We empirically investigate the distributional consequences of the Riester scheme, the main private pension subsidization program in Germany. We find that 38% of the aggregate subsidy accrues to the top two deciles of the population, but only 7.3% to the bottom two. Nonetheless the Riester scheme is almost distributionally neutral when looking at standard inequality measures. This is due to two offsetting effects: a progressive one stemming from the subsidy schedule and a regressive one from voluntary participation. Regressions of the participation decision suggest that a high level of household wealth significantly increases the probability of benefiting from the Riester scheme.