Wednesday, July 30, 2014

Surprisingly high non-success?

This article by Conley and Önder in the latest Journal of Economic Perspectives purports to show the alarmingly low rates of research productivity among the large majority of Econ PhDs from even the very top departments. From their discussion:
Only a small percentage of economics PhDs manage to produce a creditable number of publications by their sixth year after graduation. Even at the top five departments, it would be hard to argue that the bottom half of their students are successful in terms of academic research. The number of AER-equivalent papers of the median at year six is below 0.1 in all cases and is in fact zero in most. At the majority of the departments ranked in the top ten in conventional rankings (such as Coupé 2003), 60 percent of their students fail to meet this 0.1 AER-equivalent standard, and for the majority of the PhD graduates of the top 30 departments, 70 percent fail. A tenure standard of 0.1 AER-equivalent papers is roughly equal to publishing one paper in a second-tier field journal over six years.
For all you non-economists, AER is the American Economic Review, by most reckonings the top economics journal. To place the final sentence of the passage in context, I can say with considerable confidence that at my university, "one paper in a second-tier field journal over six years" would not come even close to meeting the research requirements for tenure. In other words, the typical PhD from a top-ten department would be unlikely to make tenure at my mid-level university, never mind a major R-1 institution. Depressing news, for sure.

But hold on a sec. Let's unpack the numbers a little. To count "AER-equivalent papers," Conley and Önder assign a fraction of an AER "hit" to articles in journals with lower impact factors, using a journal ranking created by Kalaitzidakis et al. I have read the Kalaitzidakis paper, and their methods make some sense. But according to their adjusted impact factors, the very top field journal in my own sub-discipline, the Journal of Economic History, receives about 5 percent* the weight of the AER. That's right, by their calculation, I would need 20 articles in JEH to have one AER-equivalent. That would keep me very busy!

And there's more. Without justification, Conley and Önder discount co-authored papers proportionally by the number of authors. So suppose you have published ten articles in JEH or equivalent journals: four solo-authored, three with a single co-author, and three with two co-authors. Co-authorship, by the way, is increasingly common in economics. This would be an enviable record for an economic historian at many schools, if not perhaps Harvard or Berkeley. For Conley and Önder, it would count as (4+3/2+3/3)*.05 = 0.325 AER-equivalents. That's right: your 10-article CV is worth one-third of a (solo-authored) AER article.

Some other even more competitive field journals, such as the Journal of Labor Economics or the Rand Journal, score higher than JEH in the rankings, but they still count for only a fraction (about 14 percent) of the AER. So you'd need seven of those to get an AER's worth.

The authors also fail to account for the possibility that some not insignificant fraction of highly accomplished PhDs in economics choose not to pursue the academic publishing route, but may instead work for organizations for which their performance is judged on the quality of internal reports and advice. Their lack of peer-reviewed publications is hardly a sign of failure.

The authors conclude that the top PhD programs are serving their students ill, and I won't necessarily argue with that. Academia is changing, and the path to tenure at a decent university seems more fraught with hard work and long odds than ever. But their analysis needs a heavy dose of realism–in particular, specification of what economists would call an objective function. If your goal is to be a tenured professor at a decent college or university, what would it take to succeed? And how likely is success if you can get yourself through a top 10 or 20 program? I don't think Conley and Önder have really answered these questions.

[* Actually Kalaitzidakis et al present several different numbers, and the number Conley and Önder use for JEH may even be under 4%, which would only strengthen my point.]

No comments:

Post a Comment