College and Research Libraries Collections at Risk: Revisiting Serial Cancellations in Academic Libraries Tina E. Chrzastowski and Karen A. Schmidt In 1993, Tina E. Chrzastowski and Karen A. Schmidt published a longi- tudinal study of serial cancellations from five midwestern academic re- search libraries. 1 The study found that actual title cancellation overlap was small (4.3%), meaning that many unique titles were being cancelled. A profile for an "at-risk" serial title, one most likely to be cancelled, was also determined. The titles most often cancelled were English-language titles in the sciences having a higher-than-average subscription price. The present study builds on the original research, adding two more years of cancellation data as well as serial order data. Results show that the rate of cancellations accelerated and that cancellation overlap increased by 55 percent in the two years following the original study. Science titles are still most likely to be cancelled, and less likely to be ordered. Also, the cost of serial cancellations in the sciences is not balanced by orders in the sciences, which are for fewer, lower-priced journals, in contrast with serials in the social sciences and humanities. Overlap in serial or- ders was less than expected. erial cancellations continue to plague our collections and re- move unique titles from li- ~lll!!!!!!iiii!!!!!!!!l!!~ brary shelves. This study documents a dramatic increase in the number of titles cancelled and the num- ber of overlapped serial cancellations be- tween FY91-92 when compared to the previous three years (FY88-90). The new research builds on a previous study con- ducted in 1992 and published in 1993.2 This study serves to continue to sound the alarm about diminishing serial collec- tions held in academic research libraries. Intuitively, it seems that little has changed in the way in which libraries have handled serial cancellations in re- cent years. The effects of the marketplace continue to play havoc with materials budgets; libraries know too little, too late in their budget years to manage many co- operative efforts to protect vulnerable serial titles; and the overall perception is that serial collections are getting leaner, monograph collections are suffering in relative silence, and the depth and rich- ness of library collecti<;ms are disappear- ing. The original research that provides the background for this study addressed the issue of diminishing serial collections and described the at-risk serial. It also discussed the need to analyze serial or- Tina E. Chrzastowski is Chemistry Librarian and Associate Professor of Library Administration , Univer- sity of Illinois at Urbana-Champaign ; e-mail: chrz@uxl.cso.uiuc.edu . Karen A. Schmidt is Acquisitions Librarian and Professor of Library Administration, University of illinois at Urbana-Champaign; e-mail: karens@uxl.cso.uiuc.edu. · · 351 352 College & Research Libraries ders that, regardless of cancellation projects, continue to be placed in research libraries and the desirability of, and prob- lems associated with, conducting a serial overlap study. The original research posed a number of questions that required further data collection and analysis. This study seeks to answer these questions: Did the low serial cancellation overlap and the over- all profile of serial cancellations found in the original research continue? Are high- priced science titles still the most vulner- able in our collections? Through the ad- ditional collection of data on serial orders, this study also attempts to answer ques- tions about what is being added to re- search collections that were also experi- encing cancellations. Which collection de- velopment trade-offs are being made? Do these trade-offs represent a discernable profile of collection subject trends? How is the overall complexion of serials col- lections changing? Hypotheses Based on the results of the first study and general perceptions about serial cancel- lations and orders, several predictions were made. These predictions relate to the different data sets collected from the five libraries. Those data sets included infor- mation on both cancellations and orders. The hypotheses for the serial cancellation and order databases are: • Cancellations will continue to reflect results July 1996 tions by subject, with a bias against the sciences. • Order overlap will be greater than cancellation overlap, meaning that librar- ies will tend to order the same titles and continue to cancel different titles. Population To compare recent data with data from the 1993 study, the same five ARL li- braries were selected as the sample group for the follow-up study. These libraries are Michigan State University, Ohio State University, the University of Illinois at Urbana-Champaign, the Uni- versity of Iowa, and the University of Wisconsin at Madison. Each of these li- braries is a member of the Committee on Institutional Cooperation (CIC), a consor- tium of midwestern academic institu- tions. The libraries of the CIC have a his- tory of shared cooperative ventures, as well as the common characteristic of own- ing and maintaining large, research-ori- ented collections in virtually every disci- pline. Table 1 shows FY91-92 serial expen- ditures for the five libraries taken from ARL statistics.3 Only two of the five li- braries experienced increased serial bud- gets (averaging only two percent), and three libraries' serial budgets decreased from FY91 to FY92 (an average of three percent). Among the five libraries, FY91-92 serial expenditures decreased an TABLEt from the original FY88-90 study, but will accelerate. • The cancellation over- lap (the same titles being cancelled by two or more li- braries) will increase. Total Expenditures for Serials, FY91-92, and Percent Change for the Five Libraries in the Study • There will be fewer serial orders than serial cancellations in both dollar amount and number of orders. • Orders by subject will not parallel cancella- FY91 Illinois $3,707,269 Iowa 2,670,813 Michigan State 2,607,306 Ohio State 4,113,097 Wisconsin 4,084,110 (From ARL Statistics) FY92 %Change $3,807,387 +3% 2,548,192 -5 2,520,018 -3 4,163,453 +1 4,035,678 -1 Average -1 Collections at Risk 353 TABLE2 for the original 1992 Percent of Total Budget Spent on Serials, FY91-92, and Percent Change for the Five Libraries in the Study study. Serial cancella- tion data for FY91-92 were requested from each of the five librar- ies in the original study. The lists of se- rial titles were input into a database with nine fields for each title: title, price of title in the year cancelled, place of publication, FY91 FY92 %Change Illinois 60.3% 57.5% -4.6% Iowa 50.7 49.8 -1.7 Michigan State 61.2 60.1 -1.7 Ohio State 73.2 69.6 -4.9 Wisconsin 57.7 57.0 -1.2 Average -2.8% (From ARL Statistics) average of one percent. This compares to a nine percent increase in serial expendi- tures for FY88-90.4 Also, despite some overall materials budget increases, each library also reported serial cancellations for these years. In fact, serial cancellations accelerated during the two years follow- ing the original study, with slightly more cancellations during the two-year period of FY91-92 (6,518 cancellations) than had occurred during the three years covered by FY88-90 (6,503 cancellations). Table 2 shows the percent of these li- braries' total FY91-92 budgets devoted to serial purchases taken from ARL statis- tics. Each library decreased the percent- age of its total materials budget allocated to purchase serials between FY91 and FY92. These reductions could be a result of efforts to balance the percentage of to- tal budgets spent on serials and mono- graphs. They also could represent at- tempts to gain control over constantly ris- ing serial prices. For FY88-90, a three- year period, serial expenditures by the five libraries accounted for an average of 57.4 percent of these libraries' materials budgets.5 By FY91-92, however, the five libraries were spending an average of 58.8 percent of their total materials budgets on serials, despite an acceleration in serial cancellations. Methodology The follow-up study used a research methodology identical to that employed language (English or non-English), year of cancellation, Li- brary of Congress (LC) call number, an indication of whether the journal was re- ceived as a gift (if reported), an indica- tion of whether the journal was a dupli- cate at that location (if reported), and which library cancelled the title. Each field was made searchable and sortable so that the data could be examined in any combination and also could be compared to the original research database (contain- ing cancellation data for fiscal years 1988, 1989, and 1990). Data on serial orders for FY88-92 were also requested from the five libraries in order to begin comparing the balance be- tween serial titles lost and gained. Order data were more difficult to provide for many of the libraries; the result was a lim- ited database of serial orders from four schools, covering fiscal years 1990, 1991, and 1992. (Wisconsin was unable to pro- vide serial order information at the time of the study.) For comparison purposes, a working portion of the cancellation da- tabase was created to match the order database. It too covers four libraries (with- out Wisconsin cancellation data), FY90-92. The authors received incomplete data from each library. In order to complete all or most of the nine fields for each title, a number of other resource tools were used. These included the OCLC database, Ulrich 's International Periodical Directory, Faxon's Librarian's Guide to Serials, the 354 College & Research Libraries University of Illinois' online catalog, Illinet Online, and MELVYL, the online catalog of the University of California li- braries. Even with this careful checking, some fields remained unfilled. When re- porting the results, the percentage of com- plete information for that field is included in the analysis. Data analyses were conducted using SPSS-PC, a software analysis program Total cancellations for the five- year period numbered 13,021. These total serial cancellations represent a 5.7 percent decrease in serial titles owned over five years based on total holdings reported by the five libraries in FY88. that is compatible with PC-File, the soft- ware used to create the different data- bases. Different analyses were possible using the searchable indexing assigned to each field. Reports were created July 1996 based on questions posed by the hy- potheses. Results: FV91-92 Cancellation Data Total Cancellations Total cancellations for the five schools rose during the two years following the ini- tial study. Cancellations FY88-90 (three years) totalled 6,503; for FY91-92 (two years), cancellations totalled 6,518. Table 3 provides a comparison overview of data collection for the two data sets. Total can- cellations for the five-year period num- bered 13,021. These total serial cancella- tions represent a 5.7 percent decrease in serial titles owned over five years based on total holdings reported by the five li- braries in FY88. However, these libraries paid 30.5 percent more in serial expendi- tures for 5.7 percent fewer titles during FY88-92. The number of overlapping (same) titles that were cancelled between librar- ies during FY91-92 totalled 474 (7.2%), an increase of 55 percent over FY88-90. Ex- TABLE3 Overview of Data Collection Results from the Five Schools for FYSS-90, FY91-92, and Combined FSS-92 FY88-FY90 FY91-FY92 FY88-FY92 (3 years) (2 years) (5 years) Number of titles 6,503 6,518 13 ,021 cancelled Number of same-title 281 474 1,084 cancellations (duplicate overlap) Percent of cancellations 95.7 % 92.8% 91.7 % representing unique titles cancelled Total dollar value of $690,225 .64 $1,031 ,584.50 $1 ,721 ,810.14 cancellations 1 Average cost of a $121.71 $193.43 $151.31 cancelled title2 1For FY88-90, approximately 7% of titles could not be assigned prices; for FY91-92, approximately 18% of titles could not be assigned prices. Therefore prices are conserva- tive estimates . 2 Average is based on the number of titles with prices: 93 % of FY88-90 titles have prices and 82% of FY91-92 titles have prices. amination of the overlap for the total five years of cancellation data shows an over- lap rate of 8.3 percent. Duplicate titles that were cancelled by four or five schools in- clude Chemical Abstracts, Books in Print, Methods of Biochemical Analysis, Public Administration Review, Progress in Ocean- ography, Soviet Electrochemistry, and Who's Who in America. Although duplicate titles still represent a small fraction of cancel- lations, this increase in overlap suggests the continuation of a trend for libraries to cut more heavily into their core collection. Serial Cancellation Prices The total price of serial cancellations for the five schools also continued to rise: cancellations totalled $690,225.64 for FY88-90,· and $1,031,584.50 for FY91-92. The dollar value of cancellations in- creased 49 percent in a one-third shorter time period. Price information for FY91- 92 was available for only 82 percent of the titles, so these totals are conservative estimates. The average cost of a cancelled serial found in the FY88-90 study was $121.71. According to annual serial surveys for 1991 and 1992, serial price increases raised the average price of a serial to $158.53 for 1991 and to $170.93 for 1992.6 The study also found an average serial price (FY91-92) of $193.43, 13 percent higher than the average reported by the annual serial surveys. These findings con- tinue the trend in these five libraries of cancelling higher-than-average-priced journals. Language and Place of Publication To simplify data collection, there were only two language categories assigned: English and non-English. More than 99 percent of the titles had language data assigned. Results indicate that 80.5 per- cent of the titles cancelled were English language and 19.5 percent were non-En- glish language. Data for FY88-90 also showed a predominance of English-lan- guage cancellations, with the ratio being Collections at Risk 355 7 4 percent English and 26 percent non- English. The authors collected data on place of publication and divided the information into two categories: U.S.-published and non-U.S.-published titles. Information on place of publication was assigned to more than 98 percent of the titles. For FY91-92, non-U.S.-published titles represented 55.4 percent of all cancellations, and U.S.-pub- lished titles represented 44.6 percent. Data from the earlier study showed similar · findings, with non-U.S.-published titles at 53 percent of cancellations and U.S.- published titles at 47 percent. Thus, there is a clear continuing trend of cancellation of nondomestic titles. Ubrary of Congress Subject. Classification More than 97 percent of the titles in the FY91-92 cancellation database were as- signed an LC subject classification. Fig- ure 1 shows LC classification distribution for the entire FY91-92 database. As was the case in the first study, the largest sub- ject group of cancelled titles was Q-class (science) serials, accounting for 21 percent of the total database. In FY88-90, Q-class accounted for 19 percent of all cancella- tions; therefore, Q cancellations increased 10 percent between FY88-90 and FY91-92. Cancellations in the science-technology- medicine call number ranges-Q (sci- ence), R (medicine), S (agriculture), and T (technology)-rose from 40 percent of total cancellations in FY88-90 to 45.6 per- cent in FY91-92. Again, FY91-92 repre- sents a one-third shorter time span than the original database. These data make clear the acceleration in the cancellation of science serials. Results: FY9G-92 Order Data Total Orders As noted earlier, librarians continue to place orders for serials in research librar- ies even during years of serious budget cuts. A look at serial orders helps to clarify the emerging patterns of serial collections and describes the overall collecting r------ --- --------- - - - - -- FIGURE 1 LC classification distribution for 6,518 serial cancellations for five libraries in study, FY91-92 600 (/) ~ ·;::: !U M ,.Q ;:J 400 ..s:: y M !U ~ (/) ~ 200 ~ ~ ~ b() ~ -- 0 0 u A B c D E F G H K L M N p Q R s T u v z \0 LC Classification lt) Cf) trends. Also, it not only helps to describe the type of serial that is most likely to be purchased but also, by default, reinforces the portrait of the at-risk serial. During FY90-92, four libraries placed 5,643 or- ders for new serials. (As noted earlier, order data were not available from Wis- consin.) As predicted by budget expen- ditures in table 1, the library that spent the most on serials, Ohio State University, ordered the highest number of serials. · Overlapping orders (i.e., titles ordered by two or more libraries) totalled 383 titles (5.8% of the total number of titles or- dered). Fewer than one percent of the titles were ordered by three or four schools, and only 4.6 percent were or- dered by two of the schools. Titles that were ordered by three or more of the li- braries were most often CD-ROM prod- ucts, including Agricola CD-ROM, Medline CD-ROM, PsychLit CD-ROM, and Current Contents on Diskette (all titles in the series). In addition, print titles ordered by three or more libraries included Harvard Papers in Botany, Journal of East Asian Linguistics, Southeastern Journal of Music Education, and Foreign Policy Bulletin. Serial Order Prices Whereas the average list price for serials for 1991 and 1992 was approximately $165, orders from the four test libraries · averaged significantly lower at $107.73 (based on 77% of the orders with an as- signed price). As discussed later, this may be a reflection of the subject content of the serials. Only nine percent of the titles or- dered cost more than $200, with the vast majority (79%) costing less than $100. Language and Place of Publication Place of publication (U.S. or non-U.S.) and language (English or non-English) were assigned to 99 percent of the serials or- dered. Serial titles ordered were over- whelmingly English (88.4%); only 11.6 percent were non-English. Of the titles ordered, 60.2 percent were published in the United States. These results reflect not Collections at Risk 357 only a selecting bias toward English-lan- guage domestic titles, but also undoubt- edly the realities of the marketplace that continue to focus on English-language publications from publishing houses with international offices (i.e., Europe and the United States). Library of Congress Classification LC classification numbers were assigned to 98 percent of the titles in the order da- tabase. The greatest activity in serial or- dering occurred in the general social sci- ence (H) classification, which accounted for 20 percent of all orders placed. Sci- ences (classes Q, R, S, and T) accounted for only 11.4 percent of new orders. Lan- guage and literature titles, in the P classi- fication, totalled 10.9 percent of all orders. Fewer than one percent of the titles were ordered by three or four schools, and only 4.6 percent were ordered by two of the schools. Classification order data by cost are dis- played in figure 2, in juxtaposition with the cost of serial cancellations by clas- sification. As indicated earlier, the lower expenditures in serials, and the acquisition of titles costing less than $100, probably reflects the generally lower costs of serials in the social sciences and hu- manities. FY9G-92 Order and Cancellation Data Compared The authors conducted a number of analyses comparing the FY90-92 order and cancellation databases. As previously stated, these files were created using or- der data provided by four of the five schools. The participating schools were the University of Illinois at Urbana- Champaign, the University of Iowa, Michigan State University, and Ohio State University. A file matching the four schools' cancellation data for FY90-92 358 College & Research Libraries TABLE4 Data Comparing Orders and Cancellations, FY90-92 Number of cancelled titles 6,025 (92.9% unique) Number of orders placed 5,643 (93.3% unique) Dollar value of cancellations* $919,422.47 Dollar value of orders placed* $607,908.18 • 26% of cancellation records for this time period could not be assigned price; 23% of order records for this time period could not be assigned price. Therefore, the dollar values are conservative estimates. also was created for comparison pur- poses. Total Cancellations/Total Orders Compared Cancellations and orders for FY90-92 are shown in table 4 and figure 2. In terms of numbers, there were approximately six percent fewer orders than cancellations. In terms of dollars, 51 percent fewer dol- lars were spent on orders than were made through cancellations, indicating either a diversion of money to cover increased costs of serial subscriptions that were maintained or a shift in budgeting to new resources. Figure 2 shows orders and cancella- tions compared by subject classification. This figure highlights the subjects in which the greatest amount of activity oc- curred, both positive and negative. Within July 1996 this study, a major shift is taking place as dollars are moved from science titles and then are added to social science and literature titles. Table Slooks specifically at orders and cancellations (both price and number) for call numbers for these disciplines. Call numbers Q (science), H (social science), K (law), and P (lit- erature) show the greatest percentage changes in figure 2. Table 5 examines these ranges and adds data for QRST (all sciences). The positive increases in the social sciences and literature call number areas are nearly as dramatic as the negative losses in the science call number areas. Analysis on a title-by-title basis also indicates the first stirrings of a shift from print sources to CD-ROM formats in the sciences. Eighty-one titles (1.4% of all or- ders) were for CD-ROMs, whereas only eight CD-ROM titles were cancelled (13% of cancelled ti~les for FY90-92). New se- rials in these collections also appear to include several indexes and reference sources on the part of the sciences. Serial orders in the humanities and social sci- ences appear to remain with print-based publications. Orders and Cancellations by Cost Figure 3 shows two pie charts compar- ing orders and cancellations broken down into three subscription price categories: less than $100, between $100 and $200, and greater than $200. An obvious move away from high-priced journals is found. This result could be expected because of TABLES Order and Cancellation Data Comparisons, FY90-92 b:y Subject with Percent Change #Cancelled #Ordered $Cancelled $Ordered H (social science) 826 1,129 (+37%) $57,593.29 $105,806.86 (+84%) K (law) 145 325 (+124%) 26,726.27 59,867.50 (+124%) P (literature) 391 613 (+57%) 17,446.60 37,964.31 (+117%) Q (science) 1,293 644 (-50%) 375,544.82 143,435.73 (-61 %) QRST (all sciences) 2,722 1,414 (-48%) 613,756.77 232,838.61 (-62%) a (J FIGURE2 Total cost of serials ordered and cancelled by LC classification for four schools in study, FY90-92 400,000.00 350,000.00 300,000.00 250,000.00 200,000.00 150,000.00 100,000.00 50,000.00 0.00 ~~~ r-. ~ rl r. B C 0 E F G H ln.n_.-~ r rJ K L M N P Q R S T U V Z LC Classification 360 College & Research Libraries July 1996 FIGURE3 No. of titles ordered and cancelled by price, FY90-92 No. of titles ordered Lese Greater Than $200 9.0% Between $100-$200 12.0% No. of titles cancelled Lee a Than $100 68.0% Between $100-$200 14.0% the large number of science cancellations (the area of the highest-priced journals). Subscriptions priced over $200 account for 18 percent of cancellations, but only 9 percent of orders. Increases are found in the lower-priced subscriptions. Subscrip- tions priced under $100 -account for 68 percent of cancellations and rise to 79 percent of orders. · Greater Than $200 18.0% Overlap in Orders and Cancellations As reported earlier, overlap (cancellations of the same titles by two or more librar- ies) for FY91-92 was 7.2 percent and 8.3 percent overall (FY88-92 data). This is a small portion of all cancellations and sug- gests that research libraries are continu- ing to chisel away at their unique publi- cations. However, the implication is that the cancellation overlap is growing, bring- ing with it an ever-smaller group of core journals that may be duplicated from li- brary to library and, in the future, may be online, full-text journals that are easily accessible. This movement toward simi- lar collections, in tandem with the elec- tronic initiatives that several of the larger publishing houses are making, indicates a research community far less deep and rich than was previously the case. The surprise, and perhaps the mitigat- ing factor in this gloomy future, is the lack of overlap in serial orders. It had been expected that orders will be highly du- plicative, but only 6.8 percent were or- dered by two or more schools. The prob- lem with drawing too strong a conclusion on these data is the shortness of time for the study. Collection development offic- ers of the CIC libraries, when asked about this low overlap figure, suggest that al- though there is no mechanism in place to check other libraries' holdings before placing an order, there is reliance on CRL holdings to avoid ordering some titles. There also might be in this a suggestion that these institutions, while maintaining very similar research programs in a broad sense, can concentrate on very different aspects of a discipline and that libraries are now buying on a "micro" level as opposed to a "macro" level because of the pressures of the marketplace. At-risk/Not-at-risk Journal Profiles Finally, these data suggest the profile of journals at risk for cancellation and those likely to be ordered. Data from the previ- ous study identified an "at-risk journal" as a higher-than-average-priced, English- language science title. The results of two more years of data do nothing to dispel this image. On the contrary, it is even more evident that science titles are at risk and that the cost of journals is driving cancellation decisions. Conversely, "not-at-risk journals" ap- pear to be those in social science fields, again in the English language, with a Collections at Risk 361 price tag of under $100. Also, although it may appear from the data that social sci- ence and humanities serials are surging ahead of the sciences, it is perhaps more likely that journal-buying habits in the social sciences and humanities have not It had been expected that orders will be highly duplicative, but only 6.8 percent were ordered by two or more schools. The problem with drawing too strong a conclusion on these data is the shortness of time for the study. had to change as much as those in the sciences, an area in which rising prices devour huge portions of library materi- als budgets. One lesson might be that a well-designed, scholarly social science or humanities journal with a reasonable price tag will continue to do well in the library market and with the community of scholars. 7 Conclusions All but one of the hypotheses were sup- ported. As predicted, the data collected as follow-up to the original 1993 study parallel its results and point to continued serial cancellations at an increased rate. The at-risk serials described in the origi- nal study continue to be expensive, En- glish language, foreign-published science titles. They appear to be at risk for can- cellation not just today, but also almost certainly in the future: There is no doubt that the sciences, with their high-priced titles and whose funds command a large percentage of not just library serial bud- gets but the entire library materials bud- gets, are definite targets for methodical reduction. Overlap of cancelled titles increased by 55 percent between FY88-90 and FY91- 92, jumping from 4.3 to 7.2 percent of to- tal cancellations (8.3% overall). This in- crease can be explained by examining how collections and resulting cancella- tions are structured. Figure 4 shows a tar- 362 College & Research Libraries get representing academic library serial collections. The center of the target is the core serial collection-those titles with high-impact factors, high use, and which are duplicated in nearly every academic library. 8 The concentric rings radiating from the core represent steps away from the core-those titles that are incremen- tally less frequently used and less fre- quently cited . The outermost ring is where cancellations begin and overlap among libraries of similar size and mis- sion is less likely. The peripheral titles of the outer rings are among the first to be cancelled. Figure 4 shows that cancella- tion overlap in the original study (FY88-90) measured 4.3 percent. As can- cellations continue and move toward the core, overlap among cancelled titles grows. This accounts for the increased overlap of 7.2 percent found in the FY91-92 data. For the five-year period FY88-92, the cancellation overlap is 8.3 July 1996 percent. The cancellation overlap percent- age will probably grow as libraries are forced to cancel and still maintain a core group of serials. As predicted, there were fewer new serial orders than serial cancellations for the FY90-92 period. This dramatic shift is symptomatic of the inability of librar- ies to keep pace with the serial publish- ing market and is strongly indicative of the effect of the marketplace on collection decisions. It also was hypothesized that orders would not parallel cancellations in terms of subject. This hypothesis also was proven true: 2.5 times more science titles were cancelled than were ordered. Con- versely, twice as many social science titles were ordered as were cancelled. Finally; the authors predicted that li- braries would tend to cancel different titles and order the same title (i.e., that overlap between orders would be greater than overlap between cancellations). The FIGURE4 Research library's journal collection, core journals (center) to peripheral journals (outlying rings). Core journal collections: high overlap between peer collections Peripheral journal collections: low overlap between peer collections ,------ FY88-FY92 overlap in cancellations (8.3%) FY91-FY92 overlap in cancellations (7 .2%) r---- FY88-FY90 overlap in cancellations (4.3%) Cancellations begin at the outer rings where overlap of journal title holdings with other similar- sized libraries is small and proceed toward the core, with overlap in holdings increasing as cancel- lations are made. surprise finding of the data collected for FY90-92 shows that serial order overlap was less than serial cancellation overlap and, therefore, the hypothesis predicting the opposite outcome is not supported. The vast majority of cancellations (92.9 %) were unique to the study group, and 93.3 percent of orders were unique to this group. These nearly identical figures ap- pear to support the continuing existence of diversity in academic library collections. However, because the data cover only a two-year period and have not been veri- fied by a follow-up study, a broader study of a longer period of time would be re- quired to support the notion of retained diversity in academic library orders. The acceleration toward cancellations of foreign materials and higher English- language cancellations may be accounted for by examining a similar trend toward science subject-classed cancellations. Many sciences titles, even those originat- ing in foreign countries, are published in English. Tony Stankus et al. studied the trend toward translating foreign-language titles into English and found that science titles published in Germany showed an increase from 1960 to 1978 in the number of articles published in English. 9 This demonstrates a shift toward English-lan- guage publication for foreign-published science journals and could account for the predominance of English-language titles from outside the U.S. being targets for cancellation, particularly given that 45.6 percent of all cancellations in the past two years of the study are science titles. Future Research The need to conduct research on entire serial collections is obvious. The best way to understand the whole picture is to look collectively at serials that are being added and lost. The problems in doing this are technical, not intellectual. When the tech- nical difficulties of sharing complete se- rial information are solved, so will be many other issues associated with serial collection development. Those issues in- Collections at Risk 363 elude broad communication of cancella- tion decisions, holdings information, and acquisition decisions. There is a continu- ing need for consistent and timely data from libraries documenting the cuts that are occurring annually. This lack of data hampers efforts to document in an or- derly way what is happening, and pre- vents libraries from making informed decisions about the future. The most important type of research that needs to be linked with this study are investigations of the effect of these serial cuts on the users of the collections. These effects can be charted by following local journal use, interlibrary use, and the traffic found on online citation/ full-text databases such as CARL UnCover. At the end of the day, the most important ques- tion is: Are users being given access to the materials they need? Summary The trend that is beginning to take shape after examining these five years of re- search is for libraries to move slowly to- ward a core list of serial titles that will satisfy the mainstream and, perhaps, to add serial titles in very discrete portions of disciplines that are unique to each in- stitution. Although the original motivat- ing thought behind this research is that research libraries were cancelling the same titles, it is more likely that these li- braries are cancelling the same kind of titles. The same prospect is true for or- ders. Research libraries belong to the same genus, but not necessarily the same species. It is important for libraries to re- flect the persona of the institutions they serve, to have the information necessary to build collectively with other research library partners to support a larger re- search agenda that is beneficial to many institutions and many researchers, and to avoid the "minimums and ho hums" that reliance on core collections can bring.10 In 1994, Anna Perrault reported on re- search documenting the decline of the monograph in research libraries. 11 Her 364 College & Research Libraries conclusions (that monographic collec- tions are becoming more similar and that the core is under attack) unhappily mir- ror the conclusions of both the present and previous serial study. Perrault's work also urges similar deductions from the results of the study reported here: that the base of information found in traditional serial publications is dwindling; that the marketplace has become the academic li- brary community's collection policy; that communication among collection devel- opment librarians will need to be quick and responsive; and that, as Perrault notes," ... resource sharing will need to be a global cooperative effort." 12 The conclusions of this study, like Perrault's research, echo in many respects the conclusions of the Mellon Foundation report, University Libraries and Scholarly Communication, wherein the dynamics of serials collections are described in terms of the effects of the marketplace and the problems associated with science-based July 1996 journals are found to be the most acute. 13 This report further notes: "In the face of this pricing crisis, libraries have re- sponded by essentially redistributing their resources, a mode of response that cannot go on indefinitely . . . there is a growing realization that no research in- stitution can hope to sustain a self-suffi- cient collection into the indefinite fu- ture."14 This fundamental statement of reality is reflected in the specifics of this study of five research libraries over ape- riod of five years. The best hope, when observing the trend of individual collec- tions losing their unique characteristics, is for these institutions to marshal the new tools available to redesign the way librar- ies interact and manage their collections. Au. note: The authors wish to acknowl- edge the work of graduate assistant Xiaoyu Zhang and the financial support of the Uni- versity of Illinois at Urbana-Champaign's Research Board and the University Library Research and Publication Committee. Notes 1. Tina E. Chrzastowski and Karen A. Schmidt, "Surveying the Damage: Academic Library Serial Cancellations 1987-88 through 1989-90," College and Research Libraries 54, no. 2 (Mar. 1993): 93-102. 2.1bid. 3. Association of Research Libraries, ARL Statistics 1990-91 and ARL Statistics 1991-92 (Wash- ington, D.C.: ARL, 1992, 1993). 4. Chrzastowski and Schmidt, "Surveying the Damage," 95 (table 1). S.lbid. 6. Kathryn Hammell Carpenter and Adrian W. Alexander, "Periodical Price Index 1992," Library Journal117 (Apr. 15, 1992): 5~2. 7. This conclusion was also reached by Walt Crawford and Michael Gorman in Future Librar- ies: Dreams, Madness, & Reality (Chicago: ALA, 1995), chapter 9, especially pp. 145-47. 8. The term impact factor refers to the ratio between citations and citable items published. For more information, see the introductory material to Journal Citation Reports (lSI publishers) or see Eugene Garfield, "Citation Analysis As a Tool in Journal Evaluation," Science 178 (1972): 471-79. 9. Tony Stankus, Rashelle Schlessinger, and BernardS. Schlessinger, "English Language Trends in German Basic Science Journals: A Potential Collection Tool," Science and Technology Libraries 1, no. 3 (1981): 5~6. 10. William E. McGrath, "Collection Evaluation-Theory and the Search for Structure," Li- brary Trends 33 (winter 1985): 241-66. 11. Anna H. Perrault, "The Shrinking National Collection: A Study of the Effects of the Diver- sion of Funds from Monographs to Serials on the Monograph Collections of Research Librar- ies," Library Acquisitions: Practice and Theory 18, no. 1 (1994): 3-22. 12. Ibid., 16. 13. Anthony M. Cummings, MarciaL. Witte, William G. Bowen, Laura 0. Lazarus, and Rich- ard H. Ekman, University Libraries and Scholarly Communication (Washington, D.C.: ARL for the Andrew W. Mellon Foundation, 1992). 14. Ibid., xxii.