The paper underlines the major role of productivity as a determinant of the worker’sretirement expectations. We propose an overlapping generation model with a continuum of heterogeneous ability workers. The labor market is endogenously segmented between worker’s having the required ability level to occupy jobs where the productivity is indexed to the technological state (complex jobs) and the rest of workers who are employed in positions whose productivity will be relatively deteriorated in case of technological change (simple jobs). We show that, for a given state of technology, workers in complex positions have a latter retirement date than those in simple positions. Furthermore, in case of a positive technological change, workers in complex positions delay retirement whereas those in simple jobs advance retirement. These findings are confirmed by our empirical approach, where we find that, after a technological change, workers who benefit from a skill upgrading training program have a higher expected retirement age