Abstract
Empirical models of labor market competition usually assume that employers set wages noncooperatively, despite frequent allegations of collusive employer behavior. We propose an identification approach for labor market collusion that relies on production and cost data, and we use it to study how employer collusion affected wage markdowns of 227 Belgian coal firms between 1845 and 1913. We are able to detect collusion through the 1897 coal cartel without ex ante knowledge of its timing and find that it explains the fast growth in markdowns after 1900. We find that the cartel decreased both wages and employment by 6% to 17%.