We test the signalling hypothesis using detailed flow-based employer–employee data from Denmark. Our main findings are: the signalling value is proportional to the proportion of employees dismissed from the workplace and the signal is uniform across occupation groups.