Conceptual Model of Student Satisfaction in Higher Education

Abstract

To try and understand the factors which influence student satisfaction in higher education, as well as consequences of it, this study tests an explanatory model of student satisfaction in higher education. The model was tested through the utilization of structural equations and showed that the variable which has the most influence in student satisfaction in higher education is the variable – image followed by value and afterwards quality perceived. The conclusions of this study also mention the existence of a negative influence from the variable – expectations. It was also noticeable that the main consequence of satisfaction was student loyalty caused by word of mouth from student to student.

Full text

Conceptual Model of Student
Satisfaction in Higher Education
HELENA ALVES & MA
´RIO RAPOSO
Universidade de Beira Interior, Covilha
˜, Portugal
ABSTRACT To try and understand the factors which influence student satisfaction in higher
education, as well as consequences of it, this study tests an explanatory model of student
satisfaction in higher education. The model was tested through the utilization of structural
equations and showed that the variable which has the most influence in student satisfaction in
higher education is the variable – image followed by value and afterwards quality perceived. The
conclusions of this study also mention the existence of a negative influence from the variable –
expectations. It was also noticeable that the main consequence of satisfaction was student loyalty
caused by word of mouth from student to student.
KEY WORDS: Satisfaction, higher education, antecedents of satisfaction, consequences of
satisfaction
Introduction
In the last two decades, the sector of Higher Education in Portugal, similar to what happened
in the USA and in the remaining countries of Europe, has suffered quite profound changes.
If until 1996, an increase occurred in the number of institutions operating in this sector
and consequently an increase in the number of students, from 1996 onwards the situation
completely inverted with a significant reduction in the number of candidates taking place,
partly caused by the priority given by the governments to the question of quality, imposing
the definition of minimum grades and due in part to the reduction in the birth rate (44%)
in the last 25 years (Santos, 1995).
According to a study from CIPES (1999), the tendency is that there will be an aggrava-
tion to this situation as it is forecasted that until 2006, having as a basis 1995, the reduction
in the number of candidates will be 26.6%.
Adding to these alterations, a reduction in public financing and an increasing tendency
towards the perspective of “value for money”, which demands larger responsibility upon
institutions of this sector, in terms of quality and efficiency (CQAEHE & CNE, 1998).
Total Quality Management
Vol. 17, No. 9, 1261– 1278, November 2006
Correspondence Address: Helena Maria Baptista Alves, Departamento de Gesta
˜o e Economia, Universidade da
Beira Interior, Estrada do Sineiro, 6200 Covilha
˜, Portugal. E-mail: halves@ubi.pt
1478-3363 Print=1478-3371 Online=06=091261– 18 #2006 Taylor & Francis
DOI: 10.1080=14783360601074315

This way, the sector of Higher Education in Portugal faces more competitive market
structures, which threaten the survival of some of the existing institutions, as these are
now obliged to compete with scarce resources, for a smaller number of potential candi-
dates, evermore disputed by the various institutions of Higher Education.
As Lin (1997) refers, despite organisations being subsidised by the state having the
tendency to ignore the needs of its targets publics, which is the case of Universities,
facing this new and more competitive context, these institutions need to incorporate a
better orientation to the market, seeking to obtain competitive advantages over its compe-
titors, as well as the construction of a positive image close to its target market.
In this way, it becomes fundamental to analyse and study student’s satisfaction in higher
education, as institutions of higher education could greatly benefit from being able to
establish lasting relationships with their pupils. A long term relationship with students
can provide an institution with a type of competitive advantage, particularly at a positive
word of mouth level concerning potential, present and future students, as well as through
the possible collaboration with the institution, especially after graduation, contributing to
the (work) placement of recent graduates.
The dissatisfaction of students, on the contrary, could have ominous consequences for
both the university and the student, namely unsuccessful students (Wiese, 1994; Walther,
2000), quitting or transferring (Chadwick & Ward, 1987; Dolinsky, 1994; Wiese, 1994;
Thomas et al., 1996; Astin, 2001) and negative word of mouth being harmful to future
applications (Chadwick & Ward, 1987; Dolinsky, 1994; Ugolini, 1999; Walther, 2000).
By reviewing literature on the matter, we can see that the formation process for
satisfaction is not very consensual whether it is in services in general or especially in
higher education. The conclusions from various studies about customer satisfaction in
services found different antecedents in the formation of satisfaction. In higher education
this reality is very similar, with the aggravation that within this sector studies concerning
satisfaction are in truth still very scarce. Thus this study intends to test a conceptual model
of the antecedents and consequences of student satisfaction in higher education.
Antecedents of Satisfaction in Higher Education
Expectations and the Quality Perceived
For some researchers (Halstead et al., 1994; Hartman & Schmidt, 1995; Rautopuro &
Vaisanen, 2000) students have weak expectations, especially in regards to intellectual
environment as this has little influence on satisfaction, therefore the variable – perform-
ance is the most influential factor in satisfaction. Contrarily, for other researchers
(Shank et al., 1995; Eskildsen et al., 1999; Patterson, 2000) the exact opposite occurs.
Whilst Anderson & Sullivan (1993) and Yi (1993) suggest that the more difficult it is to
evaluate quality received, which happens in educational services, the higher the influence
will be of expectations in the formation of satisfaction. However, Anderson & Sullivan
(1993) and Kristensen et al. (1999) suggest that the influence of expectations is completely
mediated by quality perceived.
Still on the other hand, Walker (1995) suggests that the impossibility of evaluating the
main service causes students among other aspects to focus more on class environment and
on the teachers’ presentation. In other words on the way the service is provided (functional
quality).
1262 H. Alves & M. Raposo

Thus it becomes unclear whether or not this variable has an influence in the formation
process of satisfaction in higher education, as well as the nature of this influence,
wherefore we can establish the following hypotheses:
If the education service is a service of high ambiguity of which quality is difficult to
evaluate, then:
Hyp. 1.a) Expectations have a significant influence in the formation of satisfaction.
Hyp. 1.b) The influence of expectations is direct and simultaneously indirect through
quality perceived.
Hyp. 1.c) The influence of quality perceived in satisfaction is greater, specifically
from functional quality.
Image
The research of Clow et al. (1997), applied to various service industries, showed that
image was only influential to satisfaction in some industries. However, its influence in
expectations appeared to be quite strong. Whereas in studies carried out using the
European Customer Satisfaction Index as a basis (Kristensen et al., 1999; Cassel &
Eklo
¨f, 2001), image always appears as one of the variables with the greatest influence
in the formation of satisfaction, seeing that its direct influence through expectations is
superior to its indirect influence. According to Eskildsen et al. (1999), this variable is
really the one that has the most influence on student loyalty in higher education.
In this way, the following hypotheses have been established with the objective of trying
to contribute to knowledge development of this subject in higher education:
If higher education is an extremely complex service with characteristics of trust and
experiences which make it difficult to be evaluated before and even after being
tested, increasing the importance of image as an information source, then:
Hyp. 2.a) University’s image has a significant influence in the formation of the stu-
dent’s expectations in higher education.
Hyp. 2.b) University’s image has a direct and significant influence in the formation
process of satisfaction.
Hyp. 2.c) University’s image has a direct and significant influence in the loyalty of
the student.
Value
The variable value was studied by Hartman & Schmidt (1995) and Webb & Jagun (1997),
with these researchers proving that student satisfaction was influenced by value perceived.
On the other hand, LeBlanc & Nguyen (1997) discovered that the value perceived by the
Conceptual Model of Student Satisfaction in Higher Education 1263

students can involve dimensions related to quality received, the university’s image,
emotional values and even social values.
Nevertheless, Caruana & Money (1997) and Ismail & Abdullah (2001), noticed that
value was influenced by the global perception of service quality and McDougall and
Levesque (2000) highlighted the fact that the influence of value can vary from service
to service.
In higher education, its influence is still limited in research, in view of which the follow-
ing hypotheses have been established:
If satisfaction in higher education is the global result obtained with educational
experience in all its ways, then:
Hyp. 3.a) Student satisfaction in higher education is influenced by its perception of
value.
Hyp. 3.b) The value perceived is influenced by quality perceived and student’s
expectations.
The Consequences of Satisfaction
The consequences of satisfaction in higher education do not linearly correspond to the
consequences of satisfaction in other services, since the state education service is a non
profitable one with many specific characteristics. Whilst in the greater part of services,
complaint behaviour, word of mouth actions, loyalty, repetitive purchasing behaviour
and profit are shown as consequences of satisfaction, some of these consequences do
not make sense in higher education.
In terms of higher education, the main consequences found by some researchers were:
loyalty (Webb & Jagun, 1997; Eskildsen et al., 1999), word of mouth actions (Athiyaman,
1997) and complaints (Webb & Jagun, 1997).
Thus it is also in the scope of this study to analyse those consequences that are the most
important, resulting from satisfying the student in higher education. In this way, the
following hypotheses have been established:
Hyp. 4) If the student is satisfied with his/her educational experience, then he/she
will demonstrate loyalty towards the institution.
Hyp. 5) If the student is satisfied with his/her educational experience, then he/she
will engage in positive word of mouth.
Methodology
The Model
The model to be tested (Figure 1) results from the hypotheses previously established and
illustrate the main antecedents of satisfaction, expectations, the university’s image per-
ceived by the student, quality perceived both technical and functional of the education
1264 H. Alves & M. Raposo

service, as well as the variable value. These influences can be direct or indirect through
other variables, as one can observe in the model. The model illustrates loyalty and
word of mouth actions as the main consequences of satisfaction.
Sample’s Definition
Having defined the student as the most important customer of the education service, in
order to test the proposed model it was necessary to select a sample of students in
higher education. Due to the technical impossibility of embodying all of higher education,
we chose to limit the study to state university education. Thus, the considered target popu-
lation was all students from Portuguese state universities, which belonged to CRUP
(Conselho de Reitores das Universidades Portuguesas – the Council of Rectors of
Portuguese Universities) with the exception of Universidade Aberta due to its teaching
characteristics.
Thus, this sample was randomly selected with the objective of forming fixed
sub-samples of 250 students, which embody students of various scientific areas ministered
in universities. In the end, the sample did not correspond exactly to 250 students
per university, resulting in a total of 2687 students (Table 1).
Method of Data Obtainment
Given the intended objectives expected to be reached with this research, a survey using
questionnaires was the chosen way for gathering data, thus, a questionnaire subdivided
in 7 parts was drawn up: Sample Characterization, Expectations, Quality of service,
Value, Global satisfaction and lastly, Loyalty and Word of Mouth actions.
The scales used, result in part, from scales already tested in various studies, despite
many times the verbal context being adapted to the reality of higher education. In this
way, scales of multiple items were used in the entire questionnaire, as this allows a
reduction in standard error and the dimension of the sample required (Ryan et al.,
1995), as well as measurement with greater validity subjective constructs (Hayes, 1998;
Figure 1. Conceptual model to be tested
Conceptual Model of Student Satisfaction in Higher Education 1265

Anderson & Fornell, 2000). Intervals of 1 to 10 were used in the scales, since the enlarge-
ment of the number of points of the scale allows a reduction in skewness (Fornell, 1992).
With the view of trying to reduce possible, existing errors in the questionnaire a pre-test
of the questionnaire was realised with 25 students. The results of the pre-test revealed that
vocabulary used in the questionnaire, as well as its structure, were easily understood by the
students and that it took approximately 10 minutes to fill in.
Analysis of Data
After gathering the questionnaires, which took place between April and June 2002, in
order to test the established hypotheses, it became necessary to analyse and interpret
the data. Thus, the analysis of data was realised through structural equations using the
statistical software AMOS (Analysis of MOment Structures) version 4.0.
In this analysis, the matrix of variances-covariances was used as an initial matrix and the
estimation technique was Maximum Likelihood (ML), seeing that, apart from being
the most used, it is also one of the methods which obtains the most efficient estimations,
so long as the variables used respect the assumptions of normality (Hair et al., 1998;
Garcia & Martinez, 2000), a presupposition proved in this study.
The estimation of the model was carried out in two stages, as recommended by
Anderson & Gerbing (1988). Thus, in the first stage, the measurement model is estimated,
and in the second this same model is fixed, in order to estimate the structural model. The
logic behind this reasoning, according to the mentioned authors, is that the reliability of the
indicators is better represented in two stages, avoiding interaction between the measure-
ment and structural models.
Analysis of Results
After eliminating the least significant indicators, the measurement model resulted in
the model presented in Figure 2. For the estimated measurement model, absolute and
incremental adjustment measures are those presented in Table 2 and Table 3.
Table 1. Sample’s final composition
University No. of students %
Universidade dos Ac¸ores 79 2.9
Universidade do Algarve 270 10.0
Universidade de Aveiro 279 10.4
Universidade de Beira Interior 206 7.7
Universidade de Coimbra 199 7.4
Universidade de E
´vora 186 6.9
Universidade de Lisboa 230 8.6
Universidade da Madeira 80 3.0
Universidade do Minho 203 7.6
Universidade Nova de Lisboa 205 7.6
Universidade do Porto 272 10.1
Universidade Te
´cnica de Lisboa 256 9.5
Universidade de Tra
´s-os-Montes e Alto Douro 222 8.3
Total 2687 100
1266 H. Alves & M. Raposo

An analysis of the goodness-of-fit measures presented by the measurement model
reveals that nearly all the measures present a satisfactory level of acceptability and that
the model explains 94% data variance (value of GFI).
The statistical significance of each one of the estimated coefficients can be observed
in Table 4. From the analysis of the table, it is possible to verify that all the
indicators are statistically significant to a level of significance of 0.05, thus one can
say that all the indicators are significantly related to their specific constructs. On
the other hand, it is still noticeable that all the indicators present an estimation
above 0.7, correspondent to an internal reliability value of at least 50% (Hair
et al., 1998).
Table 5 presents the composed reliability of each construct; this is the level of internal
consistency of each one of the constructs, as well as the variance explained by each one of
the constructs.
Figure 2. Final measurement model
Conceptual Model of Student Satisfaction in Higher Education 1267

As can be observed, all the constructs exceed the level of minimum reliability of 0.7
recommended by Hair et al. (1998) and Garcia & Martinez (2000), showing indications
that the specified indicators are sufficient in its representation of inherent constructs.
One can still see that the construct with the highest internal reliability is the construct
of satisfaction (93%), measured by the level of global satisfaction, by the level of corre-
spondence to expectations and by the level of correspondence to student’s current wishes/
needs.
With a reliability level of 88%, appears the construct of value, measured through the
perception of being able to obtain a good job, the perception of a degree being a good invest-
ment and the perception of employers being interested in the students of that university.
With a reliability level of 87% and 86%, appear respectively the constructs of image and
word of mouth, being the construct of image measured through the variables: a good uni-
versity to study in, an innovating university and turned to the future and a university which
provides good preparation to its students; and the construct of word of mouth through the
variables: pride in the university and the recommendation to a friend.
The constructs of quality and expectations present very similar reliabilities, in other
words, 85% and 84% respectively. The construct of quality was measured through the
Table 3. Incremental fit measures for the measurement model
Incremental fit measures Value Levels of acceptance Acceptability
Tucker-Lewis index (TLI or NNFI) 0.958 .0.90 Good
Normed fit index (NFI) 0.966 .0.90 Good
Adjusted GFI (AGFI) 0.912 .0.90 Good
Table 2. Absolute fit measures for the measurement model
Absolute fit measuresValue Levels of acceptance Acceptability
Test of X Squared (
x
2
) 1396.166 Not applicable () Not applicable ()
Degrees of freedom (DF) 114 .than the level of
significance wished forLevel of significance (p) 0.000
Goodness-of-fit index (GFI) 0.941 0 - (Awful) Good
1 - (Very good)
Root mean square residual (RMR) 0.107 Variable () Reasonable
Root mean square error of
approximation (RMSEA)
0.065 ,0.05– 0.08 Satisfactory
() – According to Hair et al. (1998) and Bagozzi & Yi (1988), whenever the sample used is higher than 200, the
x
2
test is not indicative of good or bad adjustment of the model to the data, since the value of X-squared increases
with the sample’s increase. This way, whenever the sample is above 200 this test tends to reject all models, even
when these present a good adjustment. On the contrary, whenever the sample’s dimension is lower than 100, the
test will tend to accept all models, even those that do not present an adequate adjustment.
() – According to Hair et al. (1998) and Bagozzi & Yi (1988), the value of RMSR can only be evaluated
through the comparison of existing residuals in the initial data matrix and the existing ones in the estimated
matrix by the model. For these researchers, whenever the number of normalised residuals with a value higher
than 2.58 (in absolute value) is in the proportion of 1 to 20 (5%), these can be considered random and thus
the model can be accepted. In this model, the number of residuals with a value higher than 2.58 is of 5.8%
and can be considered a reasonable acceptability.
1268 H. Alves & M. Raposo

variables: global quality of education, skills and knowledge of the teachers and course
contents, only variables focused on the core service of the university. In turn, the construct
of expectations was only measured by expectations related to the global quality of education
and by expectations concerning the university’s capacity to supply good preparation for a
career.
Table 5. Reliability and extracted variance of constructs
Construct Indicator Reliability Explained variance
Expectations EXP1 0.843 0.728
EXP2
Value V1 0.883 0.716
V2
V4
Word of mouth P1 0.862 0.757
P2
Satisfaction S1 0.930 0.816
S2
S3
Image IM1 0.866 0.685
IM2
IM4
Loyalty L1 0.828 0.707
L2
Quality Q1 0.846 0.648
Q2
Q5
Table 4. Standardized regression weights of the measurement model
Regression weights Estimate tValues P
EXP1 Expectations 0.818 44.605 0.000
EXP2 Expectations 0.887 48.782 0.000
IM1 Image 0.857 54.535 0.000
IM2 Image 0.750 44.621 0.000
IM4 Image 0.869 55.695 0.000
Q1 quality 0.910 59.350 0.000
Q2 quality 0.740 43.594 0.000
Q5 quality 0.754 44.802 0.000
V1 value 0.867 55.045 0.000
V2 value 0.850 53.287 0.000
V4 value 0.821 50.522 0.000
S1 satisfaction 0.911 60.741 0.000
S2 satisfaction 0.903 59.918 0.000
S3 satisfaction 0.896 59.054 0.000
L1 loyalty 0.878 54.752 0.000
L2 loyalty 0.802 48.257 0.000
P1 word of mouth 0.892 53.485 0.000
P2 word of mouth 0.848 57.745 0.000
For a probability level of 0.05.
Conceptual Model of Student Satisfaction in Higher Education 1269

The construct which presents the lowest reliability is the construct of loyalty with 83%,
being measured by the variables: would choose again and would choose again for a post
graduation.
In relation to the variance explained by the constructs, it can be seen that the constructs
always explain more than 50%, minimum value recommended by Hair et al. (1998) and
Garcia & Martinez (2000).
Following the strategy of modelling in two stages and after confirming the acceptability
of the measurement model, there then proceeded an estimation of the structural model.
After estimating the model and following the indications of Hair et al. (1998) and
Garcia & Martinez (2000), there then proceeded a comparison of this model with
various alternative models.
Thus, the proposed model was compared with a model where all the alternatives which
we deemed theoretically possible were established, adding to which, the relations between
image and word of mouth, image and value, image and quality and also, the relationship
between value and word of month. The remaining models were models where the relation
that gave their name was annulled. The results from comparing the proposed model with
alternative models can be observed in Table 6.
According to Hair et al. (1998) and Garcia & Martinez (2000), the comparison between
alternative models, should mainly be done and based on parsimonious fit measures. Thus it
was seen that among the various alternative models, those that presented the best good-
ness-of-fit measures were models where the relationship between value and loyalty was
eliminated and the model with all the possible alternatives.
However, it was seen that the model where the relationship between value and loyalty
was restricted to zero, presented the best measures of parsimonious fit, in other words,
Table 6. Comparison table of models
Measures
Alternative models
Proposed
model
Model
with all
possible
alternatives
Model
without
image-
quality
Model
without
image-word
of mouth
Model
without
image-
value
Model
without
value-
word
of mouth
Model
without
value-
loyalty
Absolute Fit Measures
Number of
parameters
48 52 51 51 51 51 51
GFI 0.90 0.94 0.90 0.94 0.94 0.94 0.94
RMR 0.17 0.11 0.15 0.11 0.11 0.11 0.11
RMSEA 0.09 0.06 0.09 0.07 0.07 0.07 0.06
Incremental Fit Measures
TLI 0.92 0.96 0.92 0.96 0.96 0.96 0.96
NFI 0.93 0.97 0.93 0.96 0.96 0.96 0.96
AGFI 0.86 0.91 0.86 0.91 0.91 0.91 0.91
Parsimonious Fit Measures
PNFI 0.75 0.75 0.73 0.76 0.76 0.76 0.76
PGFI 0.64 0.65 0.63 0.66 0.66 0.66 0.66
AIC 2969.93 1546.27 2776.89 1624.28 1617.19 1574.88 1551.79
1270 H. Alves & M. Raposo

attaining the same levels of absolute fit with less parameters and even when the coefficient
value between value-loyalty tended to be non-significant. Thus, it was considered to be
more adequate to represent the data in analysis the model presented in Figure 3. This
figure presents the standardised regression weights, as well as the value of the squared
multiple correlations (values located above each indicator and each construct).
An analysis to the existence of offensive estimations, revealed the non existence of
standardized coefficients above 1, nor negative error variances, nor very high standard devi-
ations. In terms of the adjustment of the model, it can be seen from the presented indexes in
Table 7, that the model presents high levels of adequacy, as all the indexes are above the
minimum recommended, with exception to the value of the RMR index, which presents a
value of 0.11. The analysis of this value was done in light of the number of normalized
residuals with a value higher than 2.58, in absolute value. It was seen that the proportion
of this type of residuals (5.3%) was slightly higher to the proportion considered random
(5%), thus its acceptability can only be considered as reasonable.
GFI’s value gives us a percentage of the data variance which is explained by the model
(Garcia and Martinez, 2000). Thus, it can be said that the model explains a quite elevated
percentage of data variance: 94%, which, indicates that its acceptability can be considered
quite good.
Figure 3. Final structural model
Conceptual Model of Student Satisfaction in Higher Education 1271

The statistical significance of each one of the estimated coefficients can be observed
in Table 8. As one can see, all the estimated coefficients are significant to a level of sig-
nificance of 0.05.
In turn, Table 9 presents the various structural equations, as well as the determination
coefficient (R
2
) for each equation. From the analysis of the determination coefficients of
the various structural equations present in Table 9, it was seen that the construct of word
of mouth is the best explained construct by its antecedents (image, perceived value and
loyalty), as it presents an explained variance (R
2
) of 93%. The construct of quality perceived
also presents a quite elevated level of variance, explained by its antecedents (86%).
In turn, the construct of satisfaction presents an explained variance of 80%, which rep-
resents a variance non-explained by these antecedents of 20%, a value considered to be
due to error or to other non-identified factors.
It was also noticeable that the construct of loyalty is not sufficiently explained by the
construct of image and satisfaction, as these only explain 62% of the variance of this con-
struct. Lastly we noticed that the construct of expectations, just as was measured, is insuf-
ficiently explained by the construct of image (only 33% of variance explained). However,
image should be taken into consideration, as its influence was shown to be significant.
Comparing the final structural model presented in Figure 3 with the initially proposed
model, it was seen that there exist some theoretical relations that were supported, but that
there also exists evidence for the existence of empirical relations. Figure 4 illustrates the
theoretical relations that were and were not supported, as well as the empirical relations
encountered. The values presented correspond to the standard coefficients, as these
provide an easier interpretation.
As one can observe in Figure 4, nearly all the proposed relations based on the revision of
the literature were supported. The only exception was the relationship satisfaction-word of
mouth, maybe because in this research the constructs of loyalty and word of mouth are sep-
arated, whilst in other studies this separation had never been done. However, some relations
Table 7. Goodness-of-fit measures for the structural model
Absolute fit measures Value Levels of acceptance Model acceptability
Test of X Squared (
x
2
) 1449.789 Not applicable Not applicable
Degrees of freedom (DF) 120 .than the level of
significance wished forLevel of significance (p) 0.000
Goodness-of-fit index (GFI) 0.94 0-(Awful) Good
1-(Very good)
Relative Goodness-of-fit index
(RGFI)
0.94 .0.90 Good
Root mean square residual (RMR) 0.11 Variable Reasonable
Root mean square error of
approximation (RMSEA)
0.06 ,0.05–0.08 Satisfactory
Incremental Fit Measures
Tucker-Lewis index
(TLI ou NNFI)
0.96 .0.90 Good
Normed fit index (NFI) 0.96 .0.90 Good
Adjusted GFI (AGFI) 0.91 .0.90 Good
Relative Adjusted GFI
(RAGFI)
0.92 .0.80 Good
1272 H. Alves & M. Raposo

were also seen, whose support was strictly empirical, but as this kind of study had not yet
been implemented in the sector of higher education in Portugal, to our knowledge, those
relations may provide important contributions to the study of satisfaction in this sector.
Thus, it was proven in this study that the construct which influences the most student
satisfaction in higher education is the construct image, as this has a direct effect of 0.45
and still indirect effects through the expectations of the students, the quality perceived
by the students and the value perceived, with a value of 0.40. This way, the total influence
of perceived image over student satisfaction is of 0.86, in other words, in terms of total
Table 8. Non standardized and standardized regression weights of the structural model
Non-standardized regression weights
Standardized
regression weights
Estimate
(tvalues)
Standard
deviation PLoading
Expectations image 0.539 (27.642) 0.019 0.000 0.578
Quality expectations 0.117 (6.961) 0.017 0.000 0.119
Quality image 0.782 (43.901) 0.018 0.000 0.851
Value image 0.581 (9.603) 0.060 0.000 0.562
Value quality 0.250 (3.700) 0.067 0.000 0.222
Value expectations 0.099 (4.438) 0.022 0.000 0.089
Satisfaction value 0.373 (15.777) 0.024 0.000 0.405
Satisfaction quality 0.158 (3.147) 0.050 0.000 0.153
Satisfaction expectations 20.120 (27.050) 0.017 0.000 20.118
Satisfaction image 0.430 (8.837) 0.049 0.000 0.452
Loyalty satisfaction 0.892 (16.046) 0.056 0.000 0.578
Loyalty image 0.341 (6.464) 0.053 0.000 0.233
Word of mouth loyalty 0.615 (28.193) 0.022 0.000 0.655
Word of mouth image 0.323 (8.733) 0.037 0.000 0.235
Word of mouth value 0.210 (6.636) 0.032 0.000 0.158
EXP1 Expectations 0.900 (36.004) 0.025 0.000 0.818
EXP2 Expectations 1.000 () 0.887
IM1 Image 0.976 (59.039) 0.017 0.000 0.859
IM2 Image 0.961 (47.291) 0.020 0.000 0.754
IM4 Image 1.000 () 0.870
Q1 quality 1.000 () 0.910
Q2 quality 0.820 (47.511) 0.017 0.000 0.741
Q5 quality 0.866 (48.739) 0.018 0.000 0.753
V1 value 1.000 () 0.869
V2 value 1.019 (55.974) 0.018 0.000 0.850
V4 value 1.000 (52.859) 0.019 0.000 0.820
S1 satisfaction 1.000 () 0.911
S2 satisfaction 1.023 (74.247) 0.014 0.000 0.903
S3 satisfaction 1.040 (72.660) 0.014 0.000 0.895
L1 loyalty 1.000 () 0.880
L2 loyalty 0.934 (49.792) 0.019 0.000 0.804
P1 word of mouth 0.838 (59.997) 0.014 0.000 0.849
P2 word of mouth 1.000 () 0.891
Probability level of 0.05, Value not calculated as it was fixed at 1 the coefficient to fix the latent variable
scale.
Conceptual Model of Student Satisfaction in Higher Education 1273

effects, if the image of the institution increases or diminishes a unit in terms of valorisa-
tion, the satisfaction will increase or diminish in proportion of 0.86.
The second construct which influences the most satisfaction is the variable value (0.41),
followed by the construct quality perceived (0.24) and lastly, the construct expectations
with a total negative effect of 0.05.
Image is also the construct which has the most influence in the whole model, as it
strongly influences the quality perceived (total effect of 0.92), 0.85 in terms of direct
effects and 0.07 in terms of indirect effects. Its direct influence is less, but also significant,
in the formation of loyalty and word of mouth actions.
In terms of the consequences of satisfaction, it was seen that the influence of satisfaction
is directly reflected upon the formation of loyalty, seeing that in this variable it has an
influence of 0.58 and indirectly upon the actions of word of mouth where it has a
weight of 0.38.
Figure 4. Theoretical and empirical connections
Table 9. Structural equations of the model
Structural
equations
Constructs
Exogenous Endogenous
Image
Customer
expectations
Quality
perceived
Value
perceived
Student
satisfaction
Student
loyalty R
2
Customer
expectations¼
0.578 0.334
Quality
perceived¼
0.851 0.119 0.856
Value
perceived¼
0.562 0.089 0.222 0.684
Student
satisfaction¼
0.452 20.118 0.153 0.405 0.795
Student
loyalty¼
0.233 0.578 0.620
Word of mouth
actions¼
0.235 0.158 0.655 0.933
1274 H. Alves & M. Raposo

It was also seen that the perception of value on the student’s part, influences their word
of mouth actions, whether it be directly with a weight of 0.16, or indirectly with a weight
of 0.15 resulting from their influence through satisfaction and loyalty.
Discussion of the Results
Hypotheses Related to the Expectations and the Quality Perceived
In relation to the influence of expectations and the quality perceived in satisfaction, the
results illustrated that the direct influence of expectations was significant, but with a
reduced weight (total effect of 20.05), in view of which and based on the empirical
evidence of Hyp. 1.a, which suggests that expectations have a significant influence in
the formation of satisfaction, gaining empirical support. Similar to the results encountered
by Eskildsen et al. (1999), this influence was negative.
Nevertheless, it was still reported by the results obtained, that essentially, the students
create expectations concerning the preparation for the career, which is in agreement with
the results obtained by Halstead et al. (1994) and Rautopuro and Vaisanen (2000).
However, contrarily to these researchers, these expectations are high, which meets the
results encountered by Shank et al. (1995), Eskildsen et al. (1999) and Patterson (2000).
Thus, if the students have well defined expectations in respect to the preparation for the
career, according to Halstead et al. (1994), the effect of expectations in satisfaction would
be an indirect influence through the process of disconfirmation. Nevertheless, it was
proven that, that influence was direct, despite being low, as well indirect, through the
quality perceived, in accordance with Anderson & Sullivan (1993) and Kristensen et al.
(1999) and by that proposed by Hyp. 1.b, from which we can say that Hyp. 1.b gained
support in these results.
In relation to the influence of quality perceived, it was seen that apart from being
significant, this was more related with the technical quality of the education service, in
other words, with that, that is received. Functional quality had no influence in satisfaction,
contrarily to what was proposed by Walker (1995), who upholds that in the impossibility
of the students evaluating the central service, they would place more emphasis on the way
the service is provided.
Thus, there seems to be evidence to reject Hyp. 1.c, which assumes that the influence of
quality perceived in satisfaction is higher on the part of functional quality. These results
could be related to the fact of the educational service being too important to the life of
a student, causing them to base their evaluations not only on the way the service is
provided.
Hypotheses Related with Image
The results encountered illustrate that from all the antecedents, the variable image is the
one which has the most influence in the formation process of satisfaction, similar to the
results encountered by Kristensen et al. (1999) and Cassel & Eklo
¨f (2001). This influence
is also significant in the formation of the student’s expectations in higher education.
Thus hypotheses 2.a and 2.b are upheld, both suggesting a direct and significant influ-
ence of image in expectations and in satisfaction.
Conceptual Model of Student Satisfaction in Higher Education 1275

In relation to the direct influence of image in student loyalty, it was proven that this
was significant, despite its influence not being as important as what was discovered by
Eskildsen et al. (1999): This way, evidence was encountered that allow support to the
validity of Hyp. 2.c.
It can also be pointed out that the variable image is the most influential, be it for quality
perceived or value perceived.
Hypotheses Related to Value
Hyp. 3.a suggests that satisfaction in higher education is influenced by the student’s
perception of value. The results show evidence to uphold this hypothesis, given that the
influence of the variable value is, after the influence of the variable image, the one of
the greatest importance. These results meet those of Hartman & Schmidt (1995) and
Webb & Jagun (1997).
In turn, Hyp. 3.b upholds that the variable value is influenced by the quality received
and by student’s expectations. The influence of expectations and quality was both signifi-
cant and positive, despite the reduced importance from which it is possible to say that
Hyp. 3.b was supported.
Hypotheses Related to the Consequences of Satisfaction
Hypotheses related to the consequences of satisfaction are related to both loyalty and to the
student’s involvement in positive word of mouth actions.
The results encountered in this study illustrated that student’s loyalty was the main con-
sequence for satisfaction, given that it was noticeable that the student’s satisfaction had a
direct influence of 0.58 in loyalty.
These results are in accordance with those found by Webb & Jagun (1997) and
Eskildsen et al. (1999). Nevertheless, the results of these investigations also indicated
word of mouth actions as a direct consequence of satisfaction. In this research and
probably due to the fact of the two constructs being separated, word of mouth actions
are influenced by loyalty, by the value perceived and by image.
This way, there exists evidence to support Hyp. 4, in other words, if a student is satisfied
with their educational experience, they will then demonstrate loyalty to that institution. In
turn, Hyp. 5, which assumes that if the student is satisfied with their educational experi-
ence then they will involve themselves in positive word of mouth actions, is partially
rejected as the student’s involvement in word of mouth actions is only influenced by
satisfaction in an indirect manner.
Limitations and Future Areas of Research
In this research it is considered a restriction, the fact of not having studied the behaviour of
complaints as a consequence of satisfaction, in spite of it being decided from the beginning
not to include it since the other two consequences were considered more important.
In view of this restriction, we propose as future areas of research, trying to repeat the
study but this time seeking to include the behaviour of complaints.
Another future area of research is to repeat this study, trying to find alternative
indicators to measure the constructs, namely indicators that present a lower individual
1276 H. Alves & M. Raposo

reliability, just as for example, the indicator of image “Innovatory university turned to the
future”, the indicator of value “Valorisation by the employers”, among others, in order to
succeed in obtaining scales of reliability above 90% for all constructs.
References
Anderson, E. & Fornell, C. (2000) Foundations of the American customer satisfaction index, Total Quality
Management, 11(7), pp. S869– 882.
Anderson, E. & Sullivan, M. (1993) The antecedents and consequences of customer satisfaction for firms,
Marketing Science, 12(2), pp. 125–143.
Anderson, J. & Gerbing, D. (1988) Structural equation modelling in practice: a review and recommended two-
step approach, Psychological Bulletin, 103(3), pp. 411–423.
Astin, A. (2001) What Matters in College? Four Critical Years Revisited (San Francisco, CA: Jossey-Bass).
Bogozzi, R. & Yi, Y. (1989) On the use of structural equation models in experimental designs, Journal of
Marketing Research, 26(3), pp. 271– 214.
Caruana, A. & Money, A. (1997) Service quality and satisfaction – the mediating role of value: an exploratory
study among customers of an audit firm, Proceedings of the MEG 31
st
Annual Conference: Marketing
Without Borders, pp. 113–123.
Cassel, C. & Eklo
¨f, J. (2001) Modelling customer satisfaction and loyalty on aggregate levels – experience
from the ECSI pilot study, Proceedings of the 6
th
TQM World Congress, Saint Petersbourg,
pp. 307– 314.
Chadwick, K. & Ward, J. (1987) Determinants of consumer satisfaction with education: implications for college
and university administrators, College and University, 62(Spring), pp. 236– 246.
CIPES (Centro de Investigac¸a
˜o de Polı
´ticas de Ensino Superior) (1999) Previsa
˜o da evoluc¸a
˜odenu
´mero de
alunos e das necessidades de financiamento - Ensino Superior 1995a 2005, Fundac¸a
˜o das Universidades
Portuguesas, Fevereiro.
Clow, K. et al. (1997) The antecedents of consumer expectations of services: an empirical study across four
industries, The Journal of Services Marketing, 11(4), pp. 230–248.
CQAEHE (Centre for Quality Assurance and Evaluation of Higher Education), CNE (Comite
´National
d’Evaluation) (1998) Evaluation Of European education: a status report, Report Prepared for the European
Commission, DG XXII, September.
Dolinsky, A. (1994) A consumer complaint framework with resulting strategies: an application to higher
education, Journal of Services Marketing, 8(3), pp. 27–39.
Eskildsen, J. et al. (1999) Benchmarking student satisfaction in higher education based on the ECSI methodology,
Proceedings of the TQM for Higher Education Institutions Conference: Higher Education Institutions and
the Issue of Total Quality. Verona, 30 – 31 August, pp. 385 – 402.
Fornell, C. (1992) A national customer satisfaction Barometer: the Swedish experience, Journal of Marketing,
56(January), pp. 6–21.
Garcia, S. & Martinez, T. (2000) Ana
´lisis de ecuaciones estructurales, in: T. L. Martinez (Coordinador) Te
´cnicas
de Ana
´lisis de Datos en Investigacio
´n de Mercados (Edit: Pira
´mide).
Hair, J. et al. (1998) Multivariate Data Analysis, 5th edn (Englewood Cliffs, N. J: Prentice-Hall).
Halstead, D. et al. (1994) Multisource effects on the satisfaction formation process, Journal of the Academy
of Marketing Science, 22(2), pp. 114 – 129.
Hartman, D. E. & Schmidt, S. L. (1995) Understanding student/alumni satisfaction form a consumers’ perspec-
tive, Research in Higher Education, 36(2), pp. 197–217.
Hayes, B. (1998) Measurement Customer Satisfaction: Survey Design, Use, And Statistical Analysis Methods,
2nd edn (Milwaukee, WI: ASQ Quality Press).
Ismail, H. & Abdullah, N. (2001) Does perception of value create customer satisfaction?, Proceedings of the
Academy of Marketing Studies, Nashville, 6(1).
Kristensen, K. et al. (1999) Measurement the impact of buying behaviour on customer satisfaction, Total Quality
Management, 10(4/5), pp. S602– 614.
LeBlanc, G. & Nguyen, N. (1997) Searching for excellence in business education: an exploratory study of
customer impressions of service quality, International Journal of Educational Management, 11(2)
pp. 72–79.
Conceptual Model of Student Satisfaction in Higher Education 1277

Lin, L. (1997) What are student education and educational related needs?, Marketing and Research Today,
25(3), pp. 199–212.
McDougall, G. H. G. & Levesque, T. (2000) Customer satisfaction with services: putting perceived value into the
equation, Journal of Services Marketing, 14(5), pp. 392 – 410.
Patterson, P. (2000) A contingency approach to modelling satisfaction with management consulting services,
Journal of Service Research, 3(2), pp. 138–153.
Rautopuro, J. & Vaisanen, P. (2000) “Keep the customer satisfied”. A longitudinal study of students’ emotions,
experiences and achievements at the University of Joensuu, Paper Presented at The European Conference on
Educational Research, Edinburgh, 20–23 September.
Ryan, M. et al. (1995) Making CSM a power tool: composite indices boost the value of satisfaction measures for
decision making, Marketing Research, 7(3), (Summer), pp. 11– 16.
Santos, S. (1995) Tende
ˆncias do ensino universita
´rio em Portugal, Semina
´rio International Os Estudos Superiores
em Macau, Macau, Janeiro, (16 January).
Shank, M. et al. (1995) Understanding professional service expectations: do we know what our students expect in
a quality education?, Journal of Professional Services Marketing, 13(1), pp.71– 89.
Thomas, M. et al. (1996) Student withdrawal from higher education, Educational Management and Adminis-
tration, 24(2), pp. 207–221.
Ugolini, M. (1999) University dropout: a problem and an opportunity, Proceedings of the TQM for Higher
Education Institutions Conference: Higher Education Institutions and the Issue of Total Quality, Verona,
30–31 August, pp. 417 – 436.
Walker, J. (1995) Service encounter satisfaction: constructualized, Journal of Services Marketing, 9(1), pp. 5 –14.
Walther, E. (2000) The Relationships Between Student Satisfaction And Student Retention In Higher Education.
Dissertation submitted to the Faculty of The Graduate School at The University of North Carolina at
Greensboro.
Webb, D. & Jagun, A. (1997) Customer care, customer satisfaction, value, loyalty and complaining behaviour:
validation in a UK university setting, Journal of Consumer Satisfaction, Dissatisfaction and Complaining
Behaviour, 10, pp. 139–151.
Wiese, M. (1994) College choice cognitive dissonance: managing student/institution fit, Journal of Marketing
for Higher Education, 5(1), pp. 35 – 47.
Yi, Y. (1993) The determinants of consumer satisfaction: the moderating role of ambiguity, Advances in Consu-
mer Research, 20, pp. 502–506.
1278 H. Alves & M. Raposo