If you needed any additional evidence to treat evidence on the benefits of nudge interventions more skeptically, check out this detective work that appears to find dishonesty in a behavioral econ experiment on how to improve honesty. The widely-cited experiment by high-status researchers claimed to find that people were more likely to report the mileage on their car odometer more honestly if they had to sign at the top of the form affirming the veracity of their report rather than at the bottom of the form.
This study fueled the Obama Administration’s Social and Behavioral Sciences Team in their effort to construct a nudgocracy run by people who are way smarter and better than everyone else. It’s for your own good that the nudgocrats shove, er… I mean…, nudge you into doing things that they’ve decided are better for you. To oppose this is to deny science.
But it turns out that nudge research, like a lot of other research in social science, is plagued by problems with replicability and applicability when tried on scale in the real world. As readers of JPGB may remember, texting nudges meant to improve educational attainment by disadvantaged students ended up reducing their likelihood of completing college. Another nudge intervention meant to increase savings by making contributions to retirement plans the default ended up reducing net savings by encouraging people to take on greater debt.
As a recent review concluded, “nudges fail more often than is reported.” Nudges fail for many reasons, including the fact that it is extremely hard, even for very smart and well-intentioned people, to anticipate how others will respond to seemingly innocuous and subtle interventions. Even worse, many failed nudges are never published, contributing to an over-confidence in the effectiveness of policies shaped by behavioral economics. Researchers simply assume they must have designed the intervention wrong when they get the unexpected result, discard the finding, and try again after tinkering with the nudge in the hopes that it will be more effective. They repeat until they get the desired result and then proclaim eureka!
But the nudge on honestly reporting odometer mileage appears to have gone wrong in a less innocent way. With some clever sleuthing, these investigators uncovered evidence that data were fabricated and otherwise manipulated to get the desired result. Given how difficult and unrewarded this type of detective work is in academia, who knows how widespread these less innocent causes for flawed research really are.
Before the nudgocrats expect us to obey them, they might want to invest considerably more in strengthening confidence in their work. Perhaps we need to shove, er, I mean, nudge them into greater humility about the policy utility of behavioral economics.