Endogenity in logistic regression Consider a logistic regression model built to assess female work participation with household income included as an explanatory variable.  Would household income be an endogenous variable in this context? If so how do I control for the endogeneity?
 A: I would say you would have omitted variable bias there if you only have variable whether woman supplies labor or not and total labor income of the household. Total income is partly composed of the earnings of woman in the family, assuming you have family with at least two adults / earners. 
One method to eliminate this is to have a sample where women are not working at the start of the observation period. They are observed for next T periods and event (0/1 coding) occurs if they start to supply labor or not. There earnings can be used since they have not contributed to that. 
Economic model could be a system of several equations where decision to supply labor is affected by both household and individual characteristics.
A: This sounds like a classic case of a particular omitted variable bias called the Heckman sample selection problem. 
Attrition is meaningful here, as the decision to not work may be based on house characteristics, and the wage (the woman would have earned had she worked or the observed wage) potentially are driven by individual characteristics. 
Notes on the Heckman Selection Issues.
