Trends at Twelve Association of Research Libraries Members, 1908-2006

Robert E. Molyneux

Abstract

This series of pages presents the charts and tables characterizing the behavior of a few variables collected from university libraries that are currently members of the Association of Research Libraries (ARL) and that reported data from 1908-2006. This analysis was done after looking at the observed trends in public libraries in the US where first results led to the speculation that the positive experience of public libraries during the period from Fiscal Year 1992 - Fiscal Year 2004 may be a result of the business cycle. Subsequent work, appearing both on Web pages and in print appear to confirm this supposition. The argument is that, during that period, US economic variables rose as did library variables such as expenditures and staff. When the economy does well, so apparently do libraries. And we can speculate that when the economy does poorly, so do libraries. Indeed, how could it be otherwise? The money that libraries spend comes from taxes and gifts and both go up in good times and fall in bad times.

The thirteen years of currently available public library data cover a period of general economic growth so there are not yet enough years in the public library series to go beyond speculation about what might happen when the economy does not grow. Work I did with the Gerould/ARL data covered many years and led me to conclude that there was a relationship between library spending and behavior and the economy. I have updated that work on these pages to provide background behind the speculation about public libraries. What seems apparent from the results reported here is that these ARL libraries follow the business cycle and that other exogenous events affect the fortunes of libraries, also.

ARL is an organization with the largest university and non-university libraries in the US and Canada. It was founded in 1932 but data on its institutions were actually collected earlier beginning in the 1907/08 academic year. These data were collected first by James Thayer Gerould (1872-1951) at the University of Minnesota, then subsequently at Princeton University after Gerould moved there. The series continued at Princeton after Gerould retired. In 1962, the Association of Research Libraries began publishing their data and followed the format of the Princeton data. The Gerould Statistics were compiled in 1986¹ and various compilations, some including the Gerould data, were published by ARL and are maintained on the ARL site today². ARL's data, then, were collected before the founding of the organization and they apparently comprise the oldest continually collected series of library data.

The 12 institutions analyzed here are those for which we have the longest series of data. It is doubtful these institutions are representative of all college or university libraries or, indeed, all ARL libraries but the graphs are evocative and, I believe, compelling.

Growth in Volumes, 1908-2006

Growth at the 12 libraries is examined using counts of volumes added and a percent using volumes added and volumes held.

Expenditures for Materials and Binding, 1908-2006

Expenditures for materials have been kept for the period and these expenditures are examined in several different ways. Although inflation over the period has reduced the dollar to where it is worth about a 1913 nickel, these libraries kept ahead of inflation in purchasing power having increased their expenditures about 26 times (in constant dollars) from 1908 to 2006.

The Libraries and the Data

This series of pages has graphs and tables of data from the 12 ARL institutions which are in the data series from the beginning. Analyzed here are trends related to library growth and expenditures for materials and binding. These libraries may or may not be typical of all colleges and universities but we do have them for the longest period. These institutions are:

This group is primarily Midwestern, perhaps reflecting the fact that Gerould was interested in having data from his peers institutions when he was at the University of Minnesota. The "Institution number" was originally developed as a three-digit number by Kendon Stubbs and David Buxton in their pioneering Cumulated ARL University Library Statistics³ and was subsequently expanded in the first edition of The Gerould Statistics to four digits. Eventually, the second series of data compiled by the Association of College and Research Libaries⁴, the HBCU⁵ statistics, and the ARL data used the four-digit institution numbers allowing easy programming of data from all of them in one file. These numbers are included here in case anyone wishes to replicate or extend this analysis.

What is presented here are the mean values by year for several variables from these 12 institution's libraries. Grouping libraries tends to damp year-to-year variations brought on by unusual values. For instance, from time to time, libraries will make large purchases that are added to the collection. Such purchases are a part of the library growth process and should be accounted for as added volumes but, these kinds of purchases tend to stand out when one looks at the data--particulary in a time series. They are growth but growth of a different kind, however. Other kinds of unusual changes can be brought about by fire or by reorganizing campuses and institutions.

Electronic collections of these data tend not to have footnotes nor include footnotes from paper copies of the data that were converted to digital form. Consider the volumes added figure for Princeton, for instance, which had an unusual pattern of volumes added resulting from what amounts to an accounting change when the university library began including the Gest Oriental Library's 133,419 volumes in its data beginning in 1936/37⁶. The library had been adding on the order of 20-25,000 volumes a year the 10 years before and after but this year it "added" 157,000 volumes from another collection at Princeton that to that time had not been included in the library's totals. That fact is a part of the growth at the Princeton university library, is unusual as the pattern shows, but the note is not included in electronic versions of the data--I know about it because I compiled the data. Now, when Princeton's data are grouped with data from other institutions, the influence of this one unusual value is lessened by including it with other more typical values for this year from other institutions. Thus, we can discern trends affecting all institutions in these groups.

The experience of Princeton shows us several things. One is that such patterns are unusual but not uncommon as anyone who examines the data systematically knows. Second, grouping like libraries is a useful technique in damping such fluctuations and compensatomg for the occasional missing values and still allowing us to discover patterns for the groups. Third, we should be wary of "correcting" seemingly incorrect values. We should follow the principle enunciated in the Cumulated ARL University Library Statistics and resist the temptation to change the data when something seems odd:

These vagaries in data collection over the years stand here as they appeared in the original annual statistics...⁷

This group of 12 libraries is smaller than the groups of 2,000 libraries found in the public libraries so the charts are not as smooth. Still, there appears to be an interesting story in these charts that will repay further analysis.

One of the common tradeoffs with data is that the more years of data one has, the fewer variables there are for analysis and for fewer--in this case--libraries. Gerould's first published data were in 1907/08 and it included five variables:

• Books in Library
• Added Last Year
• Spent for Books, periodicals & Binding
• Number Assistants
• Total Salaries

In the Gerould Statistics, Books in Library was found to be the same variable as the ARL "vols", volumes held, and Added Last Year was the same as "volsadg" or volumes added, gross. These will be used here to characterize growth at these 12 libraries, while Spent for Books, periodicals & Binding was disaggregated in 1963 into expenditures for library materials (explm) and expenditures for binding (expbnd). These two were added together here to create an equivalent to the pre-1963 number for the latter years in the analysis of expenditures for materials at the 12 libraries.

By convention, the two years of the split academic year (1912/13) are often shortened to the latter year (1913) as they are here.

Other Analysis of Trends

Other published work on trends using these data include various introductions to the annual ARL Statistics, particularly those by Kendon Stubbs⁸. Other works include those by Molyneux not previously cited (1986^9,10,11 and 1994^12,13. These deal primarily with the subject of academic library growth which is a subject with a literature that goes back over a century. A a more recent examination by Martha Kyrillidou¹⁴ has many useful and current citations and deals with a broad set of trends.

Notes

¹ Robert E. Molyneux, The Gerould Statisics. Washington DC: Association of Research Libraries, 1986. The second edition is available on the Web.
² The ARL Statistics and Measurement Program page provides links to various statistical publications of the Association including the series analyzed here.
³ Kendon Stubbs and David Buxton, Cumulated ARL University Library Statistics, 1962-63 through 1978-79, Washington DC: Association of Research Libraries, 1981.
⁴Robert E. Molyneux, ACRL Academic Library Statistics, 1978/79-1987/88. The second series of ACRL data used the same forms as the ARL series. ARL included, roughly, the largest 125 university libraries (as well as non-University research libraries such as the Library of Congress, New York Public Library, and others) in the US and Canada. The ACRL series, roughly, had the next 125 or so largest university libraries in the US and Canada.
⁵ Robert E. Molyneux, ACRL/Historically Black Colleges and University Library Statistics, 1988-89 (Chicago: Association of College and Research Libraries, 1991.)
⁶ The Gerould Statisics, p. 182.
⁷ Cumulated ARL University Library Statistics, v.
⁸ ARL Statistics, annual. ⁹ Robert E. Molyneux, "Patterns, Processes of Growth, and the Projection of Library Size: A Critical Review of the Literature on Academic Library Growth," Library and Information Science Research, 8, no. 1 (January-March 1986), pp. 5-28. The oldest trend question is about library growth and there is an extensive literature on it. This article is a review of that literature.
¹⁰ Robert E. Molyneux, "Growth at ARL Member Libraries, 1962/63 to 1983/84," Journal of Academic Librarianship, 12, no. 4 (September 1986), pp. 211-216.
¹¹ Robert E. Molyneux, "Staffing Patterns and Library Growth at ARL Libraries, 1962-63 to 1983-84," Journal of Academic Librarianship, 12, no. 5, (November 1986): 292-297.
¹² Robert E. Molyneux, "What Did Rider Do? An Inquiry Into The Methodology of Fremont Rider's The Scholar and The Future of The Research Library," Libraries & Culture, 29, no. 3, 1994, pp. 297-325. Rider's influentional Scholar and the Future of the Research Library (New York: Hadham Press, 1944) made claims about the nature of academic library growth. This article deals critically with Rider's claims and his apparent method for analyzing growth trends and suggests better alternatives.
¹³ Robert E. Molyneux, "More Hortatory than Factual: Fremont Rider's Exponential Growth Hypothesis--and the Context of Exponentialism," For the Good of the Order: Essays in Honor of Edward G. Holley. Delmus E. Williams, et al. eds. (Greenwich: JAI Press, 1994), pp. 85-117. Examines all the ARL data available at the time and, I hope, ended the notion that academic libraries grow exponentially.
¹⁴ Martha Kyrillidou, "Research Library Trends: ARL Statistics" Journal of Academic Librarianship 26, no. 6 (November 2000), pp. 427-436. A "preprint" version is available on ARL's Web site. The Web edition will make it much easier to link to the various Web resources she refers to.

September 19, 2007