Educational integration of foreign citizen children in Italy: a synthetic indicator

(1)

(2)

(3)

Abstract Social inclusion and integration are prominent topics in social sciences, and specifically in the educational field, in order to contrast the experience of in-equality conditions at school. Using data from the 2015 national survey on “Integra-tion of the second genera“Integra-tion”, the paper proposes a synthetic indicator of educa-tional integration of second-generation children in Italy. The chosen methodology is suitable to model subjective assessments on rating scales while accounting for both agreement and heterogeneity in response patterns.

Abstract L’inclusione sociale e l’integrazione rappresentano temi cruciali nelle scienze sociali ed assumono rilievo specifico nell’ambito dell’istruzione. Utiliz-zando i dati dell’Indagine Istat sull’integrazione delle seconde generazioni (ISG 2015), lo studio propone un indicatore sintetico di integrazione scolastica che con-sidera sia la percezione del tratto in esame che l’eterogeneit´a delle risposte. Key words: Model-based indicators, Educational integration, Second-generation, Foreign children, Rating data

1 Introduction

The issue of migrants’ integration is a key topic in social sciences, and it is particu-larly relevant in the educational field to counteract the experience of inequality con-ditions by children with a migratory background, especially of those born abroad. Focusing on Italy, the number of foreigners has rapidly increased in the last two

Alessio Buonomo

University of Naples Federico II, e-mail: alessio.buonomo@unina.it Stefania Capecchi

University of Naples Federico II, e-mail: stefania.capecchi@unina.it Rosaria Simone

University of Naples Federico II, e-mail: rosaria.simone@unina.it

1

(18)

2 Alessio Buonomo, Stefania Capecchi and Rosaria Simone

decades from about 350,000 residents in the early nineties to over 5 million, accord-ing to the Population Registers. The consequent growaccord-ing number of their children have attracted the interest of scholars [10].

Due to the fact that the increase of the second generations in Italy is a rela-tively recent phenomenon, the school environment is a particularly important field of study [3]. Educational and social integration can boost placement into community and labour market [1], especially in the case of immigrants, and thus be beneficial to the whole society. Indeed, studies that focus on immigrant-specific educational inequalities have proven that social and scholastic integration are fundamental in determining educational success [8, 4].

Considering theCI-CUB procedure proposed by Capecchi and Simone [2], the paper aims at introducing an indicator of integration at school for both Italian and second-generation foreign children, distinguishing by nationality. The study pro-vides a synthetic measure able to summarize subjective evaluations expressed by a large number of respondents, relying onCUBmodels framework [9].

After a brief introduction to the available data, the implemented methodology is illustrated, some results are discussed and further developments of the research are outlined.

2 Data and methods

The “Integration of the second generation” survey (ISG) has been conducted in Italy by the Italian National Statistical Institute [5] in 2015 and offers information on students of secondary schools in Italy. The overall unweighted sample amounted to 68, 127 individuals, of which 52% were students in upper secondary school. In order to deal with a more homogeneous subset, our analyses focus on opinions of students attending higher secondary state schools. Thus, the final target sample consists of 35, 427 unweighted individuals.

We investigate students’ assessment - in terms of agreement - to 20 items referred to their perceived conditions/habits gathered in four different batteries: (A) relation-ships with classmates; (B) homework and study habits; (C) relationrelation-ships with teach-ers; (D) family attitudes towards school importance. Variables of interest are com-prised in question 14, in section C of the questionnaire. Each battery is composed by 5 items, whose responses are expressed on a 5 point Likert scale (1=“strongly agree”, 5=“strongly disagree”). For interpretative purposes, the variables were re-versed (with the exception of items A 3, B 3, B 5 and D 4) and organized in order to have the same direction, from the lowest to the highest agreement.

In the target sample, male students are about 50%; 15% of respondents are under the age of 15, and 58% are aged 15-17, whereas 27% are over 17. Italian intervie-wees represent 53% of the sample. Among foreigners, the three more frequent na-tionalities are Romanian, Albanian and Moroccan (10%, 7% and 4%, respectively). Foreign children born in Italy (second generation) are 9% of the sample, whereas those born abroad are 38%. Since foreign students represent a heterogeneous group

(19)

A synthetic indicator of educational integration 3

due to different origins and characteristics, a synthetic measure of their perception of integration may be useful both for research and policy purposes.

In order to illustrate multifaceted and sometimes elusive constructs, as in the case of subjective evaluations and perceptions, the number of composite indicators is growing. In fact, to general readership it may be easier to comprehend a single synthetic measure rather than inspect results from different separate indicators [7].

When dealing with ordinal data,CUBmodels allow to consider the uncertainty component in response behaviour, usually neglected in standard approaches. In this respect, Capecchi and Simone [2] recently proposed the novelCI-CUB- Composite IndicatorCUB- to compute a model-based indicator within theCUB framework [9]. In its simplest version, aCUBmodel (acronym of Combination of Uniform and shifted Binomial) is implemented on ratings R to single questions to provide a global measure of individual assessment.CUBmodels are specified via a two component mixture of discrete distributions, one designated to measure the feeling towards the item (Binomial), and the other accounting for the inherent uncertainty (Uniform).

Then, aCUBmodel assumes that, for a given number of categories m, the random variable R is specified by the probability mass distribution:

Pr(R = r | θ) = πm − 1 r− 1

ξm−r(1 − ξ )r−1+ (1 − π)1

m, r = 1, 2, . . . , m. The parameter (1 − ξ ) of the Binomial and the weight of the uncertainty in the mixture (1 − π) provide a description of both perception towards the item (which can be indicated as “agreement” with given statements on perceived integration in the present case study), and heterogeneity of the response distribution. Since the two featuring parameters are normalized measures ranging in (0;1], different items and groups can be displayed and compared at a glance in the unit square. Therefore,

CUBmodels establish a different paradigm, providing a unifying tool to interpret and compare the responses given to a number of items, even if measured on scales with different lengths, in a parsimonious way. The possibility of assessing the veracity of the rating evaluations with respect to the analyzed items/dimensions by means of easy-to-interpret graphical outputs is indeed one of the strengths ofCUBframework. A comprehensive R library is devoted toCUB model fitting on the official CRAN repository.

In this setting,CI-CUBindicators provide a synthesis ofCUBmodels applied to K items by n respondents: the random variable Rkassociated to the k-th item

fol-lows Rk∼CUB( ˆπk, ˆξk), for k = 1, . . . , K. Then, theCI-CUBindicator is a weighted CUBmodel ˜R∼CUB( ˜π , ˜ξ ) where

˜ π = K

∑

k=1 wkπˆk, ξ =˜ K

∑

k=1 wkξˆk, (1)

with ( ˆπk, ˆξk), for k = 1, . . . , K being the estimated parameters for the K variables of

interest.

(20)

ACI-CUB can be seen as a two-dimensional composite indicator for the latent trait under investigation such as, in this case study, the perceived educational inte-gration as measured through the 4 batteries of items (A, B, C, D).

The choice of the weights’ system w_kis usually driven by specific circumstances and available information. In the present study, due to the underlying structure of the data along the 4 dimensions, we select the weights expressed by the (normalized) squared factor loadings of the 4 categorical principal components computed on the selected variables.

3 Results

While additive data-driven approaches may lead to neglect an inherent uncertainty component when present,CUBmodels separately estimated for each of the 20 items of question 14 (here not reported for space constraints) show a medium-high level of agreement, with the uncertainty being substantially not negligible for most of the items. More specifically, item C 1, referred to the perception of an equal treatment received by teachers, shows the highest level of uncertainty with a substantive agree-ment; on the other hand, item D 4 gets the highest level of agreement with a low uncertainty, meaning that family members do consider studying to be a prominent aspect of life success.

Analysing the whole dataset, the advantages of the implemented CI-CUB indi-cators are highlighted. In order to provide synthetic probabilistic measures of stu-dents’ perceived integration,CI-CUBprocedure has been implemented according to a twofold trajectory: results are presented graphically in Figure 1 where responses are aggregated by item subgroup and nationality. Notice that the parameter space is narrowed to improve readability, since a large part of the indicators lays in the uncertainty range 0.2 − 0.45, and the agreement is always higher than 0.5. To save space, details about fitting procedure are not reported.

Left panel of Figure 1 shows the synthetic indicators for each item subgroup, corresponding to the above mentioned batteries (A, B, C, D), as compared with the overallCI-CUBindicator (ALL). For the items altogether (ALL), the level of agreement is quite high with a moderate level of uncertainty.

It seems evident that items referred to family attitudes (D) are those for which the overall level of agreement is the highest, immediately followed by those related to schoolmates (A), both manifesting a low level of uncertainty; estimates for the two subgroups pertaining to study habits (B) and assessments of relationships with teachers (C) get a lower agreement, with a larger uncertainty. Moreover, this repre-sentation allows to consider the impact on the overall indicator (ALL) of deleting one subgroup of items at a time. Specifically, removing B and C subgroups, the resulting synthetic indicators (−B and −C) show higher agreement and lower un-certainty, therefore highlighting the contribution of these two subsets in terms of feeling. These results are expected given the mutual positions of the subsets in the parameter space.

(21)

A synthetic indicator of educational integration 5

Fig. 1 Synthetic Indicators by item subgroups (left) and nationalities (right)

Right panel of Figure 1 collects CI-CUBindicators by respondents’ nationality. In general, the agreement level is higher than 0.6 and the uncertainty spans be-tween 0.10 and 0.43. Specifically, respondents from FYROM and Tunisia manifest the highest levels of agreement. Chinese respondents are quite far from the others in terms of feeling. Interestingly, Italian students express one of the lowest level of agreement, followed by respondents from China and Philippines. This finding is consistent with the so-called “Immigrant optimism approach” that assumes that im-migrants are a positively selected group: individuals that leave the country of origin are supposed to be more motivated, ambitious, stronger and optimistic as compared to the national counterpart [6].

Table 1 CI-CUBestimated probabilities of the highest agreement Nationalities Pr( ˜R= 5) Nationalities Pr( ˜R= 5) China 0.165 Egypt 0.308 Philippines 0.237 Romania 0.318 Italy 0.252 Albania 0.320 Serbia 0.260 Morocco 0.356 Pakistan 0.269 Ukraine 0.356 Ecuador 0.272 Tunisia 0.368 Other 0.295 FYROM 0.415

1452

(22)

Table 1 presents the estimated probabilities to express the highest rating ( ˜R= 5), conditional to the nationality, which can be considered as an overall measure of edu-cational integration, as derived by theCI-CUBmodels. In line with evidence in Fig-ure 1 (right panel), Chinese students have the lowest estimated probability, whereas Macedonians (FYROM) have the highest. It turns out that there is not a clear geo-graphical pattern of the estimated probabilities in terms of agreement, since differ-ent and distant countries presdiffer-ent very similar estimated probabilities, see: Morocco, Ukraine and Tunisia, or Serbia, Pakistan and Ecuador.

Further developments of the study will follow two directions. From a method-ological point of view, the main effort will be devoted to derive the confidence ellipses. For the empirical analyses, more investigation is needed on a geographi-cal basis, as well as on the effects of individual characteristics, such as gender and migratory generation.

Acknowledgements The paper is included in the Research Program on “School inclusion strate-gies and social cohesion challenges of immigrate immediate descendants in Italy”, SCHOOL/GEN2 (corresponding proponent Giuseppe Gabrielli), supported by grants from University of Naples Fed-erico II (2017-2018), D.R. n. 408 07/02/2017 (CUP: E66J17000330001).

References

1. Ballantine, J. H., Hammack, F. M.: The Sociology of Education: A Systematic Analysis (7th ed.). Pearson, New York (2012)

2. Capecchi, S. Simone, R.: A Proposal for a Model-Based Composite Indicator: Experience on Perceived Discrimination in Europe. Soc. Indic. Res. 141, 95–110 (2019)

3. De Santis, G., Pirani, E., Porcu, M. (eds.): Rapporto sulla popolazione. L’istruzione in Italia. Il Mulino, Bologna (2019)

4. Hadjar, A., Scharf, J.: The value of education among immigrants and non-immigrants and how this translates into educational aspirations: a comparison of four European countries. Journal of Ethnic and Migration Studies 45, 711–734 (2019)

5. Istat, L’indagine sull’integrazione delle seconde generazioni: obiettivi, metodologia e orga-nizzazione. Roma (2017)

6. Kao, G.,Tienda, M.: Educational aspiration of minority youth. American Journal of Education 106, 349–384 (1998)

7. OECD, Handbook on Constructing Composite Indicators, Methodology and User Guide. OECD, Paris (2008)

8. Phillips, D.: Minority Ethnic Segregation, Integration and Citizenship: A European Perspec-tive. Journal of Ethnic and Migration Studies 36, 209–225 (2010)

9. Piccolo, D., Simone, R.: The class of CUB models: statistical foundations, inferential issues and empirical evidence. Stat. Methods Appl.28, 389–435 (2019)

10. Strozza, S., De Santis, G., (eds.): Rapporto sulla popolazione. Le molte facce della presenza straniera in Italia. Il Mulino, Bologna (2017)

Educational integration of foreign citizen children in Italy: a synthetic indicator

 



         

  

Contents

Specialized sessions

Accounting for record linkage errors in inference (S2G-SIS)...2

Advanced methods for measuring and communicating uncertainty

in ofﬁcial statistics ...9

Bayesian methods in biostatistics ...27

Data integration versus privacy protection:

a methodological challenge? ...40

Designing adaptive clinical trials ...59

Education: students’ mobility and labour market...66

Environmental Statistics (GRASPA-SIS) ...85

Methods for High Dimensional Compositional Data Analysis ...98

Modern Statistics for Physics Discoveries ...117

Network Modelling in Biostatistics...130

New issues on multivariate and univariate quantile regression ...137

Recent methodological advances in ﬁnite mixture modeling

with applications (CLADAG-SIS) ...150

Statistical Analysis of Satellite Data (SDS-SIS) ...163

Statistical models and methods for Business and Industry ...176

Technology and demographic behaviours (AISP-SIS)...195

Solicited Sessions

Advanced Statistical Methods in Health Analytics ...209

Advances in textual data mining ...228

Advances in the interaction between artiﬁcial intelligence and ofﬁcial

statistics ...241

Capture-recapture methods ...259

Changes in environment extremes and their impacts ...278

Copulas: models and inference ...291

Data Science: when different expertise meet ...316

Emerging challenges in ofﬁcial statistics:

new data sources and methods ...340

Factorial and dimensional reduction methods for the construction

of indicators for evaluation (SVQS-SIS)...359

Gender statistics ...384

How Deep is Your Plot? Young SIS and deep statistical learning (ySIS)..397

Measuring poverty and vulnerability ...421

New advances in applications, a Bayesian

nonparametric perspective ...428

Non-Parametric Inference and Forecasting

of Functional and Object Data ...435

Prior choice in Bayesian Modelling (SISbayes) ...446

Sequence Analysis: methods and applications ...465

Socio economic integration of migrants ...472

Statistical Analysis for mobility and transportation ...479

Statistical Methods and Applications in Social Network Analysis ...496

Statistical Methods in Psychometrics ...521

Statistical modelling in environmental epidemiology ...546

Statistical Modelling of Scientiﬁc Evidence for Forensic Investigation

and Interpretation ...557

Topological learning: interpretable representations of complex data...581

Contributed papers and Posters

Bayesian Statistics ...607

Biostatistics ...720

Environment, Physics and Engineering ...809

Finance, business and ofﬁcial statistics ...886

Machine Learning and Data Science...1023

Models and methods - Categorical, Ordinal, Rank Data ...1135

Models and methods – Regression...1206

Models and methods – Sampling ...1271

Models and methods - Theoretical Issues in Statistical Inference ...1314

Models and methods - Time Series and Longitudinal Data...1350

Population and society ...1411

Educational integration of foreign citizen

children in Italy: a synthetic indicator

Un indicatore sintetico dell’integrazione scolastica degli

studenti con cittadinanza straniera

∑

∑

1452