Studies of availability have appeared in the literature for at least seven decades. Gaskill, Dunbar, and Brown’s 1934 use survey at the Iowa State College library calculated the percentage of time both undergraduate students and graduate students “obtained what they sought” and identified eleven reasons for failure to locate the sought-after book or magazine.⁹ In 1975, Buckland’s major monograph, Book Availability and the Library User, based on research at the University of Lancaster, reported that circulation was the major barrier to book availability, and recommended variable loan periods and purchase of duplicate copies to increase availability.¹⁰

At least forty availability studies were published between the 1930s and the mid-1980s, according to a seminal literature review and analysis published in 1986 by Mansbridge.¹¹ He discovered that most investigations were based on known-item searches conducted in academic libraries, while two-thirds were based on real users, and the remaining third used a set of citations simulating user needs. This literature review will only mention the most seminal items included in Mansbridge, while focusing on research published later or not covered by him.

This section analyzes the results of the availability studies issued since Mansbridge’s literature review (the cutoff date was 1984) as well as a small number of studies published between 1980 and 1984 not included in Mansbridge—mainly Ph.D. dissertations, a format he did not address. Appendix A summarizes 46 investigations of availability based on actual clients conducting known-item searches that were published from 1980 to 2001, listing the author, publication date (or degree date for dissertations), institution, sample size, the number of successful searches, and the overall availability rate. For the purpose of consistency and accurate comparison, the availability percentages have been calculated by the author to the first decimal point based on the data reported by the original researchers. Recalculation of the initially reported percentage was sometimes necessary due to inconsistent practice in rounding off numbers by some researchers, inexplicable errors in the originally reported percentage, or cases in which the researchers derived a final percentage by multiplying the percentage results at each branch (the method used by Kantor) rather than by simply dividing the number of successes by the total number of searches. Instances in which the percentage in appendix A differs by more than half a percentage from that which was initially reported are noted in footnotes to the appendix.

Examination of appendix A reveals that overall availability ranged from 33.8 percent at the University of Münster to 83.8 percent at Cardiff University. The high rate at the latter can probably be attributed to the study’s focus on “short loan,” items—i.e., “reserve” in North American terminology. Twenty-nine of the 46 reported results showed an availability rate in the 50s or 60s percentage range. The unweighted mean availability rate (with each of the 46 percentages counting equally) is 61.3 percent, a figure remarkably similar to the 61 percent reported by Mansbridge, who used this method for calculating the mean.⁴⁰ However, Mansbridge found a larger range in the results from different studies (8 to 89 percent).⁴¹ This literature review’s weighted mean (a calculation method not used by Mansbridge that factors in the size of each investigation) was 63.1 percent (17,801.3 successes in 28,207 searches) for 43 investigations. Excluded from calculation of the weighted mean were Simpson’s two surveys at the Pikes Peak Library District (for which the raw data is unavailable) as well as Jacobs and Young’s University of Sussex research, where the 99,778 searches would badly skew the overall average. The 61 or 63 percent mean availability rates found in this review are a bit higher than the general 50 to 60 percent range cited by Bachmann-Derthick and Spurlock; Lieshout; Ciliberti et al.; and Chandler.⁴²

The three cases focusing exclusively on serial or journal articles, Bachmann-Derthick at the University of New Mexico, Roberts at East Tennessee State, and Ciliberti et al. at Adelphi University, found lower-than-average availability rates, 55.7 percent, 54.5 percent, and 44.9 percent, respectively.⁴³ Among the investigations addressing multiple formats, Chaudhry and Ashoor’s study at King Fahd University reported a 58.6 percent availability rate for journal articles, while Harris and Garner found a 54 percent serials availability rate at the University of Western Australia.⁴⁴ While these data might lead to a facile assumption of a lower general availability rate for serials (and a confirmation of the longstanding perception that serials cause difficulties), in the final analysis, the number of cases involving serials is too small to allow firm conclusions.

As indicated in appendix A’s footnotes, three studies measured the availability rate before and after librarian intervention. (Note that the vast majority of availability studies have not included librarian assistance as a factor in the equation, so only the initial result is included in appendix A’s final column.) A librarian’s help increased availability at the Cleveland Heights-University Heights library from 52.4 to 60.8 percent (Kuraim’s study), at the Cleveland Health Sciences Library from 59.6 to 63.5 percent (Rashid’s study), and from 62.8 to 68.5 percent at the King Fahd University of Petroleum and Minerals (Chaudhry and Ashoor’s study).⁴⁵ These investigations suggest, as one would intuitively think, that librarian assistance does result in somewhat higher availability.

The small number of subject-based availability studies (not tabulated in appendix A) display less variation in their availability rates than was apparent in the known-item searches, although the results were comparable. Overall availability was 56.8 percent (108 of 190 searches) and 62 percent (31 of 50 searches) at William Patterson College and 62.2 percent (153 of 246 searches) at Adelphi University or 60 percent, as reported by Ciliberti et al., if the availability rates at each of Kantor’s branches are multiplied by each other.⁴⁶

Following appendix A’s format, appendixes B through G summarize availability at each of Kantor’s branches, covering respectively, the bibliographic citation, acquisitions, the catalog, circulation, library operations, and the user. While the fractional results occasionally reported in these appendixes may seem counterintuitive, they are easily explained by the location of one volume of a multivolume title, or the use (in a few studies) of a “correction factor” to distribute proportionately among the branches failed searches for which the precise cause could not be determined. Not represented in these appendixes are the investigations that did not employ Kantor’s branching technique plus the reports by Zondi as well as Thorne and Whitlatch, where useable data were not presented.⁴⁷

Appendix B shows a consistently high level of availability at the bibliographic branch, ranging from 94.9 to 100 percent. The unweighted mean availability for the nine reported cases is 97.1 percent, while the weighted mean is 97.7 percent (6,990 of 7,154 searches were successful). Failures in this branch were usually due to incorrect citations for the author or the title.

Appendix C tabulates 33 reported cases of availability at the acquisitions branch—the branch most frequently included in studies employing Kantor’s methodology. While availability ranged from 66.1 to 97.2 percent, it exceeded 90 percent in the majority of instances (17 of 33) and was more than 80 percent on all but six occasions. It is noteworthy that the two lowest availability rates, 68 percent and 66.1 percent, appear in studies of journal articles by Roberts at East Tennessee State University and Ciliberti et al. at William Patterson College.⁴⁸ However, the findings from the other two studies focusing on journal articles, Bachmann-Derthick and Spulock at the University of New Mexico (85.2 percent) and Shaw-Kokot and Varre at the University of North Carolina at Chapel Hill (94.4 percent), are generally consistent with the other results in the appended table.⁴⁹ The unweighted mean availability rate for these 33 reported results at the acquisitions branch is 87.1 percent and the weighted mean for 32 cases (all but Revill’s 1988 study at the Liverpool Polytechnic Library Service, which reported availability percentages but no raw data for the branches) is 89.6 percent with 19,080 successes among 21,299.9 sought-after items.⁵⁰ At the sub-branch level, the predominant reason for failure was the fact the library had not acquired the item. Other causes of acquisitions failure included weeding, cancellation, the sought-after item’s location in a different branch, the item having been declared missing, and an “on order” title having not been received.

Appendix D demonstrates a high success rate at the catalog branch, with the reported availability percentages running from 86.4 to 99.6 percent. Indeed, the rate was more than 90 percent in 20 of 21 instances and more than 95 percent in more than half the cases (11 of 21). The unweighted mean availability rate was 94.7 percent and the weighted mean 96.5 percent (13,328.4 of 13,806). Appendix D’s footnotes show that librarian assistance increased availability at this branch at the Cleveland Heights-University Heights Public Library from 93.4 to 98.8 percent (Kuraim’s study) and at the Cleveland Health Science Library from 96.9 to 98.6 percent (Rashid’s study).⁵¹ Frequent reasons for catalog branch failure were inability to locate the record in the catalog as well as transcription of an incorrect call number or an incorrect location.

Despite early studies identifying circulation as a major barrier to book availability, success rates in the circulation branch, presented in appendix E, exceeded 80 percent in all but four instances, ranging from 66.2 to 100 percent. The circulation branch’s unweighted mean availability rate for 31 reported results was 87.9 percent and the weighted mean for 30 cases, excluding Revill’s 1988 study, was 87.4 percent (16,899.73 of 19,337.7).⁵² It is noteworthy that each of the three investigations of journal articles found high availability rates in this branch: Bachmann-Derthick and Spurlock (96.7 percent), Roberts (97.5 percent), and Ciliberti et al. (100 percent), possibly reflecting the fact that journals are less likely to circulate.⁵³ As would be expected, the overwhelming majority of circulation failures were attributable to the item being checked out by a patron while another reason was checkout for interlibrary loan purposes.

Appendix F’s tabulation of 31 availability rates in the library operations branch shows a range from 65.7 to 98.9 percent with more than half (18) exceeding 90 percent and only four less than 80 percent. This branch’s unweighted mean availability rate is 88.8 percent, and the mean weighted rate, for one less case with the 1988 Revill study not counted, stands at 89.9 percent (15,185.6 successes in 16,889.73 attempts).⁵⁴ Of library branch failures, major causes of error were missing items, bindery operations, reshelving operations, misshelved items, and items in technical processing.

Appendix G tabulates success rates at the user branch as defined and reported by the original investigators. This branch almost always includes user failure to locate the item on the shelf, but may also include user failure with the bibliographic citation or in the catalog. Ranging from 77.7 to 96.6 percent, the user success rate exceeded 90 percent in 18 of the 32 reported results and was less than 80 percent only once. Both the unweighted means (32 instances) and the weighted means (31 instances with 13,833 satisfactions out of 15,413.8 searches) equal 89.7 percent. A high proportion of user errors was due to the inability to locate the item on the shelf, while another cause of failure was the user not understanding the classification system.

Whereas appendixes B through G summarize success rates in each branch, appendix H analyzes failures by branch, indicating for the various studies the number and percentage of failures attributed to each branch. In order to provide a more accurate estimation of total user failure, appendix H’s user column combines the results from appendix B (user bibliographic failure), appendix D (user failure at the catalog), and appendix G (user failure at the shelf or overall user failure). The findings have been calculated by the author from the original researchers’ raw data.

Within each branch, appendix H shows wide variation in the proportion of failures attributable to the branch, ranging from 5.9 to 70.4 percent in acquisitions; 0 to 50 percent in circulation; 1.4 to 46.4 percent in library operations; and 6.7 to 42.9 percent for the user. Of 8,991.7 total failures, 25.3 percent were in the acquisitions branch; 27.1 percent in circulation; 19.0 percent in library operations; 24.7 percent by the user; and 3.9 percent were not solely attributable to any of these branches. Disregarding the four investigations (Wood, Bremer, and Saraidaridis; Kuraim; Tamura and Sakai; and Jacobs’ fall survey) that did not include all of these branches, the largest portion of failures was caused by the circulation branch in eleven cases; the acquisitions branch in eight cases; and library operations and the user in five cases each.⁵⁵ While Mansbridge concluded that circulation and library operations were the largest source of book unavailability in academic libraries, this analysis found that the highest percentages of total failures were in the acquisitions and user branches, and that the circulation and acquisitions branches were the largest cause of failures in the most cases.⁵⁶ However, it is apparent from careful review of the data that no branch emerges as the major obstacle to availability.

Appendix I summarizes the results of simulated studies that did not involve actual patrons. Note that Kuraim as well as Chaudhry and Ashoor included both real and simulated components in their research projects.⁵⁷ Availability ranged from 13 to 84 percent with an unweighted mean of 61.8 percent and a weighted mean of 60.6 percent (2,010 successes out of 3,315)—figures quite similar to the averages for real studies. One would intuitively anticipate a higher availability rate in a simulated study because user errors would be eliminated and, when a shelflist sample is being used, the items would already have been acquired by the library. Indeed, when the one outlier (the 13 percent availability in Kuraim’s sample from the American Book Publishing Record) is disregarded, the mean availability rates increase to 68.8 percent (unweighted) and 69.1 percent (weighted with 1,945 of 2,815 on the shelf).⁵⁸

Finally, many of these studies have addressed subsidiary issues beyond the scope of this analysis, such as

• whether user type (student or faculty, full or part-time student, and so on) is a variable influencing the ability to locate items;
• comparison of the performance of different libraries; and
• longitudinal comparisons within a single library.⁵⁹

This literature review and analysis of more than fifty specific investigations of availability found that the majority used Kantor’s branching method (but numerous combinations of branches), were implemented in academic libraries (with some in public and school libraries), and were for known-item searches by actual patrons. About half the studies were conducted in U.S. libraries with the remainder carried out in ten other countries.

There is considerable variation among libraries in overall availability as well as availability in different branches. One of the values of these studies is the identification of reasons for lack of availability in a specific library context. Branch-level analysis shows that availability at each branch only occasionally falls below 80 percent and frequently exceeds 90 percent. Moreover, no branch stands out as the major bottleneck or barrier to availability. This study, in conjunction with Mansbridge’s review, demonstrates that, in a print environment, patrons on average find what they are looking for only slightly more than 60 percent of the time. An obvious question concerns what the availability rate would be for electronic resources on the Web or licensed by a library.

More than a decade ago, Kaske argued that the availability concept was no longer applicable to the then current environment because user needs were often met through externally-procured print items. He advocated the development of a new availability model that would incorporate the searching of multiple libraries and the time the user could wait to obtain the item.⁶⁰

Expanding upon Kaske, this paper maintains that the traditional availability concept reviewed here can be adopted with modification for an electronic environment to measure user success in immediately obtaining sought-after items on the Web or in the proprietary e-resources licensed by a particular library. While many patrons may no longer expect the immediate gratification of finding an item on a library shelf, they may nevertheless expect immediate gratification in locating it electronically. The extent to which they are successful in doing so can serve as an objective, user-centered evaluation and performance measure in the newly emerging electronic environment. Such tests could also identify barriers to user success and facilitate the design of better electronic systems.

A detailed outline of the procedures for conducting an electronic availability study is beyond this paper’s scope. As with traditional availability studies, the method would work best for known-item searches by real users, although a simulated study would be a possibility.

Future research issues regarding electronic availability studies include

• development of a method for measuring availability of electronic resources;
• determination of an expected standard for availability rates in an electronic environment;
• comparison of electronic availability rates with those reported in the numerous studies conducted in traditional print environments; and
• identification of the major barriers to electronic availability and the relative success rates at the various barriers.

One might hypothesize the following potential barriers or branches: the item is not on the Web; the URL has changed; the item is only available in a proprietary database that is not licensed by the user’s library; the maximum licensed number of simultaneous users are logged on; technological failure, such as a server is down or the local power off; and the user is unable to locate an otherwise available item. Many of the issues that have been investigated in typical availability studies could be addressed in an electronic availability study. Examples would include

• comparative availability among different user groups;
• comparative availability among libraries;
• longitudinal comparison within a library;
• the impact of professional assistance to the user upon availability; and
• the impact of policy changes on availability.

Finally, one could design and conduct electronic accessibility tests, modeled on Orr’s DDT, to ascertain the time requited to obtain electronic documents.

While availability studies are sometimes viewed as a research approach from the 1970s and 1980s, this literature review has demonstrated their continued use on an international scale throughout the 1990s and into the twenty-first century.⁶¹ Although developed for a print environment, the issues investigated in an availability study are equally relevant to an electronic environment. Mansbridge contended that an availability study could be used for nonlibrary purposes, such as analyzing the availability of audio-visual equipment in an academic setting.⁶² If, at some future point, libraries cease to exist or are radically transformed into now unrecognizable entities, an electronic availability study could still be employed in an academic institution as an objective, user-centered evaluation measure to help assess how effectively faculty and student information needs are being met.