摘要：This paper’s purpose is to examine the importance of controlling for the comparability of the units in a sample when examining wage differentials of various types. We use U.S. CPS data for the period 1992-2000 to compare estimated wage differentials between public and private sector workers obtained using two different methodologies: Lee’s two-step method that controls for the selection bias resulting from the non-randomness of the sample, and the propensity score matching method that controls also for the comparability of the workers. Lee’s method suggests that federal workers are paid a premium, while state and local workers are underpaid compared to private sector workers. However the matching method indicates that this data is too heterogeneous to be used to compare wages across sectors. We conclude that, when the outcome under study is not only affected by some sort of selection but also requires comparable groups, the traditional methodology may not be enough.
