Neighborhood effects on unemployment ? A test à la Altonji

The aim of this paper is to test for the influence of neighborhood deprivation on individual unemployment probability in the case of Lyon (France). We estimate a bivariate probit model of unemployment and location in a deprived neighborhood. Our identification strategy is twofold. First, we instrument neighborhood type by the gender composition of household's children and the spouse's workplace. Second, we use the methodology proposed by Altonji et al. (2005), that in our case consists in making hypotheses as to the correlation between the unobservables that determine unemployment and the unobservables that influence the selection into neighborhood types. Our results show that the effect of neighborhood deprivation is not significantly different from zero in the bivariate probit with exclusion restrictions. We also show that a correlation of the unobservables as low as ten percent of the correlation of observables is sufficient to explain the positive neighborhood effect that is observed when endogeneity is not accounted for.