First 100: The Superstar Research Driving the Checks Debate Has a Problem

It’s based on ZIP code-level data, which may not be as precise as the researchers contend.

by

It’s February 8, 2021 and welcome to First 100. You can sign up to have First 100 delivered to your email by clicking here.

The Chief

Despite Larry Summers’ best efforts, it doesn’t look like there’s much in the $1.9 trillion American Rescue Plan that’s going to change from Joe Biden’s initial proposal. The Senate parliamentarian will determine the fate of the $15 an hour minimum wage (Biden is being rather defeatist in predicting it won’t make it, considering the budget impact is as much as $500 billion.) This may also doom the paid leave sections of the bill, and the $3,000-$3,600 annual child allowance, though if “giving money to families with kids” doesn’t have a direct budget impact we have to rethink the whole process.

The real variable is whether direct payments to Americans will have the same means test threshold as in prior COVID relief bills, or whether it will be narrowed, phasing out beginning at $50,000 in income for individuals rather than $75,000 ($100,000 rather than $150,000 for couples). Treasury Secretary Janet Yellen is now floating a split-the-difference $60,000.

For all the attempts to make this a matter of economics, it’s really a political question. There’s an innate fear from Democrats of some Fox News headline of a rich family getting a government check. But that neglects what should be the much bigger fear of voters who got the first two checks missing out on the third and properly blaming the new administration for it. As Sen. Jeff Merkley (D-OR) says, if Democrats want to so clearly break a campaign promise and lose all the momentum they gained in Georgia last month, that’s on them. The public doesn’t actually care about “undeserving” check recipients; this is an elite political obsession that’s going to randomly hurt a lot of people.

Read all of our First 100 Reports

Because political elites would rather see themselves motivated by facts and logic, they have attempted to shoehorn an economic rationale for tightening the means test and leaving 45 million or so Americans without a full check. They’ve zeroed in on a study from Raj Chetty which claims to prove that the second check, the $600 one approved in December and sent out in early January, was not spent in the first couple weeks by households making more than $78,000 a year.

Everyone favoring more targeted checks cites the Chetty analysis to prove their point, including the aforementioned Mr. Summers. Opportunity Insights, Chetty’s Harvard-based organization, is well-funded by very rich interests, and has the ear of official Washington.

I’ve already covered how the economics of the Chetty analysis are just wrong: whether a transfer payment is spent immediately is not indicative of whether it’s worthwhile (also debt reduction is not a poor use of funds, either). But the analysis is quietly controversial in economic circles for a different reason: This study of the spending habits of low- and middle-income households isn’t based on household-level income data?

As described in their technical appendix, the Opportunity Insights data is derived from a proprietary set of spending data from Affinity Solutions, which has a set of about 10 percent of all U.S. credit and debit card activity. (Though it’s proprietary, Opportunity Insights has made the data publicly available so researchers can replicate their findings.) That data are sorted by the ZIP code associated with the card. And then the ZIP codes are divvied up “using 2014-2018 ACS (The Census Bureau’s American Community Survey) estimates of ZIP Code median household income,” according to the appendix.

The only way you get to the conclusion that low-income people spent the check quickly and higher-income people didn’t, in other words, is by saying that ZIP codes that had lower-income people in them between three and seven years ago contained a higher level of immediate spending than ZIP codes with higher-income people during this period.

All of this time, of course, pre-dates the pandemic. As Lindsay Owens, interim executive director of the Groundwork Collaborative, told me, “There is absolutely no reason to be nickel and diming families, many of whom have seen their incomes drop since 2019, on the size of the promised checks.”

When I asked Opportunity Insights about this, they acknowledged “the distribution of incomes change within each ZIP code over time and as people move.” However, they expressed confidence that 2019 distribution of incomes is similar to the ACS ZIP code data, and that since employment rates have fully recovered at the top end of the income distribution, that the ZIP code data for the top households is fairly consistent.

ZIP code data is used fairly commonly in micro-economics, and you could make a somewhat plausible case that this proxy works. But lots of economists have problems with it. “I think the paper is unsuitable for the policy discussion,” said Claudia Sahm, a former Federal Reserve and Council of Economic Advisers economist. “It’s one paper at odds with 20 years of research.” Sahm noted that only household-level spending and income data would be appropriate to draw the kinds of conclusions the paper reaches. The sampling errors in the ACS data are pretty high. “I know the sampling error has to be in the thousands of dollars, there’s no way it’s that precise,” Sahm said.

There’s also potentially significant variation of income within ZIP codes, a point that Census economist John Voorhies made on Twitter. The Opportunity Insights data is presented implicitly as if the income variation doesn’t apply. “This means there's such severe income segregation that the only possible policy conclusion from your research is to burn the whole thing down and start over,” Voorhies noted.

Opportunity Insights, when questioned, agreed that there is some income variation within a ZIP code, and said that there are bounds for the underlying conclusions (in other words, spending from a particular income level may vary from the analysis a bit). Of course, that’s not how the paper is presented; it makes a fairly firm argument that households under $46,000 are spending the checks right away and households above $78,000 are not. More important, this is happening in the midst of a live discussion with major policy implications. “Frankly it’s irresponsible what they’re doing with the data,” Sahm said “The research is not solid, and it’s presented in a way that’s irresponsible and wrong.”

You can believe the general story, as economist Dean Baker does, that higher-income households are doing pretty good in this economy, which would lead you to the same conclusions as Chetty’s data. But applying his fine-grained analysis to a means test threshold would still be a leap of faith. And that’s before you get to the obvious point that we have no way to apply a specific income-threshold means test, because the only tax data we have is from 2019. (That will start to change as filing season opens this week; apparently once you file, the IRS will use 2020 tax data if it comes in before the relief payments go out. This puts the burden on filers to rush, or not rush, to maximize the payment, which just highlights the inexact nature of all this.)

JPMorgan Chase, using its own credit and debit card data, has also put out analyses of spending habits in this form. (Reportedly an upcoming analysis of the second relief checks will not show the same thing as Chetty’s work.) That uses household-level income statistics, and the spending series was developed for six years to fine-tune it to spit out real-time data. Opportunity Insights has only developed its tracker in the last year.

There’s been some notable pushback to the Chetty-driven concept of further means testing the checks, and not just the expected kind from the left. Reps. Lisa Blunt Rochester (D-DE) and David McKinley (R-WV) pointedly introduced a bill last Friday with the original phase-out levels; Blunt Rochester was a co-chair of Biden’s campaign. Sen. Jon Ossoff (D-GA), who campaigned on the checks, told HuffPost, “I don’t want to see any reductions in the help that we’re sending to people.”

President Biden has held back the forces of austerity on most counts early in his presidency. But he’s left open the question of means testing the checks. The fact it’s based on sketchy, untested, proxy data that the government has no real way to act on should push him over the top.

“If I knew what families really needed this money, I’d send it to them,” said Sahm, stressing the inexact nature of both Chetty’s research and IRS pre-pandemic income data. “The reality is we do not know. Stop pretending that you know.”

What Day of Biden’s Presidency Is It?

Day 20.

Today I Learned

TAP depends on your support

We’ve said it before: The greatest threat to democracy from the media isn’t disinformation, it’s the paywall. When you support The American Prospect, you’re supporting fellow readers who aren’t able to give, and countering the class system for information. Please, become a member, or make a one-time donation, today. Thank you!