The Cato Institute Misunderstands Basic Poverty Metrics

Edit: Original piece said Hughes works at Heritage, but he actually works at Cato. This was an error, though one could be excused from failing to see the difference.

Another day, another discussion of poverty that is loaded down with inaccuracies. This time, it's Charles Hughes of the Cato Institute, writing at The Federalist, who repeats falsehoods that have been circulating for some time among the self-styled smart conservative pundit sphere.

In the interests of brevity, let us discuss just this especially egregious part:

These defenders of the War on Poverty usually point to a different preferred metric, the Supplemental Poverty Measure (SPM). The SPM does account for these programs, but it differs in that it is a relative poverty measure based on the thirty-third percentile of expenditures on a basket of basic necessities. The problem with relying exclusively on a relative measure like this is that, as articulated by President Obama’s Council of Economic Advisers, “eliminating poverty defined with a relative measure may be nearly impossible, as the threshold rises apace with incomes.” Just to give an illustration, if somehow everyone’s incomes doubled tomorrow, there would be no perceivable effect on the supplemental measure, while the official poverty rate would plunge to near zero.

Instead of an either/or proposition, the two poverty measures should be used in combination because the official rate remains useful in a specific role: it conveys what percent of the population cannot earn enough to support themselves without government assistance.

Supplemental Poverty Metric
In the first paragraph, Hughes lifts (without attribution) an identical argument made by Ramesh Ponnuru in March of 2014. Here is Ponnuru:

A bigger flaw with the argument comes with Bartels’s second piece of evidence. He cites scholars at Columbia University who have concluded that Social Security, the earned income tax credit and other programs are responsible for the entirety of the decline in one measure of poverty over that period. The measure those scholars used, however, changes over time as the economy does.

It counts you as poor if you make less than a household at the 33rd percentile of household expenditures spends. The White House’s Council of Economic Advisers said in a recent report that to the extent poverty is measured in relative terms like this, ending it “may be nearly impossible.”

This measure also stacks the deck in favor of government and against markets as an anti-poverty tool — or, more precisely, in favor of redistribution rather than economic growth. If economic growth doubled the income of the poorest households but also increased that 33rd-percentile baseline for spending by a comparable amount, the poverty rate would remain unchanged.

Wow. So wonky. So sharp. Right? Nope. Here are the problems:
  1. The SPM does not set the poverty line at the 33rd percentile of household expenditures. The SPM poverty line is set at the 33rd percentile of household expenditures of food, clothing, shelter, and utilities (FCSU) multiplied by 1.2 and adjusted up or down regionally based on local cost of living. The 33rd percentile of FCSU will not move at nearly the same pace as the 33rd percentile of expenditures more generally. In fact, if you move out of hypothetical land and into reality world, what we've seen (Figure 1) is that the SPM poverty line moved in total lockstep with the official poverty line from 1967 to 2000, deviating only slightly since then. In 2012, the SPM line was around 6% higher than the OPM line.
  2. The SPM line can easily be anchored. To the extent that the SPM is relative, this poses no particular problem for "constant" measurements over time, if you feel like making them. All you have to do is pick the SPM line for a particular year, find out the nominal dollar value of the line in that year, and then use a price index to adjust that nominal value for inflation over time. This is called anchoring and it's extremely common (see my handy relative poverty calculator for an example). In fact, the Columbia team that ran the SPM back into time, has anchored data (see here), which is what most people cite to when talking about the dramatic success of the War On Poverty in cutting poverty by 40%.

Official Poverty Metric

In the second paragraph, Hughes repeats an argument from a Paul Ryan publication that Reihan Salam also made in July of 2014:

But getting more money from the government doesn’t really make you less poor.

The federal government’s official poverty measure, as Ryan’s report reminds us, doesn’t take into account benefits like the earned-income tax credit and food stamps. ... The official poverty measure is very useful because it tells us how much people are earning.

Salam's explanation of the Official Poverty Metric was so incorrect that, after I wrote a post about it at Demos shortly after, Slate edited the piece and issued a correction.

The OPM absolutely is not a market poverty metric. It absolutely is not based upon how much people are "earning." It absolutely does not tell us, as Hughes claims, "what percent of the population cannot earn enough to support themselves without government assistance."

Last week, when dealing with poverty mistakes made by David Brooks, I shared a screenshot from my (very handy) poverty calculator that shows you very clearly what the official poverty metric counts as income:

Very keen observers might notice the second checked item on the list is the largest government assistance program in the country by far. The welfare benefits it pays out to seniors, the disabled, and widows kept 22 million people out of poverty in 2013 under the OPM. If you strip out the other government benefits as well, you find that the official poverty rate "without government assistance" in 2013 was 23.1%, 27 million more impoverished people than when you include the government assistance incomes.

Hughes and Salam describe the OPM as a market poverty measure when it obviously is not. What's especially sad about this is that there are market poverty measures out there in the world. I produce a market poverty rate each year via my poverty calculator for OPM (23.1% most recently) and SPM (27.9% most recently). CBPP also produces one for SPM (28.1% most recently), which differs slightly from mine because of the use of a slightly different market income definition (spoiler: market income is not a coherent concept). And for relative market poverty, you can always rely on the trusty OECD (28.3% most recently). In any event, if Hughes wants to use a market poverty measure (something that I tend to shy away from) for his particular ideological purposes, he has many options! The OPM, however, is not one of them.

To be fair, coming to an understanding of how our poverty metrics work is not necessarily the easiest thing. It took me a number of years and ultimately painstaking replication projects before I fully understood them myself.

With that said, if you are going to be a fellow at a major think tank or offer yourself out as a smart conservative commentator who has really thought deeply on this subject, you really should know these kinds of things. A full understanding of how the measures work may make it harder for you to make your case against the greatness of modern welfare states (which are both good and cool), but at least whatever you come up with after gaining such an understanding will be not be littered with basic errors.