French people born abroad are more often housed in the public sector than French individu-
als. This observation might reflect an attempt by the public housing offices to compensate for
discrimination on the private housing sector or the fact that foreign individuals are pushed to-
ward the public housing sector due to this discrimination. As far as socioeconomic variables are
concerned, occupational status along with education explain the propensity to live in a public
housing unit. Blue-collar workers are more likely to rent a public housing unit than intermediate
professions (the reference category) by 10 points, and office workers by 7 points. The lower the
educational level is, the higher the probability of being renter in the public housing sector. Sur-
prisingly, the spouse’s educational level and not that of the household head is significant. This
probably reflects the fact that it is the possibility to have or not a second wage in the household
(low educated women having a weak incentive to take part in the labor-force) that determines
income and is considered by public housing offices during the application process.
The second column gives marginal effects estimated from the neighborhood equation.
A far as socioeconomic variables are concerned, marginal effects are very similar to marginal
effects in the public housing equation. As expected, nationality, education and professional
status determine the probability to live in a deprived neighborhood, even after conditioning
for the accommodation in the public housing sector.15 Only highly-educated individuals have
significantly different behaviors regarding tenure and neighborhood choices: they do not differ
from the reference category (high school final diploma) as far as tenure is concerned, whereas
they are less likely than the reference to locate in a deprived neighborhood. We may think
that skilled individuals are likely to apply for a public housing unit at the beginning of their
career, but that in any case they avoid deprived neighborhoods. On the contrary, demographic
variables do not explain the probability to live in a deprived neighborhood: neither the age of the
household head16 , nor the number of children have significant coefficients. This accounts for the
fact that the demographic situation of the household is among the criteria that are considered by
public housing offices, whereas they are less relevant in determining residential location choice.
Finally, the public housing variable is the more powerfull in explaining the neighborhood choice
and it has the strongest marginal effect. The introduction of this variable significantly improves
the likelihood of the model.17 Being a renter in the public sector more than doubles (marginal
effect +30 points) the probability to live in a deprived neighborhood, and as will be clear in
next subsection from the simultaneous estimation of the three probits, this estimate does not
15 Estimated coefficients do not change with the introduction of the public housing variable. Only the four-
children variable looses it significance with the introduction of the public housing variable.
16 Nor the age of the spouse introduced in a previous specification.
17The statistic of the likelihood ratio test is 570 for a χ20.5 critical value of 3.84.
14