Are starting wages reduced by an insurance premium for preventing wage decline? Testing the prediction of Harris and Holmstrom (1982)

In the model of Harris and Holmstrom (1982) labour market entrants pay an insurance premium to prevent wage decline. As employers are unable to assess the ability of an entrant, they would offer a wage equal to expected productivity of the worker's category and adjust it with unfolding information on true individual productivity. Workers are willing to accept a reduction in starting wage to prevent a reduction in their wage when their productivity is revealed to be below the expected value for their category. Harris and Holmstrom assume that entrant ability risk can be measured from wage variance of experienced workers, implying a negative effect of later variance on starting wages. Using Portuguese data covering virtually all private sector employees, we find that the prediction is unequivocally rejected. Robustness, flaws and pitfalls of the test are discussed.