Does the apple fall far from the tree? The role of parental influence on sociability among the children of immigrants

ABSTRACT

We explore, using a unique Italian survey dataset containing matching parent–child information, the determinants shaping sociability patterns among the children of immigrants. While it is well-known that long-term migration disrupts interpersonal networks, little is known about the consequences of such disruption for the offspring of immigrants. Adopting a multilevel approach that allows for the consideration of different individual and parental characteristics, we investigate such effects. Our findings reveal that the size and diversity of core discussion networks among immigrant parents are strongly associated with those of their children. Contrary to widespread belief, once the actual sociability patterns of parents are taken into account, out-group attitudes, religious affiliation, and religiosity of the parents play no significant role in determining the size and composition of the close relationships of their children. We consequently put forward the hypothesis that parents may lead mostly by example. Consequently, an adequate understanding of the structure of the core discussion networks of the children of immigrants requires paying attention to the actual interpersonal networks that children have been exposed to within their households.

KEYWORDS:

• Core discussion networks
• sociability
• children of immigrants
• interethnic relationships
• Italy

1. Introduction

Relational approaches to international migration have consistently shown the significance of personal networks in shaping, or even making possible, a wide range of migratory experiences: from the migration decision to settlement trajectories, from job searching to entrepreneurship, from processes of legal incorporation to psychological well-being. Social networks, it has been argued, provide immigrants with (material and immaterial) resources that play a crucial role in the settlement process (Tilly and Brown 1967; Portes 1995; Gold 2005; Bashi 2007; Vacca et al. 2018). The acknowledgement of their significance is such that ‘social network’ has been presented as a key concept in the contemporary sociology of migration (FitzGerald 2014; Bilecen and Lubbers 2021). The centrality of social networks in contemporary migration research, however, has been largely focused on their effectiveness as conduits for resources and information. The emphasis on networks as a source of social capital has not been accompanied, moreover, by an effort to understand the migration-related mechanisms that contribute to the formation, maintenance (and dissolution) of interpersonal ties in the first place.

Besides the centrality of social networks, another important finding in contemporary migration research shows that the consequences of international migration – be they penalties or, more rarely, advantages – may and often do persist across generations (Portes and Rumbaut 2001; Heath and Cheung 2007; Heath and Brinbaum 2014). Such an important finding, however, has been studied nearly exclusively in terms of chances for socio-economic mobility. Very little is known about interpersonal networks among the children of immigrants, and about the determinants that shape their sociability patterns. Do the disruptions in interpersonal networks and sociability practices caused by long-term geographical mobility also play a role in shaping the sociability patterns of immigrant descendants? Besides a few local studies, this question has been largely ignored in the scholarly literature.

In the following pages, we provide an inter-generational analysis of some features of interpersonal networks among the foreign population in Italy, exploring the role of parental influence. We focus on a particularly important type of strong ties, so-called core discussion networks (hereafter: core networks), the small set of people with whom individuals feel they can freely discuss important personal matters (Marsden 1987; McPherson et al. 2006). Although core networks have a long-standing tradition of analysis in social network literature, the consequences of long-term mobility on the core networks of individuals are still understudied.

We centre our analyses on two key features of the core networks among the children of immigrants: network size and the degree of ‘national’ diversity among its members. Our analysis contributes to a better understanding of the predictors of both the extension and composition of social networks among the children of immigrants in several ways.

First, we analyse the case of Italy, a Southern European country that, despite a (relatively) short migration history, has many minors with a migration background in its population (Sciortino 2016). As previous research on the determinants of inter-ethnic relationships has been usually carried out in North-Western European countries, our study provides a comparative counterpoint. Some features of the Italian migratory situation are, in fact, rather different from the better known cases of Northern European countries. Most of the foreign population in Italy – more than 5 million individuals (9 percent of the total population) – have arrived in the country in the last three decades. The primary migration systems are not rooted in a colonial heritage or guest-worker recruitment programmes, but rather in individual explorations followed by the development of informal migration chains (Colombo and Sciortino 2004). Until a decade ago, immigration to Italy was mostly composed of labour migrants, while the number of settled refugees and asylum-seekers was – and (comparatively) still is – rather low (Colombo and Dalla-Zuanna 2019).¹ Although the foreign population in Italy is overall gender-balanced, the gender patterns and migration sequences of foreign communities are often very diverse: in some cases, male pioneers are joined later by female followers; in others, particularly among migrants active in household services and care work, female migrants are the pioneers, often subsequently recruiting other women; still in others, families tend to migrate together (Gabrielli et al. 2019). Both family reunification and union formation have been prevalent processes: according to the National Statistical Institute, 64 percent of households (having at least one foreign-born member) have one or more minors in it. The analysis, moreover, provides an original perspective in reference to the existing literature concerning the children of immigrants in Italy. Such literature, in fact, has been overly concentrated on the educational outcomes of immigrant descendants, leaving other socio-cultural integration dimensions in the background (Cvajner 2011; Azzolini and Barone 2013; Di Bartolomeo et al. 2017; Triventi et al. 2022).

A second advantage of our analysis is its reliance on a large, nationally representative household survey. Previous studies on sociability among the children of immigrants and the ethnic composition of their networks have been restricted to within-school ties among adolescents (Smith et al. 2016; Smith 2018). We take into consideration a wider age range of immigrant descendants, as well as a variety of possible members of core networks, including relatives, kin, and non-school friends. A further advantage of our dataset is that it provides matching interviews with all the other members of the household. While most previous studies have been forced to rely, as far as information on parents was concerned, on the accounts of their children, we have been able to study the role of parental observed sociability practices in shaping the sociability of their children, controlling for other parental-level factors.

Although research on the ethnic composition of personal networks has often stressed the importance of third parties in the development of out-group connections by the children of immigrants (van Tubergen and Smith 2018), previous studies have almost always restricted the exploration of parental influences to racial or religious attitudes towards their friends, acquaintances, or spouses (Edmonds and Killen 2009; Carol 2014). To our knowledge, only one previous study has addressed explicitly the role of parental out-group relational practices in shaping the composition of networks among their children. It was, however, limited to within-school friendship ties (Smith et al. 2015). Thanks to the availability of matching interviews, we have been able to document how the actual structure and composition of the parental networks independently relates to both the ethnic composition and the size of core networks among their children, regardless of national and religious specificities.

A final advantage of our data is that they allow the inclusion of ‘intimate’ social isolation in the analysis, i.e. those interviewees that have no trustworthy confidant with whom they could discuss important personal matters. Such a condition, along with feelings of loneliness, is known to be comparatively more widespread among immigrants, even after a long period of residence in the country of settlement (van Tubergen 2014; van Tilburg and Fokkema 2021). Studies of loneliness have often stressed how such a condition is contagious: it occurs in clusters, and it spreads across networks (Cacioppo et al. 2009). As the chances of contagion could be quite significant across members of the same household, it is important to explore if such chances have implications for the inter-generational transmission of sociability patterns.

The article is structured as follows. Section 2 reviews the existing literature on immigrant interpersonal relationships, focusing on the nexus between migration and network size, the factors affecting network diversity, and the role of parents in shaping inter-ethnic relationships among children. Section 3 presents the study’s data and methods. Section 4 elaborates the results of the analysis. In section 5, we review and discuss the actual and potential limitations of our study, thus paving the way for the conclusions drawn in section 6.

2. Background

2.1. Core discussion network size across immigrant generations

The size of personal networks may have important consequences for the lives of individuals. People with more social ties tend to receive more and better social support and large interpersonal networks reduce the subjective feeling of loneliness (Wellman and Gulia 1999; Green et al. 2001). Several studies have shown that a decline in social connectedness figures among the consequences of geographical mobility (Perez et al. 2018). Such disruption has multiple causes. A decrease in the frequency of physical interaction and the contraction of shared experiences make it difficult to maintain relationships with kin and friends at the pre-existing level of intensity (Morosanu 2013). The more migrants settle in a new destination, the more relationships in countries of origin tend to fade away (Lubbers et al. 2010; Nisic and Peterman 2013). This loss is rarely balanced by the development of strong ties in the new location, as they are hindered by strong barriers such as poor knowledge of the local language, scarcity of economic resources, and discrimination (Priest et al. 2014; Wu and Penning 2015, Hjalmarsson and Mood 2015).

In the migration context, the weakness of strong ties must not be confused with the lack of social contacts. In fact, many forms of migration rely on (and generate) a large wave of interpersonal contacts, who represent a source of help and support during settlement. Much local research, however, stresses how these contacts often have a limited liability character. They are characterised by low levels of trust, and often fail to develop into long-lasting friendships (Ryan 2011; McGhee et al. 2015). In fact, immigrants usually experience higher levels of loneliness than natives (Koelet and de Valk 2016; van den Broek and Grundy 2017). Long-term mobility is consequently often associated with the contraction of strong ties.

This consideration appears to be confirmed by empirical studies both in Europe and the United States. The sizes of core networks among immigrants are generally small in comparison to those of the native population (Völker et al. 2008; van Tubergen 2014; Bilecen and Vacca 2021). An extended period of residence, moreover, does not always imply a substantive increase in the number of strong ties upon which the individual can rely. Studies on elderly immigrants, for example, have documented the enduring difficulties to reconstitute, even after many years, a sizable, satisfying, and mixed network of kin and friendship ties in later life (ten Kate et al. 2020; de Jong Gierveld et al. 2015). Migration, particularly when associated with low socio-economic status and poor health conditions, is consequently a strongly isolating factor (van Tilburg and Fokkema 2021).

Is the ‘isolating’ effect of geographical mobility transmitted, wholly or partially, to the offspring of immigrants? Unfortunately, research on the size of core networks among the children of immigrants is scarce. However, evidence suggests that immigrant adolescents that have spent few years in the place of settlements and have been subjected to discrimination have a higher risk of loneliness, compared both to natives and to the children of immigrants that have grown up in the settlement country (Neto 2002; Reach Madsen et al. 2016). Our study further extends the research on the drivers of network size among immigrant youth, investigating the role of parental network size, and thus exploring inter-generational determinants.

2.2. Determinants of ethnic diversity in immigrant relationships

Most research carried out in Western Europe on immigrant interpersonal networks is overly, if not exclusively, interested in their ‘ethnic’ (i.e. national or religious) composition. To have out-group ties is considered beneficial for the immigrant since natives provide migrants with valuable information linked to better economic opportunities (Lancee 2012; Kanas et al. 2012). It is often assumed that having ties with natives fosters a stronger identification with local culture (Ali and Fokkema 2015). In the same vein, it is claimed that natives who develop interethnic contacts are likely to reduce their prejudice (Pettigrew et al. 2011), although friendship-making processes are in turn forged by culturally driven selectivity (Manevska et al. 2018).

In recent years, research on inter-ethnic sociability has grown remarkably. The traditional interest in intermarriage (Kalmijn 1998) has been recently matched by the exploration of the role played by diversity in shaping various forms of interpersonal networks. Besides core discussion networks (van Tubergen 2015), researchers have investigated: networks of acquaintances and neighbours (Martinovic et al. 2009; Schaeffer 2013); relationships among schoolmates (Rivellini et al. 2011; Smith et al. 2016; Smith 2018); and the processes of dating and mating (Clark-Ibañez and Felmlee 2004; Cvajner 2011; Carol 2016). A shared finding seems to indicate that the proportion of inter-ethnic ties drops with the intensity of the relationship. Unsurprisingly, core networks appear to be the most ethnically segregated (van Tubergen and Smith 2018).

The strength of ingroup relationships is predicated upon the working of three mechanisms: homophily, the tendency for individuals to prefer relationships with individuals that are similar to them with regard to some characteristics (McPhearson et al. 2001); opportunities, the contextual constraints that shape the likelihood for each individual to meet diverse alters (Kalmijn 1998); and the intervention of third parties, when other social figures (e.g. parents, peers, kin) are able to influence individuals in their relational choices (Smith et al. 2015).

The Western European studies carried out so far have elaborated multiple factors affecting the composition of interpersonal networks, contingent upon two or more of the above mechanisms. The time of arrival plays a strong role in shaping the ethnic composition of interpersonal networks among both immigrants and their offspring. Those born in the country of settlement experience higher opportunities and preferences to befriend ethnically diverse peers (van Tubergen 2015). Similarly, Italian studies on lower secondary education pupils have shown how the children of immigrants born in Italy (or have arrived at a very young age) are much more likely to spend some of their free time with Italian friends, to have a majority of Italian peers in their core network, and to be nominated as a friend by native classmates (Dalla Zuanna et al. 2009; Rivellini et al. 2011; Gabrielli et al. 2013).

Some studies have also drawn attention to the fact that sociability patterns originating in homophilic choices can actually be contingent upon the opportunities to befriend various types of individuals. The composition of core networks is strictly contingent upon the number of co-ethnics in the context observed, e.g. the school class, the neighbourhood, the region (Kalmijn 1998; Fischer 2008; Schaeffer 2013; van Tubergen 2015; Smith 2018).

Some studies have highlighted how the composition of core networks of immigrants may be, directly or indirectly, the outcome of preferences and everyday choices by the immigrants themselves. Since ‘temporary’ immigrants tend to be less interested in investing in new, and more difficult, diverse relationships, increased network diversity is linked to the growing intention to settle (Martinovic et al. 2015).

Other studies have focused on norms, values, attitudes, and traditions as possible explanations for the homogeneity of core networks. It has been argued that immigrants who maintain ingroup traditions – and hostility towards other groups – are likely to have low levels of network diversity. Such a hypothesis has been tested, demonstrating how the diversity among interpersonal networks of immigrants varies according to national origin (Martinovic et al. 2015), the level of identification with their own ethnic group (van Tubergen 2015), attitudes toward out-group friendships (Smith 2018), and the degree of inter-ethnic affection (Martinovic 2013). Furthermore, indicators of religiosity are generally associated with a lower frequency of core out-group relationships (van Tubergen 2015; Carol 2016).

The most controversial aspect for the studies carried out in Europe has been the role of religious affiliation, particularly regarding Islam (Drouhot and Nee 2019). It has been observed that Muslim immigrants are (comparatively) more successful in maintaining ingroup conservative traditions, strong religious identities, and a high level of religiosity across generations (de Hoon and van Tubergen 2014). Research has shown that Muslim immigrants have very low rates of intermarriage (Carol 2016) and higher levels of religious homophily in friendship networks (Leszczensky and Pink 2016). For some scholars, this degree of cultural and social closure figures among the reasons for which wide sectors of the native population react to their presence with high levels of suspicion.

2.3. Parental influence on inter-ethnic relationships among children of immigrants

Notwithstanding these findings, sociability among the children of immigrants is still an underexplored research topic. Systematic research is rare, and it covers a very small number of European countries. The available information, moreover, is largely concentrated on primary-school age and adolescent immigrant youth, relying in large part on schoolmate data. Unsurprisingly, such data constrain the possibility of studying the role of third parties – and particularly parents – in influencing the extension and composition of core networks among the children of immigrants. These studies rely on the accounts of what the children of immigrants think the attitudes and preferences of their parents are. As any parent may testify, such accounts should be considered with care.

The literature has generated several hypotheses on how immigrant parents may influence the sociability of their children. They may affect it indirectly, shaping their opportunities to meet co-ethnic peers. As schools and neighbourhoods have various levels of ethnic segregation, the choices and constraints of parents may influence to a significant degree the chances for their children to meet and befriend peers of different backgrounds. As immigrant parents of higher socio-economic status are more likely to live in mixed neighbourhoods and send their children to schools in which native pupils comprise the majority, they improve the chances for their children to develop out-group ties (Smith 2018). Adolescents with highly educated immigrant parents, moreover, are more likely to enroll in academic tracks, in which natives are overrepresented (Smith et al. 2015).

Parents may also influence the features of their children’s core networks through the transmission of their own attitudes towards their out-group relationships (van Tubergen and Smith 2018). The children of immigrants may take into consideration the positive (or negative) parental preferences for their romantic relationships (Edmonds and Kellen 2009; Carol 2014), as well as for their friendships (Smith et al. 2015). Within some groups, parents are expected to monitor and actively pressure their children to maintain their social and dating life within the boundaries of the group.

Do parental influences really operate only – or mostly – through overt normative pressures and conscious elicitation of parental preferences? In fact, many socialisation processes operate outside the realm of conscious instruction. Children develop their normative frames through repeated interaction, physical copresence with adult figures, and co-participation in concrete, emotionally charged, practices and ‘rituals’ enacted within shared environments (Guhin et al. 2021). It is consequently likely that the sociability patterns and the out-group attitudes among the children of immigrants, especially when they still live in the parental home, are shaped by the patterned practices of their parents. Previous research has already stressed how the inclusion of natives in the social networks of immigrant parents has some consequences for their children’s participation in mixed groups of schoolmates (Smith et al. 2015) and the development of inter-ethnic romantic relationships (Clark-Ibàñez and Felmlee 2004). In the following section, we formulate some hypotheses on the importance of parental practices and behaviours for the study of core discussion networks among the children of immigrants.

2.4. Research hypotheses

Our hypotheses are grounded in the assumption that, as children are constantly exposed to the social practices of household members, the observation of, and participation in, such practices may influence their relational opportunities and choices. In particular, we assume that the sociability patterns of immigrant parents in the settlement context – e.g. the size and ethnic composition of their core network in Italy – may be strongly connected with both the core network size and the chances to develop out-group relationships among their children. We also propose that the impact of the sociability patterns among parents cannot be reduced to other factors known to be associated with the opportunities and preferences to meet same-ethnic peers and to develop larger networks among the children of immigrants. That is, its impacts are analytically independent from other known factors such as parental national origin, years spent in Italy, religious identity and religiosity, and socio-economic status, as well as the individual characteristics of children, such as language proficiency, employment status, and migratory generation status. Therefore, we formulate two main hypotheses:

H1: Controlling for other parental and individual characteristics, the larger the parental core discussion network, the larger the core discussion network among the children of immigrants in Italy.

H2: Controlling for other parental and individual characteristics, the higher the presence of out-group confidants in the parental core network, the higher the chances for immigrant children to have out-group relationships in theirs.

3. Data & methods

To test our hypotheses, we use data provided by the Social Condition and Integration of Foreign Citizens (SCIF), a large representative household survey of the foreign population in Italy collected in 2011–12 by the Italian National Institute of Statistics (ISTAT).² The survey is based on a random sample of households that include at least one foreign national component. Within each household, all members have been separately interviewed (N = 25,326).

The SCIF survey has a module on the core discussion networks of the interviewees. Each of them has been asked to list up to three people in Italy with whom they ‘discuss relevant matters’. For each of the listed individuals, ISTAT has collected information on the type of relationship (kin, wife/husband/partner, friend, neighbour, etc.) and national origin of the confidant (same nationality as respondent, other foreign nationality, Italian).

For our analyses, we rely on a SCIF subsample (N = 1,977),³ including all the children of international immigrants – whether born in Italy or elsewhere – aged 14-25, childless, living in the household with their parents. We excluded, as their size was too low to allow for a separate analysis, the children of mixed-couples, the descendants of immigrants from EU15 and other economically advanced non-EU countries (Canada, the United States, Japan, Australia, New Zealand, and South Africa).

We have constructed several measures of network range and composition to describe the individual and parental network structures. Network size simply refers to the number of confidants listed as members of the core network. For the children of immigrants, we have measured separately the overall, kin, and out-group size,⁴ the latter of which includes both ‘ethnic’ Italians and immigrants of a different nationality. As a measure of the size of the parental network, we have computed the average network size between mother and father.⁵ Furthermore, as measures of network composition – that is, the types of alters in an individual’s network (Marsden 1987) – we have computed for all non-isolated respondents the proportion of kin and out-group relationships respectively out of total connections (that range from 0 to 1). A similar compositional measure has been computed at the parental level, accounting for the observed core network of both parents simultaneously (up to six connections).

Table 1 offers a synopsis of these measures. They highlight how parents and children have different core network sizes: children of immigrants in Italy most commonly list two confidants they fully trust, whereas parents (both mothers and fathers) are more likely to point to only one person with whom they would discuss personal matters. Considering the core network composition, parents are more likely to have a larger proportion of kin-related members in their core network in comparison to their children. Contextually, parents have a lower incidence of out-group ties, as documented by a mean proportion less than half that of their children (0.12 vs. 0.26). Finally, migrant parents are also much more likely to have exclusively in-group members in their networks than children (more than 75% vs. 60%) and much less likely to have a core network made entirely of out-group ties (4% vs. 14%).

Table 1. Measures of network size and composition.

	Value	%	Mean	Standard Deviation	Min	Max	N
network size	0	17.2	2.45	2.99	0	50	1,977
	1	18.9
	2	29.9
	3	16.4
	4+	17.6
kin network size	0	35.0	1.15	1.02	0	3	1,977
	1	25.6
	2	28.4
	3	11.0
out-group network size	0	66.1	0.51	0.81	0	3	1,977
	1	21.1
	2	9.0
	3	3.8
proportion kin	0	21.5	0.66	0.40	0	1	1,596
	0.1-0.33	7.2
	0.34-0.65	7.5
	0.66-0.99	11.8
	1	52.0
proportion out-group	0	59.0	0.26	0.36	0	1	1,596
	0.1-0.33	12.6
	0.34-0.65	7.7
	0.66-0.99	6.5
	1	14.1
father's network size	0	18.0	2.34	3.38	0	50	1,562
	1	33.2
	2	17.6
	3	14.0
	4+	17.2
mother's network size	0	17.0	2.27	2.72	0	20	1,894
	1	30.4
	2	20.9
	3	16.5
	4+	15.1
avg. parental network size	0	14.0	2.35	2.94	0	35	1,977
	0.5-1	27.7
	1.5-2	24.3
	2.5-3	16.0
	3.5+	18.0
parental prop. kin	0	8.1	0.80	0.32	0	1	1,683
	0.1-0.33	6.5
	0.34-0.65	7.9
	0.66-0.99	13.2
	1	64.2
parental prop. out-group	0	76.0	0.12	0.25	0	1	1,683
	0.1-0.33	11.2
	0.34-0.65	5.4
	0.66-0.99	3.4
	1	4.0

Note: weighted cases.

The ‘generation’ status of the children of immigrants is associated with quite different network structures (Table 2). Children of immigrants that have themselves migrated relatively late (13 years or older) are more likely to be socially isolated, have a larger proportion of kin relations in their core network, and list fewer levels of contacts with out-group alters in comparison to those who are born in Italy or have arrived as children. It is worth mentioning that the percentage of non-isolated children without out-group contacts is only 47% among the interviewees that are born in Italy or arrived very young and around 70% among those who have arrived during adolescence. The proportion of children of immigrants with exclusively out-group ties grows from 5% (late arrivals) to 26% (early arrivals).

Table 2. Children of immigrants’ core network size and composition by migratory generation status.

		Arrived 13 + years	Arrived 6–12 years	Arrived 0–5 years or born in Italy
network size	0	19.5	17.0	14.1
	1	18.4	19.0	19.4
	2	30.3	27.9	32.2
	3	13.6	17.6	18.6
	4+	18.2	18.4	15.7
		100	100	100
proportion kin	0	13.3	22.9	30.5
	0.1-0.33	7.9	6.3	7.5
	0.34-0.65	6.6	6.8	9.4
	0.66-0.99	9.9	15.0	10.3
	1	62.3	48.9	42.3
		100	100	100
proportion out-group	0	69.3	58.1	46.6
	0.1-0.33	11.9	14.5	11.2
	0.34-0.65	8.1	7.2	8.0
	0.66-0.99	5.2	6.4	8.3
	1	5.6	13.7	25.9
		100	100	100

Note: weighted cases.

These results – highlighting the large differences between immigrant parents and their children (and between children with different lengths of settlement) have been repeatedly reported in the literature. In the following, we want to explore whether, controlling for other parental and individual factors, the differences in parental sociability are associated with the size and composition of the core networks of their dependents.

Therefore, we have selected as dependent variables two network measures for immigrant children. First, we have defined an ordinal variable of overall individual network size with 5 categories (no ties, 1, 2, 3, 4+). Second, considering the ‘ethnic’ diversity of interpersonal relationships, we have created, only for children of immigrants who have nominated at least one tie, a dummy variable that equals 1 for respondents having one (or more) out-group confidant in their core network, and zero otherwise. This dependent variable, which measures children’s core network diversity, differs from parental network diversity – measured as continuous – which is used as independent variable. The reason for this difference is to simplify interpretation of our estimates through binary response modelling and, at the same time, preserve the heterogeneity of parental network diversity including an indicator of the degree of ethnic composition of parental network, which accounts for both mother and father’s core networks (up to six ties).⁶ As shown in Table 1, around 40 percent of the children of immigrants with at least one contact in our sample has one or more out-group trustworthy partner in Italy (the proportion of the out-group is larger than 0).

As main independent variables, we have chosen two continuous measures of the structure of the parental core network: the average network size, to account for the range of parental relationships, and the proportion of out-group relationships multiplied by 10 (1 unit increase indicates an increase of 10 percentage points in the proportion), to account for diversity in the parental network composition.⁷

Furthermore, other parental-level variables have been included.⁸ We consider parental indicators of education (no education and lower secondary, upper secondary, tertiary); self-reported household poverty (yes/no); area of origin (Latin America, Eastern Europe, Asia, Middle East, North Africa, Other Africa); area of residence in Italy (North-West, North-East, Center, South and Islands); household structure (nuclear, single parent, complex), following the Laslett classification scheme (Laslett and Wall 1972) and separating nuclear household with both parents from single-parent nuclear households; and the years since their migration to the reunification (or birth) of the first child, thus obtaining a measure that is fully independent from the age of the children of immigrants. Because concerns about the alleged lack of out-group contacts of certain immigrant groups in Europe is usually associated with Islamic minorities (Carol 2016), we have also created a variable for Muslim parents, which equals 1 when at least one parent is Muslim. Additionally, we have created one indicator of parental religiosity, measuring the highest prayer frequency observed among both parents (every day, several times a week, several times a year, no answer) and one indicator of parental attitude toward the receiving country, through a dummy that equals 1 when at least one parent declares feeling ‘at home’ in Italy. Further, as individual measures, we include migratory generation (based on the child’s age at arrival: 0–5 years or born in Italy, 6–12 years, 13 + years), employment status, language proficiency (defined using information on four language competencies – reading, writing, speaking, and listening – ordered on a rating scale from 1 to 4 and then added together to create a unique scale), sex, and age (distinguishing among those within compulsory schooling –14-17– and those beyond compulsory schooling –18-25). Table A1 in the appendix shows the univariate distribution of all these variables for our subsample.

Given the hierarchical structure of our data (Figure 1), with connections belonging to children of immigrants (i), for some cases siblings, nested in the same household (j), we adopt a multilevel approach estimating two-level, random intercept models for our categorical responses to account for the inclusion of variables both at the individual and parental (or household) level (Rabe-Hesketh and Skrondal 2012).

Figure 1. Structure of the data.

First, to study parental and individual determinants of the network size among children of immigrants, we estimate a random intercept ordered logit model on the individual network size, which for each individual i and household j can be formalised as follows: (1) $\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{i}\mathrm{t}\left[\mathrm{P}\left(\mathrm{N}\mathrm{e}\mathrm{t}\mathrm{w}\mathrm{o}\mathrm{r}\mathrm{k}\mathrm{\_}\mathrm{s}\mathrm{i}\mathrm{z}{\mathrm{e}}_{\mathrm{j}\mathrm{i}}>\mathrm{k}\right)\right]={\alpha }_{\mathrm{k}}+{\vartheta }_j+{\text{β}X}_{\mathrm{j}\cdot }^1+{\mathit{\gamma }X}_{\mathrm{j}\mathrm{i}}^2\mathrm{k}=1,2,\dots \mathrm{K}-1$(1) where k refers to different levels of the ordinal dependent variable (K = 5) for which different thresholds $\alpha$ are estimated, $X^1$ is a vector of independent variables including parental indicators (network size, proportion of out-group ties, religion, religiosity, attitude toward receiving country, household structure, education, poverty, years since migration, origin, area of residence), $X^2$ is a vector of individual independent variables (sex, age, migratory generation, employment status, language proficiency), and ${\vartheta }_j$ is the random intercept.⁹

Secondly, to investigate the ethnic composition of relationships, we have estimated (only for respondents who nominated at least one tie) a random intercept logistic model on the probability for immigrant children to have at least 1 out-group relationship in their core network, which can be formalised as follows: (2) $\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{i}\mathrm{t}\left[\mathrm{P}\left(\mathrm{O}\mathrm{u}\mathrm{t}\mathrm{g}\mathrm{r}\mathrm{o}\mathrm{u}{\mathrm{p}}_{\mathrm{j}\mathrm{i}}=1\right)\right]=\alpha +{\vartheta }_j+{\text{β}X}_{\mathrm{j}\cdot }^1+{\mathit{\gamma }X}_{\mathrm{j}\mathrm{i}}^2$(2) where $X^1$ and $X^2$ are vectors of independent variables, including all covariates used in the equation (1)(1) $\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{i}\mathrm{t}\left[\mathrm{P}\left(\mathrm{N}\mathrm{e}\mathrm{t}\mathrm{w}\mathrm{o}\mathrm{r}\mathrm{k}\mathrm{\_}\mathrm{s}\mathrm{i}\mathrm{z}{\mathrm{e}}_{\mathrm{j}\mathrm{i}}>\mathrm{k}\right)\right]={\alpha }_{\mathrm{k}}+{\vartheta }_j+{\text{β}X}_{\mathrm{j}\cdot }^1+{\mathit{\gamma }X}_{\mathrm{j}\mathrm{i}}^2\mathrm{k}=1,2,\dots \mathrm{K}-1$(1) , and ${\vartheta }_j$ is the cluster-level random intercept.¹⁰

4. Results

Table 3, showing the coefficient estimates of our multilevel models, presents the main results of our analyses. We find that immigrant parents with a higher network extension have children with a larger core network size (H1) and that parents with a higher proportion of out-group relationships have children with more diverse core networks (H2).

Table 3. Coefficient estimates on the individual network size (model 1) and on the probability to have out-group relationships (model 2).

	(1) Network size			(2) Out-group
parental level
network size	1.10***	(0.07)		0.11*	(0.05)
proportion outgroup	0.01	(0.03)		0.25***	(0.05)
education (no school and lower sec.)
upper secondary	0.41*	(0.18)		−0.03	(0.27)
tertiary	1.03***	(0.29)		0.53	(0.41)
economic wellbeing (non-poverty)
poverty	0.00	(0.18)		−0.80**	(0.28)
years since migration	0.00	(0.02)		0.01	(0.02)
religion (non-Muslim)
Muslim	0.05	(0.24)		−0.04	(0.35)
religiosity: pray freq (every day)
sometimes a week	−0.17	(0.20)		0.01	(0.29)
sometimes a year	−0.29	(0.20)		−0.07	(0.29)
doesn't know	−0.32	(0.38)		0.06	(0.58)
feels at home (yes)
no	−0.07	(0.19)		0.40	(0.28)
origin (Eastern-Europe)
Latin	−0.49	(0.31)		0.01	(0.44)
Asia	0.64*	(0.26)		−1.17**	(0.40)
MENA	0.32	(0.30)		0.35	(0.44)
Other Africa	1.36***	(0.38)		0.24	(0.53)
household structure (nuclear)
nuclear – single parent	0.01	(0.20)		0.07	(0.30)
complex	0.25	(0.22)		−0.47	(0.33)
area of residence (North-West)
North-East	−0.53*	(0.25)		0.31	(0.35)
Centre	0.45	(0.26)		0.44	(0.36)
South and Islands	−0.66**	(0.22)		0.14	(0.31)
individual level
sex (males)
females	0.26*	(0.13)		0.18	(0.20)
age on interview (14-17)
18–25	0.24	(0.16)		0.50*	(0.24)
migratory generation (arrived 13+)
arrived 6–12 years	0.10	(0.17)		0.16	(0.26)
arrived 0–5 years or born in Italy	−0.17	(0.22)		0.72*	(0.33)
employmet status (student)
worker	−0.11	(0.20)		−0.79*	(0.31)
unemployed or inactive	−0.40*	(0.18)		−0.89**	(0.29)
language proficiency	0.37***	(0.10)		0.97***	(0.19)
Constant				−2.00***	(0.54)
Thresholds:
k1	−0.62	(0.36)
k2	1.48***	(0.36)
k3	4.09***	(0.39)
k4	6.15***	(0.43)
level-2 variance	4.80***	(0.66)		6.22***	(1.56)
rho	0.593			0.654
N of observations (individuals)	1,977			1,596
N of groups (households)	1,404			1,177
Observations per group
min	1			1
avg	1.4			1.4
max	6			6

Standard errors in parentheses.

* p < 0.05, ** p < 0.01, *** p < 0.001.

Note: only children of immigrants with at least one contact in model 2.

First, we highlight how some parental characteristics seem to shape the opportunity structures of their children, thus influencing the size of their individual core networks. Children of more educated parents are significantly more likely to have larger networks than children having less educated parents (lower secondary education or less). Second, some individual characteristics of the children of immigrants are positively associated with the size of their networks. Students have larger core networks in comparison to unemployed or inactive children. Although migratory generation does not show a significant impact, proficiency in Italian is significantly associated with the number of connections, controlling for other factors.

Ceteris paribus, however, there seems to be a relationship between the size of the parental network and the size of core networks among their offspring. More specifically, for each one-unit increase in the average network size of parents, we estimate the chances for their offspring to have a larger than given network size to be 3 times higher, significantly different from 0 (exp [1.10] = 3). To better visualise this effect, the model estimates of the average predicted probabilities, at different values of the parental network size, are shown in Figure 1 (left plot). With the average parental network size ranging from 0 to 3, we observe growth in the estimated probability for children to have 4 or more relationships (thus, a large network size) from a value that is closed to 0 to about 0.2.

Besides the size of the networks, the network composition of parents plays a significant independent role in the proportion of out-group ties among the core networks of their children. Our analyses confirm that the economic conditions of parents shape the relational diversity of their children’s networks. For children living in (self-declared) poor families, the relative risk to have out-group ties with respect to having only in-group relationships is lower than for non-poor families. Furthermore, the children of immigrants from Asian countries, although having, on average, larger core networks, are more likely to list only co-nationals as members.

The chances of having out-group ties in their core networks are, even controlling for these factors, higher among the children of immigrants that have migrated very young or are born in Italy, in comparison to those who have arrived in Italy in their early youth (13 years or older). Students have more out-group ties than workers or unemployed/inactive children. As expected, proficiency in Italian is strongly associated with the probability of having out-group ties, as well as with larger core network size. As we control for age at arrival (and education status), the current age of interviewees can be considered as a proxy of the years spent in Italy. Unsurprisingly, children of immigrants aged 18–25 show higher chances to have at least one out-group relationship than those aged 14–17.

Even controlling for all the factors identified in the existing literature, we can positively identify a further independent role played by the sociability pattern of the parents. In fact, the higher the proportion of out-group ties in the core networks of the parents, the higher the relative risk of their children having at least an out-group tie in their core network. We estimate that, controlling for other factors, when the parental proportion changes from 0 to 0.45, the probability for children to have at least one out-group relationship grows significantly, by about 0.15 points (Figure 2, right plot). However, for children of immigrants having both parents without any out-group confidant (the most common case) we estimate a relatively high probability (about 0.37). Finally, even controlling for parental network diversity and other factors, the larger the size of the parental network, the higher the chances for their children to have out-group relationships. Specifically, a one-unit increase in the average size of the parental network implies an 11 percent increase in the relative risk for children to have out-group ties. We may interpret this finding as a clue to the possibility that parents with more extensive and vibrant social lives, even if more limited in the opportunities for out-group contacts, make for ‘sociable’ children. This would endow them with skills and attitudes that make the subsequent activation of the opportunities for out-group connections more widespread in the social environments of the children.

Figure 2. Predicted probabilities (at the average of the covariates) and 95% confidence intervals of having a large network size at different values of the avg. parental network size (left plot) and having at least one out-group relationship (only for those who nominated at least one tie) at different values of parental network proportion of out-group ties (right plot).

5. Limitations

Thanks to the availability of a large representative sample of immigrant households containing matched data on the sociability of parents and children, we have been able to identify some mechanisms shaping the size and composition of the core discussion networks of the children of immigrants. The inter-generational transmission of sociability pattern, in our hypothesis, may contribute to explaining how disruption in the interpersonal networks experienced by migrants can influence the core discussion networks of their children. As far as the children of immigrants are concerned, moreover, such a mechanism, at least in Italy, could play a stronger role than the ‘ideational’ factors – of cultural and religious nature – on which previous studies have mostly focused.

Beside its strengths, our analysis present both actual and potential limitations worth reviewing. To be sure, we have been unable to deal with two important elements of sociability research. We have not been able to rule out the possibility of reverse causation. If children are likely to be influenced by parents, the relational choices and opportunities of parents are also likely to be conditioned by their children’s choices and opportunities. Unfortunately, neither SCIF, nor any other currently available data on Italy or other countries contain reliable longitudinal data. We are also unable to answer the question of how much the mechanisms we identify are migration specific. Are parental interpersonal networks a key feature of the sociability of their children or does such a factor acquire a special significance only, or particularly, in the context of the settlement process? To answer this query, we would need a comparative analysis of sociability patterns in immigrant and native households. To our knowledge, however, there are currently no household surveys, in Italy or elsewhere, with matched information on the sociability of parents and children covering both native and immigrant population (or operating with the necessary over-sampling of the latter). Still, it can be argued that even if such mechanism would work similarly across immigrant and native households, its impact would be higher in the case of children of immigrants, given the disruption of interpersonal networks experienced by their parents.

Other limits may derive from the importance of potential confounding factors we have not considered. First, in our analysis we have concentrated only on parents as potential ‘third parties’. We have not checked for the presence of multiple influences in the household. We have not, as example, explored the role of siblings and how they may affect the opportunities and preferences of other children. Further research is needed to investigate multiple and reciprocal influences. Second, despite our best efforts to include indicators of parental attitudes, we have been unable to explore in full the potential parental interests in keeping the relational life of their children within specific ethnic boundaries. The association between the (alleged) parental attitude to reproduce in-group traditions and the actual relational attitudes and behaviours of children is a complex one, which would require a much richer set of information on specific attitudes, which unfortunately is not currently available, at least for the Italian case.

Furthermore, we do not explore in depth the role of territorial concentration among co-ethnic and co-national immigrant groups in favouring homogeneous networks, an aspect that has been stressed, particularly in reference to Northern European countries, as among the opportunities that shape immigrant network diversity. We have been able to control in our analysis only for macro-regional fixed effects (see Table A1). We also rely, however, on the existing literature concerning the geographical distribution of migrants in Italy, which finds the distribution to be rather diffuse in the Centre-North (and some Southern urban areas), with little concentration in specific districts and municipalities (Mariani et al. 2023).

Another potential limitation of our study lies in the fact that, being based on matched parental and children information, it includes only the children of immigrants who live with their parents. This may, of course, have a role in shaping their response to parental influence. It should be considered, however, that most of the children of immigrants aged 14–25 in Italy live with their parents. According to our sample, 93% of the children of immigrants aged 14–25 with at least one parent who lives in Italy are living in the parental home.¹¹ Children of immigrants in Italy seem to follow closely the predominant living arrangements among native youth, who notoriously leave the parental home much later than their counterparts in North-Western Europe (Gabrielli and Impicciatore 2020). The potential bias in this regard appears consequently negligible.

A final possible weakness may be the presence among the children of immigrants of both adolescents and young adults, who may have very different sociability patterns. As a robustness check, we have run models separately for two different groups: children of immigrants who are still minors and fall within the compulsory education age (14–17) and those who are of age and beyond compulsory schooling (18–25). Results (shown in the Appendix) suggest that the influence of parental network measures is still substantive and significant for both groups.

6. Conclusion

Research on immigrants’ social relationships has documented how protracted geographical mobility implies significant changes in their interpersonal networks. Immigrants, besides being more often lonely, tend to have smaller core networks. This finding is hardly surprising: during migration, many friendship ties accumulated in pre-emigration lives weaken or fade, while several structural factors – lack of free time, poor knowledge of the local language, scarcity of economic resources, closure mechanisms, and discrimination – make the generation of new intimate ties difficult and rare. Many migrants, moreover, have to rely mostly on weak ties. The literature assumes, however, that the consequences of such disruption are contained within the first generation, leaving the parental transmission of sociability characteristics largely underexplored. Our analyses suggest that the consequences of the disruption play a role also for sociability among the children of immigrants, the interpersonal networks of parents shaped by migration being associated with the sociability patterns of their children, at least when they still live in the parental household.

The size and composition of core networks among the children of immigrants are often shaped by their migration history. We show how older age at arrival may have disruptive relational effects that hamper the core network diversity of young immigrants. Second, lower Italian language proficiency, an indicator of integration in the settlement context, restricts both core network size and proportion of out-group relationships. Often, differences by parental area of origin are also important. As smaller and less complex discussion networks may be less productive of social capital, these differences could easily translate in further inequalities, particularly in the educational system (Cvajner 2011).

Nonetheless, a very important, and understudied, factor is the intergenerational transfer of sociability practices from immigrant parents to their children. Our analyses have shown how the actual size and composition of the parents’ core networks are strongly associated with the size and composition of the core networks of their children.

The results of our analysis consequently point to the importance of considering the sociability of third parties, especially parents, in shaping sociability among the children of immigrants. While previous research has always stressed the importance of third parties on the ethnic composition of friendship networks, the actual mechanisms have been rarely investigated. In fact, much of the literature has assumed that the actions of third parties are mostly shaped by cultural and religious rules and operate mostly through normative pressures and situational constraint. Our analysis tells a different story.

The multilevel analysis has shown that the actual relational practices of immigrant parents are strongly connected to the relational patterns of their offspring irrespective of any other parental and individual considered factors. We have found such an association to be independent from parental education and (self-defined) family poverty, two important factors shaping the opportunity for children to meet out-group peers. We do not find, moreover, any difference based on the religious affiliation of the parents, a factor that prior European studies have identified as a main relational divide. At least in Italy, having Islamic parents does not imply differences in the structures of interpersonal networks among children. We have also investigated the possible role of parental religiosity and attitudes of the parents towards the receiving country. Even in such cases, we do not find any noticeable difference in the basic structures of core networks among the children of immigrants.

We consequently hypothesise that a nuanced understanding of the structure of core discussion networks among the children of immigrants requires moving beyond the nearly exclusive emphasis on parental attitudes and values, taking also on account the actual experience of interpersonal networks that children have been exposed to within their households.

Acknowledgements

The paper has been written within the framework of the PRIN2017 project Immigration, Integration, Settlement. Italian Style. We wish to thank for their comments and criticisms of earlier drafts the following colleagues: Giancarlo Blangiardo, Corrado Bonifazi, Martina Cvajner, Bernadette Nadya Jaworski, Debora Mantovani, Salvatore Strozza and Agnese Vitali. The remaining weaknesses are of our own production.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Ministero dell'Università e della Ricerca (MUR): [Grant Number PRIN 2017N9LCSC].

Notes

1 The demand for foreign labor in Italy is mostly concentrated in low-skilled agriculture, small and medium manufacturing enterprises, or urban and household services. They are all sectors that offer a limited chance of occupational mobility (Fellini and Guetto 2019; Panichella et al. 2021). Overall, the foreign born compose around one third of the Italian working class.

2 See https://www.istat.it/en/archivio/191097.

3 Sample size here refers to the entire subsample of individuals (children of immigrants), whereas the number of individuals with at least one tie is 1,596. Furthermore, in our subsample, the total number of ties is 3,419 and the total number of households is 1,404.

4 Respondents were asked to indicate their overall core network size in Italy through a direct question. Kin and out-group network size in Italy refer only to the first three relationships mentioned, thus ranging from 0 to 3.

5 There are no same-sex couples with children in the sample.

6 Due to the limited number of collected ties per respondent (maximum three), using a continuous measure of ethnic diversity for the children of immigrants would not be recommended.

7 For the multivariate analysis, we consider isolated parents to have a proportion of out-group relationships that equals 0.

8 For parental-level indicators, we have information on both parents, except for single-parent households. Education refers to the highest educational level of parents; origin refers to the mother’s place of birth; years since migration refers to the parent who first accessed Italy; household-level poverty has been collected from the survey’s main adult respondent.

9 Note that we have included in the model both measures of parental network size and composition simultaneously. In this way, we can account for the influence of the parental average proportion of out-group ties, regardless of the parental network size (and vice versa).

10 Models have been estimated using STATA, through commands meologit and melogit respectively.

11 The ISTAT sampling design excludes the children of immigrants 18–25 living alone with Italian citizenship. Our estimate could then well be slightly upwards biased.

Appendix

Table A1. Univariate distribution of parental and individual variables.

	N	%	Mean	Std. Dev.	Min	Max
parental-level:
education
no school & lower sec.	835	42.2
upper secondary	952	48.1
tertiary	191	9.6
yars since migration	1,977		5.59	5.14	−7	47
poverty
no	1,453	73.5
yes	524	26.5
religion
non-Muslim	1,330	67.3
Muslim	647	32.7
religiosity (prey freq.)
every day	952	48.1
sometimes a week	453	22.9
sometimes a year	459	23.2
no answer	114	5.8
origin
Latin	202	10.2
Eastern-Europe	961	48.6
Asia	325	16.5
MENA	380	19.2
Other Africa	108	5.5
area of residence
North-West	727	36.8
North-East	537	27.2
Centre	472	23.9
South and Islands	241	12.2
household structure
nuclear	1,226	62.0
nuclear – single parent	428	21.7
complex	322	16.3
individual-level:
sex
males	1,159	58.6
females	818	41.4
age	1,977		18.71	3.15	14	25
age (categorigal)
14–17	797	40.3
18–25	1,180	59.7
migratory generation
arrived 13+	740	37.5
arrived 6–12	711	36.0
arrived 0–5 & born in Italy	525	26.6
employment status
student	1,106	55.9
employed	384	19.4
unemployed or inactive	487	24.6
language proficiency	1,977			0.83	−2.92	0.89

Note: weighted cases.

Table A2. Coefficient estimates on the individual network size for the entire subsample (ordered logit regression models). Separate models by age group.

	(1) Compulsory schooling age (14-17)		(2) Non-compulsory schooling age (18-25)
parental level
network size	0.91***	(0.10)	1.22***	(0.10)
proportion outgroup	0.00	(0.04)	0.04	(0.05)
education (no school and lower sec.)
upper secondary	0.44	(0.24)	0.50	(0.25)
tertiary	0.68	(0.40)	1.28**	(0.40)
economic wellbeing (non-poverty)
poverty	0.31	(0.24)	−0.20	(0.25)
years since migraiton	0.01	(0.02)	−0.02	(0.02)
religion (non-Muslim)
Muslim	0.18	(0.30)	−0.06	(0.33)
religiosity: pray freq (every day)
sometimes a week	−0.14	(0.27)	−0.24	(0.28)
sometimes a year	−0.45	(0.27)	−0.22	(0.29)
doesn't know	−0.09	(0.51)	−0.64	(0.53)
feels at home (yes)
no	−0.33	(0.25)	0.14	(0.26)
origin (Eastern-Europe)	−0.47	(0.44)	−0.49	(0.42)
Latin
Asia	0.54	(0.33)	0.92*	(0.36)
MENA	0.06	(0.38)	0.41	(0.43)
Other Africa	1.16*	(0.49)	1.66**	(0.56)
household structure (nuclear)
nuclear – single parent	−0.10	(0.30)	0.02	(0.27)
complex	0.32	(0.32)	0.22	(0.30)
area of residence (North-West)
North-East	0.02	(0.31)	−1.00**	(0.35)
Centre	0.86*	(0.34)	0.09	(0.36)
South and Islands	−0.42	(0.28)	−0.93**	(0.31)
individual level
sex (males)
females	0.50**	(0.19)	0.21	(0.19)
migratory generation (arrived 13+)
arrived 6–12 years	−0.08	(0.32)	−0.02	(0.23)
arrived 0–5 years or born in Italy	−0.39	(0.34)	−0.08	(0.32)
employmet status (student)
worker	1.28	(0.97)	−0.18	(0.24)
unemployed or inactive	−0.27	(0.29)	−0.51*	(0.24)
language proficiency	0.38*	(0.15)	0.48***	(0.14)
Thresholds:
k1	−0.41	(0.51)	−1.24*	(0.49)
k2	1.52**	(0.52)	1.00*	(0.48)
k3	3.76***	(0.63)	3.95***	(0.53)
k4	5.76***	(0.76)	6.14***	(0.61)
level-2 variance	3.14*	(1.30)	6.35***	(1.25)
rho	0.488		0.659
N of observations (individuals)	805		1,172
N of groups (households)	711		922
Observations per group
min	1		1
avg	1.1		1.3
max	3		4

Standard errors in parentheses

* p < 0.05, ** p < 0.01, *** p < 0.001

Table A3. Coefficient estimates on the individual probability to have out-group relationships only for those who nominated at least one tie (logistic regression models). Separate models by age group.

	(1) Compulsory schooling age (14-17)		(2) Non-compulsory schooling age (18-25)
parental level
network size	0.33**	(0.12)	0.01	(0.06)
proportion outgroup	0.21*	(0.08)	0.29***	(0.07)
education (no school and lower sec.)
upper secondary	−0.32	(0.40)	0.19	(0.32)
tertiary	0.65	(0.66)	0.34	(0.48)
economic wellbeing (non-poverty)
poverty	−0.77	(0.45)	−0.93**	(0.35)
years since migration	0.03	(0.04)	−0.03	(0.03)
religion (non-Muslim)
Muslim	−0.47	(0.53)	0.36	(0.42)
religiosity: pray freq (every day)
sometimes a week	−0.07	(0.44)	0.02	(0.35)
sometimes a year	−0.63	(0.47)	0.12	(0.36)
doesn't know	−0.58	(0.89)	0.26	(0.72)
feels at home (yes)
no	−0.11	(0.42)	0.70*	(0.35)
origin (Eastern-Europe)
Latin	0.10	(0.70)	−0.07	(0.51)
Asia	−1.05	(0.61)	−1.23*	(0.51)
MENA	0.52	(0.65)	0.02	(0.54)
Other Africa	−1.56	(0.86)	1.54*	(0.69)
household structure (nuclear)
nuclear – single parent	0.15	(0.50)	0.06	(0.34)
complex	0.00	(0.51)	−0.55	(0.39)
area of residence (North-West)
North-East	0.77	(0.54)	−0.11	(0.43)
Centre	0.03	(0.52)	0.52	(0.43)
South and Islands	−0.02	(0.46)	0.18	(0.38)
individual level
sex (males)
females	0.32	(0.32)	0.09	(0.25)
migratory generation (arrived 13+)
arrived 6–12 years	0.23	(0.56)	0.12	(0.31)
arrived 0–5 years or born in Italy	0.89	(0.62)	0.70	(0.41)
employmet status (student)
worker	0.57	(1.92)	−0.82*	(0.33)
unemployed or inactive	−0.57	(0.59)	−0.81*	(0.33)
language proficiency	0.95*	(0.37)	1.00***	(0.24)
Constant:	−2.28*	(0.98)	−1.32*	(0.62)
level-2 variance	6.01	(4.11)	5.45**	(2.11)
rho	0.646		0.624
N of observations (individuals)	641		955
N of groups (households)	575		773
Observations per group
min	1		1
avg	1.1		1.2
max	3		4

Standard errors in parentheses.

* p < 0.05, ** p < 0.01, *** p < 0.001.