We investigate whether the ''match'' between student ability and college qu
ality is an important determinant of college graduation rates. We jointly e
stimate a multinomial probit model of college attendance decisions in which
the alternatives are no college and attendance at college in four quality
categories, and a binomial probit model of subsequent graduation decisions.
By allowing the error terms to be correlated across alternatives and time
periods, we identify the effects of observed factors net of their correlati
on with unobservables. We find that students of all ability levels have hig
her chances of graduating if the quality level of their college ''matches''
their observed skill level.