2 ms·

Related: http://www.fooledbyrandomness.com/FatTails.html http://www.fooledbyrandomness.com/FatTails.html "How to (not) estimate Gini indices for fat tailed va

• mortehu 10y ago
  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.