This paper explains the AKM model, a widely used method to break down wages into a worker part and a firm part. The name AKM comes from the authors Abowd, Kramarz, and Margolis (1999). Using linked employer–employee data that follow the same workers across jobs, the model asks how much of wage differences come from who the worker is versus which firm they work for.

At its core the AKM model writes a worker’s wage (in logs) as a sum of three pieces: observed characteristics (like experience), a worker-specific component αi, and a firm-specific component ψj, plus a leftover shock. In symbols the paper shows Yit(j) = Xitβ + αi + ψj + Uit. The firm component ψj is a reduced-form measure of how much a firm pays its workers on average. The worker component αi summarizes a person’s persistent earning ability.

A key idea is that wage changes when people move jobs reveal the two parts. If many workers move between the same pair of firms, the model can separate persistent differences across firms from persistent differences across workers. The paper uses a simple example with three workers moving between two firms to show that, under the model’s additive structure, moving between the same firms produces the same percentage wage change for different workers.

This matters because it gives researchers a way to measure sources of wage inequality. With estimates of αi and ψj, economists can decompose overall wage dispersion into: worker heterogeneity (differences between workers), firm heterogeneity (differences between firms), and sorting (whether higher-ability workers tend to work at higher-paying firms). The AKM approach has been applied beyond wages, for example in health economics and corporate finance, and the authors provide a practical user’s guide and an accompanying notebook to help applied researchers.