Session 5: The Micro and Macro of Labor Markets

I study hiring under uncertainty when firms can either hold permanent labor as a buffer or hire from a spot market after demand is realized. I build a model where the spot market is endogenously determined such that when one firm relies more on spot labor, it thickens the market that other firms use as well. Individual firms do not internalize that contribution, so the competitive equilibrium involves too much buffering. I evaluate the model in the market for registered nurses using the universe of daily time sheet records from skilled nursing facilities. The data support the model’s general equilibrium predictions. Markets with thicker agency markets have less rationing conditional on total nursing hours, and thicker agency markets have compressed upper tails of permanent staffing and higher overall hours in the lower part of the distribution. Estimating a facility-level newsvendor model with heterogeneous spot market frictions and rationing costs I find that the nursing labor market was more efficient post-COVID, there are substantial welfare gains from lower spot market frictions, and that the marginal external benefit for a firm that participates in the spot market instead of hiring an employee as buffer averages 7 percent of wages. This labor market underprovides flexibility because the value of availability is not fully priced.

Can profit-driven middlemen arise endogenously to correct aggregate misallocations and promote development? This paper studies the role of recruitment agencies in facilitating spatial labor reallocation in China. Using a newly constructed dataset, I find that such profit-driven labor intermediaries established branches to arbitrage away pre-existing inter-regional wage disparities. Exploiting the staggered rollout of agencies in difference-in-differences and IV designs, the analysis shows that the expansion of agency networks facilitated temporary mobility of low-skilled workers and reduced regional wage dispersion by 10%. This translates into 3% aggregate output gains by 2010, of which agencies appropriated 40% as profits. I develop a neoclassical growth model with endogenous agency dynamics, where initial misallocations create pecuniary incentives for agencies who in turn speed up regional convergence. Promoting agency workers’ transition into the permanent labor force in receiving regions can generate large efficiency gains while also ensuring these gains are more widely shared.

This paper develops a framework for empirically estimating aggregate labor supply across occupations with flexible substitution patterns and applies it to study the incidence of automation and artificial intelligence in the U.S. labor market. Central to the analysis is the distance-dependent elasticity of substitution (DIDES), where worker substitutability between occupations declines with their distance in skill space. By mapping 306 occupations into cognitive, manual, and interpersonal skill dimensions, we estimate a low-dimensional latent skill model that preserves granular substitution patterns. We show that both automation and artificial intelligence cluster in skill-adjacent occupations, constraining employment adjustment and amplifying wage effects: 20–50% of labor demand shocks pass through to wages, while mobility recovers only 20% of wage losses.

We develop a model of joint wage and mobility outcomes with two-sided heterogeneity. The framework extends the finite mixture approaches of Bonhomme, Lamadon, and Manresa (2019) and Lentz, Piyapromdee, and Robin (2023) by allowing worker types to evolve according to a hidden Markov process that may depend on the firm type the worker is matched with. We estimate the model using a variational expectation maximization (VEM) algorithm that jointly classifies worker and firm types using the full likelihood of wages and mobility outcomes, yielding substantial improvements in latent type classification relative to existing approaches. 

The model is identified on Italian register data. We estimate it using administrative matched employer–employee data from the Veneto region of Italy covering 1982–2001. The model fits key features of the data, including employment rates, life-cycle wage growth, and the evolution of wage dispersion. Worker heterogeneity accounts for the largest share of wage dispersion, while sorting between worker and firm wage types also contributes. Sorting increases over workers’ careers, but this increase is driven primarily by firm-type-dependent worker type dynamics rather than by worker mobility across firms. Workers experience stronger wage-type growth when employed at high wage-type firms, implying that the correlation between worker and firm wage types rises even without substantial reallocation; once this mechanism is accounted for, job mobility tends to dampen sorting.

Worker type dynamics also shape life-cycle wage growth and dispersion. Initial conditions explain about 60 percent of wage variation at labor market entry and decline to roughly 50 percent after twenty years. Worker wage type growth accounts for more than 85 percent of wage growth with experience, and the doubling of wage variance over the first twenty years of a career is almost entirely driven by increased worker wage type dispersion. Periods of non-employment generate persistent scarring effects by slowing or reversing worker type progression, with particularly large impacts for workers previously matched with high wage-type firms.

Job loss is one of the most costly economic risks workers face, but a firm’s layoff risk is difficult to observe. We document substantial, persistent variation in firm layoff rates, creating scope for workers to change their job loss risk through firm choice. We exploit linked survey, experimental, and administrative data from Austria to examine how unemployed workers perceive and respond to information about firm-level layoff risk. Workers believe that past layoffs predict future risk and prefer jobs at firms with lower historical layoff rates but have significant misperceptions about which firms are safer. Providing workers with information about firm layoff histories causes them to redirect their search toward historically safer employers. Using a search and matching model, we show that imperfect information distorts equilibrium outcomes: it reverses the compensating differential for layoff risk and raises the average layoff rate by allocating more workers to high-risk firms.

Many temporary foreign worker programs issue “closed” visas that effectively tie workers to a single employer, restricting worker mobility and weakening bargaining power. We study the labor market return to temporary foreign workers (TFWs) gaining permanent residency (PR), which loosens this mobility restriction. Using administrative data linking matched employer-employee data in Canada to temporary and permanent visa records from 2004–2014 along with an event- study design, we find that gaining PR leads to a sharp, immediate, and persistent increase in the job switching rate of 21.7 percentage points and an increase in earnings of 5.7 percent three years after PR. Workers also sort into high-wage firms after gaining PR, and the increase in the firm pay premium is roughly 56 percent of the total earnings gain. We find larger earnings gains for job switchers across industries, low-skilled workers, and workers from low-income countries. To guide and interpret our reduced-form results, we develop a search-and-matching model featuring heterogeneous workers and firms. Permanent residents and native-born workers search for jobs in the same labor market and engage in on-the-job search, while TFWs search separately within a segmented labor market and do not receive outside wage offers. We calibrate the model to match our reduced-form results, and we use it to simulate the long-run effects of PR and consider two counterfactual policies: (1) increasing the cost to firms of posting a TFW vacancy and (2) allowing TFWs to switch employers freely under “open” visas. We evaluate how these policies affect output, wages, profits, and overall social welfare.

Over the past decade, a broad consensus has emerged that labor markets are not perfectly competitive. The natural question is: what assumption replaces perfect competition in labor markets? We develop a scalable framework for testing which labor market conduct best fits the data. Our approach distinguishes among perfect competition, monopsonistic competition, oligopsonistic competition, and collusion, as well as partially-collusive conducts that combine oligopsonistic and collusive elements. Our approach is scalable in the sense that it can be applied on a case-by-case basis to each local labor market observed in administrative data using only five commonly-available variables: employment, wages, revenues, a market identifier, and a labor demand shifter. Applying the framework to each of 108,000 labor markets defined by industry-commuting zone pairs in U.S. Census micro-data, we find that 35 percent of employment is in markets best characterized by monopsonistic competition, 40 percent by oligopsonistic competition, 22 percent by full collusion, and 3 percent by partial collusion. By allowing each market to operate under its best-fitting conduct, we find nationally that markdowns are stronger on average and more dispersed than any single conduct assumption would imply. Collusive conduct is concentrated in markets with few competitors and more prevalent in markets with lower entry rates and less differentiated employers. Applying our conduct test to seven specific markets investigated by the U.S. Department of Justice for employer collusion, we detect collusion in five, including both cases that led to a guilty plea or conviction.

I study how labor market power distorts the broader production process, combining evidence from merger event studies and a structural oligopsony model in which wages, employment, capital, and output are jointly determined in equilibrium. Using administrative data from the U.S. coal industry, I show that rent extraction from the labor market not only lowers wages and employment, but also creates a scale effect that reduces the firm’s demand for capital, suppresses output, and diminishes aggregate productivity. These “knock-on” distortions of labor market power are four times larger, by value, than the direct welfare loss to workers. The results suggest that labor market outcomes can significantly underestimate the welfare costs of oligopsony, especially in capital-intensive industries like mining.

I ask whether monopsonistic wage-setting can distort the labor market when firms face elastic labor supply schedules. As labor supply becomes more elastic, the distance between the efficient worker allocation and the monopsonistic worker allocation will converge to a non-zero constant. The social cost of monopsonistic wage setting can also grow as labor supply becomes more elastic. I derive that social cost both in a random utility framework and in a search framework, and in particular I show that those two expressions are identical.

Monopsony power is often measured by interpreting firm wage and labor responses to shocks through static models. But when workers face frictions to changing jobs, employment adjusts gradually and workers respond to changes in the total value of a job—not just the current wage. We develop a general equilibrium dynamic monopsony model where firms contract with risk-averse workers over idiosyncratic shocks. We show that the shock-identified labor supply elasticity depends on the persistence of the shock, worker risk aversion, and the horizon over which it is estimated. These forces induce a wedge between the inverse shock-identified labor supply elasticity and the wage markdown. We estimate the model using U.S. Census employer-employee matched data. The small and persistent wage response to temporary shocks is consistent with firms insuring risk-averse workers. Search frictions explain why employment continues to rise even after wages have started to fall. We find the average worker’s wage is marked down 8.3%. By contrast, the static model approach of inverting the shock-identified labor supply elasticity implies a markdown estimate as wide as 26%. Lastly, we show that firm employment dynamics are not efficient: insurance distorts the job ladder, preventing productivity-improving job transitions from occurring.

Employer market power is typically measured and analyzed under the assumption of fixed labor market boundaries, often approximated by geographic units such as commuting zones or counties. We show that this practice misses an important margin: workers broaden the geographic scope of their job search when their home market is concentrated or cyclically weak. Using over 63 million online applications to U.S. hourly jobs, we document two stylized facts. First, the spatial breadth of search is heterogeneous across workers: about one in three searches in a market either smaller or larger than a typical commuting zone. Second, search breadth is also cyclical: when local labor markets deteriorate, an increasing share of workers searches farther from home. To interpret these facts, we propose a search-and-matching model that endogenizes labor market boundaries by combining granular firms with multi-market search. Relative to a fixed-market benchmark, the model has a set of distinct predictions: first, workers in concentrated markets reallocate search effort away from them; second, re-application to large firms is dampened; third, within-firm wages fall with distance to market boundaries, and, fourth, productivity pass-through is attenuated where cross-market search is more costly. Each prediction holds in the data. Effective labor market monopsony exposure is thus jointly determined by local market structure and workers’ search-effort reallocation.