knieval.indd 35 Use of Circulation Statistics and Interlibrary Loan Data in Collection Management Jennifer E. Knievel, Heather Wicht, and Lynn Silipigni Connaway Jennifer E. Knievel is an Assistant Professor and Humanities Reference and Instruction Librarian at the University of Colorado at Boulder; e-mail: jennifer.knievel@colorado.edu. Heather Wicht is an Assistant Professor and Electronic Resources Specialist at the University of Colorado at Boulder; e-mail: Heather. wicht@colorado.edu. Lynn Silipigni Connaway is a Consulting Research Scientist at the Office of Research, OCLC Online Computer Library Center, Inc. (OCLC); e-mail: lynn_connaway@oclc.org. The authors would like to thank the following people for their substantive and helpful assistance with this project: Edward T. O’Neill, consulting research scientist, OCLC; Cli�on Snyder, so�ware engineer, OCLC; and Justin Li�man, information technology specialist, Office of Strategic Initiatives, Library of Congress (formerly Systems Analyst, NetLibrary, a division of OCLC). The authors analyzed the holdings, circulations, and interlibrary loan (ILL) borrowing requests of the English-language monograph collection at the University of Colorado at Boulder. Data for each area were mapped to conspectus subject areas, using Library of Congress Classifications, and then compared. The resulting data and subject distributions were analyzed by overall holdings, transactions per item, percentage of col- lection circulated, and a ratio of ILL holdings to requests. The method of analysis used in this study could be fruitfully applied to other research collections to assist with remote storage, preservation, and collection development decisions. n the current economic en- vironment, it is critical for librarians to practice evi- dence-based decision mak- ing. With decreasing library budgets, especially for monographic materials, it also is essential for librarians to assess and manage collections to determine subject areas for acquisition, de-acces- sion, digitization, preservation, and remote storage. One empirical collection assessment method is to examine usage statistics, such as circulation and inter- library loan (ILL) data. Many librarians employ usage data as an indicator of the library’s materials’ relevance to user populations. Sonia Bodi and Katie Maier-O’Shea believe that librarians’ knowledge of collections is often intuitive and urge librarians to make data-supported col- lection decisions. They suggest that col- lection development focus on learning outcomes of library users rather than on strictly comprehensive collections or even curriculum support.1 This approach is impossible without solid data from vari- ous perspectives. Librarians at the University of Colo- rado (UCB) at Boulder Libraries utilized circulation statistics and ILL data in the development of collection and as- sessment criteria that accurately reflect patrons’ needs. UCB is a Research I, doctoral-granting institution with 26,000 FTE students. The libraries hold approxi- mately three million volumes. Collection assessment has become more important with the current budget reductions at UCB, a problem shared by many libraries around the state and country. The materi- als budget has suffered severe permanent and one-time cuts that make efficiency in collection management increasingly important. Like many university libraries, the UCB libraries’ shelves have been filled to capacity for some time. In 1998, UCB began a remote storage project, and to- day, approximately 425,000 volumes are stored in a shared remote storage facility in Denver. The UCB subject bibliogra- phers have used a variety of methods and tools to select materials to be placed in remote storage. One of these methods is the analysis of circulation statistics and ILL borrowing data. OCLC Online Computer Library Cen- ter, Inc. (OCLC) maintains the WorldCat bibliographic database, the Online Union Catalog, which contains more than 50 mil- lion records. WorldCat serves not only as an aggregator of bibliographic data but also lends itself to data mining and data analysis. The WorldCat database includes the holding symbol for every member library holding each item represented in the database. There are almost a billion holding locations for library resources identified in WorldCat. This study compares the UCB libraries’ circulation statistics and ILL borrowing requests between January 1, 1998, and December 31, 2002, to UCB’s WorldCat holdings data during this same time period. Four analyses are compared and discussed: the UCB’s overall hold- ings; the average number of circulation transactions per item; the percentage of the subject collections circulated; and the holdings to ILL ratio. All are compared using conspectus subject categories. These analyses can provide empirical data for collection management and remote stor- age decisions. Literature Review Reduced buying power in libraries makes the use of data in collection develop- ment of increasing importance. Difficult selection and funding decisions are more informed when they consider usage data. In 1995, Chuck Hamaker stated that col- lection for the sake of the collection is no longer possible except in the most financially insulated institutions.2 Charles B. Osburn stated in 1992 that economic stresses and strains influenced a new set of guiding principles for library and infor- mation services.3 One such principle is a shi� in emphasis from spending acquisi- tions budgets on the basis of speculation to spending on the basis of demand. As Dennis P. Carrigan observed, this shi� should increase the importance of data produced by automated systems.4 Car- rigan wrote that the recent crisis of schol- arly publishing has made it necessary for librarians to prove that their libraries are making good use of the materials they have. The 80/20 rule (80 percent of us- ers’ needs are satisfied by 20 percent of the collection) is no longer satisfactory in this fiscal climate.5 Carrigan believes that increased accountability in libraries will necessitate the use of circulation and other types of data. A 1998 Library Journal survey found that “virtually every library LJ surveyed relied on faculty when making purchasing decisions, and half of them ranked faculty as the number-one source.”6 Although faculty suggestions should be considered, they should not be the only source of input. The library must satisfy the needs of all users. According to Hamaker, “both computer science and math faculty say don’t buy books, we need journals; usage pa�erns say something quite different to the selectors.”7 He also speculated that 36 College & Research Libraries January 2006 even in ARL libraries, some selectors still have the attitude that they know best what to buy and do not need to know what is being used (and who is using it). Hamaker has advocated the use of data in collection development, stating that intuition is not the key to understanding the usage of library collections. Lynn Silipigni Connaway and Justin Littman state, “Circulation analysis is one of the traditional approaches taken to use studies and collection evaluation in libraries. The results of circulation analyses have been applied to a number of important issues, including evaluating collection acquisition policies, guiding such management decisions as allocating physical space for materials, identifying materials for offsite storage, allocating funding for materials, and suggesting approaches to deselection.”8 Although the collection and analysis of circulation data have been fairly com- mon practices for some time in libraries, an examination of the existing literature on the topic suggests that circulation data are not being widely used in com- bination with other types of data for collection development decision making. The articles that do discuss data-based collection development tend to consider only circulation data, not ILL data or a combination of both types of data. Mike Day and Don Revill stated that “circula- tion is the strongest single element we have on which to base decisions and we feel that the analysis of this data provides a useful tool; a tool that could be used by our subject librarians along with their professional judgment and user satisfac- tion surveys to evaluate and fine-tune their collections.”9 In an article describing a variety of methods of collection evalua- tion, George S. Bonn introduced a method of obtaining proportionate circulation statistics by subject class. The number of circulations in a given subject area is compared to the number of holdings in the same subject area, resulting in a ratio Bonn labeled the “use factor ” for that subject area.10 William Aguilar wrote one of the first articles that advocated the combined use of circulation and ILL statistics in collection development.11 It described a series of basic methods for comparing holdings data, circulation statistics, and ILL statistics to assess the activity of the library collection. He recommended analyzing collections by subject areas and described three primary methods for evaluating subject areas. Aguilar suggest- ed that Bonn’s use factor can be applied to determine the number of circulations relative to holdings. He further explained that Bonn’s use factor takes on greater significance as refined by Terry R. Mills’s “percentage of expected use,” which is simply the use factor multiplied by one hundred to obtain a percentage.12 Subjects above 100 percent can be defined as over- used whereas subjects below 100 percent are underused. Aguilar introduced the concept of a similar ratio to analyze ILL borrowing requests. The “ratio of borrow- ings to holdings” compares the number of interlibrary loans relative to the holdings in a given subject area. John N. Ochola has used the methods described in Aguilar’s article to conduct a pilot project at Baylor University to determine which monographs should be moved to remote storage.13 He also applied Aguilar ’s four “rules,” which advocate specific actions for subjects with low or high ILL borrowings and low or high circulation. Bonn, Aguilar, Hamaker and Carrigan all discussed the evaluation of circulation and ILL data by subject area. As Carrigan stated, “Circulation data must be ana- lyzed at the macro level (across subject areas, not title by title) to bear significance on resource allocation and guide collec- tion development.”14 Methodology Data This project utilizes three distinct sets of data from the UCB libraries. In all three sets, data are limited to books only and analyzed across subject areas, rather than Use of Circulation Statistics and Interlibrary Loan Data 37 title by title. It is important that all three data sets employ the same definition of a book so that overlap can be meaningful rather than misleading. For instance, if one set includes government documents as “books,” but another does not, compar- ing the data based on subject, circulations, and requests will no longer be valid. (For the purposes of this project, a specific definition of “books” is discussed later in this article.) The three data sets are: • The entire collection of books owned by UCB • All the books that circulated at UCB from January 1998 through December 2002 • All the books that were requested via UCB’s ILL department between Janu- ary 1998 and December 2002 A number of items are excluded from the data sets. The UCB Law Library is a separate organization from the rest of the UCB libraries and therefore is not included in the study. The University of Colorado’s other campuses in Denver and Colorado Springs also are not included. This study is limited to English-language books; foreign-language materials com- prise nearly 25 percent of the total ILL requests, constituting a large enough data set for separate study. Government documents are excluded from this study because the extent of government docu- ments collections is generally determined by depository status, rather than active collection development decisions, making comparisons of circulation, holdings, and borrowing requests irrelevant. Manu- scripts, dissertations, and theses also are excluded from the data because many unpublished items are generally held by only one institution. In particular when considering ILL borrowing requests, it is misleading to include unique items such as dissertations in the source data to compare against the holdings of a library that would not reasonably be expected to acquire the item. Musical scores, although representing a respectable proportion of borrowing requests, also are excluded be- cause they are not language material and represent a different kind of user need and behavior than this study a�empts to analyze. Wherever possible, OCLC numbers were used to obtain Library of Congress Classifications (LCCs) for all items, for the purpose of subject area des- ignation. Items without OCLC numbers are classed as “unknown.” Definition of a Book Though it seemed on early consideration that a “book” was an obvious entity, it became clear that this study needed a very specific definition of a book. The operational definition of a book used for this research was as follows: a book is an English-language printed monograph, including large print and print reproduc- tion. Several criteria were used to exclude materials in any medium except paper: 1. Bibliographic Level is m, or mono- graph, in the MARC record and MARC 007 field is blank, a, b, or d. These require- ments excluded nonbooks such as serials, government documents, dissertations, theses, manuscripts, scores, and music. 2. Items do not have a $h field in the 245 MARC field and blank, d, or r, in the 008 field. These requirements excluded items in any medium except paper, such as microfilm, and any nonbook materials except large print and print reproduc- tions. 3. Items had a 260 $b or 020 field and a MARC 300 $a field with more than forty- nine pages. These requirements limited the data set to published materials of fi�y pages or more only, and eliminated all theses, dissertations, manuscripts, pam- phlets, and other unpublished works. Conspectus Subject Categories The data for this study were analyzed by subject classifications. The conspectus concept was developed by the Research Libraries Group, beginning in the 1980s. Conspectus subject categories are broad subject classifications based on the Li- brary of Congress classification system that provide a basis for large research li- braries to assess the subject distribution of 38 College & Research Libraries January 2006 their collections. The conspectus classifi- cation has twenty-five subject areas. An existing research collection, in this case the collection at UCB, can be mapped to these conspectus categories in order to provide a picture of the overall con- centrations in that collection. With only twenty-five subject categories, it is not sufficient to use the conspectus to evalu- ate a collection in detail. However, it is very useful to provide a broad picture of a collection. Table 1 lists all the conspectus categories. From this point forward, these subjects will be referred to as conspectus subject categories.15 Holdings Data The authors pulled the UCB holdings data from holdings in OCLC WorldCat in July 2003. Again, the set was limited to books as defined above. Several elements of the UCB collection are missing from this data set, most notably, the holdings of the special collections and archives departments, most of which have not been cataloged. This does not present a significant obstacle to this study for several reasons. It is unlikely that hold- ings of special collections and archives represent a significant overlap with ILL borrowing requests because of the rarity of the items hold. They are noncirculating collections, so they could not be present in the circulation data. In addition, a dearth of electronic records in special collections and archives is a problem mirrored by nearly all major university collections for several reasons, including long-term backlogs in cataloging departments.16 Therefore, the methods and results of this study, even without the special collections and archives records, still will be entirely relevant to other collections, indeed, even more so than if the two departments were included, given the similar lack of elec- tronic records in other institutions. An important element of these results is the inclusion, rather than the exclusion, of one particular kind of record: noncir- culating items. There is no code in OCLC WorldCat to identify noncirculating ma- terials. Identifying all possible reference materials and checking them with the local catalog is prohibitively laborious. As with most libraries, the noncirculat- ing books represent a small proportion of the total collection. The main reference collection and the reference collections of all the branches combined represent approximately two percent of the total items studied, an amount sufficiently small to ignore without skewing the data, analyses, or results. OCLC numbers were used to collect LCC numbers and map them to conspec- TABLE 1 Conspectus Subject Categories 1. Agriculture 2. Anthropology 3. Art and Architecture 4. Biological Sciences 5. Business and Economics 6. Chemistry 7. Computer Science 8. Education 9. Engineering and Technology 10. Geography and Earth Sciences 11. History and Auxiliary Sciences 12. Invalid or unknown 13. Language, Linguistics, and Literature 14. Law 15. Library Science, Generalities, and Reference 16. Mathematics 17. Medicine 18. Music 19. Performing Arts 20. Philosophy and Religion 21. Physical Education and Recreation 22. Physical Sciences 23. Political Science 24. Psychology 25. Sociology Use of Circulation Statistics and Interlibrary Loan Data 39 tus subject categories and to provide a picture of UCB’s holdings by subject. Total monographic holdings at UCB were 970,780 titles. Circulation Data The authors collected bibliographic infor- mation for every book, as defined above, that circulated at UCB from January 1998 through December 2002, the final com- plete year for which data were available at the beginning of this project. These dates coincide with available ILL borrowing data as described below. The authors col- lected the bibliographic record for every title (as opposed to every copy) that had one or more circulations in the time pe- riod defined, or two or more circulations if the most recent circulation was in 2003. (See figure 1 for a flowchart demonstrat- ing inclusion of circulation data.) For example, an item that met the criteria for a book but last circulated before 1998 would not be included. An item that met the criteria for a book and last circulated some time between 1998 and 2002 was included. An item that met the criteria but last circulated in 2003 was only included if it had at least two circulations because at least one circulation occurred outside the window of the study. OCLC numbers were used again to collect LCCs, which were mapped to conspectus subject cat- egories to get a picture of which subjects circulated, and in what proportions, at UCB during the period of study. T h e U C B c i r c u l a t i o n system does not track all action dates on every item; it collects only the most recent action date and the total number of circulations since the item was entered into the catalog. Table 2 demonstrates how circula- tion data were collected and calculated. Because the number of items resembling book X is likely to be very small, the authors elected to leave FIGURE 1 Circulation Data Qualifies as “book” Cataloged before 1/1/2003 Last circulated before 1998 = not included Last circulated 1998-2002 and 1 or more circs = included Last circulated after 2002 and 2 or more circs = included TABLE 2 Circulation Data Collection and Calculations Book –1997 (before study) 1998 –2002 (during study) 2003– (after study) Total Circulations in System Collection and Calculation Implications X 4 circs 0 circs 3 circs 7 circs Inadvertently included (has 2+ circs and last circ’d in 2003; impossible to distinguish 2003 and pre-1998 circs from circs in study period) Y 0 circs 12 circs 1 circ 13 circs Correctly included (demonstrates why books with 2+ circs and last circ’d in 2003 cannot be excluded) Z 2 circs 4 circs 0 circs 6 circs Correctly included (counts 2 circs too many, but system cannot distin- guish individual circ dates) 40 College & Research Libraries January 2006 them in the data to be analyzed. Circula- tion statistics have been collected by the UCB catalog starting in 1995, so any books such as book Z that were published before 1995 and circulated in our time period will have three years of extra circulation tallies. The total number of circulations was 1,638,740. Interlibrary Loan Data The authors collected ILL borrowing data for UCB from 1 January 1998 to 31 Decem- ber 2002. The starting date was controlled by implementation of the department’s ILL tracking so�ware, which provided ac- cess to the borrowing data, and the ending date was the last full year for which com- plete data were available at the time the project began. Data were included only for requests for books as defined above; cancelled requests were excluded from the data sample. OCLC numbers a�ached to the borrowing records were used to identify the LCCs for the requested titles and mapped to conspectus subject head- ings in order to analyze the subject disper- sion of the borrowing requests. During the period of study, borrowers were not required to provide their department af- filiation or academic status (faculty, staff, and so on), and very few volunteered the information. Records that included department and status accounted for less than ten percent of the total requests. This small amount does not provide sufficient data for meaningful analysis. Those fields are now mandatory at UCB to facilitate future study. It is important to mention that UCB is a member of a statewide consortium with a shared online catalog. The consortium is composed of several library systems in the state, representing the majority of the academic libraries in Colorado and Wyoming, and many of the largest public library systems in Colorado. It is via this unified catalog that a large amount of in- state ILL borrowing requests are handled. However, lending and borrowing data are not archived in the statewide catalog, making analysis of in-state borrowing requests impossible. Therefore, one can conclude that the ILL data collected for this particular study represent materials sufficiently rare that no other member of the consortium can provide a copy. The total number of ILL borrowing requests for the period of study was 22,064. Results and Discussion The data collected for this study yielded many valuable results that are too nu- merous to discuss in full in this paper. Based on previous studies reported in the literature, the elements selected for discussion are: • Overall holdings • Average transactions per item • Percentage of items circulated in a given subject collection • Ratio comparing ILL requests to holdings in a subject area Overall Holdings The monographs examined in the study were broken down into conspectus subject categories. (See table 3 for hold- ings figures.) The largest collections are disciplines with very high publishing output.17 In addition, literature and his- tory remain fields that rely very heavily on the research monograph, which may contribute to the large size of the collec- tions.18–20 In contrast, though UCB has a very strong chemistry program, the col- lection is more dominated by serials and electronic resources than monographs. UCB has an anthropology program. The small size of the collection is likely a result of the small monographic publish- ing output in the field of anthropology.21 There is no sport science program at UCB. Given the enormous popularity of various kinds of sports and physical exercise in the Boulder and Colorado community, it is likely that most of this collection sup- ports personal or popular use rather than academic discourse. In addition, a state- wide reciprocal borrowing policy, whose statistics could not be tracked at the level of the data used for this project, make it more likely that such popular subjects in Use of Circulation Statistics and Interlibrary Loan Data 41 the Colorado community would circulate via that borrowing policy. The collection in agriculture is the smallest of all subject collections at UCB. It is reasonable to ex- pect such a small collection because there is no agriculture program at UCB. Average Transactions per Item The average transactions per item rep- resents the average number of times each book circulated. This average represents the total number of circu- lations within a given subject area, divided by the number of circulated monographs in that subject area. For example, if 100 books in a given subject area circulated, and those same 100 books tallied a total of 600 circulations, the average circulation per title would be 6.0. This figure reveals the subject areas that receive particularly heavy or light use, regardless of collection size. With a few possible exceptions, only books that circulated during the period studied were included. However, the time frame for the circulations of these books could not be defined because of limitations in the circulation system. This study examined the total number of transactions from the origin of the circulation system in December of 1994. (See table 4.) High average transactions per item could demonstrate an extremely tar- geted collection or simply an extremely small collection in a popular area. Explanation for the high number of transactions per item will require fur- ther study. Though the undergraduate enrollment in computer science was not exceptionally high, it is likely that cir- culation was increased due to the broad applicability of computer science in many other degree programs, particu- larly in engineering, which has one of the highest undergraduate enrollment numbers in the university. Although the undergraduate enrollment in sociology for the period of this study was fewer than 600 students, the materials in a sociology collection have much broader applicability in all of the social sciences and in women’s studies, which may have contributed to the high circulation in this area. The art and architecture programs at UCB have very large undergraduate enrollments, which may partially account for such high circulation figures. TABLE 3 Number of UCB Holdings in WorldCat by Subject Rank Conspectus Subject Category Holdings 1 Language, Linguistics, and Literature 193,781 2 History and Auxiliary Sciences 126,797 3 Business and Economics 85,973 4 Engineering and Tech- nology 58,377 5 Philosophy and Religion 50,024 6 Art and Architecture 46,763 7 Sociology 43,437 8 Medicine 36,501 9 Political Science 35,764 10 Invalid or unknown 35,065 11 Library Science, Gener- alities, and Reference 33,327 12 Education 33,314 13 Physical Sciences 28,497 14 Biological Sciences 26,234 15 Mathematics 21,145 16 Music 18,476 17 Geography and Earth Sciences 16,550 18 Psychology 13,406 19 Computer Science 12,958 20 Performing Arts 11,921 21 Law 11,721 22 Chemistry 9,504 23 Anthropology 7,688 24 Physical Education and Recreation 7,067 25 Agriculture 6,490 42 College & Research Libraries January 2006 third largest monographic collection, so the small transaction average is a li�le surprising. This is most likely the result of the speed with which business materials become obsolete and the difficulty—and sometimes inadvisability—of weeding quickly enough to keep up with the pace of obsolescence in the field. As mentioned above, the law collection at UCB is sepa- rate, which almost certainly accounts for the low average transactions in law. The A low average could indicate various things. It could simply be the result of not having an academic program in the area, so there is less circulation in sup- port of research and study. It also could mean that a collection has increased in obsolescence and is in need of weeding. It is difficult to interpret the meaning of this average without further study of user behavior and collection decision making. Business and economics constitutes the TABLE 4 Average Number of Transactions per Item by Subject Rank Conspectus Subject Category Circulation Transactions Circulating Items Transactions per Item 1 Music 53,855 7,230 7.4 2 Computer Science 35,378 5,202 6.8 3 Sociology 106,724 17,809 6.0 4 Physical Education and Recreation 14,432 2,409 6.0 5 Art and Architecture 106,186 17,962 5.9 6 Anthropology 19,424 3,331 5.8 7 Psychology 31,281 5,376 5.8 8 Geography and Earth Sciences 31,473 5,552 5.7 9 Engineering and Technology 108,834 19,712 5.5 10 Mathematics 45,037 8,247 5.5 11 Performing Arts 25,106 4,661 5.4 12 Physical Sciences 52,423 9,786 5.4 13 Medicine 73,555 13,846 5.3 14 Agriculture 11,109 2,145 5.2 15 Biological Sciences 40,308 7,961 5.1 16 Language, Linguistics, and Literature 280,667 56,631 5.0 17 History and Auxiliary Sciences 230,262 46,515 5.0 18 Philosophy and Religion 91,324 18,696 4.9 19 Chemistry 13,011 2,775 4.7 20 Business and Economics 102,587 23,027 4.5 21 Political Science 52,108 11,745 4.4 22 Law 15,929 3,638 4.4 23 Invalid or unknown 44,707 10,419 4.3 24 Education 37,425 8,870 4.2 25 Library Science, Generalities, and Reference 15,595 4,972 3.1 Use of Circulation Statistics and Interlibrary Loan Data 43 lower average transactions in education are likely the result of the separate depart- mental education collection. Education re- searchers and students use that collection o�en, and that use may result in fewer circulations from the main library col- lection. The lowest average transactions were in library science, generalities, and reference. In this subject, the low average could simply be an indicator that noncir- culating reference materials are clustered in this subject. Another potential reason for such a low average could be that many sources are used in-house by librarians and not checked out even when they are used. In addition, many reference mate- rials are clustered into this subject, and even if they are circulating titles, patrons may find it easier to copy or note down the information rather than check out the book for detailed study. Percentage of Items Circulated The percentage of items circulated rep- TABLE 5 Percentage of Collection Circulated by Subject Rank Conspectus Subject Categories Holdings Circulating Items % Items Circulated 1 Anthropology 7,688 3,331 43.3% 2 Sociology 43,437 17,809 41.0% 3 Computer Science 12,958 5,202 40.1% 4 Psychology 13,406 5,376 40.1% 5 Music 18,476 7,230 39.1% 6 Performing Arts 11,921 4,661 39.1% 7 Mathematics 21,145 8,247 39.0% 8 Art and Architecture 46,763 17,962 38.4% 9 Medicine 36,501 13,846 37.9% 10 Philosophy and Religion 50,024 18,696 37.4% 11 History and Auxiliary Sciences 126,797 46,515 36.7% 12 Physical Sciences 28,497 9,786 34.3% 13 Physical Education and Recreation 7,067 2,409 34.1% 14 Engineering and Technology 58,377 19,712 33.8% 15 Geography and Earth Sciences 16,550 5,552 33.5% 16 Agriculture 6,490 2,145 33.1% 17 Political Science 35,764 11,745 32.8% 18 Law 11,721 3,638 31.0% 19 Biological Sciences 26,234 7,961 30.3% 20 Invalid or unknown 35,065 10,419 29.7% 21 Language, Linguistics, and Literature 193,781 56,631 29.2% 22 Chemistry 9,504 2,775 29.2% 23 Business and Economics 85,973 23,027 26.8% 24 Education 33,314 8,870 26.6% 25 Library Science, Generalities, and Reference 33,327 4,972 14.9% 44 College & Research Libraries January 2006 Chemistry also had a generally low circu- lation, which could be partially explained by the extreme dominance of serials in library expenditures toward chemistry at UCB and the low monographic publish- ing output in the field of chemistry.22 Ratio of Holdings to Interlibrary Loan It was difficult to create a measurement of ILL activity that allows for reasonable comparison with holdings and circulation data. Actual numbers of ILL requests were helpful but needed to be considered in relation to the size of the library’s owned collection because ILL activity in some subjects, such as literature and history, was bound to be inflated by very high publishing output. The authors have expressed ILL activity as a ratio to the UCB library system’s overall holdings. This number compared unowned titles with owned titles, essentially comparing apples to oranges. However, the number does give an impression of how ILL use compares to the library system’s holdings. The nearer the ratio approaches 1:1, the more borrowing occurs in that area when compared to the holdings. For example, if the library owned 200 titles in a particu- lar subject and borrowed 100 other titles in that subject, the ratio would be 2:1, indicating that for every book borrowed, the library owns two books in the same subject. Or, for example, if the holdings to ILL ratio is 14.9:1, that indicates that for every book borrowed in that subject, the library owns approximately fi�een titles in the same subject. There was a broad range of ratios among the conspectus subject categories. Table 6 shows the hold- ings, numbers of ILL requests, and ratios for each subject. Borrowing in the agriculture collection was quite small, which was logical given the absence of an agriculture program at UCB. However, despite the fact that the literature collection at UCB had one of the highest actual numbers of borrowing requests in the entire collection, the sheer magnitude of the literature collection made the borrowing ratio one of the low- resents the portion of the total collection in that subject that circulated in the five- year window of this study. For example, a measurement of 50 percent indicates that exactly half the monographs in the subject circulated during the study. This percent- age did not reflect the popularity of any one title. Each title in this category was unique; thus, an item that circulated 100 times would be counted only once for this calculation, allowing a measurement of use that was not inflated by any particular popular title. In addition, multiple copies were counted only once, so this was an indication of circulating titles measured against held titles; any title with many owned copies would not falsely inflate the measurement of collection use. This percentage could represent a number of qualities of the collection, including the possibility that a highly circulating col- lection was very targeted, or very broadly applicable, or aggressively weeded. It was impossible to identify the exact cause of these calculations without further study. Overall, it is unlikely that usage of online materials had any measurable effect on these numbers, either high or low, because this study was limited to monographs and the UCB collection holds only 15,000 electronic books, a mere 1.5 percent of the total monograph collection in the library. Overall, 33 percent of the UCB collection circulated one or more times between January 1998 and December 2002. (See table 5 for information on holdings, items circulated, and the percentage of items circulated.) It was noticeable that the anthropology collection was substantially smaller than other top circulating collections. Psychol- ogy and sociology also were extremely popular. It is likely that the statewide reciprocal borrowing policy inflates the circulation of such subjects that are popu- lar in the greater Colorado community. Actual circulation numbers of most English and American literature were generally high, despite the fact that this conspectus subject category overall had one of the lowest circulation percentages. Use of Circulation Statistics and Interlibrary Loan Data 45 est. Perhaps the most important factor af- fecting borrowing in language, linguistics, and literature was the existence at UCB of a statewide consortial borrowing program whose ILL numbers were not stored and therefore not included in the data for this study. As a result, borrowing data con- sidered in this study were only for items not held by any other library in the state. It was likely that this factor significantly reduced the borrowing numbers in fiction and literature. The lowest borrowing ratio in the overall collection, with 144.3:1, was in library science, generalities, and refer- ence. This ratio indicates that the library owns approximately 144 books for every title borrowed in the subject. Data Summary Though this study resulted in far more data than can be thoroughly described here, some overall statements can be exemplified by the data. There was li�le overlap between the subjects with large collections and the subjects with high percentages of materials circulated. Large TABLE 6 Holdings : ILL Ratio by Subject Rank Conspectus Subject Category Holdings ILL Items Holdings : ILL Ratio 1 Agriculture 6,490 713 9.1 : 1 2 Medicine 36,501 2,496 14.6 : 1 3 Physical Education and Recreation 7,067 387 18.3 : 1 4 Engineering and Technology 58,377 2,334 25.0 : 1 5 Invalid or unknown 35,065 1,353 25.9 : 1 6 Sociology 43,437 1,673 26.0 : 1 7 Music 18,476 654 28.3 : 1 8 Psychology 13,406 392 34.2 : 1 9 Anthropology 7,688 218 35.3 : 1 10 Biological Sciences 26,234 719 36.5 : 1 11 Law 11,721 307 38.2 : 1 12 Chemistry 9,504 246 38.6 : 1 13 Geography and Earth Sciences 16,550 377 43.9 : 1 14 Philosophy and Religion 50,024 1,086 46.1 : 1 15 Computer Science 12,958 279 46.4 : 1 16 Performing Arts 11,921 222 53.7 : 1 17 History and Auxiliary Sciences 126,797 2,343 54.1 : 1 18 Art and Architecture 46,763 728 64.2 : 1 19 Mathematics 21,145 327 64.7 : 1 20 Business and Economics 85,973 1,210 71.1 : 1 21 Physical Sciences 28,497 390 73.1 : 1 22 Education 33,314 447 74.5 : 1 23 Political Science 35,764 457 78.3 : 1 24 Language, Linguistics, and Literature 193,781 2,475 78.3 : 1 25 Library Science, Generalities, and Reference 33,327 231 144.3 : 1 46 College & Research Libraries January 2006 collections, not surprisingly, were always in subjects with very high publishing outputs. Some subjects had high percent- ages of circulation, indicating an active collection. However, these subjects did not necessarily have high transactions per item, indicating that collections in which many of the books circulate are not necessarily collections in which those books circulate o�en. Only a few subjects appeared in the top or bo�om five subjects in most of the four analysis areas of overall holdings, average transactions per item, percentage of items circulated, and ratio of holdings to interlibrary loan. Those subjects were sociology, physical education and rec- reation, education, and library science, generalities, and reference. Although it was not among the largest or small- est collections, sociology had very high transactions per item, percentage of the collection circulated, and ratio of holdings to interlibrary loan. This indicates that sociology is an extremely active collection in all ways. This is possibly due to the fact that sociology books are relevant to many subjects outside sociology. Physical education and recreation was one of the smallest collections in the UCB library, probably partly due to the fact that there is no sport science program at UCB. Despite this small collection, however, physical education had one of the highest transactions per item, showing that this collection circulates frequently. Physical education also had among the highest borrowing ratios, showing that it is a very popular subject despite the lack of a related university program. Education was not among the largest or smallest subjects and appeared in the bottom of every other list. Education had few transactions per item, a very low percent- age of the collection circulated, and few borrowing requests. Education is a subject that is not extremely active in the UCB libraries because a separate collection exists in the education department. Li- brary science, generalities, and reference also did not appear among the largest or smallest collections, but like education, appeared at the bo�om of every other area of analysis, with a low average of transactions per item, low percentage of the collection circulated, and a very low borrowing ratio. This is likely the result of the fact that noncirculating materials are clustered in this area and that most users of these materials would use them in the library without checking them out. Perhaps the overall most important fac- tor demonstrated by this study is the im- portance of combining different sources of data for collection development decisions. Simply looking at collection size, circula- tion data, or ILL data creates a limited, and usually not entirely accurate, picture of a subject collection. Librarians must use multiple sources of data to create a truly useful picture of their collections. At UCB, these data are currently being used to as- sist in an off-site storage transfer project. For this project, the data were mapped to conspectus subject headings. In addition, the data were mapped to North American Title Count (NATC) subjects. The NATC divides the LC classification system into 700 categories, rather than 26. With 700 categories, the NATC is not useful for providing a broad analysis of a collection. However, since it is so minute in its subject areas, the NATC can provide an extremely detailed subject analysis of a collection. Librarians have used these combined data to single out areas of li�le and heavy use, defined by the combined circulation and ILL data. This allows the librarian to weed more heavily in some areas and less in others. The result is a collection that continues to meet the strongest need with the on-site collection. Limitations of the Study Although the results of this study have been beneficial to both the UCB libraries collection development and assessment processes, and the method can be applied to other academic and research libraries as a collection assessment model, there are limitations to the study methodology. One of the major limitations is the lack of Use of Circulation Statistics and Interlibrary Loan Data 47 time that practitioners are able to dedicate to data gathering and the limited statisti- cal expertise available for data analysis. Without the assistance of the research scientists at the OCLC Office of Research, the UCB libraries would not have been able to analyze their WorldCat holdings in order to compare the circulation statistics and ILL borrowing data by subject. This study does not include statistics of materials that are used in the library, but not checked out, because the UCB libraries do not have the resources to collect these data. The circulation sta- tistics and ILL borrowing data analyzed reflect only a five-year period. This is a small percentage of the many years of circulation statistics and ILL borrowing data that have been accumulated by the UCB libraries. Since the inception of this study, the UCB librarians have decided to capture the date the catalog record was created to track the acquisition process and to analyze this information with the circu- lation statistics and ILL borrowing data on a quarterly basis. As discussed in the methodology, the UCB libraries did not capture annual circulation data. After the data from this study were shared with the UCB librarians, a decision was made to reconfigure the integrated library system (ILS) tracking statistics to capture the number of circulations/item/year beginning in January 2004. In January of 2005, these data were captured for use in the off-site storage transfer project. In addition, discussions are under way to configure the ILS to collect circulation statistics by subject areas identified by the bibliographers. The identification of ILL borrowers by discipline and status was the original intent of the study. Because ILL request forms did not require discipline and status, less than ten percent of the ILL requests in the study period included this information; therefore, there was insufficient data to include in this study. As a result of this, the ILL request forms have been revised to require discipline and status. Although it is o�en difficult to transi- tion research into practice and to change current decision-making processes, the UCB librarians have implemented changes to the libraries’ data collection methods based on the methodologies and results of this study. Future Research and Conclusion In addition to the above-mentioned changes made to the UCB libraries’ data collection methods, the results of this re- search have provided empirical data for collection management and remote storage decision making at UCB. The results also could be utilized for the development of qualitative assessments, such as interviews with faculty, as suggested by Sonia Bodi and Katie Maier-O’Shea.23 The circulation statistics and ILL borrowing requests could be used to calculate the obsolescence of a book based on its publication date and the increase or decrease of the number of circulations during a five-year period. An analysis of the publishers of the items circulated and requested through ILL may provide librarians sufficient information to make collection manage- ment and remote storage decisions in a programmatic way. This could be accom- plished by identifying the highly circulat- ed and requested titles by publisher and subject. These could be compared to the titles owned by the library to determine which titles would be likely candidates for remote storage. Identifying the languages, other than English, and subjects with high circula- tion and ILL borrowing requests also would support collection management and remote storage decisions. Although the criteria used for the acquisition and the methods used to identify and retrieve journal articles differ from those for books, the identification of journal titles and sub- ject areas that are accessed through the library and requested through ILL would provide empirical data for the selection of database and online services. The UCB librarians will gather and analyze the circulation statistics and the 48 College & Research Libraries January 2006 ILL borrowing data on an annual basis and present the analyses and findings to the subject bibliographers. By tracking the date the catalog record was created, it will be possible to determine how soon a�er cataloging a title is circulated. The analysis of circulation and ILL borrowing data by subjects, as suggested by Hamaker, and borrower discipline and status will help the librarians make informed collection deci- sions based on users’ borrowing pa�erns. It is essential that collection develop- ment decisions include the consideration of data such as those gathered for this study. Usage data are even more im- portant in the light of remote storage facilities and the attendant collection storage decisions that have been adopted by many U.S. libraries. Although the lit- erature includes many studies utilizing circulation statistics or ILL data, very few published studies investigate the intersection of circulation, interlibrary loan, and holdings, an intersection that can provide very fruitful guidance for subject librarians. Notes 1. Sonia Bodi and Katie Maier-O’Shea, “The Library of Babel: Making Sense of Collection Management in a Postmodern World,” Journal of Academic Librarianship 31, no. 2 (2005): 143–50. 2. Chuck Hamaker, “Time Series Circulation Data for Collection Development or: You Can’t Intuit That,” Library Acquisitions: Practice and Theory 19, no. 2 (1995): 191–95. 3. Charles B. Osburn, “Collection Evaluation and Acquisitions Budgets: A Kaleidoscope in the Making,” Journal of Library Administration 17, no. 2 (1992): 3–11. 4. Dennis P. Carrigan, “Data-guided Collection Development: A Promise Unfulfilled,” College & Research Libraries 57 (1996): 429–37. 5. Richard L. Trueswell, “Some Behavioral Pa�erns of Library Users: The 80/20 Rule,” Wilson Library Bulletin 43, no. 5 (1969): 458–61. 6. Barbara Hoffert, “Book Report, Part 2: What Academic Libraries Buy and How Much They Spend,” Library Journal 123, no. 14 (1998): 144–46. 7. Chuck Hamaker, “Time Series Circulation Data for Collection Development.” 8. Justin Li�man and Lynn Silipigni Connaway, “A Circulation Analysis of Print Books and E-books in an Academic Research Library,” Library Resources and Technical Services 48, no. 4 (2004): 256–62. 9. Mike Day and Don Revill, “Towards the Active Collection: The Use of Circulation Analyses in Collection Evaluation,” Journal of Librarianship and Information Science 27, no. 3 (1995): 149–57. 10. George S. Bonn, “Evaluation of the Collection,” Library Trends 22, no. 3 (1974): 265–304. 11. William Aguilar, “The Application of Relative Use and Interlibrary Loan in Collection Development,” Collection Management 8 (1986): 15–24. 12. Terry R. Mills, The University of Illinois Film Center Collection Use Study (Urbana: Univ. of Illinois, 1981). 13. John N. Ochola, “Use of Circulation Statistics and Interlibrary Loan Data in Collection Management,” Collection Management 27 ( 2002): 1–13. 14. Carrigan, “Data-guided Collection Development,” 432. 15. Research Libraries Group, “A Brief History of the RLG Conspectus,” Web page. Available at h�p://www.rlg.org/conspechist.html. [Accessed 23 July 2004]. 16. Anna H. Perrault, “Library Groupings.” Global Collective Resources: A Study of Monographic Bibliographic Records in WorldCat, Report of a study conducted under the auspices of an OCLC/ ALISE 2001 Research Grant. Available at h�p://www.oclc.org/research/grants/reports/perrault/intro.pdf. 17. Dave Bogart, ed., The Bowker Annual Library and Book Trade Almanac, 48th ed. (Medford N.J.: Information Today, Inc., 2003), 495. 18. Charlene Kellsey and Jennifer E. Knievel, “Global English in the Humanities? A Longitudi- nal Citation Study of Foreign Language Use by Humanities Scholars,” College & Research Libraries 65 (2004): 194–204. 19. Jennifer Wolfe Thompson, “The Death of the Scholarly Monograph in the Humanities? Citation Pa�erns in Literary Scholarship,” Libri 52 (2002): 121–36. 20. John Cullars, “Characteristics of the Monographic Literature of British and American Literary Studies,” College & Research Libraries 46 (1985): 511–22. 21. Bogart, Bowker Annual Library and Book Trade Almanac, 496. 22. Ibid., 496. 23. Bodi and Maier-O’Shea, “The Library of Babel.” Use of Circulation Statistics and Interlibrary Loan Data 49