Monopsony Power in the AI Labor Market

In the AI Story

The rediscovery of monopsony in labor economics began in the late 1990s with Alan Manning's book Monopsony in Motion, which argued that labor markets are systematically less competitive than textbook models assume. Subsequent empirical work documented concentration in local labor markets, the wage-suppressing effects of non-compete clauses, and the pervasive use of no-poach agreements among employers. By the 2020s, monopsony had moved from a fringe concern to a central framework for understanding wage stagnation.

AI platforms intensify the dynamic in three ways. First, they often have strong network effects (drivers want to be where riders are and vice versa), producing natural tendencies toward concentration. Second, they typically classify workers as independent contractors, removing the protections that would otherwise constrain employer power. Third, they control the algorithmic systems that determine work assignment, pricing, and performance evaluation — giving them information and control asymmetries that traditional employers lacked.

The productivity-pay gap that emerges from this structure is a distributional rather than a technological phenomenon. The AI tools are genuinely more productive; the workers using them are genuinely more productive. But the institutional arrangements determine who captures the productivity gains. In a competitive labor market, workers would capture gains through wage competition. In the AI platform economy, workers often see productivity gains captured by platforms while their effective wages stagnate or fall. This is the distribution problem in its sharpest form.

Origin

Though Autor has not focused on monopsony as a primary research program, his work on the implications of labor market power has engaged with the monopsony framework, particularly in collaborations with Anna Stansbury and in public writing on the policy implications of AI platforms.

Key Ideas

Labor markets are not competitive. Empirical work since Manning (2003) has documented systematic employer bargaining power, including in high-skill labor markets previously assumed to be competitive.

AI platforms concentrate the asymmetry. Network effects, contractor classification, and algorithmic control combine to give AI platforms monopsony power at unprecedented scale.

Productivity and pay decouple. When employers have bargaining power, productivity gains do not automatically translate into wage gains — the institutional determinants of gain distribution matter more than the size of the gain.

Policy responses exist. Antitrust enforcement, worker classification reform, portable benefits, and sectoral bargaining are among the institutional structures that could rebalance the gain distribution.