Public-PRIVATE Pay Differentials
319
Table 6. (Continued) Marginal effects of the reduced form and switching regression
for males and females
|
Males |
Females | |||
|
Reduced |
Structural |
Reduced |
Structural | |
|
form |
form |
form |
form | |
|
Skilled non-manual |
0.141 |
-0.022 |
0.169 |
-0.496 |
|
(0.065) |
** (0.725) |
(0.091) * |
(1.658) | |
|
Skilled manual |
-0.096 |
0.111 |
0.126 |
-0.006 |
|
(0.042) |
** (0.828) |
(0.123) |
(2.013) | |
|
Wage differential |
1.322 |
2.964 | ||
|
(2.97) *** |
(P) (BC) |
(3.029)*** | ||
|
Observations |
677 |
516 | ||
Notes: The male model is based on the univariate probit specification with sample selection
correction for sector choice only and the female model on simple OLS. *,**, and *** denote
significance at 10%, 5% and 1% respectively.
The most interesting variable is certainly the predicted wage differential
between the public and private sector. Clearly, there is evidence that wage
differentials impact significantly on the sector choice. Both the male and female
effects on the expected wage differential between the public and private sector are
positive and highly significant. The marginal effect seems to be by far the largest
compared to other regressors. This is similar to what Hartog and Oesterbeek (1993)
find in their study for the Netherlands using a MLE approach and Lee (1978) for
the union/non-union decision. However, in both cases the coefficient rather than
the marginal effect is reported.
The interpretation of the marginal effect is cumbersome. Given that the predicted
wage differential is the difference in two log terms, a log percentage change in the
predicted wage differential increases the probability of choosing public sector
employment by roughly 1.3 and 2.9 % for men and women respectively.18 This is a
18 Neither Lee (1978) nor van der Gaag and Vijverberg (1988) put a meaning on the coefficient
of the predicted wage differential. On the other hand, Hartog and Oesterbeek (1993) interpret
the coefficient as selection elasticity.