 |
 |
 |
Return to Contents
|
|
|
There is always the possibility that a time series expressed in current dollars will be misleading because it will be distorted by changes in the general level of prices. In this instance, however, correcting for changes in the value of the dollar Figure 3.2 [small | large] serves mainly to reduce the average rate of increase.[3] In constant dollars, average expenditures on materials and binding increased 4.3 percent per year---a far from trivial real rate of growth.
It is revealing that the average annual rate of increase expressed in current dollars is steadier than the average annual rate of increase expressed in constant dollars.[4] This comparison supports the common-sense view that the upward pressures on the library budget have been driven primarily by forces that are not tied closely to the general price index (especially the volume of publications and the prices of those publications). Determined efforts to express every variable in constant dollars can confuse analyses of this kind.
Redoing the analysis of trends in expenditures in constant dollars does serve, however, to account for much of the difference in average rates of increase in current expenditures between the pre- and post-World War II periods noted earlier. The higher inflation rate during much of the post-World War II period undoubtedly escalated library costs, as all other costs. The average annual rate of increase in expenditures measured in constant dollars was 4.7 percent (R^2 = .88) between 1952 and 1991, as compared with 4.3 percent per year for the entire period from 1912 to the present.
This analysis also confirms an important point made earlier: the patterns characteristic of the 1960s are anomalous and not at all consistent with patterns for longer time periods. However one draws the regression lines, the observations for the 1960s are above the long-term trends (see both fig. 3.1[small | large] and fig. 3.2[small | large]). They cannot be taken as any indication of what is "normal." When we confine our analysis to the years after the 1960s, we find that expenditures on materials and binding by the Research 1 libraries have increased, on average, 8.8 percent per year in current dollars and 2.7 percent per year in constant dollars. These are much more modest rates of increase than many observers of the worlds of libraries and university finance would have expected to find, and we shall return to their meaning and interpretation later in this chapter.
If we now place on one figure Figure 3.3 [small | large] the data showing the trend in volumes added that we examined in Chapter 2 and the data showing the trend in expenditures on materials and binding expressed in constant dollars, we find that the curves first begin to diverge in the late 1950s and then to move in different directions after about 1970. From about 1970 on, expenditures on materials and binding measured in real terms continued to rise at the same time that the rate of acquisitions actually declined. (We do not present separate data for the various library composites because they all behave similarly in this regard.) In some general sense, libraries began to "pay more for less."[5] This major development is explored in detail in later sections of this study.
Over the entire period 1963-91 the overall rate of growth in total library expenditures was almost precisely the same as the overall rate of growth in expenditures on materials and binding. For the Research 1 composite, the two growth rates were 8.8 percent per year (total expenditures) and 8.79 percent (materials and binding)! For all 24 libraries the corresponding growth rates were 9.2 percent (total expenditures) and 9.12 percent (materials and binding). (This comparison implies an overall consistency in the composition of the budgets of research libraries that is correct for the materials and binding share but not correct for other components, as we shall see in Chapter 4.)
An even more interesting picture emerges when we examine average rates of change in total expenditures within each of our four composites during three subperiods: 1963-70, 1970-82, and 1982-91 (table 3.1)[small | large] . Total library expenditures rose extremely rapidly between 1963 and 1970 within all four composites (at an average annual rate of over 14 percent for all 24 libraries). Not surprisingly, library expenditures rose even more rapidly at the Private 2 and Public 2 libraries, which had lower levels of expenditures at the start of this period of expansion, than at the longer-established Private 1 and Private 2 libraries. (The respective average annual rates of increase were 13.1 percent at the twelve Research 1 libraries versus 16.3 percent at the Research 2 libraries.)
Library expenditures continued to rise faster at the Research 2 libraries than at the Research 1 libraries in both the 1970-82 and 1982-91 intervals, but the differentials were much compressed. The sharpest contraction in the rate of increase in library expenditures in the most recent period occurred within the Public 1 composite, no doubt as a result of the fiscal pressures on state budgets. The Private 1 universities experienced the slowest rate of increase in library expenditures in the 1970-82 period, presumably as a result of the severe financial pressures felt within those universities in the 1970s.
When we examine year-to-year changes in total expenditures within all 24 libraries (Figure 3.4) [small | large] , we see an abrupt change in the slope of the curve in 1970. In effect, two regimes can be distinguished: (1) the expansionary years between 1963 and 1970, when library expenditures rose at what was clearly a nonsustainable rate; and (2) the years since 1970, when library expenditures continued to rise steadily but at an average rate of just over 8 percent per year, as compared with an earlier average rate of about 14 percent. Of course, an 8 percent annual rate of increase---which implies a doubling of library expenditures every eleven years---is hardly trivial. We are reminded again of the strength of upward pressures on library budgets within each of our four composites.
The sharp rise in the libraries' share of Educational and General Expenditures (hereafter E&G) from the mid-1960s through 1971 indicates that research libraries participated more than fully in the rapid growth in university budgets that occurred during that expansionary period. On average, these libraries increased their share of E&G expenditures from under 3 percent to nearly 4 percent. (This was also the period of the most rapid expansion in doctoral education.)
During the severe retrenchment in higher education that characterized the decade of the 1970s, libraries essentially held their own with respect to share of E&G expenditures. Then, starting in 1980, the libraries' share fell every year but one during the 1980s---until it reached a low point of 3.08 percent in 1990, a level just slightly higher than the level in the mid-1960s.
Is this recent downtrend in the relative emphasis given to library budgets a function of exceptionally rapid increases in expenditures for such purposes as student aid (necessitated by a decline in the relative amount of federal funding available), student services, or central administrative functions? This proposition can be tested, at least roughly, by making use of a more narrowly defined benchmark that focuses solely on more strictly "academic" expenditures---the category called Instruction and Departmental Research on the Department of Education's survey instruments.[7]
Relating library expenditures solely to Instruction and Departmental Research (hereafter I&DR) does not change the basic pattern described above.[8] Library expenditures grew in relation to I&DR during the last half of the 1960s, held constant through 1975, and then declined---at first very sharply and then more gradually (Figure 3.6) [small | large] . In particular, we see again the same kind of steady decline during the decade of the 1980s that is evident when library expenditures are compared with all Educational and General Expenditures. (It should be noted, however, that the relative rate of decline is less rapid for this measure than for library expenditures as a percentage of E&G expenditures.) We also see stronger evidence in this figure of the emergence of a new plateau, since library expenditures have been almost constant as a percentage of I&DR from 1987 through 1990. There are other intriguing aspects of these sets of data, looked at together, but they are more relevant to a broader study of university finance during these decades than to this study of research libraries.[9]
The principal conclusion is inescapable: rather than continuing to claim a larger and larger percentage of the university budget, the typical research library has seen its share of all E&G expenditures fall steadily in recent years. The consistency of this pattern among the 24 universities included in this study is striking, as can be seen from table 3.2[small | large] , which compares the library percentages in 1979 with the comparable percentages in 1990. The plethora of minus signs in the last two columns speak for themselves. In all but 2 of the 24 universities, library expenditures declined in relation to all E&G expenditures.[10]
The same pattern is evident when we plot the year-by-year ratios of library expenditures to E&G expenditures values for each of the four library composites, showing the detail for both the 17-university data, going back to 1966, and the 24-university data from 1972 forward (Figure 3.7)[small | large] . The only general comparison to note is that the ratio falls most rapidly for the Private 1 set of libraries, which have had the highest library share in all years. More generally, we find that within each of the four composites the largest decline in the library share occurred at the library that had received the largest absolute share of its university's total expenditures in 1979 (for example, Princeton, Virginia, Northwestern, and Rutgers). Conversely, libraries that had rather low ratios of library expenditures to total expenditures in 1979 tended to see the library share decline somewhat less rapidly (though there are exceptions). The tendency for library expenditures to decline in relation to the I&DR category was also pervasive, although somewhat less consistent than in relation to E&G expenditures. In 17 of the 24 universities, library expenditures declined as a percentage of I&DR between 1979 and 1990 (table 3.3) [small | large]). The seven exceptions (Columbia, Michigan, Wisconsin, Georgetown, Southern California, Maryland, and Michigan State) were spread almost evenly among the four composites and can be explained, we believe, primarily with respect to local circumstances. Again, the plot of annual ratios for the four composites (Figure 3.8)[small | large] provides additional detail of another kind, all of which is consistent with the general pattern that we have been describing.
The 1970s and early 1980s were difficult times for higher education, and the evidence cited here suggests that retrenchment is particularly harsh on libraries. It may be easier to slow the growth of acquisitions than to take other kinds of budgetary actions such as reducing faculty positions, laying off staff members, and reducing financial aid. We do not believe, however, that this is the full explanation for the failure of the library to maintain its share of total expenditures. In our view other developments, including technological changes and steep increases in the prices of serials, led to changed attitudes toward library expenditures and their long-term place in the university budget---differences in fundamental assumptions as to what was both desirable and sustainable. These themes are developed in subsequent chapters. Before turning to these topics, however, we will provide a brief postscript to this chapter's analysis of trends in expenditures in the form of a comparison of research libraries with college libraries.
That ratio of 8:1 or 9:1 can be considered a rough upper bound of what appears to be required at the level of library investments if an institution is to commit itself to doctoral programs in a large way. We say "upper bound" because research universities also enroll many more students than do selective liberal arts colleges, and some of the difference in library expenditures noted earlier is simply a matter of scale. Also, research universities usually maintain highly expensive professional libraries in fields such as medicine and law.
It is instructive, therefore, to compare library expenditures at Princeton (which is primarily an arts-and-sciences university, with no professional schools of law, medicine, business, or education, and a total enrollment of about 6,000) with library expenditures at Oberlin and Smith, which have strong libraries and enrollments roughly half as large as enrollment at Princeton. Annual library expenditures at Princeton are four to five times greater than library expenditures at these two colleges, and this ratio is perhaps more indicative of the direct effects of doctoral education and the other features of a research university. If we were to make a generous allowance for differences in enrollment, the ratio of library expenditures at Princeton to library expenditures at Oberlin or Smith might be reduced to something like three or four to one (since no one believes that library outlays need to increase in proportion to the number of undergraduate students). The implications of research-university status for library expenditures are obvious.
While the absolute level of library expenditures is far higher at a research university than at a liberal arts college, the share of library expenditures in the overall budget of the institution is significantly higher at the typical liberal arts college. (See Appendix Tables 3.1-3.5.) This finding may come as something of a surprise to those who thought that doctoral education and a heavy emphasis on research would require disproportionate investments in library resources. The apparent explanation is that libraries at liberal arts colleges as well as at universities entail heavy fixed costs (subscriptions to a core set of journals and reference materials, for example), and the far larger enrollments at research universities permit these fixed costs to be spread over larger numbers of students. In short, there may well be substantial economies of scale.
Again, it is instructive to examine the case of Princeton, where differences in enrollment as compared with the liberal arts colleges, while still significant, are less overwhelming than at most other research universities. Even with a larger enrollment (roughly three times the enrollment of the average liberal arts college included in this analysis), Princeton allocated a larger percentage of instruction and departmental research expenditures to the library than did any of the fifteen liberal arts colleges---22 percent in 1990 at Princeton versus a high for the colleges of about 20 percent at Bryn Mawr and Haverford.[12] (Bryn Mawr, of course, has important doctoral programs and is in that sense something of a misfit among the liberal arts colleges.) The average for the fifteen colleges was about 15 percent. Princeton also allocated a slightly higher percentage of all educational and general expenditures to the library (5.4 percent in 1990) than did the typical college (5.1 percent in 1990). In this instance, economies of scale were insufficient to outweigh the additional expenses associated with a heavy emphasis on doctoral education and research. Research universities apparently have to be larger than Princeton before scale effects predominate.[13]
A primary reason for examining library expenditures at colleges is to see if their trends mirror those at the research universities.[14] Overall, library expenditures at the fifteen colleges increased by a factor of about 3.2 between 1977 and 1990, for an average annual growth rate of approximately 9.4 percent per year. This is just slightly higher than the growth rate for library expenditures within the research universities (about 8.7 percent per year).
A more revealing comparison is of trends in the library's share of all educational and general expenditures. As in the case of the universities, library expenditures at the colleges have declined in recent years as a percentage of both E&G and I&DR expenditures (Figure 3.9)[small | large] . Thus, in this sector too we see that, contrary to some popular wisdom, libraries have not absorbed a larger and larger share of the budgets of institutions of higher education.
The rate of decline in the library share of budgets at the colleges has been very modest, however, and does not appear to have become pronounced until about 1985 and thereafter. Between 1979 and 1990 the typical college in our set saw its library expenditures decline from 6.0 percent to 5.1 percent of E&G expenditures and from 16.4 percent to 15.9 percent of I&DR expenditures. The comparable figures for university libraries (tables 3.2[small | large] and table3-3[small | large]) indicate that library shares of university budgets fell somewhat more than library shares of college budgets.
[1]As we will see in Chapter 4, expenditures for materials have remained a fairly constant share (roughly one-third) of total expenditures between 1963 and 1991. Although binding expenditures declined throughout that same period, they are a very small percentage of all expenditures on materials and binding. Data for total expenditures (including salaries and other operating expenditures) are available on a consistent basis only from 1963 forward.
[2] We limit this analysis to the Research 1 libraries since data for the other libraries included in this study are available only for more recent periods. The R^2 is 0.97.
[3] We use the GNP deflator as our general index of price changes. The Consumer Price Index is not really relevant to library expenditures, and the Higher Education Price Index (HEPI, which should be a somewhat more refined measure) is not available before 1960. A comparison of the GNP deflator and the HEPI after 1960 shows that the movements of the two indices are quite similar in any case.
[4] The R^2 is 0.93 for the regression line in figure 3.2[small | large] , which measures the trend in expenditures in constant dollars, as contrasted with an R^2 of 0.97 for the regression line in figure 3.1[small | large] , which measures the trend in expenditures in current dollars.
[5] Moreover, Donald W. Koepp, university librarian at Princeton, has suggested in a conversation with Anthony Cummings that the significance of these trends may be even more dramatic at some institutions than the raw data suggest. Because of efficiencies resulting from the automation of the cataloging function described in the next chapter, arrearage in cataloging has been greatly reduced. Accordingly, rates of increase in volumes added gross may be said to be artificially higher for the period since the automation of the cataloging function, since a larger proportion of volumes acquired has been cataloged expeditiously.
[6] We use the U.S. Department of Education survey data for Educational and General Expenditures but continue to rely on the ARL data for library expenditures. For sources and definitions of the Education Department data, see Appendix A. In brief, Educational and General Expenditures include academic and administrative expenditures of all kinds, including sponsored research, maintenance of the plant, and student aid; they exclude such auxiliary activities as dormitories and food services. One advantage of using this broad grouping of expenditures is that it is less influenced than some of its components by changing definitions of the boundaries between subcategories. The percentages in figure 3.5[small | large] are averages of the percentages for the seventeen libraries for which consistent data could be obtained from 1966 through 1990. The 7 libraries (of the total group of 24 included in this study) for which these data could not be calculated back to 1966 are Berkeley, Columbia, Cornell, Maryland, Michigan, Rutgers, and Wisconsin. The missing data are almost always the Educational and General Expenditures. In the cases of a few other libraries, we had to interpolate figures for one or two years to obtain a consistent time series. These percentages can be computed for all 24 libraries for years from 1972 forward, and the 24-university averages for the 1972-90 period are very similar to the 17-university averages for the same years, except that the 24-university averages are slightly higher in absolute terms (3.20 percent of Educational and General Expenditures in 1990, as compared with 3.08 percent for the 17-university average). With very few exceptions, the year-to-year changes are nearly identical.
[7] This expenditure category consists mainly of the basic budgets of the academic departments and is therefore unaffected by changes in student aid, student services, plant maintenance, and administrative costs. Sponsored Research is another category, also separate from Instruction and Departmental Research, but it is less independent in that gains and losses in Sponsored Research funding can have significant effects on Instruction and Departmental Research by shifting portions of some salaries onto (or off) the regular departmental budget.
[8] For convenience, we express library expenditures as a percentage of expenditures on Instruction and Departmental Research, even though library expenditures are not a component of the I&DR category. This is simply one way of calculating a ratio, thereby scaling the data.
[9] The major difference between figures 3.5[small | large] and 3.6[small | large] occurs during the period 1974-76, when library expenditures remain essentially constant as a percentage of E&G but decline sharply as a percentage of I&DR. The proximate explanation has nothing directly to do with library expenditures; rather, the I&DR share of total university expenditures rose markedly between 1974 and 1976. This may be nothing more than a statistical artifact, related to some changes in the HEGIS forms and reporting in that interval. The rapid increase in the Consumer Price Index during these years could also be part of the explanation, since universities felt strong pressure to do as much as they could for salaries of faculty and staff, which make up a disproportionately large part of the I&DR category.
[10] However, at least one component of library expenditures---serials subscriptions---rose appreciably more rapidly than E&G expenditures, especially in the last decade or so. Kendon Stubbs has found that serials expenditures have even kept pace with national expenditures on research and development, which have grown much faster than E&G expenditures (Letter to Richard Ekman, October 2, 1992). Chapter 4 contains a fuller discussion of shifts in the components of library expenditures.
[11] The fifteen colleges included in this analysis are Bowdoin, Bryn Mawr, Carleton, Davidson, Grinnell, Haverford, Middlebury, Mount Holyoke, Oberlin, Pomona, Reed, Smith, Swarthmore, Trinity, and Vassar. These fifteen were chosen in part because data were available for them on a reasonably consistent basis from 1977 through 1990. In the case of these institutions, none of which are members of the Association of Research Libraries, data for both library expenditures and other expenditures come from the HEGIS/IPEDS surveys. As a result, we have information on total library expenditures only, and these data are somewhat less reliable than the comparable data collected by ARL for the research universities.
[12] At Oberlin, Total Library Expenditures were also approximately 20 percent of instruction and departmental research expenditures in 1990. However, the value for total library expenditures for this year was imputed, so we do not have as much confidence in its accuracy.
[13] It would be interesting and worthwhile to carry out a more detailed analysis of the relationship between enrollment and library expenditures within both the research universities and the liberal arts colleges in an effort to understand better the shapes of the relevant cost curves. However, even cursory inspection of the data collected for this study demonstrates how difficult it would be to control for other important variables such as the number of programs of study, the research emphasis of the institution, the balance between the humanities and the sciences, and the wealth (and perhaps the age) of the institution. In any case, these scale effects may well change markedly as more use is made of some of the new technologies discussed in Part 2 of this study.
[14] For another analysis of college libraries as compared with university libraries, see Richard Hume Werking "Collection Growth and Expenditures in Academic Libraries: A Preliminary Inquiry," College & Research Libraries 52 (1991): 5-23. The author uses data primarily from the Bowdoin List for college libraries and ARL for university libraries and finds, as we do, that the college library expenditures grew more rapidly than the university library expenditures between 1977 and 1987. (Our analysis extends to 1990.) He also analyzes both college library collections and components of expenditures within total college library expenditures.
