Sunteți pe pagina 1din 11

Chezy Ofir, Itamar Simonson, & Song-Oh Yoon

The Robustness of the Effects of


Consumers’ Participation in Market
Research: The Case of Service
Quality Evaluations
The authors propose that participation in market research can determine consumers’ experiences with and
evaluations of marketing services/products. Building on and extending a prior finding that expectation to evaluate a
service (or product) leads to more negative evaluations, this research investigates the robustness, process, and
consequences of asking consumers to form evaluations of services/products. The results of several field studies
(conducted in Israel, Korea, and the United States) show that (1) the effect of expecting to evaluate on the service
provider’s performance evaluation is enduring and lasts beyond the immediate aftermath of the service encounter;
(2) the effect of study participation on the perception of marketers reflects true perceptual change rather than
“made-up” perceptions—it requires supporting evidence and is eliminated if cognitive load interferes in the
production of that evidence; and (3) the robust bias produced by expecting to evaluate cannot be eliminated by
causing participants to consider the task effect on their evaluations or the impact of their feedback on the evaluated
marketers. The authors discuss the broader implications of this research for understanding the impact of
participation in market research on consumers’ experiences and inputs.

Keywords: market research, service expectations, customer satisfaction, task effects, consumer experience

btaining customer input is the main task of market vices, design more effective advertisements). Regardless of

O research and is essential for the development of


effective marketing strategies. Thus, marketers use
various research techniques to find out how customers eval-
whether customers receive monetary compensation for their
time and input, it is reasonable to assume that, in general,
they try to be helpful and provide useful inputs.
uate service quality and satisfaction, the importance of There are positive aspects to consumers’ desire to be
attributes, and how consumers’ brand perceptions change helpful. However, such a cooperative tendency also means
over time. At the same time, managers are often skeptical of that consumers’/respondents’ perceptions of what their role
the value of consumer studies and whether the firm should is and what being helpful entails might be key determinants
revise its current strategies, offerings, and communications of how they perform their task. Even more important, the
program on the basis of consumers’ evaluations and feed- mere participation in market research while or before expe-
back. A key goal of the current research is to assist man- riencing a product/service can affect customers’ experiences
agers and marketing researchers in understanding the fac- and subsequent satisfaction (e.g., Dholakia and Morwitz
tors that determine the validity of research findings and the 2002; Mandel and Nowlis 2008; Morwitz, Johnson, and
limitations of the data produced by the answers consumers Schmittlein 1993; Ofir and Simonson 2001, 2007).
provide to survey questions, in general, and questions Although prior research has illustrated that participation
regarding customer satisfaction, in particular. in market research can affect consumers’ perceptions, little
When providing feedback to marketers, customers is known about the robustness, underlying mechanisms, and
might be aware that the researchers are trying to learn from boundaries of these effects. Furthermore, a question that
their responses and are using the information they provide naturally arises is whether marketing researchers can con-
to improve performance (e.g., offer better products and ser- trol or eliminate such unintended effects of participation in
studies on consumers’ perceptions and evaluations by alert-
ing them to the possibility that the mere performance of a
Chezy Ofir is Academic Director of Davidson Research Center, School of
Business Administration, Hebrew University of Jerusalem (e-mail: task can affect their experience and satisfaction. The current
msofir@mscc.huji.ac.il). Itamar Simonson is Sebastian S. Kresge Profes- research examines these questions in the context of an
sor of Marketing, Graduate School of Business, Stanford University important and commonly used market research task—con-
(e-mail: Itamars@stanford.edu). Song-Oh Yoon is Assistant Professor of sumers’ evaluations of service quality. Prior research has
Marketing, Korea University Business School, Korea University (e-mail: shown that when consumers know in advance that they will
soyoon@korea.ac.kr). The authors thank the anonymous JM reviewers for be asked to evaluate a service provider, they tend to evaluate
their helpful comments and acknowledge the support of the Kmart and
Davidson Research Centers at the Hebrew University.
the service more negatively, thus misrepresenting the true
satisfaction of other customers (Ofir and Simonson 2001).

© 2009, American Marketing Association Journal of Marketing


ISSN: 0022-2429 (print), 1547-7185 (electronic) 105 Vol. 73 (November 2009), 105–114
We use this finding to gain a broader understanding of In this research, we examine the impact of a common
the problems that may arise by asking customers to provide customer research task in which a consumer is told in
feedback and inform marketers, and we examine whether advance to evaluate a service provider and then subse-
such undesirable consequences of market research dissipate quently provides his or her feedback (Ofir and Simonson
over time and can be eliminated by revising task instruc- 2001). Considering the importance of delighting customers
tions. The results of several field studies (in Israel, Korea, and enhancing customer loyalty, marketers routinely mea-
and the United States) show that (1) the negative effect of sure customers’ service evaluations and satisfaction. Fur-
expecting to evaluate on the service provider’s evaluation is thermore, in many cases, customers know in advance that
enduring and lasts beyond the immediate aftermath of the they will be asked to evaluate, either based on prior experi-
service encounter; (2) the effect of study participation on ence or because they receive the evaluation form at the
the perception of marketers reflects true rather than “made- beginning of the service encounter (e.g., when checking in
up” perceptions—it requires supporting evidence and is to a hotel).
eliminated if cognitive load interferes in the production of When told before a service encounter about an upcom-
that evidence; and (3) the robust bias produced by expecting ing evaluation task, many consumers are likely to expect
to evaluate cannot be eliminated by causing participants to that they should provide constructive suggestions for
consider the task effect on their perceptions or the impact of improvement. Merely reporting that “everything is fine”
their feedback on the evaluated marketers. We discuss the may be viewed as not doing a proper job and taking the
theoretical implications of the findings with respect to use easy way out. Conversely, identifying areas that need
improvement can (1) help enhance marketers’ performance
of market research and interpretation of consumers’ inputs
and (2) provide the customer’s point of view to a marketer
to marketers.
and/or to customer researchers regarding issues or areas in
which customers expect better performance. Indeed, most
The Effect of Participation in participants in a pilot study with 25 supermarket shoppers,
who were asked about the purpose of providing store
Market Research on Service evaluations, referred to service improvement aspects
Evaluations: Duration, Boundaries, (details of this pilot study are available on request).
and Underlying Process Customers who believe that they should identify defi-
A great deal of research in both psychology and marketing ciencies may be more inclined, perhaps unconsciously, to
has examined the role and consequences of demand charac- look for problems during the service encounter and to make
teristics and demand artifacts (e.g., Orne 1962). Much of sure they have something to say and are able to support and
this research has focused on experimental findings, espe- illustrate their feedback. In other words, to fulfill their role,
cially the impact of such effects on the internal validity and such customers may look for negative aspects. Such a
generalizability of laboratory studies. However, the notion process might be unconscious, as is typically implied by the
that experimental tasks and related cues may influence sub- notion of motivated reasoning (Kunda 1990), or conscious,
sequent perceptions and behavior is not limited to lab when respondents are aware of both their search for things
experiments and the familiar demand effects; they may also to improve and their resulting focus on negative perfor-
mance aspects.
apply to forms of actual market research and other con-
Ofir and Simonson (2001) show that expectation to
sumer initiatives (e.g., Sherman 1980). For example, cus-
evaluate a service or product produces more negative
tomers who are shown an advertisement for a new product
evaluations, which reflects the internalization of respon-
and asked if they intend to buy it may, in some cases,
dents’ perceived role, referred to as “negativity enhance-
assume that they are expected to respond favorably to the
ment” (i.e., the prominence of negative information), and
advertised product; in turn, this may affect their true evalua- their “role expectations.” However, this prior research does
tions of the evaluated new products. not address questions that are particularly important for our
Such effects of participation in market research can understanding of the impact of marketing research, in gen-
have a major impact on consumer behavior, regardless of eral, and service evaluations, in particular, on consumers’
whether consumers have well-defined hypotheses regarding experiences and perceptions. Specifically, prior research
their research roles. Even when there are no prior research does not enable us to determine whether the apparent
hypotheses and/or respondents discern the wrong hypothe- prominence of negative aspects reflects real changes in con-
ses, demand biases may be produced as long as respondents sumers’ perceptions of service providers. In particular, true
act on what they believe to be the research hypotheses or changes in perceptions, rather than mere transient artifacts,
the purpose of the study. Furthermore, whereas guessing the imply that the negative evaluations associated with expect-
precise hypothesis may often be difficult, participants in ing to evaluate will persist well beyond the research context
customer research and many other studies can have general and could be observed even several days or weeks follow-
notions about what certain tasks entail (e.g., paying close ing the service experience. This question is important for
attention to details). In particular, when asked for their marketers because a lasting effect would indicate that the
inputs regarding the performance of marketers and the ser- harm created by participation in market research can influ-
vices or products they offer, respondents will try to get ence long-term relationships with customers.
ready and be prepared to provide well-reasoned, construc- It is also important to determine whether the negative
tive feedback. bias of (expected) service evaluations can be eliminated by

106 / Journal of Marketing, November 2009


revising the task. Specifically, it might be possible to Study 1: Does the Negative Effect
decrease the bias produced by participation in market
research by alerting consumers to the notion that expecting
of Expecting to Evaluate Extend
to evaluate may affect their evaluations. The success of such Beyond the Service Encounter?
a debiasing manipulation is far from certain. Especially if Although prior research has established that expectation to
the bias produced by participation in market research is evaluate generates more negative service evaluations imme-
unconscious, eliminating it is likely to be challenging when diately after the experience (Ofir and Simonson 2001), we
consumers are unaware that they judge the service provider still do not know whether that effect lasts beyond the ser-
more negatively simply because they were asked to provide vice encounter and has longer-term implications. Beyond its
their evaluation. Conversely, if consumers recognize that practical implications, this question is important from a
they tend to be too critical, the mere mention of such task theoretical perspective because it can provide information
effects may diminish the resulting bias. as to whether the negative bias of expected evaluations is
Our assumption is that though market research partici- genuine and reflects true evaluations or whether it is merely
pants try to be helpful and to offer constructive feedback, a short-term effect that does not reflect true changes in
they are unlikely to fabricate or make up evidence when internal representation of the service experience. To exam-
none exists. That is, although expecting to evaluate leads ine this question, in Study 1, we varied both the task condi-
consumers to pay more attention to negative service aspects tion (expecting to evaluate or not expecting) and the timing
of the postservice evaluation (both immediately after the
(Ofir and Simonson 2001), they are unlikely to recognize
shopping experience and again 4–5 days later or just 4–5
that this negative focus is due to their task (e.g., Nisbett and
days later). This design enabled us to determine the robust-
Wilson 1977). Accordingly, customers are less likely to
ness of the negative bias of expected evaluations.
report more negative evaluations if their ability to support
such evaluations is constrained (e.g., if they are under cog- Method
nitive load), and we do not expect them to manipulate the
valence of information to fit their presumed task. The respondents were 120 consumers who were inter-
viewed immediately after shopping at a supermarket and/or
In summary, we predict that the negative effect of
4–5 days after shopping at that supermarket. They were ran-
expecting to evaluate will endure beyond the immediate
domly assigned to one of four conditions (n = 30 per
aftermath of the service encounter and be resistant to cor-
group), including two experimental groups and two control
rection or debiasing. Furthermore, consistent with the
groups. Shoppers assigned to the two experimental condi-
assumption that the more negative (expected) evaluations
tions were informed before entering the supermarket that
are based on the actual perceptions created during service
they would later be asked several questions about customer
evaluations, when consumers lack the cognitive resources
service. In one experimental group (Condition 1), con-
needed to perform the task and gather (mostly negative) evi-
sumers were interviewed immediately after their shopping
dence, the negative bias of expected evaluations will be less trip and again, by telephone, 4–5 days later. In the second
pronounced. In addition, we predict that the negative bias of experimental group (Condition 2), consumers were inter-
expected evaluations will be resistant to change. The fol- viewed by telephone only once, 4–5 days after shopping.
lowing hypotheses summarize our predictions: Respondents in the control groups were not informed about
H1: The negative effect of expecting to evaluate on service the upcoming evaluation task before entering the store and
evaluations extends beyond the immediate aftermath of were interviewed either immediately after finishing their
the service encounter. shopping (Condition 3) or 4–5 days later (Condition 4). As
H2: Restricting the cognitive resources during the service we explain subsequently, these control groups enabled us to
encounter decreases the impact of expecting to evaluate test two rival accounts for the results: (1) that more negative
on evaluations. delayed evaluations were produced by expressing more
H3: The negative effect of expecting to evaluate on service negative evaluations immediately after shopping and (2)
evaluations persists even after survey participants are that the more negative evaluations reflected the delay rather
alerted to a possible task effect; only explicit information
regarding the bias produced by the task decreases the
than the effect of expecting to evaluate.
effect. We used the same questionnaire in all four groups.
Specifically, we included the following measures: (1) listing
In Study 1, we test H1 by examining whether partici- of thoughts that come to mind while shopping at the store;
pants who expect to evaluate a store still hold more negative (2) six items pertaining to service, including employee
evaluations of that store after the service encounter. In courtesy, employee willingness to help, professionalism of
Study 2, we test H2 by contrasting the effect of expecting to employees, waiting time, convenience in shopping, and
evaluate under normal conditions with the effect of expect- quality of service; (3) two overall satisfaction items, satis-
ing to evaluate under a cognitive-load condition. Finally, in faction with the shopping trip and satisfaction with the ser-
Study 3, we test H3 by examining the effect of (1) asking vice; and (4) intention to recommend the store to friends
respondents to consider whether there is an effect of expect- (all measures, except for thought listing, used seven-point
ing to evaluate and what that effect might be and (2) scales). As a manipulation check, we also asked respon-
informing them about the negative effect of expecting to dents if they had expected to be interviewed about their ser-
evaluate. vice experience at the store.

Effects of Consumers’ Participation in Market Research / 107


Results MDelay only = 4.00; t(87) < 1; and for the recall measure:
As we expected, respondents in the two control groups did MDelay after immediate = –1.07 versus MDelay only = –1.03;
not expect to be questioned about their supermarket experi- t(87) < 1).
ence (average ratings in both groups were under 1.5 on a Second, a rival account for the more negative delayed
seven-point scale), whereas those in the expected- evaluations is that they reflect the delay rather than the
evaluation groups did expect to be questioned (average rat- effect of expecting to evaluate. However, the evaluations of
ings were greater than 6). participants in the delayed-evaluation control group were
As Table 1 shows and consistent with prior research, the significantly more positive than the evaluations of those in
evaluations of participants who expected to evaluate and both expected-evaluation (delayed experimental) groups
were interviewed immediately after shopping were signifi- (see Table 1). Because delayed evaluations in Conditions 1
cantly lower than the evaluations of those in the control and 2 were virtually identical and not significantly different,
group (for the combined service measure: MExpect = 4.72 we averaged and compared them with the delayed evalua-
versus MControl = 5.37; t(58) = 4.39, p < .0001; for the com- tions of the control (Condition 4) (for service: t(87) = 8.59,
bined satisfaction measure: MExpect = 4.58 versus MControl = p < .001; for satisfaction: t(87) = 6.33, p < 001; and for rec-
5.35; t(58) = 3.09, p < .003; and for recommendation: ommending the store: t(87) = 6.74, p < .001).
MExpect = 4.60 versus MControl = 5.13; t(58) = 1.96, p < .05).
Moreover, the service evaluations and likelihood of rec- Discussion
ommendation by the same respondents (i.e., Condition 1) The results of Study 1 show that the (negative) effect of
after a delay of 4–5 days were significantly lower than their expecting to evaluate on subsequent evaluations lasts
own evaluations immediately after shopping (for the service beyond the service encounter and has a longer-term effect
index: M = 4.14; paired-t(29) = 5.63, p < .0001; for recom- of consumers’ perceptions. Indeed, though tentative and
mendation: M = 4.20; paired-t(29) = 2.35, p < .03). Further- requiring replication, the results indicate that the negative
more, the difference between the number of positive and the effect may even become stronger after time has passed. This
number of negative recalled aspects of the visit also indi- finding is consistent with findings in prior research showing
cated an evaluation deterioration over time: (MDelay = –1.07 an increase in the extremity of attitude over a duration of
versus MImmediate = –.77; paired-t(29) = 2.07, p < .05). The time, which provides an opportunity to process evidence
overall satisfaction index was (lower but) not significantly consistent with the initial attitude (Sadler and Tesser 1973;
lower after 4–5 days of delay (MDelay = 4.35 versus Tesser and Conlee 1975).
MImmediate = 4.58; t(29) = 1.37, p > .1). The temporal robustness of the effect of expecting to
Next, we tested two rival accounts. First, it is possible evaluate suggests that the resulting (more negative) evalua-
that the negative delayed evaluations were produced by the tions reflect true changes in the internal representation of
more negative immediate (postshopping) evaluations of the store experience rather than an attempt to fulfill a pre-
participants who had expected to evaluate. However, the conceived evaluation role. In particular, expecting to evalu-
evaluations of those in a second experimental group, who ate the store’s service appears to change a consumer’s
expected to evaluate but were interviewed only 4–5 days actual shopping experience and to promote a more thorough
after shopping, were statistically indistinguishable from the evaluation process, which might account for the stable
evaluations of those who expected to evaluate and were effect on evaluations. If this conjecture is correct, we would
interviewed at both times (see Table 1) (for service expect conditions that limit the shoppers’ ability to evaluate
[delayed]: MDelay after immediate = 4.14 versus MDelay only = to reduce the effect of expecting to evaluate on evaluations.
4.04; t(87) < 1; for satisfaction [delayed]: MDelay after immedi- We test this prediction in Study 2 using cognitive load to
ate = 4.35 versus MDelay only = 4.27; t(87) < 1; for recom- control the available cognitive resources (Drolet, Luce, and
mendation [delayed]: MDelay after immediate = 4.20 versus Simonson 2009; Ward and Mann 2000).

TABLE 1
Study 1 Means

Expected
Evaluations: Expected
Immediate and Evaluations: Only Control: Immediate Control: Delayed
Delayed Evaluations Delayed Evaluations Evaluations Evaluations
(Condition 1) (Condition 2) (Condition 3) (Condition 4)
Service index (1–7) 4.72 → 4.14 4.04 5.37 5.35
Satisfaction index (1–7) 4.58 → 4.35 4.27 5.35 5.33
Recommendation (1–7) 4.60 → 4.20 4.00 5.13 5.30
Recall index: number of
positive aspects –
number of negative
aspects –.77 → –1.07 –1.03 .03 –.27

108 / Journal of Marketing, November 2009


Study 2: Using Cognitive Load then completed their shopping visit. When exiting the store,
they were informed of a second, unrelated product recall
Timing to Test the Process study and were asked to memorize the same nine names
Underlying the Effect of Market (followed by the store evaluation task). Thus, all four
Research on Consumers’ Service groups of respondents—control, expected evaluation, cog-
Evaluations nitive load before shopping, and cognitive load after shop-
A question that arises is whether the expectation to evaluate ping—were interviewed immediately after they existed the
creates a desire to report more negative (and constructive) store.
evaluations and whether consumers report such negative Specifically, respondents were asked to rate the super-
evaluations regardless of the actual information or evidence market on nine dimensions, all on seven-point scales. Six
they gather about the service at issue. That is, because con- items assessed specific dimensions of service quality,
sumers who expect to evaluate are determined to be con- including politeness of store employees, employees’ will-
structively negative, they will find something negative to ingness to assist shoppers, employees’ professionalism,
report regardless of the actual evidence they collect. Alter- length of wait at the checkout, convenience of product dis-
natively, it is possible that reporting negative evaluations play, and level of service (Cronbach’s α = .75). Two items
requires supporting evidence, and a failure to collect such measured satisfaction—overall satisfaction with the service
evidence will diminish the tendency to be more negative. and overall satisfaction with the shopping visit (r = .86, p <
We examine this question in Study 2 using a manipulation .0001)—and one item asked participants about their likeli-
of cognitive load (Drolet and Luce 2004; Drolet, Luce, and hood to recommend the store to friends. Respondents were
Simonson 2009). also asked whether, while shopping at the store, they paid
We examine the effect of cognitive load that is applied more attention to positive (7) or negative (1) aspects. A
either before or after the evaluation task is performed. manipulation check confirmed that those in the control
Specifically, in addition to the standard expected-evaluation group did not expect to be interviewed after shopping (M =
and control groups, some supermarket shoppers were asked 1.4 on a seven-point scale), whereas those in the expected-
to recall product names either before entering the store evaluation, cognitive-load-before, and cognitive-load-after
(after being told about the evaluation task) or immediately groups expected to be reinterviewed (higher than M = 6.6
after shopping (before reporting their evaluations). We on a seven-point scale).
expected that imposing cognitive load before shopping
would constrain consumers’ ability to gather the negative Results
evidence necessary for supporting negative evaluations. If As a replication of prior studies, shoppers in the expected-
such evaluations are reported only when pertinent evidence evaluation group evaluated the supermarket more negatively
is gathered online, the cognitive load will diminish or elimi- than those in the control group (for service: MExpect = 4.44
nate the effect of expecting to evaluate. versus MControl = 5.46; for satisfaction: MExpect = 4.07 ver-
However, even if cognitive load mitigates such negative sus MControl = 5.40; and for recommendation: MExpect =
evaluations, it may still have a neutralizing effect for other 4.27 versus MControl = 5.77; ps < .001). The main test
reasons, such as a more general effect of distraction, even focused on the contrast between the effect of cognitive load
without influencing the ability to gather (negative) evidence before and after shopping. The results indicated that cogni-
during the shopping experience. Accordingly, we contrasted tive load before entering the store statistically significantly
the effect of cognitive load before entering a supermarket decreased the negative effect of expecting to evaluate,
with the same cognitive-load manipulation that occurs whereas cognitive load after completing the store visit did
immediately after exiting the supermarket but before the not.
evaluation task. Specifically, all average ratings were significantly lower
in the cognitive-load-after group than in the cognitive-load-
Method before and control groups for service (MLoad_after = 4.85,
Participants were 120 supermarket shoppers, who received MLoad_before = 5.14, and MControl = 5.46), for satisfaction
no compensation. They were randomly assigned to one of (MLoad_after = 4.35, MLoad_before = 4.87, and MControl =
four conditions. The expected-evaluation group was 5.40), and for recommendation (MLoad_after = 5.20,
informed of the upcoming service evaluation task before MLoad_before = 6.10, and MControl = 5.77). Conversely, the
they entered the store, whereas the control group was not. evaluations of the cognitive-load-before group and control
Respondents in the cognitive-load-before-shopping group group were similar on all aspects (all ps >.1). Finally,
were also informed of the upcoming evaluation task. How- although the ratings of the cognitive-load-after group were
ever, they were then told that they would be participating in significantly lower than those of the control and cognitive-
a second, unrelated study that examined people’s ability to load-before groups, they were higher than the ratings in
recall product names. These respondents were then given a the expected-evaluation (no-load) group for service
minute and a half to memorize nine brand-product names, (MLoad_after = 4.85 versus MExpect = 4.44; t(116) = 3.06, p <
and their recall was tested after they exited the store and .01) and recommendation (MLoad_after = 5.20 versus
completed all other measures. Respondents assigned to the MExpect = 4.27; t(116) = 3.60, p < .01).
cognitive-load-after-shopping group were informed of the Next, we examined the effect of cognitive load and its
upcoming evaluation task before they entered the store and timing on the participants’ stated focus of attention. Corre-

Effects of Consumers’ Participation in Market Research / 109


sponding to the evaluation findings, the cognitive-load- vice: MLoad_Only_After = 4.86 versus MControl = 4.79; t(196) <
before group reported significantly greater focus on positive 1, p > .1; for satisfaction: MLoad_Only_After = 5.11 versus
aspects than the cognitive-load-after group (MLoad_before = MControl = 4.97; t(196) = 1.34, p > .1; and for recommenda-
5.10 versus MLoad_after = 4.30; F(1, 116) = 18.5, p < .001) tion: MLoad_Only_After = 4.96 versus MControl = 4.56; t(196) =
and was statistically indistinguishable from the control 2.97, p < .003) and significantly more positive than those in
(MLoad_before = 5.10 versus MControl = 5.13; F(1, 116) < 1, the expected-evaluation + cognitive-load-after-shopping
p > .1). Conversely, the cognitive-load-after group was sta- group (for service: MLoad_Only_After = 4.86 versus
tistically indistinguishable from the expected-evaluation MExpecting&Load_after = 4.28; t(196)= 6.40, p < .0001; for sat-
group (MLoad_after = 4.30 versus MExpect = 4.27; F(1, 116) < isfaction: MLoad_Only_After = 5.11 versus MExpecting&
1, p > .1) and significantly more negative than the control Load_after = 4.28; t(196) = 7.94, p < .0001; and for recom-
(MLoad_after = 4.30 versus MControl = 5.13; F(1, 116) = mendation: MLoad_Only_After = 4.96 versus MExpecting&
20.03, p < .001). Load_after = 3.84; t(196) = 8.30, p < .0001). Corresponding to
these results, the recall measure (number of positive less
Discussion negative aspects) was .84 in the control, –1.22 in the
The main finding of Study 2 is that cognitive load before expected-evaluation group (t(196) = 5.14, p < .0001), –1.28
the online evaluation (i.e., before the actual shopping in the expected-evaluation + load-after-shopping group, and
experience) greatly attenuates the effect of expecting to 1.08 in the cognitive-load-after-shopping group (t(196) =
evaluate, whereas cognitive load after the experience does 5.89, p < .0001). We obtained no significant difference
not. Furthermore, shoppers who received the name-recall between the control and the load-only-after-shopping
task (i.e., cognitive load) before shopping subsequently groups (shopping groups: t(196) < 1, p > .1). Finally, the
reported that their focus of attention was more positive than pattern of results for the focus-of-attention measure (i.e.,
those in the expected-evaluation group and the cognitive- the explicit measure regarding the focus to positive versus
load-after group. Evidently, cognitive load before shopping negative aspects of the supermarket) paralleled the evalua-
limits the production of negative evaluations during the tion results for the four groups (MControl = 4.54, MExpecting =
actual experience. However, after the negative evaluation 3.46, MLoad_Only_After = 4.82, and MExpecting&Load_after =
has been formed, cognitive load does not inhibit shoppers 3.84). Thus, the results of the follow-up study confirmed
from reporting their already-formed evaluations. that (1) cognitive load alone does not produce more nega-
These results indicate that expectation to evaluate tends tive store evaluations and (2) the valence of recall in the
to cause a shopper to pay attention to negative aspects of the cognitive-load-only-after-shopping group was similar to the
store’s service when cognitive resources are available. Fur- control group and significantly more positive than the
thermore, the findings of Study 2 imply that shoppers do expected-evaluation group.
not “fabricate” negative evaluations just to comply with Another follow-up study, with 120 bookstore shoppers,
their task, and they report such evaluations only if they are focused on the mediating role of consumers’ “role expecta-
able to gather supporting evidence during the shopping tions.” As the pilot study we mentioned previously shows,
experience. most consumers are likely to assume that their role is to
We extended Study 2 in a follow-up study with 200 offer constructive feedback, which highlights areas on
supermarket shoppers, who were randomly assigned to one which the service provider should improve its performance.
of four groups: control, expecting to evaluate, expecting to Accordingly, the study examined whether consumers’ per-
evaluate + cognitive load after shopping, or just cognitive ceptions regarding the type of feedback that would be most
load after shopping. We designed the fourth condition to helpful to the store mediates the effect of expecting to eval-
test whether cognitive load even without expecting to evalu- uate. In addition to the same measures of service, satisfac-
ate produces more negative evaluations. Another objective tion, and recommendation as in Studies 1 and 2, we mea-
of the follow-up study was to contrast the content of recall sured respondents’ perceptions of relative helpfulness of
from the shopping visit across the four experimental positive versus negative feedback to the store using a seven-
groups. As in Study 1, we computed for each group a recall point scale (1 = “negative feedback,” and 7 = “positive feed-
measure by subtracting the average number of positive back”). In line with Baron and Kenny’s (1986) mediation
aspects recalled from the average number of negative test, the results indicated that the consumers’ naive “role
aspects recalled. expectation” theory mediates the effect of expecting to eval-
The store evaluation results in the control, expected- uate on service evaluations. First, as a replication of the
evaluation, and expected-evaluation + load-after groups results of Study 1 and 2, there was a standard negative bias
were similar to those found in Study 2. We replicated the of expected evaluations on all three measures (for service
basic effect (for service: MControl = 4.79 versus MExpecting = quality: MExpect = 4.71 versus MControl = 5.22; t = 2.87, p <
4.06; t(196) = 8.06, p < .0001; for satisfaction: MControl = .01; for satisfaction: MExpect = 4.85 versus MControl = 5.32;
4.97 versus MExpecting = 4.05; t(196) = 8.80, p < .0001; and t = 2.10, p < .05; and for recommendation: MExpect = 4.73
for recommendation: MControl = 4.56 versus MExpecting = versus MControl = 5.36; t = 2.63, p < .01). Second, shoppers
3.58; t(196) = 7.27, p < .0001). More important, the evalua- who were expected to evaluate before entering the store
tion ratings in the cognitive-load-after-shopping group reported that negative versus positive feedback is more
(without expected evaluation) were either statistically indis- helpful to the store manager in improving future service
tinguishable from or higher than the control group (for ser- (MExpect = 3.43 versus MControl = 4.01; t = 2.00, p < .05).

110 / Journal of Marketing, November 2009


However, the significant main effect of expectation on to evaluate on evaluations; and (3) asking respondents, after
evaluation decreased significantly or marginally signifi- they are informed about the upcoming evaluation but before
cantly when we included both the task condition and the they enter the store, whether they expect the store to listen
perceived helpfulness of negative feedback as predictors. to their feedback and consider it important. This third task
The previously significant impact of expecting to evaluate variation enables us to assess the impact of consumers’ per-
on overall satisfaction was no longer significant when we ceptions regarding the impact of their service evaluations.
included the perceived helpfulness of negative feedback as We refer to the first revised task as “known effect,” to the
a predictor (t = –1.43, p > .1). In addition, Sobel tests show second as “predicted effect,” and to the third as “feedback
that the negative effect of expecting to evaluate on service use.” Consistent with the previous findings, we expect the
quality, satisfaction, and recommendation decreased signifi- shoppers’ negative bias of expected evaluations to be rather
cantly when we added perceived helpfulness of negative resistant to change, except when they are explicitly told
feedback as a predictor variable (Sobel’s Z = 1.77, p < .07; about the “bias.”
Sobel’s Z = 1.81, p < .07; and Sobel’s Z = 2.05, p < .05,
respectively). As prior research (e.g., Ofir and Simonson Method
2001, Study 3) has shown, despite the apparent mediating Participants were 175 supermarket shoppers who were ran-
impact of role expectations, the prominence of negative domly assigned to one of five conditions (n = 35 in each
information (i.e., negativity enhancement) has an indepen- condition). The expected-evaluation (Condition 1) and con-
dent, potentially greater impact on the resulting negative trol (Condition 2) groups were the same as in the prior stud-
evaluations. ies. Respondents in the known-effect group (Condition 3)
Specifically, the prominence of negative information is were also told of the upcoming evaluation task. However,
likely to interact with and be magnified by role expecta- they were then informed of a previous study that included
tions. Thus, as the results of Studies 1 and 2 indicate, con- two groups, one that had been told about the evaluation task
sumers do not just “make up” negative feedback to be help- and one that had not, which showed that the former group’s
ful. Instead, they appear to internalize their market research evaluations were significantly lower than the latter’s.
task, which unconsciously affects their shopping experi- Respondents in the predicted-effect group (Condition 4)
ence, the information they attend to, their recall, and subse- were also provided with a description of the previous study.
quent enduring store evaluations. However, instead of being told the outcome, they were
A question that arises is, Can marketers seek customer asked for their opinions as to whether the subsequent satis-
feedback and encourage consumers to pay attention to what faction with the store was the same in both groups or
they like and dislike about their service without creating a whether one of the two tended (after the store visit) to be
negative bias, which greatly undermines the value of the more satisfied than the other. Respondents assigned to the
feedback? That is, can the expected-evaluation task be feedback-use group (Condition 5) answered (immediately
revised in such a way that the negative bias in resulting after being told about the subsequent evaluation task) two
evaluations is eliminated or reduced? In Study 3, we exam- questions on seven-point scales: (1) “To what extent do you
ine the impact of three revised expected-evaluation instruc- expect the supermarket to listen to customers’ feedback?”
tions that call the respondents’ attention to the possible and (2) “To what extent do you believe that the supermar-
impact of their task and feedback. ket’s management regards customers’ feedback as important
and useful?”
Study 3: Can Predictions of Market When exiting the store, all respondents were asked the
same questions. They rated the supermarket on nine seven-
Research Task Effect Correct the point scales, including (1) six items that were designed to
Negative Bias of Expecting to assess specific dimensions of service quality (politeness of
Evaluate? store employees, employees’ willingness to assist shoppers,
Study 3 examines the impact of three revised expected- employees’ professionalism, length of wait at the checkout,
evaluation tasks, which call attention to the role of the convenience of product display, and level of service; Cron-
evaluation task and could potentially decrease or even bach’s α = .86); (2) two overall satisfaction measures (with
eliminate the effect. The evidence presented so far suggests service and the shopping visit; r = .81, p < .0001); and (3)
that the negative bias of expected evaluations is robust and their likelihood to recommend the store to friends. Respon-
not easy to reverse (with the exception of cognitive load dents were also asked whether, while shopping at the store,
before the shopping task). The question that arises is they paid more attention to positive (7) or negative (1)
whether calling study participants’ attention to the effect of aspects. A manipulation check confirmed that participants
the evaluation task on their own evaluations and on the in the control group did not expect to be interviewed after
store’s response to these evaluations can eliminate the nega- shopping (M = 1.35 on a seven-point scale), whereas those
tive bias of expected evaluations. Accordingly, the three in the expected-evaluation conditions did expect to be rein-
revised tasks tested in Study 3 include (1) explicitly inform- terviewed (higher than M = 6.5 on a seven-point scale).
ing shoppers that other consumers who were asked to eval-
uate tended to provide more negative evaluations; (2) pro- Results
viding more implicit information and asking those in the As Table 2 shows, consistent with the prior studies, the
expected-evaluation task to predict the effect of expecting evaluation ratings in the expected-evaluation group were

Effects of Consumers’ Participation in Market Research / 111


TABLE 2
Study 3 Means
Standard Expected Expected: Expected: Expected:
Control Evaluation Known Effect Predicted Effect Feedback Use

Service index (1–7) 5.60 4.56 5.25 4.49 4.46


Satisfaction index
(1–7) 5.72 4.40 5.21 4.47 4.33
Recommendation
(1–7) 5.40 3.91 5.03 3.23 4.06
Recall index:
number of positive
aspects – number
of negative
aspects .46 –1.17 .74 –1.11 –.66

lower than in the control (for service: MControl = 5.60 versus Discussion
MExpecting-no debias = 4.56; t(170) = 5.23, p < .0001; for satis- The results of Study 3 further show the robustness of the
faction: MControl = 5.72 versus MExpecting-no debias = 4.40; negative bias created when consumers are told in advance to
t(170) = 4.96, p < .0001; and for recommendation: form evaluations of marketers’ performance. The only
MControl = 5.40 versus MExpecting-no debias = 3.91; t(170) = group who received explicit information regarding that bias
4.01, p < .0001). Although the ratings were also slightly did not show it, whereas other procedures that were likely
lower in the known-effect group (i.e., the group told explic- to enhance consumers’ task involvement and cause them to
itly about the negative bias of expected evaluations) than in consider possible task effects failed to correct the negative
the control, these differences were statistically insignificant bias of expected evaluations. That is, both the predicted-
(see Table 2). The results suggest that the known-effect effect group and the feedback-use group were indistinguish-
group did not exhibit the bias, whereas the expected- able in their subsequent store and service evaluations from
evaluation condition, without any debiasing procedure, the standard expected-evaluation task. In real market
exhibited the bias. research applications, a company cannot and should not
The other two debiasing procedures failed to overcome lead respondents to address previously observed biases,
the effect of expecting to evaluate. Thus, asking consumers because such explicit instructions greatly undermine the
to predict the direction of the expected-evaluation effect did potential value of customer feedback and may simply
not diminish the negative bias. Specifically, the evaluations reflect demand effects. Thus, the observed bias produced
were similar to those of the expected-evaluation group, and by participation in a rather commonly used service/
the differences were insignificant (for service: MExpecting-no performance evaluation task is difficult to debias. These and
debias = 4.56 versus MPredicted = 4.49; t(170) < 1; for satis-
the previously reported results also highlight the process
faction: MExpecting-no debias = 4.40 versus MPredicted = 4.47; mechanisms that generate this robust bias.
t(170) < 1; and for recommendation: MExpecting-no debias =
3.91 versus MPredicted = 3.23; t(170) = 1.85, ps >.05). These
results were observed regardless of the predicted effect (i.e.,
General Discussion
This research examines the scope, boundaries, and mecha-
both the majority who expected the task to have a negative
nisms underlying the impact of participation in market
effect and those who expected a positive or no effect).
research on the evaluation of marketers’ performance.
Furthermore, asking respondents before they entered the
Although our focus is on consumer feedback regarding ser-
store whether they expected management to listen to their
vice performance, our findings have broader implications
feedback and find it useful did not debias the subsequent regarding some unintended consequences of market
evaluations (i.e., the negative bias was still observed) (for research with respect to customer loyalty and perceptions of
service: MExpecting-no debias = 4.56 versus MFeedback use = service quality. In particular, a primary determinant of the
4.46; t(170) < 1; for satisfaction: MExpecting-no debias = 4.40 value of market research is the degree to which the inputs
versus MFeedback use = 4.33; t(170) < 1; and for recommen- consumers provide in market studies/surveys can help mar-
dation: MExpecting-no debias = 3.91 versus MFeedback use = 4.06; keters improve their performance, such as offering better
t(170) < 1, ps > .1). As in the predicted-effect group, there products or better service. Thus, if consumers who partici-
was no significant difference between those within the pate in market research studies evaluate marketing perfor-
feedback-use group who expected the store management to mance differently from those who do not, (1) the provided
use the customer feedback and those who did not (p > .1). inputs may be misleading, and (2) the impact of participa-
Finally, as Table 2 shows, the recall measures corresponded tion in market research may linger beyond the task and
to the pattern of ratings in the various conditions, such that affect subsequent attitudes and purchase behavior.
lower ratings were associated with a higher percentage of Indeed, Study 1 shows not only that consumers who
negative versus positive recalled aspects. expect to evaluate a store perceive it more negatively (Ofir

112 / Journal of Marketing, November 2009


and Simonson 2001) but also that the bias persists after the reliability of the findings of such studies. Specifically,
several days, regardless of whether service evaluations are the same three steps we employed in this research can be
reported immediately after shopping or after a few days (or followed. First, researchers need to identify the beliefs, if
at both times). Indeed, there was preliminary evidence that any, that study participants form regarding their role and
the negative bias may exacerbate over time. This finding how it should be performed. Second, researchers can exam-
indicates that more negative expected evaluations are not ine the process and valence consequences of these beliefs.
made up but rather reflect the true resulting perceptions of Third, moderators of the observed effects need to be
the market research participants. investigated.
The subsequent studies reinforce this conclusion and Furthermore, a key question that needs to be addressed
demonstrate how difficult it is to correct the biased evalua- is whether and in what ways researchers might debias
tions. Study 2 shows that cognitive load while shopping participants in marketing studies. In general, the record of
interferes with the production of the negative bias, which debiasing procedures in terms of lasting effectiveness is
indicates that the resulting perceptions are “real” and rather mixed (e.g., Fischhoff 1982). In the case of the effect
require supporting evidence, which consumers under cogni- of expecting to evaluate, the simplest debiasing tactic is not
tive load fail to obtain. Conversely, cognitive load after to forewarn consumers about the upcoming evaluation task
shopping but before reporting store evaluations does not and merely interview them at the conclusion of the experi-
correct the negatively biased perceptions. ence. This approach assumes that the control group pro-
Study 3 further shows how challenging it is to debias vides unbiased and more accurate feedback, whereas those
market research participants; only participants explicitly who expect to evaluate are biased and report incorrect
told about the specific negative bias observed in a prior assessments. However, a limitation of relying on the control
study managed to avoid it. Conversely, asking respondents group (i.e., a nonforewarned group) is that people may
to consider the task effect or whether the store would listen mindlessly and superficially report a high level of satisfac-
to their feedback failed to debias. Overall, these results indi- tion; conversely, the more careful evaluation of those in the
cate that the negative bias of study participants is robust, expected-evaluation group might more accurately reflect
enduring, and difficult to debias. consumers’ true long-term responses to the evaluated ser-
We should recognize that most market research studies vice and the likelihood of switching to other service
often involve somewhat unnatural tasks that consumers do providers. Determining unambiguously which condition is
not perform except when providing inputs to marketers. For more biased or less informative can be challenging and may
example, outside the context of market research, consumers depend on the specific goals of the research.
rarely rate brands, service, or attribute importance on a For the current discussion, we continue to assume that
seven-point or any other scale. Furthermore, although con- the more negative evaluations among those who expected to
sumers may spontaneously evaluate their product satisfac- evaluate represent a bias and that forewarning consumers
tion or the performance of a service provider, the knowl- about the task is considered beneficial for obtaining careful
edge that participants prepare such evaluations as inputs to and accurate evaluations. Under these assumptions, a sim-
marketers who want to improve their offerings is likely to ple debiasing approach might be to tell participants that the
influence how these consumers approach the evaluation sole purpose of the research is to receive their honest and
task. Without an understanding of how this knowledge unbiased opinion, considering both positive and negative
affects the feedback-generation process and its resulting aspects they encounter. A downside of this approach is that
content, marketers may often be unable to determine it might stimulate the production of hypotheses about the
whether and in what way to use consumers’ inputs. objectives of the study and the way participants are
Each marketing research task may trigger different pro- expected to perform. Alternatively, the researchers could
cesses, perceptions, and naive role theories that affect the inform participants explicitly about the previously docu-
reliability of consumer inputs. In some cases, consumer role mented biases that might be created when consumers expect
perceptions and theories regarding their task may not inter- to evaluate; that is, in addition to informing participants
fere with accurate reporting of the true underlying states, if about the upcoming task, researchers can forewarn them
such exist. For example, certain findings derived from labo- about specific mistakes that expected evaluators tend to
ratory choice tasks (e.g., Dhar 1997; Huber, Payne, and make (as we did in Study 3).
Puto 1982; Kivetz 2005) appear to replicate under real- As managerial reliance on marketing research tech-
world conditions. However, in many other cases, con- niques increases, the limited understanding of the way
sumers’ beliefs about what they are supposed to do can play study participants perceive their roles and how such percep-
a key role in how the task is performed, leading to unreli- tions affect their behavior and reported evaluations becomes
able and potentially misleading results. In particular, con- more of an issue. For example, the task of portraying brands
sidering the growing evidence that consumer preferences as types of people may not be something that consumers
are often susceptible to seemingly irrelevant influences and perform spontaneously in everyday life, but such tasks are
their evaluations of consumption experiences are fuzzy and becoming increasingly popular in marketing research. Gain-
imprecise, perceived task roles can be especially influential. ing insights into consumers’ lay beliefs about this and other
Further research might examine other key marketing common market research tasks is essential to understand
research procedures in depth to determine what, if any, task whether and in what ways market research findings can be
(mis)perceptions they produce and the resulting impact on best employed.

Effects of Consumers’ Participation in Market Research / 113


REFERENCES
Baron, Reuben M. and David A. Kenny (1986), “The Moderator- Experience on the Enjoyment of That Experience,” Journal of
Mediator Variable Distinction in Social Psychological Consumer Research, 35 (1), 9–20.
Research: Conceptual, Strategic, and Statistical Considera- Morwitz, Vicki G., Eric Johnson, and David Schmittlein (1993),
tions,” Journal of Personality and Social Psychology, 51 (6), “Does Measuring Intent Change Behavior?” Journal of Con-
1173–82. sumer Research, 20 (1), 46–61.
Dhar, Ravi (1997), “Consumer Preference for a No-Choice Nisbett, Richard and Timothy Wilson (1977), “Telling More Than
Option,” Journal of Consumer Research, 24 (2), 215–31. We Can Know: Verbal Reports on Mental Processes,” Psycho-
Dholakia, Utpal M. and Vicki G. Morwitz (2002), “The Scope and logical Review, 84 (3), 231–59.
Persistence of Mere-Measurement Effects: Evidence from a Ofir, Chezy and Itamar Simonson (2001), “In Search of Negative
Field Study of Customer Satisfaction Measurement,” Journal Customer Feedback: The Effect of Expecting to Evaluate on
of Consumer Research, 29 (2), 159–67. Satisfaction Evaluations,” Journal of Marketing Research, 38
Drolet, Aimee and Mary Frances Luce (2004), “The Rationalizing (May), 170–82.
Effect of Cognitive Load on Emotion-Based Trade-Off Avoid- ——— and ——— (2007), “The Effect of Stating Expectations
ance,” Journal of Consumer Research, 31 (1), 63–77. on Customer Satisfaction and Shopping Experience,” Journal
———, ———, and Itamar Simonson (2009), “When Does of Marketing Research, 44 (February), 164–74.
Choice Reveal Preference? Moderators of Heuristic Versus Orne, Martin T. (1962), “On the Social Psychology of the Psycho-
Goal-Based Choice,” Journal of Consumer Research, 36 logical Experiment: With Particular Reference to Demand
(June), 137–47. Characteristics and Their Implications,” American Psycholo-
Fischhoff, Baruch (1982), “Debiasing,” in Judgments Under gist, 17 (11), 776–83.
Uncertainty: Heuristics and Biases, Daniel Kahneman, Paul Sadler, Orin and Abraham Tesser (1973), “Some Efforts of
Slovic, and Amos Tversky, eds. Cambridge, UK: Cambridge Salience and Time upon Interpersonal Hostility and Attraction
University Press, 422–44. During Social Isolation,” Sociometry, 36 (1), 99–112.
Huber, Joel, John W. Payne, and Christopher Puto (1982), “Adding Sherman, Steven J. (1980), “On the Self-Erasing Nature of Errors
Asymmetrically Dominated Alternatives: Violations of Regu- of Prediction,” Journal of Personality and Social Psychology,
larity and the Similarity Hypothesis Source,” Journal of Con- 39 (2), 211–21.
sumer Research, 9 (1), 90–98. Tesser, Abraham and Mary C. Conlee (1975), “Some Effects of
Kivetz, Ran (2005), “Promotion Reactance: The Role of Effort- Time and Thought on Attitude Polarization,” Journal of Per-
Reward Congruity,” Journal of Consumer Research, 31 (4), sonality and Social Psychology, 31 (2), 262–70.
725–36. Ward, Andrew and Traci Mann (2000), “Don’t Mind if I Do This:
Kunda, Ziva (1990), “The Case for Motivated Reasoning,” Psy- Disinhibited Eating Under Cognitive Load,” Journal of Person-
chological Bulletin, 108 (3), 480–98. ality and Social Psychology, 78 (4), 753–63.
Mandel, Naomi and Stephen M. Nowlis (2008), “The Effect of
Making a Prediction About the Outcome of a Consumption

114 / Journal of Marketing, November 2009

S-ar putea să vă placă și