Before diving into the causes of rising higher education spending, it's worth asking if we ought to care. It's obvious that we should care about, say, our rising spending on health care. It is very important for cancer patients to be able to afford chemotherapy; any argument that it's not fails the laugh test. But it's become somewhat fashionable of late to claim that higher education isn't really that essential. Perhaps the answer to the cost problem is for kids to simply stop going.

"For an increasing number of kids, the extra time and money spent pursuing a college diploma will leave them worse off than they were before they set foot on campus," Megan McArdle concluded in a Newsweek cover story last fall. Peter Thiel, the billionaire PayPal co-founder, has been paying smart undergraduates to drop out and start working on something, anything, other than college.

So does college raise incomes? Is it an investment good enough to make widely accessible?

Yes, it is. Period. Usually, this would be the part of the article where I note that there's disagreement and perhaps a slight weighting of evidence to one side or the other. I won't. Even McArdle and other college skeptics acknowledge that the average college graduate today will make far more over the course of his or her life than the average high-school graduate who doesn't attend college. And the bulk of the information indicates that college really is the cause. Going to college means you make more money than you otherwise would, and that benefit far, far outstrips its upfront price.

First, the raw numbers. According to the Census Bureau's Current Population Survey, in 2011 the median income of a high school grad who never went to college was $28,659; for those with some college but no degree, it was $32,036. By contrast, college graduates without advanced degrees had a median income of $49,648. Those with professional degrees had a median income of $87,356, more than three times that for high school grads.

The premium has held even as incomes have stagnated. People with bachelor's degrees don’t make much more than they did 20 years ago, but they still make way more than high school grads. The college wage premium is real, and it is huge.

But that alone doesn't tell you that college is causing those earnings differentials. Maybe it's just that naturally smart people tend both to get a lot of education and to earn a lot, but the two aren't actually related. For example, Mark Zuckerberg and Bill Gates both went to Harvard and became billionaires, the theory goes, because they were smart. Harvard didn't cause their riches. They dropped out before earning degrees, in fact. Their smarts caused both Harvard and the riches.

The evidence, however, suggests that the relationship is in fact causal. One prominent strand of research compares earnings of siblings or twins with different degrees of academic achievement. Genetically identical twins raised in the same environment are likely to have very similar ability levels to start with, which enables one to isolate the effect of education.

The economists Orley Ashenfelter and Alan Krueger (the latter of whom served as President Obama's chief economist) found that people with one more year of schooling than their twin earn, on average, 12 to 16 percent more. Cecilia Rouse (a former Obama adviser herself) found similar results with a larger version of Ashenfelter and Krueger's sample, and another study conducted by Ashenfelter with siblings, rather than twins, confirmed the findings.

Another set of studies use a method called "instrumental variables" to isolate the effect of education. That is, researchers find a third variable, aside from education levels and income, which would be expected to increase education levels but not to increase income except insofar as it increases education levels.

Krueger and Joshua Angrist, for example, used date of birth as an instrumental variable, as compulsory schooling laws mean that high schoolers whose birth date is closer to the beginning of the school year who drop out lose out on more schooling than those born at the end. If you drop out on your 16th birthday, you leave with a lot more education under your belt if that birthday’s in June than if it’s in September. They found that older students earned less, suggesting that the education they lost by dropping out earlier would have gained them money.

Rouse and Thomas Kane used distance from a college as an instrumental variable to measure the economic impact of two- and four-year colleges, as you'd expect people growing up near colleges to be be likelier to go but not necessarily likelier to earn more. They too found significant effects of both two- and four-year college degrees, though some have cast doubt on that particular instrumental variable (maybe areas with more colleges are also likelier to earn more regardless of how much education you get).

Truly experimental evidence is hard to find, as it's difficult to design a study in which people are randomly assigned to go to college or to stay home. But something close to that happened when Mathematica Policy Research ran a randomized evaluation of Job Corps, a vocational education program run by the federal government. It found significant positive impacts on earnings from the education included in Job Corps.

Something close to randomized design also happened by accident during the Vietnam War. The draft lottery meant that service in Vietnam was, to some degree, randomized, and thus, so was access to the G.I. Bill. That lets you isolate the returns to education for those returning soldiers, by comparing their earnings to those of people who got more favorable lottery numbers. Angrist and Stacey Chen analyzed the education levels and earnings of those selected into and out of service, and estimated a return to education of about 7 percent a year.

The University of British Columbia’s Thomas Lemieux and Berkeley’s David Card looked at Canada’s version of the GI Bill, noting that it provided much greater benefits to Ontario residents than Quebec natives, due to differential enlistment rates. Comparing Ontarians and Quebecois’s earnings and educational attainment, they estimated a return to education of about 15 percent.

"The results of all these studies – those using family relationships, those using [instrumental variables], and that from the Job Corps evaluation – are surprisingly consistent," Rouse writes in a summary of the research. "They indicate that the return to schooling is not dominated by an omitted correlation between ability and schooling." In plain English: they suggest that schooling does, in fact, cause incomes to increase.

There's another strand of criticism that alleges that whatever effect college has on wages is an effect of signaling. It's the degree that matters, as a way to tell employers you're smart and accomplished, rather than what you learn in pursuit of the degree, under this theory. Note, first of all, that this doesn't really negate the finding that college causes higher earnings. It just presents a slightly cynical explanation of how it is that college causes that earnings bump.

This theory makes a prediction: If signaling is leading to income gains, you'd expect the gains to education to be "non-linear". You'd expect that a one-year increase in education that brings a degree with it (say, going from three years of college to four years and a bachelor's) to have more of an effect on earnings than a one-year increase that doesn't come with a degree (say, going from dropping out of college after two years to dropping out after three years). Sure enough, that's exactly what the data show, as Thomas Hungerford and Gary Solon first demonstrated in 1987.

But signaling doesn’t explain all of the gains. Getting actual degrees seems to increase earnings on its own, but so does increased schooling with degrees to go along with it. "Completing an associate’s degree appears to be associated with a 15 to 27 percent increase in annual earnings," Kane and Rouse write. "Since estimates suggest that two years of community college credit is associated with a 10 to 16 percent increase in earnings…there appears to be some additional gain to the associate’s degree itself." But if those numbers are correct, then the credential itself accounts for well less than half the total effect. The actual schooling, apart from its signaling effects, still matters.

Perhaps the most persuasive skeptical case to be made in the face of all this evidence is that while the average student still benefits from college, that doesn't mean the marginal student does. That is, the fact that a typical college student gets a wage bump from his time in the classroom doesn't mean that a student on the margins, who may or may not be ready for college and is weighing his options, would get a similar benefit. If that's true, then encouraging more people to go to college may not make sense. This is the case McArdle makes. "Experts tend to agree that for the average student, college is still worth it today, but they also agree that the rapid increase in price is eating up more and more of the potential return," she writes. "For borderline students, tuition hikes can push those returns into negative territory."

So what does college do for "borderline" students? The best study I've seen on this comes from Seth Zimmerman, a PhD student at Yale. He compared the earnings of Florida students whose GPAs were just above and just below the Florida State University system's cutoff. There's a huge effect of being just above the cutoff (3.0, for most students), suggesting that students at the margin still gain substantially from college. Some of the students below the cutoff still attended college, but at a much lower rate than those above the threshold, which enables Zimmerman to estimate the effect on the marginal student.

Work by Nobel laureate James Heckman and his colleagues backs this up. While in some cases, the marginal effect of policies that spur marginal students to attend college is lower than the effect of college for average students, "for other policies associated with other marginal expansions, the marginal gains are substantial." So too does a paper by the University of Toronto’s Philip Oreopoulos and Uros Petronijevic. Marginal students may not gain as much as average students, but they still gain substantially in many cases.

College matters. It causes real, large increases in lifetime earnings, both for average students and those teetering between going and not going to college. And it has other benefits too. Oreopoulos and the Norwegian School of Economics’s Kjell G. Salvanes find that life satisfaction rises with increased years in college, even when you control for income. That's why keeping its price low and enrollment accessible is so vitally important.