Trust only 1/3 of the results from Pscyhology, but 2/3 of the results from behavioral economics

Dan Ariely (psychologist and behavioral economist) was forced to retract a well-cited paper based on fraudulent data. Ariely, a frequent TED Talk speaker and a Wall Street Journal advice columnist, cited the study in lectures and in his New York Times bestseller The (Honest) Truth About Dishonesty: How We Lie to Everyone — Especially Ourselves. We have blogged about Ariely's work and want to warn readers that we may have retract the blog posts, pending the outcome of an internal investigation.  

Professor Ariely's problems are part of a larger "replication crisis." It began when the journal Science found that only 1/3 of the psychology experiments underlying the most cited papers in the field of psychology  could be replicated.  A bigger and more recent study finds much the same thing.  [Behavioral Economics does better than psychology, as 2/3 of the results can be replicated.]

Contributing to the crisis is what is called HARK, Hypothesis After Results Known.  Due to the pressure to publish statistically significant results, researchers do science backward.  They begin by looking for a statistically significant result (p hacking).  Then, they specify an hypothesis to explain it.  Now, good practice requires that researchers register hypotheses before testing them.  Note that a similar problem plagues non-experimental data as well (regression fishing).

We regularly blog about deviations from the rational actor paradigm, and include some of the more useful results in our textbook, notably loss aversion (losses hurt more than gains help), and hyperbolic discounting (overweighting the present).  To the extent that these findings cannot be replicated, we could be wrong.

For the sixth edition, we will do our best to determine whether these Behavioral Economics results can be trusted.  For now, place a 2/3 weight on them being right.