计量经济学 Wage Function.ppt

* The difference between the advantaged and disadvantaged coefficient vectors is ?b = b* - b* implying b* = b* - ?b, (3) By substituting the second term into the equation on the top of this page and rearranging, we get This equation states that the mean difference in log earnings between the advantaged and disadvantaged groups can be decomposed into the effects of differences in their average endowments (the first term on the right-hand side of the equation), and the effects of discrimination, as revealed by differences in estimating coefficients (the second term). * where average endowment differences are now weighted by advantaged workers’ estimated coefficients and coefficient differences are weighted by mean characteristics of the advantaged workers. ? ?b = b* - b* implying b* = b* - ?b, (3) An equally plausible alternative procedure involves solving the first term in Equation (3) for b* rather than b*, substituting for b* into Equation (2), and rearranging. This yields a variant of Equation (4). * Note that average endowment differences in Equation (5) (the first term of right-hand side) are weighted by disadvantaged workers’ estimated coefficients, while differences in the estimated coefficients are weighted by average characteristics of the advantaged workers. The choice between use of Equations (4) or (5) to measure discrimination is essentially a classic index number problem involving a decision on which weights to employ. In practice, empirical researchers have typically employed both equations, arguing that together they bracket the estimated effects of discrimination on wages. * Note also that the above framework can easily be extended to apply to comparisons of numerous disadvantaged groups with a common advantaged group and with each other. To important caveats are worth mentioning. First, regardless of which procedure is employed, to provide an accurate measure of discrimination, the list of regressors in Equation (1) must include all r