1 – Increasing residential segregation

7 Residential segregation has been studied extensively since the emergence of urban sociology in the United States at the beginning of the 20th century (Park and Burgess, 1925; Oberti and Préteceille, 2016). Thus, an exhaustive review of the sociological literature on the causes of segregation, its intensity, or its consequences on population is beyond the scope of this article. However, suffice it to say that racial and socioeconomic segregation (measured in terms of either income or socio-occupational status) has increased in large metropolitan areas since the beginning of the 21st century, although France maintains lower levels than the United States (Safi, 2009; Reardon and Bischoff, 2011; Quillian and Lagrange, 2016; Logan et al., 2018). In particular, segregation has increased in Paris as the rich and poor tend to concentrate in different neighborhoods (Préteceille, 2006).

8 Whereas “racial segregation is a major part of the neighborhood effects story in the USA” (Sampson, 2019, p. 8), the recent rise of wealth inequalities in advanced economies (Piketty, 2013; Alvaredo et al., 2019) suggests a need to reconsider the role of social stratification and socioeconomic inequalities. Most U.S. studies on residential segregation are based on income, while most analogous French studies are based instead on socio-occupational categories. This classification, created by the Institut National de la Statistique et des Études Économiques (INSEE), combines occupation-related information with the social situation associated with employment (distinguishing between employees and self-employed by occupational skill level, etc.). Nevertheless, both approaches (income-based and socio-occupational) reveal that segregation is increasing in both the United States and France (Rhein, 1998; Maurin, 2004; Préteceille, 2006; Fleury et al., 2013), albeit the rise is higher in the United States. In this paper, we are interested in the evolution of socio-spatial stratification in cities, and we thus study the changes in residential segregation for different socio-occupational categories of households benefiting from NILs in France.

9 Several research contributions to the literature on neighborhood effects indicate that the level of socioeconomic segregation in neighborhoods has major consequences on their inhabitants, especially children (Crane, 1991; Chetty et al., 2016). Living in a deprived neighborhood affects child development and adult outcomes in the form of access to education, job opportunities, residential mobility, health, and social inclusion (Wilson, 1987; Sharkey and Faber, 2014; Sampson, 2019).

10 Most of the literature appears to link poverty concentration to either housing deprivation (slums and informal settlements) or social housing programs. These programs were initially implemented after the Second World War to address housing shortages and improve the living conditions of working-class and middle-class households until their incomes rose and they gained access to employment stability and homeownership (Chombart de Lauwe, 1959). However, scholars in the 1990s began to increasingly contest their positive impact on social mobility as mass unemployment rose in advanced economies and massive urban riots broke out in large housing estates (Tissot, 2007). Thus, many scholars have begun to consider social housing programs to be a major contributor to social and racial segregation, both in the United States (Massey and Denton, 1993; Galster, 1999) and in Europe (Pan Ké Shon, 2009; Verdugo, 2011).

2 – Place versus individual-based programs to reduce segregation

11 Consequently, public policies since the 1990s have taken two different approaches to reducing the spatial concentration of the poor and racial groups in specific neighborhoods: they intervene at either the neighborhood or individual level, thus facing the famous “place versus people” dilemma (Galster, 2017). Place-based programs aim to rehabilitate poor neighborhoods and buildings, such as Hope VI in the United States (which involves demolishing, refurbishing, and privatizing social housing) and the Programme National de Rénovation Urbaine in France (a national urban policy that aims to demolish low-quality social dwellings in the poorest neighborhoods and enact new small-scale social housing construction projects in mixed areas). In contrast, individual-based programs aim to increase “the opportunities for low-income households to reside outside of deprived neighborhoods where, presumably, opportunities for socioeconomic advancement and quality of life are enhanced, while also increasing housing affordability and quality” (Galster, 2017, p. 267). These programs include housing vouchers coupled with mobility assistance for low-income families. During the 1990s, the U.S. federal government developed Section 8 programs such as the Housing Choice Voucher (HCV), [3] while the “Moving to Opportunity” experiment proposed housing vouchers to encourage poor households to move into better neighborhoods.

12 Many studies have tried to evaluate the impact of such housing policies on their recipients. An analysis of the 100 biggest U.S. metropolitan areas in 2000 and 2008 showed that Housing Choice Voucher recipients have been suburbanizing over time along with other poor households (Covington et al., 2011). Moreover, despite the substantial income supplement from the vouchers, most voucher recipients have remained severely economically deprived (Galster, 2017).

13 Our article aims to contribute to this debate by studying the objective and subjective dimensions of residential mobility among those benefiting from an individual-based program, the NIL; and we will do so by combining data at both the individual and municipality levels.

3 – Mortgage and real estate market

14 The location choices of low- and middle-income households trying to access homeownership have become more constrained by the recent rise in housing prices relative to average earnings in many developed countries and by the existence of borrowing constraints. In the United States, the mortgage market is also considered to be a major determinant of racial segregation (Ross and Yinger, 2002) because of direct or indirect discrimination in access to credit (Massey and Denton, 1993; Munnell et al., 1996; Pager and Shepherd, 2008). More recently, some studies have revealed that the rise in subprime lending and the ensuing wave of foreclosures have affected mainly racial minorities (Rugh and Massey, 2010; Ross and Squires, 2011). In Europe, Aalbers (2011) shows frequent occurrences of mortgage redlining —the rejection of mortgage loan applications solely on the basis of residential place —thus leading to high levels of socioeconomic segregation. Despite great interest in this research, the role of the mortgage market in segregation processes has been less studied in Europe, and France is no exception (Lambert, 2016).

4 – The no-interest loan in France

15 In France, several measures have been implemented or reactivated to encourage homeownership: direct financial support to households (in the form of subsidies), tax incentives (through tax exemptions on mortgage interest for the main residence), and subsidized loans (on which the state pays the interest). The prêt à taux zéro is the most widespread of these measures. Created in 1995 to help first-time home buyers, this NIL is a complementary interest-free loan that cannot exceed 20% of the transaction amount or 50% of the household’s total loan amount for purchasing the dwelling. There is an income criterion for accessing a NIL that depends on family structure (the threshold increases with the number of children) on the one hand, and on location of the dwelling to be purchased on the other. The latter initially distinguished only between the Paris region and the rest of France (where the threshold was lower); but three types of zones were established after 2003, depending on housing market tightness. [4] Because the income criterion is not very restrictive, the NIL was accessible to 94% of tenant households outside the Paris region in 1998 (Gobillon and Leblanc, 2008). [5] According to the 2002 and 2006 housing surveys, the yearly average value of NILs was around €16,000. The NIL was first restricted to newly built homes before it was extended to existing homes in 2005.

16 The NIL is expected to affect the housing decisions of credit-constrained households on the borderline of being able to access homeownership. As banks consider it an additional down payment, it should help modest-income households to purchase a dwelling. In that perspective, it should affect manual and clerical workers to a larger extent. Still, in large cities where housing prices are high, the NIL may also allow higher-income households to purchase homes in places rather close to the city center, while low-income households may not be provided enough help to access homeownership in such attractive areas. Such a case thus makes the expected overall effect ambiguous. As for spatial effects, modest households may be able to purchase only dwellings far away from city centers, since housing prices are often lower at the periphery of French cities. By extension, one may wonder whether households relocating far away from city centers anticipate all the potential harmful effects of distance, specifically regarding accessing the labor market and consumption amenities (such as nurseries, schools, and shops).

1 – Changes in housing transitions across occupations after introducing NILs

25 Here, we compare the rates of first-time home purchases for different socio-occupational categories before and after the creation of the NIL program. For that purpose, we use the 1996, 2002, and 2006 housing surveys and restrict the sample to households who were tenants 4 years before the survey date. We report in Table 1 the differences between socio-occupational categories and managers/professionals (taken as the reference) in the proportions of households in each wave who purchased a new dwelling, purchased an existing dwelling, and remained tenants (with or without a residential move). The proportions are reported in Appendix Table D.1. First, note that the proportions of managers/professionals purchasing existing and new dwellings increased between 1996 and 2002 (respectively, from 17.5% to 23.7%, and from 4.9% to 5.4%) before stabilizing. The difference in the proportions of manual (resp. clerical) workers and managers/professionals purchasing existing dwellings in the 1996 wave is large, reaching −11.5 (resp. −12.6) percentage points for manual (resp. clerical) workers. It grows even larger over time, reaching −17.9 (resp. −18.7) percentage points for the 2006 wave (Table 1). Patterns are similar when separately considering the Paris region and other regions. When turning to the new dwelling purchases, we find that differences in percentage points are initially much smaller than for existing dwellings, reaching, respectively, −1.6 and −3.2 percentage points for manual and clerical workers in the 1996 wave. Interestingly, differences increase only very slightly over time and they amount, respectively, to −1.9 and −3.7 percentage points for the 2006 wave. Again, the patterns are similar when separately considering the Paris region and other regions. As NILs can be used only to purchase new dwellings, the results suggest that NILs may limit the increase in disparities among socio- occupational categories regarding their access to new dwellings.

26 An important issue is that household and local characteristics within socio-occupational categories may vary over time and be correlated with the propensity of becoming the homeowner of a new or existing dwelling. To deal with this issue, we use the pooled 1996, 2002, and 2006 housing surveys to estimate a multinomial logit model with the following three outcomes: (1) purchased a new dwelling during the 4 years preceding the survey date, (2) purchased an existing dwelling, and (3) remained a tenant (reference outcome). [11] The specification involves household and local characteristics, and in particular indicators for socio-occupational categories as well as their interactions with indicators of the survey wave after the introduction of NILs (2002 or 2006), taking the 1996 survey wave as a reference. The model is estimated for the whole of France and then separately for the Paris region and other regions. Details on variables introduced in the specification and estimated odds ratios are reported and discussed in Appendix E.1. From the estimated coefficients, we then compute for each socio-occupational category the evolution of the probabilities of purchasing existing and new dwellings, for given distributions of household and local characteristics (using those of the considered category in the 1996 survey, before NILs were introduced). Finally, we compute differences in these evolutions between each socio-occupational category and managers/professionals. In this way, we measure differences in evolution between socio-occupational categories, such that the composition of households within each category is held constant over time.

Table 1. Differences between socio-occupational categories and managers/ professionals in the proportions of tenants (in percentage points) for every housing transition type and use of NILs, in 1996, 2002, and 2006

Pooled Paris region Other regions 1996 2002 2006 1996 2002 2006 1996 2002 2006 Purchased an existing dwelling Manual workers −11.5 −15.6 −17.9 −11.8 −14.8 −19.2 −12.5 −16.6 −18.6 Clerical workers −12.6 −17.6 −18.7 −11.8 −17.6 −18.0 −13.7 −18.2 −19.6 Intermediate occupation −6.5 −7.9 −8.6 −7.7 −11.0 −11.4 −6.7 −7.2 −8.3 Managers/professionals 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Others −10.0 −11.4 −16.7 −9.0 −13.4 −16.2 −11.2 −11.7 −17.7 Purchased a new dwelling Manual workers −1.6 −1.4 −1.9 −1.5 −0.7 −1.4 −2.6 −3.0 −3.8 Clerical workers −3.2 −3.3 −3.7 −2.2 −1.3 −2.1 −4.3 −5.0 −5.6 Intermediate occupation 0.9 −0.6 0.7 −0.6 −1.3 −0.5 0.9 −1.2 −0.2 Managers/professionals 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Others −1.8 −2.3 −0.5 −1.8 −0.9 −0.2 −2.7 −3.9 −2.2 Without NIL Manual workers - −1.6 −2.2 - −0.7 −1.3 - −2.4 −3.2 Clerical workers - −1.8 −2.7 - −0.8 −1.5 - −2.6 −3.7 Intermediate occupation - −1.0 −0.9 - −0.3 −0.5 - −1.6 −1.7 Managers/professionals - 0.0 0.0 - 0.0 0.0 - 0.0 0.0 Others - −1.1 −0.1 - 0.6 0.0 - −2.1 −0.9 With NIL Manual workers - 0.2 0.3 - 0.0 −0.1 - −0.6 −0.6 Clerical workers - −1.5 −1.1 - −0.6 −0.6 - −2.4 −1.9 Intermediate occupation - 0.4 1.6 - −1.1 −0.0 - 0.4 1.5 Managers/professionals - 0.0 0.0 - 0.0 0.0 - 0.0 0.0 Others - −1.1 −0.4 - −1.5 −0.2 - −1.9 −1.3 Remained tenant (with or without a residential move) Manual workers 13.1 17.0 19.8 13.3 15.5 20.6 15.1 19.6 22.4 Clerical workers 15.8 20.9 22.4 14.0 18.9 20.1 18.0 23.2 25.2 Intermediate occupation 5.5 8.4 7.9 8.4 12.4 11.9 5.8 8.4 8.6 Managers/professionals 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Others 11.8 13.6 17.2 10.7 14.3 16.4 14.0 15.7 19.9

Table 1. Differences between socio-occupational categories and managers/ professionals in the proportions of tenants (in percentage points) for every housing transition type and use of NILs, in 1996, 2002, and 2006

Note: The proportions of tenants for every housing transition type and use of NILs used for the computations are provided in Appendix Table D.1. When computing the proportion of tenants purchasing a new dwelling, the 65 households with missing NIL information are dropped.
Coverage: All households who were tenants 4 years before the survey date.

27 Results on the adjusted probabilities confirm our descriptive statistics (Table 2). Before the introduction of NILs, the difference in the adjusted probabilities of purchasing an existing dwelling is large, reaching −11.5 (resp. −12.6) percentage points when contrasting manual (resp. clerical) workers with managers/professionals. After the introduction of NILs, differences are larger and amount to, respectively, −17.9 and −19.4 percentage points. Turning to the adjusted probabilities of purchasing new dwellings, we find that differences are much smaller than for existing dwellings before the introduction of NILs: −1.7 and −3.3 percentage points, respectively, for manual and clerical workers. After the introduction of NILs, they remain stable.

Table 2. Differences between socio-occupational categories and managers/ professionals in the adjusted proportions of tenants (in percentage points) for every housing transition type before and after the introduction of NILs

Pooled Paris region Other regions Before NIL After NIL Before NIL After NIL Before NIL After NIL Purchased an existing dwelling Manual workers −11.5 −17.9 −11.8 −16.7 −12.4 −21.9 Clerical workers −12.6 −19.4 −11.8 −17.4 −13.7 −23.7 Intermediate occupation −6.4 −8.5 −7.7 −10.0 −6.6 −9.7 Managers/professionals 0.0 0.0 0.0 0.0 0.0 0.0 Others −10.0 −15.2 −9.0 −14.5 −11.2 −18.6 Purchased a new dwelling Manual workers −1.7 −1.8 −1.5 −0.9 −2.8 −2.2 Clerical workers −3.3 −3.4 −2.2 −1.8 −4.5 −3.3 Intermediate occupation 0.9 −0.2 −0.5 −0.9 0.8 −0.6 Managers/professionals 0.0 0.0 0.0 0.0 0.0 0.0 Others −1.7 −1.2 −1.8 −0.9 −2.6 −1.7 Remained tenant (with or without a residential move) Manual workers 13.2 19.7 13.3 17.6 15.2 24.1 Clerical workers 15.9 22.8 14.0 19.2 18.1 27.0 Intermediate occupation 5.5 8.6 8.3 10.9 5.9 10.3 Managers/professionals 0.0 0.0 0.0 0.0 0.0 0.0 Others 11.7 16.4 10.7 15.4 13.9 20.3

Table 2. Differences between socio-occupational categories and managers/ professionals in the adjusted proportions of tenants (in percentage points) for every housing transition type before and after the introduction of NILs

Note: Diff. = difference. We compute for each socio-occupational category the evolution of the probabilities of purchasing existing and new dwellings, for given distributions of household and spatial characteristics (using those of the considered category in the 1996 survey, before NILs were introduced). They are derived from the estimates of a multinomial logit model of housing transitions for tenants, detailed in Appendix E.1.
Coverage: All households who were tenants 4 years before the survey date.

2 – Disparities in housing transitions and the use of NILs across occupations

28 We now focus on housing transitions and the use of NILs after the no-interest loan policy was implemented. For that purpose, we use only the 2002 and 2006 housing surveys and report in Table 1 differences between each socio-occupational category and managers/professionals in the proportions of previous tenants purchasing an existing dwelling, a new dwelling with NIL, and a new dwelling without NIL (proportions are reported in Appendix Table D.1). First, note that the proportions of managers/professionals purchasing a new dwelling without and with a NIL were, respectively, 3.2% and 2.2% in 2006 (Appendix Table D.1). The difference between manual (resp. clerical) workers and managers/professionals for the purchase of a new dwelling without NIL reaches −2.2 (resp. −2.7) percentage points in 2006, and it is even smaller for the purchase of a new dwelling with a NIL, at +0.3 (resp. −1.1) percentage points (Table 1). Patterns are similar for the Paris region and other regions, but, interestingly, differences are a bit starker for new dwelling purchases with and without a NIL outside the Paris region. Overall, these findings show smaller disparities (in percentage points) across socio-occupational categories in new dwelling purchases with rather than without a NIL.

29 Manual/clerical workers and managers/professionals likely have different household and local characteristics that affect their ability to purchase existing and new dwellings, as well as their access to NILs. To deal with this issue, our model now distinguishes, among households who purchased a new dwelling, between those who benefited from NILs and those who did not. We thus use the pooled 2002 and 2006 housing surveys to estimate a multinomial logit model with the following four outcomes: (1) purchased an existing dwelling, (2) purchased a new dwelling without a NIL, (3) purchased a new dwelling with a NIL, and (4) remained a tenant (reference outcome). The specification includes household and local characteristics, as before, and in particular indicators for socio-occupational categories. Details on the variables introduced into the specification and estimated coefficients are reported and discussed in Appendix E.2. From the estimated coefficients, we can compute for each socio-occupational category the probabilities of purchasing existing and new dwellings with and without a NIL, for given distributions of household and local characteristics (using those of managers/professionals). We can then compute the differences in these probabilities between each socio-occupational category and managers/professionals. In this way, we recover differences in housing choices between socio-occupational categories —which are possibly related to differences in income —but net of the impact of other characteristics. Results reported in Table 3 confirm very large differences in the purchase of an existing dwelling between manual (resp. clerical) workers and managers/professionals (respectively, −17.8 and −12.8 percentage points) and much smaller differences in the purchase of a new dwelling without a NIL (respectively, −3.3 and −1.8 percentage points). Almost no differences exist in new dwelling purchases with a NIL (respectively, −1.1 and −0.7 percentage points). Patterns for the Paris region and other regions are again similar. Overall, these results confirm that differences in new dwelling purchases between socio-occupational categories come from differences in purchases without a NIL.

30 Differences in access to homeownership across socio-occupational categories are likely to capture differences in income. We thus estimate an alternative specification using income quintiles to capture household heterogeneity in financial capacities rather than socio-occupational categories. The results are very close to those obtained with socio-occupational categories, although the disparities are larger (see Appendix Table F.1). Ultimately, one may argue that occupation is endogenous in our regressions, since it is measured at the survey date and can thus be influenced by a housing transition. Access to homeownership in a distant suburb can influence access to employment and thus make some new homeowners change jobs and possibly even the occupations associated with those jobs. Thus, we assess in Appendix Table F.2 the robustness of our results when using educational attainment, which is exogenous, rather than socio-occupational categories. In particular, we contrast individuals who have no diploma or a vocational degree with those who have a 2-year higher education degree. Looking at the results obtained from studying not only the housing transitions over the whole period [12] but also housing transitions together with the use of NILs after their introduction, we find that they are very consistent when contrasting occupations and when contrasting diplomas.

Table 3. Differences between socio-occupational categories and managers/ professionals in the adjusted proportions of tenants (in percentage points) for every housing transition type and use of NILs after the introduction of NILs

Pooled Paris region Other regions Purchased an existing dwelling Manual workers −17.8 −16.3 −17.4 Clerical workers −12.8 −11.9 −12.6 Intermediate occupation −8.5 −9.6 −7.6 Managers/professionals 0.0 0.0 0.0 Others −12.5 −13.5 −11.7 Purchased a new dwelling without NIL Manual workers −3.3 −1.7 −3.8 Clerical workers −1.8 −0.9 −2.1 Intermediate occupation −1.4 −0.5 −1.9 Managers/professionals 0.0 0.0 0.0 Others −1.3 −0.1 −1.7 Purchased a new dwelling with NIL Manual workers −1.1 −0.1 −1.3 Clerical workers −0.7 −0.4 −0.8 Intermediate occupation 0.2 −0.5 0.4 Managers/professionals 0.0 0.0 0.0 Others −1.3 −0.8 −1.3 Remained tenant (with or without a residential move) Manual workers 22.2 18.1 22.5 Clerical workers 15.2 13.2 15.5 Intermediate occupation 9.8 10.5 9.0 Managers/professionals 0.0 0.0 0.0 Others 15.1 14.4 14.8

Table 3. Differences between socio-occupational categories and managers/ professionals in the adjusted proportions of tenants (in percentage points) for every housing transition type and use of NILs after the introduction of NILs

Note: We compute for each socio-occupational category the probabilities of purchasing existing and new dwellings with and without a NIL, for given distributions of household and spatial characteristics (using those of managers/professionals). They are derived from the estimates of a multinomial logit model of housing transitions and use of NILs for tenants, detailed in Appendix E.2.
Coverage: All households who were tenants 4 years before the survey date.

31 Combined with the decrease in interest rates that allowed households with sufficient income to borrow, the increase in real estate prices during the NIL program period may have counterbalanced the specific effects of NILs and limited their redistributive effects. Additionally, given the income thresholds specified by the law, the NIL program may have benefited mostly the least affluent managers/professionals and the most affluent manual workers. This could explain why the socio-occupational differences in access to homeownership of new dwellings with NILs are smaller than the differences in access to homeownership of both existing and new dwellings without NILs. Overall, the NIL program seems to have redistributed flows of new homeowners across segments of the market. In particular, it contributed to channeling low socio-occupational groups toward new home purchases —which is one of the most complex pathways to homeownership, given the large number of public and private intermediaries involved, namely: subdividers, landowners, home builders, notaries, planning departments, and others (Bourdieu, 2000). It thus remains unclear if access to homeownership enabled by the specific terms of NILs was more favorable to manual and clerical worker households than were other pathways to homeownership from which they were increasingly excluded.

1 – The characteristics of previous and new municipalities of residence

32 Hereafter, we use NIL data to assess variations across socio-occupational categories in the evolution of municipal characteristics for households with a NIL. We first assess the extent to which access to homeownership is associated with a change in city size. Table 4 provides the proportion of first-time homebuyers with a NIL by urban unit size bracket for each socio-occupational category before and after moving. It shows that, whatever the category, the proportion of households located in urban units with more than 50,000 inhabitants decreased after the move, while the proportion of those living in rural areas increased. [13] Only 11.6% of manual workers resided in urban units with more than 200,000 inhabitants after their move, compared to 24.3% of clerical workers and 36.9% of managers and professionals.

33 We then investigate whether access to homeownership goes along with larger proportions of households in the city outskirts. For that purpose, we decompose the area outside urban units into three categories: (1) within an urban area but outside of urban units, i.e., in the outskirts of a single city; (2) within several urban areas but outside of urban units, i.e., in the outskirts of several cities; and (3) rural, outside of urban areas, i.e., farther than the outer fringes of a city. Interestingly, Table 4 shows that the proportion of households living in the outskirts of a single city is much larger after purchasing a dwelling for all socio-occupational categories, but especially for manual workers. The increase in this proportion is as large as 9.2 percentage points for manual workers, compared to 8.3 percentage points for clerical workers and 6.3 percentage points for managers and professionals. Such differences are observed to a lesser extent for the proportions of households in the outskirts of several cities and farther areas.

Table 4. Municipal characteristics before and after NIL-assisted home purchase

Variables Managers/ Professionals Clerical workers Manual workers Before After Before After Before After Urban Unit Size (1999 definition) Paris 19.1 17.4 11.7 9.6 4.3 3.0 200,000–2,000,000 25.9 19.5 21.0 14.7 14.4 8.6 50,000–200,000 14.1 10.6 16.0 11.0 13.5 8.1 20,000–50,000 6.8 5.4 8.1 5.8 8.2 4.9 < 20,000 15.0 18.1 18.0 20.4 23.2 21.7 Rural 19.2 28.9 25.2 38.4 36.5 53.6 Urban Category of the Municipality in 1999 Within an urban unit 80.8 71.1 74.8 61.6 63.5 46.4 Within an urban area but outside urban units 9.3 15.6 10.8 19.1 13.4 22.6 Within several urban areas but outside urban units 2.4 4.0 3.1 5.1 5.1 8.0 Rural, outside urban areas Unemployment rate in 1999 7.5 12.5 9.3 11.7 11.3 12.9 14.3 11.9 18.0 12.3 23.0 11.2 Household income by consumption unit (in euros) 12,795 12,488 11,815 11,519 11,073 10,833 Proportion of social housing units among principal residences in 2000 15.6 13.0 14.8 11.3 13.1 8.8 Proportion of homes in detached houses in 2000 50.4 60.6 57.6 68.6 67.4 78.6 Price/m2 of dwellings in 2000 1,372 1,324 1,222 1,170 1,071 1,031 Number of dwellings in the municipality in 2000 26,837 16,864 19,433 11,164 11,434 5,004 Number of dwellings in the urban unit in 2000 735,217 684,876 419,613 420,615 231,290 188,258 Proportion of managers/ professionals in 1999 15.3 14.0 12.7 11.2 10.1 8.8 Proportion of workers in intermediate occupations in 1999 20.7 20.6 19.8 19.4 18.3 17.6 Proportion of clerical workers in 1999 29.6 29.2 29.8 29.2 28.5 27.6 Proportion of manual workers in 1999 22.6 23.9 25.3 27.0 30.0 31.8 Proportion of other categories of workers in 1999 Number of observations 11.8 123,447 12.4 123,447 12.5 360,959 13.2 360,959 13.1 392,165 14.2 392,165

Table 4. Municipal characteristics before and after NIL-assisted home purchase

Note: For every socio-occupational category, the mean of each variable before and after NIL-assisted home purchase is estimated with great accuracy due to the large sample size, and differences between means before and after NIL-assisted home purchase are all significant at the 1% level.
Coverage: Tenants with a NIL contracted between January 1997 and December 2005.

34 As neighborhood socioeconomic composition contributes to determining access to resources (job opportunities, schools, social network, etc.), we also examine the socioeconomic profile of the residential municipalities before and after tenants access homeownership with a NIL. Table 4 shows that after managers/professionals and manual/clerical workers moved, they lived in municipalities that were less wealthy since they were characterized by lower income per consumption unit and lower housing prices. These municipalities also exhibited lower proportions of managers/professionals and workers in intermediate occupations, but a higher proportion of manual workers. This is especially true for NIL-assisted manual workers and managers/professionals. At the same time, municipalities after the move were characterized by a lower unemployment rate and a smaller proportion of social housing than municipalities before the move.

35 Regarding NIL-assisted manual workers, they accessed homeownership in small municipalities that were markedly poorer than the new municipalities of NIL-assisted managers/professionals. Moreover, these municipalities were characterized by lower proportions of managers/professionals and workers in intermediate occupations, but also by a lower proportion of social housing. This is because, to a larger extent, manual workers moved to peri-urban or distant rural areas where the proportion of social housing is structurally low, as social housing in France is concentrated mostly in urban metropolitan areas. Among new homeowners with a NIL, clerical workers occupied an intermediate position between managers/professionals and manual workers in terms of municipal income, proportion of socio-occupational categories, and proportion of social housing after moving.

2 – From objective indicators to subjective perceptions of residential mobility

36 As shown above, for manual and clerical workers accessing homeownership with a NIL, income per consumption unit and the proportion of social housing were lower in the new municipality than in the previous one. How did these socio-occupational groups perceive their destination municipalities, where fewer situations of poverty seem to exist than in their previous municipalities? Even more specifically, did lower-class households value peri-urban and rural municipalities with mainly individual accommodations as well as low building and population density? Were there differences across genders within couples, who represent the majority of first-time homebuyers?

37 Our in-depth interviews in a peripheral area of Lyon show that two determining factors in NIL-assisted households’ choice of residential location were greater housing availability and lower real estate prices in low-density areas. Lower-class households did not mention the attractiveness of the “countryside” and the “green environment” (low urban density, proximity to nature, low pollution, etc.) when referring to their new places of residence. Even if they left metropolitan areas, none of them valued the old stone houses, traditional French village architecture, or built heritage. On the contrary, the majority mostly associated living in the countryside with peasantry, a traditional economy, and a lack of modern accommodations. Moreover, no household mentioned prioritizing the technical and aesthetic characteristics of the building (type, dimension, or materials). As stated by one 52-year-old father of four and manual worker in the car industry, “I don’t care what the house is like, all I care about is the social environment.”

38 In contrast, when referring to their new neighborhoods, households rather emphasized the need for urbanizing the area (which involves building new shopping centers, amenities such as sidewalks, playgrounds for children, high schools, and providing public buses, etc.). For example, a 30-year-old young mother of four who left school at 16, grew up in a social housing project in Lyon, and married a construction worker, described her new neighborhood as “a hole,” “in the middle of nowhere,” adding:

39

Here, besides home and the school, there’s nothing to do. It’s a good thing I have my car! And a good thing that Cleyzieu is developing a lot, they’re going to build more housing estates and all that, it’ll be less like the countryside. […] But when we feel down, we’re sick of it, we say to ourselves: “Why did we ever come here to this hole?”

40 Moving away from the city center (and from large social housing projects in high-density areas) was motivated by issues such as poverty, social and racial segregation, and poorly performing schools. A father of two who recently arrived in Cleyzieu-Lamarieu even declared, “I was looking for schools where the proportion of Pierre, Paul, and Jacques was normal. [Interviewer: Meaning what?] Around 80% of the pupils.” This suggests that the main goal of the interviewed households was not to access homeownership and an individual dwelling, but instead to escape from the large public housing projects built in the 1960s in the suburbs of Lyon, although not specifically from the city.

41 In-depth interviews also enabled us to reconstruct all the steps taken in searching for a better home, which is otherwise not possible using current quantitative data. What kind of dwelling were the households originally looking for? How many dwellings did they visit? And how did they change their housing target during the search (with respect to location, housing tenure status, and housing type)? We found qualitative evidence that low-income families first tried to relocate to a better neighborhood inside the city center (whatever the housing tenure status) rather than purchasing a home right away in peripheral areas. This was especially the case for women in charge of child care. They valued proximity to their former neighborhoods and their relatives (who tended to live close by) because they provided continuous and free services that were central to their everyday lives, as well as affective support (Lambert et al., 2018). Indeed, informal family support plays a substantial role for low-income households (Schwartz, 1990), for whom nonstandard work schedules and short-term contracts are more frequent (Letroublon and Daniel, 2018). The qualitative data also show that, before looking into the private sector, former social housing tenants first applied for a better social dwelling in lower-density buildings and/or in more valued areas within the city (as they considered that social housing provides better protection from eviction and rent increases than the private housing sector does). However, due to the rise in rents and other housing costs in the public sector over the past decade (INSEE, 2017), long waiting lists to access a social dwelling, and new government subsidies for home purchases, these tenants gradually changed their residential strategies and started considering rural and peri-urban areas, even if they had neither professional nor familial connections in such areas.

42 As we followed both men and women within households over a 3-year period, we were also able to analyze gender-specific changes in living conditions and neighborhood perceptions. We observed that most lower-class individuals coming from distant municipalities rapidly perceived major disadvantages that they had not anticipated, even though differences in perception also exist across genders. A few weeks after moving in, low-skilled women showed signs of boredom or even depression (“[W]hen we moved in, the first week, I cried and thought, ‘what did [my husband] take from me?’ It was hard,” said a 32-year-old mother of two who was a nurse and married to a French phone company employee). After a few months in their new housing developments, one-third of the women in our sample had to quit their salaried jobs because they could no longer afford to commute —which was not the case for men, whose wages are traditionally higher. This suggests that physical distance to place of work had been underestimated, as well as the expenses for commuting and childcare, which had to be outsourced and ended up being unsustainable for some households. Indeed, low-skilled women in our sample frequently occupied low-paying jobs with nonstandard working hours and part-time jobs located in the first-ring suburbs of the metropolitan area. These work arrangements weighed heavily on their work–life balance while making it difficult to organize their lives far away from their relatives and former favorable locations in big cities, where the municipalities often provided free or heavily subsidized childcare services and public transport ran extensively and frequently all day from early morning to late at night. By contrast, low-income families from municipalities bordering Cleyzieu-Lamarieu (who were on average older than those from the Lyon metropolitan area) appeared to be satisfied with their new dwellings. Moving to Cleyzieu-Lamarieu had almost no impact on their jobs (mostly blue-collar, local industry jobs) or on their social relationships with friends and family. The new house was thus perceived as a sign of upward social mobility and family achievement. Finally, both partners in middle-class couples moving to Cleyzieu-Lamarieu (including those from distant municipalities) could keep their jobs, as they worked regular office hours during the day and could afford the daily commute. They considered their house to be a first step (springboard) into the homeownership market, and they hoped to soon move to a better-off area closer to the Lyon city center, where their professional and social activities were still based.

43 Both lower-class men and women who moved a long distance from their former neighborhoods appeared to have a negative perception of their residential environment, which they compared to old types of suburbs. They called their new development “flat public housing” (HLM à plat) because they viewed their neighborhood as a horizontal succession of small, low-quality, standardized dwellings similar to the vertical stacks of low-quality apartments piled up in high-rise social housing projects constructed en masse during the postwar period in France. They also criticized the great geographical distance from the city center and the lack of public transportation, exactly in the same way that the inhabitants of the first large suburban housing projects (grands ensembles) did in the 1960s (Fourcaut, 2006).