MEASURING DISCRIMINATION BY DECOMPOSING EARNINGS FUNCTIONS

An unrecognized problem is examined in a frequently used method for de composing earnings functions in the measurement of discrimination. The conventional method decomposes changes in the log of the geometric me an of the distribution, as opposed to the correct quantity, which is t he change in the log of the arithmetic mean. Using the log of the geom etric mean involves decomposing both changes in the mean and the sprea d of the distribution. But the mean and the spread of many earnings fu nctions are subject to change, so it is not quite clear what this deco mposition achieves. Two examples illustrate timewise and cross-section al comparisons.