisati, F. and P. Givord (2023), “The learning gain over one school year among 15-year-olds: An international comparison based on PISA”, Labour Economics, Vol. 84, p. 102365, https://doi.org/10.1016/j.labeco.2023.102365.
[103] Bächmann, A. (2022), “Are female-dominated occupations a secure option? Occupational gender segregation, accompanied occupational characteristics, and the risk of becoming unemployed”, European Sociological Review, Vol. 39/6, pp. 876-889, https://doi.org/10.1093/esr/jcac068.
[92] Bacolod, M., B. Blum and W. Strange (2009), “Skills in the city”, Journal of Urban Economics, Vol. 65/2, pp. 136-153, https://doi.org/10.1016/j.jue.2008.09.003.
[126] Bakker, D. et al. (2009), “Adding typology to lexicostatistics: A combined approach to language classification”, Linguistic Typology, Vol. 13/1, https://doi.org/10.1515/lity.2009.009.
[95] Balducci, M. et al. (2024), “The gender-equality paradox in intraindividual academic strengths: A cross-temporal analysis”, Psychological Science, Vol. 35/11, pp. 1246-1259, https://doi.org/10.1177/09567976241271330.
[44] Barro, R. and J. Lee (2013), “A new data set of educational attainment in the world, 1950–2010”, Journal of Development Economics, Vol. 104, pp. 184-198, https://doi.org/10.1016/j.jdeveco.2012.10.001.
[80] Bernardi, F. (2014), “Compensatory advantage as a mechanism of educational inequality”, Sociology of Education, Vol. 87/2, pp. 74-88, https://doi.org/10.1177/0038040714524258.
[21] Bernardi, F. and M. Valdés (2021), “Sticky educational expectations: A cross-country comparison”, Research in Social Stratification and Mobility, Vol. 75, p. 100624, https://doi.org/10.1016/j.rssm.2021.100624.
[93] Berry, C. and E. Glaeser (2005), “The divergence of human capital levels across cities”, Papers in Regional Science, Vol. 84/3, pp. 407-444, https://doi.org/10.1111/j.1435-5957.2005.00047.x.
[16] Bodovski, K. (2013), “Adolescents’ emerging habitus: The role of early parental expectations and practices”, British Journal of Sociology of Education, Vol. 35/3, pp. 389-412, https://doi.org/10.1080/01425692.2013.776932.
[26] Borgonovi, F. (2022), “Is the literacy achievement of teenage boys poorer than that of teenage girls, or do estimates of gender gaps depend on the test? A comparison of PISA and PIAAC”, Journal of Educational Psychology, Vol. 114/2, pp. 239-256, https://doi.org/10.1037/edu0000659.
[28] Borgonovi, F., A. Choi and M. Paccagnella (2021), “The evolution of gender gaps in numeracy and literacy between childhood and young adulthood”, Economics of Education Review, Vol. 82, p. 102119, https://doi.org/10.1016/j.econedurev.2021.102119.
[96] Borgonovi, F., Á. Choi and M. Paccagnella (2018), “The evolution of gender gaps in numeracy and literacy between childhood and adulthood”, OECD Education Working Papers, No. 184, OECD Publishing, Paris, https://doi.org/10.1787/0ff7ae72-en.
[52] Borgonovi, F. and A. Ferrara (2020), “Academic achievement and sense of belonging among non-native-speaking immigrant students: The role of linguistic distance”, Learning and Individual Differences, Vol. 81, p. 101911, https://doi.org/10.1016/j.lindif.2020.101911.
[12] Borgonovi, F. and A. Pokropek (2021), “The evolution of socio‐economic disparities in literacy skills from age 15 to age 27 in 20 countries”, British Educational Research Journal, Vol. 47/6, pp. 1560-1586, https://doi.org/10.1002/berj.3738.
[37] Borgonovi, F. et al. (2017), “Youth in Transition: How Do Some of The Cohorts Participating in PISA Fare in PIAAC?”, OECD Education Working Papers, No. 155, OECD Publishing, Paris, https://doi.org/10.1787/51479ec2-en.
[15] Breen, R. and J. Goldthorpe (1997), “Explaining differences: Towards a formal rational action theory”, Rationality and Society, Vol. 9/3, pp. 275-305, https://doi.org/10.1177/104346397009003002.
[109] Bukodi, E. and J. Goldthorpe (2018), Social Inequality and Social Mobility: Is there an Inverse Relation?, Center for Open Science, https://doi.org/10.31235/osf.io/jkqne.
[72] Canache, D. et al. (2013), “Openness, extraversion and the intention to emigrate”, Journal of Research in Personality, Vol. 47/4, pp. 351-355, https://doi.org/10.1016/j.jrp.2013.02.008.
[47] Cathles, A. et al. (2021), “Where do you come from, where do you go? Assessing skills gaps and labour market outcomes for young adults with different immigration backgrounds”, International Journal of Educational Development, Vol. 86, p. 102466, https://doi.org/10.1016/j.ijedudev.2021.102466.
[59] Chetty, R. et al. (2018), The Opportunity Atlas: Mapping the Childhood Roots of Social Mobility, National Bureau of Economic Research, https://doi.org/10.3386/w25147.
[58] Chetty, R. et al. (2014), “Where is the land of opportunity? The geography of intergenerational mobility in the United States”, The Quarterly Journal of Economics, Vol. 129/4, pp. 1553-1623, https://doi.org/10.1093/qje/qju022.
[18] Chetty, R. et al. (2022), “Social capital II: Determinants of economic connectedness”, Nature, Vol. 608/7921, pp. 122-134, https://doi.org/10.1038/s41586-022-04997-3.
[82] Christoffersen, A. (2021), “The politics of intersectional practice: Competing concepts of intersectionality”, Policy & Politics, Vol. 49/4, pp. 573-593, https://doi.org/10.1332/030557321x16194316141034.
[42] Cohen, J. (2013), Statistical Power Analysis for the Behavioral Sciences, Routledge, https://doi.org/10.4324/9780203771587.
[108] Corak, M. (2013), “Income inequality, equality of opportunity, and intergenerational mobility”, Journal of Economic Perspectives, Vol. 27/3, pp. 79-102, https://doi.org/10.1257/jep.27.3.79.
[81] Crenshaw, K. (1991), “Mapping the margins: Intersectionality, identity politics, and violence against women of color”, Stanford Law Review, Vol. 43/6, p. 1241, https://doi.org/10.2307/1229039.
[3] Cunha, F. and J. Heckman (2007), “The Technology of Skill Formation”, American Economic Review, Vol. 97/2, pp. 31-47, https://doi.org/10.1257/aer.97.2.31.
[4] Demange, P. et al. (2022), “Estimating effects of parents’ cognitive and non-cognitive skills on offspring education using polygenic scores”, Nature Communications, Vol. 13/1, p. 4801, https://doi.org/10.1038/s41467-022-32003-x.
[13] Dickson, M., P. Gregg and H. Robinson (2016), “Early, late or never? When does parental education impact child outcomes?”, The Economic Journal, Vol. 126/596, pp. F184-F231, https://doi.org/10.1111/ecoj.12356.
[73] Dohmen, T. et al. (2011), “The intergenerational transmission of risk and trust attitudes”, The Review of Economic Studies, Vol. 79/2, pp. 645-677, https://doi.org/10.1093/restud/rdr027.
[70] Echazarra, A. and T. Radinger (2019), Learning in rural schools: Insights from PISA, TALIS and the literature, OECD Education Working Papers, OECD Publishing, Paris, https://doi.org/10.1787/8b1a5cb9-en.
[87] Eckel, C. and S. Füllbrunn (2015), “Thar SHE blows? Gender, competition, and bubbles in experimental asset markets”, American Economic Review, Vol. 105/2, pp. 906-920, https://doi.org/10.1257/aer.20130683.
[85] Ellison, G. and A. Swanson (2010), “The gender gap in secondary school mathematics at high achievement levels: Evidence from the American mathematics competitions”, Journal of Economic Perspectives, Vol. 24/2, pp. 109-128, https://doi.org/10.1257/jep.24.2.109.
[5] England-Mason, G. and A. Gonzalez (2020), “Intervening to shape children’s emotion regulation: A review of emotion socialization parenting programs for young children”, Emotion, Vol. 20/1, pp. 98-104, https://doi.org/10.1037/emo0000638.
[33] Falk, A. and J. Hermle (2018), “Relationship of gender differences in preferences to economic development and gender equality”, Science, Vol. 362/6412, https://doi.org/10.1126/science.aas9899.
[10] Farkas, G. (2003), “Cognitive skills and noncognitive traits and behaviors in stratification processes”, Annual Review of Sociology, Vol. 29/1, pp. 541-562, https://doi.org/10.1146/annurev.soc.29.010202.100023.
[63] Ferreira, F., J. Gyourko and J. Tracy (2011), Housing busts and household mobility: An update, National Bureau of Economic Research, https://doi.org/10.3386/w17405.
[86] Fryer, R. and S. Levitt (2010), “An empirical analysis of the gender gap in mathematics”, American Economic Journal: Applied Economics, Vol. 2/2, pp. 210-240, https://doi.org/10.1257/app.2.2.210.
[64] Ganong, P. and D. Shoag (2017), “Why has regional income convergence in the U.S. declined?”, Journal of Urban Economics, Vol. 102, pp. 76-90, https://doi.org/10.1016/j.jue.2017.07.002.
[65] Gobillon, L. and H. Seold (2021), Spatial mismatch, poverty, and vulnerable populations, Springer Berlin Heidelberg, https://www.parisschoolofeconomics.eu/en/publications-hal/spatial-mismatch-poverty-and-vulnerable-populations/.
[75] Grönqvist, E., B. Öckert and J. Vlachos (2016), “The intergenerational transmission of cognitive and noncognitive abilities”, Journal of Human Resources, Vol. 52/4, pp. 887-918, https://doi.org/10.3368/jhr.52.4.0115-6882r1.
[11] Gruijters, R., I. Raabe and N. Hübner (2023), “Socio-emotional skills and the socioeconomic achievement gap”, Sociology of Education, Vol. 97/2, pp. 120-147, https://doi.org/10.1177/00380407231216424.
[6] Grusec, J. and M. Davidov (2019), “Parent Socialization and Children’s Values”, in Handbook of Parenting, Routledge, https://doi.org/10.4324/9780429433214-22.
[7] Guhin, J., J. Calarco and C. Miller-Idriss (2021), “Whatever Happened to Socialization?”, Annual Review of Sociology, Vol. 47/1, pp. 109-129, https://doi.org/10.1146/annurev-soc-090320-103012.
[94] Guiso, L. et al. (2008), “Culture, gender, and math”, Science, Vol. 320/5880, pp. 1164-1165, https://doi.org/10.1126/science.1154094.
[111] Hakkarainen, A., L. Holopainen and H. Savolainen (2013), “Mathematical and Reading Difficulties as Predictors of School Achievement and Transition to Secondary Education”, Scandinavian Journal of Educational Research, Vol. 57/5, pp. 488-506, https://doi.org/10.1080/00313831.2012.696207.
[39] Hanushek, E. et al. (2025), “Age and cognitive skills: Use it or lose it”, Science Advances, Vol. 11/10, https://doi.org/10.1126/sciadv.ads1560.
[1] Hanushek, E., S. Machin and L. Woessmann (eds.) (2011), The Economics of International Differences in Educational Achievement, Elsevier B.V., https://hanushek.stanford.edu/sites/default/files/publications/Hanushek%2BWoessmann%202011%20HandEconEduc.pdf.
[97] Herlitz, A. et al. (2024), “A systematic review and new analyses of the gender-equality paradox”, Perspectives on Psychological Science, Vol. 20/3, pp. 503-539, https://doi.org/10.1177/17456916231202685.
[22] Holm, A., A. Hjorth-Trolle and M. Jæger (2019), “Signals, educational decision-making, and inequality”, European Sociological Review, Vol. 35/4, pp. 447-460, https://doi.org/10.1093/esr/jcz010.
[36] Huang, C. (2012), “Gender differences in academic self-efficacy: A meta-analysis”, European Journal of Psychology of Education, Vol. 28/1, pp. 1-35, https://doi.org/10.1007/s10212-011-0097-y.
[24] Hyde, J. (2014), “Gender similarities and differences”, Annual Review of Psychology, Vol. 65/1, pp. 373-398, https://doi.org/10.1146/annurev-psych-010213-115057.
[125] IEA (2023), ICILS 2023 Database, International Computer and Information Literacy Study, International Association for the Evaluation of Educational Achievement, https://doi.org/10.58150/ICILS_2023_data.
[124] IEA (2022), ICCS 2022 Database, International Civic and Citizenship Education Study, International Association for the Evaluation of Educational Achievement, https://doi.org/10.58150/ICCS_2022_data_edition_2_including_process_data.
[48] Isphording, I. (2014), “Disadvantages of linguistic origin - Evidence from immigrant literacy scores”, Economics Letters, Vol. 123/2, pp. 236-239, https://doi.org/10.1016/j.econlet.2014.02.013.
[99] Jansen, M., M. Becker and M. Neumann (2021), “Dimensional comparison effects on (gendered) educational choices”, Journal of Educational Psychology, Vol. 113/2, pp. 330-350, https://doi.org/10.1037/edu0000524.
[104] Keane, C., H. Russell and E. Smyth (2017), “Female participation increases and gender segregation”, ESRI Working Paper, No. 564, The Economic and Social Research Institute, https://www.esri.ie/publications/female-participation-increases-and-gender-segregation.
[61] Kimbrough, E. (ed.) (2025), “Big cities fuel inequality within and across generations”, PNAS Nexus, Vol. 4/2, https://doi.org/10.1093/pnasnexus/pgae587.
[41] Kraft, M. (2020), “Interpreting Effect Sizes of Education Interventions”, Educational Researcher, Vol. 49/4, pp. 241-253, https://doi.org/10.3102/0013189x20912798.
[53] Kuperman, V. (2022), “Effects of linguistic distance on cognitive skills, health, and social outcomes in Canadian immigrants”, Frontiers in Political Science, Vol. 4, https://doi.org/10.3389/fpos.2022.874195.
[17] Lareau, A. (2011), Unequal Childhoods: Class, Race, and Family Life, University of California Press, http://www.jstor.org/stable/10.1525/j.ctt1ppgj4.
[78] Lee, H. et al. (2025), “Mechanisms of cultural diversity in urban populations”, Nature Communications, Vol. 16/1, https://doi.org/10.1038/s41467-025-60538-2.
[8] Leerkes, E., L. Bailes and M. Augustine (2020), “The intergenerational transmission of emotion socialization”, Developmental Psychology, Vol. 56/3, pp. 390-402, https://doi.org/10.1037/dev0000753.
[84] Legewie, J. and T. DiPrete (2012), “School context and the gender gap in educational achievement”, American Sociological Review, Vol. 77/3, pp. 463-485, https://doi.org/10.1177/0003122412440802.
[55] Lehtonen, M. et al. (2018), “Is bilingualism associated with enhanced executive functioning in adults? A meta-analytic review”, Psychological Bulletin, Vol. 144/4, pp. 394-425, https://doi.org/10.1037/bul0000142.
[89] Leslie, S. et al. (2015), “Expectations of brilliance underlie gender distributions across academic disciplines”, Science, Vol. 347/6219, pp. 262-265, https://doi.org/10.1126/science.1261375.
[74] Liefbroer, A. and C. Elzinga (2012), “Intergenerational transmission of behavioural patterns: How similar are parents’ and children’s demographic trajectories?”, Advances in Life Course Research, Vol. 17/1, pp. 1-10, https://doi.org/10.1016/j.alcr.2012.01.002.
[54] Li, P., J. Legault and K. Litcofsky (2014), “Neuroplasticity as a function of second language learning: Anatomical changes in the human brain”, Cortex, Vol. 58, pp. 301-324, https://doi.org/10.1016/j.cortex.2014.05.001.
[30] Malin, A. (ed.) (2014), Associations between age and cognitive foundation skills in the Nordic countries – A closer look at the data, University of Jyväskylä, The Finnish Institute for Educational Research, https://pure.vive.dk/ws/files/293064/d112.pdf.
[105] McCaughey, M. (2023), No one left behind? Gender segregation in the workplace sees women losing out, https://www.eurofound.europa.eu/en/blog/2023/no-one-left-behind-gender-segregation-workplace-sees-women-losing-out (accessed on 17 February 2025).
[77] Militaru, I. et al. (2023), “The lay of the land: Associations between environmental features and personality”, Journal of Personality, Vol. 92/1, pp. 88-110, https://doi.org/10.1111/jopy.12822.
[88] Niederle, M. and L. Vesterlund (2007), “Do women shy away from competition? Do men compete too much?”, The Quarterly Journal of Economics, Vol. 122/3, pp. 1067-1101, https://doi.org/10.1162/qjec.122.3.1067.
[51] Norman, D. and D. Bobrow (1975), “On data-limited and resource-limited processes”, Cognitive Psychology, Vol. 7/1, pp. 44-64, https://doi.org/10.1016/0010-0285(75)90004-3.
[57] Nucette, A. et al. (2024), “Can learning a new language make you better at maths? A meta-analysis of foreign language learning and numeracy skills during early adolescence”, Bilingualism: Language and Cognition, pp. 1-31, https://doi.org/10.1017/s1366728924000701.
[60] OECD (2025), To Have and Have Not – How to Bridge the Gap in Opportunities, OECD Publishing, Paris, https://doi.org/10.1787/dec143ad-en.
[91] OECD (2024), Education Policy Outlook 2024: Reshaping Teaching into a Thriving Profession from ABCs to AI, OECD Publishing, Paris, https://doi.org/10.1787/dd5140e4-en.
[110] OECD (2024), Income Distribution Database, (database), Data Explorer, http://data-explorer.oecd.org/s/17a (accessed on 12 February 2025).
[45] OECD (2024), International Migration Outlook 2024, OECD Publishing, Paris, https://doi.org/10.1787/50b0353e-en.
[23] OECD (2024), PIAAC 2nd cycle database, Survey of Adult Skills, OECD, Paris, https://www.oecd.org/en/data/datasets/piaac-2nd-cycle-database.html (accessed on 10 December 2024).
[123] OECD (2024), PISA 2022 Results (Volume III): Creative Minds, Creative Schools, PISA, OECD Publishing, Paris, https://doi.org/10.1787/765ee8c2-en.
[20] OECD (2024), PISA 2022 Results (Volume V): Learning Strategies and Attitudes for Life, PISA, OECD Publishing, Paris, https://doi.org/10.1787/c2e44201-en.
[122] OECD (2023), PISA 2022 Assessment and Analytical Framework, PISA, OECD Publishing, Paris, https://doi.org/10.1787/dfe0bf9c-en.
[128] OECD (2023), PISA 2022 Results (Volume I): The State of Learning and Equity in Education, PISA, OECD Publishing, Paris, https://doi.org/10.1787/53f23881-en.
[118] OECD (2022), PISA 2022 Database, OECD, Paris, http://www.oecd.org/en/data/datasets/pisa-2022-database.html (accessed on 5 November 2024).
[14] OECD (2021), OECD Skills Outlook 2021: Learning for Life, OECD Publishing, Paris, https://doi.org/10.1787/0ae365b4-en.
[121] OECD (2020), PISA 2018 Results (Volume VI): Are Students Ready to Thrive in an Interconnected World?, PISA, OECD Publishing, Paris, https://doi.org/10.1787/d5f68679-en.
[71] OECD (2019), Providing Quality Early Childhood Education and Care: Results from the Starting Strong Survey 2018, OECD Publishing, Paris, https://doi.org/10.1787/301005d1-en.
[27] OECD (2019), Skills Matter: Additional Results from the Survey of Adult Skills, OECD Skills Studies, OECD Publishing, Paris, https://doi.org/10.1787/1f029d8f-en.
[117] OECD (2018), PISA 2018 Database, OECD, Paris, http://www.oecd.org/en/data/datasets/pisa-2018-database.html (accessed on 5 November 2024).
[46] OECD (2018), Skills on the Move: Migrants in the Survey of Adult Skills, OECD Skills Studies, OECD Publishing, Paris, https://doi.org/10.1787/9789264307353-en.
[68] OECD (2017), “Country roads: Education and rural life”, Trends Shaping Education Spotlight 9, OECD, Paris, https://www.oecd.org/content/dam/oecd/en/publications/reports/2017/04/country-roads_3e5f7ba4/ea43a39d-en.pdf.
[107] OECD (2017), PIAAC 1st Cycle Database, https://www.oecd.org/en/data/datasets/PIAAC-1st-Cycle-Database.html (accessed on 14 November 2024).
[120] OECD (2017), PISA 2015 Results (Volume V): Collaborative Problem Solving, PISA, OECD Publishing, Paris, https://doi.org/10.1787/9789264285521-en.
[43] OECD (2016), Skills Matter: Further Results from the Survey of Adult Skills, OECD Skills Studies, OECD Publishing, Paris, https://doi.org/10.1787/9789264258051-en.
[116] OECD (2015), PISA 2015 Database, OECD, Paris, http://www.oecd.org/en/data/datasets/pisa-2015-database.html (accessed on 5 November 2024).
[35] OECD (2015), The ABC of Gender Equality in Education: Aptitude, Behaviour, Confidence, PISA, OECD Publishing, Paris, https://doi.org/10.1787/9789264229945-en.
[119] OECD (2014), PISA 2012 Results: Creative Problem Solving (Volume V): Students’ Skills in Tackling Real-Life Problems, PISA, OECD Publishing, Paris, https://doi.org/10.1787/9789264208070-en.
[115] OECD (2012), PISA 2012 Database, OECD, Paris, http://www.oecd.org/en/data/datasets/pisa-2012-database.html (accessed on 5 November 2024).
[114] OECD (2006), PISA 2006 Database, OECD, Paris, https://www.oecd.org/en/data/datasets/pisa-2006-database.html (accessed on 14 November 2024).
[113] OECD (2003), PISA 2003 Database, OECD, Paris, http://www.oecd.org/en/data/datasets/pisa-2003-database.html.
[32] OECD/Statistics Canada (2005), Learning a Living: First Results of the Adult Literacy and Life Skills Survey, OECD Publishing, Paris, https://doi.org/10.1787/9789264010390-en.
[19] Owens, A., S. Reardon and C. Jencks (2016), “Income segregation between schools and school districts”, American Educational Research Journal, Vol. 53/4, pp. 1159-1197, https://doi.org/10.3102/0002831216652722.
[79] Paccagnella, M. (2016), “Age, Ageing and Skills: Results from the Survey of Adult Skills”, OECD Education Working Papers, No. 132, OECD Publishing, Paris, https://doi.org/10.1787/5jm0q1n38lvc-en.
[49] Paradis, J. (2011), “Individual differences in child English second language acquisition”, Linguistic Approaches to Bilingualism, Vol. 1/3, pp. 213-237, https://doi.org/10.1075/lab.1.3.01par.
[101] Permanyer, I. (2009), “The measurement of multidimensional gender inequality: Continuing the debate”, Social Indicators Research, Vol. 95/2, pp. 181-198, https://doi.org/10.1007/s11205-009-9463-4.
[38] Rebollo-Sanz, Y. and S. De la Rica (2020), “Gender gaps in skills and labor market outcomes: Evidence from the PIAAC”, Review of Economics of the Household, Vol. 20/2, pp. 333-371, https://doi.org/10.1007/s11150-020-09523-w.
[25] Rebollo-Sanz, Y. and S. de la Rica (2020), From Gender Gaps in Skills to Gender Gaps in Wages: Evidence from PIAAC, Working Papers 20.09, Universidad Pablo de Olavide, Department of Economics, https://ideas.repec.org/p/pab/wpaper/20.09.html.
[69] Rentfrow, P., S. Gosling and J. Potter (2008), “A theory of the emergence, persistence, and expression of geographic variation in psychological characteristics”, Perspectives on Psychological Science, Vol. 3/5, pp. 339-369, https://doi.org/10.1111/j.1745-6924.2008.00084.x.
[100] Richardson, S. et al. (2020), “Is there a gender-equality paradox in science, technology, engineering, and math (STEM)? Commentary on the study by Stoet and Geary (2018)”, Psychological Science, Vol. 31/3, pp. 338-341, https://doi.org/10.1177/0956797619872762.
[31] Rosdahl, A. (2014), Fra 15 til 27 år: PISA 2000-eleverne i 2011/12 [From 15 to 27 years of age: The PISA 2000 students in 2011/12], Det Nationale Forskningscenter for Velfærd [The Danish Center for Social Science Research], https://www.vive.dk/da/udgivelser/fra-15-aar-til-27-aar-yvrg7kzb/.
[34] Silverman, I. (2003), “Gender differences in delay of gratification: A meta-analysis”, Sex Roles, Vol. 49/9/10, pp. 451-463, https://doi.org/10.1023/a:1025872421115.
[29] Solheim, O. and K. Lundetræ (2016), “Can test construction account for varying gender differences in international reading achievement tests of children, adolescents and young adults? – A study based on Nordic results in PIRLS, PISA and PIAAC”, Assessment in Education: Principles, Policy & Practice, Vol. 25/1, pp. 107-126, https://doi.org/10.1080/0969594x.2016.1239612.
[66] Spring, A., B. Gillespie and C. Mulder (2023), “Internal migration following adverse life events: Assessing the likelihood of return migration and migration toward family”, Population, Space and Place, Vol. 30/3, https://doi.org/10.1002/psp.2711.
[98] Stoet, G. and D. Geary (2018), “The gender-equality paradox in science, technology, engineering, and mathematics education”, Psychological Science, Vol. 29/4, pp. 581-593, https://doi.org/10.1177/0956797617741719.
[90] Storage, D. et al. (2020), “Adults and children implicitly associate brilliance with men more than women”, Journal of Experimental Social Psychology, Vol. 90, p. 104020, https://doi.org/10.1016/j.jesp.2020.104020.
[50] Takano, Y. and A. Noda (1995), “Interlanguage dissimilarity enhances the decline of thinking ability during foreign language processing”, Language Learning, Vol. 45/4, pp. 657-681, https://doi.org/10.1111/j.1467-1770.1995.tb00457.x.
[106] Torre, M. (2018), “Stopgappers? The Occupational Trajectories of Men in Female-Dominated Occupations”, Work and Occupations, Vol. 45/3, pp. 283-312, https://doi.org/10.1177/0730888418780433.
[76] Tucker, L. (2023), “Open-mindedness: A double-edged sword in education”, Theory and Research in Education, Vol. 21/3, pp. 241-263, https://doi.org/10.1177/14778785231203100.
[102] UNDP (2025), Gender Inequality Index (GII), United Nations Development Programme, https://hdr.undp.org/data-center/thematic-composite-indices/gender-inequality-index#/indicies/GII (accessed on 13 February 2025).
[2] Van de Werfhorst, H. and J. Mijs (2010), “Achievement Inequality and the Institutional Structure of Educational Systems: A Comparative Perspective”, Annual Review of Sociology, Vol. 36/1, pp. 407-428, https://doi.org/10.1146/annurev.soc.012809.102538.
[112] Wang, X. (2013), “Why Students Choose STEM Majors”, American Educational Research Journal, Vol. 50/5, pp. 1081-1121, https://doi.org/10.3102/0002831213488622.
[56] Ware, C. et al. (2021), “Does second language learning promote neuroplasticity in aging? A systematic review of cognitive and neuroimaging studies”, Frontiers in Aging Neuroscience, Vol. 13, https://doi.org/10.3389/fnagi.2021.706672.
[127] Wichmann, S. et al. (2010), “Evaluating linguistic distance measures”, Physica A: Statistical Mechanics and its Applications, Vol. 389/17, pp. 3632-3639, https://doi.org/10.1016/j.physa.2010.05.011.
← 1. Differences in the standardised effect size “Cohen’s d” benchmarks are interpreted as small effects if Cohen’s d is below 0.2, as medium for values between 0.2 and 0.5, and as large for valued of 0.5 and greater.
← 2. The Inequality of Opportunity measure is calculated using market income whereas total income inequality is calculated using disposable income, a difference that may explain discrepancies between estimates.
← 3. PISA scores have a standard deviation of around 100 score points (OECD, 2023[128]).