Hindsight Bias in the Context of a Consumption Experience

ABSTRACT - Hindsight bias occurs when the known outcomes of an event affect the memory for the prior expectations about the event. Hindsight bias can only occur when a discrepancy exists between the expectations and the experiences on an event. If hindsight bias occurs, the discrepancy between experiences and remembered expectations is less than the discrepancy between experiences and actual expectations. In a pretest-posttest design, hindsight bias was studied in the context of an involving, ambiguous consumption experience: attending an academic conference. Prior research suggests that under these circumstances a feedforward mechanism from expectations to experiences attenuates the discrepancy between expectation and experiences, hence reducing the likelihood that hindsight bias will occur. Nevertheless, both between- and within-subjects analyses reveal a consistent hindsight bias. The ability to recall prior expectations is reduced most clearly when the personal experiences are salient. Implications of the results for models of consumer decision making are discussed.

Citation:

Rik Pieters and Rami Zwick (1993) ,"Hindsight Bias in the Context of a Consumption Experience", in E - European Advances in Consumer Research Volume 1, eds. W. Fred Van Raaij and Gary J. Bamossy, Provo, UT : Association for Consumer Research, Pages: 307-311.

European Advances in Consumer Research Volume 1, 1993      Pages 307-311

HINDSIGHT BIAS IN THE CONTEXT OF A CONSUMPTION EXPERIENCE

Rik Pieters, Erasmus University, Rotterdam, and Nijenrode University, The Netherlands

Rami Zwick, Pennsylvania State University, U.S.A.

ABSTRACT -

Hindsight bias occurs when the known outcomes of an event affect the memory for the prior expectations about the event. Hindsight bias can only occur when a discrepancy exists between the expectations and the experiences on an event. If hindsight bias occurs, the discrepancy between experiences and remembered expectations is less than the discrepancy between experiences and actual expectations. In a pretest-posttest design, hindsight bias was studied in the context of an involving, ambiguous consumption experience: attending an academic conference. Prior research suggests that under these circumstances a feedforward mechanism from expectations to experiences attenuates the discrepancy between expectation and experiences, hence reducing the likelihood that hindsight bias will occur. Nevertheless, both between- and within-subjects analyses reveal a consistent hindsight bias. The ability to recall prior expectations is reduced most clearly when the personal experiences are salient. Implications of the results for models of consumer decision making are discussed.

INTRODUCTION

People are often unable to correctly remember their expectations of the outcomes of an event once the outcomes of the event become known. Typically, they tend to exaggerate in hindsight what they knew in foresight (Fischhoff, 1975; Fischhoff & Beyth, 1975). They tend to view what has happened as having been inevitable, and to view it as having appeared relatively inevitable to them even before it happened (Christensen-Szalanski & Willham, 1991; Hawkins & Hastie, 1990).

Research on this phenomenon, known as hindsight bias or the 'I-knew-it-all-along-effect', has focused on events that have objective outcomes, like the results of elections and sports games, the correct answers to almanac questions, whether or not people were laid off and so forth. Research on the incidence of hindsight bias when the outcomes are personal experiences is absent. This is remarkable since in real life, people constantly form expectations about their personal experiences of future events, like how cosy the atmosphere in a certain restaurant will be, how bitter a bar of chocolate will taste, or how humorous a movie may turn out to be.

EXPECTATIONS AND PERSONAL EXPERIENCES

Regular hindsight bias occurs when the expectations and the actual outcomes of an event are discrepant, and when the remembered expectations are in the direction of the outcomes. Hindsight bias is partly due to people's belief that the world is predictable, even controllable (Walster, 1967). When outcomes and expectations of an event are discrepant, and the outcomes are objective, the tension between the assumed predictability on the one hand, and the discrepancy on the other hand can be minimized by assimilating the remembered expectations in the direction of the outcomes. This process could be called 'backward assimilation', since the memory for the prior expectations is assimilated in the direction of the actual outcomes.

It is not obvious that hindsight bias occurs when the outcomes are personal experiences instead of objective results. When the outcomes of the event are personal experiences, i.e., subjectively interpreted results, a discrepancy between expectations and experiences can be minimized by assimilating the experiences in the direction of the expectations. This process could be called 'forward assimilation'. When forward assimilation occurs, the discrepancy between prior expectations and personal experiences is reduced, and thus the likelihood that backward assimilation occurs.

Several studies have shown the presence of forward assimilation. For instance, Olson and Dover (1979) exposed a group of homemakers, over a two-week period, to three adlike communications that claimed that a particular brand of ground coffee had no bitterness at all. Next, this experimental group and a control group, that had not been exposed to the communications, sampled the taste and aroma of the coffee. The coffee was actually prepared to be quite bitter. The analyses showed that people who had been exposed to the adlike communications, rated the actual bitterness of the coffee lower than people who had not been exposed to the messages.

Deighton and Schindler (1988) exposed subjects to a tape of a radio station containing four messages claiming that the music of the station was 'new'. Expectations concerning the 'newness' of the music of the advertised station and three other stations were assessed both before and after exposure to the tape. Two weeks after this, the actual experienced 'newness' of the music of the stations was assessed. The analyses indicate that exposure to the message of a particular radio station in combination with listening to the station boosted the experienced 'newness' of its music.

Hoch and Ha (1986) found that when making overall evaluations of brands, people tend to rely on the consumption experience itself (data-driven) when the experience is unambiguous, and that they tend to rely on prior opinions or advertising claims (hypothesis-testing) when the experience is ambiguous.

So, expectations may directly, or in combination with other variables like the amount of experience, affect the content of the personal experience. The distinction between forward and backward assimilation is presented in Figure 1.

Backward assimilation has been observed in situations where the outcomes are objective (Hawkins and Hastie, 1990). Forward assimilation is more likely to occur when the outcomes are subjective experiences instead of objective outcomes. For example, it is easier to report that the coffee does not taste very bitter (while its bitterness is in fact high) than that President Bush lost the elections (while in fact he did not). When forward assimilation occurs, the discrepancy between expectations and personal experiences is reduced, and as a consequence, the likelihood that hindsight bias occurs is reduced as well.

In sum, it is possible, but by no means obvious, that hindsight bias occurs when subjects try to remember the expectations they had before they personally experienced some consumption event.

The presence of hindsight bias in the case of personal experiences may have implications for models of consumer decision making that focus on a comparison of experiences with prior expectations, e.g., models of consumer satisfaction (Oliver and Swan, 1989), or of perceived service quality (Zeithaml, Berry and Parasuraman, 1988). If after experiencing an event, people tend to believe that they expected the experience, while in fact they didn't, the felt discrepancy between expectation and experience is less than the actual discrepancy. In that case, the issue is whether the perceived service quality or felt satisfaction of consumers is determined by the discrepancy between prior expectations and experiences, by the discrepancy between hindsight expectations and experiences, or by both.

FIGURE 1

FORWARD AND BACKWARD ASSIMILATION

TABLE 1

RESEARCH DESIGN

DOMAIN AND HYPOTHESES

The goal of the present study is to analyse the incidence of hindsight bias when the outcomes of an event are relevant personal experiences. The study was conducted in the context of the 1990 conference of the Association for Consumer Research (ACR) in New York. The conference was chosen because it constitutes a real, involving consumption experience with few tangible, objective outcomes. Outcomes of the conference are personally experienced by the participants, as Holbrook and Thayer (1985, p.598) explain:

'Conference attendees ... are at heart consumers. They seek experiences that they will find pleasurable or informative or, even better, both. The expectation of such rewards keeps them in the audience, returning year after year, forever in search of truth and enlightment.'

ACR members were surveyed either before and after, or only after the conference in New York. Their expectations, experiences and remembered expectations were assessed.

We hypothesized that the best recall of prior expectations about the conference would be demonstrated when relevant intervening experiences are absent, i.e., compared to ACR members that attended the conference, ACR members that did not attend the conference should be better able to exactly remember after the conference which expectations they had before the conference. We expected that backward assimilation would occur when personal experiences would be present, i.e., ACR members that attended the conference were expected to express a hindsight bias.

Salient information stands out because it is novel, unexpected, complex or goal-related. It tends to have a larger impact than non-salient information in decision making, due to its increased accessibility when the actual decisions are made (Fiske and Taylor, 1991). Hell et al. (1988) found in their study that poor memorability of the original expectations is a necessary condition for hindsight bias to occur. This lead us to the next hypothesis, that the incidence of hindsight bias is higher when the personal experiences are salient when people try to recall their prior expectations.

METHOD

Subjects

A total of 65 members of ACR participated in the study. Subjects were randomly sampled from the 1990 ACR membership directory and from the provisional program of the conference to ensure that a sufficient number of subjects in the sample actually attended the conference. Forty-six subjects in the final sample attended the conference, nineteen did not.

Design

Table 1 presents the research design. Forty six subjects participated in a telephone survey ('pretest') conducted one to two days before the start of the conference. A total of 27 subjects planned to attend the conference (groups 1 and 2), while 19 planned not to attend the conference (group 5). All were asked to indicate what their expectations of the conference were. One to two days after the conference all subjects in the pretest, and an additional 19 subjects who attended the conference but who were not included in the pretest, were asked to participate in the posttest. In the posttest, conducted by telephone, the personal experiences, if relevant, and hindsight expectations were measured. Although some consumer researchers were somewhat reluctant to participate in the present consumer study, only two actually refused.

The salience of the personal experiences was manipulated by the order in which the questions in the posttest were asked. One group of attendees (group 1) first answered questions about their experiences during the conference before answering questions about their prior expectations. This was expected to lead to increased salience of the personal experiences when remembering the prior expectations. A second group of subjects (group 2) received the reverse question order.

Groups 3 and 4 were included in the design to detect possible sensitization effects. A sensitization effect occurs when the expectations or experiences are affected, either in strenght or in salience, by their mere measurement (Lana, 1969; Feldman and Lynch, 1988). Sensitization threatens the internal validity of the research. Sensitization may express itself as a difference in the personal experiences of people who are and people who are not included in the pretest. Sensitization may also express itself as a difference in the experiences of people who first answer questions about their remembered expectations, before answering questions about their experiences, as compared to people who have the reverse question order. To analyse the presence of sensitization effects and to test for the existence of hindsight bias in a between-subjects manner, groups 3 and 4 were included in the design.

TABLE 2

MEAN RESPONSES IN THE 5 GROUPS

Dependent variables

In the sequel, the expectations indicated prior to the conference in the pretest will be called 'foresight expectations', while the remembered expectations as indicated in the posttest will be called 'hindsight expectations.' The foresight expectations were:

a. 'What do you expect that the average quality of the papers will be?'

b. 'What do you expect that the average quality of the presentations will be?'

c. 'What are the chances that you will be exposed to really new ideas during the conference (if you would go)?'

d. 'What are the chances that you will make new contacts during the conference (if you would go)?'

e. 'What are the chances that the conference will be very rewarding or worthwhile to you (if you would go)?'

Responses to items a. and b. could be indicated on 11-point scales from 'very good' (0) to 'very bad' (10). Responses to the other expectations could be indicated on 11-point scales from 'very unlikely' (0) to 'very likely' (10).

After the conference, the personal experiences of the attendees and the hindsight expectations of all subjects were assessed on the same aspects as in the pretest. The experience questions were formulated as follows: 'What was in your judgement ... [the average quality of the papers at the conference]?'

Before asking the hindsight questions, subjects who had participated in the pretest were told: 'We would like to know to what extent people are able to remember the expectations that they had before an experience. I will read to you the same questions we asked you last week. Please try to recall the expectations you had before the conference.' Subjects who had not participated in the pretest were told: 'I will ask you several questions about the expectations that you had before the conference. Please try to recall your expectations when answering the questions.'

The hindsight questions were formulated as follows: 'What did you expect that ... [the average quality of the papers at the conference would be]?' The response scales were the same as in the pretest. In the pretest, before the conference, several other relevant questions were included. The interviewers were graduate students who were unaware of the design and the hypotheses of the study.

RESULTS

In Table 2 the mean foresight expectations, experiences, and hindsight expectations are presented for each of the five groups involved in the study.

To test for hindsight bias in a between-subjects design it is necessary that the experiences of groups 1 and 2 on the one hand and of groups 3 and 4 on the other hand are the same. If these experiences are the same it can be assumed that their expectations before the conference were the same too, and only than it is meaningful to compare groups 3 and 4 hindsight expectations with groups 1 and 2 foresight expectations. The occurrence of a sensitization effect on the experiences would hinder between-subject analyses of hindsight bias. To test whether inclusion in the pretest sensitized the experiences of the conference attendees, a 2 (Pretest) x 2 (Question Order) MANOVA on the personal experiences was performed. Sensitization effects are absent: neither the main effects nor the interaction were statistically significant (F(5,36) = 0.977; F(5,36) = 1.272; F(5,36) = 1.127; for the pretest, question order, and the interaction respectively).

To test for hindsight bias, a discrepancy between foresight expectations and personal experiences has to be present. Without a discrepancy there can be no hindsight bias. For groups 1 and 2, the discrepancy between foresight expectations and personal experiences was analysed. For all items, except for 'Quality of Presentations' in group 1, the average experience rating is below expectations (see Table 1) ('subtractive disconfirmation': Tse and Wilton (1988)). The average absolute discrepancy between foresight expectations and personal experiences, aggregated across subjects (group 1 and 2) and items was 1.23 (n = 131). Although this discrepancy is quite small, the overall difference between foresight expectations and personal experiences, over all five items, is statistically significant (F(5,20) = 8.55, p < 0.001). Univariate analyses by item (for groups 1 and 2 combined) reveal significant differences on the items 'New Ideas', 'New Contacts', and 'Quality of Papers' (t = -3.63, p < 0.002; t = -3.36, p < 0.003, and t = -2.69, p < 0.02, respectively), a marginally significant difference for item 'Very Rewarding' (t = -1.80, p < 0.08), and no significant difference for 'Quality of Presentations' (t = 0.53).

TABLE 3

FREQUENCY DISTRIBUTION OF MEMORY TYPES IN GROUP 1 AND 2

The overall discrepancy between foresight expectations and personal experiences is reflected in the discrepancy on the level of individual subjects and items. Out of 27 subjects in groups 1 and 2 combined, respectively 22, 15, 24, 22, and 23 rated their experiences below expectations on the items 'Quality of Papers', 'Quality of Presentations', 'New Ideas', 'New Contacts', and 'Very Rewarding'. These results show that although the discrepancy between foresight expectations and personal experiences of the conference was very small for most subjects in groups 1 and 2, the personal experiences did fell significantly below the foresight expectations on most items. This means that there is significant but limited room for hindsight bias to occur.

Memory for expectations

It was expected that the best recall of expectations about the conference would be achieved when relevant intervening experiences are absent (group 5), and that the more salient the overall consumption experiences are when consumers try to recall their expectations after the event, the lower the incidence of correct recall will tend to be (group 1 vs. 2).

First, the exact recall of the foresight expectations, in hindsight was determined. Exact recall is demonstrated when the difference between foresight expectation and hindsight expectation is zero. The best recall of expectations was demonstrated by groups 2 (question order: hindsight - experience), and 5 (when relevant intervening experiences are absent). Seven out of 13 subjects in group 2, and 11 out of 19 subjects in group 5 correctly recalled 3 or more out of 5 foresight expectations. The average numbers of correct recalls in groups 2 and 5 were not significantly different from each other, being respectively 2.53 and 2.37 (F(1,44) = 0.19, p > 0.66). In contrast, the average number of correct recalls was 1.53 in group 1. This average is significantly different from both group 2 (F(1,44) = 5.58, p < 0.02), and group 5 (F(1,44) = 4.57, p < 0.04). Only one subject out of 14 in group 1 correctly recalled 3 or more foresight expectations.

These results confirm the hypothesis that the presence of relevant experiences can interfere with the ability to recall prior expectations. Furthermore, we found that the more salient the intervening experiences are, the higher the interference is. However, these results do not determine the nature of the memory errors. Subsequent analyses provide a closer look at the nature of the memory errors that were made.

Hindsight bias: a between-subjects analysis

The foresight expectations of groups 1 and 2 were compared with the hindsight expectations of groups 3 and 4. If hindsight bias exists, then the hindsight expectations of groups 3 and 4 will fall below the foresight expectations of groups 1 and 2 since the bias is in the direction of the experiences.

As expected, on four of the five experience items (except 'Quality of Papers'), the average hindsight expectations of groups 3 and 4 are below the average foresight expectations of groups 1 and 2 (see Table 1). These differences are significant for items 'New Ideas', 'New Contacts', and 'Very Rewarding' (F(1,44) = 4.27, p < 0.05; F(1,44) = 11.03, p < 0.002; and F(1,44) = 4.79, p < 0.04, respectively). The difference on the other two items is not significant.

Hindsight bias: a within-subjects analysis

The incidence of hindsight bias was investigated in more detail employing a within-subjects analysis, using the procedure described below.

First, for each individual and for each of the 5 experience items, the absolute discrepancy (ABSDIFF) between the foresight expectation and the reported personal experience was determined. Without a discrepancy between foresight expectation and actual experience, there can be no hindsight bias.

Next, three memory types were distinguished for all cases in which the absolute difference of the discrepancy between foresight expectation and personal experience was larger than zero (ABSDIFF > 0). The memory type is 'Recall' if the hindsight expectation is identical to the foresight expectation. The memory type is 'Hindsight Bias' if the hindsight expectation falls between the foresight expectation and the personal experience. The memory type is 'Other' if the memory type is neither 'Recall' nor 'Hindsight Bias'. The memory type 'Other' is found when the hindsight expectation falls outside the range formed by the foresight expectation and the personal experience. Table 3 presents the frequency distribution of the three memory types as a function of ABSDIFF.

The null hypothesis was tested that the frequency of hindsight bias is due to responding randomly to the hindsight expectations questions. For each ABSDIFF the expected number of hindsight biases (under the null hypothesis, and given the 11 point scale we used) is ALL x (ABSDIFF/11). The observed frequencies of hindsight bias (17, 9, and 11 for ABSDIFF = 1, 2, and 3 respectively) are significantly higher than expected (5.18, 2.90, and 4.09 for ABSDIFF = 1, 2, and 3 respectively) (tests for proportions: z = 4.44, p < 0.001; z = 3.96, p < 0.05; z = 4.02, p < 0.05, for ABSDIFF = 1, 2, and 3, respectively). The point-biserial correlation between ABSDIFF and memory type, in which hindsight bias (HB) was compared with the other two memory types (Recall and Other) taken together, was .43, p < .000. A logit analysis with memory type as the dependent variable and ABSDIFF as the independent variable gave similar results. These results show that hindsight bias occurs, and that the higher the discrepancy between foresight expectations and actual experiences, the higher the likelihood of its occurrence.

CONCLUSION AND DISCUSSION

The results confirm that personal experiences can interfere with the ability to correctly recall foresight expectations. Typically, there is a tendency to bias hindsight expectations in the direction of the personal experiences. This was found both in between- and within-subjects analyses. The hindsight expectations of subjects in groups 3 and 4 were significantly lower, on most items, than the foresight expectations of subjects in groups 1 and 2 although both groups reported similar experiences at the conference. This indicates the presence of hindsight bias, since for most subjects the reported experiences were below expectations. Similarly, the memory errors on the individual level were in line with the hindsight bias phenomenon.

Subjects for whom the personal experiences were salient in the posttest, due to the question-order, expressed less recall, than subjects for whom the personal experiences were not salient. These results are not due to sensitization effects, as subjects who were included in the pretest did not differ from subjects who were not included in the pretest on their personal experiences during the conference. Also, question-order in the posttest did not affect the reported experience of the conference.

Although statistically significant, the average absolute discrepancy between foresight expectations and personal experiences, aggregated across subjects was very small. This indicates that indeed, for the most part, conference attendees 'Knew it all along'. Despite this small discrepancy, a consistent hindsight bias was present. Conference attendees felt even stronger that they knew it all along, than they actually did. Although there was hardly any room to assimilate their hindsight expectations in the direction of their personal experiences, our subjects tended to do so anyway. This underlines the robustness of hindsight bias.

Hindsight bias may have interesting implications for models that assume that consumers are motivated and able to recall their foresight expectations when or after they have experienced an event. For instance, the dominant model of consumer satisfaction is the expectancy-disconfirmation paradigm (Oliver, 1980; Oliver and Swan, 1989). In this paradigm, (dis)satisfaction is determined by the level of prior expectations, the level of product performance, and the degree to which expectations are positively or negatively disconfirmed during consumption. The model rests on the assumption that foresight expectations are still available, in an unbiased form, when or after consumers experience the performance of a product. Otherwise a 'true' comparison of the foresight expectations with the personal experiences is not possible.

If, as we have found, after experiencing the product or service, consumers tend to falsely believe that they knew the outcomes in foresight, how than can they properly compare their prior expectations with their experiences to arrive at an overall satisfaction / dissatisfaction? Also, if the discrepancy between personal experiences and hindsight expectations is smaller than the discrepancy between foresight expectations and experiences, what should be measured in research on consumer satisfaction: foresight expectations, or hindsight expectations, actual expectation-disconfirmation, or felt (after the fact) expectation-disconfirmation? By answering these and related questions, more insight in consumer decision making processes, and the role of memory in these processes will be gained.

In the present study, backward assimilation was focused upon. In future research the joint effects of forward and backward assimilation in the context of personal experiences might be fruitfully studied. Also, the conditions that determine the presence and the strength of forward and backward assimilation may deserve attention.

REFERENCES

Christensen-Szalanski, J.J. and C.F. Willham (1991), The Hindsight Bias: A Meta Analysis. Organizational Behavior and Human Decision Processes 48, 147-168.

Deighton, J. and R.M. Schindler (1988), Can advertising influence experience. Psychology and Marketing 5, 103-115.

Feldman, J.M. and J.G. Lynch (1988), Self-generated validity and other effects of measurement on belief, attitude, intention and behavior. Journal of Applied Psychology 73, 421-435.

Fischhoff, B. (1975), Hindsight _ Foresight: The effect of outcome knowledge on judgement under uncertainty. Journal of Experimental Psychology: Human Perception and Performance 1, 288-299.

Fischhoff, B. and R. Beth-Marom (1975), "I knew it would happen" - Remembered probabilities of once-future things. Organizational Behavior and Human Performance 13, 1-16.

Fiske, S. and S. Taylor (1991), Social Cognition. New York: Random House, 2nd ed.

Hawkins, S.A. and R. Hastie (1990), Hindsight: Biased judgment of past events after the outcomes are known. Psychological Bulletin 107, 311-327.

Hell, W., G. Gigerenzer, S. Gauggel, M. Mall and M. Mueller (1988), Hindsight Bias: an interaction of automatic and motivational factors. Memory and Cognition 16, 533-538.

Hoch, S.J., and Y. Ha (1986), Consumer learning: Advertising and the ambiguity of product experience. Journal of Consumer Research 13, 221-233.

Holbrook, M.B. and E.L. Thayer (1985), Casey at the conference: Some reflections on the ACR experience. In: E.C. Hirschman and M. Holbrook (Eds.), Advances in Consumer Research 12. Provo, UT: Association for Consumer Research, 598-600.

Oliver, R. (1980), A cognitive model of antecedents and consequences of satisfaction decisions. Journal of Marketing Research 17, 460-469.

Oliver, R.L. and J.E. Swan (1989), Equity and disconfirmation perceptions as influences on merchant and product satisfaction. Journal of Consumer Research 16, 372-383.

Olson, J.C. and Ph.A. Dover (1979), Disconfirmation of consumer expectations through product trial. Journal of Applied Psychology 64, 179-189.

Tse, D.K. and P.C. Wilton (1988), Models of consumer satisfaction formation: an extension. Journal of Marketing Research 25, 204-212.

Walster, E. (1967), "Second Guessing" important events. Human Relations 20, 239-249.

Wasserman, D., R.O. Lempert, and R. Hastie (1991), Hindsight and causality. Personality and Social Psychology Bulletin 17, 30-35.

Zeithaml, V.A., L.L. Berry and A. Parasuraman (1988), Communication and control processes in the delivery of service quality. Journal of Marketing 52, 35-48.

----------------------------------------