/AANOHAR U.

KALWANI and CHI KIN YIM*

The authors report results from a controlled experiment designed to investigate the impact of a brand's price promotian frequency ond the depth of promotional price discounts on the price consumers expect to poy for that brand. A key feature of the work is that expected prices elicited directly from respondents in the experiment are used in the analysis, os opposed to the latent or surrogote measures of expected prices used in previous studies. As hypothesized, both the promotion frequency and the depth of price discounts are found to have a significant impact on price expectations. Evidence also supports a region of relotive price insensitivity around the expected price, such that oniy price changes outside that region have a significant impact on consumer brand choice. Further, the outhors find that consumer expectations of both price and promotional activities should be considered in explaining consumer brand choice behavior. Specifically, the presence of a promotional deol when one is not expected or the obsence of a promotional deal when one is expected may hove a significont impact on consumer brond choice. Finally, as in the case of price expectations, consumer response to promotion expectations is found to be asymmetric in that losses loom larger than gains.

Consumer Price and Promotion Expectations: An Experimental Study

The role of expectations in predicting the behavior of eeonomic agents has long been widely accepted in the economics literature (Muth 1961; Nelson 1977). In marketing, the study of the impact of price expectations on consumer choice behavior has begun to receive increasing attention In recent years. The intuitively appealing proposition that consumers form price expectations and use them in evaluating price infonnation when making a purchase has both theoretical and empirical support {Gurumurthy and Little 1989; Kaiwani et al. i990; Lattin and Bucklin 1989; Puller 1990; Raman and Bass 1988; Rinne 1981; Thaler 1985; Winer 1986). Introducing a product at a lower than regular price and then raising the price afterward to its regular level has been shovi'n to

Manohar U. Kalwani is Professor ot" Management. Krannen Graduate School of Management, Purdue University. Chi Kin Yim is Assistant Professor of Administrative Science. Jesse H. Jones Graduate School ot" Administration. Rice University. The authors thank Randy Bucklin, Jim Lattin, Dan Putler. the editor, and three anonymous JMR reviewers for their helpful comments and suggestions.

have an adverse effect on subsequent sales. The reason is that consumers come to adopt the low introductory price as a reference and consider the regular price to be unacceptably greater than the price they expect to pay (Doob et al. 1969). From a managerial viewpoint, understanding how consumers form and use price expectations in making purchase decisions is important because the failure to incorporate price expectations has been shown to result in a misestimation of price elasticity, which can lead to nonoptimal pricing decisions (Doyle and Saunders 1985). Recently, the price expectations hypothesis has been used to provide an altemative explanation for the observed adverse long-term effect of price promotions on brand choice (Kaiwani et al. 1990). Previous research has shown that repeat purchase probabilities of a brand after a promotional purchase are lower than the corresponding values after a nonpromotional purchase (Dodson, Tybout, and Sternthai 1978; Guadagni and Little 1983; Shoemaker and Shoaf 1977). Dodson, Tybout, and Sternthai evoke self-perception theory to predict that if a purchase is induced by an external cause (such as a price promotion) as opposed to an internal cause (e.g..
90 Journal of Marketing Research

Vol. XXK (Febniary 1992), 90-100

CONSUMER PRICE AND PROMOTION EXPECTATIONS

preference for the brand), repeat purchase probability of the brand will be reduced when the extemal cause is removed. Alternatively, Kalwani et al. argue that consumers form expectations of a brand's price on Ihe basis of, among other things, its past prices and the frequency with which it is price promoted. Consumers' reactions to a retail price then may depend on how the retail price compares with the price they expect to pay for the brand. Specifically, during a price promotion, they are apt to perceive a price "gain" and react positively; correspondingly, when the deal is retracted, they are apt to perceive a price "loss" and are unlikely to purchase the brand. Neslin and Shoemaker (1989) offer yet another alternative explanation for the phenomenon of lower repeat purchase rates after promotional purchases. They argue that the lower repeat purchase rates may be the result of statistical aggregation rather than actual declines in the purchase probabilities of individual consumers after a promotional purchase. Specifically, "if the promotion attracts many consumers who under nonpromotion circumstances would have very low probabilities of buying the brand, then on the next purchase occasion the low probabilities of these consumers bring down the average repurchase rate among promotional purchases" (p. 206). The behavior of households that have low probabilities of buying a brand upon the retraction of a deal can be explained readily in a price expectation framework. We suggest that the price they expect to pay for the brand may be close to the deal price and they may forego purchasing the focal brand when it is not promoted because its retail price far exceeds what they expect to pay for it. We investigated the impact of price promotions on consumers" price expectations and brand choice in an interactive computer-controlled experiment. Our work differs frotn previous research in that expected prices were elicited directly from respondents in the experiment and used in the empirical investigations of the impact of price promotions on consumers' price expectations. Further, rather than studying the impact of just a single price promotion and its retraction, we assessed the significance of the dynamic or long-term effects of a sequence of price promotions. Specifically, our objectives in the ex[lerimental study were (1) to investigate the effects of different price promotion frequencies and depths of price discounts on the brands' expected prices and (2) to test the impact of price promotions on brand choice, as mediated by consumers' expectations, in a controlled experimental setting. Our results from the laboratory experiment provide support for the hypotheses that both the price promotion frequency and the size of price discounts have a significant adverse impact on a brand's expected price. Consistent with the findings of Raman and Bass (1988) and Gurumurthy and Little (1989), we also find evidence in support of a region of relative price insensitivity around the expected price such that changes in price within that region produce no pronounced change in consumers'

perceptions. Price changes outside that region, however, are found to have a significant effect on consumer response. Further, we find that promotion expectations are just as important as price expectations in understanding consumer purchase behavior, ln particular, we find that consumers who have been exposed to frequent price promotions in support of a given hrand may come to form promotion expectations and typically will purchase the brand only when it is price promoted. We also find that, as in the ease of price expectations, consumer response to promotion expectations is asymmetric in that losses loom larger than gains. At this point, it is useful to define what we mean by the terms "expected price" and "price promotion." Following Thaler (1985), we view the price consumers use as a reference in making purchase decisions as the price they expect to pay prior to a purchase occasion. Further, the expected price may also be called the "intemal reference price" (Klein and Oglethorpe 1987) as opposed to an extemal reference price such as the manufacturers' suggested list price. Finally, we assume that a brand is on price promotion when it is offered with a temporary price cut that is featured in newspaper advertising and/ or brought to consumers" attention with a store display sign. In the following section, we delineate the hypotheses about reiationships between different price promotion schedules and brands' expected prices. Then we discuss the modeling of the effects of consumers' price expectations on brand purchase probabilities. After describing the interactive computer experiment setup for data collection, we outline the statistical analyses for the testing of each of the hypotheses and for the estimation of model parameters. We then report the findings from the experimental study, Finally, we discuss the implications of our findings and outline directions for future work, including the potential for further experimental work in the area. PRICE PROMOTION SCHEDULES. PRICE EXPECTATIONS. AND BRAND CHOICE Applying Helson's (1964) adaptation-level theory to price perceptions. Sawyer and Dickson (1984) suggest that price promotions may work in the short run because consumers may use the brand's regular price as a reference and then are induced by the lower deal price to purchase the brand. However, frequent temporary price promotions may also lower the brand's expected price and lead consumers to defer purchases of the brand when it is offered at the regular price. Impact of Price Promotion Frequency on Expected Price Our first hypothesis pertains to the impact of a brand's price promotion frequency on the price consumers expect to pay for that brand. H|: The price consumers expect to pay for a brand decreases with an increase in the observed frequency

JOURNAL OF MARKETING RESEARCH, FEBRUARY 1992

"

of price promotions of the brand. Further, the relationship between the expected price and the price protnotion frequency can be approximated by a sigmoid function.

Tversky and Kahneman (1974) have shown that people rely on a limited number of heuristic principles that reduce complex tasks of assessing probabilities and predicting values to simpler judgmental operations. In some cases, people may anchor and adjust their forecasts by starting with a preconceived point and weigh that point heavily in arriving at a judgment. When the frequency of past price promotions is "very low," consumers identify a price promotion offer as an exceptional event and may not modify the brand's expected price.' The brand's expected price then will be anchored around the regular price because of insufficient adjustment. In other cases, people may arrive at a judgment on the basis of how similar or representative the event is to a class of events. Therefore, when a brand is price promoted "too often." consumers come to expect a deal with each purchase and hence expect to pay only the discounted price on the basis of its representativeness. Clearly, given a certain level of price discount, the brand's expected price will be bounded by the regular price and the implied sale price. That tine of reasoning suggests that the relationship between the price promotion frequency and the expected price can be approximated by a sigmold function. Impact of Depth of Price Discounts on Expected Price Our second hypothesis pertains to the impact of the depth of price discounts offered during a brand's price promotions on its expected price. H^: The price con.sumers expect to pay for a brand decrease.s with an increase in the observed depth of price discounts of the brand. Further, the relationship between the expected price and the depth of price discounts can be approximated by u concave function. Whether a price discount will affect the brand's expected price depends on how consumers perceive the discount. Uhl and Brown (1971) postulate that the perception of a retail price change depends on the magnitude of the price change. They report results from an experiment indicating that 5% deviations were identified correctly 64% of the time whereas 15% deviations were identified correctly 84% of the time. Delia Bitta and Monroe (1980) find that consumers" perceptions of savings from a promotional offer do not differ significantly between 30%, 40%, and 50% discount levels. However, they find significant differences between the 10% and

30 to 50% levels. They also discuss some managers' beliefs that at least a 15% discount is needed to attract eonsumers to a sale. Apparently, small price changes may not be noticed and even a large price reduction (say, 60 or 70%) may not be assimilated to affect the brand's expected price if it is considered exceptional. Hence, the impact of the depth of price discounts on lowering the brand's expected price is likely to occur when the price discount offered by the brand is relatively large but not so large that it is seen as an exceptional event. In our research, we chose to examine price discounts ranging from !0 to 40%', a range commonly used in past research on price discounts in the consumer packaged goods categories (Berkowitz and Walton 1980; Curhan and Kopp 1986). That was also the range of price discounts available in stores in the local area where the experiment was conducted. Within that range, the findings of Uhl and Brown (1971) and Delia Bitta and Monroe (1980) suggest that it is reasonable to expect the relationship between the brand's expected price and the depth of price discounts to be concave. Consumer Price Expectations and Brand Choice Previous research provides plenty of evidence to support the hypothesis of a standard price serving as a reference level for price judgments (Helson 1964; Kahneman and Tversky 1979; Monroe 1973). When modeling consumers' evaluation of a purchase, in addition to the element of consumer surplus or acquisition utility as suggested by standard economic theory. Thaler (1985) posttjlates a transaction utility that pertains solely to the merit of the "deal" available or the perceived savings of the offer. Specifically, a negative transaction utility (when the retail price exceeds the expected price) will reduce the brand purchase probability, whereas a positive transaction utility (when the retail price is less than the expected price) will enhance the brand purchase probability. As a further refinement of how consumers respond to price changes, Raman and Bass (1988) apply assimilation-contrast theory (Sherif 1963) to postulate a region of price insensitivity around a brand's expected price such that changes in price within that region produce no marked change in perceptions. Hence, only price changes that are outside that region are assumed to have a significant impact on consumer brand choice. We identify the threshold values representing the boundaries of the region of indifference empirically on the basis of the quality of the model fit. As mentioned, there is a growing amount of empirical evidence on the role of reference price in determining consumer response to retail prices. Two recent studies by Winer (1985. 1986), based on linear probability models of consumer response to price information for the purchase of seven durable goods and three brands of coffee, reveal that a retail price higher than ihe reference price ("sticker shock") has a statistically significant negative

'Terms such as "low" and "often," which are used to describe Ihc frequency of price promotions, are all relative. Please refer lo the section discussing the ncatment levels to put such qualifiers into proper perspective.

CONSUMER PRICE AND PROMOTION EXPECTATIONS

93 wise.' The threshold levels, 6^. and 0,, are measured in percentage terms as a fraction of a brand's average nonpromotional price. We subsequently estimate the threshold parameters, 6^ and 6,. empirically through a grid search on the basis of the fit of our brand choice model. Mathematically, we represent the deterministic component of utility, V,t, as
(2)

impact on consumer purchase probability for three of the seven durable goods and two of the three brands of coffee. Gurumurthy and Little (1989), Kalwani et al. (1990), and Putler (1990) also find that the transaction utility, as proposed by Thaler (1985). plays an important role in determining consumer brand choice behavior and that consumers react more strongly to priee losses (i.e., when the retail price is higher than the expected price) than to price gains (i.e., when the retail price is lower than the expected price). Finally. Lattin and Bucklin (1989) use an exponentially smoothed measure of a reference price to model its effect on brand choice, but they find that the reference effect of price is not significant. Nevertheless, all of the authors cited, who rely on consumer purchase data to calibrate their models, only use a surrogute measure of the expected price. Our approaeh differs from previous research in that we investigate the impact of price expectations on brand choice by using a direct measure of the expected price from a controlled experiment. Specification of the Brand Choice Model We propose a binary logit model to represent the impact of a brand's expected price, in relation to its retail price, on a consumer's probability of purchasing that brand. Following McFadden (1973). we assume the consumer chooses the brand that offers more utility and the random errors on utility are independent and identically Gumbel distributed; the probability of choosing brand i is given by (I)
Pik

= a,,, +

i = 1,2, where: brand-specific constant for brand / (only one brand-specific constant is specified to avoid singularity in the maximum likelihood estimation), consumer k's preference for brand /, the retail price of brand / as observed by consumer A, a dummy variable that captures the effect of a promotion of brand / observed by consumer k.

= 1 if

> 6,.

0 otherwise. 0 otherwise.

Loss,,=

EPit - the expected price of brand / directly elicited from consumer k, and RPn = the average nonpromotional price of brand ( as observed by consumer k. Because the coefficient ai captures the effect of brand preference on brand choice, we expect it to have a positive sign. The coefficient a^ represents the direct effect of price and so is expected to be negative. The promotion variable can be interpreted as representing the shortterm effect of promotions and we expect a, > 0. Finally, we expect a perceived price gain (perceived price loss) to have a positive (negative) effect on brand choice and, hence, we expect 04 > 0 and a^ < 0. DESIGN OF THE EXPERIMENT Overview The experiment consisted of two parts. The first part was designed to study the impact of different price promotion schedules on brands' expected prices. The second part was aimed at investigating the role of consumers' price expectations in a brand choice situation.

where V,* is the deterministic component of utility that consumer k derives trom the purchase of brand /. We model the deterministic compt^nent of the utility that consumer k derives from the purchase of brand ( as a linear function of the consumer's preference of brand ( (directly measured in the experiment), the retail price of brand /, a O-I dummy variable indicating whether or not brand / is available on a promotional deal, and the perceived price gain and loss variables. In operationalizing the price gain and loss variables, we take into account two factors, (I) a postulated region of price insensitivity around the expected price and (2) the proptisition that the utility consumers derive from purchasing a brand may not vary linearly with the magnitude of the difference between the brand's expected and retail prices. In particular, we posit that the price gain variable takes a value of I if a brand's expected price exceeds its retail price by a certain threshold level, 6^., and takes the value 0 otherwise. Analogously, the price loss variable is assumed to take the value ! if a brand's retail priee exceeds its expected priee by a certain threshold level, 6,, and is assumed to take the value 0 other-

"We are indebted 10 one of the anonymous JMR reviewers for the construction of a region ot" price insensitivily around the brand's expected price. Raman and Bass (1988) and Gurumurthy and Little (1989) also report empirical evidence in support of that latitude of acceptable price difterences.

94

JOURNAL OF MARKETING RESEARCH. FEBRUARY 1992

Briefly, a group of respondents were exposed to price and promotion infonnation on two competing liquid laundry detergent brands and then were asked to put themselves in a hypothetical buying scenario in which they needed to purchase laundry detergent from a local supermarket. The laundry detergent product category was selected on the basis of the results of a pilot study and a pretest of the actual experiment. The pilot study was used to identify key product attributes for characterizing the brands in the experiment, the level of regular brand prices, and the hypothetical names of brands. Two hypothetical brands of liquid laundry detergent were used to minimize the potential respondents' tendency to invoke an image or previous experience, which can cause variability (Monroe 1977). Descriptions and performance ratings of the two brands were distinct. For instance, for one brand we emphasized versatility or high perfonnance in all temperatures and for the other brand we emphasized whitening power. The assignment of attributes to the hypothetical brands was based on two major design considerations: (1) each brand should have some distinct attributes (attributes not shared by the other brand) to allow a variation in the brand preference and (2) to allow the possibility of switching between the two brands, neither of the brands should be dominated by the other brand. The regular prices of the two brands were set at levels consistent with actual retail store data and respondents' perceptions. The higher priced brand was used as the target brand and the lower priced hrand was used as the control brand in terms of the manipulation of the price promotion frequency treatment variable (described in detail subsequently). A randomly determined 1 % variation was added to the regular price to represent minor changes in store prices and to simulate the effect of limited memory of consumers. Research Design

is price promoted frequently and a private label brand that is seldom price promoted. Sample Design and E.xperimenlal Procedures Two hundred undergraduate students enrolled in introductory marketing management classes at a large midwestern university served as respondents for the experiment. From the 200 respondents, we collected 188 sets of completed responses. Responses were collected interactively on a computer. The computer was used for stimuli presentation and response recording. The use of an interactive computer experiment was also helpful in the implementation of individualized randomization and enhancing respondent involvement in the experiment (Aronson, Brewer, and Carlsmith 1985). Purchase context, retail outlet, and package size were specified to controi for variability of price perception due to those contextual variables. The experiment was conducted in a computer laboratory. Upon arrival, the respondents were assigned randomly to one of the 16 treatment conditions. They were given a handout describing the objectives of the experiment, the phenomena represented in the experiment, and how to use the computer. A practice session at the outset of the experiment familiarized respondents with the purchase situation and the use of the computer. Two shampoo brands were used in the practice session. Respondents were debriefed after the experiment. The debriefmg was used to explore with each respondent the impact of the experimental events and to seek answers to such questions as whether the instructions were clear, whether or not the respondent was suspicious, and whether or not particular suspicions would invalidate the results. The true objective of the experiment then was revealed to the respondent. Measures Three of the variables directly measured in the experiment were (1) brand preferences, (2) expected brand prices, and (3) brand choice. Consumers' preferences for the brands were measured hy a constant-sum preference question in which respondents were asked to allocate itX) points between the two brands of laundry detergent after reading a description of the benefits and features (excluding price) and performance ratings of each brand. This measure was obtained before the respondents were exposed to any price and promotion information on the two brands. The brands' expected prices were measured by asking the respondents to answer the following open-ended question: "Based on the prices of Brand X over the past 10 weeks, how much do you expect the price of Brand X to be this week (week ID? $ " Puto (1987), Rowe and Puto (1987), and Zeithaml and Graham (1983) have used similar questions to obtain measures of expected prices. Prior to being asked to give their perceptions of brands" expected prices, the respondents were exposed to 10 weeks (week 1 through week 10) of price

The research design for testing the impact of price promotions on brands' expected prices was essentially a 4 (price promotion frequencies) by 4 (depth of price discounts) bet ween-subject factorial design. The four levels of price promotion frequencies were 1, 3, 5, and 7 price promotions over 10 weeks. The four levels of price discounts were 10, 20. 30, and 40% off the average nonpromotional price. The pretest revealed that applying the same manipulations to both the target brand and the control brand appeared to result in respondent information overload. Respondents were not able to separate the identity and the past price and promotion information (especially the frequency of price promotions) of the two brands. Brand choices seemed to be made randomly. Therefore, in the experiment, the i6 treatments were applied only to the target brand and the treatments applied to the control brand were restricted to oniy one promotion over 10 weeks and the four levels of price discounts. Recall that the target brand was also the higher priced of the two brands and, hence, our experimental setup can be viewed as a market comprising a national brand that

CONSUMER PRICE AND PROMOTION EXPECTATIONS

95

and promotion information on the two brands. The individual prices and number of price promotions respondents saw varied across the 16 treatment cells. The pattem of price promotions for each brand within a particular treatment cell was randomized for each respondent. A price promotion was presented as a "Store Special $ Off" (flashing on the computer screen in a bright color to simulate a display effect). Finally, the two brands' prices and promotion information, if any, in week 11 were shown to the respondent. Brand choice was obtained by requesting the respondent to select one of the two brands for purchase. A binomial process with the probability parameter .5 was used to determine for each of the two brands whether it would be made available on a price promotion to a given respondent during week 11. Selected other questions were asked during and after the experiment to provide data for various manipulation checks (e.g., checking whether respondents mentally processed the price and promotion information presented to them) and some demographic and shopping behavior information. DISCUSSION OF FINDINGS Manipulation checks reveal that the manipulations of the two factorsthe price promotion frequency and the depth of price discountwere successful. We find that respondents did process the price and promotion information presented to them. Impact of Price Promotions on Consumers' Price Expectations We tested our two price expectations hypotheses in two steps. First, we employed an analysis of variance to explore the effects of the two factors-{I) the frequency of price promotions and (2) the depth of the price discounts offered during a brand's promotions^on a brand's expected price. We then used an F-test for linearity (Johnston 1984, p. 54-56) to analyze further the relationship between the brand's expected price and the two factors. The results of the ANOVA are summarized in Table I. They show that, as hypothesized, consumers' exposure to different price promotion frequencies and depths of price discounts had significant effects on the brands' expected prices. However, the results do not suggest the presence of an interaction effect of the price promotion frequency and the depth of price discounts on a brand's expected price.

Table 1
ANOVA RESULTS WITH BRAND'S EXPECTED PRICE AS THE DEPENDENT VARIABLE p-vatue (two'taitedi .0001 .0001 .3382

Source Frequency of price promotions Depih of price discounts Interaction of promotion frequency and depth of price discounts Error Total

d.f.

SS

MS

3 H.06 3.69 15.84 3 7.59 2.53 10.87

9 2.38 172 40.03 187 61.06 .26 .23 M4

Next, we tested whether the relationships between the brands' expected prices and each of the two independent variables were nonlinear as hypothesized. Results of the F-tests for linearity show that we cannot reject the hypotheses of linear relationships between the brand's expected price and both the frequency of price promotions and the depth of price discounts at the a = .10 level. However, the results of a multiple comparison by the Tukey test indicate that, among all the contrasts of the brand's expected prices at different price promotion frequencies, only the contrast between the second level (i.e., 3 promotions over 10 weeks) and the third level (i.e., 5 promotions over 10 weeks) of the price promotion frequency is significant at a = .05. A plot of the average expected prices for the different levels of price promotion frequency also provides some directional support for a sigmoid relationship between the expected price and the price promotion frequency (see Figure 1).

Figure 1 EXPECTED PRICE AS A FUNCTION OF THE FREQUENCY OF PRICE PROMOTION

Kxpected Price
4.3 4.2 4,1 4,0 3.9 3.8

'Two 4 x 4 factorial analyses of variance were used in the manipulation checks to ensure that respondents processed the price and promotion information. Both ANOVAs used the actual price promotion frequency and depth of price discount in the experiment as independent variables. The recalls of the number of times a brand was on sale at a discounted price and the average amount of price discount offered on each of the brands in the experiment were the dependent variables used in the two ANOVAs, respectively.

3.7

3.5

Frequency of Price Prumotion

96

JOURNAL OF AAARKETING RESEARCH, FEBRUARY 1992

Similarly, the testing of contrasts among different levels of price discounts reveals that only the difference in the brand's expected price between the 30% and 40% discount levels is significant at a ^ .05. Funher. the difference of the differences in the brand's expected price between the 30% and 40% price discount treatments and between the 10% and 30% price discount treatments is significant at a = . 10. Again, a plot of the average expected prices for the different levels of price discounts provides some directional support for a concave relationship between the expected price and the depth of price discounts (see Figure 2). Price Expectations Model of Brand Choice Findings from the calibration of the price expectations model representing the impact of price expectations on brand choice are reported in Table 2. As shown in the first column, the brand preference variable is the most statistically significant variable in the determination of brand choice, which is consistent with the fmdings of previous research on modeling brand choice (e.g., Guadagni and Little 1983). The retail price variable is significant at a = .05 but the short-term promotion variable is not significant at conventional significance levels. For the impact of price expectations on brand choice, we find that both the price gain and the price loss variables have the hypothesized signs, but only the effect of price loss is significant at the a = .01 level. Further, we observe that the magnitude of the coefficient of the price loss variable exceeds the corresponding value for the price gain variable, a finding consistent with the loss aversion phenomenon reported in the research cited previously. For the threshold parameters, 6 ^ ^ and 6/, delineating the boundaries of the price insensitivity region, we find that

the values providing the highest adjusted U^ for the brand choice model are each equal to 6% of the brand's average nonpromotional price. This Onding implies that price changes of 5% or less of the brand's average nonpromotional price did not produce a significant change in consumers' price perceptions. The finding of a latitude of acceptable price differences around the brand's expected price is also consistent with the empirical fmdings reported by Raman and Bass (1988) and Gurumurthy and Little (1989). A comparison of the goodness of fit of our price expectations model with that of a traditional brand choice model (see column 3 in Tabie 2) by the likelihood ratio test yields a x' value of 7.596 with 2 d.f., which is significant at a = .025. We also estimated a price expectations model without threshold effects in which price gains and losses were measured as differences between a brand's expected and retail prices. A test of non-nested hypotheses (Ben-Akiva and Lerman 1985, p. 171-174) between the price expectations model with threshold effects (shown in the first column of Table 2) and the price expectations model withottt threshold effects shows that the former provides a significantly better fit to the data at the a = .025 level. These fmdings suggest that the incorporation of price expectations, with threshold effects in consumers" perceptions of price gains and losses, can contribute significantly to the explanation of consumer brand choice behavior. Combined Price and Promotion Expectations Model of Brand Choice Lattin and Bucklin (1989. p. 229-300) report fmdings from analyses of the IRI academic coffee dataset, suggesting that "consumer expectations about future promotional activity are just as important to understanding consumer choice behavior as consumer expectations of price." Though we did not set out to elicit expectations for promotional activity as we did for price, our data do permit inference of consumers' promotion expectations on the basis of their past exposures to promotional activities. The question arises: How specifically do expectations of promotional activity affect consumer brand choice? We suggest that consumers who are exposed to price promotions on a brand to a degree beyond a prescribed threshold come to expect a discount every time they buy that brand. When a brand does not offer a promotional deal that a consumer has come to expect, we refer to the choice occasion as an "unfulfilled promotion expectation event." Clearly, unfulfilled promotion expectation events will have an adverse effect on consumer brand purchase probability. Analogously, we define an "unexpected promotion event" as the choice occasion when a consumer encounters a price promotion on a brand that he or she did not expect because the brand has engaged in minimal promotional activity. Obviously, unexpected promotion events will affect brand purchase probability positively. To model the impact of promotion expectations on

Figure 2 EXPECTED PRICE AS A FUNCTION OF THE DEPTH OF PRICE DISCOUNTS

E'^xptected Pr ice 4.3 4.2 4.1 4.0 3.9 3.8 3.7 3.6 3.5

i, ,

^T- ,

'

1 '.

'

io*

20% Mm^ Depth of Price Diiscounts

40%

CONSUMER PRICE AND PROMOTION EXPECTATIONS

97

Table 2
CALIBRATION OF THREE ALTERNATE CONSUMER BRAND CHOICE MODELS

Variable Brand-specific consiam Brand preference {PREF) Retail price (P) Promotion {PROM) Gain (8, = .06) Loss (9, = .06) Unfulfilled promotion expectations event Unexpected promotion event

Price expectations model -.100 (-.284) .066 (6.049)' -1.268 (-1.758)^ .685 (.8(M) .571 (1.077) -1.510 (-2.547)'

Coefficient estimates' Price and promotion expectations model .252 (.616) .068 (5.950)' -1.223 (-1.633)" .237 (.250) .491 (.798) -.946 (-1.440)" -1.589 (-2.168)'^ .721

Traditional model -.373 (-1.146) .062 (6.096)'' -1.315 (-1.898)' 1.460 (1.950)'

Adjusted U^ L^ig likelihood -63.169 (at convergence) Log likelihood -130.310 (at zero) 'Entries in parentheses are ihe /-statistics for the estimated eoefftcienis. ''Significam at a = ,01 in the one-tailed asymplotic /-test. 'Significant at a = .05 in the one-tailed asymptotic f-test. "significant al a = .10 in the one-tailed asymptotic /-test.

.5152 .4692

(1.128) .5379 .4765

.4861 .4554 -66.967 130.310

-60.215 -130.310

brand choice, we operationalize the promotion expectation variable as

(3) if brand i promotes 50% or more of the time as observed by consumer it, otherwise.

Note that the promotion frequency threshold of 50% or more of the time is adopted in the operationalization because it provides the highest adjusted U' for the brand choice model in a grid search. A similar result is reported by Lattin and Bucklin (1989)they fmd that when consumers observe a brand to be on price promotion more than half of the time, they come to expect a price discount on it and arc reluctant to buy it when it is not price promoted. In this connection, to represent the effects of promotion expectations on brand choice in our price expectations model (equation 2). we operationalize (I) an unfulfilled promotion expectation event variable as ETP^k = I and PROM,I, ='- 0 and (2) an unexpected promotion event variable as ETP,^ - 0 and PROM^^ 1. Thus we can capture the potential asymmetric effects on brand choice of not getting an expected promotion and getting a surprise or unexpected promotion. The findings from the calibration of the combined price and promotion expectations model are reported in the second column of Table 2. A likelihood ratio test of the combined price and promotion expectations model against the price expectations model yields a x' value of 5.908 with 2 d.f. of freedom and is significant at the a = .10 level. Thus, the incorporation of the promotion expectation effects is found to contribute significantly to the

ability to explain brand choice above and beyond the modeling of just the price expectation effects. All variables in the combined price and promotion expectations model have the hypothesized signs. However, of the two promotion expectation effects arising from the unfulfilled promotion expectation and unexpected promotion events, only the effect of the unfulfilled promotion expectation event is significant at a = .05. Because the coefficient estimate for the unfulfilled promotion expectation event variable bas a larger absolute magnitude and a larger lvalue than the ctwfficient for the unexpected promotion event variable, we conclude that, as in the case of price expectations, for promotion expectations losses loom larger than gains. This fmding is consistent with prospect theory (Kahneman and Tversky 1979, p. 279), which states that the "value function for losses is steeper than the value function for gains." It suggests that though offering frequent price promotions will increase the chance of consumers observing unexpected promotions, which then may produce short-tenn sales gains, those sales increases may be offset by the sales losses resulting from consumers not getting promotions they have come to expect. Our individual-level data enable us to track purchase behavior, in the presence and absence of a price promotion, separately for respondents whom we posit have come to expect a promotion versus those whom we posit have not. Recall that, on the basis of a grid search, the promotion frequency threshold at which the respondents are posited to form promotion expectations is 50% or more of the time. We now compare the impact of the presence of a price promotion on the respondents who

98 have come to expect a promotion on the brand with the impact on those who have not. Similarly, we also compare the impact of the absence of a price promotion on the respondents who have come to expect a promotion with the impact on those who have not. Table 3 shows average purchase probabilities with and without price promotions when a promotion on the target brand is expected and when it is not expected. We see that 59% of the 51 respondents who were offered a price promotion on the target brand and expected a promotion on it purchased the brand in contrast to 79% of the 42 respondents who were offered a price promotion on the target brand but did not expect it. The difference in the purchase probabilities is significant at a = .05. Correspondingly, the absence of a price promotion on the target brand is found to cause a larger reduction of brand choice probability among respondents who expected a promotion on the target brand than among those who did not expect a promotion. Specifically, only 14% of the respondents who expected a promotion on the target brand purchased it without a price promotion in contrast to 40% of the respondents who did not expect a price promotion on the target brand. The difference in the purchase probabilities is significant at a = .01. In sum, these findings reinforce our inference from the calibration of the combined price and promotion expectations brand choice model (see column 2 of Table 2) of a loss aversion phenomenon in the case of promotion expectations. Therefore, we conclude that consumers' response to the loss of an expected promotion is stronger than their response to a surprise or unexpected promotion. CONCLUSIONS We set out to investigate the impact of different price promotion schedules on brands' expected prices. Our findings from the tests of the two hypotheses pertaining to the impact of exposure to price promotions on consumers' price expectations reveal that the impact is significant. However, our experimental findings do not

JOURNAL OF MARKETING RESEARCH, FEBRUARY 1992

Table 3
CLASSIFICATION OF BRAND PURCHASE PROBABILITIES OF THE TARGET BRAND WITH AND WITHOUT PRICE PROMOTIONS BY PROMOTION EXPECTATIONS Promotion expectations' Expect a promotion Do not expect a promotion Promotion available 59%' (=30 of 5l> 79% (=33 of 42) Promotion not available 14% ( = 6 of 42) 40% ( = 21 of 53)

'In this study, rcsptindents are assumed to expecl a brand to be available an price promotion if they have seen il offered on price promoUon 50% or more of the time. ""A total of 30 of the 51 respondents who were offered a price promotion that they expected purchased the brand.

suggest rejection of the hypotheses that the brand's expected price is a linear function of the price promotion frequency and the depth of price discounts at conventional significance levels. Nevertheless, the results provide .some directional support for nonlinear relationships between the expected price and the two elements of a price promotion schedule. Given the important implications of such potential nonlinear effects of price promotions on brands' expected prices, further research testing those nonlinear effects of price promotions should prove fruitful for the design of optimal price promotion policies. We find evidence in support of a region of price insensitivity around a brand's expected price within which price changes do not produce a significant change in consumers' price perceptions. Price differences outside that region, in contrast, are found to have a significant impact on consumer brand purchase probability. Further, we observe that the effects of price gains may be offset by the effects of price losses when the price promotion is retracted because losses loom larger than gains. The former finding on the region of price insensitivity around a brand's expected price implies, as Gurumurthy and Little (1989, p. 21) note, "marketers wishing to increase prices should nibble, not bite. Small price increases are less hazardous if they stay within the latitude of acceptance." The latter finding on the loss aversion phenomenon suggests that marketers of products such as coffee and sugar, who experience wide fiuctuations in raw commodity costs, should attempt to .smooth out price fiuctuations instead of passing along entire increases and decreases In commodity costs to consumers. Our findings suggest that the sales gains realized from a price decrease that is outside the region of price insensitivity may nol compare favorably with the sales losses incurred as a result of an equal price increase. Our findings on promotion expectations suggest that unfulfilled promotion expectation event.s among consumers who have come to expect promotions on a brand because of frequent exposure to them will have an adverse impact on the brand. Analogously, unexpected promotion events wili enhance the probability of purchasing a brand among consumers who have not been exposed to many price promotions and therefore do nol as a rule expect the brand to be available on a promotional deal. We suggest that those results are consistent with the rational expectations view that "any policy rule that is systematically related to economic conditions, for example, one observed with stabilization in mind, will be perfectly anticipated, and therefore have no effect on output or employment" (Maddock and Carter 1982, p. 43). Policy actions that come as a surprise to people, In contrast, will generally have some real effect. Clearly, the design of optimal price promotion schedules requires consideration of the fact that an increase in the use of price promotions could erode long-term consumer demand by lowering the prices that consumers anticipate paying for the brand. Price promotional deals may come

CONSUMER PRICE A N D PROMOTION EXPECTATIONS

99 "Effect of Initial Selling Price on Subsequent Sales," Journat of Personatity and Social Psychology. 11 (4). 345-50. Doyle, Peter and John Saunders (1985), " t h e Lead Effect of Marketing Decisions," Joumal of Marketing Research. 22 (February) 54-65. Guadagni, Peter M. and John D. C, Little (1983). "A Logit Model of Brand Choice Calibrated on Scanner Data," Marketing Science. 2 (Summer). 203-38. Gurumurthy, K. and John D. C. Little (1989), "A Price Response Model Developed From Perception Theories," working paper, Sloan School of Management. MIT. Helson, H, (1964), Adaptation-Level Theory. New York: Harper & Row Publishers. Inc. Johnston, J. (1984), Econometric Methods. 3rd ed, McGrawHill Botjk Company. Kahneman, Daniel and Amos Tversky (1979), "Prospect Theory: An Analysi.s of Decision Under Risk," Econometrica, 47 (March), 263-91, Kalwani, Manohar U., Chi Kin Yim, Heikki J, Rinne, and Yoshi Sugita (1990), "A Price Expectations Mtxlel of Customer Brand Choice." Journal of Marketing Research. 27 (August), 251-62. Klein. Noreen M, and Janet E. Oglethorpe (1987). "Cognitive Reference Points in Consumer Decision Making." in Advances in Consumer Research, Vol. 14, Melanie Wallendorf and Paul Anderson, eds. Provo, UT: Association for Consumer Research. 183-7. Lattin, James M. and Randolph E. Bucklin (1989). "Reference Effects of Price and Promotion on Brand Choice Behavior," Journal of Marketing Research. 26 (August). 299310. Maddock, Rtxiney amd Michael Carter (1982), "A Child's Guide to Rational Expectations," Journat of Economic Literature, 20 (March), 39-51. McEadden. Daniel (1973), "Conditional Logit Analysis of Qualitative Choiee Behavior." in Frontiers in Econometrics, P. Zarembka, ed. New York: Academic Press, Inc. Monroe. Kent B. (1973). "Buyers' Subjective Perceptions of Priee," Journat of Marketing Research. 10 (February). 7 0 80. (1977), "Objeetive and Subjective Contextual Influences on Price Perception," Consumer and Industrial Buying Behavior. Arch G. Woodside, Jagdish N. Sheth, and Peter D. Bennett, eds. New York: North-Holland Publishing Company. 287-96. Muth. John E. (1961), "Rational Expectations and the Theory of Price Movement," Econometrica. 29 (July), 315-35. Nelson, Clarence W, (1977). "Rational ExpectationsFresh Ideas that Challenge Some Established Views of Policy Making," special report. Federal Reserve Bank of Minneapolis, Neslin, Scott A. and Robert W. Shoemaker (1989). "An Alternative Explanation for Lower Repeat Kates After Promotion Purchases," Journal of Marketing Re.search. 26 (May), 205-13. Putler. Daniel S, (1990). "Incorporating Reference Price Effects Into a Theory of Consumer Choice." working paper. Economic Research Service. USDA, Puto, Christopher P. (1987), "*The Framing of Buying Decisions," Journal of Consumer Research. 14 (December), 3 0 1 15. Raman. Kalyan and Erank M, Bass (1988), "A General Test of Reference Priee Theory in the Presence of Threshold Ef-

to be "perfectly anticipated" and have much less impact on consumer response than they do when they come as a surprise to consumers. Evaluation of the tradeoff between the short-term sales gain from a price promotion and the adverse effeel on future sales because of consumers forming price and promotion expectations requires knowledge of how price promotions affect the formation of consumers" expectations under different market conditions. We cannot, from the findings of our study, make more definitive recommendations about the optimal price promotion schedule. Such recommendations must come from analysis that explicitly considers cost data and information about competitive promotional activities. Our study and the others cited previously have begun to atte.st to the important role of price and promotion expectations in determining consumer purchasing behavior. Therefore it is critical for us to study how price and promotion expectations are formed. Experimental studies such as ours, we believe, afford an opportunity to conduct systematic investigations of how such expectations are formed. In our study, the direct measurement of brands' expected prices enabled us to examine the impact of different price promotion schedules on price expectations. However, we did not measure consumers' promotion expectations and had to rely on a grid search procedure to infer the promotion frequency threshold at which consumers come to form promotion expectations. In future research, a probabilistic measure of consumers' promotion expectations should be employed to obtain further insights into the formation of promotion expectations and their impact on brand choice behavior. REFERENCES Aronson. Elliot. Marilynn Brewer, and J. Merrill Carlsmith (1985), "Hxperinienlation in Social Psychology," The Handhooti of Sociat Psychology. Gardner Lindzey and Elliot Aronson, eds. New York: Alfred A. Knopf, I n c . Chapter 8. Ben-Akiva. Moshe and Steven R. Lcmian (1985). Discrete Choice Analysis: Theory and Application to Travel Demand. Cambridge. MA: MIT Press. Berkowitz, Eric N, and John R. Walton (1980). "Contextual Influences on Consumer Price Responses: An Experimental Analysis," Journal of Marketing Research. 17 (August), 34958. Curhan. Ronald C. and Robert J, Kopp (1986), "Factors Influencing Grocery Retailers' Support of Trade Promotions." Report No. 86-104. Boston: Marketing Science Institute. Delia Bina. Albert J. and Kent B. Monroe (1980), "A Multivariate Analysis of the Perception of Value Erom Retail Price Advertisements," m Advances in Consumer Research, Vol. 8, Kent B. Monroe, ed, Ann Arbor. Ml: Association for Consumer Research, 161-5. Dodson, Jtie A., Alice M, Tybout, and Brian Sternthai (1978), "Impact of Deals and Deal Retraction on Brand Switching," Journat of Marketing Research. 14 (February), 72-81, D(K)b, Anthony N,, J, Merrill Carlsmith. Jonathan L. Freedman. Thomas K. Landauer. and Saleng Tom. Jr. (1969),

100 fects," working paper. College of Business Administration, Universiiy of Florida. Rinne, Heikki J. (1981). "An Empirical Investigation of the Effects ol Reference Prices on Sales." unpublished PhD dissertation, Purdue University. Rowe. Debra and Christopher P. Puto (1987), "Do Consumers' Reference Points Affect Their Buying Decisions?" Advances in Consumer Research, Vol. 14, Melanie Wallendorf and Paul Anderson, eds. Provo. UT; Association for Consumer Research, 188-92. Sawyer. Alan G. and Peter H. Dickson (1984). "Psychological Perspectives on Consumer Response to Sales Promotion." Research on Sates Promotion: Collected Papers. Report No. 84-104, Katherine E. Jocz, ed. Boston: Marketing Science Institute. 1-21. Sherif, Carolyn W. (1963), "Social Categorization as a Function of Latitude of Acceptance and Series Range." Journal of At^normal Psychology, 67. 148-56. Shoemaker, Robert W. and F. Robert Shoaf (1977), "Repeat Rates of Deal Purchases." Journat of Advertising Research, 17 (April). 47-53,

JOURNAL OF MARKETING RESEARCH, FEBRUARY 1992

Thaler, R. (1985), "Mental Accounting and Consumer Choice." Marketing Science. 4 (Summer). 199-214. Tversky. Amos and Daniel Kahneman (1974), "Judgment Under Uncertainty: Heuristics and Biases," Science, 185, 11243!. Uhl. J. N. and Harold L. Brown (1971). "Consumer Perception of Experimental Retail Food Priee Changes," Journat of Consumer Affairs. 5 (Winter). 174-85. Winer. Russell S. (1985). "A Price Vector Model of Demand for Consumer Durables: Preliminary Development," Martaeling Science. 4 (Winter), 74-90. (1986). "A Reference Price Model of Brand Choice tor Frequently Purchased Products." Journat of Comumer Research. 13 (September), 250-6. Zeithaml. Valarie A. and Karen L. Graham (1983). "The Accuracy of Reported Reference Prices for Professional Services," Advances in Con.sumer Research, Vol. 14. Richard P. Bagozzi and Alice M. Tybout. eds. Ann Arbor, MI: Association for Consumer Research, 607-11.
Rcprini No. JMIU91107

You ccai have it all...

What you 're reading now plus the important articles in 800 other business and management magazines, in a matter of minutes. The ABI/INFORM^*" business database and your computer give you access to article summaries from magazines worldwide. Call the publishers of ABI/INFORM at 800/6262823, today

Vr.