Behavioral economists Richard Thaler and Cass Sunstein are known for advocating a sort of “paternalistic libertarianism.” The basic idea is to nudge people toward more rational behavior through non-coercive means. Indeed, their book on the subject is entitled “Nudge“.
While there is much to be said in favor of this idea, especially when compared with more coercive governmental alternatives, this approach is not without risks. Joshua Madsen and Jonathan Hall studied the effect of electronic highway signs designed to frighten motorists into driving more carefully. The Economist reports that their study found some unintended consequences:
The study focused on Texas, where the year’s cumulative death toll from road accidents was displayed on highway signs one week in four. The authors found that, between 2010 and 2017, there were more accidents in the weeks when death counts were shown. Most excess crashes happened in the kilometre after a sign, but for several kilometres there was still an elevated risk (see top chart). . . .
The authors think that the sombre messages may be distracting drivers.
Luckily, the story has a happy ending:
The study highlights how seemingly innocuous “nudges”, used by governments to try to change behaviour, can backfire.
Luckily America’s government has given a nudge of its own. Last year the Federal Highway Administration released a memo clarifying that it was inappropriate to use electronic highway signs to display death tolls.
READER COMMENTS
gwern
Jul 29 2022 at 9:09pm
The most important lesson here is about the value of randomized experiments, even for things which seem “obvious”. Surely even the skeptics involved in this experiment merely thought that the benefit would be too small to be worth bothering with, rather than what it actually turned out to be!
Michael Sandifer
Jul 30 2022 at 11:14am
This is interesting, but I question such easy interpretations. Who knows what is really going on here? Perhaps distraction is the most plausible hypothesis for the reported results, but it could be that, at least for some drivers, the number of deaths shown was seen as low, and encouraged faster, less careful driving.
I understand that traffic deaths are way up in the US over the past few years, and this is also true in Texas, but they still represent a tiny fraction of drivers each year. If I were driving as saw a sign like this, I suspect it wouldn’t influence my behavior at all, but I could be wrong. Perhaps the mere novelty of it would distract me and cause a problem. I’m a terrible driver anyway, so whatever happened to me would not necessarily be representative.
Dylan
Jul 30 2022 at 12:49pm
I’m with Gwern in saying this really underscores the importance of RCTs, but also with Michael, in that I’d want to understand the design of the trial a little better before jumping to a particular conclusion. There’s a lot that could be going on and designing a good trial isn’t easy (was this a prospective or retrospective look at the data for instance? Was the sign turned on only some weeks for the sake of the trial, or was it perhaps only turned on during holiday weeks?)
Finally, a broader word about nudges. I think it is important to realize, in many/most of the cases where nudging is suggested, there is no “no nudge” option. Think of the common example of opting new hires into the 401K, the old option of opting everyone out by default was just as much of a nudge as opting people in.
Scott Sumner
Jul 30 2022 at 2:46pm
“Think of the common example of opting new hires into the 401K, the old option of opting everyone out by default was just as much of a nudge as opting people in.”
Yes, and to be clear I’m not always opposed to nudges.
David Seltzer
Jul 30 2022 at 6:24pm
“Yes, and to be clear I’m not always opposed to nudges.” Agreed, as long as a nudge from the government doesn’t become a shove.
robc
Aug 1 2022 at 3:31pm
I don’t think the opt-out can really be called a nudge, though. I see your point, but the default condition is to give someone all their money, so that isn’t really a nudge.
nobody.really
Aug 2 2022 at 5:10pm
Seems like a semantic distinction.
Employers set employees up with a default x% of their income put into a 401(k), and employees have the option to change the default. Evidence reveals that the amount employees invest in their 401(k) accounts is greatly influenced by the default amount established by the employer.
You can choose to attach some significance to the figure 0% that you don’t attach to the figure 8%–but, as far as I can tell, the analysis remains the same.
Bob
Jul 30 2022 at 6:03pm
To the people praising RCTs of government policies: they obviously violate several Nuremberg codes. No voluntary consent, there was a prior reason to believe this highway message test would lead to excess death, etc. So… just something to think about.
Also, behavior is rational in the context of a goal and beliefs. A rain dancer is behaving rationally, he just holds false beliefs. A speeder late for a meeting is also behaving rationally. Usually, “nudging towards more rational behavior” is just doublespeak for “nudging towards behavior that I approve of”.
Matthias
Jul 30 2022 at 9:26pm
Surely the RCT only violates things that the full policy implementation would also violate?
Bob
Jul 31 2022 at 1:43pm
People are going to want to normalize these kinds of policy experiments in the name of scientific progress. I worry that this is a slippery slope. To the extent that we have government coercion, it should not be done to learn about humans but because it is believed to be the best policy.
Jon Murphy
Jul 31 2022 at 7:38am
I’m with Matthias. I don’t see how the Nuremberg Code applies. Ignoring the fact the code is not legislation anywhere, it only applies to medical research. Ordinary policy isn’t covered.
Bob
Jul 31 2022 at 1:45pm
I thought it applied more broadly to human experimentation, but maybe I’m wrong.
nobody.really
Aug 2 2022 at 5:25pm
I largely share this view. Thaler’s premise is that we have some kind of social science indicating that certain behavior (reducing smoking, for example) would produce better outcomes (for the individual, or the policy designer, or both), so the policy-designer implements the policy to promote those interests while still affording the individual the option of making a different choice.
But yes, nudging involves a policy designer trying to influence another person’s behavior–and yes, policy designers are not godlike in either insight or benevolence. But we should also recognize that policies DO influence people’s behavior whether or not anyone intends that result. If I’m inevitably going to be influenced by policies, I’d rather be influenced for the better. I take little comfort in the idea that I’m being influenced out of the policy-designer’s ignorance or indifference.
Johnson85
Aug 1 2022 at 11:59am
The descriptor “paternalistic libertarianism” seems inaccurate to me. This may be unfair, but every example I think I have ever seen related to their work involves either something that doesn’t really impact that involvement of the state, or it increases it, but with a “nudge” rather than a requirement. I never see their work cited as a reason for replacing a government mandate with a nudge. “Light handed statism” or “Lighter handed interventionists” aren’t very catchy phrases, but seem closer?
Scott Sumner
Aug 1 2022 at 12:50pm
I’m actually fine with either term—your suggestion or their suggestion.
robc
Aug 1 2022 at 3:30pm
An example, that could exist, from above. If the nudge to opt-in to a 401k was a replacement for Social Security, that would be an example. But the nudge alone is just adding a nudge to a shove.
Comments are closed.