Immigrants’ children’s transition to secondary school in Italy

Abstract

Choosing a secondary school represents an important step in the lives of students in Italy, in that it has a strong bearing on their ultimate educational achievement and labor force trajectory. In this paper, we analyze the effect of generational status and length of residence on the transition to secondary school among immigrants living in Italy. Using data from the ITAGEN2 follow-up, we analyze scholastic results from the middle school final exam and the choice of secondary school among the adolescents in Italy. Children of immigrants are more likely to have inferior outcomes on the middle school exam and to enroll in vocational and polytechnic schools. Our multivariate results indicate that, after controlling for the family’s human capital and other key background factors, immigrant students show greater propensity to choose a vocational path. Differences between immigrants and natives in secondary school tracks are also manifest when previous scholastic results are taken into account.

1 Introduction

In the mid-1980s, the number of immigrants living in Italy began to steadily increase. Initially, the flow of immigrants was moderate, and consisted mostly of movement from North Africa, several Sub-Saharan countries, and the Philippines. With the fall of the Berlin Wall in 1989, migration into Italy reached significant levels, due in large part to the arrival of numerous individuals from ex-communist European countries (especially Albania and Romania). In early 2008, the number of foreigners (including illegal immigrants without residence permits) living within the nation’s borders was estimated at 4 million, i.e. 7% of the total population (Cesareo, 2007).

In addition, during the 15 year period of 1993–2007, foreigners grew at least 250-thousand in number annually (including the number of births from foreign couples). During the same period, 500-thousand children were born each year to couples with at least one foreign parent; meaning that one third of Italy’s demographic renewal occurred thanks to immigrants, who have contributed significantly to slowing the aging of the population, (Billari and Dalla Zuanna, 2008).

Rapid growth in the number of immigrant youth has brought about profound modifications in the educational system. In fact, the population in school of non-Italian nationality rose from 60 thousand in 1997 to over 600 thousand in 2009; these numbers do not include universities. During the same period, the percentage of foreigners in school increased from 0.7% to 7.0%. According to data from the Italian Ministry of Education, non-Italian students appear to be more vulnerable than natives in the educational system. They are, in other words, more likely to achieve lower scholastic outcomes, and to have higher dropout rates and lower levels of school attainment, (Ministry of Education, 2009).

School is the first important formal organization that children encounter on their own and, of course, a major conduit to social integration. Compared to other major European countries, Italy is characterized by lower educational attainment and lower intergenerational mobility (Brunello and Checchi, 2005). Due to the recency of immigration into Italy, there is a lack of extensive studies which document the educational attainment and social mobility of second generation immigrants. Study of the educational choices of immigrants’ children is thus an important step in formulating appropriate integration policies.

The Italian educational system - beginning with its earliest stages - is organized as follows. Kindergarten starts at the age of 3 and ends at the age of 6, and attendance is optional; it is organized and financed either by the central government or by local councils. Compulsory education spans the ages of 6 to 15 years of age and includes: primary school (from 6 to 11, called scuola primaria), middle school (from 11 to 14, called scuola secondaria del primo ciclo), and the initial year of upper secondary school (from 14 to 17 or 19, depending on the chosen track, called scuola secondaria del secondo ciclo). Different options are available for upper secondary school, and can be described as tripartite (we also use this classification in the empirical analysis): (1) a generalist academic orientation provided by high schools (5 years, called licei, with further distinction made between schools oriented towards the humanities, scientific activities, languages, and pedagogical sciences), (2), a technical oriented education provided by Polytechnic schools (5 years, called istituti tecnici, with further differentiations by specialization) and (3), a vocational training offered by local schools organized at the regional level (3 years, called istituti di formazione professionale). Until 1969, only one track granted the possibility of college admission, while another track granted admission only in a limited number of disciplines. The remaining track prepared individuals for specific jobs (such as primary school teachers or construction site supervisors). The current system is still characterized by a number of options, but now all grant eligibility for college admission, condition on completing five years of secondary schooling (i.e. students from vocational schools may enroll if they attend two integrative years). However, each of these tracks still predicts very different outcomes in terms of any additional schooling acquired and labor market performance. More than 88% of students who graduate from licei enroll in a university as opposed to 17.8% of students coming from the vocational track. The choice of the type of secondary school to attend is typically made at the age of 13, during the final year of middle school. There is no admission exam to enter any secondary school. The educational system is dominated by public schools. The proportion of students attending a private school during the 2007–2008 scholastic year was 5.1% for middle schools and 7.0% for secondary schools. Students and their families receive counseling from teachers, in some cases supported by psychologists, who use students scores achieved during compulsory schooling as one of the principal guidelines for orientation.

In this paper, we focus on a specific moment during the educational career of a student: the transition from primary school to secondary school. Official statistics show that immigrant youths have a higher risk of enrolling in vocational schools compared to natives. In fact, the percentage of foreign students enrolled in the first year of secondary school (9th grade) is 10.6% in vocational schools compared to 2.5% in high schools. For example, in the Veneto region during the 2007–2008 scholastic year, among newly enrolled students in Enaip-Acli courses (the major institution in the region offering training courses), almost a third were immigrants’ children and half of them arrived in Italy after their 14th birthday (Dalla Zuanna et al., 2009a). The importance of focusing on the typology of secondary school is thus motivated by two reasons: first, the choice of school influences future educational attainments and, in the case of the vocational track, it can prevent access to a college education; second, the higher incidence of foreign students in vocational schools may be a signal of segregation in the school system.

Because of the early age of track differentiation across schools, the Italian system tends to privilege parents’ choice more than ability of students, leading to low intergenerational mobility (Checchi and Flabbi, 2007). In such a context, immigrant families might be disadvantaged if they lack country-specific human capital and appropriate knowledge of the Italian educational system. The aim of this paper is to provide a greater understanding of how immigrants and second-generation youths progress through the education system, in order to highlight possible obstacles to successful integration. In particular, we ask: are immigrants’ children able to progress through the educational system on par with other Italian children, ceteris paribus? Does the school system provide them with the necessary instruments to overcome initial differences between immigrants and natives?

To shed light on the transition to secondary school, our study takes into account the scholastic trajectory of students from the end of middle school (8th grade) to the first year of secondary school (9th grade). Using individual data for both Italians and immigrant adolescents, we are able to examine the determinants of school results at the end of middle school and to follow the same student’s choice of secondary school. The use of longitudinal data allows us to disentangle the process of adaptation to the school system and the educational career of the student (Glick and White, 2003; Portes and Rumbaut, 2005; White and Glick, 2009).

Literature on the assimilation and integration of second generation immigrants is extensive, but much attention has focused on the United States. Scholars have proposed different theoretical models to explain the position of second generation immigrants in society. According to a straight-line assimilation perspective, differences between natives and second generation immigrants may attenuate monotonically with time across generations (Neidert and Farley., 1985). A monotonic process is attributed to the migration-adaptation process itself, and does not depend on the country of origin, once socioeconomic background has been controlled for. Other scholars have suggested that integration paths may diverge for different immigrant groups (Rumbaut and Portes, 2001; Zhou, 1997). In this alternative perspective, adaptation to the host society varies among different ethnic groups, even when controlling for compositional characteristics. The process of assimilation is therefore linked to the receptivity of the host society that may favor the upward mobility of some ethnic groups, but fail to support that of others, especially those from historically disadvantaged ethnic subgroups. This disadvantage may persist across generations, causing what has been defined as “downward assimilation”, where the gap between immigrants and natives increases with the second generation, independently of whether they were born in the host country.

The number of European studies on the adaptation of second generation immigrants has recently in-creased, often drawing on analogous studies conducted in the US (Thomson and Crul, 2007). Although the theoretical framework is the same, second generation immigrant groups in Europe are, on the whole, ethnically very different compared to those in the US. Within the American literature on second generation immigrants, US-born children of Mexican and Asian immigrants play a key role, while in Europe the composition of second-generation groups is more heterogeneous and varies among countries. The parents of the largest groups of second generation immigrants in Europe either come from ex-colonies or were recruited, as in the case of Italy, as labor migrants. The study of second generation immigrants in Europe often emphasizes the importance of the host society and context for integration pathways (Crul and Vermeulen, 2003; Doomernik, 1998). In particular, national variations in institutional arrangements play a key role in defining distinct patterns of integration across Europe. Differences in educational systems and the ways in which the transition to the labor market is formalized (Crul and Vermeulen, 2006) play an important part in the successful integration of second generation immigrants. In this paper we attempt to describe the process of assimilation of immigrants’ children within the Italian educational system.

The paper is organized as follows. In the next section we present our research question and our identification strategy. In section 3 we describe our data, methods and empirical analysis. After the descriptive analysis (section 4), we present our results in section 5. Section 6 concludes the paper. We discuss selection problems and the consequent strategy adopted to reduce eventual sources of bias in the appendix.

2 The role of generational status and country of origin in the transition to secondary school

Immigrants in school often differ from native children in terms of performance and attainment. It is not clear, however, whether such differences necessarily persist in the presence of statistical controls for compositional effects. Characteristics such as socioeconomic status, educational achievements and family composition are closely tied to immigration status (Glick and White, 2003). In our analyses we are interested in distinguishing between the effects of these factors. Our first aim is to test whether the disadvantage of being an immigrant remains after controlling for family characteristics. Traits such as family structure and family socioeconomic status have been shown to influence academic performance (Bankston and Caldas, 1997; Caldas and Bankston, 1997). Once we have described the determinants of scholastic outcomes, we endeavor to investigate their role in the educational trajectories of students. From a meritocratic perspective, school should nullify any initial differences between immigrants, second generation immigrants, and natives. According to this hypothesis, access to secondary school would be influenced only by previous outcomes achieved during the scholastic career. In our analysis, we examine whether generational status plays a role in the choice of secondary school, controlling for outcomes at the end of middle school. We also investigate the role of compositional characteristics. The effect of family structure and socioeconomic status may, in fact, influence not only the outcome but also intentions of educational attainment existent within families (Checchi and Flabbi, 2007). Moreover, we explore whether there are significant variations in either the outcome or the trajectories of immigrants from different countries. There is some evidence that immigrants from different groups have access to specific resources in term of “social capital”, related to their country of origin (Bauer and Riphahn, 2007; Borjas, 1992; Fekjaer, 2007). Differentials in the educational outcomes of children from different immigrant groups (after controlling for generational status and socioeconomic background), may highlight critical situations among disadvantaged communities or point to successful situations of adaptation to the host society. The heterogeneity in terms of country of origin does not necessary indicate segmented assimilation, but may emphasize different strategies of educational investment among immigrant groups.

3 Data and Methods

Data were drawn from the ITAGEN2, a survey of students living in Italy and attending middle school during the 2005–2006 school year. ITAGEN2 is the first nation-wide extensive survey on children with at least one foreign parent, and focuses on the determinants of social integration. Wave I includes a sample of 6,368 foreigners and 10,537 natives (Barban and Dalla Zuanna, 2010; Dalla Zuanna et al., 2009b). The subjects live in 44 provinces and attend 228 different middle schools. The schools were randomly chosen among those with a foreign student body consisting of +10% of the total (in five of the Central and Northern regions: Lombardy, Veneto, Tuscany, Marches and Lazio) and +3% of the total (in four of the Southern regions: Campania, Apulia, Calabria and Sicily). In each school, three entire classes were interviewed (one from each level of middle school) as were all of the immigrants in attendance. In schools with more than 60 foreign students, data for a greater number of classes were collected in order to improve the sample of natives. For each school, a mean of 64 Italians and 51 immigrants was interviewed. The Wave I interview focused primarily on the characteristics of the family, the migratory process, the children’s use of time, and their opinions and aspirations for the future. Information on scholastic achievement was not collected during the in-school Wave I interview.

The 2008 wave was the first follow-up. Data were collected by means of a CATI interview among an ITAGEN2 subsample in five regions: Veneto, Marches, Apulia, Calabria and Sicily. The target population includes 1,389 immigrants’ children¹ and 1,589 Italians. The follow-up survey took place two years after the first interview. Almost two thirds of the initial sample of the students had therefore completed middle school. The follow-up questionnaire included a set of questions concerning scholastic attainment and achievement. The response rate was 70% among Italians and 47% among foreigners. The great majority of the non-responses is attributable to discontinued telephonic contacts rather than to refusals. In order to gain supplementary data on scholastic outcomes, we performed (in the Veneto and Apulia only) an additional survey in the schools. We also collected data on the final middle school exam for 364 students. Data from the baseline survey, linked to the follow-up data and the supplementary survey, allow us to trace the educational career of students who attended the 7th or 8th grade during the 2005–2006 scholastic year. Table 1 shows the pattern of available final exam outcomes and the choice of secondary school; data refer only to students attending 7th and 8th grade during the baseline survey.

Table 1.

Pattern of availability of data for Middle School exam and secondary School. Baseline survey and follow-up ITAGEN2. Data restricted to the target population of the follow-up survey

Pattern of Availability	Natives		Immigrants		Total
	n	Percentage	n	Percentage	n	Percentage
Baseline data	1,589	100	1,389	100	2,978	100
Final Exam Available of which from register data	1,137	64	830	53	1,967	58
	177		187		364
Secondary school Available	1,116	70	659	47	1,775	60

Source: ITAGEN2 survey

To compensate for the unequal selection in the follow-up interview, we adopted a weighting strategy. In the appendix, we discuss in detail the different sources of selection and the correction adopted.

3.1 Variables

Dependent variables

Scholastic Outcome

As a measure of scholastic performance, we collected the results of the final exam taken by pupils at the end of their third year of middle school. The middle school final exam is a compulsory ministerial exam and it is required in order to enroll in secondary school. The exam consists of both written and oral components. The customary age at which the exam is taken, without previous grade retention or delay in ad-mission, is between 13 and 14. The exam has six possible outcomes: Not Admitted, Failure, Sufficient, Good, Very Good, and Excellent. In the case of a “Not Admitted” or “Failure” result, students must repeat the last grade, after which they may take the exam the following year.

Secondary School

In the Italian school system, at the end of middle school (8th grade) students must choose between a number of different secondary school options. This choice is one of the most important factors influencing the level of education that they ultimately attain. Secondary schools can be divided in three groups: vocational schools, polytechnic schools and high schools. Although secondary schools are not compulsory, the minimum age at which a student may drop out the school system is 15. In our analysis, we collect data for secondary school attendance at the moment of the follow-up survey. For those who dropped out before the interview, we collect information on the last secondary school attended.

Independent variables

Generational Status

Generational status was categorized by the parents’ birthplace and by length of residency in Italy. We identified second generation immigrants as those youth born in Italy to at least one foreign-born parent. We separated “recent immigrants” (in Italy less than 5 years at the time of the baseline survey) from “pre-school immigrants” (in Italy at least 5 years at the time of the first survey; i.e. arrived in Italy before school age or at early elementary school age). We defined natives as those respondents born in Italy of both Italian parents.

Country of Origin

We included measures of ethnicity in our models in order to control for the possibility that the country of origin, not migration status per se, results in differential academic achievement and educational pattern. The country of origin was recorded, for both immigrants and second generation immigrants, as the mother’s birthplace. We identified respondents as belonging to the six largest communities present in the sample: Italy, Albania, Yugoslavia, Morocco, Tunisia, Macedonia, and China. For respondents not included in the previous categories, we distinguished between “Others from developed countries” (i.e. European Union (12 countries), United States, Canada, Japan, and Israel) and “Others”.

Other covariates

Other measures included in the analyses represent students’ demographic characteristics, access to human capital, family environment and college aspirations. In place of a standardized measure of socio-economical status (SES), we utilized the education level of the parents, homeownership, and family size. With specific regard to parents’ level of education, we constructed a series of dummy variables indicating the highest education level between the two parents. In the “low education” category we include parents who finished school before the age of 14; “medium education”, those who finished school between the ages of 15 and 19; and “high education”, those who finished when they were 20 or older; the last category includes cases for which information about the education of both parents is missing. Homeownership is indicated with a dummy variable equal to one when the family of the respondent owns the house where she/he is living. Number of siblings is also represented by a series of dummy variables: (1) no siblings, (2) one or two siblings (reference group), and (3) three or more siblings (e.g Vernez et al., 1996). Our measure of the respondent’s aspirations for college is a dummy variable equal to one where she/he expresses, at the moment of the baseline survey, willingness to attend college.

3.2 Methods

In order to investigate the transition to secondary school we ran two different sets of regression models. First, we modeled the outcome of the final exam taken by students at the end of middle school. As the outcome is expressed in four different categories, we utilized an ordinal regression model (Agresti, 2002). In this model, the outcome variable is expressed in ordered categories (Four categories, from Sufficient to Excellent). The assumption underlying this approach is that there is a specific ranking of the different categories. We designated the lowest grade as the reference category. The estimates express a measure of how much a covariate increases the probability of achieving a higher grade. In the second set of models, we ran a multinomial logit regression (MLN) to describe the choice of secondary school. In this case we did not assume a specific rank in the categories of the outcome variable. We obtained different estimates for the effect of same covariate on the probability of attending a typology of secondary schools compared to the reference category, “vocational school”. Estimates can be expressed in terms of relative risks. In both models we utilized the same set of covariates. We constructed “nested models” to test different specifications using the log-likelihood ratio test. The geographical heterogeneity of outcomes is taken into account including a “province fixed effect” in the models.

4 Descriptive analysis

We begin our analysis by providing some descriptive statistics of the baseline sample. In table 4 we show a summary of the differences according to generational status. Immigrant children are slightly older than other groups at the end of middle school. This is due to school administrators who may opt to enroll immigrants in lower grades if language proficiency is not satisfactory. We do not observe strong differences based on parents’ level of education. As pointed by Reyneri (2004), immigrant families are quite similar to Italians in terms of years of schooling. Homeownership rates vary significantly among groups. The proportion of homeowners increases considerably with the amount of time spent in Italy by the children, likely due to gradual stability in the family’s migratory experience (Barban and Dalla Zuanna, 2010). Nevertheless, we observe a notable gap in homeownership between immigrants’ children born in Italy and their native counterparts. Immigrants’ children are also more likely to live in larger size families. The average number of siblings decreases with the length of the children’s migratory experience. Recent immigrants live with an average of 2.03 siblings, preschool immigrants with 1.90 and second generation immigrants with 1.85. In comparison, native children live with 1.45 siblings on average. Similar to other studies (Dalla Zuanna et al., 2009a; Rumbaut, 1996; St-Hilaire, 2002), we do not find considerable differences between the ambitions and aspirations of immigrant children and those of Italians. Also, the proportion of children who think they will go on to college is similar across groups. Data on scholastic results and secondary school enrollment (Tables 2 and 3) show significant differences among groups. All immigrants’ children are more likely to achieve lower results on the final exam. In particular, over half (52%) of recent immigrants received the lowest grade; we find a similar the result among preschool immigrants (46%), even if the latter experienced the entire schooling process in Italy. Scholastic achievement is also lower (38.5% sufficient) among those who were born in Italy compared to the Italian children (22% Sufficient).

Table 4.

Characteristics of the sample for baseline sample by generational status

Variable	Recent Immigrants	Preschool Immigrants	Second Generation	Natives
Sex (%)
Male	53.4	53.0	51.0	48.9
Female	46.6	47.0	49.0	51.1
Age at the Exam (mean)	15.1	14.7	14.2	14.0
Mother’s country of origin (%)
Italy	7.2	7.7	17.2	100.0
Albania	16.0	31.4	6.9	-
Yugoslavia	3.9	6.1	2.2	-
Morocco	3.9	9.3	4.2	-
Tunisia	7.4	3.9	3.4	-
Macedonia	11.9	12.8	9.8	-
China	7.0	5.1	18.1	-
Others from developed countries	1.9	1.4	12.0	-
Others	40.9	22.5	26.2	-
Parent’s Education (%)
Low	25.7	23.3	23.8	29.9
Medium	32.9	36.0	34.3	35.8
High	23.4	19.6	27.2	23.2
Unknown	18.1	21.1	14.7	11.2
Household posession (%)
Rentals	81.3	71.5	57.4	18.0
Owners	18.7	28.5	42.7	82.0
Siblings (%)
No siblings	12.1	8.5	9.3	11.1
1–2 siblings	58.2	66.8	65.7	78.3
3 or more siblings	29.4	24.7	25.0	10.6
College’s aspiration (%)
Yes	48.1	43.5	55.6	52.2
No	19.3	22.5	14.5	18.7
Don’t know	32.7	34.0	29.9	29.1
Number of Cases	487	494	408	1,589

Source: ITAGEN2 baseline survey

Note: Only the respondents of the target population of the follow-up

Table 2.

Outcomes of the Middle School final exam. Unweighted and Weighted frequencies.

Variable	Recent Immigrants	Preschool Immigrants	Second Generation	Natives
Unweighted frequencies
Outcome Middle school’s Final Exam (%)
Sufficient	56.2	47.9	39.1	22.2
Good	24.4	26.2	24.7	31.3
Very Good	9.7	15.0	20.5	23.9
Excellent	9.7	10.9	15.8	22.6
Total	100.0	100.0	100.0	100.0
Number of Cases	258	267	215	1,084
Weighted frequencies
Outcome Middle school’s Final Exam (%)
Sufficient	52.4	46.0	38.5	22.0
Good	27.1	27.1	26.5	31.6
Very Good	10.0	15.6	20.1	23.9
Excellent	10.5	11.4	14.6	22.5
Total	100.0	100.0	100.0	100.0

Source: ITAGEN2 follow-up survey

Table 3.

Secondary school. Attainment by nativity. Unweighted and Weighted frequencies.

Variable	Recent Immigrants	Preschool Immigrants	Second Generation	Natives
Unweighted frequencies
Secondary school (%)
Vocational	53.2	38.3	23.4	15.5
Polytechnic	29.8	33.3	38.0	34.1
High school	17.0	28.4	38.6	50.4
Total	100.0	100.0	100.0	100.0
Number of Cases	141	162	158	833
Weighted frequencies
Secondary school (%)
Vocational	52.7	35.7	23.5	14.3
Polytechnic	31.1	37.0	39.8	35.7
High school	16.2	27.3	36.7	49.9
Total	100.0	100.0	100.0	100.0

Source: ITAGEN2 follow-up survey

Similar trends are observed in terms of the choice of a secondary school. While almost half of Italians’ children are going on to high school, this proportion decreases among immigrants’ children, who instead privilege polytechnic and vocational schools. Vocational school is the most frequent choice among recent immigrants, only a small fraction of whom enroll in high school.

5 Results

5.1 The middle school exam score

In order to investigate scholastic achievement at the end of middle school, we ran three ordinal logit regression models. The outcome variable is the result of the middle school exam, taken at the end of the 8th grade. The lowest grade (Sufficient) is the reference category (Table 5).

Table 5.

Parameter estimates predicting the outcome of final middle school exam. Ordinal logistic regression model

	Model 1	Model 2	Model 3
Gender (vs. Male)
Female	0.594*	0.609*	0.665*
Generational Status (vs. Native)
Second generation	−0.654*	−0.349	−0.250
Preschool immigrant	−0.960*	−0.615*	−0.388
Recent immigrants	−1.220*	−1.108*	−0.677*
Country of origin (vs. Italy)
Albania		−0.258	−0.256
Yugoslavia		−0.631*	−0.721*
Macedonia		−1.477*	−0.956*
China		0.899*	1.308*
Morocco		−1.147*	−0.708*
Tunisia		−0.019*	−0.584*
Other developed countries		−0.156	−0.395
Others		0.097	−0.115
Parents’ education (vs. low)
High			1.044*
Medium			0.618*
Unkown			−0.201
Household possession (vs. rentals)
Homeowners			0.618*
Siblings (vs. 1–2 siblings)
No siblings			0.341*
3 or more siblings			−0.369*
μ1	−0.255*	−1.869*	−0.891
μ2	1.050*	0.454	1.617
μ3	2.129*	0.725	1.858*
Province fixed effect	yes	yes	yes
Observations	1,967	1,967	1,967
Wald χ2	262.34**	334.47**	514.33**

Source: ITAGEN2 follow-up survey

Note:

^*5% significance level

^**Log-likelihood test with the previous model (2) (3); or the null model in (1)

The first model includes gender and generational status as the sole variables. The results indicate that second-generation immigrants and immigrants have inferior results to those of natives. In particular, lower performances are associated with shorter lengths of stay in Italy. Model 2 adds the country of origin, indicated by the mother’s birthplace, to the analysis. Respondents originally from Yugoslavia, Macedonia, Morocco and Tunisia are significantly more likely to have lower outcomes compared to Italians. The addition of country of origin as a covariate weakens the effects of generational status. In this model specification, in fact, second generation immigrants do not differ significantly from natives. In the final model (model 3), we included the socioeconomic status of the family. As indicators we utilized the education level of the parents, the size of the family, and a dummy variable indicating homeownership. These variables are generally good predictors of scholastic achievement, as demonstrated by numerous studies on education (Haveman and Wolfe, 1995). We find that the scholastic achievement of children with less educated parents, who live in a rented house, and have more than 3 siblings, is significantly lower. The introduction of compositional variables softens the relation with generational status, providing evidence that only recent immigrants have lower results compared to native Italians. The coefficients associated with second generation immigrants and pre-school immigrants attenuate with the inclusion of the socioeconomic status variables. The inclusion of the last block of variables underscores the negative pattern among children with parents from Yugoslavia, Morocco, Tunisia and Macedonia, but also indicates a positive coefficient for students originally from China. Controlling for the other variables, Chinese students seem to have higher levels of achievement than natives.

With the inclusion of socioeconomic status and background characteristics, only recent immigrants differ from natives. This could mean that what really matters is the unobservable experience acquired with length of residence in Italy (e.g. linguistic proficiency or assimilation into the Italian school system), more than place of birth. On the other hand, the results indicating country of origin suggest different ethnic groups enact diverse strategies. The negative effect of the students from Macedonia, Yugoslavia, Morocco and Tunisia do not point to a relation with the duration of residence in Italy of the ethnic community. For example, if forms of discrimination are related to the history of a particular ethnic community, we might expect a lower coefficient for communities that came to Italy recently, in particular the Asians. On the contrary, however, among the Asiatic communities we observe a strong positive effect of China. Controlling for other variables, in particular immigration status and socioeconomic status, children of Chinese origin have higher outcomes than Italians. This result is consistent with studies conducted in other countries (Chiswick, 2004; Glick and White, 2004; Louie, 2001; Portes and Hao, 2004).

5.2 The transition to secondary school

After investigating the determinants of scholastic achievement at the end of middle school, we explored the transition to secondary school. The choice of secondary school is one of the most critical steps in youths’ educational paths, as it is strongly associated with post-secondary school attainment. Our principal aim is to determine whether generational status and ethnicity have an effect on the transition to secondary school. According to our hypothesis, in the absence of other effects, academic achievement is the only predictor. We used a multinomial logit regression (MLN) to predict the probability of enrolling in a polytechnic school or a high school as opposed to a vocational school. Using a multinomial model, we did not make assumptions about the ordering of the different categories, and we estimated different coefficients for each category. In the first model (model 1, table 6) the sole predictor is the final exam taken at the end of middle school. Under the hypothesis of complete meritocracy, previous scholastic achievement should be the only predictor for the choice of scholastic track. In this perspective, families would decide the appropriate scholastic path for their children by considering only their prior performance. The estimates show that previous outcomes are strong predictors of secondary school choice: the higher the previous outcome the greater the propensity to enroll in high school rather than a polytechnic or vocational school. In the second model, we added the same individual and background characteristics used in the previous analysis. As shown in table 6 (model 2), gender, parental education level and homeownership have a further effect on the choice of secondary school. Girls are more likely to avoid polytechnics in favor of either high school and or vocational school. Consistent with recent results (Flabbi, 2001), parental level of education influences the choice of school net of previous achievements.

Table 6.

Parameter estimates predicting Secondary school choice. Multinomial logit regression model

	Model 1		Model 2		Model 3		Model 4
	P	HS	P	HS	P	HS	P	HS
Outcome (vs. Sufficient)
Good	1.30*	1.83*	1.16*	1.47*	1.14*	1.39*	1.13*	1.38*
Very Good	2.01*	3.72*	1.78*	3.16*	1.76*	3.10*	1.73*	3.11*
Excellent	4.57*	7.10*	4.43*	6.55*	4.61*	6.84*	4.71*	6.91*
Gender (vs. Male)
Female			−0.19	0.73*	−0.24	0.69*	−0.26	0.67*
Parents’ education (vs. low)
High			0.56*	1.66*	0.74*	1.91*	0.75*	1.88*
Medium			0.85*	0.78*	0.98*	0.95*	1.00*	0.91*
Unknown			0.50*	0.47*	0.65	0.72	0.64*	0.69
Household possession (vs. rentals)
Home owners			0.71*	1.31*	0.35	0.63*	0.35*	0.67*
Siblings (vs. 1–2 siblings)
No siblings			0.27	0.47*	0.31	0.61*	0.33	0.69*
3 or more siblings			−0.04	−0.37	0.20	−0.06	0.22	0.00
Generational Status (vs. Native)
Second generation					−0.06	−0.26	0.02	−0.58
Preschool immigrant					−0.77*	−1.24*	−0.68*	−1.88*
Recent immigrants					−1.19*	−2.22*	−1.19*	−2.61*
Country of origin (vs. Italy)
Albania							−0.14	1.08*
Yugoslavia							−0.22	0.26
Macedonia							−0.10	0.46
China							0.09	0.00
Morocco							−0.54	0.51
Tunisia							0.23	0.52
Other developed countries							−0.48	0.47
Others							0.00	0.24
Constant	−0.80	−1.75	−0.92	−3.24*	0.04	−1.91	−1.70	0.28
Province fixed effect	yes	yes	yes	yes	yes	yes	yes	yes
Observations	1,775		1,775		1,775		1,775
Wald χ2**	772.54*		1027.66*		1124.66*		1145.98

Source: ITAGEN2 follow-up survey

Reference category: Vocational school; P polytechnic; HS High School

Note:

^*5% significance level

^**Log-likelihood test with the previous model (2) (3) (4); or the null model in (1)

In the final model (model 3) we added generational status to the covariates in order to test the effect of the latter on the choice of educational path. Our results indicate that immigrants have a significantly lower probability of enrolling in polytechnic and high schools compared to natives and second generation students. Adding country of origin to the regression model (model 4) does not significantly improve the explanation of secondary school choice. The likelihood ratio test, where we compare the performance of the regression model with the previous one, suggests using the model without country of origin.

6 Discussion

Italian Constitutional Law declares that the state should remove any obstacle of social or economic nature that impedes social mobility and that the highest degrees of education should be attained on the basis of merit². Our results suggest that the goals of this law are not always met. The highest degrees of education are partially determined by family background and country of origin. Immigrants’ children are more likely to achieve lower scholastic results and pursue shorter educational career paths with respect to their native counterparts. We simulated the probabilities of attending different typologies of secondary school for students who achieved a score of “Very good” at the end of middle school using an MLN regression model (Table 7). Recent immigrants have about half the probability of enrolling in high school compared to their native counterparts. Analogously, among students with lower grades, the probability of attending vocational school is 0.72 among recent immigrants and only 0.39 among natives.

Table 7.

Probabilities of secondary school choice given the result “Very Good” at the end of middle school. Simulations from estimates of MLN regression model. Standard Errors in parenthesis.

Immigrant generation	Vocational school	Polytechnic school	High school
Natives	0.05 (0.04)	0.31 (0.12)	0.63 (0.14)
Second generation immigrants	0.06 (0.05)	0.35 (0.15)	0.59 (0.18)
Preschool immigrants	0.12 (0.06)	0.39 (0.13)	0.49 (0.16)
Recent immigrants	0.28 (0.12)	0.39 (0.08)	0.33 (0.14)

Estimated probability from model 3, table 6

There appears to be an imbalance in educational opportunities to the disadvantage of immigrants’ children, even if they have spent their entire educational career in the host country. The Italian education system may be leaving behind potentially superior students of different origins while continuing to favor native Italians. From a meritocratic perspective, access to secondary school would be influences only by previous outcomes achieved during the scholastic year. However, our results show that this does not always happen in Italy. After controlling for previous scholastic results, Italian families seem to encourage their children to pursue higher educations. On the contrary, immigrant families’ children choose shorter educational paths that will likely exclude attending college.

The Italian school system is characterized by universal access and, since the majority of the schools are public, promotes relationships between social classes. In the majority of elementary and middle schools, pupils come from different social classes and share the same teachers. Despite this, our analyses show that differences in school attainment and achievements persist after the completion of compulsory schooling. In addition, we observe a further disadvantage on those students who have an immigrant background. One possible reason could be that children attend school for relatively few hours each day compared to other countries. Since - with some exceptions - all educational activity is carried out during the morning, the rest of the day is the family’s responsibility³. Children who come from families with low levels of education may be at more of a disadvantage if they receive less support or encouragement from their families to keep up with homework. In addition, children of immigrant parents may have less support if their parents are not proficient in Italian. Even if many immigrants’ children are born in Italy, a considerable number also arrive at young age and some of them have already attended school in their country of origin. Immigrant students need to quickly learn Italian as second language. Generally, schools do not provide extra classes in which foreign students can learn the new language.

In this paper, we examine the transition to secondary school. Our results suggest that, in addition to social class, immigrant status plays a role in the scholastic attainment of students. One possible explanation could be that young immigrants and their families have different preferences in terms of investing in education. They may prefer to start working earlier, or may be less interested in acquiring country-specific skills. Especially in case of those returning to their country of origin (or departing for another country), an investment in higher education may be considered too country-specific and therefore not profitable. Another possible explanation may be that the choice of a secondary school is affected by poor counseling received during the last year of middle school. Immigrants may be discriminated against and teachers may suggest that they enroll in a vocational or polytechnic school even if they have the same scholastic results of natives. Unfortunately, we do not have enough information to test these hypotheses. Our analysis is therefore not able to give a complete explanation of the differences in scholastic track among immigrants. We also cannot claim a causal effect of generational status on the choice of educational track since immigrant groups might be selected based on unobservable variables that we are unable to take into consideration in our regression models. Despite these challenges, our results are among the first to show differences in the acquisition of human capital on the part immigrants and their descendants living in Italy.

Appendix

A Sample selection

A common problem in studies of education is the selection of the sample and the presence of missing values due to attrition. In our case, we face three main sources of possible selection bias. First, we may not observe the outcomes of students who failed a grade and thus had not completed middle school at the time of the follow-up. Second, students who took the exam might have left school or changed the kind of school they were enrolled in after middle school. In other words, students attending secondary school might, in fact, have changed typology between the first and the second year. Third, some students were not interviewed in the follow-up. This last source of selection is the most sizable and troublesome. Table 1 shows that attrition is particularly significant among immigrants and second generation immigrants.

In response to the first two sources of selection, we restricted our analysis to students who passed the exam and enrolled in a secondary school. Moreover, we excluded from the sample students who never enrolled in a secondary school. The number of these cases is very limited, given that school is compulsory up to age 15 in Italy, and the middle school exam is usually taken at the age of 13⁴. Since we do not have the complete history of secondary school attendance, we classified students by the last school attended. These restrictions to the sample mean that our analysis is valid only for the subsample of students who completed middle school and enrolled in a secondary school.

Dealing with the attrition presents more of a challenge. In order to control for bias induced by the loss of data during the follow-up, we adopted a weighting strategy. First, we investigated whether the probability of being included in the follow-up varies according to several characteristics of the sample. In particular, we were interested in testing whether attrition is attributable to observable variables linked to the stability of the migratory process. To this end, we ran a probit model using inclusion in the follow-up as the dependent variable. As covariates we considered: generational status, gender, region of residence, number of siblings, indicators of scholastic proficiency, indicators of support and distance from relatives, home-ownership. If the subsample of the follow-up was a completely random selection of the target population, we expected that the effect of the covariates would not differ significantly from zero. This condition, called MCAR “Missing Completely At Random” (Little and Rubin, 2002), was not empirically supported. In fact, immigrants and those living in rented homes in the south of the country in families with less support from relatives are less likely to respond to the follow-up interview.

In order to compensate for unequal selection probabilities and response rates, it is common in most public-use surveys to weight the response cases. “Inverse probability weighting” is an extension of inverse weighting methods used in survey sampling and in missing data problems (Hogan and Lancaster, 2004). If the sample used to draw inferences is not a simple random sample (SRS), but rather one in which members of sub-populations are over- or under-sampled (e.g., recent immigrants or preschool immigrants), the inverse probability weights are computed by assigning to each unit its inverse probability of being sampled. In an SRS of n size, the sampling probability is 1/n, such that the relative weights are equal to 1. The weights themselves can be interpreted as quantifying the number of non-sampled members of the population represented by the sampled unit. For example, if the weight of an observed unit is 1/4, then this unit represents information from four members of the target population. In our analysis we utilized as weights the inverse of the predicted probability obtained by the probit selection model mentioned above. All the response units obtained in the supplementary survey were weighted 1. Table 2 and 3 illustrate the weighted and unweighted frequencies of the final exam and the choice of secondary school. Descriptive statistics give an indication that selection overestimates the proportion of students attending high school, particularly for immigrants. The comparison between weighted and unweighted frequencies suggests that students with lower scholastic achievements are more likely to be excluded in the follow-up sample.

The use of an inverse probability weighting method helped in obtaining approximate unbiased estimates of population means and totals (see e.g. Winship and Radbill, 1994). Although this is a standard practice, it is not without problems. As pointed out by DuMouchel and Duncan (1983), while weighting may be appropriate for estimating population means and totals, relying only on weighted estimates may be dangerous in regression problems.

One condition for unbiased estimates in regression models is that observations are “missing at random” (MAR). Under this condition, attrition is independent of missing responses, conditionally on observed responses and model covariates. This assumption is a component of the ignorability condition, which allows valid inference procedures to be based on the likelihood function only for the observed values. In other words, if attrition depends on a set of covariates, and we include in our regression model those covariates, the inference procedures are valid for the entire population. In the regression models presented in the following paragraphs, we assumed the MAR condition. To check the sensitivity of our models, we used a Heckman selection model to describe the exam outcomes. The Heckman selection approach is designed to help adjust for this nonrandom exclusion from the sample under a non-ignorable condition as well (Heckmann, 1974; Winship and Mare, 1992). We compared estimates from the Heckman selection model and a simple linear regression model. Given that substantial differences between the estimates did not occur, we argue that the ignorability condition is satisfied and the missing process is captured by the inclusion of covariates.

A.1 Selection Model

To compensate for the unequal selection in the follow-up interview, we adopted a weighting strategy. Using a probit regression, we modeled the probability of being included in the follow-up interview. Our model specification uses as independent variables several individual and background characteristics collected during the baseline survey. A description of the variables not presented in the previous sections follows. Table 8 shows the estimated coefficients utilized to construct the sample weights.

Table 8.

Parameter estimates predicting selection in the follow-up sample. Probit regression model

Variable	Coefficient	(Std. Err.)
Gender (vs. Male)
Female	0.095	(0.059)
Generational Status (vs. Native)
Second generation	−0.141	(0.094)
Preschool immigrant	−0.278**	(0.091)
Recent immigrants	−0.398**	(0.096)
Geographic region (vs. Veneto)
Marches	−0.019	(0.082)
Apulia	−0.217**	(0.081)
Calabria	−0.294*	(0.131)
Sicily	−0.414**	(0.096)
Distances from relatives
Less than 1 kilometer	0.000	(0.064)
Support from relatives (vs. Granparents)
Uncles	−0.019	(0.081)
Other relatives	−0.240*	(0.095)
Non relatives	−0.069	(0.113)
Nobody	−0.183*	(0.080)
How are you doing in school? (vs. I’m among the best in my class)
I’m doing pretty well	−0.155	(0.085)
I’m doing ok	−0.286**	(0.093)
I’m not doing so well	−0.369**	(0.118)
Secondary school aspiration (vs. No school)
High school or Polytechnic	0.560**	(0.197)
Vocational	0.637**	(0.197)
Don’t know	0.457*	(0.200)
Household possession (vs. rentals)
Home owners	0.325**	(0.071)
Siblings (vs. 1–2 siblings)
No siblings	0.093	(0.094)
3 or more siblings	−0.264*	(0.113)
Does the family you live with owns 50 books? (vs. No)
Yes	−0.136*	(0.061)
Intercept	0.063	(0.271)

Source: ITAGEN2 follow-up survey

Significance levels

^*5%

^**1%

Distances from the relatives

As a measure of family ties, we collected information on the distance that grandparents and uncles/aunts live from the respondent. The categories were collapsed into two cases: a distance of less than 1 kilometer if the closest grandparent or uncle/aunt lives within that distance; a distance of over 1 kilometer if they live farther or in case that none of the grandparents or uncles/aunts is alive.

Support from relatives

Respondents were asked about support that the family receives from relatives. The question was ex-pressed as follows: “If they need something, who do the adults you live with usually turn to?” Possible choices include: grandparents, uncle(s) or aunt(s), other relatives, someone who is not relative, no one.

How are you doing in school?

To collect respondents’ perception of their own scholastic achievements, children were asked how they are doing in school. Possible answers include: I’m among the best in my class, I’m doing pretty well, I’m doing Ok, or I’m not doing so well.

Secondary school intentions

Respondents were asked if they think they will go on to secondary school and which school they think they will attend.

Does the family you live with owns 50 books?

As measure of education and consumption of cultural goods, respondents were asked whether the family they live with owns at least 50 books (non-scholastic texts).