Abstract: What is the optimal local industry composition when specialization increases expected productivity but also concentrates exposure to sectoral shocks? I study this question using a dynamic model that treats local industry composition as a portfolio with endogenous returns. Agglomeration externalities generate increasing returns to scale, stochastic industry productivity generates risk, and captive local demand determines how shocks pass through the local economy. These features generate a mean-variance trade-off: optimal ex ante specialization depends on both the expected gains from agglomeration and the costs of concentrated sectoral risk. I estimate the model with U.S. metro-industry data and use neural networks to solve the planner’s problem and decentralized equilibrium for the full set of U.S. MSAs and two-digit NAICS industries. The resulting optimal allocations show that risk reduces the optimal level of ex ante absolute and relative specialization compared to a no-shock benchmark, even in the absence of risk aversion per se. Moreover, the welfare gains from optimal policy are significantly smaller in the stochastic model than in the no-shock benchmark. These results imply that standard models of agglomeration that omit sectoral risk may overstate the optimal levels of absolute and relative specialization as well as the welfare gains from optimal policy.

Abstract: This paper tests whether volatility in inflation and GDP growth was higher at the ZLB across the period from 1990:1 to 2019:3 as predicted by conventional models. Contrary to the theoretical predictions, volatility was not higher at the ZLB.

Abstract: This paper investigates self-reported wedges between how much people work and how much they want to work at their current wage. More than two-thirds of full-time workers in German survey data are overworked—usual hours exceed desired hours. We combine this evidence with a simple model of labor supply to assess the welfare consequences of tighter weekly hours limits via willingness-to-pay calculations. According to counterfactuals, the optimal length of the workweek in Germany is 37 hours. Introducing such a cap would raise welfare by 0.6–1.2% of GDP. The gains from a shortened workweek are largest for workers who are married, female, college-educated, white-collar, middle-aged, and high-income. An extended analysis integrates a non-constant wage-hours relationship, falling capital returns, and a shrinking tax base.

Abstract: This paper studies employer retaliation against on-the-job search as a source of endogenous labor market power. I nest a strategic retaliation choice in a Burdett–Mortensen wage-posting model with endogenous search, heterogeneous firms and workers, and unobservable experience-good amenities. Because amenities are unobservable, firms do not know which workers are searching. Instead, on-the-job search generates noisy evidence, with false positives coming from non-searchers. Firms choose the evidentiary threshold for imposing a proportional reduction in wages and productivity, subject to a trade-off between higher retention from reduced search and lower recruitment from reduced job values. In the calibrated model, retaliation appears in equilibrium as the optimal choice of high-productivity firms, covering just over half of the workforce. Search decreases both at retaliating firms and at non-retaliating firms, as the returns to search at the bottom of the job ladder decline. This loosens the connection between wages and the recruitment and retention margins, leading the endogenous wage offer distribution to shift down and markdowns to rise, especially at retaliators. Stationary wages, productivity, and worker and aggregate welfare all decline, while firm value increases. Retaliation thus allows firms to generate endogenous labor market power by modifying workers’ search incentives.

Abstract: This paper investigates the role of non-Ricardian beliefs about future tax incidence in mediating the aggregate economic response to an unfinanced tax cut. Using a standard one-asset HANK model with an endogenous fiscal rule, I simulate a ten-year tax cut under systematically biased beliefs. Behavioral agents forecast their own future incidence as a fixed multiple of the FIRE benchmark and attribute offsetting incidence to households far from them in income and asset space, so that beliefs remain consistent with the government’s long-run budget constraint. Systematic optimism operates like a large wealth effect, generating lower labor supply and savings relative to FIRE. Rising interest rates lead to lower investment and a revaluation of existing government debt, worsening the fiscal impact of the tax cut. Output and welfare decrease. Systematic pessimism reverses these signs. Simulations of the Reagan, Bush, and Trump tax cuts under optimism produce results consistent with the muted aggregate expansions following these policies. A decomposition shows that the effect of biased beliefs is highly sensitive to the existing stock of government debt, highlighting the importance of the revaluation channel.

Abstract: This paper investigates the impact of ambiguity aversion on worker search behavior, job tenure, and markdowns. I embed ambiguity aversion in a standard Burdett–Mortensen wage-posting model with endogenous search effort, heterogeneous firms, and imperfectly observable amenities. Workers receive repeated signals about the amenity at their current job, but only one signal about each external offer. They are averse to ambiguity over the difference in amenities between their current job and potential outside offers. This aversion leads to stricter acceptance criteria for outside offers, reducing the value of on-the-job search. Workers search less and move less often. The combination of high early returns to learning and long-run selection matches the empirical hump-shaped relationship between quit rates and tenure. Lower worker search and offer acceptance rates weaken the connection between wages and the recruitment and retention margins, leading to a downward shift in the equilibrium offer distribution and an increase in markdowns. Compared to full information and Bayesian learning benchmarks, ambiguity aversion lowers search, wages, productivity, and both worker and aggregate welfare, while raising markdowns, average tenure, and firm welfare.

Abstract: We use gradient-boosted trees to estimate U.S. county-quarter real GDP for 2013–2024 using BEA county-year real GDP data, QCEW county-quarter payroll, employment, and establishment counts, and spatial characteristics such as population and area. In held-out cells, the model produces an R² of 0.990 and a WAPE of 5.31% when comparing annual-aggregated predicted county-quarter GDP to observed county-year values. The resulting panel has broad applicability to questions in the field of regional economics, with the potential to improve statistical power in empirical and calibration applications.