Leaving the nest in immigrant neighbourhoods: gender and origin differences in France

ABSTRACT

This article investigates patterns of leaving the parental home in immigrant-dense neighbourhoods by gender and immigrant origin. We draw on a unique large sample, individual-level panel, the Permanent Demographic Sample (1990–2013), matched with neighbourhood-level census data, to track three types of transitions out of the parental home: leaving for an unmarried union, marriage, or independent living. The findings show that growing up in an immigrant-dense neighbourhood is associated with a decreased likelihood of leaving the parental home net of individual, family and contextual controls. Yet patterns vary by gender and origin. French majority youth, Southern European origin women and Sub-Saharan African men are more likely to remain in the parental home when they originate in an immigrant-dense neighbourhood. For others, particularly North African women, growing up in an immigrant-dense neighbourhood is linked to more frequent departures from the parental home to enter marriage. Opposite patterns are found for French majority and Asian origin women. Variation in home-leaving by neighbourhood environments is generally more pronounced for women. We discuss these trajectories in light of socioeconomic disadvantage and normative constraints in immigrant areas and residential sorting.

Introduction

Leaving the parental home is a salient moment in the life course that marks autonomy from the family of origin, a degree of financial independence, and a step towards employment, marriage and family formation. Numerous studies have shown that home-leaving trajectories vary substantially by racial/ethnic background in the U.S. (Britton 2013; Lei and South 2016) and in Europe (Kleinepier, de Valk, and van Gaalen 2015; Ferrari and Pailhé 2017; Bolt 2002; Zorlu and Mulder 2011). Theoretical frameworks explain these differences by highlighting the role of diverging cultural norms that shape decisions about home-leaving, in particular for women (Impicciatore 2015; De Valk and Billari 2007), or material constraints that prevent disadvantaged groups from acquiring residential independence (Treas and Batalova 2011).

Yet, whether in Europe or elsewhere, few studies have investigated the role of neighbourhoods in shaping when youth leave the parental home and for what types of living arrangements (Zorlu and Van Gaalen, 2016; Zorlu and Mulder 2011; Ravanera, Rajulton, and Burch 2003). Given the importance of ethnic residential segregation (Tintori, Alessandrini, and Natale 2018), and a wealth of evidence suggesting that neighbourhoods shape individual socioeconomic outcomes (Ellen and Turner 1997; Sampson 2012; Sharkey and Faber 2014) and demographic behaviours such as marriage and fertility (Galster et al. 2007; Hill and Johnson 2004; Rahnu et al. 2020; South and Crowder 2010; Wilson and Kuha 2018), neighbourhoods could be a potent determinant of home-leaving trajectories, particularly for the children of immigrants. Residential environments are key sites where socialisation occurs and where cultural norms are transmitted. Neighbourhoods also provide opportunities for skills acquisition and socioeconomic mobility. Reduced contact with natives and a greater presence of immigrants within local institutions such as schools, religious organisations and peer networks likely exert an influence on home-leaving decisions that underpin ethnic group and gender differences. Further, youth growing up in immigrant-dense neighbourhoods may also face specific structural constraints on job and housing markets that hinder the transition to adulthood.

This article investigates patterns of leaving the parental home in immigrant-dense neighbourhoods, focusing on differences linked to gender and immigrant origin. Children of immigrants represent a sizeable fraction of the French population, about 15% in 2008¹ (Beauchemin, Hamel, and Simon 2018). Recent research documents that immigrants and their children live more often in disadvantaged urban areas and are less likely than natives to move out of these spaces over the life course (McAvay and Safi 2018; McAvay 2018). Still little is known about how patterns of leaving the nest vary by residential environments and differ across gender and origin.

We draw on a unique large-sample, individual-level panel, the Permanent Demographic Sample (1990-2013), matched with neighbourhood-level census data, to track three types of transitions out of the parental home: leaving for an unmarried union, marriage, or independent living. The analysis identifies whether home-leaving trajectories vary by the share of immigrants in the neighbourhood. The findings show that growing up an immigrant-dense neighbourhood is associated with a lower likelihood of leaving the parental home. This negative relationship persists net of individual, family and contextual-level controls and holds across several alternative specifications. Yet, home-leaving patterns in immigrant areas vary by gender and origin. French majority youth, Southern European origin women and Sub-Saharan African men are more likely to remain in the parental home when they originate in an immigrant-dense neighbourhood. For other groups, particularly North African women, growing up in an immigrant-dense neighbourhood is linked to more frequent departures from the parental home to enter marriage. Opposite patterns are found for French majority and Asian origin women. Variation in home-leaving by neighbourhood environments is generally more pronounced for women.

While we frame our analysis within theories of neighbourhood effects, our empirical strategy does not allow us to argue in favour of a purely causal effect of neighbourhoods. Residential self-selection may also account for different home-leaving patterns in immigrant neighbourhoods, particularly if families that hold more traditional norms about marriage and family, or who have fewer resources to trigger their children's residential independence, sort into these neighbourhoods. We highlight both neighbourhood effects and residential sorting mechanisms in the conclusion. Despite this limitation, our findings emphasise the importance of considering residential environments in interaction with gender and origin when investigating trajectories out of the home, factors which prior studies have only explored separately (Zorlu and Van Gaalen, 2016; Zorlu and Mulder 2011; Ravanera, Rajulton, and Burch 2003).

Background

Home-leaving determinants

Past research on home-leaving has emphasised the role of cultural factors, material resources and institutional constraints. Family values, attitudes, and aspirations about family relationships (Aassve, Arpino, and Billari 2013; Billari and Liefbroer 2007), as well as the broader context of childhood socialisation, familialism or religiosity (Liefbroer and Elzinga 2012; Barber 2000; Lehrer 2004; Goldscheider and Goldscheider 1999), all shape the timing and type of trajectory out of the parental home. Given contrasting socialisation processes for boys and girls within families and pervasive differences in age norms for men and women, the timing and patterns of leaving the parental home differ by gender (Bernhardt et al. 2007; Oesterle, Hawkins, and Hill 2011; Winkler-Dworak and Toulemon 2007), with women generally leaving the nest earlier than men, especially for union formation (Chiuri and Del Boca 2010).

Material resources can be a barrier to leaving home if limited career opportunities and financial resources reduce children’s access to independent housing (Berzin and De Marco 2010; Bynner 2005; Mulder and Clark 2000; Sironi and Furstenberg 2012; Iacovou 2010). On the other hand, low-quality housing conditions could be an incentive to move out (Ermisch 1999). Further, institutional factors such as the economic context, credit and housing markets, educational systems, labour market regulations, and public policies all shape the transition to adulthood (Mulder and Clark 2000; Thévenon 2015).

The role of neighbourhoods

Although local residential environments have received little attention in the literature on home-leaving, the neighbourhood might have a direct, independent effect on individuals’ propensity to move out. Two mechanisms, namely the cultural and structural influence of residential environments, may be particularly crucial to home-leaving trajectories.

First, because they are contexts of socialisation (Sharkey and Faber 2014; Wilson and Kuha, 2018), neighbourhoods influence young people’s values, expectations and preferences about the future course of their adult lives and shape personal beliefs about independent living, cohabitation and marriage. Living in immigrant-dense neighbourhoods implies a more intense degree of socialisation within one’s ethnic group. Youth behaviours are also informed by greater exposure to the group within local institutions such as schools, religious organisations and peer networks. The effect of peer networks may be bolstered when school placement overlaps with place of residence. Hence, growing up in an environment with many immigrants may reinforce the influence of group-specific cultural norms regarding the timing of life events, marriage and family formation.

Second, material resources, institutions and amenities within neighbourhoods create structural constraints that shape individual outcomes (Sharkey and Faber 2014). Residents of socioeconomically disadvantaged neighbourhoods have lower access to educational and job opportunities, transportation networks, and public services, which in turn impact educational and employment prospects as well as opportunities to socialise outside the neighbourhood. Reduced socioeconomic opportunities might discourage young people from achieving economic independence and leaving the home. Home-leaving may further be hindered by job and housing market discrimination or redlining against residents of stigmatised neighbourhoods (Bunel, L’Horty, and Petit 2016; L’Horty et al. 2019; Pager and Shepherd 2008; Aalbers 2005). While some studies emphasise the positive socioeconomic returns of living in ethnic enclaves due to better labour market networking and matching (Edin, Fredriksson, and Åslund 2003), the common overlap between ethnoracial segregation and poverty concentration means that ethnic minorities are more likely to be impacted negatively by spatial disadvantage.

Research that looks at local contextual effects during the transition to adulthood is scarce and focuses mainly on neighbourhood outcomes after leaving the parental home, not at departure patterns (Swisher, Kuhl, and Chavez 2013; Lagrange 2016). To the best of our knowledge, only a few studies analyze how local residential environments (the neighbourhood share of immigrants or non-Western immigrants) influence departures from the parental home and the types of destinations chosen (Zorlu and Van Gaalen, 2016; Zorlu and Mulder 2011; Ravanera, Rajulton, and Burch 2003), with contrasting findings. In Canada, no significant relationship between residential context is found (Ravanera, Rajulton, and Burch 2003), while in the Netherlands, higher shares of immigrants impact home-leaving, but in different ways across origin groups (Zorlu and Van Gaalen, 2016; Zorlu and Mulder 2011). Turkish and Moroccan backgrounds leave home at a faster rate in immigrant dense neighbourhoods, while natives are less likely to leave for unions (Zorlu and Mulder 2011).

Immigrant origin disparities in home-leaving

Coming from an immigrant family affects the entry into adult life, and in particular the departure from the parental home. Extant studies generally show a lower propensity to leave home among minority youth compared to Whites or natives (Lei and South 2016; Bolt 2002) as well as striking differences in subsequent living arrangements (Goldscheider and Goldscheider 1999; Zorlu and Mulder 2010; Kleinepier, de Valk, and van Gaalen 2015; Gabrielli and Impicciatore 2020). These disparities reflect a greater sense of familial obligation (Fuligni and Pedersen 2002) and specific expectations about the appropriate timing and sequencing of life course events (De Valk and Liefbroer 2007) among immigrant families. Specific structural factors such as the low socio-economic position of migrant families also impact home-leaving trajectories (De Valk and Billari 2007; Ferrari and Pailhé 2017).

There exists considerable diversity across ethnic groups in home-leaving patterns (Lei and South 2016; Nielsen 2014; Zorlu and Mulder 2011; Gabrielli and Impicciatore 2020). This may be explained because groups come from countries with different value orientations. In France, ethnic groups have different attitudes towards marriage and family (Pailhé 2015; Collet and Santelli 2016) and are more or less culturally distant from the mainstream French model, in which formal marriage has lost ground to cohabitation and independent living due to increased affirmation of individual autonomy and secularisation (Prioux 2009). Second generation North Africans leave the parental home later (Ferrari and Pailhé 2017) and are more likely to leave for direct marriage compared to French natives (Pailhé 2015). Direct endogamous marriage remains the main route of leaving home for Turkish second generations (Milewski and Hamel 2010; Hamel et al. 2018), a group which is considered to have low cultural integration in France given their high degree of language maintenance, strong religiosity and strong attachment to the country of origin. While there are similarities between Southern European and French native marital practices, the Southern European model is more familialist, with stronger family ties and obligations (Reher 1998; Dalla Zuanna 2001). Leaving home occurs at later ages and is more closely linked to partnership formation (Holdsworth 2000; Billari and Liefbroer 2010; Mencarini et al. 2017).

Beyond cultural mechanisms, children of European immigrants tend to have more favourable labour market positions, with higher income and lower unemployment than other immigrant origin groups, which may favour early departures (Dos Santos 2005; Meurs, Pailhé, and Simon 2006). In contrast, lower socioeconomic status as well as labour, housing and credit market discrimination faced by non-Europeans could hinder the transition to independent living (Quillian et al. 2019; L’Horty et al. 2019). Cultural and structural mechanisms are moreover intertwined: When faced with discrimination or disadvantaged situations, descendants of immigrants may actively oppose the host society’s values and norms (Wimmer and Soehl 2014; Drouhot 2021).

Interplay between gender, immigrant origin and neighbourhood environment

Expectations about gender roles in migrant populations can also lead to diverging patterns of departure from the parental home in immigrant neighbourhoods between men and women. In some immigrant communities, traditional gender role expectations may be pervasive (Pessin and Arpino 2018), with strong expectations that women place family needs before their own, while men must be in a position to provide household income (Hamel, Moguérou, and Santelli 2011). In Turkish, North and Southern African patriarchal societies, the social control over women is particularly rigid (Milewski and Hamel 2010; Streiff-Fénart 2006). Gender relations are more equal in South-East Asia societies, where women’s economic contribution is important and women have greater autonomy in choosing a partner. Patriarchal values are less pronounced and gender inequalities are lower in Southern Europe, but gender roles are still shaped in more traditional ways than in France (Anxo et al. 2011). In neighbourhoods with high shares of immigrants, strong group homogeneity may contribute to the reinforcement of these cultural beliefs and gender norms. Social pressure may be more potent for descendants of immigrants to conform to the values of the origin country, especially for women.

Hypotheses

In light of the above, we test the following hypotheses. Due to socialisation effects and structural constraints within neighbourhoods, we predict that:

As the neighborhood immigrant share increases, youth will be less likely to move out of the parental home (H1a).

When they do leave, youth originating in immigrant-dense neighborhoods will be less likely to live independently or cohabit and more likely to enter more traditional unions such as marriage (H1b).

Yet these patterns will vary by origin and gender:

The negative association between living in an immigrant neighborhood and home-leaving will be stronger for second generation immigrants compared to natives, especially of non-European origin (H2). Women living in areas with higher share of immigrants will be more likely than men to marry directly and less likely to cohabit or live on their own, especially among non-European origins (H3).

Data and methods

Data come from the French Permanent Demographic Sample (EDP), an on-going individual-level panel produced by the National Institute of Statistics and Economic Studies (INSEE) since 1968. EDP compiles data over time on individuals from each consecutive census as well as from civil registries on births, marriages and deaths. The sampling design relies on days of birth² to ensure a representative sample of the population. We use the most recent complete waves of EDP (1990, 1999, 2008, 2013³) due to the unavailability of the neighbourhood-level IRIS code prior to 1990.

The data contain a wide range of socio-demographics concerning individuals and their households. Given its large sample size and broad time frame, EDP is a rich source for investigating departures from the parental home across neighbourhoods. A geographical ID code is provided for each respondent at each panel wave, indicating the municipality and neighbourhood (IRIS⁴) of residence. This enables the data to be matched with neighbourhood and municipality-level characteristics constructed from the French census. The data are also one of the few large-scale statistical sources in France in which children of immigrants can be identified by detailed national origin. Several prior studies have used EDP data to investigate residential segregation and residential mobility among immigrant populations (Rathelot and Safi 2014; McAvay 2018).

Measuring home-leaving trajectories

EDP does not report the date or age at which individuals leave home. However, departures from the parental home can be measured by identifying whether individuals change their position within the household and change their place of residence between panel waves. While this measurement is less precise in terms of the exact timing of home-leaving, it may be considered a more objective observation compared to retrospective declarations typically used in survey data.

To measure home-leaving, the analysis is restricted to EDP children⁵ in t who are observed at the next panel wave (t+1). This results in a total sample of 149,736 individual/time observations and 128,454 individuals.⁶ If the EDP child is observed again in t+1 as a child in the same household, the child is considered as still living in the parental home. If, however, the EDP child is observed again in t+1 as an adult in a new household in a new place of residence, the child is considered to have left the home.⁷ Households refer to non-institutional settings and collective housing, including student residences.

The dependent variable is a 4-level categorical measurement of home-leaving trajectories. The variable is coded 1 if no departure has occurred, i.e. the EDP respondent is still a child in t+1 in the same place of residence. The other three categories refer to EDP respondents who have left the parental home. Using information on marital status in t+1, we further distinguish these home leavers by types of trajectory. Respondents may leave the parental home to enter an unmarried cohabitation (2), for marriage (3), or for independent living (4). Independent living refers to adults living alone or with roommates. Adults observed in t+1 as single parents are excluded from the analysis, as they presumably entered some form of cohabitation after leaving the home that is not observed in the data. This exclusion represents less than 1% of the sample of children in t.

The model can be written as follows: ${\mathrm{Y}}_{\mathrm{i}\mathrm{j},\mathrm{t}+1}={\alpha }_{\mathrm{j}}+{\text{β}}_{1\mathrm{j}}\mathrm{X}{1}_{\mathrm{i},\mathrm{t}}+{\text{β}}_{2\mathrm{j}}\mathrm{X}{2}_{\mathrm{i}}+{\text{β}}_{3\mathrm{j}}\mathrm{X}{3}_{\mathrm{i}}+{\text{β}}_{4\mathrm{j}}\mathrm{X}{4}_{\mathrm{i},\mathrm{t}}+{\mathrm{e}}_{\mathrm{i}\mathrm{j}}$

Where Y_i,t+1 is the living arrangement j in t+1, X1 is the neighbourhood immigrant share measured in t, X2 is gender, X3 is immigrant origin, and X4 are covariates measured in t.

This design results in a sample of EDP children that is heterogeneous in terms of age. We therefore only include children aged between 19 and 31 in t+1 and all models control for age and age-squared. Moreover, analyzing transitions between panel waves results in two potential biases.

The first is due to unequal timing between observations. Because of the change in the periodicity of the census collection, some individuals are observed 9 years after t (i.e. between 1990 and 1999), while others may be observed 5 years later (i.e. between 1999 and 2004) or less (i.e. between 2005 and 2009). We address this in a robustness check included in Table A2. The second is due to attrition, or the loss of individuals over time in panel designs. This issue is particularly salient for the immigrant origin population, as loss of respondents over time may be due to the out-migration of immigrant families leaving France (Solignac 2018). If individuals residing in immigrant-dense neighbourhoods are more likely to emigrate, this could affect the estimation of the neighbourhood immigrant share on home-leaving. While we cannot formally identify whether respondents leave the panel due to out-migration or natural attrition, we can assess the degree to which the loss of respondents over time influences our findings. To do so, we replicate the main model including attrition as an outcome of the categorical dependent variable, namely whether a respondent present in t was absent from the panel in t+1. The dependent variable is thus coded as follows: stays in home (1); leaves the parental home to enter an unmarried cohabitation (2); leaves for marriage (3); leaves for independent living (4); leaves the panel (5). This robustness test is included in Table A2 and discussed below.

Independent variables

Neighbourhood immigrant share: Our main independent variable of interest is the share of immigrants in the neighbourhood during childhood (measured in t). This measurement is theoretically relevant as the impact of the local environment experienced in childhood is believed to have effects on outcomes in adulthood (Sharkey and Faber 2014; Wilson and Kuha, 2017). We follow several prior studies that use the share of immigrants at a given spatial scale as an independent variable in individual behaviours such as residential mobility or home-leaving (Crowder, Hall, and Tolnay 2011; Zorlu and Mulder 2010; Crowder and South 2008).

The neighbourhood scale used is the IRIS. Commonly used in French urban research, IRIS are infra-municipality units of between 1,800 and 5,000 inhabitants, which make them somewhat smaller than U.S. census tracts on average. The IRIS share of immigrants is calculated using the French census and then matched to EDP respondents using geographical ID codes. The immigrant share refers to the proportion of the immigrant population out of the entire IRIS population. While we use a continuous measurement in the main models, the variable is cut into quantiles for some descriptive analyses and robustness checks.

Immigrant origin: Like most French statistical sources, EDP does not include a variable identifying French-born descendants of immigrants. However, when EDP individuals are children, it is possible to determine whether their parents are foreign-born. We consider any individual as having an immigrant origin if they have at least one foreign-born parent⁸ and then draw on the country of birth of that parent to construct the following categorical variable⁹: Western Europe, Eastern Europe, Southern Europe, Southeast Asia¹⁰, Turkey, North Africa, Sub-Saharan Africa and French majority. If both parents are foreign-born, the father’s country of birth is assigned as children typically inherit the paternal surname. If this is not reported, the mother’s country of birth is used. By this definition, immigrant origin respondents may be second generations born in France or immigrants born abroad who arrived during childhood. For this reason, all models control for nativity, namely a dummy indicating whether the respondent was born in France. For sake of concision, all immigrant origin groups are referred to as ‘second generation immigrants’ (G2). The French majority are children whose parents are both French-born.

Second generations make up 13% of the sample (Table 1). Southern European and North African G2 represent the largest groups, reflecting the composition of the French migrant population (Beauchemin, Hamel, and Simon 2018). Immigrant origin children live in neighbourhoods with substantially higher immigrant shares (12% on average) than the French majority population (5% on average).

Table 1. Descriptive Statistics on All Covariates.

	French majority sample	Immigrant origin sample
Main covariates
Immigrant share in neighbourhood	0.05	0.12
Western Europe	–	0.06
Eastern Europe	–	0.03
Southern Europe	–	0.37
North Africa	–	0.40
Asia	–	0.04
Turkey	–	0.05
Sub-Saharan Africa	–	0.06
Born in France	0.99	0.81
Female	0.47	0.48
Individual covariates
Age	16.70	16.56
Employment status
Students	0.50	0.49
Unemployed	0.04	0.05
Active	0.14	0.12
Inactive	0.31	0.34
Lived in same neighbourhood in t-1
Yes	0.61	0.57
No	0.39	0.43
Year of observation
1990	0.47	0.53
1999	0.26	0.26
2004–2008	0.27	0.21
Parental covariates
Age	45.26	47.91
Number of children	2.26	3.31
Type of household
Single parent (male)	0.02	0.01
Single parent (female)	0.13	0.08
Couple	0.85	0.90
Occupation
Unemployed/inactive	0.06	0.14
Blue collar	0.19	0.41
White collar	0.24	0.19
Farmers/artisans/small business owners	0.14	0.09
Intermediary professions	0.23	0.11
Managers	0.14	0.07
Education
No education	0.10	0.47
Primary/middle school	0.19	0.16
Professional certificate	0.37	0.19
High school diploma	0.16	0.08
University	0.18	0.10
Housing
Owner-house	0.65	0.38
Owner-apartment	0.05	0.06
Renter-house	0.08	0.06
Renter-apartment	0.05	0.11
Public housing	0.13	0.36
Other	0.03	0.03
Number of rooms
2 or less	0.01	0.03
3	0.09	0.12
4	0.29	0.35
5 or more	0.61	0.50
Contextual covariates
Share married in the neighbourhood	0.44	0.41
Unemployment rate in neighbourhood	0.10	0.13
Share of renters in municipality	0.34	0.46
City size
<100,000 inhabitants	0.68	0.44
>100,000 inhabitants	0.23	0.31
Paris region	0.10	0.25
N	129,706	20,030
Total	87%	13%

Source: Permanent Demographic Sample (1990-2013)

Sample: EDP individuals who are children in t and subsequently observed in t+1 aged between 19 and 31.

Covariates

Individual-level controls: We control for a number of lagged covariates measured in t when the EDP respondent was a child. Individual-level controls include gender, age, age-squared, employment status (studying, unemployed, employed, inactive) and period of observation.

Family-level controls: These include parental education and occupation¹¹, parental age, total number of children in the household, and type of household (couple, female single parent, male single parent). As housing conditions are known to trigger home-leaving, we also include housing type, tenure and the number of rooms.

Contextual-level controls: Including additional neighbourhood and municipality covariates enables us to better specify the net correlation between the neighbourhood immigrant share and the likelihood of leaving the parental home. We use the neighbourhood unemployment rate in t, or the proportion of unemployed persons out of the working population of the IRIS, a commonly used measure of socioeconomic disadvantage in French urban research. We also control for the share of married persons aged 18 or older in the neighbourhood out of the total population in t. This covariate captures local-level preferences for marriage that may shape individual behaviour. Finally, two municipality-level controls in t are included to measure the impacts of urban location and the housing market: the number of inhabitants and the share of renters out of the total population.

Residential location is not randomly determined; individuals sort into neighbourhoods based on their level of economic resources as well preferences for certain residential environments. In order to better disentangle the impact of the neighbourhood from residential sorting, we control for a wide range of these factors – education, occupation, housing tenure, etc. – that have been shown in past studies to influence the residential location of immigrant and native families in France (Rathelot and Safi 2014; McAvay 2018).

Of course, other unobservable characteristics contribute to residential choices, such as preferences. We also control for a variable indicating whether the EDP respondent lived in the same neighbourhood at the previous census date to capture some effects of residential preferences by distinguishing between long-term residents and recent movers. Yet, to the extent that preferences and other unobservables are correlated to home-leaving decisions, the neighbourhood immigrant share is still potentially endogenous to the dependent variable. For this reason, we are cautious in the interpretation of our findings and do not argue that we are estimating a purely causal effect of the neighbourhood. Descriptive statistics for all covariates are included in Table 1.

Estimation strategy

Model 1 is a multinomial logistic regression estimating trajectories out of the parental home including all independent variables. We then aim to identify whether home-leaving patterns vary by neighbourhood immigration in different ways by gender and immigrant origin. To assess this, we run a second model (Model 2) that includes a three-way interaction between gender, immigrant origin and the neighbourhood immigrant share. All models are estimated using robust standard errors clustered at the individual level.¹² To facilitate interpretation, predicted probabilities of the four types of outcomes according to the neighbourhood immigrant share for both men and women, separately for each origin group, are displayed graphically. We use the default setting for the margins command in Stata, which calculates the average marginal effect (AME) of the neighbourhood immigrant share on the outcome when gender and origin are set at specific values. All other independent variables in the model are held constant at their observed values.

Results

Overall, 42% of the sample leave the parental home between panel waves (Table 2). Moving out for unmarried cohabitation is the most frequent destination of home-leavers (18%), followed by independent living (15%) and marriage (9%). Yet, these trends vary by levels of neighbourhood immigration, gender and immigrant origin. The share of children moving out is lower in immigrant-dense neighbourhoods: 65% of children from neighbourhoods with the strongest immigrant concentrations (>95th percentile) remain in the parental home, compared to 53% of children originating in low-share immigrant neighbourhoods. Women are more likely to leave the home than men, and enter marriage or unmarried cohabitation at greater rates. The French majority and European origin children of immigrants have the highest rates of leaving the nest, in contrast with children of non-European origin. Around three-quarters of Sub-Saharan African, North African, and Asian second generations are still living at home in the next panel wave. Turkish second generation immigrants stand out due to high rates of leaving for marriage and the lowest likelihood of any group to enter an unmarried cohabitation or live independently.

Table 2. Departure Pathways by the Neighbourhood Immigrant Share, Immigrant Origin and Gender

	% Stayed in home	% Unmarried Cohabitation	% Marriage	% Independent living
Quantiles of the Neighbourhood Immigrant Share in Childhood
<25th	53	21	10	16
25–50	57	19	8	16
50–75	59	17	8	15
75–95	60	16	10	14
>95	65	13	10	12
French majority	56	19	9	16
Western Europe	57	15	8	19
Eastern Europe	63	14	9	15
Southern Europe	58	18	12	12
Asia	72	11	6	11
Turkey	68	4	22	6
North Africa	74	6	11	9
Sub-Saharan Africa	77	8	4	11
Women	51	21	12	15
Men	63	15	7	15
Total	58	18	9	15

Source: Permanent Demographic Sample (1990-2013).

Sample: EDP individuals who are children in t and subsequently observed in t+1 aged between 19 and 31.

Model 1 tests these main independent variables of net of individual, family and contextual controls. Results for the key independent variables are summarised in Table 3, and full model results are included in Table A1.

Table 3. Departure Pathways by Neighbourhood Immigrant Share, Immigrant Origin and Gender (Coefficients from Model 1)

		Model 1 Base outcome: Stayed in home
		Unmarried cohabitation	Marriage	Independent living
Neighbourhood immigrant share (t)		−1.142***	−0.784***	−0.959***
		(0.159)	(0.192)	(0.157)
Women		0.836***	1.196***	0.396***
		(0.016)	(0.022)	(0.016)
Ref: French majority
Western Europe		−0.122	−0.076	0.016
		(0.097)	(0.130)	(0.085)
Eastern Europe		−0.443***	−0.302†	−0.109
		(0.133)	(0.166)	(0.119)
Southern Europe		−0.283***	−0.101*	−0.148***
		(0.038)	(0.047)	(0.040)
North Africa		−1.455***	−0.253***	−0.387***
		(0.060)	(0.061)	(0.049)
Asia		−0.562***	−0.362†	−0.146
		(0.136)	(0.192)	(0.128)
Turkey		−1.523***	1.197***	−0.463***
		(0.168)	(0.125)	(0.135)
Sub-Saharan Africa		−0.752***	−0.366*	−0.162
		(0.124)	(0.176)	(0.102)
N		146,593	146,593	146,593

Source: Permanent Demographic Sample (1990-2013).

Sample: EDP individuals who are children in t and subsequently observed in t+1 aged between 19 and 31.

Controls: age, age-squared, employment status, nativity, parental age, parental occupation, parental education, number of children in the household, number of rooms, type of household, housing type and tenure, lived in same neighbourhood in t-1, city size, neighbourhood unemployment rate, neighbourhood share of married persons, municipality share of renters, period of observation.

Table shows coefficients. Robust standard errors in parentheses

*** p < 0.001, ** p < 0.01, * p < 0.05, † p < 0.10.

Net of controls, higher shares of immigrants in the neighbourhood is significantly negatively associated with all types of departures. The negative coefficient is particularly strong when it comes to the likelihood of moving in with an unmarried partner. We ran a number of additional specifications to further test the robustness of the negative correlation between residing in immigrant neighbourhoods and home-leaving. Results are summarised in Table A2. First, we account for the possibility that out-migration of immigrant-origin youth from the panel could bias the findings. We therefore predict home-leaving and attrition in a single multinomial logit model. Accounting for attrition as an outcome slightly increases the magnitude of the coefficients for unmarried cohabitation and marriage, while remaining statistically significant, suggesting that out-migration or attrition lead us to underestimate the findings. Second, we also tested two alternative measurements of the neighbourhood immigrant share: (a) controlling for the neighbourhood immigrant share coded in deciles to test for non-linear effects and (b) controlling for the share of co-ethnics¹³ in the neighbourhood in lieu of the share of immigrants. Finally, we ensured that the relationship between neighbourhood immigration and leaving the parental home is not driven by differential timing between panel waves or certain sub-populations. The negative association between the neighbourhood immigrant share and leaving the home for unmarried cohabitation is robust across all specifications. Only home-leaving by the share of co-ethnics follows a slightly different pattern: it is positively (but insignificantly) correlated with leaving for marriage.

Table 3 also documents gender and immigrant origin disparities in home-leaving net of other factors. Gender differences persist, all things being equal, with women more likely to leave home than men. Significant differences are also found between most second generation immigrants, particularly when it comes to non-European origins leaving for unmarried cohabitation.

Full model results in Table A1 further show how individual and family factors relate to home-leaving. Youth who are unemployed are less likely than students to move out over the period, while having a job increases the odds of departure. Parental socioeconomic status also contributes to residential independence: youth with highly educated parents or who belong to the managerial class have greater chances of leaving to live independently.

We now focus on whether patterns of home-leaving in immigrant-dense neighbourhoods vary by gender and immigrant origin. Figures 1–5 illustrate predicted probabilities of the four types of outcomes according to the neighbourhood immigrant share for both men and women, separately for each origin group. Patterns differ substantially across origins. For some, namely Turkish, Eastern European, and Western European second generations, the neighbourhood immigrant share has no significant association with home-leaving patterns (figures not shown). For others, namely the French majority, Southern European second generation women, and Sub-Saharan African men, the probability of remaining in the home tends to increase with the share of immigrants in the neighbourhood, while the likelihood of leaving for unmarried cohabitation and independent living declines.

Figure 1. Trajectories out of the Parental Home by the Neighbourhood Immigrant Share and Gender among the French Majority Youth.

Source: Permanent Demographic Sample (1990-2013). Sample: EDP individuals who are children in t and subsequently observed in t+1 aged between 19 and 31. Predicted probabilities are estimated controlling for all the covariates included in Model 1.

Figure 2. Trajectories out of the Parental Home by the Neighbourhood Immigrant Share and Gender among Southern European Origin Youth.

Source: Permanent Demographic Sample (1990-2013). Sample: EDP individuals who are children in t and subsequently observed in t+1 aged between 19 and 31. Predicted probabilities are estimated controlling for all the covariates included in Model 1.

Figure 3. Trajectories out of the Parental Home by the Neighbourhood Immigrant Share and Gender among North African Origin Youth.

Source: Permanent Demographic Sample (1990-2013). Sample: EDP individuals who are children in t and subsequently observed in t+1 aged between 19 and 31. Predicted probabilities are estimated controlling for all the covariates included in Model 1.

Figure 4. Trajectories out of the Parental Home by the Neighbourhood Immigrant Share and Gender among Sub-Saharan African Origin Youth.

Source: Permanent Demographic Sample (1990-2013). Sample: EDP individuals who are children in t and subsequently observed in t+1 aged between 19 and 31. Predicted probabilities are estimated controlling for all the covariates included in Model 1.

Figure 5. Trajectories out of the Parental Home by the Neighbourhood Immigrant Share and Gender among Asian Origin Youth.

Source: Permanent Demographic Sample (1990-2013). Sample: EDP individuals who are children in t and subsequently observed in t+1 aged between 19 and 31. Predicted probabilities are estimated controlling for all the covariates included in Model 1.

Further, growing up in immigrant-dense neighbourhoods is associated with leaving for marriage in different ways across groups. North African origin women are significantly more likely to leave the home for marriage as the share of immigrants in the neighbourhood increases. This is also true, to a lesser extent, for Southern Europeans. Among Sub-Saharan African origin youth, there are notable gender differences: women who grow up in immigrant neighbourhoods are more likely to marry, while men are less likely to, yet these trends fall short of statistical significance. As for the French majority, women are less likely to leave home for marriage as the immigrant share in their neighbourhood increases, while for men, the association is null. This is also the trend for Asian second generations.

Finally, it is notable that neighbourhood immigration is generally more decisive to home-leaving for women than for men. For the French majority and Southern European second generations, as the neighbourhood immigrant share increases, women tend to be more likely to stay in the parental home. While this is also true for men, the pattern is stronger for women, to the extent that gender differences in leaving the home diminish in high immigrant share neighbourhoods among these groups. Further, as neighbourhood immigration rises, the probability of leaving the home for marriage increases at a greater rate for North African origin women compared to men of the same origin. Finally, the negative association between neighbourhood immigrant shares and marriage among French majority and Asian origin women illustrates the stronger role of context in shaping women’s departure patterns.

Discussion and conclusion

This article explored the link between local residential environments and home-leaving. Using a rich, large-sample individual-level panel from France, matched with neighbourhood-level census data, we explored how pathways out of the parental home (leaving for unmarried cohabitation, marriage and independent living) varied by levels of neighbourhood immigration, immigrant origin and gender.

The results showed that children growing up in immigrant-dense neighbourhoods were less likely to leave the home, confirming H1a. Types of trajectories out of the home also varied across residential environments. While we did not find evidence that, overall, originating in areas with high shares of immigrants is linked to entering more traditional unions such as marriage, youth from these areas were still less likely to move out for unmarried cohabitation – a less traditional union that is currently the most pervasive living arrangement in France (Prioux 2009). Controlling for the share of co-ethnics further points in this direction, having observed a strong negative relationship between co-ethnic-dense areas and leaving for unmarried cohabitation in particular. Together, these findings provide partial evidence in favour of H1b.

Growing up in immigrant-dense neighbourhoods could directly shape home-leaving through structural and cultural mechanisms. Neighbourhood effects research from multiple contexts shows lower access to higher education and job opportunities for youth from deprived neighbourhoods, which could impede residential independence (Levy, Owens, and Sampson 2019; Nieuwenhuis, Kleinepier, and van Ham 2021; Aeberhardt, Rathelot, and Safi 2015). Youth potentially face greater levels of job, housing or credit discrimination based on their place of residence. The strong correlation between immigrant-dense neighbourhoods and spatial disadvantage in France implies that youth from these areas likely face such structural barriers. However, it is noteworthy that the negative coefficient of neighbourhood immigration persists even after controlling for neighbourhood unemployment, which was also negatively correlated with some trajectories out of the parental home. Home-leaving patterns are thus not shaped solely by the socioeconomic features of the environment but also specifically by its ethnoracial composition, suggesting that specific cultural processes related to in-group norms are also at play. Greater interaction with immigrants and their descendants in local institutions, schools and peer networks may trigger a stronger degree of co-ethnic socialisation, intensifying the transmission of norms about home-leaving and union formation and shaping preferences and expectations about the timing of these trajectories.

Nonetheless, while we control for a wide range of factors that influence residential location, our empirical strategy does not allow us to rule out the possibility that selection into immigrant neighbourhoods based on unobserved characteristics drives at least part of this presumed neighbourhood effect. Migrant families with low cultural or socioeconomic integration likely maintain more traditional norms about their children’s residential independence and union formation, and these families may also choose to live in proximity to co-ethnics or other immigrants. A lack of socioeconomic resources would further simultaneously influence residential sorting into deprived areas, while impeding children’s trajectories out of the home towards higher education or employment. In this case, our models would overestimate the negative association between neighbourhood immigration and remaining in the parental home.

We further hypothesised that the link between neighbourhood immigration and departures from the home would be particularly pronounced for second generation immigrants (H2). Evidence for this hypothesis was mixed due to substantial variation in the link between neighbourhood environments and home-leaving patterns by origin and gender. In line with expectations, we found specific patterns of home-leaving in immigrant neighbourhoods for the most disadvantaged immigrant origin groups and those with more traditional value orientations. Yet, we also found variation among the French majority linked to local environments. Finally, we found some evidence that neighbourhood patterns in home-leaving were more salient for women (H3).

Again, these patterns might be due to residential sorting or direct neighbourhood-level mechanisms. Non-European origin migrants and their descendants are more socioeconomically disadvantaged relative to other migrant groups, are more likely to live in deprived neighbourhoods, and face higher levels of job and housing market discrimination than European second generations (McAvay and Safi 2018; Meurs, Pailhé, and Simon 2006). These structural disadvantages would simultaneously influence residential location and hinder the transition to higher education, employment and financial and residential independence. This could explain the reduced home-leaving of Sub-Saharan African men. Further, prior research shows that traditional normative constraints, such as familialism and standards about acceptable partnership formation, are also stronger among North African, Sub-Saharan African and Turkish origins, and to a lesser extent Southern Europeans, shaping preferences to stay with parents longer and leave to enter marriage directly (Pailhé 2015; Hamel et al. 2018). If traditional values are associated with residential choices, sorting could explain some of the findings for these origin groups, such as the higher rates of departure for marriage among North African origin, and the decreased departure for unmarried cohabitation among Southern European origin women, in immigrant-dense neighbourhoods.

Still, these findings can also be interpreted in line with differential effects of neighbourhoods along origin and gender lines. On the one hand, the residential independence of Sub-Saharan African men may be inhibited in particular by a combination of origin- and address-based discrimination. On the other, North African and Southern European origin girls may be more constrained by traditional cultural norms in these environments. This interpretation is consistent with prior research suggesting that parental monitoring of girls is stronger in poor segregated neighbourhoods (Deville 2007). These within-origin gender disparities are, moreover, perhaps the least interpretable in terms of residential sorting, as it is unlikely that families of the same origin group sort differently into neighbourhoods based on their child’s gender. Rather, girls and boys are likely impacted differently by the influences of their local environment, in line with past studies on gender differences in neighbourhood effects (Sharkey and Faber 2014).

The unexpected result that the home-leaving patterns of the French majority vary by residential context may be explained by both mechanisms as well. The lower likelihood of the French majority to leave the home in immigrant neighbourhoods for union formation (whether unmarried or married) is in line with prior studies (Zorlu and Mulder 2011) and resonates with research showing that local immigration influences natives’ partner choice and fertility due to social interactions with different ethnic groups (Rahnu et al. 2020). Moreover, evidence from testing studies in France shows that natives from neighbourhoods with large immigrant shares also face job market discrimination based on place of residence (L’Horty, Duguet, and Du Parquet 2012). Yet, residential selection mechanisms could also come into play for the French majority if families living in these areas are particularly negatively selected in terms of unobserved socioeconomic characteristics.

Future studies could address some of the empirical limitations of this analysis. More precise data on the timing of home-leaving, along with multiple time-repeated observations on youth’s neighbourhood environments, could allow for improved causal designs to better disentangle direct neighbourhood effects from residential sorting mechanisms. Residential mobility and changes in living arrangements within inter-census periods cannot be tracked, which may involve other steps in the transition to adulthood (i.e. entry into higher education or the job market, couple formation and dissolution, return to the parental home, etc.). Despite these limitations, this paper shows that the concentration of immigrant populations has marked effects on their entry into adulthood, which may limit future opportunities and affect their overall life course. French urban policy has sought for decades to curb residential segregation by implementing social mix policies and requiring a minimum share of public housing in all municipalities. Yet these programmes have been undermined by the failure of many richer municipalities to comply with the law (Maaoui 2021). Accelerating scattered-site programmes that aim to build public housing in rich municipalities is therefore essential to promote not only economic but also social integration (Bolt 2009). Further, while France has implemented spatial affirmative action programmes to remedy the negative effects of living in deprived areas, still no policy frameworks exist for addressing ethnic/racial inequalities, in line with France’s colorblind political model (Sabbagh 2011). Our findings highlight the need for policies that address the intersection of spatial and ethnoracial inequalities in shaping life course trajectories.

Acknowledgements

This article is dedicated to the memory of Matthieu Solignac. We are also grateful to the members of the 2019 Alp-Pop Conference, colleagues at the Institut national d'études démographiques, the anonymous reviewers, and the editor for their suggestions. This work has been funded by the French National Research Agency (grant ANR-16-CE41-0007-01).

Disclosure statement

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

Notes

1 Including both French-born and foreign-born children of immigrants.

2 From 1968 to 1999, individuals born on the first four days of October were included in the panel; since 2006, 16 birthdays are used in January, April, July and October. This ensures coverage of approximately 1/100th of the population until 1999 and 1/25th since 2006.

3 The periodicity of EDP follows that of the French census. From 1968 until 1999, the French census was conducted on the entire population every 7–9 years. As of 2004, it is conducted every year on a portion of the population such that a full census is completed every five years. Likewise, 5 years must be aggregated to obtain a complete EDP wave. We compile years 2004–2008 and 2009–2013 to form the most recent panel waves and control for period of observation in all models. Prior studies show that the change in census collection does not bias mobility studies at a national level (Pan Ké Shon 2007).

4 French acronym for “aggregated units for statistical information.” IRIS were not implemented until 1999, prior to which the infra-municipality division used was the îlot. We used the îlot/IRIS correspondence table provided by INSEE to match the 1990 îlots with the 1999 IRIS.

5 Childhood and adulthood status is identified using a variable defining the individual’s position within the household.

6 Repeated individual observations are possible as individuals may be observed as children at more than one date.

7 There is a small risk that the same individual will be counted twice in the census, in particular if an individual resides simultaneously in two households. This could be the case for students who may be counted both in university dwellings as well as in the parental home. However, estimations of double counts are low, representing less than 2% of observations in the new census (Toulemon, Durier, and Marteau 2018). These cases are most frequent among students and children under the age of 18. Thus, the age restriction we implement in the analysis allows us to reduce the potential number of double counts.

8 In France, both country of birth and nationality at birth are typically used to define immigrant status. This is because return colonials and expats are not considered immigrants; despite having been born abroad, they have French citizenship at birth. To avoid confounding children of return colonials or expats with the immigrant origin sample, we exclude children whose parents are French citizens at birth and born outside of France.

9 These represent the largest immigrant origin groups in France. Sample sizes of children of migrants from other parts of the world are too small and are removed from the analysis.

10 Laos, Viet-Nam and Cambodia

11 In France, social class is measured using a socio-professional categorization that incorporates occupation with social class hierarchies. For parental occupation and education, we take the highest level achieved by the father or the mother.

12 Models using robust standard errors clustered at the neighborhood level were also estimated and gave consistent results.

13 This variable measures, out of the total neighborhood population, the share of immigrants in the neighborhood of the same origin group as the individual. For instance, for an individual of North African origin, it is equal to the share of North African immigrants in the neighborhood.

Appendix

Table A1. Multinomial Logistic Regression Model Predicting Departure Pathways Out of the Parental Home (Model 1)

	Model 1 Base outcome: Stayed in home
	Unmarried cohabitation	Marriage	Independent living
Neighbourhood immigrant share (t)	−1.142***	−0.784***	−0.959***
	(0.159)	(0.192)	(0.157)
Women	0.836***	1.196***	0.396***
	(0.016)	(0.022)	(0.016)
Immigrant origin/Ref: French majority
Western Europe	−0.122	−0.076	0.016
	(0.097)	(0.130)	(0.085)
Eastern Europe	−0.443***	−0.302†	−0.109
	(0.133)	(0.166)	(0.119)
Southern Europe	−0.283***	−0.101*	−0.148***
	(0.038)	(0.047)	(0.040)
North Africa	−1.455***	−0.253***	−0.387***
	(0.060)	(0.061)	(0.049)
Asia	−0.562***	−0.362†	−0.146
	(0.136)	(0.192)	(0.128)
Turkey	−1.523***	1.197***	−0.463***
	(0.168)	(0.125)	(0.135)
Sub-Saharan Africa	−0.752***	−0.366*	−0.162
	(0.124)	(0.176)	(0.102)
Individual covariates
Age	1.178***	1.828***	0.485***
	(0.028)	(0.052)	(0.022)
Age-squared	−0.025***	−0.037***	−0.009***
	(0.001)	(0.001)	(0.001)
Employment status/Ref: Students
Unemployed	−0.129**	−0.421***	−0.281***
	(0.041)	(0.052)	(0.049)
Active	0.241***	0.205***	−0.108***
	(0.027)	(0.033)	(0.032)
Inactive	−0.092**	−0.137*	0.010
	(0.033)	(0.058)	(0.030)
Born in France	0.275***	0.262***	0.097†
	(0.057)	(0.065)	(0.052)
Parental covariates
Age	−0.025***	−0.025***	−0.006***
	(0.001)	(0.002)	(0.001)
Occupation/Ref: Managers
Unemployed/Inactive	0.138**	0.028	−0.362***
	(0.045)	(0.062)	(0.045)
Blue collar	0.035	−0.017	−0.550***
	(0.035)	(0.048)	(0.034)
White collar	0.045	−0.040	−0.351***
	(0.033)	(0.047)	(0.031)
Farmers/artisans/small business owners	−0.113**	−0.039	−0.278***
	(0.036)	(0.048)	(0.033)
Intermediary professions	−0.011	−0.052	−0.218***
	(0.031)	(0.043)	(0.027)
Education/Ref: University
No education	0.087*	0.073	−0.773***
	(0.037)	(0.051)	(0.037)
Primary/middle school	0.082*	0.041	−0.573***
	(0.033)	(0.045)	(0.031)
Professional certificate	0.110***	−0.008	−0.485***
	(0.029)	(0.042)	(0.026)
High school diploma	0.073*	−0.033	−0.260***
	(0.031)	(0.045)	(0.027)
Number of children	0.022**	0.048***	−0.037***
	(0.008)	(0.010)	(0.008)
Number of rooms/Ref: 2 or less
3	−0.054	−0.136	−0.158*
	(0.077)	(0.105)	(0.077)
4	0.007	−0.081	−0.091
	(0.075)	(0.101)	(0.075)
5 or more	0.046	0.013	−0.026
	(0.075)	(0.102)	(0.075)
Type of household/Ref: Couple
Single parent (male)	0.205***	−0.017	0.231***
	(0.050)	(0.072)	(0.054)
Single parent (female)	−0.076**	−0.288***	0.147***
	(0.027)	(0.039)	(0.027)
Housing/Ref: Owner-house
Owner-apartment	−0.071†	−0.091†	−0.039
	(0.040)	(0.051)	(0.038)
Renter-house	0.207***	0.030	0.063*
	(0.030)	(0.044)	(0.031)
Renter-apartment	0.094*	−0.012	0.067†
	(0.038)	(0.053)	(0.037)
Public housing	0.107***	0.065†	−0.030
	(0.028)	(0.038)	(0.030)
Other	0.088†	0.039	0.072
	(0.047)	(0.064)	(0.046)
Lived in same neighbourhood in t-1	−0.181***	−0.195***	−0.117***
	(0.017)	(0.023)	(0.017)
Contextual covariates
City size/Ref: <100,000 inhabitants
>100,000 inhabitants	−0.293***	−0.155***	−0.369***
	(0.021)	(0.028)	(0.021)
Paris region	−0.465***	−0.512***	−0.692***
	(0.032)	(0.042)	(0.032)
Neighbourhood unemployment rate (t)	0.202	0.483**	−0.258*
	(0.126)	(0.160)	(0.131)
Share married in neighbourhood (t)	−0.302†	−0.096	−0.249
	(0.158)	(0.218)	(0.161)
Share of renters in municipality (t)	0.361***	0.318***	0.758***
	(0.063)	(0.085)	(0.063)
Period/Ref: 1990
1999	−0.581***	−1.405***	−0.555***
	(0.019)	(0.027)	(0.020)
2004	−1.500***	−3.185***	−1.282***
	(0.057)	(0.116)	(0.059)
2005	−1.675***	−3.273***	−1.535***
	(0.046)	(0.091)	(0.049)
2006	−2.021***	−4.011***	−1.694***
	(0.049)	(0.120)	(0.051)
2007	−2.177***	−4.410***	−1.853***
	(0.051)	(0.135)	(0.053)
2008	−2.352***	−4.544***	−2.007***
	(0.032)	(0.079)	(0.033)
Constant	−12.234***	−20.870***	−5.034***
	(0.292)	(0.522)	(0.243)
N	146,593	146,593	146,593

Source: Permanent Demographic Sample (1990-2013).

Sample: EDP individuals who are children in t and subsequently observed in t+1 aged between 19 and 31.

Table shows coefficients. Robust standard errors in parentheses.

*** p <0.001, ** p < 0.01, * p < 0.05, † p < 0.10

Table A2. Robustness Checks of the Neighbourhood Immigrant Share Coefficient from Different Estimation Samples

	Base outcome: Stayed in home
	Unmarried cohabitation	Marriage	Independent living
Estimation samples				N
Restricted to low parental SES	−1.047***	−0.429†	−0.630**	76,883
	(0.199)	(0.240)	(0.214)
Restricted to large cities	−0.693**	−0.523*	−0.515*	51,702
	(0.222)	(0.260)	(0.226)
Restricted to those observed 5 years after t	−2.185***	−1.234**	−1.269***	65,435
	(0.303)	(0.475)	(0.306)
Restricted to those observed 9 years after t	−0.711***	−0.572**	−0.782***	81,158
	(0.185)	(0.212)	(0.184)
Controlling for attrition (out-migration) as an outcome in t+1	−1.204***	−0.813***	−0.955***	243,387
	(0.156)	(0.185)	(0.155)
Alternative measurements of the neighbourhood immigrant share
Quantiles/Ref: Q1			146,593
Q2	−0.119***	−0.154***	−0.072**
	(0.022)	(0.031)	(0.023)
Q3	−0.223***	−0.240***	−0.167***
	(0.024)	(0.033)	(0.024)
Q4	−0.273***	−0.238***	−0.223***
	(0.028)	(0.039)	(0.029)
Q5	−0.338***	−0.283***	−0.234***
	(0.049)	(0.062)	(0.049)
Share of co-ethnics in the neighbourhood	−1.856**	0.394	−0.821	19,831
	(0.630)	(0.498)	(0.567)

Source: Permanent Demographic Sample (1990-2013).

Sample: EDP individuals who are children in t and subsequently observed in t+1 aged between 19 and 31.

Table shows coefficients. Robust standard errors in parentheses.

Models include all the covariates of Model 1.

*** p < 0.001, ** p < 0.01, * p < 0.05, † p < 0.10