Publications

Objectives Previous studies have reported inconsistent findings regarding the relationship between COVID-19 restrictions and suicide rates, particularly concerning business and social restriction policies. This study aimed to address this gap by analyzing detailed US county-level restriction and suicide death data. Study Design Data from the US Centers for Disease Control and Prevention (CDC) were obtained for county-level suicide rates by race, sex, and age from 2016 to 2023. Yale School of Management-Tobin Center State and Local COVID Restriction Database provided data on COVID-19 social and business restrictions. These datasets were combined with other relevant data on county-level demographics, gross domestic product (GDP), unemployment, and population density. Methods Poisson interrupted time-series regression was employed to assess whether these restrictions were associated with changes in suicide rates during pandemic (2020 and 2021) and post-pandemic (2022 and 2023) periods. Results During the pandemic restriction era of 2020–2021, stricter business capacity limits were linked to lower suicide rates overall (Poisson coefficient: -0.90 [95% CI -1.54, -0.25, p = 0.006]), and in particular among males (Poisson coefficient: -1.13 [-1.94, -0.32, p = 0.006]). The estimated coefficient was not statistically significant for females. Among age groups, individuals aged 25–34 and 35–44 experienced lower suicide rates in counties with tighter restrictions, while other age and sex groups did not show similar trends. Additionally, no statistically significant correlations were found across racial groups. In contrast, social gathering restrictions had a less consistent relationship with suicide rates; while those aged 15–24 experienced an increase in rates under tighter restrictions, those aged 25–34 had a decrease. No other demographic groups yielded statistically significant coefficient estimates. Conclusions Results underscore the importance of considering differential effects of business and social restrictions on suicide rates, and to tailor interventions to address the unique needs of specific populations during public health crises.

Local, state, and federal governments offer firms combinations of tax incentives and financial intermediation to help attract or retain jobs and investment for their constituents. Tax breaks are often implemented as place-based policies (PBPs), for which firms must allocate resources to a particular locality to maximize subsidy amounts. Tax instruments underlying PBPs can take many forms, including tax breaks for specific firms in critical sectors, broad-based subsidies for hiring and capital expenditures, industrial policies which operate through intergovernmental development plans, and local revitalization programs targeting neighborhoods which appear to be distressed based on measures such as unemployment and poverty rates. While there is a large body of research examining the equity-efficiency tradeoffs inherent in PBPs based on aggregated real economic outcomes and via quantitative spatial models, less is known about how firms alter their production processes and corporate strategy in responding to policy nudges. Data limitations, especially in contexts with small, privately held firms, prevent comprehensive analyses of these margins of adjustment. On the labor side, firms can use subsidies to engage in labor hoarding; for multi-plant firms, PBPs induce firms to shift the spatial distribution of worker skills within the firm's internal network, with implications for regional inequality. Firms also alter their investment plans over time, across space, and between physical and intangible capital inputs; funds obtained through place-based programs may substitute for external financing sources. Metrics for scoring PBPs aimed at firms range from ex post partial equilibrium cost per job or general equilibrium NPV calculations to ex ante criteria based on compatibility of firms' incentives with policymakers' objectives and the scope for welfare losses from inter-jurisdictional tax competition. Large variation in the same metric across existing studies focusing on the same type of corporate tax instrument underscores challenges in extrapolating the successes and failures of any one PBP into general policy design principles.

Public disclosure regimes seek to align regulated agents' incentives and improve downstream decision-making, but adverse signals can conflate useful information with noise arising from luck and circumstance. We study this tradeoff in the context of registered advisers' personal financial distress disclosures, examining their labor-market effects and informational value. We find greater job separation following disclosures (13.9% and 17.6% increases relative to baseline separation rates following bankruptcy and other financial disclosures, respectively), comparable to the effects of adviser misconduct disclosures, suggesting that downstream actors infer broad performance concerns even from weakly informative signals. Using FOIA-obtained termination-reason data, we show that these distress-related separations largely reflect involuntary discipline rather than voluntary exit. Exploiting the regime's fixed ten-year disclosure limit, we find a discrete improvement in advisers' labor-market mobility once older disclosures cease to be publicly visible, suggesting that ongoing public visibility shapes labor-market outcomes. We also find evidence that these disclosures contain some information about adviser quality but also substantial noise arising from external shocks. Financial disclosures predict future misconduct, but much more weakly than misconduct disclosures, and distress events linked to medical emergencies and prior job loss have little predictive value despite substantial labor-market penalties. Consistent with this interpretation, increases in local housing wealth reduce disclosure likelihood, highlighting the role of shocks in generating distress. Black and Hispanic advisers exhibit higher disclosure rates and greater sensitivity of disclosure risk to housing-wealth shocks, implying unequal exposure to the costs of mandatory disclosure while benefits accrue more broadly to clients and markets. Taken together, our findings suggest that disclosure regimes can generate collateral consequences when beneficiaries cannot fully distinguish informative signals from distress arising from luck and circumstance.

We study the role of information and access in capacity-constrained selection problems with fairness concerns. We develop a statistical discrimination framework, where each applicant has multiple features and is potentially strategic. The model formalizes the tradeoff between the (potentially positive) informational role of a feature and its (negative) exclusionary nature when members of different social groups have unequal access to this feature. Our framework finds a natural application to policy debates on dropping standardized testing in admissions. Our primary takeaway is that the decision to drop a feature (such as test scores) cannot be made without the joint context of the information provided by other features and how the requirement affects the applicant pool composition. Dropping a feature may exacerbate disparities by decreasing the amount of information available for each applicant, especially those from nontraditional backgrounds. However, in the presence of access barriers to a feature, the interaction between the informational environment and the effect of access barriers on the applicant pool size becomes highly complex. Furthermore, we consider an extension with two schools and costly tests, where strategic students decide whether to take the test or not. Our theoretical results reveal that the students’ test-taking behavior can be nonmonotonic. We characterize the two-school policy equilibria and show that each school’s optimal decision to drop the test critically depends on the other school’s test policy. Finally, using calibrated simulations, we demonstrate the presence of practical instances where the decision to eliminate standardized testing improves or worsens all metrics.

Motivated by our collaboration with a major refugee resettlement agency in the United States, we study a dynamic matching problem where each new arrival (a refugee case) must be matched immediately and irrevocably to one of the static resources (a location with a fixed annual quota). In addition to consuming the static resource, each case requires postallocation service from a server, such as a translator. Given the time-consuming nature of service, a server may not be available at a given time, thus we refer to it as a dynamic resource. Upon matching, the case will wait to avail service in a first-come-first-serve manner. Bursty matching to a location may result in undesirable congestion at its corresponding server. Consequently, the central planner (the agency) faces a dynamic matching problem with an objective that combines the matching reward (captured by pair-specific employment outcomes) with the cost for congestion for dynamic resources and overallocation for the static ones. Motivated by the observed fluctuations in the composition of refugee pools across the years, we design algorithms that do not rely on distributional knowledge constructed based on past years’ data. To that end, we develop learning-based algorithms that are asymptotically optimal in certain regimes, easy to interpret, and computationally fast. Our design is based on learning the dual variables of the underlying optimization problem; however, the main challenge lies in the time-varying nature of the dual variables associated with dynamic resources. To overcome this challenge, our theoretical development brings together techniques from Lyapunov analysis, adversarial online learning, and stochastic optimization. On the application side, when tested on real data from our partner agency and incorporating practical considerations, our method outperforms existing ones, making it a viable candidate for replacing the current practice upon experimentation.

In late 2018, GigaCloud CEO Larry Wu (Yale MBA, 2002) sat reviewing his company’s operating metrics with senior managers at the company’s headquarters in Los Angeles. Wu and his team had begun selling furniture from Asia to the US in 2014. As a pioneer in cross-border e-commerce, Wu’s company, GigaCloud, procured directly from Asian manufacturers, taking advantage of the rich SKU assortment and competitive price to attract a wide array of U.S. customers.

All went well until mid-2018 when demand suddenly dipped amid higher interest rates and rising tariffs. Not only did demand decline, but the Asian manufacturers also started to sell directly to other U.S. retailers to stay profitable during the challenging time. Operating income slipped; the company went from earning hundreds of thousands of dollars per month to going into the red by August.

Reviewing the sales data and the company’s cash position, the senior management team agreed to sell inventory at a loss to pull GigaCloud through the crisis. Then, looking to the future, Wu proposed a rather radical pivot: build a B2B service business in addition to the company’s existing B2C business by attracting Asian manufacturers to offer their wares through the GigaCloud website to other U.S. retailers. Many of GigaCloud’s senior management team were opposed to the proposal, believing the service would cause GigaCloud’s retail business to collapse. These managers thought GigaCloud could weather the storm and return to the business as usual – nothing fundamental needed to change.

However, Wu believed the business would inevitably run into trouble down the road, and it was the right time to make some fundamental changes. Wu and his team debated the sharp change in the business model late into the night. Even if the company decided to shift its business model, how would GigaCloud implement the change?
 

Conversational recommender systems powered by generative AI can enhance personalization by facilitating information elicitation through follow-up questions. However, engaging in these conversations imposes a communication cost on users. As platforms with different objectives and monetization models deploy these systems, a central question is: how does the platform’s objective and sellers’ strategic response shape the design of these systems in terms of their elicitation strategy? We develop a parsimonious model of conversational elicitation in which interaction generates noisy preference information and imposes a communication cost borne by the user. A user-welfare-maximizing platform elicits more information when accurate niche matching yields large gains, even when niche users are rare. In contrast, under a conversion objective, for the same setting, the optimal strategy is to immediately recommend the same mainstream option to all users with no or minimal preference elicitation because the incremental conversion benefit from improved matching is bounded, while communication costs are borne by all users. When prices are endogenous and the platform earns a commission, increased elicitation is again optimal because improved screening raises equilibrium prices and platform revenue; however, these price responses can counteract consumer benefits and reduce user welfare. The model also highlights that the optimal elicitation intensity increases with preference heterogeneity, helping explain why conversational systems ask more in highly differentiated categories than in low-heterogeneity ones. We complement the theory with a dataset of long-form product queries that vary in length and informational content. Using our dataset and LLM-based user simulation, we quantify how additional information impacts user decisions and demonstrate that the magnitude of this impact depends on the degree of preference heterogeneity. Additionally, this dataset provides a testbed for measuring the (incremental) value of preference elicitation and may be of independent interest.

Public discourse often framed artificial intelligence (AI) as a threat to jobs and livelihoods. Yet in the social-services sector, AI held the potential to raise pay and ease workloads for frontline care teams. This insight was not lost on the team at Lawrence Hall.

Founded in 1865, Lawrence Hall had grown into a leading child and family services agency in Chicago, providing foster and residential care for youth. Originally an orphanage, it evolved through eras of social and technological change into a trauma-informed non-profit with programs for young people across the city. 

Central to its model: residential facilities where young people (ages 8–21) lived and received round-the-clock support. Best practice in child welfare was to keep youth safely in family or family-like settings, reserving residential care as a last-resort intervention. Yet in the U.S., more than 30% of teenagers in foster care were housed in group or institutional settings. Once in residential care, the average youth had little chance at adoption. Of those older than 13, nearly half remained in state custody until “aging out” of the system at age 21. 

For youth in residential care, enduring, supportive adult relationships are among the strongest predictors of positive outcomes. Yet the child welfare system was plagued by employee turnover. Residential staff managed heavy caseloads and administrative demands, while earning modest wages in resource-constrained facilities. The ensuing cycle of departures came at a substantial cost to children and the agencies that served them.

Generative AI stood to shift this dynamic by reducing administrative burden for both the care team and back-office departments. The result would be a double dividend: care teams could spend more quality time with the young people they serve, while money saved from back-office automation boosted their compensation.

The executive team—Kara Teeple (Chief Executive Officer), Sean McGinnis (Chief Program Officer), and Devan Hughes (Chief Financial Officer)—met to strategize. Some sectors could afford speculative AI experimentation, but foster agencies could not. Per-resident reimbursement rates were capped by state policy, and any additional costs had to be covered by grants or individual donors. The agency would need to select a narrow set of high-yield use cases.

The executive team identified residential facilities as an optimal starting point. This setting was home to the full spectrum of agency personnel, from case managers and therapists to child-care aides and administrative staff. Insights from this environment could later be applied across Lawrence Hall’s broader continuum of care. The executive team resolved to pilot AI tools at two residential programs: a campus for youth aged 8–17 and a transitional living program for adolescents aged 17–21.

But which personnel or departments were best suited for the pilot? Should the rollout differ for each program? To answer this, the team needed to map costs to each program, identify which AI deployments were likely to deliver the greatest efficiency, and model resource reallocation across the care team.

This paper studies how mandatory greenhouse gas disclosure affects new business formation. We find a significant increase in business entry following the implementation of the Greenhouse Gas Reporting Program in affected industries, relative to unaffected controls. We propose two channels. First, through a production channel, disclosure pressures incumbent firms to reduce emissions by scaling back production or reallocating resources toward cleaner technologies, weakening incumbents’ competitive positions and creating space for new entrants. Second, through an information channel, public disclosure of previously proprietary emissions data helps potential entrepreneurs identify viable entry opportunities. We present evidence consistent with both channels. Incumbent firms reduce economic activity and experience declines in profitability, and entry is concentrated in industries facing greater emissions reductions and public scrutiny. Additionally, regulatory and industry commentary highlights concerns over the proprietary nature of disclosed emissions data. Overall, our findings reveal an unintended yet economically meaningful consequence of environmental disclosure mandates.

We study and develop an algorithmic framework for incorporating "ex-ante" constraints—constraints on outcomes that hold only on average—into stateful sequential search problems with costly inspection. Our framework encompasses the classical Weitzman's Pandora's box [Weitzman,1978] as well as its extensions to joint Markovian scheduling [Dumitriu et al., 2003; Gittins, 1979], which model richer processes such as multistage search with multiple layers of inspection. Ex-ante constraints are particularly motivated by social considerations in algorithmic hiring, where they can adjust outcome distributions to promote equity and access. While most work in the algorithmic fairness literature in computer science and economics has focused on incorporating such constraints into machine learning tasks like classification and regression, far less attention has been devoted to operational problems such as sequential search, with their unique intricacies. Our work aims to bridge this gap. Building on the optimality of index-based policies in the unconstrained versions of these problems, we show that optimal policies under a single ex-ante constraint (e.g., demographic parity) retain an index-based structure but require (i) dual-based adjustments of the indices and (ii) randomization between two such adjustments via a "tie-breaking rule," both easy to compute and economically interpretable. We then extend our results to multiple affine constraints by reducing the problem to a variant of the exact Carathéodory problem and providing a polynomial-time algorithm that constructs an optimal randomized dual-adjusted index-based policy satisfying all constraints simultaneously. For general affine and convex constraints, we develop a primal-dual algorithm that randomizes over a polynomial number of dual-based adjustments, yielding a near-feasible, near-optimal policy. These results rely on the key observation that a suitable relaxation of the Lagrange dual function for these constrained problems admits index-based policies akin to those in the unconstrained setting. Finally, through a numerical study, we investigate the implications of imposing socially aware ex-ante constraints and their socially desirable outcomes.

Do corporate control transactions discipline the labor force? Consistent with synergies, new disclosures of employee misconduct in the investment advisory industry drop by between 17 and 22 percent following mergers. Both targets and acquirers have better pre-merger misconduct records than the industry’s average firm and, within the subsample of merging firms, there is assortative matching on misconduct. Merger events facilitate further reductions in misconduct through separations of target firm employees with high misconduct. Many of these employees remain in the industry, suggesting that consolidation plays an important role in the redistribution of misconduct across firms.

Apportionment is the act of distributing the seats of a legislature among political parties (or states) in proportion to their vote shares (or populations). A famous impossibility by Balinski and Young (2001) shows that no apportionment method can be proportional up to one seat (quota) while also responding monotonically to changes in the votes (population monotonicity). Grimmett (2004) proposed to overcome this impossibility by randomizing the apportionment, which can achieve quota as well as perfect proportionality and monotonicity — at least in terms of the expected number of seats awarded to each party. Still, the correlations between the seats awarded to different parties may exhibit bizarre non-monotonicities. When parties or voters care about joint events, such as whether a coalition of parties reaches a majority, these non-monotonicities can cause paradoxes, including incentives for strategic voting. In this paper, we propose monotonicity axioms ruling out these paradoxes, and study which of them can be satisfied jointly with Grimmett’s axioms. Essentially, we require that, if a set of parties all receive more votes, the probability of those parties jointly receiving more seats should increase. Our work draws on a rich literature on unequal probability sampling in statistics (studied as dependent randomized rounding in computer science). Our main result shows that a sampling scheme due to Sampford (1967) satisfies Grimmett’s axioms and a notion of higher-order correlation monotonicity.

To meet voluntary climate targets, firms often complement internal decarbonization efforts by purchasing carbon credits in the voluntary carbon market (VCM), which finance projects that reduce emissions elsewhere. However, these emissions reductions are difficult to verify, and growing evidence of overcrediting has cast doubt on the VCM's potential to genuinely offset emissions. We investigate how the VCM's defining features shape its climate effectiveness. Our model captures three central elements: adverse selection, as high-quality projects that truly reduce emissions are costlier yet difficult to distinguish from low-quality ones; imperfect third-party certification, as projects are screened based on a noisy signal of quality; and buyer preferences for non-carbon attributes, as some firms value credits that generate observable social or economic co-benefits beyond reducing emissions. We show that the market fails to sustain trade if certification is sufficiently noisy, as quality uncertainty erodes buyer confidence and triggers a market-for-lemons collapse. However, demand for co-benefits can sustain markets that would otherwise collapse. Yet in such cases, the market remains active but yields limited carbon abatement, as most traded credits are low-quality. We then examine policy and market design interventions reflecting recent developments in practice, such as penalizing buyers for greenwashing and offering credit portfolios. We show that these measures can be counterproductive for carbon mitigation if certification remains inaccurate. Accordingly, we demonstrate that the certifier’s incentives for accuracy can be strengthened by modifying its fee structure so that its revenue is tied to the market value rather than the volume of credits.

Pandemic restrictions targeted non-essential businesses that required in-person contact to operate. We document the impact of these restrictions county-by-county on directly regulated businesses and on businesses not subject to restrictions. Through 2023, a one standard deviation increase in restrictions increased business closures by 3% and reduced net job creation by 18%. These adverse effects extended beyond regulated firms to those exempt from regulations. We attribute this transmission at least in part to the loss of face-to-face interactions.

Making sense of massive, individual-level data is challenging: marketing researchers and analysts need flexible models that can accommodate rich patterns of heterogeneity and dynamics, work with and link diverse data types, and scale to modern data sizes. Practitioners also need tools that can quantify uncertainty in models and predictions of consumer behavior to inform optimal decision-making. In this paper, we demonstrate the promise of probabilistic machine learning (PML), which refers to the pairing of probabilistic modeling and machine learning methods, in pushing the frontier of combining flexibility, scalability, interpretability, and uncertainty quantification for building better models of consumers and their choices. Specifically, we overview both PML models and inference methods, and highlight their utility for addressing four common classes of marketing problems: (1) uncovering heterogeneity, (2) flexibly modeling nonlinearities and dynamics, (3) handling high-dimensional and unstructured data, and (4) addressing missingness, often via data fusion. We also discuss promising directions in enriching marketing models, reflecting recent developments in representation learning, causal inference, experimentation and decision-making, and theory-based behavioral modeling

Problem definition. Expanding infrastructure in transport, energy, and digital sectors is important for achieving the economic and sustainability goals of many developing economies. The values of projects involving such infrastructure expansion can be interdependent due to inherent network effects. However, institutional constraints and limited coordination capacity often force governments to award projects through separate auctions. Brazil’s National Logistics Plan (PNL) exemplifies this setting, where interdependent transport investments are procured separately despite strong network effects across projects. We study how a government buyer should design separate per-project procurements when the overall value depends on the resulting infrastructure network, and bundling is not feasible. Methodology/results. We analyze: (i) parallel procurement, in which auction rules are fixed in advance; (ii) sequential procurement, in which their sequence is fixed, but later rules can adapt to earlier executed projects. In both settings, we show that the optimal mechanism retains a simple form akin to canonical results in auction theory: each project is awarded to the lowest virtual-cost bidder if its virtual cost falls below a project-specific threshold that reflects its expected contribution to the equilibrium-induced network. Beyond the optimal mechanism, we propose a simple and interpretable ``credit-aware'' heuristic that adjusts each project's threshold by a simple one-step estimate of its expected network contribution. To demonstrate the importance of network effects in procurement design, we construct a calibrated case study of a major railroad corridor expansion under the PNL. Modeling the Brazilian infrastructure as a hub--spoke network, and the expansion as two key added segments, we estimate their network effect by solving a congestion game aligned with current practice. We show that joint commissioning of these segments adds 16.4% additional economic value, and the common practice of ignoring network effect in procurement design leaves 30–36% of attainable surplus unrealized. Our credit-aware heuristic recovers about 90% of the optimal surplus. Managerial implications. As developing economies invest in infrastructure expansion, it is important to evaluate these projects holistically, as network effects can be strong and first-order. Even when institutional constraints require projects to be procured separately, network effects can still be incorporated effectively through careful adjustments to existing mechanisms, and even simple changes can yield substantial economic and environmental gains.

Nature provides essential inputs to the economy through ecosystem services. In Asia and the Pacific, accelerating biodiversity loss is eroding these services, increasing vulnerability to shocks, and creating new risks for investors and governments.

Women in masculine-typed roles often experience their gender identity as a barrier to proving themselves by the ideal-worker norms of their male-dominated occupations. Yet, these women often internalize these experiences, blaming themselves for their struggles. They rarely identify as members of a disadvantaged identity group and often distance themselves from other women at work. How and when might such women externalize their struggles as gendered and collective? Drawing on data from a qualitative field study of staff working in many masculine-typed roles across various male-dominated occupations at a U.S. public-lands management organization, I develop grounded theory suggesting when and how some women might come to reappraise some of their struggles as rooted in the gendered cultures of their occupations rather than in their own deficiencies or idiosyncratic circumstances. I find that “parallel-peer connections” between similarly situated women outside their local tokenized work groups can spark transformative mindset shifts when these encounters occur under the right conditions: during a window of sensemaking about a career impasse and in a less competitive context that is conducive to sharing confidences. Some women credited these shifts with prompting them to shed years of self-doubt and to promote gender equality at work. This study contributes to our understanding of supportive workplace relations among tokenized women and mindset shifts at work.

In 2015, Yara International faced simultaneous financial, legal, and reputational pressures as industry conditions deteriorated. Newly appointed CEO Svein Tore Holsether concluded that restoring performance required more than cost-cutting and reputation repair, and instead moved to future-proof the firm by embedding sustainability into Yara’s purpose, strategy, and operating model. This case examines the rationale, execution, and 2025 outcomes of Yara’s purpose-led transformation amid evolving market and societal constraints.

The design of algorithms for political redistricting generally takes one of two approaches: optimize an objec- tive such as compactness or, drawing on fair division, construct a protocol whose outcomes guarantee partisan fairness. We aim to have the best of both worlds by optimizing an objective subject to a binary fairness constraint. As the fairness constraint we adopt the geometric target, which requires the number of seats won by each party to be at least the average (rounded down) of its outcomes under its worst and best possible partitions of the state.