Though inequality in the US has unquestionably risen since the early 1970s, by essentially any measure – I have a review of this evidence with L. Martinez (2008) here – Gordon argues that conceptual problems, such as comparing measures deflated with different deflators, imply a much lower rate of increase in inequality and a much smaller gap between productivity growth and income growth that is often reported. The paper usefully reviews much of the research since Piketty and Saez (2003), but is in many parts disingenuous. For instance, when adjusting for the location of college graduates, the education premium falls by two-thirds due to increasing college graduate concentration in expensive cities. Nonetheless, a useful measure of inflation shouldn’t consider a one-bed, one-bath apartment in New York the same as a one-bed, one-bath apartment in rural Michigan, as the New York apartment clearly embeds amenities not included in the price index. Similar stretches of logic are far too common in the working paper; it is most useful as a survey of the literature along with some sound advice about how to carefully use CPS and Income Tax data.
http://faculty-web.at.northwestern.edu/economics/gordon/SFONBER_Combined_090902.pdf
A Fine Theorem 