Optimal Redistribution with Institutional Reference Points

Standard optimal tax models typically ignore reference-dependent behavior induced by institutional thresholds. This paper incorporates loss aversion into a Mirrlees op timal income tax framework to analyze how such exogenous reference points, unlike social comparisons, alter optimal redistribution. I show that institutional loss aversion calls for globally higher marginal tax rates and a quantitatively large expansion of the lump-sum transfer. To accommodate behavioral bunching at the reference point, I em ploy an ironing approach and derive a modified optimal tax formula that remains valid in the presence of a mass point. Simulations calibrated to the U.S. economy imply that the optimal lump-sum transfer increases by 19–32% and yield welfare gains equivalent to 5.8–7.5% of consumption. These results are robust under both paternalistic and non-paternalistic welfare criteria.