In the early 1990s the Dutch labour unions and employer organisations agreed to transform the generous and actuarially unfair early retirement (ER) schemes into less generous and actuarially fair schemes that reward individuals for postponing retirement. The starting dates of these new ER programs varied by industry sector. In this study, we exploit this variation in starting dates to estimate the causal impact of the policy reform on early retirement behaviour. We use a large administrative dataset, the Dutch Income Panel 1989–2000, to estimate hazard rate models for the retirement age. We conclude that the policy reform has indeed induced workers to postpone retirement. Both the wealth effect (lower ER wealth) and the substitution effect (lower implicit taxes on retirement postponement) are significant, the latter being more substantial.