College and Research Libraries


Research Notes 

Earnings Determinants 
of Library Faculty of the 
University System of Georgia 

W. Ken Farr and R. Neil Scott 

This study analyzes the effects of selected factors on earnings of library 
faculty employed at senior colleges of the University System of Georgia 
with a master of library science degree. The factors are also dissected 
to explore for any differences in their impacts on earnings by gender. 
Major findings are that earnings increase with experience, becoming a 
library director, greater supervisory responsibility, and higher academic 
rank, whereas a decrease in salary can be expected upon changing 
jobs. Results from the study also suggest that male and female library 
faculty earnings are determined in the absence of gender discrimina-
tion. A surprising finding is that intellectual contributions and additional 
graduate education are not directly rewarded with significant increases 
in earnings. 

D 
he major objectives of this 
paper are twofold. First, a 
model is developed to explore 
the influence of human capi-

tal, institutional, and personal character-
istics on the compensation of library fac-
ulty employed by senior colleges of the 
University System of Georgia. A multiple 
regression analysis is employed to find 
the statistically significant factors that 
influence the earnings of members of this 
population. Second, even though the 
study is conducted within the limitations 
of a segment of the larger universe of aca-
demic libraries in the United States, a 
major contribution of this paper is that it 
can serve as a model for other colleges and 

universities (or groups of such institu-
tions) that wish to study the structure of 
their own compensation systems. 

Other benefits of conducting studies 
such as this one are that results can be 
used as a basis from which to lobby to 
correct compensation discrepancies and/ 
or to develop a formula-based salary scale 
to reward activities and attributes of li-
brary faculty in a more predictable man-
ner. Researchers constructed an example 
of such a formula at Lamar University.1 

Readers interested in a more basic, 
descriptive analysis of the original sur-
vey from which the data of this study 
were gathered are referred to the authors' 
previous article in Southeastern Librarian. 2 

W. Ken Farris an Associate Professor of Economics at the J. W. Bunting School of Business, Georgia 
College, Milledgeville. R. Neil Scott is an Associate Professor and Coordinator of Information Services at 
the Russell Library, Georgia College, Milledgeville. 

77 



78 College & Research Libraries 

Review of the Literature 
In a nondiscriminating competitive labor 
market, marginal productivity theory 
suggests that wage differences depend 
largely upon differences in the produc-
tivity of labor. Other things being equal, 
higher incomes are generally expected to 
be associated with factors (e.g., education 
and experience) that enhance a person's 
labor productivity and vice versa. How-
ever, in real-world labor markets, market 
imperfections may also contribute to in-
come differences between individuals in 
specific labor markets because an 
individual's factor endowments may be 
evaluated differently based on, for ex-
ample, gender or race. In addition, wage 
differences of individuals may also re-
flect other factors that limit market par-
ticipation, such as specific job qualifi-
cations. 

Previous studies have found that 
wages in the academic setting are, in gen-
eral, significantly determined by such in-
stitutional factors as intellectual activities, 
job performance, experience, and admin-
istrative responsibilities.3-10 However, ear-
lier studies focused attention primarily on 
"traditional" academic faculty, thus the 
role that these factors play in determin-
ing the earnings of library faculty is un-
clear as rewards for factor endowments 
are often market specific. The authors, 
however, believe that it is reasonable to 
assume, a priori, that similar factors 
would also be important in explaining 
earnings variations in the academic labor 
market for library faculty. Support for this 
argument is found in human capital 
theory which asserts that, through time, 
individuals accumulate skills that en-
hance the productivity of our labor ser-
vices, based upon our education levels 
and work experience. 

This assertion is further supported by 
the "Standards for Faculty Status for Col-
lege and University Librarians" adopted 
and promoted by ACRL. 11 In recognition 
of these standards, which encourage 
equal treatment of academic library fac-

January 1996 

ulty with other comparably ranked in-
structional faculty, librarians employed 
by the University System of Georgia hold 
faculty status and are fully recognized by 
the Board of Regents as members of the 
"Corp of Instruction" at each institution.12 

Another basis for this assumption may be 
found in results from a dissertation which 
concluded that many of the same factors 
that influence academic faculty salaries in 
other disciplines are similar for academic 
librarians employed at medium-sized 
state-supported universities in the 
midwestern United States. 13 

Description of the Survey and Data 
The data used in this study were collected 
during the summer and fall of 1992 as 
part of a survey of all library faculty em-
ployed by the fourteen senior colleges of 
the University System of Georgia (a popu-
lation of 95). The authors gathered data 
from a tailored questionnaire designed to 
obtain descriptive information while 
maintaining personal confidentiality. The 
authors limited the data set to librarians 
with academic rank who possess an ALA-
accredited master's degree in library sci-
ence (MLS) since the ALA master's de-
gree is the minimum education level for 
those accepted for employment as a li-
brary faculty member in the University 
System of Georgia. The data were current 
for the budget year ending June 30,1992. 
The final sample consisted of sixty-eight 
usable questionnaires, reflecting a re-
sponse rate of 72 percent. 

Primary Model Specification 
The traditional human capital model 
forms the basis for the factors included 
to explain income variations of library 
faculty. In addition to factors indicating 
investments in human capital (HC), this 
conventional model is expanded to cap-
ture the effects of administrative respon-
sibilities (A), rank (R), and personal char-
acteristics (P). Also, since other studies 
have shown that earnings may vary by 
gender, this possibility is also explored.14-20 



The formal model is expressed as fol-
lows: 

where each component is explained as 
follows: 

The term y;, the dependent variable, is 
the natural logarithm of the annual (12-
month) salary of the ith library faculty 
member. The reason for the logarithmic 
form of salary is because of its prepon-
derance in similar academic studies as 
there is no strong theoretical support for 
the exact functional form of the depen-
dent variable. In an effort to improve re-
sponse rates, wage information was gath-
ered using categorical choices and the 
midpoint of each category was used as 
an approximation of salary. 21 

The explanatory variables begin with 
HC; which represents three human capi-
tal measures of the ith library faculty 
member. The first is experience, a continu-
ous variable that measures the number 
of years since receiving the MLS degree. 
The second is publications, a weighted 
composite continuous variable that rep-
resents the cumulative intellectual output 
of the ith library faculty member during 
his or her career. A weighting scheme is 
used to reduce the number of indepen-
dent variables and increase the degrees 
of freedom in the regressions models. The 
weights assigned to specific intellectual 
activities are: published books by a fac-
tor of 5; published academic research ar-
ticles by a factor of 2.5; academic research 
papers presented at professional organi-
zation meetings by a factor of 1.25; book 
reviews by a factor of .625; and any other 
"intellectual" activity reported in the sur-
vey by a factor of .3125. Although any 
weighting scheme is subjective and there 
is always the problem of evaluating qual-
ity versus quantity, the authors believe 
that the above scale is consistent in terms 
of the importance assigned to intellectual 
activities for academic librarians. In ad-
dition, weighting schemes such as the one 

Earnings Determinants 79 

described have been shown to be useful 
in similar empirical studies. 22,23 No at-
tempt was made to adjust this variable to 
differentiate between single and joint au-
thorship. The third and final human capi-
tal variable is used to capture the effects 
of a specialized subject degree. This is a 
dichotomous variable with a value of one 
if the ith library faculty member has one 
or more graduate subject degrees in ad-
dition to the MLS, zero if otherwise. This 
variable is included to capture benefits 
associated with additional degrees where 
it is expected that a supplemental 
degree(s) would enhance the operation of 
the library in that field . Each of the above 
variables is expected to have a positive 
impact on salary since each reflects an 
increase in human capital. 

The reason for the logarithmic form 
of salary is because of its preponder-
ance in similar academic studies as 
there is no strong theoretical support 
for the exact functional form of the 
dependent variable. 

The next explanatory component is A; 
which represents two administrative mea-
sures. The first is a dichotomous variable 
that takes a value of one if the ith library 
faculty member is the director of the li-
brary, zero if otherwise. This variable is 
included to capture the expected positive 
influence on salary of an individual ulti-
mately responsible for the operation of a 
college library. The second administrative 
variable is used to measu.re supervisory 
responsibilities of the ith library faculty 
member. This is a weighted composite 
continuous variable where the number of 
professional library faculty supervised is 
weighted by a factor of 1, the number of 
support staff supervised is weighted by 
a factor of .5, and the number of student 
assistants supervised is weighted by a 
factor of .25. This weighting scheme is 
again used to reduce the number of ex-
planatory variables and increase the de-



80 College & Research Libraries 

grees of freedom. It is also used to attach 
importance to the type as well as the num-
. ber of individuals supervised. As before, 
any weighting scheme is subjective and 
subject to debate. However, the authors 
believe that this scheme, as designed, is 
consistent with the importance assigned 
to levels of supervisory responsibility for 
library faculty. A positive impact on earn-
ings is expected from both of these ad-
ministrative measures. 

The term Ri is included to represent the 
· three academic ranks of assistant profes-
sor, associate professor, and full profes-
sor. Each is a dichotomous variable where 
a value of one is assigned according to 
the ith library faculty member's rank, zero 
if otherwise. These variables are included 
to capture rewards that are given for an 
amalgam of achievements accumulated 
by an individual during his or her career 
which are not explicitly accounted for in 
the model or which individually would 
not have a significant impact on wages. 
In association with an increase in rank 
often comes an increment in one's salary. 
It is therefore expected that each of these 
variables would have a positive impact 
on salary. 

The final explanatory component of 
the model is Pi which represents two per-
sonal characteristic measures. The first is 
the number of library employers (exclud-
ing the first employer) of the ith library 
faculty member during his or her years 
of employment since receiving the MLS 
degree. This continuous variable is in-
cluded based upon the hypothesis that in-
dividuals who change their employer 
more frequently, do so to enhance their 
salary. While economic theory does not 
provide direct evidence of the likely ef-
fect of this variable, other studies have 
found mobility to be a significant posi-
tive earnings determinant in acade-
mia. 24,25 The second personal characteris-
tic variable is gender, a dichotomous vari-
able that takes a value of one if the ith 
library faculty member is male, zero if 
otherwise. This variable is included to 

January 1996 

assist in exploring for gender bias in li-
brary faculty earnings. If differences ex-
ist in starting pay, other things constant, 
and/ or if differences exist in how factor 
endowments are evaluated for salary, 
gender discrimination may be the cause. 
Conversely, discrimination cannot be ar-
gued if differences in earnings are ex-
plained by factor endowments. A meth-
odology to explore for this possibility is 
developed in the next section. 

The component ei is a random distur-
bance term included to capture any un-
explained variation in earnings not ac-
counted for by the explanatory variables. 

Model Decomposition 
To investigate for differences in library 
faculty salaries by gender, begin by de-
composing the earnings model into its 
component parts, as suggested by Alan 
Blinder.26 To simplify this procedure, first 
rewrite equation (1) in more general terms 
as: 

where yi is the salary received by the ith 
individual, zji represents factor endow-
ment values taken by the jth explanatory 
variables of the ith individual, a and ~i 
represent unknown parameter coeffi-
cients to be estimated using least squares, 
and ei is a stochastic disturbance term. 
Equation (2) can be rewritten to represent 
a male or female sample as: 

- m ~ m "" n A'"z- m 
Y = a + ~i=l Pi i 

" f 
-t " t "" n A z-t 
Y =a +~i=I P i i 

(3) 

(4) 

where a and Pi represent estimates ofthe 
unknown parameters in equation (2), 
them and f superscripts refer to results 
from a sample of males or females, and 
the bar superscripts refer to sample 
means. 



Total differences in the average salaries 
of males and females can therefore be 
shown as: 

II m - m - f 
- m - f II m II f ~ n A z z 
y - y = (a - a) + ~i=t P i ( i - i ) + 

n _
1 

11 m 11 [ 

:Li=l z i <Pi - ~j ) (5) 
where ( - m _ y-1), which is referred to as the 
total eflect of the average salary differ-
ences, decomposes into the sum of three 
components: the constant effect, the en-
dowment effect, and the coefficient effect, 
respectively. To explain each component, 
first assume that salaries of female library 
faculty are determined in the absence of 
discrimination. This can be expressed as: 

II m- t 
- h II m A z 
Y =a +Pi i (6) 

whereyhrepresents female mean salaries 
when factor endowments of females are 
evaluated using the rewards structure for 
males. Earnings variations that are ex-
plained by differences in average factor 
endowments can then be shown as: 

which is the endowment effect shown in 
equation (5). Any remaining difference in 
total earnings between male and female 
library faculty is then referred to as the 
residual effect, defined as the difference 
between ( y" - y1 ) and shown as: 

- 11 m 
11 ! 

y"- y1 = (am- o/) + :Li:l ~ (~i - p i) (8) 
where the residual effect is the sum of the 
constant and coefficient effect in equation 
(5). The constant effect represents earn-
ings differentials unrelated to factor en-
dowment levels or the evaluation of said 
endowments for income determination. 
The coefficient effect, conversely, repre-
sents earnings variations as a result of the 
differences in how factor endowments are 
evaluated. This suggests that even if both 

- t 
male and female library faculty had Z i 

Earnings Determinants 81 

factor endowments, earnings differences 
would still exists. Given this, an argument 
for traditional gender discrimination may 
be made if the constant and/ or coefficient 
effect is significant and of the appropri-
ate sign, which in this context is expected 
to be positive since it is generally assumed 
that males earn more than females in com-
parable positions. 

Using the methodology of John Jack-
son and James T. Lindley, a statistical test 
for significance of the residual effect can 
be accomplished by combining the male 

To explain each component, first 
assume that salaries of female 
library faculty are determined in the 
absence of discrimination. 

and female sample data on the variables 
in equation (1). 27 A dichotomous variable 
to represent gender is included, as well 
as a complete set of gender interaction 
terms. The residual effect can be tested 
assuming the null hypothesis of no sig-
nificant differences in earnings by gender 
using the following test statistic: 

u= __ <s_s_E_R _-s_s_E~u_) _/_J __ _ (9) 
SSEU I ( T- K) 

where SSER is the sum of squares of the 
least squares errors obtained from the re-
stricted model excluding the gender and 
interaction terms, and SSEu is the sum of 
squares of the least squares errors ob-
tained from the unrestricted model which 
includes the gender and interaction terms. 
J represents the number of joint hypoth-
eses being tested; K, the number of pa-
rameters estimated in the unrestricted 
model; and T, the sample size. The test 
statistic u follows an F distribution with J 
and (T - K) degrees of freedom. 

The components of the residual effect 
can also be tested for statistical signifi-
cance. This is done by testing for the co-
efficient effect using the same procedure 
as above, except the gender dummy is 



82 College & Research Libraries 

added to the restricted model. The con-
stant effect is tested using at-test on the 
parameter estimate of the gender variable 
in the unrestricted model. 

Since some regression equations are es-
timated with pooled cross-sectional data 
using both male and female library faculty, 
correction for heteroscedastic errors is nec-
essary. Further, since tests for gender bias 
depend on regression errors, proper model 
specification becomes critical. Therefore, a 
RESET test was conducted and the model 
judged to be properly specified. 28-30 Finally, 
the presence of multicollinearity was tested 
using the matrix decomposition procedure 
suggested by David A. Belsley, Edwin Kuh, 
and Roy E. Welsch, and judged to be of no 
consequence for any of the estimated re-
gression equations.31 

The empirical evidence suggests that 
the greatest addition to earnings 
comes from being a library director. 

Presentation and Analysis of Results 
Table 1 presents parameter estimates of 
five separate regressions related to equa-
tion (1). All five are necessary to accom-
plish the objectives of the study since each 
contains information that is needed for 
statistical analysis. 

Overall, each equation is statistically 
significant at the a = .01 level, with the 
lowest estimated R2 of .85 for the female 
model. This suggests that the specified 
model does an acceptable job of explain-
ing wage variations of library faculty. 
Exploring for significant differences in 
salary by gender begins by reviewing the 
parameter estimates from the male and 
female regressions (see columns 1 and 2). 
These suggest that several common fac-
tors significantly influence salary, includ-
ing experience, appointment as library di-
rector, supervisory responsibility, and the 
rank of associate professor. However, dif-
ferences are also shown to exist. Addi-
tional significant variables for the males 
regression include publications, the rank 

January 1996 

of assistant professor, and the number of 
past employers. An additional significant 
variable for the females regression is the 
rank of full professor. Other differences 
include the intercepts of the two regres-
sions, which suggest a lower starting sal-
ary for females, and the magnitude of the 
coefficients on the explanatory variables, 
which imply differences in how unit 
changes in these variables impact salary. 
While differences appear to exist, the 
question becomes: are the visible differ-
ences statistically significant? 

The answer to this question is given at 
the bottom of table 1 where the estimated 
mean salary of men ( }"") is shown to be 
only 4.63 percent higher than for women 
( y' ), which was defined earlier to be the 
total effect. Roughly 28 percent of this dif-
ference is explained by factor endow-
ments (the endowment effect), while the 
remaining 72 percent (the residual effect) 
is a result of other causes. Using equation 
(9), a test of statistical significance on the 
residual effect is performed and found to 
be insignificant. In addition, a test of the 
differences in starting pay (the constant 
effect) is also shown to be insignificant. 
Moreover, the evaluation of factor endow-
ments that determine salary (the coeffi-
cient effect) is not significantly different 
between the sexes. Results of these tests 
suggest that there are no significant dif-
ferences in the earnings of library faculty 
by gender that cannot be explained in eco-
nomic terms. As such, the two groups can 
be "pooled" and treated as identical 
within the context of the regression 
model. The appropriate regression for 
further analysis then becomes the pooled 
model without gender dummy or inter-
action terms (see column 3). 

Regression (3) results in table 1 are sum-
marized in dollar equivalents in figure 1. 
These estimates show that an individual 
with an MLS from an ALA -accredited pro-
gram holding the rank of instructor in a 
senior college of the University System of 
Georgia with zero values for the explana-
tory variables would have had a starting 



Earnings Determinants 83 

TABLEt 

Earnings Model Parameter Estimates 
(t-values are given in ~arentheses) 

(1) (2) (3) (4) (5) 
Pooled 

Pooled Pooled Model 
Model No Model Gender 

Males Females Gender Gender and Males Females 
Variable Model Model Dummy Dummy Interaction Means Means 
Constant 10.1469 10.0703 10.0835 10.0692 10.0703 

(205.6)" (224.9)" (333.0)" (310.4)" (224.9)" 

Human Capital Variables 
Experience 0.0110 0.0075 0.0087 0.0084 0.0075 12.600 12.875 

(3.466)" (2.303)b (4.183)• (3.981)• (2.303)b 

Publications 0.0052 -0.0013 0.0012 0.0013 -.0013 5.906 7.129 
(2.290)b ( -0.996) (1.124) (1.175) (-.996) 

Subject -0.0595 0.0067 -0.0027 -0.0129 0.0067 0.550 0.292 
Degree ( -1.654) (0.150) (-0.114) (-0.509) (0.150) 

Administrative Variables 
Director 0 .3 611 0.3840 0.3324 0.3423 0.3840 0.150 0.167 

(5.232)" (6.311)• (8.556)• (8.627)• (6.312)• 

Supervisory 0.0112 0.0138 0.0102 0.0092 0.0138 6.163 3.250 
Responsibility (3.021)b (2.871)• (4.780)" (4.082)" (2.872)• 

Rank Variables 
Assistant 0.1150 0.0492 0.0847 0.0897 0.0492 0.400 0.479 

Professor (2.183)" (0.999) (2.727)• (2.865)• (0.998) 

Associate 0.1595 0.1607 0.1674 0.1792 0.1607 0.250 0.250 
Professor (2.194)" (2.505)b (3.831)• (4.005)• (2.506)b 

Full 0.0636 0.4280 0.2387 0.2465 0.4280 0.050 0.042 
Professo r (0.449) (3.461 )• (3.003)" (3.089)" (3.461)" 

Personal Variables 
Number of -0.0539 -0.0094 -0.0172 -0.0160 -0.0094 2.350 2.708 

Employers (-2.242)b (-1.150) (-2.303)b (-2.126)b ( -1.150) 

Gender 0.0307 0.0767 0.294 0.706 
Dummy (1.224) (1.150) 

Gender Interaction Variables 0.0035 
Gender-Experience (0.777) 

0.0065 
Gender-Publications (2.476)b 

-0.0662 
Gender-Subject Degree (-1.153) 

-.0229 
Gender-Director (-0.249) 

-0.0026 
Gender-Supervisory (-0.424) 
Responsibility 

0.0657 
Gender-As sistant (0.911) 
Professor 

-0.0012 
Gender-Associate (-0.013) 
Professor 

-0.3644 
Gender-Full ( -1.937)c 
Professor 

-0.0445 
Gender-Number (-1.751)" (Cont.) 
of Employers 



84 College & Research Libraries January 1996 

TABLE 1 cont. 
(1) (2) (3) (4) (5) 

Pooled 
Pooled Pooled Model 

Model No Model Gender 
Males Females Gender Gender and Males Females 

Variable Model Model Dummy Dummy Interaction Means Means 

F-value 36.0400" 23 .7940" 50.9930" 46.2800" 28.5470" 
R-squared 0.97010 0.84930 0.88780 0.89030 0.91870 
M.S.E. 0.00436 0.01353 1.00000 1.00000 1.00000 
S.S .E. 0.04356 0.51423 0.71417 0.70541 0.55779 
T 20 48 68 68 68 
Mean In Salary 10.36952 10.32420 10.33753 10.33753 10.33753 

Differences in Logarithms 
Percentage Differences 

in Nominal Terms 

ym = 10.3695 
yh = 10.3569 
y = 10.3242 

Total Effect = ym- yf = .0453 
Endowment Effect= ym - yh = .0126 
Residual Effect = l- yf = .0327 

4.63 
1.27 
3.32 

Constant Effect = Gender coefficient in the full Interaction Model = .076649 

Coefficient Effect = Residual Effect - Constant Effect = .0327 - .076649 = -0.043949 

u for Residual Effect= [( .71417- .55779) I 10] I [( .55779 I ( 68- 20 )] = 1.3457 
F critical= 2.0346 (a.= .05) 

u for Coefficient Effect= [( .70541 - .55779) I 9] I [( .55779 I (68- 20 )] = 1.41147 
F critiCal = 2.0817 (a. = .05) · 

• Significant at the a= .01 level 
b Significant at the a= .05 level 
c Significant at the a = .1 0 level 

salary of $23,945 for fiscal year 1992. 
Among the human capital factors in-

cluded in the model, only experience is 
found to be statistically significant. Regres-
sion estimates show that earnings increase 
with each year of experience by a nominal 
.87 percent, or $208 per year. Both of the 
administrative variables are found to sig-
nificantly improve salaries. The position of 
library director is estimated to enhance 
earnings by a nominal 39.43 percent, or 
$9,441. Responsibility for directly supervis-
ing employees is estimated to increase sal-
ary by $245 for each professional library 
faculty member supervised, $123 for each 
support staff member supervised, and $61 
for each student supervised. 

All academic ranks are found to be sta-
tistically significant. Salary increases 
nominally with the rank of assistant pro-
fessor by approximately 8.8 percent, or 
$2,116, over the salary of an instructor. 

Likewise, the rank of associate professor 
is associated with a salary increase of a 
nominal 18.2 percent, or $4,363, over the 
rank of instructor. For a full professor the 
increase is 26.9 percent, or $6,455. 

The last factor found to be statistically sig-
nificant is the number of employment 
changes. While many other studies found this 
variable to positively increase earnings, esti-
mates show that each employment change 
since receiving the ALA-accredited MLS de-
gree causes salary to decrease nominally by 
1.7 percent, or $409. This suggests that library 
faculty change jobs for reasons ranked higher 
in importance than earnings. Moreover, re-
gardless of the reason for the job change, the 
library faculty in this study had to accept sig-
nificantly lower salaries when changing em-
ployers. 

Summary and Conclusions 
This earnings study of library faculty em-



ployed by senior colleges of the Univer-
sity System of Georgia found no evidence 
to suggest that gender discrimination 
plays any role in determining salary. In 
other words, the same set of factors is sig-
nificant in determining salary for both 
males and females, as well as how they 
are evaluated in setting salaries. 

The significant factors found to in-
crease earnings for library faculty are ex-
perience, administrative responsibilities, 
and faculty rank. The empirical evidence 
suggests that the greatest addition to earn-
ings comes from being a library director. 
This conclusion is not unexpected since 
the individual in this position is ulti-
mately responsible for the overall day-to-
day operations of the library. However, 
results also indicate that significant in-
creases in salary can also be achieved by 
those individuals who are not library di-

Earnings Determinants 85 

rectors by assuming additional supervi-
sory responsibilities. 

Experience is another factor found to 
increase earnings significantly. Interest-
ingly, the evidence suggests that earnings 
increase linearly with experience, vis-a-vis, 
as theoretically expected in a nonlinear 
fashion. Because human capital theory sug-
gests that with greater experience comes 
greater productivity, the study's empirical 
evidence is unclear as to the exact economic 
interpretation of experience's impact on 
earnings. Does salary increase with expe-
rience because of increased productivity or 
simply because of longevity? This issue 
warrants further study. 

The significance of faculty rank vari-
ables explains the rewards given to indi-
viduals based on a composite of factors 
not specifically included in the study or 
which individually do not affect salary. 

FIGURE 1 

1992 Fiscal Year Salary Estimate of College Librarians in the 
University System of Georgia Based on the Results of This Study 

I. Base pay of ALA-accredited MLS academic library 
faculty member with rank of instructor: $23 ,945 

II. Add: $208 for each year of 
post-MLS experience: ($208 X_ Years) = + __ 

$9,441 if hold position of library director: ($9,441) = + __ 
Number of people directly supervised: 

Faculty rank: 

($245 X_ faculty) = + 
($123 X_ support staff)= + 
($61 X_ student assistants)= + 

($2, 116 for assistant professor) 
($4,363 for associate professor) 
($6,455 for professor) 

Total = 

III. Subtract: Number of post-MLS employers, excluding 
present employer: 

($409 X_ employers) = 

Predicted Salary 

+ __ _ 

+ __ _ 

$ __ 



86 College & Research Libraries 

Two factors which were included in this 
study and not found to be significant in 
salary determination were publications 
and graduate academic degrees in addi-
tion to the MLS. While these factors indi-
vidually may not significantly affect sal-
ary, they may be important to specific ad-
ministrators who make promotion deci-
sions; hence increments given for higher 
ranks can indirectly reflect these factors. 

The study also shows that library fac-
ulty who change employers find that each 
move significantly lowers their salary. 
This finding is somewhat surprising since 
similar studies of other groups have 
shown this factor to affect earnings posi-
tively. This suggests that library faculty 
change jobs for reasons ranked higher in 
importance than to seek a higher salary. 
Two peculiarities of this market may help 
explain why this is true. First, in the la-
bor market for library faculty, positions 
are often advertised and funded based on 
minimum qualifications. However, it is 
not. uncommon to find a large number of 
applicants for each vacant position. As a 
result, employers often find the applicant 
pool to include many candidates with 
qualifications far exceeding the minimum 
requirements of the position advertised. 
This allows an opportunity for the em-
ployer to fill the position with a person 
who has a current salary, which is based 
on the qualifications of the individual in 
his or her current position, that is higher 
than the amount allocated for the posi-
tion the individual is seeking. This means 
that if the individual accepts the offered 
position, he or she will have to accept a 
reduction in earnings. 

A second peculiarity related to this find-
ing can be found in the fact that the aca-

January 1996 

demic library labor market is dominated 
by females (70 percent in the sample for 
this study). Numerous studies have shown 
that females tend to have lower reserva-
tion wages than men. That is, the minimum 
earnings that a female will accept for a 
given position is generally lower than those 
for men. These facts help explain, based on 
the results of this study, the obvious will-
ingness of library faculty to accept lower 
salaries when changing employers. Brenda 
Major and Janice J. Kirkland provide are-
view of the literature, with an examination 
and discussion of gender differences re-
lated to personal entitlement with respect 
to payment for work performed.32,33 

A final observation concerns empirical 
evidence that suggests that intellectual ac-
tivities are not directly rewarded with sig-
nificant increases in earnings. This finding 
is contrary to previous studies of academic 
faculty disciplines and probably affects the 
amount of effort exerted by library faculty 
to contribute in this area. (As noted earlier 
in the paper, promotion increments can 
capture the impact of intellectual contribu-
tions on earnings.) Given ACRL' s desire to 
encourage college administrators to treat 
academic library faculty equally to com-
parably ranked faculty in other disciplines, 
this finding suggests that the reward struc-
ture for research activities may not be in 
place to support the promotion provisions 
of the "Standards." However, whether the 
lack of reward for intellectual activities is 
peculiar to library faculty at the senior col-
leges of the University System of Georgia 
is unclear. It may be that the research ac-
tivities of faculty in all disciplines at these 
institutions are not directly rewarded. Ad-
ditional research is needed to provide fur-
ther insight into this aspect. 

Notes 

1. Jeanie M. Welch, "Suggested Guidelines for Salary Determination in an Academic Library," 
in Energies for Transition : Proceedings of the Fourth National Conference of the Association of College 
and Research Libraries (Chicago: ACRL, 1986), 152-56. 

2. R. Neil Scott, W. Ken Parr, E. Lorene Flanders, and Bonnie E. Spiers, "Professional and 
Salary Characteristics of Librarians Employed by Senior Colleges of the University System of 
Georgia," Southeastern Librarian 44, no. 3 (fall1993): 46-51 . 



Earnings Determinants 87 

3. Josef M. Broder and Rod F. Ziemer, "Determinants of Agricultural Economics Faculty 
Salaries," American Journalof Agricultural Economics 64 (May 1982): 301-3. 

4. Charles D. Delorme Jr., R. Carter Hill, and Norman J. Wood, "Analysis of a Quantitative 
Method of Determining Faculty Salaries," Journal of Economic Education (fall 1979): 20-25. 

5. M. A Ferber, ''Professors, Performance, and Rewards," Industrial Relations (Feb. 1974): 69-77. 
6. D. A Katz, "Faculty Salaries, Promotions, and Productivity at a Large University," Ameri-

can Economic Review (June 1973): 469-77. 
7. Matt L. Riggs, Ronald G. Downey, Patricia E. Mcintyre, and Donald P. Hoyt, "Using Dis-

criminant Analysis to Predict Faculty Rank," Research in Higher Education 25, no. 4 (1986): 365-76. 
8. Howard P. Tuckman and Jack Leahey, "What Is an Article Worth?" Journal of Political 

Economy 83, no. 5 (1975): 951-67. 
9. Howard P. Tuckman, James H. Gapinski, and Robert P. Hagemann, "Faculty Skills and 

the Salary Structure in Academe: A Market Perspective," American Economic Review (Sept. 
1977): 692-702. 

10. William C. Weiler, "Integrating Rank Differences into a Model of Male-Female Faculty 
Salary Discrimination," Quarterly Review of Economics and Business 30, no. 1 (1990): 3-15. 

11. "Standards for Faculty Status for College and University Librarians," College & Research 
Libraries News (May 1992): 317-18. 

12. Policy Manual: Board of Regents, University System of Georgia, 2nd ed, (Atlanta: University 
System of Georgia, 1982): 19-30. 

13. Ung Chon Kim, "A Statistical Study of Factors Affecting Salaries of Academic Librarians 
at Medium-Sized State-Supported Universities in Five Midwestern States," (Ph.D. diss., Indi-
ana Univ., 1980). 

14. Andrew P. Barkley, "Earnings of Kansas State University Agriculture Graduates: 1978-
88," American Journal ofAgricultural Economics (Feb. 1992): 215-22. 

15. Alan S. Blinder, "Wage Discrimination: Reduced Form and Structural Estimates," Journal 
of Human Resources (fall1973): 436-55. 

16. Ronald Oaxaca, "Male-Female Wage Differentials in Urban Labor Markets," International 
Economic Review (Oct. 1973): 693-709. 

17. John D. Jackson and James T. Lindley, "Measuring the Extent of Wage Discrimination: A 
Statistical Test and a Caveat," Applied Economics 21 (Apr. 1989): 515-40. 

18. James T. Lindley, Mary Fish, and John Jackson, "Gender Differences in Salaries: An Appli-
cation to Academe," Southern Economics Journal (Oct. 1992): 241-59. 

19. Kenneth Rosenzweig, Elizabeth Gustafson, and Lawrence Hadley, "An Economic Analy-
sis of the Earnings of Industrial Accounts," Journal of Economics and Finance 18, no. 1 (spring 
1994): 125-38. 

20. Weiler, "Integrating Rank Differences," 3-15. 
21. Warren P. Preston, Josef M. Broder, and Maria Cristina P. Almero, "Temporal Analysis of 

Income Earned by Former Agriculture Students," American Journal of Agricultural Economics 
(Feb. 1990): 13-23. 

22. Broder and Ziemer, "Determinants of Agricultural Economics," 301-3. 
23. Delorme, Hill, and Wood, "Analysis of a Quantitative Method," 20-25. 
24. Broder and Ziemer, "Determinants of Agricultural Economics," 301-3. 
25. Katz, "Faculty Salaries, Promotions, and Productivity," 469-77. 
26. Blinder, "Wage Discrimination," 436-55 . 
27. Lindley, Fish, and Jackson, "Gender Differences in Salaries," 241-59. 
28. James B. Ramsey, "Tests for Specification Errors in Classical Linear Least-squares Regres-

sion Analysis," Journal of the Royal Statistical Society, series B (1969): 350-71. 
29. Jerry G. Thursby and Peter Schmidt, "Some Properties of Test for Specification Error in a 

Linear Regression Model," Journal of the American Statistical Association 72, no. 359 (Sept. 1977): 
635-41. 

30. Jerry G. Thursby, "Alternative Specification Error Tests: A Comparative Study," Journal of 
the American Statistical Association 74, no. 365 (Mar. 1979): 222-25. 

31. 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). 

32. Brenda Major, Dean B. McFarlin, and Diana Gagnon, "Overworked and Underpaid: On 
the Nature of Gender Differences in Personal Entitlement," Journal of Personality and Social Psy-
chology 47, no. 6 (1984): 1399-1412. 

33. Janice J. Kirkland, "Equity and Entitlement: Internal Barriers to Improving the Pay of 
Academic Librarians," College & Research Libraries 52 (July 1991): 375-80.