Models of Consumer Satisfaction Formation

Page created by Francisco Chavez
 
CONTINUE READING
Models of Consumer Satisfaction Formation
DAVID K. TSE and PETER C. WILTON*

                                              "File authors extend consumer satisfaction literature by theoretically and empiri-
                                           cally (1) examining the effect of perceived performance using a model first proposed
                                           by Churchill and Surprenant, (2) investigating how alternative conceptualizations of
                                           comparison standards and disconfirmation capture the satisfaction formation pro-
                                           cess, and (3) exploring passible multiple comparison processes in satisfaction for-
                                           mation. Results of a laboratory experiment suggest that perceived performance ex-
                                           erts direct significant influence an satisfaction in addition to those influences from
                                           expected perfarmance and subjective disconfirmatian. Expectation and subjective
                                           disconfirmation seem ta be the best conceptualizations in capturing satisfaction for-
                                           mation. The results suggest multiple comparison processes in satisfaction formation.

       Models of Consumer Satisfaction Formation:
         An Extension

   Recent attempts to understand and model consumer                         ambiguities. First, though Churchill and Surprenant (1982)
satisfaction formation have produced several important                      found perceived performance to be a determinant of C S /
findings. First, it is generally agreed that postconsump-                    D, most CS/D models have not incorporated a direct
tion consumer satisfaction/dissatisfaction (CS/D) can be                    link from this construct to CS/D. Second, researchers
defined as the consumer's response to the evaluation of                     have not converged on the exact conceptualization of the
the perceived discrepancy between prior expectations (or                    comparison standard and disconfirmation constructs. For
some other norm of performance) and the actual perfor-                      example, the comparison standard has been conceptual-
mance of the product as perceived after its consumption                     ized as expected (e.g., Oliver 1980), ideal (e.g., Sirgy
(Day 1984). Second, researchers have suggested that C S /                    1984), or normative performance (e.g.. Woodruff, Ca-
D is influenced by a pre-experience comparison standard                     dotte, and Jenkins 1983). Similarly, disconfirmation has
(Bearden and Teel 1983; Cardozo 1965; Day 1977;                             been modeled as the result of subtractive functions (e.g.,
Liechty and Churchill 1979; Miller 1977; Oliver 1977,                       LaTour and Peat 1979) between product performance and
1980; Woodruff. Cadotte, and Jenkins 1983) and dis-                         some comparison standards or as the subjective evalu-
confirmation, that is, the extent to which this pre-ex-                     ation (e.g., Oliver 1980) of this discrepancy. Finally,
perience comparison standard is disconfirmed (e.g., An-                     some researchers have suggested consumers may engage
derson 1973; Bearden and Teel 1983; Day 1977; Howard                        in multiple comparisons in CS/D formation (e.g., Oliver
and Sheth 1969; LaTour and Peat 1979; Maddox 1981;                          1985), but this proposition has not been assessed em-
Oliver 1977, 1980; Olshavsky and Miller 1972; Swan                          pirically.
and Combs 1976; Swan and Trawick 1981; Wilton and                              The purpose of our note is to (1) examine theoretically
Tse 1983; Woodruff, Cadotte, and Jenkins 1983).                             and empirically the role of perceived performance using
   Within this consensus, however, there are three major                    Churchill and Surprenant's model (1982), (2) compare
                                                                            the effects of altemative disconfirmation and comparison
                                                                            standard conceptualizations, and (3) investigate the pos-
   "David K. Tse is Assistant Professor, Marketing Division, Uni-           sibility of multiple comparison standards in satisfaction
versity of British Columbia. Peter C. Wilton is Chief Operating Of-         formation.
ficer, MP (Myer Pacific) Ventures Inc.
   The study was partially supported by a grant from the Humanities
and Social Science Council, University of British Columbia. The au-         Perceived Performance in Satisfaction Formation
thors thank the anonymous JMR reviewers, and as well as Gerald
Gom and Charles Weinberg, for their comments.                                 Theoretical support can be found for including per-
                                                                            ceived perfonnance as a direct determinant of CS/D. In

                                                                      204

                                                                                                          Journal of Marketing Research
                                                                                                          Vol. XXV (May 1988), 204-12
Models of Consumer Satisfaction Formation
CONSUMER SATISFACTION FORMATION                                                                                        205

particular. LaTour and Peat (1979) argue that under cer-       models of consumer preference and choice (see, e.g.,
tain conditions the disconfirmation construct alone may        Holbrook 1984), represents the optimal product perfor-
fail to explain CS/D formation adequately. For exam-           mance a consumer ideally would hope for. It reflects what
ple, consumers forced to buy an inferior brand (say, if        performance "can be." It may be based on previous
their preferred brand is out of stock) may not necessarily     product experiences, learning from advertisements, and
experience disconfirmation of a pre-experience compar-         word-of-mouth communication (Liechty and Churchill
ison standard, but may nonetheless be dissatisfied be-         1979; Miller 1977).
cause of its inferior perfonnance. Triers of new brands           Expected product performance, deriving from expec-
who experience unfavorable disconfirmation of a high           tancy theory (Tolman 1932), represents a product's most
pre-experience standard {generated, say, through adver-        likely perfonnance. It is the most commonly used pre-
tising) may still be satisfied with the brand if it has more   consumption comparison standard in CS/D research. The
of the desired attributes than do competing brands.            construct refiects what performance "will (probably) be"
    Additional support is provided by studies in cognitive     (Liechty and Churchill 1979; Miller 1977). It is affected
dissonance (e.g., Festinger 1957; Holloway 1967), which        by the average product performance (Miller 1977) and
 suggest that the dissonance reduction strategy adopted        advertising effects (Olson and Dover 1979).
 by an individual after a disconfirming consumption ex-            Altemative approaches to modeling the comparison of
 perience will depend on the psychological costs of al-        product performance against a pre-experience standard
 temative reduction strategies. After a very bad (or good)     have been discussed by Swan and Trawick (1981). The
 product experience, the psychological costs of adjusting      subtractive disconfirmation approach (e.g., LaTour and
 the product performance cognition in line with a pre-         Peat 1979), deriving from comparison level theory (Thi-
 experience anchor may exceed the costs of not adjusting       baut and Kelley 1959), assumes that the effects of a post-
 the perfonnance cognition but modifying the pre-expe-         exfwrience comparison on satisfaction can be expressed
 rience anchor. In this case, product perfonnance per-         as a function of the algebraic difference between product
 ception will dominate in the postconsumption evalua-          performance and a comparison standard. This approach
 tions and hence the construct is important in CS/D            has considerable support from studies in industrial and
 formation.                                                    cognitive psychology (e.g., Anderson 1981; Locke 1977),
     A consumer's consumption motive also suggests a di-       where simple algebraic mies of psychological variables
 rect link between perceived pertbrmance and CS/D. A           have been found to represent human infomiation pro-
 study by Cohen and Houston (1972) on the cognitive            cesses adequately over a wide variety of situations.
 consequences of brand loyalty indicates that dissonance           As an altemative approach, subjective disconfirmation
 reduction is only one possible postconsumption process.        (e.g., Churchill and Surprenant 1982; Oliver 1980) rep-
 \f learning from experience is an important consumption        resents a distinct psychological construct encompassing
 motive (especially with new products), then whenever a         a subjective evaluation of the difference between product
 product performs well a consumer is likely to be satis-        perfonnance and the comparison standard; that is, sub-
 fied, regardless of the levels of the pre-experience com-     jective disconfirmation encompasses the set of psycho-
 parison standard and disconfirmation. Clearly, to capture      logical processes that may mediate perceived product
 a diversity of consumption experiences, a comprehen-           performance discrepancies. Such processes are likely to
  sive CS/D model should incorporate perceived perfor-          be important in situations in which product performance
  mance .                                                       cannot be judged discretely.
                                                                   An important distinction between the two approaches
                                                                is drawn by Oliver (1980), who suggests that subtractive
Alternative Comparison Standards and                            disconfirmation may lead to an immediate satisfaction
Disconfirmation Models                                         judgment, whereas subjective disconfirmation represents
                                                                an intervening "distinct cognitive state resulting from the
   Three approaches to conceptualizing a pre-experience         comparison process and preceding a satisfaction judg-
comparison standard have been suggested in CS/D lit-            ment" (p. 460). Hence, subjective disconfirmation is likely
erature. Equitable perfonnance, borrowing from equity           to offer a richer explanation of the complex processes
theory (Adams 1963), represents a normative standard            underlying CS/D formation. Further, if CS/D is mod-
for performance based on implicit relationships between         eled with direct effects from perceived perfomiance, a
the individual's costs/investments and anticipated re-          comparison standard, and disconfirmation simulta-
wards. It represents the level of performance the con-          neously, specifying disconfirmation as a subtractive
sumer ought to receive, or deserves, given a perceived          function of the remaining two independent variables will
set of costs (Liechty and Churchill 1979; Miller 1977;          induce overspecification of the CS/D model. Subjective
Wcwdruff. Cadotte, and Jenkins 1983). The construct is          disconfirmation. as a distinct evaluative construct, is free
likely to be affected by the price paid, effort invested        of such confounding. In subsequent analysis subtractive
(Jacoby 1976), and previous product experiences                 disconfirmation is not used to predict CS/D judgment
(Woodruff, Cadotte, and Jenkins 1983).                          together with comparison standard and perceived per-
                                                                formance.
   Ideal product performance, deriving from the ideal point
Models of Consumer Satisfaction Formation
206                                                                    JOURNAL OF MARKETING RESEARCH, MAY 1988

Multiple Comparison Processes in CS/D Formation                on a 5-point bipolar scale ranging from "very poor" to
  Though most empirical operationalizations of CS/D            "very good"), equitable product performance (perfor-
have focused on a single, unique comparison process,           mance that should be "reasonably expected if you in-
several researchers have suggested the possibility of          vested $50 of your own money in purchasing the product
multiple comparison processes in CS/D formation. Sirgy          . . ." on a 5point bipolar scale ranging from "very poor"
(1984), for example, has identified at least four concep-      to "very good"), product attitude (6-point bipolar "over-
tually distinct comparisons that may underlie CS/D for-        all opinion" scale ranging from "strongly dislike" to
mation. More recently, Oliver (1985), Wilton and Ni-           "strongly like"), and finally purchase intent. For the first
cosia (1986), Forbes, Tse, and Taylor (1986), and Tse          three constructs, measures were taken for both the over-
and Wilson (1986) have conceptualized CS/D as a post-          all product and for the 26 individual attribute levels.
choice process involving complex, simultaneous inter-             Subjects next were exposed to the expectation manip-
actions that may involve more than one comparison stan-        ulation and were asked to evaluate expected product per-
dard. In light of these propositions, a process of multiple   formance at both the overall product and attribute levels
comparisons (which might occur either simultaneously           (performance that " . . . you would expect to see, con-
or sequentially) is likely to have reasonable empirical        sidering everything you have seen or read so far") on 5-
support.                                                       point, bipolar scales ranging from "very poor" to "very
                                                               good" product attitude and purchase intent. Support for
                                                               these various operationalizations is given by Liechty and
                 RESEARCH DESIGN                               Churchill (1979), Miiler (1977), and Oliver (1980).
   The preceding propositions were tested by a 2-expec-           Then subjects received the good or poor perfonnance
tation by 2-product factorial experiment that explicitly       player (depending on the treatment to which they were
allowed for measurement and estimation of altemative           randomly assigned) plus a set of discs and written in-
model specifications within the basic expectancy-discon-       structions on how to operate the player. They were told
firmation paradigm. Subjects were instructed to act as         to use it for as long as they wished. Subjects subse-
potential consumers in a test market trial of a new elec-      quently evaluated perceived performance (". . . your
tronic, hand-held, miniature record player. The design         objective evaluation of the [product's] performance . . .
required subjects to report their psychological states at      regardless of your level of satisfaction with litl . . . "
each of three stages during the experiment: before and         on a 5-point bipolar scale ranging from "very poor" to
after exposure to an expectation manipulation and after        "very good" performance), subjective di.sconfirmation
product perfonnance manipulation.                              (". . . overall . . . how close did the 1 product! come
                                                              to your expectations . . .," on a 5-point bipolar scale
Treatments                                                     ranging from "very much poorer than expected" to "very
   Expectation of product performance was manipulated          much better than expected"), satisfaction (". . . con-
by giving subjects a written evaluation of the product by      sidering everything, how satisfied are you with the
an independent consumer testing laboratory. Half of the        [product]?" on a 5-point biptilar scale ranging from "very
subjects (chosen at random) were gi\en favorable eval-        dissatisfied" to "very satisfied"), future expectation,
uations of the overall product and on 14 specific product     product attitude, and purchase intent. Support for the
attributes (e.g., ease of operation, sound clarity, etc.).    operationalizations of perceived performance and satis-
The remaining subjects received unfavorable evaluations        faction is found in articles by Churchill and Surprenant
of the overall product and its attributes.                    (1982) and Oliver (1980). There were 15 subjects in each
   Product performance was manipulated by two ver-            of the two favorable expectation conditions and 16 sub-
sions of the product: one an unmtxlified, advanced model      jects in each of the two unfavorable expectation condi-
(representing good performance), the other an earlier         tions for a total of 62 subjects.
model further modified by the experimenters to guar-
antee poor product performance (representing poor per-
fonnance). Subjects were assigned randomly to the treat-                              RESULTS
ment conditions.
                                                              Manipulation Checks
Procedure and Measurement                                        Manipulation checks of the expectation and product
   Subjects (student volunteers from advanced marketing       treatments suggest they were highly effective. Subjects
management courses at a westem university) entered the        in the favorable expectation condition expressed signif-
laboratory and were directed to individual soundproof         icantly higher expected performance (/» < .001), more
cubicles. After reading a one-page description of the         positive attitudes {p < .001), and stronger purchase in-
product and the assigned task, subjects provided re-          tentions (p < .05) than subjects in the unfavorable ex-
spronses on the pre-expectation manipulation constructs:      pectation condition. In addition, subjects in the good
ideal product performance (anticipated performance of         performance condition perceived the product to perform
a player that "has exactly the combination of attributes      better (p < .001) and were more satisfied (p < .(X)l)
you would like to see in a miniature record player . . . "    than those in the poor peri'ormance condition.       , _
Models of Consumer Satisfaction Formation
CONSUMER SATISFACTION FORMATION                                                                                                      207

   The discriminant and convergent validity of altema-                      models, and (3) multiple comparison standards in C S /
tive comparison standard operationalizations was as-                        D formation.
sessed by examining the correlation between a compar-                       Effect of Perceived Performance in Satisfaction
ison standard's sum of attribute ratings and its overall
product rating, according to the criteria of the multitrait,                Formation
multimethod procedure (Campbell and Fiske 1959). The                           Our preceding discussion suggests that perceived per-
correlation matrix is at the top left comer of Table 1,                     formance is important to CS/D fomiation. Table 2 com-
which also gives the pairwise product-moment con^ela-                       pares Churchill and Surprenant's (1982) CS/D model,
tion coefficients for other constructs measured in the study.               which incorporates perceived perfonnance in addition to
The intrastandard correlation coefficients (along the di-                   expectation and subjective disconfirmation, with other
agonal) are all statistically significant and higher than the               CS/D models. This model (r^ = .730) outperforms all
corresponding interstandard correlations, suggesting the                    other models, with the improvement in r^ over the best
measures have good discriminant and convergent valid-                       two-variable model (e.g., Oliver 1980 model with r^ =
ity.                                                                        .597) significant at p < .01.
   Discriminant validity checks for the remaining con-                         However, this model raises possible interpretive prob-
stmcts were obtained by comparing their frequency dis-                      lems because of potentially high collinearity in our study,
tributions. Though not reptorted here in detail, the results                as perceived peri'ormance is highly correlated with sub-
of a series of Kolmogorov-Smimov one-sample and Wil-                        jective disconfirmation (r = .73, last row in Table I).
coxon matched pairs, ranked signs tests (Siegel 1956)                       Collinearity among independent variables can produce
indicated systematic differences for 25 of 30 possible                      infiated standard errors, negatively conrelated beta esti-
distribution pairs. Hence, our operationalizations app>ear                  mates, and smaller estimated coefficients in comparison
to be valid.                                                                with models using the independent variables separately.
   Let us now examine the effects of (1) perceived per-                     In our study, though the standard en-or for perceived per-
formance using Churchill and Surprenant's model, (2)                        formance in the proposed three-variable model increases
altemative comparison standards and disconfinnation                         marginally over the errors observed for either a single-

                                                                    Table 1                              -
                                                         CONSTRUCT ASSOCIATIONS
                                                         (product moment correlations)

                                                                                Modet construct
                                                                                             Disconfirmation
                                       Comparison standard'                                  Perceived Perceived     Perceived

                               Expectation             Ideat       Equity      Subjective      minus
                                                                                              expected       minus
                                                                                                             ideal    minus
                                                                                                                     equitabte  Perceived
                                                                                                                               performance
                                                                            disconfinnation
    Modet construct            (I)      (2)      (I)        (2>     (I)           (1)           (I)           (I)       (1)        (0
Comparison standard'
  Expecution         (1)
                        (2)    .72**
  Ideal                 (1)    .04     -.04
                        (2) -.17       -.13     .42**                       i
  Equity                (1)    .13       .19    .18        -.02
                        (2) -.09       -.01     .26**        .20    .57**
Dikconfirmation
  Subjective
    disconflrmation     (1)    .11             -.07                 .07
  Perceived-expeclcd
    performance         (1)   -.53**           -.23*               -.10          .56**
  Perceived-ideal
    performance         (1)    .18             -.64**              -.08          .61**         .64**
  Perceived-equitable
    performance         (1)    .13             -.29**              -.57**        .56**         .62**         .78**
  Perceived product
     performance        (1)    .25*            -.23*             -.00            .73**         .68**         .89**     .82**
  Satisfaction          (1)    .39**           -.24*               .08           .71**         .41**         .75**     .62**      .81**
  '(I) Overall producl-level rating. (2) Average of 26 individual attribute ratings.
  *p < .05.
  **p < .01.
Models of Consumer Satisfaction Formation
208                                                                                     JOURNAL OF MARKETING RESEARCH, MAY 1988

                                                                    Table 2
                          ALTERNATIVE MODELS OF THE DIRECT DETERMINANTS OF SATISFACTION LEVEL

                                                                                 Determinant(s)'
              Model                         Suhtractive               Subjective                                      Perceived
           proponent(s)                   disconfirmation          disconfirmation         Expectation               performance            r^
 Cardozo (1965)
 Olshavsky and                                                           .71                       na                     oa               .500
   Miller (1972)                                                        (.095)
 Anderson (1973)

 Oliver (1985)
 Bearden and                                      m                      .70                      .33                     M                .597
   Teel (1983)                                                          (.087)                   (.088)
Churchill and                                                             lU                       na                    .81               .650
  Surprenant (1982)'                                                                                                    (.068)
 LaTour and                                 .04 (.012)                   •Mi'                      na                     te '            .146
   Peat (1979)"                             .02 (.009)                    M                        na                     •a              .077
                                      •     .03 (.008)     „              mt                       na                     sa              .195
Sirgy (1984)'                               .52   (.060)                  m                        na                     •Hi             .557
                                            .31   (.089)                  oa                       na                     ba              .168
                                            .45   (.074)                  na                       na                     u               .381
                                     -      .18   (.026)                  Bi                       na                     Oft             .451
Combined model                                    Q>                     .r?                      .23                    .55              .730
  (Churchill and                                                        (.102)                    (.07)                 (.104)
  Surprenant) 1982
   'Bela coefficients with standard error in parentheses.
   ''na: not included in the proposed model.
   'In the Churchill and Surprenant study, two models were proposed. For video disc players, only perceived performance was significant whereas
for house plant.s. perceived performance, subjective disconfirmation, and expectation were significant. The former is tested here and the latter is
tested in the combined model.
   "The functional form of LaTour and Peat's model tested is;
                                               .^ Importance; x (Perceived Performance, - Comparison Standard,)
                                 Satisfaction — 2j                 '
                                                                       Comparison Standard,
The disconfirmations are estimated by using ideal expectation and equity, respectively.
   'The disconfirmations are estimated by using perceived performance minus ideal, expectation, equity, and the sum of these differences, re-
spectively.

or two-variable model, the signs for all coefficients in                       plete CS/D model separately for each standard, em-
the proposed three-variable model remain positive and                          ploying a subjective disconfirmation measure as proposed
subjective disconfirmation remains highly significant.                         in our theoretical discussion. The results are reported in
Thus, though collinearity is present, it is not sufficiently                   Table 3. Note that equity fails to produce a direct effect
severe to misspecify the CS/D model by omitting per-                           on any of the three dependent variables (model 3). Con-
ceived perfonnance or subjective disconfirmation (Chat-                        sequently, its indirect effects on satisfaction are also in-
terjee and Price 1977).                                                        significant. This result, which suggests eliminating the
   The importance of perceived performance in CS/D                             comparison standard altogether, contradicts previous
formation is evident from Table I. It explains 65% of                          findings in CS/D research. A more reasonable inference
the variation in satisfaction, outperforming any other single                  is that equity is not a good operationalization of the com-
predictor model, including disconfirmation, and the two-                       parison standard.
variable model proposed by Oliver (1980), shown in Ta-                            Results for expectation and ideal are more meaningful
ble 2. In addition, perceived performance exerts a sig-                        (the insignificant chi square values for both models 1 and
nificant indirect effect on satisfaction through its influ-                    2 suggest that neither model can be rejected as a good
ence on subjective disconfirmation (beta is .74 at p <                         representation of the data), though their effects are very
.001, Table 3).                                                                dissimilar. For expectation, the effect on satisfaction is
                                                                               direct and positive; for ideal, it is indirect (through per-
Alternative Comparison Standards and
                                                                               ceived performance) and negative. This result is intu-
Disconfirmation Models                                                         itively appealing. If one is prepared to accept both as-
  To determine both direct and indirect effects of alter-                      similation/contrast explanations in CS/D formation, the
native comparison standards, we path analyzed the com-                         results suggest that ideal as an anchor may tend to evoke
Models of Consumer Satisfaction Formation
CONSUMER SATISFACTION f=ORMATION                                                                                                               209

a contrast effect on the evaluation of the experience,                         tive disconfirmation to predict CS/D is .56 (r - .75,
whereas expectation may evoke an assimilation effect.                          last row of Table 1), in comparison with .73 for sub-
This result suggests that the appearance of assimilation                       jective disconfirmation with perceived performance and
and/or contrast effects may depend on the specific in-                         expectation. Hence, in a comparison of the subjective
ternal anchor employed. Note also that all comparison                          and subtractive approaches to model disconfirmation, the
standards show only weak influence on subjective dis-                          former would be superior. The subtractive approach con-
confirmation. In each case, the effect of the standard ap-                     tains an inherent confound when predicting satisfaction
pears to be dominated by perceived performance. This                           and does not capture all the determinants of CS/D for-
issue is discussed shortly.                                                    mation.
   The last row in Table I reports the perfomiance of
                                                                               Multiple Comparison Standards in Satisfaction
alternative disconfirmation models in predicting CS/D.
As expected, subjective disconfinnation performs well.                         Formation
The effect of this variable on CS/D is significant and                            The last row in Table 1 suggests more than one com-
positive, though not as large as the effect of perceived                       parison standard may be involved in CS/D formation,
performance. Together with pierceived performance and                          because both expectation (r = .39) and ideal (r = — .24)
expectation, subjective disconfirmation achieves an r of                       relate individually to satisfaction. Model 4 of Table 3
.73. In comparison with the fully recursive model with                         assesses possible simultaneous effects of these two com-
all relevant paths included, this model achieves a high                        parison standards in CS/D. The path coefficients ob-
goodness of fit {Q = .940) and an insignificant chi square                     tained support the proposition. Expectation exerts sig-
statistic, suggesting this disconfirmation model fits the                      nificant influence in both subjective disconfirmation and
data well.                                                                     satisfaction, whereas ideal is significant in affecting per-
   In contrast, subtractive disconfirmation does not ap-                       ceived performance in CS/D formation. This model has
pear to yield a better representation of satisfaction than                     a high goodness of fit (Q = .919) and an insignificant
subjective disconfirmation. The r^ for the best subtrac-                       chi square.

                                                                      Table 3
     SUBJECTIVE DISCONFIRAAATION A N D ALTERNATIVE COMPARISON STANDARDS IN SATISFACTION FORA\ATION
                                                                   (path analysis)''

                                                                              Independent variables
                               *                                                                  Perceived          Objective
                                            Comparison                    Subjective               product           product
         Dependent variable                  standard                  disconfirma tion         performance           quality*"          r'
Model I. Standard = expected performance
 Satisfaction                                    .23*                       .27**                  .55**                 na              .73
 Subject disconfirmation                         M                           na                    .74**                 na              .55
  Perceived perfomiance                          ns                          na                     na                 .69**             .47
                                                      Q = 940: X^ 2 d.f. = 3.69, p < .15
Model 2. Standard = ideal performance
 Satisfaction                                    OS                                                .63**                 na              .68
 Subjective disconfirmation                      m                           u                     .74**                 na              .55
 Perceived performance                         -.24*                  na                            na                 .69*"
                                                   Q = 945: X^ 2 d.f. = 3.38. p < .18
Modet 3. Standard = equitable performance
 Satisfaction                                    ns                            .24*                .63**                 na              .68
  Subjective disconfirmation                     OS                              oa                .74**                 na              .55
  Perceived performance                          ns                              na                 na                 .69**             .47
                                                      Q=      • 977:
                                                                     x'. 3 d.f. = L39./> < .70
Model 4. Multiple standards
                                    Expected          Ideal
  Satisfaction                       .23*               ns                  .27**                  .55**                 na              .73
  Subjective disconflrmation           ns               ns                    na                   .74**                 na              .55
  Perceived performance               .18*         -.25**                     na                  na                   .67**             .56
                                                     Q = 919: x^ 3 d.f. = 4.97. p < .18
   'Table shows the estimated path coefficients for the reduced model only. Coefficients estimated but proving statistically insignificant in the
fully recursive model are omitted and denoted ns; na indicates not applicable.
   "A dichotomous variable representing the product treatment.
    "p < .05.
    •*p < .01.
Models of Consumer Satisfaction Formation
210                                                                              JOURNAL OF MARKETING RESEARCH, /AAY 1988

   The validity of this paradigm requires that the two                   out apparently would continue to be supported if method-
standards be independent. The validity checks described                  adjusted correlations were used.
before strongly support their independence. In addition,                    Another issue is the strength of the product manipu-
the reliability coefficient when expectation and ideal are               lation. Because perceived performance is central to C S /
pooled is low at .07 (Hotelling T^ = 91.10). Thus, ex-                   D formation, it may dominate all other determinants in
pectation and ideal appear to represent different con-                   predicting CS/D if the product manipulation is too strong,
structs contributing separately to the CS/D formation                    as reported by Churchill and Surprenant (1982) using video
process.                                                                 disc players. In our study, perceived performance is found
   In addition, the single-comparison standard models                    to exert more influence than other model elements in
(Table 3, models 1 and 2) are nested within, and can be                  {HEdicting CS/D. Nevertheless, the other elements (e.g.,
considered reduced forms of, the multiple-comparison                     expectation) do have significant direct influences on C S /
standard model (model 4). Testing the difference be-                     D (models I and 4, Table 3). In future studies research-
tween these two types of models when expectation is the                  ers must be cautious in balancing the strengths of prod-
standard produces a Q index of .748 and a chi square                     uct and expectation manipulations in their designs.
with 2 d.f. of 17.88 (p < .001). When ideal is the stan-
dard, the Q index is .790 and the chi square is 14.38                    Implications
with 2 d.f. ip < .001). Thus, the single-standard model
fails to represent the underlying processes adequately in                   Our study provides strong theoretical and empirical
connparison with a multiple-standard paradigm.                           support for extending the expectation and disconfinna-
                                                                         tion model of CS/D to include direct influences from
                                                                         perceived performance. Several theoretical frameworks
                          DISCUSSION
                                                                         discussed before suggest that perceived performance
Methodological Issues                                                    cognition may outweigh expectation in determining C S /
                                                                          D judgments across a variety of consumption situations.
   As in prior CS/D studies, the distribution of responses               For example, subjects may modify their expectations when
on some variables in our study is skewed and nonhom-                     their expectations are disconfirmed by the product treat-
ogeneous. Whether deviations from normality or non-                      ment. If this happens, their responses may show both
homogeneous variable distributions should disqualify the                 (1) a weaker association between satisfaction and ex-
study's inferences depends on both sample size and the                   pectation and (2) a greater propensity to modify the ex-
size of the correlations assuming equivalent distribu-                   pectation after the product trial than they would if ex-
tions. On the basis of the criteria proposed by Nunnally                 pectations were confirmed. In our study, the association
(1978, p. 142), it appears that correlation coefficients                 between expectation and satisfaction proves to be much
obtained from 62 subjects should be considered only                      lower and the shift in pre-/post-product-trial expecta-
moderately robust.'                                                      tions much larger for subjects in the disconfirmation
   The r^'s reported here are higher than those in other                 condition than for subjects in the confirmation condition
studies (e.g., Bearden and Teel 1983; Oliver 1980), sug-                 (bivariate correlations of —.05 (p > .10) and .78, re-
gesting the possibility of common methods variance in                    spectively; mean shifts on a 5-point bipolar expectation
the measures. Though the discriminant validity results                   scale of 1.0 and .35, respectively). Thus, because ex-
for comparison standards reported before suggest this is                 pectation and product performance appear to assume dis-
unlikely, the issue can be explored further by applying                  tinctly different roles in CS/D formation, the effect of
the multitrait, multimethod procedure to the other mea-                  perceived perfonnance on C S / D should be modeled sep-
sures. By the common method used throughout the study                    arately.
(overall product-level measures on each construct), the
correlations between satisfaction and (1) subjective dis-                   Another important finding is the presence of multiple
confirmation and (2) perceived performance are .71 and                   comparison standards in C S / D formation. Though mul-
.81, respectively. Replacing the overall satisfaction                    tiple comparisons have been suggested conceptually by
measure with an average attribute rating estimate causes                 previous researchers (Sirgy 1984) and supported by a
these correlations to drop to .63 and .71, respectively.                 broad process conceptualization of CS/D (Oliver 1985;
Hence, it appears that, though the common method em-                     Wilton and Nicosia 1986), our study provides the initial
ployed throughout the study slightly elevates the ob-                    empirical evidence. This process conceptualization ex-
tained coefficients, the relationships observed through-                 plicitly allows for simultaneous interactions or temporal
                                                                         changes in key satisfaction determinants. Possibly, for
                                                                         example, the postexperience comparison can be de-
                                                                         scribed best not as a state variable, but as a continuous
   'In this study the ideal measure was skewed positively, with as       process involving different comparison standards as the
expected a small but nonzero variance. The lack of variability in the    effects of product usage decay. Though testing of such
ideal can explain the high correlation (.89) between perceived per-      possibilities requires longitudinal observations not avail-
formance and perceived perfonnance minus the ideal (i.e., subtraclive
disconfirmation). Further research is needed to investigate other pos-
                                                                         able in our study, further investigation could be expected
sible measures for the ideal that would have more variability.           to prove rewarding.
Models of Consumer Satisfaction Formation
CONSUMER SATISFACTION FORMATION                                                                                            211

                       REFERENCES                                  liam L. Wilkie, ed. Ann Arbor. MI: Association for Con-
                                                                   sumer Research, 431-7.
                                                                 Liechty, M. and Gilbert A. Churchill. Jr. (1979). "Conceptual
Adams. J. Stacy (1963). "Towards an Understanding of In-           Insights into Consumer Satisfaction with Services." in Ed-
 equity," Journal of Abnorrrml and Social Psychology, 67           ucators' Conference Proceedings. Series 94. Neil Beckwith
  (October), 422-36.                                               et al., eds. Chicago: American Marketing Association, 509-
Anderson. Nomian H. (1981). Information Integration The-            15.
  ory: A Case History in Experimental Science. Vol. I. New       Locke. E. A. (1977). "The Nature and Causes of Job Satis-
  York: Academic Press. Inc.
                                                                   faction." in Handbook of Industrial and Organizational Psy-
Anderson. Ralph E. (1973), "Consumer Dissatisfaction: The          chology. Marvin D. Dunnette, ed. Chicago: Rand McNally
  Effect of Disconfirmed Expectancy on Perceived Product           College Publishing Company, 1297-350.
  Performance," Journal of Marketing Research. 10 (Febru-
  ary). 38-44.                                                   Maddox. R. Neil (1981), "Two-Factor Theory and Consumer
                                                                   Satisfaction: Replication and Extension," Journal of Con-
Bearden. William D. and Jesse E. Teel (1983). "Selected
                                                                   sumer Research. 8 (June). 97-102.
  Determinants of Consumer Satisfaction and Complaint
                                                                 Miller, John A. (1977). "Studying Satisfaction, Modifying
  Reports," Journal of Marketing Research. 20 (November),
  21-8.                                                            Models, Eliciting Expectations, Posing Problems and Mak-
                                                                   ing Meaningful Measurements." in Conceptualization and
Campbell. Donald T. and Donald W. Fiske (1959), "Conver-
                                                                   Measurement of Consumer Satisfaction and Dissatisfaction.
  gent and Discriminant Validation by the Multitrait-Multi-
                                                                   H. Keith Hunt, ed. Cambridge. MA: Marketing Science In-
  method Matrix." Psychological Bulletin. 56 (March), 8 1 -
   105.                                                            stitute. 72-91.
                                                                 Morrison. D. F. (1976), Multivariate Statistical Methods, 2nd
Cardozo, Richard N. (1965). "An Experimental Study of Con-
  sumer Effort. Expectations and Satisfaction," Journal of         ed. New York: McGraw-Hill Book Company.
  Marketing Research. 2 (August), 244-9.                         Nunnally, Jum C. (1978). Psychometric Theory. 2nd ed. New
Chatterjee, S. and B. Price (1977), Regression Analysis by          York: McGraw-Hill Book Company.
  Example. New York: John Wiley & Sons. Inc.                     Oliver. Richard L. (1977), "A Theoretical Reinterpretation of
Churchill, Gilbert A.. Jr. and Carol Surprenant (1982). "An         Expectation and Disconflmiation Effects on Post-Exposure
  Investigation into the Determinants of Consumer Satisfac-         Product Evaluations: Experience in the Field." in Consumer
  tion." Journal of Marketing Research. 19 (November). 4 9 1 -      Satisfaction. Dissatisfaction and Complaining Behavior,
   5(M.                                                             Ralph L. Day, ed. BUwmington: Indiana University. 2 - 9 .
Cohen, Joel B. and Michael J. Houston (1972), "Cognitive               — (1980). "A Cognitive Mt>del of the Antecedents and
  Con.sequences of Brand Loyalty." Journal of Marketing Re-         Consequences of Satisfaction Decisions." Journal of Mar-
  search. 9 (February), 97-9.                                       keting Research, 17 (November), 460-9.
Day, Ralph L. (1977). "Towards a Process Model of Con-                     (1985), "An Extended Perspective on Post-Purchase
  sumer Satisfaction," in Conceptualization and Measurement         Phenomena: Is Satisfaction a Red Herring?", unpublished
  of Consumer Satisfaction and Dissatisfaction. H. Keith Hunt,      paper presented at 1985 Annual Conference of the Associ-
  ed. Cambridge, MA: Marketing Science Institute, 153-83.           ation for Consumer Research. Las Vegas (October.)
         (1984). "Modeling Choices Among Altemative Re-          Olshavsky, Richard W. and John A. Miiler (1972), "Con-
  sponses to Dissatisfaction," in Advances in Consumer Re-          sumer Expectation. Product Perfomiance and Perceived
  search. Vol. n , Thomas C. Kinnear, ed. Ann Arbor, MI:            Product Quality," Journal of Marketing Research. 9 (Feb-
   Association for Consumer Research, 469-9.                        ruary), 19-21.
Festinger. Leon A. (1957). A Theory of Cognitive Dissonance.     Olson, Jerry C. and Philip Dover (1979), "Disconfirmation of
   Stanford. CA: Stanford University Press.                         Consumer Expectations Through Product Trial." Journal of
Forbes, J.D.. David K. Tse, and Shirley Taylor (1986), "To-         Applied Psychology. 64 (April), 179-89.
   ward a Model of Consumer Post-Choice Response Behav-          Siegel. S. (1956), Nonparametric Statistics for Behavioral Sci-
   ior," in Advances in Consumer Research. Vol. 13, Richard         ences. New York: McGraw-Hill Book (Tompany.
   L. Lutz. ed. Ann Arbor. MI: Association for Consumer Re-      Sirgy. M. Joseph (1984), "A Social Cognition Model of Con-
   search. 658-61.                                                  sumer Satisfaction/Dissatisfaction: An Experiment," Psy-
                                                                    chology and Marketing, 1 (Summer). 27-44.
Holbrook, Morris B. (1984). "Situation-Specific Ideal Points
   and Usage of Multiple Dissimilar Brands," in Research in      Sobel. Robert A. (1971), "Tests of Performance and Post-Per-
   Marketing. Vol. 7, Jagdish N. Sheth. ed. Greenwich, CT:          formance of Satisfaction With Outcomes," Journal of Per-
   JAl Press. Inc.. 93-112.                                         sonality and Social Psychology. 19 (July). 213-21.
 Holloway, Robert J. (1967), "An Experiment on Consumer          Swan, John E. and Linda J. Combs (1976), "Product Perfor-
   Dissonance." Journal of Marketing. 31 (January), 39-43.          mance and Consumer Satisfaction: A New Concept," Jour-
 Howard. John A. and Jagdish N. Sheth (1969), The Theory of         nal of Marketing, 40 (June). 25-33.
   Buyer Behavior. New York: Wiley Marketing Series.                       and I. Frederick Trawick (1981). "Disconfirmation of
Jacoby. Jacob (1976). "Consumer Psychology and Industrial           Expectations and Satisfaction with a Retail Service," Jour-
   Psychology: Prospects for Theory Corporation and Concep-         nal of Retailing. 57 (Fall), 49-67.
   tual Contribution," in Handbook of Industrial and Organi-     Tolman, E. C. (1932), Purposive Behavior in Animals and
   zational Psychology. Marvin D. Dunnette. ed. Chicago: Rand       Men. New York: Appleton-Century.
   McNally College Publishing Company. 1031-62.                  Thibaut, J. W. and Harold H. Kelley (1959). The Social Psy-
 LaTour. Stephen A. and Nancy C. Peat (1979). "Conceptual           chology of Groups. New York: John Wiley & Sons, Inc.
   and Methodological Issues in Consumer Satisfaction Re-        Tse. David K. and Peter C. Wilton (1985), "History and Fu-
   search," in Advances in Consumer Research, Vol. 6, Wil-          ture of Consumer Satisfaction Research." in Historic Per-
Models of Consumer Satisfaction Formation
212                                                                      JOURNAL OF MARKETING RESEARCH, MAY 1988

 spective in Consumer Research: National and International        sponse to Communication and Product Experiences," in Ad-
 Perspectives (Proceedings of the Association for Consumer        vertising and Consumer Psychology. Arch Woodside and
 Research. National University of Singapore). Jagdish N. Sheth    Larry Percy, eds. Lexington, MA: Lexington Books, 3 1 5 -
 and Chin-Tiang Tan, eds. 251-6.                                  32.
Wilton. Peter C. and M. Nicosia (1986), "Emerging Para-          Woodruff, Robert B., Emest R. Cadotte, and Roger L. Jen-
 digms for the Study of Consumer Satisfaction," European          kins (1983), "Modeling Consumer Satisfaction Processes
 Research. 14 (January), 4 - 1 1 .                                Using Experience-Based Norms," Journal of Marketing Re-
       and David K. Tse (1983), "A Model of Consumer Re-          search, 20 (August), 296-304.

                  -^.«^.
                                                                                            Call toll free
                                                                                         1-800-828-6107

             ^^;f.°;;:ie so
You can also read