Hal Varian of UCB, writing in The New York Times, has put together a well-written but disastrously hyperlinked reexamination of the increase in income disparities in the U.S. during the Clinton years.
Many Theories on Income Inequality, but One Answer Lies in Just a Few Places
According to research being led economist James K. Galbraith at the University of Texas Inequality Project the tech boom led to concentrations of wealth in an incredibly small geographic area, leaving much of the country unchanged throughout the decade.
U.S. national figures show increasing income disparity over the course of the Clinton presidency. The increase in the Clinton era, however, was not due to the rich of American grabbing and keeping more of the nation's wealth for themselves. If one makes a single small adjustment to the data - replacing the economic growth experienced in 4 of America's 3,100 counties - Santa Clara, San Mateo and San Francisco Counties of California and King County, Washington - with the national rate of economic expansion, then the radical growth in income disparity in the years 1992 to 2000 simply fades away.
See the slide show:
Measuring Inequality - a practical workshop on theory and technique
Slide 61 shows the result of the thought experiment--though the presentation is worth going through for the interesting things it has to say about the structural differences in the growth of inequality in Russia and China.
And the discussion starting on page 11 of the paper:
Income Distribution and the Information Technology Bubble
So now how many people (please raise your hands) believe that one could find 4 municipalities of Japan which would have an equally powerful effect on national statistics--that is, if we replaced their growth since 1992 with the national average, one would see the shakai kakusa disappear?
What are those four municipalities?
Ok, let's see:
Tokyo Metropolitan District, Minato City...
P.S. The Inequality Project has a nice list of papers on rises in inequality around the world.
Nothing on Japan, though.