Very cool, especially the use of a normalized semilog slope to quantify income disparity, and your final gradient chart on a log scale. That Excel decile gradient chart is a nice alternative to the boxplot (even Fabrice’s variant) when higher resolution is desired, and much easier to interpret than the Economist graphic.

Possibly the advantage of tracking mean income by decile is that it preserves the ability to exactly calculate an overall mean simply from the 10 decile points? If we used purely robust methods and tracked the median of each decile as you suggested, such data wouldn’t let us exactly reconstruct the overall mean — or even median. For practical purposes, either of those approaches allows a pretty decent reconstruction of the distribution, except for the ~ 5% tails.

Nice. The boxplots lose resolution more in the way the stacked bars against a linear income scale lose resolution: everything at the lower end is compressed. You could apply a log scale, and the disparity in income would be easy to compare.

Ran – That’s 1….