Following pension reforms aiming to reduce the demographic burden on pay-as-you-go pension schemes, Germany and many other countries introduced substantial tax benefits to set incentives for third pillar private saving. However, there is a frequently expressed concern that those who need private retirement savings the most do not save much in tax-favored retirement accounts despite high incentives, while those who are sufficiently well-off make frequent use of the favorable tax treatment. As a result, the tax benefits benefit the well-off and do not fulfill their objective of increasing the retirement savings of poorer households. The aim of the project is to study the distributional effects of such tax-favored retirement accounts using the German Taxpayer Panel, income-tax data from administrative records available for the years 2001-2014. In these data we observe everything that is necessary to identify eligible individuals in the German Riester scheme and to calculate their exact tax incentives. Further, we couple these data with simulations of hypothetical tax benefits that tax payers, who are not actually saving in the Riester scheme, could receive if they started saving. This in turn allows us to simulate counterfactual distributions of pension savings where everybody saves through the tax-favoured accounts and compare these counterfactual distributions to the actually observed distribution. In addition, we are particularly interested in the dynamics and diffusion of the retirement scheme over time as well as in the individual dynamics and characteristics of those saving in the Riester scheme. The panel structure of the data set allows us to study whether people stop to contribute or whether they fail to adjust savings to changes in income and thus fail to optimize over time.