In this paper, we compute distributions of rates of return by cohort for the Social Security retirement system, using a combination of historical data and stochastic forecasts of productivity and mortality rates. Since our forecasts of productivity and mortality are stochastic, the rate-of-return estimates themselves are stochastic; that is, we compute an entire distribution of rates for each cohort. We repeat these calculations under a variety of policy scenarios designed to bring the trust fund into future solvency with roughly 50% probability. This allows us to examine the impact of various schemes on different cohorts. Policies which delay reform the longest and impact taxpayers the greatest dramatically concentrate the impact of reform on the youngest cohorts. Reforms which are more immediate and focused on retirees tend to spread the cost of reform across generations more evenly.