Sufficient Statistics for Measuring Forward-Looking Welfare

We provide a sufficient-statistic method for measuring forward-looking welfare in money-metric terms. The method adapts the money metric to dynamic environments with uncertainty, incomplete markets, and borrowing constraints. Under time separability and using households with negligible idiosyncratic income risk as a reference group, we infer the welfare-relevant value of future consumption from consumptionsaving behavior, given an estimate of the intertemporal elasticity of substitution (EIS). In simulations disciplined by PSID income dynamics, the method tracks true welfare more accurately than net-present-value calculations. Applying the method to PSID households from 2005 to 2019, we estimate dynamic cost-of-living indices below the static CPI benchmark. This is because, controlling for age and real wealth, unconstrained households in 2019 spend a greater share of wealth on consumption than comparable households in 2005. With an EIS below one, this implies a lower perceived cost of future consumption relative to present consumption. We also illustrate how the resulting welfare measure can be used as an outcome in reduced-form work by estimating welfare losses associated with job loss.