June 27, 2013
Thomas Edsall has a lengthy blogpost on a new measure of income developed by Cornell University Professor and AEI fellow Richard Burkhauser. Burkhauser’s measure reverses the widely reported finding that inequality has increased substantially over the last three decades.
While Edsall went to great lengths to include extensive comments from other economists (including me) on Burkhauser’s methodology and concluded himself that Burkhauser’s methodology doesn’t measure up, readers may still be led to believe that there is more ambiguity on this issue than is actually the case. This is because Burkhauser’s measure is so peculiar and counter-intuitive, that it is unlikely that many readers would understand what he has in fact done.
Burkhauser does address a legitimate question — the treatment of capital gains. Usually economists calculate inequality by both taking income without counting realized capital gains (sales of stock, houses, or businesses) and also including the gains. The latter will generally show higher degrees of inequality since wealthy people are likely to have realized capital gains, whereas middle and lower income people are not.
This approach does pose a problem since the decision to sell an asset is an arbitrary one and does not necessarily reflect when the gain actually took place. Also, a lower capital gains tax rate will encourage people to sell their assets more frequently, which by itself would lead to larger reported income. So a methodology that includes realized capital gains is problematic.
However Burkhauser’s response, to include unrealized gains, makes no sense in a serious measure of income. The reason is that asset prices (especially stock, but in recent years housing as well) are hugely volatile. For people who have substantial assets, the movement in these prices in any given year will often swamp their other income. Gary Burtless and I both made this point in our comments.
An implication of Burkhauser’s methodology is that our measure of inequality would depend hugely on the exact year we picked for our analysis. In his study, the base year for most of his analysis is 1989, a year in which the S&P 500 rose by more than 27 percent. This hugely increased the earnings of the top quintile in his base year. As a result, the change from 1989 forward would be guaranteed to be small. By contrast, if Burkhauser had chosen 1987, when the S&P fell more than 6 percent, he would have a much lower base. This would make the growth in income for the top quintile appear much larger.
To see this, imagine the average income, not counting capital gains, for the top quintile is $200,000 in both 1987 and 1989. Suppose they own $1 million in stock on average. In Burkhauser’s methodology their income in 1989 is $470,000. Their income in 1987 is $140,000. (Number corrected, thanks Yoram.) We would be telling a very different story about the growth of income inequality over the next two decades if we opted to choose 1987 as our base year rather than the year picked by Burkhauser. (The year 1989 is often chosen as a base year because it is a business cycle peak. That makes sense in a measure that is primarily reflecting earnings growth which tends to peak at the peak of the cycle. It makes no sense when taking a measure that is moved primarily by capital gains.)
There could be an argument for taking unrealized capital gains averaged over a longer period, which is not the methodology that Burkhauser chose. By this methodology we would average the capital gains for households over the period being investigated and add the annual amount to their income. This would be a considerably more defensible methodlogy, but it still would give very misleading results because of the housing bubble.
Depending on the exact starting point, the run-up in house prices could easily add 10-20 percent to the annual income of low and middle income households. Of course this is all reversed after the bubble burst in 2007. So this methodology would allow to say that there was little if any rise in inequality in the years between 1989 and 2007, but then a huge surge in inequality since 2007.
That is probably not a helpful way to think about trends in income. It makes more sense to keep income, excluding capital gains, separate from measures of wealth. This would allow us to see that the wealth of low and moderate income families did increase substantially up to the bursting of the housing bubble and then fell sharply.
There are some other serious issues with the Burkhauser methodology (for example the switch from defined benefit pension to defined contribution pensions would make earnings appear to rise), but there is one common item that is worth noting and criticizing. He counts the cost of government provided benefits as income. This has the effect of substantially raising the income of the bottom two quintiles because of Medicare and Medicaid.
The reason these benefits matters so much is that health care is expensive in the United States. The cost of Medicare for a couple is now close to $25,000. That will be very large relative to the income of people in the bottom two quintiles. While these benefits are quite important, the reason they are so expensive is that we pay way more for our health care than any other country. Our per person costs are more than twice the average of other wealthy countries. This is not because we get better care but because we pay more to providers.
This means that by including the cost of government provided health care as income for beneficiaries, if we raise the pay of cardiologists from $400k a year to $500k a year (thereby raising the value of Medicare and Medicaid benefits) we make the bottom two quintiles look better off. A simple alternative would be to price the value of care in accordance with the cost in reference countries with comparable health outcomes.
Getting back to the basic point, the whole Burkhauser project of trying to find a perverse methodology that will make rising inequality go away, is more than a bit annoying. This would be like an astronomer looking for quirks in planetary motion and using this as a basis for arguing that the sun actually revolves around the earth.
There undoubtedly are quirks in the motions of the planets, however no reputable astronomer will use them as a basis for disputing that the earth orbits the sun. Unfortunately economics does not have the same standards. Economists who want to argue the equivalent of the sun orbiting the earth will find substantial funding and an audience for their work.
Comments