Simon Kuznets conducted extensive research relating inequality to growth, culminating into a report published in March 1955 titled ‘Economic Growth and Income Inequality’.
Trends in Income inequality data are for the US, England and Germany, which he admitted proved to be a “paltry sample.”
In the US, in the distribution of income among families, the shares of the 2 lowest quintiles rose from 13⅟₂ percent in 1929 to 18⁰/₀ after the second world war. Conversely, the share of the top quintile declined from 55 to 44⁰/₀ and the top 5% dwindled from 31 to 20%.
In the UK, the share of the top 5% of family units decreased from 46% in 1880 to 43% in 1910 or 1913, to 33% in 1938 and then to 24% in 1943. The share of the lower 85% remained relatively constant from 1980 to 1913, between 41 and 43% but then rose to 46% in 1929 and 55% in 1947.
In Prussia, income inequality increased slightly between 1875 and 1913- the shares at the top quintile rose from 48% to 50%, the top 5% started from 26% to 30%; the share of the lower 60% lower stayed the same.
In Saxony, the contrast between 1880 and 1913 was relatively minor: the share of the 2 lowest quintiles declined from 15% to 14⅟₂ %. The share of the third quintile rose from 12% to 13%, the fourth quintile from 16⅟₂% to about 18%, that of the top quintile declined from 56⅟₂% to 54⅟₂ percent, and of the top 5% from 34 to 33%.
In Germany, as a whole, relative income inequality plummeted fairly sharply from 1913 to 1920s, in part largely to the annihilation of large fortunes and property incomes during the war and the subsequent inflation; but then began to return to pre-war levels during the depression of the 1939.
Kuznets maintained that even though approximations are made concerning the data, differences averaging 2 or 3 points would be redundant. Thus, the data was quite reliable, according to his perspective.
One must judge by the general weight and consequences of the evidence – which unfortunately is limited to only a few countries.
The Kuznets’ hypothesis has faced a lot of scrutiny over the years, most notably through empirical tests using cross-sectional data for various country sets. Most researchers who attempted these tests were aware that the best approach would be through time series data analysis.
Oshima (1950) was the first one to concur with Kuznets’ theory that inequality tends to widen in the early phases of growth then to narrow in the later phases of growth. Another prominent economist to come forward with his own approbation, Ahluwalia (1976) sampled 60 developing and developed countries to reach a firm conclusion. He used GDP figures measured in US dollars at 1970 prices and divided the population within each country into five quintiles; 20% of the population with the lowest income share pitted against the quintile with the highest income share. The results obtained from the regression showed that for all the quintiles save for the highest, the share of income falls with a rise in per capita GNP but increases thereafter. In closing, he maintained that the regressions provided significant weight to Kuznets’ assertion of a U-shaped pattern linking inequality and per capita income.
Anand and Kanbur (1993) sought to verify Ahluwalia’s claims using different functional forms. They initially found that the distributions weren’t comparable with regard to income definitions, population units and the gamut of the survey. They came up with a rather inconsistent data set and re-estimated the relationship put forth by Kuznets, opining that the U-shaped hypothesis was in fact the opposite of the real relationship.
Milanovic (1994) came forward with an augmented Kuznets’ Hypothesis in which the size of the income distribution takes into account factors that are considered as given in the short run, such as the level of income and regional heterogeneity in a country and social choice variables. He used a sample of 80 countries for his analysis and found that the relationship between income and inequality is quadratic, that is, inequality rises with income and then ebbs out. He conjectured that social choice variables reduce the Gini coefficient by some 13 points and when income increases, society tends to favour social policies that will lessen inequality.
Deninger and Squire (1996) successfully proffered a comprehensive and vastly improved data set on inequality. Their data didn’t support the notion of the inverted U-shaped relationship between income and inequality when they evaluated it on a country-to-country basis. In about 90% of the countries they tackled, they found no conclusive support for Kuznets’ hypothesis. Using decades of growth episodes as the evidence for the purported link, they didn’t find a conclusive relationship between growth and changes in aggregate inequality. The reason they came to that conclusion is because the changes in the Gini coefficient were too small to be taken into account.
Conversely, they also found that there was a strong correlation between aggregate growth and changes in income in all of the quintiles except the top one. For most of the growth phases in their sample, they derived this result. Even if the growth in income happened in conjunction with increases in inequality, the members of the lowest quintile experienced a rise in income.
Bruno, Ravallion and Squire (1996) used data from 63 surveys over a time period of 11 years (1981-1992) for 44 countries to reuse a number of the specificities to test the hypothesis found in most of the literature. They didn’t find the U-shaped relationship as conjectured by Kuznets and they couldn’t even reject the null hypothesis that the regression coefficients were jointly zero. They debated that the cross-sectional finding of the U-shaped link from many of the previous tests of the hypothesis that mainly used data from 1950-1970, was not robust. According to them, over time, sets of the distributed data became blurred and ambiguous.