Research

This paper studies the welfare and distributional effects of dynamically priced highway toll lanes. To quantify the equilibrium effects of tolling, we develop and estimate a model of driver demand, the road technology, and the pricing algorithm. The demand model features heterogeneous drivers who choose their departure time under imperfect information about travel times and prices and then, once the uncertainty is resolved, choose whether to take the priced (faster) or unpriced (slower) lanes. We estimate the model using data on toll transactions, historical traffic conditions, and driver characteristics for I-405 in Washington State. We find that tolling a subset of highway lanes increases welfare in aggregate, as well as for drivers in all income quartiles, especially when the newly priced lanes were previously carpool-only. A key component of these gains is the “option value” of tolling: even drivers who infrequently take the priced lanes benefit from having the option to pay for speed when traffic is worse than expected. Moreover, the largest gains accrue to drivers in the bottom income quartile, primarily due to the spatial distribution of lower- and higher-income drivers rather than preference heterogeneity. Lower-income drivers have longer I-405 commutes and—thanks to the design of the pricing algorithm—often face an advantageous tradeoff between time savings and price. Finally, we show how simple revisions to the pricing algorithm can increase aggregate welfare and help achieve redistributive goals.

How does the location of affordable housing affect tenant welfare, the distribution of assistance, and broader social objectives such as racial and economic integration? Using administrative data on households living in units funded by the Low-Income Housing Tax Credit (LIHTC), we first show that tenant characteristics such as income, race, education, and family structure vary widely across neighborhoods, despite common eligibility thresholds. To quantify the welfare implications, we develop and estimate a residential choice model in which households choose from both market-rate and affordable housing options, where the latter are priced below-market and must be rationed. Moving a new development to a neighborhood with less poverty and better access to good schools and jobs increases aggregate tenant welfare and reduces both racial and economic segregation. However, it is also more costly to provide and disproportionately benefits more moderate-need, non-Black/Hispanic households. This change in the distribution of assistance arises in part because of the rationing process: households that only apply for assistance in opportunity-rich neighborhoods crowd out other households willing to apply anywhere. Relative to the choice of where to build, policy levers available post-construction—such as lowering the eligibility thresholds—have only limited effects on outcomes.

Revealed preference arguments are commonly used when identifying models of both single-agent decisions and non-cooperative games. We develop general identification results for a large class of models that have a linearly separable payoff structure. Our model allows for both discrete and continuous choice sets. It incorporates widely studied models such as discrete and hedonic choice models, auctions, school choice mechanisms, oligopoly pricing and trading games. We characterize the identified set and show that point identification can be achieved either if the choice set is sufficiently rich or if a variable that shifts preferences is available. Our identification results also suggest an estimation approach. Finally, we implement this approach to estimate values in a combinatorial procurement auction for school lunches in Chile.

The years following the Great Recession were challenging for forecasters. Unlike other deep downturns, this recession was not followed by a swift recovery, but generated a sizable and persistent output gap that was not accompanied by deflation as a traditional Phillips curve relationship would have predicted. Moreover, the zero lower bound and unconventional monetary policy generated a policy environment without precedents. We document the actual real-time forecasting performance of the New York Fed dynamic stochastic general equilibrium (DSGE) model during this period and explain the results using the pseudo real-time forecasting performance results from a battery of DSGE models. We find the New York Fed DSGE model’s forecasting accuracy to be comparable to that of private forecasters—and notably better, for output growth, than the median forecasts from the FOMC’s Summary of Economic Projections. The model’s financial frictions were key in obtaining these results, as they implied a slow recovery following the financial crisis.

Unlike many other goods allocated through centralized mechanisms, affordable housing is durable: who receives a unit today affects the supply of units available to reallocate in the future. We build a dynamic model of the allocation mechanism that endogenizes the arrival rate of vacant units. Households in the model make decisions on both whether to apply and, if allocated a unit, whether to move out each period. Policy changes that affect the move-out rate (e.g., giving households more choice when applying or allowing tenants to swap units) lead to a tradeoff between providing longer stays in subsidized housing to fewer households or shorter stays to more households. Optimal policy depends on dynamic considerations such as how match quality, need, and any treatment effects on households evolve over time. To take the theory to the data, we are working on a data-sharing agreement with New York City, where units are allocated through centrally-run lotteries.

Empirical research suggests that women see fewer employment ads than men online, at least in part because they’re actually more attractive to many advertisers. Perversely, greater competition for women’s attention may cause them to see fewer of the ads that they may value more. In this paper, I propose an alternative ad allocation mechanism that is robust to this phenomenon, which I call market-induced discrimination. First, I solve for the mechanism which maximizes producer (advertiser and platform) surplus subject to a fairness constraint. The optimal fair mechanism is a second-score auction in which job advertisers’ bids on men (women) are effectively taxed (subsidized). Next, I extend the model to include consumer preferences and surplus. I show that under certain realistic conditions, the fair mechanism achieves a greater total social surplus than the (producer surplus-maximizing) VCG mechanism. I use this result to argue that the fair mechanism is a tractable alternative to the socially optimal mechanism.