Does Behavioral Economics Undermine the Welfare State?

That’s the title of a post by Mike Konczal, who answers it in the negative. The question comes from Karl Smith and is based on a paper by Bryan Caplan and Scott Beaulier. The paper argues that welfare programs expand the set of choices available to people; while that is all good according to traditional economics, if we think that people are inclined to make bad choices (“behavioral economics”), then welfare programs give people more opportunity to make bad choices and hurt themselves. This is particularly a problem because, they claim, “there are good empirical reasons to think that behavioral economics better describes the poor than it does the rest of the population” (p. 4). In other words, if poor people are more irrational, then giving them more choices will hurt them more than other people.

Let’s start with that last claim. What could it even mean that “[some academic subfield] better describes [one group of people] than it does the rest of the population”? It seems to me there’s a category error here. Behavioral economics describes human beings, and the major population used in most experiments is undergraduates at prestigious universities. If the findings of the research are biased in any way, that’s the bias.

But what Caplan and Beaulier really mean to say is this: “Existing literature provides good reasons to think that the deviations of the poor from the standard neoclassical model are especially pronounced.  Their judgmental biases are more extreme, and their self-control problems more severe, than those of the rest of the population” (p. 12). So basically they boil down all of behavioral economics to the proposition that people behave irrationally (admittedly, this is what is most prominent in the popular literature), and then they say that the poor are more irrational than “normal” people. (The normative standpoint is theirs, not mine. Check out this clause: “deviant behavior is much more pronounced among the poor.”)

OK, this is offensive. But still, is there something to it? Not much, I would say. This is their evidence:

  • Poor people are more likely to exhibit heavy alcohol use.
  • Poor people are more likely to be obese.
  • Poor people are more like to smoke and use illegal drugs.
  • Poor people are more likely to have sex earlier and have children while in their teens.
  • Poor people are more likely to commit crimes.

Since all of these behaviors involve (they claim) an inability to make rational judgments about the tradeoffs between present benefits and future costs, this means that the poor are more prone to the judgment problems that are often cited in popular accounts of behavioral economics.

There is so much nonsense here it’s hard to figure out where to start. First of all, some of the factual claims aren’t even supported. For example, on obesity, they cite a study finding that obesity declines with educational level. Then they assert: “Given the strong correlation between education and income, there is little doubt that the poor have more trouble maintaining a healthy body weight.” Yes, given the above, it is likely that there is a negative correlation between obesity and income; but it is also likely that that negative correlation is explained by education, not by poverty. Couldn’t they have made the effort to find a study with the facts they need instead of just assuming?*

Second, and more importantly, if we assume for the purposes of argument that more poor people are obese, it’s an enormous leap to say that this is because they make worse judgments. There are many other possible explanations. Here are a few, each of which I find more compelling than the behavioral one:

  • It’s more expensive to eat healthy food than unhealthy food.
  • Many poor people work long hours or have limited child care options, which makes it harder to buy and cook healthy food from the grocery store.
  • Poor people have worse health care than rich people.

You can add your own.

This criticism can be easily made of several of their points. Take crime, for example. The most obvious counterargument is that crime is more prevalent among the poor because they are poor. Caplan and Beaulier think they can deal with this by saying that few inmates received illegal income prior to being arrested. But this is only a small part of the puzzle. A major reason for high incarceration rates among the poor is recidivism, and a major reason for recidivism is the difficulty of finding employment and otherwise integrating into society when you have a criminal record. People without stable support networks are more likely to commit crimes and have less to lose, regardless of their ability to make judgments. The failure to control for anything at all should make their point invalid.

Along the way, they also cite Levitt and Dubner to support the proposition that since drug dealers don’t make much money, “it is easy to see the appeal of crime to those who overestimate their chance of becoming a gang leader, or who simply have poor impulse control” (p. 14). But what Levitt and Dubner really say is this (Freakonomics, paperback edition, p. 102):

“To the kids growing up in a housing project on Chicago’s south side, crack dealing seemed like a glamour profession. For many of them, the job of gang boss — highly visible and highly lucrative — was easily the best job they thought they had access to.”

So far, so good for Caplan and Beaulier. But they continue:

“Had they grown up under different circumstances, they might have thought about becoming economists or writers. But in the neighborhood where J.T.’s gang operated, the path to a decent legitimate job was practically invisible. Fifty-six percent of the neighborhood’s children lived below the poverty line . . . Seventy-eight percent came from single-parent homes. Fewer than 5 percent of the neighborhood’s adults had a college degree; barely one in three adult men worked at all. The neighborhood’s median income was about $15,000 a year, well less than half the U.S. average. During the years that Venkatesh lived with J.T.’s gang, foot soldiers often asked his help in landing what they called “a good job”: working as a janitor at the University of Chicago.”

Levitt and Dubner’s real points are these: First, people turn to drug dealing because there are no good options. Second, drug dealers are just like people pursuing other glamour professions.

OK, enough for that point. What else? Well, there are measurement problems, although they probably aren’t that big. Poor people are more likely to use illegal drugs and commit crimes in part because society defines those drugs as illegal (as opposed to, say, prescription pain-killers) and enforces the laws more strictly against poor people. But that’s probably not a huge factor.

In any case, the idea that poor people are more likely to exhibit bounded rationality is a huge assertion backed up by not much of anything. This isn’t even a standard correlation-causation problem. This is a correlation where their preferred cause is well down the list of possible causes.

But . . . that still leaves the theoretical point. Is it possible that expanding people’s choice sets enables them to make bad decisions? Of course. Even if you’ve never read a behavioral economics paper in your life, just look at any chapter of Predictably Irrational. But does that justify this conclusion?

“Most obviously, if government assistance to the disadvantaged amplifies the ill effects of their judgmental biases and self-control problems, it strengthens the case for reducing the size of welfare benefits, limiting their duration, restricting eligibility, and even abolition.”

Let’s take that sentence one clause at a time. Does government assistance necessarily amplify judgmental biases? Not at all. As Konczal noted, there’s a lot of picking and choosing going on in this paper. Take retirement savings, for example. Mountains of research shows that people are very bad at savings and investment decisions, because of all the usual culprits: loss aversion, hyperbolic discounting, etc. So Social Security is a good thing: it forces people to save, and doesn’t let you invest it all in tech stocks, and forces you to wait until you’re in your sixties to get anything back. (There’s also a redistributive component, but even without that component it would still be a good thing for the above reasons.)

Then there’s the topic du jour: the individual mandate. Because of optimism bias, some people will not buy health insurance even when they should, with negative consequences for them and for society as a whole.

At most, even if we accept the first clause of Caplan and Beaulier’s sentence for purposes of argument, it should affect the way we deliver welfare benefits, not their existence or size (which should be determined on other grounds). If the problem is choice, then we should just reduce choice. For example, we could eliminate every single welfare program and replace all of them with (1) Medicare for everyone and (2) a flat $10,000 (indexed) cash payment to every person in the country. (If you want to switch to a flat tax or a consumption tax to raise the necessary revenue, I’m willing to take that trade, as long as nothing eats into (1) or (2).) That way we would achieve the social goals of welfare with none of the choice problems Caplan and Beaulier complain about. Even leaving aside my fantasy, we should be happy about things like unemployment insurance and the EITC, which both give you an incentive to work.

Finally, on Konczal’s broader question: Like any field, behavioral economics gives you lots of opportunity to pick and choose, and if you’re willing to be superficial or unscrupulous, you can justify lots of policy positions with it. But on balance I think it cuts in favor of the welfare state. In its popular form, behavioral economics is all about judgment errors that hurt people’s own self-interest, which provide support for the idea that government should attempt to correct for those errors. Forced saving is one example. Robust disclosures are another (although not one I’m particularly optimistic about). Now, there is a valid argument about how effectively government policy can correct for those errors, and at what cost, so maybe we should be humble about the ability of Cass Sunstein to come up with the right correction for every problem. But in principle that seems like the simplest inference to draw.

More fundamentally, behavioral economics throws into question all of the foundations of traditional microeconomics, on which the entire theory of the free market is based. One of the findings of the research is that not only do people not make decisions that are consistent with their preferences, but their preferences are arbitrary in the first place, even under conditions of perfect information. (See “Coherent Arbitrariness” by Ariely, Loewenstein, and Prelec, for example.) If people can be tricked into valuing something more highly because, say, the last two digits of their social security numbers are high, how much confidence should we have that market prices reflect absolute value to consumers?

* This may seem harsh, especially coming from a blogger. Frankly, I would be more sympathetic to Caplan and Beaulier if they had put their ideas in the form of a blog post instead of a “paper,” with the level of carefulness the latter used to imply.

Originally published at The Baseline Scenario and reproduced here with permission.