Existing research consistently finds that unemployment insurance (UI) benefits delay job finding with limited effects on job quality, but focuses on changes in UI generosity while holding fixed access to re-employment services. Using employer-employee matched data from Washington State and a fuzzy regression discontinuity design around the eligibility threshold for UI, we find that benefit receipt minimally delays re-employment but substantially improves labor market outcomes. UI increases cumulative hours worked by approximately 15 full-time weeks and earnings by $14,000 in the two years following job loss, representing 37% and 50% increases, respectively. These gains are driven by improved job quality, as recipients experience longer tenure and higher wages with their next employer. Effects are larger for workers living near public employment offices, suggesting that access to re-employment services enhances search productivity. Expanding UI access by lowering the eligibility threshold is much more cost-effective than raising benefit levels or extending potential duration, as workers benefit from more stable, higher-paying re-employment that partially offsets its cost.

Standard models of unemployment insurance (UI) focus on how benefit generosity affects the average claim duration, while assuming perfect take-up. Yet, benefit receipt is highly incomplete with estimates of take-up among eligible workers below 50% in the United States. In this paper, we show take-up is an important margin of response: If benefits become more generous, more workers claim benefits in addition to claimants remaining on benefits for longer. Using a sample of likely eligible workers, we leverage a regression kink design to identify the causal effect of weekly benefit level on take-up and total benefits paid. Our results suggest a 10% increase in the weekly benefit leads to a 4.7% increase in take-up, which drives a 6.2% increase in total benefits paid. Previous work that focused only on claim duration did not account for this and thus underestimated the fiscal externality from raising benefit levels. These findings have important implications for policy: accounting for endogenous take-up reduces the optimal benefit level by 29% and lowers the value of additional spending to raise benefits by 27%.

We develop a new survey analyzing Generative AI use in the labor market to assist in measuring its economic effects. We find that Generative AI tools like large language models (LLMs) are most commonly used in the labor force by younger individuals, more highly educated individuals, higher income individuals, and those in particular industries such as customer service, marketing and information technology. We find that LLM adoption among U.S. workers has increased rapidly from 30.1% as of December 2024 to 36.7% as of September 2025, and adoption in the U.S. remains the highest among advanced economies. We also estimate the effects of Generative AI exposure on several labor market outcomes, finding that more exposed occupations have experienced larger wage increases since the November 2022 public release of ChatGPT while finding no significant effects in job openings or total jobs.

We develop a method to jointly measure the response of worker search effort (micro effect) and vacancy creation (macro effect) to changes in the duration of unemployment insurance (UI) benefits. To implement this approach, we exploit an unexpected cut in UI durations in Missouri and provide quasi-experimental evidence on the effect of UI on the labor market. The data indicate that the cut in Missouri significantly increased job-finding rates by raising the search effort of unemployed workers and the availability of jobs. The latter accounts for around one-third of the total effect.

We investigate the role of firm pay premiums in explaining the large, persistent earnings losses of displaced workers. They first estimate that long-run earnings for displaced workers from 2002 to 2008 in Ohio are depressed by 22%. Drawing upon empirical approaches from the displaced worker and firm heterogeneity literature, the authors then estimate that one-quarter of this earnings loss can be explained by the forfeiture of a favorable employer-specific pay premium. Such firm rents are more salient for those laid off from manufacturing firms, explaining half of their lost earnings. Nevertheless, this study adds to early evidence that firm rents do not explain the majority of earnings losses sustained by displaced workers in the United States.

Displaced workers suffer large and persistent earnings losses. These losses can be mitigated by returning to school, yet the extent to which such workers enroll in post-secondary education in response to displacement is poorly understood. Using employer-employee-student matched administrative data from Ohio, we provide the first direct evidence of workers’ enrollment responses following mass layoffs in the United States. We estimate that for every 100 displaced workers, only 1 is ever induced to enroll in a public college. This effect is concentrated almost entirely among displaced manufacturing workers, who enroll at a rate of 2.5 per 100. Workers with relatively low earnings at their layoff firms are the most likely to enroll in public institutions.

The study of labor unions in the United States is significantly constrained by the coarseness of the Current Population Survey (CPS), whose geographic granularity is limited to the state level or large metropolitan areas. This absence of fine geography means that recent approaches to studying labor market shocks that use shocks to within-state labor markets cannot incorporate union density as an outcome, covariate, or source of heterogeneity. This paper overcomes these problems by marshaling numerous survey and administrative data sources to construct a novel dataset on union membership and density at the local labor market level for the period 1980-2020. We validate our granular density measure across regions and over time with use cases such as the 1980s recession and the China shock.

This paper develops and evaluates the concept of revenue-neutral seasonal consumption taxation in a tourism-dependent economy. Building on a model with heterogeneous elasticities and tax salience, we establish two theoretical results: (i) when the peak season is dominated by tourists with less elastic and less salient demand, there exists an efficiency-dominant seasonal tax schedule that yields the same revenue as a uniform rate while reducing deadweight loss; and (ii) such schedules necessarily shift a greater share of the tax burden from residents to tourists. We calibrate the framework using monthly sectoral sales data for Maine and analyze alternative tax schedules along the revenue-neutral locus. The results show clear U-shaped efficiency curves and systematic incidence shifts, confirming the theoretical propositions. Quantitatively, applying the efficiency-dominant seasonal rates reduces annual deadweight loss by about $0.4 million, lowers residents’ tax burden by roughly $19 million (about $14 per capita), and raises tourists’ burden by a corresponding amount, all while holding total state revenue constant. Aggregate pre-tax business sales increase slightly (about $12 million, or 0.06%), suggesting no adverse effects on local firms. Taken together, the findings demonstrate that seasonal taxation can simultaneously enhance efficiency, relieve residents, and preserve fiscal capacity by leveraging the seasonal influx of visitors. The analysis highlights the practical potential of seasonal tax design as an alternative to uniform sales taxation in tourism-intensive economies.