The Wrong Way To Think About Inequality

In the last month, a new paper from Michael Norton and Dan Ariely has drawn some attention to the issue of wealth inequality in America. The paper, called "Building a Better America -- One Wealth Quintile at a Time," finds that Americans underestimate the inequality of America's wealth distribution and express a preference for a more equal distribution. Indeed, while the top quintile of Americans hold about 84% of national balance sheet wealth, survey respondents believe the figure is just 59% and would prefer a figure of 32%. The authors use the paper to argue for more redistributive policies -- or rather, for the insertion of these public preferences into policy debates.

The paper has been widely discussed in the blogosphere (for example by Matt Yglesias on the left and Reihan Salam on the right.) I am unimpressed with the paper for a few reasons, and generally think we should be cautious about the idea that America needs sharply more wealth redistribution (especially to the radical degree that would be implied in the paper.) I do think that there are valid reasons to be concerned about distribution of resources, or to favor more egalitarian policies, but that they are not implied by this paper.

I have three key objections to the paper. First, in asking respondents to develop ideal wealth distributions, the authors told survey respondents to imagine they would be "randomly assigned" to one of the wealth quintiles -- implying that effort plays no role whatsoever in wealth accumulation. Second, there is little reason to believe that the public is good at evaluating ideal distributions of wealth -- as demonstrated by the impossibility of the preferred wealth distribution found in the paper. Third, the paper focuses on wealth distribution, when income distribution is a better metric for inequality.

The first objection undermines the authors' finding that Americans would prefer a highly equal distribution of wealth, where the top quintile's wealth holdings (32%) would outstrip the bottom quintile's (11%) by less than three-to-one. (In fact, the bottom two quintiles in America hold approximately zero net wealth.) In forming this ideal distribution, respondents were told they would be "randomly assigned to a place in the distribution... from the very richest to the very poorest."

Essentially, they were being told to discount the possibility that they would work to improve their lot in the distribution, if they so chose. They also were not advised that the economic policies required to achieve such an equal wealth distribution would shrink the economy overall. This question framing may have helped point respondents to endorse a wealth distribution that could not be produced by a set of policies observed in any country with a high degree of human development.

This leads to by second objection. I noted that respondents expressed a preference for a wealth distribution with 11% of wealth in the hands of the bottom quintile. In a section of the paper called "Americans Prefer Sweden," the authors note that over 90% of respondents prefer a wealth distribution modeled on Sweden's, including 11% of wealth for the bottom 20% of people, to that of the United States.

Except that Sweden's bottom quintile doesn't actually hold anywhere close to 11% of that country's wealth. If you notice footnote 2 of the paper, you'll see this comment: "We used Sweden's income rather than wealth distribution because it provided a clearer contrast to the equal and United States wealth distributions; while more equal than the United States' wealth distribution, Sweden's wealth distribution is still extremely top heavy."

That is, the authors took a completely different measure of inequality and presented it as a chart of wealth distribution by quintile -- the chart does not represent a wealth distribution actually observed in a country. In fact, it is unlikely that any advanced country has a wealth distribution with anywhere close to 11% of wealth held by the bottom quintile.

There is a reason that the bottom two quintiles of households have essentially no net worth, in countries all around the world: people with low incomes tend to consume their incomes rather than saving them. They do this partly because saving is a luxury relative to their consumption options, and partly because they expect higher incomes in later years and are smoothing consumption over their lifetimes. This is true even in countries with significantly lower income inequality than the United States, such as Germany.

Unlike income distributions, the World Bank doesn't produce international comparisons of wealth allocation. But the paper that Norton and Ariely cite for wealth distribution statistics has figures for a number of countries. Of the figures it contains, the highest wealth share for a bottom quintile in an advanced country is 2.1 percent, in Japan -- a far cry from 11 percent. (China's bottom quintile holds 2.8 percent of that country's wealth.)

Other advanced countries closely track the minus 0.1 percent figure in the United States, which means that bottom-quintile households have slightly more debt than assets: minus 0.2 percent in Germany, 0 percent in Australia. (There is something screwy with the figures for Denmark and Sweden, which show the least-wealthy household quintiles having sharply negative net wealth; that seems unlikely, and the source data are in a language I don't speak.)

A drop in income inequality would largely serve to increase consumption by poor households (a perfectly reasonable policy goal), not to increase their wealth. No plausible set of tax-and-transfer policies could produce the wealth distribution advocated in this paper. The only way you could get the bottom quintile's wealth share into double digits would be to force these households to save large shares of their income that they would prefer to consume.

For example, if we achieved Sweden's income distribution (the most equal among the world's advanced countries) we would also need to have equal saving rates in each quintile in order for the bottom quintile to hold 11 percent of wealth. This would not be desirable: low-income households get more utility from saving less and consuming more.

So, what should we make of the fact that 92 percent of study participants thought a wealth distribution with 11 percent in the hands of the bottom quintile looked better than the actual wealth distribution in the United States? I'd say the key takeaway is that members of the public are not good at looking at pie charts of wealth distribution and deciding which represents a good society. It's a bit like asking people what's the best mix of materials to use when making a jumbo jet -- how on earth should they know?

Finally, I think this study would have been a lot more interesting if it had asked about income distributions rather than wealth distributions. Net worth misses a lot of important but off-balance sheet assets and liabilities that we hold. One is human capital -- a recent graduate of medical school is likely to have a negative net worth but is not "poor" by any reasonable definition. Another important asset is the expectation of receiving future government benefits, including Social Security and Medicare. If these factors are included, America's wealth distribution becomes significantly less skewed.

And except for very rich people, the income statement is a far better predictor of living standard than the balance sheet. Think about a middle-class family of four with annual expenses of about $60,000 after taxes. In determining whether the family could meet those costs, would you first ask about family assets or family income? Over the next year, most people will rely much more on human capital than on balance sheet wealth to support themselves, which makes balance sheet measures a weak indicator of need.

There is an important discussion to be had about income inequality and the desirable level of progressivity in government policies. But this paper, which points toward an outcome that could only be achieved with extremely undesirable policies, does little to inform that debate.