Nudging, shoving, and manhandling

I’ve been puzzled why Richard Thaler’s “nudge” idea attracts such hostility from some people to my political left (including very smart people such as Henry Farrell and Cosma Shalizi). The worst thing you can say about nudging as I understand it is that it’s not very powerful; other than that, nudging is like chicken soup: it can’t do any harm.

So I’m grateful to Tyler Cowen for clarifying matters for me. Either Cowen or I badly misunderstands Thaler’s idea; if Cowen is right, you can add me to the list of anti-nudgers. But I’m pretty sure the Cowen is wrong about what Thaler says, and certain that his account confuses things that ought to be distinguished.

Nudging, as I understand it, involves changing “choice architecture” – altering the way options are presented or the time choices are made, or changing the “default outcome” if no option is explicitly chosen – in order to bring people’s actual choices more closely in line with their true preferences, as measured by the choices they would make with full information after serious reflection. That is, nudging is simply the opposite of temptation.

One of the defining features of a nudge (understood this way) is that it doesn’t narrow the range of outcomes available to the chooser. For example: presented with a menu of retirement-savings options, many employees will pick none of them, in part because of the psychological costs of decision-making and the fear of getting it wrong (“analysis paralysis”). This can be true even in the case when inaction is clearly the worst option (e.g., when the employer is picking up all of the cost). In that case, a nudge strategy would be to make enrollment in the plan that seems to experts most appropriate for the largest number of employees the default option: i.e., what happens if an employee just doesn’t fill out the form.

Crucially to the definition of a nudge, an employee who doesn’t want that option can costlessly (other than the effort of making the decision) switch to another, or none at all. As long as there’s no deception involved, and the people designing the choice architecture know what they’re doing and have the welfare of the people making the choices in mind, nudging seems to me almost entirely benign. A program that doesn’t limit freedom of choice can’t properly be said to reduce liberty, so replacing “opt-in” with “opt-out” should be thought of as facilitative rather than coercive. The same is true of, e.g., putting the salad bar first in the cafeteria line.

However, Cowen’s understanding of nudgery has a much harder edge. He gives examples where a choice less preferred by the government (or whoever is setting up the system) is made materially less attractive or more expensive, such as legally complicated and expensive divorce procedures, or abortion restrictions that force women to travel inconvenient distances. Cowen even wants to call restrictive immigration laws “nudges,” because would-be immigrants who can’t get visas can always forge documents or sneak across the border!

In my view, that sort of cost-imposing policy is radically distinct from “nudging;” Steve Teles calls it “shoving.” I don’t doubt that some such “shoves” are justified on paternalistic grounds: taxation to reduce cigarette consumption is an example. (Shoves are often justifiable on non-paternalistic grounds, such as taxes to reduce air pollution.) But such strategies aren’t always benign; people who keep smoking in the face of heavy tobacco taxes wind up just as sick as they would have otherwise, and poorer. And of course for those with limited means making something expensive can amount to barring it entirely.

Now, I agree with Cowen that the “shoving” for paternalistic reasons he wants to label as “nudging” is often preferable to more drastic means of protecting people from their own bad decisions: means that we might call “manhandling.” A tax on cigarettes is more respectful of liberty, and less prone to generate bad side effects, than an outright prohibition would be. But – in contrast to nudging – shoving is like manhandling in making those who don’t take the hint worse off. Changing incentives isn’t the same thing as changing choice architecture, and requires much stronger justification.

Nudging is no panacea, because changing choice architecture can only go so far in changing choice. Some people will continue to fall into behavioral traps even if the traps are clearly marked. And when the intervention is on non-paternalistic grounds, a considerable amount of shoving, or even manhandling, may be both justified and required. But there is clearly some room for improving outcomes at low cost and without diminishing liberty from the use of pure nudges. It’s therefore worthwhile to distinguish clearly – as it seems to me Cowen’s analysis does not – between generically benign nudges and the less benign alternatives he wants to include under the same label.