No More Limited Mobility Bias: Exploring the Heterogeneity of Labor Markets

We propose a bootstrap method to correct limited mobility bias in the variance components of AKM models. Our method handles multiple corrections without increasing computational cost and works with any symmetric estimator of the error covariance matrix, including the one from Kline, Saggio, and Sølvsten (2020). Monte Carlo simulations show our method corrects the bias and is faster than alternative methods. Using French administrative data, we apply our method in four ways: (i) a full-sample variance decomposition of log wages under different assumptions on the error structure, (ii) correcting thousands of labor markets to study the relationship between market size and worker-firm or worker-coworker sorting, (iii) analyzing gender differences and (iv) life cycle patterns in wage components. In all cases, the corrections are important to interpret the results.