Jump to ContentJump to Main Navigation
Unequal AttainmentsEthnic educational inequalities in ten Western countries$

Anthony Heath and Yaël Brinbaum

Print publication date: 2014

Print ISBN-13: 9780197265741

Published to University Press Scholarship Online: May 2015

DOI: 10.5871/bacad/9780197265741.001.0001

Show Summary Details
Page of
date: 30 May 2020

Ethnic Penalties and Premia at the End of Lower Secondary Education

Ethnic Penalties and Premia at the End of Lower Secondary Education

Chapter:
(p.63) 3 Ethnic Penalties and Premia at the End of Lower Secondary Education
Source:
Unequal Attainments
Author(s):

Anthony Heath

Catherine Rothon

Publisher:
British Academy
DOI:10.5871/bacad/9780197265741.003.0003

Abstract and Keywords

This chapter investigates the grades and test scores of second-generation minorities at the end of compulsory schooling (around age fifteen). We document the differences in overall achievement both between minorities and between countries, showing that Chinese and some other Asian groups out-perform the majority group while Turkish, North African, Caribbean, Sub-Saharan African groups and many European minorities all perform less well. Whilst the disadvantage of many European minorities can largely be explained by their disadvantaged socioeconomic backgrounds, socioeconomic background does not explain the disadvantages experienced by minorities with Turkish and North African backgrounds, meaning that these groups experience significant ‘ethnic penalties’ even after controlling for their parent's socioeconomic circumstances. Conversely, the Chinese experience substantial ‘ethnic premia’. We then investigate explanations for these ethnic premia and penalties, concluding by considering the extent to which different countries offer more or less favourable environments for the educational success of the children of immigrants.

Keywords:   educational inequality, ethnic minorities, test scores, ethnic penalties, ethnic premia

Introduction

IN THIS CHAPTER WE FOCUS ON YOUNG PEOPLE'S educational performance towards the end of lower secondary education, at around age 15–16, which in most of our countries is towards the end of compulsory education. There are several reasons, both theoretical and practical, for beginning with this stage. In most, although not all, of our countries compulsory schooling ends at age 16 and in many educational systems, as we saw in Chapter 2, young people then have to make important decisions whether to stay on at school or not, and what kind of course to follow in upper secondary education.

We also know from a great deal of previous research that young people's decisions about which course to pursue in upper secondary education depend to a very substantial extent on how well they have performed educationally in lower secondary school. Performance at around age 15 is thus likely to have an important or indeed decisive influence on future educational trajectories. On the standard rational choice theories of educational decision-making, the decision whether or not to continue will depend in part on the perceived probability of success at the next stage. So students who have performed badly might be quite rational if they were to decide not to continue. Of course, the issue is much more complex than that since students' decisions how hard to work in lower secondary education might be influenced by their eventual plans for education in upper secondary or tertiary education, and there are many other factors apart from educational achievement at age 15–16 that influence continuation decisions. But educational research has always found a strong relationship between test scores at this age and continuation decisions (e.g. Jackson et al. 2007).

(p.64) Previous research has also shown that there are quite large differences in the educational success of children from different ethnic origins at age 16 in Western countries. Broadly speaking, the children of migrants from East and South Asia perform rather well, often outperforming their peers in the majority group, while the children of migrants from less-developed countries, especially from Turkey or North Africa, tend to be less successful (e.g. Kao & Thompson 2003; Heath et al. 2008). There thus seems to be considerable variation in the success or otherwise of the children of immigrants in Western countries.

There are many theories about the reasons why young people from particular ethnic origins might perform better or worse on tests during lower secondary education. In Western countries social class has repeatedly been found to be associated with the test performance of children from the majority groups (e.g. Marks 2005a), and there is every reason to expect social class to be important in explaining minority performance too, particularly as some minorities, such as Turkish ‘guest workers’ in Belgium, Germany or the Netherlands, were largely recruited into low-skill jobs at the bottom of the class hierarchy (Castles & Kosack 1973). We thus expect to find that part at least of the observed gross disadvantages of ethnic minority students whose parents came as guest workers can be explained by their disadvantaged socioeconomic circumstances. However, as we shall see in the next section, there is considerable disagreement about the extent to which, and for which minority groups, socioeconomic background can explain ethnic inequalities. There are also some major issues in determining whether social class and related characteristics mean the same thing among the migrant parents as they do for native families, and whether social class background is equally critical in different Western countries. We turn to these problems later in this chapter.

One major task of this chapter is to determine as best we can how far socio-economic background can explain the pattern of ethnic minority advantage and disadvantage, and in turn to identify which minority groups experience ethnic ‘penalties’ or ‘premia’ when compared with members of the majority group in similar socioeconomic circumstances. In other words, are there ethnic-specific advantages or disadvantages over and above those arising from general processes of social class stratification?

Our general expectation is that the more similar the origin and destination countries are culturally and institutionally (both in terms of language and in terms of schooling systems and practices), the less likely it is that the children of migrants from the origin country will experience disadvantages over and above those due to their disadvantaged socioeconomic circumstances. In other words, we would expect minorities from culturally similar countries (for example Norwegians or Danes in Sweden) to perform similarly to their majority-group peers with similar social class backgrounds. In these cases we (p.65) expect ethnic disadvantage (or advantage), after controlling for social class background, to be minimal. Hence, following Heath & Brinbaum's (2007) theory of cultural dissonance, we expect to find that ethnic penalties in addition to socioeconomic penalties occur primarily among groups who come from more dissimilar national origins.

Previous research has also suggested that there are considerable differences between destination countries in the educational success of their ethnic minorities. In particular it has been suggested that the children of migrants in the classic countries of immigration, such as the USA and Canada, do better educationally than those in European countries which have only become countries of immigration more recently (e.g. see OECD 2006). This has led some writers to conclude that the classic countries of immigration have developed better mechanisms for integrating minorities. Various mechanisms have been suggested such as the ease of gaining citizenship, an inclusive ‘civic’ conception of national identity rather than an exclusive ‘ethnic’ one, the absence of racial prejudice and discrimination, or multiculturalism in the school curriculum. Reitz (1998) has used the expressive term ‘warmth of the welcome’ to capture this idea (see also Portes & Zhou's (1993) discussion of contexts of reception. We might perhaps expect more inclusive or multiculturalist countries to exhibit less ‘institutional racism’ within schools and hence to elicit more positive educational engagement on the part of minority children.

While descriptively there can be no doubt that the children of immigrants in Canada perform rather well in comparison with their majority-group peers, while the children of immigrants in Belgium and Germany do much less well, it is less clear how much this has to do with the integration policies of the destination countries. A major issue is that the ethnic composition of the second generation will differ very considerably in different countries. Thus in Belgium and Germany, as Lessard-Phillips and her colleagues noted in Chapter 2, the largest immigrant groups are those from less-developed countries such as Turkey or North Africa whereas in Canada there are larger numbers from the highly achieving East Asian groups. A great deal of the cross-national differences may therefore be due to these ‘compositional’ differences and not to integration policies or practices in the countries of destination

Another possibility is that it is not so much the integration policies as the immigration regime that accounts for the cross-national variation. As explained in Chapter 2, Canada has long had a ‘point system’ of entry permits, and this is likely to mean that migrants to Canada will be more ‘positively selected’ than those migrating to other Western countries (Feliciano 2005a). They will typically be quite highly educated, and they may also be unusually determined and ambitious, especially if they come from less-developed countries and have had to overcome major barriers in order to reach the West. Conversely, migrants who came as guest workers may well be neutrally or even negatively (p.66) selected, often being drawn from poorer areas of the origin country and typically coming through national programmes.

It should be emphasised that the degree of positive selection need not mirror the socioeconomic standing of the migrants in the country of destination. Since the educational and occupational profiles of the countries of origin and destination (especially in the parents' generation) can be very different, it is perfectly possible for a migrant group to appear socioeconomically disad-vantaged when compared with the majority in the country of destination, and yet at the same time to be relatively advantaged, i.e. positively selected, in comparison with non-migrants in the country of origin.

While the degree of positive or negative selection will apply to the migrant (i.e. parental) generation, it can also have major implications for the success of their second-generation children. We assume that positively selected parents will be particularly ambitious for their children to succeed educationally, and indeed migration is often a ‘family project’ with the aim of securing a more promising future for their children (Zeroulou 1988; Van Zanten 1997; Heath et al. 2008). Feliciano's research on the second generation in the USA confirms that the degree of positive selection of migrant groups can help explain the differences in their children's levels of educational attainment (Feliciano 2005b).

A helpful way to look at the range of issues that arise when comparing ethnic inequalities in different countries (following Van Tubergen et al. 2004) is to think of them in terms of ‘origin effects’, ‘destination effects’ and ‘community effects’. Origin effects will arise if there are distinctive features of an ethnic minority which lead it to do well or badly in whichever destination country its members have migrated to. This could occur if the origin country has a less-developed educational system which has not prepared its citizens for life in a more advanced Western society. In contrast, destination effects could occur if the Western country holds out a warmer (or cooler) welcome to migrants, whichever country they come from. By community effects we refer to those arising from a particular combination of origin and destination countries. Linguistic similarity is a typical example of this; for example, the Irish migrating to the USA will speak English just like the majority group, whereas the Irish migrating to Germany will have to learn a new language. Linguistic similarity thus depends on the combination of origin and destination, and is not a property of either origin or destination country on its own. Similarly, post-colonial migrants from Jamaica to England and Wales may find adjustment easier than do migrants from the Congo, whereas the situation may be reversed for migrants from the same two countries to Belgium.

Our central questions, therefore, are which origin groups experience ethnic penalties or ethnic premia, which countries provide the most favourable (p.67) environments for minorities' educational success, and whether there are any distinctive patterns that apply to particular ethnic groups in particular countries.

Previous research

There has been some previous cross-national work, primarily using data from the OECD's Programme for International Student Assessment (PISA), which has also attempted to explore origin and/or destination effects. The main PISA report, Where Immigrant Children Succeed (OECD 2006), essentially explored cross-national differences (destination effects) drawing on the 2003 round of PISA which focused on achievement in mathematics. Because of the relatively small sample sizes, the official OECD report simply distinguished first-generation from second-generation ‘immigrant’ students and did not explore ‘origin’ effects. That is, they pooled all second-generation minorities into a single category, ignoring the substantial differences between them. Broadly speaking, the study concluded that the first generation generally showed very large disadvantages in mathematics achievement, disadvantages that persisted in many countries in the second generation although they were typically reduced in magnitude. The differences between ‘immigrant’ and ‘native’ students were found to be most pronounced in Austria, Belgium, Denmark, France, Germany, the Netherlands and Switzerland, while ‘immigrant’ students performed at similar levels to ‘native’ students in the traditional settlement countries of Australia, Canada and New Zealand.2

Marks (2005b) has also used the PISA data to explore cross-national and generational differences in reading and mathematics test scores (finding very similar results for reading and maths). Within the second generation, Marks distinguished young people who spoke the relevant national language at home (e.g. German-speakers in Germany) from those who spoke a non-national language (e.g. Turkish speakers in Germany). While he found a general pattern for speakers of the national language to score better on test scores than speakers of other languages, the results varied considerably from one country to another, probably reflecting the precise composition of the two categories. (For example, in the Netherlands, Norway, Switzerland and Sweden there was little difference between the majority and minority-language speakers, while there were enormous differences between the two groups in the USA in their (p.68) performance.) Marks also found, similarly to the OECD study, that among the minority-language speakers differences between second-generation and majority-group students were largest in Austria, Belgium, Denmark and Germany, but were also very large in the USA. They were smallest in Australia and Canada, but not in New Zealand.

Again in a similar vein to the OECD study, Marks found that socioeconomic background accounted for much or all of the gap between second-generation minority and majority students in several countries (France, Germany and the Netherlands, for example) but only around half the gap in Belgium, Switzerland and England and Wales, and even less of the gap in Norway and Sweden. Why this should be the case was not explored further but, as we shall see when we do our own analysis, it probably reflects the different socioeconomic profiles of the migrants in the different countries.

A further major series of studies of cross-national differences using PISA data on test scores have been carried out by Jaap Dronkers and his colleagues (Levels & Dronkers 2008; Levels et al. 2008; Dronkers et al. 2012). They go rather further than both the OECD (2006) and the Marks (2005b) studies in trying to distinguish ethnic-origin differences within the second generation. They differentiate according to broad groupings of world regions (primarily basing their coding on the country of birth of the students' parents). They distinguished, for example, Northern European, Western European, Eastern European, Southern European, North American, South and Central American, North African, Southern and Central African, Western Asian, Eastern Asian, Southern Asian and Southeast Asian regional origins. This was a big advance on the distinctions that Marks and the OECD were able to make although, to be sure, some of these broad regional groupings are still very diverse in their ethnic composition. Nevertheless, it enabled Levels & Dronkers (2008) to make a much more systematic comparison of origin and destination effects than had been previously attempted. Unfortunately, however, this came at a major cost: they were unable to make these distinctions in many important immigrant-receiving countries and thus they had to drop from their analysis Canada, France, the Netherlands, Sweden, Norway, England and Wales and the USA. There is thus little overlap with the countries which are the focus of our study.

Levels & Dronkers' main conclusion was that there were both origin and destination effects. With respect to origin, they found that second-generation minorities from Southern Europe, Southern and Central America, North Africa and Western Asia had substantially lower maths scores than comparable members of the majority groups. Controls for social background generally tended to reduce the magnitude of these origin differences, but typically by only a third or less—much less than in the analyses conducted by Marks. For all these groups the disadvantages remained substantively large and (p.69) statistically significant after controls. (Only for South Asians were the disadvantages wholly explained by socioeconomic background, while the advantages of the Northern European and North American groups were also largely explained by socioeconomic background.)

With respect to destination, Levels & Dronkers also found that some countries seemed to provide less favourable environments for immigrant integration. Particularly notable in this respect were Belgium, Denmark and Switzerland. ‘We could argue that these new migration societies (Belgium, Denmark and Switzerland) have strong insider/outsider distinctions, possibly strengthened by their social security systems and their labour market regulations, and that this insider/outsider distinction is not yet blurred by a tradition of migration, former colonies or lost territories and a volatile history’ (Levels & Dronkers 2008: 1423).

Perhaps most importantly, Levels & Dronkers emphasise the crucial need to distinguish the different origins of the minorities:

Analysing migrants' integration in host societies without properly taking into account these origin effects will indeed lead to flawed results. Depending on the composition of the migrant population in a certain society, results may be too optimistic or too pessimistic. One could be tempted to explain test-country effects by certain policies (selective immigration policies in Australia) or educational systems (early selection in Germany) instead of the various compositions of migrant populations (2008: 1422).

We shall do our best to take such ‘origin effects’ into account in our own work.

In summary, then:

  • There is an emerging consensus that origin differences matter, and that it is essential to disaggregate the second-generation into specific origin groups because of the great variation in the different minorities' educational achievements.

  • There is also an emerging consensus that these origin differences can in part be explained by socioeconomic background, although how much can be explained and for which groups remains uncertain.

  • There is an emerging consensus that destinations matter, and several scholars have hypothesised that ‘new’ migration societies provide less favourable environments for immigrant integration than the more established or classical countries of immigration, although the mechanisms involved are still unclear.

  • A major unresolved issue, however, is whether these apparent destination effects are actually due to the warmth of the welcome and better arrangements for integration, or are simply the consequence of immigration policies of varying degrees of selectivity.

In this chapter we begin by documenting the ‘gross differences’ between different second-generation minorities in our ten Western countries. We then (p.70) establish the extent to which socioeconomic background can account for these gross differences, and for which groups in which countries ethnic penalties or ethnic premia remain after socioeconomic controls. Finally we explore how far these ethnic penalties or premia can be explained by processes of positive or negative selection and which countries provide more or less favourable environments for minority education after taking selectivity into account.

Gross differences

Our first step, then, is to document the gross differences in educational achievement at around the age of 15. By ‘gross differences’ we mean the overall differences in test scores before including controls for socioeconomic background. In Table 3.1 we show these gross differences for the main ethnic groups that we are able to distinguish in our ten countries.3

To improve the comparability of these measures we standardise them, setting the scores of the majority population at zero for ease of comparison. We then show the coefficients for the various minority groups which we have been able to distinguish. Essentially these coefficients show us, in a way that is reasonably comparable between countries, how much better or worse the various ethnic minorities are performing relative to their majority groups. Positive scores indicate that overall the minority scores more highly than the majority group, whereas negative scores indicate the reverse. These coefficients can be interpreted in terms of the area under the normal curve. That is to say, a coefficient of +0.50 indicates that the minority's average score is approximately at the 70th percentile; conversely a coefficient of −1.00 indicates that the minority's average score is around the 15th percentile.

We make the finest distinctions between origin groups that the data permit. In many cases we have been able to distinguish specific national-origin minorities. For example, in France the Portuguese second generation can be distinguished as a separate group, while in Germany we can distinguish separately Italians and Greeks. In other cases, however, because of small sample sizes, we have had to group origins into broader categories such as Iberian or Southern European categories. We need to remember, too, that in some cases (p.71)

Table 3.1. Gross differences in achievement between minority and majority groups in the country of destination: standardised scores

Country of origin: Asian

Country of destination

Belgium

Canada

England and Wales

Finland

Germany

Netherlands

Sweden

Switzerland

USA

Chinese/East Asian

+0.35

+0.54

+0.05

+0.34

+0.56

Indian/South Asian

+0.18

+0.22

+0.11

Southeast Asia

+0.25

−0.01

Filipino

−0.01

Pakistani

−0.23

Bangladeshi

−0.07

Iranian

+0.06

Iraqi

−0.31

Turkish

−1.20

−1.06

−0.53

−0.43

−1.05

West Asian (other)

−0.41

Country of origin: European

Canada

Finland

France

Germany

Sweden

Switzerland

USA

West European

+0.11

−0.08

 

−0.27

Finnish

−0.37

 

Danish

−0.43

 

Norwegian

−0.30

 

Polish

−0.18

+0.04

 

Former Soviet Union

−0.21

−0.59

 

Former Yugoslavian

−0.90

−0.75

−0.26

−1.21

Albanian/Kossovar

−1.87

Italian/South European

−0.37

−0.77

−0.26

−0.85

Portuguese/Iberian

−0.30

−0.91

Greek

−0.45

 

Caribbean

−0.22

−0.46

+0.01

−0.33

−0.49

North African

−1.03

−0.63

−0.47

−0.73

−0.29

Sub-Saharan African

−0.16

−0.53

−0.45

 

−0.30

Chilean

−0.59

Other South American/Latino

−0.27

−0.66

Mexican

−0.87

Note: coefficients in bold are significant at the 0.05 level. In the USA, we have allocated the second-generation white group to the West European row and the black group to the Caribbean row.

(p.72) (p.73) national origin should not be equated with ethnic background. This is most notably the case in Sweden where many of the young people whose parents came from Turkey will actually be ethnic Kurds.

To ease presentation we have split Table 3.1 into three panels, distinguishing Asian, European and African/American origin countries respectively. The first impression from Table 3.1 is that the Asian-origin minority groups in the top panel are the most successful, many scoring well above their respective majority groups. In contrast, the European-origin groups in the second panel are surprisingly unsuccessful with many significant negative scores. On average, though, it is the groups in the third panel, from Africa and the Americas, who have the lowest achievements. Within each of the panels of the table, however, we see large differences both between ethnic groups (origin differences) and between countries of residence (destination differences). We next explore these differences in more detail.

Asian-origin groups

First, we see that the most successful group is the Chinese, who invariably have higher average test scores than the majority groups. However, it is also worth noting that Chinese are present in sufficient numbers for analysis only in Canada, England and Wales, Finland, Sweden and the USA. In these five countries of destination, the second-generation Chinese students outperform the majority groups by around one third of a standard deviation. It is striking how similar the coefficients are in the different countries, despite the differences in the measures of achievement used (PISA test scores of literacy in Canada, overall GCSE points in England and Wales, overall teacher-assigned grades in Sweden). To get some sense of what this means in practice, we typically find in these countries that the gap between students from higher and intermediate social class backgrounds (e.g. someone from a professional or higher managerial social background compared with someone from a white collar or small business background) also comes to around one third of a standard deviation.

Second, we see that Southeast and South Asians such as minorities of Indian background (apart from Pakistanis in England and Wales) are relatively successful on these various tests, although not as consistently so as the Chinese, and their advantage over the majority group is also typically somewhat smaller than that of the Chinese. Again, these groups are found in a similar range of countries as the East Asians.

Apart from minorities of Iranian and Iraqi background in Sweden, whose performance is quite close to the Swedish average, our main West Asian group is the Turkish minority. As many previous scholars have shown, this group scores well below the destination-country average in all five countries where (p.74) they are present in sufficient numbers for analysis. However, it is important to recognise that these are a somewhat different range of countries from those where our most successful group of Chinese is to be found: the Turkish are found only in Belgium, Germany, Netherlands, Sweden and Switzerland. So it is premature to draw any firm conclusions about either origin or destination effects.

European-origin groups

Turning next to the European-origin groups, we find that these minorities are surprisingly unsuccessful in these tests and, although their results are quite diverse, most European minority groups perform less well than the majorities in the countries of destination. This result applies in Canada, Finland and Sweden, as well as in France, Germany and Switzerland. It is especially interesting that minorities in Sweden from neighbouring highly developed Nordic countries such as Denmark, Finland and Norway perform significantly worse than the Swedish majority. This strongly suggests that cultural similarity or dissimilarity is far from the whole story.

To be sure, many of these European-origin groups are the descendants of guest workers—relatively low-skilled migrants from southern Europe recruited to meet the labour shortages in the more flourishing north European economies of Belgium, Netherlands and Germany. And in the case of Sweden many of the European migrants would also have been labour migrants. Their parents were in this sense likely to be neutrally or negatively selected, and the second-generation students represented in Table 3.1 will typically have parents in low-skilled jobs. We therefore expect many of these differences between European groups to be explained by social background. We turn to this later in the chapter.

Groups with origins in Africa and the Americas

The third panel of Table 3.1 looks at the groups who in Canada would be termed ‘visible minorities’. Again, almost all the coefficients are negative, indicating that these groups perform less well than their respective majority groups, but the magnitude of the disadvantage is on average no greater than that for the much less visible groups of European origin in the middle panel of the table. The black groups, though below the majority-group average, are interesting for not being as far behind as those from North Africa.

(p.75) Net differences after controlling for the socioeconomic background of the second generation

Our next step, then, is to control for social background. As we noted earlier, in most Western countries we find that performance on test scores is strongly related to social background, particularly to parental social class and educational level. In some of our countries, such as England and Wales, parental social class appears to be the more important driver of children's educational performance, while in others such as the Netherlands it is the parents' own education that seems to matter most. In all our countries, however, one or other of these factors has strong and statistically significant relationships with young people's test scores. It therefore makes sense to see if the low test scores of the more disadvantaged minorities, such as some of the European groups and those from Africa and the Americas, can be explained by their disadvantaged socioeconomic backgrounds.

To be sure, there are a number of problems when attempting to control for minorities' social background. First, there are well-known differences between countries in the strength of the association. Typically, social background seems to be more weakly associated with children's educational achievement in Canada, Finland and Sweden than it is in Belgium, Germany or Switzerland. This means that we have to be sensitive to these cross-national differences when exploring how far social background can account for minorities' educational performance.

Second, there is some evidence that the first, migrant generation tend to experience downwards mobility, and to obtain lower returns on their education, when they migrate to another country. This suggests that their social class may be artificially low and may not give an appropriate indication of the level of educationally relevant resources in the minorities' homes. How important this is depends to some extent on one's theory of why parental class matters. If one's theory is that it is the lack of economic resources that accounts for the effects of family social class, then there is no problem: unskilled workers will lack economic resources irrespective of the particular reasons for their location in these kinds of job. However, if one's theory is that social class operates via parental socialisation (which is the general consensus when considering the ‘primary’ effects of social class), then there is more of a problem.4

Third, partly as a consequence of these problems, minorities' social backgrounds may have a weaker association with their children's educational achievements than they do in the Western majority groups. This is an (p.76) assumption which we can actually test, and our research suggests that, with one or two notable exceptions such as the Chinese, social background has rather similar relationships with children's test scores among the minority groups as it does among the majority groups (see Heath & Brinbaum 2007). This suggests that it is not entirely misguided to attempt to control for minorities' social background. Indeed, it could also be argued that it would be misguided not to control, since we know that some migrant groups in Europe, particularly those of the guest workers, were specifically recruited from relatively disadvantaged backgrounds in their countries of origin to fulfil low-skilled work in Western countries. It would be remarkable if their disadvantaged social backgrounds did not go some way towards explaining their children's low test scores, just as it does among disadvantaged groups within the majority populations. We therefore undertake controls for socioeconomic background (parental class, parental education and family composition) in order to see whether this can account for the observed gross differences in test scores between minority and majority groups.

Unlike previous researchers in this field, we have the statistical power to run country-specific models rather than assuming that socioeconomic background has similar importance in all ten countries. That is to say, we are able to take account of the fact that socioeconomic background has a weaker effect on educational achievement in some countries than it does in others. In all ten of our countries, family composition, parental education and parental social class are significantly associated with the students' test scores but the size of the effects in our samples does differ across countries in ways that are familiar from previous research in these countries (e.g. Breen et al. 2010). In particular, we find that the effects of social background are slightly smaller in Canada and Finland than they are elsewhere, and they are slightly larger in Belgium and Germany. This in turn means that we do not expect controls to make quite as much difference in the case of Canada or Finland as they do in the case of Germany.

When controlling for social background we use linear regression (OLS) models since these are the appropriate techniques when modelling normally distributed outcomes such as test scores. What we expect to find here is that the magnitude of the negative ethnic coefficients will have declined considerably in those cases where the parental generation had lower socioeconomic profiles than the majority group in the country of destination, which will typically have been the case for the various guest worker groups. However, we should note that, if the gross difference was positive and the minority in question had a disadvantaged socioeducational background, then the effect of the controls will actually be to increase the ethnic ‘effect’. And if minorities have similar social class and educational backgrounds to the majority group in their country of destination, then no amount of socioeconomic controls will (p.77)

Ethnic Penalties and Premia at the End of Lower Secondary Education

Figure 3.1. Gross and net differences in achievement between minority groups of Asian origin and the majority groups in the country of destination: OLS coefficients.

(p.78)
Ethnic Penalties and Premia at the End of Lower Secondary Education

Figure 3.2. Gross and net differences in achievement between minority groups of European origin and the majority groups in the country of destination: OLS coefficients.

(p.79)
Ethnic Penalties and Premia at the End of Lower Secondary Education

Figure 3.3. Gross and net differences in achievement between minority groups of African and South American origin and the majority groups in the country of destination: OLS coefficients.

(p.80) do anything to explain any educational disadvantage that they experience. More generally, controls for social background will only explain (in a statistical sense) the gross ethnic differences if (a) the minority differs from the majority with respect to the control variables and (b) if the control variables are predictive of the outcome, in our case educational test scores.

The dark grey bars in Figures 3.13.3 show the gross differences while the light grey bars show the (statistically significant) net differences after controlling for parental education, parental class, gender and family composition. Where a coefficient was not significantly different from zero, the bar is left unshaded. As with Table 3.1, we divide the results into three panels.

Asian-origin groups

Beginning as before with the East Asians, we see that in fact the controls for social background make little difference to the magnitude of East Asian success in the tests. In some cases (e.g. Finland) the ethnic coefficient actually becomes larger after controls, whereas in others (such as the USA and Canada) they become somewhat smaller. This is because in the USA and Canada the East Asian parents were actually rather better educated than the majority-group parents, and hence their high levels of education can partly explain their children's educational success. In England and Wales, Finland and Sweden, on the other hand, the East Asian parents were slightly less well-educated than the majority group, and hence after controls the ‘effects’ become even greater. Nevertheless, the overall picture remains very close to what it was before and the controls have made little real difference. The picture is one of very substantial East Asian success in all five countries where we can make the comparison, with net ethnic coefficients of around +0.4.

The South and Southeast Asians came somewhat behind the East Asians with respect to gross differences in achievement, and this remains the case after controls for social background. However, the story is strikingly different for the groups with a Turkish background. In every case, the net disadvantage is sharply reduced from the gross disadvantage, and in the case of the Turkish minority in Sweden almost disappears. This is of course because these groups have particularly disadvantaged socioeconomic backgrounds. But even after controls, young people of Turkish heritage still have very low scores, with net coefficients as high as −0.6 in some cases (though effectively zero in Sweden, where many of this group may actually be Kurdish refugees rather than Turkish guest workers as in the other countries).

European-origin groups

(p.81) The second generation of European heritage had surprisingly low scores on the tests shown in Table 3.1. We now see that many of these gaps, especially for those whose parents came from Western or Southern Europe, can be explained by social background. In every single case we see that, where there was gross disadvantage (a negative coefficient) in Table 3.1, this has been substantially reduced or entirely eliminated after the controls. Only for a few minorities, and not in all countries, do we find substantial continuing net disadvantages. Net disadvantages remain among the Italians in Germany and Switzerland (but not in Canada or Sweden), the minorities from the former Yugoslavia in Finland, Germany and Switzerland (but not in Sweden), and the Portuguese in Switzerland (but not in France).

Two comments are perhaps warranted. First, the South European groups who continued to have large net disadvantages in test scores are predominantly the children of guest workers, whereas those groups who have more or less achieved parity with the majority groups tend to be the children of voluntary migrants. Second, we should note that these groups of European heritage young people are coming from relatively similar countries to those where they now live. There will be no major cultural or institutional barriers to overcome. We can also have confidence that, in the case of the European migrants, our measures of socio-educational background are comparable to those of the destination countries.

The one major exception to this story is that of the Albanians in Switzerland, where huge ethnic penalties remain after controls for social background. This is predominantly a Muslim group who will therefore be culturally rather different from the Swiss majority groups.

Minority groups from Africa and the Americas

Table 3.1 showed that some, although not all, of the groups from less-developed countries had very large gross disadvantages in test scores; some as high as −1.0. In most cases these are reduced by the controls (by around half), especially among the guest workers from North Africa. Thus young people of North African heritage seem slightly less disadvantaged after controls for socioeconomic background, with net coefficients around −0.3.

The picture is somewhat better for young people of Caribbean heritage where the net coefficients are around −0.2 (with around half the gross gap being explained by the controls). And the other minorities from less-developed countries, especially Sub-Saharan Africa, tend to have even smaller net coefficients.

(p.82) In summary, then, we find that:

  • controls for social background do not do much in the case of East Asians, who continue to outperform the majority groups with net ethnic coefficients (ethnic premia) around +0.4.

  • after controls, South and East Asian coefficients either increase or remain the same, at around +0.3.

  • in the case of groups of European heritage, controls generally explain half or more of the gross gaps, with the net coefficients averaging around zero for North and West Europeans but remaining very substantial, around −0.5, for some East and South European groups in some countries.

  • controls explain around half the gap for ‘visible minorities’ from less-developed country origins, but still leave large net disadvantages with coefficients around −0.6 for young people of Turkish heritage and somewhat smaller net disadvantages, around −0.2, for the black groups.

A somewhat clearer picture of destination differences is also beginning to appear. Net disadvantages are typically smallest in Canada, the USA, Sweden, France and England and Wales, and remain largest in Belgium, Germany and Switzerland, with the Netherlands and Finland in between (interestingly, a pattern rather similar to the one found for the labour market; see Heath & Cheung 2007).

Overall then, the controls for social background have changed the magnitude but have not fundamentally changed the profile of ethnic advantage and disadvantage in test scores, though for some (but by no means all groups) they have explained around half the gross disadvantages. And the extent to which the controls explain the disadvantages is, unsurprisingly, closely in line with the extent to which the parental generation occupied disadvantaged socio-educational situations.

Nevertheless, although the range of net coefficients is smaller than that of the gross coefficients, there is still very substantial variation to explain, both positive and negative, and both between origin groups (the ethnic minorities) and between destination countries. To explain these remaining ethnic premia and penalties, a promising line of enquiry, which has been suggested by scholars such as Feliciano, is the degree of positive or negative selection involved in the migration of the first generation. We turn to this in the final section of this chapter.

Selectivity

The controls for socioeconomic background simply take account of the extent to which the ethnic minorities differ from the majority group in the country of destination in their social profile. However, as Laurence Lessard-Phillips and (p.83) her colleagues explained in Chapter 2, minorities may also differ from non-migrants in the country of origin and in this sense may be ‘positively’ or ‘negatively’ selected. The degree of positive or negative selection may thus be an additional explanation for educational success, or otherwise, of the children over and above their socioeconomic profile in the country of destination.

In particular, positive selection is a potential explanation for the ethnic premia obtained by some of the East and South Asian groups. Although India and mainland China are now developing quite rapidly, at the time when many of the parents of our second generation migrated they were still relatively poor and predominantly agricultural economies. The people who migrated were thus quite exceptional, as Chapter 2 showed, in their educational levels and were probably exceptional in the drive and determination that enabled them to overcome the many obstacles to migration to the West. The parents' drive and determination may thus help explain why their children have been so successful. Controlling for socioeconomic status in the Western country of destination does not take this exceptional drive and determination into account (since the parents may well have suffered discrimination or downwards mobility on arrival in the West), but we can use the selectivity index described in Chapter 2 to test this hypothesis. Selectivity could also be important in explaining some of the successes of young people whose parents were refugees, since Chapter 2 demonstrated that refugees also tended to be positively selected (contrary to conventional wisdom).

Positive selection is thus most likely to be able to explain, in part, the ethnic premia of the East and South Asian groups and some of the refugee groups. In contrast we do not expect it to play any role in the case of migrants within Western Europe, such as the various Nordic groups who migrated to Sweden. This is because most Western European countries tend to be at similar levels of economic development to each other with fairly similar socioeconomic profiles. Such groups who are disadvantaged socioeconomically in the country of destination were almost certainly equally disadvantaged in the country of origin, and thus we do not expect selectivity to add anything to the controls for socioeconomic background that we have already included.

It is also quite possible that taking account of selectivity will work in the opposite direction for some disadvantaged groups. For example, if some of the North African groups were positively selected, then this would imply that their educational difficulties in the West are even more serious than our analyses so far would have led us to expect. In other words, given their parents' drive and determination, these young people might be expected to have done rather better, not worse, than their parents' disadvantaged socioeconomic profiles in the West would have predicted.

As we noted earlier, positive selection could also be important in explaining some of the cross-national differences, especially the apparent success of (p.84) minorities in Canada. The Canadian points system is likely to have resulted in migrants who were on average positively selected, and it is quite conceivable that it is the restrictive Canadian immigration regime rather than its inclusive integration policies that accounts for the educational success of their second-generation minorities.

We will undertake a formal statistical examination of the overall effects of selectivity on the full range of educational outcomes in Chapter 9. In this section, however, we conduct an exploratory analysis, focusing in particular on the extent to which positive selection can explain some of the ethnic premia and the cross-national differences which we have described in this chapter.

In Figure 3.4 we plot the net disadvantages after controls against the NDI measure of positive or negative selection described in Chapter 2. As we can see, there is a great deal of scatter (reflecting in part the measurement error involved with both measures but also reflecting the large origin differences that we have described earlier) but there is nonetheless a fairly clear relationship, with positive selection being associated with ethnic premia and negative selection with ethnic penalties. (The detailed statistical analysis in Chapter 9 confirms that this bivariate relationship is statistically significant.)

The first thing to notice is that the line of best fit passes beneath the origin point. What this means is that young people from an ethnic group which was neutrally selected (which was often the case for voluntary and post-colonial migrant groups) will nevertheless be disadvantaged educationally when compared with their Western peers in similar socioeconomic circumstances. This is an important result, since it highlights the impediments to educational success in Western educational systems for the children of voluntary migrants. They do not appear, at least at this stage of the educational career, to be competing on a level playing field.

We also find that, as expected, positive selection can help to explain some of the East and South Asian education success. In Canada, England and Wales and the USA selectivity can explain about a quarter of the ethnic premia (although selectivity has little explanatory role in Sweden where these Asian groups were not especially selected). Negative selection can also explain some of the ethnic penalties experienced by the Mexicans in the USA, North Africans in France and Caribbeans in England and Wales.

We can also use the scatterplot to re-estimate national differences. If the ethnic coefficients for a particular destination country are above the line of best fit, this suggests that minorities in that particular country are generally doing better than would (p.85)

Ethnic Penalties and Premia at the End of Lower Secondary Education

Figure 3.4. The relationship between test scores (net effects) and selectivity.

(p.86) be expected given their degree of selectivity (and their socioeconomic profile). Conversely if the ethnic coefficients are generally below the line, this suggests that minorities are doing worse than would be expected, and that the country provides a more difficult environment for the educational integration of the second generation.

What we find is that, in a number of countries—Canada, France and the USA—the ethnic coefficients are either close to the line or scattered around the line of best fit. In Sweden the net coefficients are predominantly above the line, and this also applies (though not quite so clearly) to England and Wales. In contrast, Belgium, Germany and Switzerland are fairly clear cases where the ethnic coefficients are predominantly well below the line, while Finland and the Netherlands are somewhat less clear cases.

This analytical strategy allows us, albeit imperfectly, to deal with the problem that different minority groups are present in different countries. By taking account of the socioeconomic profile and degree of selectivity of each minority, we are doing the best we can to compare like with like and to take account of the fact that Canada has had a highly selective immigration regime while Germany has recruited guest workers. At the very least, the analysis casts considerable doubt on the notion that the ‘new’ countries of immigration necessarily provide a less favourable environment for minority integration than do the classic countries of immigration. ‘New’ Sweden is unambiguously ahead of ‘classic’ Canada in this respect.

Finally, we can look for clues to whether there are any ‘community effects’, i.e. groups who fare much better, or much worse, than might have been expected given the general origin and destination effects described above. Somewhat surprisingly there are relative few clear-cut cases of distinctive ‘deviant cases’.5 As inspection of Figures 3.1, 3.2 and 3.3 reveals, the most common pattern is for members of a specific origin group to have rather similar net coefficients whichever country they live in. We also need to take account of the finding that in some countries like Sweden, minorities consistently have better grades than their co-ethnics in other countries. In addition, given general problems of measurement error, one needs to be very careful when identifying unusually high or low achievement. The most striking examples are the relatively low scores of Former Yugoslavs in Switzerland, the relatively low scores of Chinese in Canada; and the relatively high scores of young people of Turkish background in the Netherlands (and Sweden) and of North Africans in France.

We must be very careful not to over-interpret these cases. There is always a major risk of ‘false positives’ (or false negatives) when one is looking at a large number of estimates. But it is reassuring that the recent TIES project which compared the second generation in Belgium, France, Germany, the (p.87) Netherlands, Sweden and Switzerland came to somewhat similar conclusions about the cross-national differences in the educational achievements of the Turkish second generation (Crul et al. 2012).6

Part of the explanation for these exceptional cases may be that the minorities are not strictly comparable. For example, the former Yugoslavs in Switzerland include a large number of Muslim Bosniaks and Kossovars, whereas the former Yugoslavians in Germany may have a larger proportion of Croatians and Serbs. This might also apply to young people of Turkish background in Sweden (as we noted earlier, many being ethnic Kurds) or of North Africans in France (some being Harkis or pied noirs). More detailed ethnographic studies are needed to understand these more subtle aspects of within-group diversity.

On the substantive side, one possibility is that the history of a particular minority in a particular country may be important. It may well be that the longer a group is established in the Western country, the more it converges towards the Western norms of achievement. The third generation may thus be less distinctive than the second generation, but in addition the second generation in a long-established community may be affected by the behaviour patterns of their group's third-generation members. North Africans in France are a particularly long-established group, and this may perhaps account for their closing the gap with the majority group.

Another possibility is that there may be distinctive ways in which a majority group treats a particular minority. Thus it is possible that Muslim minorities in Switzerland are disproportionately treated as outsiders (witness the controversy over building mosques in Switzerland) or that there have been special provisions for incorporating favoured minorities (as for example with Cuban refugees in the USA).

Conclusions

While we have made some progress in documenting and explaining second-generation ethnic penalties and premia in test scores, we must not pretend that we have solved all the problems. While it is clear that both socioeconomic profile and degree of selectivity of the first generation matter for the educational success of the second generation, some major puzzles remain. Thus our measure of selectivity has helped but is far from fully explaining the (p.88) educational success of the East Asian young people. This might be, in part, because of ‘noise’ in our measure of selectivity. Measurement error generally tends to reduce the explanatory power of one's predictors, and we would be the first to admit that the measure of selectivity (based on sample surveys from origin and destination countries) will be subject to error. But it seems clear that East Asian minorities typically perform even more highly in Western educational systems than would be expected given their positive selectivity.

East Asian success is by no means a new finding, and a considerable literature has developed attempting to explain East Asian ‘exceptionalism’ (for a recent overview see Sakamoto et al. 2009). One line of enquiry focuses on distinctive East Asian values, especially those deriving from Confucianism, although this line of argument has some difficulty coping with the success of South Asian Hindu and Sikh groups. Another line of enquiry focuses on the strength of the family and on parenting practices, where many of the South and East Asian groups will tend to have stronger and more traditional family structures (although this would probably also apply to West Asian groups, which are not nearly so successful educationally).

Another important possibility is that there may be what sociologists of education term ‘contextual effects’. For example, there is considerable evidence that a young person's performance is affected not only by their own social class background but also by the socioeconomic position of their peers. Since many, but by no means all, ethnic minorities have quite high levels of ‘bonding’ social capital, the crucial contextual influence on young people of minority background may not be the social class composition of their community but its ethnic composition. Thus Chinese young people may in a sense be ‘pulled up’ by the norms, knowhow and educational aspirations of their co-ethnic peers, while some more disadvantaged minorities may be held back by the lower aspirations and expectations of their communities. We can think of these, perhaps, as ‘multiplier’ or ‘spillover’ effects where members of successful communities become even more successful than their individual characteristics would have led one to expect, and vice versa for members of less successful communities. Group processes of this kind are probably the most plausible explanations for the remaining ‘origin’ differences that we have been unable to explain— although we should emphasise that they may well be processes developed in the country of destination rather than ones brought with them from the country of origin.

We also have considerable work to do explaining the country differences that we have identified. Why do minorities living in Sweden appear to have more favourable grades than those living in Belgium, Germany or Switzerland? First of all, we need to remember that some of these destination differences may be due in part to other variables that we have been unable to capture. The (p.89) size and history of the minorities could well be important, as could the group processes to which we have just alluded. But it is also plausible that institutional differences between countries in their educational arrangements have an important role to play.

We will carry out a more detailed analysis of these institutional differences in Chapter 9, but one promising line of enquiry focuses on the age at which tracking and selection occurs within schools. Broadly speaking, early tracking which sorts young people into different groups can be expected to have contextual effects, so that the progress which a student makes will be affected by the performance of his or her peers in the same track. If we compare two pupils of borderline ability, one assigned to a lower track and one assigned to the higher track, we would expect to find that the one in the higher track ends up making more progress than her peer assigned to the lower track. If minorities are disproportionately assigned to lower tracks (perhaps because of initial language difficulties or unfamiliarity with Western education), then they may make less progress than in more comprehensive systems.

More generally, educational systems with early selection may give little opportunity for ambitious young people to overcome initial educational weaknesses and move upwards out of their assigned track, while open comprehensive systems such as those of Sweden, England and Wales and the USA may give more opportunity for positively selected and ambitious members of the second generation to move ahead.

References

Bibliography references:

Boudon, R. (1974), Education, Opportunity, and Social Inequality (New York, Wiley).

Breen, R, Luijkx, R., Müller, W. & Pollak, R. (2010), ‘Long-term Trends in Educational Inequality in Europe: Class Inequalities and Gender Differences’, European Sociological Review, 26: 31–48.

Castles, S. & Kosack, G. (1973), Immigrant Workers and Class Structure in Western Europe (Oxford, Oxford University Press).

Crul, M., Schneider, J. & Lelie, F. (2012), The European Second Generation Compared: Does the Integration Context Matter? (Amsterdam, Amsterdam University Press).

Dronkers, J., Van der Velden, R. & Dunne, A. (2012), ‘Why are Migrant Students better off in Certain Types of Educational Systems or Schools than in Others?’ European Educational Research Journal, 11(1): 11–44.

Feliciano, C. (2005a), ‘Educational Selectivity in U.S. Immigration: How do Immigrants Compare to those Left Behind?’ Demography, 42(1): 131–52.

Feliciano, C. (2005b), ‘Does Selective Migration Matter? Explaining Ethnic Disparities in Educational Attainment among Immigrants' Children’, International Migration Review, 39(4): 841–71.

Heath, A.F. & Brinbaum, Y. (2007), ‘Explaining Ethnic Inequalities in Educational Attainment’, Ethnicities, 7: 291–305.

(p.90) Heath, A.F. & Cheung, S.-Y. (2007), Unequal Chances, Ethnic Minorities in Western Labour Markets, Proceedings of the British Academy 137 (Oxford, Oxford University Press).

Heath, A.F., Rothon, C. & Kilpi, E. (2008), ‘The Second Generation in Western Europe’, Annual Review of Sociology, 34: 211–35.

Jackson, M., Erikson, R., Goldthorpe, J.H. & Yaish, M. (2007), ‘Primary and Secondary Effects in Class Differentials in Educational Attainment: The Transition to A-level Courses in England and Wales’, Acta Sociologica, 50: 211–29.

Kao, G. & Thompson, J.S. (2003), ‘Racial and Ethnic Stratification in Educational Achievement and Attainment’, Annual Review of Sociology, 29: 417–42.

Levels, M. & Dronkers, J. (2008), ‘Educational Performance of Native and Immigrant Children from Various Countries of Origin’, Ethnic and Racial Studies, 31: 1404–25.

Levels, M., Dronkers, J. & Kraaykamp, G. (2008), ‘Immigrant Children's Educational Achievement in Western Countries: Origin, Destination, and Community Effects on Mathematical Performance’, American Sociological Review, 73(5): 835–53.

Marks, G.N. (2005a), ‘Cross-national Differences and Accounting for Social Class Inequalities in Education’, International Sociology, 20: 483–505.

Marks, G.N. (2005b), ‘Accounting for Immigrant–Non-immigrant Differences in Reading and Mathematics in Twenty Countries’, Ethnic and Racial Studies, 28: 925–46.

OECD (2006), Where Immigrants Succeed: A Comparative Review of Performance and Engagement in PISA 2003 (Paris, OECD).

Portes, A. & Zhou, M. (1993), ‘The New Second Generation: Segmented Assimilation and its Variants among post-1965 Immigrant Youth’, Annals of the American Academy of Political and Social Science, 530 (November): 74–96.

Reitz, J.G. (1998), Warmth of the Welcome: The Social Causes of Economic Success for Immigrants in Different Nations and Cities (Boulder, CO, Westview Press).

Sakamoto, A, Goyette, K.A. & Kim, C.H. (2009), ‘Socioeconomic Attainments of Asian Americans’, Annual Review of Sociology, 35: 255–76.

Van Tubergen, F., Maas, I. & Flap, H. (2004), ‘The Economic Incorporation of Immigrants in 18 Western Societies, Origin, Destination and Community Effects’, American Sociological Review, 69: 704–29.

Van Zanten, A. (1997), ‘Schooling Immigrants in France in the 1990s: Success or Failure of the Republican Model of Integration?’ Anthropology & Education Quarterly, 28: 351–74.

Zeroulou, Z. (1988), ‘La réussite scolaire des enfants d'immigrés: l'apport d'une approche en termes de mobilisation’, Revue française de sociologie, 29: 447–70.

Table A3.1. Data sources for Chapter 3.

Name of data source

Year of data collection

Population

Sample size for Chapter 3 analyses

Belgium

PISA

2003

15-year-old students

8,162

Canada

Youth in Transition Survey

2002

15-year-old students

24,639

England and Wales

Youth Cohort Study (YCS)Cohort 10 Sweep 1

2000

Young people who reached minimal schoolleaving age in the 1998–9 school year

16,134

Finland

Administrative data (linked Registers)2000–5

2000–5

Students in 9th grade in 2000–4

22,422

France

Panel-95

Parental background:1998, Educational out comes 2002

Representative sample of students who entered lower secondary education (college) in the 1995–6 school year

9,909

Germany

PISA-E (German enlargement)

2000

15-year-old students

30,822

Netherlands

VOCE 1999 (Secondary education cohort study)

1999 with yearly follow ups

Students entering upper secondary education in 1999 (around age 12)

18,167

Sweden

STAR database (register data)

1998–2003

All residents of Sweden who completed compulsory education

565,922

Switzerland

PISA-E (Swiss enlargement of PISA 2000 in German- and French-speaking Swiss cantons)

2000

15-year-old students

11,815

USA

Education Longitudinal Study 2002 (ELS 2002)

2002

Nationally representative sample of 10th graders

14,295

(p.92) (p.93)

Table A3.2 Measurement of variables.

General guidelines

Test score

Definition of ethnicity

Definition of parentaloccupational class

Definition of parental education

z-standardised test scores or grades

Target person born in destination country, both parents born in origin country

Highest occupational class of either parent, based on EGP class scheme

Combinations of highest educational qualification of parents, distinguishing between lower secondary or less, full secondary and tertiary

Country-specific details

Belgium

Reading test scores (PISA)

According to general guidelines

Highestoccupational class of either parent according to PISA measures (white collar/blue collar high skilled/lowskilled)

According to general guidelines

Canada

Reading test scores (PISA)

According to general guidelines

According to general guidelines

According to general guidelines

England and Wales Finland

GCSE points score Average of teacher-assigned grades

Includes 1.5 generation Children of immigrants who are born in Finland or who had at least 10 years residence before 9th grade

According to general guidelines According to general guidelines (with some minor deviation from the EGP class schema)

According to general guidelines According to general guidelines

France

Test scores in French at the end of lower secondary: in 1998 (for the children on time, or in 1999 or 2000 if they have repeated years)

According to general guidelines

According to general guidelines

According to general guidelines

Germany

Reading test scores (PISA)

According to general guidelines

According to general guidelines

According to general guidelines

Netherlands

Score on the standardised test taken at the first year of secondary school (age 12–13)

Official Dutch definition of second generation immigrant: child born in the Netherlands but at least one parent born abroad. If parents are from different origin countries child is classified according to father's country of origin

Head of household is employee, self-employed, inactive or has unknown labour market status

According to general guidelines

Sweden

Sum of grades in 16 best subjects at the end of compulsory school

Country of birth of biological (or adoptive) parent(s). For mixed origin: the country cultural closer to Sweden is used

According to general guidelines

Dominance method if both parents are present

Switzerland

Reading test scores (PISA)

According to general guidelines

According to general guidelines

According to general guidelines

USA

Reading test score in 10th grade

Target person bornin the USA with atleast one parent born abroad or in Puerto Rico

According to general guidelines

According to general guidelines

(p.94)

Notes:

Proceedings of the British Academy, 196, 63–93. © The British Academy 2014

(1) Georg Lorenz kindly prepared the data and ran the analyses for Switzerland. He is currently completing a doctorate at Otto-Friedrich-Universität, Bamberg.

(2) The OECD terminology of immigrant and native students is rather unfortunate as the second generation, born in the destination country, are natives in the usual sense of the term. In this volume we shall be using a terminology of ‘minorities’ and ‘majority groups’, not ‘immigrants’ and ‘natives’.

(3) We have had to use somewhat different measures of attainment in different countries: in some cases we can use results on the tests of literacy that are employed by the PISA studies (Belgium, Canada, Germany, Switzerland) and which are typically taken by a sample of young people at the age of 15; in France, Netherlands and the USA we also use test scores, although not from the PISA tests. In Finland and Sweden we use teacher-assigned grades taken at the end of lower secondary education; in England and Wales we use aggregated scores in the public examinations of the General Certificate of Secondary Education (GCSE) and in the Netherlands we use test scores earlier during lower secondary education.

(4) Following Boudon (1974) it could well be that socialisation is more important for test scores at the end of lower secondary education while economic factors are more important in continuation decisions.

(5) From a technical point of view, the measure of selectivity that we have used in this chapter can also be thought of as a ‘community effect’ since its value depends on the particular combination of origin and destination country.

(6) The TIES study surveyed minorities and a comparison group of the majority population in two cities in each country. The study also included Austria and Spain. The main measure of educational achievement, however, was highest level of education attended rather than test scores.