Sufficient Statistics for Measuring Forward-Looking Welfare

We propose a sufficient-statistics method to measure a money-metric index of dynamic welfare. The method relies on two assumptions: (i) individuals’ preferences are separable between present and future consumption, and (ii) a subset of households bear no undiversifiable idiosyncratic risk. For this group, we infer the value of the future relative to the present from consumption-saving behavior. We then pin down the welfare of all other households by matching their budget shares. Our method accommodates incomplete markets, life-cycle motives, non-rational expectations, and non-exponential discounting—without specifying functional forms. Implementation needs only time series of prices, repeated cross-sectional income-balance-sheet- spending data, and an estimate of the intertemporal-substitution elasticity. In simulations, our method tracks true welfare better than net-present-value calculations, especially under credit constraints. Applied to the PSID, our measure differs sharply from a static index. Our estimates can be used to study the welfare effect of shocks without enumerating every margin and forecasting every horizon through which those shocks operate.