We consider the implications of narrow choice sets on welfare estimation in
a random utility model of recreation demand We hypothesize that careful fo
rmulation of the choice set focusing on the sites of policy interest and th
eir closest substitutes will give reasonably accurate welfare estimates. We
use nearby sites as close substitutes and treat more distant sites as aggr
egate alternatives in our application to fishing in Maine. We find that the
welfare estimates are rather sensitive to narrowing choice sets in this ma
nner The sensitivity largely tracts variation in the estimated travel cost
coefficient across the different models considered. (JEL Q26).