Intentions versus Consequences

I’m gratified to hear that Richard Thaler doesn’t support all the policies that appear in his earlier work on libertarian paternalism. The inclusion of those policies sounded a lot like an endorsement to me, but I’ll take him at his word that he personally doesn’t back them.

And yet my concerns are not assuaged. As I indicated in my previous post, this discussion isn’t about the personal beliefs of Richard Thaler. It’s about a broader movement of which he is only a part. Thaler may indeed be the most libertarian of paternalists, but others are far less restrained. Policymakers, even less so.

Moreover, there is still slippery-slope potential in Thaler’s position. Keep in mind that slippery slopes are not merely about policies (i.e., policy A leads to policy B). Slippery slopes often result from intellectual frameworks, or ways of thinking. My point, from the beginning of this exchange, has been that Sunstein and Thaler (among others) have framed the debate about paternalism in an unfortunate way — one that will tend to encourage more serious intrusions down the line.

What is dangerous about their framing of the issue? I will highlight two features.

First, there is the failure to draw important distinctions between public and private, and between coercive and voluntary. Most of the new paternalists simply don’t mention these distinctions at all. Thaler and Sunstein, in their pre-Nudge articles, present all the harder paternalist policies (the ones Thaler now says he doesn’t actually support) as consistent with libertarian paternalism. So it doesn’t really matter whether Thaler himself supports those policies; he has constructed a framework that includes them. He provides no clear standard by which to draw the line between soft and hard paternalism. This feature of new paternalism makes it vulnerable to slippage.

Even in his most recent post, Thaler underscores this point: “We are merely stating the obvious, that all policies have costs and that there is no bright line between libertarian paternalism and what might be called ‘almost libertarian’ paternalism.” Why is there no bright line? Only because they have bizarrely defined ‘coercion’ in terms of the cost of exercising choice, without regard to who imposes the cost and how. Under this definition, there really is no line between (a) a fat tax imposed by the government, and (b) a self-imposed cost from storing fatty foods in a high cupboard.

Admittedly, bright lines are sometimes hard to draw. Even the brightest line may look fuzzy on close inspection. But in this case, the line is not so difficult. Self-imposed costs are not coercive. Costs that result from the laws of nature are not coercive. Costs that result from the free choices of others about how to use their property (such as a store owner’s decision not to stock Twinkies) are not coercive. But costs imposed as taxes and penalties by the government on private, voluntary transactions are indeed coercive. Unfortunately, the new paternalist framework tends to ignore, or even actively erase, this line.

(Just to be clear, I don’t think the cost of exercising choice is irrelevant. For a given policy that meets the criterion of coercion, a higher-cost version of that policy is generally more coercive. A $10 Twinkie tax is more coercive than a $1 Twinkie tax; ten years in prison for drug use is more coercive than one.)

Second, there is the matter of illegitimately choosing among preferences. I hesitate to bring this up again because this is where Thaler believes I’ve been most unfair to him. He states repeatedly that he only wants to help people “by their own standards.” The problem is that his analytical approach makes it unavoidable that someone must do the choosing — if not Thaler and his coauthors, then somebody else. The logic of their approach reveals that they are indeed choosing among preferences, whether they intend to or not.

Recall that behavioral economists have (often) demonstrated that people are irrational by means of experiments showing that people are internally inconsistent. They choose “A” under one framing of a decision, but “B” under a different (but analytically equivalent) framing of the same decision. In principle, such conflicts of preferences don’t reveal true preferences. Arguably, they reveal the absence of true preferences; this is a point that Sunstein and Thaler make repeatedly in their published work. We simply do not know whether people “really” prefer A or B, and there might not even be an answer to that question. So when the planners arbitrate by favoring the A-preference over the B-preference, they are necessary invoking some external standard. Maybe it’s their own preferences; maybe it’s socially approved preferences; maybe it’s politically correct preferences; or maybe they’re just flipping a coin.

Even if we suppose that true underlying preferences exist, we don’t know what they are. This is what Mario Rizzo and I have called “the knowledge problem of paternalism.” To successfully implement welfare-improving paternalist policies (such as optimal fat taxes or even optimal default rules), the planner must possess an insurmountably large amount of information: what people’s true preferences are, which biases prevent their achievement, the extent of those biases, how they interact with other biases, how much people have already self-corrected their biases, how much “unlearning” of self-correction will occur in response to policy, and how heterogeneous the public is on all of these dimensions.

The knowledge problem constitutes a direct argument against new paternalism; it casts doubt on whether even the wisest planners can implement it properly. But the knowledge problem also bears on the slippery slope risk. What will planners do in the absence of all that information needed to craft good paternalist policy? Friedrich Hayek, in his critique of central planning, hit upon the answer. Lacking information about their subjects’ preferences, planners end up privileging their own preferences — or, in a democratic system, those of biased voters and motivated special interests.

To summarize: Some new paternalists, like Thaler, may support only the mildest of measures. But their analytical approach will, if generally adopted, set in motion slippery-slope processes that are largely independent of their intentions. So I would like to see Prof. Thaler and other new paternalists do more than repudiate specific policies; I would like to see them remedy the objectionable features of their intellectual framework.

Also from this issue

Lead Essay

  • In his lead essay, Glen Whitman argues that “soft” or “libertarian” paternalism lends itself to ever more intrusive regulations in a variety of ways. Some of these are due to the very cognitive biases that advocates of soft paternalism have identified in their own research. For example, behavioral economists note that people exhibit extremeness aversion — a phenomenon in which they avoid what appear to be extreme positions. Yet introducing some amount of paternalism will make more paternalism, rather than less, appear to be the plausible middle ground. The dividing line between “soft” and “hard” paternalism is difficult to find, and Whitman offers many reasons why policymakers will tend toward more and more intrusive paternalism over time.

Response Essays

  • Richard Thaler argues that libertarian paternalism does not face any of the dangers that Glen Whitman suggests. Given that policymakers and administrators are commonly forced to set up default rules anyway, it only makes sense to set these up in ways that help the governed individuals to realize their true goals. Given that the problem of choice architecture is unavoidable, Thaler argues that this is a reasonable, mainstream solution.

  • Jonathan Klick argues that not only is the slippery slope real, but a host of other difficulties lie ahead for libertarian paternalists. While we might find academic behavioral economists sensible regulators, in reality regulations are made by lobbyists and Congressional staffers, whose incentives don’t necessarily lie in the direction of good policy. In the long term, too, allowing people to shift the burden of decision to others will habituate people to not deciding — and stunt their abilities.

  • Shane Frederick explores some of the unanswered questions of soft paternalism. He admits that Whitman is right to doubt whether we can postulate a given preference set as the “true” one for all time, for a given individual, and by implication for a given society. Still, he argues, this doesn’t mean we must refrain from all opinions about the behavior of others. Why not express such opinions in our choice architecture, if they are sincerely held and not terribly hard to opt out of?