You are on page 1of 27

Housing Studies

ISSN: 0267-3037 (Print) 1466-1810 (Online) Journal homepage: http://www.tandfonline.com/loi/chos20

Building social mix by building social housing?


An evaluation in the Paris, Lyon and Marseille
Metropolitan Areas

Emre Korsu

To cite this article: Emre Korsu (2015): Building social mix by building social housing? An
evaluation in the Paris, Lyon and Marseille Metropolitan Areas, Housing Studies, DOI:
10.1080/02673037.2015.1114075

To link to this article: http://dx.doi.org/10.1080/02673037.2015.1114075

Published online: 24 Nov 2015.

Submit your article to this journal

Article views: 41

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=chos20

Download by: [Laurentian University] Date: 16 March 2016, At: 09:45


Housing Studies, 2015
http://dx.doi.org/10.1080/02673037.2015.1114075

Building social mix by building social housing? An evaluation


in the Paris, Lyon and Marseille Metropolitan Areas
Emre Korsu
Laboratoire Ville Mobilité Transports (LVMT)/Research Unit City Mobility Transport, University
Paris-Est Marne-la-Vallée, Ecole des Ponts ParisTech – IFSTTAR – UPEMLV, Ecole des Ponts ParisTech (UMR
LVMT), Marne-la-Vallée, France
Downloaded by [Laurentian University] at 09:45 16 March 2016

ABSTRACT ARTICLE HISTORY


In France, social housing is perceived as an instrument for promoting Received 11 September 2014
social mix. In particular, there is an expectation that introducing social Accepted 8 October 2015
housing into wealthy areas will bring in low-income households KEYWORDS
and lead to greater coexistence between lower and higher socio- Social mix; segregation;
economic groups. However, several factors tend to hinder the pro-mix social housing; housing
effects of social housing: financial constraints that reduce the number policy; neighbourhoods;
of new buildings, especially in high-income neighbourhoods; Not in Paris; Lyon; Marseille; France
my backyard attitudes in wealthy areas; allocation practices by social
landlords who seldom rent dwellings in expensive neighbourhoods
to poor households. Previous experiments with social housing have
often proved disappointing in their impact on social mix. What about
today’s experiments? Has the social housing built in recent times
increased social mix? The empirical evaluation we carried out in Paris,
Lyon and Marseille shows that recent social housing developments
have stimulated social mix but the impact measured is very small.

Introduction
The world’s cities are changing fast but some things do not change. Social segregation is
still alive and well in most urban settlements and continues to rank high among the issues
that urban policy seeks to tackle. The general feeling is that the situation has become worse
and that cities are even more segregated today than a few decades ago. Urban governments
in most cities, both in France and in other countries, are keen to challenge this trend and
frequently advocate social mix as an ideal to be pursued (Andersen, 2002; Arthurson, 2002;
Bolt et al., 2010; Busch-Geertsema, 2007; Galster, 2013; Goodchild & Cole, 2001; Musterd
& Andersson, 2005; Musterd et al., 2003; Norris & Shiels, 2007).
The scientific literature specifies three broad strategies that can be deployed in social mix
policies: dilution (moving wealthier people into poorer areas), dispersal (moving poorer
people into wealthier areas) or diversity (ensuring a mix of different income groups in new
developments) (Busch-Geertsema, 2007; Melis et al., 2013). Most of the tools that public
authorities can employ to implement these strategies are linked to public intervention in the

CONTACT  Emre Korsu  emre.korsu@enpc.fr 


© 2015 Taylor & Francis
2    E. Korsu

housing market. On the supply side, the construction of public/social housing financed and
managed by the authorities is potentially an effective way to increase social mix in urban
areas. So is the implementation of programmes designed to stimulate private and/or non-
profit developers to build affordable housing for low-income households outside poverty
areas or to create mixed-income housing developments in poor areas, through subsidies or
tax incentives (such as Low-Income Housing Tax Credits in the United States – Anderson
et al., 2003; Wallace, 1995; Williamson, 2011). On the demand side, direct financial support
for low-income households (such as rent vouchers and allowances for tenants) also has the
potential to influence social mix in a positive direction, as it increases the ability of these
households to move to fairly wealthy areas.
Most nations in the western world have tried out these strategies over the past decades.
Nevertheless, the weight given to the different strategic options, the specific tools used and
the way the policies have changed over time, vary across countries. For instance, the policy
of providing direct financial aid to help low-income households to move out of deprived
Downloaded by [Laurentian University] at 09:45 16 March 2016

neighbourhoods are currently more widespread in the United States than in European
countries. In Europe, social mix policies in the United Kingdom and the Netherlands have
been markedly orientated towards increasing the diversity of housing provision in poverty
areas, through regeneration programmes that include the demolition and/or sale of public
housing and the construction of housing for owner-occupation. Policies that seek to increase
social mix through more mixed tenure can also be found in Scandinavian countries such as
Sweden or Finland, but they do not so much specifically target poor areas as all urban areas.
Social mix policies in France have much in common with those of other European coun-
tries, but the French context is also distinctive, with strong emphasis still placed on the role
of social housing. The main purpose of social housing is to provide affordable homes to
low-income groups, but in France it is also recognised in law that social housing has a role
to play in anti-segregation policy (Driant & Lelèvrier, 2006; Tissot, 2005).
While the United States stopped investing in public/social housing in the mid-1960s,
France and many European countries continued to build, although at a slower pace than in
the immediate aftermath of the Second World War. When located in middle- or upper-class
areas, social housing gives low-income households the opportunity to live in wealthier areas
than they can afford on the free market, paving the way for more social mix in such neigh-
bourhoods. Theoretically, social housing can also reinforce social mix in low-income areas
by attracting middle-class residents into these areas. In France as in many other European
countries, social housing is not exclusively restricted to poor households, and most French
middle-class families are eligible for it. Good quality social housing can therefore be a way
to keep middle-income households in low-income areas or to attract new middle-income
residents.
That being said, in practice, there are adverse factors that tend to counteract the poten-
tially pro-mix impact of new social housing developments: budget constraints that make
social housing hard to build, especially in wealthy areas; strong political resistance by local
government in areas with primarily middle/upper-class populations; the view of social land-
lords that they have legitimate reasons for not allocating valued social housing in middle/
upper-class neighbourhoods to low-income households; etc.
In the past, paradoxically, social housing fact often turned out to work against social
mix. Many social housing complexes built following the Second World War in Europe
or in the US quickly became areas of concentrated poverty and actually bred segregation
Housing Studies   3

(Butler & Hamnett, 2011; Dunleavy, 1981; Hirsch, 2000; Madoré, 2004; Massey & Denton,
1993). What, then, about the new social housing that public authorities produce today? Is
recently built social housing increasing the social mix in France’s urban neighbourhoods?
The objective of the research presented here is to provide empirical evidence on this issue.
We tried to answer the question by analysing the geography and occupation of social housing
units built between 1999 and 2008 in France’s three biggest cities, Paris, Lyon and Marseille.
Although our research focuses on social housing in the specific French context, the find-
ings may provide insight into policies conducted in all countries that continue to build social
housing and also seek to increase social mix, even where new social housing developments
are not explicitly intended for that purpose. This includes countries like the Netherlands,
the United Kingdom, Denmark, Sweden, Finland, Germany or Spain—in these countries,
social housing still accounted annually for 10–20 per cent of new housing in the late 2000s,
as it did in France (Pittini & Laino, 2011). The impact of new social housing development
on social mix is a relevant question in those countries too, so the French experience may
Downloaded by [Laurentian University] at 09:45 16 March 2016

be of some interest for public actors there. Many of the obstacles outlined above are also
likely to have an adverse effect on the effectiveness of social housing as a means of increasing
social mix in these countries.1
More generally, the basic question guiding our research may be extended to many other
kinds of public intervention. The United States, for instance, has policies that seek to develop
affordable housing in wealthy suburbs, but it is questionable whether these attempts are
successful in increasing social mix, given that they encounter ‘Not in my backyard’ (NIMBY)
attitudes (Scally, 2014) and are combated through exclusionary zoning practices. Similarly,
the impact of mixed tenure developments in neighbourhoods with deprived public housing
estates, an approach tried in many western countries, is open to similar questions. Whether
or not public intervention in the housing market has been successful in increasing social
mix in recent years is therefore a question that can be asked about a range of policies
implemented in different countries. We believe that the broad methodology designed for
our research could be used, with appropriate adjustments, for future evaluations of policy
in any given context.
The paper opens with a brief review of the social housing sector in France and a more
in-depth discussion of the reasons why the impact of recent social housing developments
has been small, or non-existent, in generating social mix. The second section describes the
methodology and the data used. The subsequent three sections summarise our main results.
The last section contains our concluding remarks.

Building social mix by building social housing: a steeplechase


France has a long tradition of social housing. Nationwide, 4.2 millions social housing units,
17 per cent of the whole housing stock, are rented to households. With 70 social housing
units per 1000 inhabitants, France occupies a median position in the European rankings,
behind countries such as the Netherlands (155 units per 1000) but ahead of countries like
Germany or Italy (15 units per 1000 in the latter case) (Lévy-Vroelant & Tutin, 2007).
The 1990s were a turning point in France with regard to the connection between social
housing and social mix (Tissot, 2005). In response to urban riots that broke out in the
suburbs of Lyon and Paris in 1990–1991, the socialist government adopted the Urban
Development Act (Loi d’Orientation sur la Ville) which introduced the principle of a ‘right
4    E. Korsu

to the city’ and set as a primary goal the balanced development of private and social housing
in city districts. From that point onwards, social housing was officially expected to play a
key role in the promotion of social mix in urban areas. The French left put the issue of social
mix centre stage once again with the Solidarity and Urban Renewal Act (Loi Solidarité et
Renouvellement Urbain) adopted in 2000. The flagship measure introduced in this period
was the 20 per cent threshold for social housing (recently increased to 25 per cent) appli-
cable to all urban municipalities above a set population threshold. The law required such
municipalities to develop a social housing stock that accounts for at least 20 per cent (now
25 per cent) of their total housing stock. Municipalities that fall below this threshold are
fined if they fail to make sufficient efforts to catch up. Many of these municipalities are
wealthy areas, inhabited mostly by middle- and high-income households, so increasing
the number of social housing units will inevitably have a positive impact on social mix.
While the intention of new social housing is to contribute positively to social mix in
French cities, there are many forces that run counter to the desired pro-mix effect. What
Downloaded by [Laurentian University] at 09:45 16 March 2016

follows is a discussion of those forces.

Diminishing public investment in social housing


France’s substantial social housing stock is mainly a legacy of the past. In recent decades, the
scale of new construction has fallen sharply, in France as in other countries, as a result of
reduced public investment (Whitehead & Scanlon, 2007). In the French case, annual output
decreased from 100 000 new units per year during the 1960s–1970s to fewer than 50 000
per year in the 2000s (Lévy-Vroelant & Tutin, 2007). This is a fundamental factor that is
likely to limit the impact of new social housing on social mix: the fewer the constructions,
the smaller the impact.

Economic and political barriers outside poor areas


New social housing is more difficult to produce outside low-income areas because of higher
land prices. The lowest prices are generally found in deprived neighbourhoods where the
poorest households live, while prices elsewhere are higher and rise steadily with proximity
to the most coveted areas, occupied by the highest income groups. These high prices dis-
courage builders, as they inflate construction costs and this effect is greatest in the wealthiest
neighbourhoods.
Politics add to the invisible walls built by land economics. In many municipalities where
the population profile is in the middle or upper socio-economic category, local governments
are unwilling to increase the supply of social housing. Most residents of these areas are
happy with the prevailing social homogeneity and hostile to any social housing which, in
their opinion, would ruin this social ‘harmony’ by introducing ‘unsuitable’ neighbours. In
these circumstances, municipal authorities which dared to go against the NIMBY social
consensus would damage their re-election chances (Madoré, 2004).
Resistance by local authorities, which have powerful allies at the heart of national polit-
ico-administrative structures (many mayors, especially in big cities, are also members of
parliament or senators), is probably the main reason why the legal provisions which estab-
lished the 20  per cent threshold have so far failed to make much impact. The original
financial penalties were quickly revised downward and fairly low thresholds were adopted.
Housing Studies   5

In consequence, many municipalities did not make the investments needed to reach the
20 per cent threshold and opted to pay the fines. In 2004, 30 per cent of municipalities in the
Paris Metropolitan Area (the Île-de-France region) required by the legislation to invest in
social housing (55 municipalities out of 185), had built none during the period 2001–2004.
A total of 151 municipalities (82 per cent) failed to reach the three-year construction target,
taking into account only housing units started and completed during the period (Bilek
et al., 2007). The recalcitrant municipalities included iconic localities like Neuilly-sur-Seine
or Le Raincy, two bastions of the inner suburban Paris bourgeoisie.

Social housing in middle- or upper-class areas is not primarily for the poor
Building social housing outside low-income areas increases the social mix only if these
homes are allocated to households from the lower social strata. This condition is often not
met. Several factors are at play here.
Downloaded by [Laurentian University] at 09:45 16 March 2016

In France, the income ceilings that regulate access to social housing are fairly high, and
consequently a significant proportion of households are eligible. There are three ranges
of social housing built in France. The intermediate range, with ‘standard’ social housing,
accounts for the largest proportion of the stock, at around 80 per cent. Seventy-one per
cent of the population is eligible to live in these dwellings, which means that 71 per cent of
the population earn less than the threshold above which access is denied to standard social
housing. The higher range, which accounts for 10 per cent of the stock, is clearly intended
for middle-income households. The lower range, so-called ‘very social housing’, built to
provide priority accommodation for the poorest households, represents only 10 per cent
of the total social housing stock, and even then one-third of the population is eligible for
such housing (Lévy-Vroelant & Tutin, 2007).
When they build social housing, local authorities in municipalities with a population
profile in the middle or upper socio-economic categories tend to develop intermediate or
higher range products and are less likely to invest in the lower range. By way of example,
we can cite the case of the upper-class municipalities in the Hauts-de-Seine, the wealthiest
département in the Paris Metropolitan Area after Paris. According to data provided by
the AORIF (the public organisation for social landlords and related entities in the Paris
MA), the municipalities of Saint-Cloud, Levallois-Perret, Asnières-sur-Seine, Neuilly-sur-
Seine, Meudon, Vaucresson, Sceaux, Garches, Chaville, Courbevoie—all of which rank
among the top 20 per cent municipalities of the Paris MA in terms of mean per capita
income—altogether built 1150 new social housing units in the 2007–2008 period, but only
15 per cent of these were lower-range products intended as accommodation for poor house-
holds (social housing financed through PLA-I loans), as compared with 28 per cent for the
Paris MA as a whole.2 In some cases, the purpose of this strategy is to circumvent the legal
constraints of the 20 per cent threshold: municipalities build social housing, but they avoid
the ‘very social’ products much more likely to be occupied by poor households. With higher
standard products, the new dwellings can be allocated to middle-class households. This
situation puts low-income households in competition with the middle-classes for social
housing developed in middle/upper-class areas.
In this competition, the odds are against working-class households. The allocation sys-
tem generally works against them.3 The social landlords (‘bailleurs sociaux’) who build and
administer the social housing stock, see the ‘occupancy’ of their dwellings as crucial. They
6    E. Korsu

fear tenants who are ‘bad payers’, because they lead to unpaid rents. Conflicts among ten-
ants adversely affect the relations between landlord and households and cause significant
management costs. They also generate financial costs when they prompt some tenants to
leave, resulting in vacancies and rent shortfalls. Vacancy can become chronic in buildings or
housing estates that develop a bad reputation. In fact, demand for many stigmatised social
housing estates is almost non-existent and these landlords experience the most difficulty
in finding solvent tenants. These issues prompt social landlords to develop ‘occupancy’
strategies in order to minimise the risks. The main strategy is systematically to rule out all
‘risky’ applications for social dwellings in high value buildings and neighbourhoods and to
steer them towards the low value portion of the stock. This protects the positive image of
the high value areas and ensures high, creditworthy, low-risk demand, at the cost of ‘sac-
rificing’ the most rundown, low-value areas, where the most ‘problematic’ households are
accommodated (Epstein & Kirszbaum, 2003; Simon, 2003).4 The degree of risk associated
with an application is often assessed on the basis of assumptions about ethnic identity or
Downloaded by [Laurentian University] at 09:45 16 March 2016

family characteristics (single-parent families, families with multiple children). An applicant


with particular characteristics is considered to be a threat to the ‘social balance’ in a given
building because past experience shows that this ‘type’ of tenant housed in this ‘type’ of
environment generates conflicts (Simon (2003) calls this ‘probabilistic discrimination’). In
most cases, such practices are driven by positive intentions such as the desire to maintain
a ‘balanced’, ‘peaceful’ community, but they are nevertheless discriminatory and based on
multiple prejudices (Epstein & Kirszbaum, 2003; Sala Pala, 2005; Simon, 2003).5 They are
also strictly speaking illegal, since the criteria they rely on are not recognised by the law
(Tissot, 2005).
This mechanism is likely to be reinforced by self-censorship among applicants who, antic-
ipating the unlikelihood of success, do not even dare to apply for homes in wealthy areas. In
addition, there may also be a ‘conservative’ tendency that prompts low-income applicants
to prefer to live in familiar neighbourhoods, in other words usually low status areas.

New social housing in lower-class areas: to rent or not to rent to the middle classes
While economic and political obstacles restrict the development of social housing in wealthy
areas, other factors contribute to construction in poor areas. For a number of years, the
French State has been undertaking large-scale regeneration programmes in old social
housing estates, most of them in France’s poorest areas. The programmes, financed by the
National Urban Renewal Agency (Agence Nationale de la Renovation Urbaine), often involve
the demolition of the most rundown buildings. The essential objective of the programmes is
to replace each unit demolished by a new one in the same place. As the objective is also to
increase the tenure mix, not all the rebuilt residential units are social housing. Nonetheless,
the renewal programmes result in the large-scale development of new social housing in
low-income areas. Many of these homes are allocated to low-income households from the
demolished buildings, with the result that they do little to improve the social mix.
Social housing developments in poor areas might possibly contribute positively to social
mix if they could attract new middle-class households to these neighbourhoods. However,
this possibility too is hampered by adverse factors. There is clearly strong middle-class
demand for social housing, especially in big cities, but this demand is mainly for social
housing outside poor areas, with the exception (as in Paris) of a few fairly low-income
Housing Studies   7

Table 1. Location of new social housing (NSH) units and socio-professional characteristics of heads of
households living in these dwellings in 2008 (per cent).
Paris Lyon Marseille
All All
NSH All housing NSH housing NSH housing
Socio-professional status of the head of the household
Farmer 0.0 0.1 0.0 0.4 0.0 0.2
Cratsman, tradesman or small business owner 2.7 4.2 3.1 4.7 3.2 4.5
Senior manager or superior intellectual profession 11.6 21.4 5.3 15.5 4.4 11.9
Middle manager or equivalent 23.0 17.4 19.8 17.5 18.8 15.3
Office worker with a high educational achievement 9.2 4.7 6.8 4.2 7.2 4.3
Office worker with a low educational achievement 19.6 9.4 18.0 7.7 20.6 9.4
Blue-collar worker 21.2 12.7 28.3 14.7 24.0 12.8
Retired person 7.9 24.6 11.9 27.9 14.0 31.5
Other inactive person 4.9 5.6 6.8 7.5 7.8 10.1
All 100.0 100.0 100.0 100.0 100.0 100.0
Downloaded by [Laurentian University] at 09:45 16 March 2016

Location
Upper-class neighbourhoodsa 15.7 22.7 15.8 24.1 21.7 21.5
Working-class neighbourhoodsb 36.8 45.8 25.9 39.9 21.0 38.8

Number of NSH units 48 248 6 900 4 061  


Note: NSH units are social dwellings in buildings built between 1999 and 2008.
Source: French National Census 2008 – INSEE.
a
The two quintiles of neighbourhoods/municipalities/cantons where senior managers and members of superior
intellectual professions are the most represented and which gather 40 per cent of the households of the metropolitan
area with a head belonging to this socio-professional group.
b
The quintile of neighbourhoods/municipalities/cantons where blue-collar workers are the most represented and which
gather 20 per cent of the households of the metropolitan area with a head belonging to this socio-professional group.

areas close to the city centre. Negative opinions about the quality of dwellings and the
social environment in poor neighbourhoods tend to reduce middle-class demand for social
housing in these areas.
Moreover, the idea of allocating social housing in low-income areas to middle-class
households raises a dilemma. It potentially conflicts with the goal of providing affordable
housing for households experiencing economic and social hardship. Since the 1990s, the
pressure on social housing has intensified, as the number of households unable to find
an adequate home on the private housing market has grown dramatically. As previously
mentioned, however, social landlords tend to reserve a substantial share of social dwellings
outside poor neighbourhoods for middle-class households. If the same were to occur in
poor areas as well, there would not be enough space for applicants from lower social strata
undergoing economic and social hardship. This is an option that would seem particularly
difficult to justify under current conditions.

Methodology
To determine whether the newly built social housing is increasing the social mix or not, we
carried out an empirical evaluation based on a methodology specifically conceived for this
purpose. In this section, we outline the key components of this methodology.
Our approach is based on data from the 2008 census produced by the National Bureau
of Statistics – Institut National de la Statistique et des Études Économiques (INSEE). The
so-called ‘Detail Files’ (Fichier Détail Individus) contain information on a very large sample
8    E. Korsu

of housing units in France (social/non-social, location, characteristics of occupants) and


on the completion date of buildings. Using this last piece of data, we can define ‘recently
built social housing units’ as social housing in buildings completed between 1999 and 2008.
This set is not exclusively confined to new social housing produced during this period.
It also includes units in older buildings, not originally part of the social housing stock,
which have been acquired by public authorities, possibly renovated and converted to social
housing. This kind of new social housing unit could not be specifically identified in the
data-set we used, so our evaluation does not cover them. They seem to represent a significant
number of dwellings, especially in the inner cities.6 Overall, however, the large majority of
additional social housing in this period was new construction; in the Paris MA, for instance,
according to data from the Observatoire du Logement Social en Ile-de-France, 82 per cent
of the new social housing units put into use in 2012 were in new constructions completed
during that year.
The evaluation relates to the Metropolitan Areas (‘aires urbaines’, as defined by INSEE) in
Downloaded by [Laurentian University] at 09:45 16 March 2016

the three major French cities studied. The Paris Metropolitan Area (MA) had a population
of 12.1 million in 2008, the Lyon MA 2.1 million and the Marseille MA 1.7 million. There
were 48 000 recently built social housing units in the Paris MA, 6900 in the Lyon MA and
4000 in the Marseille MA (cf. Table 1 for descriptive statistics), in each case representing
less than 1  per cent of the total number of occupied dwellings (‘résidences principales’)
(respectively, 0.98, 0.94 and 0.59 per cent).
The location of housing units is provided at neighbourhood scale in the case of densely
populated municipalities, at the scale of municipalities (‘communes’) or ‘cantons’ (admin-
istrative groupings of several municipalities) otherwise.
‘Social mix’ is a complex notion which is not easy to define statistically. It refers to a
situation where people of different backgrounds and life histories, with different social,
economic, cultural and demographic identities, live in close contact with each other and
share the same urban spaces, ideally in the hope that this spatial proximity will lead to
intense and peaceful inter-group relations. Social mix is very much a political concept, since
in many countries the socially mixed city is valued symbolically/normatively as a model
associated with shared values such as inclusion, justice, fraternity, solidarity, tolerance, etc.
and is normatively assumed to be ‘healthier’ and more ‘liveable’ than a city where visible
and/or invisible boundaries separate people by race, income or social status (Bridge et al.,
2012; Rose, 2004). It is perceived as a counter-model to the segregated city and idealised
as a city without ghettos. From a scientific perspective, things are more complicated. Many
social scientists consider that the normative rationale behind the idealisation of the social
mix cannot be taken for granted. Theoretically, socio-spatial proximity is not a guarantee
of a united, fair, friendly, tolerant urban community. Injustice, inequality and exclusion
can exist in a city where dissimilar people live close together. Different social, cultural and
religious groups can coexist in the same territory without any inter-group interaction,
ignoring each other, building no social bonds. Worse, such forms of coexistence can be
conflictual and may fuel hostility and intolerance. The challenges are also substantial on the
empirical side. Degrees of cross-group social interaction, mutual trust, friendship, tolerance
and social cohesion are hard to measure, especially at the scale of an entire city. This is the
essential reason why most quantitative research on social mix has so far failed to do more
than simply measure the spatial proximity between groups, skipping the question of the
social meanings of the observed proximity. But even this elementary task raises several
Housing Studies   9

methodological complications: the degree of spatial proximity between groups depends on


the spatial scale at which it is measured and it is not easy to determine which scale is best;
the construction of the social groups for comparison is problematic; there are multiple ways
to measure spatial proximity between groups and hence competing statistical indices, etc.
Our approach to ‘social mix’ follows the standard methodology used in most previous
quantitative research on this issue and we are consequently exposed to the same meth-
odological limitations as those. We define ‘social mix’ in the weak sense, simply as the
cohabitation of people from different social groups in different areas of a city. The degree of
cohabitation between groups is measured by two statistical indices, calculated at a scale of
mix (neighbourhood/municipality/canton) which depends on the scale used in identifying
the location of housing units. Both of these indices figure among those most commonly
used in studies on social mix/segregation:
• A dissimilarity index, measuring the symmetry/asymmetry in the residential location
Downloaded by [Laurentian University] at 09:45 16 March 2016

patterns of two different social groups. The value of this index varies between zero—a
perfect mix (which corresponds to a situation where the geography of residential loca-
tion of a group j is perfectly identical to that of a group k)—and one, total separation,
meaning that all members of a group j live in areas where there are no members of
group k, and vice versa;
• An exposure index, corresponding to the mean proportion of households from group
k in the ‘vicinity’ (at neighbourhood, municipality or ‘canton’ scale) of households
from group j.
The evaluation is carried out in two ways: (i) an aggregated approach that measures the
overall impact of new social housing on social mix; (ii) a disaggregated approach that looks
at individual household units and measures the marginal impact of each of them.
The social groups are mainly defined by socio-professional status. We measure the level
of social mix between five socio-professional categories: senior managers and higher intel-
lectual professions (subsequently referred to as senior managers); middle managers and
equivalent (subsequently referred to as middle managers); office workers with high edu-
cational attainment (HEA)—i.e. with a college degree; office workers with low educational
attainment (LEA)—i.e. without a college degree; blue-collar workers. Each household is
classified according to the socio-professional status of the head of the household. The ‘office
workers’ category was split into two on the basis of educational attainment in order to sep-
arate those clearly belonging to the middle classes (in terms of wage level, cultural capital,
lifestyle, job security) from those culturally and socio-economically closer to the working
classes.

Measuring global impact


To measure the overall impact of new social housing, we compared the situation observed
in 2008 with a hypothetical situation (H0), corresponding to the situation that would be
observed in 2008 under a twin hypothesis: (i) no social housing is built after 1999; (ii) house-
holds that in reality occupy the social housing built after 1999 are instead distributed around
the city in exactly the same pattern as the other households present in 1999, according to
socio-professional status. The strategy is to compare the level of social mix observed in 2008
following the input of new social housing units built after 1999, with the hypothetical level
10    E. Korsu

that would be observed if those new social housing units did not exist and if the house-
holds occupying them were instead housed according to the same distribution as other
households. If the observed level of social mix is higher than the hypothetical level, then
it may be considered that the global effect of new social housing on social mix is positive,
since social mix is greater with these dwellings than without. As to the difference between
the two levels of social mix, this provides information on the magnitude of the total effect.
This approach is applied with each of the two indices. To calculate the hypothetical value
under H0, we simulated a virtual situation by redistributing households from each group j
living in the new social housing in such a way that their spatial distribution was identical
to the pattern observed for the rest of the households from group j. In this virtual state,
the total number of households from group j living in each neighbourhood i is calculated
using the following formula:

HHji = HHjio + Σi HHjis ∗ (HHjio ∕Σi HHjio )


Downloaded by [Laurentian University] at 09:45 16 March 2016

where HHji is the number of households from group j living in neighbourhood i. HHjio is
the number of households from group j living in neighbourhood i and occupying non-new
social housing. HHjis is the number of households from group j living in neighbourhood i
and occupying new social housing.
The H0 hypothesis indirectly makes some assumptions about the way the urban system
would have ‘reacted’ if the social sector built no new housing in the period. In truth, it is
impossible to know what precisely would have happened if no social housing were built
between 1999 and 2008. There are multiple possibilities which could have combined in
several ways: a number of new households would not have been formed, another set of
households would not have moved, other households would have moved out of the Paris/
Lyon/Marseille MA, the private sector would have built more housing, the vacancy rate
in the housing market would have fallen, etc. The basic assumptions of the H0 hypothesis
are twofold. It assumes first that the trend in the number of households in the three MAs
would not have been altered if no social housing were constructed between 1999 and 2008.
Second, it assumes that the housing market, no matter how, would have reacted in such a
way that the spatial distribution of social groups in 2008 would not be different from the
distribution actually observed in 2008 for households living in dwellings other than the new
social units constructed between 1999 and 2008. These assumptions are naturally open to
question and we discuss this issue further in the final section of the paper.
As mentioned above, the way in which new social housing units affect social mix
depends on the social identity of tenants. But this population is not fixed once and for all.
It evolves, as some households leave and are replaced by new ones. This implies that the
impact on the social mix produced by the same set of dwellings is potentially variable over
time. The method we designed measures this impact at a time t given the social charac-
teristics of tenants at that moment. At another moment, the impact measured might have
been different. In theory, what we face here is a methodological problem, but we believe
that its practical consequences for our findings should be limited for the following reason.
The aim of our research is to evaluate the impact of new social housing on social mix here
and now. Are these dwellings promoting a social mix now? How things will turn out in
the long run is a question beyond the scope of our research. In the short run, the impact
of the new social dwellings on social mix is likely to be fairly steady, because residential
Housing Studies   11

mobility in the social housing sector is slow—in 2006, tenants remained in the same
dwelling for an average of 11 years. This implies that any measurement at a given time t
has a high probability of being a reliable reflection of the short-term impact of new social
housing units on social mix.

Measuring the disaggregated micro-impact of each housing unit


A complementary evaluation is carried out in terms of the micro-impacts—positive or
negative—produced by social dwellings individually. The basic idea is as follows. When a
given social housing unit, with its specific combination of occupancy and location, produces
a marginal effect on the value of the index in the direction of a greater social mix, then it
can be considered that this dwelling contributes positively to social mix.
The value of the dissimilarity index shifts downward (meaning greater social mix) when
the spatial distributions of the two groups being compared tend to converge. Now, consider
Downloaded by [Laurentian University] at 09:45 16 March 2016

a social dwelling occupied by a group j household in a given area. If the location of this
household has a convergent effect on the respective distributions of groups j and k, then
one can consider that this social dwelling has a positive impact on the social mix between
groups j and k. This can be translated statistically as follows. We consider that a social
housing unit located in a neighbourhood i and occupied by a group j household produces
a positive micro-impact on the social mix between groups j and k if we have:

HHjio ∕Σi HHjio < HHki ∕Σi HHki

In this case, an additional group j household in neighbourhood i increases the proportion


of group j households living in neighbourhood i, leading to a decrease in the value of the
dissimilarity index j/k.
In the case of the exposure index, we assume that the micro-impact of a given social
dwelling located in a neighbourhood i, occupied by a group j household, on the social mix
between groups j and k, depends on the proportion of group k households in the neigh-
bourhood i. If this proportion is higher than the average value for group j households,
we consider that this dwelling positively affects the cohabitation between groups j and k,
since it brings a group j household into an area where group k is better represented than
the average for group j households. An example may help to clarify this point. Suppose
that in a given city, low-income households live in areas where on average 15 per cent of
their neighbours are high-income households. A new social housing unit is allocated to
a low-income household in an area where the share of high-income households is 30 per
cent. We consider that this dwelling produces a positive impact on the social mix because
it allows a low-income household to live in an area where low-income households have
more high-income neighbours than the average for low-income households. Statistically,
the micro-impact of a social dwelling located in neighbourhood i and occupied by a group
j household on the social mix between groups j and k is positive if we have:

HHki ∕Σk HHki > Σi ((HHki ∕Σk HHki ) ∗ HHjio )∕Σi HHjio

At the end of the procedure, we can calculate the proportion of housing units that produce
a positive/negative marginal effect.
12    E. Korsu

Evaluating impact by reference to a potential


The new social housing built in the early 2000s represents only a small portion of the total
housing stock. This fact may fundamentally limit its global impact on the social mix. To
take the example of the Paris MA, 50 000 new social units cannot be expected to bring
about a profound change in the level of segregation prevailing in a huge urban area with
five million dwellings.
A way to incorporate this point into the evaluation is by reference to the maximum
impact that this much social housing could potentially have. This means comparing the
observed impact on social mix with a potential maximum: how much impact did the new
social housing produce on social mix and how much could have been produced at best?
The difference between the two is an alternative way to measure the efficiency of new social
housing in promoting social mix.
The potential maximum can be calculated by simulating different hypothetical scenarios
Downloaded by [Laurentian University] at 09:45 16 March 2016

representing virtual situations where new social dwellings are located and allocated in such
a way as to maximise their impact on the social mix. For the sake of the research, we ruled
out overly unrealistic scenarios. For instance, the impact on the social mix between senior
managers and blue-collar workers would probably have been highest if all new social hous-
ing were built in high-income areas and if all of those units were allocated to blue-collar
families. We preferred to use more realistic scenarios.
We designed three virtual scenarios, one in which the location of new social housing
units is modified (Max1), another where it is the allocation of new dwellings that is modi-
fied (Max2), and a third where both parameters are modified simultaneously (Max3). The
aim is to measure the potential reached in situations where either the location parameter
or the allocation parameter is modified and also to introduce a gradation in the test sce-
narios (Max3 is a more demanding scenario than the other two, since both parameters are
adjusted).
The first scenario we designed (Max1) simulates a hypothetical situation where the new
social housing occupied by group j households is redistributed spatially in the same pattern
as all households from group k. To give an example, we simulated a situation where all
new social housing allocated to blue-collar households is relocated in such a way that its
spatial distribution becomes the same as the spatial distribution of all households headed
by a senior manager. This scenario stages a virtual situation where the number of social
dwellings occupied by blue-collar households in an area is proportional to the number of
local households where the head is a senior manager. The aim is to measure the additional
social mix between groups j and k that could potentially have been obtained through better
location of dwellings occupied by group j households.
A second scenario (Max2) simulates a situation where the new social housing is allocated
almost exclusively to households from low-income socio-professional groups. There are
three versions of this scenario. In the first (Max2a), all new social housing is allocated to
households where the head is a blue-collar worker, except dwellings occupied by house-
holds where the head is a retired person or a non-retired person not in work. The addi-
tional blue-collar households placed in the new social housing units are ‘picked up’ from
different areas of the city in proportion to the number of blue-collar households living in
each area. The non-blue-collar households that are ‘taken out’ of the new social housing
are ‘relocated’ in such a way that their spatial distribution is identical to the distribution
Housing Studies   13

observed for other households in the same socio-professional group. In the second version
of this scenario (Max2b), the same procedure is applied to households where the head is an
office worker with LEA (all new social housing occupied by households where the head is
not a retired person, a non-retired person not in work, or a blue-collar worker, is allocated
to households where the head is a LEA office worker), and then to households headed by a
HEA office worker (Max2c). This second scenario attempts to measure the point that would
have been reached if the quantity of new social housing reserved for households from the
lower social strata were ‘maximised’.
The third scenario (Max3a, Max3b and Max3c) is an aggregate of the two previous sce-
narios. Both the allocation and location of new social housing are modified simultaneously.
All new social dwellings are allocated to households from the lower social strata (following
the procedure described in scenario Max2) and are then relocated in the city with a spatial
pattern identical to that observed for group k households.
On the basis of these scenarios, we can measure the level of ‘achieved potential’, i.e. the
Downloaded by [Laurentian University] at 09:45 16 March 2016

point reached between the minimum (hypothetical situation with no social housing built
between 1999 and 2008) and the maximum (according to the Max scenarios).
%achieved potential = (OBS − MIN)∕(MAX − MIN) ∗ 100
where OBS is the observed value for the index used. MIN is the value calculated in H0. MAX
is the value calculated in scenarios Max1/2/3.

Global impact on social mix: positive but small


Overall, the impact on social mix of the social housing units built between 1999 and 2008 in
the Metropolitan Areas of Paris (48 000 units), Lyon (6900 units) and Marseille (4000 units)
was positive but quite modest. They increased cohabitation between all socio-professional
groups but in marginal proportions. The values of the 20 dissimilarity indices corresponding
to the real situation in 2008 in each of the three MAs are all smaller than the values calculated
under the H0 hypothesis—except in one case in the Lyon MA—which suggests a systematic
pro-mix impact. In every case, however, the difference is less than 1.5 per cent (Table 2).
The results obtained with the exposure index confirm these findings (Table 3). It shows
that in each MA, for every socio-professional group, the new social housing reduced with-
in-group cohabitation and increased the cohabitation with other groups, including those
most distant in the social space. Again, however, the impact is very small. To give an exam-
ple, in the Paris MA, blue-collar households live in areas where, on average, 18 per cent of
households belong to this same socio-professional category. This is 0.25 per cent less than
would have been observed had no social housing been built between 1999 and 2008 (H0).
Similarly, households headed by a senior manager or equivalent account for an average of
15 per cent of neighbours of blue-collar workers, which is 0.25 per cent more than under
the H0 hypothesis.

Room for a greater social mix


Could more have been achieved? As described above, we attempted to answer this question
by exploring hypothetical scenarios designed to estimate the maximum that could have
been done with the new social housing built in the 2000s.
14    E. Korsu

Table 2. Observed level of social mix in 2008 and hypothetical level under the H0 hypothesis—dissim-
ilarity index.
Paris Lyon Marseille
OBS Delta obs/H0 b
OBS Delta obs/H0 OBS Delta obs/H0
2008a (%) 2008 (%) 2008 (%)
B-CW vs.
SM&SIP 0.486 −0.27 0.398 −0.38 0.422 −0.35
Middle managers 0.271 −0.65 0.254 −1.06 0.267 −0.38
HEA office workers 0.329 −0.67 0.371 −0.54 0.311 −0.05
LEA office workers 0.176 −0.94 0.183 −1.25 0.169 −0.01
       
LEA office workers vs.
SM&SIP 0.419 −0.47 0.339 −0.34 0.388 −0.49
Middle managers 0.223 −1.07 0.203 −0.77 0.237 −0.73
HEA office workers 0.238 −1.08 0.249 −0.78 0.234 −0.38
B-CW 0.176 −0.55 0.183 −0.22 0.169 −1.45
Downloaded by [Laurentian University] at 09:45 16 March 2016

       
HEA office workers vs.
SM&SIP 0.289 +0.03 0.283 −0.04 0.292 −0.20
Middle managers 0.200 −1.06 0.223 −0.75 0.199 −0.51
LEA office workers 0.238 −0.68 0.249 −0.76 0.234 −0.72
B-CW 0.329 −0.29 0.371 −0.40 0.311 −0.71
       
Middle managers vs.
SM&SIP 0.267 −0.06 0.189 +0.48 0.207 −0.06
HEA office workers 0.200 −0.78 0.223 −0.53 0.199 −0.20
LEA office workers 0.223 −0.64 0.203 −0.77 0.237 −0.13
B-CW 0.271 −0.03 0.254 −0.19 0.267 −0.08
       
SM&SIP vs.
Middle managers 0.267 −0.24 0.189 −0.47 0.207 −0.08
HEA office workers 0.289 −0.38 0.283 −0.31 0.292 −0.04
LEA office workers 0.419 −0.16 0.339 −0.27 0.388 −0.05
B-CW 0.486 −0.09 0.398 −0.21 0.422 −0.03
Abbreviations: B-CW – blue-collar workers; SM&SIP—senior managers and superior intellectual professions; HEA office
workers—office workers with high educational achievement (college degree); LEA office workers—office workers with
low educational achievement (no college degree).
Source: French National Census 2008 – INSEE.
a
Observed value of the dissimilarity index in 2008.
b
Difference between the observed value of the dissimilarity index and the hypothetical value calculated under H0
hypothesis.

The most demanding of the scenarios (Max3) simulates a situation where all new social
housing units are allocated to lower social groups (first blue-collar workers’ households,
then LEA office workers, and lastly HEA office workers) and are virtually ‘relocated’ to areas
where households from higher social strata are well represented. The findings reveal that
the global impact on the social mix in this virtual scenario is significantly greater than the
real impact (Table 4). If the Max3a scenario had really happened, the level of social mix
between blue-collar workers (B-CW) and other groups would have risen by 6.7 per cent in
the Paris MA, 5.2 per cent in the Lyon MA and 3.6 per cent in the Marseille MA, compared
with the H0 hypothesis. The increase in the social mix is similar in the case of cohabitation
between LEA office workers and other groups (Max3b) and even greater in the case of
cohabitation between HEA office workers and other groups (Max3c: around 10 per cent in
the Paris MA, 8 per cent in Lyon).
Housing Studies   15

Table 3. Observed level of social mix in 2008 and hypothetical level under the H0 hypothesis—exposi-
tion index.
Paris Lyon Marseille
% OBS Delta % OBS 2008 Delta Obs/H0 % OBS 2008 Delta Obs/H0
2008a Obs/H0b (%) (%) (%)
B-CW – mean social composition of neighbourhoods of residence
SM&SIP 14.9 +0.21 12.1 +0.21 9.3 +0.25
Middle managers 17.4 +0.14 16.7 +0.13 14.4 +0.09
HEA OW 4.7 +0.22 3.8 +0.30 4.0 +0.07
LEA OW 11.6 +0.06 8.7 +0.06 10.5 −0.02
B-CW 17.9 −0.25 19.2 −0.36 16.0 −0.02

LEA OW – mean social composition of neighbourhoods of residence


SM&SIP 16.6 +0.34 13.3 +0.25 9.9 +0.30
Middle managers 17.4 +0.23 17.1 +0.16 14.6 +0.12
HEA OW 5.1 +0.17 4.4 +0.15 4.3 −0.02
Downloaded by [Laurentian University] at 09:45 16 March 2016

LEA OW 12.0 −0.36 9.1 −0.39 11.0 −0.06


B-CW 15.7 −0.08 16.6 −0.17 14.3 +0.04

HEA OW – mean social composition of neighbourhoods of residence


SM&SIP 21.3 +0.03 16.1 −0.13 12.2 +0.03
Middle managers 17.5 +0.19 18.1 +0.06 15.7 +0.04
HEA OW 5.9 −0.24 5.5 −0.24 5.2 −0.05
LEA OW 10.0 +0.11 7.9 +0.14 9.5 +0.07
B-CW 12.5 +0.12 13.3 +0.19 12.0 +0.18

Middle managers – mean social composition of neighbourhoods of residence


SM&SIP 21.1 +0.02 16.1 −0.07 12.9 +0.01
Middle managers 18.5 +0.04 18.5 +0.01 16.6 +0.00
HEA OW 4.8 +0.16 4.4 +0.11 4.4 +0.01
LEA OW 9.4 +0.12 7.5 +0.09 8.9 +0.03
B-CW 12.7 +0.05 14.0 +0.07 12.0 +0.06

SM&SIP – mean social composition of neighbourhoods of residence


SM&SIP 27.9 −0.07 19.3 −0.12 16.2 −0.01
Middle managers 17.1 +0.05 18.2 +0.03 16.5 +0.01
HEA OW 4.7 +0.10 4.4 +0.06 4.3 +0.02
LEA OW 7.3 +0.13 6.6 +0.09 7.8 +0.02
B-CW 8.8 +0.11 11.5 +0.10 10.0 +0.04
Notes: In 2008, blue-collar workers’ households in the Paris MA live in neighbourhoods where 14.9% of households, on
average, are SM&SIP households. This is 0.21% more than the weight calculated under the H0 hypothesis.
Abbreviations: B-CW—blue-collar workers; SM&SIP—senior managers and superior intellectual professions; HEA OW—of-
fice workers with high educational achievement (college degree); LEA OW—office workers with low educational achieve-
ment (no college degree).
Source: French National Census 2008 – INSEE.
a
Observed proportion of the socio-professional group k in neighbourhoods of residence of households of group j in 2008.
b
Difference between the observed proportion of a socio-professional group k and the hypothetical proportion calculated
under H0 hypothesis.

These results can be interpreted in two ways. On the one hand, they suggest that a greater
improvement in social mix could have been achieved, even with the limited number of
social housing units built in the 2000s. On the other hand, they also reveal that—given
the number of social housing units that public authorities are currently able to build over
a decade—a radical increase in the level of social mix cannot be expected, even assuming
very favourable conditions in the location and allocation of new dwellings.
If we refer to the most demanding of the scenarios (Max3), the percentage of the
potential achieved looks modest, less than 25 per cent in most cases (Table 5). In the case of
16    E. Korsu

Table 4. Differences (%) between levels of social mix under Max scenarii and under the H0 hypothesis –
dissimilarity index.
Paris Lyon Marseille
Max1 Max2 Max3 Max1 Max2 Max3 Max1 Max2 Max3
B-CW vs. Max2a Max3a Max2a Max3a Max2a Max3a
SM&SIP −1.6 −2.2 −6.7 −1.8 −1.5 −5.2 −1.1 −1.4 −3.6
Middle managers −1.6 −2.6 −6.7 −1.8 −2.6 −5.2 −1.1 −1.2 −3.6
HEA OW −1.6 −2.4 −6.7 −1.8 −2.4 −5.2 −1.1 +0.1 −3.6
LEA OW −1.6 +1.2 −6.7 −1.8 −2.8 −5.2 −1.1 +1.7 −3.6

LEA OW vs. Max2b Max3b Max2b Max3b Max2b Max3b


SM&SIP −2.0 −1.6 −6.9 −2.2 −0.7 −6.5 −1.3 −1.4 −3.4
Middle managers −2.0 +0.0 −6.9 −2.2 −0.7 −6.5 −1.3 −2.0 −3.4
HEA OW −2.0 −0.8 −6.9 −2.2 −1.3 −6.5 −1.3 +0.2 −3.4
B-CW −2.0 +3.4 −6.9 −2.2 +0.2 −7.4 −1.3 −1.4 −3.0

HEA OW vs. Max2c Max3c Max2c Max3c Max2c Max3c


SM&SIP −1.9 +2.8 −9.6 −1.5 −0.4 −7.8 −1.0 −1.2 −4.6
Downloaded by [Laurentian University] at 09:45 16 March 2016

Middle managers −1.9 +3.6 −9.6 −1.5 −3.1 −7.8 −1.0 −0.5 −4.6
LEA OW −1.9 +2.1 −9.9 −1.5 −3.7 −8.3 −1.0 −1.2 −4.6
B-CW −1.9 +1.1 −10.0 −1.5 −3.6 −8.3 −1.0 −2.6 −4.6

Middle managers vs.


SM&SIP −1.3 nc. nc. −1.1 nc. nc. −0.7 nc. nc.
HEA OW −1.3 nc. nc. −1.1 nc. nc. −0.7 nc. nc.
LEA OW −1.3 nc. nc. −1.1 nc. nc. −0.7 nc. nc.
B-CW −1.3 nc. nc. −1.1 nc. nc. −0.7 nc. nc.

SM&SIP vs.
Middle managers −0.5 nc. nc. −0.3 nc. nc. −0.2 nc. nc.
HEA OW −0.5 nc. nc. −0.3 nc. nc. −0.2 nc. nc.
LEA OW −0.5 nc. nc. −0.3 nc. nc. −0.2 nc. nc.
B-CW −0.5 nc. nc. −0.3 nc. nc. −0.2 nc. nc. 
Notes: For scenarii Max1, 2 & 3, see Methodology ‘Evaluating the impact regarding a potential’. The value of the dissimilarity
index measuring the level of segregation between B-CW and SM&SIP is 1.6% lower under the Max1 scenario compared
to the value under the H0 hypothesis.
Abbreviations: B-CW—blue-collar workers; SM&SIP—senior managers and superior intellectual professions; HEA OW—
office workers with high educational achievement (college degree); LEA OW— office workers with low educational
achievement (no college degree).
Source: French National Census 2008 – INSEE.

the Marseille MA, for instance, the additional social mix between blue-collar workers and
middle managers brought about by the new social housing (compared with the baseline
H0 hypothesis, in the dissimilarity index) is only 12 per cent of the surplus mix that
would be achieved in the Max3 scenario. This means that with the 4000 social housing
units built between 1999 and 2008 in the Marseille MA, it would have been possible
to get eight times more social mix than was actually achieved. The few cases where the
percentage of the achieved potential is greater are specific situations where the estimated
‘maximum’ itself is very near to the estimated ‘minimum’ under the H0 hypothesis. The
higher percentage of achieved potential is therefore a consequence of the low starting
potential.
The location and occupancy of the new social housing units in the three MAs help us to
understand the reasons for the large gap between the actual impact on social mix and the
potential impact that could have been achieved. In the Paris and Lyon MAs, only 15 per cent
of the social housing built between 1999 and 2008 was located in the top two quintiles of
neighbourhoods/municipalities/cantons inhabited by senior managers and people in the
Housing Studies   17

Table 5. Share of achieved potentiala (%) in reference to the Max scenarii—dissimilarity index.
Paris Lyon Marseille
Max1 Max2 Max3 Max1 Max2 Max3 Max1 Max2 Max3
B-CW vs. Max2a Max3a Max2a Max3a Max2a Max3a
SM&SIP 16.6 15.0 5.0 20.9 37.3 10.6 31.8 27.7 10.6
Middle managers 40.0 18.3 7.1 58.6 37.8 18.8 34.4 35.0 11.7
HEA OW 40.9 29.9 10.7 30.1 38.1 17.5 4.6 <0 19.4
LEA OW 57.7 <0 9.3 68.8 43.3 23.4 0.6 <0 28.5

LEA OW vs. Max2b Max3b Max2b Max3b Max2b Max3b


SM&SIP 22.8 37.2 8.4 15.7 93.6 9.4 37.8 38.0 15.2
Middle managers 52.3 >100 17.0 35.3 >100 20.5 56.7 36.3 21.1
HEA OW 52.8 >100 14.8 35.3 >100 20.9 29.7 <0 19.1
B-CW 27.1 <0 9.0 10.2 >0 16.5 >100 72.7 34.2

HEA OW vs. Max2c Max3c Max2c Max3c Max2c Max3c


SM&SIP <0 <0 2.5 2.8 87.5 4.2 20.0 20.0 5.2
Middle managers 56.2 <0 11.6 49.9 38.7 15.3 51.1 78.7 8.0
Downloaded by [Laurentian University] at 09:45 16 March 2016

LEA OW 35.9 <0 10.4 50.2 36.6 16.3 72.7 55.4 14.2
B-CW 15.1 <0 7.2 26.8 25.2 11.1 71.2 27.1 15.1

Middle managers vs.


SM&SIP 4.9 nc. nc. <0 nc. nc. 8.6 nc. nc.
HEA OW 60.2 nc. nc. 49.6 nc. nc. 28.3 nc. nc.
LEA OW 49.6 nc. nc. 71.8 nc. nc. 17.4 nc. nc.
B-CW 2.4 nc. nc. 18.3 nc. nc. 10.6 nc. nc.

SM&SIP vs.
Middle managers 45.7 nc. nc. >100 nc. nc. 36.3 nc. nc.
HEA OW 71.8 nc. nc. 95.6 nc. nc. 19.5 nc. nc.
LEA OW 29.6 nc. nc. 84.1 nc. nc. 22.0 nc. nc.
B-CW 17.4 nc. nc. 65.1 nc. nc. 14.7 nc. nc.
Note: For scenarii Max1, 2 & 3, see Methodology ‘Evaluating the impact with reference to a potential’.
Abbreviations: B-CW—blue-collar workers; SM&SIP—senior managers and superior intellectual professions; HEA OW—
office workers with high educational achievement (college degree); LEA OW—office workers with low educational
achievement (no college degree).
Source: French National Census 2008 – INSEE.
a
Share of ‘achieved potential’ = (OBS − MIN)/(MAX − MIN) * 100 with: OBS: observed value for the dissimilarity index; MIN:
value calculated under the hypothetic situation H0; MAX: value calculated under scenarii Max1, Max2 or Max3.

superior intellectual professions (representing 40 per cent of all households in this group
in the MA),7 and Marseille MA performs only slightly better with 21 per cent. A much
larger proportion of new social dwellings were built in poverty areas: 59 per cent of new
social housing in the Paris and Lyon MAs was located in the two neighbourhood quintiles
with the highest proportion of blue-collar workers in their population—54 per cent in the
Marseille MA. It would seem that the location of new social housing developments was
more closely aligned with the geography of low-income households than with the geography
of the upper socio-professional categories. If this had been less the case, the impact on the
social mix would most probably have been greater.
The social characteristics of tenants living in the newly built social housing confirm that
a significant number of these dwellings, especially those in upper-class areas, were allocated
to middle-class households. Overall, households headed by a senior manager or middle
manager (or equivalents), occupy 35 per cent of all new social housing in the Paris MA,
25 per cent in the Lyon MA, 23 per cent in the Marseille MA (see Table 1). These percent-
ages rise considerably as we move into high-status areas. In the Paris MA, half of the 4000
freshly built social housing units in the top quintile of the wealthiest areas were rented to
18    E. Korsu

households headed by senior managers and middle managers—similar observations can be


made for Lyon and Marseille. In contrast, a large proportion of the new social housing built
in lower-class areas was let to lower-class households. In the Lyon MA, for instance, 40 per
cent of new social dwellings in the top quintile of areas with the most blue-collar households
were rented to blue-collar households and therefore did not increase cohabitation between
this group and others. This suggests that the allocation processes—for whatever reasons,
good or bad—also played their part in the relatively small real impact of new social housing
developments on the social mix, compared with the Max3 scenario.
With more modest scenarios, the fraction of the potential achieved reaches more sig-
nificant levels, in some cases topping 50 per cent. The Max1 scenario describes a virtual
situation where all new social housing occupied by group j households is ‘re-located’ in
a spatial pattern identical to that of group k households. The level of social mix actually
achieved in Paris, Lyon and Marseille between B-CW and middle managers and between
LEA office workers and middle managers is comparable with the levels that would have been
Downloaded by [Laurentian University] at 09:45 16 March 2016

reached in Max1. In the Paris MA, the impact of new social housing on cohabitation between
blue-collar workers and middle managers—measured with the dissimilarity index—is 40 per
cent of the outcome that Max1 would have produced. This fraction climbs to 60 per cent
in the Lyon MA and 35 per cent in the Marseille MA. We obtained 41 per cent of potential
in the Paris MA for the social mix between B-CW and HEA office workers, 30 per cent in
Lyon (but only 5 per cent in Marseille). For cohabitation between LEA office workers and
middle managers, the figure was 52 per cent in Paris, 35 per cent in Lyon, 57 per cent in
Marseille. But the performances are worse when it comes to improving the mix between
B-CW and senior managers and superior intellectual professions (SM&SIP). In the Paris
MA, the fraction of potential is only 17 per cent for the B-CW/SM&SIP mix and no more
than 23 per cent for the LEA office workers/SM&SIP mix. Lyon fared no better than Paris
(21 per cent and 16 per cent, respectively). Marseille scores better (32 and 38 per cent,
respectively), but mainly because the potential maxima were low.
The global impact of social housing allocated to households where the head is a senior
manager or works in the superior intellectual professions confirms the difficulty of mixing
households from the lower and upper social strata. In all three cities, these dwellings had a
positive effect on the social mix between blue-collar workers and LEA office workers, but the
fraction of potential achieved is low in Paris and Marseille: 17 and 15 per cent, respectively,
for the SM&SIP/B-CW mix, 30 and 22 per cent, respectively, for the SM&SIP/LEA office
workers mix. This means that if the new social dwellings occupied by SM&SIP were located
with the same geographical distribution as households occupied by B-CW or LEA office
workers, the impact on social mix would have been several times greater. The fractions of
achieved potential are higher in Lyon, but mainly as a consequence of low potential maxima
under Max1—which is itself due to the small number of new social housing units allocated
to SM&SIP households in Lyon (363 out of 6900). However, it would seem that the social
housing allocated to SM&SIP households was more efficient in increasing cohabitation with
middle-class households than with working-class households.
The second scenario (Max2) simulates a situation where the priority in the allocation
of new social housing is assigned to households from lower social strata. When compared
with this scenario, the real impact of new social housing on B-CW/SM&SIP and on B-CW/
Middle managers mixes appears rather low. In Paris, the impact on these two types of social
mix is only 20 per cent of the impact that would have been achieved under Max2a, i.e. if all
Housing Studies   19

new social housing, except that occupied by nonworking people, were allocated to B-CW
households. The proportion of potential achieved is greater in the case of LEA office workers/
SM&SIP and LEA office workers/middle managers mixes, but this seems mainly because
the potential maximum attained under Max2b is relatively lower.

The disaggregated evaluation


The second way to evaluate the impact of new social housing on social mix is to adopt a
micro perspective and to measure the marginal effect produced by each new social dwelling.
This approach seeks to distinguish housing units with a positive impact from those with a
negative impact.
When we use the dissimilarity index, the share of social housing units with a positive
impact varies according to the socio-professional groups concerned, but revolves around
50 per cent (Table 6). In each of the three cities, the proportion of social dwellings that
Downloaded by [Laurentian University] at 09:45 16 March 2016

increase social mix is within a range of 45–55 per cent in half of the 20 cases explored
(5 social groups matched in pairs). It would seem that in one in two cases, the new social
housing units increased cohabitation between the pairs concerned as often as they pro-
duced the opposite effect. To give an example, in the Marseille MA, 49 per cent of the one
thousand new housing units allocated to B-CW households produced a positive marginal
impact on social mix between B-CW and middle managers, while 51 per cent affected it
negatively. We have here a sort of zero-sum game where the pro-mix marginal impact of
half of the new social dwellings is offset by the anti-mix marginal impact of the other half.
As a general rule, the proportion of new social dwellings with a positive impact on
social mix is greater when the socio-professional groups considered are fairly close in
the social structure. Housing units with a pro-mix impact are more frequent in the case
of cohabitation between LEA office workers and HEA office workers or middle manag-
ers than in the case of cohabitation between LEA office workers and SM&SIP. In Paris,
60 per cent of social dwellings allocated to households headed by LEA office workers
have a positive effect on the mix between LEA office workers and middle managers, but
only 43 per cent produce a positive impact on the mix between LEA office workers and
SM&SIP. The same is true for the new social housing allocated to B-CW households in
Paris and Lyon (but not in Marseille): these dwellings are more efficient in increasing
cohabitation with office workers or middle managers than with SM&SIP. Similar obser-
vations can be made about new social housing attributed to middle-class households. A
larger share of these dwellings produced a pro-mix impact to increase the cohabitation
between SM&SIP and middle managers or between these two groups and HEA office
workers. Conversely, negative marginal impacts were more frequent in the case of cohab-
itation with LEA office workers or B-CW households. In Marseille, 43 per cent of new
social dwellings occupied by SM&SIP households positively impacted the cohabitation
with middle managers, whereas only 30 per cent of the same housing units had a positive
impact on the SM&SIP/B-CW mix.
In Paris and Marseille, social housing units allocated to SM&SIP households contributed
the least to social mix. However, the opposite is true in Lyon. This suggests that in the first
two cities, allocating part of the new social housing units to households from the upper-
middle-classes achieved poor returns in terms of social mix, while in the third city, this
option worked more effectively.
20    E. Korsu

Table 6. Share of NSH units with a pro-mix impact (%)—dissimilarity index.


Paris Lyon Marseille
B-CW vs.
SM&SIP 34.8 45.4 47.8
Middle managers 49.5 60.1 48.7
HEA OW 46.5 39.3 41.5
LEA OW 55.9 53.1 45.7

LEA OW vs.
SM&SIP 43.1 45.8 51.9
Middle managers 60.4 56.1 60.2
HEA OW 53.1 45.7 50.2
B-CW 51.1 49.3 64.1

HEA OW vs.
SM&SIP 46.3 36.8 43.4
Middle managers 56.6 47.0 59.4
LEA OW 55.4 54.6 58.7
Downloaded by [Laurentian University] at 09:45 16 March 2016

B-CW 44.4 44.5 57.1

Middle managers vs.


SM&SIP 42.9 39.6 47.6
HEA OW 59.5 55.1 50.1
LEA OW 48.5 50.7 38.5
B-CW 39.4 44.3 36.8

SM&SIP vs.
Middle managers 48.8 65.7 42.8
HEA OW 51.6 64.1 38.8
LEA OW 40.0 55.5 34.2
B-CW 34.8 53.0 29.7
Note: 34.8% of new social housing units allocated to B-CW households produce a positive marginal impact on the social mix
between B-CW and SM&SIP—with the dissimilarity index.
Abbreviations: NSH—New social housing; B-CW—Blue-collar workers; SM&SIP—Senior managers and superior intellectual
professions; HEA OW—office workers with high educational achievement (college degree); LEA OW—office workers with
low educational achievement (no college degree).
Source: French National Census 2008 – INSEE.

The results of the evaluation are more encouraging with the exposure index (Table 7). The
share of new social housing units which positively affect social mix gravitates around a mean
of 60 per cent. This means that in most cases, dwellings with a marginal pro-mix impact,
which bring group j households into areas where the frequency of group k neighbours is
greater than on average for group j, are a substantial majority. In Lyon, almost 70 per cent of
new social dwellings occupied by B-CW households are located in areas with more middle
managers than the average area occupied by B-CW households. The results are similar in
Paris and Marseille. The share of dwellings with a marginal pro-mix impact is even greater
in the case of the LEA office workers/middle managers mix. In the Lyon MA, new social
housing is also quite successful in giving B-CW and LEA office workers access to neigh-
bourhoods where SM&SIP are better represented than usually in the vicinity of these two
groups. Sixty per cent of new social dwellings inhabited by B-CW or LEA office workers
are located in areas where the proportion of SM&SIP households is above the average for
B-CW or LEA office workers.
The new social housing allocated to the middle classes also seems to foster cohabitation
with the lower social strata. In each of the three cities, some 65 per cent of dwellings allocated
to SM&SIP households are located in areas containing more LEA office workers than the
Housing Studies   21

Table 7. Share of NSH units with a pro-mix impact (%)—exposition index.


Paris Lyon Marseille
B-CW vs.
SM&SIP 52.2 60.2 54.0
Middle managers 65.3 68.9 64.4
HEA OW 59.5 51.9 59.6
LEA OW 49.6 49.7 40.7

LEA OW vs.
SM&SIP 54.8 60.2 55.2
Middle managers 68.6 74.8 67.3
HEA OW 55.1 48.3 45.3
B-CW 42.7 32.3 42.4

HEA OW vs.
SM&SIP 46.9 44.7 45.7
Middle managers 66.4 54.1 46.7
LEA OW 51.5 61.8 56.9
Downloaded by [Laurentian University] at 09:45 16 March 2016

B-CW 47.3 50.8 60.9

Middle managers vs.


SM&SIP 48.9 43.7 50.6
HEA OW 63.3 53.2 48.5
LEA OW 57.4 61.2 45.8
B-CW 47.0 54.7 50.2

SM&SIP vs.
Middle managers 69.6 60.1 62.5
HEA OW 69.8 52.6 51.1
LEA OW 65.1 67.9 65.3
B-CW 53.7 79.1 64.6
Note: 52.2% of new social housing units allocated to B-CW households produce a positive marginal impact on the social mix
between B-CW and SM&SIP—with the exposition index.
Abbreviations: NSH—new social housing; B-CW—blue-collar workers; SM&SIP—senior managers and superior intellectual
professions; HEA OW—office workers with high educational achievement (college degree); LEA OW—office workers with
low educational achievement (no college degree).
Source: French National Census 2008 – INSEE.

mean level for SM&SIP households. The percentages are equally high for the SM&SIP/B-CW
pairing in Lyon and Marseille.

Concluding remarks
At the end of this evaluation, it is reassuring to find that the new social housing built in the
2000s increased cohabitation between social groups in Paris, Lyon and Marseille, but the
scale of this pro-mix impact is certainly disappointing. With hindsight, this finding looks
like common sense: the additional social dwellings built each year are so few, compared with
the existing stock, that they cannot be expected to bring about a profound change in urban
social structure. This suggests that, under present circumstances, it is somewhat unrealistic
to count on social housing policy to reduce segregation substantially in France’s big cities.
It would seem fairly clear that socio-spatial separation cannot be overcome solely through
the construction of social housing. Given the somewhat slow pace of construction and the
‘sub-optimal’ impact on the social mix, it would take decades to make a significant dent in
segregation through social housing policy. This conclusion probably holds true for other
countries, where policies employed to increase social mix may also have only a marginal
impact on the existing housing stock.
22    E. Korsu

We believe that our basic conclusions hold true despite the fact that the assumptions
underlying our evaluation are open to debate. The impact of recent social housing devel-
opments on social mix was measured by comparison with an H0 scenario containing the
following hypothesis: if no new social housing were constructed in the period, the urban
system would have evolved in such a way that the geographical distribution of households
which in reality occupy the new social housing units would be the same as that of other
households with similar socio-professional characteristics. Other scenarii could of course
be imagined. The important point, however, is that the H0 scenario, since it reproduces the
segregated pattern of the real city, does not entail an overestimation of the level of social
mix that would exist if no social housing were constructed in the period in question. This
implies that if the H0 scenario introduces a bias, the nature of that bias is conservative with
regard to our conclusions, since the risk of seriously underestimating the impact on social
mix with an evaluation based on the H0 scenario would seem to be small.
In reality, it would appear that the development of new social housing has a modest
Downloaded by [Laurentian University] at 09:45 16 March 2016

impact in giving a number of low-income households the possibility of living in areas where
middle/upper-class neighbours are slightly more common than is usual for these groups. In
2008, of the 10 000 B-CW households which rented a new social dwelling in the Paris MA,
around 5000–6000—approximately 1 per cent of total B-CW households—lived in areas
where SM&SIP, middle managers or HEA office workers were more highly represented than
in the average neighbourhood of residence for B-CW households: this modest outcome is
the type of benefit brought about by almost a decade of social housing construction. It does
not change the world, but nor should such outcomes be sneezed at.
Nonetheless, more could have been done, even with the limited scale of social housing
construction in recent years. The additional social mix generated by the new social dwell-
ings is fairly small compared with the increase that could have been obtained if more of
these housing units were allocated to low-income households and if those dwellings had
been better located. One can take this result as encouraging: it suggests that the pro-mix
impact of new social housing can be substantially improved. However, there is another way
to look at it. First, our simulations show that even with highly favourable assumptions, the
additional social mix generated by new social housing is not particularly large (at best, and
only in some cases, an increase of 10 per cent). Second, the capacity of the public authorities
to improve the pro-mix impact of social housing remains open to doubt.
One of the findings that emerged from our assessment is that, as a general rule, new social
housing was more successful in bringing together households in the lower socio-profes-
sional strata with those in the middle socio-professional strata, than with those in the upper
socio-professional strata. Contrary to what might initially be thought, this may be a fairly
positive outcome: many social scientists believe that cohabitation between social groups is
more likely to produce positive outcomes if the groups sharing the same living space are not
socially too different from each other.8 In any case, cohabitation in the same neighbourhoods
between households from the lower and middle social strata still generates social mix, and
there is no obvious reason why this form of social mix should be deemed inferior.
Another finding is the significant share of new social dwellings occupied by B-CW and
LEA office workers, which has a negative impact on social mix, locating these households
in low-income areas where their own group is already strongly represented. This is symp-
tomatic of the social housing dilemma caused by the potential antagonism between the
principles of ‘right-to-housing’ and social mix (Epstein & Kirszbaum, 2003; Genestier &
Housing Studies   23

Bacqué, 2004; Simon, 2003; Tanter & Toubon, 1999). These dwellings enable low-income
households to get affordable housing at a time when house prices and rents are rising sharply
on the private market. But at the same time, they maintain the spatial separation between
the lower social strata and the rest of society.
The same rationale applies to the new social housing built in wealthy areas and allocated
to middle-class households. In Paris, Lyon and Marseille, 5000 new social housing units
were occupied by middle managers and 2000 by senior managers and superior intellec-
tual professions, negatively impacting the social mix between these groups and blue-collar
workers. It would seem that allocating a substantial share of new social housing units to
middle-class tenants works against a social mix with populations from lower socio-pro-
fessional categories.

Notes
Downloaded by [Laurentian University] at 09:45 16 March 2016

1. 
For instance, the issue of discriminatory practices in the allocation of social housing—with
landlords showing a systematic propensity to channel demand from applicants with specific
ethnic or demographic characteristics towards specific areas—seems as significant in the
UK as in France (Sala Pala, 2006). It is also worth recalling the Gautreaux case in the United
States back in the late 1970s, when the Chicago Housing Authority was convicted of racial
discrimination in public housing policy.
It should be noted that the proportion of lower-range products (financed through PLA-Iloans)
2. 
in new constructions went up sharply in the late 2000s. Nationwide, this share increased from
10 per cent in 2000 to more than 20 per cent after 2011.
3. 
Social housing in France is built by social landlords who own and administer the properties.
The investments are financed mainly by the State, by local governments and by private
companies. According to the level of funding, these financing bodies gain the right to reserve
a quota of dwellings—and are therefore called ‘reserving organisations’. This right allows each
reserving organisation to propose its own candidates for dwellings within its own quota.
The allocation procedure occurs as follows. For each unoccupied dwelling in its quota, the
reserving organisation selects and ranks a set of applications. The landlord’s administrative
departments review the selected applications, sometimes with the collaboration of partners
(municipality, civil society bodies). The applications are then forwarded to the allocation
committee. Each time this committee reaches a consensus, a dwelling is allocated to an
applicant household.
4. 
French social landlords are not alone in employing this strategy. Variants of the same basic
strategy can be found in other European countries (cf. Busch-Geertsema, 2007).
As pointed out by the French Haute Autorité de Lutte contre les Discriminations et pour l’Egalité
5. 
(2011), the existence of such practices is indirectly proven by several studies which bring to
light phenomena such as: the concentration of immigrant groups in specific neighbourhoods
or buildings; the overrepresentation of immigrants in social housing located in deprived
neighbourhoods; poorer quality housing offers to immigrant applicants; longer periods on
waiting list for immigrant applicants.
The Atelier Parisien d’Urbanisme (2011) estimated that there were 26 000 non-social housing
6. 
units acquired by the municipality of Paris and converted into social housing between 2001
and 2010 (12 000 of them after renewal) while 23 000 new social housing units were built
during the same period.
7. 
The quintiles are based on a ranking of neighbourhoods, municipalities and cantons according
to the share of a given socio-professional group in the local population. Each quintile consists
of a group of neighbourhoods/municipalities/cantons representing 20  per cent of the
households in the metropolitan area for a given socio-professional group. The two quintiles
of neighbourhoods/municipalities/cantons where senior managers and people in superior
intellectual professions are the most represented account for 40 per cent of the households in
24    E. Korsu

the metropolitan area headed by a person belonging to this socio-professional group. For the
sake of simplicity, in the remainder of this paper we will talk about ‘neighbourhood quintiles’
as a simplified formula instead of ‘quintiles of neighbourhoods/municipalities/cantons’.
8. 
There is some evidence suggesting that low-income groups take more advantage of social
mix when in the proximity of middle-income groups than high-income groups. In this latter
case, the social differences are so large that they considerably reduce inter-group dialogue
and exchange—consistent with the findings of Atkinson & Kintrea (2002), who observe that
the propensity to look for social contact with others decreases as inter-personal differences
increase. Also, the wealthiest neighbourhoods often provide few amenities (shops, services,
jobs) available to low-income people, so living in such neighbourhoods can be uncomfortable
for these groups. Comparatively, middle-class neighbourhoods are often more welcoming.
According to Bacqué et al. (2011), one of the most common forms of trans-class solidarity
is between middle classes and working classes. The authors consider that the fraction of the
middle classes defined more by cultural than by economic capital (such as middle managers
involved in artistic activities, social services, education or health) is particularly disposed to
develop social relations with people from lower socio-economic groups.
Downloaded by [Laurentian University] at 09:45 16 March 2016

Disclosure statement
No potential conflict of interest was reported by the author.

References
Andersen, H. (2002) Can deprived housing areas be revitalised? Efforts against segregation and
neighbourhood decay in Denmark and Europe, Urban Studies, 39, pp. 767–790.
Anderson, L. M., St. Charles, J., Fullilove, M. T., Scrimshaw, S. C., Fielding, J. E., Normand, J. & The
Task Force on Community Preventive Services. (2003) Providing affordable family housing and
reducing residential segregation by income. A systematic review, American Journal of Preventive
Medecine, 24(3), pp. 47–67.
Arthurson, K. (2002) Creating inclusive communities through balancing social mix: A critical
relationship or tenuous link?, Urban Policy and Research, 20(3), pp. 245–261.
Atelier Parisien d’Urbanisme. (2011) Les chiffres du logement social à Paris, Notes de 4 pages, 46,
pp. 1–5.
Atkinson, R. & Kintrea, K. (2002) Area effects: What do they mean for British housing and regeneration
policy?, European Journal of Housing Policy, 2(2), pp. 147–166.
Bacqué, M.-H., Fijalkow, Y., Launay, L. & Vermeersch, S. (2011) Social mix policies in Paris: Discourses,
policies and social effects, International Journal of Urban and Regional Research, 35(2), pp. 256–273.
Bilek, A., Costes, N. & Monmousseau, F. (2007) La loi SRU incite-t-elle les maires à construire
du logement social?, Les enseignements d’une analyse principal-agent, [Does the Solidarity and
Urban Renewal act incite mayors to build social housing? Lessons from a principal-agent analysis],
Economie Publique, 20(1), pp. 207–237.
Bolt, G., Phillips, D. & Van Kempen, R. (2010) Housing policy, (de)segregation and social mixing:
An international perspective, Housing Studies, 25(2), pp. 129–135.
Bridge, G., Butler, T. & Lees, L. (2012) Mixed Communities. Gentrification by Stealth? (Bristol: Policy
Press).
Busch-Geertsema, V. (2007) Measures to achieve social mix and their impact on access to housing
for people who are homeless, European Journal of Homelessness, 1, pp. 213–224.
Butler, T. & Hamnett, C. (2011) Ethnicity, Class and Aspiration. Understanding London’s New East-
End (Bristol: The Policy Press).
Driant, J.-C., & Lelèvrier, C. (2006) Le logement social: mixité et solidarité territoriale [Social housing:
social mix and territorial solidarity], in: H. Lagrange & M. Oberti (Eds) Emeutes urbaines et
protestations [Urban riots and protests]. Une singularité française, pp. 177–193 (Paris: Presses de
Sciences-Po).
Housing Studies   25

Dunleavy, P. (1981) The Politics of Mass Housing in Britain, 1945–1975. A Study of Corporate Power
and Professional Influence in the Welfare State (Oxford: Clarendon Press).
Epstein, R. & Kirszbaum, T. (2003) L’enjeu de la mixité sociale dans les politiques urbaines [The social
mix issue in urban policy], Regards sur l’actualité [News at a glance], 292, pp. 1–14.
Galster, G. C. (2013) Neighbourhood social mix: Theory, evidence and implications for policy and
planning, in: N. Carmon & S. Fainstein (Eds) Policy, Planning and People. Promoting Justice in
Urban Development, pp. 307–336 (Philadelphia, PA: University of Pennsylvania Press).
Genestier, P. & Bacqué, M.-H. (2004) Comment loger les plus pauvres si l'on démolit les HLM?
[How can we house the poor if we demolish the social housing buildings?], Mouvements, 32(2),
pp. 126–134.
Goodchild, B. & Cole, I. (2001) Social balance and mixed neighbourhoods in Britain since 1979:
A review of discourse and practice in social housing, Environment and Planning D: Society and
Space, 19(1), pp. 103–121.
Haute Autorité de Lutte contre les Discriminations et pour l’Egalité (HALDE). (2011) Accès au
logement social  : garantir l’égalité [Access to social housing: ensuring equality] (Paris: Haute
Autorité de Lutte contre les Discriminations et pour l'Egalité [High Authority for the Fight against
Downloaded by [Laurentian University] at 09:45 16 March 2016

Discrimination and for Equality].


Hirsch, A. (2000) Choosing segregation. Federal Housing Policy between Shelley and Brown. In From
Tenements to the Taylor Homes, in: J. Bauman, R. Biles, & K. Szylvian (Eds) Search of an Urban
Housing Policy in Twentieth Century America, pp. 206–225 (Pennsylvania, PA: The Pennsylvania
State University Press).
Lévy-Vroelant, C. & Tutin, C. (2007) Social housing in France, in: C. Whitehead & K. Scanlon (Eds)
Social Housing in Europe, pp. 70–89 (London: London School of Economics and Political Science).
Madoré, F. (2004) Ségrégation sociale et habitat [Social Segregation and Housing] (Rennes: Presses
Universitaires de Rennes).
Massey, D. & Denton, N. (1993) American Apartheid. Segregation and the Making of the Underclass
(Cambridge: Harvard University Press).
Melis, G., Marra, G. & Gelormino, E. (2013), Housing and Social Mix Report Equity Action Work
Package 6 – Litterature Review, November.
Musterd, S. & Andersson, R. (2005) Housing mix, social mix, and social opportunities, Urban Affairs
Review, 40(6), pp. 761–790.
Musterd, S., Ostendorf, W. & De Vos, S. (2003) Neighbourhood effects and social mobility: A
longitudinal analysis, Housing Studies, 18(6), pp. 877–892.
Norris, M. & Shiels, P. (2007) Housing inequalities in an enlarged European Union: Patterns, drivers
and implications, Journal of European Social Policy, 17(1), pp. 59–70.
Pittini, A. & Laino, E. (2011) Logement social européen 2012. Les rouages d’un secteur [European
Social Housing 2012: workings of a sector] (Brussels: CECODHAS Housing Europe’s Observatory).
Rose, D. (2004) Discourses and experiences of social mix in gentrifying neighbourhoods: A Montreal
case study, Canadian Journal of Urban Research, 13(2), pp. 278–316.
Sala Pala, V. (2005) Le racisme institutionnel dans la politique du logement social [Institutional racism
in social housing policy], Sciences de la société [Social sciences], 65, pp. 87–102.
Sala Pala, V. (2006) La politique du logement social au risque du client ? Attributions de logements
sociaux, construction sociale des clients et discriminations ethniques en France et en Grande-
Bretagne [The Social housing policy and the client. Social housing allocation, social construction
of clients and ethnic discrimination in France and the United Kingdom], Politiques et management
public [Policy and public management], 24, pp. 77–92.
Scally, C. P. (2014) The nuances of NIMBY: Context and perceptions of affordable rental housing
development, Urban Affairs Review, 49(5), pp. 718–747.
Simon, P. (2003) Le logement social en France et la gestion des ‘populations à risques’ [Social housing
in France and the management of “high-risk populations”], Hommes et Migrations [People and
migrations], 1246, pp. 76–91.
Tanter, A. & Toubon, J.-C. (1999) Mixité sociale et politiques de peuplement: genèse de l’ethnicisation
des opérations de réhabilitation [Social mix and occupation policies: the origins of the ethnicization
of renewal operations], Sociétés contemporaines [Contemporary societies], 33(1), pp. 59–86.
26    E. Korsu

Tissot, S. (2005) Une « discrimination informelle » ? [An “informal discrimination”], Actes de la


recherche en sciences sociales [Acts of the research in social sciences], 159(4), pp. 54–69.
Wallace, J. (1995) Financing affordable housing in the United States, Housing Policy Debate, 6(4),
pp. 785–814.
Whitehead, C. & Scanlon, K. (2007) Social Housing in Europe (London: London School of Economics
and Political Science).
Williamson, A. R. (2011) Can they afford the rent? Resident cost burden in low income housing tax
credit developments, Urban Affairs Review, 47(6), pp. 775–799.
Downloaded by [Laurentian University] at 09:45 16 March 2016

You might also like