Economist Stan Liebowitz has a longstanding interest in the difficulties of flagging published research errors. Recently he wrote on the related topic of dishonest authorship:
While not about direct research fraud, I thought you might be interested in this paper. It discusses the manner in which credit is given for economics articles, and I suspect it applies to many other areas as well. One of the conclusions is that the lack of complete proration per author will lead to excessive coauthorship, reducing overall research output by inducing the use of larger than efficient-sized teams. Under these circumstances, false authorship can be a response to the warped reward system and false authorship might improve research efficiency since it might keep actual research teams (as opposed to nominal teams) from being too large to produce research efficiently. One of the questions I rhetorically ask in the paper is whether anyone has ever been ‘punished’ for having their name included on a paper for which they did not perform their share of the work (I think the answer is “no” in economics).
Liebowitz may well be right when it comes to the incentive structure. There is very little cost to including collaborators. Much of my most successful research has been collaborative (with varying percentages of contributions from the collaborations). Many times I’ve added a collaborator after most of the work had already been done, just because I think it makes the work stronger.
But . . . Liebowitz argues that excessive collaboration can harm research, if people would get more done working on their own. I have a different perspective in that I think even a small collaboration by a coauthor can make an article or book much stronger. Given that this seems to hold in statistics, where we publish dozens of papers a year, I’d think it would be even more the case in economics, where researchers take years to publish a single article.
True, collaboration did not save economists Carmen Reinhart and Kenneth Rogoff from making three serious errors in a single article, but—no joke—I think that adding a third collaborator might have made a difference and allowed them to have avoided their errors.
Look at it this way. Either Reinhart and Rogoff had a research assistant do their calculations or they did it on their own.
- If it was a research assistant who ran the numbers, it seems quite possible to me that being a coauthor would’ve been a big deal for the RA, and he or she would’ve been super-careful to get the spreadsheet calculation right and (if the person was a high-quality RA of the sort that Reinhart and Rogoff could certainly afford) might even have caught the other problems with the analysis. (True, Reinhart and Rogoff still don’t seem to want to admit there are other problems, but of course they’re in defensive mode now. They might have been more open during the research process.)
- If Reinhart and Rogoff did the analysis . . . well, we’ve already seen that they didn’t have full control over their data. Maybe they were too busy to take care to do things right. Having a coauthor with real responsibility could’ve made all the difference.
Liebowitz also writes:
As a measure of research success for senior faculty, department chairs, on average, rely most heavily on the journal of publication, with a weight of 40%. Citations, by way of contrast, were in third place, receiving a weight of 26% . . . This is analogous to a sports team picking its starting lineup based on the ex ante performance of players on athletic tests intended to predict success on the field, instead of using the ex post actual performance in games.
This is an issue too, highly relevant to the question of incentives for untenured academic researchers, but to me somewhat distant from the larger concerns about research quality.
Don’t get me wrong—I don’t think coauthorship is a panacea for the problems of sloppy research. But for an important project, I see real advantages to having multiple people taking responsibility for the result.
P.S. Here’s Table 1 of Liebowitz’s paper:
I think you’ll agree that it should be a graph.