The transmission of income shocks to consumption is typically estimated at the household level because within-household allocations are difficult to observe. Yet household-level estimates can conceal differences in individual exposure when spouses share risk imperfectly. Using spouse-level Norwegian administrative expenditure data and plausibly exogenous firm-driven income variation, we estimate the pass-through elasticity of spousal income shocks to household, personal, and joint expenditures. Household consumption responds more to women’s permanent shocks than to men’s, indicating imperfect risk-sharing. These income shocks pass strongly into the affected spouse’s personal consumption, with substantially higher pass-through for women than for men. At the same time, the effect of one spouse’s shock on the other’s personal consumption is very small. Joint expenditures, including groceries and utilities, exhibit greater pooling across spouses’ shocks. These responses are difficult to reconcile with standard models in which spouses pool all resources, and survey data we collect show financial separation is widespread among Norwegian couples. We develop a limited-commitment model with separate budgets that can reproduce our findings. Relative to full risk-sharing, our model implies lower insurance gains from marriage and higher welfare gains from public insurance.
Revise and Resubmit, Quarterly Journal of Economics
How should employer 401(k) matching formulas, which allocate $250 billion annually, be designed to raise employee saving and better target employer contributions? We combine administrative data from 1.8 million Vanguard participants with a custom survey eliciting choices under counterfactual formulas. We use these hypothetical choices to predict contributions under a broad class of formulas, including those no employer offers, and validate our survey-based predictions against actual contributions in the administrative data. We find that employee contributions are inelastic to the match rate and that non-elective contributions do not crowd out employee saving. These facts imply that the most commonly adopted formulas are dominated: at the same employer cost, alternatives with lower match rates, higher caps, and non-elective contributions can raise employee contributions while reducing inequality. We estimate that redesigning safe harbors (i.e., formulas favored by regulation) to maximize saving could generate at least $6.7 billion in additional annual employee contributions without increasing employer costs or inequality.
Revise and Resubmit, Journal of Finance (2nd Round)
Vesting requirements are a common yet understudied feature of defined-contribution retirement plans. Using administrative recordkeeping data, we find that 30% of separations occur during participants' vesting periods. The resulting forfeitures of employer contributions are concentrated among lower-income participants and make the distribution of 401(k) compensation significantly more regressive. Firms do not enjoy offsetting efficiency benefits: employing both cross-plan and within-plan identification strategies, we find no evidence that vesting exerts a causal retention effect. A linked survey shows informational frictions to be a key mechanism, as a majority of respondents do not know their current plan's vesting rules.
Expense shocks—large, unexpected household outlays—are a common source of liquidity risk. This paper studies how households absorb them by linking an original survey of more than 3,000 households to high-frequency transaction panel data. Between January 2025 and March 2026, 65% of households reported experiencing a shock, and the median shock equals 46% of monthly household income. Using a difference-in-differences design, we find that expense shocks moderately reduce food expenditures by 2.1% on average, with stronger effects among lower-income households. Consumption responses vary sharply with shock size: shocks worth half to one month of household income reduce food expenditures by 8.1%, while the effects of smaller and larger shocks are small and statistically indistinguishable from zero. This non-monotonicity partly reflects financing: households cover small shocks with cash and large shocks with loans, while medium-sized shocks are too large for liquid buffers but too small for costlier financing.
Work in progress
Elite Social Status and Protection against Economic Risk