Testing For Perceptual Underestimation of 9-Ending Prices

Robert M. Schindler, Rutgers University-Camden
Thomas Kibarian, Innovation and Information Consultants, Inc.
ABSTRACT - Although perceptual underestimation of 9-ending prices is one of the most common explanations for the tendency of price setters to choose retail prices which end in the digit 9, two studies using an immediate recall task fail to provide evidence for the existence of any substantial perceptual underestimation. However, the pattern of recall errors found in these studies does suggest that the rightmost digits of a price receive less consumer attention than the leftmost digits. These results direct further research toward more sophisticated conceptions of price perception and how it may be affected by the price setter's choice of rightmost digits.
[ to cite ]:
Robert M. Schindler and Thomas Kibarian (1993) ,"Testing For Perceptual Underestimation of 9-Ending Prices", in NA - Advances in Consumer Research Volume 20, eds. Leigh McAlister and Michael L. Rothschild, Provo, UT : Association for Consumer Research, Pages: 580-585.

Advances in Consumer Research Volume 20, 1993      Pages 580-585

TESTING FOR PERCEPTUAL UNDERESTIMATION OF 9-ENDING PRICES

Robert M. Schindler, Rutgers University-Camden

Thomas Kibarian, Innovation and Information Consultants, Inc.

ABSTRACT -

Although perceptual underestimation of 9-ending prices is one of the most common explanations for the tendency of price setters to choose retail prices which end in the digit 9, two studies using an immediate recall task fail to provide evidence for the existence of any substantial perceptual underestimation. However, the pattern of recall errors found in these studies does suggest that the rightmost digits of a price receive less consumer attention than the leftmost digits. These results direct further research toward more sophisticated conceptions of price perception and how it may be affected by the price setter's choice of rightmost digits.

INTRODUCTION

Retail price setters show a marked tendency to set prices just below round numbers (Friedman 1967; Kreul 1982) and in particular to choose prices which end in the digit 9 (Twedt 1965). This pricing technique has often been termed "psychological pricing," apparently because of the belief that 9-ending prices act on the consumer via "special psychological effects" (Mason and Mayer 1990, p. 442).

Although a number of such special psychological effects of 9-ending prices have been hypothesized, the one which is probably the most common and persistent is that a 9-ending leads consumers to perceptually underestimate the level of a price (e.g., Georgoff 1972, pp. 4-6; Lambert 1975; Schindler and Warren 1988; Simon 1989, p. 183). This will be termed the underestimation hypothesis. This view maintains that consumers' tendency to minimize effort often leads to the use of truncation rather than rounding when perceiving prices. Thus, a consumer would perceive a price such as $29.99 by dropping off the rightmost 9's and thinking of it as "around $20" or "twenty and some dollars" rather than rounding it up to $30.

The underestimation hypothesis is certainly not the only plausible psychological mechanism for an effect of 9-ending prices. It is possible that 9-endings carry a symbolic meaning to the consumer (Schindler 1991) or that consumers use round numbers as cognitive reference points and thus perceive a 9-ending price as one where the price setter has knocked off a penny (Kreul 1982). However, the commonness and persistence of the underestimation hypothesis suggests that it deserves some careful examination.

The difficulty of measuring perception directly has often led memory measures to be used as surrogates of perception in advertising research (e.g., Ostlund 1978). A similar approach might be useful for testing the underestimation hypothesis. At this early stage of research, a recall measure would be more appropriate than a recognition test since it would not require specifying ahead of time expectations as to the exact underestimated price each consumer will perceive.

The presumption of the use of a recall measure is that when consumers recall a price which they have seen, they will recall it as they perceived it. If the consumer's perceptual processes fail to encode one or more of the rightmost digits of a price and instead represent them with a default value such as zero or some other low number, then the perception of a 9-ending price would be an underestimate of the true price. That underestimate is then what the consumer would be expected to recall. Thus, if the underestimation hypothesis is true, the average recalled price for a 9-ending price should be considerably more than one cent lower than the average recalled price for the equivalent (i.e., one-cent higher) 0-ending price.

Schindler and Wiman (1989) have conducted such a recall test and have found that at least some types of 9-ending prices will tend to be underestimated when they are recalled. However, this result is unconvincing as evidence for the underestimation hypothesis because Schindler and Wiman tested the subjects' price recall a full two days after they were exposed to the price. This delayed recall procedure raises the possibility that the subjects may have perceived the 9-ending prices accurately, but then dropped off the rightmost digits over the two-day period before the recall test.

More convincing evidence for the underestimation hypothesis would be the demonstration that 9-ending prices are underestimated relative to the equivalent 0-ending prices even when a price is recalled within a few minutes of having been seen. This immediate recall procedure does not remove the possibility that rightmost digits are being dropped during the period following the perception of the price. But reducing the post-perception period to minutes versus days would be expected to sharply reduce the degree to which this would occur. Demonstration that the underestimation of 9-ending prices persists under immediate recall conditions would provide much stronger evidence that this underestimation occurs at the time of perception rather than at some later time.

STUDY 1

This study was an attempt to replicate the Schindler and Wiman finding but using an immediate recall measure rather than delayed recall. Three pairs of advertisements were prepared so that one member of the pair displayed a 0-ending price while the other member of the pair was identical except that it displayed the 9-ending price one cent lower than the 0-ending price. Each subject was shown one of these six advertisements and, within three minutes, was asked to recall the price which was contained in that ad. The underestimation hypothesis would predict that the mean recalled price for the 9-ending prices would be substantially lower than the mean recalled price for the 0-ending prices.

Method

Design. Each subject saw one of three different advertisements in either its 9- or 0-ending form. Thus, the three-level between-subjects variable, Advertisement, was crossed with the two-level between-subjects variable, Price Ending. Recalled price was the dependent variable.

Although the purpose of the experiment was to determine whether 9-ending prices are underestimated, the 0-ending condition was included in order to control for the possibility of a general tendency to underestimate prices.

Subjects. The participants in this study were women recruited from church groups and PTAs in middle-income suburban neighborhoods. Data were collected from 145 individuals, but four subjects failed to follow instructions and their data were discarded. Thus, the results reported here are based on 141 subjects.

Materials. Each subject received a packet consisting of two booklets attached by a paper clip. The first booklet consisted of two pages. The first page was a cover sheet with introductory instructions. The second page showed a reproduction of a newspaper advertisement. Below the ad was a question which asked whether the price in the ad was "in line with what stores in this area would charge for this item." The subjects responded by checking one of three possible responses. The sole purpose of this question was to insure that the subjects look at the price in the advertisements.

TABLE 1

MEAN RECALLED PRICES IN EACH CONDITION OF STUDY 1

Each of the three ads used featured only one product and showed only one price. The ads were selected from among real newspaper ads which featured a product of some relevance to middle-class women and which prominently displayed a price. An attempt was made to choose a set of advertised products which represented the diversity of items likely to be of interest to the subjects. One of the three ads selected featured a dress, another featured a sports coat, and the third featured patio furniture. To reduce the possibility that subjects might happen to know the advertised price before seeing the ad, the ads were selected from out-of-town newspapers and were chosen so that both the stores and (if identified) the brands advertised lacked national reputations.

Each ad was altered to produce two versions: one with a 0-ending price, and one with the corresponding 9-ending price (one cent less than the even price). For example, in one version of the patio furniture ad, the price was $300.00. In the other version, otherwise identical to the first, the price was $299.99. In order to prevent the possibility that the price in one version might look more realistic than the other ad's price, the price was redrawn in both versions of the ad.

Since there were two versions of each of three ads, there was a total of six ads: the dress at $49.99, the dress at $50.00, the sports coat at $119.99, the sports coat at $120.00, the patio furniture at $299.99, and the patio furniture at $300.00. Since only one of these ads appeared on the second page of the first booklet, there were six versions of the first booklet.

The second booklet contained one of the three advertisements that had been used in the first booklet. However, for this booklet, these ads were altered so that the price was replaced by a blank line. Each subject's second booklet contained the same advertisement that her first booklet contained. For example, if a subject's first booklet had the advertisement for the dress at either $49.99 or $50.00, her second booklet had the same advertisement for the dress, but without the price. The subjects were asked to write, on the blank line, the exact price they had seen in the first booklet.

This second booklet included a cover sheet and was sealed to prevent the subjects from seeing its contents while they were viewing the first booklet. Also, no one was allowed to begin the second booklet until everyone completed the first booklet and it had been collected. This prevented subjects from looking back at the first booklet as they were responding to the second one.

Procedure. The subjects were run in groups ranging in size between twenty and fifty. The six different packet types were distributed randomly to the subjects in each group, so as to insure random assignment to the experimental conditions. Each group was told that the study concerned "how a store which is entering a new area should advertise." Subjects were told that they would see an advertisement for a store which they were probably not familiar with and should "look at this ad as you would if you saw it in your local newspaper," and answer the question below the ad. The subjects were given no indication that they would later be asked to recall anything concerning the ad they were about to see.

All subjects in a group began the first booklet at the same time. After everyone completed the first booklet, it was collected. The subjects were then told about the contents of the second booklet and were instructed to write, on the blank line, the "exact price" they had seen in the first booklet's ad. In all groups, the time between having seen the ad containing the price in the first booklet and writing the recalled price into the second booklet was less than three minutes.

Results

The mean recalled prices in each of the six experimental conditions can be seen in Table 1. A two-way, between-subjects analysis of variance indicated that there was a highly significant main effect of Advertisement F[2,135] = 10876.0, p < .001), no significant main effect of Price Ending (F[1,135] < 1), and no significant interaction between Price Ending and Advertisement (F[2,135] < 1). Replicating this analysis with both the actual and percent recalled-actual price differences also yields no significant Price Ending effect or Price Ending by Advertisement interaction (all F's < 1). Clearly, there is no substantial underestimation of the 9-ending prices in this immediate recall task.

However, while an analysis of the recalled prices indicates that there was no substantial underestimation of the 9-ending prices, examination of the recall errors reveals evidence for a considerable tendency to drop off the cents digits of the 9-ending prices (see Table 2). Of the 71 subjects who saw the 0-ending prices, 2.8% recalled a price which was an underestimate of the price they had seen. By contrast, of the 70 subjects who saw the 9-ending prices, far more (55.7%) recalled an underestimate of the actual price (X2[1] = 47.8, p<.001). Further, almost all of the underestimates produced by the 9-ending subjects involved the cents digits of the price. Twenty-eight percent of the 9-ending subjects dropped four cents from the correct price (e.g., recalling $49.99 as $49.95), 16% dropped 99 cents from the correct price (e.g., recalling $49.99 as $49.00), and 10% dropped either one to three cents or 5-49 cents from the correct price. Interestingly, not one of the 70 subjects who saw a 9-ending price recalled the even price which was one penny higher.

TABLE 2

DISTRIBUTION OF RECALLED PRICES FOR 9- AND 0- ENDING PRICES

Although the majority of the 9-ending subjects underestimated the cents digits of the price, the mean extent of this underestimation was only 35 cents. It is easy to see how this small difference would be overwhelmed by the small number of larger recall errors which were made and thus fail to substantially affect the mean recalled prices.

Analysis of the overestimation errors must remain very limited, since only five such errors were made. Four of these five overestimation errors occurred to the $119.99/$120.00 prices. Two subjects recalled $120.00 as $129.00, one subject recalled $120.00 as $150.00, and one subject recalled $119.99 as $199.00. The fifth overestimation error consisted of $50.00 being recalled as $50.98. Note that the extreme leftmost digit was recalled correctly in each case, and that it appears that the digit 9 occurs more often than chance would predict in the recall errors for the rightmost digits.

Discussion

This study failed to support the prediction of the underestimation hypothesis that there would be substantially greater underestimation of 9-ending prices than there would be for the equivalent 0-ending prices. However, the pattern of errors, overestimation errors as well as underestimation errors, does offer a bit of support for the underestimation hypothesis. This pattern of errors indicates that the leftmost digits of the prices were recalled more accurately than the rightmost digits. This is consistent with the possibility that consumer perceptual processes give more attention to the leftmost digits of a price.

STUDY 2

The results of Study 1 may have been influenced by the artificiality of the experimental situation. The subjects were not actually shopping for the items pictured in the ads, and were aware that they were participating in a consumer behavior study as they examined the advertisements.

Study 2 was an attempt to remedy these problems by investigating immediate recall of prices in a field setting. As has been done with past research (e.g., Conover 1986; Dickson and Sawyer 1990), a grocery shopping situation was found to be an appropriate and convenient setting. Supermarket shoppers were stopped as they were leaving the store and asked the prices of some of the items they had just purchased. The underestimation hypothesis would predict greater underestimation of those items whose prices ended in the digit 9 than those items whose prices did not end in the digit 9.

Method

Design and materials. The study focused on the recall of prices in twelve categories of grocery products (coffee, milk, eggs, flour, canned tuna, margarine, sugar, jelly, apple juice, trash bags, toilet paper, and dish detergent). The twelve categories selected were commonly purchased and contained a mix of 9-ending and non-9-ending prices at the time of the study (since very few supermarket prices ended in the digit 0, all prices ending in the digits 0-8 were grouped together to form the alternative to 9-ending pricing). Several large signs listing each of these twelve product categories were prepared.

Sample. The respondents were shoppers of a major supermarket chain in a large metropolitan area. They were selected at random from shoppers leaving the store. Shoppers were sampled at eight different outlets of the chain on seventeen separate occasions. These occasions included daytime, evening, and weekend time periods.

A total of 1511 recalled prices were obtained. The vast majority of the respondents provided data on only three or fewer items. However, due to an error, the number of prices recalled per respondent was not systematically recorded. Thus, for the purposes of statistical analysis, it was necessary to assume that the 1511 recalled prices approximated independent observations. It was judged that such an approximation would not interact meaningfully with the hypotheses of the study.

Procedure. The interviewer approached shoppers leaving the store and asked if they had purchased any of the products listed on the sign. If so, they were asked if they would be willing to answer a few brief questions about these purchases. Approximately two-thirds of the eligible shoppers who were approached agreed to participate.

Each respondent was asked to recall the exact price of the items which he or she had purchased which were among the product categories listed on the sign. When the recalled prices had been recorded by the interviewer, the respondent was asked to find the actual item purchased in the shopping bag so that the brand and size could be recorded by the interviewer and so the actual price of the item could be determined.

The respondents were first exposed to the prices of the grocery items by the shelf tags. They were again exposed to the prices when they appeared on the checkout screen as the items were scanned. Most interviews occurred within four minutes of this second exposure.

Results

The mean actual and recalled prices for the 9-ending prices and the non-9-ending prices can be seen in Table 3. T-tests showed no significant difference in recall between the 9- and non-9-ending prices whether recalled-actual differences (t[1509] < 1) or percent recalled-actual differences (t[1509] < 1) were used.

TABLE 3

MEAN ACTUAL AND RECALLED PRICES IN STUDY 2

TABLE 4

DISTRIBUTION OF RECALLED PRICES FOR 9- AND NON-9- ENDING PRICES

The pattern of recall errors of the 9- and non-9-ending prices can be seen in Table 4. There was no significantly greater tendency to produce an underestimate for 9- than for non-9-ending prices (X2[1] < 1) in this immediate recall task. The percentage of recall underestimates for the 9-ending prices (18.4%) was higher than the percentage of recall underestimates for the forty-seven 0-ending prices which occurred in the data (12.8%). However, because of the small number of 0-ending prices, this difference was not statistically significant (X2[1] < 1).

Although there were no significant differences between the likelihood of underestimating 9- and non-9-ending prices, there was a greater likelihood of producing an overestimate in the recall of non-9-ending prices than in the recall of 9-ending prices (X2[1] = 40.5, p < .001). Further examination of this difference showed it to be due entirely to a greater likelihood of overestimates of 9 cents or less (2.3% of all 9-ending recalls vs. 19.2% of all non-9-ending recalls). In fact, most (74%) of these incremental overestimations were due to the subjects recalling the extreme rightmost digit of the non-9-ending prices as the digit 9 rather than as one of the digits 0 through 8. It is the small average size of these incremental overestimations (4.3 cents) which led them to be overwhelmed by the small number of much larger errors and thus fail to substantially affect the mean recalled price.

Discussion

The results of Study 2 agree with those of Study 1 in that they provide no evidence that 9-ending prices are substantially underestimated in an immediate recall task. These results also agree with those of Study 1 in the finding of a nonrandom pattern of recall errors. However, the nature of the nonrandom pattern of errors differs between the two studies. In Study 1, the differences between 9- and 0-ending prices concerned the proportion of underestimates. In Study 2, the differences between 9- and non-9-ending prices concerned the proportion of overestimates.

There are numerous differences between the two studies which may have contributed to these differences in the distribution of recall errors. It is likely that the small proportion of 0-ending prices in the Study 2 data and the nonexistence of prices ending in the digits 1 through 8 in Study 1 strongly influenced the overall error distribution. It is also possible that the greater length of the prices in Study 1 than in Study 2 (4-5 digits vs. 2-3 digits) may have played a role in the large proportion of recall underestimates to the 9-ending prices of Study 1. The longer prices may have increased the subjects' motivation to minimize effort and give less consideration to the rightmost digits.

The use of a natural situation in Study 2 may also have affected the results. Unlike the participants in an artificial experimental situation, the respondents of Study 2 were able to draw on general knowledge that most prices of supermarket items do in fact end in the digit 9. This knowledge creates an expectation of 9-ending prices which is capable of serving as a source of bias in the recall task (Helgeson and Beatty 1987).

Although the pattern of recall errors differed between the two studies, in both studies the vast majority of errors involved only the rightmost two digits. In this respect, the results of both studies are consistent with the presumption of the underestimation hypothesis that the rightmost digits receive less consumer attention.

GENERAL DISCUSSION

Both of the two studies reported here failed to show any substantial underestimation of 9-ending prices in an immediate recall task. Although a study employing a much larger sample size than that used in the present studies would certainly be more sensitive to very small differences in mean recall level, there is no reason to believe that such a larger study would find differences larger in extent than those, insignificant from a practical standpoint, found in the present studies.

This result contradicts at least a simple form of the underestimation hypothesis, that price perception invariably involves perceiving rightmost 9's as 0's or other low-value digits. It also suggests that the recall underestimation found in Schindler and Wiman's (1989) delayed recall task may have been due more to processes of memory than perception. For example, the consumer's memory trace of the rightmost digits of a price may tend to fade more rapidly than the memory trace of a price's leftmost digits.

It is interesting that, in both studies, the overestimation errors provided evidence for a bias toward recalling the digit 9 for the rightmost digits. This bias was most apparent in Study 2, where the presence of a meaningful context was able to clearly guide consumer expectations. Although such a bias may affect only how the consumer generates a response for an unrecallable digit during the retrieval process (Bettman 1979), it also raises the possibility that the perceptual default value of an unencoded digit may not always be 0 or some other low number. In other words, under some circumstances, prices which do not end in the digit 9 may be perceived as if they did, and thus be overestimated. In either event, the apparent sensitivity of recall to the context of the price information highlights the importance of taking contextual variables into account in research on price-ending effects.

Even though these two studies were consistent in their failure to find substantial underestimation of 9-ending prices, there is an important reason why they should not be used as grounds for a general rejection of the underestimation hypothesis. Both studies provided evidence that consumers give less attention to the rightmost digits of a price. The preponderance of small underestimates of 9-ending prices in Study 1 and the large number of small overestimates of non-9-ending prices in Study 2 both indicate this. These errors derive from biases in the processes used to estimate values of the rightmost digits. But note that it is these digits which are usually the ones which need to be estimated. The leftmost digits are usually recalled correctly.

Conclusions and Directions for Future Research

Although these studies both failed to show any substantial difference between 9-ending prices and non-9-ending prices in level of recalled price, they both do provide evidence for the mechanism which is thought to be behind the hypothesized underestimation effect. This suggests that future research on the underestimation hypothesis should focus on this mechanism, the tendency to give less attention to the rightmost digits of a price.

One approach to further study of this attentional mechanism would be to conduct studies under conditions where this effect would be expected to be greatest. In both of the studies reported here, consumers were tested under conditions where they were likely to have given the price at hand their full attention. In Study 1, they knew they were in an experiment, and in Study 2, they had just selected and purchased the items which were being tested. It is possible that the conditions of these studies allowed the subjects to devote so much attention to the prices that the effects of the tendency to give less attention to the rightmost digits were largely swamped by the generous amount of attention available. Perhaps a substantial underestimation effect would be found only under conditions where the total amount of attention the consumer expends on a price is very limited. Although several studies of price ending effects have attempted to create such limited attention situations (Lambert 1975; Alpert et al. 1981; Schindler and Warren 1988), only the Schindler and Warren study produced consistent evidence for a price-ending effect under these conditions.

Another approach to the study of this mechanism would be to apply new methods to measure the effects of differences in the amount of attention consumers give to the digits of a price. For example, it may be the case that the consumer's perception of a price may not occur instantaneously, but rather may develop in stages over a brief, but measurable time period (e.g., Posner and Mitchell 1967). The earliest stages of this perceptual process may provide a preliminary impression of the price in which the rightmost digits are ignored. While this fleeting impression may usually be elaborated and corrected by the later stages of perception, it may exert an enduring "first-impression effect" on the consumer's evaluation of the price.

The results of the two immediate recall studies reported here do constitute evidence against a simple form of the underestimation hypothesis. But they also provide some support for the mechanism thought to underlie the hypothesis. Thus, these results should not be used to reject the underestimation hypothesis. Rather they should serve to direct further research toward the formulation and investigation of more sophisticated versions of this plausible and interesting mechanism which may lie behind the effects of "psychological pricing."

REFERENCES

Alpert, Mark I., John E. McGrath, and Judy I. Alpert (1984), "Magic Prices: An Extension," in Proceedings and Abstracts of the Thirteenth Annual Meeting of the Western Regional Conference, American Institute for Decision Sciences.

Bettman, James R. (1979), "Memory Factors in Consumer Choice: A Review," Journal of Marketing, 43 (Spring), 37-53.

Conover, Jerry N. (1986), "The Accuracy of Price Knowledge: Issues in Research Methodology," in Advances in Consumer Research, Vol. 13, ed. Richard J. Lutz, Ann Arbor, MI: Association for Consumer Research, 589-93.

Dickson, Peter R., and Alan G. Sawyer (1990), "The Price Knowledge and Search of Supermarket Shoppers," Journal of Marketing, 54 (July), 42-53.

Friedman, Lawrence (1967), "Psychological Pricing in the Food Industry," in Prices: Issues in Theory, Practice, and Public Policy, eds. Almarin Phillips and Oliver E. Williamson, Philadelphia: University of Pennsylvania Press, 187-201.

Georgoff, David M. (1972), Odd-Even Retail Price Endings, East Lansing, MI: Michigan State University.

Helgeson, James G. and Sharon E. Beatty (1987), "Price Expectations and Price Recall Error: An Empirical Study," Journal of Consumer Research, 14 (December), 379-386.

Kreul, Lee M. (1982), "Magic Numbers: Psychological Aspects of Menu Pricing," The Cornell Hotel and Restaurant Administration Quarterly, 23, (August), 70-75.

Lambert, Zarrel V. (1975), "Perceived Prices as Related to Odd and Even Price Endings," Journal of Retailing, 51 (Fall), 13-22, 78.

Mason, J. Barry and Morris L. Mayer (1990), Modern Retailing: Theory and Practice, 5th edition, Homewood, IL: BPI/Irwin.

Ostlund, Lyman E. (1978), "Advertising Copy Testing: A Review of Current Practices, Problems and Prospects," Current Issues and Research in Advertising, 87-105.

Posner, Michael I. and R. F. Mitchell (1967), "A Chronometric Analysis of Classification," Psychological Review, 74, 392-409.

Schindler, Robert M. (1991), "Symbolic Meanings of a Price Ending," in Advances in Consumer Research, Vol. 18, eds. Rebecca H. Holman and Michael R. Solomon, Provo UT: Association for Consumer Research, 794-801.

Schindler, Robert M. and Lori S. Warren (1988), "Effect of Odd Pricing on Choice of Items from a Menu," in Advances in Consumer Research, Vol. 15, ed. Michael J. Houston, Provo, UT: Association for Consumer Research, 348-353.

Schindler, Robert M. and Alan R. Wiman (1989), "Effect of Odd Pricing on Price Recall," Journal of Business Research, 19 (November), 165-177.

Simon, Hermann (1989), Price Management, Amsterdam, The Netherlands: Elsevier Science Publishers.

Twedt, Dik W. (1965), "Does the '9 Fixation' in Retail Pricing Really Promote Sales?," Journal of Marketing, 29 (October), 54-55.

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