This paper presents an analysis of how firing costs impact the wages. The analysis exploits the changes of the employment-at-will introduced in different states from 1977 to 1997 in the United States, to evaluate how these increases in firing costs affected the wages of individuals with varying levels of education. A quasi-experimental approach reveals negative effects of these reforms at both the top and bottom of the wage-education distribution, indicating a polarized impact. The calibration of a standard search and matching highlights the crucial role of the probability of match-specific productivity changes in explaining the strong negative effect of firing costs on wages at both ends of the wage distribution.