Spending Better for Children through Social Policy

Report

5 October 2026

4. The value for money of social policies in promoting educational and health outcomes

Copy link to 4. The value for money of social policies in promoting educational and health outcomes

Abstract

This chapter assesses the “value for money” of social policies in promoting the educational and self-reported health outcomes of 15‑year‑old adolescents. It examines how spending across policy areas, income transfers, as well as the timing of family cash benefits and education spending across childhood, affects the effectiveness of social expenditure and influences future spending pressures. It first reviews key trends in adolescents’ PISA performance and self-reported health, highlighting declines since the late 2000s and the persistence of strong socio‑economic inequalities in both the risks of low academic and health outcomes and the chances of achieving higher levels. The analysis points to the need both to prevent poor outcomes and to expand opportunities for adolescents from low and middle socio‑economic background to achieve high levels of academic performance, as a way to combine greater fairness with stronger preparation for increasingly technology-intensive societies. It highlights the importance of tailoring support to the needs of disadvantaged children and stresses that preventing disadvantage from accumulating early in life, while sustaining support throughout childhood and adolescence, is essential to improve educational and health outcomes and strengthen opportunities for social mobility later in life.

Children’s educational and health outcomes constitute essential dimensions of their immediate well-being and can have profound implications for their future life trajectories. A substantial body of research demonstrates that cognitive development and educational attainment in childhood strongly shape later educational pathways, field-of-study choices, labour market outcomes, and broader social and economic outcomes in adulthood (OECD, 2025[1]). Similarly, childhood health conditions are closely associated with adult health status, including the prevalence of chronic diseases, functional limitations, and overall life expectancy (Mallo and Wolfe, 2020[2]; Flores, García-Gómez and Kalwij, 2020[3]).

Educational and health trajectories are also strongly shaped by socio‑economic background, with inequalities emerging early in childhood and persisting over time. These disparities are driven in part by differences in families’ economic and social resources, which influence the quality of the home learning environment, housing conditions, nutrition, and parents’ capacity to support children’s development. Socio‑economically disadvantaged families are also more likely to live in deprived neighbourhoods characterised by poorer housing quality, greater environmental risks, and more limited access to high-quality services, including education, healthcare, childcare, green spaces, and recreational and sports infrastructure (OECD, 2025[4]; 2025[5]). Together, household resources and neighbourhood characteristics interact to shape children’s opportunities and are strongly associated with educational achievement and health outcomes across the life course (Nieuwenhuis and Hooimeijer, 2016[6]; Edwards, 2005[7]; Troost, van Ham and Manley, 2023[8]; Robinette, Charles and Gruenewald, 2017[9]; Rebouças, Falcão and Barreto, 2021[10]; Palloni et al., 2009[11]; Pearce et al., 2019[12]; Pickett et al., 2022[13]).

This chapter complements the analysis presented in the previous chapter on child poverty by examining the impact of social policies on children’s educational and health outcomes. Social policies may contribute to improvements in education and health indirectly through the reduction of economic disadvantage, while also exerting direct effects, for example through the provision of healthcare services or by promoting early literacy and numeracy development within early childhood education and care programmes.

Taken together, the two chapters provide a more comprehensive understanding of how social policies shape child well-being across multiple dimensions, including poverty, educational achievement, and self-reported health. They also highlight the potential complementarities and trade‑offs policymakers face when allocating resources across different policy areas, choosing between modes of support such as cash transfers and services, and determining which population groups should be prioritised for support.

The remainder of this chapter is structured as follows: Section 4.1 describes recent trends in children’s educational and health outcomes across selected measures, and their variation across socio-economic groups. Section 4.2 highlights key adaptations to the analytical value for money estimation framework to account for the cumulative effect of past social policies on adolescents’ educational and health outcomes in the present. The following two sections describe the results of the analysis for educational and health outcomes, respectively, and section 4.5 concludes with a discussion of the main policy challenges in designing social policies that promote children’s educational and health outcomes and their implications for social spending pressures.

4.1. Trends in children’s educational and health outcomes

Copy link to 4.1. Trends in children’s educational and health outcomes

Raising student academic performance while ensuring fairness

Education has long been regarded as a key driver of social mobility and equality of opportunity, with evidence showing that educational expansion and broader access to secondary and tertiary education have contributed to upward mobility across many advanced economies. Research comparing learning progression during the school year and summer holidays also suggests that schools partially compensate for inequalities in children’s home learning environments, as socio-economic achievement gaps tend to widen more when schools are closed (Box 4.1). However, the equalizing potential of education remains limited by inequalities that emerge early in childhood, well before formal schooling begins. Children from disadvantaged backgrounds often enter school with lower cognitive and socio‑emotional skills, and students from high socio-economic status families are more likely to overcome poor academic performance through access to greater economic, social and cultural resources, including tutoring, institutional support and second educational chances. Overall, while education remains an important mechanism for promoting upward mobility, schooling alone is often insufficient to fully offset broader socio-economic inequalities that shape children’s life chances from an early age.

At the same time, many OECD countries have experienced a stagnation or decline in students’ average academic performance following an earlier period of improvement. In many cases, this deterioration has been driven primarily by an expansion of the lower tail of the achievement distribution, with a growing share of students failing to reach baseline proficiency levels. However, several countries have also recorded substantial declines in the proportion of high-performing students (OECD, 2023[14]). These trends have raised concerns about future shortages of highly skilled workers and about countries’ capacity to sustain innovation and remain competitive in increasingly technology-intensive economies. In a context of rapid technological change, demand is expected to grow particularly strongly for individuals with advanced STEM skills, which are widely seen as essential for innovation and productivity growth in frontier sectors such as artificial intelligence, cybersecurity and other cutting-edge digital technologies (European Commission, 2020[15]). Strong reading, scientific and mathematical literacy are also increasingly important not only for labour market success but also for active participation in society and democratic life. These skills enable individuals to critically assess information, distinguish evidence from misinformation, and engage constructively in public debate.

Against this background, a central policy challenge is to reconcile efficiency and fairness: raising overall levels of academic performance while ensuring that children from low- and middle‑socio-economic backgrounds have equal opportunities to reach high levels of achievement. Greater fairness and social cohesion require providing support that reduces the share of children falling behind academically, while ensuring that low-SES children – who make up a large proportion of low performers – receive the stronger support they need to realise their potential. It also requires expanding opportunities to achieve high levels of performance, particularly for low- and middle‑SES children, who are far less likely than their more advantaged peers to become high achievers. This may further spur economic growth, as there is evidence that it depends not only on average cognitive skills, but also on the share of high-performing students who are more likely to become scientists, engineers, entrepreneurs and highly skilled workers in innovation-intensive sectors (Hanushek and Woessmann, 2008[16]; 2015[17]).

The strength and nature of social stratification in educational trajectories vary substantially across countries depending on institutional contexts, influencing the extent to which socio-economic background affects the risk of poor educational or health outcomes, as well as the likelihood of achieving high levels of performance and well-being (Dräger et al., 2024[18]). For this reason, assessing equality of opportunity requires going beyond average outcomes and examining how children’s chances of performing high academically are associated with their family background. At the same time, reducing the share of low-performing students remains most critical, both to improve individual and collective educational outcomes and to limit the risks of school disengagement, early school leaving, and later exclusion from employment, education or training.

Box 4.1. Education as the great equalizer?

Copy link to Box 4.1. Education as the great equalizer?

Education has long been regarded as the central mechanism through which modern societies can promote equality of opportunity and foster upward social mobility. Yet cross-country evidence presents a more complex picture. While countries with higher income inequality also tend to display lower intergenerational mobility – consistent with the “Great Gatsby Curve” (Corak, 2013[19]; OECD, 2018[20]) – the relationship between economic inequality and inequality in educational attainment by social background appears much weaker (Bernardi and Plavgo, 2019[21]). In other words, countries with relatively similar levels of educational inequality may nevertheless display very different levels of social mobility. This suggests that education remains an important channel of mobility, but not necessarily a sufficient equalizer of life chances on its own.

Evidence from the social stratification literature shows that inequalities in educational attainment by social background declined substantially across many high human development countries during the second half of the 20th century (Breen et al., 2009[22]; Bernardi and Ballarino, 2016[23]; Barone and Ruggera, 2018[24]). Educational expansion, particularly at the secondary level, enabled broader access to qualifications that were previously restricted to more advantaged groups. Greater participation in higher education also contributed to increased social mobility, as the link between parental background and occupational attainment tends to weaken among individuals with tertiary education (Breen and Jonsson, 2007). These developments were particularly important for cohorts entering adulthood in the decades following the Second World War, when educational expansion coincided with strong economic growth and relatively low-income inequality.

Research comparing learning progression during the school year and summer holidays also suggests that schooling exerts an equalizing effect. Studies exploiting this “natural experiment” find that achievement gaps by socio-economic background tend to widen more rapidly during school holidays than during the school year, when children are exposed to formal schooling (Downey and Condron, 2016[25]; von Hippel, Workman and Downey, 2018[26]). While the equalizing effect observed during the school year is generally modest, these findings nevertheless indicate that schools partially compensate for inequalities in children’s home learning environments.

At the same time, several factors limit the equalizing potential of education systems. First, inequalities emerge early in life, well before formal schooling begins. Children from disadvantaged socio-economic backgrounds often enter school with significantly lower average literacy, numeracy and socio‑emotional skills than their more advantaged peers (Heckman, 2006[27]; Bradbury and Corak, 2015[28]). As a result, schools frequently operate on top of already substantial developmental gaps rather than starting from equal conditions.

Second, families with higher socio-economic status are often better able to shield their children from the consequences of poor academic performance. Research on “compensatory advantage” shows that students from advantaged families are more likely to continue progressing through the education system despite weak school performance, benefiting from access to private tutoring, remedial schooling, stronger parental guidance and broader social and cultural resources (Bernardi, 2014[29]; Bernardi and Triventi, 2020[30]; Yastrebov, Kosyakova and Kurakin, 2018[31]). Similar mechanisms operate in the labour market: individuals from advantaged backgrounds who fail to obtain high educational credentials are nevertheless more likely to avoid downward occupational mobility and access relatively secure and well-paid employment (Bernardi and Ballarino, 2014[32]).

Academic performance is declining but fewer youths are NEETs

Data from the Programmes for International Student Assessment (PISA) provide information on 15‑year‑olds academic achievement. Together, the share of students who are high achievers in PISA and those who are low performers – defined as those attaining Level 5 or 6 (high) or below Level 2 (low) in at least one of the three core subjects of reading, mathematics, or science – offer information on performance across the learnings distribution. Overall, the data suggest a decline in academic performance from 2009 to 2025, with the deterioration being particularly pronounced at the bottom of the distribution.

Across the OECD, more than four in ten children (43%, see Figure 4.1, Panel A) perform below baseline proficiency levels in PISA assessments, and this share has increased in nearly all countries since the late 2000s. However, cross-country differences remain substantial. In the 2025 assessment, the proportion of low-performing students was lowest in Japan and Estonia, where it remained below 25%, and highest in several Latin American countries, including Colombia, Mexico and Costa Rica, where it reached over 70% (OECD, 2026[33]). Across the OECD, the share of low-performing students increased by nearly 40% between 2009 and 2025, rising from 31% to 43%. This deterioration has been observed in almost all OECD countries, with the notable exceptions of Türkiye (+7 p.p.). The increase in low performance has been particularly marked in Iceland, Finland, the Netherlands, Norway and Greece, where the share of students performing below baseline proficiency levels rose by more than 20 p.p. over the period.

The share of 15‑year‑old students who are high achievers in PISA varies widely across OECD countries, ranging from nearly 30% in Korea and Japan to below 1% in Costa Rica and Mexico (Figure 4.1, Panel B). On average across the OECD, the proportion of high-performing students is over three times smaller than the share of low performers. In most OECD countries, the share of high performers declined between 2009 and 2025, falling on average from 15.1% to 11.9%. However, trends differ substantially across countries. The sharpest declines in rates of high achievers were observed in Finland, Iceland, Belgium, Switzerland, France and Germany, with reductions of around ten p.p. that reach 15 p.p. in Finland. By contrast, a small number of countries recorded increases in the share of high-performing students over the same period, including Türkiye, the United States and the United Kingdom, where increases of around 3 to 5 p.p. were recorded.

The share of young people who are neither in employment, education nor training (NEET) captures a population that is disengaged both from the education system and from the labour market. This group is of particular concern because transitions into stable employment are often more difficult for NEET youth (Ripamonti, 2023[34]; Rahmani and Groot, 2023[35]; Thompson, 2017[36]; Bynner and Parsons, 2002[37]), and periods of disengagement at young ages can generate long-lasting “scarring effects” on future employment prospects, earnings trajectories, and broader social outcomes. NEET youth also face elevated risks of poverty, poor health, and social exclusion.

Across the OECD, 13% of 15‑29 year‑olds are NEETs (Figure 4.1, Panel C). Encouragingly, the data show declining NEETs rates in most countries, down from an OECD average of 16% in 2010. The largest reductions, exceeding ten p.p., were recorded in Latvia, Israel, Ireland and Türkiye. Only Luxembourg observed a substantial increase in NEETs rates by 3.5 p.p. during this period. Lithuania, the United Kingdom, Israel, Belgium, Colombia and Iceland experienced increases of around one to four p.p. since 2017, although their current rates remain below 2010 levels. Today, the lowest rates are found in the Netherlands, Norway, Iceland, Sweden, Slovenia, Czechia, Ireland, Germany, Latvia and Australia, ranging from 6% to 10%, which compare to rates above 20% in Türkiye, Colombia and Costa Rica. Against a background of large and growing rates of low academic performers at age 15, the reductions in NEETs rates observed in most countries point towards education, labour market and social policies that were broadly effective at supporting those most at risk of disengagement.

Figure 4.1. Academic performance is deteriorating, but youth NEET rates are decreasing
Copy link to Figure 4.1. Academic performance is deteriorating, but youth NEET rates are decreasing

Socio‑economically disadvantaged children experience worse education

Promoting social mobility requires addressing the socio-economic factors that that both increase the risk low academic performance and hamper the ability of children from disadvantaged backgrounds to reach high levels of academic performance.1

Figure 4.2, Panel A shows that students from families with low socio‑economic status have a high probability to be low performers in the PISA test, ranging from less than 35% in Japan to around 90% in Colombia and Mexico. This is due to both, the relatively high share of low performers overall (43% across the OECD) and the highly stratified educational outcomes. Over half (58%) of all 15‑year‑olds in the bottom 25% of socio‑economic status are low performers in PISA, compared to 38% of adolescents with medium socio‑economic status and 24% of students with the highest 25% of socio‑economic status. In every OECD county, adolescents with low socio‑economic status are at least 60% more likely to be low performers compared to socio‑economically advantaged students, an absolute difference that is above 20 p.p. in all countries. Low academic achievers are not drawn exclusively from the most disadvantaged groups: nearly half (49%) come from medium socio-economic backgrounds, whereas disadvantaged students accounted for 36% of low performers in 2025.

Symmetrically, on average, only 5% of 15‑year‑olds from the bottom quarter of the socio-economic distribution are high performers, compared with 12% among students from middle socio-economic backgrounds and 23% among those from the top quarter (Figure 4.2, Panel B). In all OECD countries, students from high socio-economic backgrounds are at least three times more likely to be high achievers than their disadvantaged peers, while the gap exceeds 10 p.p. in all countries except Mexico, Colombia, Greece, Chile and Iceland, where overall shares of high performers are comparatively low. In most countries the gap in high performance between students from high and middle socio-economic backgrounds is larger than the gap between middle‑ and low-status students, suggesting that highly advantaged family environments may provide particularly strong conditions for reaching top levels of achievement.

The idea that promoting upward educational mobility among disadvantaged students necessarily comes at the expense of overall school system performance is not supported by cross-country evidence. At the national level, countries with larger shares of high-performing socio-economically advantaged students also tend to be those in which disadvantaged students are more likely to achieve high levels of performance (correlation coefficient among OECD countries r = 0.82). This suggests that, at country level, policies aimed at increasing the academic achievement of disadvantaged students do not necessarily weaken the performance of more advantaged groups.

Moreover, it is important to note that the group of high-achieving 15‑year‑olds is not composed exclusively – or even predominantly – of socio-economically advantaged students. On average across the OECD in 2025, high performers were distributed almost equally between students from high and middle socio-economic backgrounds, who accounted for 44% and 47% of high achievers respectively. This largely reflects the greater size of the middle socio-economic group, which includes students ranked between the 25th and 75th percentiles on the PISA index of economic, social and cultural status. This pattern is broadly observed across most OECD countries. The main exceptions are Mexico, Colombia and Chile, where socio-economically advantaged students account for nearly 70% of high performers.

Figure 4.2. Adolescents’ educational performance is strongly linked to socio-economic status
Copy link to Figure 4.2. Adolescents’ educational performance is strongly linked to socio-economic status

Promoting adolescent health and reducing socio‑economic inequalities

In parallel to trends in educational outcomes, similar concerns have emerged regarding child health, both in relation to possible deteriorations over time and to the persistent influence of social determinants on inequalities in child development and health outcomes.

Adolescents’ health outcomes have deteriorated

Health is a multidimensional concept, encompassing not only the absence of disease, but also individuals’ physical and mental capacity to develop and function well. As such, it is difficult to capture fully through a single synthetic indicator. In practice, health is often measured using self-reported assessments, where individuals are asked to rate their own health status. While these measures provide valuable information on how people perceive their health and well-being, they also have important limitations, as they reflect subjective perceptions rather than objective measures of physical health or medical conditions.

For 11‑to‑15‑year‑old adolescents, a number of self-reported health indicators are provided by the Health Behaviour in School-aged Children (HBSC) survey (2026[40]). Figure 4.3 displays the results of how 15‑year‑olds view their health globally, tracking both the rates of adolescents who experience “excellent” health as well as those at the other end of the spectrum who view their own health as “fair” or “poor”. Additionally, Figure 4.3, Panel C comprises an indicator of adolescent psychological well-being, which assesses the share of 15‑year‑olds who report having experienced at least two of a list of eight symptoms (e.g. headaches, difficulties getting to sleep, feeling low) more than once a week over the past six months. Overall, the data point to a deterioration in adolescents’ health outcomes over the past decade, particularly through increased prevalence of negative health outcomes and especially low mental well-being.

Figure 4.3, Panel A shows that 28% of 15‑year‑olds across the OECD view their own health as “excellent”, with wide variations across countries: Israel stands out as the country with the highest ratio (46%) whereas in Poland and Latvia, fewer than one in six adolescents report the same. Despite this variety, a common trend across countries is the decline in adolescents reporting “excellent” health from 2013-2014, by over 3 p.p. on average, down from 31%. While over three‑quarters of countries with available data saw a decrease, the largest reductions were felt in Greece and Spain with over ten p.p. Notable exceptions are Iceland, Finland and Czechia, which registered substantial increases of around 4 to 8 p.p. during the same time frame. The COVID‑19 pandemic may have played a role in this decrease in positive health for some individual countries, but no strong difference appears in the average decline between 2013-2014 and 2017-2018 compared to 2017-2018 and 2021-2022.

The share of adolescents who rate their own health as “fair” or “poor” is slightly smaller on average across the OECD (22%, Figure 4.3, Panel B). It ranges from less than 15% in Sweden, France and Finland to 35% or more in Hungary, Latvia and Poland. For this indicator, the deterioration of adolescent health is more marked. The proportion of 15‑year‑olds reporting “fair” or “poor” health increased by an average of 5 p.p. between 2013-2014 and 2021-2022. This upward trend was particularly pronounced in Hungary, Canada, Ireland and Poland where increases exceeded ten p.p. By contrast, Luxembourg, Finland, Iceland and Sweden recorded declines over the same period, of up to three p.p. Most of the deterioration occurred between 2017-2018 and 2021-2022. During this period, 84% of countries with available data experienced increases in the prevalence of poor self-reported health among 15‑year‑olds, compared with fewer than 70% of countries between 2013-2014 and 2017-2018. Moreover, the average increase exceeded 4 p.p. between 2017-2018 and 2021-2022, whereas the corresponding rise between 2013-2014 and 2017-2018 was only 0.8 p.p. This pattern may reflect the effects of the COVID‑19 pandemic, which could have had a particularly adverse impact on adolescents already experiencing poorer health.

Figure 4.3. Adolescents’ reports of low health are rising
Copy link to Figure 4.3. Adolescents’ reports of low health are rising

The prevalence of adolescents reporting multiple subjective health complaints also shows marked trends (Figure 4.3, Panel C). In 2021-2022, more than half (53%) of all 15‑year‑olds reported experiencing at least two symptoms more than once a week during the previous six months. The prevalence was high across all OECD countries, ranging from 41% in Spain and Slovenia to nearly 70% in Greece. Moreover, the share of adolescents experiencing multiple subjective health complaints increased substantially over time. On average across the OECD, rates rose by almost 3 p.p. between 2013-2014 and 2017-2018, followed by a further increase of more than 11 p.p. between 2017-2018 and 2021-2022. Over the entire period, all countries recorded significant increases, ranging from 6 p.p. in Slovenia and Spain to more than 20 p.p. in Greece, Portugal, Germany, Estonia, Austria and Hungary. Although the COVID‑19 pandemic likely intensified these trends, deteriorating adolescent mental health was already evident prior to the pandemic. Between 2013-2014 and 2017-2018, all countries except Spain, the Netherlands and Ireland experienced increases in the proportion of 15‑year‑olds reporting multiple subjective health complaints.

Social determinants of child health inequalities

Social determinants of health have long been recognised as major drivers of childhood health outcomes, accounting for a large share of childhood illness, developmental difficulties and premature mortality (Moore et al., 2015[41]; Palloni et al., 2009[11]; Singh, Siahpush and Kogan, 2010[42]; Pickett et al., 2022[13]; Spencer, 2018[43]; Coller and Kuo, 2015[44]). For children, the social determinants of health encompass the social, economic and environmental conditions in which they are born, grow up, live and learn, as well as the institutions and systems that promote health and respond to illness (Coller and Kuo, 2015[44]). These determinants influence child health through complex interactions between structural factors – such as household income, parental education and employment conditions – and more proximal factors, including living conditions, nutrition, stress exposure and health-related behaviours (Pearce et al., 2019[12]; Spencer, 2018[43]). Because these determinants largely operate outside the healthcare system itself, they are strongly shaped by broader social and economic policies.

Clear socio-economic disparities emerge in adolescents’ self-reported health. In every country, 11‑, 13‑ and 15‑year‑olds from the 25% most affluent households are more likely to report “excellent” health than their peers in the bottom quarter of family affluence. On average across the OECD, 37% of adolescents in the top quarter of the family affluence scale report excellent health, compared to 32% of students in the middle two quarters and 27% of 15‑year‑olds in the bottom quarter (Figure 4.4, Panel A). The smallest discrepancies by socio-economic status are found in countries where the prevalence of excellent self-reported health is high overall. In Slovenia, Switzerland, Israel and Greece, highly affluent adolescents are at most 15% or seven p.p. more likely to self-report excellent health than the least affluent students. By contrast, the ratio exceeds 80% and represents a difference of at least 15 p.p. in Canada and the United Kingdom (Wales and Scotland) – all jurisdictions where the overall rates of “excellent” self-reported health are among the lowest.

In terms of the prevalence of self-reports of “fair” or “poor” health, 15% of 11‑, 13‑ and 15‑year‑olds from highly affluent families state these, compared to 17% of medium affluent students and 22% of their peers from the least affluent families. This pattern is observed in all OECD countries except Slovenia, the Slovak Republic and Norway where differences by family affluence are not statistically significant (Figure 4.4, Panel B). The largest inequalities are again found in countries where the overall self-reported health is comparatively poor. For example, in Canada and Ireland, highly affluent students are 20 and 17 p.p. less likely to report “fair” or “poor” health than the least affluent adolescents, respectively. Both countries have comparatively high rates of students reporting these negative health outcomes overall (around 30%).

Multiple subjective health complaints are widespread across all levels of family affluence, and their socio-economic gradient is much less pronounced (Figure 4.4, Panel C). On average across the OECD, 45% of adolescents from both the top quarter and the middle two quarters of the family affluence scale report experiencing at least two out of a list of eight psychosomatic symptoms (e.g. stomach-ache, feeling nervous, difficulties getting to sleep) more than once a week over the past six months. The same was true for 49% of students in the bottom quarter of the family affluence scale. Countries with the largest inequalities are Canada, Estonia and Hungary where the least affluent adolescents are over 25% or 11 p.p. more likely to report multiple subjective health complaints than the most affluent students. Nearly 40% of OECD countries with available data show no statistically significant difference by family affluence score and in Switzerland, the opposite pattern is observed: the most affluent children are significantly more likely than their least affluent peers to report multiple subjective health complaints (47% and 43%, respectively).

Figure 4.4. Social inequalities in low and high self-reported health
Copy link to Figure 4.4. Social inequalities in low and high self-reported health

The evidence presented above shows that children’s educational and health outcomes vary substantially across countries, while trends since the early 2010s point to a deterioration in outcomes across a large number – if not all – OECD countries. At the same time, persistent socio-economic disparities remain evident in all countries. Children from low- and middle‑socio-economic backgrounds are generally more exposed to a range of social, economic and environmental risk factors, while also having more limited access to protective and enabling resources. This increases their likelihood of poor educational and health outcomes and reduces their chances of reaching high levels of achievement and well-being (OECD, 2021[45]).

The strong socio-economic disparities observed in adolescents’ educational and health outcomes suggest that poverty-reduction measures could also yield additional indirect benefits for children’s learning and well-being. In this regard, the social policies identified in Chapter 3 as effective in reducing child poverty and material deprivation are likely to generate positive spillover effects on children’s educational and health outcomes. Moreover, the persistence of strong social gradients at both the lower and upper ends of educational performance and health outcomes indicates that both dimensions need to be considered simultaneously, as the barriers associated with avoiding poor outcomes may differ from those associated with achieving high levels of performance and well-being. Strengthening support for children from disadvantaged socio-economic backgrounds is particularly important, given their substantially lower likelihood of attaining high academic and health outcomes.

Importantly, low educational outcomes are not confined to children from the most disadvantaged backgrounds. A substantial share of low-performing students comes from middle socio-economic groups, suggesting scope for strengthening support for a broader range of children and families. Such efforts may help prevent low- and middle‑income children from falling behind educationally while also increasing their chances of reaching higher levels of academic performance, which are increasingly regarded as essential in economies characterised by rapid technological change and growing demand for advanced skills. The following analysis explores these relationships further.

4.2. Appling the value‑for-money framework to educational and health outcomes: The methodology

Copy link to 4.2. Appling the value‑for-money framework to educational and health outcomes: The methodology

This chapter applies the value‑for-money framework developed in Chapter 3 (Section 3.3, “The methodology”) to a set of children’s educational and health outcomes. The approach retains the same two complementary dimensions of value for money: the effectiveness of social spending in improving child outcomes and its implications for the long-term evolution of social expenditure. The two dimensions are estimated separately and linked through the role of per capita social spending. The technical details of the framework, including the estimated equations and the variables included in each specification, are presented in Box 4.2.

The outcomes considered in this chapter capture complementary dimensions of educational achievement, labour-market engagement and adolescent health. They include the share of PISA high achievers, the share of PISA low achievers, the NEET rate, the share of 15‑year‑olds reporting excellent health, the share of 15‑year‑olds reporting fair or poor health, and the share of 15‑year‑olds reporting multiple subjective health complaints.

The main adaptation of the approach in this chapter relates to the age‑specific nature of the outcomes. Whereas Chapter 3 examines poverty and material deprivation among children as a whole, this chapter focusses primarily on outcomes measured at age 15, with the NEET rate covering young people aged 15‑29. The analysis therefore accounts for the time lag between exposure to a policy and the age at which its potential effects on outcomes are observed. In particular, policies that affect children earlier in life – such as ECEC, family cash benefits for younger children and education spending during earlier years – are introduced with lags designed to capture the period during which the cohort under consideration could have been exposed to them. For example, ECEC enrolment and spending are lagged to reflect participation when the cohort was of the relevant age, while education spending is similarly aligned with the cohort’s educational pathway.

The estimation strategy also differs somewhat across outcomes because of data availability. For educational outcomes, the analysis uses dynamic ordinary least squares (DOLS), as in Chapter 3, incorporating leads and lags of first-differenced variables to account for short-run dynamics and potential sources of endogeneity. For adolescent health outcomes, the available country-year observations are more limited and do not support the DOLS specification. These outcomes are therefore estimated using a two‑way fixed-effects ordinary least squares model. DOLS estimates are nevertheless used as robustness checks where feasible and produce broadly similar results.

The specifications additionally account for factors that are particularly relevant to educational and health outcomes. These include primary and secondary education expenditure per student and, for the education models, mean years of schooling among the population aged 25 and over. The models also retain the poverty-depth measure used in Chapter 3, while the determinants of social expenditure remain unchanged from the earlier framework.

The simulations used to assess value for money follow the same logic as those described in Chapter 3. First, the analysis estimates the change in each outcome and in social expenditure associated with an increase in social spending or GDP per capita, respectively, while keeping the policy mix unchanged. It then repeats the simulation while increasing one policy characteristic at a time. Value for money is measured by the difference between these two scenarios, capturing the additional improvement in the outcome and the additional spending pressure associated with changing the policy design, conditional on overall growth in social spending or GDP.

As in Chapter 3, two types of policy changes are considered. For most policy characteristics, the simulations assume a one‑standard-deviation increase in the policy variable. For policies measured as a share of total social expenditure, a second set of simulations assumes that one additional p.p. of total social expenditure is allocated to the policy area. The latter provides a common monetary basis for comparing alternative spending allocations.

Box 4.2. The value‑for-money methodology applied to educational and health outcomes

Copy link to Box 4.2. The value‑for-money methodology applied to educational and health outcomes

The empirical framework

The value‑for-money framework estimates two relationships separately. The first relates child outcomes to per capita social expenditure, policy design and other determinants of child outcomes:

${C O}_{i t} = α_{i} + μ_{t} + {T S}_{i t} β + P_{i, t - x}^{T} γ + {T S}_{i, t - x} P_{i, t - x}^{T} δ + {X O}_{i, t - x}^{T} η + \sum_{k = - 1}^{1} Δ {L L O}_{i, t - x + k}^{T} ζ_{k} + ε_{i t} (1)$

where ${C O}_{i t}$ denotes the child outcome in country $i$ and year $t$ , ${T S}_{i t}$ is per capita total social expenditure, $P_{i, t - x}$ is the vector of policy characteristics at the time $t - x$ that corresponds to the period during which they could have had an impact on the cohort under analysis, and ${X O}_{i, t - x}$ captures other core drivers of child outcomes. For example, when estimating educational and health outcomes of 15‑year‑olds, ECEC enrolment rates from 11‑15 years ago are used to reflect the cohort’s participation in these services. In addition to enrolment in and spending on ECEC services, family cash benefits by children’s age are lagged to represent the entitlements adolescents likely benefitted from throughout their childhood. Similarly, primary and secondary education expenditures from earlier periods are introduced to align with the educational pathways. Preventive health spending and spending on active labour market policies (ALMP) are lagged by three years to reflect the time needed for benefits to materialise. All other variables are assumed to impact child outcomes predominantly in the current period ( $t$ ), as in Chapter 3 (see the results tables in Annex 4.C for detailed information on variable lags where applicable).1 The interaction terms ( ${T S}_{i, t - x} P_{i, t - x}^{T})$ allow the effectiveness of policy design to vary with the size of the welfare state.

The second equation models the long-run determinants of per capita social expenditure, which are assumed to impact social expenditure predominantly in the current period ( $t$ ), as in Chapter 3:

${T S}_{i t} = ρ_{i} + σ_{t} + {G D P}_{i t} θ + P_{i, t}^{T} κ + {G D P}_{i, t} P_{i, t}^{T} λ + {X S}_{i t}^{T} φ + \sum_{k = - 1}^{1} Δ {L L S}_{i, t + k}^{T} ξ_{k} + υ_{i t} (2) .$

where ${G D P}_{i t}$ denotes GDP per capita net of social spending and ${X S}_{i t}$ includes the structural demographic and health-related drivers of social expenditure. The interaction terms allow the relationship between policy design and social spending to vary according to economic development.

The two equations are estimated independently. Educational outcomes are estimated using DOLS, incorporating one lead, the contemporaneous value and one lag of the first differences of the explanatory variables ( $Δ {L L O}_{i, t - x + k}^{T} a n d {L L S}_{i, t + k}^{T})$ .2 Because the available data for adolescent health outcomes are more limited, these models are estimated using two‑way fixed-effects OLS, without the dynamic terms. Where feasible, DOLS specifications are estimated as robustness checks.

Outcomes and explanatory variables

Child outcomes ( ${C O}_{i t}$ ) comprise:

share of PISA high achievers, attaining Level 5 or 6 in at least one of reading, mathematics and science;
share of PISA low achievers, attaining below Level 2 in at least one of the three core subjects;
NEET rate among 15‑29 year‑olds;
share of 15‑year‑olds reporting excellent health;
share of 15‑year‑olds reporting fair or poor health; and
share of 15‑year‑olds reporting multiple subjective health complaints.

The vector of child-outcome drivers ( ${X O}_{i, t - x}^{T}$ ) includes:

log per capita total social expenditure, in constant USD PPP;
log primary education expenditure per student; this captures general government spending on primary educational institutions at central, state and local level across private and public institutions.
log secondary education expenditure per student, which covers general government spending on secondary educational institutions at central, state and local level across private and public institutions.
mean poverty gap before taxes and transfers (as, for instance, deeper forms of poverty before taxes and benefits can diminish the effect of social policies on children’s educational and health outcomes, for example when cash transfers are insufficient in size to cover families’ lack of financial resources to provide a healthy home environment that stimulates learning); and
mean years of schooling among the population aged 25 and over, included in the education specifications only to capture structural aspects of children’s wider environment outside the school that may support their development.

The policy variables ( $P_{i, t - x}^{T}$ ) are those identified in the chapter as potentially affecting the effectiveness of social expenditure in improving educational and health outcomes. They largely correspond to the policy characteristics examined in Chapter 3. Age‑specific policies are introduced with lags reflecting the period during which the cohort under consideration could have been exposed to them (see the description of the empirical framework above).

The determinants of social expenditure ( ${T S}_{i t}$ ) are unchanged from Chapter 3 and comprise the log of GDP per capita net of social spending, log life expectancy at birth and the dependency ratio. See Chapter 3 for their definitions and interpretation.

Constructing the value‑for-money measures

The value‑for-money measures are calculated from the estimated coefficients of equations (1) and (2), using the same simulation approach as in Chapter 3.

For the outcome dimension, the analysis first simulates a half-standard-deviation increase in log per capita social expenditure while holding the policy mix fixed at its average. It then repeats the simulation while increasing one policy characteristic by one standard deviation. The difference between the two simulations represents the incremental change in the child outcome associated with the policy change, conditional on the increase in social expenditure.

For the spending-pressure dimension, the equivalent simulations increase GDP per capita by half a standard deviation, first with the policy mix unchanged and then together with a one‑standard-deviation increase in the policy characteristic. The difference between the two simulations represents the additional spending pressure associated with the policy change, conditional on economic growth.

For policy variables expressed as a share of total social expenditure, a second set of value‑for-money simulations replaces the one‑standard-deviation policy change with a one‑p.p. increase in the expenditure share. This provides a common monetary scale for comparing alternative allocations of additional social resources across policy areas.

The simulated half-standard-deviation increases in GDP per capita and social expenditure correspond to approximately 11 years of average growth based on trends since the early 2000s.

1. Although adolescents’ educational and health outcomes are shaped by the cumulative effects of their activities, environments and public policies throughout their lives, estimations do not include variables for the same policy across multiple time periods. Specifications with additional lags for drivers and policies were tested but discarded due to issues of multicollinearity and small sample sizes.

2. Lead and lags of first differenced variables are restricted to a total of three periods because they impose substantial demands on the data, while the available time series remain relatively short and sample sizes comparatively small.

Social policy leavers to improve children’s educational and health outcomes

Policy features ( $P_{i, t - x}$ ) capture key design characteristics of social expenditure systems that are assumed to influence the effectiveness of social spending in improving child outcomes as well as the spending pressures in response to GDP growth. The specific indicators included in the analysis are outlined below, and further details on variable definitions and data sources are provided in Annex 4.A and on the data imputations applied in Annex 4.B.

Social services can enhance children’s educational and health outcomes through both direct and indirect mechanisms. Directly, services such as ECEC can foster cognitive and socio‑emotional development during critical stages of childhood, while healthcare services for children contribute to improved physical and mental health outcomes. Indirectly, social services may strengthen children’s well-being outcomes by supporting parents, for example through measures that promote parental health, facilitate labour market participation, improve family income, and reduce the risk of poverty and its associated adverse effects on child development. The selected set of policy indicators is largely equivalent to the policy variables covered in Chapter 3.2

Acknowledging the strong influence of families’ economic resources as shown earlier in this chapter, we analyse role of social transfers on children’s educational and health outcomes. A core subset of measures of income transfers from Chapter 3 is used.3

4.3. The value for money of social expenditure: Results

Copy link to 4.3. The value for money of social expenditure: Results

The value for money of social expenditure in improving children’s educational outcomes

The results offer insights into the policy levers within social policy to support children’s educational achievement4 and thereby complement the analysis in Chapter 3. Looking first at the drivers of educational outcomes:

A core result is the negative relationship between economic inequality and educational performance. The mean poverty gap, which captures the distance of the equivalised income of the average poor person from the relative poverty line before taxes and transfers, is a strong predictor for lower PISA scores. Countries with lower relative market incomes of the poor tend to have lower shares of PISA high performers and higher shares of PISA low performers and NEETs (coefficients of ‑2, 2 and 2, respectively, for a 10‑p.p. change in the mean poverty gap). When adding country fixed effects and analysing within-country variation only, the effect becomes even stronger: A reduction in equivalised market income of the average poor person by ten p.p. of the relative poverty threshold is associated with a reduction in the share of high performers by 3 p.p., an increase in the rate of low performers by 3 p.p. and a rise in NEETs by 2 p.p. This implies that poverty reduction is a key channel through which social spending may affect children’s educational and employment outcomes.
Greater levels of public expenditure on education per student over students’ lifetimes is associated with higher PISA test performance when the estimation is based on pooled cross-country variation rather than within-country changes over time. Importantly, the timing of educational expenditure appears to play a significant role. Earlier investment is more strongly associated with reductions in low academic performance, whereas sustained and later investment may be more effective in fostering high achievement and preventing disengagement from education and employment. Specifically, a one‑percent increase in past per-student expenditure on primary education is associated with a 6‑p.p. reduction in the share of low performers in PISA, whereas a one‑percent increase in spending on secondary education does not significantly affect the proportion of low performers (Annex Table 4.C.7).
Moreover, there is some indication that past spending on both primary and – to a greater extent – secondary education may support higher performance, although the coefficients are not consistently statistically significant across specifications (Annex Table 4.C.2). In addition, countries with greater levels of per student expenditure on secondary education exhibit significantly lower NEETs rates four to six years later. This suggests that investment during adolescence may play an important role in sustaining educational engagement and facilitating successful transitions into further education or employment.
When the analysis focusses on within-country variations (Annex Table 4.C.1, Annex Table 4.C.6, and Annex Table 4.C.11), the associations between educational spending and adolescents’ educational outcomes are no longer statistically significant. This may reflect the fact changes in per-student education spending within countries over the past two decades have generally been much smaller than the differences observed across countries,5 and may therefore not have been large enough within the period covered to generate clear and statistically identifiable effects on educational outcomes.
There is some evidence that higher levels of educational attainment among the adult population are related to higher PISA test scores among adolescents. Across specifications, a within-country increase in the mean years of schooling by one year is associated with reductions in the rate of low performers between 2 and 5 p.p., although the effect is not significant in the baseline specification. In addition, one additional year of mean schooling among adults is associated with about one‑p.p. higher rates of PISA high performers across countries (Annex Table 4.C.2) but this relationship disappears once country fixed effects are added. No clear relationship is found between educational attainment among the adult population and rates of youths who are NEETs.
Total social expenditure also influences adolescents’ educational outcomes and youth NEET rates, but its effectiveness depends on the design of social policies, as illustrated in Figure 4.5. Several mechanisms may explain these associations. On the one hand, policies that reduce child poverty – either by enhancing parental labour market activity and earnings or directly through income transfers – can lessen economic inequality, enhance resources available in low-income families, and consequently yield improvements in children’s skills. On the other hand, investments in areas such as the provision of ECEC and child healthcare can support children’s healthy development and learning directly.

Figure 4.5 displays the value for money simulations of social policies with respect to children’s educational outcomes, following the methodology outlined in the previous section and based on the regression estimates in Annex 4.C that include country and time fixed effects. The green quadrants indicate that an increase in a given policy by one standard deviation is associated with better current or future educational outcomes as well as with greater spending moderation. Red quadrants identify both worse educational outcomes and greater spending pressures and the remaining two quadrants represent trade‑offs between improved child outcomes and fiscal moderation. Notably, and contrary to the interpretation in Chapter 3, simulated changes in the effectiveness of social spending on adolescents’ educational outcomes are directly linked to an amelioration or worsening of outcomes when social spending has no significant association with educational outcomes on its own. In other words, the red line and the 0‑line coincide.6

Figure 4.5. The value for money of social policies to support educational outcomes, for average levels of per capita social expenditure and GDP per capita
Copy link to Figure 4.5. The value for money of social policies to support educational outcomes, for average levels of per capita social expenditure and GDP per capita

The role of services in promoting educational outcomes

In-kind social services can support educational performance by mitigating the detrimental effects of poverty and enhancing families’ earnings capacity as well as by supporting children’s development directly. More specifically:

The combination of take‑up of and spending on ECEC services is a powerful enhancer of later educational achievement
The effects of ECEC coverage for children aged 3‑5 and spending on ECEC are highly interdependent. Allocating greater shares of total social spending to ECEC services is linked to an increase in the share of high achievers by 0.4 p.p. by itself. However, high shares of spending coupled with a greater ECEC coverage at ages 3‑5 are associated with even starker improvements in adolescents’ PISA performance. A joint one‑standard-deviation increase in the share of social spending dedicated to ECEC (i.e. by 2.3 p.p. of social spending) and enrolment at ages 3‑5 (+ 18 p.p.) yield a value for money improvement of 1.4 p.p. in the share of high performers and of ‑4.2 p.p. in the rates of low performers in PISA.7 Such a simultaneous expansion in the coverage at ages 3‑5 and the weight of ECEC in the spending mix is not linked to any changes in spending pressure.
Results for younger children (ages 0‑2) provide further suggestive evidence of the importance of accompanying a relative growth in spending on ECEC with broad coverage in order to boost later educational outcomes. By itself, ECEC enrolment at age 0‑2 is associated with no change in rates of high performers and with +1.1 p.p. higher shares of low performers, and greater impacts on low performance are observed for higher levels of per capita social expenditure. This positive association may reflect both the unequal distribution of access and the heterogeneous quality of services for children under age three. Children from low-income families are substantially less likely to participate in ECEC at these ages and, when they do, may on average access lower-quality services than more advantaged children (see Chapter 2). At the same time, socio‑economically disadvantaged children are more likely to become low performers later in adolescence. By contrast, participation in pre‑primary education from age three tends to involve much broader coverage of disadvantaged children and is more often delivered through structured early education systems governed by curricula and more homogeneous quality standards.
Nevertheless, greater enrolment for children under age three is also associated with a 2.4%-reduction in the projected social spending increase that accompanies economic growth. This likely reflects the important role that childcare provision for very young children plays in supporting parental – particularly maternal – employment, thereby helping to strengthen household income and reduce longer-term demands on social protection systems (see Chapter 3).
Overall, the results point to the provision of high-quality ECEC services with broad coverage as a powerful tool to enhance children’s educational performance in a lasting way. This comes in addition to their potential to reduce child poverty by facilitating parental employment (see Chapter 3).
Health spending is linked to reductions in NEETs rates and spending pressure
Dedicating higher shares of social spending to health expenditure appears to be one of the most impactful policy tools to reduce NEETs rates. This effect is driven by non-preventive health spending, which suggests that greater emphasis on curative healthcare effectively removes barriers for youth to participate in education and work.
One possible explanation is that a significant minority of NEETs experience physical or mental health problems, while the NEET population spans a broad age range that includes many young adults who are more likely than children or younger adolescents to benefit directly from curative healthcare. This may also explain why a higher share of social expenditure devoted to healthcare is not associated with better PISA test outcomes among adolescents. In fact, greater spending on non-preventive healthcare is associated with slightly higher rates of low achievement in PISA, possibly reflecting the fact that much of this expenditure is directed towards addressing the healthcare needs of older age groups rather than supporting children’s development and learning.
Moreover, health spending is found to help moderate the growth of social expenditure as economies develop, as already noted in Chapter 3. This likely reflects its role in improving population health, thereby reducing future demand for costly care while supporting higher labour market participation and productivity.
Other in-kind spending can help reduce NEETs rates
Allocating a greater share of social spending on in-kind services in incapacity-related, housing, and other social spending areas can contribute to lowering NEETs rates (‑1.1 p.p. in Figure 4.5). The effect diminishes for more generous social protection systems, possibly indicating that these supports are highly complementary to other social policies in promoting participation in education, training or employment among youth. On the other hand, a greater use of such in-kind services in the policy mix shows no significant link with adolescents’ performance in PISA.
However, as discussed in Chapter 3, in-kind expenditure in these policy areas is associated with higher social spending pressures (+5.8% of additional spending), which may suggest that these services often complement rather than replace other forms of support.
ALMPs have no significant link with educational outcomes
Within-country variations in the share of social spending dedicated to ALMP are not significantly related to adolescents’ performance in the PISA test or NEETs rates,8 although a greater policy orientation towards ALMPs is lined to spending moderation, especially in wealthier economics (see Chapter 3).

The role of income transfers in promoting educational outcomes

Direct income transfers can enhance families’ available resources and promote children’s educational performance in addition to offering protection from poverty (see Chapter 3 on the latter). The evidence suggests that early spending on families is protective against low adolescent academic performance whereas later income support, especially on middle‑income households, supports high academic performance.

Higher income support systems are associated with high educational performance
A higher level of average income support is associated with greater rates of PISA high performers in high spending contexts, with no change in the rates of low achievers. Additional analysis suggests that this is driven by average levels of benefits rather than taxes (see Annex Table 4.C.1): for a given level of average taxes, an increase in average benefits by one standard deviation is linked to an increase in high performers by 1.3 p.p. whereas reducing the average level of taxes for a given level of benefits shows no significant effect. At the same time, greater use of redistribution through benefits is associated with additional spending pressure (see Annex Table 4.C.19).
For PISA low performers and NEETs rates, neither within-country changes to the average size of income support nor average levels of benefits and taxes separately appear to matter.
Countries with more progressive income redistribution systems tend to show higher shares of students with high PISA scores but further shifts towards narrowly targeted transfers may weaken overall performance.
Cross-country estimates (i.e. without country fixed effects) suggest that larger shares of benefits allocated to the bottom 30% by one standard deviation (around 10 p.p.) are associated with a 4.5‑p.p. increase in the share of high performers and a reduction in low performers by 3.6 p.p. Similarly, a rise in taxes paid by the bottom 30% by one standard deviation (3.5 p.p.) are associated with increased rates of low performers by 1.8 p.p., an effect that increases further for above‑average levels of social spending (interaction coefficient of +2), and fewer high achievers in contexts where total social spending per capita is high (interaction coefficient of ‑1.3) (see Annex Table 4.C.2 and Annex Table 4.C.7).
However, when restricting the analysis to within-country variation, a different pattern emerges: an increase in the share of income benefits received by the bottom 30% by one standard deviation is linked to a 3.1‑p.p. reduction in the share of high performers, and no significant association is found for low performers. In other words, although countries with income support systems that are, on average, more concentrated on lower-income households exhibit stronger educational outcomes, further increases in the degree of targeting within countries over time appear to be associated with a lower share of high-performing students. On the other hand, the targeting of taxes is no longer related to changes in rates of high or low performers. Hence, financial support to middle‑income families appears to be particularly salient to promote educational achievement at the top. This may be achieved at the cost of increased social spending pressure since lower benefits targeting is associated with greater social expenditures in extensive welfare systems, as already discussed in the analysis in Chapter 3.
Greater shares of social expenditure on family cash benefits are associated with better academic performance and their timing across childhood matters
The share of social expenditure that is distributed to families in cash is weakly linked to reduced rates of PISA low performers in the same period. In the value‑for-money simulations, increasing the share by one standard deviation (an additional 4.5 p.p. of total social expenditure allocated to family cash benefits) is associated with a 0.7‑p.p. reduction in the share of low-performing students and no significant relationship with the rate of high performers. This effect, however, becomes stronger as per capita social spending increases, suggesting that cash benefits are more effective in improving educational outcomes when embedded within more comprehensive welfare systems. Additional results by socio‑economic status (Box 4.3 and Annex 4.C) further indicate that family cash spending during early and middle childhood substantially reduces the share of low-performing students from low-SES backgrounds in high-spending contexts (the estimated interaction coefficients exceed 5 p.p.).
Further, spending a greater share of total social expenditure on families in cash in high social spending contexts is related to lower NEETs rates five years later (‑0.7 in the value for money simulation, an effect that also increases in per capita social spending levels).
This contrasts with the above results on general income redistribution mechanisms through taxes and benefits, which appear effective at promoting high educational performance but not at mitigating low achievement. One potential explanation might be that labelling cash benefits as intended for children can reinforce spending on children, which becomes especially impactful in households where economic resources are scarce (see Chapter 3).
When analysing family cash benefits by age groups, past spending on family benefits for early and middle childhood is strongly linked to decreases in the rates of adolescent low performers in high spending contexts, whereas contemporaneous cash benefits on adolescents have no impact. For shares of PISA high performers, past spending during middle childhood is related to increases in high spending contexts. While contemporaneous family cash benefits for adolescents are found insignificant in the baseline value for money calculations (see Figure 4.5), alternative regression specifications employed as robustness checks signal a strong positive relationship between a greater policy orientation towards cash benefits for adolescents and increased shares of PISA high performers.9 Overall, orienting a greater share of social spending towards cash benefits provided during early and middle childhood appears to prevent poor educational outcomes among 15‑year‑olds, likely by preventing children from falling behind and containing educational gaps between children. On the other hand, continued support provided later in childhood and adolescence is associated with higher levels of achievement, possibly because the provision of additional resources helps provide a home learning environment that is conducive to high educational performance.
In line with the idea that cash benefits for young children are particularly protective against negative later educational outcomes, family cash benefits paid during adolescence10 have no effect on NEETs rates among 15‑29 year‑olds eight years later, while past spending on family benefits across all ages are associated with reductions in NEETs rates.
Overall, the evidence suggests that cash spending on families is generally linked to additional spending pressure (see the value for money simulations in Chapter 3 for more detail). At the same time, age‑specific cash spending on families is not significantly associated with changes in spending pressures. One potential explanation could be that age‑specific benefits are largely substitutive, and countries may direct additional resources towards particular age groups without changing the total allocation of family cash benefits.
Last but not least, additional results discussed in the Box 4.3 and reported in Annex 4.C suggest that promoting social mobility among low- and middle‑SES children depends on placing greater emphasis on specific policy areas, particularly access to ECEC services and income support. At the same time, concentrating a larger share of income support benefits on households in the bottom 30% of the income distribution does not appear to improve their chances of high performance. This may reflect the fact that stronger targeting is often accompanied by lower benefit levels for lower-income and middle‑class households overall.

Box 4.3. Promoting high academic performance among low- and middle‑SES students

Copy link to Box 4.3. Promoting high academic performance among low- and middle‑SES students

Additional analyses of the policy features associated with high academic performance among adolescents from low- and middle‑socio-economic status (SES) backgrounds point to several policy dimensions that appear particularly important for supporting upward educational mobility. The findings show that higher shares of social spending on early childhood education and care (ECEC) coupled with higher enrolment rates among children aged 3 to 5, are associated with greater chances that middle‑SES adolescents achieve high performance in PISA at age 15. Positive associations of both relative spending and enrolment are also observed for low-SES students, although they are not statistically significant, possibly because of the comparatively limited variation over time in the share of disadvantaged students reaching top performance levels. Increased preschool participation is not significantly associated with higher performance among high-SES students, potentially because children from more advantaged families may already benefit from home learning environments and parental investments that partly substitute to formal ECEC participation.

Income support also appears to play an important role in supporting higher achievement among comparatively disadvantaged students. For high levels of social expenditure, larger average income transfers are associated with higher probabilities that low- and middle‑SES adolescents – but not high-SES adolescents – become high performers, likely reflecting the higher income support received especially by low- and middle‑income families in more generous income transfer systems. However, stronger targeting of benefits towards the bottom 30% of the income distribution is not associated with significant increases in the share of low- and middle‑SES students reaching high-performance levels. For middle‑SES students, the association is even negative, which may reflect that tighter targeting has sometimes been accompanied by lower levels of support for middle‑SES families (Mcknight, 2015[46]).

The timing of family cash support also appears to matter. Increased spending on family cash benefits during early childhood is particularly strongly associated with higher chances of low-SES students reaching high-performance levels. Spending later in childhood shows clearer associations for middle‑SES students, while the estimated effects for low-SES students remain positive but statistically imprecise, possibly because of the relatively small share of disadvantaged adolescents attaining top PISA scores. At the same time, indicators of higher spending on family cash benefits at every stage of childhood are also associated with greater chances of high-SES students performing high. While the mechanisms cannot be directly identified, this may reflect the greater capacity of advantaged households to translate additional income into investments that enhance educational performance, such as access to higher-quality schools or enriched home learning environments, or paying for private tutoring.

The value for money of social expenditure on children’s health outcomes

The tables in Annex 4.C display the regression estimates from the analysis on health outcomes. The results are again shown with and without country fixed effects in addition to time fixed effects to analyse within-country associations as well as differences in levels across countries. The results offer insights into the policy drivers of children’s health outcomes, and they complement the previous analyses on child poverty and educational outcomes. Looking first at the drivers of adolescent health outcomes:

Market inequalities at the bottom of the income distribution have no clear link to children’s health outcomes. There is some evidence across specifications that within-country increases in the mean poverty gap before taxes and transfers seem detrimental to the share of adolescents experiencing their own health as “excellent” and to rates of 15‑year‑olds reporting multiple subjective health complaints (with coefficients around ‑0.2 and +0.3 p.p., respectively) but no consistent patterns emerge.
On average, social expenditure has large negative effect on the share of 15‑year‑olds reporting multiple subjective health complaints: an increase in the log of per capita social expenditure by half a standard deviation – representing the average historical social expenditure growth over one decade approximately – lowers the rate of adolescents reporting multiple subjective health complaints by nearly 7 p.p. However, total social expenditure is not related to changes in the share of adolescents reporting their own health as “excellent”, nor to those rating it as “fair” or “poor”. This implies that the overall effectiveness of social spending depends critically on its policy features to improve adolescent self-reported health whereas for the share of 15‑year‑olds who report multiple subjective health complaints, policy design plays a smaller role in the strong overall relationship. The impact of social spending on children’s health outcomes may be both direct – for example, by providing preventive and curative healthcare to children – and operate indirectly by reducing children’s exposure to poverty and enhancing family resources that support a healthy lifestyle and limit exposure to harmful environments.

Figure 4.6 displays the value for money simulations of social policies with respect to children’s health outcomes based on the fixed effects regression estimates in Annex 4.C. As before, the green quadrants indicate that an increase in a given policy by one standard deviation is associated with better health outcomes over time, as well as with greater spending moderation. Red quadrants identify both worse outcomes and greater spending pressures and the remaining two quadrants represent trade‑offs between improved child outcomes and fiscal moderation.11

Figure 4.6. The value for money of social policies to foster children’s health outcomes, for average levels of per capita social expenditure and GDP per capita
Copy link to Figure 4.6. The value for money of social policies to foster children’s health outcomes, for average levels of per capita social expenditure and GDP per capita

The role of services in promoting health outcomes

A range of in-kind social services can support adolescent health directly by supporting their development throughout childhood, possibly mitigating some of the adverse effects of child poverty, and indirectly by supporting their parents, including to enhance their earnings capacity.

Higher spending on preventive healthcare is associated with an increased share of adolescents reporting excellent health.
Preventive health expenditure has the largest impact on improvements in positive health outcomes among all policy measures. Increasing its share in total social expenditure by one standard deviation (0.9 p.p.) is expected to yield an additional 6‑p.p. increase in 15‑year‑olds reporting their health as excellent in the value for money simulation (Panel A, Figure 4.6). This represents an additional 20% of adolescents reporting excellent health for the average OECD country. Greater effects are predicted for higher-than average levels of per capita social expenditure. On the other hand, preventive health spending does not result in significant changes in adolescents’ negative health outcomes, namely the share of 15‑year‑olds reporting their own health as “fair” or “poor”, or the proportion reporting multiple subjective health complaints. This could indicate that curative health measures are more relevant for children with poor health, although the analysis shows no significant association between the share of social expenditure allocated to non-preventive health expenditure and any of the three adolescent health measures.
These results may underestimate the effect of healthcare on children’s health due to substantial measurement issues. Only a small share of health spending goes to children, with much of it being spent on the elderly, especially for non-preventive healthcare (OECD, 2025[47]). Disproportionate increases in health spending on older population groups could therefore disguise the relationship between healthcare and adolescent health outcomes, reducing its magnitude or leading to insignificant results.
As already suggested in Chapter 3, prioritising health spending in the social policy mix tends to substantially moderate social spending pressure (11% reduction in additional spending as economies develop). This probably reflects the role of generous public health systems in improving population health, thereby reducing future demand for costly care while supporting higher labour market participation and productivity among working-age adults. While the largest gains in spending moderation stem from non-preventive spending, increases in preventive health spending – holding non-preventive spending constant – are associated with particularly beneficial effects on the evolution of social spending in advanced economies (‑2.5% lower additional spending for the average country as depicted in Figure 4.6, which rises to 7.4% for countries with high GDP as shown in Annex Figure 4.A.3).
Increases in ECEC spending come with improvements across all of children’s later health outcomes
While across countries, changes in the share of social expenditure dedicated to ECEC are not related to adolescents’ later health outcomes, greater spending on ECEC services comes with improvements in all three health outcomes in high spending environments when focussing on within-country variation: For the average OECD country, a one‑standard-deviation increase in ECEC spending improves the share of 15‑year‑olds reporting their health as excellent by an additional 1.3 p.p., reduces the share of adolescents declaring their health as “fair” or “poor” by 1 p.p. and lowers the rate of 15‑year‑olds experiencing multiple subjective health complaints by 1.5 p.p. in the value‑for-money simulation (Figure 4.6).
In addition to spending on ECEC, greater enrolment in ECEC (ages 3‑5) further enhances the protective effects against adolescents’ poor health outcomes: For the average OECD country, the value for money simulation reduces the share of 15‑year‑olds reporting their own health as “fair” or “poor” both by prioritising ECEC spending in the policy mix (‑1 p.p. for a one‑standard-deviation increase) and separately by increasing enrolment rates among 3‑5 year‑olds by a standard deviation (‑0.7 p.p.). Augmenting spending on ECEC services and enrolment levels at ages 3‑5 simultaneously by one standard deviation is over three times more impactful at lowering the share of 15‑year‑olds who report multiple subjective health complaints than spending on ECEC alone, and raising coverage by itself has no significant impact at all.
Enrolment at lower ages (0‑2) also provides suggestive evidence of protective effects of ECEC attendance against later negative health outcomes. A one‑standard-deviation increase is associated with lower rates of multiple subjective health complaints (‑1.9 p.p. for the average OECD country), with greater effects observed when per capita social expenditure is high. Multiple subjective health complaints are experienced by nearly one in two 15‑year‑olds on average across the OECD, irrespective of family affluence. In contrast, rates of “fair” or “poor” self-reported health are less frequent and more concentrated among children from less affluent families (22% vs. 14% among highly affluent children on average, see Figure 4.4). The significant protective link between ECEC enrolment under 3 and lower adolescent rates of multiple subjective health complaints but not poor self-rated health is therefore consistent with the child population that attends these services and does not mean that ECEC services for younger ages are less effective at mitigating later negative health outcomes than at ages 3‑5. Instead, the results could be interpreted as suggestive evidence of benefits of ECEC attendance also for more advantaged children.
As discussed before, a greater prioritisation of ECEC services in the policy mix is not related to increases in social spending pressure. On the contrary, expanding coverage of education and care services, especially among children under age 3, can help ease longer-term pressures, as it supports parental employment and fosters early human capital development, thereby reducing future reliance on income support and other costly social interventions as economies grow.
ALMP are associated with better self-reported health in adolescents
Increasing the share of social spending dedicated to active labour market programmes by one standard deviation increases spending efficacy, yielding higher shares of children describing their own health as “excellent” (+3.1 p.p.), lower rates of children reporting their health as “fair” or “poor” (‑2.3 p.p.) and lower rates of 15‑year‑olds reporting multiple subjective health complaints (‑1.1 p.p.).
Other in-kind spending is associated with better health outcomes in higher-spending welfare states
Devoting a greater share of social expenditure to in-kind spending on incapacity-related, housing and other social policy areas is linked to a 4.2‑p.p. reduction in rates of adolescents who report multiple subjective health complaints for the average OECD country. At the same time, the simulation predicts a worsening of adolescent self-reported health at average levels of spending (‑1.5 p.p. for “excellent” health, and +1.2 for “poor” or “fair” health). However, as per capita social expenditure levels grow, such in-kind spending becomes more valuable. It is linked to improved health outcomes across all three measures in high spending contexts (see Annex Figure 4.A.4). The strong dependence on spending levels suggests important complementarities of these types of supports with other policies that promote adolescent health. However, these benefits are associated with substantial spending pressures as discussed before.

The role of income transfers in promoting health outcomes

Providing additional financial resources to families can improve children’s health in numerous ways. For example, it may improve their access to nutritious food, sports and leisure activities, the quality of their housing, and possibly family relationships by reducing stress, which typically comes with financial strain. Promising policy levers are:

Targeting income transfers towards lower-income households can strengthen adolescent health
Although changes in the average size of countries’ income support systems are not significantly associated with any of the three adolescent health outcomes, targeting transfers towards lower-income households appears to represent an effective policy lever for promoting adolescent health. Reducing the share of taxes paid by the bottom 30% is linked to improvements in self-reported health by 0.8 p.p. for excellent health, and by ‑1.6 p.p. for the share of adolescents rating their own health as “fair” or “poor”. Similarly, raising the share of benefits given to the bottom 30% by one standard deviation enhances social spending effectiveness at reducing the share of children reporting multiple subjective health complaints in the simulations (‑3 p.p. for the average OECD country), with greater impact at higher spending levels.
Greater targeting of benefits is linked to spending moderation in wealthier countries. As detailed in Chapter 3, a greater targeting of benefits is typically accompanied by lower overall volumes of benefits.
Cash spending on families appears beneficial for adolescents’ health
A one‑standard-deviation increase in social expenditure allocated to family cash benefits is associated with a decline in the share of adolescents reporting multiple subjective health complaints by ‑1.6 p.p. in the same time period. In addition, family cash benefits are linked to a rise in 15‑year‑olds who declare their own health as “excellent” by +1 p.p., with higher impacts at higher levels of spending. Effects on the share of children experiencing their own health as “fair” or “poor” are not significant, although coefficients are negative.
When assessing the impact of cash benefits for families throughout children’s lives, past transfers during early childhood appear effective at mitigating the prevalence of low self-reported health in adolescence. Cash benefits during early childhood further increase adolescent self-reports of excellent health along with spending during middle childhood, which indicates the importance of sustained investment during childhood to promote positive health outcomes. On the other hand, family benefits paid out during adolescence are negatively linked to 15‑year‑olds describing their own health as excellent. This highlights the critical importance of early and middle childhood in shaping healthy development and may suggest that increased spending on family benefits during adolescence is insufficient to compensate for limited investment earlier in life. The data reveal no clear age‑spending patterns for multiple subjective health complaints.
As described in the discussion on educational outcomes and in Chapter 3, the evidence points to a positive association between an expansion in the use of cash spending on families and greater social spending pressures.

Comparing the value for money of equivalent investments (in monetary terms) across social policy domains

The analysis presented thus far has focussed on comparing estimated effects of policy changes on children’s educational and health outcomes in terms of policy expansions by one standard deviation of the observed sample variation. This approach offers the advantage of rendering policies expressed in different units broadly comparable. To analyse the results in terms of equivalent monetary investments, we can rescale the estimated effects on child outcomes and spending pressures for a subset of policies that refer to social expenditure shares. This may align more closely with the trade‑offs policymakers face when deciding on long-term spending allocations. Figure 4.7 presents the expected benefits of an equivalent investment (one additional p.p. of total social expenditure) allocated to a given policy area in the average OECD country.

Figure 4.7. Allocations across policy areas shape spending effectiveness at improving children’s educational and health outcomes and social expenditure levels
Copy link to Figure 4.7. Allocations across policy areas shape spending effectiveness at improving children’s educational and health outcomes and social expenditure levels

Estimated effects of a one‑p.p. increase in the share of social expenditure allocated to different policy domains on children’s educational and health outcomes and total social spending levels

Comparing adolescents’ outcomes for equivalent expenditure reallocations across spending domains identifies two spending areas with large effect sizes. First, investing the additional p.p. of social expenditure to expanding ECEC coverage is associated with improved educational performance (1.7 p.p. fewer low performers and 0.4 p.p. more high performers, which represent 4% and 3% improvements relative to the OECD averages in 2025, respectively) and ‑1.6 p.p. fewer 15‑year‑olds reporting multiple subjective health complaints.

Second, preventive health stands out for its strong relationship with excellent self-reported health: increasing the share of total social expenditure allocated to preventive health by one p.p. is associated with a 6.6‑p.p. increase in the share of adolescents reporting excellent health (equivalent to an increase by 24% from 2021‑2022 levels).

Figure 4.7 also highlights differences in the expected impact of increases in spending on family cash benefits depending on the stage of childhood targeted. Investment directed towards children in middle childhood (ages 6‑11) is associated with the largest improvements in educational outcomes: an additional one p.p. of total social expenditure to programmes targeting this age group is associated with 1.5 p.p. fewer low-performing students and 0.8 p.p. more high-performing students in PISA assessments. More generally, the estimated effects of additional spending on children aged 0 to 5 are smaller in magnitude than those associated with increases in spending targeted at older children. A higher baseline level of expenditure may help explain why additional spending appears to yield smaller marginal effects at younger ages compared with later states of childhood where average expenditure is lower: on average, OECD countries devote approximately twice as much to family cash benefits during early childhood as they do during middle childhood and adolescence, largely reflecting the prominence of childbirth-related leave benefits. In addition, variation in family cash expenditure during early childhood (as measured by its standard deviation) is approximately twice as large as during subsequent developmental periods, potentially reflecting differences in the composition and objectives of spending programmes across countries, which may influence the overall results. Nevertheless, a greater share of cash benefits targeted towards young children remains associated with improvements across a broad range of educational and health outcomes.

For the average OECD country, dedicating one additional p.p. of social expenditure to ALMP is associated with improved health among adolescents, with 1.3 p.p. of 15‑year‑olds reporting “excellent” health, one‑p.p. fewer adolescents indicating their own health as “fair” or “poor”, and 0.5‑p.p. fewer 15‑year‑olds reporting multiple subjective health complaints.

Rescaling reduces the magnitude of other in-kind benefits by a factor of 3.5. Shifting an additional p.p. of expenditure to services in the areas of invalidity related, housing and other social policies is thus linked to a reduction in adolescents with multiple subjective health complaints by 1.2 p.p. As before, increases in other spending in kind are found to be less effective at promoting self-reported health than the average composition of social expenditure.

4.4. Implications for social policy

Copy link to 4.4. Implications for social policy

The examination of recent trends shows that a large share of 15‑year‑old adolescents across the OECD have experienced declines in both academic outcomes and self-reported health. Combined with the persistence of strong social stratification in children’s educational and health outcomes, these trends contribute to unequal starting conditions early in life (Clarke and Thévenon, 2022[48]; OECD, 2021[45]). This is particularly reflected in our analysis by the strong positive within-country association between increases in the mean poverty gap – a measure of poverty depth – and both the share of young people who are NEET and the proportion of low performers in PISA, alongside a negative association with the share of high-performing students. Taken together, these findings suggest considerable scope for social, education and health policies to address the underlying drivers of these inequalities and potentially help reverse deteriorating trends in child educational and health outcomes.

Designing policies capable of tackling entrenched socio-economic inequalities requires addressing the multiple and interconnected factors that shape gaps in children’s educational and health outcomes. On the one hand, children from low-income families are more likely to experience poor home learning environments, including more limited access to educational resources, fewer opportunities to study in quiet and adequate conditions, and higher levels of financial strain and family stress that can reduce parents’ capacity to support learning. Parents in disadvantaged households may also be less familiar with learning-support practices and have less time available to engage with their children because of demanding, unstable or non-standard working conditions. On the other hand, disadvantaged children are also more likely to grow up in neighbourhoods characterised by lower-quality schools and public services, fewer educational and recreational opportunities, and greater exposure to social and environmental risks, all of which can negatively affect educational achievement, health and broader well-being outcomes.

Encouragingly, a wide range of policy levers exists to address these inequalities in early life and help offset the disadvantages associated with growing up in unequal environments. These include early childhood education and care, education system policies, income support, health services and broader social policies. However, strengthening their impact requires overcoming a number of interrelated challenges and addressing them in a co‑ordinated and integrated manner, so that policies reinforce one another and can collectively produce meaningful and lasting improvements in children’s developmental trajectories.

Leveraging ECEC to improve children’s education and health

Scaling up ECEC services requires sustained public investment to expand availability, affordability, and quality, which places upward pressure on social spending given the labour-intensive nature of the sector and its limited scope for productivity gains. However, ensuring that these investments translate into broader and more equitable coverage is crucial to ease longer-term spending pressures and to allow children and families to fully benefit from these services.

The potential of ECEC to start preparing children for life is widely emphasised in policy discussions. However, the evidence on its impact on developmental, educational and health outcomes is mixed. The literature suggests that the long-term benefits of ECEC depend on specific conditions, with the age at enrolment and the quality of services emerging as key factors (OECD, 2025[49]; 2011[50]).

This nuanced picture is also reflected in the findings presented above. The associations between childcare coverage and adolescents’ educational and health outcomes differ depending on the age group considered. Higher enrolment in pre‑primary education and care for children aged 3‑5 and dedicating greater shares of social expenditure to ECEC are associated with better academic performance at age 15 – reflected in lower shares of low performers and higher shares of high performers in PISA – as well as with greater protection against poor health outcomes in adolescence. By contrast, weaker and more mixed results are found between enrolment of children under age 3 in formal childcare and later PISA test scores or health outcomes.

The absence of clear beneficial effects for younger children suggests scope to strengthen the role of ECEC in laying solid foundations for physical and cognitive development. It also reflects that the provision of ECEC services for children under age 3 has not yet delivered the expected impact at scale, due to several overlapping factors. First, as noted in Chapter 2, participation in formal childcare remains lower among children from disadvantaged backgrounds, despite evidence that they stand to benefit the most. In addition, the quality of provision – crucial for positive outcomes – is more heterogeneous for services offered to younger children than for pre‑primary education, meaning that not all forms of formal care services deliver comparable benefits for younger children. More generally, the evidence on the impact of early enrolment in ECEC services for children under age 3 is quite mixed: earlier enrolment does not systematically translate into better outcomes, and effects vary depending on the outcomes considered (Box 4.4).

By contrast, the positive association between enrolment of 3‑5 year‑olds and performance at age 15 aligns with evidence from PISA microdata, which shows that children who attend ECEC for at least two years tend to achieve higher scores, particularly in mathematics (OECD, 2025[49]). This likely reflects both the broader coverage of these services – including among disadvantaged children, with in 2023 on average 89% of 3‑ to 5‑year‑old children being covered – and the more structured nature of pre‑primary education (OECD, 2026[51]).

Moreover, the findings suggest that higher ECEC coverage is more likely to translate into improvements in children’s later educational and health outcomes when accompanied by increased levels of spending. For education, the benefits associated with pre‑primary participation appear stronger in contexts with higher investment, highlighting the importance of combining expanded access with sufficient resources to ensure service quality. Similarly, for health outcomes, higher shares of social expenditure devoted to ECEC are associated with better outcomes across all indicators, even at comparable levels of coverage, suggesting that investment in high-quality ECEC can also play an important role in supporting healthy child development.

While further research is needed to better understand the underlying mechanisms, these findings are consistent with evidence from individual-level studies (e.g. Morrissey (2019[52])), which identify multiple pathways through which high-quality ECEC can affect child health. These include direct positive effects through access to health screening, care, nutrition and other health-promoting activities; potential negative effects through increased exposure to infections; and indirect effects via higher household resources linked to parental employment, as well as through improved educational trajectories that support healthier behaviours and outcomes over time.

Overall, the literature confirms that high-quality ECEC can improve cognitive, social and emotional skills in both the short and long term, with stronger effects for disadvantaged children. However, these benefits are not automatic: they depend on programme design, quality, and context (Box 4.4). Four main policy implications emerge.

First, investing in the quality of ECEC services is critical. Evidence from intensive, high-quality targeted programmes – such as the Perry Preschool, Chicago Child-Parent Centers and the Abecedarian project – shows substantial long-term gains in education, health and earnings, although these interventions were highly resource‑intensive and targeted, limiting their generalisability. Evidence from larger-scale (or universal) programmes show more mixed results, highlighting the importance of quality in achieving positive outcomes at scale.
Second, intensity and duration of ECEC attendance matter. Given the cumulative nature of learning, more time spent in ECEC can support skill development, particularly for disadvantaged children, provided programmes are of high quality. While intensive participation (e.g. 30 hours or more per week) is associated with better cognitive outcomes, especially for vulnerable children, evidence on socio‑emotional outcomes is less clear, and very early, high-intensity enrolment may carry some risks.
Third, increasing participation among disadvantaged children is essential. Since benefits are strongest for children from lower socio‑economic backgrounds, policies should prioritise expanding access in deprived areas and addressing financial, social and cultural barriers that may limit take‑up, even where services are available. Affordability of ECEC services is key. For instance, recent evidence from the United States suggests that relatively modest spending on free childcare for non-college‑educated single mothers can significantly reduce investment gaps in children’s development, partly because these households reinvest savings from childcare support (Caucutt et al., 2026[53]). By contrast, achieving similar effects for two‑parent households would require substantially higher investment.
Fourth, sustaining the impact of high-quality ECEC requires smooth transition across stages of the system and strong co‑ordination with other services. For ECEC to have lasting effects on child development, it is essential to ensure smooth transitions within ECEC and from ECEC to primary and subsequent levels of education. This implies aligning curricula and learning expectations, adapting pedagogies to children’s developmental stages, maintaining rich teacher-child interactions, and ensuring effective co‑ordination between settings. Sustaining early gains therefore requires not only preparing children for school, but also adapting primary education – and subsequent investments – to build on the skills acquired in the early years.

Box 4.4. Impact of ECEC on child development depends on quality and intensity of services provided, and inclusion of disadvantaged children

Copy link to Box 4.4. Impact of ECEC on child development depends on quality and intensity of services provided, and inclusion of disadvantaged children

The impact of early childhood education and care (ECEC) on child development critically depends on the quality and intensity of provision, as well as on the extent to which services effectively reach disadvantaged children.

Because of the cumulative nature of learning, more time spent in ECEC can support skill development, provided programmes are able to adapt content and practices to children’s evolving competencies (Li et al., 2020[54]; Melhuish and Gardiner, 2020[55]). Evidence suggests that more intensive participation (i.e. more hours) can yield positive effects – particularly for children from low socio‑economic backgrounds – when provision is of high quality. High-intensity ECEC (around 30 hours or more per week) is associated with stronger cognitive and language development, especially for vulnerable children. By contrast, evidence on socio‑emotional outcomes is less conclusive, and extended hours at very young ages (especially under age 2) may be associated with some longer-term behavioural risks. Overall, the literature indicates that intensive, high-quality ECEC generates the largest gains for disadvantaged children.

For very young children, however, findings are more mixed (Duncan et al., 2022[56]). Full-day or high-intensity participation at early ages often shows limited or ambiguous benefits, particularly for non-cognitive outcomes. While theories of skill formation and early inequality point to the potential value of early enrolment, empirical evidence suggests that “earlier” is not always “better,” and effects depend strongly on quality. Enrolment between ages 2 and 3 tends to be beneficial, whereas evidence for younger children is more mixed (Melhuish et al., 2015). Concerns about very early enrolment (especially under age 1) relate mainly to social, emotional and health outcomes, particularly when combined with long hours. For cognitive and language development, participation in the first year of life can have neutral or even negative effects, although high-quality ECEC between ages 1 and 3 is generally associated with positive outcomes in language, early numeracy and motor skills (Carbuccia et al., 2020[57]). For example, evidence from the French ELFE cohort shows positive effects on language skills, no impact on motor skills, and some negative effects on behaviour at age 1 (Berger, Panico and Solaz, 2020[58]). Importantly, positive effects – especially on language – are concentrated among disadvantaged children. More broadly, the literature consistently finds stronger impacts for children from low-income and low-educated families, while starting age appears less relevant for children from more advantaged backgrounds.

In addition to quality, the scale and intensity of provision matter for long-term impacts on inequalities. Evidence from intensive, targeted programmes in the United States – such as the Perry Preschool, Chicago Child-Parent Centers and the Abecedarian project – shows substantial long-term gains in education, health and earnings (McCoy et al., 2017[59]). However, these programmes were highly intensive, long in duration, and focussed on disadvantaged children, which limits their generalisability. For example, the Abecedarian programme provided year-round services over the first five years of life. These interventions were also costly, with per-child expenditures far exceeding typical spending levels in OECD countries (Whitaker et al., 2026[60]). More recent large‑scale programmes targeting broader populations have produced more mixed results, reinforcing the importance of quality in achieving positive outcomes at scale (Burchinal et al., 2024[61]).

Evidence from European ECEC systems, which tend to rely more on universal provision, further highlights the central role of quality. A meta‑analysis of 17 longitudinal studies across nine European countries finds small but persistent positive effects of process quality on language, literacy and mathematics outcomes (Ulferts, Wolf and Anders, 2019). Another meta‑analysis confirms that ECEC quality is a key determinant of lasting impacts and finds little evidence of fade‑out over time (Van Huizen and Plantenga, 2015[62]). In addition, gains are consistently concentrated among children from lower socio‑economic backgrounds.

Universal preschool programmes for children aged 3‑5 may have stronger impacts than targeted programmes for two key reasons (Duncan et al., 2022[56]). First, they can generate positive peer effects, as children benefit from interacting with a broader and more balanced group of classmates. Second, when entire cohorts enter school with higher and more homogeneous skill levels, teachers can build more effectively on these foundations. This reduces disparities in early skills and allows teaching to be pitched at a higher level for all students, fostering dynamic complementarity between preschool and later learning. By contrast, when skill gaps are large at school entry, teachers often focus on bringing lower-performing students up to a minimum level, which can limit overall learning gains and contribute to the fade‑out of early advantages.

While ECEC can provide a supportive environment for children’s early development and learning, it cannot on its own ensure that the benefits of early investment are sustained. To deliver lasting gains, ECEC policies need to be complemented by broader investments in education and health, so that early advantages translate into improved outcomes in adolescence and adulthood.

Sustaining investment in education across the school-age years

Investments in education are associated with improved educational outcomes at age 15. This finding reinforces earlier work by Egert et al. (2020[63]), which shows that higher public spending on education is linked to increases in average years of schooling – a proxy for human capital. Using an approach to identify factors increasing the effectiveness of spending on education that similar to the one used here for social policies, the authors also identify several policy features that amplify the returns to education spending, including higher participation in pre‑primary education, greater institutional autonomy, lower student-teacher ratios, a later age of tracking in secondary education, and fewer financial barriers to tertiary education.

A key contribution of the present analysis is to show that the level and allocation of education spending matter for different dimensions of performance. Countries with higher per capita investment in primary education, all else equal, tend to have fewer low performers in PISA, while spending on secondary education is more closely associated with a greater share of high performers. However, when the analysis is restricted to within-country trends over time, no statistically significant associations are found. This likely reflects limitations in the available data for identifying robust time‑series relationships. More substantially, it suggests that variations in per capita education spending within countries over the past two decades have not been consistently large enough to generate clear changes in the shares of low and high performers.

From a policy perspective, reducing the share of low performers and increasing the share of high performers are complementary objectives that help combine greater fairness with stronger responses to the future demand for highly skilled workers. Lowering the number of low performers is essential for improving life chances and promoting social inclusion, while increasing the share of high performers can be a goal in itself to support productivity and competitiveness in high-skill sectors. Achieving these objectives may therefore require targeted investments addressing different groups. In particular, boosting the number of high achievers may involve not only supporting those at risk of low performance, but also investing in students who would otherwise perform at intermediate levels. Combining improvements in overall academic performance – including increasing the share of students achieving high levels of performance – with greater fairness therefore requires ensuring that policies particularly benefit the most disadvantaged students – both by reducing their risk of falling behind and by strengthening their chances of reaching high levels of academic performance.

Aligning support to the stronger needs from low- and middle‑ SES children

The importance of cash transfers in helping families meet the costs of raising children – and their impact on poverty, with knock-on effects on children’s health and development – is well documented, as highlighted in Chapter 3. That chapter also shows that, while increasing pressure on social spending, a higher share of expenditure allocated to cash transfers is effective in reducing income poverty, but has more limited effects on severe material deprivation. The findings in the present chapter further indicate that the design of income support is a key determinant of its effectiveness in enhancing children’s educational and health outcomes.

Family-related benefits, in particular, appear to have a clearer association with adolescents’ educational and health outcomes. The results suggest that a higher share of social expenditure devoted to cash benefits for families is associated primarily with a reduced share of low performers in PISA. This is consistent with the predictability and the “framing effect” of family-targeted transfers highlighted in the literature (e.g. Page (2024[64])), which shows that the impact of income on child outcomes depends on its regularity and source. For example, windfall gains such as lottery winnings tend to have little effect on children’s outcomes, whereas structured transfers – especially those explicitly targeted at families with children – are more likely to generate positive impacts.

Linking benefits to the presence of children may create incentives for parents to allocate a larger share of resources to child-related investments than they would with equivalent unlabelled income. In line with this, Waldfogel (2025[65]) shows that cash support to families tends to increase child-related consumption across countries, regardless of whether transfers are earmarked or conditional, pointing to a broader responsiveness of household spending when additional resources are available in the presence of children.

Spend early to prevent, later to promote

The timing of cash benefit support across childhood also matters, with earlier spending playing a stronger role in preventing poor outcomes, and sustained spending during later years being more closely associated with promoting higher achievement and well-being. Specifically, increases in cash transfers during early and middle childhood are associated with reductions in the share of low-performing students in PISA, while higher transfers targeted at middle childhood and to some extent adolescence are linked to greater shares of high performers at age 15. This pattern is consistent with findings on education spending: higher investment in primary education is associated with fewer low performers, whereas higher spending in primary education and, more strongly, secondary education are linked to a greater share of high performers.

Similar dynamics are observed for children’s health outcomes. Greater spending during early childhood is linked to lower rates of children reporting fair or poor health at age 15 whereas sustained spending during early and middle childhood is associated with a higher proportion of 15‑year‑olds reporting excellent health. Echoing recent work by Hendren and Sprung-Keyser (2020[66]) and Page (2024[64]), these findings suggest that early childhood is not the only stage at which cash assistance can make a difference. Rather, they point to a more nuanced relationship between the timing of income support and educational and health outcomes in adolescence than is implied by a simple “skills-beget-skills” framework.

Explaining how cash benefit support contributes to higher school performance and better health is not straightforward. However, evidence shows that the costs of raising children increase markedly during adolescence in many countries, often reaching levels comparable to those of adults (Rapp and Thévenon, 2025[67]). This helps explain the importance of income support for families with adolescents, particularly in covering expenses related to education and broader developmental needs. Even when transfers are not specifically earmarked for education, they can ease budget constraints, enabling families to invest in a healthier and more supportive learning environment – such as better housing conditions, access to leisure, sports and social activities – all of which are important for adolescent well-being and engagement in learning. Further analysis of household spending patterns among families with adolescents would help clarify how income support is allocated and shed light on the mechanisms linking cash transfers to improved educational outcomes.

While the level and timing of spending on cash benefits are important determinants of effectiveness, other institutional features also play a key role in shaping their impact on children’s educational and health outcomes. In particular, eligibility conditions can be designed to better reach specific groups – including harder-to-reach populations such as Indigenous communities – and to link benefit receipt to behaviours that support child development (Banerjee et al., 2024[68]). These conditions may include requirements related to school enrolment, minimum academic performance, or participation in health checks, thereby increasing the likelihood that transfers translate into improved outcomes for children.

For example, the Chile’s Beca Indígena programme, in place since 1991, illustrates how conditional financial support can lead to substantial improvements in both educational attainment and adult labour market outcomes (Lucas, McEwan and Irribarra, 2025[69]). The programme provides renewable grants to Indigenous students enrolled in upper-primary, secondary, and tertiary education. Eligibility is based on certified Indigenous ancestry, school enrolment, minimum academic performance, and household income below a proxy-means threshold. Grants are sizeable relative to household income and can be renewed across successive levels of education. Since its inception, the programme has expanded significantly – from around 300 beneficiaries in 1991 to over 92 000 in 2021 – creating variation in exposure across cohorts. This allows for an assessment of its long-term effects. Among the most exposed Indigenous cohorts (born between 1992 and 2000), schooling increased by 0.6 years and labour earnings by around 20%. Scaling these results suggests that an additional year of grant exposure leads, on average, to a 0.3‑year increase in schooling and a 10% increase in earnings.

The effectiveness of such programmes depends on key contextual and design features (Lucas, McEwan and Irribarra, 2025[69]). In the case of Chile, institutional stability ensured consistent programme rules over more than three decades. The renewability of grants across educational stages provided sustained incentives to remain in education, while academic performance requirements encouraged student effort and achievement. These features were further reinforced by broader contextual conditions, including sustained investments in education quality and favourable labour market dynamics associated with economic growth, which increased demand for skilled workers.

Tailoring the scope and intensity of support to the needs of low- and middle‑SES children

Achieving both higher educational and health outcomes across the population and greater equality of opportunity requires tailored support for children with the greatest needs. This implies following the principle of proportionate universalism, whereby policies are universal in scope but provide more intensive support to the most disadvantaged children according to their level of need.

The findings of the present analysis highlight the important role of preschool attendance and family cash benefits in substantially reducing the share of low-performing students from low- and middle‑SES backgrounds, while also increasing their chances of reaching high levels of performance and, in turn, improving prospects for social mobility. The effects of family cash benefits during early and middle childhood also appear particularly strong for low-SES children in preventing low performance, suggesting both greater needs and higher returns to income support among disadvantaged groups.

Larger average income support is associated with significantly higher chances for low- and middle‑SES students to achieve high performance. By contrast, stronger targeting of support towards households in the bottom 30% of the income distribution does not appear to increase effectiveness, although it may help contain social spending pressures. One likely explanation is that stronger targeting often coincides with lower overall levels of income support. The results even suggest lower chances of high academic performance among middle‑SES students, possibly because they are disproportionately affected by benefit reductions while also representing the largest pool of potential high achievers: across the OECD, middle‑SES students account on average for 43% of high performers and around half of low performers.

At the same time, a major challenge remains in expanding opportunities for low-SES students to achieve high levels of academic performance. Across the OECD, only around 5% of 15‑year‑olds from the bottom quarter of the socio‑economic distribution are high performers, compared with 13% among students from middle socio‑economic backgrounds and 28% among those from the top quarter.

The mixed effects of targeting income support towards the lowest-income households are not entirely surprising. As shown in Chapter 3, directing a larger share of benefits to these groups can help reduce income poverty and contain pressures on social spending, but tends to have more limited effects on severe material deprivation and, likely, on the broader home learning environment, which extends beyond the poorest households. The results presented in this chapter further suggest that increasing the share of benefits allocated to the bottom 30% of the income distribution is associated with a decline in the share of high performers. This may reflect the limits of a narrowly targeted approach when it is not accompanied by stronger support for low- and middle‑income families, who constitute an important pool of potential high achievers.

At the same time, stronger targeting appears to generate benefits for health outcomes. Increasing the share of benefits directed to the bottom 30% is associated with reductions in the share of adolescents reporting multiple subjective health complaints, with stronger effects observed at higher levels of spending. Similarly, reducing the tax burden on lower-income households is associated with improvements in adolescents’ self-reported health.

Overall, these findings suggest that while targeting income support towards lower-income groups can help contain spending pressures and improve certain health outcomes, a broader distribution of support that also reaches low- and middle‑income households may be more effective in reducing severe material deprivation, strengthening home learning environments, and sustaining higher educational performance.

Investing in health to strengthen adolescent health and contain long-term social spending

Increased health spending is associated with some improvements in adolescents’ health, with preventive health expenditure standing out as the component most strongly linked to increases in the share of children reporting excellent health. By contrast, the absence of a clear association between curative (non-preventive) health spending and adolescent health outcomes may reflect data limitations, particularly the difficulty of identifying the share of such expenditure that directly benefits children. Strengthening the estimation of child-specific health expenditure data would help clarify how different types of health spending translate into improvements in child health outcomes, particularly given that such estimates are not routinely produced (Morgan and Mueller, 2023[70]).

Interestingly, while no association is found between health spending and performance in PISA tests, the results suggest that higher health expenditure is linked to lower rates of young NEETs. Although this relationship warrants further investigation to better understand the underlying mechanisms, it is consistent with a broader body of research showing that NEET status is associated with long-term socio‑economic disadvantages, including lower educational attainment, higher unemployment risks, and poorer mental health, alongside low self-esteem and social exclusion (OECD, 2025[71]; Gunnes et al., 2025[72]). These findings underline the importance of health support as part of a broader policy approach to help disengaged youth reconnect with education and employment.

Finally, a higher share of social spending allocated to health appears, overall, to help moderate the long-run growth of social expenditures, likely by maintaining the employability of the working-age population and preventing the need for more costly interventions as health conditions worsen.

The potential of ALMPs to enhance children’s health

Last, there is some indication that spending on ALMPs, although not primarily designed for this purpose, may contribute to improving adolescent health. This finding warrants further investigation to identify more precisely the types of programmes and mechanisms at play, as children are not the direct beneficiaries of these measures. One possible explanation is that ALMPs improve parents’ employment prospects and job stability, thereby strengthening household resources and reducing financial stress. This, in turn, may ease parental pressure, support more positive parenting practices, and foster a more stable and supportive home environment – conditions that are conducive to good health among children. While this interpretation would need to be confirmed using individual-level data, it is consistent with the broader social objectives that ALMPs are intended to achieve in some countries (see Chapter 1).

References

[68] Banerjee, A. et al. (2024), “Social Protection in the Developing World”, Journal of Economic Literature, Vol. 62/4, pp. 1349-1421, https://doi.org/10.1257/JEL.20241646.

[24] Barone, C. and L. Ruggera (2018), “Educational equalization stalled? Trends in inequality of educational opportunity between 1930 and 1980 across 26 European nations”, European Societies, Vol. 20/1, pp. 1-25, https://doi.org/10.1080/14616696.2017.1290265.

[58] Berger, L., L. Panico and A. Solaz (2020), “The Impact of Center-Based Childcare Attendance on Early Child Development: Evidence from the French Elfe Cohort”, No. 254, INED, Paris.

[29] Bernardi, F. (2014), “Compensatory advantage as a mechanism of educational inequality: A regression discontinuity based on month of birth”, Sociology of Education, Vol. 87/2, pp. 74-88, https://doi.org/10.1177/0038040714524258.

[23] Bernardi, F. and G. Ballarino (2016), Education, Occupation and Social Origin: A Comparative Analysis of the Transmission of Socio-economic Inequalities, Edward Elgar Publishing, https://hdl.handle.net/1814/41005.

[32] Bernardi, F. and G. Ballarino (2014), “Participation, equality of opportunity and returns to tertiary education in contemporary Europe”, European Societies, Vol. 16/3, pp. 422-442, https://doi.org/10.1080/14616696.2012.750729.

[21] Bernardi, F. and I. Plavgo (2019), “Education as an equalizer for human development?”, UNDP (United Nations Development Programme), New York, https://hdr.undp.org/content/education-equalizer-human-development.

[30] Bernardi, F. and M. Triventi (2020), “Compensatory advantage in educational transitions: Trivial or substantial? A simulated scenario analysis”, Acta Sociologica, Vol. 63/1, pp. 40-62, https://doi.org/10.1177/0001699318780950.

[28] Bradbury, B. and M. Corak (2015), Too Many Children Left Behind: The U.S. Achievement Gap in Comparative Perspective, Russell Sage Foundation, https://www.jstor.org/stable/10.7758/9781610448482.

[22] Breen, R. et al. (2009), “Nonpersistent inequality in educational attainment: Evidence from eight European countries”, American Journal of Sociology, Vol. 114/5, pp. 1475-1521, https://doi.org/10.1086/595951.

[61] Burchinal, M. et al. (2024), “Unsettled science on longer-run effects of early education”, Science, Vol. 384/6695, pp. 506-508, https://www.science.org/doi/10.1126/science.adn2141.

[37] Bynner, J. and S. Parsons (2002), “Social Exclusion and the Transition from School to Work: The Case of Young People Not in Education, Employment, or Training (NEET)”, Journal of Vocational Behavior, Vol. 60/2, pp. 289-309, https://doi.org/10.1006/jvbe.2001.1868.

[57] Carbuccia, L. et al. (2020), Revue de littérature sur les politiques d’accompagnement au développement des capacités des jeunes enfants [Rapport de recherche], Laboratoire interdisciplinaire d’évaluation des politiques publiques, https://sciencespo.hal.science/hal-03455799.

[53] Caucutt, E. et al. (2026), “Child Skill Production: Accounting for Parental and Market-Based Time and Goods Investments”, Journal of Political Economy, Vol. 134/1, pp. 150-209, https://doi.org/10.1086/738480.

[48] Clarke, C. and O. Thévenon (2022), “Starting unequal: How’s life for disadvantaged children?”, OECD Policy Papers on Well-being and Inequalities, No. 6, OECD Publishing, Paris, https://doi.org/10.1787/a0ec330c-en.

[44] Coller, R. and A. Kuo (2015), “Social Determinants of Child Health”, Child Health, pp. 79-110, https://doi.org/10.1093/MED/9780199309375.003.0005.

[19] 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.

[25] Downey, D. and D. Condron (2016), “Fifty Years since the Coleman Report: Rethinking the Relationship between Schools and Inequality”, Sociology of Education, Vol. 89/3, pp. 207-220, https://doi.org/10.1177/0038040716651676.

[18] Dräger, J. et al. (2024), “The relevance of tracking and social school composition for growing achievement gaps by parental education in lower secondary school: a longitudinal analysis in France, Germany, the United States, and England”, European Sociological Review, Vol. 40/6, pp. 964-980, https://doi.org/10.1093/ESR/JCAD076.

[56] Duncan, G. et al. (2022), “Investing in Early Childhood Development in Preschool and at Home”, NBER Working Paper Series, Vol. 29985, https://doi.org/10.3386/w29985.

[7] Edwards, B. (2005), “Does it take a village? An investigation of neighbourhood effects on Australian children’s development.”.

[63] Égert, B., J. Botev and D. Turner (2020), “The contribution of human capital and its policies to per capita income in Europe and the OECD”, European Economic Review, Vol. 129, p. 103560, https://doi.org/10.1016/J.EUROECOREV.2020.103560.

[15] European Commission (2020), “European Skills Agenda for Sustainable Competitiveness, Social Fairness and Resilience”, https://employment-social-affairs.ec.europa.eu/policies-and-activities/skills-and-qualifications/european-skills-agenda_en (accessed on 21 May 2026).

[3] Flores, M., P. García-Gómez and A. Kalwij (2020), “Early life circumstances and labor market outcomes over the life cycle”, Journal of Economic Inequality, Vol. 18/4, pp. 449-468, https://doi.org/10.1007/s10888-020-09446-7.

[72] Gunnes, M. et al. (2025), “Young adults not in education, employment, or training (NEET): a global scoping review”, BMC Public Health, Vol. 25/1, https://doi.org/10.1186/S12889-025-24781-Y.

[17] Hanushek, E. and L. Woessmann (2015), The Knowledge Capital of Nations: Education and the Economics of Growth, The MIT Press, https://www.jstor.org/stable/j.ctt17kk9kq.

[16] Hanushek, E. and L. Woessmann (2008), “The Role of Cognitive Skills in Economic Development”, Journal of Economic Literature, Vol. 46/3, pp. 607-68, https://doi.org/10.1257/JEL.46.3.607.

[40] HBSC (2026), Health Behaviour in School-aged Children (HBSC) Survey, https://www.hbsc.org/.

[27] Heckman, J. (2006), “Skill formation and the economics of investing in disadvantaged children”, Science, Vol. 312, pp. 1900-1902, https://doi.org/10.1126/science.1128898.

[66] Hendren, N. and B. Sprung-Keyser (2020), A Unified Welfare Analysis of Government Policies, Oxford University Press, https://doi.org/10.1093/qje/qjaa006.

[54] Li, W. et al. (2020), “Timing in Early Childhood Education: How Cognitive and Achievement Program Impacts Vary by Starting Age, Program Duration, and Time Since the End of the Program”, Ed WorkingPaper: 20-201, https://doi.org/10.26300/5tvg-nt21.

[69] Lucas, A., P. McEwan and D. Irribarra (2025), “The Long-Run and Intergenerational Effects of Conditional Cash Transfers: Evidence from Chile’s Indigenous Grants”, NBER Working Paper, No. 33798, https://doi.org/10.3386/W33798.

[2] Mallo, M. and B. Wolfe (2020), The Influence of Early Life Health Conditions on Life Course Health, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w27174.

[59] McCoy, D. et al. (2017), “Impacts of Early Childhood Education on Medium- and Long-Term Educational Outcomes”, Educational Researcher, Vol. 46/8, pp. 474-487, https://doi.org/10.3102/0013189X17737739.

[46] Mcknight, A. (2015), “A fresh look at an old question: is pro-poor targeting of cash transfers more effective than universal systems at reducing inequality and poverty?”, CASE, No. 191, Centre for Analysis of Social Exclusion, LSE, London, https://researchonline.lse.ac.uk/id/eprint/103977/1/McKnight_a_fresh_look_at_an_old_question_published.pdf (accessed on 10 April 2018).

[55] Melhuish, E. and J. Gardiner (2020), Study of Early Education and Development (SEED): Impact Study on Early Education Use and Child Outcomes up to age five years, Department for Education, https://assets.publishing.service.gov.uk/media/5e4e5c1ed3bf7f393c09149e/SEEDage5briefFEB.pdf (accessed on 27 April 2026).

[41] Moore, T. et al. (2015), “Early childhood development and the social determinants of health inequities”, Health Promotion International, Vol. 30/suppl 2, pp. ii102-ii115, https://doi.org/10.1093/heapro/dav031.

[70] Morgan, D. and M. Mueller (2023), “Understanding international measures of health spending: Age-adjusting expenditure on health”, OECD Health Working Papers, No. 162, OECD Publishing, Paris, https://doi.org/10.1787/043ed664-en.

[86] Moritz, S. and T. Bartz-Beielstein (2017), “imputeTS: Time Series Missing Value Imputation in R”, The R Journal, Vol. 9/1, p. 207, https://doi.org/10.32614/rj-2017-009.

[52] Morrissey, T. (2019), “The Effects Of Early Care And Education On Children’s Health”, Health Affairs Brief, https://doi.org/10.1377/hpb20190325.519221.

[6] Nieuwenhuis, J. and P. Hooimeijer (2016), “The association between neighbourhoods and educational achievement, a systematic review and meta-analysis”, Journal of Housing and the Built Environment, Vol. 31/2, pp. 321-347, https://doi.org/10.1007/s10901-015-9460-7.

[51] OECD (2026), PF3.2 Enrolment in childcare and pre-school, OECD Family Database, https://webfs.oecd.org/Els-com/Family_Database/PF3_2_Enrolment_childcare_preschool.pdf (accessed on 4 September 2026).

[33] OECD (2026), PISA 2025 Results (Volume I): Future-Ready Students, PISA, OECD Publishing, Paris, https://doi.org/10.1787/73451bc5-en.

[4] OECD (2025), “Building child-friendly neighbourhoods: Empowering communities with data-driven action”, OECD Papers on Well-being and Inequalities, No. 33, OECD Publishing, Paris, https://doi.org/10.1787/4f46a260-en.

[71] OECD (2025), Education at a Glance 2025: OECD Indicators, OECD Publishing, Paris, https://doi.org/10.1787/1c0d9c79-en.

[47] OECD (2025), Health at a Glance 2025: OECD Indicators, OECD Publishing, Paris, https://doi.org/10.1787/8f9e3f98-en.

[1] OECD (2025), OECD Skills Outlook 2025: Building the Skills of the 21st Century for All, OECD Publishing, Paris, https://doi.org/10.1787/26163cd3-en.

[49] OECD (2025), Reducing Inequalities by Investing in Early Childhood Education and Care, Starting Strong, OECD Publishing, Paris, https://doi.org/10.1787/b78f8b25-en.

[5] OECD (2025), “The importance of monitoring neighbourhood conditions for children’s well-being and development”, OECD Papers on Well-being and Inequalities, No. 34, OECD Publishing, Paris, https://doi.org/10.1787/143ae959-en.

[73] OECD (2024), PISA 2022 Technical Report, PISA, OECD Publishing, Paris, https://doi.org/10.1787/01820d6d-en.

[14] 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.

[45] OECD (2021), Measuring What Matters for Child Well-being and Policies, OECD Publishing, Paris, https://doi.org/10.1787/e82fded1-en.

[20] OECD (2018), A Broken Social Elevator? How to Promote Social Mobility, OECD Publishing, Paris, https://doi.org/10.1787/9789264301085-en.

[50] OECD (2011), Doing Better for Families, OECD Publishing, Paris, https://doi.org/10.1787/9789264098732-en.

[39] OECD (n.d.), Education at a Glance Database, https://data-explorer.oecd.org/s/5dh.

[74] OECD (n.d.), OECD Child Well-being Data Portal, https://www.oecd.org/en/data/datasets/income-and-wealth-distribution-database.html.

[83] OECD (n.d.), OECD Education at a Glance, https://data-explorer.oecd.org/s/51d (accessed on 14 September 2026).

[75] OECD (n.d.), OECD Education Database, https://data-explorer.oecd.org/s/5ca (accessed on 14 September 2026).

[79] OECD (n.d.), OECD Family Database, https://www.oecd.org/en/data/datasets/oecd-family-database.html (accessed on 14 September 2026).

[78] OECD (n.d.), OECD Health expenditure and financing, https://data-explorer.oecd.org/s/51c (accessed on 14 September 2026).

[84] OECD (n.d.), OECD Health Statistics, https://data-explorer.oecd.org/s/515 (accessed on 14 September 2026).

[85] OECD (n.d.), OECD Historical population data, https://data-explorer.oecd.org/s/516 (accessed on 14 September 2026).

[80] OECD (n.d.), OECD Income Distribution Database, https://www.oecd.org/en/data/datasets/income-and-wealth-distribution-database.html (accessed on 14 September 2026).

[81] OECD (n.d.), OECD National Accounts, https://data-explorer.oecd.org/s/2lm (accessed on 14 September 2026).

[76] OECD (n.d.), OECD Social Expenditure Database, https://data-explorer.oecd.org/s/5cb (accessed on 14 September 2026).

[38] OECD (n.d.), PISA Database, https://www.oecd.org/en/about/programmes/pisa/pisa-data.html (accessed on 8 September 2026).

[77] OECD/Eurostat/WHO (2017), A System of Health Accounts 2011: Revised edition, OECD Publishing, Paris, https://doi.org/10.1787/9789264270985-en.

[64] Page, M. (2024), “New Advances on an Old Question: Does Money Matter for Children’s Outcomes?”, Journal of Economic Literature, Vol. 62/3, pp. 891-947, https://doi.org/10.1257/JEL.20231553.

[11] Palloni, A. et al. (2009), “Early childhood health, reproduction of economic inequalities and the persistence of health and mortality differentials”, Social Science and Medicine, Vol. 68/9, pp. 1574-1582, https://doi.org/10.1016/j.socscimed.2009.02.009.

[12] Pearce, A. et al. (2019), “Pathways to inequalities in child health”, Archives of Disease in Childhood, Vol. 104/10, pp. 998-1003, https://doi.org/10.1136/archdischild-2018-314808.

[13] Pickett, K. et al. (2022), “The social determinants of child health and inequalities in child health”, Paediatrics and Child Health, Vol. 32/3, pp. 88-94, https://doi.org/10.1016/J.PAED.2021.12.003.

[35] Rahmani, H. and W. Groot (2023), “Risk Factors of Being a Youth Not in Education, Employment or Training (NEET): A Scoping Review”, International Journal of Educational Research, Vol. 120, p. 102198, https://doi.org/10.1016/j.ijer.2023.102198.

[67] Rapp, S. and O. Thévenon (2025), “Parenting on a budget: The cost of raising children in changing family configurations and its policy implications”, OECD Papers on Well-being and Inequalities, No. 36, OECD Publishing, Paris, https://doi.org/10.1787/bd0b594e-en.

[10] Rebouças, P., I. Falcão and M. Barreto (2021), “Social inequalities and their impact on children’s health: a current and global perspective”, Jornal de Pediatria, Vol. 98/Suppl 1, p. S55, https://doi.org/10.1016/J.JPED.2021.11.004.

[34] Ripamonti, E. (2023), “School-to-work transition: putting non-cognitive skills in context. The case of NEET and suggestions for policy”, International Journal for Educational and Vocational Guidance, Vol. 25/2, pp. 335-356, https://doi.org/10.1007/s10775-023-09635-6.

[9] Robinette, J., S. Charles and T. Gruenewald (2017), “Neighborhood Socioeconomic Status and Health: A Longitudinal Analysis”, Journal of community health, Vol. 42/5, p. 865, https://doi.org/10.1007/S10900-017-0327-6.

[42] Singh, G., M. Siahpush and M. Kogan (2010), “Neighborhood socioeconomic conditions, built environments, and childhood obesity”, Health Affairs, Vol. 29/3, pp. 503-512, https://doi.org/10.1377/hlthaff.2009.0730.

[43] Spencer, N. (2018), “The social determinants of child health”, Paediatrics and Child Health, Vol. 28/3, pp. 138-143, https://doi.org/10.1016/J.PAED.2018.01.001.

[36] Thompson, R. (2017), “Opportunity structures and educational marginality: the post-16 transitions of young people outside education and employment”, Oxford Review of Education, Vol. 43/6, pp. 749-766, https://doi.org/10.1080/03054985.2017.1352502.

[8] Troost, A., M. van Ham and D. Manley (2023), “Neighbourhood effects on educational attainment. What matters more: Exposure to poverty or exposure to affluence?”, PLOS ONE, Vol. 18/3, p. e0281928, https://doi.org/10.1371/JOURNAL.PONE.0281928.

[82] UNESCO (n.d.), UNESCO UIS Methodology for Estimation of Mean Years of Schooling, https://databrowser.uis.unesco.org/view#s=&geoMode=countries&geoUnits=&timeMode=range&view=table&chartMode=multiple&chartHighlightSeries=&chartHighlightEnabled=true&indicatorPaths=UIS-EducationOPRI%3A0%3AMYS.1T8.AG25T99.

[62] Van Huizen, T. and J. Plantenga (2015), “Universal Child Care and Children’s Outcomes: A Meta-Analysis of Evidence from Natural Experiments”, U.S.E. Discussion Paper Series, No. 15-13, Utrecht School of Economics, https://www.uu.nl/sites/default/files/rebo_use_dp_2015_15-13.pdf.

[26] von Hippel, P., J. Workman and D. Downey (2018), “Inequality in Reading and Math Skills Forms Mainly before Kindergarten: A Replication, and Partial Correction, of “Are Schools the Great Equalizer?””, Sociology of education, Vol. 91/4, p. 323, https://doi.org/10.1177/0038040718801760.

[65] Waldfogel, J. (2025), Child Benefits: A Smart Investment for America’s Future, Russell Sage Foundation, https://doi.org/10.7758/LIWL3023.

[60] Whitaker, A. et al. (2026), “Why are preschool programs becoming less effective?”, Journal of Policy Analysis and Management, Vol. 45/1, p. e70031, https://doi.org/10.1002/PAM.70031.

[31] Yastrebov, G., Y. Kosyakova and D. Kurakin (2018), “Slipping Past the Test: Heterogeneous Effects of Social Background in the Context of Inconsistent Selection Mechanisms in Higher Education”, Sociology of Education, Vol. 91/3, pp. 224-241, https://doi.org/10.1177/0038040718779087.

Annex 4.A. Sample description

Copy link to Annex 4.A. Sample description

The sample is based on available data for 32 OECD countries from 1997 to 2021, excluding Australia, Chile, Colombia, Costa Rica, Japan and Mexico due to extensive missing data across variables and over time. Detailed information on the indicator construction is provided in Annex Table 4.A.1and summary statistics are presented in Annex Table 4.A.2. For variables that include imputed values, the statistics represent average values across the five datasets that result from the imputation process. For further information on the imputations, see Annex 4.B. The regression analyses are restricted to the period 2002-2021. Data for the earlier years (1997 to 2001) serve to incorporate time lags.

Indicator	Definition	Data source
Child Outcome variables
Share of PISA high achievers	The percentage of 15‑year‑old students who attained Level 5 or 6 in at least one of the three core PISA subjects (reading, mathematics and science). For more details on the construction of the PISA proficiency scales and proficiency levels, see the PISA 2022 Technical Report (OECD, 2024[73]) and the corresponding Technical Reports from earlier rounds. PISA test results are used from 2009, 2012, 2015, 2018 and 2022. Data for years between surveys are imputed using Stineman interpolation.	OECD Child Well-being Data Portal (OECD, n.d.[74])
Share of PISA low achievers	The percentage of 15‑year‑old students who attained below Level 2 in at least one of the three core PISA subjects (reading, mathematics and science). For more details on the construction of the PISA proficiency scales and proficiency levels, see the PISA 2022 Technical Report (OECD, 2024[73]) and the corresponding Technical Reports from earlier rounds. PISA test results are used from 2009, 2012, 2015, 2018 and 2022. Data for years between surveys are imputed using a Stineman interpolation.	OECD Child Well-being Data Portal (OECD, n.d.[74])
NEETs	Percentage of population aged 15‑29 years not in education, employment or training. Children and young people are classified as “NEET” if they had neither received formal education and/or training in the regular educational system in the four weeks prior to being surveyed, nor were either working for pay or profit for at least one hour or had a job but were temporarily not at work during the survey reference week. Education or training corresponds to formal education; therefore, someone not working but following non-formal studies is considered NEET.	OECD Education Database (OECD, n.d.[75])
Share of 15‑year‑olds reporting excellent health	The percentage of 15‑year‑old school age children who responded with “excellent” when they were asked “Would you say your health is …?” and presented with the response options “Excellent”, “Good”, “Fair” and “Poor”. Data are taken from the HBSC surveys 2013‑2014, 2017‑2018, and 2021‑2022 and data for years between surveys are imputed using a Stineman interpolation. Values for subnational jurisdictions in Belgium and the United Kingdom were aggregated to form a national average.	OECD Child Well-being Data Portal (OECD, n.d.[74])
Share of 15‑year‑olds reporting low health	The percentage of 15‑year‑old school age children who responded with “poor” or “fair” when they were asked “Would you say your health is …?” and presented with the response options “Excellent”, “Good”, “Fair” and “Poor”. Data are taken from the HBSC surveys 2013‑2014, 2017‑2018, and 2021‑2022 and data for years between surveys are imputed using a Stineman interpolation. Values for subnational jurisdictions in Belgium and the United Kingdom were aggregated to form a national average.	OECD Child Well-being Data Portal (OECD, n.d.[74])
Share of 15‑year‑olds reporting multiple subjective health complaints	The percentage of 15‑year‑old school age children who report having experienced at least two symptoms of the following list more than once a week in the past six months: 1) ”Headache”, 2) ”Stomach-ache”, 3) ”Backache”, 4) ”Feeling low”, 5) ”Irritability or bad temper”, 6) ”Feeling nervous”, 7) ”Difficulties in getting to sleep” and 8) ”Feeling dizzy”. Data are taken from the HBSC surveys 2013‑2014, 2017‑2018, and 2021‑2022 and data for years between surveys are imputed using a Stineman interpolation. Values for subnational jurisdictions in Belgium and the United Kingdom were aggregated to form a national average.	OECD Child Well-being Data Portal (OECD, n.d.[74])
Policy variables
Spending on health services	The variable refers to the share of public and mandatory private and voluntary private expenditures on health in the total social expenditure (including spending on families, health, survivor and incapacity-related benefits, housing, active labour market programmes, unemployment, and other social policy areas, but excluding spending on old age and pensions). The variable is denoted in per capita USD PPP and constant 2015 prices.	OECD Social Expenditure Database (OECD, n.d.[76])
Spending on preventive health services	This indicator approximates the share of public and mandatory private and voluntary private expenditures on preventive care in the total social expenditure (including spending on families, health, survivor and incapacity-related benefits, housing, active labour market programmes, unemployment, and other social policy areas, but excluding spending on old age and pensions). The variable is denoted in per capita USD PPP and constant 2015 prices, and some values have been imputed. It is constructed by multiplying the indicator “Spending on health services” described above with the share of preventive care in government/compulsory schemes, voluntary schemes/household out-of-pocket payments and non-resident financing schemes in total expenditure on health. Following the Classification of Healthcare Functions (ICHA-HC), preventive care refers to primary prevention (health measures aimed at avoiding diseases and risk factors, e.g. vaccinations) and secondary prevention (detection of disease and therapy as early as possible, e.g. via screening). It includes information, education and counselling programmes; immunisation programmes; early disease detection programmes; healthy condition monitoring programmes; epidemiological surveillance and risk and disease control programmes; and preparing for disaster and emergency response programmes. For more detailed information, see Chapter 5 in A System of Health Accounts 2011: Revised edition (OECD/Eurostat/WHO, 2017[77]).	OECD Social Expenditure Database (OECD, n.d.[76]) and OECD Health expenditure and financing (OECD, n.d.[78])
Spending on non-preventive health services	This indicator approximates the share of public and mandatory private and voluntary private expenditures on non-preventive care in the total social expenditure (including spending on families, health, survivor and incapacity-related benefits, housing, active labour market programmes, unemployment, and other social policy areas, but excluding spending on old age and pensions). The variable is denoted in per capita USD PPP and constant 2015 prices. It represents the difference between the indicators “Spending on health services” and “Spending on preventive health services”. Accordingly, some values rely on imputed values for spending on preventative health services.	OECD Social Expenditure Database (OECD, n.d.[76]) and OECD Health expenditure and financing (OECD, n.d.[78])
ECEC enrolment 0‑2	The per cent of children aged 0 to 2 enrolled in formal care institutions. Some values have been imputed.	OECD Family Database, Indicator PF3.2 (OECD, n.d.[79])
ECEC enrolment 3‑5	The per cent of children aged 3 to 5 enrolled in early childhood education and care services. Some values have been imputed.	OECD Family Database, Indicator PF3.2 (OECD, n.d.[79])
Spending on ECEC	The variable refers to the share of public and mandatory private and voluntary private expenditures on ECEC services in the total social expenditure (including spending on families, health, survivor and incapacity-related benefits, housing, active labour market programmes, unemployment, and other social policy areas, but excluding spending on old age and pensions). The variable is denoted in per capita USD PPP and constant 2015 prices.	OECD Social Expenditure Database (OECD, n.d.[76])
Spending on ALMP	The variable refers to the share of public and mandatory private and voluntary private expenditures on active labour market programmes in the total social expenditure (including spending on families, health, survivor and incapacity-related benefits, housing, active labour market programmes, unemployment, and other social policy areas, but excluding spending on old age and pensions). The variable is denoted in per capita USD PPP and constant 2015 prices.	OECD Social Expenditure Database (OECD, n.d.[76])
Other social spending in kind	The indicator comprises the public and mandatory private and voluntary private expenditures paid in kind for incapacity-related, housing and other social policy areas as a share of total social expenditure (cash and in-kind in the areas of families, health, survivor and incapacity-related benefits, housing, active labour market programmes, unemployment, and other social policy areas, but excluding spending on old age and pensions). The variable is denoted in per capita USD PPP and constant 2015 prices. The indicator excludes from the numerator spending on active labour market programmes, families and health (they enter as separate policy variables), unemployment (countries do not report in-kind benefits in this area), survivors (they mainly cover in-kind benefits related to the funeral and are thus less relevant to child outcomes) and old-age benefits (outside the scope).	OECD Social Expenditure Database (OECD, n.d.[76])
Cash transfers to families	The variable refers to the share of public and mandatory private and voluntary private expenditures allocated to families in cash in the total social expenditure (including spending on families, health, survivor and incapacity-related benefits, housing, active labour market programmes, unemployment, and other social policy areas, but excluding spending on old age and pensions). The variable is denoted in per capita USD PPP and constant 2015 prices.	OECD Social Expenditure Database (OECD, n.d.[76])
Cash transfers to families during early childhood (ages 0‑5)	This indicator approximates the public and mandatory private and voluntary private expenditures that is allocated in cash to families with children aged 0‑5 as a share of total social expenditure (including spending on families, health, survivor and incapacity-related benefits, housing, active labour market programmes, unemployment, and other social policy areas, but excluding spending on old age and pensions). The indicator is denoted in per capita USD PPP and constant 2015 prices. It is constructed by multiplying the indicator “Cash transfers to families” described above with age‑spending profiles. The latter capture the average family cash benefits for a child at ages 0‑5 as a share of the average cash benefits for children at ages 0‑17 and contain imputed values for some years.	OECD Social Expenditure Database (OECD, n.d.[76]) and OECD Family Database, Indicator PF1.6 (OECD, n.d.[79])
Cash transfers to families during middle childhood (ages 6‑11)	This indicator approximates the public and mandatory private and voluntary private expenditures that is allocated in cash to families with children aged 6‑11 as a share of total social expenditure (including spending on families, health, survivor and incapacity-related benefits, housing, active labour market programmes, unemployment, and other social policy areas, but excluding spending on old age and pensions). The indicator is denoted in per capita USD PPP and constant 2015 prices. It is constructed by multiplying the indicator “Cash transfers to families” described above with age‑spending profiles. The latter capture the average family cash benefits for a child at ages 6‑11 as a share of the average cash benefits for children at ages 0‑17 and contain imputed values for some years.	OECD Social Expenditure Database (OECD, n.d.[76]) and OECD Family Database, Indicator PF1.6 (OECD, n.d.[79])
Cash transfers to families during adolescence (ages 12‑17)	This indicator approximates the public and mandatory private and voluntary private expenditures that is allocated in cash to families with children aged 12‑17 as a share of total social expenditure (including spending on families, health, survivor and incapacity-related benefits, housing, active labour market programmes, unemployment, and other social policy areas, but excluding spending on old age and pensions). The indicator is denoted in per capita USD PPP and constant 2015 prices. It is constructed by multiplying the indicator “Cash transfers to families” described above with age‑spending profiles. The latter capture the average family cash benefits for a child at ages 12‑17 as a share of the average cash benefits for children at ages 0‑17 and contain imputed values for some years.	OECD Social Expenditure Database (OECD, n.d.[76]) and OECD Family Database, Indicator PF1.6 (OECD, n.d.[79])
Volume of redistribution	This indicator is calculated as the average amount per equivalent household member of public transfers received (public social security, excluding employment-related social insurance transfers and private transfers received) minus taxes on income and wealth paid and of contributions paid by working-age individuals to public social security schemes, expressed as a percentage of GDP per capita. Some values have been imputed.	OECD Income Distribution Database (OECD, n.d.[80]) and OECD National Accounts (OECD, n.d.[81])
Average size of taxes	The average taxes per equivalent household member paid by working-age individuals, as a share of average disposable income per equivalent household member of working-age individuals. Some values have been imputed.	OECD Income Distribution Database (OECD, n.d.[80])
Average size of benefits	The average public transfers (public social security) per equivalent household member received by working-age individuals, as a share of average disposable income per equivalent household member of working-age individuals. Some values have been imputed.	OECD Income Distribution Database (OECD, n.d.[80])
Share of taxes paid by the bottom 30%	The taxes per equivalent household member paid by working-age individuals in the bottom 30% of post-transfer, equivalised household income, as a share of taxes per equivalent household member paid by all working-age individuals. Some values have been imputed.	OECD Income Distribution Database (OECD, n.d.[80])
Share of benefits received by the bottom 30%	The public transfers (public social security) per equivalent household member received by working-age individuals in the bottom 30% of pre‑transfer income, as a share of public transfers per equivalent household member received by all working-age individuals. Some values have been imputed.	OECD Income Distribution Database (OECD, n.d.[80])
Child outcome drivers
Total social expenditure (TS)	Total social spending refers to countries’ public, private mandatory and private voluntary expenditures in-cash and in-kind in eight different policy areas: health, incapacity-related benefits, unemployment, active labour market programmes, family, housing, survivors, and other social policy areas (e.g. social assistance). It excludes spending on old age. It is denoted in constant USD PPP (base 2015 = 100) and in logarithmic form.	OECD Social Expenditure Database (OECD, n.d.[76])
Mean poverty gap before taxes and transfers	The mean poverty gap before taxes and transfers, i.e. the difference between the average equivalised pre‑tax income of individuals living in a working-age household below the relative poverty line and the relative poverty line, as a share of the relative poverty line. Some values have been imputed.	OECD Income Distribution Database (OECD, n.d.[80])
Mean years of schooling	Average number of completed years of education of a country’s population aged 25 years and older, excluding years spent repeating individual grades. Some values have been imputed.	UNESCO UIS Methodology for Estimation of Mean Years of Schooling (UNESCO, n.d.[82])
Spending on primary education	General government spending (at central, state and local level) for primary educational institutions across all educational institutions (both public and private), in constant US dollars per student, PPP converted (price base 2020). Some values have been imputed.	OECD Education at a Glance (OECD, n.d.[83])
Spending on secondary education	General government spending (at central, state and local level) for secondary educational institutions across all educational institutions (both public and private), in constant US dollars per student, PPP converted (price base 2020). Some values have been imputed.	OECD Education at a Glance (OECD, n.d.[83])
Total social spending drivers
GDP per capita	GDP per capita net of total social spending per capita (as defined in “total social expenditure” above), denoted in USD PPP in constant prices and in logarithmic form.	OECD National Accounts (OECD, n.d.[81])
Life expectancy	Years of life expectancy at birth, denoted in logarithmic form.	OECD Health Statistics (OECD, n.d.[84])
Dependency ratio	The population aged less than 20 years or 65 years or over as a ratio of the population aged 20‑64.	OECD Historical population data (OECD, n.d.[85])

Variable	Unit	Standard deviation	Mean	Minimum	Maximum
Share of PISA high achievers	Percentage of children	5.13	15.91	1.58	32.25
Share of PISA low achievers	Percentage of children	7.81	29.93	11.09	59.33
NEETs	Percentage of children	6.52	14.57	4.90	43.60
Share of 15‑year‑olds reporting excellent health	Percentage of children	7.72	29.08	13.16	54.19
Share of 15‑year‑olds reporting low health	Percentage of children	5.24	18.26	8.86	36.79
Share of 15‑year‑olds reporting multiple subjective health complaints	Percentage of children	13.23	25.05	7.88	67.28
ALMP	Percentage of social expenditure (excl. old age spending)	2.33	3.33	0.00	19.52
ECEC enrolment, ages 0‑2	Percentage of children	18.12	30.35	0.19	76.20
ECEC enrolment, ages 3‑5	Percentage of children	17.74	80.69	31.71	100.10
Spending on ECEC	Percentage of social expenditure (excl. old age spending)	2.26	3.96	0.00	10.66
Health services	Percentage of social expenditure (excl. old age spending)	9.37	44.67	23.33	81.03
Spending on preventive health services	Percentage of social expenditure (excl. old age spending)	0.88	1.35	0.01	7.27
Spending on non-preventive health services	Percentage of social expenditure (excl. old age spending)	8.87	43.19	22.24	78.41
Other in-kind spending	Percentage of social expenditure (excl. old age spending)	3.49	5.44	0.19	17.97
Volume of redistribution	Percentage of GDP	0.05	‑0.05	‑0.18	0.08
Average size of taxes	Percentage of average disposable income	10.80	28.42	0.02	56.55
Average size of benefits	Percentage of average disposable income	6.18	16.47	3.18	34.82
Taxes paid by bottom 30%	Percentage of total taxes paid	3.52	9.34	1.33	26.77
Benefits received by bottom 30%	Percentage of total benefits received	10.49	33.70	12.00	58.00
Cash transfers to families	Percentage of social expenditure (excl. old age spending)	4.46	8.72	0.25	25.01
Cash transfers to families during early childhood (ages 0‑5)	Percentage of social expenditure (excl. old age spending)	2.66	4.21	0.11	18.78
Cash transfers to families during middle childhood (ages 6‑11)	Percentage of social expenditure (excl. old age spending)	1.23	2.19	0.11	6.19
Cash transfers to families during adolescence (ages 12‑17)	Percentage of social expenditure (excl. old age spending)	1.26	2.12	0.10	6.19
Social expenditure	USD PPP (base year = 2015)	3 448.01	6 293.26	626.84	18 794.87
Social expenditure, logged	Log of USD PPP (base year = 2015)	0.65	8.57	6.44	9.84
Mean poverty gap before taxes and transfers	Percentage of the relative poverty line (50% of equivalised median disposable income)	8.39	61.11	35.57	80.62
Mean years of schooling	Years	1.59	11.75	5.86	14.26
Spending on primary education	USD PPP (base year = 2020)	4 100.73	9 286.35	1 173.67	23 762.94
Spending on primary education, logged	Log of USD PPP (base year = 2020)	0.52	9.02	7.07	10.08
Spending on secondary education	USD PPP (base year = 2020)	4 399.16	10 552.20	2 184.32	26 547.73
Spending on secondary education, logged	Log of USD PPP (base year = 2020)	0.46	9.17	7.69	10.19
GDP per capita	USD PPP (base year = 2020)	15 585.90	39 239.12	10 923.12	110 467.82
GDP per capita, logged	Log of USD PPP (base year = 2020)	0.38	10.51	9.30	11.61
Life expectancy at birth	Years	3.13	79.07	69.70	84.20
Life expectancy at birth, logged	Log of years	0.04	4.37	4.24	4.43
Dependency ratio	Ratio	7.13	66.03	48.15	92.67

Note: These summary statistics are based on available data for 32 OECD countries excluding Australia, Chile, Columbia, Costa Rica, Japan and Mexico from 1997 to 2021. For variables that include imputed values, the statistics represent average values across the five datasets that result from the imputation process. For further information on the imputations, see Annex 4.B.

Annex 4.B. Data imputations

Copy link to Annex 4.B. Data imputations

As for the panel analysis in Chapter 3, variables used in the regression analysis in this chapter contain incomplete time series. Again, Stineman interpolation between observed values, implemented via the R package imputeTS (Moritz and Bartz-Beielstein, 2017[86]), is applied to indicators that are unlikely to change substantially over time and where only few observations are missing. The approach was used for Mean years of schooling, Spending on primary education, Spending on secondary education in addition to the variables already mentioned in Annex 3.C, namely those that capture the share of social expenditure allocated on average to cash benefits for a child at ages 0-5, 6-11 and 12-17. Moreover, we impute observation values for the following outcome variables between survey rounds: Share of PISA high achievers, Share of PISA low achievers, Share of 15‑year‑olds reporting excellent health, Share of 15‑year‑olds reporting low health, Share of 15‑year‑olds reporting multiple subjective health complaints.

In a second step, the same multiple imputation method described in Annex 3.C was applied to the variables Mean years of schooling, Spending on primary education, Spending on secondary education, and to the variable describing the percentage of expenditure on health that is spent on preventive care, which is later use to construct the indicators Spending on preventive health services and Spending on non-preventive health services. Further, the variables ECEC enrolment 0‑2, ECEC enrolment 3‑5, Volume of redistribution, Average size of taxes, Average size of benefits, Share of taxes paid by the bottom 30%, Share of benefits paid by the bottom 30%, and Mean poverty gap before taxes and transfers contain the same imputed values as those used for the regression analysis in Chapter 3.

Annex 4.C. Regression results

Copy link to Annex 4.C. Regression results

In the regression estimations presented below, policy variables and the two main drivers – the log of total social expenditure per capita and the log of GDP per capita – are standardised using all available observations, including imputed values for variables where applicable, for the 32‑country sample over the period 1997-2021. First differenced and lagged variables are constructed after standardisation, and the regression analysis is restricted to the period 2002-2021. For specifications in which the dependent variable is available for a shorter period, the estimation sample is reduced correspondingly.

Educational outcomes

Copy link to Educational outcomes

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	1.174	0.805	0.769	1.291	1.395*	1.348	0.579	1.245*	0.941
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	0.535	0.617	0.587	0.521	0.7	0.528	0.903	0.935	0.62
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		‑0.268	‑0.14
		0.017	‑0.059
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		1.561	1.878
		0.574	0.259
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	0.168		‑0.659	0.236	0.549	0.2	‑0.996	0.198	0.006
	0.828*		0.669	0.868*	0.567	0.588	0.876	‑0.081	0.176
ECEC enrolment 0‑2 (lagged 13‑15 years) x spending on ECEC (lagged 11‑15 years)			‑0.33
ECEC enrolment 3‑5 (lagged 9‑10 years) x spending on ECEC (lagged 11‑15 years)			1.472*
Spending on health services x TS	0.771	2.154**	1.352		0.356	0.788	3.895**	1.472	1.826
Spending on health services x TS	1.189	1.638**	1.852**		0.879	0.92	1.138	1.176	‑0.086
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				‑0.376
				0.013
Spending on non-preventive health services x TS				0.607
Spending on non-preventive health services x TS				0.903
Other social spending in kind x TS	‑1.784	‑1.365	‑2.167**	‑1.668	‑2.632***	‑1.541	‑2.923**	‑3.288***	‑2.644
Other social spending in kind x TS	1.13	1.222	1.282*	1.035	0.77	0.503	1.3	1.986*	2.666
Volume of redistribution x TS	0.532	0.77	0.396	0.677			0.75	0.729	0.618
Volume of redistribution x TS	0.759*	0.542	0.849*	0.778			1.188*	0.606	0.58
Size of benefits					1.31*	‑0.356
Size of taxes					‑1.257	‑0.614
Share of taxes paid by the bottom 30% x TS	‑0.199	‑0.271	‑0.913	‑0.212		‑0.215	‑0.903	‑1.07	‑0.454
Share of taxes paid by the bottom 30% x TS	‑0.387	‑0.316	‑0.326	‑0.445			‑0.645	‑0.641	‑0.452
Share of benefits received by the bottom 30% x TS	‑3.128*	‑1.911	‑1.311	‑3.167*		‑5.055***	‑0.369	‑3.469**	‑5.263***
Share of benefits received by the bottom 30% x TS	‑0.489	0.263	0.116	‑0.317			‑0.406	‑0.034	‑0.337
Size of benefits x share of benefits received by the bottom 30%						‑0.178
Size of taxes x share of taxes received by the bottom 30%						‑0.361
Cash transfers to families x TS	0.771	1.528*	1.673**	0.795	0.507	1.083*
Cash transfers to families x TS	0.569	0.456	0.482	0.409	0.976*	0.566
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							0.974
							1.347
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								1.105
								1.969*
In-cash spending on families, ages 12‑17 x TS									1.189
In-cash spending on families, ages 12‑17 x TS									‑0.073
Total social expenditure per capita (logged)	1.953	2.101	3.229	2.118	1.212	2.852	9.14**	1.139	0.236
Primary education spending (logged, lagged 7‑9 years)	0.197	0.405	2.421	0.372	1.851	0.096	2.233	2.104	1.384
Secondary education spending (logged, lagged 2‑4 years)	‑3.281	‑0.978	‑1.498	‑3.712	‑2.685	‑1.133	2.875	‑5.421	‑3.981
Mean poverty gap	‑0.261***	‑0.171	‑0.209	‑0.257***	‑0.329***	‑0.24**	‑0.158	‑0.249***	‑0.374***
Mean years of schooling	‑0.805	‑0.76	‑0.442	‑0.75	‑1.052	‑1.085	0.236	‑0.69	‑0.346
Adjusted R²	0.367	0.442	0.526	0.367	0.316	0.386	0.598	0.511	0.464
Number of countries	32	32	32	32	32	32	27	28	28
Number of observations	336	275	275	335	335	335	182	289	239
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

Note: Coefficients in bold are the ones used for the value for money calculations. “x” signifies an interaction between two variables. “x TS” means the interaction between the policy variable in the line above and the log of per capita total social spending. The log of total social spending per capita is standardised along with the policy variables to facilitate the interpretation of results.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	2.223***	2.55***	2.777***	1.801***	2.771***	2.156***	3.012***	1.705***	1.333
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.351**	‑1.726***	‑1.852**	‑0.951	‑1.453	‑1.541***	‑2.132***	‑0.418	‑0.208
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		‑0.718	‑0.752
		‑0.616	‑0.893
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		1.458***	2.518***
		0.719*	0.702*
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	‑0.807		‑1.674**	‑0.845	‑0.84	‑0.496	‑1.287	‑0.418	‑1.179
	‑0.565		‑0.202	‑0.507	‑0.459	‑0.55	0.372	0.696	1.445
ECEC enrolment 0‑2 (lagged 13‑15 years) x spending on ECEC (lagged 11‑15 years)			‑0.509
ECEC enrolment 3‑5 (lagged 9‑10 years) x spending on ECEC (lagged 11‑15 years)			0.973
Spending on health services x TS	0.086	0.662	0.415		1.542	0.42	‑0.099	‑1.51	‑2.175
Spending on health services x TS	‑1.201	‑0.019	‑0.38		‑2.461*	‑0.871	‑1.48	1.491	1.778
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				0.767
				‑2.275*
Spending on non-preventive health services x TS				‑0.941
Spending on non-preventive health services x TS				0.319
Other social spending in kind x TS	‑1.231	‑1.894**	‑1.494	‑1.751	0.31	‑0.913	‑0.64	‑0.079	‑0.655
Other social spending in kind x TS	‑1.064	‑0.51	‑0.183	‑0.444	‑2.081*	‑1.208	‑0.812	‑2.935***	‑2.236**
Volume of redistribution x TS	0.021	0.44	0.683	‑0.265			‑1.195	‑0.563	‑0.731
Volume of redistribution x TS	1.634**	1.134	0.64	2.112***			2.614**	1.74**	1.901**
Size of benefits					‑2.733***	1.628
Size of taxes					‑0.549	‑1.249**
Share of taxes paid by the bottom 30% x TS	0.37	‑0.104	‑0.069	0.107		‑0.413	‑0.665	‑0.573	‑0.125
Share of taxes paid by the bottom 30% x TS	‑1.253*	‑0.726	‑1.062*	‑0.975			‑0.555	0.856	0.423
Share of benefits received by the bottom 30% x TS	4.457***	3.734***	4.123***	4.423***		5.089***	3.859***	3.606***	4.041***
Share of benefits received by the bottom 30% x TS	0.339	2.481*	1.65	0.642			‑0.196	0.81	0.09
Size of benefits x share of benefits received by the bottom 30%						‑0.877*
Size of taxes x share of taxes received by the bottom 30%						‑0.89
Cash transfers to families x TS	‑0.367	0.985	0.686	‑0.617	‑0.224	‑0.52
Cash transfers to families x TS	‑0.066	0.892	0.344	0.223	0.312	0.581
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							‑1.301***
							‑1.33***
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								‑1.638**
								2.654***
In-cash spending on families, ages 12‑17 x TS									‑1.441*
In-cash spending on families, ages 12‑17 x TS									2.339**
Total social expenditure per capita (logged)	‑3.345***	‑2.585**	‑3.186***	‑3.424***	‑0.777	‑2.937**	‑4.573***	‑2.828**	‑1.558
Primary education spending (logged, lagged 7‑9 years)	2.254	3.218***	3.969***	2.007	2.233*	2.016	2.863***	0.969	0.691
Secondary education spending (logged, lagged 2‑4 years)	4.498	4.83	3.521	3.64	1.005	4.594	5.756**	2.699	2.686
Mean poverty gap	‑0.221**	‑0.309***	‑0.31***	‑0.237**	0.176**	‑0.245*	‑0.219***	‑0.022	‑0.06
Mean years of schooling	0.873**	1.063**	1.181*	1.02**	1.15**	0.784**	1.268*	1.149**	0.851
Adjusted R²	0.708	0.772	0.784	0.723	0.556	0.691	0.832	0.775	0.76
Number of countries	32	32	32	32	32	32	27	28	28
Number of observations	336	275	275	335	335	335	182	289	239
Country fixed effects	No	No	No	No	No	No	No	No	No
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

Note: “x” signifies an interaction between two variables. “x TS” means the interaction between the policy variable in the line above and the log of per capita total social spending. The log of total social spending per capita is standardised along with the policy variables to facilitate the interpretation of results.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	0.194	0.18	0.074	0.37	0.294	0.181	0.608	0.329	0.112
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	0.381	0.478	0.323	0.285	0.335	0.266	0.343	0.544	0.151
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		0.766	0.913
		0.148	0.238
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		0.66	0.953
		0.08	0.013
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	0.156		‑0.71	0.188	0.349	0.095	‑0.077	0.686	0.292
	0.352		‑0.081	0.307	0.227	0.22	‑1.115*	‑0.667	‑0.222
ECEC enrolment 0‑2 (lagged 13‑15 years) x spending on ECEC (lagged 11‑15 years)			‑0.026
ECEC enrolment 3‑5 (lagged 9‑10 years) x spending on ECEC (lagged 11‑15 years)			0.754
Spending on health services x TS	‑0.286	0.441	0.097		‑0.38	‑0.111	1.685	0.374	0.588
Spending on health services x TS	0.73*	0.789	0.76		0.593	0.585	‑0.114	0.653	0.277
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				‑0.205
				0.535
Spending on non-preventive health services x TS				‑0.108
Spending on non-preventive health services x TS				0.606
Other social spending in kind x TS	‑1.185	‑0.718	‑1.047	‑1.287	‑1.498	‑0.85	‑2.263**	‑2.204**	‑2.206
Other social spending in kind x TS	0.126	0.265	0.251	0.233	‑0.053	‑0.272	0.974	0.698	1.146
Volume of redistribution x TS	0.502	0.657	0.494	0.722			0.858	0.579	0.105
Volume of redistribution x TS	0.793**	0.549	0.657*	0.697*			0.458	0.799**	1.165*
Size of benefits					0.687	‑0.487
Size of taxes					‑0.336	0.067
Share of taxes paid by the bottom 30% x TS	0.013	‑0.035	‑0.139	0.045		‑0.16	0.04	‑0.389	‑0.251
Share of taxes paid by the bottom 30% x TS	‑0.162	‑0.072	0.009	‑0.242			‑0.23	‑0.401	‑0.411
Share of benefits received by the bottom 30% x TS	‑1.592	‑0.813	‑0.584	‑1.312		‑3.247***	‑1.73	‑2.105***	‑2.833***
Share of benefits received by the bottom 30% x TS	0.015	0.104	0.045	0.123			‑0.612	0.526	0.04
Size of benefits x share of benefits received by the bottom 30%						‑0.249
Size of taxes x share of taxes received by the bottom 30%						‑0.196
Cash transfers to families x TS	0.208	0.569	0.719	0.267	0.186	0.581
Cash transfers to families x TS	0.151	0.187	0.202	0.051	0.466	0.233
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							0.961
							1.496*
Cash transfers to families, ages 6‑11 (lagged 4‑6years) x TS (lagged 4‑6 years)								0.508
								1.145
In-cash spending on families, ages 12‑17 x TS									0.067
In-cash spending on families, ages 12‑17 x TS									‑0.094
Total social expenditure per capita (logged)	‑0.952	0.361	0.44	‑0.61	‑1.048	‑0.024	0.625	‑2.227	‑2.037
Primary education spending (logged, lagged 7‑9 years)	0.431	0.604	1.678	0.148	1.108	‑0.149	1.296	0.691	‑0.821
Secondary education spending (logged, lagged 2‑4 years)	‑0.162	0.858	1.552	‑0.555	0.212	1.513	1.148	‑2.895	‑1.89
Mean poverty gap	‑0.259***	‑0.223**	‑0.247**	‑0.252***	‑0.298***	‑0.241**	‑0.132	‑0.255***	‑0.338***
Mean years of schooling	‑0.999	‑1.005	‑0.835	‑0.917	‑1.149	‑1.284*	0.581	‑1.133	‑1.048
Adjusted R²	0.358	0.461	0.486	0.352	0.307	0.352	0.638	0.5	0.431
Number of countries	32	32	32	32	32	32	27	28	28
Number of observations	333	272	272	332	332	332	181	286	236
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	1.068	0.647	0.533	1.254	1.344*	1.257	0.568	1.213	0.73
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	0.635	0.669	0.556	0.625	0.821	0.606	1.183	1.138*	0.804
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		0.307	0.607
		0.324	0.24
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		1.678	2.257*
		0.576	0.3
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	0.472		‑0.863	0.615	0.944	0.55	‑0.661	0.658	0.31
	1.049**		0.256	1.092**	0.804	0.82	1.115	0.012	0.351
ECEC enrolment 0‑2 (lagged 13‑15 years) x spending on ECEC (lagged 11‑15 years)			0.027
ECEC enrolment 3‑5 (lagged 9‑10 years) x spending on ECEC (lagged 11‑15 years)			1.938**
Spending on health services x TS	1.285	3.111***	2.116**		1.115	1.637	5.943***	2.255**	2.477*
Spending on health services x TS	1.092	1.456**	1.63*		0.604	0.664	0.97	1.045	‑0.674
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				‑0.309
				0.297
Spending on non-preventive health services x TS				1.118
Spending on non-preventive health services x TS				0.715
Other social spending in kind x TS	‑1.435	‑0.671	‑1.492	‑1.28	‑2.478**	‑1.061	‑2.409**	‑2.95**	‑2.462
Other social spending in kind x TS	0.666	0.706	0.656	0.552	0.194	‑0.027	1.132	1.374	1.885
Volume of redistribution x TS	0.688	0.796	0.485	0.894			‑0.045	0.692	0.846
Volume of redistribution x TS	0.906**	0.618	0.969**	0.88*			1.201*	0.82*	1.046
Size of benefits					2.005**	‑0.013
Size of taxes					‑1.147	‑0.404
Share of taxes paid by the bottom 30% x TS	0.192	0.182	‑0.361	0.199		0.086	‑1.263	‑0.955	‑0.071
Share of taxes paid by the bottom 30% x TS	‑0.386	‑0.272	‑0.183	‑0.49			‑0.639	‑0.741*	‑0.513
Share of benefits received by the bottom 30% x TS	‑4.181**	‑2.722	‑1.998	‑4.141**		‑6.154***	‑0.726	‑4.413***	‑6.028***
Share of benefits received by the bottom 30% x TS	‑0.454	0.227	0.063	‑0.129			‑0.407	0.124	‑0.391
Size of benefits x share of benefits received by the bottom 30%						‑0.189
Size of taxes x share of taxes received by the bottom 30%						‑0.358
Cash transfers to families x TS	0.933	1.648*	1.986**	1.044	0.645	1.299**
Cash transfers to families x TS	0.632	0.573	0.62	0.469	1.155**	0.665
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							0.824
							1.18
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								1.183
								2.486**
In-cash spending on families, ages 12‑17 x TS									1.05
In-cash spending on families, ages 12‑17 x TS									0.042
Total social expenditure per capita (logged)	2.374	3.573	4.107	2.875	1.504	3.694	11.267***	1.873	0.016
Primary education spending (logged, lagged 7‑9 years)	‑0.447	‑0.009	2.697	‑0.435	1.565	‑0.844	0.725	0.605	‑0.346
Secondary education spending (logged, lagged 2‑4 years)	‑3.605	‑1.086	‑1.254	‑4.45	‑3.439	‑1.664	2.476	‑6.019	‑3.447
Mean poverty gap	‑0.272***	‑0.181	‑0.22	‑0.264***	‑0.385***	‑0.284**	‑0.193**	‑0.251***	‑0.389***
Mean years of schooling	‑1.14	‑1.266	‑0.867	‑1.135	‑1.603	‑1.592	0.006	‑1.106	‑0.748
Adjusted R²	0.403	0.461	0.552	0.409	0.327	0.415	0.595	0.553	0.509
Number of countries	32	32	32	32	32	32	27	28	28
Number of observations	333	272	272	332	332	332	181	286	236
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	1.994	1.355	1.66	1.815	2.291	2.375	‑0.974	1.621	1.639
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	0.464	0.697	0.79	0.596	0.863	0.62	0.966	0.984	0.711
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		‑1.417	‑1.248
		‑0.371	‑0.502
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		2.774	2.862
		1.023	0.141
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	‑0.219		0.215	‑0.279	0.187	‑0.248	‑3.142**	‑1.265	‑0.629
	0.859		1.941	1.025	0.418	0.484	2.369	0.07	‑0.222
ECEC enrolment 0‑2 (lagged 13‑15 years) x spending on ECEC (lagged 11‑15 years)			‑1.102
ECEC enrolment 3‑5 (lagged 9‑10 years) x spending on ECEC (lagged 11‑15 years)			1.666
Spending on health services x TS	0.776	2.315	1.658		‑0.371	0.072	4.902	1.583	1.699
Spending on health services x TS	2.023	3.036**	3.486**		1.775	1.82	2.253	2.158	0.508
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				‑0.639
				‑1.453
Spending on non-preventive health services x TS				0.091
Spending on non-preventive health services x TS				1.752
Other social spending in kind x TS	‑3.449*	‑3.469*	‑4.599**	‑3.103	‑4.529***	‑3.558**	‑4.838**	‑5.392***	‑3.27
Other social spending in kind x TS	3.33**	3.293*	3.699***	2.98**	2.869**	2.412*	2.823	5.068***	5.767**
Volume of redistribution x TS	0.03	0.965	0.022	‑0.138			1.807	0.271	0.088
Volume of redistribution x TS	0.313	0.559	0.863	0.632			1.696	‑0.693	‑1.377
Size of benefits					0.718	‑0.869
Size of taxes					‑2.519*	‑1.657
Share of taxes paid by the bottom 30% x TS	‑1.192	‑1.292	‑2.9	‑1.33		‑0.831	‑1.374	‑2.137	‑1.027
Share of taxes paid by the bottom 30% x TS	‑0.681	‑0.635	‑1.021	‑0.584			‑1.032	‑0.737	‑0.291
Share of benefits received by the bottom 30% x TS	‑3.127	‑1.527	‑0.993	‑3.874		‑5.217*	‑0.544	‑4.231	‑7.097**
Share of benefits received by the bottom 30% x TS	‑1.349	0.169	0.068	‑1.43			‑0.954	‑0.832	‑0.873
Size of benefits x share of benefits received by the bottom 30%						‑0.192
Size of taxes x share of taxes received by the bottom 30%						‑0.7
Cash transfers to families x TS	1.011	1.967*	1.927*	0.773	0.512	1.193
Cash transfers to families x TS	0.622	0.368	0.258	0.415	0.89	0.404
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							2.854**
							2.598*
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								3.688**
								2.543
In-cash spending on families, ages 12‑17 x TS									2.978**
In-cash spending on families, ages 12‑17 x TS									‑0.071
Total social expenditure per capita (logged)	4.127	2.093	6.094	3.224	2.645	4.103	15.146**	4.169	3.456
Primary education spending (logged, lagged 7‑9 years)	0.963	1.937	3.564	2.262	3.391	2.18	4.168	6.334	3.644
Secondary education spending (logged, lagged 2‑4 years)	‑6.223	‑1.733	‑4.77	‑5.711	‑4.811	‑3.278	8.6	‑7.165	‑6.952
Mean poverty gap	‑0.278*	‑0.131	‑0.161	‑0.287	‑0.289	‑0.19	‑0.074	‑0.289*	‑0.438**
Mean years of schooling	‑0.479	‑0.097	‑0.16	‑0.404	‑0.277	‑0.36	0.638	‑0.027	0.037
Adjusted R²	0.251	0.333	0.409	0.265	0.256	0.291	0.566	0.385	0.396
Number of countries	32	32	32	32	32	32	27	28	28
Number of observations	333	272	272	332	332	332	181	286	236
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.165	‑1.477	‑1.463	‑1.261	‑1.443	‑1.476	‑3.207***	‑1.832**	‑0.353
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.502	‑2.423**	‑1.827*	‑1.581	‑1.775*	‑1.658	‑1.876*	‑2.388***	‑2.043**
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		‑0.036	‑0.103
		2.286*	1.837
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		‑1.724	‑3.484**
		‑0.487	0.206
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	1.353		2.482*	1.267	1.062	1.032	1.161	0.351	1.27
	‑0.226		0.073	‑0.272	‑0.046	‑0.249	0.803	0.574	0.105
ECEC enrolment 0‑2 (lagged 13‑15 years) x spending on ECEC (lagged 11‑15 years)			1.21
ECEC enrolment 3‑5 (lagged 9‑10 years) x spending on ECEC (lagged 11‑15 years)			‑3.209***
Spending on health services x TS	‑1.278	‑1.081	‑0.716		‑1.705	‑1.827	‑1.649	‑2.165	‑3.348**
Spending on health services x TS	0.749	0.512	1.177		1.47*	1.03	0.412	0.623	3.634***
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				‑1.086
				‑0.888
Spending on non-preventive health services x TS				‑1.751
Spending on non-preventive health services x TS				1.206*
Other social spending in kind x TS	2.446	3.449***	4.788***	2.272	2.993**	2.344	5.168***	3.301***	1.138
Other social spending in kind x TS	‑1.524	‑1.698	‑2.498**	‑1.4	‑0.984	‑1.025	‑0.475	‑1.282	‑0.688
Volume of redistribution x TS	0.563	‑0.29	0.136	0.621			‑3.569***	‑0.19	0.389
Volume of redistribution x TS	‑0.396	‑0.397	‑0.519	‑0.452			‑0.601	‑0.492	0.715
Size of benefits					‑1.189	‑0.967
Size of taxes					0.261	‑0.152
Share of taxes paid by the bottom 30% x TS	0.011	0.992	1.613	0.193		‑0.043	0.639	1.08	0.276
Share of taxes paid by the bottom 30% x TS	0.375	0.369	0.31	0.415			0.442	0.874	0.961
Share of benefits received by the bottom 30% x TS	2.408	1.009	0.412	2.589		2.583	‑0.307	3.243*	5.778**
Share of benefits received by the bottom 30% x TS	0.988	‑0.008	0.146	0.347			‑0.196	‑0.703	‑0.242
Size of benefits x share of benefits received by the bottom 30%						‑0.688
Size of taxes x share of taxes received by the bottom 30%						0.312
Cash transfers to families x TS	0.552	0.267	‑0.004	0.135	0.713	0.685
Cash transfers to families x TS	‑1.33*	‑0.085	‑0.02	‑1.376*	‑1.641***	‑1.31**
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							‑1.477
							‑3.315***
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								0.7
								‑3.692***
In-cash spending on families, ages 12‑17 x TS									‑1.664
In-cash spending on families, ages 12‑17 x TS									‑0.984
Total social expenditure per capita (logged)	1.715	4.104	1.724	0.476	1.64	0.168	‑0.658	1.459	4.858
Primary education spending (logged, lagged 7‑9 years)	‑1.629	‑3.076	‑4.881	‑1.659	‑2.975	‑2.089	‑2.336	‑2.109	‑3.506
Secondary education spending (logged, lagged 2‑4 years)	2.657	4.531	3.484	3.358	2.665	2	1.842	11.601**	7.556
Mean poverty gap	0.265**	0.116	0.129	0.244*	0.357**	0.34**	0.045	0.237**	0.361**
Mean years of schooling	‑1.842	‑3.169**	‑3.898***	‑1.681	‑1.599	‑1.697	‑5.092***	‑1.068	‑1.287
Adjusted R²	0.269	0.357	0.429	0.274	0.267	0.286	0.654	0.449	0.455
Number of countries	32	32	32	32	32	32	27	28	28
Number of observations	336	275	275	335	335	335	182	289	239
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑6.295***	‑6.505***	‑6.826***	‑5.304***	‑6.666***	‑5.829***	‑5.576***	‑4.924***	‑4.716***
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	3.787**	4.147***	4.036***	2.335**	2.949*	3.659***	3.518**	2.9**	3.269**
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		2.207	2.475
		0.724	1.26
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		‑0.869	‑1.778
		‑1.284*	‑1.091
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	1.18		1.397	0.595	1.662	1.885*	3.437***	1.769	2.853**
	‑0.295		‑1.323	‑0.55	‑1.406	‑0.623	‑3.187**	‑3.524**	‑4.269***
ECEC enrolment 0‑2 (lagged 13‑15 years) x spending on ECEC (lagged 11‑15 years)			1.364
ECEC enrolment 3‑5 (lagged 9‑10 years) x spending on ECEC (lagged 11‑15 years)			‑0.843
Spending on health services x TS	‑2.629	‑3.555	‑3.516		‑3.983*	‑2.284	3.187*	2.646*	2.822
Spending on health services x TS	3.803	3.845*	4.087		4.46**	3.016	‑1.391	‑2.208	‑1.423
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				‑2.838***
				2.639
Spending on non-preventive health services x TS				0.953
Spending on non-preventive health services x TS				0.159
Other social spending in kind x TS	0.906	1.243	1.073	2.751**	‑0.884	0.455	0.649	0.538	0.616
Other social spending in kind x TS	2.041	1.286	1.04	0.673	3.181*	1.37	1.116	2.633**	2.264
Volume of redistribution x TS	‑0.668	‑1.226	‑1.151	‑0.736			‑1.354	‑1.993	‑2.014
Volume of redistribution x TS	0.223	0.622	1.153	0.095			1.218	2.64*	2.559*
Size of benefits					2.273	‑4.983***
Size of taxes					0.599	1.095
Share of taxes paid by the bottom 30% x TS	1.765**	2.535***	2.464***	2.175***		2.585***	2.234**	2.999***	2.973***
Share of taxes paid by the bottom 30% x TS	2.003**	0.948	1.086	1.303*			0.334	‑1.025	‑0.78
Share of benefits received by the bottom 30% x TS	‑3.61**	‑3.264*	‑3.343	‑3.445**		‑7.703***	‑6.306***	‑4.539***	‑4.089***
Share of benefits received by the bottom 30% x TS	‑2.395	‑3.921*	‑2.947	‑2.663*			2.145	‑0.373	‑0.837
Size of benefits x share of benefits received by the bottom 30%						0.65
Size of taxes x share of taxes received by the bottom 30%						1.426
Cash transfers to families x TS	‑1.923**	‑2.779***	‑2.699***	‑1.41	‑1.152	‑0.212
Cash transfers to families x TS	1.47	1.62	1.806	0.314	1.74	1.662
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							‑0.6
							0.881
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								‑0.065
								‑2.026
In-cash spending on families, ages 12‑17 x TS									‑0.251
In-cash spending on families, ages 12‑17 x TS									‑0.986
Total social expenditure per capita (logged)	2.635*	1.334	0.599	2.139	‑0.609	3.937**	1.645	0.265	‑1.715
Primary education spending (logged, lagged 7‑9 years)	‑6.067***	‑7.113***	‑7.316***	‑5.726***	‑4.878**	‑5.042***	‑4.656**	‑3.131*	‑3.421**
Secondary education spending (logged, lagged 2‑4 years)	‑2.421	‑1.798	0.185	‑1.861	1.973	‑3.038	‑1.182	‑0.077	0.45
Mean poverty gap	0.248**	0.319**	0.304**	0.253**	‑0.206	0.513***	0.218**	0.064	0.076
Mean years of schooling	‑0.915	‑0.736	‑0.961	‑1.167**	‑1.028	‑0.943*	‑0.048	‑0.826	‑1.126
Adjusted R²	0.693	0.752	0.756	0.739	0.525	0.701	0.786	0.759	0.764
Number of countries	32	32	32	32	32	32	27	28	28
Number of observations	336	275	275	335	335	335	182	289	239
Country fixed effects	No	No	No	No	No	No	No	No	No
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	0.729	‑0.279	0.066	0.305	0.416	0.683	‑2.655*	‑0.641	1.576
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.931*	‑3.537***	‑2.604*	‑1.785	‑2.166*	‑2.121*	‑1.097	‑2.728***	‑2.239*
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		‑1.085	‑1.287
		3.602*	2.992
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		‑1.695	‑3.303
		0.226	0.933
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	3.739*		4.293**	3.573*	3.112	3.226	2.409**	1.173	3.125*
	0.403		0.44	0.52	0.678	0.335	3.457***	2.036**	1.05
ECEC enrolment 0‑2 (lagged 13‑15 years) x spending on ECEC (lagged 11‑15 years)			0.858
ECEC enrolment 3‑5 (lagged 9‑10 years) x spending on ECEC (lagged 11‑15 years)			‑2.676*
Spending on health services x TS	‑1.752	‑0.826	‑1.066		‑1.083	‑1.698	‑1.586	‑2.468	‑4.633**
Spending on health services x TS	2.017*	0.981	2.079		2.358*	1.777	1.856	1.204	4.582***
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				‑0.872
				‑1.95
Spending on non-preventive health services x TS				‑2.624
Spending on non-preventive health services x TS				2.456**
Other social spending in kind x TS	1.377	1.982	3.142	1.608	2.643	1.157	5.583***	2.626	1.152
Other social spending in kind x TS	‑0.721	‑1.021	‑2.07	‑0.918	‑0.319	‑0.536	‑1.159	‑0.428	‑0.314
Volume of redistribution x TS	1.055	‑0.498	0.675	0.751			‑6.209***	0.011	1.619
Volume of redistribution x TS	‑1.238	‑0.93	‑1.071	‑0.951			‑0.464	‑1.603	‑1.012
Size of benefits					‑0.664	0.789
Size of taxes					0.019	‑0.687
Share of taxes paid by the bottom 30% x TS	0.05	1.211	1.808	0.125		0.091	0.248	0.995	0.129
Share of taxes paid by the bottom 30% x TS	0.508	0.536	0.411	0.722			0.565	1.067	0.789
Share of benefits received by the bottom 30% x TS	4.541	2.196	2.721	3.882		5.876*	‑1.432	5.762**	8.824***
Share of benefits received by the bottom 30% x TS	0.211	‑1.122	‑1.097	‑0.434			0.46	‑2.357	‑2.008
Size of benefits x share of benefits received by the bottom 30%						‑1.003
Size of taxes x share of taxes received by the bottom 30%						‑0.202
Cash transfers to families x TS	0.886	0.546	0.298	0.296	1.151	0.857
Cash transfers to families x TS	‑1.46	‑0.161	‑0.025	‑1.49	‑2.166**	‑1.733*
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							‑1.27
							‑5.1***
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								2.38
								‑5.463***
In-cash spending on families, ages 12‑17 x TS									‑0.137
In-cash spending on families, ages 12‑17 x TS									‑0.915
Total social expenditure per capita (logged)	3.41	7.213	2.626	1.679	4.138	1.025	5.644	3.24	5.087
Primary education spending (logged, lagged 7‑9 years)	‑2.032	‑2.218	‑3.907	‑0.983	‑3.726	‑1.399	‑1.679	1.401	0.511
Secondary education spending (logged, lagged 2‑4 years)	‑1.261	6.571	4.487	0.189	‑1.898	‑3.814	2.386	13.813**	7.347
Mean poverty gap	0.427**	0.198	0.225	0.399**	0.48*	0.43*	0.028	0.379**	0.572***
Mean years of schooling	‑1.413	‑2.767	‑3.508	‑1.413	‑1.108	‑1.088	‑6.28***	0.282	‑0.02
Adjusted R²	0.275	0.321	0.372	0.284	0.263	0.307	0.656	0.483	0.462
Number of countries	32	32	32	32	32	32	27	28	28
Number of observations	333	272	272	332	332	332	181	286	236
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.772	‑2.012*	‑2.063**	‑1.779	‑1.995	‑2.109*	‑3.795***	‑2.315***	‑0.899
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.303	‑1.841*	‑1.45	‑1.454	‑1.7	‑1.481	‑1.823*	‑2.19**	‑1.895**
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		0.273	0.043
		2.09*	1.69
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		‑1.947	‑3.487***
		‑0.411	0.16
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	0.567		1.651	0.476	0.319	0.232	0.575	‑0.238	0.532
	‑0.789		‑0.425	‑0.87	‑0.671	‑0.822	‑0.461	‑0.124	‑0.354
ECEC enrolment 0‑2 (lagged 13‑15 years) x spending on ECEC (lagged 11‑15 years)			1.387
ECEC enrolment 3‑5 (lagged 9‑10 years) x spending on ECEC (lagged 11‑15 years)			‑3.444***
Spending on health services x TS	‑0.642	‑0.769	0.21		‑1.342	‑1.291	‑1.593	‑1.969	‑2.715
Spending on health services x TS	0.348	0.57	0.74		1.234	0.752	‑0.322	0.484	3.504**
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				‑1.527
				‑0.958
Spending on non-preventive health services x TS				‑1.128
Spending on non-preventive health services x TS				0.866
Other social spending in kind x TS	2.981*	4.043***	5.538***	2.703	3.522**	2.772*	5.469***	3.857***	1.238
Other social spending in kind x TS	‑1.37	‑1.958	‑2.429**	‑1.232	‑0.425	‑0.595	‑0.368	‑1.178	‑0.143
Volume of redistribution x TS	0.674	‑0.094	0.238	0.827			‑2.932***	0.054	0.535
Volume of redistribution x TS	‑0.251	‑0.175	‑0.411	‑0.347			‑0.572	‑0.346	0.986
Size of benefits					‑1.171	‑1.336
Size of taxes					0.122	‑0.195
Share of taxes paid by the bottom 30% x TS	‑0.736	‑0.141	0.619	‑0.512		‑0.88	‑0.057	0.341	‑0.563
Share of taxes paid by the bottom 30% x TS	0.312	0.241	0.218	0.305			0.2	0.806	1.05
Share of benefits received by the bottom 30% x TS	2.334	1.693	0.46	2.707		2.178	‑0.09	3.315	6.049**
Share of benefits received by the bottom 30% x TS	1.47	0.239	0.589	0.809			‑0.108	‑0.239	0.296
Size of benefits x share of benefits received by the bottom 30%						‑0.795
Size of taxes x share of taxes received by the bottom 30%						0.516
Cash transfers to families x TS	0.493	0.515	0.137	0.06	0.47	0.63
Cash transfers to families x TS	‑1.31*	0.291	0.326	‑1.435**	‑1.604**	‑1.229**
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							‑1.329
							‑3.25***
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								0.78
								‑3.379**
In-cash spending on families, ages 12‑17 x TS									‑2.214*
In-cash spending on families, ages 12‑17 x TS									‑1.177
Total social expenditure per capita (logged)	1.376	2.534	1.286	0.154	1.498	‑0.087	‑3.406	0.455	5.271
Primary education spending (logged, lagged 7‑9 years)	‑1.123	‑1.321	‑4.121	‑1.291	‑2.949	‑1.89	‑0.389	‑0.837	‑2.656
Secondary education spending (logged, lagged 2‑4 years)	3.174	4.413	3.532	3.753	3.369	3.057	3.427	10.747**	7.606
Mean poverty gap	0.301**	0.112	0.155	0.278**	0.386**	0.38**	0.052	0.289**	0.431***
Mean years of schooling	‑2.054*	‑3.462**	‑4.039***	‑1.735	‑1.7	‑2.024	‑5.045***	‑1.123	‑1.501
Adjusted R²	0.314	0.41	0.498	0.327	0.298	0.324	0.671	0.478	0.506
Number of countries	32	32	32	32	32	32	27	28	28
Number of observations	333	272	272	332	332	332	181	286	236
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.545	‑1.52*	‑1.722*	‑1.218	‑1.682	‑1.819	‑1.546**	‑1.608*	‑1.131
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑0.826	‑1.56*	‑0.949	‑1.111	‑1.005	‑0.914	‑1.903**	‑1.526**	‑1.231*
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		‑0.271	‑0.386
		1.306*	0.989
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		‑2.689**	‑4.659***
		‑1.163	‑0.283
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	0.392		1.2	0.468	0.264	0.437	0.214	0.301	0.302
	0.138		0.816	‑0.035	0.248	0.153	‑0.199	0.452	0.786
ECEC enrolment 0‑2 (lagged 13‑15 years) x spending on ECEC (lagged 11‑15 years)			0.564
ECEC enrolment 3‑5 (lagged 9‑10 years) x spending on ECEC (lagged 11‑15 years)			‑2.653***
Spending on health services x TS	‑1.817*	‑1.559	‑1.917		‑1.79	‑2.043*	‑3.561**	‑1.835*	‑2.332**
Spending on health services x TS	0.483	‑0.221	0.672		0.786	0.835	0.422	0.407	2.296***
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				‑0.82
				1.063
Spending on non-preventive health services x TS				‑1.758*
Spending on non-preventive health services x TS				0.716
Other social spending in kind x TS	1.047	2.227**	2.881***	0.53	1.27	1.287	2.097***	1.411	‑1.028
Other social spending in kind x TS	‑2.335**	‑1.458	‑2.377**	‑1.777*	‑2.474***	‑2.182**	0.026	‑2.165*	‑1.117
Volume of redistribution x TS	0.298	‑0.414	‑0.285	0.842			‑2.719**	‑0.31	‑0.241
Volume of redistribution x TS	0.396	0.109	0.264	‑0.054			0.219	0.681	1.916*
Size of benefits					‑0.664	‑0.252
Size of taxes					0.617	0.229
Share of taxes paid by the bottom 30% x TS	1.291	2.427*	2.753**	1.589		1.161	1.449*	1.941*	1.227
Share of taxes paid by the bottom 30% x TS	0.347	0.288	0.288	0.212			0.517	0.662	0.739
Share of benefits received by the bottom 30% x TS	1.013	‑1.337	‑1.428	2.078		1.251	1.181	1.486	2.863
Share of benefits received by the bottom 30% x TS	0.62	0.345	0.148	0.221			‑0.714	‑0.074	0.025
Size of benefits x share of benefits received by the bottom 30%						0.139
Size of taxes x share of taxes received by the bottom 30%						0.313
Cash transfers to families x TS	0.193	0.033	‑0.205	0.081	0.65	0.326
Cash transfers to families x TS	‑0.796	‑0.25	‑0.128	‑0.813	‑0.741	‑0.591
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							‑1.136
							‑1.504
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								‑1.527
								‑1.703
In-cash spending on families, ages 12‑17 x TS									‑2.471***
In-cash spending on families, ages 12‑17 x TS									‑0.741
Total social expenditure per capita (logged)	‑0.705	2.655	0.347	‑0.776	‑1.183	‑1.243	‑3.776	0.152	3.593
Primary education spending (logged, lagged 7‑9 years)	‑2.214	‑6.046*	‑5.86	‑3.408	‑2.217	‑2.58	‑3.475	‑5.798	‑5.429
Secondary education spending (logged, lagged 2‑4 years)	2.836	‑1.756	‑1.429	2.35	2.974	2.785	‑3.801	6.834	7.074
Mean poverty gap	0.011	0	‑0.066	0.009	0.072	0.041	0.05	0.014	0.029
Mean years of schooling	‑1.302	‑2.494**	‑2.891***	‑1.136	‑1.292	‑1.24	‑2.254*	‑1.309	‑0.968
Adjusted R²	0.199	0.312	0.395	0.217	0.183	0.199	0.649	0.284	0.341
Number of countries	32	32	32	32	32	32	27	28	28
Number of observations	333	272	272	332	332	332	181	286	236
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	0.278	0.904	0.856	0.248	0.263	0.271	0.248	0.177
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑0.255	0.516	0.321	‑0.352	‑0.362	‑0.357	‑0.465	‑0.074
ECEC enrolment 3‑5 (lagged 16‑18 years) x TS (lagged 16‑18 years)			0.368
			0.375
Spending on ECEC (lagged 16‑18 years) x TS (lagged 16‑18 years)		1.153
		0.196
Spending on health services x TS	‑1.943**	‑1.104	‑0.487		‑2.218**	‑2.208**	‑2.234***	‑1.392
Spending on health services x TS	0.846*	0.787	0.856		1.015**	0.918*	1.234**	0.318
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				0.467
				0.495
Spending on non-preventive health services x TS				‑2.346***
Spending on non-preventive health services x TS				0.823
Other social spending in kind x TS	‑2.158**	‑1.797	‑0.609	‑2.368***	‑2.038**	‑1.916**	‑2.1**	‑2.48***
Other social spending in kind x TS	2.097**	0.795	‑0.163	2.389***	2.179***	1.953**	2.417***	2.165**
Volume of redistribution x TS	0.343	0.096	‑0.203	0.142				0.155
Volume of redistribution x TS	0.218	‑0.035	‑0.14	0.226				‑0.275
Size of benefits					0.519	0.426	0.972
Size of taxes					‑0.634	‑0.605	‑0.749
Share of taxes paid by the bottom 30% x TS	‑0.815	‑0.086	0.088	‑0.811	‑0.646	‑0.647		‑1.28*
Share of taxes paid by the bottom 30% x TS	‑0.237	‑0.53	‑0.082	‑0.188				‑0.596
Share of benefits received by the bottom 30% x TS	‑0.957	‑3.07*	‑4.651**	‑0.871	‑0.743	‑0.796		‑1.021
Share of benefits received by the bottom 30% x TS	0.599	1.357	1.452*	0.761				0.761
Size of benefits x share of benefits received by the bottom 30%						‑0.293
Size of taxes x share of taxes received by the bottom 30%						‑0.044
Cash transfers to families (lagged 4‑6 years) x TS (lagged 4‑6 years)	‑0.34	‑1.497	‑1.84	‑0.052	‑0.503	‑0.394	‑0.485
	‑1.412**	‑1.328	‑0.933	‑1.649***	‑1.485**	‑1.508**	‑1.672***
Cash transfers to families, ages 12‑17 (lagged 7‑9 years) x TS (lagged 7‑9 years)								‑0.333
								‑0.516
Total social expenditure per capita (logged)	2.196	0.402	2.586	2.37	1.448	1.576	1.537	2.785*
Primary education spending (logged, lagged 13‑15 years)	‑1.239	2.598	5.832	‑0.718	‑1.412	‑1.655	‑1.705	0.172
Secondary education spending (logged, lagged 4‑6 years)	‑2.814	‑0.765	‑0.385	‑2.791	‑2.607	‑2.79	‑3.114	‑2.47
Mean poverty gap	0.212***	0.186*	0.179**	0.217***	0.205**	0.208**	0.168*	0.196**
Mean years of schooling	0.389	0.019	0.62	0.421	0.294	0.267	0.222	‑0.324
Adjusted R²	0.553	0.477	0.489	0.559	0.567	0.567	0.565	0.586
Number of countries	30	30	30	30	30	30	30	26
Number of observations	253	171	175	253	253	253	253	218
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.147	‑0.523	‑0.771	‑1.519*	‑0.994	0.125	‑1.353*	‑1.56*
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	0.914	1.257*	0.98	1.829**	0.814	0.81	0.82	1.544**
ECEC enrolment 3‑5 (lagged 16‑18 years) x TS (lagged 16‑18 years)			0.014
			‑0.967
Spending on ECEC (lagged 16‑18 years) x TS (lagged 16‑18 years)		0.866
		0.245
Spending on health services x TS	0.865	0.577	0.095		0.118	1.257	‑0.154	0.734
Spending on health services x TS	‑0.467	‑0.165	‑0.09		0.644	‑0.655	1.352	0.102
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)				2.081***
				‑1.207
Spending on non-preventive health services x TS				‑1.865
Spending on non-preventive health services x TS				1.691
Other social spending in kind x TS	2.429*	1.844	1.562	0.927	1.598	1.623	1.259	3.318*
Other social spending in kind x TS	‑2.205	‑2.756**	‑2.007	‑0.899	‑0.988	‑1.649	‑0.821	‑3.176
Volume of redistribution x TS	1.636*	0.832	1.42	1.426*				1.854**
Volume of redistribution x TS	‑0.993	‑0.436	‑0.669	‑0.528				‑1.308
Size of benefits					1.291	1.099	2.509*
Size of taxes					‑0.995	‑1.344**	‑1.356*
Share of taxes paid by the bottom 30% x TS	‑0.616	‑0.593	‑0.517	‑0.836	‑0.674	‑0.836*		‑0.619
Share of taxes paid by the bottom 30% x TS	‑0.804	‑0.697	‑0.885	‑0.202				‑0.164
Share of benefits received by the bottom 30% x TS	‑3.136***	‑3.486***	‑2.85**	‑3.042***	‑1.77	‑2.316**		‑3.912***
Share of benefits received by the bottom 30% x TS	1.481	2.036*	1.399	0.947				1.758
Size of benefits x share of benefits received by the bottom 30%						‑1.67***
Size of taxes x share of taxes received by the bottom 30%						0.275
Cash transfers to families (lagged 4‑6 years) x TS (lagged 4‑6 years)	‑0.859	‑1.069	‑1.182	‑1.159**	‑0.967	‑0.93*	‑0.859
	0.043	0.975	0.807	0.14	‑0.258	‑0.732	‑0.114
Cash transfers to families, ages 12‑17 (lagged 7‑9 years) x TS (lagged 7‑9 years)								‑1.328
								1.085
Total social expenditure per capita (logged)	‑0.59	‑1.439	‑1.119	‑0.044	‑0.556	1.012	‑1.201	‑0.1
Primary education spending (logged, lagged 13‑15 years)	‑5.609*	‑3.854	‑7.105**	‑7.083***	‑5.81*	‑7.645***	‑4.021	‑8.87***
Secondary education spending (logged, lagged 4‑6 years)	2.599*	1.796	3.074*	2.405**	1.734	2.126*	1.401	3.458**
Mean poverty gap	0.186**	0.147	0.208	0.21***	0.166	0.204***	0.032	0.322***
Mean years of schooling	‑0.698	0.19	0.084	‑0.296	‑0.74	‑0.26	‑1.286*	‑0.565
Adjusted R²	0.707	0.71	0.698	0.77	0.701	0.78	0.679	0.765
Number of countries	30	30	30	30	30	30	30	26
Number of observations	253	171	175	253	253	253	253	218
Country fixed effects	No	No	No	No	No	No	No	No
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

Health outcomes

Copy link to Health outcomes

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)	3.751***	3.89**	3.891***	3.325***	3.298***	3.265**	4.201***	3.575**	2.195*
	4.188**	4.548**	3.875*	3.376*	3.543*	3.235*	3.98*	5.188**	1.879
Spending on non-preventive health services x TS	0.387	‑0.249	‑0.325	‑0.512	‑0.369	‑1.021	‑0.341	‑0.24	‑2.152
Spending on non-preventive health services x TS	‑0.414	‑0.594	‑0.679	0.246	0.098	0.265	‑0.613	‑0.243	‑1.032
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		‑1.104	‑1.266
		1.245	‑0.043
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		‑0.649	0.675
		‑0.216	‑0.488
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	‑0.021		‑0.942	‑0.064	‑0.098	0.116	‑0.302	0.409	2.128
	2.574***		1.639	2.757***	2.595***	2.65***	3.769***	2.824**	1.724*
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 0‑2 (lagged 13‑15 years)			0.621
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 3‑5 (lagged 9‑11 years)			1.49
Other social spending in kind x TS	‑4.263***	‑3.752**	‑4.419***	‑5.282***	‑4.95***	‑5.24***	‑3.346***	‑3.766***	‑2.918***
Other social spending in kind x TS	5.544***	4.172***	5.56***	6.814***	6.347***	6.873***	4.515***	5.023***	3.332***
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	1.758***	1.215*	1.369*	1.699***	1.741***	1.667***	1.991***	2.254***	1.313**
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	2.707***	1.37	2.333**	2.524***	2.461***	2.653***	2.689***	2.767**	1.923*
Volume of redistribution x TS	0.117	0.147	‑0.14				0.049	0.315	‑0.89
Volume of redistribution x TS	‑0.854	‑1.035	‑0.608				‑0.555	‑0.625	0.595
Size of taxes				‑0.639	‑0.76	‑0.776
Size of benefits				‑0.245	‑0.473	‑0.21
Share of taxes paid by the bottom 30% x TS	‑0.782**	‑0.536	‑0.587	‑0.853**	‑0.833**		‑0.634*	‑0.565*	‑0.873**
Share of taxes paid by the bottom 30% x TS	‑0.006	0.127	0.163				‑0.052	‑0.337	‑0.822
Share of benefits received by the bottom 30% x TS	0.009	‑0.306	‑0.239	0.372	0.114		‑0.713	‑1.307*	‑1.514**
Share of benefits received by the bottom 30% x TS	1.835	2.089	1.971				1.868	2.751**	1.192
Size of taxes x share of taxes received by the bottom 30%					‑0.157
Size of benefits x share of benefits received by the bottom 30%					‑0.357
Cash transfers to families x TS	0.319	0.167	0.289	‑0.174	‑0.107	‑0.256
Cash transfers to families x TS	1.935**	1.44	1.588	1.953**	1.774**	2.02**
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							0.839
							2.471***
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								2.03*
								‑0.264
In-cash spending on families, ages 12‑17 x TS									‑1.086***
In-cash spending on families, ages 12‑17 x TS									‑1.458**
Total social expenditure per capita (logged)	5.522	1.851	4.815	4.416	4.384	3.588	3.534	3.871	‑0.503
Mean poverty gap	‑0.133	‑0.105	‑0.161**	‑0.118	‑0.096	‑0.12	‑0.149*	‑0.163**	‑0.074
Adjusted R²	0.455	0.359	0.472	0.444	0.446	0.427	0.503	0.492	0.46
Number of countries	29	29	29	29	29	29	24	25	27
Number of observations	249	247	247	249	249	249	214	217	171
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	No	No	No	No	No	No	No	No	No

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)	‑5.864***	‑5.166***	‑5.185***	‑6.228***	‑6.642***	‑6.081***	‑5.341***	‑6.91***	‑7.024***
	0.371	0.582	0.652	0.022	‑0.263	0.522	0.168	3.03	2.528
Spending on non-preventive health services x TS	6.808***	6.055***	2.293	6.266***	8.015***	5.08**	6.844***	4.318**	6.378**
Spending on non-preventive health services x TS	‑3.054	‑4.767	‑4.791*	‑2.205	‑4.276	‑1.865	‑8.111***	‑4.616	‑4.562
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		0.97	‑0.279
		‑4.292**	‑3.726**
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		0.142	5.179**
		0.362	‑0.235
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	0.613		‑1.498	0.385	1.471*	0.523	‑2.405*	‑1.745*	‑1.024
	‑1.532		‑3.679	‑1.84	‑1.544	‑1.934	2.375	0.029	‑2.761*
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 0‑2 (lagged 13‑15 years)			0.336
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 3‑5 (lagged 9‑11 years)			5.906**
Other social spending in kind x TS	1.357	0.731	‑0.518	1.838	1.672	1.269	9.089***	2.473	4.013**
Other social spending in kind x TS	0.866	0.485	1.354	0.456	‑2.081	‑0.243	‑10.926***	0.095	‑1.095
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.801	‑1.531	‑1.395	‑2.164	‑0.464	‑2.237	‑2.836***	‑2.62**	‑2.525*
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	2.004	2.494	2.257	2.074	2.49*	2.495	6.23***	3.576***	3.742**
Volume of redistribution x TS	‑2.141	‑0.429	‑0.049				‑0.945	‑0.193	‑0.268
Volume of redistribution x TS	2.669*	1.307	0.772				1.465	1.813	2.016
Size of taxes				1.579	1.65	1.032
Size of benefits				0.282	‑0.846	1.684
Share of taxes paid by the bottom 30% x TS	‑0.941	‑0.212	‑0.942	‑1.123	‑1.574*		‑0.955**	‑1.486**	‑1.523*
Share of taxes paid by the bottom 30% x TS	1.174	1.159	0.485				0.659	1.481	2.188
Share of benefits received by the bottom 30% x TS	‑3.248	‑2.628	1.556	‑2.166	‑2.04		‑7.48***	‑4.715***	‑5.409***
Share of benefits received by the bottom 30% x TS	1.37	1.656	‑0.399				7.438***	1.641	2.545
Size of taxes x share of taxes received by the bottom 30%					2.22*
Size of benefits x share of benefits received by the bottom 30%					‑2.171**
Cash transfers to families x TS	‑0.762	1.03	‑0.055	‑0.973	‑0.133	‑1.391
Cash transfers to families x TS	‑1.597	‑2.286	‑1.961	‑1.214	‑0.932	‑0.138
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							0.912
							0.666
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								4.68***
								‑6.743***
In-cash spending on families, ages 12‑17 x TS									1.661
In-cash spending on families, ages 12‑17 x TS									‑4.949**
Total social expenditure per capita (logged)	2.423	0.283	‑4.48	1.378	1.378	0.388	0.69	5.215**	2.807
Mean poverty gap	‑0.07	‑0.088	‑0.18	‑0.155	‑0.161	‑0.243*	‑0.414**	‑0.292**	‑0.093
Adjusted R²	0.467	0.507	0.581	0.45	0.525	0.423	0.698	0.675	0.609
Number of countries	29	29	29	29	29	29	24	25	27
Number of observations	249	247	247	249	249	249	214	217	171
Country fixed effects	No	No	No	No	No	No	No	No	No
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	No	No	No	No	No	No	No	No	No

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)	‑1.251	‑1.543	‑1.552	‑1.305	‑1.201	‑1.202	‑1.302	‑1.168	‑0.608
	‑1.578	‑2.627	‑2.278	‑1.109	‑1.222	‑0.859	‑2.881	‑2.81	‑0.283
Spending on non-preventive health services x TS	‑0.608	‑0.86	‑0.688	‑0.501	‑0.842	0.42	0.119	0.441	0.846
Spending on non-preventive health services x TS	‑1.705	‑1.901	‑1.871	‑1.937	‑1.726	‑1.895	‑1.258	‑1.831	‑2.227
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		1.27	1.384
		‑1.756	‑0.955
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		1.777	0.887
		‑1.494*	‑1.245
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	0.158		0.646	0.161	0.388	‑0.121	0.894	‑0.016	0.219
	‑1.924**		‑0.501	‑1.907**	‑1.781**	‑1.718*	‑3.13*	‑1.892*	‑1.97**
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 0‑2 (lagged 13‑15 years)			‑1.011
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 3‑5 (lagged 9‑11 years)			‑0.919
Other social spending in kind x TS	3.363***	2.669*	2.996**	3.538***	3.313**	3.418***	3.717***	4.134***	4.529***
Other social spending in kind x TS	‑4.342***	‑3.412**	‑4.116***	‑4.597***	‑4.256***	‑4.711***	‑4.813***	‑4.988***	‑4.771***
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.551**	‑0.513	‑0.531	‑1.576**	‑1.578**	‑1.527**	‑0.941*	‑1.394**	‑0.152
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.476*	‑1.252**	‑1.752**	‑1.578*	‑1.521*	‑1.793*	‑2.279***	‑2.352***	‑2.89***
Volume of redistribution x TS	‑0.353	‑0.564	‑0.366				‑1.14	‑0.866	1.122
Volume of redistribution x TS	1.005	0.593	0.269				0.913	0.615	‑1.584
Size of taxes				0.607	0.749	0.827
Size of benefits				‑0.4	‑0.325	‑0.302
Share of taxes paid by the bottom 30% x TS	1.645***	1.296***	1.26***	1.447***	1.389***		1.408***	1.459***	1.21*
Share of taxes paid by the bottom 30% x TS	0.652	0.362	0.273				0.672	0.713	1.205
Share of benefits received by the bottom 30% x TS	‑1.068	‑1.353	‑1.434	‑1.106	‑0.815		‑0.9	‑0.599	0.238
Share of benefits received by the bottom 30% x TS	0.042	0.41	0.441				0.579	0.017	0.539
Size of taxes x share of taxes received by the bottom 30%					0.671
Size of benefits x share of benefits received by the bottom 30%					0.213
Cash transfers to families x TS	‑0.222	‑0.789	‑0.821	‑0.004	‑0.104	0.08
Cash transfers to families x TS	‑0.438	0.023	‑0.031	‑0.443	‑0.411	‑0.525
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							‑1.507**
							‑1.599
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								0.5
								‑0.917
In-cash spending on families, ages 12‑17 x TS									0.533
In-cash spending on families, ages 12‑17 x TS									0.644
Total social expenditure per capita (logged)	‑1.862	‑2.531	‑4.186	‑1.356	‑1.358	0.1	‑1.267	‑0.547	‑1.928
Mean poverty gap	‑0.014	‑0.013	0.023	‑0.014	‑0.042	‑0.013	0.007	0.017	‑0.006
Adjusted R²	0.317	0.365	0.418	0.316	0.323	0.248	0.429	0.411	0.464
Number of countries	29	29	29	29	29	29	24	25	27
Number of observations	249	247	247	249	249	249	214	217	171
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	No	No	No	No	No	No	No	No	No

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)	1.27	1.306*	1.604**	1.205	2.243***	1.205	0.043	0.982	0.464
	0.441	0.428	‑0.391	0.319	1.194	1.453	0.305	‑1.765	‑0.073
Spending on non-preventive health services x TS	‑0.644	‑1.119	0.144	‑1.687	‑3.958***	‑0.963	‑2.643	‑1.441	‑1.3
Spending on non-preventive health services x TS	‑2.522	‑1.003	‑1.036	‑1.709	‑0.242	‑1.954	2.444	‑0.738	‑1.899
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		2.137***	3.241***
		2.045**	1.897**
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		‑0.633	‑2.982***
		‑0.472	0.052
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	0.025		0.147	‑0.635	‑1.545**	‑0.45	1.254	0.928	1.369***
	0.004		‑0.018	0.239	‑0.105	0.599	‑2.549**	‑0.308	0.732
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 0‑2 (lagged 13‑15 years)			0.728
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 3‑5 (lagged 9‑11 years)			‑2.814**
Other social spending in kind x TS	1.235	1.416	2.492**	1.406	1.105	1.538	‑3.577**	‑1.135	‑0.627
Other social spending in kind x TS	‑3.987**	‑4.764***	‑5.195***	‑4.157**	‑2.453	‑3.692**	3.544*	‑1.657	‑2.019
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	1.054	1.358*	1.016	1.212	0.1	1.329	1.498**	1.567**	1.027
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑2.709**	‑3.088***	‑3.134***	‑2.354*	‑2.627**	‑2.143	‑5.041***	‑3.915***	‑4.711***
Volume of redistribution x TS	‑0.83	‑1.337	‑1.394				‑0.986	‑0.532	0.821
Volume of redistribution x TS	‑0.011	‑0.415	0.179				0.41	‑0.847	‑2.081***
Size of taxes				1.393*	1.322*	1.194
Size of benefits				1.002	1.762**	‑0.98
Share of taxes paid by the bottom 30% x TS	‑0.211	‑0.33	‑0.016	‑0.323	‑0.095		0.318	0.183	‑0.658
Share of taxes paid by the bottom 30% x TS	‑0.817	‑1.171*	‑0.901				‑0.705	‑1.343**	‑0.494
Share of benefits received by the bottom 30% x TS	‑0.286	‑0.111	‑1.968	2.667*	2.837**		3.091**	1.755**	1.553**
Share of benefits received by the bottom 30% x TS	3.384**	1.954	3.124*				‑1.139	2.162**	3.1***
Size of taxes x share of taxes received by the bottom 30%					0.557
Size of benefits x share of benefits received by the bottom 30%					2.294***
Cash transfers to families x TS	2.832***	2.582***	2.85***	1.619**	0.589	2.033***
Cash transfers to families x TS	0.468	1.437	0.648	0.324	1.001	‑0.363
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							0.967
							‑0.181
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								1.149
								1.746*
In-cash spending on families, ages 12‑17 x TS									1.588***
In-cash spending on families, ages 12‑17 x TS									2.379**
Total social expenditure per capita (logged)	‑1.46	‑2.004*	‑1.254	‑3.928***	‑4.383***	‑1.779	‑1.165	‑3.336*	‑1.565
Mean poverty gap	‑0.106	‑0.037	0.001	‑0.171	‑0.119	0.012	0.126	‑0.082	‑0.057
Adjusted R²	0.498	0.554	0.588	0.46	0.55	0.432	0.562	0.576	0.618
Number of countries	29	29	29	29	29	29	24	25	27
Number of observations	249	247	247	249	249	249	214	217	171
Country fixed effects	No	No	No	No	No	No	No	No	No
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	No	No	No	No	No	No	No	No	No

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)	‑1.062	‑1.273	‑1.037	‑0.145	‑0.017	‑0.135	0.221	‑0.389	‑0.012
	0.085	‑1.494	‑0.028	2.239	1.333	2.283	3.312	1.324	0.229
Spending on non-preventive health services x TS	‑2.88	‑2.707	‑2.894	‑0.475	‑1.052	‑0.305	3.91*	3.691	‑1.25
Spending on non-preventive health services x TS	2.011	1.481	2.019	0.454	1.425	0.576	‑0.939	‑0.043	‑2.597
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		‑0.269	0.023
		‑3.731**	‑2.919
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		0.211	‑1.597
		‑1.089	‑0.855
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	1.477		2.755	1.666	1.617	1.689	1.242	0.303	2.141
	‑2.969*		‑3.573*	‑3.699**	‑2.876**	‑3.663**	‑5.778***	‑4.39***	‑2.963**
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 0‑2 (lagged 13‑15 years)			3.202
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 3‑5 (lagged 9‑11 years)			‑3.314*
Other social spending in kind x TS	3.084	1.461	3.408	5.717*	3.827	5.625*	4.255**	4.666**	4.23***
Other social spending in kind x TS	‑8.473***	‑6.813**	‑9.408***	‑11.907***	‑9.235***	‑11.928***	‑8.995***	‑10.033***	‑5.942***
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑0.533	‑0.088	‑0.172	‑0.362	‑0.665	‑0.357	‑1.002	‑0.981	0.105
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑2.299*	‑1.945	‑2.465	‑2.06	‑1.757	‑2.061	‑4.37***	‑4.598***	‑2.112***
Volume of redistribution x TS	1.635	1.292	1.567				5.296***	5.027***	1.189
Volume of redistribution x TS	1.169	0.399	0.156				‑2.509*	‑1.335	‑1.452
Size of taxes				‑1.117	‑0.447	‑1.118
Size of benefits				0.694	1.978	0.948
Share of taxes paid by the bottom 30% x TS	‑0.187	‑0.687	‑0.335	0.112	‑0.029		0.088	0	0.619
Share of taxes paid by the bottom 30% x TS	0.205	‑0.189	0.126				0.573	0.414	1.743*
Share of benefits received by the bottom 30% x TS	1.434	0.704	0.974	‑0.828	0.296		1.906	1.298	0.649
Share of benefits received by the bottom 30% x TS	‑5.858**	‑5	‑4.65				‑4.573*	‑4.813**	‑0.326
Size of taxes x share of taxes received by the bottom 30%					0.453
Size of benefits x share of benefits received by the bottom 30%					2.144**
Cash transfers to families x TS	‑1.577*	‑1.712	‑2.368*	‑0.178	‑0.458	‑0.262
Cash transfers to families x TS	‑1.597	‑1.941	‑2.289	‑2.199	‑1.081	‑2.159
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							1.097
							0.064
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								‑0.498
								‑2.133
In-cash spending on families, ages 12‑17 x TS									0.198
In-cash spending on families, ages 12‑17 x TS									1.353
Total social expenditure per capita (logged)	‑13.694**	‑14.225**	‑18.031***	‑11.015*	‑10.988*	‑10.791*	‑7.459	‑6.191	‑6.812
Mean poverty gap	0.18	0.15	0.207	0.193	0.092	0.19	0.312**	0.286**	0.063
Adjusted R²	0.258	0.259	0.309	0.199	0.233	0.204	0.33	0.331	0.367
Number of countries	29	29	29	29	29	29	24	25	27
Number of observations	249	247	247	249	249	249	214	217	171
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	No	No	No	No	No	No	No	No	No

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Spending on preventive health services (lagged 2‑4 years) x TS (lagged 2‑4 years)	0.65	0.603	0.813	0.599	1.53**	0.601	‑0.77	0.555	0.702
	1.394	1.061	0.889	1.359	2.149	1.837	3.416	‑0.307	‑0.339
Spending on non-preventive health services x TS	‑1.163	‑1.294	‑1.113	‑1.615*	‑3.622***	‑1.325	‑3.698***	‑1.846*	‑1.778
Spending on non-preventive health services x TS	‑0.968	‑0.023	‑0.168	‑0.64	0.622	‑0.742	4.534***	0.403	‑1.366
ECEC enrolment 0‑2 (lagged 13‑15 years) x TS (lagged 13‑15 years)		1.997***	2.358***
		2.045***	2.055***
ECEC enrolment 3‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)		‑1.309**	‑1.931*
		‑0.074	0.079
Spending on ECEC (lagged 11‑15 years) x TS (lagged 11‑15 years)	0.242		0.158	‑0.059	‑0.849	0.019	2.095*	0.996	1.676***
	0.533		‑0.449	0.584	0.282	0.737	‑2.567**	0.043	0.59
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 0‑2 (lagged 13‑15 years)			0.602
Spending on ECEC (lagged 11‑15 years) x ECEC enrolment 3‑5 (lagged 9‑11 years)			‑0.762
Other social spending in kind x TS	1.789	2.727***	2.987**	1.976	1.694	2.026	‑3.326**	0.211	‑0.741
Other social spending in kind x TS	‑3.458**	‑4.368***	‑4.591***	‑3.686**	‑2.224	‑3.501**	4.211**	‑1.786	‑2.139*
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	0.34	0.519	0.414	0.393	‑0.559	0.439	0.386	0.284	0.76
Spending on ALMP (lagged 3 years) x TS (lagged 3 years)	‑1.919**	‑2.358***	‑2.398**	‑1.779*	‑2.013**	‑1.683*	‑3.392***	‑2.326**	‑3.98***
Volume of redistribution x TS	‑0.629	‑1.205*	‑1.144				‑0.803	‑0.276	0.193
Volume of redistribution x TS	0.274	‑0.006	0.323				0.972	‑0.307	‑1.187
Size of taxes				0.841	0.777	0.747
Size of benefits				0.636	1.286	‑0.184
Share of taxes paid by the bottom 30% x TS	‑0.012	‑0.406	‑0.336	‑0.15	0.042		0.398	0.306	‑0.453
Share of taxes paid by the bottom 30% x TS	‑0.397	‑0.572	‑0.44				0.326	‑0.678	‑0.475
Share of benefits received by the bottom 30% x TS	‑0.418	‑0.541	‑1.004	1.097	1.256		2.504***	0.905	1.56*
Share of benefits received by the bottom 30% x TS	1.594	‑0.291	0.247				‑2.751	0.77	2.836**
Size of taxes x share of taxes received by the bottom 30%					0.592
Size of benefits x share of benefits received by the bottom 30%					2.012***
Cash transfers to families x TS	2.265***	1.75***	1.788**	1.7***	0.791	1.868***
Cash transfers to families x TS	0.819	0.719	0.51	0.831	1.459*	0.558
Cash transfers to families, ages 0‑5 (lagged 9‑11 years) x TS (lagged 9‑11 years)							1.518*
							‑0.167
Cash transfers to families, ages 6‑11 (lagged 4‑6 years) x TS (lagged 4‑6 years)								1.072
								1.421
In-cash spending on families, ages 12‑17 x TS									1.384**
In-cash spending on families, ages 12‑17 x TS									2.507***
Total social expenditure per capita (logged)	‑1.444	‑2.17*	‑2.197	‑2.749**	‑3.162**	‑1.856	‑0.148	‑3.104**	‑2.137
Mean poverty gap	‑0.015	0.084	0.096	‑0.058	‑0.01	0.018	0.267***	0.01	‑0.05
Adjusted R²	0.389	0.443	0.437	0.389	0.458	0.387	0.393	0.332	0.598
Number of countries	29	29	29	29	29	29	24	25	27
Number of observations	249	247	247	249	249	249	214	217	171
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	No	No	No	No	No	No	No	No	No

Total social expenditure

Copy link to Total social expenditure

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Spending on ALMP (lagged 3 years) x GDP (lagged 3 years)	‑0.035	‑0.034	‑0.033	‑0.043**	‑0.032	‑0.017	‑0.01	‑0.017	‑0.015	‑0.016
Spending on ALMP (lagged 3 years) x GDP (lagged 3 years)	‑0.046*	‑0.051**	‑0.041**	‑0.04**	‑0.043**	‑0.044**	‑0.045**	‑0.022	‑0.026	‑0.023
ECEC enrolment 0‑2 x GDP	0.047		0.03	0.046	0.05	0.059*	0.041	‑0.001	‑0.004	‑0.001
ECEC enrolment 0‑2 x GDP	‑0.047***		‑0.038***	‑0.043***	‑0.034**	‑0.045***	‑0.043***	‑0.02	‑0.027**	‑0.03**
ECEC enrolment 3‑5 x GDP	‑0.022		‑0.009	‑0.015	‑0.015	‑0.046**	‑0.054***	‑0.02	‑0.029	‑0.026
ECEC enrolment 3‑5 x GDP	0.02		0.017	0.015	0.008	‑0.001	0.001	0.015	0.016	0.016
Spending on ECEC x GDP		‑0.022	‑0.022
Spending on ECEC x GDP		0.016	0.007
Spending on ECEC x ECEC enrolment 0‑2			0.007
Spending on ECEC x ECEC enrolment 3‑5			0.007
Spending on health services x GDP				‑0.11***
Spending on health services x GDP				‑0.014
Spending on preventive health services (lagged 2‑4 years) x GDP					0.017
					‑0.049**
Spending on non-preventive health services x GDP					‑0.087***
Spending on non-preventive health services x GDP					‑0.004
Other social spending in kind x GDP	0.058**	0.064**	0.06**	0.03	0.039*	0.033	0.016	0.062***	0.067***	0.069***
Other social spending in kind x GDP	‑0.004	‑0.026	‑0.016	‑0.015	‑0.011	‑0.003	0.008	‑0.083***	‑0.083***	‑0.085***
Volume of redistribution x GDP	0.029	0.014	0.015	0.045*	0.034			0.061***	0.056***	0.056***
Volume of redistribution x GDP	‑0.046***	‑0.029	‑0.032*	‑0.065***	‑0.062***			‑0.067***	‑0.056***	‑0.059***
Size of taxes						0.105***	0.133***
Size of benefits						‑0.048**	‑0.03
Share of taxes paid by the bottom 30% x GDP	‑0.029	‑0.026	‑0.023	‑0.018	‑0.011		‑0.027	0.005	0.003	0.003
Share of taxes paid by the bottom 30% x GDP	0.002	0.015	0.015	0.008	0.006			0.029**	0.027*	0.026*
Share of benefits received by the bottom 30% x GDP	0.001	0.006	0.004	0.017	‑0.013		0.059	0.086*	0.087*	0.082*
Share of benefits received by the bottom 30% x GDP	‑0.051***	‑0.036**	‑0.033*	‑0.056**	‑0.065***			‑0.047***	‑0.038**	‑0.037**
Size of taxes x share of taxes received by the bottom 30%							0.026**
Size of benefits x share of benefits received by the bottom 30%							‑0.014
Cash transfers to families x GDP	0.062***	0.057***	0.062***	0.033	0.032	0.031	0.03
Cash transfers to families x GDP	‑0.002	0.017	0.006	0.007	‑0.003	‑0.018	‑0.013
Cash transfers to families, ages 0‑5 x GDP								0.014
Cash transfers to families, ages 0‑5 x GDP								‑0.001
Cash transfers to families, ages 6‑11 x GDP									‑0.003
Cash transfers to families, ages 6‑11 x GDP									‑0.032
In-cash spending on families, ages 12‑17 x GDP										0.004
In-cash spending on families, ages 12‑17 x GDP										‑0.032
GDP per capita (logged)	0.178***	0.178***	0.172***	0.229***	0.199***	0.264***	0.274***	0.08	0.091*	0.094*
Life expectancy (logged)	0.38	0.238	‑0.111	‑0.948	‑1.094	0.875	0.604	2.278	2.222	2.219
Dependency ratio	‑0.008	‑0.006	‑0.008	‑0.003	‑0.002	‑0.002	‑0.003	‑0.006	‑0.007	‑0.007
Adjusted R²	0.619	0.636	0.652	0.658	0.686	0.657	0.675	0.616	0.632	0.627
Number of countries	32	32	32	32	32	32	32	28	28	28
Number of observations	482	482	482	482	478	478	478	333	333	333
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Leads and lags for core and policy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

Note: Coefficients in bold are the ones used for the value for money calculations. “x” signifies an interaction between two variables. “x GDP” means the interaction between the policy variable in the line above and the log of GDP per capita. The log of GDP per capita is standardised along with the policy variables to facilitate the interpretation of results.

Annex 4.D. Value for money at different levels of social expenditure and GDP

Copy link to Annex 4.D. Value for money at different levels of social expenditure and GDP

← 1. NEETs data are not available by socio‑economic status.

← 2. The indicators include spending on active labour market programmes (ALMP), ECEC enrolment among children aged 0‑2 and 3‑5, spending on ECEC, spending on health services, preventive and non-preventive healthcare spending, and other social spending provided in kind. Definitions of these indicators are provided in Chapter 3. Where relevant to the timing of the expected effects, the variables are lagged: ALMP spending is lagged by three years to reflect the time required for programmes to translate into employment outcomes; ECEC enrolment and spending variables are lagged to match the previous enrolment of the cohorts considered; and preventive healthcare spending is lagged by two to four years to reflect the time required for effects on adolescent health outcomes to materialise. Other social spending in kind refers to in-kind spending across incapacity-related, housing and other social policy areas.

← 3. The indicators include the volume of redistribution, measured as equivalised net benefits received from public transfers (benefits net of taxes paid) by the average individual living in a working-age household, expressed as a percentage of GDP per capita; the share of benefits received by the bottom 30%, measuring public social security cash benefits received by working-age individuals in the bottom 30% of the post-transfer, equivalised household income distribution, per equivalent household member, as a share of benefits received by all working-age individuals; the share of taxes paid by the bottom 30%, defined analogously as the share of taxes paid by working-age individuals in the bottom 30% of the post-transfer, equivalised household income distribution; and the share of cash transfers to families, measured as the share of total social expenditure allocated in cash to families. Cash transfers to families are lagged by five years when the outcome is the NEET rate, reflecting the time elapsed since many of the individuals concerned were eligible for such benefits. Three additional indicators capture the share of total social expenditure provided in cash to children and their families during early childhood (ages 0‑5), middle childhood (ages 6‑11) and adolescence (ages 12‑17). Spending during early and middle childhood is lagged to capture exposure to benefits in previous years, while spending during adolescence is similarly lagged when estimating the NEET rate. For the benefit and tax distribution indicators, households are ranked according to their post-transfer equivalised income because of data availability. This may underestimate the redistributive effect of benefits or taxes where these themselves cause individuals to move into or out of the bottom 30% of the income distribution.

← 4. The tables in Annex 4.C display the regression estimates from the country-level panel analysis on educational outcomes as laid out in the previous section. In addition to time fixed effects, results are shown both with and without country fixed effects to analyse within-country associations as well as differences in levels across countries.

← 5. In our sample, the between-country variance of the average logged primary education expenditure per student is nearly four times larger than the within-country variance and it is nearly ten times larger for logged secondary education expenditure per student.

← 6. Annex 4.D provides complementary simulations that illustrate the value for money of social policies in supporting children’s educational outcomes for countries at different levels of per capita social expenditure and GDP per capita.

← 7. Results for the rates of NEETs among 15‑29 year‑olds yield no significant associations with the provision of ECEC 16 to 18 years ago, neither for enrolment rates at ages 3‑5 nor spending shares on ECEC services. This may be a consequence of the large cohort of NEETs under consideration and the limited available time series data on ECEC, which restricts the possibility to match youth outcomes to the characteristics of ECEC services these individuals might have experienced during their early years.

← 8. A possible explanation of the absence of association found between spending on ALMPs and academic performance at age 15 and NEETs rates could be that ALMPs operate through improvements in parental employment and income stability, which help reduce severe material deprivation by alleviating immediate financial constraints and improving living conditions. However, these gains may not be sufficient on their own to translate into measurable improvements in more distal outcomes such as PISA performance, which are shaped by a broader set of factors and accumulate over a longer period. Educational outcomes depend not only on parents’ labour market attachment, but also on the quality of schooling, family relationships, and the continuity and timing of support received throughout childhood. As a result, while ALMPs appear effective in reducing material hardship (see Chapter 3), their contribution to adolescent educational outcomes may be more indirect, gradual and difficult to detect empirically.

← 9. These results are available on request.

← 10. The impact of earlier cash benefits was not tested to due limitations in the available time series data, which prevent the construction of sufficiently long lags.

← 11. Annex 4.D provides complementary simulations that illustrate the value for money of social policies in supporting children’s health outcomes for countries at different levels of per capita social expenditure and GDP per capita.

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment

Governance

Health

Industry, business and entrepreneurship

Regional, rural and urban development

Science, technology and innovation

Society

Taxation

Trade

Energy

Nuclear energy

Transport

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment

Governance

Health

Industry, business and entrepreneurship

Regional, rural and urban development

Science, technology and innovation

Society

Taxation

Trade

Energy

Nuclear energy

Transport

Countries A - C

Countries D - I

Countries J - M

Countries N - R

Countries S - T

Countries U - Z

Regional and global engagement

Countries

Countries A - C

Countries D - I

Countries J - M

Countries N - R

Countries S - T

Countries U - Z

Regional and global engagement

Publications

Publications

Featured publications

Data

Data

Featured data

News & events

News & events

Featured events

About OECD

About

Engage with us

Work with us

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment