A brief history of inequality in modern economics

by
Editor
My articles My siteMy booksMy videos
Follow on:Twitter

from James Galbraith and Jaehee Choi and issue 92 of RWER

In the years following World War II the division of labor between neoclassical micro- economics and pseudo-Keynesian macroeconomics was pioneered at MIT and disseminated worldwide from there. Macro held a narrow strip of economic territory: unemployment, inflation, interest rates and money supply, the business cycle, the rate of growth and their interrelations through the quantity theory, the Phillips Curve and Okun’s Law. The personal distribution of income fell squarely into the microeconomics of labor markets, governed by supply and demand for various levels of skill, alongside such ad hoc matters as firm-size effects, industry-specific labor rents, imperfect competition and efficiency wages. A theory of changing inequality was offered for developing countries by Simon Kuznets in 1955, positing a rise in inequalities in the early stages of development but a decline later on. For the rich,  the Kuznets evolution was supposedly complete, the Cobb-Douglas distribution theory with Hicks Neutral Technical change predicted stable functional shares, and national income accounts appeared to bear this out. So the functional distribution – the division between wages, profits and rent – was hardly spoken of.

Beginning in the late 1970s and early 1980s, circumstances began to force a change. An early hearing on rising inequalities at the Joint Economic Committee (1982)¹ pointed an accusing finger at right-wing policies, and this message was restated by Bluestone and Harrison (1988), who laid the blame on de-industrialization and the war on unions, conspicuous features of the Reagan and Thatcher years. The point seemed obvious enough, but there was a subtle difficulty. The severing of micro from macro made it conceptually difficult for many economists to tie the Reagan Recession of 1981-82 and its UK counterpart major drivers of deindustrialization – to a distributional outcome. Instead the emphasis fell on specific anti-worker political actions – in the US these included the firing of air traffic controllers, deregulation of trucking, a radical-right National Labor Relations Board. Still this was a minor muddle compared with what was to come.

It was only in the early 1990s that mainstream economics began a concerted search for a less-contentious explanation of rising inequality, rooted in the labor market analysis to which distribution issues had been consigned. Given the evolving preference of applied micro- economists for data based on surveys of household characteristics – however limited these may be by survey-takers’ fixation on race, gender, age, education and a handful of similarly simple categories – the evidentiary basis for a labor market analysis of inequality was remarkably thin. It consisted of little more than widely-separated surveys of earnings, stratified by worker characteristics, and largely confined to a small handful of wealthy countries.

Bound and Johnson (1992) set the template for neoclassical investigation. Rising in- equality was a matter of changing relative demand for skills, a characteristic unobservable in practice but usually approximated by the number of years spent in school. Demand being driven by technology, the underlying cause had to be a “bias” in the character of technological change. The remedy to the resultant inequality could only be an increased supply of skill – more years in school. This remedy had the peculiar feature that if enough people pursued it, the advantage accruing to each would diminish until it disappeared. Education was economically worthwhile, but only if it is restricted – a truism that is nevertheless in its way subversive. The labor economists Goldin and Katz (2008) eventually produced a thick book on this theme, from which the ugly class politics of the 1980s had disappeared.

The discipline of economics is such that to have purchase with the profession, any argument counter to “skill biased technological change” had to adapt the same broad framework of  labor market supply-and-demand. Such an alternative was presented by Wood (1994), who argued that North-South trade in manufactures would expand the effective supply of unskilled workers in the Global North, driving down their wages in rich countries but raising them among the poor (where Wood argued factory workers form an intermediate skill class) thus moving inequality in opposite directions in the two hemispheres. Wood’s argument gained an audience briefly but was ultimately dismissed by the mainstream; among other things the encouragement it would have given to skeptics of free trade made it politically incorrect.

In the mid 1990s an analysis based loosely on the Kuznets hypothesis revived, thanks in part to efforts at the World Bank to begin to compile a comprehensive global data  set  of inequality measures, along with income measures prepared by the Penn World Tables and Purchasing Power Parity (PPP) estimates of the relative purchasing power of different national currencies. Fairly soon after the publication of the landmark Deininger and Squire (1996) data set there were multiple efforts to trace the growth (or decline) of inequality on the world scale, resolving roughly into three conceptual measures as described by Milanovic (2005): inequality between countries pure and simple (Concept I), inequality between countries weighted by population (Concept II), and inequality across individuals or households irrespective of nationality (Concept III). The diversity of concepts brought with it new sources of uncertainty in the result and indeed inconsistent – on more precisely, divergent – conclusions depending on the concept deployed. Thus, while inequality between countries (Concept I) tended to rise, inequality between countries (Concept II) fell. The difference was largely due to the rise in average Chinese incomes. Meanwhile Concept III inequality could be calculated only by merging data sets from different countries, a task of heroic proportions; the extensive data requirements meant that only few years (initially just three) could be brought to fruition. Changes in Concept III inequality from one period to the next generated the famous “elephant curve” showing sharp gains for those at the very top of the global income scale, substantial gains for the lower middle (mostly Chinese and Indian) masses, and stagnation for the incomes of the middle classes in the already-wealthy countries. These numbers too were driven largely by national-average movements (mainly the rise of average incomes in China) rather than by measures of inequality per se.

At the other end of the measurement-method scale, the Luxembourg Income Study set out to blend and homogenize household and personal income surveys so as to permit detailed and accurate welfare comparisons – but with the limitation that such surveys are sparse, restricted mainly to the wealthy countries and for the most part to recent years. What one gains in fine detail on household characteristics one loses on the capacity for extensive international and historical comparison. In these matters, there are different ways to process a finite body of data but, methodologically speaking, there is no free lunch.

In this cacophony of facts and semi facts, Kuznets’ straightforward and intuitive hypothesis  did not fare well. Indeed, most researchers citing Kuznets were not much interested in his narrative of intersectoral shifts; rather they sought inverted–U curves anywhere they might find them and made that the test of Kuznets’ thesis, irrespective of whether there existed an underlying framework of early-to-late transition from agriculture to industry and from rural to urban life.

For many researchers by then, the relation of inequality to income level was no longer of prime interest. Debates over development, education, industrial policy (the East Asian Miracle) and economic growth directed attention toward the link between initial levels of inequality and later growth rates. Two competing strands emerged. One held that low levels  of inequality were good for growth (Birdsall et al., 1995) – citing Korea, Taiwan, Post-Mao China but largely ignoring East Germany and the USSR – while the other advanced the opposite thought, that income and savings must first be concentrated before investment and growth will follow (Forbes 2000). A fair summary of these debates is that by choosing periods, countries, data sources and econometric techniques with sufficient care, either argument can be made. But whatever the result, this literature bore only a slight resemblance or relation to Kuznets. An exception is the work of Deaton (2015), who argues that improvements in human welfare must start by increasing inequalities along the relevant dimension, whether life expectancy, infant mortality, years of education or any other index. Only after an improvement has taken root somewhere first, will it be adopted broadly and so eventually inequalities along that dimension will decline. read more