We provide a decision‐theoretic analysis of bandit experiments under local asymptotics. Working within the framework of diffusion processes, we define suitable notions of asymptotic Bayes and minimax risk for these experiments. For normally distributed rewards, the minimal Bayes risk can be characterized as the solution to a second‐order partial differential equation (PDE). Using a limit of experiments approach, we show that this PDE characterization also holds asymptotically under both parametric and non‐parametric distributions of the rewards. The approach further describes the state variables it is asymptotically sufficient to restrict attention to, and thereby suggests a practical strategy for dimension reduction. The PDEs characterizing minimal Bayes risk can be solved efficiently using sparse matrix routines or Monte Carlo methods. We derive the optimal Bayes and minimax policies from their numerical solutions. These optimal policies substantially dominate existing methods such as Thompson sampling; the risk of the latter is often twice as high

The 2009 CARD Act limited credit card lenders' ability to raise borrowers' interest rates on the basis of new information. Pricing became less responsive to public and private signals of borrowers' risk and demand characteristics, and price dispersion fell by one‐third. I estimate the efficiency and distributional effects of this shift toward more pooled pricing. Prices fell for high‐risk and price‐inelastic consumers, but prices rose elsewhere in the market and newly exceeded willingness to pay for over 30% of the safest subprime borrowers. On net, average traded prices fell and consumer surplus rose at all credit scores. Higher consumer surplus was partly driven by a fall in lender profits, and partly by the Act's insurance value to borrowers who could retain favorable pricing after adverse changes to their default risk. The relatively high level of pre‐CARD‐Act markups was crucial for realizing these surplus gains.

This paper studies assortative matching in a non‐stationary search‐and‐matching model with non‐transferable payoffs. Non‐stationarity entails that the number and characteristics of agents searching evolve endogenously over time. Assortative matching can fail in non‐stationary environments under conditions for which Morgan (1995) and Smith (2006) show that it occurs in the steady state. This is due to the risk of worsening match prospects inherent to non‐stationary environments. The main contribution of this paper is to derive the weakest sufficient conditions on payoffs for which matching is assortative. In addition to known steady state conditions, more desirable individuals must be less risk‐averse in the sense of Arrow–Pratt

We estimate valence measures of candidates running in U.S. House elections from data on vote shares. Our identification and estimation strategy builds on ideas developed for estimating production functions, allowing us to control for possible endogeneity of campaign spending and sample selection of candidates due to endogenous entry. We find that incumbents have substantially higher valence measures than challengers running against them, resulting in about 3.5 percentage‐point differences in the vote share, on average. Eliminating differences in the valence of challengers and incumbents results in an increase in the winning probability of a challenger from 6.5% to 12.1%. Our measure of candidate valence can be used to study various substantive questions of political economy. We illustrate its usefulness by studying the source of incumbency advantage in U.S. House elections.

We recover valuations of time using detailed data from a large ride‐hail platform, where drivers bid on trips and consumers choose between a set of rides with different prices and wait times. Leveraging a consumer panel, we estimate demand as a function of both prices and wait times and use the resulting estimates to recover heterogeneity in the value of time across consumers. We study the welfare implications of personalized pricing and its effect on the platform, drivers, and consumers. Taking into account drivers' optimal reaction to the platform's pricing policy, personalized pricing lowers consumer surplus by 2.5% and increases overall surplus by 5.2%. Like the platform, drivers benefit from personalized pricing. By conditioning prices on drivers' wait times and not on consumers' data, the platform can capture a significant portion of the profits garnered from personalized pricing, and simultaneously benefit consumers.

Policymakers often choose a policy bundle that is a combination of different interventions in different dosages. We develop a new technique— treatment variant aggregation (TVA)—to select a policy from a large factorial design. TVA pools together policy variants that are not meaningfully different and prunes those deemed ineffective. This allows us to restrict attention to aggregated policy variants, consistently estimate their effects on the outcome, and estimate the best policy effect adjusting for the winner's curse. We apply TVA to a large randomized controlled trial that tests interventions to stimulate demand for immunization in Haryana, India. The policies under consideration include reminders, incentives, and local ambassadors for community mobilization. Cross‐randomizing these interventions, with different dosages or types of each intervention, yields 75 combinations. The policy with the largest impact (which combines incentives, ambassadors who are information hubs, and reminders) increases the number of immunizations by 44% relative to the status quo. The most cost‐effective policy (information hubs, ambassadors, and SMS reminders, but no incentives) increases the number of immunizations per dollar by 9.1% relative to the status quo.