Over at the Conversable Economist yesterday (March 2), Timothy Taylor, as he does virtually every day, has an excellent post. This one is on various measures of the U.S. poverty rate over time. It’s titled “Winning the ‘War on Poverty’–and Now What?

Here’s an amazing graph that he takes from Richard V. Burkhauser, Kevin Corinth, James Ewell, and Jeff Larrimore, “Evaluating the Success of President Johnson’s War on Poverty: Revisiting the Historical Record Using a Full-Income Poverty Measure,” December 2019, IZA DP No. 12855.

 

As you can see, a more-inclusive–and quite justified–measure of income shows a much lower poverty rate than the official one reported. It’s low single digits versus low double digits. That’s a pretty big difference. (That is, if you can see. The graph is a little hazy, so to see it better, go to Tim’s post referenced above.)

The various adjustments, for size of household, for a better measure of cost of living (the Personal Consumption Expenditures price index instead of the Consumer Price Index), for after-tax income including cash benefits, and for the market value of non-cash benefits, all make sense. And the 3.3 percent poverty above doesn’t yet adjust with the PCE instead of the CPI.

Also, Tim has one of the nicest, most-succinct explanations of why the CPI rather than the PCE is still widely used. Here it is:

But one interesting fact about the CPI is that after it is calculated, it is never revised–not even when the US Bureau of Labor Statistics later makes changes in the technical formulas used to adjust the CPI. Because various contracts and laws depend on the CPI, this lack of adjustment makes some sense. But if you want to know how the poverty rate should have been adjusted over time, it makes sense to use the most current methods for calculating inflation. One prominent measure of buying power that is adjusted over time is called the Personal Consumption Expenditures price index.

I have two critical comments, not of Tim’s post, but of the Burkhauser et al article that he draws on.

First, Burkhauser et al say that in including government medical benefits (presumably that’s largely Medicaid), they are valuing them at market. That’s tricky. To value them at market values, you can’t simply value them at cost. If you give me something that you paid $1.00 for, there’s no assurance that I value it at $1.00. Indeed, in a June 2015 NBER study (#21308), health economists Amy Finkelstein, Nathaniel Hendren, and Erzo F.P. Luttmer estimated that beneficiaries of Medicaid valued $1.00 of spending at only 20 to 40 cents. Burkhauser et al do address the Finkelstein et al point in a footnote. I didn’t totally understand the footnote. But the bottom line is that they seem to be saying they can ignore the Finkelstein et al point. They keep referring to market value of health insurance, even in that footnote, but I do not think that term means what they think it means. They seem to be assuming that the market value = cost. We know that that’s false.

Fortunately, even without their valuation of health insurance, the poverty rate is about 6.9 percent. Notice also their Figure 5 on page 18, where, using the PCE rather than the CPI, they get a poverty rate of only 2.3 percent.

Second, Burkhauser et al conclude that Lyndon Johnson’s War on Poverty worked. They write:

These results demonstrate that even though there remain individuals who are struggling financially based on today’s living standards, President Johnson’s War on Poverty—based on economic standards when he declared that war—is largely over and a success. This observation is apparent using our Full-income Poverty Measure since, unlike previous poverty measures, it is anchored to President Johnson scientifically arbitrary but policy relevant judgment with respect to the poverty population in 1963 and sets its initial thresholds accordingly. It also adjusts these nominal thresholds each year to hold them constant in real terms, and uses a full measure of real income. While this conclusion stands in stark contrast to conventional wisdom (and the poverty rates based on the Official and Supplemental Poverty Measures), it should not be surprising given the substantial resources contributed to transfer programs that are not counted by other poverty measures, as well as the overall economic growth that has occurred since the 1960s.

We need more evidence than the evidence they give. Look at what was happening to the poverty rate in the year or two before the LBJ programs kicked in. It was falling like a rock. Why wouldn’t we assume that continued economic growth would cause it to keep falling? You might say that the prima facie evidence for the War on Poverty’s success is “the substantial resources contributed to transfer programs” that Burkhauser et al highlight in the quote above. But we know that people respond to incentives. The implicit marginal tax rates for people receiving direct welfare, food stamps, housing subsidies, and Medicaid are sometimes 70 percent or more. When they earn income, they can lose 70 cents or more of benefits for every additional dollar they earn past some threshold. So of course many people won’t work to earn money when they can get these transfers and they are heavily penalized, by loss of transfers, for working. If we look at their incomes net of transfers, we are tempted to say that without the transfer programs, their incomes would have been the same as the data show. But that is highly likely to be false. Without those transfer programs, people would have made more income and possibly, in many cases, a great deal more.