Thursday 21 January 2021

Revisiting the Aaron A. Aardvark problem

Back in 2017, I posted about alphabetic discrimination in economics. Essentially, because the default in economics is for authors to be listed alphabetically, and there is some advantage to being the first-named author on a co-authored paper, authors with surnames earlier in the alphabet have an advantage. If your name was Aaron A. Aardvark, you're going to be significantly advantaged over Zachary Z. Zebra.

I recently read a 2018 paper by Debraj Ray (New York University) ® Arthur Robson (Simon Fraser University) that offers an alternative, which they term 'certified random order'. As they explain:

Here is a simple variant of the randomization scheme which will set it apart from private randomization. Suppose that any randomized name order is presented with the symbol ® between the names: e.g., Ray ® Robson (2018) is the appropriate reference for this paper. Suppose, moreover, that such a symbol is certified by the American Economic Association, for example, simply acknowledging that this alternative is available.

The paper itself is quite mathematical and not for the faint of heart. Ray ® Robson show that:

...® can be introduced not as a requirement but as a nudge, because our results predict that it will invade alphabetical order in a decentralized way. It may provide a gain in efficiency. But, more important, it is fairer. Random order distributes the gain from first authorship evenly over the alphabet. Moreover, it allows “outlier contributions” to be recognized in both directions: that is, given the convention that puts “Austen ® Byron” (or “Byron ® Austen”) on center-stage, both “Austen and Byron” and “Byron and Austen” would acquire entirely symmetric meanings.

The latter bit of that quote is important, and solves a problem I had not considered before. Under the current convention, if a paper is Austen and Byron, that conveys no information about which author did most of the work (if one of them did), but a deviation from the convention (i.e. Byron and Austen) does. In the case of Byron and Austen, it tells you that Byron did most of the work. There is no way for Austen to receive that credit if they did most of the work, since Austen and Byron would be seen as simply following convention. The certified random order approach (denoted by ®) solves that problem.

You would think that authors with surnames early in the alphabet would be against anything that disrupts their dominance. But the 'Byron and Austen' problem may be enough to encourage Aaron A. Aardvark to accept certified randomness.

Despite the mathematical advantages, it doesn't seem like many other authors have taken up the certified random order approach though, in the two-plus years since publication of the article. In fact, I still see some authors noting random order of their names by a footnote on the first page of the article. It can take some time for new conventions to take hold - I wonder if this one will ever do so.

Read more:

No comments:

Post a Comment