In this paper we make a systematic presentation of returns to education in Austria for the period 1981-1997. We use consistent cross-sections from the Mikrozensus and find falling returns over time. These falling returns are not caused by changes in the sample design and reduced willingness to reveal personal incomes in the survey. Moreover, it is shown that especially returns to university education have fallen. If the focus is not on mean returns, but if we apply quantile regression techniques, interesting patterns emerge: returns are falling the most in the lowest quantiles, but remain almost constant in the highest quantiles. The overall picture of falling returns is consistent with a rise in the supply of highly-educated workers in the recent two decades.