Do Ethicists Steal More Books? More Data
I've finished collecting data relevant to the question of whether ethics books are more likely to be missing from libraries than non-ethics books in philosophy. You might think ethics books would vanish at a lower rate, if the people interested in them were influenced by the contents! (See this post for some earlier discussion.)
I was led to gather these data by what I call The Problem of the Ethics Professors -- the fact, or apparent fact, that ethics professors seem to behave no better than the rest of us. This is perhaps one small, imperfect way of assessing the presupposition behind that question. If I'm wrong, and ethicists really do behave better than the rest of us, perhaps this will reveal itself in their library habits?
It doesn't. For this analysis, I constructed a list of ethics books in philosophy and a comparison list of non-ethics books. The ethics list was derived from all the ethics books reviewed in Philosophical Review from 1990-2001 -- mostly technical work in philosophy principally of interest to advanced graduate students and professors -- combined with all the books originally published after 1959 that appear it at least five different bibliographies in the ethics entries in the Stanford Encyclopedia of Philosophy (including political philosophy, legal philosophy, and history of ethics, but excluding philosophy of action and moral psychology). The comparison list was composed of about 1/3 of the non-ethics books from the same issues of Phil Review (randomly selected, and excluding philosophy of action, moral psychology, and philosophy of religion), combined with books appearing in at least five different SEP entries on philosophy of mind or language (excluding philosophy of action and moral psychology).
I then looked at the holdings of these books at the libraries in UC Berkeley, UC Irvine, UCLA, UC Riverside, UC San Diego, UC Santa Cruz, Harvard, Stanford, Princeton, Columbia, Cornell, Michigan, and Texas, as well as all the COPAC libraries in Britian, which includes all the major university libraries. (I won't get into the principles of library selection.) A paid research assistant helped me with some of this!
I sorted the information into five categories: On shelf; checked out but not overdue; overdue one year or less; "missing" or more than one year overdue (which I will interpret as also missing); uninterpretable (e.g., "record unavailable").
Here are the uninterpreted numbers:
Ethics books:
Total holdings: 14,551
Total out or missing: 3,708
Total overdue or missing: 780
Total missing: 305
Non-ethics books:
Total holdings: 9,584
Total out or missing: 1,764
Total overdue or missing: 176
Total missing: 113
Obviously more ethics books are overdue and missing, but of course more are also held and checked out. The most interesting figures, I think, are these:
Overdue or missing, as a percentage of those off shelf:
Ethics: 21.0%
Non-ethics: 10.0%
Missing, as a percentage of those off shelf:
Ethics: 8.2%
Non-ethics: 6.4%
An ethics book is more than twice as likely to be overdue, given that it is off the shelf, and about 25% more likely to be missing!
The first difference is statistically significant at p < .001, the second at p = .02, using a simple one-proportion test (two-tailed, of course!).
Now obviously there are confounds. I'm trying to work on straightening those out, and I'll hopefully report more on them next week. I'd be interested to hear suggestions for further analyses.
Here are three worries I have:
(1.) Ethics books are more checked out than non-ethics books, and there is a correlation in the data between number checked out and percentage of those off shelf that are missing or overdue.
(2.) Older books are more likely to be missing than more recent books, and the weighted average age of the ethics books (weighted by number of checkouts) is about two years earlier than that of non-ethics books (1989 vs. 1991).
(3.) It might be those pesky law students! How will things look if I exclude philosophy of law texts or exclude results from law schools?
As far as I can tell from preliminary analysis, controlling for (2) or (3) slightly reduces the effect, but a statistically significant difference remains. Controlling for (1) seems to eliminate the second effect, but not the first.