baseline specification.22 However, the raise of neighborhood effect remains limited, with a max-
imum increase by only 0.18 points (that is, 15% of the baseline estimate) in the specification
considering previous occupational status as unknown. Although this is only an informal way
of assessing the effect of unobserved heterogeneity, these additional results suggest that our es-
timation method provides a reliable estimation of the neighborhood effect. This effect may be
reasonably thought of as being a little higher than 1 point probability.
In summary, living in the 35% of neighborhoods that have been identified as having the
worst combination of social characteristics in our data analysis step increases the probability
of being unemployed by slightly more than 1%. The change of neighborhood type amounts
to a decrease in neighbors’ unemployment rate by 8.7%. By way of comparison, Topa (2001)
found, in the case of Chicago in 1990, that an increase by 8% in the employment of neighboring
tracts would increase employment rate by 1.3 %. As far as public housing is concerned, our
results indicate that only an indirect effect exists, according to which being housed in the public
sector increases unemployment probability by 3%. These effects can be compared with marginal
effects of individual characteristics. For instance, the neighborhood effect is about as low as
two-thirds that of spouse’s foreign nationality, and it is twice as low as the effect of having the
lower education level rather than having graduated from high school (Table 4, column 5). These
effects can also be compared with differences in observed unemployment rates. On average,
observed unemployment rate is by 4.9 points higher in deprived than in other neighborhoods
(Table 7). According to our results, 1.18 probability points of this gap, that is a bit more than
20%, would be the consequence of neighborhood effects, the remaining part ensuing from spatial
sorting. This results holds within each sub-category as defined by the two residential variables.
For instance, the unemployment probability of individuals outside the public housing sector but
living in deprived neighborhoods would decrease by 18% if they were located in another type of
neighborhood.
4.4 Policy simulations
Our results give support to a law that was recently passed in France, aimed at achieving a more
even distribution of public housing units in order to counter potentially harmful effects of public
housing location (Loi SRU “Soiidarité et Renouvellement urbain”, 2000). Our methodology
allows us to go a step further and to assess the potential effect of a change in the spatial
distribution of public housing units in French cities. Let us recall that for this purpose, it is
22Detailed results are available from the authors on request.
19