(Bloomberg View) -- College tuition has gone up and up, and there are two popular explanations. But they aren't equally scrutinized. Liberals really want to believe one of them and disprove the other.
A very good essay by Jason Delisle discusses the hypotheses and faults the liberal media for accepting one too eagerly and critiquing only the conservative idea. He's onto something.
The two theories are called the Disinvestment Hypothesis and the Bennett Hypothesis. The first is exactly what it sounds like: the theory that educational costs are rising because state governments aren't putting as much money into the system as they used to.
Liberals are very fond of eyeballing state appropriations, eyeballing tuition, and declaring that they’ve found the main causal mechanism behind rising college costs. But changing the amount of state assistance doesn’t change costs, only who pays them. And college administrators can always find stuff to do with new money other than giving students a discount.
Moreover, when you look closely at the graphs, the correlation between falling state assistance and rising tuition is weak. Tuition has generally risen, and state spending on education has generally fallen, but they haven't moved together. And when I’m looking for causation, I usually want to see a tighter fit between two data sets.
Now, let’s turn to the favorite theory of conservatives: the Bennett Hypothesis. William Bennett, Ronald Reagan’s secretary of education, theorized that increasing student loans would just cause schools to increase tuition, allowing schools rather than students to benefit. I confess, I find this theory very plausible: When you subsidize something, you usually get more of it. And since what we are subsidizing is tuition, rather than education, I would expect tuition is what would rise.
And rise it has, soaring as the volume of student loans and the government repayment subsidies have gone up.
But, you know, causation’s tricky. (Anyone who relies on gross eyeballing to prove a connection needs to spend more time on this site. You should spend more time on that site even if you don’t engage in such statistical quackery, simply because it’s hilarious.) A lot of things have been happening in the economy that could also have been pushing college costs higher. The economy has lost a lot of the manufacturing jobs that used to provide relatively remunerative and steady work to people without college degrees, which in turn has made going to college more attractive to more people. There's also the possibility of reverse causation: The political process responds to felt needs among voters, so student loans may have gotten bigger and more common because tuition was going up.
The real value of Delisle’s essay is that it shows the very different treatment that these two hypotheses get from liberal media. When it comes to the Bennett Hypothesis, liberals want to see really high-quality evidence, and they’re careful to stress the distinction between correlation and causation. When it comes to the Disinvestment Hypothesis -- yep, looks like a correlation, and that's that.
Lest you think I’m picking on liberals, I’m not. (Delisle kind of is, and fair enough). I’m fairly confident that you could find conservatives doing the exact opposite: accepting the Bennett Hypothesis at face value, while demanding really good evidence before they believe in disinvestment.
And this brings me to the discussion of how bias usually operates in the real world of public intellectuals. It is not a conspiracy to make up facts. It is not even a conspiracy to set the bar higher for the other side than for yours. Rather the most powerful effect of our ideological priors is to determine what we are motivated to disconfirm.
When I was on my book tour, I used to offer a simple mathematical series: 2, 4, 6, 8 … and ask audience members to guess the rule that I was using for including numbers in the set. I told them they could throw out a number and I’d tell them whether it was included or not; they could guess any time they wanted.
Most people try 10 … 12 … 14 and decide that the rule was “even numbers.” But occasionally I’d get an engineer or someone in the group who would try negative numbers or odd numbers, and these types were the only ones who ever guessed the rule: “any positive integer.”
Why am I telling you this? Because it illustrates how much more powerful disconfirming evidence is than confirming evidence. You should always try to test hypotheses with examples that you expect to fail the test, as well as ones you expect to confirm it. That is a much faster, and better, way of getting to the truth than looking only at stuff that fits your model.
Unfortunately, humans are bad at this. When we like a hypothesis, we don’t want it to fail. We forget to look for disconfirming evidence, and start looking only at the things that fit our theory.
And that’s why, for all the bad rap that partisanship and ideology get, they’re actually tremendously valuable in moving toward the truth. The idea of perfectly neutral arbiters looking for “just the facts, ma’am” is an illusion; we are all human, fallible, and more than occasionally blind. Ideological diversity within a group means that even if the individuals are blind in different spots, at least the collective has a decent panoramic view.
That base, irrational, often angry “I know that’s wrong!” feeling that people get when reading an op-ed by the other team is actually the start of something wonderful: the search for disconfirming evidence that can falsify bad theories (the other team's, of course), and refine good ones (yours, of course). So that bit by bit, jab by jab, we get closer to the whole picture.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
Megan McArdle is a Bloomberg View columnist. She wrote for the Daily Beast, Newsweek, the Atlantic and the Economist and founded the blog Asymmetrical Information. She is the author of “The Up Side of Down: Why Failing Well Is the Key to Success.”
For more columns from Bloomberg View, visit http://www.bloomberg.com/view.