The purpose of the paper is to assess the theory that the downside risk insurance provided by more generous welfare states generates long run efficiency gains, which counterbalance the short run efficiency losses caused by work disincentives in these states (Feldstein 1974, 1976; Sinn 1995, 1996). Testing downside risk theory requires long term data, so the paper makes use of the three longest running national socio-economic panel surveys. These are the American Panel Study of Income Dynamics (PSID, 1968-), the German Socio-Economic Panel (1984-) and the Dutch Socio-Economic Panel (SEP, 1984-). The paper focusses on prime age households (heads 25-59) and assesses their participation in and returns to adult education and job training. Our results indicate some support for the theory in so far as Dutch and German prime age adults, living in more generous welfare states, were much more likely than Americans to take the risk of foregoing current earnings and investing in additional education. In all three countries returns on investment were substantial.