This paper investigates the differential change in subjective well-being among Hungarians 0–3 years and 8–11 years following voluntary and involuntary retirement. Controlling for baseline individual characteristics is important to circumvent possible endogeneity problems between retirement and subjective well-being; however, voluntary and involuntary retirees correspond to considerably different sets of observed confounders, and thus regression models may be subject to interpolation and extrapolation bias. Here, we use genetic matching to improve the comparability of these two subgroups and approximate the conditions of a controlled experiment in which voluntary retirement is the treatment variable. The same regression model applied to the matched and the non-matched data leads to different results. However, the results obtained through the matching procedure are superior in terms of subgroups comparability and model performance. These results show that voluntary retirees have a higher level of subjective well-being than involuntary retirees not only in the short but also in the long-term, the latter contradicting our expectation that the two groups would converge over time.