According to the individualization of the life-course hypothesis, a largest set of institutionalized options available should facilitate individuals in realizing their preferences for the timing of life-course transitions. This study contributes to the literature by considering differences by gender and education (and their interaction) in the ability to fulfil preferences for early retirement across welfare states. We use longitudinal data from the SHARE survey as it includes information on preferences for early retirement expressed before actual retirement. Wwe therefore avoid biases driven by the assessment of preferences for the timing of retirement after it occurred. We estimate discrete time regression models and find that positive preferences for early retirement are associated with an actual anticipation of retirement with respect to the statutory age. Although the size of the effect is small, it remains statistically significant even after models are adjusted for a number of potential confounding factors. However, no differences by gender and education (or their interaction) exist in the strength of the correspondence between preferences and behaviour. Finally, these results do not vary across welfare regimes. This evidence suggests that the timing of the transition to retirement is only marginally shaped by preferences and it is still strongly institutionalized, being a by-product of “agency-within-structure”” mechanisms rather than an individualized process.