Cognitive Dissonance - Reconsidering An Important And Well - Established Phenomenon In Consumer Behaviour Research

Cognitive Dissonance – Reconsidering An Important
And Well-Established Phenomenon In Consumer Behaviour Research

Thomas Salzberger, Monika Koller, University of Economics and
Business Administration Vienna, Austria

Abstract

Although cognitive dissonance is a well-established construct in consumer behaviour
research, applications are relatively scarce in current marketing research projects. The reasons
are diverse. First, dissonance is often seen as merely a transitory phenomenon. Second,
problems of measurement as well as difficulties in administering data collection often get in
the way of empirically addressing cognitive dissonance. The current study builds upon recent
contributions to the measurement of cognitive dissonance. By applying an innovative
measurement model, it is demonstrated that obstacles of measuring dissonance can be
overcome. In a longitudinal setting, data are collected at two points in time allowing for
insight into the developing of dissonance over time. Dissonance turns out to be an important
factor in the formation of satisfaction. About 10% of respondents are prone to develop a
substantial level of dissonance – dissonance which does not self-dissolve but requires specific
marketing activities.

Keywords: Cognitive dissonance, satisfaction, measurement, Rasch model

Theoretical Background
The phenomenon of cognitive dissonance, originally stated by Festinger in 1957, has been
quickly adopted by consumer behaviour research. Described as a psychologically
uncomfortable state that arises from the existence of contradictory (dissonant, non-fitting)
relations among cognitive elements (Festinger 1957) cognitive dissonance revealed high
exploratory power in explaining the state of discomfort buyers are often in after they made a
purchase. While research experienced its peak in the mid-1970s (Sweeney et al., 1996),
cognitive dissonance ceased to be a frequent topic on the research agenda thereafter.
Customer satisfaction, complaint behaviour, and loyalty superseded cognitive dissonance as
constructs of key interest in consumer research. The reasons do not lie so much in a lack of
interest but are rooted in considerable difficulties in the measurement of cognitive dissonance
and, in particular, in its timing. The current study wants to contribute to overcoming both
problems and to assess the relationship of cognitive dissonance on the one hand and
satisfaction and loyalty on the other.

According to Montgomery and Barnes (1993, p. 204) “[v]ery little research has been
conducted with post purchase dissonance” since the late 1970s. Oliver (1997) lamented that
the reduction in interest was inexplicable and unfortunate. In 2000, Sweeney et al. (p. 369)
stated that “there is no well established scale to measure [cognitive dissonance]”. In
reassessing previous work, Sweeney et al. (1996) concluded that the measurement of
cognitive dissonance suffers from several substantial problems. For example, many
measurement approaches refer to consequences of dissonance (e.g., activities aiming at a
reduction of dissonance) rather than to the level of dissonance itself. Korgaonkar and Moschis
(1982) operationalise dissonance by four dichotomous items which are solely based on
behaviour suited for reducing cognitve dissonance. Others invoke cognitive dissonance theory
but do not operationalise dissonance itself. Clow et al. (1998), e.g., refer to cognitive
dissonance as an explanatory construct for shifting expectations but do not attempt to measure
cognitive dissonance. Another critical issue is the timing of measurement. Since cognitive
dissonance is assumed to be a transitory phenomenon that is not stable over time, a
ANZMAC 2005 Conference: Consumer Behaviour
290

---

retrospective approach of measuring dissonance, when the product has already been in use for
an extended period of time, is very unlikely to reflect accurate levels of post-purchase
dissonance. From a theoretical point of view, it is interesting to assess the impact of post-
purchase dissonance on satisfaction and, ultimately, on loyalty. The retrospective approach
yields distorted insight because dissonance will be adjusted and, in most cases, reduced over
time. Interpersonal variation will thereby be levelled-out.

Constructs and Scale development

Cognitive dissonance, satisfaction and loyalty represent the pivotal constructs in the study.
The scale for measuring cognitive dissonance is based on work by Sweeney et al. (2000) and
Soutar and Sweeney (2003) who developed a multi-dimensional instrument comprising an
emotional component, a dimension termed concern over the deal and a third one named
wisdom of purchase. Another source of the item pool is previous work by Montgomery and
Barnes (1993). Table 1 provides a technical overview of the central constructs.

Satisfaction is closely related to dissonance. Sweeney et al. (1996) demonstrated empirically
that dissonance and satisfaction are distinct constructs with cognitive dissonance being an
important antecedent to satisfaction (see, e.g., Sweeney et al., 1996; Oliver, 1997; Hausknecht
et al., 1998; Soutar and Sweeney, 2003). In our study, satisfaction with the product is
measured on a general (overall) level without reference to specific product features. The
indicators have been based largely on Westbrook and Oliver (1991) supplemented by newly
generated items. An attribute-oriented approach is not feasible in the context of this study
because of the diversity of the products under consideration (ranging from dvd players to
vacuum cleaners).

Finally, loyalty as a consequence of satisfaction (see, e.g., Kasper, 1988; Anderson and
Sullivan, 1993; Mooradian and Oliver, 1997) represents a central construct of interest in the
study. Loyalty has been conceptualised as referring first, to the shop where the product has
been bought, and, second, to the brand of the product. Moreover, future purchase intentions
and recommendation (word-of-mouth) have been separated. Thus, four distinct constructs of
loyalty have been considered and evaluated separately.

Table 1: Central constructs of the study
# of items
Construct Time Initiallya After
Examples (partly abbreviated)
purificationb
Cognitive dissonance
t
I wonder if I have made the right choice. / I
0 28
19
(9+10)c
think I had rather spent the money on
18
13
t
something else. / I ask myself whether I
1
(14+4)d
(9+4)e
really should have bought the product.
Satisfaction
Possessing the product is a good thing. / The
t1 20 14 product is the best I could have bought. / The
product has all the features I expected.
Loyalty



- word of mouth shop
9
7
… recommend shop to friends …
- word of mouth brand
t
… recommend brand to friends …
1
7
4
- shop
6
5
... will buy other consumer durables …
- brand (cross-selling)
4
4
… will buy other products of same brand
a Items administered in the questionnaires at t0 (time of purchase) and t1 (three months after purchase)
b Items in the final scale after common scale evaluation of data collected at t0 and at t1
c 9 items are identical at t0 and t1, 10 items are unique to t0
d 14 items from t0, 4 new items unique to t1
e 9 items are identical at t0 and t1, 4 items are unique to t1

ANZMAC 2005 Conference: Consumer Behaviour

291

---

Hypotheses

First, dissonance at t0 (i.e., at the time of purchase) is assumed to be inversely related to
satisfaction at t1 (i.e., three months after purchase). Second, the same relationship should hold
for dissonance at t0 and loyalty at t1. Third, dissonance at t1 and satisfaction at t1 should
correlate inversely. Fourth, satisfaction and loyalty should be positively correlated. Finally,
dissonance at t1 is supposed to be lower than dissonance at t0, i.e. a decline of dissonance of
time is assumed to occur. In terms of causality, the first two hypotheses assume a causal
relationship, which is highly plausible because of the temporal course of measurement. The
third and the fourth hypothesis are confined to t1. In this case an alternate impact on the
constructs under scrutiny is very likely. Consequently, these relationships are interpreted in a
non-causal way.

Empirical Design
The study was undertaken in two outlets of consumer electronics and household goods.
Customers buying a good worth at least 80 A$ qualified for being interviewed immediately
after the purchase (n=250). At this point in time (t0), cognitive dissonance was the main
construct of interest. Three months later (at t1), those respondents who identified themselves
were contacted again (n=125). This period of time was chosen to allow for consumers having
sufficient time to experience product usage. The second questionnaire was administered by
mail focusing on satisfaction and loyalty. However, cognitive dissonance was measured again
in order to investigate the developing of dissonance over time. If dissonance were a transitory
phenomenon only relevant shortly after the time of purchase, it should be dissolved three
months later. Respondents who did not use the product for themselves were excluded from the
analysis leading to a sample of n=110.

Methodology

The measurement models were investigated based on item response theory (see Singh 2004
and Ewing et al. 2005 for introductions in the marketing context). Specifically, the
polytomous Rasch model (Rasch 1980, Andrich 1988) was applied using RUMM 2020
(Andrich et al. 2003). The model places items and persons onto the same latent interval-scaled
dimension (see Ewing et al. 2005 for details). Compared to standard approaches rooted in
classical test theory, this model requires the data to adhere to a more stringent model making
measurement more trustable. The quality of the scale is primarily assessed by investigating
the data model fit by a chi square test, which compares response patterns expected under the
model to actual patterns. This test can be applied to each individual item as well as to the
whole set of items in the scale. Besides, a statistic called person separation index can be
interpreted very similar to traditional reliability.

Empirical evidence
The measurement models

The evaluation of the items measuring cognitive dissonance was carried out simultaneously
for all items and both points in time of data collection. This allows for investigating the
stability of the items’ functioning over time and ultimately for a direct comparison of
dissonance levels at t0 and at t1. In the final scale, nine items have been administered at both
points in time, 10 items have been presented only at t0 whereas four items have been newly
introduced at t1. The latter refer explicitly to the period of time between the time of purchase
ANZMAC 2005 Conference: Consumer Behaviour

292

---

and the time of the second data collection. The data fit the Rasch model highly satisfactorily
( =84.46, df=69, p=.10) indicating that a unidimensional measure of cognitive dissonance
seems reasonable. The person separation index amounts to .82 suggesting that the items
provide enough information to locate the respondents on the latent dimension of cognitive
dissonance and that the power of the test of fit is adequate.
The measurement of general satisfaction is based on a set of 14 items (out of 20 items in the
questionnaire) fitting the Rasch model quite well ( =38.32, df=26, p=.03) with a person
separation index of .85. The critical level for all analyses has been fixed to 0.01 which is
still very conservative given the large number of tests of fit for individual items.

The operationalisation of four types of loyalty turned out to be successful, as well, with the
number of items in each scale ranging from four to seven (see table 1) and p-vales of the
overall fit statistics between .04 and .46.

Potential Mere-Measurement Effects

Mere-measurement effects have been discussed recently in the case of satisfaction (Dholakia
and Morwitz, 2002) and purchase intentions (Morwitz and Fitzsimons, 2004). Indeed, the
measurement of cognitive dissonance could have an impact on the level of dissonance. Either
could measurement enhance actual dissonance or it might boost dissonance reducing
processes and thereby decrease the level of dissonance. Alternatively, both forces could be
relevant simultaneously leading to larger interpersonal variation by reducing low levels but
enhancing higher levels of dissonance. In the present study, we attempted to get at least some
evidence by introducing two dissonance items at the beginning of the questionnaire where
respondents were not so much aware of the topic of the study. The responses to these items
have been compared to responses to identical questions in the whole set of dissonance items.
However, no significant differences could be found in terms of means and variance.
Consequently, there is no indication of a strong mere-measurement effect in the data.

Descriptive results

Generally, the respondents reported rather low levels of dissonance at the time of purchase (t0)
as well as after the product usage (t1). At t0, only 7.4% of the consumers felt considerable
cognitive dissonance while 76.4% were affected by dissonance very slightly or not at all1.
Interestingly, the sort of product had no impact on the level of dissonance experienced.
Likewise there are no significant differences in the price of the product for different levels of
dissonance. After the phase of product usage, the share of consumers feeling comparatively
strong dissonance increased to 9.1%. At the same time, the proportion of respondents
featuring a very low level of dissonance also rises to 81.8%. Consequently, the variance of
dissonance scores increased while the mean decreased only weakly and insignificantly (see
H5).

Testing the hypotheses

H1 postulated that low cognitive dissonance after the purchase (at t0) leads to a higher level of
satisfaction after the product has been used for several months. The standardised regression
coefficient of the Rasch measures amounts to -.40 (n=110, p<.001) affirming H1.2 In order to
eliminate potential outlier effects, Cook’s distance D (Fox, 1991) has been considered to
identify influential cases. After eliminating those respondents with D > 4/(n - k - 1), where n
is the number of cases and k is the number of independents, the coefficient decreased only
slightly to -.36.
ANZMAC 2005 Conference: Consumer Behaviour

293

---

H2 states that cognitive dissonance ultimately exerts an inverse influence on loyalty.
Cognitive dissonance is significantly but rather weakly related to repurchase intentions (same
shop r=-.22, n=110, p=.01; same brand r=-.18, n=110, p= .03). However, when accounting for
satisfaction as a mediator of the relationship, the partial correlation turns insignificant
suggesting that there is no direct impact of dissonance on repurchase intentions. In terms of
word-of-mouth, only the correlation of dissonance and recommendation of the shop is
significant (r=-.20, n=110, p=.02). However, this correlation is not significantly different from
0 either, once satisfaction has been accounted for. Consequently, H2 has to be rejected. The
impact of dissonance on loyalty is entirely mediated by satisfaction (see H4).

H3 assumes dissonance measured at t1 and satisfaction to be correlated inversely. Indeed, the
correlation is r=-.61 (n=110, p<.001). The correlation decreased after accounting for
influential cases (based on Cook’s distance, see H1) only slightly to r=-.57. Given that
satisfaction is generally rather high, the increase in the strengths of this association from t0 to
t1 indicates that dissonance and satisfaction converge on a high level.

H4 suggests satisfaction and loyalty to be positively correlated. H4 has universally been
confirmed. Satisfaction is related to repurchase intentions (same shop r=.44, p<.001; same
brand r=.38, p<.001) and to word-of-mouth (recommending the shop r=.42, p<.001;
recommending the brand r=.47, p<.001).

Finally, H5 refers to the level of cognitive dissonance over time. Although cognitive
dissonance degrades slightly over time, the difference is not significant (one-tailed paired
sample t-test, t=0.53, df=109, p=.30). Interestingly, the correlation between t0 and t1 is only
.51 (n=110, p<.001) implying that dissonance changes over time but does not degrade
uniformly for all respondents. Rather, dissonance and satisfaction converge (see also the
results for H3).

Discussion
Technically, the measurement models meet stringent requirements of measurement as
specified by the Rasch model. Thus, the study demonstrates that the problems of measuring
cognitive dissonance can be overcome. Although the level of cognitive dissonance is rather
low at both points in time, those consumers who feel dissonance immediately after the
purchase do not fully reduce their dissonance. On the contrary, dissonance seems to
differentiate over time, i.e. a low level decreases during product usage while a higher level
even increases. Since dissonance after the purchase explains more than one third of the
variance of satisfaction, which in turn has a significant impact on loyalty, cognitive
dissonance should not be considered a transitional phenomenon that can be ignored. In the
context of this study, about 10% of the consumers are vulnerable to developing cognitive
dissonance suggesting that there is need for post-purchase marketing activities aiming at
helping consumers reduce dissonance and avoiding the negative impact of dissonance on
satisfaction.

The low level of dissonance at the time of purchase points out that measuring dissonance
immediately after the purchase might possibly be too early. It is not implausible that the
purchase causes preliminary enjoyment while sceptical thoughts might develop during the
days after the purchase. Insofar, it seems that the limitation of recalling dissonance a few days
after purchase, rather at the time of purchase, Soutar and Sweeney (2003, p. 243) concede in
their study, is actually no limitation at all.

ANZMAC 2005 Conference: Consumer Behaviour

294

---

One question asked consumers whether there were moments of regret between the time of
purchase and the second time of data collection (i.e. between t0 and t1). About 23% of the
respondents after all did not fully disagree with this statement and for all of them cognitive
dissonance increases over time. However, 19% remain in the area of moderate dissonance3
suggesting that the proportion of consumers prone to develop dissonance might actually be
even higher than the estimate of about 10%. This limitation of the timing of measurement
may not be overcome easily since contacting consumers a few days after the purchase is
harder to administrate than interviewing them at the time of purchase. However, a promising
way to shed more light on the whole time period starting with the purchase and following
right through the product usage phase would be a diary study. In such a study respondents are
requested to answer statements repeatedly and also provide qualitative data at several points
in time. This would allow for an even more comprehensive insight into the temporal course of
the development of dissonance than the current study with two measurements can provide.

References

Anderson, E., and Sullivan, M., 1993. The antecedents and consequences of customer
satisfaction for firms. Marketing Science. 12 (2), 135-141.

Andrich, D., 1988. A General Form of Rasch´s Extended Logistic Model for Partial Credit
Scoring. Applied Measurement in Education. 1 (4), 363-378.

Andrich, D., Sheridan, B.S., and Luo, G., 2003. Rumm2020: Rasch Unidimensional
Measurement Models. RUMM Laboratory Perth, Western Australia.

Clow, K.E., Kurtz, D.L., and Ozment, J., 1998. A Longitudinal Study of the Stability of
Consumer Expectations of Services. Journal of Business Research. 42, 63-73.

Dholakia, U.M., and Morwitz, V.G., 2002.The Scope and Persistence of Mere-Measurement
Effects: Evidence from a Field Study of Customer Satisfaction Measurement. Journal of
Consumer Research. 29 (2), 159-167.

Ewing, M., Salzberger, T., and Sinkovics, R., 2005. An Alternate Approach to Assessing
Cross-Cultural Measurement Equivalence in Advertising Research. Journal of Advertising.
Spring 2005, 34 (1), 17-36.
Festinger, L., 1957. A Theory of Cognitive Dissonance. Stanford: Stanford University Press.

Fox, J., 1991. Regression Diagnostics. Thousand Oaks, CA: Sage Publications. Quantitative
Applications in the Social Sciences Series No. 79.

Hausknecht, D.R., Sweeney, J.C., Soutar, G.N., and Johnson, L.W., 1998. “After I had made
the decision, I …:” Toward a scale to measure cognitive dissonance. Journal of Consumer
Satisfaction, Dissatisfaction and Complaining Behavior. 11, 119 – 127.

Kasper, H., 1988. On problem perception, dissatisfaction and brand loyalty. Journal of
Economic Psychology. 9 (3), 387-397.

Korgaonkar, P.K., and Moschis, G.P., 1982. An Experimental Study of Cognitive Dissonance,
Product Involvement, Expectations, Performance and Consumer Judgement of Product
Performance. Journal of Advertising. 11 (3), 32-44.

ANZMAC 2005 Conference: Consumer Behaviour

295

---

Montgomery, C., and Barnes, J.H., 1993. POSTDIS: A short rating scale for measuring post
purchase dissonance. Journal of Consumer Satisfaction, Dissatisfaction and Complaining
Behavior. 6, 203-216.

Mooradian, T., and Olver, J., 1997. “I can’t get no satisfaction”: The impact of personality
and emotion on postpurchase processes. Psychology & Marketing. 14 (4), 379-393.

Morwitz, V.G., and Fitzsimons, G.J., 2004. The Mere-Measurement Effect: Why Does
Measuring Intentions Change Actual Behavior? Journal of Consumer Psychology. 14(1&2),
64–74.

Oliver, R. L., 1997. Satisfaction. New York: McGraw-Hill.

Rasch, G., 1980. Probabilistic Models for Some Intelligence and Attainment Tests. Chicago:
MESA, Reprint of 1960, Danish Institute for Educational Research.

Singh, J., 2004. Tackling measurement problems with Item Response Theory: Principles,
characteristics, and assessment, with an illustrative example. Journal of Business Research, 57
(2), 184-208.

Soutar, G., and Sweeney, J., 2003. Are There Cognitive Dissonance Segments?. Australian
Journal of Management. 28 (3), 227-249.

Sweeney, J., Hausknecht, D., and Soutar, G., 2000. Cognitive Dissonance after Purchase: A
Multidimensional Scale. Psychology and Marketing. 17 (5), 369-385.

Sweeney, J., Soutar, G., and Johnson, L., 1996. Are Satisfaction and Dissonance the same
construct? A preliminary analysis. Journal of Consumer Satisfaction, Dissatisfaction and
Complaining Behavior. 9, 138-143.

Westbrook, R., and Oliver, R. L., 1991. The dimensionality of consumption emotion patterns
and consumer satisfaction. Journal of Consumer Research. 18 (1), 84-91.

Footnotes

1 Whether a respondent is deemed affected by dissonance in a considerable way, moderately or only slightly/not
at all, depends on the Rasch person score in reference to the average of the item locations (considerable implies a
person score that is higher than the item average).
2 All constructs have also been evaluated based on traditional test theory (i.e., factor analysis, Cronbach’s alpha
reliability; results not presented here but available from the authors). Interestingly, satisfaction is correlated -.21
with dissonance (uncertainty component) and -.25 with dissonance (emotional component). This suggests that
the strength of association of cognitive dissonance and satisfaction assessed by standard procedures of
measurement may be underestimated in this case.
3 Moderate dissonance is defined as the area between the average of the Rasch item locations and one logit unit
below.
ANZMAC 2005 Conference: Consumer Behaviour

296

---