Integration of choice probabilities in logit

We find for logit with income effects a function that generates choice probabilities via Roy's identity. We show that it possesses all the properties to qualify as an indirect utility, it has a closed-form expression for the practically interesting translog specification of the systematic utilities, it reduces in the case without income effects to the expectation of the maximum utility. We discuss the use of the findings in welfare measurement and the extension to observed heterogeneity and mixed logit.