Earnings differentials between married and unmarried men have been dec
lining since the late 1960s. We consider two possible explanations for
this decline: changes in the nature of selection into marriage; and c
hanges in role specialization within marriage. Our analysis of changes
in marriage differentials within cohorts supports only a small contri
bution of changes in selection. There is some evidence that difference
s in human-capital investment between married and unmarried men have f
allen over time, but this effect has apparently been largely offset by
increases in the return to that human capital.