This paper exploits discontinuities and randomness in the college admissions in Sweden in 1982, to estimate the economic return to college in the 1990s. At the time, college admissions were highly selective and applicants were ranked with respect to their formal merits. Admissions were given to those ranked higher than some threshold value. At the margin, applicants were sometimes randomly assigned to college. Exploiting this Regression-Discontinuity design, individuals who were admitted in 1982 are estimated to have about 0.20 years longer college education in 1996. However, the earnings effects for applicants at the margin of admission are insignificant. Controlling for the college admission determinants, the OLS-estimates of the return to college is 1.4 percent in 1991-96. The IV-estimates are not significantly different from the OLS counterparts. (Contains 4 figures and 6 tables.)