1 Introduction
A rapidly growing stream of research in the social interactions literature focuses on neigh-
borhood effects, that is, the impact of neighbors’ characteristics and behaviors on individual
socio-economic outcomes.1 Indeed, theoretical and empirical evidence suggests that interactions
among neighbors are likely to affect individual labor-market outcomes through peer effects and
role models in the human capital acquisition process, attitudes towards work, and dissemina-
tion of information on job opportunities. Arnott and Rowse (1987) show that less-able learners
exerts negative externalities on the learning process of other students. Benabou (1993) argues
that the cost of education acquisition may be influenced by education decisions of neighbors.
Wilson (1987) explains that the lack of successful role models among older adults in deprived
neighborhoods may influence youths’ motivations and attitudes. The role of social networks on
information about job opportunities has also been highlighted, especially for low-skilled workers
who often resort to informal search modes such as personal contacts. As a consequence, the
percentage of employed individuals in the neighborhood may influence other residents’ access to
job opportunities (Topa, 2001; Bayer et al., 2005). Finally, the stigmatization of deprived neigh-
borhoods may lead employers to discriminate workers on the basis of their residential location
(Zenou and Boccard, 2000).
Measuring neighborhood effects raises the issue of location choice endogeneity, which gen-
erates correlated effects (Moffitt, 2001; Durlauf, 2004). Indeed, urban economics has recognized
for long that individuals with similar socio-economic characteristics, labor-market outcomes, and
unobservable traits tend to sort themselves into certain areas of the urban space. Therefore,
studies that do not control for the endogeneity of neighborhood choice will yield biased results.
The inadequate correction for this bias has been put forward to explain the great divergence
of results obtained by empirical studies and is one of the major focuses of recent research on
neighborhood effects.
This paper aims to test for the existence of neighborhood effects on unemployment. Our
identification strategy is twofold. First, it consists in correcting for endogenous selection into
neighborhoods by estimating simultaneously two non-linear models of unemployment and neigh-
borhood choice. More precisely, in a preliminary data analysis step, we classify neighborhoods
as deprived or not deprived and then estimate a simultaneous probit model of unemployment
and type of neighborhood. Second, the large share of public housing units in France and their
1 See Durlauf and Young, 2001 for a review of the social interactions literature and Durlauf, 2004 for neighbor-
hood effects.