Alongside employment and citizenship, access to housing is a key factor of immigrant integration. Yet the spatial concentration of immigrants and their descendants in certain types of housing – public housing in particular – and in certain neighbourhoods, may lead to spatial segregation. Drawing on data from the Trajectories and Origins survey (TeO) Haley McAvay focuses on the residential integration of immigrants in France. She reveals the mechanisms underlying disparities in housing tenure and analyses the probability of living in a residential area with a high concentration of immigrants by country of origin and immigrant generation.
1Over the past decade, the residential segregation of immigrants has received increasing scholarly attention in France. Although quantitative studies remain sparse, due in part to a lack of adequate data, extant research has revealed high levels of segregation between non-European immigrants and French natives (Pan Ké Shon and Verdugo, 2015; Préteceille, 2009; Safi, 2009; Verdugo, 2011). However, the degree to which spatial incorporation is occurring in housing and neighbourhoods remains largely undocumented. This empirical blind spot is surprising given that France’s history of post-colonial immigration is deeply linked to the development of the public housing sector  and the emergence of disadvantaged urban neighbourhoods. While the public housing construction boom of the post-war period was originally intended to provide middle-class households with affordable housing, throughout the 1970s and 1980s, these estates increasingly began to absorb poor immigrants and their families. While urban settlement policies were decisive in these location patterns (Bernardot, 1999), for some immigrants, entering the sector was also part of a residential strategy to improve living conditions and plan the transition towards homeownership (Boudimbou, 1993; Dietrich-Ragon, 2011). Non-European immigrants in particular became increasingly concentrated in the lower-quality segments of the sector located in peripheral urban areas (David, 2010; Pinçon, 1981; Simon, 1999).
2This article uses individual-level data from the 2008 Trajectories and Origins (TeO) survey to explore patterns of immigrants’ housing and neighbourhood outcomes. Unlike most studies of segregation which rely on census data, TeO offers rich information on country of birth, migratory background, nationality and experiences on housing markets, providing a unique opportunity to assess central mechanisms within the two major theories of immigrant segregation: spatial assimilation and place stratification.
3The analysis has two broad aims. The first is to confront these perspectives to immigrant homeownership and public housing access in France, focusing on the effects of age at migration, generation and geographical origin. Are housing tenure disparities between immigrants and the mainstream population  smaller among first-generation immigrants who migrated early in life? Are the housing outcomes of second-generation immigrants similar to those of the French mainstream population? And how does immigrant origin intervene in housing outcomes? The second aim is to explore the links between housing tenure and neighbourhoods across immigrant origins, focusing on whether living in public housing coincides with living in immigrant neighbourhoods, for non-Europeans in particular.
I – Spatial assimilation and place stratification in France
4The incorporation of immigrants has been a longstanding concern in the United States since the sociologists of the First Chicago School forged the spatial assimilation theory. According to this theory, the concentration of immigrants within ethnic enclaves is a natural starting point in a gradual process of acculturation and integration culminating in mobility out of segregated neighbourhoods (Alba and Nee, 2003; Park and Burgess, 1921; Park et al., 1925). The theory puts emphasis on length of stay and socioeconomic mobility as the primary driving forces of desegregation. As immigrants adjust to the host society over time and integrate the labour market, they will convert their economic gains into improved residential positions, leaving impoverished immigrant areas, notably through access to homeownership. Generational dynamics are of key theoretical importance to this perspective, with the assumption that successive immigrant generations (second generation) will have improved residential outcomes compared to earlier waves of migrants (first generation). Spatial assimilation can thus be understood as a largely mechanical process whereby immigrants living in France shed their distinctiveness over time and come to occupy similar housing and neighbourhoods as mainstream groups. While empirical evidence in the United States has generally supported this theoretical model, persistent inequalities among minorities have revealed the decisive role of race/ethnicity in shaping housing and neighbourhood outcomes, beyond individual-level factors (Alba and Logan, 1993; Massey and Denton, 1985, 1993; Painter et al., 2001; Pais et al., 2012; South et al., 2008).
5The place stratification perspective focuses on the role of race-based structural factors that are understood to facilitate or hinder minorities’ residential opportunities (Alba and Logan, 1991; Charles, 2003; Logan and Alba, 1993; Massey and Denton, 1985, 1993). This literature has highlighted how indirect and direct discrimination by a variety of institutions bolsters residential segregation by channelling minorities’ demand for housing towards lower-value real-estate markets in less desirable neighbourhoods (Bobo and Zubrinsky, 1996; Charles, 2003; Farley and Allen, 1987; Massey and Denton, 1993; Yinger, 1997). Place stratification also pays attention to the role of diverging residential preferences across groups: whites’ avoidance of non-white neighbourhoods, as well as minorities’ preferences for living in proximity to ethnic networks both contribute to enduring patterns of residential segregation.
6Initial evidence of immigrants’ locational and housing outcomes from France sheds light on the importance of disparities linked to immigrant origin. Safi (2009) calculates dissimilarity indexes at the municipality level to show that non-European immigrants have substantially higher levels of segregation with respect to French natives than European immigrants. These findings are largely echoed by Préteceille (2009) and Verdugo (2011) using similar methods. Pan Ké Shon and Verdugo (2015) point to higher spatial concentration of non-Europeans, despite decreasing segregation levels between French natives and immigrants over time. Verdugo (2011) finds that residential segregation has increased specifically among non-Europeans living in public housing. These patterns appear to be further sustained by restricted out-mobility among immigrants from co-ethnic areas (Rathelot and Safi, 2014) and disadvantaged neighbourhoods (Pan Ké Shon, 2010). While little is known about second-generation immigrants, Pan Ké Shon (2011) documents that second generations of Africans and Turkish origin are still over-represented in underprivileged neighbourhoods, although less concentrated than first-generation immigrants.
7The few studies on housing show similar patterns of stratification linked to immigrant origin. Simon (1995) documents that high proportions of European and Asian immigrants access homeownership, contrary to those of North African, sub-Saharan African and Turkish origin, a large share of whom live in public housing. These trends have been confirmed in more recent studies. Using longitudinal data, Gobillon and Solignac (2015) show that while the evolution of the housing gap over twenty-five years is consistent with spatial assimilation, the disadvantage of African immigrants with respect to natives does not decrease much over time. Other research shows that non-Europeans are disproportionately sorted into the public housing sector even after controlling for socioeconomic and life cycle factors (Fougère et al., 2011; Pan Ké Shon and Scodellaro, 2011; Verdugo, 2011).
8Are these origin-based disparities the reflection of differential demand for housing among immigrants and natives, or rather differential treatment on housing markets? Extant research from France points to how discrimination and preferences, both highlighted by the place stratification perspective, may be at play.
9On the one hand, entering the public housing sector may be a veritable residential strategy for immigrants. Some findings suggest that the public sector provides immigrants with opportunities to improve housing conditions, especially for those moving from expensive but low-quality housing on the private market (Boudimbou, 1993; Dietrich-Ragon, 2011). Public housing also represents a pathway to homeownership by increasing the savings capacity of households (Goffette-Nagot and Sidibé, 2016; Trevien, 2013). For immigrants who intend to become homeowners in the country of origin, public housing might be a viable choice in order to increase savings and future investments at home. The presence of ethnic networks in areas with substantial public housing opportunities could reinforce immigrants’ demand for the sector in immigrant neighbourhoods as well.
10On the other hand, qualitative studies investigating institutional discrimination within the French public housing sector give reason to believe that beyond preferences, structural constraints are also shaping immigrants’ residential outcomes. This research has shown that the public housing sector widely performs racial profiling in housing assignments, and that these exclusionary processes are inherent to the way public housing agencies operate under national urban policy initiatives relating to social diversity (mixité sociale) (Bourgeois, 2013; Sala Pala, 2013; Simon, 1999, 2003b; Simon and Kirszbaum, 2001; Tanter and Toubon, 1999; Tissot, 2005). The concept of social diversity emerged as an urban policy imperative in the early 1990s with a series of laws explicitly aiming to combat ghettoization. Although the policies do not officially claim to be concerned with the spatial distribution of minorities, non-European immigrants and their offspring are informally understood as the populations whose concentration should be avoided (Bourgeois, 2013; Sala Pala, 2013; Simon, 2003b). Yet paradoxically, such assignment practices actually reinforce segregation. These studies show that due to their real or perceived low socioeconomic status and cultural distance from the mainstream, populations of non-European origin are construed as potentially problematic tenants and neighbours, a threat to the prestige and value of real estate and neighbourhoods. Public housing agents thus informally draw on race/ethnicity to categorize “bad” (i.e. non-European) and “good” (i.e. French or European) candidates, sorting non-Europeans into low-quality housing where disadvantaged minorities are already present in large numbers. Cases of institutional discrimination have increasingly been brought before the French courts against public housing authorities for having excluded African households (Bourgeois, 2013; Sala Pala, 2013).
11Discrimination on the private real estate market is also a driving force of immigrant segregation and interacts with exclusionary practices in the public housing sector. Evidence from audit studies shows that discrimination on the private housing market is particularly acute against African renters, who are substantially less likely to be chosen for an apartment compared to mainstream applicants with identical characteristics (Bonnet et al., 2015; Bunel et al., 2017; HALDE, 2006).
12Three hypotheses guide this analysis. Following spatial assimilation, we expect to see origin-based disparities in housing decline with length of stay and generation. Individual-level factors should further absorb most of the difference between immigrants and the French mainstream population. If, however, groups have unequal access to housing markets, as place stratification posits, net differences in housing tenure will persist. This could mean lower access to homeownership among non-Europeans, and if the latter’s demand for housing is channelled toward the public sector, greater concentration in public housing. Moreover, we expect that non-Europeans living in public housing will have a greater odds of living in neighbourhoods with high shares of immigrants than other groups.
II – Data and methods
13Data come from the Trajectories and Origins (TeO) survey conducted by the French Institute for Demographic Studies (INED) and the French National Institute for Statistics and Economic Studies (INSEE). TeO is a nationally representative study conducted in 2008 on approximately 21,000 individuals residing in metropolitan France. The survey is one of the rare French data sources that contains detailed information on migratory characteristics, origin, housing tenure, and experiences on housing markets (Beauchemin et al., 2018). The data also include contextual variables pulled from the French census describing the immigrant composition of respondents’ neighbourhoods (IRIS ) and municipalities.
14Using information on country of birth and nationality at birth of respondents and their parents, the sample is categorized by immigrant generation and origin. Four generational categories are used (G1, G1.5, G2 and G2.5). First-generation immigrants are defined as individuals born outside of France without French citizenship at birth, and are distinguished by age at arrival: those who arrived in France after the age of 16 are referred to as G1 and those who arrived before age 16 are defined as G1.5. Any observed differences between G1 and G1.5 could be interpreted as effects of cultural and social integration, as the G1.5 generation would have been educated at least partly in France, may have a greater familiarity with the language, stronger social ties in France, etc., all factors which may widen homeownership opportunities for this group. Alternatively, G1.5 immigrants having arrived in France as children with their parents, differences with respect to G1 immigrants may be due to family status, as G1.5 would have had greater eligibility for entering public housing than G1 immigrants who arrived in France alone.
15Second-generation immigrants are French-born citizens whose parent(s) immigrated to France. Individuals with two immigrant parents are referred to as G2. Those with one immigrant parent and one French mainstream parent are defined as G2.5.  These categories thus capture the effects of nativity, age at migration and mixed background.
16Five immigrant origin categories have been defined on the basis of immigrants’ national origins (for G1 and G1.5) or the national origins of second-generation immigrants’ parent(s) (G2 or G2.5).  These categories pool together all respondents with an immigrant background regardless of generation and include: North Africa (Algeria, Morocco and Tunisia), sub-Saharan Africa, Turkey,  Asia (Cambodia, Laos and Vietnam) and Europe (Italy, Spain, and Portugal). Other origins are removed from the analysis. 
17Finally, the French mainstream population comprises French-born citizens whose parents are also both French citizens by birth.
18The total sample size is 11,319 individuals (Table 1) with a sizeable over-representation of first- and second-generation immigrants. Once weighted, 83% of the sample are members of the French mainstream population, 8% are first-generation immigrants, and 9% second-generation. People of North African and European origin represent the largest immigrant origin groups. Appendix Table A.1 provides summary statistics for all variables used in the analysis.
Generations and origins of the sample population
Generations and origins of the sample population(a) Country of origin of immigrants (for G1 and G1.5), country of origin of parent(s) for the second generation (G2 or G2.5).
Note: Once weighted, the French mainstream represent 83% of the sample.
19We first apply a multinomial logit model (Model 1) predicting housing tenure in three categories: homeowner, renter on the private market and public housing occupant. The model results are interpreted as relative risk ratios  of being a homeowner or living in public housing compared to renting. While the focus is on homeownership and public housing, this is not meant to imply a hierarchy of housing tenures. The choice of these outcomes is motivated by the theoretical and empirical literature assessed above: spatial assimilation theory stresses the transition to homeownership as part of the integration process, while previous literature from France has suggested a link between immigrant spatial concentration and public housing. An interaction between immigrant generation and origin is included, resulting in 20 groups including the mainstream population. The following individual and contextual controls are included: gender, age, education, household income per capita, marital status and family size, current unemployment, homeownership in a country other than France, the share of public housing/homeowners in the municipality, city size and département fixed effects.
20The second analysis, Model 2, draws on a simultaneous equations model, namely a bivariate probit model for categorical outcomes, in order to estimate housing tenure and neighbourhood attainment together. Since this model requires a binary dependent variable, housing tenure is dichotomized to indicate living in public housing (y1 = 1) versus renting or owning on the private market (y1 = 0). The neighbourhood outcome is measured using a dummy variable indicating residence in an IRIS with a high concentration of immigrants (y2), coded 1 when individuals live in the 10% of neighbourhoods concentrating the highest shares of immigrants in France and 0 otherwise.  The model allows for correlation of the error terms (Cameron and Trivedi, 2005). Furthermore, this estimation strategy makes it possible to explore different combinations of outcomes by calculating joint probabilities derived from the model (notably Pr(y1 = 1 and y2 = 1) and Pr(y1 = 0 and y2 = 1)). The bivariate probit model specifies two unobserved latent variables as:
22and the observed outcomes as:
24Model 2 includes the same independent variables as those in Model 1.
III – Results
1 – Disparities in housing tenure linked to generation and origin
25Table 2 reports housing tenure by immigrant generation and origin. Unsurprisingly, the lowest rate of homeownership is found among G1 immigrants, at 24%, while the mainstream population has the highest rate, at 54%. Yet, signs of generational assimilation into homeownership are also observed. The homeownership rate of immigrants who arrived in France before age 16 (G1.5) is substantially higher than that of G1 immigrants (42% versus 24%). Second-generation immigrants, especially those with one French mainstream parent (G2.5), also show a greater propensity towards homeownership than first-generation immigrants (G1).
26Generational trends in public housing occupancy further support the hypothesis that immigrants assimilate into homeownership and out of public housing. The French mainstream population has the lowest concentration in the public housing sector (13%). A gap of 8 percentage points separates them from second-generation immigrants with one French mainstream parent (G2.5). While second-generation immigrants with two immigrant parents (G2) live more frequently in public housing than G2.5 immigrants (32% versus 21%), they are less likely to do so than G1 immigrants (44%). Finally, the disparity of 11 points between G1.5 and G1 immigrants attests to lower concentration in public housing as a function of age at arrival.
Housing tenure by immigrant generation and origin (%)
Housing tenure by immigrant generation and origin (%)Note: Table shows weighted row percentages.
Interpretation: 54% of the French mainstream population are homeowners. 13% live in public housing.
27Table 2 also displays variation in housing tenure by immigrant origin. Disparities in housing tenure linked to origin are more pronounced than those linked to immigrant generation. Immigrants of sub-Saharan African origin and their descendants display the lowest homeownership rates (13%) and the highest concentration in public housing (51%), with a larger gap than that found between the French mainstream population and G1 immigrants overall. Immigrants from North Africa and their descendants also have high rates of public housing (43%) and low rates of homeownership (27%) relative to other groups, yet the difference are still much less marked than for people of sub-Saharan African origin. In contrast, half of immigrants from Asia and Europe and their descendants are homeowners and their presence in public housing is small (less than 20%). The first and second generations from Turkey hold an intermediate position, with a somewhat higher proportion of public housing tenants (39%) than homeowners (34%).
28These generational and origin disparities were tested using a multinomial logit model to predict housing tenure, all other variables being held constant (Appendix Table A.2). First, the results show the salience of factors related to the life cycle, socioeconomic status and residential context. Significant positive correlations are found for age, being married and having children and both accessing homeownership and living in public housing. Higher education and income are also positively correlated with homeownership and negatively correlated with public housing occupancy. Finally, the relative risk of homeownership decreases in large cities, while public housing occupancy increases.
29To grasp assimilation dynamics, Figure 1 displays the relative risk ratios of owning one’s home and living in public housing for all ethnic and generational groups. In terms of access to homeownership, compared to the descriptive statistics in Table 2, controlling for other factors considerably shrinks homeownership disparities with the mainstream population. A substantial gap is now only found for G1 and G1.5 immigrants from sub-Saharan Africa, who are about half as likely to be homeowners as the mainstream population. GI immigrants from Europe and North Africa also still face a barrier to homeownership, although this difference disappears among G1.5 immigrants. Controlling for other factors completely absorbs the homeownership disadvantage of G1 and G1.5 immigrants from Asia and Turkey; in G1.5, these origin groups actually have significantly higher homeownership opportunities than the mainstream population. By the time the second generation is reached, no significant disparities linked to origin are observed for any group.
30Immigrant origin appears to play a greater role than immigrant generation when it comes to living in public housing. Unlike the gap in homeownership which declines across generations, origin-based disparities in public housing persist. This is only true for groups of North African, Sub-Saharan African and Turkish origin, however. These groups are at least twice as likely to live in public housing than the mainstream population, regardless of immigrant generation. For all other groups, public housing outcomes are similar to or even lower than those of the mainstream population.
Relative risk ratios (RRR) of homeownership and public housing occupancy by immigrant generation and origin
Relative risk ratios (RRR) of homeownership and public housing occupancy by immigrant generation and originNote: Graphs display relative risks ratios compared to private renters.
Interpretation: Compared to the French mainstream population, G1 immigrants from sub-Saharan Africa are about twice as likely to live in public housing rather than rent on the private market.
2 – Different levels of social housing demand and housing market discrimination across immigrant-origin groups
31Are these residual effects of immigrant origin a result of preferences for public housing or housing market discrimination, or both? TeO provides information on public housing requests and experience of discrimination which help shed some light on these mechanisms. Demand for public housing  is indeed highest among both generations of North African and sub-Saharan African origin: between 15% and 23% of people in these groups have applied in the last 12 months (Table 3). Demand is also high among Turkish immigrants of both generations (around 13%), versus less than 7% among immigrants from Europe and Asia and the mainstream population. The differences are similar for the sub-sample of renters.
Demand for social housing and reported discrimination by immigrant origin
Demand for social housing and reported discrimination by immigrant origin(a) “During the past 12 months, have you or your partner applied for or renewed an application for public housing (including a request for transfer)?”
(b) “During the past five years, were you ever refused housing, either rented or purchased, without a valid reason?” This question is only asked to persons who have applied for housing or have moved over the last 5 years.
Note: Weighted percentages.
32Among people who had moved or applied for housing over the last five years, members of ethnic minorities more often report experience of discrimination on the housing market. North African and sub-Saharan African first-generation immigrants are considerably more likely to report experience of discrimination, at almost 20%. Exposure to discrimination does not appear to weaken substantially among the second generation. Second-generations of Turkish and Asian origin are also more likely to report experience of discrimination. Average rates, however, are much lower among the mainstream population, and among people of European and Asian origin.
3 – Relation between housing tenure and neighbourhood characteristics
33We will now explore whether housing and neighbourhood characteristics intertwine in different ways across immigrant origin groups by examining exposure to neighbourhoods with a high concentration of immigrants across housing tenures. Table 4 displays for homeowners, renters and public housing residents, the share of each group living in such neighbourhoods (at least 14% immigrants in the IRIS, cf. Appendix Table A.1.).
Neighbourhood outcomes by origin and housing tenure
Neighbourhood outcomes by origin and housing tenureInterpretation: 61% of people of North African origin who live in public housing reside in neighbourhoods with a high concentration of immigrants.
Note: Weighted percentages.
34Overall, a higher proportion of public housing occupants (36%) live in immigrant areas than renters (17%) or owners (10%). Yet, interestingly, the likelihood of living in an immigrant area varies substantially by group within the public housing sector: 24% of the mainstream population and 33% of public housing residents of European origin live in neighbourhoods with a high concentration of immigrants. This is the case for 60% of people of Asian, North African and Turkish origin and up to 70% of people of sub-Saharan African origin living in public housing. Disparities linked to origin are less pronounced among renters and owners.
35Comparing tenure-based differences in neighbourhood outcomes within the same origin groups is also revealing. Among people of North African and sub-Saharan African origin, renters are slightly more concentrated in immigrant areas than owners. Among the former, 41% of renters and 34% of owners live in such neighbourhoods, versus 50% and 40%, respectively, among the latter. Public housing residents of North African and sub-Saharan African origin much more frequently live in neighbourhoods with a high concentration of immigrants, however (more than 60%). Among public housing residents of Turkish and Asian origin, the pattern is similar, though the disparity with respect to other tenures is not as strong. Those of Asian origin, notably, are more likely to live in immigrant neighbourhoods as homeowners than as renters. Among people of European origin and the mainstream population, stratification across housing tenures is far more moderate.
36Bivariate probit models are used to test these descriptive findings net of controls by estimating public housing tenure (y1) and residence in an immigrant neighbourhood (y2) together (Appendix Table A2, Model 2). Two types of residential scenarios estimated by the model are shown in Figure 2: living in public housing in an immigrant neighbourhood ^Pr(y1 = 1; y2 = 1)h and renting or owning in the private sector in an immigrant neighbourhood ^Pr(y1 = 0; y2 = 1)h. 
37The results confirm the descriptive findings that living in public housing does not always coincide in the same way for all groups with living in spaces that concentrate immigrants. Among all immigrant generations, those of North African and sub-Saharan African origin have the highest net probability of this outcome, ranging from around 14% (G2.5 North Africa) to 29% (G2 sub-Saharan Africa). The chances that first-generation (G1) Turkish immigrants will live in public housing in immigrant neighbourhoods is likewise high. In contrast, people of Asian and European origin, regardless of generation, as well as the mainstream, have the lowest net probability of living in public housing in immigrant neighbourhoods, at around 10% or lower. The latter groups are indeed more likely to live in immigrant neighbourhoods in private housing than in public housing.
Two types of housing and neighbourhood outcomes by immigrant origin and generation
Two types of housing and neighbourhood outcomes by immigrant origin and generationInterpretation: G1 immigrants from North Africa have about a 19% net probability of living in private housing in an immigrant neighbourhood.
38This article used a unique survey on immigrants and their descendants in France to offer new findings on spatial incorporation in housing and neighbourhoods. While most extant research offers aggregate-level studies of segregation, this analysis draws on individual-level data to assess whether key mechanisms in the spatial assimilation and place stratification perspectives shape spatial attainment.
39The results point to both spatial assimilation and place stratification dynamics, partially confirming the first two hypotheses. Following spatial assimilation (first hypothesis), homeownership disparities between immigrant origin groups and the mainstream population have largely decreased with successive generations, and are partially absorbed by individual controls. Immigrant origin only appears to be a salient factor for immigrants of sub-Saharan African origin, for whom the lower rate of homeownership persists among G1 and G1.5 immigrants, before declining in the second generation. Origin appears to play a more determinant role in public housing occupancy. In this case, people of North African, sub-Saharan African and Turkish origin have the highest probability of living in public housing net of controls, and generational differences are not prominent. The predominance of immigrant origin in access to public housing tends to confirm our second hypothesis. Finally, in line with our third hypothesis, the link between living in public housing and living in immigrant neighbourhoods is more salient for people of North African, sub-Saharan African and Turkish origin.
40What is behind the residual effect of immigrant origin? The housing outcomes of Turkish origin immigrants revealed here notably diverge from Simon’s (1995) preliminary typology. While higher homeownership rates distinguish this group from African minorities, an equally substantial share continues to live in public housing. Turkish homeownership appears to be strongly determined by socioeconomic status; after controlling for these factors, people of Turkish origin have a higher relative risk than the mainstream of owning their homes. Prior research has shown that the integration patterns of Turkish immigrants are marked by a low degree of acculturation contrasted with a relatively high degree of socioeconomic integration (Safi, 2008; Simon, 2003a). Research suggests that ethnic support networks may provide resources to Turkish immigrants, possibly contributing to better job integration and greater homeownership. At the same time, people of Turkish origin, particularly the second generation, are not sheltered from inequality and discrimination (Meurs et al., 2006; Simon, 2003a), which may help account for their over-representation within public housing.
41The specific residential positions of people of African origin coincide with evidence from prior research of discrimination on both the private and public housing markets (Bonnet et al., 2015; Bunel et al., 2017; HALDE, 2006; Sala Pala, 2013), as well as with subjective experiences of discrimination documented in these analyses. However, the interpretation in terms of discrimination is complicated by the fact that a number of factors which influence residential outcomes remain unobserved. An alternative explanation is that the allocation of public housing to people of African origin in immigrant neighbourhoods may result from their specific housing preferences. If entering the public housing sector is a residential choice (Bonnal et al., 2013; Boudimbou 1993; Dietrich-Ragon 2011), such preferences will reinforce spatial sorting patterns. The geographical concentration of people of African origin in a few major urban areas in France, where housing prices are high, makes the public housing particularly financially attractive in these regions. The findings presented here indeed evidenced the high demand for public housing among African immigrants and their descendants.
42The residual effect of immigrant origin may also reflect the impact of ethnic networks. Research in the United States has shown that remaining in ethnic enclaves may be beneficial to integration (Logan et al., 2002). Similarly, evidence from France shows that for some groups, the presence of ethnic networks helps employment prospects (Toma, 2016). Ethnic networks also provide word-of-mouth information about housing options that may reinforce spatial clustering. Differential strategies during the public housing allocation procedure could also be responsible for diverging residential outcomes. Prior research into the allocation procedure shows that immigrants are more likely to accept the first housing offer they receive. Yet, initial offers tend to be for low-quality units in low-demand neighbourhoods where immigrants are already living. Public housing agents, aware that immigrants are less selective, may be in turn more likely to propose low-demand housing (Schmutz, 2015). Nonetheless, it is likely that preferences for public housing or immigrant neighbourhoods and discrimination do not operate as distinct mechanisms. Prior experience of discrimination on the private housing market may trigger higher demand for public housing, while broader feelings of social exclusion and hostility may foster a desire to remain in immigrant neighbourhoods. 
43Selective return migration patterns also potentially account for the differences between immigrants and the mainstream population. Prior evidence shows that the departure of immigrant renters from France over time biases cross-sectional estimations of immigrant homeownership (Gobillon and Solignac, 2015). Moreover, entering the public housing sector may be a residential strategy in line with a return migration project, as immigrants who plan to leave France in the future might opt for the less expensive public housing sector rather than invest in homeownership (Boudimbou, 1993). Unfortunately, the impact of return migration on spatial outcomes requires longitudinal data and cannot be properly measured with the survey used here.
44All in all, this article contributes new findings on spatial assimilation dynamics in housing in France. Yet further research is needed on these questions. Qualitative studies investigating residential preferences would help better disentangle the role of choice and constraint in immigrants’ locational outcomes. Quantitative research, drawing on longitudinal data with more precise neighbourhood measures (i.e. the share of immigrants by national origin), in which housing and neighbourhood outcomes can be tracked over the life course, would help determine the factors triggering mobility out of immigrant areas and into homeownership, as well as the durability of these disparities over time.
Description of the sample by immigrant generation and region of origin (%)
Description of the sample by immigrant generation and region of origin (%)Note: Table shows weighted averages.
Regressions predicting housing tenure (Model 1, multinomial logistic), and the fact of living in public housing and in a neighbourhood with a high concentration of immigrants (Model 2, bivariate probit)
Regressions predicting housing tenure (Model 1, multinomial logistic), and the fact of living in public housing and in a neighbourhood with a high concentration of immigrants (Model 2, bivariate probit)Note: Model 1 gives relative risk ratios. Model 2 gives coefficients. Standard errors in parentheses.
Significance levels: *** p<0.01, ** p<0.05, * p<0.10.
In France, in 2011, public housing represented about 18% of the total housing stock (Trevien, 2013). The public housing sector is not reserved only for the poor (Trevien, 2013; Whitehead and Scanlon, 2007). Income eligibility requirements are broad enough to include housing for middle-class households. Yet, the sector is highly socioeconomically segregated, with different categories of housing based on household income. Public housing also varies considerably across geographical areas in terms of quality, supply and demand.
The “mainstream” population comprises French nationals born in France and whose parents were also born French.
IRIS (an acronym of lots regroupés pour information statistique) are geographical units comparable to US census tracts. This territorial division was introduced by INSEE starting with the 1999 census. Most IRIS units contain between 1,800 and 5,000 inhabitants. All French municipalities of more than 10,000 inhabitants (and the majority of those with more than 5,000 inhabitants) are broken down into IRIS units. When information at the IRIS level is not available due to small municipality size, contextual variables at the municipality level are used instead.
The G2.5 category refers to individuals with two parents whose origin can be identified. In the case where the origin of one parent is unknown, the individuals are categorized according to the origin of the available parent.
When second-generation immigrants have two migrant parents with different national origins, the national origin of the father is used.
Given the low number of respondents of Turkish origin in the G2.5 category (26 individuals), the G2 and G2.5 categories of Turkish immigrants are combined.
These national origins were chosen as they represent the largest immigrant groups in France. The categorization broadly reflects geographical regions and is in line with other studies on immigrant integration in France (Meurs et al., 2006; Safi, 2008; Simon, 2003a).
Obtained by exponentiating the multinomial logit coefficients (ecoef), these are commonly interpreted as odds ratios.
For data privacy reasons, TeO does not contain the exact proportions of immigrants within each IRIS. Rather, this information is coded in deciles indicating where each respondent’s IRIS of residence falls within the immigrant share distribution for all IRIS units. See Table A.1 for the variable’s distribution.
This demand concern applications for housing, but also include requests for changes within the public housing sector made by public housing residents during the last twelve months. Excluding individuals who have been public housing residents for more than one year changes the rates but not the discrepancies between groups.
Probabilities are obtained using Stata’s margins command, which computes the average probability pr(y) for each category of the independent variable of interest, holding the other independent variables constant.
For example, see the work of Launay (2012) on the experience of public housing residents who enter affluent neighbourhoods through social mix policies but feel like intruders in these mostly white areas.