 |
 |
 |
Return to Contents
|
|
|
Much has been written on this subject in an attempt to understand underlying patterns and thereby, perhaps, to project future library size and anticipate library needs. Ever since Fremont Rider's well-known assertion, "It seems, as stated, to be a mathematical fact that, ever since college and university libraries started in this country, they have, on the average, doubled in size every sixteen years," [1] there have been numerous attempts to substantiate or disprove his thesis, particularly with regard to the concept of exponential growth. (Rider's assertion is equivalent to proposing a constant average annual rate of growth of approximately 4.5 percent.)
Our analysis of this aspect of library growth differs in some respects from earlier studies. [2] First, earlier studies were necessarily focused on more limited time spans than is this study, which includes annual data for years from 1912 through 1991. Rider's analysis, for example, was based on the number of volumes held at various libraries in six specific years between 1831 and 1938. A series of studies at Purdue University attempted to predict values of several variables through 1980 on the basis of data collected from 1950-51 through 1971-72. Baumol and Marcus (1973) analyzed Purdue data for the period 1950-51 to 1968-69. Other studies (Drake 1977; Wyllys 1978; Leach 1976) analyzed data from the early 1960s through the mid-1970s. [3] Molyneux analyzed data from 1962-63 through 1983-84 in an attempt to show that library growth had not been exponential during the decade of the 1970s.
A second difference has to do with purpose. In contrast to most of the earlier studies, our purpose is not to project future library growth. We do not start with the premise that there is some automatic or inexorable force that drives library growth independent of other variables, both internal to the university library system and external to it. The data themselves show how dangerous it is to make projections based solely on historical patterns. The patterns of growth that have characterized libraries in the past 30 years in no way reflect growth rates up to 1960. Moreover, we expect that changing technology will alter fundamentally the future structure of libraries---with major implications for library growth. (Part 2 of this study is devoted almost entirely to this topic.) Nonetheless, trends in library growth are useful in providing a context for examining some of the factors underlying major developments within the library world---both historical and prospective---and that is the reason for our interest in them.
The annual percentage increases for both composites can be seen most clearly by examining the bottom panel of figure 2.1, which contains the same data plotted on a semi-log scale (with, therefore, a straight line representing equal percentage increases from year to year). The two composites grew at different rates: Public 1 universities had an average annual rate of increase of 4.5 percent, whereas the collections at the Private 1 universities grew 3.2 percent per year on average. [6] It is an interesting curiosity (nothing more than that, in our view) that the 4.5 percent average annual growth rate for the Public 1 universities is precisely equivalent to Rider's assertion that libraries double in size every sixteen years. The average annual growth rate of 3.2 percent for the Private 1 universities implies a doubling of collections every 22 years. Of course, the actual annual rates of increase were not constant (the observations do not all lie exactly on the regression line). Annual growth rates for both the Private 1 and the Public 1 composites peaked in the mid- to late 1960s and then fell slowly throughout the 1970s. From about 1940 forward, the absolute size of the difference in scale between the two sets of libraries is fairly consistent, which is one reason why in much of the subsequent analysis we group the Private 1 and Public 1 libraries into a single Research 1 composite.
Once again, data are available in years before 1963 only for the libraries represented in the Research 1 composite. The growth in this measure was remarkably steady between 1912 and the late 1950s, except for the expected dips that occurred during the two world wars Figure 2.2.[small | large] Then, during the 1960s, the number of volumes added annually rose at a record rate of about 10 percent per year---more than doubling, from just under 70,000 volumes in 1960 to a peak of 148,330 in 1970. The retrenchment of the 1970s, which afflicted all of higher education, dramatically reversed this trend.
These sharp fluctuations in the annual numbers of volumes added illustrate clearly that the fortunes of the major libraries have been affected markedly, as one would have expected, by broad trends in higher education. One would expect collections to expand most rapidly when enrollments are rising rapidly, colleges and universities are expanding their offerings, and resources are relatively plentiful. And these were, of course, precisely the defining characteristics of higher education in the United States during the 1960s. [8]
The unprecedented growth in doctoral programs and in doctorates conferred was an additional source of extra pressure on libraries during the 1960s. [9] In essentially all fields of study, doctoral programs can be offered only if library resources are at least reasonably adequate, and any university considering the addition of one or more doctoral programs or even the inclusion of more subfields within an existing program must anticipate pressures for significant growth in library holdings. The number of distinct doctoral programs is a better index of pressures on a research library than is enrollment. It is hardly a coincidence that universities with large numbers of active doctoral programs are the same universities that have large---and growing---collections.
The strong interconnection between graduate education and faculty scholarship and research makes this relationship an even tighter one. Research collections designed to serve expanding graduate programs and faculty who are themselves deeply committed to scholarly and research agendas face unremitting pressures to keep growing; they cannot afford to fail to continue to build their holdings. (The implications of graduate programs for library collections are also reflected in the pronounced differences in library expenditures between universities and colleges, which we note later.)
It is hardly surprising, then, that there is a close correspondence between trends in doctorates conferred and trends in library volumes added Figure 2.3[small | large] . [10] Both doctorates conferred annually and annual volumes added gross increased rapidly throughout the decade of the 1960s before a precipitous fall in the early 1970s. The peak year for volumes added gross was a bit earlier (1970) than the peak year for Ph.D.s conferred (1972), and this slight lag is what one might expect given the duration of graduate study; most students who entered graduate programs during the mid- to late 1960s would not have received their degrees until the early to mid-1970s.
Generally speaking, the 1970s and most of the 1980s were years of retrenchment for both graduate programs and libraries. However, reductions in the annual number of volumes added gross were evidently much more modest than the reductions in doctorates conferred. The apparent asymmetry is real. It is generally easier to contemplate reductions in the sizes of entering cohorts of graduate students (in part because external factors, such as declines in the numbers of strong applications and the decreasing availability of financial aid, can be limiting factors) than it is to contemplate reductions in acquisitions. Decisions to invest in building a strong library collection in a certain field usually are---and should be---made for the long run. The figures showing year-to-year movements in volumes added illustrate well that decisions to step up the level of acquisitions are not readily reversed. There is, if you will, a kind of ratchet effect at work here, and any acquisitions policy designed to curtail the growth of a collection is likely to be hotly contested by faculty members and may well prove very difficult to implement.
There is one final observation to be made concerning the general comparison between levels of library acquisitions and numbers of doctorates conferred. The modest recovery at the end of the 1980s in number of Ph.D.s conferred was matched, at least in part, by an even more modest recovery in the annual number of volumes added gross. However, as we will attempt to explain later, other factors---including the increasingly severe fiscal problems of universities and the rapidly rising prices of serials---were important in determining the rates at which collections could be augmented.
Since the mid-1980s, the Private 1 composite has shown evidence of a recovery, with volumes added increasing each year between 1985 and 1990 (and then declining slightly in 1991). The Public 1 composite, however, has shown little evidence of any persistent recovery in the rate of acquisitions. Considerable year-to-year fluctuations continue to characterize this group of libraries, with a particularly significant drop in 1991. This pattern may result from unusually volatile funding from state governments during much of the 1980s; it illustrates the difficulty that state universities have experienced in planning their acquisitions budgets.
Over longer periods of time the curve showing annual volumes added by the Private 1 libraries has been even more volatile. During the expansion of the 1960s, available resources tended to grow faster at the Private 1 universities than at the Public 1 universities, just as they subsequently tended to fall faster. Whatever the full range of reasons for this pattern, there is no doubt that it is real. The differences just described are not due to the behavior of one or two large libraries in either the Private 1 or the Public 1 composite. The pattern of sharp increases and decreases in annual rates of acquisitions was reported by all six of the Private 1 libraries, and the top half of table 2.1[small | large] summarizes the extent to which the average growth rates for this composite differed during expansion and contraction from the average growth rates for the Public 1 libraries. [11]
This difference between private and public libraries in the magnitude of swings in rates of acquisitions breaks down, however, when we consider the Research 2 category (bottom half of table 2.1)[small | large]. In fact, between 1963 and 1970 the Public 2 composite increased its annual level of acquisitions faster than any of the other three components (12.2 percent per year). The Private 2 composite recorded the second largest increase in the rate of acquisitions during the 1960s (nearly 9 percent per year), and much of the explanation for such large growth rates at both sets of Research 2 libraries no doubt has to do with the fact that these universities were expanding at more rapid rates (especially at the graduate level) than the longer-established Research 1 institutions. The Public 2 and the Private 2 libraries began in 1963 from a base of volumes added that was about half the size of the base for the Public 1 libraries and about 40 percent of the base for the Private 1 libraries.
This proposition is difficult to test with ARL data, since members must achieve a certain scale in order to be considered for ARL membership. [12] Thus, by the time that their data are included within the ARL universe, libraries are at least reasonably well established and may have already passed through an initial period of rapid expansion. It is possible, however, to compare rates of growth for charter (1932) members of ARL [13] with rates of growth for those that joined ARL after 1956 see Figure 2.5[small | large] . The percentage growth rate has been larger for the younger libraries, as we would have expected, since this set of libraries started from a smaller average base. However, when the curve for the younger libraries is shifted to take account of its smaller base (see the dashed line on fig. 2.5), the two curves are astonishingly similar. This demonstrates that changes in the absolute numbers of volumes added have been nearly identical. The parallel patterns of expansion, modest contraction, and then a leveling off from the 1960s through the 1970s and 1980s for both sets of libraries is strong evidence that common forces (internal and external) have overwhelmed any natural stages of development related to age.
The greater volatility among the larger libraries reflected in these patterns may be due to the factors mentioned earlier in discussing the high volatility of the Private 1 composite during the 1960s and 1970s (since all the Private 1 institutions except Princeton fall into the larger category). In addition, some reduction in the rate of increase in acquisitions may have seemed more realistic for the larger libraries during the 1970s. Their holdings and their continuing rates of acquisition may have seemed large enough to permit them to withstand a period of slower growth without feeling that the integrity of the collections was being threatened. Smaller libraries may have thought that they had less margin for adjustment.
[1] Fremont Rider, The Scholar and the Future of the Research Library (New York: Hadham Press, 1944), 8.
[2] The best critical overview of this literature is Robert Molyneux's article "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 (January-March 1986): 5-28. In this article the author asserts that "library growth has not been modeled well and that no successful method of projecting growth has been developed." He cites three reasons for the failure to develop reasonable projections: (1) most writers apparently assumed that library growth occurs in one particular pattern; (2) no understanding of the processes of growth underlying the patterns was developed; and (3) the methods used by the writers did not allow for very different patterns of growth during different periods.
[3] Rider, Scholar and the Future. W. J. Baumol and M. Marcus, Economics of Academic Libraries (Washington, D.C.: American Council on Education, 1973). M.A. Drake, Academic Research Libraries: A Study of Growth (West Lafayette, Ind.: Purdue University Libraries and Audio-Visual Center, 1977). R. E. Wyllys, "On the Analysis of Growth Rates of Library Collections and Expenditures," Collection Management 2 (1978): 115-128. S. Leach, "The Growth Rates of Major Academic Libraries: Rider and Purdue Reviewed," College and Research Libraries 37 (1976): 531-542.
[4] ARL uses the following definition of a volume: "a physical unit of any printed, typewritten, handwritten, mimeographed, or processed work, contained in one binding or portfolio, hardbound or paperbound, which has been cataloged, classified, and made ready for use." Sarah M. Pritchard and Eileen Finer, comps. ARL Statistics 1990-91 (Washington, D.C.: Association of Research Libraries, 1992), 63.
[5] Nor do data for all twelve of the Research 1 libraries go back to 1912. Data on volumes added gross for the University of North Carolina and the University of Virginia begin in 1922 and 1923, respectively. To maintain consistency in the number of institutions represented in the Public 1 composite over the entire time period, we imputed data for these two institutions for the years in which data were missing. The data were imputed to reflect the pattern of growth of the average of the other four missing years. The levels of the imputed data for each "missing" institution were based on volumes added gross in the first year for which data were available for the respective libraries (1922 for U.N.C., 1923 for the U.Va.).)
[6] The average annual rates of increase were determined by fitting a least-squares regression line to the natural logarithms of the values of each curve and then determining the (constant) annual percentage increase implied by the slope of the regression line.
[7] This measure does not reflect volumes that may be lost, stolen, or deaccessioned for some other reason.
[8] For a summary of broad trends in enrollment and an analysis of changing patterns of degrees conferred, see Sarah E. Turner and William G. Bowen, "The Flight from the Arts and Sciences: Trends in Degrees Conferred," Science 250 (October 26, 1990): 517-521. For a more general discussion of trends in higher education in the 1960s and early 1970s, see Earl F. Cheit, The New Depression in Higher Education: A Study of Financial Conditions at 41 Colleges and Universities (New York: McGraw Hill, 1971).
[9] See Bowen and Rudenstine, In Pursuit of the Ph.D., especially chapters 2-5.
[10] Data on trends in the number of doctorates conferred are taken from special tabulations from the National Research Council. The average represented here is a simple average of the Ph.D.s conferred at our Research 1 universities in six fields only (English, history, economics, political science, mathematics, and physics).
[11] See Appendix Table 2.1 for detail on growth rates for individual libraries in all four composites.
[12] ARL has developed an elaborate quantitative formula based on five variables (total volumes, total staff, serials held, volumes added annually, and total expenditures) by which a library's level of activity and scale are evaluated. To be considered for membership, a library must be above a defined threshold for four years. (The threshold was based on the activity and scale of the ARL charter members. The intent of the four-year requirement is to require a long-term commitment to the development of the library by the parent institution.) The parent institution must have a minimum of 32 Ph.D. programs. Extensive narrative documentation is required, and an ARL committee conducts a site visit to the library that is applying for membership.
[13] New York University library is included in the 1932 grouping even though it was not a member of ARL until 1936.
[14] For the purposes of this analysis, the following composites were constructed, based on the number of volumes held in 1963: smaller (<=1,000,000)---Georgetown, Iowa State, Maryland, Boston, Washington State, Washington University, Michigan State; medium (1,000,000 to 2,000,000)---Florida, Southern California, Rutgers, Iowa, Virginia, New York, North Carolina, Wisconsin, Northwestern, Princeton; larger (>=2,000,000)---Chicago, Stanford, Cornell, Berkeley, Columbia, Michigan, Yale.
