The varying researcher assumptions about model specification in nested-logit random utility models can have significant effects on model estimates and corresponding welfare measures. Using a three-level nested-logit model of Oregon marine sport fishing, we estimate and compare several models with varying assumptions of choice sequence and income effects.