Firms increasingly rely on recruiters to find talent. Recruiters are typically paid using refund contracts, which specify a payment upon a successful candidate suggestion and hire, and a refund if a candidate is hired but leaves for any reason during an initial period of employment. We study how recruiters and refund contracts shape talent selection. When a firm needs to fill a position, it engages a recruiter who observes private signals about a candidate’s productivity and decides whether to suggest this candidate to the firm. There is variation in both the candidates’ productivity and in the quality of information available about productivity. We characterize the unique equilibrium and show that refund contracts induce artificial risk aversion in both the recruiter’s suggestion strategy and the firm’s hiring strategy relative to a first-best benchmark. This risk aversion leads to candidates with lower expected productivity but more informative signals (“safe bets”) being favored over candidates with higher expected productivity but less informative signals (“diamonds in the rough”). Our findings imply that delegated recruitment generates statistical discrimination.

In the US public sector, there are many examples where a few government workers earn large amounts from overtime. Is this government inefficiency driven by insider influence, or an efficient reflection of worker preferences? We study the Los Angeles Department of Transportation, where several traffic officers earned around $100,000 in overtime pay despite an ex ante equal system called “the wheel.” We show that trade via informal networks achieves 93.6% of the maximum possible allocative efficiency. This illustrates a general idea: endogenous trading networks can create the conditions for a version of the Coase Theorem to hold.

Using millions of task assignments from salon management software, I find significant establishment-level dispersion in labor productivity and internal task specialization and a strong association between the two that is unexplained by establishment size. The 25% most specialized salon-quarters are on average 68% more productive than the bottom 25%. To rationalize these facts, I identify and estimate a model where competing firms assign tasks to workers with multidimensional skills in light of firm-specific organization costs. I show that accounting for task specialization can qualitatively change the productivity implications of economic shocks. Without internal reorganization, immigration of low-wage workers into Los Angeles County provides a competitive advantage to less productive salons, replacing specialized with generalist jobs and reducing labor productivity by 1.0%. With internal reorganization, all types of salons adjust to incorporate immigrant skills, prices fall and market shares rise at most salons, and specialized jobs are created increasing labor productivity by 1.4%.

Whether tips should be treated differently than wages under tax and minimum-wage laws depends on whether they incentivize service quality. I develop a test for incentive relevant tipping, capable of detecting both quality-based social norms and forward-looking behavior, that consists of a simple regression of first-time tip percentages on an indicator for whether a customer returns in the future. I implement the test using more than 200,000 first-time tips from the beauty industry, a setting where the test is high-powered because the customer return rate is 30.7%. Across 2,953 workers, I fail to reject incentive-irrelevant tipping in 98.5% of cases

Consumer reviews reflect both product quality and price, with more favorable reviews for a lower-priced product. We study whether this review behavior induces a firm to manipulate reviews by underpricing its product below consumers’ willingness to pay. We introduce a model with a privately informed firm repeatedly selling its product to rational consumers who learn the product quality from past value-based reviews and the current price. We characterize the necessary and sufficient condition for underpricing, which depends on the relative amount of vertical versus horizontal quality differentiation. This condition implies that underpricing need not occur even if the firm is perfectly patient. Rating manipulation via underpricing, when it occurs, unambiguously benefits consumers.

In this paper, I study how voluntary labor supply decisions within an organization impact workplace injury using novel data on the payroll and workers’ compensation claims of Los Angeles traffic officers. I use the leave taken by coworkers as an instrument to estimate the causal effect of daily labor supply decisions on workplace injury. Self selection via voluntary labor supply reduces injuries by 48 percent compared to the underlying injury rate. I decompose selection into a predictable component that could be accomplished via direct assignment on observables by a manager, and a private component known only to the individual worker. I show the vast majority of the effect is driven by the private component, implying decentralized overtime assignment mechanisms like shift auctions are an effective way to reduce organizational injury rates.

This paper provides a microfoundation for the Shannon mutual information as a measure of management in organizations that must complete many tasks. We propose a model where an organization must assign an arbitrarily long sequence of tasks of different types to workers with different skills. Assigning tasks to the best suited worker results in faster production and a lower wage bill but requires greater management, measured as the number of contingency plans that must be developed. We show management is initially increasing returns to scale, but eventually become constant returns to scale and equal to the number of tasks multiplied by a constant which is based on Shannon mutual information. This occurs because the optimal design of large organizations involves core competencies: the organization develops a relatively small number of contingency plans that perform well on a relatively small number of distinct task sequences. As optimal organizations grow, the number of core competencies often become a vanishingly small fraction of the total number of possible sequences, but the contingency plans account for the realized sequence with probability approaching one.

This paper uses 10 years of public records documenting the almost free trade of work within a government organization. I estimate a dynamic equilibrium model where workers build a portfolio of shifts that determines their hours beyond the standard work week, considering both their static labor-leisure trade-off as well as dynamic considerations to save and consume. Each workers’ hours beyond their standard work week is determined by frictional trading with coworkers. The model yields worker-specific structural preferences for working hours, that allow the computation of individual labor supply elasticities under the current organizational constraints as well as counterfactual labor supply elasticities that would arise if a worker could continuously and freely choose their hours at any market wage. The curvature of each worker’s utility function with respect to hours is identified by variation in the initial assignment of additional work, which is determined by a quasi-exogenous rotating list. Sometimes workers are endowed with a 16 hour shift and sometimes a 1 hour shift, and the data record whether the shift is traded away. I compare my estimates to those in the literature, and argue that the microeconomic labor supply elasticities reported in the literature likely incorporate significant organizational frictions. The organization if racially and gender diverse, and I show how worker preferences and elasticities vary across demographic groups. I conclude by showing how the preferences of an organization’s workers combine with internal institutions to amplify the reduced-form labor supply responses of some workers and dampen others.

This paper aims to implement a survey of firms and workers to study the way hours are determined in the US labor market. In particular, we will survey a nationally representative sample of 1,000 managers and 1,000 workers. A novel innovation of our study is that we conduct a dyad survey in which we interview both workers and their direct managers within the same firm, allowing us to study their interaction. This unique feature of the survey enables us to measure both parties’ preferences for and beliefs about the worker’s hours, and test how these demand and supply forces interact to determine actual hours worked.

In this project, we study how healthcare firms organize the production of care by assigning tasks to physicians according to their skills. Our goal is to understand the determinants of physician specialization within firms and how these organizational choices affect the quality of care delivered to patients. In particular, we aim to quantify the roles of coordination costs, market competition, and reimbursement incentives in shaping the organization of medical work in the Medicare market. More broadly, this project contributes to the Russell Sage Foundation’s interest in understanding how organizational structures and market institutions shape the delivery of essential services and economic outcomes in the United States.

Asset forfeiture is the seizure of property that is believed to be connected to criminal activity, typically the sale of illicit drugs. Despite the dismantling of the California Bureau of Narcotic Enforcement, the legalization of marijuana, and legislative reform the value of assets seized in California has not decreased. To understand why, we digitize government reports and hand collect the membership of task forces in order to link each seizure to individual teams of state, local and federal agencies. We document that between 2002 and 2009, law enforcement agencies seized $3.17 billion. While federal agencies unilaterally seized 58% in terms of revenue, local agencies initiated 60.9% of seizures without federal involvement. Collaborative seizures involving multiple agencies are larger: the median federal-local collaborative seizure is almost three times the median federal unilateral seizure. Among seizures with local involvement, 75% of revenue came from collaborative seizures. We map the network of collaborations, and show that it features a core-periphery structure with 99.97% of potential revenue accruing to agencies in the core. We design and plan to estimate a network formation model where agencies choose who to collaborate with and how much effort to contribute. We plan to use the model to understand how different reforms impact both total asset forfeiture and the amount of law enforcement collaboration.