College and Research Libraries BARBARA MOORE, TAMARA}. MILLER, AND DON L. TOLLIVER Title Overlap: A Study of Duplication in the University of Wisconsin .,_ System Libraries T It has been commonly believed by many in the profession as well as many of those who fund library programs that collection duplication is exceedingly high across comparable libraries that support similar user needs. Devising efficient ways of assessing duplication , or overlap, among libraries was diffi- cult prior to the availability of archival tapes from bibliographic utilities. Most collection overlap studies were conducted drawing samples from card catalogs or order slips. This paper outlines the results of a study of collection overlap for the University of Wisconsin System libraries. OCLC archival tapes covering the time period July 1977-]une 1979 provided two years of cataloging data. More than 392,000 monograph records created by the cataloging activities of eleven libraries were compared to determine duplication rates. Findings based on analyses of these cataloging records are discussed. EcoNOMIC coNDITIONS and inflation have created an environment of increasing costs and dwindling budgets, forcing university and library administrations to rethink library program priorities. Today, few, if any, uni- versity libraries enjoy a real growth in pur- chasing power Inflation is affecting not only the actual costs of books but also the dollars required to handle, process, and house mate- rials. A sense of alarm has developed and uni- versities have taken a variety of steps to meet the challenge. One strategy for lessening the impact of inflation on libraries is to reduce duplicate Barbara Moore is systems librarian , Library and Learning Resources, University of Wisconsin- Whitewater. Tamara ]. Miller is head, Library Systems, University Libraries, University of Tennessee-Knoxville. Don L. Tolliver is assistant director for planning and budget, University Li- braries, University of Michigan , Ann Arbor. The authors would like to thank Atlee Svanoe, Univer- sity of Wisconsin- Whitewater, who spent many long hours. writing and running the computer pro- grams. 14 I book purchases. This policy can be adminis- tered rather easily in single collection li- braries. However, such a policy is somewhat more difficu t to implement across a number of libraries, Nonetheless, libraries are now examining c operative collection develop- ment agreements. Cooperative collection de- velopment allows two or more libraries to establish purchasing agreements that reduce duplication in specific areas. Under such a plan, some materials are purchased and housed in only one library and are made available to other member libraries") The assumptions supporting this approach are based on the belief that collection dupli- cation is exceedingly high from library to li- brary. Thus by monitoring and, when feasi- ble, controlling duplicate book purchases, the impact of budgetary downturns and in- flation may be lessened. Of course, it also has been argued that some unspecified level of duplication is necessary to provide accept- able library service at each university loca- tion. The University of Wisconsin (UW) System created a Library Planning Study Committee _- (LPSC) in 1978 to undertake a comprehen- sive study of the libraries in the UW System. S<:>me of the issues addressed by the LPSC were the need for adequate housing for grow- ing library collections; the potential for coop- - erative library activities; the feasibility of co- ordinated acquisitions; and the need for resource sharing. In order to fully address these issues, the LPSC required detailed in- formation about the current overlap in ac- , •· quisitions of the UW System libraries. Much of the work of this overlap study was spon- sored by the LPSC. LITERATURE REVIEW Collection overlap studies among libraries have largely been conducted through sam- . piing techniques. Information has been gath- ered from card catalogs, order slips, or archi- - val tapes provided by a bibliographic utility . .. Nhile there have been several studies of title overlap, only Nugent has assessed both cur- -rent acquisitions and total holdings. He found a 40 percent title overlap across the holdings of six New England state university libraries. This figure increased to 47 percent when only current imprints were examined. 1 Cooper, Thompson, and Weeks investi- gated the degree of title overlap among nine University of California campuses. Their study found that 25 percent of the titles held at Berkeley were duplicated on at least one of three other northern campuses, and 44.9 per- cent of the University of California at Los Angeles titles were duplicated on at least one of four other southern campuses. 2 In a study of Canadian addictions li- braries, Dingle-Cliff and Davis noted that 76 percent of the seventy-one titles with im- prints of 1969 or later were found in at least two libraries. 3 Altman found a 52 percent ti- tle overlap among thirty-one school li- braries. 4 She stated, "This finding refutes the assumption that school library collections are basically similar. "5 By sampling orders placed from libraries in the London University System, Urquhart and Schofield found an overlap of 15 per- cent. 6 Parker also examined the amount of overlap in acquisitions of five member li- braries of the Consortium of Universities of Metropolitan Washington. He reported that 61 percent of the orders were duplicated with Title Overlap I 15 a rate of 2.24 copies per title. 7 All of the above studies relied on samples drawn from card catalogs or order records. Evans, Gifford, and Franz reported that the State University of New York (SUNY) Cen- tral Administration Office of Library Ser- vices used the machine-readable biblio- graphic records (OCLC magnetic tapes) of four member libraries to assess title overlap. This study was part of an overall analysis of OCLC tapes completed to determine collec- tion development practices. Unique OCLC record numbers were compared for a twelve- week cataloging period. It was found that 86.7 percent of 25,622 titles on the tapes were owned byonlyonelibrary. Only 13.3 percent of the titles were held by more than one library. 8 UNIVERSITY OF WISCONSIN SYSTEMS LIBRARIES The University of Wisconsin System has approximately 150,000 students enrolled at thirteen degree-granting universities and fourteen freshman-sophomore centers lo- cated throughout the state of Wisconsin. All but two of the thirteen libraries at degree- granting t,miversities were using the OCLC system as a means of cataloging in 1979. The cataloging transactions for these eleven li- braries were routinely recorded on magnetic tape and maintained by the Wisconsin Li- brary Consortium. This wealth of data pro- vided an excellent opportunity for an ex- panded overlap analysis similar to the SUNY study. All eleven of the UW libraries included in the study serve undergraduate students. In addition, most offer graduate programs, with UW-Madison and UW-Milwaukee pro- viding extensive doctoral programs. Al- though the nine nondoctoral universities of- fer similar core course programs, many have focused curriculum specialties, e.g., busi- ness, applied technologies, agriculture, home economics, etc. The magnetic tape cataloging data was complete from the date each library began using OCLC. However, four of the libraries studied were not online during portions of the two-year period studied. Table 1 details the total number of transactions, number of ti- tles, and the dates covered for each of the UW libraries included in the study. 16 I College & Research Libraries • January 1982 TABLE 1 UNIVERSITIES AND TITLES INCLUDED IN TITLE OvERLAP STUDY University of Wisconsin Campus Eau Claire *Green Bay *LaCrosse tMadison Milwaukee Oshkosh *Parkside *Platteville *River Falls Stout Whitewater Total records Total titles Monograph Records on OCLC Tapes 7/l/77 -6/30/79 32,764 57,346 11,380 129,732 64,986 26,393 11,768 9,016 7,597 20,701 23,333 392,016 Records Used after Collapse Regardless of Imprint 30,639 54,648 11,004 113,740 57,758 24,875 10,769 8.824 7,145 18,441 19,710 357,553 267,979 Records Used after Collapse with Imprints 1976-79 20,252 9,407 7,089 71,122 36,313 19,607 7,315 4,061 4,355 11,195 13,843 204,559 124,774 •Online status : Green Bay: retrospective conversion project under way; LaCrosse: online Dec. 1977- ; Parkside: identifiable retrospective records eliminated prior to collapse programs; Platteville: online Feb. 1978- ; River FalJs: online June 1978- . tincludes the Wisconsin State Historical Society from Dec . 1977- ; Instructional Materials Center, Health Sciences and Law Libraries. METHODS OCLC archival tapes for the UW System for the time period July 1, 1977, through June 30, 1979, were used. Two years of cataloging data were examined to minimize the effect of purchasing and/or cataloging delays. The twelve-week time period used in the SUNY study was viewed as too short, as the authors of the SUNY report had warned: "The pres- ence of 'unique' records (those held by exactly one institution) is suspect because it is unusu- ally high and similar for all institutions. It is due to the short time period analyzed, the difference in cataloging practices at the four institutions, and the relative timeliness of the different selection and acquisition pro- cesses."9 The archival tapes contain a full biblio- graphic description, OCLC unique control number, and a library identification code for each cataloging transaction (produce, up- date, replace, and cancel) completed by each library. The tapes do not contain uniform data for each library. Local practice for cata- loging various types of materials and special projects (retrospective conversion of old cata- loging and reclassification) are reflected on the OCLC tapes. Thus, all identifiable retro- spective conversion records were excluded. Also, serial records were excluded from the study since several UW System libraries do not catalog these materials. The UW System study differs from most of the previously cited studies in that govern- ment documents, musical scores, and audio- visual and instructional materials were in- cluded. While the inclusion of serials would likely effect the results, it was felt that the large sample size would reduce the impact of these nonstandard items. Indeed, only 2 per- cent of the titles analyzed were audiovisual or instructional materials. All titles, regardless of imprint date, were used for the first analysis. A second analysis was done using a smaller group of titles with publication dates from 1976 to 1979. The ex- amination of current imprints diminished the possibility of artificially high rates of unique titles due to cataloging backlogs, unidenti- fied retrospective cataloging, reclassifica- tion, and time lags in purchasing materials. While the analysis of current imprints pro- vided a better indication of the overlap in current a~quisitions among the eleven UW libraries, it ignored those materials that were currently purchased for special, rare book, and other collections made up of older im- prints. Thus, the results of the two analyses represent two extremes of overlap. The following items were extracted from each archival tape record: the unique OCLC control number, the three-character library identification code, and the first two letters of the LC classification number (discipline code). All eleven libraries use the Library of Congress classification scheme. The disci- pline code was taken from either the local LC call number (090) or national LC call num- ber (050) field. Of the total records processed, 6.1 percent did not contain an LC call num- ber and were excluded only for the analysis by discipline. These twelve-byte records were sorted by OCLC control number and by library loca- tion within OCLC control number. A col- lapse program was run to eliminate all but the most recent occurrence of a record with the same OCLC control number and library location code. The collapse program com- bined all UW-Madison locations, including the Wisconsin State Historical Society, thereby eliminating the duplicates found among these campus libraries. Further, the collapse program eliminated all duplicate records created by a single library. These rec- ords may represent multiple copies, correc- tions to previous cataloging, or cancel trans- actions (to withdraw a title). The results of the collapse are found in table 1. Title overlap frequency programs were then run to compare the numbers of remain- ing location codes for each title. Five catego- ries of overlap comparisons were completed: (1) among all eleven university libraries; (2) between the two university libraries serving doctoral and nondoctoral universities; (3) among the nine university libraries serving nondoctoral institutions; (4) between univer- sity libraries serving doctoral and nondoc- toral universities; and (5) among all univer- sity libraries by four disciplines. All comparisons were done for all titles re- gardless of imprint date. Only the first three Title Overlap I 17 comparisons were repeated for titles having an imprint date of 1976-1979. FINDINGS Comparison of the 267,979 records created by the cataloging activity of the eleven UW libraries from July 1977 through June 1979 revealed that only 18.16percent of these titles had two or more location codes and 1. 05 per- cent had six or more location codes. The per- centage of titles overlapped increased to 31.99 percent for two or more locations and to 2.24 percent for six or more locations when only current imprints were compared. Table 2 details the overlap among titles for both comparisons. The actual overlap in currently cataloged materials falls between these two figures, 18.16 percent and 31.99 percent, for titles held in two or more locations. The actual number of titles with six or more locations differed by only eleven titles between the two comparisons. The low overlap among all ti- tles cataloged confirms the SUNY findings. 10 These findings indicate that the level of ti- tle duplication among all the UW libraries was much lower than previously assumed. Since this study relates to currently cataloged material, it does not necessarily serve as an indication of total collection overlap among the libraries studied. The data for each individual UW library compared with the rest of the UW System libraries studied are given in table 3. The per- cent of overlap was higher for any single li- brary than for the group as a whole. The larg- TABLE2 CoMPAmsoN AMONG ALL UNIVERSITIFS Titles Comparison of All Titles Held* By one univ . 219,327 By two univ. 28,863 By three univ. 9,132 By four univ. 5,012 By five univ. 2,833 By six or more univ. 2,812 Comparison of Current Imprints held (1976-79)l By one univ . 84,860 By two univ. 20,967 By three univ. 8 ,443 By four univ. 4,898 By five univ. 2,805 By six or more univ. 2,801 •Three titles were held in all eleven university libraries. tNo titles were held in all eleven university libraries . Percent 81.84 10.77 3.41 1.87 1.06 1.05 68.01 16.80 6.77 3.93 2.25 2.24 Cumulative Percent 92.61 96.02 97.89 98.95 100.00 84.81 91.58 95.51 97.76 100.00 18 I College & Research Libraries • January 1982 TABLE3 PERCENT OF UNIQUE TITLES IN EACH UNIVERSITY LIBRARY All Titles Cataloged Current Imprints University of Titles Percent Titles Percent Wisconsin Campus Cataloged Unique Cataloged Unique Eau Claire 30,639 45 .95 20,252 28.57 Green Bay* 54,648 80.03 9,407 22.42 LaCrosse 11 ;004 46.63 7,089 27.56 Madison 113,740 72.85 71,122 61.91 Milwaukee 57,758 57 .99 36,313 39.37 Oshkosh 24,875 34.30 19,607 22.62 Parkside 10,769 37.07 7,315 18.65 Platteville 8,824 53.32 4,061 21.79 River Falls 7,145 46.13 4,355 24.75 Stout 18,441 57.45 11,195 40.39 Whitewater 19,710 45.22 13,843 31.76 •The figures for the comparison of titles regardless of imprint date for UW-Green Bay included retrospective conversion records that were unable to be identified and eliminated. est library, UW-Madison, had the lowest overlap rate with 38.09 percent of its current imprints duplicated elsewhere in the UW System. The smaller libraries (14,000 titles cataloged or less) tended to have considerably larger overlap rates for current imprints, with an average of 73.24 percent. This con- firms Davis and Shaw who found that, "the larger the library from which the sample is drawn, the more likely it is to have material that is either esoteric or of only local inter- est."11 Since the two largest libraries studied are also the two libraries that serve doctoral- degree-granting universities, a series of com- parisons were run with the doctoral and non- doctoral groups. The overlap comparison between the two groups shows that 11.63 percent of all titles cataloged during the pe- riod were held by at least one library in each group. Table 4 presents these results. Comparisons were also done within each group. Between the doctoral universities (UW-Madison and UW-Milwaukee), 10.85 percent of all titles cataloged were dupli- cated. When only current imprints were con- TABLE4 CoMPAIUSON BETWEEN DoCTORAL AND NoNDOCTORAL UNIVERSITIES Titles held by doctoral Universities and by nondoctoral Titles Percent Universities (overlap) 31,170 11.63 Titles held by either doctoral Universities or by nondoctoral Universities (unique to either group) 236,809 88.37 sidered, the figure increased to 17.41 per- cent. Among the nine nondoctoral universities, 17.29 percent of all titles cata- loged were duplicated by at least two li- braries in the group. Current imprint overlap was 34.85 percent. (See tables 5A, 5B.) As expected, the overlap among the smaller, nondoctoral group was higher than the overlap between the larger university li- braries. This difference may be the result of the similarity in basic freshman and sopho- more courses offered at the nondoctoral uni- versities as well as the diversity of the aca- demic programs at the doctoral institutions. The results, however, may be simply a func- tion of current purchase volume. Comparisons were also done for all eleven libraries in four basic academic subject areas. Analysis was done on all titles cataloged dur- ing the period, eliminating only those records without LC call numbers. Discipline codes were assigned as given in table 6. The general literature material showed the most duplica- tion, with 24.13 percent of the titles held in two or more libraries, followed by science with 23.07 percent, social science with 19.89 percent and history with 17.08 percent dupli- cation. These four disciplines had similar lower levels of overlap than anticipated for basic subject areas. Thus, overlap was surprisingly low in all five comparisons for the UW System li- braries. In an effort to further understand the nature of these results, a limited analysis was done to determine the bibliographic charac- teristics of the high versus low overlapped ti- tles. A sample of 249 unique titles (0.3 per- cent of the unique titles) and a sample of 84 of Title Overlap I !9 TABLE5A CoMPARISON BETWEEN DocroRAL UNIVERSITIES Titles Held By both UW-Madison and UW-Milwaukee By only one doctoral university Total All Titles Cataloged Titles Percent 16,788 137,922 154,710 10.85 89.15 100.00 TABLE5B Current Imprints Titles Percent 15,933 75,569 91,502 17.41 83 .59 100.00 COMPARISON AMONG NINE NONDOCTORAL UNIVERSITIES All Titles Cataloged Cumulative Current Imprints Titles Held Titles Percent Percent Titles Percent Byoneuniv. 119,463 82.71 39,071 65 .15 By two univ. 15,335 10.62 93.33 11,588 19.32 By three univ. 5,413 3.75 97.08 5,127 8.55 By four univ. 2,413 1.67 98.75 2,381 3.97 By five univ. 1,118 0.77 99.52 1,110 1.85 By six or more univ. 697 0.48 100.00 694 1.16 Total 144,439 59,971 TABLE6 CoMPARISON AMONG ALL UNIVERSITIES IN FouR DISCIPLINE AREAs Discipline Area Titles Percent History (D , E and F) Titles Held In one univ. 22,986 82.92 In two univ. 2,567 9.26 In three univ. 890 3.21 In four univ. 559 2.02 In five univ. 353 1.27 In six or more univ. 365 i.32 Total 27 ,720 Social Science (H) 'fitles Held In one univ. 39,088 80.11 In two univ. 5,437 11.14 In three univ. 1,871 3.83 In four univ. 1,030 2.11 In five univ. 661 1.35 In six or more univ. 706 1.45 Total 48,793 Science (Q) Title8 Held In one univ. 14,210 76.93 In two univ. 2,513 13.61 In three univ. 931 5.04 In four univ. 439 2.38 In five univ. 188 1.02 In six or more univ. 190 1.03 Total 18,471 General Literature (PN) Titles Held In one univ. 3,725 75.87 In two univ. 614 12.51 In three univ. 245 4.99 In four univ . 160 3.26 In five univ. 73 1.49 In six or more univ. 93 1.89 Total 4,910 Cumulative Percent 84.47 93.02 96.99 98.84 100.00 Cumulative Percent 92.18 95.39 97.41 98.68 100.00 91.25 95.08 97.19 98 .54 99.99 90.54 95.58 97 .96 98.98 100.01 88.38 93.37 96.63 98.12 100 .01 20 I College & Research Libraries • January 1982 the highly overlapped titles (3.0 percent of the current imprints held by six or more li- braries) was drawn for the comparison across all institutions. Full OCLC bibliographic records, not locally edited versions, were ex- amined for each title. Each of these records was analyzed by seven factors: (1) format of the material, (2) language of publication or performance, (3) subject as determined by LC call number, (4) type of material content, (5) publisher, (6) series, and (7) source of original cataloging. For language, publisher, and source of cat- aloging there were major differences be- tween the high and low overlap groups. The high overlap titles were all written in En- glish, 40 percent of the high overlap titles were published by university presses, and all but one of the high overlap titles was origi- nally cataloged by the Library of Congress. The uniquely held items included 22 per- cent written in foreign languages. University press publications accounted for only 7 per- cent of the unique titles. The source of cata- loging for the unique items was much more varied than for the high overlap group. Twenty-four percent, or 61 titles, was origi- nally cataloged by an OCLC member li- brary. Twenty-seven of these titles were orig- inally cataloged by the UW System library holding the item. These 27 titles were pri- marily local materials, e.g., theses, state and local documents, and locally published or produced audiovisual materials and mono- graphs. There were no large differences in the sub- ject area of the materials in the high and low overlapped titles. Both groups had a large number of titles in the areas of social science and general literature. Of the high overlap titles, 20 percent were in the social sciences and 13 percent in literature. This order was reversed for the unique titles with 16 percent in literature and 14 percent in the social sci- ences. Philosophy/psychology/religion was the third largest subject area with 11 percent of the high overlap titles. All other subjects had 8 percent or less of the titles for both samples. Table 7 details series and type of contents. No differences were readily apparent. Gov- ernment documents and biographies were the most common types of contents. It is difficult to assess the effect of material TABLE 7 BREAKDOWN BY SERIES AND TYPE OF CONTENTS FOR HIGH OVERLAP AND uNIQUE TITLES Series statement present Type of contents Government publications Textbooks Addresses, ·essays , lectures Conference publications Biography Fiction Dictionaries Juvenile Bibliographies Indexes Percent of High O verlap Titles 22.62 0 1.19 9.52 4.76 11.90 0 1.19 0 3.57 1.19 Percent of Unique Titles 35.34 13.25 2.41 4.42 3.21 4.42 5.62 0 1.61 2.01 0 format on overlap. The high overlap titles were all printed items. Three percent of the unique titles were nonprint materials. Since academic libraries do not acquire audiovi- sual materials in great numbers, they are more likely to be unique titles. The UW- Stout, which acquires a large amount of non- print material in applied technologies , had the second highest percent of unique mate- rials. SuMMARY This analysis of title overlap placed the levels of duplication among all eleven univer- sities at a lower level than previously as- sumed. The actual rate of duplication falls between 18 and 32 percent for the entire UW System. These overlap rates were determined by listing all titles studied and counting the ac- tual number of titles with two or more loca- tions. These findings validate the findings of the SUNY study, but differ substantially from many of the overlap studies cited earlier. A close reading of those studies show that, in some cases, the overlap figures given are ac- tually an average of the individual overlaps in several libraries and not a comprehensive count of overlapped titles. There were no clear trends in any of the four basic academic subject areas. However, the larger libraries and those smaller libraries with special subject emphasis seem to be col- lecting material that is highly unique. A co- operative acquisitions program built on exist- ing subject emphasis should result in larger collections of unique materials. The highly overlapped titles do not appear to be a core of needed reference materials. The materials most likely to be frequently du- plicated are English language materials pub- lished by university presses. These materials are standard items that seem to be routinely . cataloged by the Library of Congress. The purchase of these materials by six or more of the uw libraries may be an indication of both the pertinence of the material published by university presses and the reliance on these publishers by book selectors. University press titles are prime candidates for inclusion in a systematic cooperative acquisition program. Although the overlap found in this study was smaller than anticipated, the numbers of titles duplicated was still substantial. The 10.77 percent of all titles cataloged that were held in two separate locations represented 57,726 separate purchases of 28,863 titles: Title Overlap I 21 The 1.06 percent held by five UW libraries represented 14,165 purchases of 2,833 titles. Further analysis of the titles that were du- plicated would be helpful in developing co- operative acquisition policies. A more de- tailed study of the subject areas represented by high and low overlap titles , extending be- yond the basic disciplines, might reveal cur- rent trends in purchasing that could be used to serve as a basis for focusing collection- development responsibilities among the UW libraries. Additional study of existing collections, as well as material currently added to the col- lections, is needed to identify the core of ma- terial that should be present in each library. Identifying some level of desirable ~r re- quired duplication will clarify those collec- tion areas in which duplication can be sub- stantially reduced. IU:FERENCES s 3~9A l. William R. Nugent, "Statistics of Collection Overlap at the Libraries of the Six New En- gland State Universities, " Library Resources & Technical Services 12:31 (Winter 1968). 2. William S. Cooper, Donald D. Thompson, and Kenneth R. Weeks, "The Duplication of Monograph Holdings in the University of Cali- fornia Library System," Library Quarterly 45:262-63 Guly 1975). 3. Susan Dingle-Cliff and Charles H. Davis, "Collection Overlap in Canadian Addictions Libraries," Special Libraries 70:79 (Feb. 1979). 4. Ellen Altman, "Implications of Title Diversity and Collection Overlap for Interlibrary Loan among Secondary Schools," Library Quarterly 42:185 (April1972). 5. Ibid., p.185 . 6. John A. Urquhart and J. L. Schofield, "Over- lap of Acquisitions in the University of London Libraries: A Study and a Methodology," Jour- nal of Librarianship 4:37 Gan. 1972). 7. Ralph Parker, A Feasibility Study for a ]oint Computer Center for Five Washington , D.C. University Libraries; Final Report (Washing- ton, D.C.: Consortium of Universities of Met- ropolitan Washington, D.C., 1968), p.18 . 8. Glyn T. Evans, Roger Gifford, and Donald R. Franz, Collection Development Analysis Us- ing OCLC Archival Tapes; Final Report (U.S. Department of Health, Education and Wel- fare, Office of Education, 1977), p.19. 9. Ibid . , p.18. 10. Ibid., p.19. 11. Charles H. Davis and Debora Shaw, "Collec- tion Overlap as a Function of Library Size: A Comparison of American and Canadian Pub- lic Libraries," Journal of the American Society for Information Science 30:21 Gan. 1979). /