Re:Income inequality today may be higher today than in any o...
by mortehu on Saturday July 02, 2016 @09:20AM
(#12022386)
Quote from the summary of the article:
Consider 10 separate countries, cities, or other units
of equal size, with population 1000. Assume the wealth in
each unit follows a power law distribution, say a ParetoLomax,
all with the exact same parameters. Assume a
tail exponent of 1.1. The average Gini index as obtained
by direct measurement will be ≈ .71 per country. Now
aggregate them into a single country. The composite
Gini — as traditionally and currently measured — will
be ≈ .75, that is 6% higher – for the same sample.
This inconsistency implies not only that the Gini cannot
lend itself to comparisons between units of different size
but that intertemporal assessments are vitiated by the
population changes.
http://www.fooledbyrandomness.com/FatTails.html
"How to (not) estimate Gini indices for fat tailed variables":
https://dl.dropboxusercontent.com/u/50282823/GINI-1.pdf