Sie sind auf Seite 1von 9

WILLIAM O. BEARDEN and JESSE E.

TEEL*
Data obtained from 375 members of a consumer panel in a two-phase study of
consumer exfjeriences with automobile repairs and services were used to examine
the antecedents and consequences of consumer satisfaction. The results support pre-
vious findings that expectations and disconfirmatian are plausible determinants of
satisfaction, and suggest that complaint activity may be included in satisfaction/
dissatisfaction research as suggested by earlier descriptions of consumer complain-
ing behavior.
Selected Determinants of Consumer Satisfaction
and Complaint Reports
Consumer satisfaction is important to the marketer
because it is generally assumed to be a significant de-
terminant of repeat sales, positive word-of-mouth, and
consumer loyalty. Satisfaction is important to the indi-
vidual consumer because it reflects a positive outcome
from the outlay of scarce resources and/or the fulfill-
ment of unmet needs (Day and Landon 1977; Landon
1977).
Growing recognition of consumer satisfaction as a
critical construct in marketing and consumer behavior
has generated substantial research interest in the pro-
cesses preceding judgments of satis faction/dissatisfac-
tion and the consequences of those decisions. This re-
search has consisted primarily of experimental
Investigations of consumer expectations (e.g., Anderson
1973; Cardozo 1965) and surveys of problem incidences
and reports of redress-seeking activity (e.g., Andreasen
1977; Day and Ash 1979).' In an effort to integrate the-
ory and prior empirical research in the area of consumer
satisfaction, Oliver (1980a) reported the development
and test of a model encompassing interrelationships
among expectations, disconfirmation. satisfaction, and
the traditional criteria of attitudes and intentions. Our
study is an attempt to extend this research effort by in-
corporating consumer complaining behavior into a the-
oretical model of consumer satisfaction.
'For an excellent review of these experimental studies, see LaTour
and Peal (1979).
*WilHam O, Bearden and Jesse E. Teel are Associate Professors
of Marketing, University of South Carolina.
COMPLAINING BEHAVIOR^
Complaining behavior research has been largely lim-
ited to retrospective self-reports of behavior which oc-
curred prior to assessment of the antecedents of satis-
faction/dissatisfaction. Though this limitation stems
from the inherent shortcomings of cross-sectional data
collection designs typically used in consumer satisfac-
tion research, surveys of sources of consumer dissatis-
faction and reports of complaint actions after unsatisfac-
tory purchase experi ences have enhanced our
understanding of consumer problems and consumer ef-
forts to seek redress.
Unsatisfactory purchases, though varying by product
or service category, appear to be prevalent. For exam-
ple, Andreasen and Best (1977) report that as many as
one in five purchase experiences results in some dissat-
isfaction. Similarly, Day and Bodur (1978) and Day and
Ash (1979) report frequent incidences of dissatisfaction
for services and durable products.
Reaction to dissatisfaction in terms of complaining
behavior and redress seeking, however, varies consid-
erably. Frequently, consumers do not take action to al-
leviate marketplace problems. Day and Bodur (1978, p.
265) found that reported cases of extreme dissatisfaction
for which no action of any kind was taken were 49.6%
^Day (1980) has classified consumer responses to dissatisfaction
into three categories: (1) redress seeking in which a specific remedy
is sought, (2) complaining or communications for reasons other than
redress seeking, and (3) boycotting or personal decisions to discon-
tinue usage or patronage. For purposes of our study, consumer com-
plaint reports are used in the generic sense to include a range of these
activities.
21
Journal of Marketing Research
Vol. XX (February 198.1), 21-8
22 JOURNAL OF MARKETING RESEARCH, FEBRUARY 1983
for nondurable products, 29.4% for durable products,
and 23.2% for services. Similarly, Andreasen and Best
(1977, p. 96) report that well over half of all the non-
price purchase difficulties uncovered in their study pre-
cipitated no action. These results are corroborated by the
earlier national survey of Warland, Hemnann. and Wil-
lits (1975) and a recent regional study reported by Shup-
trine and Wenglorz (1981).
These findings are discouraging for several reasons.
First, failure to express dissatisfaction prevents the con-
sumer from achieving redress from an unpleasant mar-
ketplace experience. Second, limited action on the part
of consumers may mask marketplace problems which
the firm could and/or should correct. Further, wide-
spread failure to express complaints limits the usefulness
of complaint data as a basis for policy. For example,
complaint data have been suggested as being useful for
analysis of consumer discontent over time and across
products (e.g., Gronhaug and Amdt 1980). Munns
(1978) has advocated providing complaint data to con-
sumers as prepurchase information. However, the lim-
ited research on reactions to dissatisfaction and the lack
of representative data limit these potential applications.
Our study was designed to further understanding of
consumer satisfaction by integrating complaint behavior
into an explanation of consumer satisfaction. The frame-
work examined (Figure 1) reflects a revision of Oliver's
(1980a, p. 463) original model to include complaint be-
havior. Satisfaction is depicted as a function of con-
sumer expectations operationalized as product attribute
beliefs (Olson and Dover 1979) and disconfirmation.
Expectations and disconfirmation are presented as being
unrelated, additive, and exogenous to the system. As is
consistent with the multiattribute explanation of atti-
tudes, expectations/beliefs about product attributes are
included as determinants of attitudes which precede in-
Figure 1
THEORETICAL ANTECEDENTS AND CONSEQUENCES OF
CONSUMER SATISFACTION
Figure 2
CONSUMER COMPLAINT BEHAVIOR"
"Based on Day and Landon (1977).
tentions. Satisfaction/dissatisfaction is, in tum, pre-
sumed to infiuence subsequent attitudes, intentions, and
complaint behavior.
The hypothesized relationships between satisfaction
and expectations and disconfirmation are based on
Oliver's (1980a, b) interpretation of Helson's (1964)
adaptation level theory which posits that one perceives
stimuli in relation to an adapted standard. Expectations
are assumed to perform the function of an adaptation
level in that they define the standard against which sub-
sequent performance is judged. Disconfirmation is as-
sumed to serve as a major force causing movement away
from the standardthe net effect of which is satisfaction
or dissatisfaction (Oliver 1981. p. 28).
Though a substantial body of research supports the
hypothesized causal chain between beliefs, attitudes, and
intentions, the effects of disconfirmation in conjunction
with expectations on satisfaction have received less at-
tention. Some support for the additive and unrelated as-
sumption about expectations and disconfirmation was
provided by the zero order correlations reported by
Oliver (1980a). However, disconfinnation may also in-
fiuence satisfaction and postpurchase attitudes through
an interactive relationship with expectations. In an effort
to help clarify this issue, an alternative version of the
model incorporating disconfirmation as a moderator
variable was tested (Sharma, Durand, and Gur-Arie
1981, p. 293).
All of the relationships with the exception of the path
between satisfaction and complaining are hypothesized
to be positive. For this negative path, satisfaction is as-
sumed to be related inversely to complaint behavior. The
nature of the satisfaction/dissatisfaction and complaint
process as conceptualized by Day and Landon (1977) is
shown in Figure 2. Though many factors may inhibit or
encourage complaint expressions, dissatisfaction is rec-
ognized as a primary determinant of legitimate consumer
complaints. As shown, the various actions that con-
sumers take may be subdivided into private and public
responses. Private actions include decisions to stop fur-
ther purchases and warnings to friends; public actions
DETERMINANTS OF CONSUMER SATISFACTION 23
include redress-seeking efforts directed toward the seller
and complaints to third-party consumer affairs institu-
tions.
The potential for other individual and situational fac-
tors to infiuence complaining behavior should be ac-
knowledged. Many of these factors have been examined
in the consumer satisfaction/dissatisfaction literature
and include socioeconomie characteristics (Bourgeois
and Barnes 1979), personality differences (Fomell and
Westbrook 1979), the costs and benefits from complain-
ing (Richins 1982), propensity to complain (Day and
Landon 1977), and importance of purchase (Gronhaug
1977). Many of these infiuences on complaining behav-
ior have been reviewed by Landon (1977) and Day and
Landon (1977).
METHOD
Data were collected in a two-phase longitudinal study
with a four-month measurement interval from members
of a t200-family. two-state consumer panel. Expecta-
tions, attitudes, intentions, satisfaction, and complaint
reactions were collected for the use of automobile repair
service outlets.^ The results described are based on the
375 user adult respondents from an original sample of
749 households participating in both waves.
As part of the initial data collection, respondents were
requested to provide the name and address of an auto
service outlet which might be used during the coming
months. Beliefs, attitudes, and intentions were then col-
lected with respect to that outlet. These variables were
measured again in a followup mailing which also as-
sessed reports of complaint behavior. Panel members
excluded from the final sample were respondents pro-
viding incomplete responses to one or both mailings,
nonusers of the originally identified repair service facil-
ity, and those respondents who were not familiar with
a repair outlet. Names and addresses of the auto busi-
nesses were also collected in the followup mailing for
help in verification and matching of surveys. Fifty-four
percent of the respondents were men; 93% were white.
The median family income category was $25-30,000
and the respondents' age was 49.94 years. The 749 sam-
ple members were compared with the 451 nonpartici-
pants by means of the permanent demographic file main-
tained on each panel member and updated annually. No
significant differences were found in terms of demo-
graphic characteristics.
Operational Measures
Expectations, attitudes, and intentions were assessed
by using multi-item measures commonly employed in
attitude research (e.g., Fishbein and Ajzen 1975). Dis-
confirmation of expectations and satisfaction were as-
sessed by procedures similar to those described by
Oliver (1980a). An index of complaint behavior was
developed on the basis of the research of Day and his
colleagues into the range of actions consumers may take
after an unsatisfactory purchase experience. All scales
were worded to refiect the automobile repair service con-
text. An initial version of the questionnaire was tested
by a mail survey of the general population of a medium-
sized metropolitan area.
Expectations were collected for the six service attri-
butes identified by Adier and Hlavacek (1978): (1) rea-
sonableness of costs, (2) location, (3) quality of repairs,
(4) reputation of firm, (5) speed of service, and (6)
friendliness of management. Each was operationalized
as a seven-item bipolar statement labeled "likely"-"un-
likely" and scored -1-3 to 3. As is consistent with the
multiattribute depiction of belief structure and the pro-
cedures suggested by Oliver (1980a), expectations were
formed as a sum of the six belief scores. Evaluation di-
mensions were not included because of the assumptions
of unit positive evaluation for each attribute and over-
time stability of attribute evaluations. Attitudes toward
patronizing the outlet were assessed by three seven-place
scaled statements labeled " good" - " bad, " " wi s e" -
"foolish, " and "beneficiaP' -' ^' harmfuI." Statements
were similar to: "My taking vehicles for needed repairs
to (business) is. . . ." Behavioral intentions were op-
erationalized via two seven-place scaled statements la-
beled "likely"-"unlikely" and "probable"'-"improba-
bl e . " Similar items were used in the followup
questionnaire to represent both attitudes and intentions.
In the followup analyses, disconfirmation was as-
sessed by a single measure labeled "better than"-
"worse than" reflecting each respondent's overall re-
actions to the disconfirmation of expectations. Satisfac-
tion was defined via four "agree"-"disagree" state-
ments similar to those used by Oliver (1980a)."' The
satisfaction items were similar to the following two ex-
amples: (1) "Our choice to use the repair outlet was a
wise one. " or (2) "If I had it to do all over again, I
would feel differently about using the business."
An index of complaint activity was constructed to rep-
resent complaint behavior. The operational measure of
complaint behavior consisted of a scale containing re-
sponses to the personal and direct complaint alternatives
identified and investigated in prior studies of consumer
reactions to marketplace dissatisfaction (Day and Ash
1979; Day and Bodur 1978; Day and Landon 1977).
These actions ranged from "warned family and friends"
'Problems with automobile repairs and service are one of the most
frequently mentioned consumer complaints. Because of yearly ex-
penditures now in excess of $30 billion and the perva.sive u.se, im-
portance, and complexity of automobiles, consumer problems with
auto repairs and service are worthy of study in themselves (Survey
1979; Webbink 1978).
"Data were collected for the complete set of six items used by
Oliver (1980a, p. 463) to operationalize satisfaction. Two of these
items were omitted on the basis of the results of a reliability analysis
(coefficient alpha) which indicated that the two items lowered the
overall reliability of the scale and were apparently perceived as am-
biguous by the respondents.
24 JOURNAL OF MARKETING RESEARCH, FEBRUARY 1983
Table 1
INDICATOR CORRELATIONS, MEANS, AND STANDARD DEVIATIONS"
y
y
y
^ 4
'y
' '
y
y
y
y
y
y,
.70
.70
.66
.70
.50
.49
.42
.44
.48
.49
.50
.43
.48
-.25
.74
.22
y:
.56
.71
.85
.83
.49
.47
.42
.43
.48
.54
.54
47
.54
-. 19
.70
?l
y.
.55
.81
.63
.69
.43
.36
.30
.33
.42
.42
.46
.38
.42
-. 24
.63
IS
y*
.46
.68
69
.90
4^
.36
.30
.36
.38
.42
.42
38
41
- 15
6?
y j
.43
.69
.71
.75
.46
.42
.39
.35
.42
.43
.42
.39
.44
-. 21
.64
.18
y*
11
19
19
.20
.26
.81
.73
.76
.70
.73
68
.66
7?
- 44
.54
SO
yj
.31
.40
.39
.24
.29
.83
.90
.77
.68
.72
.69
.69
.74
-. 36
.50
.45
y
.30
.41
.38
.27
.28
.80
.83
.77
.68
.70
.67
70
.74
-. 34
40
19
y^
.34
.31
.33
.24
21
.70
.69
.75
.61
.66
65
62
.69
-. 37
45
4?
y,o
.37
.45
.47
.30
.28
.68
.65
.65
.59
.90
.82
87
.84
-. 36
49
.36
y j i
.44
49
.51
.36
.34
7?
.73
7?
.64
.85
.83
90
QO
-. 33
48
19
y. i
.34
.33
.43
.29
.29
.59
.59
.58
.52
.73
.77
80
.81
-. 25
50
11
y]3
.34
.34
.40
.38
.38
.65
.60
.63
.60
.65
.75
.71
.93
-. 32
40
.37
y,4
.31
.33
.41
.38
.38
.61
.56
.59
.54
.61
.71
.73
.85
-. 36
45
.42
y.s
-. 11
-. 11
-. 09
-. 08
- 01
-. 23
-. 25
-. 32
-. 31
-. 34
-. 31
-. 27
-. 26
-. 26
11
.24
.65
.66
.66
.45
.52
.50
.46
.45
.37
.61
.59
.48
.46
.43
-. 28
78
Xi
15
.20
.25
.15
08
44S
.44
.46
.44
.48
.52
.45
.45
.41
- 14
.19
Initial
Mean
5.74
5.96
5.90
6.26
6.28
5.89
5.94
6.14
6.01
5.96
5.95
6.06
5.99
6.04
.52
9.23
4.37
SD
1.43
L28
1.30
1.2!
1 18
1.70
1.67
1.57
1.76
1.33
1.46
1.35
1.64
1.59
.93
7.12
1.55
Repli-
cation
Mean
5.79
6.16
6.21
6.42
6.46
6.01
5.95
6.04
5.91
5.95
5.90
5.95
6.02
5.98
.52
9.77
4.44
SD
1.49
1.14
.98
1.04
.91
1 54
1.54
1.57
1.74
1.44
1.51
1.38
1.62
1 6?
94
7.04
1.40
"Correlations below the diagonal are based on the initial sample data; correlations above the diagonal are based on the replication sample.
to "contacted lawyer or took some legal action." The
construct was operationalized as a Guttman scale in an
effort to reflect increasing intensity of possible com-
plaint actions. The coefficients of reproducibility and
scalability were .98 and .78, respectively/
RESULTS
The 375 user respondents were split randomly into
initial {n = 188) and replication (n = 187) samples. The
models for both samples were estimated by the system
estimation procedures developed by Joreskog and Sorbom
(1978) and reviewed by Aaker and Bagozzi (1979). Con-
struct indicator intercorrelations, means, and standard
deviations for both groups are reported in Table 1.
Initial Sample
Overall analysis. Analysis of the hypothesized model
(see Figure I) for the initial sample resulted in a chi
square statistic of 305.13 (108 d.f., p < .01).^ Exami-
nation of the first derivatives of the endogenous indicator
error terms and tests of the model allowing additional
paths among constructs suggested that the fit of the
'The items included, in order of increasing complaining intensity,
were: (1) wamed family and friends, (2) returned vehicle for rework
and/or complained to management. (3) contacted manufacturer, (4)
contacted Better Business Bureau, state office of consumer affairs,
or private consumer agency, and (5) took some legal action.
''As is consistent with procedures used in multiwave studies, the
indicator coefficients for the two identical measures of attitude and
of intentions were restricted to be equal over the two waves. These
restrictions result in exactly equal estimates for the unstandardized
parameter values: however, the standardized coefficients may differ.
Further, the structural error terms and the errors in measurement for
the two identical attitude and the two intentions measures were al-
lowed to he correlated.
model to the data could be improved. Specifically, the
two exogenous variablesexpectations and disconfir-
mationwere allowed to be correlated and four paths
among error terms within three of the constructs were
freed. Two correlated errors were allowed among the
four satisfaction indicators and one each within the time
1 and time 2 attitude constructs. The resulting chi square
is 263.82 (103 d.f., p < .01). These results are shown
in Figure 3.
Normally, a significant chi square value is taken as
evidence that the model is an inadequate fit or poor rep-
resentation of the dat a/ Bentler and Bonett (1980) sug-
gest computation of an incremental fit index for com-
paring a theoretical model with a relevant null model.
For purposes of our study, the null model chosen for
comparison was that of no relationships in which only
the error term paths were allowed to be free. This is
analogous to the test of the hypothesis that the vector of
regression coefficients is the zero vector in a single cri-
terion variable multiple regression analysis.
Tests of this null model based on the initial sample
resulted in a chi square value of 3599.43 (136 d.f.). This
finding suggests that the model does result in a substan-
tial improvement over the assumption of no relation-
ships. This conclusion is supported by a normalized in-
'The split-half samples in this study were 188 and 187 respondents
for the initial and replication studies, resfwctively, and could not be
described as especially large. However, the eight constructs and 17
indicators result in a somewhat more complex framework than typi-
cally used in most marketing applications. Though Bagozzi (1981)
has suggested that LISREL is appropriate if the sample size minus the
degrees of freedom exceeds 50, the effects on the distribution of chi
square from small sample sizes and the use of nonmultivariate nor-
mally distributed data are currently unknown.
DETERMINANTS OF CONSUMER SATISFACTION 25
Fi gur e 3
PATH ANALYSIS RESULTS"
'Standardized coefficients for the iniitial sample analyses are shown above each arrow; the chi square statistic is 263.82 (103 d.f., p < .01).
Coefficients for the replication sample are shown in parentheses; the resulting chi square statistic is 179.94 (103 d.f., p < .01). All indicator
coefficients in both analyses are significant.
"Significant in both analyses.
'Significant in only the replication {p < .01).
cremental fit index of .93 and a non-normed fit index
of .90 (Bentler and Bonett 1980, p. 599). It is further
supported by the average residual between the original
and reproduced correlations of . 05! in comparison with
an average input correlation of .521.
Measurement model. Each of the indicators is signif-
icant as a reflection of the unobservable construct it was
designed to assess. These coefficients and their error
terms are also shown in Figure 3. The measurement
model was analyzed further by the procedures suggested
by Bagozzi (1980) and Fomell and Larcker (1981). The
resulting indicator and construct reliabilities and shared
variance estimates for both the initial and replication
analyses are reported in Table 2.^ In general, these es-
timates suggest adequate indicator and construct relia-
bilities. With the exception of one indicator in the atti-
tude measures, all indicator reliabilities are above .70.
The construct reliability estimates range from .88 to .95.
The shared variance estimates representing the amount
of variance captured by the construct in relation to the
"Individual indicator reliabilities for those constructs measured by
multiple items are computed from model estimates derived from the
LISREL analysis. Also, on the basis of the responses of 35 under-
graduate students to a three-week administration, test-retest reliability
estimates for summated versions of the expectations, attitude, and in-
tention measures were .70, .76, and .96, respectively.
amount of variance due to measurement error range from
.70 to .90.
Structural model. Eight of the 10 hypothesized paths
are significant (p < .05) and each is consistent with the
hypothesized direction of influence. Expectations are
related positively to attitudes as anticipated, and across
time to satisfaction. Similarly, disconfirmation is a sig-
nificant predictor of satisfaction. Satisfaction, in tum,
is significantly correlated with postpurchase attitudes.
As is consistent with the normal depiction of attitude
structure, the strongest relationships are between atti-
tudes and intentions within time periods. Last, satisfac-
tion is related negatively to complaint behavior as hy-
pothesized. The two nonsignificant paths involve the
relationship between disconfirmation and followup atti-
tudes and between the two intentions measures.
Replication Sample
Overall analysis. Analysis of the model allowing the
exogenous variables to be correlated and the within-con-
struct correlated errors was replicated on the remaining
half of the sample. The resulting chi square for the rep-
lication analysis is 179.94 (103 d.f., p < .01), Test of
the null model resulted in a chi square value of 2750.63
(136 d.f.). The normalized and non-normed incremental
fit indices are .93 and . 91, respectively. The average
residual between the original input correlation matrix
(averaging .454) and the reproduced correlations is .043.
26 JOURNAL OF MARKETING RESEARCH, FEBRUARY 1983
Construct
Attitude (1)
Intentions (1)
Satisfaction
Attitude (2)
Intentions (2)
No. of
indicators
3
2
4
3
2
Table
RELIABILITY AND SHARED
Indicator
reliability range
Initial
.64-.83
.90-.90
.76-.96
.71-.90
.90-92
Replication
.66-.8I
.74-. 76
.66-.83
.72-.88
.88-92
2
VARIANCE
Initial
.88
.95
.95
.94
.96
ESTIMATES
Construct
reliabitiry
Replication
.87
.85
.93
.90
.92
Shared
Initial
.70
.90
.82
.84
.92
variance
Replication
.70
.75
.76
.79
.85
Measurement model. As in the initial sample analysis,
each of the indicator coefficients is significant. Exami-
nation of their reliabilities (see Table 2) and the mag-
nitude of their error variances indicates that the individ-
ual reliability estimates are above .60. The reliability
estimates for the constructs with multiple indicators
range from .85 to .93. The shared variance estimates
range from .70 for the first attitude construct to .85 for
the second intentions variable.
Structural model. In the replication, nine of the 10
proposed causal paths are significant. Again, each is
consistent with the hypothesized direction of influence.
The eight significant paths based on the initial sample
analyses are also significant in the replication and are
comparable in magnitude to the initial sample results.
The path between the two intentions measures is again
not significant.
In sum, the results support the previous findings that
expectations and disconfirmation appear to be plausible
determinants of satisfaction. The traditional directional
relationships between beliefs, attitudes, and intentions
are also confirmed. Moreover, satisfaction is related
negatively to complaint activity as hypothesized.
Additional Tests
Disconfirmation as a moderator. In an effort to ex-
amine the effect of disconfirmation on the relationships
between expectations and satisfaction and postpurchase
attitudes, we tested an altemative model incorporating
disconfirmation as a quasi moderator (Sharma, Durand,
and Gur-Arie 1981, p. 295) of these relationships for
both the initial and replication samples. A quasimoder-
ator variable is allowed to affect a criterion variable di-
rectly and also indirectly through an interactive relation-
ship with another variable. For this test, the model of
Figure 3 was reformulated with an additional exogenous
variableexpectations multiplied with disconfirmation.
The resulting chi square values are 301.41 (116 d.f.) and
194.87 (116 d.f.). The paths from the interaction vari-
able to satisfaction and to postpurchase attitudes are not
significant for the initial sample (p < .05). The corre-
sponding path to satisfaction is significant in the repli-
cation sample, but the path to postpurchase attitudes is
not. Given the ambiguity of these results, we cannot rea-
sonably either reject or accept disconfirmation as a mod-
erator of the influence of expectations on consumer sat-
isfaction and postpurchase attitudes, and this possibility
should be considered for future research.
All-respondent analysis. With the caveat that corre-
lations among constructs may be caused by all measures
assessing the same construct, correlations between pur-
ported measures of intentions, attitudes, and attitude
components have been accepted as evidence of nomo-
logical validity (Peter 1981). Consequently, a reduced
model incorporating only expectations, attitudes, and in-
tentions was tested on all 749 user and nonuser respond-
ents. Again the indicators for identical constructs were
constrained to be equal over time and correlated mea-
surement errors were allowed.
The magnitude and signs of the path coefficients
among the constructs closely parallel the relationships
shown in Figure 3. That is, the relationships between
expectations, attitudes, and intentions within time pe-
riods are positive, relatively large, and significant. The
indicator coefficients are again significant and provide
evidence of high construct and indicator reliability. The
correlations of the indicators with the constructs are
higher in all cases for the construct each set of indicators
was designed to measure than those for constructs mea-
sured by different indicators. Though the resulting chi
square value is comparable to the initial and replication
user analyses (272.44, d.f. = 36, p < .01), given the
smaller number of degrees of freedom, the overall fit of
the model is somewhat worse. This finding is not un-
expected given the omission of the satisfaction construct
and the probable lower degree of involvement or famil-
iarity with the repair service outlet for the nonusers.
DISCUSS/ON
These results provide additional support for the find-
ings of Oliver (1977. 1980a) on expectations and dis-
confirmation as determinants of consumer satisfaction.
In both analyses, expectations and disconfirmation are
related positively to satisfaction. Satisfaction, in tum,
appears to partially mediate the relationship hetween
disconfirmation and followup attitudes.
Though our results are largely consistent with the re-
search hypotheses, several caveats are in order. First,
the study is limited by its focus on only a single type of
service and the use of a single-item, overall measure of
disconfu^mation. The reliability of this measure is un-
known because no test-retest analysis was performed.
DETERMINANTS OF CONSUMER SATISFACTION 27
Second, in spite of measurement at two time periods
separated by four months, several constructs were mea-
sured simultaneously on the same questionnaire and are.
hence, subject to common response bias. Further, the
complaint behavior measure is a unidimensional variable
(i.e., from no actions to severe actions) and the corre-
lation of this measure with a bidirectional satisfaction
measure may have been attenuated. Last, the large sig-
nificant chi square values suggest some model misspe-
cification which in this case is probably due to omission
of constructs likely to influence such a complex behavior
as complaining (e.g., costs and benefits of complaining,
individual circumstances) and/or measurement error.
The study does, however, overcome some of the
shortcomings of previous consumer satisfaction/dissat-
isfaction research. The use of panel data helped avoid
the tendency to sample only extreme cases (typically
found in complaint files or self-reported in single sur-
veys covering the preceding 12 months). Further, the
integration of two existing and complementary theories,
i.e., explanations of both consumer satisfaction and
complaint behavior (topics which are normally not in-
cluded simultaneously in empirical research) circum-
vents further "overtheorizing"" an area already criticized
for that tendency (Russo 1979). In contrast to Oliver's
inoculation study, the focus on automobile services and
repairs represents a situation in which consumers select
from a number of altemative outlets and pay for ser-
vices, and so more closely reflects normal consumer ac-
tivity. The availability of altemative outlets also pro-
vides the consumer with more ways to respond to an
unsatisfactory experience (e.g., changing outlets).
Implications for Satisfaction Research
Several issues pertaining to expectations and discon-
firmation as determinants of satisfaction remain to be
addressed. For the relationship between expectations and
satisfaction, several possibilities seem plausible. First,
satisfaction as operationalized may be simply another
form of attitude. In this case, the expectations sat-
isfaction link represents another approximation of the
general depiction of attitude structure. Second, because
users were the focus of the study, their responses may
have been biased by attempts to respond consistently
with the first wave of data collection. One could also
argue that higher levels of satisfaction are reported in
efforts to rationalize earlier high expectations. Oliver
(1980b) has suggested that certain dogmatic tendencies
resulting from involvement or commitment may make
some individuals rely more heavily on expectation lev-
els.
For disconfinnation, it is apparent that there is still a
need for further research into why consumers displace
themselves at various points along the disconfirmation
continuum. Problems remain in the operationalization of
disconfirmation. As suggested by Day (1980) and Oliver
(1980a), issues of the relative appropriateness and effi-
cacy of overall versus individual attribute disconfirma-
tion measures badly need resolving. It should also be
noted that, though tests of our model using disconfir-
mation as a moderator variable did not improve face
validity of the results or appreciably the overall fit, we
found some evidence that disconfirmation and expecta-
tions are correlated constructs.
The consumer satisfaction paradigm examined appears
capable of encompassing more concrete behavioral cri-
teria in a manner that is consistent with previous theories
of consumer complaining (e.g., Andreasen 1977; Day
1977; Day and Landon 1977; Landon 1977). Specifi-
cally, satisfaction is found to be related negatively to an
index of complaint reports in both analyses. Because of
the positive contributions to both consumers and mar-
keters that are possible from early expressions of dis-
satisfaction, understanding the determinants of com-
plaint behavior seems important. The ability of a single
satisfaction variable to moderately explain complaint
behavior is encouraging. Much as interpersonal and
other affective influences have been shown to affect con-
sumer satisfaction (cf. Westbrook 1980), both intemal
characteristics (e.g., perceived risk, assertiveness) and
extemal structural factors (e.g.. economic constraints,
the benefits and costs from complaining) may also in-
hibit or encourage complaint behavior (Richins 1980,
1982). Future efforts need to incorporate these variables
into further research on the determinants of consumer
satisfaction and complaining behavior.
Additionally, several other issues related to consumer
complaining behavior are unresolved and in need of con-
sideration. For example, the role of the relative intensity
of satisfaction and the interaction of (cf Day 1980) dis-
satisfaction with the benefits (or the potential payoffs)
from complaining need addressing. Further, the con-
sumer complaining behavior construct itself warrants
additional conceptualization and measurement attention.
REFERENCES
Aaker, David A. and Richard P. Bagozzi (1979), "Unobserv-
able Variables in Structural Equation Models with an Ap-
plication in Industrial Selling," Journal of Marketing Re-
search. 16 (May), 147-58.
Adlcr, Lee and James D. Hlavacek (1978), "Key Repair Ser-
vice Factors for Consumer Durable Goods," Journal of
Marketing Research, 15 (November), 634-8.
Anderson, Rolph E. (1973), "Consumer Dissatisfaction: The
Effect of Disconfirmed Expectancy on Perceived Product
Performance," Journal of Marketing Research, 10 (Febru-
ary) 38^W.
Andreasen, Alan R. (1977), "A Taxonomy of Consumer Sat-
isfaction/Dissatisfaction Measures," Journal of Consumer
Ajfairs. U (Winter), 11-23.
and Arthur Best (1977), "Consumers Complain
Does Business Respond,"" Harvard Business Review. 55
(July-August), 93-101.
Bagozzi, R. B. (1980), Causal Models in Marketing. New
York: John Wiley & Sons, Inc.
(1981), "Evaluating Structural Equation Models with
Unobservable Variables and Measurement Error: A Com-
28 JOURNAL OF MARKETING RESEARCH, FEBRUARY 1983
ment," Journal of Marketing Research. 18 (August), 375-
81.
Bentler. P. M. and Douglas G. Bonett (1980), "Significance
Tests and Goodness of Fit in the Analysis of Covariance
Structures," Psychological Bulletin. 88, 588-606.
Bourgeois, J. C. and James G. Bames (1979), "Viability and
Profile of the Consumerist Segment," Journal of Consumer
Research, 5 (March). 217-28.
Cardozo, Richard N. (1965). "An Experimental Study of
Consumer Effort, Expectation and Satisfaction," Journal of
Marketing Research. 2 (August), 244-9.
Day, Ralph L. (1977), "Extending the Concept of Consumer
Satisfaction." in Advances in Consumer Research, Vol. 4.
W. D. Perreault. ed. Atlanta: Association for Consumer Re-
search, 149-54.
(1980), "Research Perspectives on Consumer Com-
plaining Behavior. " in Theoretical Developments in Mar-
keting, C. W. Lamb and P. M. Dunne, eds. Chicago:
American Marketing Association. 211-15.
and Stephen B. Ash (1979). "Consumer Response to
Dissatisfaction with Durable Products," in Advances in
Consumer Research, Vol. 6, W. L. Wilkie, ed. Miami:
Association for Consumer Research. 438-44.
and Muzaffer Bodur (1978). "Consumer Response to
Dissatisfaction with Services and Intangibles.'" in Advances
in Consumer Research, Vol. 5, H. Keith Hunt, ed. Ann
Arbor: Association for Consumer Research, 263-72.
and E. Laird Landon (1977), "Toward a Theory of
Consumer Complaining Behavior," in Consumer and In-
dustrial Buying Behavior. A. G. Woodside, J. N. Sheth,
and P. D. Bennett, eds. New York: North-Holland. 425-
37.
Fishbein, Martin and leek Ajzen (1975), Belief. Attitude. In-
tention, and Behavior: An Introduction to Theory and Re-
search. Reading, MA: Addison-Wesley Publishing Co.
Fomell, Claes and David F. Larcker (1981), "Evaluating
Structural Equation Models with Unobservable Variables
and Measurement ETTOT,^'Journat of Marketing Research.
18 (February), 39-50,
and Robert A. Westbrook (1979). "An Exploratory
Study of Assediveness, Aggressiveness, and Consumer
Complaining Behavior. " in Advances in Consumer Re-
search. Vol. 6, W. L. Wilkie, ed. Miami: Association for
Consumer Research, 105-10.
Gronhaug, Kjell (1977), "Exploring Consumer Complaining
Behavior: A Model and Some Empirical Results. " in Ad-
vances in Consumer Research. Vol. 4, W. D. Perreault, ed.
Atlanta: Association for Consumer Research. 159-65.
; and Johan Amdt (1980). "Consumer Dissatisfaction
and Complaining Behavior as Feedback: A Comparative
Analysis of Public and Private Delivery Systems. " in Ad-
vances in Consumer Research. Vol. 7, Jerry Olson, ed. San
Francisco: Association for Consumer Research, 324-8.
Helson, Harry (1964). Adaptation Level Theory. New York:
Harper and Row.
Joreskog. Karl and Dag Sorbom (1978), LISREL IV: Analysis
of Linear Structural Relationships hy the Method of Maxi-
mum Likelihood. Chicago: National Educational Resources,
Inc.
Landon, E. Laird (1977), "A Model of Consumer Complaint
Behavior," in Consumer Satisfaction. Dissatisfaction, and
Complaining Behavior, Ralph L. Day, ed. Bloomington:
Indiana University, 31-5.
LaTour. Stephen A. and Nancy C. Peat (1979), "Conceptual
and Methodological Issues in Consumer Satisfaction Re-
search," in Advances in Consumer Research. Vol. 6, W.
L. Wilkie. ed. Miami: Association for Consumer Research,
431-7.
Munns, Joyce M. (1978). "Consumer Complaints as Pre-pur-
chasc Information: An Evaluation of Better Busine.ss Bureau
Reports to Consumers. " Journal of Consumer Affairs, 12
(Summer), 76-87.
Oliver, Richard L. (1977), "Effect of Expectation and Dis-
confirmation on Postexposure Product Evaluations: An Al-
temative Interpretation," Jourtial of Applied Psychology,
62, 480-6.
(1980a). "A Cognitive Model of the Antecedents and
Consequences of Satisfaction Decisions," Journal of Mar-
keting Research, 17 (November), 460-9.
1980b), "Theoretical Bases of Consumer Satisfaction
Research: Review. Critique, and Future Direction," in The-
oretical Developments in Marketing, C. W. Lamb and P.
M. Dunne, eds. Chicago: American Marketing Association,
206-10.
(I98I). "Measutt;ment and Evaluation of Satisfaction
Processes in Retail Settings," Journal of Retailing, 57
(Fall). 25-47.
Olson. Jerry C. and Philip A. Dover (1979), "Disconfirmation
of Consumer Expectations Through Product Trial," Journal
of Applied Psychology, 64, 179-89.
Peter, J. Paul (1981), "Construct Validity: A Review of Basic
Is.sucs and Marketing Practices," Journal of Marketing Re-
.search. 18 (May). 133-45.
Richins, Marsha L. (1980), "Consumer Perceptions of Costs
and Benefits Associated With Complaining," in Refining
Concepts and Measures of Consumer Satisfaction and Com-
plaining Behavior. H. Keith Hunt and Ralph L. Day, eds.
Bloomington: Indiana University. 50- 3.
(1982), "An Investigation of Consumers' Attitudes
Toward Complaining," in Advances in Consumer Research.
Vol. 9, Andrew Mitchell, ed. St. Louis: Association for
Consumer Research (forthcoming).
Russo, J. Edward (1979), "Consumer Satisfaction/Dissatis-
faction: An Outsider's View," in Advances in Consumer
Research, Vol. 6. W. L. Wilkie, ed. Miami: Association
for Consumer Research, 453-5.
Sharma, Subhash, Richard M. Durand, and Odcd Gur-Arie
(1981), "Identification and Analysis of Moderator Vari-
ables," Journal of Marketing Research. 18 (August). 291-
300.
Shuptrine. F, Kelly and Gerhard Wenglorz (1981), "Compre-
hensive Identification of Consumers' Marketplace Problems
and What They Do Abtiut Them,'" in Advances in Con-
sumer Re.search, Vol. 8. Kent Monroe, ed. Washington:
Association for Consumer Research, 687-92.
"Survey of Automobile Repair Services" (1979), Office of
the Secretary of Transportation, Division of Consumer Af-
fairs, Washington, DC.
Warland, Rex H. , Robert O. Herrmann, and Jane Willits
(1975), "Dissatisfied Consumers: Who Gets Upset and Who
Takes Action," Journal of Consumer Affairs. 9 (Winter).
148-63.
Webbink. Douglas W. (1978). "Automotive Repair: Does
Regulation or Consumer Information Matter," Journal of
Consumer Research, 5 (December). 206-9.
Westbrook. Robert A. (1980), "Intrapersonal Affective Influ-
ences Upon Consumer Satisfaction with Products. " Journal
of Consumer Research. 1 (June), 49-54.

Das könnte Ihnen auch gefallen