Open Collections

Abstract

This dissertation comprises three self-contained chapters. Chapters 1 and 3 share a common theme: how firms interact within the same local labor market. Chapter 2 addresses the properties of neural network solutions of dynamic economic models. Chapter 1 examines how the expansion of large, multi-establishment (national) employers influences the wage and hiring policies of smaller, local firms. Using administrative data from Brazil covering firms’ and establishments’ data across different locations and occupations, I conduct a matched event study to assess how local employers respond to national employers’ large, idiosyncratic labor demand shifts. The findings reveal that during the expansion period, national employers increase wages in large cities by 8 percent, and they simultaneously raise wages by 5 percent and expand employment in other locations. This expansion pressures local employers in smaller locations to increase wages by 2 percent, resulting in a 1.5 percent wage growth for incumbent workers. Despite local employers reducing employment, workers are not adversely affected because they reallocate to national employers. In Chapter 2, we analyze the properties of deep learning approximations to dynamic structural models. We show that, due to their inherent inductive bias toward finding flat solutions to functional equations, such approximations can automatically satisfy long-run boundary conditions. This highlights the potential of deep learning to simplify the verification of an equilibrium’s long-term behavior in structural models. Additionally, the inductive bias provides a foundation for modeling forward-looking behavioral agents with self-consistent expectations. Chapter 3 develops a unified theoretical framework to quantify how heterogeneous firms in local labor markets respond to industry-level productivity shocks. We extend the standard monopsony model of Card et al. (2018) to a general-equilibrium setting, incorporating cross-employer spillovers that allow firms to adjust endogenously to shifts in workers’ outside options arising from productivity changes in other local industries. Using German administrative data, we estimate the model by employing national industry wage premiums as proxies for industry-specific productivity shocks. Our analysis reveals strong outside-option effects, a high degree of monopsony power, and relatively limited variation in that power level across commuting zones.