THE EFFECT OF STATING EXPECTATIONS ON CUSTOMER SATISFACTION AND SHOPPING EXPERIENCE. Chezy Ofir School of Business Administration, Hebrew University

Transcription

1 1 THE EFFECT OF STATING EXPECTATIONS ON CUSTOMER SATISFACTION AND SHOPPING EXPERIENCE Chezy Ofir School of Business Administration, Hebrew University Itamar Simonson Graduate School of Business, Stanford University February 2006 Chezy Ofir is Associate Professor of Marketing, School of Business Administration, Hebrew University, Mount Scopus, Jerusalem 91905, Israel Itamar Simonson is Sebastian S. Kresge Professor of Marketing, Graduate School of Business, Stanford University The authors acknowledge the support of the Davidson and Kmart Research Centers at the Hebrew University and the Stanford Graduate School of Business, as well as the assistance of students who participated in data collection and analysis.

2 2 The Effect of Stating Expectations on Customer Satisfaction and Shopping Experience ABSTRACT Customers expectations are key determinants of their consumption experiences, satisfaction, and loyalty. The authors examine alternative theoretical predictions regarding the impact of stating expectations before purchase on post-purchase perceptions of the shopping experience and the firm. In particular, the authors suggest that asking customers to articulate their expectations can backfire and lead to lower post-purchase evaluations of the shopping and consumption experience. A series of field experiments indicate that, compared to a control group, stating prepurchase expectations leads customers to focus on negative aspects of the shopping experience and perceive the same performance more negatively. The tendency to rate shopping experiences less favorably produced by stating pre-purchase expectations is inconsistent with confirmation bias as well as assimilation, contrast, and positivity effects. The last study contrasted the impact of stating expectations about the next store shopping experience with the impact of evaluating the store s past performance. The results show that, although (pre-purchase) expectations were indistinguishable from evaluations of the store s past performance, the former led to lower postpurchase (than pre-purchase) evaluations whereas the latter tended to generate higher postpurchase evaluations. The article concludes with a discussion of the theoretical and practical implications of this research. Key words: Customer expectations, negativity, surveys, satisfaction.

3 3 It is now widely accepted that exceeding customer expectations is key to customer satisfaction, delight, and loyalty (e.g., Kotler 2000, p. 36). Accordingly, it is critical for marketers to try to find out in advance what their customers expectations are, because a failure to meet or exceed those expectations could lead to dissatisfaction and defection. In some cases, customers have well-formed expectations, for example, when they have a great deal of experience with the service or product at issue. Other times, expectations may be ill-defined, in which case asking customers to state expectations might help formulate or even create them (e.g., Bettman, Luce, and Payne 1998; Schwarz and Bohner 2001). Furthermore, whether or not the measurement of expectations affects their content and clarity, the need for customers to state expectations prior to a consumption experience is likely to make them more accessible during the experience (Feldman and Lynch 1988). As discussed subsequently, these and related issues have been examined in the context of research on the effects of measuring intentions, judgments, and satisfaction (e.g., Dholakia and Morwitz 2002; Fitzsimons and Williams 2000; Kardes, Allen, and Pontes 1993; Morwitz and Fitzsimons 2004). In the present research, we examine the impact of measuring customers expectations just prior to a shopping or consumption experience (e.g., before entering a store or a bank) on the subsequent evaluation of that experience and the firm. Prior research suggests that measuring customers satisfaction with the firm s past performance induces more favorable evaluations of the firm and has a positive effect on purchase behavior and loyalty (Dholakia and Morwitz 2002). As Dholakia and Morwitz point out, this finding suggests the intriguing possibility that satisfaction surveys can be used strategically to strengthen customer relationships. Since evaluations of satisfaction with the firm s past performance are likely to provide the basis for

4 4 expectations regarding future performance, one might predict that measuring expectations will have a similar effect on subsequent evaluations and behavior as measuring satisfaction. However, research also indicates that, compared to customers who are unaware of an upcoming service evaluation task, customers who are forewarned before a service encounter that they would be asked subsequently to evaluate that service provide more negative post-experience evaluations (Ofir and Simonson 2001). If one assumes that expecting to evaluate a service or product spontaneously generates expectations (for a related argument, see, e.g., Fitzsimons and Williams 2000), then stating expectations before a consumption experience might actually generate more negative (i.e., lower) evaluations of the firm following that experience. In the next section, we explore the different predictions that can be derived from prior research about the effect of stating expectations on subsequent evaluations. We then show that measuring expectations tends to generate lower evaluations. We also present evidence regarding the evaluation processes triggered by stating expectations, which indicates that the measurement of expectations causes customers to pay more attention to negative aspects and to encode performance characteristics more negatively. In the final study, we directly contrast the effect of measuring prior satisfaction with the effect of stating expectations before purchase. Consistent with the evidence presented here as well as the work of Dholakia and Morwitz (2002), we find that, although stated expectations about future performance are indistinguishable from judgments of past performance, measuring expectations generates lower subsequent evaluations whereas assessing past performance tends to produce more positive evaluations. We conclude with a discussion of the theoretical and practical implications of the results.

5 5 INFLUENCES OF STATING EXPECTATIONS: CONFIRMATION, POSITIVITY, AND/OR NEGATIVITY ENHANCEMENT As is often the case when analyzing the impact of a particular condition on response, the difficulty is not in identifying potentially relevant theories, but rather, determining which of several candidate theories is most applicable. In particular, existing theories could support different predictions regarding the effect of stating expectations about a consumption experience, including (a) stated expectations will confirm and reinforce existing beliefs about the experience being considered, which, in turn, will affect post-experience evaluations in accordance with the stated expectations, (b) stating expectations will generate higher post-experience evaluations, and (c) stating expectations will generate lower post-experience evaluations. One assumption that we can make rather confidently is that the act of stating expectations for a service (or product) experience, just prior to the consumption experience or service encounter, makes the stated expectations more accessible and salient during that and possibly subsequent experiences (for related arguments, see, e.g., Dholakia and Morwitz 2002; Feldman and Lynch 1988; Fitzsimons and Williams 2000; Kardes et al. 1993; Morwitz and Fitzsimons 2004). One might also expect that the mere act of stating expectations enhances involvement with the subsequent shopping experience, leading to more central-route processing of information and potentially more accurate judgments (e.g., Petty and Cacioppo 1984). Furthermore, prior research indicates that, while stated expectations may be influenced by preexisting expectations and other information stored in memory, they are often constructed and become clarified when the need to articulate them arises (e.g., Bettman et al. 1998; Schwarz and Bohner 2001). However, these prior findings regarding accessibility, involvement (or ELM), and

6 6 preference construction do not lead to directional predictions with respect to the impact of stating expectations on the valence of post-experience evaluations. Assimilation and confirmation bias (e.g., Nisbett and Ross 1980; Sanbonmatsu, David, and Kardes 1998) offer one possible account for the effect of stating expectations on postexperience evaluations. Specifically, stating expectations may reinforce prior beliefs and lead to post-purchase evaluations that confirm and are even more polarized than prior expectations (e.g., Tesser 1978). Alternatively, we might observe a contrast effect or regression to the mean, such that those with positive priors will subsequently have less favorable evaluations and those with lower expectations will later evaluate the store more favorably. If we assume that the measurement of satisfaction with past performance is equivalent to the measurement of expectations regarding future performance, then we might observe a positive effect of stating expectations on subsequent evaluations. Specifically, as Dholakia and Morwitz (2002) argue, users of a firm s service (e.g., a store or a bank) are likely to include a relatively high proportion of satisfied customers, which is often the reason they are that firm s customers in the first place. Accordingly, asking those customers to state their expectations for the firm causes them to articulate their favorable evaluations and thus further reinforces their positive attitude, leading to more positive post-experience evaluations. It is noteworthy that participants in the Dholakia and Morwitz studies were aware that the satisfaction study was conducted on behalf of the firm being evaluated, which might have contributed to the positive impact of the survey on subsequent customer behavior (see also Dholakia, Morwitz, and Westbrook 2005). However, recent research by Williams, Fitzsimons, and Block (2004) suggests that such mere measurement effects also occur, and might even be more pronounced, when the organization sponsoring or conducting the research is not identified.

7 7 Yet another possible effect of stating expectations is to generate lower post-experience evaluations. Ofir and Simonson (2001) showed that customers who expect to evaluate a service before the actual experience tend subsequently to evaluate that service more negatively than customers who find out about the need to evaluate only after the experience. They offered two possible explanations for this effect, referred to as negativity enhancement and role expectations. The latter account is specific to the expected evaluation task and is based on the notion that consumers believe that their evaluation role calls for constructive criticism, which generates more negative evaluations. Conversely, the negativity enhancement account suggests that a general tendency to overweigh negative aspects (e.g., Fiske 1980; Kanouse 1984; Tversky and Kahneman 1991) becomes more pronounced when evaluations are constructed online. As a result, any task condition that stimulates more online rather than memory-based evaluations (e.g., Hastie and Park 1986) would be expected to promote more negative evaluations (e.g., of service providers). Consistent with this account, an explicit need to form and state performance expectations before a service encounter might have a similar negative effect on subsequent evaluations, because it is likely to encourage comparisons between the stated expectations and actual performance. Furthermore, in their Study 1, Ofir and Simonson (2001) showed that stating expectations about a service also had an independent negative effect on subsequent evaluations of the service. However, they did not pursue that finding and focused instead on the effect of expecting to evaluate. Thus, just as negativity enhancement accounts for the negative effect of expecting to evaluate, it leads to the prediction that stating expectations prior to a service encounter will generate more negative subsequent evaluations of the service experience. In summary, prior research could be used to support very different predictions regarding the impact of stating expectations prior to a service encounter on post-experience evaluations of

8 8 that service. Specifically, (a) an assimilation effect and confirmation bias suggest that the (positive or negative) expectations stated prior to the experience will largely determine and make more extreme the post-experience evaluations; (b) a contrast effect suggests that those stating positive (negative) expectations will tend to evaluate the service more negatively (positively) after the experience; (c) if measuring expectations is equivalent to measuring satisfaction with past performance, then stating expectations could lead to more positive evaluations; and (d) negativity enhancement suggests that stating expectations will produce more negative subsequent evaluations. These rival predictions are examined in the studies described next. STUDY1 In Study 1, we contrast the post-purchase evaluations of two customer groups: shoppers who stated their expectations for a shopping experience before entering a supermarket and a control group that was interviewed only when exiting the supermarket. As discussed above, given conflicting theoretical accounts, we used this study as a preliminary investigation. Method The participants were 80 shoppers who were interviewed at the entrance to a supermarket and received no compensation. Those randomly assigned to the stated expectations group (n=40) were interviewed both before and after shopping at the store, whereas the control group (n=40) was interviewed only after shopping. When first interviewed, respondents were informed that the survey was part of a university research project and were assured of confidentiality. Those in the stated expectations group were asked to indicate their expectations regarding the service they would receive at the supermarket on that day. Specifically, they stated their expectations for the store on nine dimensions, all on a 7-point scale. Six items were designed to assess specific dimensions of expected service quality, including politeness of store employees,

9 9 employees willingness to assist shoppers, employees professionalism, length of wait at the checkout, convenience of product display, and level of service. Two items measured directly expected satisfaction: overall satisfaction with the service and overall satisfaction with the shopping visit. Finally, participants were asked about the likelihood that they would recommend the store to friends (for an alternative scale, see Parasuraman, Zeithaml, and Berry 1988). Upon exiting the store, those in the stated expectations group and those assigned to the control group were unexpectedly asked to evaluate their experience and the store based on the visit they had just completed. They rated the store using the exact same items as those used in the stated expectations group before entering the store. Respondents were also asked whether, while shopping at the store, they paid more attention to positive (7) or negative (1) aspects. Respondents were randomly assigned to conditions as they were about to enter the supermarket. Interviewers were able to track shoppers assigned to each group, who entered the supermarket through the same entrance/exit. A concern may arise regarding the condition assignment procedure and the possibility that participants in the stated expectations and control groups differed in systematic ways. We addressed these concerns in four ways. First, interviewers were blind to the purpose of the study and the research hypotheses. Second, to disguise the fact that those in the stated expectations group would be re-interviewed when exiting the store, the interviewer thanked them at the conclusion of the pre-shopping interview as if they completed their task. Third, we included the following question at the end of the post-shopping interview: To what extent did you expect that we would interview you about your shopping experience at the store? (1=did not expect at all; 7=definitely expected). In addition, in Study 2 we replicated the test of the effect of stating expectations using a more traditional lab experiment in which random assignment of subjects to conditions was straightforward.

10 10 Results Neither the stated expectations group (M=2.0, on a 7-point scale) nor the control group (M=1.7) had expected to be interviewed following the shopping visit (F(1,78)=1.2, p>.1). The average pre-visit service expectations in the stated expectations group, across the six specific service items (α=.91), was 5.4 on a 7-point scale. The average of the two satisfaction items (α=.98) was 5.9, and the average rating of the likelihood of recommending the store was 5.9. Following the shopping visit, those in the stated expectations group evaluated the supermarket more negatively than those in the control group: for the combined service measure, MAfter=3.9 versus MControl=4.5 (F(1,78)= 6.2, p<.01); for the combined satisfaction measure, MAfter=4.0 versus MControl=4.8 (F(1,78)=6.2, p<.01); and for the recommendation measure, MAfter=3.7 versus MControl=4.7 (F(1,78)=5.3, p<.01). Moreover, the post-purchase evaluations of those in the stated expectations group were lower than their pre-shopping expectations on all dimensions (at the p<.01 level). 1 It is noteworthy that the pre-purchase expectations in the stated-expectations group were higher than the post-purchase evaluations of the control group; however, no such (or any) pattern was observed in subsequent studies. Finally, when asked during the post-purchase survey whether they had paid more attention to negative (1) or positive (7) aspects of their visit, the ratings of customers in the stated expectations group indicated that they had paid more attention than those in the control group to negative aspects of their shopping visit: MStated Exp=4.3 versus MControl=5.2 (F(1,78)=9.9, p<.01). 1 A median-split based on stated expectations further shows that, after the shopping experience, those with higher prior expectations were as negative as those with more negative priors. A similar pattern was observed in the subsequent studies. Details can be obtained from the authors.

11 11 Discussion The main finding of Study 1 was that stating expectations just prior to a shopping experience led to more negative evaluations of the service and store following that experience, relative to both a control group that evaluated the store only after the shopping experience and relative to the pre-experience expectations. All expectations/evaluations, both before and after the experience, were measured on the same dimensions. These findings are inconsistent with accounts based on confirmation bias or assimilation, which would imply that stating positive expectations before entering a store would make subsequent evaluations more positive, rather than more negative. Furthermore, the results of Study 1 indicate that, unlike satisfaction surveys, stating expectations does not appear to create more favorable attitudes towards the firm or store. The results of Study 1 do not allow us to distinguish between the negative effect of stating expectations suggested by negativity enhancement (Ofir and Simonson 2001) and a contrast effect. That is, although a median-split analysis showed that both the relatively high and low stated expectations groups expressed more negative evaluations after the shopping experience, a contrast effect could have still contributed to the obtained results. This issue was addressed in a follow-up study with 90 consumers in which we tested the effect of stating expectations before entering a supermarket that had a rather negative reputation (due to past service problems). Thus, given low expectations, a contrast effect would suggest that post-purchase evaluations would be more positive than stated expectations, whereas negativity enhancement predicts that the negative effect of stating expectations on post-experience evaluations would still be observed. In this follow-up study, we added a recall measure taken immediately after participants exited the store. Specifically, respondents were asked, What do you remember from today s visit to the supermarket? Coders, who were blind to the

12 12 hypotheses, classified the responses as positive, negative, or neutral. We also included a third condition in which respondents were only interviewed before entering the store and asked an open-ended question: What are your expectations for today s shopping visit at this supermarket? Answers to this question were designed to provide insights into the spontaneous expectations of shoppers that were not influenced by the provided closed-ended items. The results of this follow-up study showed that the post-experience evaluations in the stated expectations group were again lower than in the control group on all measures (p<.01). Furthermore, an analysis of the open-ended responses showed that, compared to the control group, those in the stated expectations group recalled more negative (M=2.2 vs. M=1.1) and fewer positive aspects (0.8 vs. 2.0) of the shopping experience. These results were consistent with the participants own ratings of their focus of attention: on a 7-point scale (where 1=paid attention more to negative aspects), the average rating in the control group was 4.9 compared to 3.4 in the stated expectations group (F(1,58)=19.4, p<.01). These results support the negativity enhancement account and are inconsistent with the explanation based on the operation of a contrast effect. Finally, respondents in the (pre-shopping) expectations-only group stated more positive (1.6) than negative (0.9) expectations. A possible concern about the studies described so far is that, despite the attempt to assign respondents randomly to conditions and the reliance on interviewers who were blind to the research purpose and hypotheses, the participants in the control and stated expectations groups differed on other dimensions (e.g., their prior evaluations of the store). Accordingly, in Study 2 we used a procedure that ensured that the assignment of respondents to conditions was random. We also attempted to have tighter control over the actual content of the considered information. Furthermore, Study 2 examined another rival (or complementary) account according to which

13 13 stating expectations also changes how participants encode the same items of information. That is, one might conjecture that, in addition to any shift in the focus of attention resulting from stating expectations, that task may affect how customers interpret the observed performance of the store (see, e.g., Schwarz and Bohner 2001). Method STUDY 2 The respondents were 60 Israeli consumers, who were interviewed at their homes or offices. They were asked to evaluate a new (fictitious) drugstore (referred to as Pharm ) chain that was being established, based on a newspaper review of that chain s first store. Respondents were randomly assigned to the stated expectations (n=30) or control group (n=30) and were informed that the research was part of a university study. Those in the stated expectations group were told in advance that they would be shown a newspaper review of the first store of the new Pharm chain. They were also informed that they were randomly selected for the survey in order to gauge customers likely evaluations of the new chain. Before reading the article, those respondents stated their expectations with respect to service and the overall expected evaluation of the store (both on a 7-point scale). It should be noted that, although respondents had no real basis for stating expectations, the use of the (abbreviated) English term Pharm and the existence of a popular drugstore chain called SuperPharm were likely to produce relatively high expectations (compared to typical Israeli pharmacies that tend to offer less variety and poorer service). Respondents in the control group were given the same introduction but were not asked to state expectations for the store. The translated text of the article appears in Appendix A. We designed the store review such that it contained thirty characteristics of a drugstore, with the same number of positive, negative, and neutral characteristics. Specifically, 18 pretest respondents from the same

14 14 population that participated in Study 2 were asked to rate characteristics of a Pharm store on a 1 (negative) to 7 (positive) scale. Items with an average rating of over 4.75 were classified as positive characteristics, items with ratings between 4.75 and 3.25 were classified as neutral, and items with average ratings of less than 3.25 were considered negative. After respondents finished reading the article, it was removed from sight. They were then given four minutes to list as many of the characteristics of the drugstore as they could remember from the article. Next, respondents rated the degree to which each of the 30 drugstore characteristics mentioned in the article was negative (1) or positive (7). The pre-designated positive, negative, and neutral characteristics were intermixed on the list. Respondents then rated the service of the reviewed store on four (7-point) items (cleanliness, professionalism of employees, product variety, service quality), overall store evaluation on two items (negative/ positive and bad/good), and whether they would recommend the store to friends. Finally, respondents indicated if they had paid more attention to negative (1) or positive (7) aspects. Results As expected, based solely on the mention of a new Pharm chain, respondents in the stated expectations group had relatively high expectations: 5.3 on service and 5.6 for the store s overall score. After reading the store review, respondents in both the stated expectations and control groups listed in four minutes all the store characteristics they could recall. Those in the control group remembered an average of 4.1 (pre-designated) positive features, 2.9 negative features, and 2.9 neutral features. Conversely, the average number of positive, negative, and neutral recalled characteristics in the stated expectations group was 1.7, 3.9, and 1.8, respectively. Thus, the control group remembered more positive (and neutral) features and fewer negative features ( F(1,58)=41.8, p<.01, and F(1,58)=6.1, p<.01, respectively); negative features

15 15 represented 54% of the features recalled in the stated expectations group, compared to just 29% in the control group. These results were consistent with the participants own rating regarding their focus of attention, which indicated that those in the stated expectation group focused more on negative aspects (M=3.3, where 7= paid attention more to positive aspects) than those in the control group (M=5.6, F(1,58)=77.4, p<.01). Next, respondents rated each of the 30 store characteristics mentioned in the review on a negative (1) to positive (7) scale. In all cases, the characteristics were rated more negatively in the stated expectations group (in all cases, p<.01). For example, (a) the statement a store membership card can be purchased for a small fee received a rating of 2.6 in the stated expectations group and 6.1 in the control group; (b) the statement, the store manager has an academic degree received an average rating of 5.0 in the stated expectations group and 6.6 in the control group; and (c) the statement the store does not accept checks received an average rating of 2.3 in the stated expectations group and 4.2 in the control group. Thus, the same store features were rated systematically more negatively in the stated expectations group. Furthermore, the average ratings on the provided service items (α=.95), store overall (α=.94) and recommendation items followed a similar pattern. The average service (M=3.2 vs. M=5.7; F(1,58)=127.7, p<.01), store overall (M=3.2 vs. 5.9; F(1,58)=161.8, p<.01), and likelihood of recommendation (M=3.1 vs. M=5.9; F(1,58)=156.7, p<.01) ratings were all lower in the stated expectations group. It is also noteworthy that the lower evaluations produced by stating expectations was observed both among those with high expectations and those with low expectations (based on a median-split of prior expectations), which is inconsistent with a simple contrast effect of stated expectations on subsequent evaluations. Discussion

16 16 The results of Study 2 provide further insights into the lower post-purchase evaluations of customers who had first stated their expectations. Specifically, the results show that, not only do stated expectations lead to a focus on negative aspects of the available information, but stating expectations also affects the encoding and interpretation of the same items of information. That is, in addition to remembering more negative features mentioned in the store review, those who had stated expectations about the store subsequently rated the same store features lower (i.e., more negative), compared to the control group. Thus, stating expectations evidently triggers a critical, negatively skewed evaluation mode and focus of attention. A question that arises about the negative impact of stating expectations on subsequent evaluations is whether the mere act of stating expectations, even regarding an unrelated entity, triggers a negative bias (i.e., lower evaluations). Accordingly, we examined in Study 3 the impact of stating expectations about unrelated firms, one that was known to be associated with favorable expectations and another known to be associated with unfavorable expectations. Method STUDY 3 Study 3 included four experimental groups, consisting of a control and three different stated expectations groups. The respondents were 160 supermarket shoppers, who were randomly assigned to one of the four conditions (n=40 per group). In addition to the standard stated expectations group, the other two groups stated expectations (also before entering the supermarket) either about the duty-free shop in Israel s main airport or about a service center of the largest Israeli communication company ( Bezek ). The duty-free shop was known to have favorable associations even among those who had not shopped at that store, whereas the communication company had been involved in a severe labor dispute (related to the privatization of the company) and was generally associated with rather poor customer evaluations.

17 17 We reasoned that, if the mere act of stating expectations leads to lower evaluations even when the expectations refer to an unrelated target, then all stated expectations groups should evaluate the supermarket equally negatively after. An alternative prediction is that stating expectations about the supermarket visit lowers subsequent evaluations, whereas expectations about other entities will either have no effect or will lead to a corresponding contrast effect. According to the latter account, those who state expectations about the subsequently evaluated target will be more negative than the control group, whereas the sign of the difference between the other stated expectations groups and the control will depend on whether the evaluated supermarket is perceived more or less favorably than the reference point evaluated earlier. All three stated expectations groups indicated their expectations (regarding the supermarket, the duty-free shop, or the communication company s service center) using virtually the same items that were used in Study 1; there were just two slight changes with respect to the telephone company s service center: length of wait at the checkout was changed to length of wait at the service desk, and convenience of product display was changed to product availability. The measures, referring to service, satisfaction, and likelihood of recommending the store, were also used to evaluate the supermarket after participants completed their shopping. Before evaluating the supermarket, they answered an open-ended question regarding what they could recall from the just completed shopping experience. Finally, respondents indicated whether they had expected to be asked about their shopping experience at the supermarket and indicated whether they paid more attention to negative (1) or positive (7) aspects while shopping. Results All four groups had not expected to be questioned about their supermarket shopping experience (the average ratings in all groups were under 2.0 on a 7-point scale; see the Table).

18 18 Again, in the standard stated expectations group, expectations regarding the supermarket were higher than the subsequent post-purchase ratings by the same customers (for Service, pairedt(39)=12.2, p<.01; for Satisfaction, paired-t(39)=11.6, p<.01); and for Recommendation, pairedt(39)=14.0, p<.01). The post-visit evaluations were also lower than those provided in the control group (p<.01 for Service, Satisfaction, and Recommendation). As predicted, the stated expectations regarding the duty-free shop were the highest (over 6 on a 7-point scale), followed by the supermarket (between 5.8 and 5.9), with the communication company receiving substantially lower expectation ratings ( ). The postvisit evaluations showed that those in the telephone company group rated their supermarket visit most favorably, followed by the control group, duty-free group, and finally, the supermarket stated expectations group, which was the most negative. These results suggest that both a contrast effect and a negative effect of stating expectations regarding the target (supermarket) contributed to the pattern of findings. Omnibus one-way ANOVAs applied to all post supermarket visit responses (i.e., service, satisfaction, and recommendation measures) were significant at the p<.01 level. A set of contrasts supported the conclusions regarding the differences between groups. An analysis of the content of participants recall from the supermarket visit revealed a similar pattern. Those in the supermarket stated expectations group remembered the highest number of negative aspects, whereas the control group recalled the highest number of positive aspects; the other two groups were either equal to the control group or between the control and supermarket stated expectations groups (see Table). Finally, participants own reports (on a 7- point scale) regarding their focus of attention while shopping showed that those in the (supermarket) stated expectation group were most focused on negative aspects (M=2.5),

19 19 followed by the duty-free group (M=3.1), the control group (M=4.8), and the communication company group (M=5.3). Discussion Consistent with the earlier findings, the results of Study 3 indicate that, although a contrast effect relative to a reference point can affect evaluations of the target experience, a contrast effect cannot account for the robust negative effect of stating expectations about the target on post-experience evaluations. For example, the relative proportion of recalled negative aspects was the highest in the supermarket stated expectations group and was higher than in the duty-free shop group, despite the very positive stated expectations for that shop. This finding is also consistent with the results of the follow-up study reported earlier, in which even a supermarket associated with relatively negative expectations was evaluated less favorably among shoppers who had stated their expectations. Furthermore, consistent with the earlier studies, the results of Study 3 show that stating expectations does not produce a confirmation bias. The finding that stating expectations about a service provider before a service encounter generates more negative post-experience evaluations may appear inconsistent with prior findings indicating that asking customers to evaluate their satisfaction with a firm s past performance has a positive effect on their relationship with the firm (Dholakia and Morwitz 2002; Dholakia et al. 2005). That is, if one makes the assumption that expectations from a service encounter are based on prior experience with the service provider, then a task requiring customers to state their expectations should have a similar effect to a task involving evaluations of the firm s past performance. However, whereas stating (future) expectations generates negativity enhancement, past performance evaluations may trigger a process of confirming or seeking support for those

20 evaluations. In the next study, we examine whether the seemingly conflicting effects of considering future versus past performance can occur within the same study. 20 Method STUDY 4 The respondents in Study 4 were 201 supermarket shoppers, who received no compensation for their participation. They were told that the study was part of the research of university graduate students and were assured that their responses would remain anonymous. Participants were randomly assigned to one of four groups, including a control (n=52), a stated expectations group (n=49), and two groups in which respondents evaluated (before shopping) the supermarket s past performance. Those in the stated expectations group were asked to indicate your expectations regarding the service that you will receive during your visit today at the supermarket. They then provided (7-point scale) ratings on nine dimensions, including: (a) the politeness of the checkout employees, (b) the willingness of employees to assist shoppers, (c) the professionalism of employees, (d) the length of wait at the checkout, (e) the display arrangement, (f) the level of service, (g) overall satisfaction from the service, (h) overall satisfaction from the store visit, and (i) whether, after the visit to the supermarket, they would recommend to others to shop at that store. As indicated, there were two groups in which respondents evaluated the store s past performance before entering the store. Respondents in one of these two groups ( complete past evaluation; n=52) provided ratings on essentially the same nine items used in the stated expectations group. For each of the first seven (service-focused) items, these respondents were simply asked, Please evaluate this supermarket with respect to They were next asked What was your level of satisfaction in previous visits to this supermarket? Regarding the last item, respondents were asked, Would you recommend to your friends to shop at this supermarket?

21 21 The final group ( partial past evaluation; n=48) evaluated the supermarket s past performance (before entering the store) on just three general items: (a) the level of service, (b) satisfaction with service, and (c) overall satisfaction in previous visits to that store. Respondents in all four groups were asked the same questions after exiting the supermarket. First, they were asked to indicate everything they could recall from their just completed visit to the supermarket. Next, they evaluated the supermarket based on the service they received during their shopping visit on that day, using the exact same items as those used before entering the store in the stated expectations and complete past evaluation groups (including the three general items used in the partial past evaluation group). In addition, respondents indicated (on 7-point scales): (a) the degree to which they had paid attention to positive or negative aspects while shopping; (b) the degree to which they had expected to be interviewed after completing the store visit. Results Again, respondents in all four conditions indicated that they did not expect to be interviewed after completing their supermarket shopping (average ratings of on a 7-point scale). We next compared the stated expectations and the (past) supermarket evaluations (in the complete past evaluations group) before entering the store, which involved the same items and scales (see Figure). This comparison revealed that the expectations were statistically indistinguishable from the past evaluations. Specifically, for the combined service measure (for expected service, α=.95; for past service, α=.89), the average ratings were MExpected=5.2 and MPast=5.2 (F(1,99)<1, p>.1); for the combined satisfaction measure (for expected satisfaction, α=.86; for past satisfaction, α=.89), the average ratings were MExpected=5.2 and MPast=5.3 (F(1,99)<1, p>.1); and for the recommendation measure, the average ratings were MExpected=5.2

22 22 and MPast=5.1 (F(1,99)<1, p>.1). These findings are consistent with the assumption that a firm s past performance provides the basis for expectations regarding future performance. Also, on the comparable items, the complete and partial past evaluations of the store were statistically indistinguishable; that is, the service evaluations (t(98)=1.1, p>.1) and satisfaction evaluations (t(98)=1.0, p>.1) in the two groups were similar. Further, post-purchase evaluations of respondents in the complete and partial past evaluation groups were similar (F(1,197)<1, p>.1). Consistent with the earlier studies, the post-purchase evaluations in the stated expectations group were lower than their expectations (see Figure): for the combined service measure, MBefore=5.2 versus MAfter=4.3 (paired-t(48)=4.2, p<.01); for the combined satisfaction measure, MBefore=5.2 versus MAfter=4.2 (paired-t(48)=3.7, p<.01); and for the recommendation measure, MBefore=5.2 versus MAfter=4.1 (paired-t(48)=3.8, p<.01). Also, the post-visit evaluations were lower than the corresponding evaluations in the control group: for the combined service measure, MStated=4.3 versus MControl=5.1 (F(1,197)=11.0, p<.01); for the combined satisfaction measure, MStated=4.2 versus MControl=5.1 (F(1,197)=10.9, p<.01); and for the recommendation measure, MStated=4.1 versus MControl=4.8 (F(1,197)=3.7, p<.06). Remarkably, however, although the measured (pre-purchase) past evaluations were indistinguishable from the stated expectations, the former task had a very different effect on post-purchase evaluations. Unlike the (negative) effect of stating expectations, providing past evaluations of the store tended to produce more positive post-purchase evaluations. In the partial past evaluation condition, both the service (MPartial-past=5.3) and satisfaction (MPartial-past=5.6) post-purchase ratings were more positive than the corresponding pre-purchase (past) evaluations ( both M s=5.1)(paired-t(47)=2.0, p<.05 and paired-t(47)=2.7, p<.05, respectively). In the complete past evaluation group, post-visit service (M=5.5) was rated higher than (pre-visit)

23 23 service past evaluations ( M=5.2, paired-t(51)=1.9, p<.06). For the satisfaction (MPre-visit=5.3 vs. MPost-visit =5.5) and recommendation (MPre-visit=5.1 vs MPost-visit=5.3) measures, the differences were in the same direction but did not reach statistical significance (paired-t (51)=1.4 and 1.0, respectively; p>.1). Furthermore, post-visit evaluations in both the partial and complete past evaluation groups were higher than the post-visit evaluations of the stated expectation group. The contrasts between the stated expectation group with the average of the two past evaluation groups were all statistically significant (for service, F(1,197)=25.1, p<.01; for satisfaction, F(1,197)=32.5, p<.01; and for recommendation, F(1,197)=12.4, p<.01). In addition, across both past-evaluation groups, satisfaction ratings were higher than in the control condition (F(1,197)=3.9, p<.05). For service and recommendation, these contrasts had the same sign but were not statistically significant (F(1, 197)=1.5, p>.1 and F(1,197)=1.8, p>.1). We next examined the recall data from the shopping visit. The control and two past performance evaluation groups were statistically indistinguishable and remembered an average of more positive than negative aspects, whereas the stated expectations group remembered an average of 1.0 more negative than positive aspects. Responses to the closedended item regarding the focus of attention on negative (1) or positive (7) aspects during the visit led to a similar conclusion. The average rating in the stated expectations group was 4.43, compared to 5.23 in the control, 5.21 in the complete past evaluation group, and 5.29 in the partial past performance evaluation group; a contrast of the average of the control and the two past evaluation groups with the stated expectations group was statistically significant (p<.01). Discussion There are cases in which results that confirm and replicate seemingly conflicting prior results are surprising nonetheless, and Study 4 illustrates such a case. Although the present

24 24 research has shown that stating expectations tends to produce lower post-purchase evaluations whereas Dholakia and Morwitz (2002) showed that evaluating satisfaction has the opposite effect, we were skeptical that both effects can be demonstrated in the same study. However, Study 4 showed that pattern, while also confirming the assumption that (pre-purchase) past performance evaluations were indistinguishable from the stated expectations. Process data confirmed that the two tasks trigger different evaluation processes, perhaps reflecting the forward-looking perspective of stated expectations and backward-looking focus of past performance evaluations. It should be noted that the robustness of the positive effect of past evaluations on subsequent evaluations needs further examination, and we still do not fully understand the factors that underlie this effect. Still, the results of Study 4 suggest that evaluating past performance and stating expectations about future performance create different mindsets. GENERAL DISCUSSION Customers expectations are key determinants of their consumption experiences, satisfaction, and loyalty. Therefore, knowing in advance what customers expect is critical for the success of marketing strategies. The present research, however, suggests that measuring expectations can backfire, particularly if the stated expectations are still accessible during the consumption experience. Specifically, contrary to the standard assumption that expectations are either confirmed or disconfirmed based on any discrepancy between expectations and actual experience, the present research reveals a tendency for negative disconfirmation of stated expectations. Next, we discuss the theoretical and practical implications of the findings. Review of Findings and Theoretical Implications The conclusion that stating expectations before a consumption experience leads to lower evaluations of that experience appears robust. In the studies described above, those who had first

25 25 stated their expectations later reported more negative evaluations of their experiences, relative to both a control group and their own expectations. As shown in the follow-up to Study 1, this result occurred even in the case of a supermarket associated with negative expectations, which is an unusual situation, since customers tend to shop at establishments they prefer. Thus, the findings of this research suggest that stated expectations tend not to be met, and in all likelihood, the mere fact that expectations were explicitly stated almost guaranteed that post-experience evaluations would be lower than they would have been had expectations not been articulated. As discussed earlier, different theories could be potentially relied upon to predict and explain the effect of stating expectations on subsequent evaluations. An account based on confirmation bias was rejected, because positive stated expectations should have resulted in more positive post-experience evaluations, not more negative ones. And, there is clearly no positivity effect associated with measuring expectations. This conclusion is, in all likelihood, not due to the fact that our studies were presented to respondents as part of a university study that is not affiliated with the store being evaluated. In particular, using the notion of persuasion knowledge (Friestad and Wright 1994), Williams et al. (2004) showed that the mere measurement effect of intent questions might actually be stronger without the mention of any sponsor. However, it is quite possible that consumers are less likely to perceive a persuasive intent when asked to state expectations (or evaluations of past performance) than when they are asked to indicate their intention to buy. These differences in perceptions of and response to questions concerning intentions versus expectations can be further examined in future research. The theory that appears to account best for the observed findings, including the process measures, is based on the notion that stating expectations triggers an evaluation process in which negative aspects receive more attention and ambiguous dimensions are encoded more negatively

26 26 than in the absence of salient expectations. This negative bias goes beyond the generalization that negative aspects receive more attention (e.g., Kanouse 1984) and losses loom larger than gains (Tversky and Kahneman 1991), which should apply regardless of whether expectations were stated before experience. Instead, stating expectations appears to generate a more active, online evaluation process, which exacerbates the negative bias of evaluations (e.g., Ofir and Simonson 2001). That is, after stating their expectations, customers are more likely to consider how actual performance compares with those expectations, and this process is characterized by an emphasis on negative aspects. The negativity enhancement account can readily explain the finding that stating expectations leads to better recall of negative aspects and poorer recall of positive aspects of the consumption experience, under the reasonable assumption that recall reflects the focus of attention while shopping. However, the finding of Study 2 that stating expectations causes customers to view the same features more negatively suggests that the impact of stating expectations goes beyond focus of attention. In particular, negativity enhancement might involve a motivational component that is triggered by stating expectations, as if customers seek to find negative disconfirmation of their expectations. Although the present research does not provide much insight regarding such a process, a possible mechanism that can be explored in future research is that customers spontaneously or unconsciously use negative disconfirmation as a means to improve marketers performance. While not inconsistent with prior research (Dholakia and Morwitz 2002), perhaps the most surprising finding we observed was that of Study 4, in which stating expectations about future performance and evaluating past performance had the opposite effects on subsequent evaluations of a shopping experience. What we find particularly remarkable is that these