Here is the paper written by Roy van der Weide and myself on inequality and growth in the US. It is obviously a topic of key importance (does inequality help or hinder growth?) and the literature of the 1990s and early 2000s came to a dead-end mostly because of inadequate data and because they looked only at synthetic measures of both inequality (overall Gini) and growth (mean growth, like in GDP per capita).
Roy and I “unpack” both growth and inequality, and, for example, look at the growth rates at different percentages of income distribution in function of inequality overall, inequality among the poor and inequality among the rich. We use six large US microcensuses (American Community Surveys) that cover the period 1960-2010 at decennial intervals. We find that high inequality now lowers the future growth rate of the poor, and raises that of the rich. The summary of our data, methodology and findings was published by VoxEU.
This empirical question has recently acquired added relevance because of the slowdown of growth in rich countries and the simultaneous rise in inequality. It has inspired extensive debates over the years, and also today the opinions among economists (and politicians) about the importance of inequality for future growth are still very much divided. The responses to our study too have been very diverse. Compare for example a recent op-ed in the Salon to a recent op-ed in Forbes.
For clarity sake, let’s start with the definition of income we use. Our income data are gross income or what is called here in definitions given by Current Population Survey, Post-Social Insurance income, that is market income (the same concept as used by Piketty and Saez) plus government cash social transfers (meaning social security, unemployment benefits and whatever transfer you receive from the government in form of cash).
Post-Social Insurance Income does not include direct taxes paid by households nor in-kind benefits like food stamps received mostly by the poor. Earned Income Tax Credit is not included because its inclusion requires two steps: first, to find out who, among the surveyed population is eligible for it, and second, to find out who indeed took it. This can be done only through tax simulations. But note that food stamps, at their money equivalent, are included in CPS disposable income, as they are in LIS data for the US available here (up to and including for year 2013).
I agree with Worstall that it would be better if our data included all non-cash transfers, like food stamps and housing vouchers. But Tim Worstall is not correct that this would alter our results and that the rate of income growth of the poor would be higher than what we find in the paper because food stamps have been in existence since 1961 (reminder: our data start in 1960) and are thus not included throughout the duration of the study. In other words, there was no change in what the income concept we use covers.
Let’s move to other non-included benefits. Medicaid, that people often mention in discussions these days, is not included, no more than government-funded medical or education services are included in other countries. It is just extremely difficult to do imputation of public health and education benefits (you have to have simultaneous estimation of people’s incomes and use of health and education and knowledge of what services are subsidized and what not). Moreover the imputation for health services depends on who has been sick during the survey year and has received such government-funded help. But if I was not sick, having a potential access to such services, is still surely good for me, but it will not be recorded in the survey.
Nora Lustig and a number of associates have recently been working on such more comprehensive concept of income that would include among transfers all government-funded health and education and would also deduct indirect taxes (like sales taxes) that are normally regressive. (In order to do the latter, you need to have an integrated income and consumption survey to know who purchases what and how much in sales taxes they pay.)
The bottom line regarding the inclusion of these benefits In the US.
1) Tim Worstall is right that the inclusion of some in-kind transfers (food stamps and housing vouchers) in income would provide a better estimate of income.
2) He is probably not right that this has an influence on our results because these was no change in the treatment of these transfers during the entire period that our paper covers.
3) The inclusion of other in-kind services like health and education should also be (ideally) done, but this is not an easy matter. Note also that while their inclusion, might show that inequality increase in the United States was less than we currently think (because spending on such health services for the poor increased over time), the inclusion of the same health and education benefits in international cross-sectional data, would lead to a relative decline of the US median income because such benefits (which are currently not included anywhere) are much greater in European countries than in the US.