suffer from any endogeneity bias.
The third and fourth columns of Table 4 give marginal effects for the unemployment equa-
tion. We find very conventional results regarding individual determinants. Young individuals
are more often unemployed, and the probability to be unemployed declines until the age of 44,
after which it increases again. Individuals without any diploma or with only a short vocational
training are more likely not to find a job, whereas people who were previously independent
workers or executives are less unemployed than others. Marginal effects do not change much
with the introduction of the two residential variables (column 4 compared to column 3), with the
exception of the blue-collars’ marginal effect that looses its significancy. This result means that
blue-collar workers seem more likely to be unemployed than technicians and supervisors, but
that they in fact do not differ, when controlling for their tenure and location. Probit estimates
show that unemployment probability increases both with location in a deprived neighborhood
and with accommodation in the public sector, the latter being even more important than the
former. However, these estimation results very likely suffer from an endogeneity bias.
4.2 Simultaneous probit model estimates
Table 5 presents the results of the simultaneous probit model. Coefficients of the public housing
and neighborhood equations being very similar to the simple probit results, we do not comment
them here. For the same reasons, we do not comment exogenous variables affecting unemploy-
ment propensity.
The correlation coefficient between the error terms of the neighborhood and the unem-
ployment equations (ρ12) is significantly different from zero at the 5% level, showing as expected
that the neighborhood type is endogenous in the unemployment equation and that coefficients
estimated from a simple probit are biased. The other two correlation coefficients are not sig-
nificant, suggesting that the public housing variable is not endogenous in the unemployment
equation, nor in the neighborhood equation. The latter result can be interpreted as showing
that households are not deterred from applying to a public housing unit by the spatial distri-
bution of the public housing sector and is coherent with the fact that the application process
allows households to express spatial preferences.
While the coefficient of public housing in the simple probit of unemployment is highly
significant (see Table 4), correcting for the endogeneity of neighborhood and taking into account
its strong ties with public housing eliminate any effect of tenure on unemployment probability.
15