Substantive income effects are incorporated in a logit or nested-logit model by assuming that utility is a piece-wise linear spline function of residual income. Specific income data are not required, only income by category. Expected compensating variation is easily and accurately approximated...