We propose and estimate a structural retirement model with leisure preference heterogeneity using high-quality Danish register data with information about individual private pension wealth. We apply a non-parametric estimation technique to measure the heterogeneity in leisure preferences, and our estimates suggest that leisure preferences are widely distributed among the population. We use the model to study the extent to which increasing private retirement savings dampens the effect of retirement reforms targeted to increase labor supply. Our findings suggest that the effects of an increase in the normal retirement age by one year increases the average retirement age by 0.2 year if individuals hold zero or low wealth. The corresponding effect is more than halved once individuals hold savings equivalent to four or more years of pre-retirement earnings.