Publications

This paper identifies which investments in school facilities help students and are valued by homeowners. Using novel data on school district bonds, test scores, and house prices for 29 U.S. states and a research design that exploits close elections with staggered timing, we show that increased school capital spending raises test scores and house prices on average. However, impacts differ vastly across types of funded projects. Spending on basic infrastructure (such as HVAC) or on the removal of pollutants raises test scores but not house prices; conversely, spending on athletic facilities raises house prices but not test scores. Socio-economically disadvantaged districts benefit more from capital outlays, even conditioning on project type and the existing capital stock. Our estimates suggest that closing the spending gap between high- and low-SES districts and targeting spending towards high-impact projects may close as much as 25% of the observed achievement gap between these districts.

In collaboration with Feeding America, we aim to redesign Real-Time—a tool on its food sourcing and rescue platform, MealConnect—that facilitates the connection of ad-hoc, time-sensitive food donations to local agencies (e.g., meal programs) through an offer process. In making offer decisions, Real-Time and similar food rescue platforms face a challenge in balancing efficiency and equity due to heterogeneity in response rates across agencies: offering to many agencies upfront improves efficiency (the likelihood of acceptance) but may disadvantage those with lower response rates. On top of this, the ad hoc nature of donations introduces future uncertainty, adding to the challenge of achieving the dual goals of efficiency and equity. Motivated by these challenges, we study a sequential offer scheduling problem in which donations arrive sequentially and are connected through a multi-stage offer process. The goal is to maximize an objective that balances efficiency and equity, promoting allocations proportional to agencies needs. We first develop a dynamic programming (DP)-based algorithm that optimally solves the one-donation problem and yields an intuitive nested offer schedule. Moving beyond one donation and motivated by the ad hoc nature of donations, we take a robust approach to designing sequential offer policies that do not rely on any knowledge about future donations. We design a penalty-based offer policy that solves a modified one-donation DP by properly penalizing current allocation to hedge against future uncertainty. We establish strong (and optimal in asymptotic regimes) performance guarantees for our policy. We further demonstrate the importance of hedging: a greedy policy that solves each one-donation DP without accounting for future arrivals fails to achieve a comparable guarantee. Numerical results on real data from Feeding America’s MealConnect platform demonstrate that our proposed approach significantly improves both efficiency and equity relative to current practice and several benchmarks.

This study examines how and why the effectiveness of business and social restrictions evolved during the first year of the COVID-19 pandemic, using weekly county-level data from the United States in 2020. Panel data regressions with county-level fixed effects reveal that business restrictions became less effective the longer they remained in place. Residents progressively spent less time at home under sustained restrictions, with this decline accelerating when neighboring counties maintained more lenient policies. The analysis further shows that any initial correlation between policy stringency and COVID-19 fatalities, whether positive or negative, dissipated over time. Estimates indicate this erosion occurred within 20 weeks for many interventions. To explain this declining effectiveness, several hypotheses are tested. The data most strongly support "lockdown fatigue," whereby individuals became progressively less willing to modify their behavior in response to government restrictions over time. This explanation better accounts for the observed patterns than alternative mechanisms, such as improvements in mitigation technology or adaptive behaviors. Point estimates suggest that business restrictions became disassociated from deaths after four to five months, yet many jurisdictions maintained them considerably longer. More broadly, the findings indicate that certain policies were either ineffective from day 1 or counterproductive, calling into question their use during the pandemic response.

AI is increasingly used to guide high-stakes educational decisions, yet its effectiveness depends on whether people follow its advice. We present the first large-scale field experiment with adolescents in this context, conducted across 14 public high schools. We compare the adoption of identical college-application recommendations from human counselors versus an AI-based algorithm. Contrary to the common assumption that objective, data-based recommendations favor AI, we find that algorithm aversion intensifies when recommendations are based on objective criteria (e.g., grades, admission chances) and dissipates when criteria are more subjective. We find that student perceptions of the recommender's intent strongly drive this aversion, consistently across scenarios and statistical approaches; perceptions of alignment with personal goals, ability, and comprehension also play significant roles. The results further reveal substantial heterogeneity in recommendation adoption rates. We observe stronger aversion among female students, students from rural schools, lower-GPA students, and those with stronger prestige-seeking attitudes. Using an optimization approach, we demonstrate how a policymaker can navigate the heterogeneity in recommendation adoption rates to optimally prioritize the assignment of human versus AI-based algorithmic recommenders, under varying social priorities and limited capacity of human counselors. We find that a targeting policy relying on few readily available student and school features can approximate the first-best, personalized targeting policy effectively. These insights underscore the importance of understanding student preferences and trust in an AI system's intent, and of adopting hybrid approaches that blend human guidance with AI tools to design effective recommendation systems.

Many industries, including the NFL with the Rooney Rule and law firms with the Mansfield Rule, have adopted interview-stage diversity interventions requiring a minimum representation of disadvantaged groups in the interview set. However, the effectiveness of such policies remains inconclusive. In light of this, we develop a framework of a two-stage hiring process, where rational firms, with limited interview and hiring capacities, aim to maximize the match value of their hires. The labor market consists of two equally sized social groups, m and w, with identical ex-post match value distributions. Match values are revealed only post-interview, while interview decisions rely on partially informative pre-interview scores. Pre-interview scores are more informative for group m, while interviews reveal more for group w; as a result, if firms could interview all candidates, both groups would be equally hired. However, due to limited interview capacity and information asymmetry, we show that requiring equal representation in the interview stage does not translate into equal representation in the hiring outcome, even though interviews are more informative for group w. In certain regimes, with or without intervention, a firm may interview more group w candidates but still hire fewer. At an individual level, we show that strong candidates from both groups benefit from the intervention as the candidate-level competition weakens. For borderline candidates, only group w candidates gain at the expense of group m. To understand the impact of non-universal interview-stage interventions on the market, we study a model with two vertically differentiated firms, where only the top firm adopts the intervention. We characterize the unique equilibrium and demonstrate potentially negative effects: we show that in certain regimes the lower firm hires fewer group w candidates due to increased firm-level competition for them, and further find examples where overall fewer group w candidates are hired across the market. At an individual level, while superstar candidates in both groups benefit, surprisingly the impact on borderline candidates may reverse: the lower firm may replace borderline group w candidates with borderline group m candidates in its interview set, effectively reducing the chance of those borderline group w candidates being hired. Overall, our findings highlight challenges in diversifying the labor market at early hiring stages due to information asymmetry, filtering, and competition. Beyond our context, our natural framework of a market with two-stage hiring may be of independent interest.

Governments use their countries’ economic strength from financial and trade relationships to achieve geopolitical and economic goals. We provide a model of the sources of geoeconomic power and how it is wielded. The source of this power is the ability of a hegemonic country to coordinate threats across disparate eco- nomic relationships as a mean of enforcement on foreign entities. The hegemon wields this power to demand costly actions out of the targeted entities, including mark-ups, import restrictions, tariffs, and political concessions. The hegemon uses its power to change targeted entities’ activities to manipulate the global equilib- rium in its favor and increase its power. A sector is strategic either in helping the hegemon form threats or in manipulating the world equilibrium via input-output amplification. The hegemon acts a global enforcer, thus adding value to the world economy, but destroys value by distorting the equilibrium in its favor.

We develop a tractable dynamic contracting framework to study bank bail-in regimes. In the presence of a repeated monitoring problem, the optimal bank capital structure combines standard debt, which induces liquidation and provides strong incentives, and bail-in debt, which restores solvency but provides weaker incentives. When there are fire sales, optimal policy entails joint regulation: a bail-in regime reduces standard debt while leverage regulation reduces total debt. Bail-ins replace bailouts as a recapitalization tool.

We develop an economic model of autonomous vehicle (AV) ride-hailing markets, in which uncertain aggregate demand is served with a combination of a fixed fleet of AVs and a flexible pool of human drivers (HVs). Dispatch efficiencies increase with scale because of density effects. We analyze market outcomes in this setting under four market configurations, defined by two dispatch platform structures (common platform versus independent platforms) and two levels of supply competition (monopoly AV versus competitive AV). A key result of our analysis is that the lower cost of AVs does not necessarily translate into lower prices; the price impact of AVs is ambiguous and depends critically on both the dispatch platform structure and the level of AV supply competition. In the extreme case, we show that if AVs and HVs operate on independent dispatch platforms, there is a monopoly AV supplier, and labor supply elasticity is sufficiently high, then prices are even higher than in a pure-HV market. Indeed, to guarantee consistently lower prices (relative to a pure HV market) in all scenarios and under all supply and density elasticities, a common dispatch platform between AVs and HVs is required. Furthermore, competitive AVs lead to lower prices than monopoly AVs in every such scenario. Our results illustrate the critical role that market configuration plays in realizing potential welfare gains from AVs.

This paper examines the welfare implications of third-party informational interme- diation. A seller sets the price of a product that is sold through an intermediary, who discloses information about the product to consumers. In a model where the inter- mediary is consumer-minded—has a payoff that depends on both the seller’s revenue and the consumer surplus, we show that total welfare may decrease in the Pareto sense, as the intermediary’s consumer-mindedness increases. Furthermore, we show that consumer-mindedness emerges endogenously when a revenue-maximizing interme- diary is forward-looking and the consumer base is increasing in past consumer surplus.

Traditional measures of mutual fund skill focus on portfolio activity or performance but offer little insight into how managers select stocks. To decipher this managerial skill, we employ Large Language Models (LLMs) to measure the semantic distance between mutual fund prospectuses and the strategic priorities outlined in firms' 10-K filings, which captures difficult-to-understand information that requires managerial expertise to process. Funds with high traditional skill measures outperform unskilled peers only when they invest heavily in distant stocks. Moreover, distant investments made by skilled funds predict future stock returns and enhance market efficiency. Our findings reveal a novel mechanism behind managerial skills.

This article develops a method to automatically discover and quantify human-interpretable visual characteristics directly from product image data. The method is generative, and can create new visual designs spanning the space of visual characteristics. It builds on disentanglement methods in deep learning using variational autoencoders, which aim to discover underlying statistically independent and interpretable visual characteristics of an object. The impossibility theorem in the deep learning literature indicates that supervision with ground truth characteristics would be required to obtain unique disentangled representations. However, these are typically unknown in real world applications, and are in fact exactly the characteristics we want to discover. Extant machine learning methods require ground truth labels for each visual characteristic, resulting in a task requiring human evaluation and judgment to both design and operationalize. In contrast, this method postulates the use of readily available product characteristics (such as brand and price) as proxy supervisory signals to enable disentanglement. This method discovers and quantifies human-interpretable and statistically independent characteristics without any specific domain knowledge on the product category. It is applied to a dataset of watches to automatically discover interpretable visual product characteristics, obtain consumer preferences over visual designs, and generate new ideal point designs targeted to specific consumer segments.

Emergency Department (ED) crowding is recognized as a national crisis. Load-leveling is the process of transferring patients between campuses in the same hospital system for admission to redistribute patient capacity. Our study evaluates the impact of this load-leveling on ED throughput from the transferred patient perspective and on ED operations to evaluate operational benefits and drawbacks.

This paper examines which municipal disclosures provide informational value to investors. Using the entire universe of post-issuance financial and event disclosures from 2009 to 2022, we find that most municipal bonds do not trade in the weeks following a disclosure. However, some disclosures do provide enough new information to increase trading. Investors trade more on credit-relevant disclosures, such as adverse credit event disclosures, and less on required annual financial statements. Trading after disclosures also increases more when a bond is large or risky. Moreover, we find that credit rating agencies do respond to disclosures, lending support to the idea that some disclosures have informational value. In further analyses, we find that trading before the disclosure, lack of timeliness, illiquidity, and information processing constraints contribute to the limited trading on the average disclosure. The findings suggest that reconsidering a one-size-fits-all approach to regulating post-issuance municipal disclosures may be worthwhile.

We propose a novel, theory-based approach to the reinforcement learning problem of maximizing profits when faced with an unknown demand curve. Our method, rooted in multi-armed bandits, balances exploration and exploitation across various prices (arms) to maximize rewards. Traditional Gaussian process bandits cap- ture one informational externality in price experimentation – correlation of rewards through an underlying demand curve. We extend this framework by incorporating a second externality, monotonicity, into Gaussian process bandits by introducing monotonic versions of both the GP-UCB and GP-TS algorithms. Through reduction of the demand space, this informational externality limits exploration and experimentation, out- performing benchmarks by enhancing profitability. Moreover, our approach can also complement methods such as partial identification. Additionally, we present algorithm variants that account for heteroscedastic noise in purchase data. We provide theoretical guarantees for our algorithm, and empirically demonstrate its improved performance across a broad range of willingness-to-pay distributions (including discontinuous, time-varying, and real-world) and price sets. Notably, our algorithm increased profits, especially for distribu- tions where the optimal price lies near the lower end of the considered price set. Across simulation settings, our algorithm consistently achieved over 95% of the optimal profits.

We examine political affiliation’s role in venture team formation and success. Using data from Crunchbase and L2 on 1,125 US-based startups, we investigate political homophily in team assembly and its association with startup outcomes. Our analysis reveals strong political homogeneity in founding teams: teams with similar political views form more frequently than diverse teams, even after controlling for founders’ gender, age, location, and industry. This political homophily relates to venture performance. Startups with politically heterogeneous founding teams are more likely to shut down. Across additional performance measures (capital funding, employee size, Crunchbase rankings), we observe directionally consistent associations with worse outcomes, though these secondary findings vary in robustness. These findings highlight the dual role of founders’ political affiliations: their relationship with team composition and startup performance.

We examine how political heterogeneity in groups affects individual performance in settings where people work alongside others. Leveraging the random assignment of golfers to groups in Professional Golfers’ Association Tour tournaments, we find that golfers score 0.2 strokes better per round when playing in politically homogeneous versus heterogeneous groups. This corresponds to a five-rank improvement before the tournament cut and an additional $13,000–$23,400 in tournament earnings. The effect intensifies during periods of high societal political polarization and diminishes when polarization is low. We propose that politically heterogeneous groups create a more stressful and less psychologically safe environment, reducing focus and leading to reduced performance. Consistent with this mechanism, analyses of shot-level data reveal that this effect is strongest during driving and putting shots when players are in close physical proximity. Our study contributes to the understanding of how political heterogeneity in groups affects individual performance in competitive settings, with implications for managing ideological differences in organizations.

This paper studies access to services across geographical regions, using both theoretical and empirical analyses. We model and examine the effects of economies of density in ridesharing markets. Our model predicts that (i) economies of density skew access to rideshareing service away from less dense regions, (ii) the skew will be more pronounced for smaller platforms (i.e., “thinner markets”), and (iii) rideshare platforms do not find this skew efficient and thus use prices and wages to mitigate (but not eliminate) it. We show that these insights are robust to whether the source of economies of density is the supply-side or the demand-side. We then calibrate our model using ride-level Uber data from New York City. We devise an identification strategy based on relative flows of rides among regions which allows us to infer unobsrevable potential demand in different boroughs. We use the model to simulate counterfactual scenarios providing insights on platform optimal pricing with and without spatial price discrimination, the role of market thickness, the impact of prices/wages on access to rides, and the effects of minimum-wage regulations on access equity across regions.

This paper proposes a new measure of technological obsolescence using detailed patent data. Using this measure, we present two sets of results. First, firms' technological obsolescence foreshadows substantially lower growth, productivity, and reallocation of capital. This finding applies mainly for obsolescence of core innovation and embodied innovation, and it is stronger in competitive product markets. Second, in stock markets, high-obsolescence firms under-perform low-obsolescence firms by 7 percent annually. Using analyst forecast data, we show this is due to a systematic overestimation of future profits of obsolescent firms. The measure contains incremental information about firm innovation relative to measures focusing on new innovation.