Large firms have wage-setting power over their current workforce, but the need to ensure reliable inflows of new workers constrains that power. This tension is particularly important in markets with specialized human capital, where trained workers are captive, but wage-sensitive prospective workers pay high fixed costs to train. We develop a dynamic oligopsony model in which large firms internalize how the wages they pay shape the future supply of trained workers to their sector, extending standard monopsony models that study firms' short-run market power over a fixed pool of workers. Estimating the model on the Danish pharmacist labor market from 2001 to 2019, we show that relative to a myopic benchmark, where firms maximize short-run profits, accounting for entry reduces markdowns by 12.5%, increasing wages by $2,469 per worker. However, accounting for entry only closes the employment gap relative to the competitive benchmark by 16.2%. Because the trained labor pool is shared across firms, investment is primarily undertaken by large firms in concentrated markets---the same firms and markets where monopsony distortions are greatest. In our counterfactuals, we show that entrant wage subsidies, offered by individual firms or an employer organization, expand future supply, improving the allocation of workers across sectors, but result in lower wages for the incumbent workforce.

Abstract: Employers facing limited labor market competition may suppress wages below socially optimal levels. Unions can counteract wage suppression through collective bargaining, though they may also push wages above the socially optimal level and lead to job rationing. To assess these forces, we estimate a structural model of labor supply, labor demand, and Nash-in-Nash bargaining over wages between teacher unions and school districts in Pennsylvania's K-12 public school system from 2013 to 2020. We compare negotiated equilibrium wages and employment to pure oligopsony and social planner scenarios. On average, pure oligopsony reduces wages 16 percent below the social optimum, while collective bargaining raises wages 9 percent above the optimum. Averages masks substantial district-level heterogeneity driven by bargaining power variation. Twenty-seven percent of schools have negotiated salaries below th    e social optimum due to cross-district externalities, where high salaries at one school cause hiring reductions that increase labor supply elsewhere.

We show that combining two common sources of monopsony power, differentiated worker preferences and oligopsony, mechanically attenuates the wage response to changes in market structure when using common logit models. Differentiated preferences, when modeled as logit shocks, give each firm residual market power that increases as its labor share falls, counteracting the pro-competitive effect of additional firms. In the limiting case of perfectly elastic labor demand with pure logit labor supply and Cournot competition, they offset exactly implying that adding more firms to a market has no effect on market power. We discuss three modeling choices that partially mitigate this tension: Bertrand strategic interactions, downward-sloping product demand, and individual-level heterogeneity.