For the design of pension reform it is crucial to disentangle the employment effects related to the substitution and the income effect. In this paper we provide causal evidence about the importance of the income effect which in general has been assumed to be small or non-existent. We exploit a pension reform in Germany that raised pension benefits related to children. For the identification we exploit the discontinuity induced by the reform: only mothers with children born before 1.1.1992 were affected by the pension reform. Children born after this cut-off date did not change the pension income. We use a difference-in-differences estimator based on administrative data from the German pension insurance which includes the full individual employment history. We find that income effects are significant and economically important. We show that the policy led to a reduction in employment of affected females. In addition we are able to provide evidence for effect heterogeneity on different dimensions, by treatment intensity, age of the mother and pre reform pension wealth.