Keywords

1 Introduction

Socioeconomic inequality in young people’s academic achievement has become one of the key academic and political issues of the twenty-first century. Indeed, public policymakers across the globe are now seeking to raise the cognitive skills of young people from disadvantaged backgrounds, and to narrow the gap in achievement between this group and their more affluent peers. There are at least three reasons why this is now seen as such a pressing issue. The first is economic efficiency. In a competitive world, it is vital that each country is making the most of its human resources. Yet, if young people from poor backgrounds are failing to reach their academic potential, then this is unlikely to be the case. The second reason is social justice. Individuals do not pick the family and socioeconomic position they are born into. Rather, it is luck of the draw. Hence many would deem it to be “unfair” and inequitable if life chances are to a large extent determined by a factor, such as family background, that is largely outside of one’s control. The final reason is the persistence of inequality. Many view education as a key driver of economic inequality and intergenerational mobility (Economic and Social Research Council, 2012; Goldthorpe, 2014). Consequently, persistence in educational inequalities will translate into continuing inequalities in later life. This is not only in terms of labor market outcomes (occupation and income), but also other wider factors that education is thought to influence, such as well-being and health (Chou, Liu, Grossman, & Joyce, 2010).

Figure 1.1, drawn from Jerrim and Macmillan’s (2015) research, helps to formalize this argument by illustrating the link between parental education, their offspring’s education, and their offspring’s later lifetime outcomes. It also illustrates the three broad mechanisms that are thought to drive the parent–child relationship in educational achievement. The first is the biological channel of heredity transfers—genetic differences in individuals’ academic potential that may be transmitted across generations. A growing body of research is highlighting the importance of genetics for our understanding of socioeconomic gaps in educational achievement (Ayorech, Krapohl, Plomin, & von Stumm, 2017), though the bio-molecular work in this area is still somewhat in its infancy (Jerrim, Vignoles, Lingam, & Friend, 2015). The second mechanism is non-financial resources. This encompasses a whole host of factors throughout childhood which, although not costing much money, differ (on average) between high and low socioeconomic parents. Examples include breastfeeding, reading and interactions with the child, helping regularly with homework, and parenting styles, each of which are plausibly linked to children’s educational achievement (e.g., Sacker, Kelly, Iacovou, Cable, & Bartley, 2013). Finally, parents with lower levels of education will have fewer financial resources to invest in their children’s education. They are consequently less likely to have access to the necessary educational materials that their children need to achieve high outcomes in school. Possible examples include access to books/computers, attending lower quality schools, and being unable to afford private tutors. Along with macro-economic forces (e.g., income inequality), public investment (e.g., government expenditure on education), and institutional structures (e.g., the design of the education system), these three forces combine to generate significant disparities in educational achievement by family background that can be observed across the developed world.

Fig. 1.1
figure 1

Source Jerrim and MacMillan (2015)

Conceptual framework linking parental education to educational achievement and later lifetime outcomes.

Full size image

Figure 1.1 also serves as motivation as to why it is important to consider socioeconomic differences in educational achievement from an international comparative perspective. In order to judge whether inequality in achievement is large or small in any given country, it is necessary to have a yardstick to measure it against. For instance, is a correlation of 0.5 between parent and child years of schooling weak or strong? Drawing comparisons to other countries of a similar level of development provides an important and insightful context against which we can judge such results. Relatedly, Fig. 1.1 also has highlighted how heredity is thought to be one of the three key intergenerational mechanisms driving the intergenerational transmission of education. Yet, although this may help to explain parent–child links within a single country, it is difficult to see why this would cause differences between countries. In other words, the role of heredity transfers in generating intergenerational inequalities is likely to be approximately equal across nations. Hence, when considering why socioeconomic inequality in academic achievement is stronger in one country than another, we can largely rule this hereditary mechanism out. This then leaves factors that can be influenced by public policy—such as parental investments, macro-economic conditions, and institutional structures—as the remaining drivers of any cross-national differences. Indeed, as previous research has shown (e.g., Hanushek & Wossmann, 2006), cross-national comparisons also provide a natural way for one to consider how key institutional structures, such as the design of education systems, influences inequality in young people’s outcomes.

2 Measurement Issues Regarding Socioeconomic Background

As the previous section has highlighted, there are important reasons to study inequality in educational achievement within an international comparative framework. There are, however, also important challenges, particularly with regards to the measurement of educational achievement across multiple countries and two generations. We provide an overview of these issues here, with a focus upon the measurement of family background. Although challenges also exist with respect to the robustness and international comparability of measures of children’s academic achievement, we refer readers to chapters in previous edited volumes that have addressed this matter in detail (e.g., Goldstein, 2017).

The first decision one has to make when studying socioeconomic inequalities is which measure (or measures) of family background to use. Three main indicators are widely used in the literature: parental education, parental occupation, and (permanent) family income. Each has its advantages and disadvantages. For instance, while family income is easy to understand and interpret by a wide audience, and is arguably the most cross-nationally comparable, young people are unable to report it accurately, and it thus must be captured from parents directly. This means that it can be limited in terms of availability. On the other hand, young people generally can report parental occupation and parental education reasonably well (Jerrim & Micklewright, 2014), with these indicators therefore available within most datasets. Yet they suffer from a host of other measurement issues, as we shall discuss below.

An alternative to using just a single indicator is to combine several measures into a scale. This has been the preferred approach of the Organisation for Economic Co-operation and Development (OECD) in the Programme for International Student Assessment (PISA) study. This has the advantage of better capturing the multidimensional nature of any one indicator alone (Marks, 2011). However, such composite indicators are often difficult to interpret and communicate, while having also been criticized for their cross-national comparability (Rutkowski & Rutkowski, 2013). Additionally, composite indicators utilize cut scores to determine “low” versus “high” SES which varies largely across countries and reminds us of the important distinctions that exist between absolute versus relative poverty (Ravillion, 2016).

Throughout this volume, we have made a pragmatic choice of parental education (the highest level out of the child’s mother and father) to be the preferred measure of socioeconomic position (wherever possible). Although we recognize that previous research has suggested that different family background indicators produce similar, but not identical, orderings of countries in terms of socioeconomic inequalities in student performance (Marks, 2011), we have decided to focus upon parental education for a number of reasons. First, this information is routinely collected in most social surveys across the world. Consequently, it is available in most national and international data sources within our countries of interest. Second, despite criticisms (Schneider, 2013), the International Standard Classification of Education (ISCED) framework provides (to some extent) a harmonized framework that allows for comparisons across surveys and international jurisdictions. This is not always true of the alternatives, such as with parental occupation or composite measures, which are sometimes recorded in datasets following national-specific categorizations. Third, as Fig. 1.1 has already demonstrated, there are clear mechanisms by which higher levels of parental education may cause their offspring to have higher levels of achievement at school. Fourth, the meaning of parental education is widely understood as a valid measure of family background among public policymakers and non-specialist audiences. Finally, it has also been the preferred measure in other cross-national research into socioeconomic inequalities (e.g., Bradbury, Corak, Waldfogel, & Washbrook, 2015) meaning that the work presented in this volume is consistent with much of the wider evidence base.

Yet it is also important that we highlight the potential challenges with parental education as a measure of socioeconomic position, and the care that readers of this volume will need to exercise when interpreting the results. As we shall illustrate in more detail below, the distribution of parental education varies markedly across countries. Consequently, a different proportion of the population will be classified as coming from a “disadvantaged” background depending upon the country. Whether this is a desirable property of a family background measure is open to debate. The reason for such large differences across countries is likely due, at least in part, to differences in the prestige of vocational qualifications across nations. For instance, while some countries have well-established vocational routes leading to highly regarded educational qualifications (e.g., Germany) other countries do not (e.g., England and the United States). Hence, despite the usefulness of the ISCED framework, there nevertheless remain some questions over whether one is truly comparing like-with-like.

Another important issue with respect to parental education is measurement error. Many surveys, including the large-scale international assessments, rely upon young people to provide proxy reports of their mother’s and father’s education level. However, as Jerrim and Micklewright (2014) illustrate, agreement between parent and child reports is far from complete. Moreover, cross-national patterns of socioeconomic inequality can vary in important ways, depending upon whose reports are used. In a similar manner, missing data can also be a problem, either because children are unwilling or unable to answer questions about their parents’ education level, or because parents fail to complete the background questionnaire. Such issues may be particularly relevant for particular sub-groups. For instance, the educational qualifications of immigrants often do not easily fit into national reporting frameworks, and may, therefore, be particularly prone to non-response and miss-report. Each of the above, therefore, has the potential to impact upon the robustness of the conclusions that we can draw.

To conclude this section, we highlight these issues by illustrating the distribution of parental education across countries. Children have been grouped in low (ISCED 0–2), average (ISCED 3–5B), and high (ISCED 5A and above) parental education groups, along with those where this information is missing. Figures are presented data from the 2015 round of the Trends in International Mathematics and Science Study (TIMSS) fourth grade (age 9/10), TIMSS eighth grade (age 13/14), and PISA (age 15/16) studies. Note that in TIMSS fourth grade, information on parental education is reported by parents in response to a background questionnaire, while in TIMSS eighth grade it is reported by participating children acting as proxy respondents.

A number of gaps appear in the TIMSS results due to countries either not participating in the study (Germany, Spain, Finland, and the Netherlands for the eighth-grade sample) or not participating in the home background questionnaire where information on parental education is reported (England and the United States in the case of the fourth-grade sample). Moreover, even where countries do participate, there continue to be serious problems with respect to missing parental education data. For instance, more than half of the fourth-grade sample is missing information on parental education in Australia and the Netherlands, mainly due to parents not returning the background questionnaire. Likewise, more than a third of the eighth-grade sample in Australia, Canada, England, and Sweden are missing parental education data, due to children either skipping this question or reporting that they “don’t know” their mothers’ and fathers’ education level. Similar issues emerge with other socioeconomic background information in TIMSS, and in other international studies such as the Progress in International Reading Literacy Survey (PIRLS). There are, consequently, major limitations with using international resources such as TIMSS and PIRLS for studying the educational achievement of children from disadvantaged socioeconomic backgrounds, due to the serious risks posed by survey non-response.

In contrast, information on parental education is much more complete in PISA for most of the countries included in this volume (with the notable exception of Germany). Yet the distribution of parental education in the PISA study also helps to illustrate how the proportion falling into each of the different groups varies significantly across countries. For instance, whereas more than half of children report that at least one of their parents hold a degree in Finland, less than a quarter do in the Netherlands. Likewise, around a quarter of parents in Spain complete only basic education (ISCED level 0–2), compared to less than 5% of observations falling into this category in Finland, England, and Canada. Together, this helps to reiterate the point that, despite our use of comparable data and an internationally harmonized measure of educational qualifications, the size and composition of low parental-education groups across countries varies quite substantially.

3 Parent–Child Education Links in TIMSS and PISA

Despite the important caveats with the parental education measures in the large-scale international assessments documented above, it is nevertheless important to consider what they can tell us about the educational achievement of socioeconomically disadvantaged school children, and how this has changed over time. Table 1.2, therefore, illustrates the average mathematics scores of low education pupils according to the TIMSS/PISA 2015 studies (upper panel), and the size of the achievement gap relative to the high parental education group (lower panel). Lighter (darker) shading refers to “better” (“worse”) performance relative to other countries at a given age. Note that, when reading this table, comparisons should only be made between countries at a single age, as it is not possible to directly compare scores between the various PISA and TIMSS studies. In other words, direct comparisons can be made when reading Table 1.2 vertically, but not when reading across horizontally.

Starting with the top panel, Germany stands out as a country where the low parental education group performs relatively well compared to the other countries. However, readers should interpret this finding in light of Table 1.1, and the fact that this group is larger (and hence likely to be somewhat less selective) than elsewhere. Sweden, on the other hand, is a country where children with low educated parents have comparatively poor mathematics skills. This is particularly true at age 15/16, based upon the PISA data, where both Sweden and the United States have lower levels of mathematics achievement than other countries. Otherwise, relatively few consistent patterns emerge, with the magnitude of most cross-national differences being relatively small. For instance, at age 9/10, Australia, Spain, Finland, and Italy are separated by just 10 TIMSS test points—roughly equivalent to an effect size of 0.1 standard deviations or less. The same holds true for Australia, Canada, Spain, Italy, and the Netherlands at age 15/16 with respect to the low parental education group’s PISA scores. Our overall interpretation of the upper panel of Table 1.2 is that, on the whole, cross-national differences in the average mathematics skills of socioeconomically disadvantaged children are relatively small (at least with respect to the 10 countries included within this volume).

Table 1.1 The distribution of parental education across countries
Full size table
Table 1.2 The mathematics achievement of children from low parental education backgrounds
Full size table

The lower panel of Table 1.2 turns to the gap in achievement between the “low” (ISCED level 0–2) and “high” (ISCED 5A/6) parental education groups. Similar findings emerge with respect to Germany and Sweden; the achievement gap tends to be comparatively small in the former and large in the latter (with the exception of the TIMSS results at age 13/14). There are also perhaps some surprising findings; achievement gaps in Finland do not stand out as particularly small, and are actually larger than in some of the other comparator countries. Likewise, across all three surveys, the magnitude of the mathematics achievement gap in Italy does not stand out as being particularly large (though, as Table 1.1 has already illustrated, Italy also has a greater proportion of children within the low parental education category than elsewhere). The other notable result is that socioeconomic inequality is quite pronounced in Australia relative to the other countries according to results from the two TIMSS studies, but this is not the case in PISA. Overall, the lower panel of Table 1.2 does provide some evidence that social inequality in educational achievement does to some extent vary across our 10 countries of interest.

To conclude this section, we consider how the mathematics skills of children from low parental education backgrounds have changed over time. As the survey with the most complete data in terms of both country coverage and available information on parental education, we have based this analysis upon PISA data alone. These results can be found in Table 1.3, with the top panel referring to average mathematics scores of the low parental education group, and the lower panel the gap in achievement between children from low and high parental education backgrounds. Note that the shading should now read across the table horizontally (i.e., it aids with comparisons made within each country over time), with darker cells indicating “worse” performance (lower average scores and larger achievement gaps).

Table 1.3 How is the relationship between parent and child education changing over time? Evidence from PISA mathematics
Full size table

In terms of average scores, there has been some striking declines over the 12-year period considered. These have most notably occurred in the Netherlands, Finland, Canada, Sweden, and Australia, where there has been at least a 30 PISA test point drop between 2003 and 2015. (Note however that, for the Netherlands, response rates also tended to be lower in earlier waves of PISA, which could be having an impact upon the trends in this particular country.) In other nations, such as England, Spain, and the United States, the performance of this group has remained stagnant, with no obvious sign of progress having been made. Indeed, it is only really Germany where mathematics skills of the low parental education group has improved substantially over the last decade, with average scores in 2012/2015 around 40 points higher than in 2003/2006/2009 (this is roughly equivalent to a year of additional schooling; see Organisation for Economic Cooperation and Development, 2010, p. 167). Nevertheless, across the 10 countries considered, this seems to be the exception rather than the rule; rather than improving the mathematics skills of low socioeconomic status pupils over time, several of our 10 countries of interest are either showing no signs of progress or have gone into reverse.

Turning to the lower panel of Table 1.3, the gap in mathematics achievement between the high and low parental education groups seems to have increased in some countries, but fallen in others. Prominent examples where there has been a narrowing of achievement gaps include the United States, Germany, and (to some extent) Italy. Indeed, Germany has moved from having among the largest difference in children’s mathematics achievement between the high and low parental education groups to among the smallest, at least out of the 10 countries considered. Sweden, the Netherlands, and Finland have, in contrast, moved in the other direction. Whereas the relationship between parental education and PISA mathematics scores was relatively weak in these nations in 2003, it has become much stronger by 2015. Again, the situation in some of the other countries has remained largely unchanged (e.g., England, Spain) or with no clear pattern to the results (e.g., Australia). Nevertheless, the recent experience of Germany and Italy does suggest it is possible to raise disadvantaged children’s academic achievement and to narrow socioeconomic gaps in young people’s skills. It is unfortunate, however, that several Western countries actually seem to be moving in the opposite direction.

4 The Structure and Contents of This Volume

The analysis presented in the previous section has highlighted that, although the major international large-scale assessments such as PISA, PIRLS, and TIMSS have some advantages, they also have important limitations with respect to improving our knowledge of educational achievement among low socioeconomic status pupils. Several key issues stand out. First, there are significant issues with either missing parental education data, or potential measurement error due to children acting as proxy respondents for their parents, as previously discussed. Second, even in countries where data are available, the youngest pupils within international surveys are age 9/10, and almost at the end of their primary school education. Yet a wide body of evidence documents how large socioeconomic gaps emerge very early in life (Cunha, Heckman, Lochner, & Masterov, 2006), and can be observed as young as age 3 (Jerrim & Vignoles, 2013). Therefore, in many ways, the international surveys only start collecting data after the point when much of the damage has already been done. Third, relatedly, none of the international studies follow the same group of children over time. Consequently, although they may be able to provide a single snapshot of young people’s skills, they are unable to provide any information with regards to socioeconomic differences in developmental trajectories. Finally, as cross-sectional data, such studies can generally provide basic correlational evidence only. They are unable to reveal the wide set of factors likely to determine the poor educational outcomes of disadvantaged children, or provide much in the way of meaningful advice to education policymakers.

It is these limitations which have helped motivate the need for this volume. Rather than relying upon data from large-scale international assessments, this volume takes a somewhat different approach. Research teams from across 10 industrialized countries have been brought together to provide a series of case studies investigating socioeconomic inequalities in educational achievement from across a wide array of national contexts. This includes a diverse set of nations, ranging from those whose performance and equality according to PISA have been widely lauded (e.g., Canada, Finland) through to those whose international large-scale assessment scores are comparatively low, particularly among low parental education groups (e.g., Sweden, Italy). Although each chapter follows a similar structure, and utilizes parental education as the preferred measure of socioeconomic status (where possible), authors have also been free to exploit the full richness of the data and evidence available within their country, and have been encouraged to draw upon their detailed knowledge of their education system and subject expertise. This volume, therefore, seeks to provide readers with the latest empirical and policy evidence regarding how to improve educational achievement of young people from disadvantaged backgrounds, drawn from across the western world.

The volume is divided into three sections. Part I, including this introductory chapter, provides an overview of the topic of socioeconomic inequality and student outcomes, including methodological challenges associated with cross-cultural research on this issue. Particular attention has been devoted to explaining the strengths and limitations of PISA, TIMSS, and PIRLS for this purpose, including an investigation of what these resources tell us about the academic skills of young people from low socioeconomic backgrounds. The following chapter will consider some of the international trends related to the association between education policies and disadvantaged student populations.

Part II provides national profiles from scholars in nine countries (England, Germany, Italy, Spain, Netherlands, Sweden, Finland, Canada, and Australia). These countries have been selected because they represent Western industrialized nations that possess a range of datasets, many of which overcome some of the significant limitations with international achievement studies. These countries also vary widely in terms of their academic achievement results, education systems, and successes at addressing achievement gaps for socioeconomically disadvantaged student populations.

In order to promote a coherent approach and for the sake of comparability, each of the national profiles will be organized around four sections. An introductory section will provide a brief overview of the structure of compulsory school systems within a given country. The reader will gain an understanding of the general organizational and institutional features of the compulsory school system. This section also explains governance and administrative processes utilized to develop and refine education policies. The second section will describe the relative proportion of students who come from lower SES backgrounds within the national context. Although parental education will be the preferred measure (where possible), authors have been left to decide the most appropriate definition of the “low socioeconomic status” group using this variable within their own national context. Authors then outline the defining features associated with the disadvantaged student population, with particular attention given to explaining associated characteristics and mediating variables (i.e., gender, ethnicity, migrant status, single-parent households, and regional differences). Section 3 of the national profiles then describes the educational outcomes and choices of low SES children. These may include grades, grade repetition, graduation/dropout rates, aspirations, and standardized achievement scores, depending on the availability of data and relevance for the country context. Authors will also discuss the existing limitations of the available data and evidence within their particular national context. The final section of the national profiles then offers an analysis of the formulation, implementation, and effectiveness of education policies that are relevant for children with socioeconomic disadvantages.

Authors will also provide an explanation of the evolution of education policies as well as any refinements made to key institutional features (i.e., tracking provisions). Consequently, chapters will discuss the inherent linkages between children’s background, educational outcomes, institutional features, and policy developments within an overarching cultural, social, and political context. On the basis of this discussion, the readers will have a clear indication of what kind of policies work the best, and what should be the way forward for the country with regards to improving educational outcomes and closing the achievement gaps of lower SES student populations.

Note that the aim of these profiles is not for results from individual countries to be directly compared. Rather, we hope that they help to facilitate thought, discussion, and debate among readers, and lead policymakers to consider whether what has “worked” in other education systems might usefully be applied in other national contexts.

The final part of this volume (Part III, the conclusion) then synthesizes findings from the national profiles about the role of institutional features, education policies, and societal-level forces that influence educational inequities. The conclusion also proposes future areas of inquiry stemming from the national profiles.