If the federal government has failed at making college affordable (and it’s pretty clear that, insofar as that was a goal, it has), it’s not for a lack of trying. There’s a plethora of programs that have been passed in recent decades designed to target exactly this problem. Perhaps the most famous is the Pell Grant program, but there’s also the American Opportunity Tax Credit, Stafford, PLUS, and Perkins loans, and the Lifetime Learning Tax Credit. Then there’s the tax-exempt status of scholarships and interest from state and local government bonds used to finance public higher education, as well as the deduction for charitable contributions to colleges and universities.

Friend of the blog Mike Konczal and The Atlantic’s Jordan Weissmann have both noted that if you tally up the cost of all these programs, it’s probably enough to make college free, or at least get close. How could we be spending this much on subsidies and not getting a more affordable end product?

According to one school of thought, that question gets the matter exactly backwards. It’s not that the subsidies aren't making colleges affordable. It’s that the subsidies are actually making college more expensive.

This theory was first floated by then-Secretary of Education William Bennett during the Reagan administration, and it’s no wonder. The implications dovetail perfectly with Bennett and Reagan’s anti-spending impulses (Bennett first laid out the theory in a New York Times op-ed titled “Our Greedy Colleges”). The hypothesis goes like this. Let’s suppose that a college is totally free. No tuition, no fees, no nothing. Then the federal government comes around and says, “We’re giving every college student $5,000 a year in tuition.” The college would have to be run by idiots to not immediately jack tuition up to $5,000. That’s free money!

The Bennett hypothesis states that this happens basically whenever the government subsidizes higher ed. Every $1 increase in subsidies is matched by a $1 increase in price. The money is pocketed by the university, not the students. The economist Albert Hirschman once categorized this type of conservative argument as the “argument from perversity,” that a proposed policy change will actually have the opposite effect of its intended one. But in this case at least, there’s a ring of plausibility to it.

So what do the data say? Thankfully, there’s a rich literature that’s emerged in Bennett’s wake, trying to nail that down. Unfortunately for policymakers, that literature is very mixed. And it suggests that what may be most important is the form the aid takes, rather than that it exists at all.

There’s a 2001 study by the National Center for Education Statistics, which found no Bennett effects for any type of school (though that study has some methodological troubles). Then again, two very recent studies by UC San Diego’s Nicholas Turner  (now at the Treasury Department) and Columbia’s Lesley Turner (now at Maryland) found that aid triggers spikes in tuition; the latter found that schools capture 16 percent of Pell Grant’s value, ultimately.

So no consensus there. Maybe if we differentiate by school type, we’ll make more progress?

Nope. In one study, delightfully named “For Whom the Pell Tolls,” Larry Singell and Joe Stone at the University of Oregon found no evidence for the hypothesis for in-state tuition at public universities but found that private universities increase tuition (and public universities increase out-of-state tuition) in response to increases in Pell Grants. The University of Missouri’s Bradley Curs and UC - Riverside’s Luciana Dar find the same thing.

But the University of Rochester’s Michael Rizzo and Cornell’s Ronald Ehrenberg found the exact opposite in their own study of public flagship schools. “We find substantial evidence that increases in the generosity of the federal Pell grant program, access to subsidized loans and state need-based grant aid awards lead to increases in in-state tuition levels,” they write. “However, we find no evidence that nonresident tuition is increased as a result of these programs.” Similarly, in their book “The Student Aid Game,” then-Macalester College president Michael McPherson and then-USC dean Morton Schapiro (who is now president of Northwestern), both higher ed economists, found that private schools don’t increase tuition in response to new subsidies, but public schools do.

Harvard economist Bridget Long’s research suggests both camps are wrong. She found that the Georgia HOPE scholarship (which provides aid for tuition at either public or private schools) did, in fact, increase tuition regardless of school type, though more so at private schools. She found in another paper that four-year institutions don’t increase tuition in response to subsidies, but that public two-years do.

Meanwhile, George Washington University’s Stephanie Cellini and Harvard’s Claudia Goldin find that aid increases match tuition increases basically one-for-one at for-profit colleges (like, full disclosure, the one that the Washington Post Company owns). The research literature is so thin on for-profits that I believe this is the only study that addresses it, which makes it all the riskier to draw conclusions from it. But it’s very different from any findings on public or private schools.

Phew. Got all that?

To paraphrase a wise man, “Out of the crooked timber of non-experimental social science, no straight thing was ever made.” So it is with the Bennett hypothesis. There just is no consensus in the field, one way or another. What evidence there is is based on imperfect regression analyses, and even that evidence yields no firm conclusion.

But it really matters if the Bennett hypothesis is true. If it is, then it’s pointless to keep pouring public money into higher education. That just exacerbates the problem. But if it’s false, that concern goes away, and the case for public intervention is considerably stronger. So how should we proceed in the face of this kind of uncertainty?

One answer, embraced by Long in testimony offered to the Senate Finance Committee in 2006, is to treat the Bennett hypothesis as unproven. In this case, the evidence for rejecting the theory that government subsidies don’t affect tuition is remarkably weak, so we should accept that they have no effect. As Long put it, “Concerns about colleges raising tuition prices in response to federal aid appear to be largely unwarranted.”

Another is to formulate a better theory. That’s what Andrew Gillen, a senior fellow at Education Sector, sought to do with his paper “Introducing Bennett Hypothesis 2.0.” The problem with the original hypothesis, Gillen states, isn’t that it’s wrong so much as that it failed to take into account two factors that blunt its impact. The first is that not all types of aid have the same effect. Aid that’s geared toward low-income students only won’t cause as big an increase in tuition, for example. The second is that tuition caps and price discrimination (colleges charging different students different prices) also weaken the link between aid and cost increases.

But there’s also a factor the Bennett hypothesis ignores that strengthens its impact, Gillen says. That’s the Bowen hypothesis, which we talked about in the last segment. If schools capture more money from federal aid, then they’re going to spend it on something. They’re non-profits; they have to. But those costs - professor salaries, administrative staff, amenities, etc. - don’t just go away. They increase every year thereafter. That means that accepting a one-time bump in funding from government subsidies requires schools to boost tuition for years on end to pay for the incurred costs.

Gillen is the first to state that even his revised theory doesn’t have rock-solid empirical support; to do that, you’d need a random trial, which is tough to get. But it does explain a lot of things about the existing evidence. For example, Gillen’s theory makes it clear why you don’t see increases in tuition after increasing Pell Grants. Those narrowly target needy students, which makes them less susceptible to Bennett effects.

Indeed, Gillen’s theory points the way toward a policy approach that would work whether he’s right or people who reject the Bennett theory altogether are right. Gillen has proposed the creation of “super-Pell Grants,” as a replacement for federal student loans and other subsidies.

They’d work like this: The government would calculate how much a decent education ought to cost, given the cost of basic inputs like buildings and faculty. Then it calculates what each family can be expected to contribute to tuition based on their income (the “expected family contribution,” in financial aid parlance). Then it should subtract the latter from the former and give it as a lump sum grant to students to spend where they like.

That’s controversial, especially as Gillen proposes replacing all government subsidies to public schools with super-Pell Grants. But it would make the system much more progressive than it is today, and, in the long run, cheaper, even if you don’t think the Bennett hypothesis is at work.

Putting policy implications to one side, however, one thing unites the literature on the Bennett hypothesis. All but one study I’ve seen found some evidence of a price response to increases in aid, for some section of the higher ed market. That, and the theoretical plausibility of the relationship, suggests that at the very least the effect is worth worrying about, and tailoring subsidies to take it into account, as Gillen proposes, is a good response. We need aid - people couldn’t afford school without it. We just need to provide it in a way that doesn’t backfire.