ABSTRACT
I investigate heterogeneity across occupational characteristics in the effect of retirement eligibility on mental health in the United Kingdom. I use K-means clustering to define three occupational clusters, differing across multiple dimensions. I estimate the effect of retirement eligibility using a Regression Discontinuity Design, allowing the effect to differ by cluster. The effects of retirement eligibility are beneficial, and greater in two clusters: one comprised of white-collar jobs in an office setting and another of blue-collar jobs with high physical demands and hazards. The cluster with smaller benefits mixes blue- and white-collar uncompetitive jobs with high levels of customer interaction. The results have implications for the distributional effect of raising the retirement age.