The quality of an educational system is strictly related to its capability to enhance the performances of students and to reduce inequality related to their initial conditions. Many researchers document the persistence in Italy of marked differences across educational institutions and geographical areas with respect to this specific task (Almalaurea, 2021; Priulla et al., 2021).
The family, the school environment and the peers have also an important influence on educational choices and outcomes, as well as in the transmission of inequalities (Moogan et al., 1999). Starting from the transversal priorities of the PNRR related to ensuring equal generational, gender and territorial opportunities, the project aims to study some facets of the mechanisms of reproduction of inequalities within the Italian educational system, with particular attention to the study of the role that the school and academic environment play in shaping interactions between peers, aiming at reducing initial disadvantages and gender gaps. In this context, intangible assets such as the so-called soft skills (as leadership, creativity, self-efficacy, and risk-ropensity) are recognised as key variables, fundamental for the development of an innovative mind-set of young people, and consequently their personal development (Sica et al., 2019, 2022). Soft skills can change in line with different contexts, constraints and school types and ethos that potentially shape behavioural expression (Chell & Athayde, 2011). The study intends to define how and in which contexts the effect of peers, observed at micro and meso level, interacts in the process of reproduction of inequalities in learning outcomes, in the persistence and transmission of intergenerational and gender disparities at school and in the subsequent transition to the university; to attain these objectives, we will take into account the geographical differences and the heterogeneity in the teaching programs. Finally, a further aim is to detect which schools and universities have been able to develop institutional arrangements and practices in such a way to perform as social equalisers, identifying those that have also triggered virtuous and egalitarian processes. The information provided at national level by the Italian surveys carried out by INVALSI in high schools will be used to i) assess divergences in learning outcomes at grade 13, ii) detect inequalities issues, and iii) investigate the role of soft and disciplinary skills. The MOBYSU.IT dataset and the INVALSI data will be used to assess school-university transition and to monitor university students’ performance. Ad hoc surveys will be carried out to investigate how soft skills and social support along with peer effect explain individual performances. The survey will be conducted administering a questionnaire to samples of students attending either high school or university in different regional contexts by adopting social network research designs.
- State of the art
The distribution of educational skills in a society is a key component of inequality and the relationship between individual skills and family background is pivotal to social and intergenerational enhancement (OECD, 2012). The quality of an educational system could
be measured considering the capability of institutions to improve the performances of their students and to reduce inequality. Therefore, equity deals with the capability of the educational system to offer equal opportunities to students from different demographic, ethnic and socioeconomic backgrounds and to reduce performance disparities (Field et al., 2007; Agasisti & Longobardi, 2014; OECD, 2016).
Even though in the literature there is no consensus on exactly what “peers” mean (Weihua, 2011), a host of studies provides evidence that peers’ composition and interaction influence educational and social outcomes at all grades in different manners. We can refer to peers as friends, classmates, colleagues, neighbours and so on. Sacerdote (2011) highlights that the definition of “peer effect” includes all externalities that are produced by sharing the same educational environment. In this sense, the peer effect encompasses the spill over strictly related to peers’ ability, to their family’s socioeconomic status and to their parents’ involvement in the school management and the balance between the overall composition of the peers’ group.
In the educational setting, adopting the lens of social network analysis, the peer influence on individual performances cannot be reduced only to the effect of being part of the same learning environment. To consider the individual level, the effects of classmates, friends and best friends are all related to social influence (Mercken et al., 2010), social learning and social support mechanisms (Aleni Sestito et al., 2008). One of the potential sources of inequality relies on the different possibilities of building social capital and benefiting from it of various social groups. The key competencies in this category are required for students to learn, live and work with others. In literature, the terms, such as social competencies, social skills or soft skills (OECD, 2005), are associated with this kind of competencies as a mix of individual innate potentials and social constructions (InvalsiOpen, 2020). The peer effect should be carefully evaluated in this process. Such effects, mediated by the social networks in which the students are embedded, must be considered in analysing individual performances and scholastic careers because they affect individual disciplinary skills and, more deeply, soft skills.
The identification of peer effect has been a widely questioned issue as it inevitably raises self-selection circles, as students with similar backgrounds tend to select similar educational institutions and to make similar choices. According to Manski (1993), the social influence between peers can be related to three kinds of effects: endogenous, contextual, and correlated (Fletcher, 2015).
The peer effect is considered the endogenous one and it is related to how peers’ behaviour affects individual choices through social interactions.
This effect can be estimated only by accounting for the contextual effect, which captures the influence of exogenous factors as peers’ socioeconomic background on students’ and families’ behaviours, and the correlated effects, which can cause spurious correlations between students’ behaviours that are linked to external factors, such as the institutional environment faced by the students (Bramoullé et al., 2009. Peer effect is also pivotal looking at gender segregation in higher education in Italy. Barone & Assirelli (2020) highlight that peers’ behaviours influence student choices in terms of selection of the field of study; nonetheless, individuals’ educational choices are also driven by the school environment, counsellor services and teachers’ recommendations and are correlated with the preferences of classmates. According to Engberg & Wolniak (2010), the high school environment is one of the drivers of college choice decisions and it is influenced by students’ endowment in terms of cultural and social capital. Looking at a gender perspective, peers’ behaviours influence student choices also in terms of track. In this respect, an environment with gender-normative ideas pushes girls out of the STEM pipeline (van Der Vleuten et al., 2018), so that previous studies have identified an association between students’ gender and dropout, with female students more likely to drop out than males in STEM disciplines (Isphording & Qendrai, 2019). On the other hand, students tend to adjust their preferences to those of their friends, and female students tend to retain their STEM preferences when other females in their classroom do so (Raabe et al., 2019).
- Detailed description of the project: methodologies, objectives and results that the project aims to
achieve and its interest for the advancement of knowledge, as well as methods of dissemination of the results achieved
The project aims to investigate the role that peers’ environment at class, school, or geographical level has in affecting student learning outcomes, postsecondary educational choices and gender- oriented behaviours. Even more, the project intends to evaluate if and how the interactions within the school environment and the students’ endowment in terms of social relationships and soft skills can mitigate the influence of disadvantaged backgrounds (e.g., socioeconomic status) in affecting secondary and tertiary education performances and educational choices. The information provided by the INVALSI survey, which evaluates student skills in different learning areas at grade 13, will be used to carry on an analysis of the determinants of divergences in school/classes and territories, in terms of quality and equity, whereas the information provided by the MOBYSU.IT (see below in this section) dataset will be used to assess school-university transition and to monitor university students’ performance. To deepen the interplay of social relationships, social support, disciplinary and soft skills, peer effect measured at micro level will be investigated relying on both the INVALSI survey carried out in high schools and ad hoc surveys on samples of students enrolled either at high school or university in the regional contexts involved in the project. More specifically, this project will focus on the following main three objectives:
A. HIGH SCHOOLS
This objective is twofold:
A.1. PEER EFFECTS IN HIGH SCHOOLS. To identify the peer effect and its influence in the reproduction of divergences in educational achievement in high schools. The project aims to assess divergences between schools and territories and their determinants, with a focus on schools which operate in unfavourable conditions in terms of geographical location or students’ background composition.
A.2 PEER EFFECT ON SOFT SKILLS AND THEIR INTERPLAY WITH DISCIPLINARY SKILLS. Within this aim we intend to assess the peer effect mediated through relationships among students and social support on the soft and disciplinary skills at individual level. This task will be pursued by using both the INVALSI survey containing information on soft skills (InvalsiOpen, 2020) and OECD data, as well on data coming from ad hoc surveys on samples of high school students aiming at connecting the network data on friendships, support, advice relations among students, with their soft skills, vocational identity, their choices about future, psychological characteristics, and disciplinary skills. The results could provide hints on which social network structures promote equity in high school and which ones deepen the inequalities.
B. THE TRANSITION FROM HIGH SCHOOL TO UNIVERSITY
This objective is twofold:
B.1 SCHOOLS AND PEERS IN THE PROCESS OF TRANSITION FROM HIGH SCHOOLS TO UNIVERSITY. This objective includes an analysis of differences in the profiles of students enrolled in the last class of the high schools and those who have enrolled in Italian universities, looking for differences among students’ socioeconomic conditions, territories, and types of high schools. In this step, the research will pay attention to disentangle the peer effect from other micro, meso and macro effects, by focusing on the divergences in the propensity to enrol at the university among students who have faced similar institutional and geographical environments.
B.2 PEER EFFECT AND GENDER-ORIENTED BEHAVIOURS IN THE TRANSITION FROM HIGH SCHOOL TO UNIVERSITY. To assess if and how the presence of peers of the same sex influences students’ transition from high school to the university. A comparison will be made in their propensity to enrol at the university between gender-unbalanced degree programs and more gender-balanced ones.
C. UNIVERSITY CAREER
This objective is threefold:- C.1 PEER EFFECT AND UNIVERSITY CAREER. To assess the influence of the high school context, as the environment where peersinteract, in students’ university career. In this step, the analysis will focus on the effects that high school and peers’ interaction havein determining divergences in the outcome indicators used for monitoring students’ performances at the university (e.g., dropout, graduation, regularity of university careers) with the main aim of quantifying and disentangling the three main effects (contextualeffects, correlated effects, and peer effects) that interact in this process.
- C.2 PEER EFFECT AND GENDER BIAS IN UNIVERSITY STUDIES. To assess if and how the presence of peers of the same sex influencesstudents’ academic successful or unsuccessful career, in terms of graduation, dropout or churn across the different fields of study. In addition, a comparison will be made between gender-unbalanced courses and gender-balanced ones.
- C.3 PEER EFFECT AND SOCIAL INFLUENCE IN UNIVERSITY STUDIES. To assess how the emergence of informal student communities and different kinds of social support coming from peers affect the academic engagement and hence the academic performance. This task will be pursued by promoting ad hoc surveys based on egocentric network studies on a sample of students enrolled at university.
The expected results with respect to each of the above listed objectives are:
A. HIGH SCHOOLS
A.1 PEER EFFECTS IN HIGH SCHOOLS. To identify high schools that have adopted practices able to foster resilience and promote quality and equity in education and to assess if there exists a multiplicative effect from the interaction between peers. As a result, we expect:
– to build up a system of quality and equity indicators to map schools;
– to identify profiles of schools which have reached high performances besides the adverse exogenous conditions.
A.2 THE ROLE OF SOFT AND DISCIPLINARY SKILLS IN PROMOTING EQUITY IN HIGH SCHOOLS. The main purposes are both to determine the interplay of social relationships and social support, soft skills, individual performance and choices on future perspectives, and to assess their role in reducing or reproducing inequalities.
B. THE TRANSITION FROM HIGH SCHOOL TO UNIVERSITY
B.1 SCHOOLS AND PEERS IN THE PROCESS OF TRANSITION FROM HIGH SCHOOLS TO UNIVERSITY. The analysis aims to detect high schools that have been able to reduce the inequalities of access to the higher education system. Two outcomes are expected:
– to identify the size of peer effect in students’ university choices;
– to map schools such as to identify those who have a higher transition rate from high schools to university than expected on the light of their characteristics in terms of geographical location, typology and social composition of students.
B.2 PEER EFFECT AND GENDER-ORIENTED BEHAVIOURS IN THE TRANSITION FROM HIGH SCHOOL TO UNIVERSITY. To identify schools and university degree programs that show gender balanced enrolments, and to verify if it may depend on a same-sex peer effect from high school, the presence of a committee on gender budgeting in the athenaeum, or the geographical location of both the high school and the university itself.
C. UNIVERSITY CAREER
C.1 PEER EFFECT AND UNIVERSITY CAREER. To identify universities and degree programs, which have been able to reduce inequalities in students’ tertiary education success opportunities. The expected outcomes are:
– to assess the peers’ effect in determining successful university careers;
– to build up a system of equity indicators to map degree programs and universities, which have been able to reduce inequalities in tertiary education opportunities.
C.2 PEER EFFECT AND GENDER BIAS IN UNIVERSITY STUDIES. To identify universities and degree programs, which have been able to reduce gender inequalities in students’ university career, both at the micro- (i.e., same-sex peer effect) and at the macro- (i.e., presence of a committee on gender budgeting in the athenaeum) level.
C.3 PEER EFFECT AND SOCIAL INFLUENCE IN UNIVERSITY STUDIES. To assess the effect of different ego-centred network profiles on student careers and student academic engagement to be part of the university community and learning community.
DATA AND METHODS
To fulfil the above-mentioned goals, we will use different data archives:
I. MOBYSU.IT database, realised thanks to the Italian Ministerial grant PRIN 2017 “From high school to job placement: micro-data life course analysis of university student mobility and its impact on the Italian North-South divide”, has been built up using the following data sources:
a. ANAGRAFE NAZIONALE STUDENTI (ANS)
Administrative data on the population of students that enrolled in an Italian university between 2010 and 2020. The ANS data contains information on university students’ career, individual characteristics, and high school background and follows students’ careers from their enrolment until the academic year 2019/2020.
For each year of students’ careers, the following information are recorded: the university, the city that hosts the university, the type of degree (e.g. bachelor, master), an indicator on whether the university is an e-learning institution, the specific program chosen and the disciplinary group (i.e. the so called “classi di laurea” and ISCED-F 2013 classification), the ECTS earned by students, whether students have opened also other careers in the Italian university system, information on whether students are changing program, or university, the year of graduation and the final grade. With respect to students’ individual characteristics, the ANS provides information on students’ date of birth, citizenship, sex, and city of residence; students’ high school background in terms of high school curriculum, year of diploma, final grade and the exact school attended by the student along with information on whether the school was private and on where it was located. The ANS database allows us to precisely identify students’ trajectories from high school to university and the events which occur during their university careers as well as peer groups of students in high school and university.
b. INVALSI
Micro data on the high school careers of students that have obtained their high school diploma in Italy between 2019 and 2020, and that have enrolled in an Italian university in academic year 2019/2020 and 2020/2021. These data are linked to the ANS by using an anonymous identifier (exact match). For each student we observe: the high school identifier (linked with ANS), the Economic and Social Status indicator (ESCS), students’ grades in standardised and classical tests, an indicator on whether each student is regular in his/her career, the level of education and professional status of parents, the number of computers in the high school, and the number of rooms dedicated to computers and informatic courses. For students that have concluded high school in the school year 2018/2019 this information is at two points in time: high school grade 10th and 13th (second and fifth year). In detail, the ESCS indicator is a measure of students’ socio-economic background that depends on parents’ educational and professional status as well as on whether students have specific goods (e.g., encyclopaedias) at their home. This indicator is constructed by following the standard procedure of OCSE-PISA and it is available at individual, classroom and high school level. This last element also allows us to infer the socioeconomic status of students that, while attending the same high school, have not enrolled at university. Moreover, the ESCS indicator is a key element to understand the education inequalities based on students’ individual socio-economic status. With respect to students’ performances in standardised tests, we observe students’ grades in English (reading and listening), Italian, and Maths.
c. HIGH SCHOOLS DATABASE RELATED TO ANS
This is the high school database that includes aggregate data on all Italian high schools between 2015 and 2020. This database provides information on the high school identifiers (linked with ANS), the type of high school (e.g., lyceum, technical, etc.), the number of students (grouped by sex) admitted at the final exam and those who obtained the diploma, and the number of foreign students divided by citizenship. This database, combined with ANS and INVALSI, allows us to infer both the number of students that have not enrolled in an Italian university for each school as well as some characteristics of these students (e.g., how many students with a final grade of 60 have not enrolled).
d. OPEN DATA USTAT
This open data portal, managed by the Italian Minister of University and Research, contains several information on universities’ characteristics such as: the number of professors grouped by disciplinary groups and nationality, fees revenues, and the number of scholarships and places in dormitories provided etc.
e. GEOGRAPHICAL REFERENCES
This dataset contains distances between students’ city of residence and each observed university or high schools by using the data gathered from ISTAT and google maps. II. INVALSI SURVEYS: micro data related to the INVALSI grade 13th survey gathered in the last class of high schools in Italy (INVALSI-G13).
This dataset contains information described in source (b) but extended to the entire population of high school students. The student questionnaire also contains information on students’ perception of classroom and school climate, students’ attitude towards mathematics and reading.
III. PRIMARY DATA COLLECTION. Even though it is possible to define groups of students using administrative archives to analyse the peer effect mediated by social influence mechanism, ad hoc surveys will be conducted to define network data descending from the observed ties among units.
Social network data collection will be conducted to reconstruct the kinds of network support to identify successful scholastic and university careers. Study design, research ethics, communication, and population definition will be considered to manage practical challenges related to the network data collection process. More specifically, egocentric network studies design will be adopted based on samples of students either attending high schools or universities in different regional contexts. In the questionnaire, students (egos) will list their own alters, and describe the types of relationships they share with them, the relationships among alters, the attributes of egos and alters are collected as well. The point of asking about alter–alter ties is relevant to assess network closure around students. Alter–alter ties, indeed, constitute a whole network with the boundary defined by actors linked to students. For the questionnaire design, we can adopt two possible approaches: the name generator, where students must elicit the names of the persons in their social networks, or the role relation approach, where network members are represented by role relations (e.g., parents, friends, roommates). Multiple dimensions of social ties will be measured on various kinds of support for tangible and intangible sharing resources (instrumental, emotional). The questionnaires will also contain a set of psychometric scales useful to measure interesting soft skills (like creativity, adaptability, self-efficacy), the perceived social support, the academic engagement, and the vocational identity
METHODOLOGIES
MULTILEVEL METHODS FOR ASSESSING DIVERGENCES IN PERFORMANCES AND EQUITY ISSUES: Multilevel models are the main instrument to deal with complex clustered data (students nested in classes, schools, degree programs, universities) in education and to consider the circumstance that belonging to specific groups (hierarchical or cross-classified) can affect an observed outcome variable. The task of measuring the peer effect on quality and equity of education institutions will be pursued by adopting a multilevel approach with fixed and random effects (random intercepts and slopes at school level or class level) for assessing the value-added of peers attending the same institution to prevent divergences in students’ learning outcomes and in future educational choices and career. Namely, differences in quality (performance) can be captured by the introduction of random intercepts at
school/class/degree program level, whereas differences in equity can be captured by allowing the slopes of variables which are linked to disadvantaged conditions (e.g. the family background) to vary between relevant clusters of students (Sulis et al., 2020).
The flatter the slopes, the lower is the effect of disadvantaged conditions on the outcome (e.g. learning outcome, high school-university transition, graduation). These indicators of quality and equity can be related to group compositional factors to assess which factors (e.g. peers’ characteristics, classroom resources, management, resources, teacher collaboration, school climate) have a relevant effect on observed divergences and to test as these effects vary between classes/schools/degree programs/universities and territories. The use of cross-classified multilevel models will be adopted to detect the role played by both the high school and the university in determining the capability of both institutions of origin and of arrival to boost student learning outcomes, apart from factors related to students’ socioeconomic and cultural conditions.
METHODS FOR DISENTANGLING PEER EFFECTS: Instrumental variables and two steps approaches are widely used in the literature to disentangle peer effect and assess the presence of a multiplicative effect of peers’ behaviour (Sacerdote, 2011). A prominent approach that allows to identify the endogenous peers effect is the one proposed by Bajari et al (2010) to estimate the peer effect in stock market analysts’ recommendations with strategic interactions and applied in several domains. This methodology models peer interactions by assuming that peers take their choices simultaneously, where each individual forms expectations on his/her peers’ choices based on their observable characteristics. Operatively, peer effect can be estimated using a two steps procedure. In the first step, peers’ expected choices are estimated through a (sieve nonparametric) regression approach that approximate students’ choice probability with a set of optimal transformation of attribute data on students and peers by means of flexible polynomials. In the second step, the average predicted probability estimated in the first stage for each peer group is included in a discrete choice model (i.e., conditional or multinomial logit) as a measure of students’ beliefs regarding peers’ interaction. The estimated peer effect can be obtained by the coefficient associated with this last variable. Moreover, correlated effects are accounted for by using a full set of fixed effects at clustering level.
METHODS AND MODELS FOR EGO-CENTRED NETWORKS AND SOCIAL INFLUENCE MECHANISMS
Network data on social interactions and support among students and variables measured at individual level will be analysed by the identification and definition of network statistics able to describe the peer effects, the social embeddedness, and the social support.
The extension of multidimensional data techniques for the treatment of relational and attribute data will be considered with a particular attention to the methods related to the multiplex network (Bródka et al, 2018) to determine which kind of relationships is mostly connected to soft and disciplinary skills. In addition, clustering methods will be adopted to identify ego-network typologies (Pelle & Pappadà, 2021) and how they affect individual performance. We will use statistical models that account for the variation in ego-alters relations (multilevel models) or deal with social influence mechanisms affecting individual outcomes, such as ERGMs models (Lusher et al., 2013) as well as Network Autocorrelation Models (Doreian, 1980, Leenders, 2002). In addition, we aim to explore the use of statistical models able to deal with jointly network data, psychometric measures, individual covariates, contextual covariates and different individual outcomes. The different roles (exogenous, mediator, moderator or endogenous) of all these variables should be assessed.
OBJECTIVES AND METHODOLOGIES
For attaining the objective A.1, we intend to use two-step procedures and multilevel regression models. Our focus, namely the response variable, is the educational outcome at the high school. For the A.2 objective, ego-centred social network analysis will be performed.
To achieve the B.1 and B.2 objectives, we intend to use two-step procedures and multilevel models, where the two possible outcomes of the binary response variable are to be enrolled at university or not. We will estimate cross-classified models, to quantify among those who enrol – their propensity to enrol in the different courses, according to their school of origin, the presence of same-sex peers, the gender-balance in a course and the athenaeum in which they enrolled. As for the C.1 and C.2 objectives, we intend to deal with multilevel, discrete-time, competing risk event-history models, where each university student is followed up to: dropout; graduation; course switch; censoring. For attaining the objective C.3, we intend to use ego-centred social network analysis.
INTEREST FOR THE ADVANCEMENT OF KNOWLEDGE
This research project has several implications which directly involve citizens, governing bodies of secondary and tertiary education institutions. The main outcomes of the projects will be used in pursuing the following four delivery outputs. First, to provide a complete geographical picture of inequalities in the secondary and tertiary education systems in Italy; this picture will be needed to address an efficient use of educational resources according to the emerging priorities and to re-define activities for promoting quality and equity and removing initial disadvantages in educational opportunities. Second, to give information on the role played by peers in reducing inequalities and promoting a more egalitarian system. Third, to assess the effectiveness of policies addressed to reduce gender-oriented behaviours in educational choices. Fourth, to assess how policies addressed to the development of soft and social skills can generate positive externalities.
The project will put a special attention to the identification of institutions which potentially have adopted practices and educational policies acting as social equaliser functions in reducing disadvantaged conditions.
The main findings of the projects will provide tools for identifying:
– effect that peer interaction has on divergences in high schools learning outcomes, transition rates from high schools to universities and university performance indicators;
– schools, degree programs and universities capabilities able to reduce inequalities in educational opportunities;
– institutions able to reduce gender-oriented behaviours in students’ educational choices and outcomes;
– institutions which have boosted the development of soft/disciplinary skills addressed to promote a more egalitarian education system.
METHODS OF DISSEMINATION OF THE RESULTS ACHIEVED
The RUs will be engaged in monitoring the effectiveness of the developed activities and in promoting the dissemination of main findings. The research team will share an on-line platform containing information on activities, methodologies, relevant findings, relevant decisions.
The members will be involved in monthly meetings (using online platforms) that will ensure the exchange of information on the state of advancement and the sharing of decisions, strategies of analysis, and tools for the communication of the results.
The dissemination of intermediate and final results will be guaranteed through:
– Participation in national and international conferences;
– Submission of scientific papers at national and international conferences;
– Submission of main results to leading international peer-reviewed journals;
– Involvement in intermediate meetings of scholars and external experts to stimulate the debate and the critical contribution to the research project;
– Organisation of an intermediate and a final workshop for the dissemination of the results;
– Dissemination of main results and published papers using innovative ways of dissemination, as by financing insertions of events and results in social media, video summaries, video abstracts and highlights;
– Organisation of round tables with the involvement of policymakers and members of institutional boards of educational agencies;
– Informative publications devoted to Italian stakeholders (teachers, school directors, university boards);
– Involvement of institutional actors and experts of secondary and tertiary education in order to raise attention.