On the Origins of the Kuznets Curve

Today, I’m blogging from the Harvard University Archives. I’m five boxes deep into the Simon Kuznets collection, for my second visit related to my dissertation research. I just came across a letter that, while not especially important to my dissertation, seemed like it might be of wider use to anyone interested in the history of the “Kuznets Curve.” Simon Kuznets, in addition to spearheading the first official US national income statistics in 1932-1934, is probably most famous for this curve/hypothesis.* In his presidential address to the American Economics Association in December of 1954, later published as Economic Growth and Income Inequality, Kuznets argued that income inequality and economic growth might have an inverted-U shaped relationship. That is, as economic growth/development increases, inequality first goes up and then comes back down.

Kuznets based this empirical story on a few decades worth of data from the US and UK, and asserted in the conclusion that: “The paper is perhaps 5 per cent empirical information and 95 per cent speculation, some of it possibly tainted by wishful thinking.” (1955: 26) Since 1955, the paper has generated an entire literature asking whether or not the curve is real, whether it’s reversed in the past 30 years to create a new pattern, and so on. Much of the early research supporting the Kuznets curve came from cross-national studies which mapped inequality against current GDP/capita and showed the expected inverted-U shaped relationship. These studies, like Kuznets’, lacked longitudinal data. The recent consensus seems to be that the Kuznets curve does not hold within countries (e.g. Gallup 2012). That is, within a particular country, the relationship between economic growth and income inequality does not follow the expected path. Here are two graphs from Gallup 2012 that help sell the argument. The first is the relationship between log income and inequality, which looks like a Kuznets curve (though as Gallup notes, the fit is not quadratic for unlogged data):
Gallup 2012 Figure 3

The second is the author’s estimates of within-country income inequality trajectories (each line is one country). Note the chaos, and the absence of any coherent story, with some tendency towards a U-shape (not a upside-down U!):
Gallup 2012 Figure 8

Gallup further notes that Kuznets’ original data only had half of the original story. Yes, Kuznets shows a decrease in inequality in the US and UK in the late 19th century to post-WWII period. But Kuznets has no data on the initial upsurge in inequality! That is, Kuznets only has the falling half of the curve. Gallup graphs Kuznets’ income inequality data against historical gdp data now available from the work of Angus Madison to give some sense of what Kuznets was seeing:
Gallup 2012 Figure 1
Gallup (2012: 6) summarizes:

Kuznets showed that inequality fell in two high-income countries as they grew richer after World War I, but he had no evidence of rising inequality at low income levels. Ironically, Kuznets’ prediction that inequality will rise during the early stages of development, for which he had no evidence, is better remembered than his prediction that inequality will fall at higher incomes.

What did Kuznets have, if not data? He had a model. Specifically, Kuznets motivated his proposed empirical story (rising then falling inequality) with a two sector model of income inequality. Kuznets argued that agricultural incomes were likely as or more equal than industrial/urban incomes, but industrial incomes were much higher on average. Thus, as workers moved from agricultural to industrial labor, inequality would increase (low agricultural wages to higher industrial wages), until so few workers were in agriculture that a tipping point was reached and inequality came back down. This story falls apart, however, if agricultural incomes in the pre-development period are actually more unequal than industrial ones, because “if agricultural incomes are more unequal than industrial wages, perhaps due to unequal land ownership, movement out of agriculture into industry could reduce inequality right away.” (Gallup 2012: 7)

Ok, so why did Kuznets think that agricultural incomes were more equal than industrial ones? An archival document I just perused before starting this post sheds some light on the answer: Kuznets had some intuitions backed by absolutely no data. Data on income distribution are still poor now, and were even worse back in the 1950s, with historical data being extremely limited to a handful of studies (some conducted by Kuznets himself). These studies did not readily break incomes down by urban/rural or industrial/agricultural divides. Kuznets wrote to one of his collaborators, Selma Goldsmith (then at the Office of Business Economics, what is now the Bureau of Economic Analysis) to ask about the data she had been working on to produce official income distribution data for the United States (see, e.g., Goldsmith et al. 1954). In a letter dated August 15, 1954, Kuznets writes about his current work on economic growth and income inequality:

In my model, I somehow began with the assumption that inequality within the ag. sector is lesser than within the non-ag—although I had no evidence whatsoever. … the distributions that are available (in your D. of C. publication and elsewhere) show inequality among the ag. group (i.e. farm operator families) to be greater than among the non-ag—certainly in recent years. … Yet one would expect that with the much wider range of income opportunities in the non-ag sector, the inequality in size distribution would be appreciably wider in the latter. … Do you have any explanation for it?**

Let’s just stress that again: Kuznets had absolutely no evidence that inequality within the agricultural sector was less than the industrial one. Kuznets guesses that this discrepancy between his theoretical intuition and the data has something to do with the data not quite matching what they want – there’s higher year-to-year variability in farm incomes, but in the long-run the inequality within families might be less than appears in the data, or something like that. He’s not sure. And that uncertainty, among many others, is reflected in his strong caveats at the end of presidential address.

Years later, Kuznets himself would abandon the curve and caution against looking for historical laws of development, especially in cross-sectional data (Moran 2005). And yet the debate over it lives on, as does the search for laws of economic growth and development.

*In 1971, Kuznets won the Nobel Prize in economics, “for his empirically founded interpretation of economic growth which has led to new and deepened insight into the economic and social structure and process of development.”
**Papers of Simon Kuznets, Harvard University Archives, HUGFP88.10 Misc. Correspondence, Box 4.