Using new transaction data, I find considerable deviations from consumption smoothing in response to large, regular, predetermined, and salient payments from the Alaska Permanent Fund. On average, the marginal propensity to consume (MPC) is 25% for nondurables and services within one quarter of the payments. The MPC is heterogeneous, monotonically increasing with income, and the average is largely driven by high-income households with substantial amounts of liquid assets, who have MPCs above 50%. The account-level data and the properties of the payments rule out most previous explanations of excess sensitivity, including buffer stock models and rational inattention. How big are these ‘mistakes’? Using a sufficient statistics approach, I show that the welfare loss from excess sensitivity depends on the MPC and the relative payment size as a fraction of income. Since the lump-sum payments do not depend on income, the two statistics are negatively correlated such that the welfare losses are similar across households and small (less than 0.1% of wealth), despite the large MPCs.