The national statistics agencies of Canada, Sweden and the United Kingdom have annual series for gini coefficients going back to 1980 or so. Although cross-country comparisons of the levels should be taken with a grain of salt, the trends in these three countries are remarkably similar:
It is a Well-Known FactTM that inequality has been increasing over the past couple of decades, but it would appear that the greatest part of this increase took place before 1995 or so. Over the past ten years, inequality in market income has remained fairly stable.
Update: My initial, off-the-top-of-my-head conclusion was that these inequality patterns were driven by technology (eg., the familiar skill-biased technical change story) rather than country-specific things such as trade patterns, labour market structure, etc. But since I didn't have anything stronger than an off-the-top-of-my-head impression to back it up, I didn't mention it. But this study offers some support to this notion:
[T]he causes of rising inequality are primarily structural and related to new technology rather than to trade or institutions.
What's with the Swedish data around 1990?
Posted by: Adam | November 25, 2007 at 04:15 PM
Not sure - I don't understand Swedish. I'm guessing that there was a series break, which is why I plotted the graph as two separate lines, overlapping during 1989-1990.
Posted by: Stephen Gordon | November 25, 2007 at 04:23 PM
Great chart! (But I do cringe when people edit the Oy axis, out of generic-data-display/Tufte-is-king concerns.)
Posted by: Gabriel | November 25, 2007 at 05:43 PM
Great chart! I would not have guessed that the GINI coefficient for Canada had been relatively flat for the last 12 years or so.
Posted by: Mike Moffatt (About.com) | November 25, 2007 at 07:31 PM
LOL.. My copying Gabriel's first comment was entirely unintentional.
Posted by: Mike Moffatt (About.com) | November 25, 2007 at 07:32 PM
Ummm... Elephant in the room 1960-1990 female participation rate. I suppose it is not PC to mention it.
Posted by: reason | November 27, 2007 at 04:13 AM
Possibly, especially if it's accompanied with 'assortive marriage' (in which high-earners marry each other). It'd be a good idea to check female participation rates.
Update: I just checked to see if female participation rates stopped rising in 1995 or so. They haven't; the trend is still positive.
Posted by: Stephen Gordon | November 27, 2007 at 06:27 AM
Leo Johnson did an analysis of Income Tax Statistics that basically said that income distribution peaked in 1948 and has been downhill since then. He didn't compute Gini, he measured income by decile for tax filers. The bottom decile dropped in both absolute and relative terms from 1948.
http://books.google.com/books?id=EdqAfRb95vsC&pg=PA351&lpg=PA351&dq=leo+johnson+income+distribution&source=web&ots=YnDgVt5BNW&sig=L4rS_7daDxEoiWufLPhdDaqm7Sk#PPA352,M1
Is a reference from a book from Cy Gonick.
A version of his work could conceivably be found in the CRTC archives in the evidence presented to the Bell Canada "B" Rate Case from about 1972.
Posted by: Jim Rootham | November 28, 2007 at 12:35 AM
That didn't work try this http://tinyurl.com/2jqpzb
Posted by: Jim Rootham | November 28, 2007 at 12:40 AM
I'm still not convinced that supply side factors are not important. The proportion of well educated women entering the work force has also been increasing.
Posted by: reason | November 28, 2007 at 05:27 AM
"What's with the Swedish data around 1990?"
According to statistic Sweden the tax reform of 1990/91 affected the comparability between the time period before and after the tax reform. So for the years 1989 and -90 they calculated the numbers both according to the old and the new taxation system
Posted by: David | November 29, 2007 at 03:29 PM
Thanks very much for that, David.
Posted by: Stephen Gordon | November 29, 2007 at 09:40 PM
These estimates are biased by a wrong approach. Standard ways to calculate gini ratio adopted in many countries are based on an arbitrary portion of population. One has to bear in mind thgat there are MANY definitions of income and thus many gini ratios. Publsihed ar those based on the narrowest definition provided by internal revenue (tax) offices. Due to the smallest coverage they are the most biased.
So, one has to develop a better methodology.
SOme details are presented in my papers on Gini in the USA, which show no change in Gini related by personal incomes as measured by the Census Bureau in Current Population Surveys.
Kitov, I., (2007). Modeling the evolution of Gini coefficient for personal incomes in the USA between 1947 and 2005, Working Papers 67, ECINEQ, Society for the Study of Economic Inequality. www.ecineq.org/milano/WP/ECINEQ2007-67.pdf
Kitov, I., (2007). Comparison of personal income inequality estimates based on data from the IRS and Census Bureau, MPRA Paper 5372, University Library of Munich, Germany, http:// mpra.ub.uni-muenchen.de/5372/01/MPRA_paper_5372.pdf
Posted by: KIO | December 11, 2007 at 04:56 AM
These estimates are biased by a wrong approach. Standard ways to calculate gini ratio adopted in many countries are based on an arbitrary portion of population. One has to bear in mind thgat there are MANY definitions of income and thus many gini ratios. Publsihed ar those based on the narrowest definition provided by internal revenue (tax) offices. Due to the smallest coverage they are the most biased.
So, one has to develop a better methodology.
SOme details are presented in my papers on Gini in the USA, which show no change in Gini related by personal incomes as measured by the Census Bureau in Current Population Surveys.
Kitov, I., (2007). Modeling the evolution of Gini coefficient for personal incomes in the USA between 1947 and 2005, Working Papers 67, ECINEQ, Society for the Study of Economic Inequality. www.ecineq.org/milano/WP/ECINEQ2007-67.pdf
Kitov, I., (2007). Comparison of personal income inequality estimates based on data from the IRS and Census Bureau, MPRA Paper 5372, University Library of Munich, Germany, http:// mpra.ub.uni-muenchen.de/5372/01/MPRA_paper_5372.pdf
Posted by: KIO | December 11, 2007 at 04:57 AM
You should really use after tax income as it is closer to what people actually end up with. this ignores the role of social programs.
Posted by: Bob | September 14, 2008 at 02:22 PM