We analyze several statistical assumptions used in empirical models on
public-private sector wage structures. Based on data for Germany, whi
ch contain a large range of background variables usually not available
in other studies, we investigate the sensitivity of results to variou
s specification and identification assumptions. The standard switching
regression model is extended to allow for endogeneity of education le
vel, experience, and hours worked. These extensions lead to considerab
ly different parameter estimates. We compare conditional and unconditi
onal wage differentials between the public and the private sector for
the various specifications. These differentials are sensitive to ident
ification assumptions, but robust across specifications which do and d
o not allow for endogeneity of education, experience, and hours worked
. (C) 1998 Elsevier Science B.V. All rights reserved.