Consumption smoothing at retirement: Average and quantile treatment effects in the regression discontinuity design

Abstract Standard economic models predict that individuals smooth consumption over the life cycle. In contrast, there exists controversial empirical evidence showing that consumption declines at retirement. This paper investigates whether there is evidence for this so-called Retirement Consumption Puzzle in Switzerland. Baseline regression discontinuity estimates of average treatment effects are complemented by quantile treatment effects, where all estimates take the potential endogeneity of retirement into account. The findings suggest that disposable income significantly decreases after retirement, although there is substantial treatment effect heterogeneity. The reduction in income transmits to a negative but considerably less pronounced effect on overall consumption expenditures, indicating that households simultaneously adjust their savings. The results further show that food consumption at home is not or even positively affected by retirement, whereas expenditures in restaurants and hotels significantly decline.