A retirement model defined within a push/pull framework allowing for heterogeneity in leisure preferences is estimated. The proposed estimation method applies nonparametric estimation techniques and the estimation results show evidence of bimodular population distributions of the leisure preference parameter k.
The model is estimated separately for ten gender- and education specific groups. The results suggest that men to a larger extent are pushed, while women are pulled into retirement. The estimated distributions of the pull parameter k tend to be split in a large group with relatively low values of k and a smaller group with significantly higher k values.
The estimated model assumes different interest rates such that the interest rate on deposits is constant while the interest rate on debt increases with the debt takers age. The Endogenous Grid Method (EGM) is applied to solve for optimal consumption paths. By explicitly accounting for data censoring, a potential selection bias and loss of useful information is avoided.