Educational Achievement Gaps Between Immigrant and Native Students in Three New Immigration Countries

Transcription

1 Educational Achievement Gaps Between Immigrant and Native Students in Three New Immigration Countries Davide Azzolini a, Philipp Schnell b,c, and John Palmer d a University of Trento b University of Amsterdam c Austrian Academy of Sciences d Princeton University Work in Progress: June 7, 2011 Abstract We expand previous research on immigrant-native educational gaps by exploring the patterns and the dynamics of the phenomenon in Ireland, Italy and Spain. Within the European context, these three countries represent new immigration destinations and, as such, they have been covered by relatively little empirical research so far. However, over the past ten years these three countries have experienced huge migratory inflows, and now host important numbers of school-age children of immigrants. By assessing and comparing immigrants educational outcomes within and between these countries, we also aim to test the extent to which certain hypotheses circulating in the international literature apply to these new contexts. We analyze both mathematics and reading skills using PISA data from 2003, 2006 and Our main findings are that first-generation students are the most disadvantaged group and that their performances are only marginally and not linearly associated with age at arrival. At the same time, second-generation students seem to perform better than first-generation students but still not at the level of natives. These results vary both between countries and across time. Regarding between-country differences, immigrants in Ireland tend to perform much better than immigrants in the two Mediterranean countries, where their the gap relative to natives persists even after accounting for both individual, familyand school-level factors. We argue that the higher socioeconomic status of immigrants in Ireland could be responsible for this relative advantage. Regarding trends across time, we detected a pattern of convergence among the three countries: the magnitude of the gap increased in all new countries, but it did so more markedly in Ireland. Finally, the mechanisms underlying the observed immigrant-native gaps differ between the two Mediterranean countries 1

2 and Ireland. While in the former, socioeconomic background seems to be the most relevant factor contributing to the gap, in Ireland the key-mechanism is language spoken at home, while socioeconomic background plays the opposite role from that played in Italy and Spain. 1 Introduction The great shift of migration flows from old to new destinations in the last three decades has been one of the most striking demographic developments in recent European history. As a result of these novel migration patterns, a growing number of immigrants have entered the educational systems and are now coming of age in countries that had, until recently, been known as net senders of migrants and have now become net receivers. The experiences of these new immigrant destinations are of great significance since those countries were caught ill prepared to manage the process of incorporation of new immigrants and their children. The immigrant populations in these countries are still young and education has become the key aspect in the process of social integration, in terms of both human capital formation and its later payoffs in the labour market. This study compares the educational achievements of children of immigrants in new immigration countries, namely Italy, Spain and Ireland. Research in these countries has been scarce so far, thus restricting the empirical basis of the theoretical explanations of the immigrant-native gaps mainly to European countries with long lasting immigration histories. This paper aims at partially correcting this gap, by providing a sound descriptive inquiry on the patterns and dynamics of the phenomenon and by highlighting similarities and differences between the three countries. The key questions that we ask in this article are: How well do children of immigrants perform in new immigration countries? Is there evidence of an educational disadvantage for immigrant children? What accounts for the observed differences within and between the three new immigration countries? To what extent do the hypotheses circulating in the international literature also apply to these new immigration contexts? And how do the achievement gaps develop over time? Our empirical analyses are based on standardized test scores in mathematics and reading derived from the PISA exams given in 2003, 2006 and These data allow us to investigate both cognitive and linguistic aspects of students achievement and examine trends over time. The paper begins with a short review of the literature on the immigrant-native educational achievement gaps in countries with older immigration traditions (Part 2). It then considers the general patterns of immigration to the three new immigration countries, the key features of their educational systems, and the existing country-specific literature on their immigrants educational performance (Part 3). We summarize our main research questions and assumptions (Part 4) before proceeding with a description of the data and the analytical strategy (Part 5). Our result section (Part 6) starts with a descriptive comparison of educational achievements of different immigrant generations as well as natives at three points in time for each country. Next, we present multilevel regression estimates of the immigrant-native gap and consider the extent to which the observed differences between the compared groups within countries can be explained by three main sets of variables: language spoken at home, family background and school characteristics. Finally, we focus specifically on first-generations, examining the association between age 2

3 at arrival and achievement and comparing first-generation performance across the three countries. We conclude by discussing our empirical findings and providing suggestions for future research (Part 7). 2 Accounting for the immigrant-native educational gap: an overview of the literature The achievement gap between immigrants and natives is a well-established regularity in several western countries (Schnepf, 2004; Marks, 2005). However, immigrants performance and their relative positions compared to their native classmates differ substantially, both within and between countries. Within countries there is higher dispersion of educational outcomes among immigrants than among natives (Schnepf, 2008). In the first place, research indicates that immigrant children s generational status is associated with educational achievement. In general, second-generation children tend to outperform first-generation children (OECD, 2006). This pattern is consistent with the straight-line assimilation theory, according to which the children of immigrants tend to adapt to the host society and, thus, to experience upward mobility compared to their parental generation (Alba & Nee, 2003). At the same time, as noted above, it has been shown that immigrant students are a highly heterogeneous group whose members might take differentiated paths of assimilation (i.e. segmented assimilation framework) (Haller, Portes, & Lynch, 2011; Portes & Zhou, 1993; Portes, Fernandez-Kelly, & Haller, 2009). Significant shares of second-generation children are at risk of experiencing downward mobility if the human capital of their parents is low and the mode of incorporation into the host society is hostile. second-generation pupils can also experience upward assimilation and even outperform native-born students (Chiswick & DebBurman, 2004), which could be explained partially by the so called immigrant optimism hypothesis, according to which immigrant parents conceive of their migration process as an upward mobility project and therefore invest substantial resources in the education of their children. For instance, this pattern holds true for many Asian students in the United States (Kao & Tienda, 1995; Portes et al., 2009) and has also been found for Indians in the UK and Northern Europe (Heath, Rothon, & Kilpi, 2008; Heath & Brinbaum, 2007). Along with generational status, another source of heterogeneity in educational outcomes is age at migration (Chiswick & DebBurman, 2004; Rumbaut, 2004; Myers, Gao, & Emeka, 2009). In general, the younger a student is when entering the host country for the first time, the higher the subsequent educational achievement. Two mechanisms appear to explain this trend: The first is that exposure to the host country culture and educational system enhances children s and parents acculturation and language acquisition, which positively influence educational outcomes. The second mechanism relates to the specific age at migration: Since cognitive and linguistic development is not constant over age, it is generally argued that the sooner a child enters a host country s educational system the better are his or her chances of achieving high education levels. Analogously, this also holds true for language acquisition which is quicker and more complete if the student arrives early. However, some studies cast doubts on the linearity of the relationship, arguing that arriving at specific ages (adolescence period) could exert more negative effects on education than arriving later (Chiswick & DebBurman, 2004; Myers et al., 2009). 3

4 Which other factors account for immigrants performances and their gap compared to natives? Most empirical findings point to the prominent contribution of family and school level factors in accounting for substantial parts of these educational gaps. Regarding family level factors, research on inequality in educational opportunities has largely documented the role played by family socioeconomic background - usually indexed through variables like parental occupation, income, and educational level (Breen & Jonsson, 2005). These factors partially account for the observed gap between children of immigrants and native born students, because immigrant families tend to be highly concentrated in the lower social strata (Heath et al., 2008). However, these traditional explanations do not always work in the same way for natives as for immigrants, and across national-origin groups (Jackson, Kiernan, & McLanahan, 2010). For instance, quite often immigrants have lower occupational returns to education compared to natives, especially if they hold foreign qualifications. This over-education determines a weakening of the explanatory power of parental education on children s educational outcomes (Heath et al., 2008). This is why is often useful to use more direct measures of socioeconomic, educational resources available in the home. This way it is possible to better explore the specific mechanisms through which socioeconomic background correlates with children learning outcomes often remain elusive (Duncan & Magnuson, 2005). Socioeconomic background is indeed associated with an ample array of other family characteristics, whose effects are not easily distinguishable from one another. For instance, low socioeconomic resources are associated with weak family structures (i.e., unstable families, overcrowded houses) and with shortages of cognitively stimulating resources (Lahaie, 2008; McLanahan & Percheski, 2008). Turning to explanations related to the school context, research points to a negative correlation between school immigrant composition and educational achievement: classes with higher percentages of immigrant children display on average lower educational outcomes. However this correlation is often a reflection of the socioeconomic status of the school, because immigrants are not randomly distributed across schools, and their families are not randomly distributed across neighbourhoods. On the contrary, immigrant families self-select into the most socioeconomically deprived neighbourhoods; as a result, their children are more likely to attend schools with lower average socioeconomic composition and lower quality (Portes & Hao, 2004; Fekjær & Birkelund, 2007; Hanushek, Kain, & Rivkin, 2009; Cebolla-Boado & Medina, 2010). Both family and school level factors are relevant determinants of between-country differences in immigrants performances. In the first place, because of selection and selfselection of immigrants into the different countries (Feliciano, 2005; Schnepf, 2008; Lozano & Steinberger, 2010). Also the organization of the education systems has been frequently found to be associated with the magnitude of the immigrant-native gaps. Countries with a school-type tracked and less standardized educational systems tend to have higher levels of inequality of educational opportunity by socioeconomic and immigrant background (Crul & Vermeulen, 2003; Werfhorst & Mijs, 2010). 4

5 3 Setting the scene: Ireland, Italy and Spain in comparison 3.1 From emigration to immigration countries Within the European context, Ireland, Italy and Spain have traditionally been emigration countries marked by declining populations and high rates of out migration. In the second half of the twentieth century, Italy and Spain became source countries of the guest worker recruitment needed by Northern European countries while the majority of Irish emigrants who left their home country in the early twentieth century went to the United Kingdom. During the last two decades this situation has reversed dramatically within all three countries, making of them three new and important destinations of immigration flows within Europe (Bauer, Lofstrom, & Zimmermann, 2000; Kreienbrink, 2008; OECD, 2006; Barrett, 1999; Bover & Vellila, 1999; Colombo & Sciortino, 2004). The beginning of the 1990s marked the start of the new phase of immigration to Ireland, Italy, and Spain. The South-European countries experienced a rise in the number of migrants from the South-North migration trend originating primarily from Sub-Saharan and other African regions. After the fall of the Iron Curtain, also East-West migration from Central and Eastern Europe accounted for a large part of the positive net immigration trend. Recently, the number of refugees seeking for asylum rose within Italy and Spain. Ireland experienced economic growth from the mid 1990s onwards (known as the Celtic Tiger boom) which resulted in increasing immigration. From the mid 1990s to the early 2000s, the majority of immigrants were returning Irish nationals, asylum applicants dominated from Nigerian and Romanian nationals, as well as non-eu nationals. After the 2004 EU enlargement, the inflow converted to EU flows primarily driven by nationals from the new EU member states. As a consequence of these movements, during the 2000s the presence of nonnational immigrants in the three new immigration countries increased at much higher rates as compared to old immigration countries like Germany, the Netherlands, United Kingdom, Sweden or France (see Figure 1). On average, stocks of non-nationals residing in Italy, Spain and Ireland increased by 4.6 times compared to only 0.2 times in the remaining EU-15 countries. There are substantial differences among the new immigration countries as well. Spain shows an increase of 7 times whereas the immigrant population in Italy and Ireland increased by 3 times. The reduced but still significant net immigration rate from 2008 onwards can largely be explained by economic contraction and associated decreased inflows. In 2009, immigrants made up a sizable portion of the total population within all three countries. Spain shows the highest percentage rate (12.3%), followed by Ireland (8.6%) and Italy (7.0%). Moreover, according to recent data, Ireland, Italy and Spain host about one third of the total immigrant population among EU-15 countries. 1 While sharing similar patterns of immigration, Ireland, Italy and Spain differ with regard to immigrant selectivity. The various phases of recent migration inflows led to sensible differences in the national-origin composition of the immigrant population residing in the three countries. Differences are particularly marked between Ireland, on one side, and Italy and Spain, on the other. Ireland shows smaller incidences of nonnationals from outside EU-27 (around 18%) compared to Italy (71%) and Spain (60%). 1 Official numbers are derived from Eurostat. Immigrant population is defined as foreign citizens 5

6 800 Ireland Spain Old European immigration countries Italy New Immigration Countries Figure 1 Non-nationals living in Ireland, Italy, Spain and in the remaining EU-15 countries ( ). Source: Eurostat and OECD; base year: Note: Old immigration countries are Austria, France, Germany, Netherlands, Sweden and the United Kingdom. New immigration countries are Italy, Ireland and Spain. More precisely, the main immigrant groups in Ireland are Eastern Europeans from Poland, Hungary, and the Czech Republic, followed by EU-15 nationals and immigrants from Asia and the United States of America. 2 Thus, Ireland is still quite an ethnically homogeneous country. Around 95 per cent indicated in the latest census that their ethnicity was White, while Black, Asian and other ethnicities accounted for just 1 per cent (Population statistics, 2010; Quinn, 2010). On the contrary, since the early 2000s the main nationality groups residing in Italy have been Albanians, Moroccans, Romanians, Eastern Europeans and immigrants from Asia. In Spain the main nationalities are Romanians, Moroccans, Americans, citizens from Ecuador, Bolivia and Columbia as well as nationals from the United Kingdom (Estadística, 2008). This different composition in terms of country of origin reflects, at least in part, a different selectivity of immigrants by human capital. Figure 2 shows that Ireland attracted higher educated immigrants compared to those who migrated to Italy and Spain. Moreover, even within the same broad nationality group, immigrants living in Ireland display higher educational levels. These facts clearly point to a different selection of immigrants by country of destination (Barrett, 1999). Although the national and socioeconomic composition of the immigrant population varies across the compared countries, the vast majority of non-nationals in Italy, Spain and Ireland are first-generation and were born outside the host country. Secondgeneration immigrants only make up a small proportion of the foreign population, this representing a peculiarity of new immigration countries (Portes, Aparicio, Haller, & Vick- 2 Unfortunately, official statistics in Ireland do not breakdown the Non-Irish population further into differentiated nationality groups 6

7 100% Ireland Spain Italy 80% 60% 40% 20% 0% Africa Asia Europe North America Oceania Latin America Figure 2 Percentage of tertiary-educated immigrants by nationality group in Ireland, Italy and Spain (2001). Source: strom, 2010). Driven by the increasing inflows of immigrants in the last decades, all three countries have constantly developed integration policies and regulation. According to the latest results of the MIPEX index which measures the achievement of those policies in terms of equal rights, responsibilities and opportunities for immigrants, differences between the three compared countries remain small (Huddleston, Niessen, Ni Chaoimh, & White, 2011). 3 In sum, the three compared countries share a number of commonalities in the historical experience of immigration. They turned recently from emigration to immigration countries and reached an inflow peak in mid In general, the inflow of foreign nationals is much higher as compared to the remaining European countries with long lasting immigration experience. The vast majority of non-nationals are first-generation immigrants, who are still relatively young. This is especially reflected in the foreign-born school aged population which is also overwhelmingly made up by first-generation immigrants. Differences across the compared countries appear with respect to selectively of particular national groups. 3 The MIPEX index measures policies and their implementation in 31 countries. In total, six policy areas are included: labor market mobility, family reunion, political participation, long-term residence, access to nationality and anti-discrimination. The index aims at measuring the achievement of equal rights, responsibilities and opportunities for all residents (Huddleston et al., 2011, p. 7). 7

8 3.2 The educational systems in Italy, Spain and Ireland The Italian education system is divided into five levels: non-compulsory pre-primary education (from 3 to 6 years), primary education (from 6 to 11), lower secondary education (from 11 to 14), upper secondary education (from 14 up to 19 years), and tertiary education. Pupils start compulsory education at the age of 6 in primary school (scuola elementare), which lasts five years. Lower secondary education in Italy lasts three years. At the age of 14 regular students are supposed to attend upper secondary schools. Upper secondary education includes academically oriented and generalist schools (licei and istituti d arte), vocational schools (istituti professionali) and an intermediate and technical type of school (istituti tecnici). At the end of secondary education, students take a final examination which is centrally regulated and therefore the same for each pupil within schools across Italy. Beside these three tracks, a further branch is represented by regional training courses (formazione professionale di base) which, contrary to the previous ones, do not allow students to access tertiary education (Barone & Schizzerotto, 2008). Compulsory education ends at the age of 16. In the Italian educational system, private schools only make up a small fraction of the total number of schools. In 2007/08, students enrolled in private schools (scuole paritarie) made up 5.5% of the entire school population (Miur, 2009). In Spain, pupils enter compulsory (general) education when they turn age 6. The Spanish compulsory period is divided into Primary Education (Educación Primaria) and Secondary Compulsory Education (Educaión Secundaria Obligatoria). Primary education is the first compulsory stage and covers six years of education, divided into three two-year cycles. It is normally completed with the age of 12. Afterwards, students continue on to compulsory secondary education which lasts until the age of 16. This stage is divided into four courses and is ordinarily completed after four years. The compulsory period in Spain is a completely comprehensive system. However, the Spanish school market is significantly segmented (Cebolla-Boado & Medina, 2010). Private schools (colegios concertados) make up a considerable portion of all secondary schools (26.3% of the students enrolled in compulsory education in ). Admission criteria to these schools assign priority to those pupils who live in the same district where the school is located. Students who achieve the Graduado en Educación Secundaria degree after lower secondary education are allowed to pass on to post compulsory secondary education (Ortiz, 2008). Upper secondary education in Spain is currently branched into an academic and a vocational training track. The former lasts two years, from 16 to 18, and gives access to university after an exam while the latter was established in 1990 during the educational reforms of the Spanish educational system as an alternative track to the academic orientated one. In Ireland, school attendance is compulsory from the age of six onwards with pupils first entering primary school. 4 The majority of primary schools are run by religious orders and are generally state aided. At the age of 12 students enter the second stage of compulsory education. Participation in full-time education in Ireland is compulsory until the age of 16 leading up to the national Junior Certificate examination. This test examination is nationally standardized and besides individually chosen subjects all pupils study English and Mathematics. The grades achieved in the Junior Certificate determine 4 Many children in Ireland, however, enter full-time education at the age of four or five by enrolling in reception classes within primary schools. In 2005, almost every second four or five-year-old was enrolled in reception classes (Smyth, 2008, p. 301) 8

9 the access to particular subjects and levels within the next educational stage. Students either continue in a Transition year or may proceed directly into a two-year upper secondary school which leads to the Leaving certificate. There are three different types of leaving certificates to be obtained in upper secondary education. The Established Leaving Certificate (LCE) is academic orientated in focus, while the Leaving Certificate Vocational (LCVP) trains students in vocational and technological skills. The last type of school provides the Leaving Certificate Applied (LCA) which is intended to catch less academic-orientated students who are potentially at risk of dropping out of school. Students are awarded points on the basis on their Leaving Certificate grades which are taken into consideration in the allocation process for places in tertiary educational institutions. Table 1 summarizes the key characteristics of the three compared educational systems, highlighting similarities and differences in terms of vertical progression and horizontal differentiation. In short, all three countries have the same starting age and the same age of compulsory schooling. However, the three countries differ with regard to the degree of differentiation of the system. In Italy, students are sorted into four different tracks at the age of 14. This choice represents in Italy a topical moment for the educational careers, since depending on the school track attended future educational and professional outcomes change dramatically. The Spanish educational system is instead a comprehensive one, since students are not tracked into separate schools before the end of compulsory education while Ireland can be situated between the two other countries. Table 1 Comparison of key features of secondary education in Ireland, Italy and Spain. Ireland Italy Spain Age at entering compulsory education Years spent in school until first tracking Age when first selected Number of tracks at secondary level Age when leaving compulsory education Children of Immigrants at school Over the past 10 years the presence of children of immigrants in Italian schools has increased by almost six times, making up around 7% of the whole student body today. However, their presence is significantly lower at upper secondary education compared to the lower educational levels (Miur, 2009, 2010). This is partially due to demographic factors, but also due to the fact that immigrants display lower school attendance rates and higher risk of dropping out of school as well as experiencing delay. Focusing on upper secondary education only, immigrants display higher probability of enrolling in vocational schools and lower propensity to choose general and pre-academic schools (Barban & White, 2009; Azzolini & Barone, 2011). In addition to these aspects, research findings converge in pointing out that immigrants lag behind in marks and achievement development, especially in subjects like writing and reading (Mantovani, 2008). This gap is a consequence of a double disadvantage faced by children of immigrants. On one hand, immigrants are more likely than natives to grow up in socioeconomically deprived families 9

10 and this accounts for a substantial part of their educational disadvantage. On the other hand, immigrants face the specific hurdles given by their migration background. High heterogeneity is found with regard to country of origin, e.g., students from Northern Africa, China and Indian Subcontinent display particularly low educational attainment while students from Eastern European countries perform almost similar as natives. Moreover, as far as immigrant generational status is concerned, most empirical findings indicate that second-generation students outperform first-generations. However, second-generation in Italy is still a very young population which are in large part in primary education years, therefore new studies will be needed in the future (Miur, 2009). Also in Spain, the share of foreigners in schools has increased massively over the past decade and now it has set at similar levels as in Italy. In general, immigrants display lower school attendance at upper secondary education and higher propensity to enrol in public schools compared to natives. Analogously to Italy, there is compelling evidence of a pronounced achievement gap of children of immigrants at both primary and secondary level (Zinovyeva, Felgueroso, & Vázquez, 2008; Cebolla-Boado & Medina, 2010; Calero & Waisgrais, 2009). Previous studies on PISA data indicate that first-generation are more disadvantaged than second-generations (Zinovyeva et al., 2008; Calero & Waisgrais, 2009). Moreover, there is multiple evidence that the immigrant-native gap has been increasing over the past years (Zinovyeva et al., 2008). Indeed, previous studies (carried out on 1990s data) found no significant gap (Buchmann & Parrado, 2006). African students are found to be the lowest performing group in Catalonian schools (Gutiérrez-Domènech & Adserà, 2009). Since the beginning of 2000 the total number of immigrant origin students has increased at a higher rate in the public sector compared to the private one, indicating difficulties to access access private and publicly funded private schools. Moreover, there is also evidence of a negative association between immigrant concentration and achievement (Zinovyeva et al., 2008; Cebolla-Boado & Medina, 2010; Calero & Waisgrais, 2009). Portes and colleagues (2010) also point to low aspirations and expectations of children of immigrants in Spanish secondary schools. In Ireland, the percentage of children of immigrants within secondary schools is relatively low while in contrast, the pattern among primary schools is quite different indicating a growing generation of children of immigrants which are still in the beginning of their educational career. This pattern is comparable to Italy and Spain confirming a peculiarity of new immigration countries. By 2008, children of immigrants made up more than 20 per cent of the student body in every tenth primary school (Smyth, Darmody, McGinnity, & D., 2009). Following residential patterns, children of immigrants are highly represented in urban schools and those already catering for more disadvantaged populations. According to recent PISA data, differences in educational achievements were most pronounced between the native student population and first-generation students in They are likely to have most difficulty in terms of school performance, as they have directly experienced the challenges of immigration while their second-generation counterparts score almost equally high in achievement tests as the Irish student population (OECD, 2010b). In Ireland, the great majority of the recent immigration waves do not have English or Irish as their first language. Several studies have identified that language issues among students for whom English is a foreign language is one of the biggest challenges. While there is no observed difference in the socioeconomic background of students by immigrant status in Ireland, low levels of proficiency in the language of the host country is likely to affect immigrant students academic achievement (Devine, Kenny, & Macneeka, 2004; Keogh & Whyte, 2003; Smyth et al., 2009; Vekic, 2003). Difficulties may 10

11 already arise during primary school as children are introduced into the main subjects. Achievement problems even increase in the course of the educational career (especially in senior classes) when the reading level of texts and the emphasis on written work enhance. Immigrant students with lower test scores in the early cycles of the educational career are more often found to be reliant on family assistance in upper secondary education (Byrne & Smyth, 2010) which sets immigrants parents under pressure given their minor language abilities and generally low knowledge of the Irish educational system. 4 Research questions In this paper we make three kinds of comparisons: We compare immigrant and native performance within the three countries, we compare immigrant performance across the countries, and finally we examine the trends over time. We ask five sets of questions. In the first place, we are interested in assessing whether average scores are sensitive to some key features of individuals migration backgrounds. In line with the above-cited literature, we break down our sample by immigrant generational status. In addition, we also investigate the association between achievement test scores and language spoken at home. Secondly, we ask what part of the immigrant-native gap is accounted for by the different family socioeconomic backgrounds of immigrants and native households. Since socioeconomic resources are likely to be unevenly distributed between natives and immigrants, we expect to find that a substantial portion of the gap is indeed accounted for by family socioeconomic background. Next, we shift our focus to the school level in order to investigate which features of the national education systems (i.e., tracking, standardization, and school composition) account for the gaps. We expect the estimated gap to decrease in highly tracked and less standardized systems as well along with the segmentation of the school system in terms of socioeconomic and immigrant family backgrounds. Given the nature of the data that we use, we compare the above results across the three countries, aiming to highlight similarities and differences. Moreover, we compare results across the three available time-points in order to understand trends in the phenomena. Finally, given the specificity of new immigration countries, we focus our analyses on first-generation children. We analyze the paths along which these children assimilate, looking at the association between age at arrival and achievement, and we compare their performances across the three countries. 5 Data, Variables and Analytical Strategy 5.1 Data Our empirical analyses are based on data from the Programme for International Student Assessment (PISA) collected in Ireland, Italy and Spain in the years 2003, 2006 and This is an examination and survey administered to representative samples of 15-year-old students enrolled in any educational institution in selected countries around the world. PISA assesses students learning abilities in three domains: reading, mathematics and science. Apart from assessing students skills, individual, family and school background 11

12 information are collected through questionnaires administered to students and school officials. PISA samples are derived from a complex, two-stage stratified sampling procedure, with schools containing 15-year-old students selected in the first stage and individual students selected in the second. The data include a set of probability weights and 80 sets of replicate probability weights at each level. 5 The samples used in this article contain some missing values in our main dependent and independent variables of interest. We applied a listwise deletion method, deleting all cases with any missing values on the variables used. Table 2 provides an overview on the the total sample size of the final subsamples for each wave per country after the listwise deletion. Table 2 Sample sizes used in regression analysis (total N per country and year). Ireland Italy Spain ,285 10,993 9, ,253 20,800 19, ,017 29,573 24, Dependent Variables: Mathematics and Reading competencies As dependent variables we use students scores in mathematics and reading literacy tests. Reading literacy is defined as an individuals capacity to understand, use, reflect on and engage with written texts. More precisely, reading literacy assessment focuses on the skills of students in understanding different text formats (e.g., continuous and non-continuous texts, different prose forms); accessing and retrieving information, forming a broad general understanding of the text, interpreting it, reflecting on its contents and reflecting on its form and features. The tests are aimed at measuring students capability apply their reading literacy to different situations (e.g., a novel, personal letters, official documents, etc.). Mathematical literacy is concerned with the ability of students to analyse, reason, and communicate ideas effectively as to pose, formulate, solve, and interpret solutions to mathematical problems in a variety of situations (e.g., personal, educational, occupational, public and scientific). The tests cover concepts like quantity, space and shape, change and relationships, and uncertainty numbers, algebra and geometry. Test scores are measured on a common scale allowing us to compare student achievement across countries and in time. 6 However, students are not administered the entire set of items of each test. On the contrary, they are assigned only in a sub-group of them. As a consequence, student performance estimates are imputed as five plausible values which represent the range of abilities that a student reasonably has. In order to capture the range we carried out our analysis on all five plausible values simultaneously. 5 These weights are based on the probability of units at each level being selected, adjusted for nonresponse. Replicate weights are generated using the Balanced Repeated Replication (BRR) method (OECD, 2009). PISA reports the school level weights and the weights of both levels combined; student weights may be derived simply by dividing the latter by the former. 6 OECD mean is set to 500 and standard deviation to 100 in the year when the domain was the major domain. 12

13 5.3 Independent variables Immigrant Generational Status The sampled student body has been classified by immigrant generational status. Information on the place of birth (abroad vs. host country) of the students has been combined with that of their parents. We used a strict definition of immigrant generational status by identifying first- and second-generation immigrants as those individuals with both parents born abroad. More precisely, the sample has been broken down into following categories: natives (defined as native-born children whose parents are both native-born); first-generations (foreign-born children with both parents born abroad), second-generations (native-born children whose parents are both foreign-born). In addition, children of mixed parentage are considered as a distinct category because of their quantitative relevance in the three countries and their specificity compared to children of exclusively immigrant parentage (Brinbaum & Cebolla-Boado, 2007). 7 Thus, immigrant generational status enters our analysis as a categorical variable with four distinct categories. The distribution of the different groups is roughly similar across countries revealing an already mentioned common trait of new immigration countries. Overall, the amount of school aged children of immigrants is small, especially when looking at second-generation immigrants. Nevertheless, the the all three groups of children of immigrants increases over time. Whereas first-generation students accounted for only 2% in Italy and Ireland and 3% in Spain in 2003, six years later (2009) their share climbed to 7-8% and to 4% in Ireland, Spain and Italy respectively. The share of second-generation immigrants also increased, but at a much lower rate (around 1% per wave and country). As far as the size of mixed-parentage children, it has been increasing in all three countries, and it represents a particular weight in Ireland where it reaches 16% in 2009, compared with 6% in Italy and Spain Language Spoken At Home We include language spoken at home as a dummy variable, which takes on the value 0 if the student declares he usually speaks the host-country language (or a national dialect) and on the value 1 if he usually speaks a test language. 8 As expected, secondgeneration immigrants show higher rates speaking the host language at home compared to first-generations in all three countries. However, great differences across countries exist as well. Ireland and Spain have much higher rates of immigrants talking a foreign language at home compared to Italy. Moreover, it is also evident that the percentage of immigrants who declare to speak the host language in Ireland has been continuously decreasing over time, confirming the above described pattern of increased presence of non-return immigrants. 7 Children of mixed parentage are put in a unique category regardless of their place of birth, because those born abroad represent only a small fraction and do not differ significantly from those born in the receiving country and thus do not alter our estimates. 8 Due to high percentage of missing values on language information in Italy, instead of deleting missing values, we define a category for missing values. 13

14 5.3.3 Family Socioeconomic Background We measure family socioeconomic status and the availability of cultural and educational resources at home through four variables: In the first place we use parental education level in order to capture the human capital possessed by the family. This variable has been coded in a categorical way following the ISCED classification ranging from ISCED 1 to ISCED 5a-6. Secondly, we use the highest occupational status of parents by including the international socioeconomic index of occupational status (ISEI) (Ganzeboom & Treiman, 1996). Third, to further capture the availability of cognitively stimulating resources at home, we include an index which summarizes cultural and educational resources (e.g., availability of a place to study, a personal computer, books at home), possessions and indicators of wealth at home. Finally, we add a binary variable which allows us to adjust for variations in the family structure. The variable takes on the value 1 for nuclear families and on the value 0 for non-nuclear family types. Differences between the three compared countries are most pronounced when looking at the socioeconomic status of the parental generation. On average, Ireland and Spain have higher rates of highly educated parents (ISCED 5a-6). However, what is more relevant is the educational level of immigrant parents and their comparison with native parents. Whereas in Spain and in Italy immigrant households display roughly similar educational attainment as natives, this does not hold true for Ireland where parents of first-generation immigrants have significantly higher educational attainment compared with native parents. The same patterns observed for parental education apply to highest parental occupational status: In Ireland immigrant parents display higher labor market outcomes and relevant resources compared to natives, while the exact opposite occurs in the two Mediterranean countries, where parents of first-generations display particularly low occupational attainment School Characteristics Drawing on the above cited literature on education systems and equality of educational opportunity and on the descriptions of the three national systems, we include a set of variables at the school level. In the first place, we include information in order to account for the degree of school-type differentiation and to adjust for the different likelihood of students to enroll in one school over another. More precisely, we include a categorical variable which indicates the specific programme at secondary education in which the student is enrolled. This variable is not included in Spain because it is a comprehensive system. 9 In Italy this variable is coded as following: Academic schools, Technical schools, Vocational schools, Vocational training, Lower secondary schools. The distribution of students among these different tracks indicate the existence of a very pronounced segregation of immigrants into the shorter and less prestigious tracks (vocational schools and training courses) over the most academically oriented ones (licei). Such a segregation based on immigrant status is not to be found in Ireland since differentials between natives and children of immigrants with regard to school type are rather small. A second indicator of horizontal differentiation of the education systems is school ownership. We include school ownership as a dummy variable (public vs private school). Substantial differences exits between natives and children of immigrants. However, these 9 In Spain, individuals enrolled in higher secondary school (Bachilerado) account for less than 0.05% and so they are deleted. Moreover, we did not include grade because of collinearity with school programme. 14

15 are not consistent across countries: while immigrants in Spain have much lower probability to enrol in a private institution, no differences are found in Ireland while immigrants in Italy are more often found in private schools. Third, we include a measure of socioeconomic school composition by taking the weighted average of the highest parental occupational status. There is a similar pattern across countries in 2009, since immigrants tend to attend schools with lower socioeconomic status compared to natives, this is especially true in Italy and Spain, even though in Ireland a pronounced decline in socioeconomic status of schools attended by first-generation immigrants is found across waves. Fourth, we control for the proportion of first-generation immigrants enrolled in each class. On average, school immigrant concentration is higher in Spain and Ireland compared with Italy Additional Control Variables Finally, we further control for gender and age as well as for the size of area of residence (rural area, small town, town, city, large city) Analytical strategies In what follows, we first analyse the mean reading and mathematics scores of natives and the three groups of immigrants (first-generations, second-generations, and children of mixed-parentage) (6.1). We estimate mean scores following the approach recommended in the PISA documentation, using the five plausible values, final sampling weights, and 80 replicate sampling weights provided with the data. The use of plausible values is aimed at capturing an unbiased and continuous measure of student proficiency from discrete exam scores, while the weights account for the sampling structure of the survey and provide design-based measures of uncertainty (OECD, 2009). By using this approach, we also ensure that our estimates are comparable with the official OECD figures. As a second step, we regress reading and mathematics scores on immigrant generational status using hierarchical linear models with levels for individual students and their schools (6.2). In these models, we treat student performance as a normally distributed random variable drawn from a distribution with mean µ for student i in school j modeled as: µ ij = α j + X T λ (5.1) where α j is a random intercept estimated for each school, and X T λ is a transposed matrix of covariates and their estimated coefficients. We fit a series of model specifications, progressively adding covariates to assess how variations in language spoken at home, family socioeconomic background and school characteristics account for immigrant-native achievement gaps. We fit these models using the student and school level probability weights adjusted according to the approach suggested in Pfefferman, Skinner, Homes, Goldstein, and Rasbash (1998). 11 We fit every model five times, using each of the plausible values as the dependent variable, and then averaged the resulting parameter estimates. We did not use the replicate weights to estimate uncertainty in the regression models, opting instead to use model-based uncertainty estimates. Next, we 10 A descriptive overview of all variables and their distribution are summarized in Appendix B 11 Specifically, we calculated the Pfefferman et al. (1998)adjusted weights using software described in Chantala, Blanchette, and Suchindran (2006). 15

16 restrict our analysis to first-generation students and investigate the relationship between test scores and age at migration. Results are displayed by plotting the marginal effects of age at migration, adjusted for the full set of variables described in the previous section (6.3). Finally, we explore cross-national differences in the achievements of first-generation immigrants (6.4). We proceeded as following. As a first step, we took the predicted scores of four profiles of first-generation students obtained from the full model (thus including all the above described variables) within the three countries. We defined the four idealtype profiles of immigrants by crossing their socioeconomic status (high vs. low) and the school track attended (high- vs. low-performing). More precisely, we defined socioeconomic status as low for households with no highly educated parents, and placed at the 25th percentile in both the parental ISEI and the home possession indices. Accordingly, we labeled as high-socioeconomic status those families with at least a tertiary educated parent and placed at the 75th percentiles in the two indices mentioned above. Regarding school type, we empirically defined vocational tracks as low-performing tracks, and for high-performing we picked academic tracks. Of course, in Spain we did not distinguish between tracks, since the education system is a completely comprehensive one. All other characteristics were held at constant values, meaning that we picked boys who live in nuclear families which reside in towns (15,000 to 100,000) and where the most frequently spoken language is not the official language of the host country. The remaining continuous variables were set to their mean values. As a second step, we re-estimated the predicted scores for the just described profiles of immigrants in Italy and Spain by applying to them the values of the two socioeconomic variables (ISEI and home possessions) of their counterparts in Ireland. This way we obtain new predicted scores, adjusted for the different socioeconomic resources between the countries Empirical Results 6.1 Average competencies of natives and children of immigrants Figure 3 displays average scores in reading and mathematics of each immigrant group of students as defined by their immigrant generational status in 2003, 2006 and Overall, natives tended to perform better than first-generation immigrants, and to some extent better than second-generation immigrants, although there are substantial variations between the three countries. Irish native students display higher average competencies in both mathematics and reading compared to their Italian and Spanish counterparts. However, Ireland natives average score in both subjects sharply decreased between 2006 and 2009, while in Italy it increased and in Spain it remained substantially unchanged. These developments led to an increasing convergence between the three countries. The achievement outcomes of children of immigrants show even more striking variations across countries. However, the uncertainty of the estimates is also much higher than that of the native estimates. This is largely due to the small sizes of the immigrant sub-samples, although it may also result, in part, from higher educational dispersion 12 This statistical exercise resembles the Blinder-Oaxaca technique for decomposing continuous dependent variables. In order to take into account the multilevel structure of the data and to compare precise profiles of immigrants across countries, we did not use the Stata routine oaxaca 16