1. ARTICLE IN PRESS
Economics of Education Review 25 (2006) 543–553
www.elsevier.com/locate/econedurev
Cost efficiency in the university: A departmental
evaluation model
Vı´ ctor M. GimenezÃ, Jose Luis Martı´ nez
´ ´
`noma de Barcelona, 08193 Bellaterra, Barcelona, Spain
Departament d’Economia de l’Empresa, Universitat Auto
Received 4 March 2004; accepted 3 May 2005
Abstract
This article presents a model for the analysis of cost efficiency within the framework of data envelopment analysis
models. It calculates the cost excess, separating a unit of production from its optimal or frontier levels, and, at the same
time, breaks these excesses down into three explanatory factors: (a) technical inefficiency, which depends on the quality of
the factors consumed, the type of organization and the factor of human behaviour; (b) the availability of the fixed factors
along with their level of utilization and the factors mix; and finally (c) the scale or size of the unit of production. The
empirical application is carried out on the departments of the Autonomous University of Barcelona. The results show that
departmental costs could be reduced on average by more than 13.46% in the long term.
r 2005 Elsevier Ltd. All rights reserved.
JEL classification: C61; H52; I21
Keywords: Costs; Efficiency; Input output analysis; Economics of scale
1. Introduction tools for allocating scarce economic resources more
efficiently among departments. At the same time,
The environment in which Spanish universities such instruments should also prove useful in
operate has changed significantly since 2001, when providing university administrators with unequivo-
the Organic Law for Universities introduced re- cal criteria for evaluating, and subsequently im-
forms into the sector. Quality improvement became proving, the performance of the operational units,
a priority requirement, and this, combined with the which in this case are the departments.
budgetary constraints already faced by the univer- In this study, we propose an instrument for
sities for some years, has resulted in a new and more departmental evaluation in terms of costs that can
complex framework for universities to operate in. be used to determine improvement targets both in
Henceforth they will require new administrative costs and production and in quality levels. The
instruments capable of providing institutions with instrument presented is a new model of cost
efficiency based on Data Envelopment Analysis
ÃCorresponding author. Tel.: +34 935811209; (DEA) methodology initially developed by Charnes,
fax: +34 935812555. Cooper, and Rhodes (1978). The specific objectives
´
E-mail address: victor.gimenez@uab.es (V.M. Gimenez). of this model are to determine: (a) the overall
0272-7757/$ - see front matter r 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.econedurev.2005.05.006
2. ARTICLE IN PRESS
544 ´nez, J.L. Martınez / Economics of Education Review 25 (2006) 543–553
V.M. Gime ´
deviation in costs of each department under certain behaviors that are economically undesir-
analysis, understood as the difference between the able—since they do not minimize costs—and can be
observed cost and the optimal long-term cost considered technically efficient. Thus, this paper
assuming an optimal scale; (b) the factors into presents a model of cost efficiency in preference to
which this deviation can be sub-divided, distinguish- one based exclusively on measuring technical
ing between technical inefficiency, the incorrect efficiency.
allocation of factors, the level of utilization of the DEA models constitute an excellent instrument
available fixed factors or the appropriateness of the for university evaluation that is supported in
scale adopted; (c) targets for the different inputs and various studies, although the objectives of these
outputs that would lead to cost efficiency in the studies generally differ from ours. Among the
departments; and (d) which inefficiencies can be authors that have written about it we can mention
corrected in the short term, and which in the Beasley (1990, 1995), Athanassopoulos and Shale
long term. (1997), Sarrico, Hogan, Dyson, and Athanassopou-
One of the main advantages of frontier models is los (1997), Johnes and Johnes (1993), Madden,
their ability to evaluate the overall efficiency of a Savage, and Kemp (1997), Ahn (1987), Glass,
group of units on the basis of the inputs consumed Mckillop, and O’Rourke (1998), Emrouznejad and
and the outputs produced. For this reason, they are Thanassoulis (2005), Caballero, Galache, Gomez, ´
particularly appropriate for application to sectors Molina, and Torrico (2001, 2004), Post and Spronk
with complex productive processes, such as (1999), Li and Reeves (1999), Korhonen (2000) and
universities, where there is a lack of information Korhonen, Tainio, and Wallenius (2001).
about output prices indicating the criteria The remainder of this article is organized as
for evaluating the performance of each decision follows. The second section describes the evaluation
making unit (DMU). In such circumstances, fron- model on which this work is focused. The char-
tier models offer a valuable, objective tool for acteristics of the sample and a description of the
evaluating the public sector, which is further variables are presented in Section 3. The following
supported by extensive literature (see, for example, section outlines the most relevant results obtained in
Bifulco & Bretschneider, 2001; Drake & Simper, our application of the model to all the departments
2003; Grosskopf, Margaritis, & Valdmanis, 1995; in the Autonomous University of Barcelona (UAB).
Pedraja-Chaparro & Salinas-Jimenez, 1996; Ray, Finally, the most significant conclusions are pre-
1991). sented.
However, merely estimating technical efficiency
as a criterion of evaluation and control is not
2. The model
generally sufficient in the public sector, even though
administrators in the sector have traditionally
With the aim of making the theoretical model as
neglected the conventional monetary evaluation
general an application as possible, we will consider
criteria of the market economy, such as profit and
the case of k ¼ 1; . . . ; K university departments. The
profitability. The public nature of the sector does
productive process of a department k is character-
not necessarily mean that economic criteria should
ized by the production of a group of i ¼ 1; . . . ; I
not or cannot be employed. In this respect,
outputs yk ¼ ðyk;1 ; . . . ; yk;I Þ with quality levels in
governments are increasingly administering the
q ¼ 1; . . . ; Q dimensions Qk;q ¼ ðQk;1 ; . . . ; Qk;Q Þ
public sector more efficiently, effectively and eco-
starting from f ¼ 1; . . . ; F inputs adjustable in the
nomically, regardless of their political allegiances.
long term xk;f ¼ ðxk;1 ; . . . ; xk;F ; Þ and v ¼ 1; . . . ; V
Their aim is to cut public spending without
inputs adjustable in the short term xk;v ¼
eliminating services or prejudicing quality. The use
ðxk;1 ; . . . ; xk;V Þ and input prices wk;f ¼ ðwk;1 ; . . . ;
of monetary-based criteria in the public sector
wk;F ; Þ and wk;v ¼ ðwk;1 ; . . . ; wk;V Þ, respectively. Con-
generally implies an analysis from the cost perspec-
sequently, the observed total costs of a department
tive, since selling prices are seldom involved and, if
k for a level of production (yk) and of quality (Qk,q),
so, they are usually pre-established. The increasing
is represented by
demand for control and efficiency in the adminis-
tration of monetary resources has thrown doubt on X
V X
F
technical efficiency as the most appropriate evalua- TCk ¼ wk;v xk;v þ wk;f xk;f .
tion criterion. In fact, it is now well known that v¼1 f ¼1
3. ARTICLE IN PRESS
´nez, J.L. Martınez / Economics of Education Review 25 (2006) 543–553
V.M. Gime ´ 545
The model breaks down total cost inefficiency X
V
(TCVk) of department k into three factors: technical ¼ wk;v ðxlr-vrs À xlr-crs Þ
k;v k;v
v¼1
inefficiency (TEVk), fixed factors utilization (FCVk)
and scale (SVk). This is expressed as follows: X
F
þ wk;f ðxlr-vrs À xlr-crs Þ.
k;f k;f ð5Þ
TCV ¼ TC À TClr-crs ¼ TEV þ FCV þ SV
k k k k k k f ¼1
¼ t
ðTCk À TCk Þ þ ðTCk Àt
TClr-vrs Þ
k The deviation in the utilization of fixed factors
(FCVk) is the deviation in cost due to differences
þ ðTCk -vrs À TClr-crs Þ,
lr
k ð1Þ
between the cost that unit k should achieve if it
would be efficient in the long term, maintaining a
where
similar scale, and the minimum cost if it would be
X
V X
F technically efficient.
TCt ¼
k wk;v xt þ
k;v wk;f xk;f (2) !
v¼1 f ¼1 X
V X
F
FCVk ¼ ðTCt
k À TClr-vrs Þ
k ¼ wk;v xt
k;v þ wk;f xk;f
v¼1 f ¼1
is the total cost that the DMU k should achieve if it !
X
V X
F
were technically efficient; xt is the variable inputs
k;v À wk;v xlr-vrs þ wk;f xlr-vrs
k;v k;f
level associated with technical efficiency; v¼1 f ¼1
X
V X
F X
V
TClr-vrs ¼
k wk;v xlr-vrs þ
k;v wk;f xlr-vrs
k;f (3) ¼ wk;v ðxt À xlr-vrs Þ
k;v k;v
v¼1 f ¼1 v¼1
X
F
is the total cost that the DMU k should achieve if it þ wk;f ðxk;f À xlr-vrs Þ.
k;f ð6Þ
were efficient in the long term—i.e., if it were able to f ¼1
change the fixed and variable input mix, while If a high level of substitutability among the
maintaining a similar scale (assuming variable different inputs existed, FCVk would also show the
returns to scale); xlr-vrs and xlr-vrs are the associated
k;v k;f impact that wrong decisions in the composition of
levels of variable and fixed inputs, respectively; the mix of factors can have on cost minimization. In
X
V X
F other words, by considering input prices, it is
TClr-crs ¼
k wk;v xlr-crs þ
k;v
lr-
wk;f xk;fcrs (4) possible for FCVk to detect a non-optimal mix of
v¼1 f ¼1 factors. Consequently, in this situation it could also
be considered an indicator of allocative inefficiency.
is the cost that the DMU k should achieve if it were Finally, the technical inefficiency deviation
cost efficient in the long term and able to adapt its scale (TEVk) measures the increase in total costs caused
to the optimal size; xlr-crs and xlr-crs are the associated
k;v k;f by an excessive consumption of factors. This may
levels of variable and fixed inputs, respectively. occur in situations of management incompetence,
The scale deviation (SVk) shows the excess in errors of organization or lack of incentives, which
costs due to differences between the average cost of can be explained partially by the lack of competi-
the activity that minimizes the costs globally and the tion, according to Leibenstein’s (1966) X-efficiency
frontier value relative to the level of production of theory. It is expressed as
the DMU k. This deviation only includes effects of !
X
V X
F
scale and does not assume any type of inefficiency in TEVk ¼ ðTCk À TCt Þ ¼ wk;v xk;v þ wk;f xk;f
k
the utilization of the factors, either fixed or variable. v¼1 f ¼1
The mathematical expression of SVk is !
X
V X
F
À wk;v xt
k;v þ wk;f xk;f
SVk ¼ ðTClr-vrs À TCk -crs Þ
k
lr
v¼1 f ¼1
!
X
V X
F X
V
¼ wk;v xlr-vrs
k;v þ wk;f xlr-vrs
k;f
¼ wk;v ðxk;v À xt Þ.
k;v ð7Þ
v¼1 f ¼1 v¼1
!
X
V X
F A prerequisite for calculating the above devia-
lr-
À wk;v xk;vcrs þ wk;f xlr-crs
k;f tions is the gathering of the intermediate optimal
v¼1 f ¼1
total costs, denoted as TC. For each one, we use an
4. ARTICLE IN PRESS
546 ´nez, J.L. Martınez / Economics of Education Review 25 (2006) 543–553
V.M. Gime ´
ad hoc DEA model. TCt is calculated by solving the
k s:t:
following linear program for each department: P
K
V xlr-vrs À
k;v Zs Á xs;v ¼ 0 v ¼ 1; . . . ; V ;
P P
I s¼1
min gk À S s;v þ Sk;i P
K
v¼1 i¼1 xlr-vrs À
k;f zs Á xs;f ¼ 0 f ¼ 1; . . . ; F ;
s:t: s¼1
P
K P
K
gk Á xk;v À zs Á X s;v À Sk;v ¼ 0 v ¼ 1; . . . ; V ; Àyk;i þ zs Á ys;i À S k;i ¼ 0 i ¼ 1; . . . ; I; (9)
s¼1 s¼1
P
K P
K
xk;f À zs Á X s;f ¼ 0 f ¼ 1; . . . ; F ; ÀQk;q þ zs Á Qs;q ¼ 0 q ¼ 1; . . . ; Q;
s¼1 s¼1
P
K P
K
Àyk;i þ zs Á ys;i À S k;i ¼ 0 i ¼ 1; . . . ; I; zs ¼ 1;
s¼1 s¼1
P
K xlr-vrs ; xk;fvrs ; zs ; Sk;i ; S k;q X0:
k;v
lr-
ÀQk;q þ Z s Á Qs;q ¼ 0 q ¼ 1; . . . ; Q;
s¼1
Linear program (9) determines the levels of both
P
K
Z s ¼ 1; the variable inputs ðxlr-vrs Þ and the fixed inputs
k;v
s¼1 ðxlr-vrs Þ that minimize the long-term total costs. The
k;f
gk ; Z s ; Ss;v ; Sk;i ; difference between the optimum cost obtained by
(8) and by (9) can be assigned to adjustments in the
(8) long run changeable costs and/or changes in the mix
1
where e is a non-Archimedean constant. The linear of possible substitutable costs. In this way, it can
program (8) is very similar to other usual formula- reflect aspects relating to the fixed capacity utiliza-
tions of DEA programs for calculating technical tion and/or allocative ones respectively.
efficiency (see Cooper, Seiford, Tone, 2000, for Finally, the minimum total cost ðTClr-crs Þ, assum-
k
further details). The objective is to determine the ing a technological environment with constant
maximum reduction of the variable inputs main- returns to scale (CRS), is given by:
taining the observed levels of production and
X
V X
F
quality. Quality has been introduced as an output TCk -crs ¼ min
lr
wk;v xlr-crs þ
k;v
lr-
wk;f xk;fcrs
(Adler Berechman, 2001; Dismuke Sena, 2001; v¼1 f ¼1
Olesen Petersen, 1995). Their associated con- !
X
I X
Q
straints have an equal sign, under the supposition À S k;i þ Sk;q
that university quality cannot be modified in the i¼1 q¼1
short term. Variable returns to scale (VRS) are also
supposed, in order to eliminate the potential scale s:t:
effect, thus ensuring comparisons between depart- P
K
ments of similar size. xlr-crs À
k;v zs Á xs;v ¼ 0 v ¼ 1; . . . ; V ;
s¼1
Once the optimal value gà is calculated, the
k P
K
variable inputs level associated with technical xlr-crs À
k;f zs Á xs;f ¼ 0 f ¼ 1; . . . ; F ;
efficiency can be calculated as xt ¼ gà Á xk;v :The
k;v k
s¼1
frontier level of the long-term costs ðTClr-vrs Þ
k P
K
Àyk;i þ zs Á ys;i À S k;i ¼ 0 i ¼ 1; . . . ; I;
without scale changes is obtained from the follow- s¼1
ing program: P
K
ÀQk;q þ zs Á Qs;q À Sk;q ¼ 0 q ¼ 1; . . . ; Q;
X
V X
F
TCk -vrs ¼ min wk;v xlr-vrs þ wk;f xlr-vrs
lr s¼1
xlr-crs ; xlr-crs ; zs ; S k;i ; Sk;q X0:
k;v k;f
v¼1 f ¼1 k;f k;v
!
X
I X
Q
(10)
À Sk;i þ S k;q
i¼1 q¼1 The principal difference between (9) and (10) is
that (10) permits the use of reference units of very
1
The constant e is an infinitesimal 0oo1=N for all positive different sizes to that analyzed. The reference units
integer N. For calculating we usually assign to it a value of 10À6. in (9) are those departments that obtain an output
5. ARTICLE IN PRESS
´nez, J.L. Martınez / Economics of Education Review 25 (2006) 543–553
V.M. Gime ´ 547
equal to or greater than the one analyzed at a lower university administrators. After several interviews
cost, while in (10), the reference units could also be guided by the authors, and taking into account the
departments with a lower output in absolute terms available information, the following outputs were
at a lower cost or a higher output at a greater cost. finally selected as the most representative of the
In both cases, the model could choose them as productive process at UAB:
reference units, once they were rescaled to obtain an
equal or greater output at an equal or lower cost
than the one analyzed. This means that both New ‘‘research segments’’ awarded. This variable
programs refer to the long run situation, because measures the quantity and quality of the scientific
they admit the possibility of adjusting the fixed production of the departments. ‘‘Research seg-
factors to achieve greater cost efficiency. However, ments’’ are the result of the evaluation process
from (10) we can deduce results that would infer made by a government committee that evaluates
changes in the long run, albeit of a more structural the research carried out by Spanish university
nature, because they may imply substantial changes staff. This evaluation is carried out by subcom-
in the size of the analyzed departments in order to mittees grouped according to academic affinities,
obtain the lowest average cost. in order to ensure that the evaluation is
conducted by specialists in each scientific field.
Six-year ‘‘research segments’’ are evaluated on
3. Sample and variables the basis of the quantity and scientific quality of
their publications and salary complements are
The empirical application of the model described granted to lecturers in consequence. Obviously, a
above was carried out on departments of the professor can be awarded more than one
Autonomous UAB. The UAB consists of a total ‘‘research segment’’ in the course of his/her
of 46 departments, although complete information academic career. Hence, departments with more
was available for only 42 of these. senior lecturers would probably have a larger
The data obtained corresponds to the 1996–98 number of ‘‘research segments’’ than depart-
period. A period exceeding 1 year was chosen ments with younger staff. In order to avoid the
essentially because scientific research requires a influence of this fact in the results, the number of
development process lasting for more than one new ‘‘research segments’’ awarded over the three
year. Specifically, a three-year period was used year period of the study as a percentage of those
because that is the duration of most of the projects eligible instead of total ‘‘research segments’’
of more than 1 year supported by the different awarded was considered. There is also wide-
institutions and organisms that fund a large part of spread recognition in Spain that the number of
public research in Spain. This longer period also ‘‘research segments’’ conceded differs signifi-
avoids the distortion of the results through potential cantly among disciplines. For this reason, the
anomalies in the data for a single year. number of ‘‘research segments’’ has been divided
After reviewing the literature on nonparametric by the percentage of segments conceded over the
frontier models applied to higher education, it was number of applications for segments in each
concluded that no agreement has been reached on scientific field.2
which inputs and outputs should be used to evaluate Teaching load. To measure the teaching load, the
efficiency in university departments. However, there departments used the total number of credits
is a common denominator in the majority of the granted. For a department, ‘‘k’’ that conducted
Pj
studies reviewed: the departments’ main activities ‘‘j’’ subjects, the calculation will be i¼1 si ci ,
are teaching and research. A dual approach to these where si is the number of registered students for
two items has been selected: in the first place, a subject ‘‘i’’ and ci are the credits3 for the subject
quantitative approach (production); and in the ‘‘i’’. In this way the number of students that
second place, a qualitative approach, based on the receive instruction as well as the duration of the
conviction that both views should be considered in course are considered.
an attempt to obtain a reliable image of reality. 2
These percentages were obtained from the 2002 report of the
In order to choose the outputs, a list of the most National Research Activity Evaluation Committee. The full
frequently used outputs encountered in the litera- report can be seen at http://www.univ.mecd.es.
3
ture examined was presented to a group of One credit is equal to 10 class hours.
6. ARTICLE IN PRESS
548 ´nez, J.L. Martınez / Economics of Education Review 25 (2006) 543–553
V.M. Gime ´
Quality teaching. There are different ways of potentially adjustable over the short or long term.
determining teaching quality (see Astin, 1985; The inputs selected were the following:
George, 1982), but taking into account the
scarcity of information and the difficulty of Expenditure on temporary hired teaching and
adopting other approaches, in this work teaching research staff. This includes all spending relating
quality was measured by the opinions of its to the contracting of such staff in the department.
users—i.e. the students. The reason for using this This input is considered variable (adjustable in
indicator was the difficulty encountered in the short to medium term).
obtaining homogeneous information from the Operational expenditure. This item measures the
departments. The average scores obtained by normal operational costs of the department, such
each department in the satisfaction questionnaire as photocopies, office supplies, telephone, fax,
that students complete at the end of each computer maintenance, expenses for visiting
semester were used to determine the opinion of researchers, expenses for reading of doctoral
the students about the quality of their learning. theses, etc. Expenses particular to departments
that conduct experiments such as instruments,
Another alternative would have been to use a maintenance of laboratories, specialized equip-
compound measure of quality and production, like ment, etc. have been excluded. This input is also
in the case of research, for teaching activity. considered variable.
Designing an index involves assigning explicit Expenditure on permanent teaching and research
weights to each variable. Unfortunately, as there staff. This item includes spending related to
are no official sources for such weights either in the permanent staff, who are civil servants with a
university or in Spain in contrast to the case of working life-long contract. As this spending
research activity, it was considered more reliable to cannot be adjusted in the short or medium term,
use a DEA model to assign weights to each this input is considered fixed (adjustable solely in
department, based on Pareto’s efficiency criteria. the long term).
The origin of the question lies in whether any trade-
off quantity/quality is acceptable a priori for the Departments can make decisions about the
teaching activity, or whether the weights assigned to permanent or temporary hired teaching and re-
each one of those outputs should be restricted search staff mixed with different strategic implica-
(Pedraja-Chaparro, Salinas-Jimenez Smith, tions on cost and quality. Likewise, a part of the
1997). Our opinion is that in our case it is not operational expenditure depends on this mix,
necessary to make this restriction. The departments because it directly affects printing expenses, paper,
do not have the capacity to decide how to prioritize doctoral thesis reading, congress assistance, etc.
one of these variables over the other. In this sense, it Therefore, in this case, deviation on fixed factors
must be taken into account that the departmental utilization will also reflect a number of aspects very
teaching load is imposed by the number of students closely related to allocative efficiency and the
to be served, hence, individual departments have optimum mix between permanent and temporary
very little room for manoeuvring (Table 1). professors.
With regard to the inputs, the main budget items The inputs correspond to the division of costs
on which the departments allocates funds have been instead of physical inputs, as it appears in the
selected, distinguishing between those that are formulation of the model. As a consequence, we are
Table 1
Descriptive statistics of inputs and outputs, 1996-98
Input/output Max. Min. Mean Std. dev.
Costs permanent staff (h) 5,943,164.50 774,611.28 2,774,205.22 1,383,666.78
Costs temporary hired staff (h) 2,561,641.50 216,316.83 929,524.69 585,391.73
Operational costs (h) 592,781.29 81,333.60 285,386.30 180,413.87
New research segments awarded (% of those eligible) 87.03 31.87 58.68 24.18
Teaching load 401,977.50 13,239.00 139,822.96 87,968.05
Teaching quality 7.00 5.43 6.17 0.41
7. ARTICLE IN PRESS
´nez, J.L. Martınez / Economics of Education Review 25 (2006) 543–553
V.M. Gime ´ 549
assuming that the purchase prices of the different Table 2 depicts the average values of the cost
inputs are the same for all departments. In our case, deviations of the model. Overall, one can see that
this assumption does not simplify reality, but the average total cost deviation (TCV) is 13.46%; in
adjusts to it. All departments of the university must other words, there is a potential cost saving of this
make their purchases through suppliers that have amount in the long term, after making the necessary
previously been officially approved by the purchas- scale adjustments. However, from the management
ing department of the university. For this reason, perspective, it is useful to differentiate how much of
the price of purchases should be similar for all the this potential saving would be achievable in the
departments. In the case of the costs of the teaching short or long term. The technical deviation indicates
staff, the assumption is again plausible. Wages for the cost excess that can be corrected in the short
professors at Spanish public universities are deter- term, while the sum of the deviation due to the
mined by the government (as happens in other utilization of fixed capacity, together with the scale
countries). deviation, indicates excess costs that can be reduced
It is well known that DEA methodology is not in the long term.
appropriate for analyzing heterogeneous units. The Specifically, it can be seen that the long-term
evaluation of departments belonging to the same factors are the causes of a cost excess of 10.90% on
university cannot be exempt of this problem, as they average. By way of contrast, the factors of short-
are subject to departmental-specific constraints that term inefficiency—or, in other words, factors
make it difficult to compare them. It is noteworthy attributable to more operational aspects—are only
that departments with a high level of experimental responsible for 2.56% of the total cost excess of the
research require larger and more specialized equip- departments, which suggests that improvements
ment to carry out their normal research and should focus on structural rather than temporary
teaching tasks, implying the need for greater factors. In the case under analysis, the structural
financial resources for investment and maintenance factors are closely linked to the composition of the
than other departments with less experimentation. research and teaching staff and the size of the
In fact, this aspect (with different objectives) has departments. Among the long-term factors leading
been highlighted in other studies focusing on to cost inefficiency, the most significant is the
Spanish universities, such as that of Caballero et utilization of fixed capacity (FCV) with 7.43%,
al.(2004). It would be unfair to ignore this rather than scale deviation (SV).
circumstance in the evaluation of the departments, A detailed analysis at departmental level showed
which conduct the most experiments using special that departments with a higher proportion of
equipment. This has been the reason why these temporary hired staff generally achieve better
kinds of costs have been excluded from operational results. Table 3 shows that the staff from the most
costs. A second aspect that contributes to the varied efficient departments is composed, on average, of
results is the scientific activities of the departments. 57.33% temporary staff, while in the more ineffi-
It is well known that scientific production of cient departments, this percentage is 48.56%. The
different areas of knowledge is not directly compar- application of the Mann–Whitney non-parametric
able. Likewise, departments with more senior test confirmed that the difference between the two
professors have greater opportunities for under- figures is statistically significant (p-value ¼ 0.023).
taking high quality research tasks. We believe this In Spanish universities there are two kinds of
problem is resolved on its own and by the profiles for temporary professor. On one hand,
construction of the output ‘‘new research segments there are full-time contractual positions for those
awarded’’.
Table 2
Descriptive statistics of cost deviations (% of observed total cost)
4. Results
TEV FCV SV TCV
The results presented in this section are con- Mean 2.56 7.43 3.47 13.46
strained by the conditions laid down by the UAB Maximum 24.06 38.87 27.33 42.75
vice-chancellor’s office: the presentation of the Minimum .00 .00 .00 .00
Std. dev. 5.43 9.13 6.46 21.81
results in aggregate form was an essential prerequi-
Number of efficient departments 26 19 15 15
site for supplying the information requested.
8. ARTICLE IN PRESS
550 ´nez, J.L. Martınez / Economics of Education Review 25 (2006) 543–553
V.M. Gime ´
Table 3
Academic staff structure in the long-term for efficient and inefficient departments (% of total staff)
Mean Max. Min. Std. dev.
Efficient departments (n ¼ 19) Permanent staff (PS) 41.67 67.39 27.08 10.35
Temporary staff (TS) 57.33 72.92 32.61 10.35
% TS full time 68.97 82.56 43.28 28.46
% TS part time 31.03 49.58 21.41 28.46
Inefficient departments (n ¼ 23) Permanent staff (PS) 57.44 77.86 29.49 12.42
Temporary staff (TS) 48.56 67.07 30.87 12.42
% TS full time 49.67 65.72 39.38 27.78
% TS part time 50.33 75.33 45.94 27.78
professors that want to develop their careers entirely affirmative response can be deduced. However, we
within the university. On the other hand, there are believe that we should be cautious in this assertation
part-time contractual positions for those professors because other factors which we define here below
whose professional activity is developed outside the can exist, which can influence the obtained conclu-
university but collaborate with the university in sion and should be object of a future study, which is
teaching tasks, contributing their professional ex- outside the scope of this article. The first of these
perience. factors is the fact that temporary professors work
Within the part-time contractual positions, the hard in order to consolidate their position. Perhaps
departments have the autonomy to decide what their performance would not be the same if this
personnel structure they prefer. It has even been would not be the case. The second factor is the
observed that they are able to employ monetary possibility that these results mean that a system of
resources initially destined for permanent place- adequate incentives for permanent professors does
ments in order to cover new temporary places. not exist in the Spanish universities. Higher educa-
Table 3 breaks down the results for each of these tion salaries in Spain are under national or regional
sub-categories, expressing the percentage over the governmental control, giving rise to relative simi-
total temporary staff. larity among identically ranked professors. The
The analysis of Table 3 reveals that in the efficient monetary incentives for teaching or research, in
departments, the temporary hired staff is mainly relation to the effort required to achieve them, are
made up of professors hoping to consolidate their extremely limited, especially in the case of the latter.
professional career in the university. Specifically, So, it would seem that the results point towards the
you can observe that 68.97% of the temporary staff need to create a more efficient incentive and
is full time, contrary to the inefficient departments, motivation system than the current one, thus
where this percentage is approximately 50%. Again, guaranteeing that professors will maintain a high
the Mann–Whitney test confirms that the difference performance level once they have obtained a
between both calculations is statistically significant permanent position.
with a p-value of 0.043. There may be various In recent years some members of successive
reasons for this, and to determine them would governing boards at UAB have expressed their
require a more detailed study that is beyond the concern for the need to re-think the current
scope of current research. However, the first departmental structure of the university, with the
explanation of this result that comes to mind might aim of balancing the size of departments. This
be that in these departments there is strong research casts some light on this question, for the
competition among the temporary staff to earn results show that an inappropriate size or scale in
stable posts. Such competition may lead to an departments leads to a mere 3.47% excess in costs.
increase in the production of scientific publications Consequently, a restructuring of these characteris-
at a reduced cost, since these researchers receive tics would not probably appear to be a priority from
lower salaries than the permanent staff. the cost savings perspective.
These results posed the following question: In fact, the intensity vectors obtained from the
Should the university contract more temporary model (10) indicate that the particularly inefficient
than permanent professors? From the results, an university departments are mainly those that have
9. ARTICLE IN PRESS
´nez, J.L. Martınez / Economics of Education Review 25 (2006) 543–553
V.M. Gime ´ 551
PK Ã
expanded in size, which suggests that average costs where zà represents the optimum value
s¼1 zs ys;i , s
are lower in smaller departments. In the literature, of the variables of intensity in (9).
contradictory results can be observed, but most The results depicted in Table 4 show that once
sources conclude that there are increasing returns to costs are minimized in the long term, there is on
scale. However, the majority of the studies analyze average a margin for increasing current teaching
economies of scale at university rather than depart- levels by .39%. This reduced percentage was
mental level. There is little research on returns of predictable a priori, since the size of the teaching
scale for departments. Some examples of the staff in each department has traditionally been
research that has found that universities can reap determined in Spanish universities in accordance
benefits from scale economies are Koshal and with the amount of teaching needed (Caballero
Koshal (1999), Cohn, Rhine and Santos (1989), et al., 2004).
Dundar and Lewis (1995) and Hashimoto and Cohn Similarly, potential significant increases in teach-
(1997). Nevertheless, other studies that also take ing quality have not been detected once long term
universities as the analysis unit reached the opposite costs have been minimized (only 3.32%). However,
conclusion (Avkiran, 2001). the output associated with scientific activity is the
As stated in the introduction, the Spanish one that presents the greatest need for improve-
university system is currently immersed in a process ment. The average increase that should be reached
of cost rationalization, accompanied by a reduction in the new research segments awarded is 48.67%,
in the number of students in many disciplines, as a which should notably improve the quantity and
result of the low national birth rate sustained quality of scientific production. Hence, the results
through time, a characteristic that can only be indicate that teaching activity is found at compar-
inverted as a secondary consequence of immigra- able levels of efficiency among the different depart-
tion. Given these circumstances, universities are ments of the UAB while the opposite occurs in
turning their priorities towards quality improve- research. One explanation for this result, could be
ment and increased production of scientific re- that Spanish universities have made a huge effort
search, without the injection of additional financial during the last decades to satisfy the growing
resources in the system. As a result, the model demand for university education. In many cases,
proposed in this paper may prove to be a useful tool this has led to conditional hiring policies, which give
in determining where the quantity and quality of precedence to professors who can cover teaching
research production should be increased, without needs without taking research activity into account.
modifying or reducing costs.
In this context, and assuming cost minimization,
it may be helpful to examine the potential increases 5. Conclusions
achievable in the long term in the different outputs,
especially scientific activity. Teaching quality has This paper has presented a general model for the
also become an objective of improvement for evaluation of the cost efficiency of any type of
university management. In Table 4 we show organization, based on data envelopment analysis
descriptive statistics that correspond to the increases methodology. The main contribution of the model
that should be reached in the different outputs over is that it quantifies the difference between the
the long term. The target increase for each output in observed cost of the units under analysis and the
each department is calculated in the expression cost that they would achieve assuming long-term
cost minimization and optimal scale. The model
therefore outlines a target that is achievable in the
Table 4 long term. However, in order to make this objective
Potential increases in outputs associated with long-term cost easier to attain in successive stages, the model
minimization (% of current level) attributes the difference between each of the two
costs to different factors, each of which can be
New research Teaching Quality
segments awarded load teaching corrected in the short or long term. To be more
precise, technical inefficiency is the explanatory
Mean 48.67 .39 3.32 factor of the total cost deviation attributable to
Max. 212.43 12.29 21.87
the short term, while the factors attributable to the
Std. dev. 39.58 2.31 4.61
long term are the inappropriate utilization of the
10. ARTICLE IN PRESS
552 ´nez, J.L. Martınez / Economics of Education Review 25 (2006) 543–553
V.M. Gime ´
capacity available and an inappropriate scale from generally implies segregating the sample into sub-
the cost minimization perspective. samples, reducing the discriminating power of the
The empirical application of the model on 42 DEA. The solution applied in this case has made
departments at UAB showed that the management possible to achieve some meaningful insights.
of their economic resources is generally correct in Another limitation that we found was in measuring
the short term (2.56% optimal reduction possible), the teaching quality in the different departments.
and that improving it would mainly require adjust- The use of evaluations by the students is not ideal
ments only achievable in the long term. These and could be improved, but unfortunately the
assertions are based on the fact that long-term authors did not have access to a better homogenous
factors largely explain the total cost deviation of measure of teaching quality.
13.46%. Finally, we can mention that the future line of
Analysis of the results referring to the long term investigation will consist of extending the dynamic
revealed two interesting questions that should, in setting of the model of deviations in costs that we
the authors’ opinion, be considered by the uni- have presented in this article.
versity authorities:
Firstly, departments with a large proportion of
temporary professors striving to make a university Acknowledgments
career were shown to be more efficient. At first
sight, the solution would be to increase the This research received the financial support from
proportion of temporary staff in order to improve the Research Projects SEC2003-047707 and
efficiency. However, we should be cautious in the SEC1999-0843-C02-01 from the Spanish Govern-
interpretation of this conclusion as it can be due to ment’s Science and Technology Department
other factors, for example the inexistence of a
sufficient system of incentives for the professors,
which already have permanent positions, as we have References
noted in the previous section.
Secondly, the results show that an inappropriate Adler, N., Berechman, J. (2001). Measuring airport quality
size or scale in departments leads to a mere 3.47% from the airlines’ viewpoint: An application of data envelop-
excess in costs. Consequently, the resizing of ment analysis. Transport Policy, 8, 171–181.
Ahn, T. S. (1987). Efficiency and related issues in higher education:
departments would not appear to be a priority A data envelopment analysis approach. Doctoral Thesis. The
from the cost savings perspective. However, if it University of Texas at Austin.
were implemented, administrators should bear in Astin, A. W. (1985). Achieving educational excellence. San
mind the fact that the particularly inefficient Francisco: Jossey-Bass.
Athanassopoulos, A., Shale, E. (1997). Assessing the compara-
university departments are mainly those that have
tive efficiency of higher education institutions in the UK by
expanded, while average costs are lower in smaller means of data envelopment analysis. Education Economics,
departments. From the perspective of improving 5(2), 117–134.
outputs over the long term, the results suggest that Avkiran, N. K. (2001). Investigating technical and scale
while teaching activity develops in an efficient way efficiencies of Australian Universities through data envelop-
in general, a clear opportunity and need exist to ment analysis. Socio-economics Planning Sciences, 35, 57–80.
Beasley, J. E. (1990). Comparing university departments. Omega,
increase the quantity and quality of scientific 8(2), 171–183.
production. Beasley, J. E. (1995). Determining teaching and research
One difficulty in any research that evaluates efficiencies. Journal of Operational Research Society, 46,
different university departments is the handling of 441–452.
the heterogeneous nature of the sample. In our case, Bifulco, R., Bretschneider, S. (2001). Estimating school
efficiency: A comparison of methods using simulated data.
we have not used control variables, as we could Economics of Education Review, 20(5), 417–429.
have done, for example, for the level of experiments ´
Caballero, R., Galache, T., Gomez, T., Molina, J., Torrico, A.
conducted. In its place, we have chosen to use (2001). Efficient assignment of financial resources within a
variables that, apart from obviously representing university system. Study of the university of Malaga.
European Journal of Operational Research, 133, 298–309.
the productive process, were directly comparable
´
Caballero, R., Galache, T., Gomez, T., Molina, J., Torrico, A.
among departments. The advantage of using this (2004). Budgetary allocations and efficiency in the human
method is directly associated with the reduced size resources policy of a university following multiple criteria.
of the initial sample. Using control variables Economics of Education Review, 23, 67–74.
11. ARTICLE IN PRESS
´nez, J.L. Martınez / Economics of Education Review 25 (2006) 543–553
V.M. Gime ´ 553
Charnes, A., Cooper, W., Rhodes, E. (1978). Measuring the of data envelopment analysis. Oxford Economic Papers, 45,
efficiency of decision making units. European Journal of 332–347.
Operational Research, 2, 429–444. Korhonen, P. (2000). Some thoughts on incorporating preference
Cohn, E., Rhine, S. L. W., Santos, M. C. (1989). Institutions of information in data envelopment analysis based on produc-
higher education as multi-product firms: Economies of scale tion possibility. In Proceedings of the fourth international
and scope. Review of Economics and Statistics, 71, 284–290. conference on multi-objective programming and goal program-
Cooper, W., Seiford, L., Tone, K. (2000). Data envelopment ming: Theory and applications. MOPGP’ 00, Ustron, Polonia.
analysis: A comprehensive text with models, applications, Korhonen, P., Tainio, R., Wallenius, J. (2001). Value efficiency
references and DEA-solver software. Dordrecht: Kluwer analysis of academic research. European Journal of Opera-
Academic Publishers. tional Research, 130, 121–132.
Dismuke, C., Sena, V. (2001). Is there a trade-off between Koshal, R. K., Koshal, M. (1999). Economies of scale and
quality and productivity? The case of diagnostic technologies scope in higher education: A case of comprehensive uni-
in Portugal. Annals of Operations Research, 107, 107–116. versities. Economics of Education Review, 18, 269–277.
Drake, L., Simper, R. (2003). The measurement of English and Leibenstein, H. (1966). Allocative efficiency vs. ‘X-efficiency’.
Welsh police force efficiency: A comparison of distance American Economic Review, 56, 392–415.
function models. European Journal of Operational Research, Li, X. B., Reeves, G. R. (1999). A multiple criteria approach to
147(1), 165–186. data envelopment analysis. European Journal of Operational
Dundar, H., Lewis, D. R. (1995). Departmental productivity in Research, 115, 507–517.
American universities: Economies of scale and scope. Madden, G., Savage, S., Kemp, S. (1997). Measuring public
Economics of Education Review, 14, 199–244. sector efficiency: A Study of economics departments at
Emrouznejad, A., Thanassoulis, E. (2005). A mathematical Australian universities. Education Economics, 5(2), 153–168.
model for dynamic efficiency using data envelopment Olesen, O. B., Petersen, N. C. (1995). Incorporating quality
analysis. Applied Mathematics and Computation, 160(2), into data envelopment analysis: A stochastic dominance
363–378. approach. International Journal of Production Economics, 39,
George, M. D. (1982). Assessing program quality. In R. Wilson 117–135.
(Ed.), Designing academic program review. New directions for Pedraja-Chaparro, F., Salinas-Jimenez, J. (1996). An assess-
higher education. San Francisco: Jossey-Bass. ment of the efficiency of Spanish courts using DEA. Applied
Glass, J. C., Mckillop, D. G., O’Rourke, G. (1998). A cost Economics, 28, 1391–1403.
indirect evaluation of productivity change in UK universities. Pedraja-Chaparro, F., Salinas-Jimenez, J., Smith, P. (1997). On
Journal of Productivity Analysis, 10, 153–175. the role of weight restrictions in data envelopment analysis.
Grosskopf, S., Margaritis, D., Valdmanis, V. (1995). Estimat- Journal of Productivity Analysis, 8(2), 215–230.
ing output substitutability of hospital services: A distance Post, T., Spronk, J. (1999). Performance benchmarking using
function approach. European Journal of Operational Research, interactive data envelopment analysis. European Journal of
80, 575–587. Operational Research, 115, 472–487.
Hashimoto, K., Cohn, E. (1997). Economies of scale and scope Ray, S. (1991). Resource-use efficiency in public schools: A study
in Japanese private universities. Education Economics, 5, of Connecticut data. Management Science, 37(12), 1620–1629.
107–116. Sarrico, C. S., Hogan, S. M., Dyson, R. G., Athanassopoulos,
Johnes, G., Johnes, J. (1993). Measuring the research A. (1997). Data envelopment analysis and university selection.
performance of UK economics departments: An application Journal of Operational Research Society, 48, 1163–1177.