I recently came across an article published last year in the journal Library Philosophy and Practice titled “Conventional Wisdom or Faulty Logic? The Recent Literature on Monograph Use and E-book Acquisition.” Written by Amy Fry of Bowling Green State University, the article critiques what Fry sees as an “inaccurate and misleading” invocation of the conventional wisdom on academic library book circulation by proponents of radical change to library collection-building practices. (I should probably disclose here that I am one of the villains Fry cites in the article.)
The particular piece of conventional wisdom at issue here is the idea that in academic libraries, roughly 40% of librarian-selected books circulate rarely or never, and that this fact represents an indictment of traditional collection-development practices. “A closer look at the literature,” Fry contends, “reveals that the data being cited to support this claim has been communicated, for the most part, in an inaccurate and misleading way by its proponents and that a great deal of data exists, in fact, to refute it.” Fry provides a review of recently-published arguments that draw to some degree on this common belief about circulation patterns in academic libraries, and then cites a list of studies that she believes disprove that belief.
Fascinatingly, however, the very data she provides end up supporting the conventional wisdom with an almost spooky exactness.
Let’s back up for a moment, though, and look at the source of the conventional wisdom. Where have so many people gotten the idea that 40% of the books in academic libraries rarely or never circulate?
The urtext for this belief is a 1979 book based on research conducted by Allen Kent et al. at the Hillman Library of the University of Pittsburgh. Variously referred to as the Kent Study, the Pittsburgh Study, or the Pitt Study, this research project famously found that 39.8% of that library’s books “had never circulated during the first six years on the library’s shelves,” and, furthermore, made a statistical argument that it was unlikely those uncirculated books would ever be checked out in the future.
The Kent Study is indeed cited frequently in the library literature on collection development, and Fry is right to note that it is often referenced by those promoting the substantial (or even complete) replacement of traditional collection development practices with patron- or demand-driven acquisition models. She is also correct in noting that the Kent Study represents data that is now several decades old, and that many other circulation studies have been conducted both before and since.
But does the data produced by those other studies refute the findings of the Kent Study? Based on the evidence she provides in her article, the answer seems to be no — quite the opposite. Let’s take a closer look.
Fry shares data from circulation studies conducted at 18 institutions or consortia: one regional consortium, four ARL libraries, and 13 smaller academic libraries. The data are summarized below. (For those following along, please note that her article misidentifies the University of Denver as an ARL library and Louisiana State as a non-ARL library; those errors are corrected in the summary below.):
- 66% of the 1.3 million books purchased by libraries in the CARLI consortium between 2003 and 2008 circulated during those years. (This means that 34% did not circulate during that period.)
- 69% of books purchased on approval by Penn State Libraries circulated in the first three years after purchase. (Thus, 31% did not circulate.)
- 60% of books purchased on approval by UIUC Libraries circulated in the first three years after purchase. (40% did not circulate.)
- 62.5% of print books purchased by the Kent State University Libraries between July 2009 and December 2011 had circulated by December 2012. (37.5% did not circulate.)
- 45% of books cataloged at Louisiana State University between January 1 and August 31, 1991, had circulated by January 1992. (55% did not circulate.)
- 58% of 2005 imprints owned by the University of Denver library had circulated by 2009. (42% did not circulate.)
- 45% of books purchased by Muhlenberg College in 1940-41 had circulated by September 1942. (55% did not circulate.)
- A 1994 study at the University of Illinois Chicago (UIC) Health Sciences Library found that 58% of books acquired between 1987 and 1989 had circulated by 1989. (42% did not circulate.)
- A random sample of roughly 1% of the collection of the University of Cape Town’s library found that 84% of them had circulated at least once. (16% had never circulated.)
- 84% of books acquired by the University of New Mexico Health Sciences Center Library in 1993 had circulated within 4 years. (16% had not circulated.)
- In 1993, an examination of four health sciences libraries found that 86%, 82%, 79%, and 62% of books published between 1980 and 1992 had circulated. (14%, 18%, 21%, and 38% of these books, respectively, had never circulated.)
- 40% of books acquired in September of 1995 by the library at Western Michigan University circulated within five months; 54% of books acquired in November 1994 circulated during the first sixteen months after acquisition. (60% and 46%, respectively, had not circulated.)
- A 2000 study at the UIC Health Sciences Library found that 81% of books acquired by that library in 1994-1995 had circulated within the first three years. (19% had not circulated.)
- 44% of books acquired by Lingnan University circulated during the first year after acquisition; 67% circulated during the first six years. (56% and 33%, respectively, did not circulate.)
- 37% of books acquired between 2003 and 2008 by the Asbury Theological Seminary in Kentucky had circulated by 2012. (63% had not circulated.)
- 30% of books acquired between 2003 and 2008 by the Asbury Theological Seminary in Florida had circulated by 2012. (70% had not circulated.)
Now obviously, this is a highly diverse data set; notice the variety of libraries involved (research libraries, college libraries, health sciences libraries, theological libraries, etc.), the different types of collections examined, the different amounts of time measured, etc. On the one hand, this suggests that we’re dealing with incommensurable data – how can you compare the implications of data showing how many books have ever circulated at Cape Town University with those of data showing how many books circulated between 2003 and 2008 at a theological seminary in Kentucky?
However, it’s the very diversity of these studies that makes their combined findings so remarkable. Despite the radical diversity of these data sets—so many different kinds of libraries measuring circulation of different numbers of books over different periods of time—both the mean and the median of the aggregated non-circulation findings match almost exactly the 39.8% figure found by the Kent Study:
Here’s another interesting (and maybe alarming) fact: if you remove health sciences libraries — where circulation patterns might reasonably be expected to differ substantially from those in conventional academic libraries — from the data set, the data actually show substantially higher non-circulation rates than those found by the Kent Study: for non-medical, non-ARL academic libraries, the mean non-circulation value is 54.71% and the median is 56% — both roughly one-third higher than the non-circulation rate found in the Kent Study. (And, oddly enough, exactly the rate of non-circulation that Christopher Stewart describes — without providing adequate support for the assertion — as being typical of ARL libraries in a 2011 report.)
Now, it’s important to note that the diversity of these findings does matter in and of itself and has significant implications for particular libraries. Obviously, the fact that the mean and median values of this data set both tend to vindicate the Kent Study doesn’t mean that any individual library should automatically assume that 39.8% of its books aren’t being used. (And no one, to my knowledge, has ever argued that all academic libraries of every kind show the exact same pattern of circulation and non-circulation.) However, it is truly striking that when you tally the values of this very diverse data set, you find the Kent Study’s findings right smack in the middle of the distribution.
But all of this begs a more important question: what do these findings mean for your particular library? I would suggest that for any individual library, these general trend data should prompt local examination. A good internal question would be: “Given that the available data point to a mean/median of non-circulation among academic libraries of roughly 40%, what do the circulation data look like in my own library?” The answer to that question — and the attendant implications for collection development practices — will vary from situation to situation. But if you look at academic libraries in the aggregate, it appears that the Kent Study is being strongly vindicated.