Endogeneity is a common issue in logit models, where the instruments used must be related to the variable being studied but not to the error term. A new test for validating these instruments was compared to a simpler alternative using simulations. Both tests performed similarly, but their power decreased when instruments were highly correlated. Real data from Lisbon and Santiago showed that correcting for endogeneity improved the models, but the instruments were not valid for the Portuguese data.