Allen.indd


           
        
            
        
         
            
         
          

           
             

           
            
           
          

             
         

              
     

     
   

    
    

     
        

       
         

      
     

     
     

      
        
       

     

Toward a Formula-Based Model for 
Academic Library Funding: Statistical 
Significance and Implications of 
a Model Based upon Institutional 
Characteristics 

Frank R. Allen and Mark Dickie 

This study tests the hypothesis that a positive relationship exists between 
academic library funding (dependent variable) and selected institutional 
variables taken as indicators of the demand for library services at the 
university (enrollment, number of doctoral programs, doctoral degrees 
awarded, number of faculty, select other institutional characteristics). The 
research employs 11 years of longitudinal data from 113 members of the 
Association of Research Libraries to create a multiple regression model. 
Empirical results indicate that operational indicators of the demand for 
library services are positively associated with funding, and most of the as-
sociations are statistically significant at the five percent level or less in two 
tail tests.In a corollary finding, libraries associated with private universities 
in the United States spend 21 percent more than their public counterparts, 
while Canadian university libraries spend 21 percent less than U.S. public 
university libraries.The presence of a medical school is associated with 
an 8.6 percent greater expenditure, and the presence of a law school is 
associated with a 12.3 percent greater expenditure.The study suggests 
that this formula may be useful as a tool for library funding and assessment 
of adequacy of library budgets. 

any academic libraries in the 
United States have tradition-
ally experienced funding in 
an incremental or haphazard 

fashion. This served libraries adequately 
until the la er two decades of the 20th 
century. The challenge in recent years has 
been to find a model to support the sea 

change of issues facing academic libraries, 
including serials inflation, new technology 
initiatives, growth in academic degree 
programs, and changing usage pa erns. 
Models that may have worked decades 
ago and models that may be effective in 
our universities at large no longer serve 
libraries. How does the 21st-century 

Frank R. Allen is Associate Director for Administrative Services, University of Central Florida Librar-
ies, and Mark Dickie is Professor, Department of Economics, College of Business Administration, at the 
University of Central Florida; e-mail: fallen@mail.ucf.edu and Mark.dickie@bus.ucf.edu, respectively. 

170 

mailto:Mark.dickie@bus.ucf.edu
mailto:fallen@mail.ucf.edu
http:tests.In


      
      

      
       

        
   

    
      

       
     

      
     

     
     

       
       

        
      

       
       

       
      

        
       

       
     

          
  

        
     

      
      
    

     
   

      
 

     
      

       
       

       
       
         

      
       

      
        

      
        
      

        
       
       

       
       

       
       
      

        
      

       
       

    
       
       
    

       
       

      
      

      
       

         
      

       
      

      
      
      

         
      
     

        
     
     

        
     

     
       
      

Toward a Formula-Based Model for Academic Library Funding  171 

academic library petition for funds? What 
new funding models can libraries employ? 
What benchmarks are used to determine 
if a library is adequately funded? Where 
can we go to identify other models that 
might serve academic libraries? 

Subjective funding mechanisms for 
libraries can impair short- and long-term 
planning. How can a library director or col-
lection manager make informed resource 
allocation decisions when there is li le 
predictability for the materials budget? 
Funding uncertainties are exacerbated by 
the explosion of published literature, pack-
aged deals that encourage libraries to buy 
what they do not need, and redundancy 
of content format. It is impossible to know 
when the library is adequately funded 
because there is no defined target toward 
which to aim. Libraries may benefit from 
an element of objectivity worked into the 
funding mechanism, one that would center 
the task of allocating dollars to the library. 
The library might then be more resigned 
to going about the business of allocating 
its scarce economic resources internally, 
knowing that at least it got its share of the 
university pie. 

It is the thesis of this study that the 
library’s budget should be more program 
driven and less susceptible to the econom-
ic and political vagaries of the institution. 
A formula-based budget may provide 
more stability and render the library less 
vulnerable to disproportionate cuts. A 
program-driven budget may reinforce 
the notion of the library as an academic 
as opposed to an administrative campus 
unit. Many libraries do not have much 
traction when negotiating for funding. A 
program-driven approach may build in 
growth drivers based upon the health of 
the institution as a whole. The economic 
well-being of the library would rise and 
fall along with the larger university. 

Literature Survey 
Few published studies address funding 
mechanisms for the academic library as 
a whole. It is important to distinguish 
between models for funding of the entire 

library (the focus of this study) versus 
models for the allocation of the materials 
budget, not the focus of this study. In a 
1992 article, David Baker revealed findings 
from a survey on resource allocation in 
university libraries in the United Kingdom. 
His major conclusions were that there is a 
movement toward using formulae and unit 
costs as the basis of allocation of central 
(state) funds to universities; however, there 
is li le sign that such approaches are being 
adopted by universities once the funds are 
received. He found virtually no evidence of 
a formula approach to the method of fund-
ing university libraries.1 In a 1993 article, 
Frank Goudy presented data from 1971 to 
1990 to show that the ACRL Libraries stan-
dard that the library’s appropriation shall 
be 6 percent of total institution budget has 
never been realized. According to Goudy, 
the “6% rule” was wishful thinking that 
never existed.2 The authors note that ACRL 
subsequently dropped this standard, rely-
ing on a more generalized statement that 
the library budget should be appropriate to 
the library’s objectives. (See: www.ala.org/ 
ala/acrl/acrlstandards/standardslibraries. 
htm.)3 Kendon Stubbs lamented in a 1994 
study the decline in research libraries’share 
of university funds. Based upon survey 
data from 88 ARL libraries, the percent-
age of university E&G funds allocated 
to these libraries fell steadily from 3.91 
percent in 1982 to 3.32 percent in 1992. He 
postulated several causes for this decline, 
but did not discuss funding formulas.4 Rolf 
Griebel discussed in a 1995 examination 
of German university libraries a financial 
approach known as the “Bavarian Model 
Budget.” The model uses a methodology 
that identifies the tier to which a library and 
university belong and then specifies what 
percentage of newly published literature 
a library in this tier should purchase. He 
concluded that German libraries are un-
derfunded by approximately 30 percent, 
according to this model.5 In a 1996 opinion 
paper, Murle Kenerson criticized formulas 
that the Tennessee Higher Education Com-
mission (THEC) uses for the distribution of 
monies to support higher education in Ten-

http:www.ala.org


 

     
    

    
      

     
        

      
     

     
      

     
      

       
        

      
       

       
       

        
       

     
       

     
        

     
     

       
        

     
  

      
      

      
         

      
       

       
       
      

      
     

     
     

      
       

    
     

       
      

     
       

     
     

    
       

     
       

     
      

      
       
       

      
      

      
    

      
      

         
      

        
        

       
       

       
       

     
         

    
       

       
        

      
        

       
      

       
        

        
    

       
     

       
       

       
      

         
      

       
        

     
       

172 College & Research Libraries March 2007 

nessee, including academic libraries. The 
formulas incorporate full-time equivalent 
(FTE) enrollment and performance-based 
approaches using student test scores and 
placement of graduates. Kenerson argued 
that FTE may be appropriate for funding of 
classroom instruction; but, at the author’s 
home institution of Tennessee State Uni-
versity, “half-time” students may require 
as many library resources as full-time 
students. “Grade point averages, improved 
test scores and similar mechanisms are 
of li le consequence in assessing the real 
value of library services and in securing a 
fair share of performance funding dollars.” 
Kenerson found that even when the library 
was allocated a certain stated amount of 
funding in the university budget, the full 
sum was never received by the library.6 In 
a 1999 article Jin-Chuan Ma argued that 
Chinese university libraries are slipping 
behind in their ability to support their in-
stitutions’ research agenda. Ma suggested 
that, in addition to receiving 5 percent of 
the university’s operational budget, the 
current model, university libraries should 
also receive a percentage of the university’s 
research and grant funding, to be used for 
acquisition of information resources to 
support that research.7 

Libraries are somewhat of a microcosm 
of a university, with large staffing, opera-
tional, and program-driven costs. It may 
be of interest to look at the literature on 
funding mechanisms of universities as a 
whole. The picture here is also somewhat 
muddled. In a 2002 article, Peter Facione 
likened the fiscal process of most American 
colleges and universities to the “controlled 
economy approach” used by the former 
Soviet Union. Central commi ees project 
income and expenses, establish strict 
guidelines for expenditures of resources, 
and take back unexpended balances at 
year’s end. This approach does li le to 
reward strategic planning, program re-
view, and new program development.8 
In a 1985 paper, Jim Tolhurst discussed 
resource allocation and budgeting in U.K. 
and Australian universities. He reported 
on an earlier U.K. study that summarized 

university funding practices as routinely 
employing: (1) li le relationship between 
long-term objective and resource alloca-
tion; (2) allocation of resources based upon 
incremental allocation from an historical 
base; (3) a division of resources between 
the academic and nonacademic sectors 
based upon historical cost-share basis; (4) 
resource allocation that does not appear 
to take into account the relative strengths 
of departments.9 In a 2003 article, Nancy 
Cantor and Paul Courant discussed recent 
trends in higher education finance that 
carry disquieting consequences. The use of 
“bo om-line departmental budgeting” in 
which units retain the revenues associated 
with their activities, but are also responsi-
ble for the costs of those activities, works to 
the detriment of university “public goods,” 
a prime example of which is the university 
library. Public goods such as the library are 
vulnerable under this model because it is 
o en difficult to establish an exact return 
on their investment, yet their presence is 
thought to have a significant impact on 
the intellectual and cultural well-being 
of the community as a whole.10 In a 2003 
paper, Gunapala Edirisooriya advocated 
an entirely new approach to funding of 
higher education in the United States. The 
author’s premise was to create a reserve for 
higher education by reducing the amount 
of funding for higher education by X% per 
year, beginning far enough out to allow 
universities to prepare, and increasing this 
percentage each year until a sizable reserve 
is established. The reserve is then used to 
provide a stable source of revenue for both 
the state and higher education.11 

In summary, there appear to be few 
published models for academic library 
funding. Furthermore, of the few to be 
postulated, some of those have not been ap-
plied. At least two studies have commented 
upon the decline of academic library fund-
ing in the United States as a percentage of 
university funding, which one might argue 
gives further impetus for this study. The 
literature also reveals a concern at a larger 
level over university funding mechanisms, 
suggesting that most are short on objective 

http:plied.At
http:education.11
http:whole.10


       
      

  
       

        
     

  

 
       

    
     

    
    

     

    
 

      
     

      
        

     
       
      

     
      

     
     

        
      

      
      

        
        

          
 

     
  

     
     

       
      

       
      

    
     
     

     

   

     

    

    
   

         
       

       
     

       

       
       
    

      
       

      
     

       
        

   
     

        
       

      
        

     
        

       
         

        
       

       
 

 

Toward a Formula-Based Model for Academic Library Funding  173 

rigor and long on political influence or 
incrementalism. One example of a novel 
budgeting technique, responsibility-cen-
tered management, is a move in an interest-
ing direction but denudes the library to an 
administrative support function devoid of 
academic mission. 

Hypothesis and Methodology 
The purpose of this study is to illustrate 
a model in which funding of an academic 
library depends on institutional char-
acteristics. The model assumes that the 
funding an academic library receives is 
influenced partly by the demand for the 
library’s services from the university’s 
students, faculty, and programs. One 
reason for adopting this assumption is 
to explore the plausibility of a model that 
relies on variables external to the library 
as drivers of funding. This represents a 
significant departure from arguments that 
the library’s funding, or more specifically 
petitions for increases in funding, be pred-
icated upon its existing size, holdings, 
serials inflation—in essence, inertia. 

To implement the model empirically, we 
turn to the annual statistical survey of the 
Association of Research Libraries (ARL) 
as the source for operational indicators of 
the theoretical construct “demand for the 
library’s services from students, faculty, 
and programs.” Specifically, we use X1 
= undergraduate enrollment to indicate 
demand from undergraduate students; X2 
= graduate enrollment and X3 = number of 
Ph.D. degrees awarded annually to reflect 
demand from graduate students; X4 = 
number of teaching faculty to indicate de-
mand from faculty; X5 = number of Ph.D. 
fields, X6 = presence of medical school (1 
if present, 0 if not), and X7 = presence of 
law school (1 if present, 0 if not) to reflect 
demand from graduate and relatively 
expensive professional programs of the 
university. Although these variables are 
imperfect measures of the theoretical 
construct and, as discussed in more detail 
later, there are other potential determinants 
of library funding that might usefully be 
considered in future research, the variables 

selected represent measurable indicators 
of institutional characteristics related to 
the demand for library services. 

Our research hypothesis is that a 
positive relationship exists between total 
library expenditures (dependent vari-
able – Y) and the independent variables 
just described (X1 through X7).We use 
multiple regression to quantify this rela-
tionship. Recognizing that there may be 
systematic differences in funding between 
private and public institutions in the U.S. 
and Canadian institutions, we include 
two additional qualitative independent 
variables: X8 = U.S. private university (1 
if U.S. private, 0 if not) and X9 = Canadian 
university (1 if Canadian, 0 if not). Finally, 
we include a time trend in the regression 
as a rough way of partially accounting for 
factors outside the university that may 
increase the costs of meeting a given level 
of demand for the library’s services (such 
as serials inflation or rising salaries). 

To test the hypothesis we analyze 11 
years of data from 1992 through 2003 for 
113 libraries that are members of the ARL, 
yielding a total of 1,190 observations. ARL 
libraries not examined include nonuniver-
sity libraries and a small number of insti-
tutions that were dropped due to missing 
data. The ARL data set is selected because 
this is a widely recognizable and complete 
longitudinal data set that represents a 
relatively homogeneous population. 

We regress the natural logarithm of total 
real library expenditures (Y) on the logs of 
undergraduate enrollment, graduate en-
rollment, number of teaching faculty, num-
ber of Ph.D. fields and number of Ph.D.s 
awarded annually. The use of a logarithmic 
rather than linear model allows the esti-
mated marginal impact of a change in an 
independent variable to diminish. For ex-
ample, the impact on cost from adding the 
90th Ph.D. program is probably less than 
the impact on cost of adding the 20th. Also 
included in the regression are a time trend, 
a constant term and indicator variables for: 
presence of a medical school, presence of 
a law school, public/private status, and 
Canadian/American affiliation. 



 

       
    

     
        
    

      
     

        
      
   

     
      

      
      

     
       

      
   

     
        

       
        

      
      
 

    
     

     
     

     
    

        
      

    
    

     
     

    
     

    
     
    

      
   

    
   

      
    

    
  

       
       

      
    

      
      

     
       

       
      
      

     
      

      
      
       

       
      

         
      

       
     

      
     

      

 

174 College & Research Libraries March 2007 

TABLE 1
 Coefficient of Measurables 

Independent Variable 

Regression 
Coefficient 
(t-ratio) 

Log of Undergraduate Student Population 0.037 (1.908) 

Log of Graduate Student Population 0.071 (4.284) 

Log of Number of Ph.D.s Awarded 0.049 (3.511) 

Log of Number of Ph.D. Fields Offered 0.004 (0.291) 

Log of Number of Faculty 0.071 (6.051) 

Medical School (=1 if present, 0 if not) 0.086 (5.319) 

Law School (=1 if present, 0 if not) 0.123 (5.421) 

Private (=1 if U.S. private, 0 otherwise) 0.212 (3.773) 

Canadian (=1 if Canadian, 0 otherwise) –0.216 (–2.882) 

Trend (=1,2,…,11 by year) 0.024 (34.613) 

Constant 14.637 (60.837) 

Lagrange multiplier test vs. OLS 4,170.88 

R-squared 0.489 

Sample Size 1190 

Note: The dependent variable is the natural logarithm of 
total real expenditures. 

A “random effects” model is used to 
account for unobserved library-specific 
factors that persist through time. Ex-
amples of such factors would be a large 
library endowment, costly special collec-
tions or archives, multiple branches, or 
any other unmeasured variable specific 
to a library with an ongoing impact on 
spending. By accounting for effects of 
persistent, unobserved library-specific 
factors, the random effects model rec-
ognizes that the observations are not 
all independent, since each library is 
observed repeatedly over the years. This 
feature allows for more efficient estima-
tion (that is, lower standard errors) than 
would be obtained by estimators that 
ignored unmeasured library-specific 
factors. Intuitively, the model assumes 
that the funding response to a change in 
an independent variable is the same for 
all libraries, but the base level of funding 
may vary due to library-specific factors. 
The model is estimated by generalized 
least squares. 

Results 
The estimated model is sum-
marized in table 1. Coefficients 
of the operational indicators of 
demand for library services take 
the expected positive sign, and 
most are statistically significant 
at the five percent level or less in 
a two-tail test. The t-ratio for un-
dergraduate enrollment is 1.908, 
slightly below the five-percent 
critical value, and there appears 
to be no significant association 
between funding and the num-
ber of Ph.D. programs a er con-
trolling for other independent 
variables. The model produces a 
coefficient of determination (R2) 
of .489, indicating a modicum of 
correlation between the inde-
pendent variables and library 
funding. A “Lagrange multi-
plier” test of the random effects 
model against an ordinary 
least-squares model that ignores 
persistent library-specific fac-

tors yields a chi-square test statistic with 
one degree of freedom of 4,171, providing 
strong support for the importance of ac-
counting for library-specific effects. 

The coefficient of a logarithmic variable 
measures the estimated “elasticity,” or the 
percentage change in total real expendi-
ture associated with a one percent change 
in the independent variable. All of the 
estimated elasticities are well below unity, 
indicating that one percent changes in 
independent variables are associated with 
much smaller than one percent increases 
in library funding. For example, the larg-
est elasticity estimates of .071 indicate 
that a one percent increase in graduate 
student enrollment or in the number of 
teaching faculty is associated with about 
a 7/100 of one percent increase in total real 
expenditure. That these estimates, as well 
as the elasticity for Ph.D.s awarded, are 
larger than the elasticity for undergradu-
ate enrollment should not be surprising 
given the relatively greater investment 
in library resources required by research 



 
     

     
      

      
        

     
        

    
     

       
       

     
       

      
      

      
    

      
      

       
     
       

  

    
     

     

      
     

   
     

    
 

        

     
       

      
 

     

      
       

   

    

      

 

Toward a Formula-Based Model for Academic Library Funding  175 

and graduate education. Coefficients of 
indicator variables when multiplied by 
100 approximate the percentage change 
in total real expenditure associated with 
presence of the indicator. Thus, presence 
of a medical school is associated with 8.6 
percent greater expenditure, and presence 
of a law school is associated with 12.3 
percent greater expenditure. Libraries 
associated with private universities in 
the United States spend about 21 percent 
more than their public counterparts in the 
United States, while Canadian university 
libraries spend about 21 percent less on av-
erage than U.S. public university libraries. 
The coefficient of the trend variable repre-
sents the average year-to-year growth in 
expenditures when holding all indepen-
dent variables constant and reflects the 
influence of factors like serials inflation 
or real increases in salaries. According to 
the model, expenditures increase on aver-
age by 2.4 percent annually, holding other 
independent variables constant. 

Researchers seeking a more parsimo-
nious model might consider whether all 
three of the independent variables mea-
suring graduate enrollment, number of 
Ph.D. degrees awarded and number of 
Ph.D. fields should be included, as they 
correlate highly with one another. Pearson 

correlation coefficients between pairs of 
these variables range from 0.65 to 0.76. 
While Pearson correlation is useful for 
assessing linear relationships between 
pairs of variables, the “condition num-
ber” of the data matrix indicates whether 
more general linear relationships between 
multiple variables are problematic.12 The 
condition number computed for the data 
matrix (the ratio of the largest to the small-
est characteristic root of the normalized 
cross-product matrix) is 10.59, well below 
the value of 20 that Belsley, Kuh, and 
Welsch suggest as indicating a potential 
collinearity problem. 

Application to a Specific Institution 
Table 2 shows an application of the for-
mula to an actual set of data for a rapidly 
growing university in the south that is 
not an ARL member. Based on changes 
in the independent variables, the formula 
produces a relatively modest increase of 
$364,903 in library funding. There is, how-
ever, also an underlying rate of growth in 
university library budgets independent 
of growth in these campus independent 
variables. This trend growth, produced 
from the aforementioned trend variable, 
is what the library would have experi-
enced on average with no growth in the 

TABLE 2 
Application of Model to a University 

Trend 0.024 

Base: $10,000,000 

Year 
2002 

Year 
2003 

% 
Change X100 Elasticity 

% 
Impact 

Change 
in 

Funding 

Faculty 976 1,050  0.076 7.58 0.0710% 0.0054 $53,832 

Undergrads 22,054 25,799  0.170 16.98 0.0370% 0.0063 $62,830 

Grad Students 2,066 2607  0.262 26.19 0.0710% 0.0186 $185,920 

Number Ph.D.s 87 97  0.115 11.49 0.0490% 0.0056 $56,322 

Ph.D. Fields 20 23  0.150 15.00 0.0040% 0.0006 $6,000 

Total change $364,903 

Trend growth $240,000 

Overall change $604,903 

As percent 6.05% 

http:problematic.12


 

    
     

     

        
      

         
      

       
       

        
        

       
       

     
         

   
        
       

     
     

        
      

      

      
       

     
       

      
  

    

     
        

 

    
       
     
        

      

    

 

176 College & Research Libraries March 2007 

FIGURE 1 
Actual and Model Expenditures 

independent variables. In the example 
above, the increase from applying the 
trend variable yields $240,000. The total 
increase suggested from the formula is 
$604,903, a 6.05 percent rate of growth. 

Application to Expenditure Growth 
over Time 
A second application of the model is to 
compare the time path of actual expen-
ditures of a library or set of libraries to 
the time path of expenditures predicted 
by the model. Figure 1 illustrates this 
type of comparison for the 99 libraries 
having complete data for each of the 11 
years. The solid line plots the average of 
the 99 libraries’ total real expenditures in 
each year, while the dashed line shows 
the average real expenditure predicted 
by the model in table 1. To simplify the 
comparison, the model-predicted expen-
diture is adjusted so that it matches the 
actual expenditure in the first year. As 
shown, the average library experienced 
erratic transitory swings in its expendi-
ture growth rate, namely the dip in the 
mid 1990s and the pronounced slowing 
in year 2002–2003. In contrast, a hypo-

thetical library with the same starting 
level of expenditures in 1993 but funded 
according to the model (represented 
by the dashed line) would experience a 
much steadier and more predictable rate 
of budget growth. 

Application across Libraries 
The model can be applied in a third way 
by comparing a library’s actual spending 
against the hypothetical spending that 
the model suggests. Table 3 provides this 
comparison for the ARL libraries that 
formed the data set for this study. The Ac-
tual column shows the library’s average 
annual total expenditures, as reported in 
the ARL statistics, for the 11-year period, 
in 2003 U.S. dollars. The Model column 
shows the library expenditures predicted 
by the model, based upon the averages of 
the independent variables for that institu-
tion for the 11 years of data. The difference 
between these two produces an Over Pre-
dicted or Under Predicted amount based 
upon the methodology of the model. To 
argue that these terms are synonymous 
with “overfunded” and “underfunded” 
would be presumptuous; however, the 



      
     

       
      

       

      
     
     

        
      

       
      

      
     

      
       

      
      
      

       
       

       
      
        

       
   

     
      
   

   

      

     

      

      

      
       

     
    

  
 

  
       

     

     
     

   
      

     
      

      
       

         
      

      
      
     

       
      

        
           

    
         

      
    

        
      

     
     

Toward a Formula-Based Model for Academic Library Funding  177 

suggestion is worth exploring on a com-
parative basis. Static results that are of 
limited value for one institution might 
be useful in comparing a library against a 
cohort group of peers. For example, three 
of the ten libraries with the largest excess 
of actual over model-predicted funding 
are at Ivy League institutions (Harvard, 
Yale, Princeton) and two are in the UC 
system (Berkeley and UCLA). Only one 
Ivy League library (Brown) and two of 
seven of the included UC-system libraries 
(Davis and Riverside) have actual funding 
below model-predicted funding 

Limitations to the Model 
There are a number of limitations to this 
research. There is no separate enumeration 
of independent sources of funding such as 
endowments, or of important cost drivers 
such as archives, special collections, and 
multiple library branches (although the 
net effect of all persistent library-specific 
factors is implicitly accounted for in the 
random effects model). The analysis does 
not account for the myriad of preserva-
tion and digitization efforts underway in 
libraries. The model may be more useful 
for growing institutions and less useful for 
institutions in a steady state of existence. 
Last, the formula does not explicitly esti-
mate the impact of serials inflation but only 
accounts for it indirectly through the trend 
variable.Practical applications of formula-
based allocation schemes should account 
more directly for the funding necessities 
born from serials inflation. 

Further Study 
If the goal were to make this model truly 
practical and more applicable, one might 
consider other possible independent 
variables. Libraries can argue that there 
should be a correlation between research 
grant money flowing into the institution 
and the library’s budget. A more ambi-
tious extension would be to measure the 
amount of scholarly productivity for the 
universities under study and use this to 
create an independent variable that would 
factor the number or quality of scholarly 

publications into the model. Both of these 
ideas suggest that the greater the univer-
sity’s rate of research and publication, the 
more money should flow into the library 
to support research. This approach turns 
the notion of the size of the library as a 
measure of its “goodness” on its head. 
At least one study in the library litera-
ture suggests that a university library’s 
collection helps shape the university’s 
reputation. Lewis Liu provides empirical 
evidence that the library’s ARL ranking 
correlates closely with its U.S. News and 
World Report ranking.13 This may be true; 
however it should be noted that U.S. News 
and World Report includes library funding 
per student as a metric in its ranking of 
colleges and universities. In any case, the 
extensions to the analysis proposed here 
suggest the opposite relationship: that a 
university that is demonstrating growth 
in research activity should support its 
faculty by building a stronger library 
through increased financial support. 
This is a subtle but important distinc-
tion, which moves the debate away from 
measuring the goodness of the library by 
input measures (volume count, number of 
serials, expenditures, etc.) to a model that 
suggests that the library should be strong 
to reflect the level of research taking place 
at the university. 

Conclusion 
The purpose of the study is not to create 
a one-size-fits-all model for funding. The 
model illustrates a foundation for what 
could evolve into funding strategies based 
upon measurable inputs. Easily measurable 
inputs or drivers can potentially shield the 
library from arbitrary cuts. An institution 
can devise its own set of measurable inputs. 
Is the model a good thing or a bad thing for 
libraries? Libraries with exceptional influ-
ence on campus may find no benefit to such 
an approach. Libraries faced with political 
disadvantages or distinct underfunding 
relative to peer institutions may be able to 
use this approach to their benefit. 

Does increased library funding even 
necessarily convey additional benefit to 

http:ranking.13


 

 

 

178 College & Research Libraries March 2007 

TABLE 3 
Actual Expenditures, Model Expenditures, and Over or Under Predicted as 

Percentage of Model: Average Year 
Thousands of 2003 US Dollars 

University Actual Model Difference Percent Over 
(+) or Under (–) 
Predicted 

ALABAMA 10,870 16,512 –5,642 –34% 

ALBERTA 18,402 15,061 3,342 23% 

ARIZONA 22,400 19,898 2,502 12% 

ARIZONA STATE 21,972 18,461 3,510 19% 

AUBURN 10,399 13,835 –3,436 –25% 

BOSTON 14,848 24,051 –9,203 –38% 

BOSTON COLLEGE 15,884 19,655 –3,771 –19% 

BRIGHAM YOUNG 16,073 17,318 –1,245 –7% 

BRITISH COLUMBIA 22,254 15,891 6,362 41% 

BROWN 15,711 15,798 –87 0% 

CALIFORNIA, BERKELEY 42,217 19,251 22,966 119% 

CALIFORNIA, DAVIS 18,982 19,598 –616 –3% 

CALIFORNIA, IRVINE 16,691 15,695 996 6% 

CALIFORNIA, LOS ANGELES 39,396 21,871 17,526 81% 

CALIFORNIA, RIVERSIDE 10,322 12,427 –2,105 –17% 

CALIFORNIA, SAN DIEGO 20,720 15,960 4,760 30% 

CALIFORNIA, SANTA BARBARA 14,553 14,124 429 3% 

CASE WESTERN RESERVE 11,414 21,098 –9,684 –46% 

CHICAGO 23,378 22,871 507 2% 

CINCINNATI 17,082 18,828 –1,746 –9% 

COLORADO 17,526 15,710 1,816 12% 

COLORADO STATE 11,630 14,058 –2,427 –18% 

COLUMBIA 35,062 25,059 10,003 40% 

CONNECTICUT 19,677 18,347 1,330 8% 

CORNELL 34,170 22,372 11,798 52% 

DARTMOUTH 14,118 15,702 –1,584 –10% 

DELAWARE 13,040 13,574 –534 –4% 

DUKE 25,969 22,013 3,956 18% 

EMORY 24,691 20,702 3,988 18% 

FLORIDA 22,267 21,247 1,020 4% 

FLORIDA STATE 12,363 17,506 –5,143 –29% 

GEORGE WASHINGTON 17,475 22,471 –4,996 –23% 

GEORGIA 19,672 18,200 1,473 8% 

GEORGIA TECH 8,820 13,854 –5,034 –36% 

HARVARD 83,090 25,112 57,978 230% 

HAWAII 13,023 13,942 –919 –6% 



 

 

Toward a Formula-Based Model for Academic Library Funding  179 

TABLE 3 
Actual Expenditures, Model Expenditures, and Over or Under Predicted as 

Percentage of Model: Average Year 
Thousands of 2003 US Dollars 

University Actual Model Difference Percent Over 
(+) or Under (–) 
Predicted 

HOUSTON 12,779 16,443 –3,664 –22% 

HOWARD 12,269 20,098 –7,829 –39% 

ILLINOIS, CHICAGO 16,297 16,979 –682 –4% 

ILLINOIS, URBANA 28,473 19,592 8,881 45% 

INDIANA 27,385 18,103 9,282 52% 

IOWA 20,907 19,064 1,844 9% 

IOWA STATE 15,062 15,038 24 0% 

JOHNS HOPKINS 24,128 19,165 4,963 26% 

KANSAS 16,973 19,109 –2,136 –11% 

KENT STATE 10,974 13,540 –2,566 –19% 

KENTUCKY 17,996 18,321 –325 –2% 

LAVAL 12,378 15,297 –2,919 –18% 

LOUISIANA STATE 11,367 17,027 –5,661 –33% 

LOUISVILLE 16,519 19,189 –2,671 –14% 

MCGILL 15,639 15,546 93 1% 

MCMASTER 9,557 11,599 –2,043 –17% 

MANITOBA 10,688 13,550 –2,862 –21% 

MARYLAND 18,769 16,424 2,345 14% 

MASSACHUSETTS 12,337 14,357 –2,019 –14% 

MIT 14,808 18,253 –3,446 –19% 

MIAMI 15,430 21,635 –6,205 –29% 

MICHIGAN 39,918 22,588 17,330 76% 

MICHIGAN STATE 17,916 16,816 1,099 6% 

MINNESOTA 30,511 20,646 9,865 48% 

MISSOURI 13,001 18,298 –5,297 –29% 

MONTREAL 16,296 18,180 –1,884 –10% 

NEBRASKA 12,112 16,331 –4,220 –26% 

NEW MEXICO 18,547 17,400 1,147 7% 

NEW YORK 29,357 25,690 3,666 15% 

NORTH CAROLINA 25,944 20,314 5,630 28% 

NORTH CAROLINA STATE 19,014 14,934 4,080 26% 

NORTHWESTERN 21,047 24,191 –3,145 –13% 

NOTRE DAME 15,925 18,264 –2,339 –14% 

OHIO 12,241 15,715 –3,474 –22% 

OHIO STATE 25,372 22,242 3,130 14% 

OKLAHOMA 12,206 17,147 –4,941 –29% 



 

 

180 College & Research Libraries March 2007 

TABLE 3 
Actual Expenditures, Model Expenditures, and Over or Under Predicted as 

Percentage of Model: Average Year 
Thousands of 2003 US Dollars 

University Actual Model Difference Percent Over 
(+) or Under (–) 
Predicted 

OKLAHOMA STATE 10,368 15,262 –4,894 –32% 

OREGON 13,140 15,369 –2,229 –14% 

PENNSYLVANIA 28,942 24,911 4,031 16% 

PENNSYLVANIA STATE 33,508 20,321 13,188 65% 

PITTSBURGH 22,283 19,753 2,530 13% 

PRINCETON 29,226 16,088 13,138 82% 

PURDUE 14,251 16,331 –2,080 –13% 

QUEEN’S 10,395 13,161 –2,767 –20% 

RICE 13,502 14,528 –1,026 –8% 

ROCHESTER 12,658 17,292 –4,634 –27% 

RUTGERS 28,018 18,619 9,399 51% 

SASKATCHEWAN 9,055 12,983 –3,928 –30% 

SOUTH CAROLINA 15,956 18,292 –2,336 –13% 

SOUTHERN CALIFORNIA 24,929 25,123 –194 –1% 

SOUTHERN ILLINOIS 12,657 15,706 –3,048 –19% 

STANFORD 53,723 24,130 29,593 121% 

SUNY-ALBANY 10,530 12,988 –2,459 –19% 

SUNY-BUFFALO 16,153 18,823 –2,670 –14% 

SUNY-STONY BROOK 12,237 15,959 –3,722 –23% 

SYRACUSE 12,154 19,253 –7,099 –37% 

TEMPLE 13,181 18,841 –5,659 –30% 

TENNESSEE 15,399 17,681 –2,282 –13% 

TEXAS 30,418 20,665 9,753 47% 

TEXAS A&M 20,744 18,776 1,968 10% 

TEXAS TECH 15,337 19,073 –3,736 –20% 

TORONTO 38,465 17,260 21,205 123% 

TULANE 11,607 20,490 –8,883 –43% 

UTAH 19,466 18,058 1,408 7% 

VANDERBILT 17,031 21,965 –4,934 –22% 

VIRGINIA 24,683 18,668 6,015 32% 

VPI & SU 12,178 15,427 –3,249 –21% 

WASHINGTON 29,972 21,701 8,271 39% 

WASHINGTON STATE 12,231 13,996 –1,765 –13% 

WASHINGTON U.-ST. LOUIS 21,984 20,676 1,308 5% 

WATERLOO 9,444 10,658 –1,213 –11% 

WAYNE STATE 17,683 18,427 –744 –4% 



     
        
      

        
       

      
      

       
       
     

        
   

      

    
     

    
     

     

 

  

             

 

 

 

 

 

 

 

 
 

  

 

             

Toward a Formula-Based Model for Academic Library Funding  181 

TABLE 3 
Actual Expenditures, Model Expenditures, and Over or Under Predicted as 

Percentage of Model: Average Year 
Thousands of 2003 US Dollars 

University Actual Model Difference Percent Over 
(+) or Under (–) 
Predicted 

WESTERN ONTARIO 12,497 13,113 –616 –4%
	

WISCONSIN
	 32,378 21,869 10,510 48%
	

YALE
	 47,174 23,299 23,876 102%
	

YORK
	 14,115 11,805 2,310 20% 

the institution? Assuming that the institu-
tion’s budget process is a zero sum game, 
gains in library funding reduce funding 
elsewhere. It is well beyond the scope of 
this study to compare the marginal utility 
of dollars steered toward the library ver-
sus other programs on campus. However, 
in the cases of obvious underfunding one 
might argue that the utility of marginal 
dollars allocated to the severely under-

funded library may be high (i.e., a good 
investment by the institution). 

Last, the model may also serve to 
move the debate for funding away from 
traditional input measures and toward 
a broader set of institutionally based 
output indicators. This approach may 
be favorably received as universities as 
a whole move more toward outcomes-
based planning. 

Notes 

1. David Baker, “Resource Allocation in University Libraries,” The Journal of Documentation 
48 (Mar. 1992): 1–19. 

2. Frank W. Goudy, “Academic Libraries and the Six Percent Solution: A Twenty-Year Financial 
Overview,” Journal of Academic Librarianship 19 (Sept. 1993): 212–15. 

3. Association of College and Research Libraries, “Standards for Libraries in Higher Educa-
tion,” College & Research Libraries News 65 (Oct. 2004): 534–43. Available online from www.ala. 
org/ala/acrl/acrlstandards/standardslibraries.htm. [Accessed 15 September 2005]. 

4. Kendon Stubbs, “Trends in University Funding for Research Libraries,” ARL: A Bimonthly 
Newsle er of Research Library Issues and Actions 172 (Jan. 1994). 

5. Rolf Griebel, “University Library Budgets – Model and Reality,” New Review of Academic 
Librarianship 2 (1996): 59–67. 

6. Murle E. Kenerson, “Performance Funding and Full-Time Equivalence: Implications for 
Funding in Academic Libraries,” 13 (1996). ERIC, ED398927. 

7. Jin-Chuan Ma, “Fund Allocations for Information Resources in China’s Key Universities,” 
College & Research Libraries 60 (Mar. 1999): 174–78. 

8. Peter A. Facione, “The Philosophy and Psychology of Effective Institutional Budgeting,” 
Academe 88 (Nov./Dec. 2002): 45–48. 

9. Jim Tolhurst, “Resource Allocation and Budgeting,” Journal of Tertiary Educational Admin-
istration 7 (Oct. 1985): 143–55. 

10. Nancy Cantor and Paul N. Courant, “Scrounging for Resources: Reflections of the Whys 
and Wherefores of Higher Education Finance,” New Directions for Institutional Research 119 (Fall 
2003): 3–12. 

11. Gunapala Edirisooriya, “State Funding of Higher Education: A New Formula,” Higher 
Education Policy 16 (Mar. 2003): 121–33. 

12. David A. Belsley, Edwin Kuh, and Roy E. Welsch, Regression Diagnostics: Identifying Influ-
ential Data and Sources of Collinearity. (New York: John Wiley and Sons, 1980). 

13. Lewis G. Liu, “The Economic Behavior of Academic Research Libraries: Toward a Theory,” 
Library Trends 51 (Winter 2003): 277–92. 



   
   

 

      
      

      

    
 

  

 

     
     
    
 

         
       

A unique source of historical reference for nearly 250 years 
— delivering analyses and context for today’s headlines 

“A handy, convenient refer-
ence for a bird’s-eye view of 
what happened in the world 
during a given year…The 
strength of the publication 
lies in analysis, interpreta-
tion, and overviews.” 
– American Reference Books 
Annual 1999 

“This timely compendium 
covers the major economic, 
social, and cultural events… 
Cited in Sheehy and BCL 3.” 
– Reference & Research 
Book News 

“Continues its tradition of 
excellence… Published since 
1758, this source has proven 
its usefulness over time.” 
– American Reference Books 
Annual 2004 

“The writing is good and 
the authors are clearly 
experts. Exhaustive and 
thorough.” – American 
Reference Books Annual 2001 

For this new edition, over 80 contribu-
tors – headed by the eminent historian 
Philip M.H. Bell – combine their expert 
knowledge with singular clarity in 
style. They cover: 

 209 countries – more than the total 
membership of the UN. Provides 
quick-reference data for every country 
in the world. 

 Trends in: science and the conse-
quences of global climate change; the 
international economy; religion, law, 
and the arts. 

Special prepublication offer! 

$210 or £150, plus shipping, 
until 12/31/06; thereafter, $233 
or £166.40. Available through 
your vendor. 

The Annual Register, 247th Edition: 
World Events. Edited by D.S. Lewis & 
Wendy Slater. Overview by Philip 
M.H. Bell, with 83 author/contributors. 
ISBN 1-60030-069-3. 626 pp., incl. 4-
color photographs, maps, tables, and 
index, hardbound. 

CSA: Cambridge Information Group USA / UK 
7200 Wisconsin Ave, Bethesda, MD 20814 USA 

journals@csa.com, +1-301-961-6798