On publication bias in economics, from the comments

I’m not surprised by this finding.

Economics has a much higher bar for identification and often relies on observational data, so there are always going to be many possible problems with an empirical exercise that could be used as justification to turn down a paper. In this environment it becomes extremely important to have stars in your regression table to avoid one of these justifications being seized on as the damning explanation for why your test is no good.

Procedurally, what economists like to see is a paper that tests the predictions of a well-specified theory and comes back with statistically significant results in every single case plus a smattering of increasingly arbitrary robustness tests that can be found in the appendix. In many cases the theory itself is one that appeals conceptually to one or more of the referees (i.e. is consistent with their prior work) which further makes a null finding hard to sell.

That is from Infovores.


Comments for this post are closed