Research Article
Research Productivity and Its Relationship to Library
Collections
Sandra L. De Groote
Professor and Head of
Assessment and Scholarly Communications
University of Illinois at Chicago,
University Library
Chicago, Illinois, United
States of America
Email: sgroote@uic.edu
Beyza Aksu Dunya
Clinical Assistant
Professor, University Library, University of Illinois at Chicago
Assistant Professor, Bartin University College of Education
Chicago, Illinois, United
States of America
Email: baksu2@uic.edu
Jung Mi Scoulas
Clinical Assistant Professor
and Assessment Coordinator
University of Illinois at
Chicago, University Library
Chicago, Illinois, United
States of America
Email: jscoul2@uic.edu
Mary M. Case
University Librarian and
Dean of Libraries
University of Illinois at
Chicago, University Library
Chicago, Illinois, United
States of America
Email: marycase@uic.edu
Received: 14 Feb. 2020 Accepted: 7 July 2020
2020 De Groote, Akusa Dunya, Scoulas, and Case. This
is an Open Access article distributed under the terms of the Creative Commons‐Attribution‐Noncommercial‐Share Alike License 4.0
International (http://creativecommons.org/licenses/by-nc-sa/4.0/),
which permits unrestricted use, distribution, and reproduction in any medium,
provided the original work is properly attributed, not used for commercial
purposes, and, if transformed, the resulting work is redistributed under the
same or similar license to this one.
DOI: 10.18438/eblip29736
Abstract
Objective
–
The purpose of this study was to explore in the current academic library
environment, the relationship between library collections data (collections’
size, expenditures, and usage) and faculty productivity (scholarly output). The
researchers also examined the degree to which new and existing library metrics
predict faculty productivity.
Methods
–
Demographic data (e.g., faculty size, student size, research and development
expenditures), library budget data (e.g., collection expenditures), collection
use data (e.g., full-text article requests and database searches), and
publication output for 81 doctoral granting universities in the United States
were collected to explore potential relationships between research
productivity, collection use, library budgets, collection size, and research
expenditures using partial correlations. A hierarchical multiple regression was
also used to ascertain the significance of certain predictors of research
productivity (publications).
Results
–
A correlation existed between the number of publications (research
productivity) and library expenditures (total library expenditures, total
library material expenditures, and ongoing library resource expenditures),
collection size (volumes, titles, and ebooks), use of
collection (full-text article requests and total number of references in the
articles), and research and development expenditures. Another key finding from the hierarchical
multiple regression analysis showed that full-text article requests were the
best predictor of research productivity, which uniquely explained 10.2% of the
variation in publication.
Conclusion
–
The primary findings were that
full-text article requests, followed by library material expenditures and
research expenditures, were found to be the best predictor of research
productivity as measured by articles published.
Introduction
In this study, the authors examined the relationship between academic
library collections and research output at research intensive doctoral granting
academic institutions. With shrinking library budgets and increasing costs for
the online resources licensed or purchased by libraries, it is becoming more
challenging to provide access to the information resources needed by
researchers at academic institutions. Librarians are increasingly spending more
time trying to determine how best to spend their limited budgets as they
consider what new resources to purchase, what resources to maintain, and what
resources to cancel. Academic libraries face challenges demonstrating the need
for appropriate funding to continue to meet the information needs of researchers.
Academic libraries have evolved greatly since the migration to
electronics resources (e.g., online journals, databases). Many libraries have
increased the number of journal titles available to users through the licensing
of “big deal” journal packages (packages are selected for needed journal titles
but also typically include journal titles that a library would not necessarily
choose). Journal collections have moved from a print to an online format and
Abstracting and Indexing (A&I) tools for finding journal literature, once
found only in print, have also moved online and increased in numbers. As a
result, the way in which researchers seek and obtain information from the
library has changed dramatically.
With resources online, new sources of usage data are also available. For
example, libraries can obtain data on how many full-text article downloads
occur and how many database searches are run, a direct reflection of patron
activities. In databases such as Scopus
and Web of Science, data can be more readily obtained, compared to a
print-based environment, on how many articles were published by an institution,
how many references were included in those articles, and how many times those
articles were cited. How libraries are asked to report collections and
collections usage has also changed. Between 2011 and 2012, the Association of
Research Libraries (ARL) ceased asking for data on certain metrics (monographs
purchases, total current serials, expenditures for monographs, expenditures for
serials), and began asking for data on such metrics as number of electronic
books (ebooks), ongoing resource purchases,
collection support, number of successful full-text article requests (journals),
and number of regular searches (databases) for their ARL Annual Library
Statistics. One question that arises is whether any of the new metrics
available can illustrate a relationship between faculty productivity and the
use or availability of library resources.
The purpose of this study was to explore what new and existing library
metrics demonstrate a relationship with faculty productivity. This study
explored the relationship between research productivity, as defined by the
number of journal articles published by an institution, and (a) the size of
library collection budgets, (b) the size of the collection (e.g., number of
titles held, number of volumes held), (c) use of the collection (number of
successful full-text article requests, number of database searches, references
included in publications), and (d) other library and institutional
expenditures.
Literature Review
Several researchers have quantitatively examined the relationship
between research productivity and various metrics reflecting resources
available to an institution, including library collections. Rushton and Meltzer
(1981) studied 169 leading universities from the United States, the UK, and
Canada, exploring the relationship between total publications and total faculty
size, total student size, revenue, number of volumes in the library, and number
of current library periodical subscriptions. They found that when universities
were high on one measure, they were high on others as well. Specific to the
American universities, there was a positive correlation between the number of
volumes in the library and total publications and the number of current journal
subscriptions and total publications. Revenue was found to be the principal
factor that could predict the result of the other variables being examined,
concluding that “the quality and wealth of a university are clearly related”
(Rushton & Meltzer, 1981, p. 301). Dunbar and Lewis (1998) also explored
quantitative factors influencing and contributing to research performance at
doctoral institutions in the United States, using data from the 1993 National
Research Council study. They explored 30 doctoral programs in the biological
sciences, physical sciences and mathematics, social and behavioral sciences,
and engineering from top Carnegie classification research universities. They
found programs with more faculty were more productive, although when some
programs got beyond a certain size, productivity declined. In addition, more
full professors and more financial research support within a department also
resulted in increased research productivity. With respect to the library,
except for engineering and social sciences, departmental research productivity
was found to have a significant positive relationship with library
expenditures. In a more recent study, Rawls (2015) used ARL
Annual Library Statistics prior to the 2011/2012 change of variables. Rawls
utilized ARL library expenditure variables and other data including total
number of faculty publications, faculty size, research expenditures, and grant
awards from a five-year period (2005–2009) to explore the relationship between
faculty research productivity and library investment. Rawls found that research
productivity was positively correlated with library investment. More
specifically, electronic library resource expenditures correlated positively
with an increase in productivity.
Surveys of academic researchers have also been conducted to examine
researchers’ use of the library and their information seeking behavior, and to
assess the value of the library’s collection as it relates to their research
and other academic activities. Results of a survey of randomly selected faculty
at four large state universities in Texas found only 2.6% reported that the
library resources were not important in their research efforts and over 68%
indicated library resources were of “considerable” or “very high” importance to
their research (Cluff & Murrah,
1987). The findings also suggested that the larger the university, the more
likely faculty would report dependence on the library for research purposes.
Faculty members from seven universities (five U.S. and two Australian) were
surveyed in 2004/2005 about their scholarly article reading habits (Tenopir, King, Spencer, & Wu, 2009). Faculty were asked
to recall how many scholarly articles they had read in the past 30 days and the
source of the articles they read. Faculty members in more research-oriented
positions reported reading more for research purposes (62%) compared to the
amount reported by teaching-oriented faculty (49%). Also, the more a faculty
member published, the greater they reported their reading was for research
purposes. Faculty members in research-oriented positions also reported that 58%
of their reading materials were provided by the library, in contrast to faculty
in teaching-oriented positions, who reported that 37.8% of their reading
materials were provided by the library. A more recent survey of provosts
assessed their perceived value of the academic library (Murray & Ireland,
2018). Respondents from universities falling under the Carnegie Classification
as “Research Very High” institutions perceived the academic library as very
involved (84.21%), somewhat involved (10.53%), or marginally involved (5.26%)
in faculty research productivity.
Longitudinal survey data collected over a 30-year period has also
provided insight toward researcher behavior over time and their use of the
library. Tenopir, King, Edwards, and Wu (2009)
concluded that faculty are reading more articles than they had in the past and
that faculty were relying more on the library to provide access to articles
rather than the personal subscriptions they had relied upon in the past. The
availability of online articles resulted in faculty using more methods to
identify the articles to read, including browsing the table of contents of
online journals and searching for articles using Internet search engines,
full-text databases, and online A&I databases, compared to how they
searched for articles in the print environment.
The value of using literature in grant proposals, grant reports, and
articles has also been studied, in part to examine the impact of increased
access to journals available in the electronic collections of academic
libraries. A 2009 online survey, sent to faculty at seven different institutions
in seven different countries, examined faculty citing behavior (Tenopir et al., 2011). On average, 90% of the respondents
indicated that citations were “important”, “very important” or “essential” as
part of the grant writing process. Approximately 69.6% of respondents reported
citing 10 or more references in grant proposals, and 82.2% of respondents
reported citing at least one reference in final grant reports. Seventy-five
percent reported accessing more than half of the articles through their library’s
electronic collection, and over 50% reported accessing 75% of the articles from
the library’s electronic collection. Using a Return on Investment (ROI) model,
Kaufman (2008) explored the connection between use and investment in the
library and funded grant proposals. Grant applicants at University of Illinois
at Urbana-Champaign were surveyed regarding the role of the library in their
research and grant processes. Ninety-five percent of respondents indicated that
references were important for obtaining grants, and approximately 75% of
respondents noted that 75% of the references used in grant proposals were
provided through the library. The resulting ROI calculation found that for
every dollar invested in the library, there was a return on investment of $4.38
in grant funding (Kaufman, 2008).
Using various methodologies, the studies noted above illustrated that
academic library collections, such as journal articles, books, and databases,
are important sources of information for use in research, teaching, and grant
proposals. However, while surveys assessing information seeking behavior,
library collections use, and the value of the library provide valuable insight
and allow for flexibility in terms of the questions asked, they rely on memory
and perception to provide the data from which the findings are drawn. While
researchers in older studies have shown a relationship between research
productivity and collection size (Rushton & Meltzer, 1981) or collection
expenditures (Dundar & Lewis, 1998; Rawls, 2015),
these studies were based on data gathered prior to or during the transition of
the library from a print based to an online environment. Academic library
collections have changed, user information seeking behavior has changed, and
the measures for tracking library usage have also changed and evolved. Newer
literature exploring quantitative data in relation to research productivity is
limited. Exploring older metrics in the present library environment and
exploring newer library metrics, such as collections usage data, in relation to
research productivity can contribute to validating the impact academic
libraries have on scholarly output. This study will explore faculty
productivity and its relationship with library resource usage, library budgets,
and collection size. Because past researchers have noted the relationship with
faculty productivity and overall institutional support (Rushton & Meltzer,
1981), research support (Dunbar & Lewis, 1998), and library expenditures
(Dunbar and Lewis, 1998; Rawls, 2015), we also explore overall research
expenditures with respect to faculty productivity and library expenditures.
Aims
This study aimed to examine the relationship between library
collections, budgets, and use and research productivity among academic research
institutions in the United States that were both members of the ARL and
designated as doctoral universities according to the Carnegie Classification of
Institutions of Higher Education.
Methods
To identify the list of institutions to include in the study, the list
of ARL institutions was downloaded from the ARL website (Association of
Research Libraries, 2016) and the list of universities designated as Very High
Research Activity or High Research Activity Doctoral Institutions was
downloaded from the Carnegie Classification of Institutions of Higher Education
website using the basic classification feature to select the Doctoral
Universities (Carnegie Classifications, 2016). This data was entered into a
single spreadsheet, where a total of 104 academic research institutions were
identified for inclusion in the study.
Data regarding library collection size, budget, and use, research
productivity (journal publications), and institutional demographic data were
also obtained from various resources and entered into the spreadsheet. Because
ARL surveys its ARL member libraries on an annual basis related to multiple
data points including staffing collections, expenditures, services, and usage,
the ARL Annual Library Statistics was an ideal resource for library related
data. Scopus, an online indexing and abstracting database produced by Elsevier,
was selected to provide the number of publications for each institution
included in the study. Scopus has been reported to be the most comprehensive
article-level index of scholarly articles (Laakso
& Bjork, 2012). In order to have a metric for the overall expenditures of
an institution, the Research and Development Expenditures from the Higher
Education Research Development (HERD) survey was selected, as this is the
primary source of information on research and development expenditures at U.S.
universities (National Science Foundation, 2017). Further details on how the
data were collected is provided below.
Data Collection
ARL Annual Library Statistics Data
Data related to library collections size, use, expenditures, and
additional institutional demographic information reported by each institution
to ARL Statistics (https://www.arlstatistics.org) was obtained for 2015, 2016, and 2017. Because data and resources
might fluctuate somewhat from year to year, instead of examining a single year
of data, the last three available years of data were collected. The three years
of data were then averaged for each variable to obtain the final data used in
the analyses. The variables collected included:
·
Total materials expenditures – includes one-time
purchases (non-subscription, one-time purchases such as books, software,
backfiles, et al.), ongoing library resources expenditures, and other
collections support
·
Total library expenditures – the total expenditure of
all library funds (includes total library materials, total salaries and wages,
and other operating expenditures but excludes fringe benefits)
·
Ebooks – total number of ebooks available in the collection
·
Volumes held – total number of print
only items and ebooks
·
Titles held – total number of print and electronic
serials, monographs, manuscripts, dissertations and theses, archives,
microforms, and computer files held in the collection (excludes duplicates)
·
Number of successful full-text article requests for
journal articles (defined by the COUNTER Code of Practice – www.projectcounter.org)
·
Number of regular searches – number of database
searches as defined by the COUNTER Code of Practice – www.projectcounter.org)
·
Total full-time students (undergraduates and
graduates) and total faculty (full-time members of the instruction/research
staff)
·
Size of students and faculty were included as control
variables that represents the institutional size, which may have an impact on
research productivity and use of the resources.
Research Productivity Data
Research productivity in this study is referring to the number of
journal publications produced by an institution. In August of 2018, we searched
the Scopus database using the
affiliation field and entering the name of each institution included in the
study. Abbreviations from a sample of institutions were tested to ensure they
would link to the full names of the institutions. If multiple variations of the
same institution were displayed in the results of the institutional names, all
relevant versions were selected to provide the total number of publications.
The search results were limited by “Document type” to articles or review, to
retrieve the number of journal article publications published in 2016, 2017,
and 2018 at each institution.
References Data
As another measure of potential use of library collections, the total
number of references used in the publications studied were also obtained. For
each list of institutional publications presented in Scopus, the number of
total references included in these publications were obtained by clicking all
the publications in a list and selecting “View References.” The Scopus system
limits reference lists to a maximum of 2000. When institutions had more than
2000 publications, results were grouped using some of the limiting features of
the system to obtain numbers for the full set. The total number of references
included in the institutional publications was also entered into the
spreadsheet. Because the number of references was not displayed per article but
per set of articles, it is possible an article was cited multiple times, but
would only be displayed once in the list of references. Thus, this data may
underreport the number of references included in the studied articles when
references to articles were cited by multiple articles within a set of
publications.
As with other data included in the study, the average number of
publications and the average number of total references were obtained by
averaging the three years of data. Because of the time delay between writing a
manuscript and it being published, it is likely that much of the literature
searching and use of the library is done in the year previous to an article
being published. Therefore, ARL data from 2015 to 2017 was collected while
publication data was obtained from 2016 to 2018 to better approximate and
coincide with potential library usage. For example, while it is not a perfect
assumption, given the time to write and publish an article, if an article was
published in 2018, there is logic in assuming that in many cases the literature
review and use of library collections occurred in 2017, or potentially earlier.
Since the total number of references included in all publications at an
institution is impacted by the number of articles written, the average number
of references per article was obtained by dividing the total number of
references included in the publications (all three years) by the total number
of publications (all three years).
Research Expenditures
Research and Development Expenditures for each institution for 2015-2017
were obtained from the HERD, where universities report research expenditure and
sources of revenue (National Science Foundation, 2017). The data were entered
into the spreadsheet and the average over the three years was calculated.
It was not always possible to match up institutional data from the four
data sources. For example, some universities have multiple locations and it was
not always clear if data sets covered all locations or a specific location. In
other situations, it appeared that medical colleges’ libraries often had
separate budgets, and data might not have included data from the full
institution. In other cases, full data was not available for all the years. In
situations where the limits of the data were not clear or if data were missing
(except for collections use data), the institution was dropped from the study.
As a result, 81 institutions remained in the study. Table 1 summarizes the
average numbers of all the variables used in the analysis to demonstrate the
overall data patterns of a total of 81 ARL member libraries.
Data Analysis and Research Questions
All data were analyzed using SPSS 26. Multiple statistical tests were
employed to examine direct and indirect impacts of the library on faculty
productivity. The analyses included partial correlations and hierarchical
multiple regression. Partial correlation is a measure of strength and direction
of the linear relationship between two variables, while controlling for the
effect of one or more variables (covariates). Partial correlation allows
finding a unique relationship between two variables while eliminating the
influence of a third variable, which may drive the relationship. Hierarchical
multiple regression, a form of multiple regression in which independent
variables are entered into the regression “in the order specified by the
researcher based on the theoretical grounds” (Pallant,
2012, p.149), is used to predict the value of one dependent variable after
controlling for another, in this case, for faculty size. As shown in the
literature, there are several factors that influence faculty productivity. As
part of the institutional factors, faculty size (e.g., Dunbar & Lewis,
1998) and overall wealth (e.g., Rushton & Melzer, 1981) were found to be
linked to research productivity. Among library factors, library expenditure is
well known to correlate with publications (Dunbar & Lewis, 1998; Rawls,
2015). However, library expenditures is a broad
category and contains expenses beyond just materials (operating budgets,
salaries). To compare our findings to past studies, we also compared library
expenditures data to productivity. In addition, we explored narrower categories
of library expenditures including the overall materials budgets and ongoing
library resource expenditures. Moreover, our current study expanded on the
previous literature by examining whether factors related to library resource
use (full-text article requests, database searches) contribute to faculty
productivity.
Table
1
Descriptive Statistics for All Variables
(Institution Size and Expenditures, Library Budgets, Collection Size,
Collection Use, and Research Productivity)
|
N |
Minimum |
Maximum |
M |
SD |
|
|
Institution Demographics |
|||||
|
Total
full-time students |
81 |
6,253 |
69,939 |
25,285 |
11,674 |
|
Total faculty |
81 |
659 |
4,481 |
1,792 |
821 |
|
Institutional Expenditures |
|||||
|
HERD |
81 |
38,244 |
71,840,290 |
1,381,179 |
7,933,921 |
|
Library Budgets |
|||||
|
Total library
expenditures |
81 |
10,349,703 |
116,533,712 |
31,751,291 |
17,713,781 |
|
Total library materials
expenditures |
81 |
4,606,644 |
47,791,377 |
14,339,988 |
7,292,618 |
|
Ongoing
resource expenditures |
80 |
3,865,090 |
20,754,521 |
10,230,944 |
3,494,542 |
|
Size of Library Collections |
|||||
|
Volumes held |
81 |
1,941,116 |
20,837,233 |
5,849,571 |
3,433,706 |
|
Titles held |
81 |
970,064 |
14,863,477 |
4,653,645 |
2,583,329 |
|
Ebooks |
81 |
134,801 |
3,291,347 |
1,205,576 |
570,905 |
|
Use of the Collection |
|||||
|
Full-text
article requests |
75 |
192,686 |
12,752,344 |
4,100,529 |
2,779,468 |
|
Regular
searches |
73 |
636,732 |
7,8174,661 |
9,303,921 |
12,823,470 |
|
Total number
of references in the publications |
81 |
20,430 |
861,817 |
195,173 |
131,808 |
|
Number of references per article |
81 |
40 |
51 |
46 |
2 |
|
Research Productivity |
|||||
|
Total publications |
81 |
458 |
19,171 |
4,306 |
2,976 |
Table
2
Statistical
Tests
|
Statistical Test |
Variables |
Research Question |
|
Partial
Correlations |
|
|
|
Collection
use and research productivity |
·
Successful
full-text article requests, number of regular database searches, number of
references included in publications, average number of references per
publication (IVs) ·
Total
number of articles published (DV) ·
Total
number of full-time students and faculty, Research and development
expenditures (HERD), Total materials expenditures (Covariates) |
Holding
number of full-time students, faculty, total materials expenditures, and
research and development expenditures constant, what is the relationship
between use of the collection and total number of articles published in an
institution? |
|
Library
budgets and research productivity |
·
Total
materials expenditures; total library expenditures; ongoing resource
expenditures (IVs) ·
Total
number of articles published (DV) ·
Total
number of full-time students and faculty, Research and development
expenditures (Covariates) |
Holding
number of full-time students, faculty and research and development
expenditures constant, what is the relationship between library budgets and total
number of articles published in an institution? |
|
Collection
size and research productivity |
·
Volumes
held, titles held; ebooks (IVs) ·
Total
number of articles published (DV) ·
Total
number of full-time students and faculty, Research and development expenditures
(Covariates) |
Holding
number of full-time students, faculty, and research and development
expenditures constant, what is the relationship between library collection
size and total number of articles published in an institution? |
|
Research
expenditures and research productivity |
·
Research
and Development expenditures (IV) ·
Total
number of articles published (DV) ·
Total
number of full-time students and faculty, Total library expenditures, Total
materials expenditures, Ongoing resource expenditure (Covariates) |
Holding
number of full-time students, faculty, total library expenditures, total
materials expenditures, and ongoing resource expenditures constant, what is
the relationship between Research and Development expenditures and total
number of articles published in an institution? |
|
Library
budgets and research expenditures |
·
Total
library expenditures, Total materials expenditures, Ongoing resource
expenditure (IVs) ·
Research
and Development expenditures (HERD) (DV) ·
Total
number of full-time students and faculty (Covariates) |
Holding
number of full-time students and faculty, what is the relationship between
Research and Development expenditures and Library expenditures? |
|
Hierarchical
Multiple Regression |
|
|
|
|
·
Total
number of faculty, Research and development expenditures, Library material
expenditure and Full-text article requests (IVs) ·
Total
number of publications from 2016 to 2018 (DV) |
Controlling
for the possible effect of total number of faculty, is the set of variables
(research and development expenditures, library materials expenditures and
full-text article requests) still able to predict a significant amount of the
variance in total number of publications? |
Table 2 outlines which tests were used to address the research
questions. Before running the statistical analyses, tests of assumptions were
run to confirm it was appropriate to run the proposed analysis including the
possibility of multicollinearity using cut-off points for tolerance value of
less than .10 or VIF value of above 10 guided by Pallant
(2010). There is no violation of the multicollinearity assumption.
Results
Partial Correlations Among Collection Use, Budgets, Collections,
Research Expenditures, and Research Productivity
Results from partial correlations are displayed in Table 3. As the
number of faculty and students, amount of research and development
expenditures, and library materials expenditures were likely to influence the
number of publications and use of the collection, their contribution to the
relationship was eliminated through partial correlation. The first partial
correlation explored collection use and research productivity. It revealed a
moderate positive relationship between successful full-text article requests
and the number of articles published at an institution, r (69) = .504, p < .001, suggesting higher use of
the library (successful full-text article requests) is associated with research
productivity. Not surprisingly, there was a strong positive correlation between
the total number of references (average over 3 years) included in the
publications and the total number of publications per institution, controlling
for the number of full-time students and faculty, total library materials
expenditures, and research and development expenditures, r (75) = .994, p <
.001. One would expect as the number of total publications increased, so would
the number of references included in those
publications. However, when the average number of references per article was
compared to the total number of publications, there was a weak negative
correlation, r (75) = -.279, p = .014. This suggests the more
references used, the fewer publications or, the more articles published, fewer
references will be included. There was not a significant correlation found
between the number of publications and the number of regular database searches,
r (67) = -.200, p = .100.
The second set of partial correlation analyses
explored library expenditures and research productivity. These analyses
demonstrated a strong positive relationship between articles published at an
institution and total library materials expenditures (r (76) =.661, p < .001) and total library
expenditures (r (76) =.748, p
< .001), but only a moderate positive correlation between articles published
at an institution and ongoing resource expenditures (r (75) =.551, p < .001). The higher the
expenditures allocated for a library, the higher the numbers of publications
were produced at an institution. This demonstrates a significant relationship
between library expenditures and research productivity.
Table
3
Partial
Correlations Among Collection Use, Library Budgets, Collection Size, Research
Productivity, and Research Expendituresa
|
|
df |
r |
p |
|
Collection
Use and Research Productivity |
|
|
|
|
Article requests |
69 |
.504 |
< .001* |
|
Database searches |
67 |
-.200 |
.100 |
|
References included in publication |
75 |
.994 |
< .001* |
|
Average number of references per publication |
75 |
-.279 |
.014 * |
|
Library
Budgets and Research Productivity |
|
|
|
|
Ongoing resources expenditures |
75 |
.551 |
< .001* |
|
Materials expenditures |
76 |
.661 |
< .001* |
|
Library expenditures |
76 |
.748 |
< .001* |
|
Collection
Size and Research Productivity |
|
|
|
|
Volumes held |
76 |
.708 |
< .001* |
|
Titles held |
76 |
.646 |
< .001* |
|
Ebooks |
76 |
.282 |
.012* |
|
Research
Expenditures and Research Productivity |
|
|
|
|
Research and development expenditures |
73 |
.323 |
.005* |
|
Library
Budgets and Research Expenditures |
|
|
|
|
Ongoing resource expenditures |
76 |
.294 |
.009* |
|
Material expenditures |
77 |
.135 |
.236 |
|
Library expenditures |
77 |
.126 |
.269 |
a Strength of correlations as
indicated by Dancey and Reidy (2011) for absolute
value of r –
|.10|
< r < |.30| weak, |.40| < r < |.60| moderate, |.70|
< r < |.90| strong
*
Indicates significant p value.
The partial correlations exploring collection size and
research productivity suggested that articles published at an institution
correlated positively with the size of library collections (volumes held, r (76) = .708, p < .001; titles held, r (76)
= .646, p < .001; and ebooks, r (76) =
.282, p = .012). This means the
greater the size of a library’s collection (volumes, titles and ebooks), the greater the number of publications produced at
the institution. However, the strength of correlations for ebooks
were weak r < .30, as Dancey and Reidy
(2011) previously found.
The partial correlation revealed a positive moderate relationship
between research development expenditures and the number of articles published
at an institution, controlling for the number of full time students and
faculty, total library expenditures, total library materials expenditures, and
ongoing resource expenditures, r (73) = .323, p = .005, suggesting the higher the research and development
expenditures obtained by an institution, the higher the number of publications
produced at the institution.
The last partial correlation suggested that the amount of research and
development expenditures correlated positively with ongoing resource
expenditures (r (76) = .294, p = .009); however, the strength of the
correlation is weak r < .30. This means the greater the amount of
ongoing resource expenditures, the greater the amount of research and
development expenditures. There was not a significant correlation found between
other library expenditures and the amount of research and development
expenditures (total materials expenditures, r
(77) = .135, p = .236; total
library expenditures, r (77) = .126, p = .269).
Predicting Research Productivity (Publications)
Next, a four-stage hierarchical multiple regression was conducted to
examine the degree to which research and development expenditures and the
library related collection measures (library materials expenditures and
full-text article requests) affected research productivity, after controlling
for the influence of the institutional size (total number of faculty). The
sample size of 81 was considered adequate given four independent variables
subjected to the analysis: a ratio of 15 cases for every independent variable (Tabachnick & Fidell, 2001).
Total number of faculty was entered at stage one of the regression
to control for faculty size. The research and development expenditures variable
was included at stage two; the library materials
expenditures variable was entered at stage three, and full-text article
requests was entered at stage four. Institutional funding (research and
development expenditures), library related variables (i.e., library materials
expenditures and full-text article requests) were entered in this order because
research and development expenditures represents the institutional funding size,
library materials expenditures represents library collections, and library
usage (full-text article requests) is followed by it. Library collection size
was not included in this analysis because collection size is highly correlated
with library materials expenditures. The reason for choosing full-text article
requests, rather than the number of regular database searches at stage four, is
that this variable has a higher correlation with publications in the partial
correlation described above. We did not use total library expenditures as the
variable here because it contains expenditures beyond the collection, such as
salaries and operational expenses. We wanted to explore the unique contribution
of the collections expenditures, as reflected through
the use of materials expenditures.
Table 4 indicates the significance of each of the four ANOVA models.
While all four models were significant at p < .001, the F
value was largest for Model 4 with four predictors (total number of faculty,
research and development expenditures, library materials expenditures, and
full-text article requests), meaning that Model 4 as a whole is the most
significant (F (4, 70) = 45.932, p <.001) as a predictive model.
The hierarchical multiple regression showed that in Model 1, the total
number of faculty contributed significantly to the regression model, F
(1,73) = 23.063, p < .001 and accounted for 24% of the variation in
publications (Table 4). In Model 2, research and development expenditures
explained an additional 9.8% of the variation in publications, after
controlling for total number of faculty; this change in R² was significant, F
change (1,72) = 10.676, p = .002. After introducing library materials
expenditures in Model 3, the total variance explained by the model as a whole
(which includes faculty, research and development expenditures and library
materials expenditures) was 62.2%, F (3,71) =38.911, p < .001.
The library materials expenditures explained an additional 28.4% of the
variance in publications, after controlling for total number of faculty and
research and development expenditures. The change in R² was highly significant,
F change (1,71) = 53.239, p < .001. This result clearly showed
that library materials expenditures contributed significantly to the total
number of publications. Finally, including full-text article requests in the
fourth and final model explained an additional 10.2% of the variation in
publications, after controlling for total number of faculty, research and
development expenditures, and library materials expenditures. This change in R²
was also significant, F change (1,70) = 25.960, p = .001,
indicating that full-text article requests has a significant effect on the
publications. When all four independent variables were included in stage four
of the regression model, the total number of faculty was not a significant
predictor of publications. As shown in Table 5, the best predictor of
publications was full-text downloads (β =.509), which uniquely explained 10.2%
of the variation in publications. In order of the next important predictors of
publications they were library materials expenditures (β =.341) and research
and development expenditures (β =.150). Together the four independent variables
accounted for 72.4% of the variance in publications.
Table
4
ANOVA
Results of the Four Model-Hierarchical Regression Analysisa
|
|
|
Sum of Squares |
df |
Mean square |
F |
|
Model
1b |
Regression |
164658876.662 |
1 |
164658876.662 |
23.063* |
|
|
Residual |
521175432.352 |
73 |
7139389.484 |
|
|
|
Total |
685834309.014 |
74 |
|
|
|
Model
2c |
Regression |
231958134.313 |
2 |
115979067.157 |
18.398* |
|
|
Residual |
453876174.701 |
72 |
6303835.76 |
|
|
|
Total |
685834309.014 |
74 |
|
|
|
Model
3d |
Regression |
426452545.391 |
3 |
142150848.464 |
38.911* |
|
|
Residual |
259381763.622 |
71 |
3653264.276 |
|
|
|
Total |
685834309.014 |
74 |
|
|
|
Model
4e |
Regression |
496623237.92 |
4 |
124155809.48 |
45.932* |
|
|
Residual |
189211071.094 |
70 |
2703015.301 |
|
|
|
Total |
685834309.014 |
74 |
|
|
a Dependent variable: total number
of publications
b Predictors: total number of
faculty
c Predictors: total number of
faculty, research and development expenditures
d Predictors: total number of
faculty, research and development expenditures, library materials expenditures
e Predictors: total number of faculty,
research and development expenditures, library materials expenditures,
full-text article requests
* Indicates p is
significant at < .001.
Table
5
Summary of Hierarchical Regression Analysis for
Variables Predicting Publicationsa
|
|
β |
t |
R2 |
DR2 |
F |
DF |
|
Model
1 |
|
|
.240 |
.240 |
23.063*** |
23.063*** |
|
Total number of faculty |
.490 |
4.802*** |
|
|
|
|
|
Model
2 |
|
|
.338 |
.098 |
18.398*** |
10.676** |
|
Total number of faculty |
.506 |
5.269*** |
|
|
|
|
|
HERD |
.314 |
3.267** |
|
|
|
|
|
Model
3 |
|
|
.622 |
.284 |
38.911*** |
53.239*** |
|
Total number of faculty |
.173 |
2.377* |
|
|
|
|
|
HERD |
.227 |
3.514** |
|
|
|
|
|
Library materials expenditures |
.595 |
7.296*** |
|
|
|
|
|
Model
4 |
|
|
.724 |
.102 |
45.932*** |
25.960** |
|
Total number of faculty |
.052 |
.636 |
|
|
|
|
|
HERD |
.150 |
2.284* |
|
|
|
|
|
Library materials expenditures |
.341 |
3.965*** |
|
|
|
|
|
Full-text
article requests |
.509 |
5.095*** |
|
|
|
|
a Dependent variable: Total number of publications.
* p
< .05. ** p < .01. *** p < .001.
Discussion
The purpose of this study was to obtain information concerning the
relationship between the use of library collections and research productivity
in the electronic era. The findings illustrated the strength of this
relationship and document the contributions that today’s academic library has
on an institution’s research success. Because previous literature exploring
quantitative library metrics with research productivity is limited and older,
this study also bridges a gap.
Like previous work (Rushton & Melzer, 1981), this study found a
correlation between research productivity and library expenditures, collections
held, and research and development expenditures. As Rushton and Melzer
concluded, the overall wealth of an institution likely contributes to faculty
productivity because a strong infrastructure of support is likely to be in
place. More research is needed to better understand and to uncover underlying
factors. Similar to Rawls’ (2015) exploration of the ARL data from 2005 to
2009, this study of 2015 to 2018 data also found productivity was positively
correlated with library expenditures. We also found that total materials
expenditures and ongoing library resource expenditures were also correlated,
but not as strongly as total library expenditures. Distinct from studies of the
past, this study examined usage data and found a more direct link between use
of the collection (full-text article requests) and research productivity. The
greater the research productivity (journal article publications), the greater
the use of the library’s collection, as demonstrated through full-text article
requests.
Based on the findings from the partial correlations and literature
review, we further examined if a set of variables (research and development
expenditures, library materials expenditures, and full-text article requests)
were still able to predict a significant amount of the variance in total number
of publications after controlling for the possible effect of total number of
faculty. The primary findings from the hierarchical multiple regression
analysis was that full-text article requests were found to be the strongest
predictor of research productivity as measured by articles published, followed
by the library material expenditures and research expenditures. Even when
controlling for the total number of the faculty, research expenditures, and
library materials expenditures, full-text article requests uniquely explained
10.2% of the variation in publications. These findings provide strong evidence
that funding libraries supports faculty research success. The findings
demonstrated not just that an investment in library collections correlated with
productivity, but that the use of the library collections positively
contributed to faculty productivity. Given the cyclical nature of research,
faculty productivity likely leads to further faculty success, through additional
research development and expenditures. Libraries can use this information to
communicate the library’s impact on faculty productivity with various
stakeholders. Libraries should also explore their faculty’s research agendas
and the use of the journal collection through full-text article requests, to
assist with future collection development decisions so they are in line with
the needs of the faculty. Examining both existing metrics (e.g., collection
size or collection expenditures) and new metrics (e.g., the use of the
collection) in the current study expands on the existing literature and
confirms that the use of library collections has a great impact on research
productivity.
One of the unexpected findings of this study is that the number of average
references used per publication decreased the more productive an institution
was. One speculation for this is that the more productive faculty are, the less
likely they may be to search broadly for articles. It could also be that as
experts in their field, they are able to be more selective in the publications
that they choose to cite to address their findings. Alternatively, it might be
the case that as productivity increases, the articles produced are more
narrowly focused or cutting edge, and fewer relevant resources are available
for citing. Further exploration at the author level is needed to understand
this finding.
Limitations
There were some limitations to the data collected. In this study, we
largely explored research productivity as it related to journal articles, and
library usage as it related to journal article usage (database searches,
full-text article requests, and number of references in journal articles).
Therefore, disciplines that do not produce journal articles or are not reliant upon
them for research are excluded from this study. Although Scopus is the most
comprehensive journal literature database, it does not index all journals, nor
books or book chapters. Thus, the publication and reference data obtained from
Scopus was limited to journal publications indexed in Scopus. The above-mentioned
factors tend to bias the data toward those disciplines (e.g., STEM, social
sciences) whose research is reported primarily through the journal literature.
In addition, it was not possible to limit the examination of the data collected
in this study to specific disciplines. For example, ARL Annual Library
Statistics are reported in aggregate for each academic institution, although a
broad category of health sciences is available. With respect to disciplines and
publications, only the institutional affiliation is indexed using standardized
terminology within the Scopus
database, thus making it difficult to retrieve comprehensive publication data
from a college or department. This means the findings of this study will apply
broadly to institutions but will not provide insight into correlations or
relationships within specific disciplines.
The “number of full-text article requests” and the “number of regular
searches” were obtained from vendors
that provide “COUNTER” statistics. COUNTER statistics were developed to provide
consistent and credible data regarding the usage of databases and journals. However,
not all vendors provide this data, so the numbers provided to ARL from each
institution were likely not complete. This study was also reliant upon the
accuracy of the reported survey data collected and used in the study (ARL
Annual Library Statistics, HERD); while institutions attempt to report the most
accurate information, there is always the potential for error or incomplete
data reporting.
Conclusion
As found in past studies, research productivity correlated positively
with library expenditures. We also found that the use of the collection had a
relationship with research productivity. Even more important, full-text article
downloads uniquely explained approximately 10% of variation in research
productivity, over and above other factors including research and development
expenditures and library expenditures. Full-text article downloads were a
better predictor of research productivity than research and development
expenditures or library expenditures. This finding suggests that the use of
collections has more impact on the articles published than the total collections
dollars libraries spend. Collections developed to fit with the current research
agendas of faculty may impact their productivity. This finding may support
library decisions surrounding expenditures and future selections of resources
related to research support. This may also be important information for
academic libraries at other Carnegie levels that are building support for their
research programs.
References
Association of Research Libraries. (2016). ARL
statistics [Data file]. Retrieved
from https://www.arlstatistics.org/analytics
Carnegie Classification (2016). Custom listings –
basic classification: Carnegie classification of institutions of higher
education [Data file]. Retrieved from http://carnegieclassifications.iu.edu/
Cluff, E. D., & Murrah, D. J. (1987). The
influence of library resources on faculty recruitment and retention. Journal
of Academic Librarianship, 13(1), 19-23.
Dancey, C. P., & Reidy, J. (2011). Statistics without maths for psychology (5th ed.). Prentice Hall/Pearson.
Dundar, H., & Lewis, D. R. (1998). Determinants of research productivity
in higher education. Research in Higher Education, 39(6),
607-631.
Kaufman, P. T. (2008). The library as strategic
investment: Results of the Illinois return on investment study. Liber
Quarterly, 18(3-4), 424-436. https://doi.org/10.18352/lq.7941
Laakso M., & Bjork B. C. (2012). Anatomy of open access publishing: A
study of longitudinal development and internal structure. BMC Medicine, 10(124).
https://doi.org/10.1186/1741-7015-10-124
Murray, A., & Ireland, A. (2018). Provosts'
perceptions of academic library value & preferences for communication: A
national study. College & Research Libraries, 79(3), 336-365.
https://doi.org/10.5860/crl.79.3.336
National Science Foundation. (2016). Higher
education research development survey (HERD) [Data file]. Retrieved from https://www.nsf.gov/statistics/srvyherd/
Pallant, J. F. (2011). SPSS survival manual: A step by step guide to data
analysis using the SPSS program (4th ed.). Allen & Unwin.
Rawls, M. M. (2015). Looking for links: How faculty
research productivity correlates with library investment and why electronic
library materials matter most. Evidence Based Library and Information
Practice, 10(2), 34-44. https://doi.org/10.18438/B89C70
Rushton, J. P., & Meltzer, S. (1981). Research
productivity, university revenue, and scholarly impact (citations) of 169
British, Canadian and United States universities (1977). Scientometrics, 3(4), 275-303. https://doi.org/10.1007/bf02021122
Sutter, W. N. (2012). Introduction to educational
research: A critical thinking approach (2nd ed.). Sage Publications.
Tabachnick, B. G., & Fidell, L. S. (2001). Using
multivariate statistics (4th ed.). Boston, MA: Allyn and Bacon.
Tenopir, C., King, D., Edwards, S., & Wu, L. (2009). Electronic journals
and changes in scholarly article seeking and reading patterns. Aslib Proceedings: New Information Perspectives,
61(1), 5–32. https://doi.org/10.1108/00012530910932267
Tenopir, C., King, D. W., Spencer, J., & Wu, L. (2009). Variations in
article seeking and reading patterns of academics: What makes a difference? Library
& Information Science Research, 31(3), 139-148. https://doi.org/10.1016/j.lisr.2009.02.002
Tenopir, C., Mays, R., & Wu, L. (2011). Journal article growth and reading
patterns. New Review of Information Networking, 16(1),
4-22. https://doi.org/10.1080/13614576.2011.566796