The real case for universal programs. By Matthew Yglesias

The real case for universal programs. By Matthew Yglesias

February 2, 2021

Hey folks — this morning I wanted to introduce everyone to a new member of the team, Claire Cantrell, who came on board this week as editorial assistant to help make sure the posts here are as good as they can be and to keep the trains running on time. Claire is great, and I’ll let her say hi in her own words:

Hi everyone! My name’s Claire and you’ll be seeing me around Slow Boring periodically alongside Matthew and Marc. I graduated from college last year with a degree in International Affairs and currently live in D.C. (okay, technically Arlington), though I was born and raised in Austin, Texas. My concentrations in my studies were Latin America and geography, and I studied abroad for five months in Chile (I also did three shorter-term programs in Copenhagen, Cape Town, and Monteverde, Costa Rica). I have a senior cat named Huey, I will try pretty much anything once, and cheese is far and away my favorite food in the world… and those are the basics about me. I can’t wait to get to know everyone!

If you need even more Yglesias takes, my latest Bloomberg column argues that to get bipartisan legislation, Joe Biden needs to do… nothing. Presidential involvement is extremely counterproductive to bipartisanship, and bipartisan bills are much more likely to emerge if he just doesn’t get involved once the COVID-19 relief is done.

Now for the main topic!

(Robert Nickelsberg/Getty Images)

There’s a longstanding debate about the merits of universal vs. targeted or means-tested government programs. This is often structured around an idea that I was taught in my American Society & Public Policy class back in college: “programs for the poor become poor programs.” Fitzhugh Mullan expressed a similar idea with a similar slogan in a in a 1987 Health Affairs article titled, “Poor People, Poor Policy: Government Programs for the Disadvantaged.”

The idea here is that the case for universal programs is fundamentally political. They are more popular and therefore more sustainable, and the participation of non-poor people in the program mean elected officials actually pay attention to the quality of the service — libraries are nicer than homeless shelters because middle class people use them and care about how nice they are.

I think this argument is wrongheaded on a number of levels. It overstates the political weakness of means-tested programs, understates the political challenge of expensive programs, and I think relies on a somewhat fuzzy concept of what universality means.

A better argument relies on a different kind of distinction — categorical programs (which are not necessarily universal per se) are better because they provide superior program administration. Means-testing essentially forces you to choose between making it too easy for people to get benefits (false positives or outright fraud) and inviting political blowback, or else you insulate yourself from blowback by erecting administrative burdens that are so high you’re actually not helping many of the people in the targeted group. Politics is difficult, so for various reasons you might want to do a means-tested program anyway — but it’s going to be a relatively ineffective program compared to a categorical one.

The old argument is wrong

In retrospect, I think the old argument that means-tested programs are politically vulnerable was based on overreading the specific politics of Aid to Families with Dependent Children and “welfare reform” in the 1980s and 1990s.

On a parallel track in the ‘80s, Medicaid expanded a fair amount.

In the 1990s, welfare reform was paired with the creation of S-CHIP.

Nobody talks about this anymore, but in the 2002 farm bill that George W. Bush signed, SNAP benefits were significantly expanded in a way that’s never been reversed.

Under Obama we got further SNAP expansion, further Medicaid expansion, and the ACA benefits.

Over the past generation we’ve also seen a proliferation of tax credits that, while often structured to exclude the very poorest, are definitely means-tested.

The basic picture is that looking back, AFDC’s political vulnerability was a more complicated story than “well, the beneficiaries were poor so they lacked political clout.” In retrospect, I would say AFDC’s problems were:

It wasn’t just means-tested and it wasn’t just available to non-workers; it was specifically structured to phase out so rapidly when you had any earnings at all, thus creating a huge disincentive to work that conservatives didn’t want to fix by making the program more expensive with a more gradual phase-out

There was a racialized panic associated with the high levels of violent crime in the 1980s and 1990s that has greatly abated during the Great Awokening.

The country was in a generalized austerity mode in the mid-1990s. Taxes went up in 1990 and again in 1993 along with cuts in discretionary spending. When macroeconomic austerity is happening, something is going to get cut.

We’ve seen since then that means-tested cash benefits that are tied to work (TANF, EITC) are politically sustainable, as are means-tested non-cash benefits that are not tied to work (SNAP, WIC). Then with the pandemic survival checks, we sent out cash benefits that were not tied to work and it seemed fine politically. Many people in the discourse have categorized these checks as “universal” rather than means-tested. That would have been convenient for certain people’s arguments, but it’s not true — I wasn’t eligible for the $1,200 or the $600 and I wouldn’t be eligible for the supplemental $1,400 in Biden’s plan. And that’s mostly fine.

The checks were means-tested

But while the means-testing of the stimulus checks was not politically ruinous, I do think it was substantively weird.

The reason that my family wasn’t eligible for the $1,200 check is that our income in the previous year was too high. But at the time the bill was written, my wife had been forced out of the workforce and Vox Media had cut my salary 25% as a pandemic cost-cutting measure. We hadn’t suddenly become hard-luck cases, but our income actually had dropped down into the eligibility range. The whole thing was weird, not so much because I desperately needed the money but because ostensibly the whole point was to address the public health emergency. The program didn’t do anything to take into account the extent to which the emergency changed people’s economic circumstances.

The real answer, of course, is that expanded Unemployment Insurance was the real emergency response (and it was a huge success), and the checks were a little bit of an oddity. Right before the pandemic struck, there was a burst of media activity (some of it instigated by myself) promoting Claudia Sahm’s idea that we should fight recessions with direct fiscal transfers to households.

I think that’s an excellent idea in general (that’s why I was promoting it), though it’s actually a little bit poorly suited to the particular circumstances of the pandemic.

But it landed on the policy agenda anyway, and I’m glad that it did, both because it helped people in need and also because it served as a great test case for some points about the welfare state. For starters, it proved to be incredibly popular. That popularity can’t be because it wasn’t means-tested, because it was means-tested. What I think made it so great was threefold:

The money went out rapidly

The eligibility criteria were easy to understand

The great majority of eligible people, especially those reachable for comment by news networks and pollsters, got the money automatically, with real administrative difficulties left only for non-filers, who are mostly really poor

Pamela Herd and Donald Moynihan have a great recent book called Administrative Burden: Policymaking by Other Means (hear them discuss it on The Weeds). The appeal of the checks, I think, is that they were a huge administrative burden success story. It was smooth sailing, in part because the Trump administration wanted the checks to go smoothly rather than doing Republican Party program sabotage stuff. In contrast, the UI expansion, though substantively a huge success, was an administrative nightmare with tons of problems and stress and hassle.

As a broad principle, “programs that are well-administered will be more popular than programs that turn into bureaucratic nightmares” is not that profound of an insight. But I think it does a much better job of explaining why categorical programs are usually better.

Categorical eligibility makes the difference

One thing worth saying here is that it’s rare for a program to be truly universal.

Social Security and Medicare are the paradigmatic universal programs, but some people die before becoming eligible. Public schools are available to all children, but not every family has children.

I think the best way to think about these programs is that eligibility for them is categorical — it applies to broad swathes of people based on easily defined characteristics — which facilitates low administrative burdens. All the kids are supposed to be in school, and all the elderly are supposed to be on Medicare. There is some paperwork involved, but it is relatively easy because there is a clear consensus about who is supposed to be signed up.

By contrast, the actual participation rates with the main means-tested programs are dismal:

For AFDC’s successor program, TANF, participation rates are only about 25%, which is not just due to funding shortfalls.

For Medicaid, it’s about 75% for childless adults (in Medicaid expansion states) and 80% for parents (in all states). For kids, it’s better at 94%.

SNAP is 84%.

The WIC program providing supplemental nutrition to pregnant women and infants has substantial long-term benefits to the kids but only a 51% participation rate.

The Earned Income Tax Credit has a 78% participation rate.

For the non-EITC programs, there’s a big state government role in administration and typically big gaps between the states. The size of those gaps confirms that it’s not like it’s impossible to improve participation rates for means-tested programs, but I do think it’s challenging for fairly fundamental reasons. Claiming EITC seems to be a big audit flag for the IRS, which likes to do these cases because they’re fairly straightforward compared to auditing rich people. But also because the IRS, with good reason, doesn’t want to get beaten up in the press or Congress for letting tons of people fraudulently claim EITC benefits.

The big problem is that with a targeted program, you really kind of do need to choose between inclusion errors and exclusion errors, and given the politically explosive nature of benefits fraud, there are strong incentives to under-enroll people. Now maybe if you just think of this as pure charity that’s fine, and we don’t care that only 45% of eligible pregnant women are enrolled in WIC. But again, to quote my WIC study, “relative to their siblings, prenatal WIC participants have a lower incidence of ADHD and other common childhood mental health conditions and of grade repetition,” so actually getting low-income pregnant women to use the benefits would be good.

Sheer proliferation of different means-tested programs with slightly different eligibility rules and benefit structures also introduces some weird design features into the system, where de facto marginal tax rates can become extremely high at particular points. You could imagine abstracting from the whole current mess a world that:

Gives cash benefits to kids and the elderly

Gives wage subsidies to low-wage workers

Collects regressive consumption taxes to generate a lot of revenue

Layers a progressive income tax (with benefits considered taxable) for progressivity on top of that

Then you would have a world in which tax rates rise monotonically with income, and where program eligibility is really easy to determine and virtually nobody is left out. The strengths of this system are substantive. The weakness is in fact political — you’d need a lot of explicit tax increases.

People don’t like taxes

A decade ago, carbon taxes were the true progressive alternative to squishy moderates who wanted to talk about green jobs. Today, they are the neoliberal alternative to true progressives who want to talk about a Green New Deal.

Under either scenario, the pro-tax people are correct because for any given level of proposed green spending, if you finance that spending with emissions taxes you get a much larger reduction in carbon dioxide emissions. It’s true that you could (and under present economic circumstances, should) make the budget deficit larger. But that could come from eliminating other taxes that don’t have emissions-reducing impacts, or from spending money on useful stuff that’s not green.

This analysis is totally correct, but as I believe basically every advocate who genuinely works on energy issues now agrees, it just doesn’t work. The kind of high, genuinely universal carbon tax that would have a meaningful impact on emissions is just a total political impossibility. So if you start with “we need to do carbon pricing” as the fixed point of your analysis, you wind up compromising down in terms of scope and what the actual price is so much that you basically gut the impact. Danny Cullenward and David Victor explain this dynamic in some detail in a great interview over at David Roberts’ Substack.

But this is not just an intramural fight in green circles.

The exact same dynamic plays out over and over again. Taxing “the rich” is popular, but some people take a pretty narrow view of who “the rich” are for these purposes. Even clearly progressive proposals like curbing 529 plans or capping the SALT deduction face huge backlashes based on the fact that they sweep up a certain number of middle-class people. Indeed, these kind of ideas arguably fall into a unique kill zone where, precisely because they are progressive, they mobilize the full political clout of the super-rich but then augment them with a group of more sympathetic people. More generally, you can see that Democrats have started slicing the salami ever thinner. Obama started out vowing to raise taxes only on households that earn over $250,000 but ended up settling for a $400,000 threshold. Biden has now adopted the same $400,000 as his “the rich” cut point, and we’ll see where he gets bargained to when it comes down to it.

It’s telling, though, that just in November, the state legislature in Massachusetts overwhelmingly rejected a proposal to tax unearned income to avoid MBTA cuts. Washington state has become much bluer over the years, but the state’s Democrats don’t try to create an income tax. And ballot initiatives for carbon pricing keep failing there, too. Trying to minimize explicit tax increases is a perfectly good political reason to opt for means-tested programs, even if the programs are inferior on other grounds.

It’s politics all the way down

That of course raises the question of whether programs really need to be paid for at all. For now at least, interest rates are super low and you can definitely just deficit finance whatever you want.

Now in technical legislative terms, the problem is that you can’t create a permanent new program using budget reconciliation unless you offset it with tax increases. You could end the filibuster, of course. Or you could do what Bush did with his tax cuts and just schedule them to expire randomly after nine years in order to conform to the rules. The risk with that, though, is that while some of the Bush tax cuts were extended, some of them were not.

Fundamentally, as an advocate or analyst, I don’t think there’s a unitary “right” answer here. Elected officials should care less about the deficit. The public should be less averse to taxation. I try to persuade people of both groups of these things. Everyone should also be more attentive to questions about program design and administrative burden. That’s why I interviewed Herd and Moynihan for The Weeds. To an extent, if you think long and hard about administrative burdens, you reach the conclusion that categorical programs with easy eligibility criteria are best. But for other reasons that may not always be realistic, you need to do the program design spadework.

There’s not really an unequivocal answer to “what should you do?” In some ideal sense, the whole structure of the legislative process and public opinion would be totally different and we could adopt an idealized solution. But they’re not, so we can’t. So you get kludgy solutions because people are trying to help people as best they can under constraint.