How Not To 'Improve' Income Trend Estimates
This article originally appeared on Forbes.
There’s a new paper by a German economist named John Komlos claiming that when he “improves” income estimates from the Congressional Budget Office, “most Americans are scarcely better off than their parents were a generation ago, except those at the very top, and considering relative incomes almost all are actually worse off.” Wrong, wrong, wrong.
Although it’s not entirely clear what he’s done, he does indicate he is working from CBO’s latest spreadsheet. I’ve used this spreadsheet many, many times, and his findings immediately rang false to me. I whipped up my own estimates in 20 minutes, and they consistently show what they always have shown—substantial gains that people today would not be indifferent about losing. Komlos’s “high end” estimates are not that different from mine, but he should have used them as his “low end.” Instead, his low-end estimates depend on a discredited methodology that is known to be too pessimistic. And that alone lets him make apocalyptic statements about a hollowed-out, floundering middle class.
First, let’s go over Komlos’s “improvements.” CBO includes as income the amount that is taxed away from workers and investors by the corporate income tax, allocating some of the tax as income to workers and some as income to investors. Komlos subtracts those out. It’s unclear if he also deducts corporate taxes from what’s left of income. If he subtracts corporate income taxes from households’ income but he doesn’t credit them with the income taxed away, then he is essentially reducing this form of income twice—by not including it as income and by taking it away as taxes.
Second, Komlos converts the income from the employer’s share of payroll taxes to present values, to account for the fact that this income cannot be spent (in the form of Social Security and Medicare benefits) until a worker retires. But since people tend to get more in benefits from these programs than they put in, the future income that should be discounted is higher than the employer’s share of payroll taxes today. A far better choice would be taking the income of the retirees from private savings, from Social Security, and from Medicare and deflating it to account for the increase in the cost of living since the income originally earned was saved and the time value of money. That is, one should adjust on the back-end, later in life, not on the front-end.
There’s possibly a bigger problem with his treatment of employer payroll taxes though. If he has deducted the full amount of payroll taxes—from employee and employer—from income at the same time he discounts part of the income that is taxed away, then the problem is akin to the way he treats corporate income taxes.
As it turns out, these first two “improvement” don’t really matter all that much. A third “improvement” Komlos makes matters more. He subtracts employer and government health insurance from income. Ignoring employer-sponsored health insurance is inconsequential he says, though I’m not so sure about that. To justify ignoring Medicaid and Medicare he claims that the benefits they provide haven’t increased “meaningfully” over time.
This is a bold claim—one that should be accompanied by an argument for why cost-of-living adjustment fails to account for pure price increases in health care that are unrelated to quality improvements. If our price indices are doing their job, then they should winnow rising income from health insurance down to the part that reflects real improvement. Further, I’d ask Komlos why he thinks these Bureau of Economic Analysis and Bureau of Labor Statistics economists are wrong to conclude that, actually, our price indices tend to strongly overstate inflation (i.e., to understate the “income” or utility provided by increased health care purchasing power).
And about price indices…Komlos’s fourth “improvement” is to show results using the “CPI-U” to adjust income growth for inflation in addition to the “PCE,” which is what CBO uses. This improvement is no improvement at all—it’s a big error.
The PCE (“personal consumption expenditures deflator”) is produced by the Bureau of Economic Analysis in the Commerce Department. My readers will know that the case for using the PCE is awfully strong. The reason is that it attempts to fully account for “consumer substitution.” When prices of some things go up, consumers can switch to other purchases, and the loss in utility they experience is not as great as it would be if they could not switch.
The CPI-U, produced by the Bureau of Labor Statistics, only accounts for any substitution bias going back to 1999. The estimates before that make no attempt at all to do so. Even the correction since 1999 is only partial. Since then, the CPI-U recognizes that consumers can buy more red delicious apples when gala apples become more expensive, but it does not recognize that consumers can buy more oranges when apples become pricier.
Now, Komlos could have used a third index—the “CPI-U-RS”—which carries the post-1999 improvement and others back to 1978. It’s far better than the CPI-U, which is why serious scholars of income trends who insist on using one of the CPIs use the CPI-U-RS. The major federal statistical agencies switched away from the CPI-U for their analyses in the late 1980s and early 1990s.
In short, Komlos should not be using the CPI-U. He appears to be unaware of the problem, because his you-say-potato discussion of price indices does not even mention substitution bias in arguing that well, we just don’t know which index is better. At any rate, Komlos is wrong to say that the Census Bureau and the Bureau of Labor Statistics use the CPI-U. The Census Bureau uses the CPI-U-RS for its historical income analyses, and the Bureau of Labor Statistics prefers the CPI-U-RS and yet another index—the chained CPI.
The existence of the chained CPI is important because it is the one index produced by BLS that attempts to fully account for substitution. It only goes back to 1999, unfortunately, but as I have shown, it tracks the PCE more closely than it tracks the CPI-U-RS or CPI-U. Any other points about the merits of the CPI-U over the PCE are irrelevant given that the chained CPI is the most sophisticated index produced by BLS and closely follows the PCE. In fact, research suggests that all of these indices overstate inflation, meaning that using the PCE produces relatively conservative estimates of income growth. The income trend estimates produced using it should be considered the “low estimates,” not those from using the CPI-U (which should be considered the “definitely too low estimates”).
Komlos reports that the average household income of the bottom fifth rose by 18 to 39 percent from 1979 to 2011, and by 7 to 26 percent among the middle fifth. When I computed my own estimates, excluding health insurance and using the PCE, I found the increase at the bottom to be 35 percent and that in the middle to be 19 percent.
Since these are both in Komlos’s ranges—and closer to his “high estimates”—the way he deals with corporate and payroll taxes doesn’t really affect the results that much. Using the CPI-U drops the income growth estimates way down, but these estimates should be wholly disregarded.
These estimates, like Komlos’s, assign no value to health insurance provided by an employer or the federal government. That can’t be right—no one is indifferent between having health insurance or not having it. Let’s compromise—let’s add to income half the value that CBO assigns health insurance benefits. If you do that, the income of the bottom fifth rose by 42 percent ($6,300) and that of the middle fifth rose by 27 percent ($11,600). Those are both above Komlos’s “high estimates.”
These estimates are affected by the aging of the population. While Komlos doesn’t use them, there are tables in the CBO spreadsheet that allow one to produce separate estimates for the nonelderly and elderly populations. Nonelderly households in the bottom fifth saw their income rise by 36 percent ($6,000) and those in the middle fifth by 30 percent ($13,200). Among the elderly, the bottom fifth’s income went up by 24 percent ($3,000) and the middle fifth’s by 30 percent ($10,800).
Komlos reports the changes in income over time in terms of annual growth rates. This causes him to deem the 0.7 percent per year growth in the middle fifth’s income as “reinforcing the general impression of a floundering middle class.” “Floundering” is an interesting adjective to describe a group that is, by his estimates, richer by $8,100 than its counterpart in 1979. But as my estimates indicate, the real gain for the middle class was over $10,000.
If you are indifferent between having the income of your counterpart in 1979 and having an additional ten grand, raise your hand. It is more appropriate to conclude that middle-class income growth has slowed and greatly lags growth at the top. But the slower middle-class growth rates began in the 1970s, predating the rise in income concentration by a decade. Today, median household income is essentially at its historical peak. If income had grown faster, it would be higher still, but let’s not pretend that we’re floundering, and let’s not blame inequality for a growth slowdown that most industrialized nations have experienced.
Finally, the question of whether adults in the current generation are better off than their parents—and by how much—is a different one than whether the median household today is richer—and by how much—than the median household in 1979. Using median trends to make inferences about upward mobility is a recipe for disaster. Using longitudinal data, I found that among adults in their early 40s in 2005, the median household income was higher by 39 percent than the median among their actual parents when they were adolescents. That’s pretty consistent with what CBO shows for the change in the median.
But that does not mean the median change experienced by these adults (compared with their own parents’ income) was a 39 percent increase. Some adults experience upward mobility while others do worse than their parents. But smaller intergenerational increases in income at the bottom can translate into bigger percentage increases than larger absolute gains at the top. An adult with $10,000 has twice the income of their parents who had $5,000; another with $500,000 has just 25 percent more than his parents with their $400,000. The median change experienced by today’s adults relative to their own parents was a 93 percent increase.
So, no, today’s adults are not “scarcely” better off than their parents—they are quite a bit better off. This upward mobility has occurred in an environment of slow income growth below the top, but one healthy enough to raise middle-class incomes significantly. Any other conclusion is unsupported by the best evidence.
Scott Winship is the Walter B. Wriston Fellow at the Manhattan Institute for Policy Research. You can follow him on Twitter here.
Interested in real economic insights? Want to stay ahead of the competition? Each weekday morning, e21 delivers a short email that includes e21 exclusive commentaries and the latest market news and updates from Washington. Sign up for the e21 Morning eBrief.