Examining Alternative Operational Measures of Internal Reference Prices

William O. Bearden, University of South Carolina
Ajit Kaicker, University of South Carolina
Melinda Smith de Borrero, University of South Carolina
Joel E. Urbany, University of South Carolina
ABSTRACT - Both Winer (1988) and Klein and Oglethorpe (1987) have noted that consumer internal reference price (IRP) is a multi-dimensional construct. Internal reference prices can either be estimated indirectly using scanner data (Winer 1986; Lattin and Bucklin 1989; Kalwani et al. 1990) or measured by asking consumers directly (e.g., Zeithaml and Graham 1983; Bearden and Urbany 1989; Urbany and Dickson 1991). The current paper considers the latter, with a focus on defining different reference price concepts, discriminating between indicators of transaction utility and acquisition utility, and examining how well those indicators explain purchase intentions.
[ to cite ]:
William O. Bearden, Ajit Kaicker, Melinda Smith de Borrero, and Joel E. Urbany (1992) ,"Examining Alternative Operational Measures of Internal Reference Prices", in NA - Advances in Consumer Research Volume 19, eds. John F. Sherry, Jr. and Brian Sternthal, Provo, UT : Association for Consumer Research, Pages: 629-635.

Advances in Consumer Research Volume 19, 1992      Pages 629-635

EXAMINING ALTERNATIVE OPERATIONAL MEASURES OF INTERNAL REFERENCE PRICES

William O. Bearden, University of South Carolina

Ajit Kaicker, University of South Carolina

Melinda Smith de Borrero, University of South Carolina

Joel E. Urbany, University of South Carolina

ABSTRACT -

Both Winer (1988) and Klein and Oglethorpe (1987) have noted that consumer internal reference price (IRP) is a multi-dimensional construct. Internal reference prices can either be estimated indirectly using scanner data (Winer 1986; Lattin and Bucklin 1989; Kalwani et al. 1990) or measured by asking consumers directly (e.g., Zeithaml and Graham 1983; Bearden and Urbany 1989; Urbany and Dickson 1991). The current paper considers the latter, with a focus on defining different reference price concepts, discriminating between indicators of transaction utility and acquisition utility, and examining how well those indicators explain purchase intentions.

Since first proposed by Emory (1970) and Monroe (1973), the notion that consumers use internal reference prices (i.e., price information stored in memory) in judging product value has become well established. This notion has strong theoretical (cf. Helson 1964; Kahneman and Tversky 1979) and empirical (Winer 1986; Lattin and Bucklin 1988; Kalwani, Rinne, Sugita, and Yim 1990) support, in spite of the fact that many consumers may hold expectations with uncertainty (Urbany and Dickson 1991).

The idea that consumers have and use internal reference prices is neither new nor revolutionary. The more important question currently is which kinds of reference prices exist, are measurable, and have the most impact on purchase decisions (see Winer 1988; Klein and Oglethorpe 1987). Internal reference prices have been used in the prediction of consumer brand choice by comparing the price the consumer sees in the current market to the price that s/he expected to be offered (i.e., by calculating price discrepancies). The current research examines how different types of reference price discrepancy terms perform in predicting purchase intentions. We first consider the notion of price discrepancy in more detail.

Price Discrepancy and Consumer Choice

The classic term "sticker shock" may best describe how internal reference prices can influence consumer decision-making. That is, a price that is surprisingly high (higher than the consumer expects) creates disutility and reduces the probability of purchasing that particular brand. Likewise, a price which is lower than expected produces positive utility and increases the likelihood of purchase. These effects have been demonstrated empirically, most impressively in studies using reference prices estimated from scanner data to predict brand choice (Winer 1986; Lattin and Bucklin 1988; Kalwani, Rinne, Sugita, and Yim 1990). In such studies, internal reference prices are estimated as a function of actual previous market prices and, therefore, represent the prices that panel consumers expect to see charged in the market on a particular purchase occasion.

The price discrepancy terms used in these and other studies (Urbany and Dickson 1991; Bearden and Urbany 1989) have most likely represented what Thaler (1985) calls transaction utility. Transaction utility represents the "value of the deal" reflected in a given asking price and is based upon a comparison of the asking price and P*, an expected or "just" price. Two issues are raised in the current paper:

1. Are certain measures of P* better than others?

2. Do discrepancy terms based upon reservation price concepts add significantly to the prediction of purchase intentions?

Acquisition Utility

While price discrepancy terms which reflect transaction utility have been found to be significant predictors of purchase intentions and behavior, it seems that transaction utility itself is a fleeting (if important) source of utility for consumers. For example, positive transaction utility often is simply a perceived windfall which either encourages a consumer to enter the market or to choose one brand over another. However, the consumer's choice is ultimately based upon some evaluation of the product's ability to produce value and create utility relative to the price to be paid (e.g., "get" relative to "give") (Ahtola 1984). Thaler (1985) proposed that the construct acquisition utility captured this value concept. Acquisition utility is the surplus of utility (in dollar terms) over price paid (Thaler 1985, p. 205) or, alternatively, the ratio of perceived benefits to perceived sacrifice (Monroe 1990, p. 74). Further, Monroe (1990) suggests that acquisition utility, like transaction utility, can be represented in a price discrepancy term: (Pmax - P), where Pmax represents reservation price (see also Thaler 1985, footnote 6).

In the present research, we consider the relative explanatory power of separate price discrepancy terms representing transaction and acquisition utility. Before presenting the research methodology, we consider various operational measures of the internal reference price concepts which can be used in calculating transaction utility and acquisition utility.

Potential Estimates of P*

P* is the internal reference price concept designed to reflect the price the consumer expects to be charged or believes is just or fair (Thaler 1985). The current study asks subjects to estimate internal reference prices for a particular apartment, using a variety of approaches. The measures of P* and Pmax are presented below in verbatim form.

Perceived Price. Winer (1988) includes a number of internal reference price concepts under this category, although "perceived price" appears to relate primarily to the consumer's expectation of the price that sellers will charge for a given product on a given purchase occasion. We use three alternative (and probably substitutable) measures for this construct. These estimates have been used as well in a number of advertised reference price studies (e.g., Biswas and Blair 1991; Lichtenstein and Bearden; Urbany, Bearden, and Weilbaker 1988).

NORMAL: What do you think is the NORMAL rent for this apartment?

EXPECTED: What is the rent you would EXPECT to pay?

AVERAGE: Overall, what do you think is the AVERAGE rent for apartments like this?

"Just" or "Fair" Price. Winer (1988) separates fair prices from perceived, noting that the definition of fair price is slippery. The fair price is apparently based in part upon the consumer's perceived price, but also seems to require judgment against some standard of fairness (see Thaler 1985; Kahneman, Knetsch, and Thaler 1986). As such, the previous estimates of perceived price (i.e., shown above) may be different (and possibly higher) than those for the fair price measure:

FAIR: What do you think is a FAIR total monthly rent?

Potential Estimates of Pmax

Conceptually, Pmax is intended to indicate utility in dollar terms -- that is, the dollar amount the consumer would be indifferent to in choosing between cash and the product or the most one would be willing to pay. Operationally, we use three measures:

MOST: What is the MOST you would pay and still consider the apartment worthwhile to rent?

INDIFFERENT: From the price range given on the right, identify that price at which you would be INDIFFERENT between renting the apartment or looking further for an apartment.

HIGHEST TO SEARCH: Let's assume that you are currently living and settled in this apartment and it meets your needs. Assume that the rent is expected to change. If the apartment owners/managers actually decided to change apartment rents, what is the HIGHEST monthly rent that would make you SEARCH FOR AND MOVE to another apartment?

Research Expectations

Our research attempts to discriminate between price discrepancy measures which should indicate transaction utility (TU) and those which should indicate acquisition utility (AU). In addition to measuring price expectations, we also measured respondent perception of TU and AU using several scaled items. As such, we expect that the TU price discrepancy variables will correlate most strongly with the TU scale and that the AU price discrepancy variables will correlate most strongly with the AU scale.

We also examine below how the TU and AU discrepancy terms perform in predicting purchase intentions. Our original expectation was that the AU construct, in reflecting the more permanent "utility" that the consumer would obtain through the purchase, would have a stronger impact on purchase intentions.

METHOD

Data were collected from 125 graduate student subjects via an interactive computer simulation. Seventy-six of the participants were male; the average age for the sample was 25.2 years. The research was conducted under the guise of research involving the evaluation of residential and apartment properties for university students. Prior to exposure to several apartment descriptions, data were collected regarding prior renting experience, rents paid, and related search experiences. One hundred and eighteen of the subjects (94 percent) were current renters; half the sample rented two bedroom apartments. Seventy percent of the respondents rated their general familiarity of apartments in the University area as at least "4" on a seven-place scale. The critical apartment (i.e., the location for which the various price estimates and judgments were elicited) was described as: having two bedrooms; being close to campus; in a safe, nice area; within a new building; and having a dishwasher. Using the operational statements reviewed above, the various price estimates were obtained in the following fixed order: NORMAL, MOST, EXPECTED, INDIFFERENT, AVERAGE, FAIR, and HIGHEST TO SEARCH. Following these estimates, the respondents were exposed to similar questions for another but different apartment. These latter tasks served also to separate the initial estimates from subsequent scaled-item judgments of perceived acquisition value, transaction value, and willingness-to-rent.

After providing price estimates for the second apartment and an additional set of filler questions, the study participants were provided with the initial apartment description in the context of a newspaper advertisement, including a monthly rent. Subjects were asked to assume they were in the market for an apartment and noticed this advertisement in the Sunday newspaper. The advertised prices were manipulated to vary across seven conditions. These conditions were configured in a 1X7 design to represent prices either 15, 30, or 45 percent above or below or equal to the average of each subject's initial NORMAL and EXPECTED price estimates. As such, the prices were varied at the individual level based upon each subject's initial expectations. The cell sizes ranged from 17 to 19.

TABLE 1

MEANS AND STANDARD DEVIATIONS FOR DIFFERENT INTERNAL REFERENCE PRICE ESTIMATES AND SCALED VARIABLE MEASURES

Operational Measures. Again, the "raw price estimates" were obtained using the question wording as previously described. Difference scores, in which the individual price is subtracted from the respondents' price estimates, were computed to represent the price based estimates of acquisition value and transaction value. That is, the advertised price was subtracted from the estimates of EXPECTED, AVERAGE, NORMAL, and FAIR to reflect four alternative price-based estimates of transaction value (P* - P). Likewise, the advertised price was subtracted from the initial estimates of MOST, INDIFFERENT, and HIGHEST TO SEARCH to obtain three price-based individual estimates of acquisition value (Pmax - P). A series of four, three, and four seven-place scaled statements were also included to operationalize transaction value, acquisition value, and willingness-to-rent, respectively. The responses to the indicators for the latter three constructs were summed to from an overall index for each variable. The wording of the statements was varied in direction to inhibit response bias; each scale position was labeled to suggest equal intervals across response alternatives. The four transaction scaled statements were each worded to prompt the respondent explicitly to compare the expected price to the advertised amount. For example, the items were prefaced by: "Compared to what I expect this apartment normally rents for, the advertised price appears to be..." These statements were then followed in the computer exercise on separate screens by the following bipolar adjective sets: high-low, outrageous-reasonable, expensive-inexpensive, and overpriced-underpriced. The summated scale version of acquisition value was operationalized as the response to three items: 1) Overall, the offer for this apartment is ...Very Poor Value-Very Good Value; 2) Overall, this apartment is a good value for the money...Agree-Disagree; and 3) The apartment is an excellent buy for the money..Agree-Disagree. Willingness-to-rent (cf. willingness-to-buy) was measured as the sum of four items similar to the following two examples: 1) "How willing would you be to rent this apartment...Very Unwilling-Very Willing"; and 2) "Given the offer described, the likelihood that I would rent this apartment is...Very Low-Very High". As shown in Table 1, the coefficient alpha estimates of internal consistency reliability were 0.94, 0.81, and 0.94 for the scaled measures of transaction value, acquisition value, and willingness-to-rent, respectively.

RESULTS

The means and standard deviations for the seven price estimates along with the means for the three scaled measures are presented in Table 1. Three of the initial estimates of P* (i.e., NORMAL, EXPECTED, and AVERAGE) have essentially the same mean and standard deviation. The fairness measure produced a numerically higher mean and variance than the other three measures--suggesting a wide range of interpretations of the term "fair." There is a fairly wide range between the three reservation measures of Pmax. The lowest estimate came from the "indifference" measure for which respondents were placed in a search context. For the "highest to search" estimate (and the estimate that possessed the highest mean), subjects were asked to assume they were already renting. Of interest also, the "most" estimate was not that different from three of the the P* estimates.

TABLE 2

INTERCORRELATIONS AMONG PRICE ESTIMATES

Intercorrelations among the initial estimates of the "raw" scores along with the intercorrelations among the difference scores are summarized in Table 2. For the raw score estimates (shown above the diagonal in Table 2), some very modest evidence of discriminant validity of the P* estimates versus the Pmax estimates was provided. Additional evidence regarding the very different effects captured by the fairness measure was also provided. The high correlations among the discrepancy scores for (P*-P) and (Pmax-P) clearly show that, while the Pmax estimates may be higher (see Table 1), the Pmax or reservation price difference scores are closely correlated with the P* estimates.

Correlations of the price-based estimates of transaction value [i.e., (P*-P)] and acquisition value [i.e., (Pmax-P)] with the scaled measures of transaction value, acquisition value, and willingness-to-rent are depicted in Table 3. In support of the general research procedures employed in the study and as would be expected, the initial estimates (which were obtained prior to the provision of the apartment rent) were not correlated with the scaled variables. In contrast, and as expected, all of the discrepancy variables were significantly correlated (p<.01) with the scaled measures of acquisition and transaction value. However, all the results were not as anticipated. Specifically, the acquisition value estimates [i.e., (Pmax-P)] were not more highly correlated with the scaled measure of acquisition value, but were actually more strongly related to the four-item scaled measure of transaction value. It is possible that the respondents were unable to get or develop a solid feel for the overall utility of the apartment that might be obtained with a more complete description or a picture.

The relative predictive ability of the various measures of acquisition value and transaction value were further examined in a series of simple and multiple regression analyses. These results are presented in Table 4. In these tests, the measure of willingness-to-buy was used as the dependent variable throughout. The Table includes both standardized regression coefficients and simple correlations along with estimates of the amount the variance explained decreased when each variable was omitted from its corresponding two predictor variable multiple regression equation. These latter estimates assist in interpreting the relative importance of predictor variables when multicollinearity problems are likely.

These regression results present a number of interesting findings. The scaled measures of acquisition value and transaction value explain a considerable amount of the variance in willingness-to-rent (i.e., adjusted R-square = 0.73). However, the scaled measures shared considerable similarity in measurement methods, despite the items being interspersed among a larger number questions and the varied direction of the item wording. As shown, analysis of the three scaled variables revealed that acquisition value appeared to be the most important correlate with willingness-to-rent as anticipated. A different conclusion emerges when the price-based predictions are evaluated. That is, the discrepancy estimates of transaction value [i.e., (P*-P)] appear most important. In other words, the relative importance of TU and AU differ somewhat depending upon the types of measures used.

TABLE 3

CORRELATIONS OF ESTIMATED PRICE VARIABLES WITH SCALED MEASURES OF TRANSACTION VALUE, ACQUISTION VALUE, AND WILLINGNESS TO RENT

A number of other findings appear noteworthy as well. First, there was considerable overlap among the predictors. Little incremental variance was explained by the addition of a second variable in most of the equations. And, for the price discrepancy estimates, the fairness estimate contributed very little to the overall judgments of the advertised apartment. This finding is surprising in view of Thaler's (1985) suggestion that "fairness" represents a primary measure of P*.

DISCUSSION

This paper presented results from a simple experiment which examined the correlations between a series of price discrepancy variables and willingness to buy. In the research, the advertised price was subtracted from seven price estimates to form four and three price discrepancy measures of transaction value (P* - P) and acquisition value (Pmax - P), respectively. Relationships with scaled measures of transaction value and acquisition value were also investigated. Caveats are in order regarding the use of student subjects, the simulated nature of the research, and the failure to separate in time the initial estimates from the overall scaled measures. [Our intent for future research is to vary the order of estimate elicitation.] Efforts were made, however, to use a plausible context for which subjects were familiar and a number of filler tasks did separate the initial estimates from the advertisement exposures.

The most obvious result is that, except for the FAIR estimate of P*, the TU and the AU price discrepancy terms are relatively indistinct, especially in their influence or correlation with intentions. We would argue that previous research has operationalized price discrepancy terms which seem to represent transaction utility primarily (e.g., Kalwani et al. 1988; Lattin and Bucklin 1988; Winer 1986). To the extent that P* and Pmax overlap or are highly correlated, however, the (P* - P) term may be sufficient. That is, adding the AU term may not add incrementally to explained variance. While these studies have referred to their price expectation measures generally as "reference prices," they may be capturing elements of reservation price as well as price expectations.

TABLE 4

REGRESSION RESULTS: PREDICTING WILLINGNESS-TO-RENT FROM ESTIMATES OF TRANSACTION AND ACQUISITION VALUE

The above conclusion is drawn from our correlational analysis. However, limited incremental gain from adding the AU term to the prediction of willingness to buy may hold even when P* and Pmax are conceptually and empirically distinct. All that is required is that P* and Pmax be highly correlated. In this study, the "highest-to-search" measure produced the highest estimate of Pmax (and, therefore, was the most distinct from P*), yet was still highly correlated with the estimates of P*.

Situations can be envisioned when P* and Pmax are identical, and hence, the price discrepancy terms of AU and TU would be identical. For example, any situation in which a consumer believes that the most s/he is willing to pay is equal to the market price (i.e., there is no need to pay more) would equate P* and Pmax. Even if this is not the case, it is still likely that the expected market price provides a foundation for judging Pmax, and as a result, the two will be highly correlated. To the extent this is true, operationalizations of Thaler's model using price discrepancy terms may be accomplished most effectively using just the TU discrepancy terms.

To understand whether (P* - Pmax) will always be highly correlated, however, we need to make sure that Pmax is being properly measured. It seems that the measure of Pmax should allow for maximum distinction between itself and P*. Clearly, future research is warranted. For example, we are not certain what the respondents were considering in their responses to the "fair" estimate. The unexpected relatively high estimate for FAIR and the very large variance indicate differences in judgments for this internal reference price estimate.

Lastly, the results are derived from a single study for a single alternative evaluation, and hence, our findings are offered only as an initial effort. And, even if the conclusion were eventually reached that the AU price discrepancy term adds little to the "easier to measure" TU price discrepancy term, this would not suggest that the AU concept itself is not useful. In fact, although, intercorrelated, the scaled measures of AU which measure perceptions of the "value for the money" directly were correlated with purchase intentions.

REFERENCES

Bearden, William O. and Joel E. Urbany (1989), "Reference Price Effects on Perceptions of Perceived Offer Value, Normal Prices, and Transaction Utility," in AMA Educators' Proceedings, 54, Chicago: American Marketing Association.

Biswas, Abhijit and Edward A. Blair (1991), "Contextual Effects of Reference Prices in Retail Advertisements," Journal of Marketing, 55 (July), 1-12.

Kalwani, Manohar U., Heikki J. Rinne, Yoshi Sugita, and Chi-Kin Yim (1988), "A Price Expectations Model of Customer Brand Choice," Journal of Marketing Research, 27 (August), 251-262.

Klein, Noreen M. and Janet E. Oglethorpe (1987), "Cognitive Reference Points in Consumer Decision-Making," in Melanie Wallendorf and Paul Anderson (eds.), Advances in Consumer Research, Vol. 14, Provo, UT: Association for Consumer Research, 183-187.

Lattin, James M. and Randolph E. Bucklin (1988), "The Dynamics of Consumer Response to Price Discounts," MSI Working Paper 88-111, Cambridge, MA: Marketing Science Institute.

Lichtenstein, Donald R. and William O. Bearden (1989), "Contextual Influences on Perceptions of Merchant-Supplied Prices," Journal of Consumer Research, 16 (June), 55-66.

Thaler, Richard (1985), "Mental Accounting and Consumer Choice," Marketing Science, 4.3, Summer, 199-214.

Urbany, Joel E., William O. Bearden, and Dan A. Weilbaker (1988), "The Effect of Plausible and Exaggerated Reference Prices on Consumer Perceptions and Price Search," Journal of Consumer Research, 15 (June), 95-110.

Urbany, Joel E. and Peter R. Dickson (1991), "Consumer Normal Price Estimation: Market vs. Personal Standards," Journal of Consumer Research, 17 (June), 45-51.

Winer, Russell S. (1988), "Behavioral Perspective on Pricing: Buyers' Subjective Perceptions of Price Revisited," in Timothy Devinney (ed.), Issues in Pricing: Theory and Research, Lexington, MA: Lexington Books, 35-57.

Winer, Russell S. (1986), "A Reference Price Model of Brand Choice for Frequently Purchased Products," Journal of Consumer Research, 13, September, 250-256.

Zeithaml, Valerie and K. Graham (1983), "The Accuracy of Reported Reference Prices for Professional Services," in A. Tybout (ed.), Advances in Consumer Research, Vol. 10, Provo, UT: Association for Consumer Research, 607-611.

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