In previous posts, we discussed how to categorize authors by field and then how to rank them within fields. These discussions are still open and I can still be convinced to change the procedure. Today, I would like to make a proposal regarding the ranking of institutions within fields.
We have several options regarding how to count people from an institution for a specific field. In the examples, I assume that author A has 10, or 50%, of his works in macroeconomics, B 3 or 30%, C 10 or 20%, D 1 or 10% and A 0%. I also assume that the 25% and 5 rule applies, as discussed in the post on categorizing authors. Thus, under these rules, authors A, B, and C are considered macroeconomists. The options are:
- Count fully all authors considered within the field: A+B+C.
- Count all authors considered within the field proportionally to their involvement in the field: 0.5A+0.3B+0.2C.
- Count all authors, irrespective whether they qualify as specialists, proportionally to their involvement in the field: 0.5A+0.3B+0.2C+0.1D.
My preference is for option 3. The reasons are the following. The first option fails to properly differentiate between strong specialists and marginal ones. This may also have been a concern when ranking authors, but the issue there was the high volatility of weights at the author level. At the institutional level, this is less of a concern as several authors are aggregated. Note also that only the top 20% institutions will be listed anyway, thus I expect all of them to have several authors within the field. Thus I prefer option 3 over option 1. Then I prefer option 3 over option 2 because it allows to count for authors that may not be specialists but still may contribute to enriching the field. Think for example when a prospective graduate student compares programs. While she cares about the specialists of her field of interest, she may also care about those faculty on the fringes of the field.
Of course I am open to suggestions and can still be swayed to to change my opinion. I plan on implementing this for next month.
Option 3 is the best among the three. Note that because papers can be in several NEP fields, author weights do not add up to one.
Option 4 is better still: Only count those papers, citations, and downloads of papers in macroeconomics. This would lead to distortions until virtually all papers at I/R are classified.