not necessary to distinguish endogenous and contextual effects, as both types of causalities are
involved in the relocation of public housing units.
Table 7 displays predicted probabilities of unemployment that are issued from the baseline
two probit model (results displayed in column 5 of Table 6). For each subsample, we give
the observed unemployment rate and average predicted probabilities of unemployment if these
individuals were to be located either in deprived or in other neighborhoods. The neighborhood
effect is higher for public tenants and for individual in deprived neighborhoods. These figures
show that relocating the public housing renters who live in deprived neighborhoods would reduce
their individual unemployment probability from 14.2 to 11.0%.
To be more specific, assume that the location of other housing units remains identical.
Then, given the initial distribution (Table 2), achieving an even distribution of the public hous-
ing tenants between both types of neighborhoods would imply transferring 65% of public housing
units (that is, 43% of the public housing stock) from deprived to other neighborhoods. Suppos-
ing this new distribution to be implementable and assuming the unemployment rate of public
housing renters in non-deprived neighborhood does not change, the overall unemployment rate
of public housing tenants in Lyon’s city would decrease from 12.5% to 11.2%.23 This reduction
is limited and in any case, such a change in the distribution of public housing units would be
very costly.
Furthermore, it must be acknowledged that this simulation suffers from limitations. In
particular, our simulation does not take into account the fact that with the change in the
distribution of public housing, the percentage of public housing renters (characterized by low
levels of education, higher unemployment rates, ...) in the rest of the city would be about twice
as high as it is currently. Assessing more precisely the consequences of such a change would imply
to estimate a continuous relationship between social composition and unemployment probability.
5 Conclusion
The objective of the present paper was to examine how unemployment probabilities are influ-
enced both by accommodation in the public housing sector and location in a deprived neigh-
23This predicted average unemployment rate is calculated by applying the predicted unemployment probability
conditional on living in a non-deprived neighborhood to 65% of public housing renters living in deprived neigh-
borhoods, and the observed unemployment rates to the 35% remaining and to the public renters who are not in
a deprived neighborhood.
20