Employment to population ratios differ markedly across OECD coun- tries, especially for people over 55. Social security features also differ markedly across the OECD, particularly with respect to replacement rates, entitlement ages and earnings tests. I conjecture that differences in social security features explain many differences in employment to population ratios at older ages. I assess my conjecture quantitatively with a life cycle general equilibrium model of retirement. At ages 60-64 the correlation between my model’s simulations and observed data is .67. Replacement rates and the earnings tests are key features.