Perceived Correlates of Store Price Image: an Application of the Bootstrap

B. Kemal Buyukkurt, Concordia University
Meral D. Buyukkurt, Concordia University
[ to cite ]:
B. Kemal Buyukkurt and Meral D. Buyukkurt (1986) ,"Perceived Correlates of Store Price Image: an Application of the Bootstrap", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 42-47.

Advances in Consumer Research Volume 13, 1986      Pages 42-47


B. Kemal Buyukkurt, Concordia University

Meral D. Buyukkurt, Concordia University

[This study was funded by a research grant from Marketing Department, Graduate School of Business, Indiana University.]

The findings of this nonexperimental study suggests that consumers perceive correlations between store attributes and store price image, and may use the former to predict the latter. Within a multiple regression context, the propensity to heuristically use such perceived predictors seem to be related to (1) perceived difficulty of basing the price image judgment exclusively on self acquired price samples, (2) time pressures, (3) self assessment of experience as a shopper, and (4) the extent to which price variation is expected across stores. Also, an application of the bootstrap, a resampling procedure, is presented which provides a nonparametric estimate of the expected error of prediction for the estimated regression equation.


Tn a comprehensive study published more than a decade ago, Brown and Oxenfeldt (1972) reported that consumers perceived covariation between various store attributes and store price levels, and used the information regarding the former to predict the latter. One of the two major objectives of this study is to examine whether and to what extent, perceptions as such still exist today. The second objective is to relate the propensity to use such still exist today. The second objective is to relate the propensity to use such perceived correlations as predictors of price image to individual and task related variables. Both types of information should be valuable to managers in charge of pricing decisions at the retail level, since price image is theorized (Berry 1969 Hansen and Deutscher 1977-78, Kelly and Stephenson 1567), and reported to affect store choice (Progressive Grocer 1981, 1982).


Perceived Covariates of Store Price Image

Judgements of covariation between events is a critical aspect of cognitive behavior. Knowledge regarding which attributes or events are related and the degree of their relationship helps individuals to infer causal relationships between the variables observed in the past, control some events in the present through manipulation of a set of covariates, and predict the future or present value of a criterion given the values of some diagnostic predictors (Bettman, John and Scott 1984, Crocker 1981, Nisbett and Ross 1980, Hogarth 1980). Indeed, many psychological and socio-psychological theories are based on the notion that individuals detect covariation between cues, variables and events. Examples are Brunswik's Lens Model (e.g., Castellan 1973), Andersons' Functional Measurement (1981), and Kelley's Attribution Theory (1973).

Similarly, Brown and Oxenfeldt's (1972) findings suggest that consumers assume covariation between various store attributes and store price levels and use the information regarding the store attributes to predict store price levels. This large scale survey covered five cities (Havertown, Greensboro, New York, San Francisco, St. Louis), 60 different food retailing outlets and included more than 1,000 interviews with housewives regarding the price of 80 product items at different stores in their neighbourhood.

Respondents in each of the five cities were asked to rank perceived price levels given 11 store attributes. These attributes included whether the store was new, untidy, in a large shopping center, engaged in lots of advertising, had wide assortment, promoted "loss leaders', gave trading stamps, had an expensive interior, was open late, provided extra services and was small. A high degree of similarity existed across five communities in terms of rankings as summarized by a high coefficient of concordance (.92).

The authors concluded that consumers expected extra costs incurred by a store to be reflected in higher prices: in descending order of importance extra services offered by the store, late hours of operation, having expensive interiors and giving out trading stamps are perceived to be cues related to high prices. On the other hand, store attributes which may be regarded as indicators of large volume of operations were predictors of lower prices: being located in a large shopping center, lots of advertising and having a wide assortment of products were correlated with low prices. Chile small size was perceived to be the strongest correlate of high prices, untidiness and being a "new" store in the community were generally ranked at the top of indicators of low prices. The authors stated that, "These findings sketch the beginning of a model which links operating characteristics of stores to perception via cues involving expenses and volume" (Brown and Oxenfeldt 1972, p. 49).

Heuristic Use of Perceived Covariates as Predictors

Brown and Oxenfeldt (1972) argued that consumers heuristically use store attributes in their price image judgments as part of a task simplification strategy possibly because of the complexity of the price structure at stores. They found that, for a basket of 80 product items, a given store may have higher than average prices for some items and lower than average prices for other items in the basket. Therefore, they implied that, overall price image of the store may be cognitively difficult to asses only by sampling prices from the marketed assortments, but easier to predict from non-price store attributes perceived to be correlated with store price image. Consequently, consumers can rely on the perceived correlates if they believe a more careful decision rule such as examining the prices of a self-selected basket of items is not likely to improve the outcome of the judgment process substantially (Svenson 1979) and if the cost of the first strategy outweighs the expected benefits from it (Christensen-Szalanski 1978). The use of such predictors are encouraged also when they are perceived to be causally related to the criterion of interest as in this context (Azjen 1977). Brown and Oxenfeldt concluded (1972, p. 42):

"This deductive process of applying a broad generalization to specific situations occurred in several phases of our research. It is one of the more pervasive results. Misperceptions often occur when the environmental situation is contrary to intuitive or logical generalizations. The generalization, not the facts, gives rise to perception. The facts seldom alter a logical conclusion."

Similar store profile effects have been reported by Wheatley and Chiu (1977) in consumer judgments of product quality. Consumer predictions of one attribute value from a correlated attribute have also been observed in price-quality tradeoffs (Levin and Johnson 1984) and product evaluations (Huber and McCann 19825.

In general, use of judgmental heuristics are found to be encouraged by factors which induce cognitive complexity such as time pressure, distractions, and information overload (e.g., Hogarth 1980, Newell and Simon 1972, Payne 1976, Wright 1974). Such factors may operate in the judgment task studied here and therefore motivate prediction of price image from perceived correlates.. Distractions such as crowding,noise and indoor advertising within a store may aggravate the complexity due to price structure mentioned above. Also, some consumers feel more time pressure than others and/or perceive the comparison of the price images of different stores based on self sampled prices to be more difficult than the comparison of the price images based on perceived correlates of price image. Therefore, perceived time pressures may lead to higher propensity to use the perceived correlates as predictors of price image.

As alternatives to predicting store price image exclusively from perceived correlates, the prediction based on self acquisition and cognitive integration of a price sample independent of the perceived correlates or in conjunction with them requires examination of a basket of prices. In such a task, the magnitude of the deviations of the sampled prices from some subjective reference prices (Monroe 1973) need to be judged and cognitively integrated into an overall judgment about the observed basket and then generalized to the whole store as the price image of the store. Thus, the comparison task and the final judgment will involve some uncertainty due to uncertainty associated with the subjective reference prices since their recall is expected to be less than perfect especially for infrequently purchased and low priced items.

Since decreases in the uncertainty regarding the subjective reference prices is likely to alleviate the perceived difficulty of basing the price image judgment on a price sample, those individual related variables which enable or stimulate more attention to and better recall of prices are expected to lead to lower propensity to use the perceived correlates as predictors of price image. Self assessments of experience as a shopper, accuracy of recall of prices, overall sensitivity to prices, perceived budget constraint, and the extent to which shopping is enjoyed can be suggested as a possible list of such individual related variables.

Finally, if the consumer believes that there are no significant price differences from one store to another (say, in food retailing) then the propensity to heuristically use the perceived correlates should degrease because the need to compare the stores in question with regard to price subjectively does not exist.

Given the above discussion, Table 1 summarizes the expected relationships between the propensity to use the perceived correlates of store image as predictors of price image and the individual related variables. In order to build on the previous research by Brown and Oxenfeldt (1972), the shopping context is limited to grocery purchases. The individual related variables are presented in Likert scale form as they appear on the questionnaire.


A mail survey was conducted to collect the data of the study. The questionnaire included the Likert scales exhibited in Table 1 and the items in Table 2. The second set of items attempts to measure the direction and the degree of perceived correlation between each of various store attributes and store price image on a 5-point scale ranging from -2 on the left to +2 on the right. By presenting the poles of the scale as the characteristics of two different stores and by instructing the subjects to indicate the store they would expect to have a "higher" or "somewhat higher" grocery bill, both the magnitude and the direction of correlation is measured through a single response. The central response category reflects perceived lack of correlation between that attribute and the overall price image of the store. The store attributes in the table are based on the Brown and Oxenfeldt study (1972) and informal discussions with eight experienced grocery shoppers.

One potential problem with the instrument as it is presented in Table 2 is that the store description presented on the same side of the page may be perceived as a profile of a store and the attributes examined earlier in the questionnaire may affect the responses to the later items in the questionnaire. This may happen despite the two warnings on the questionnaire to evaluate each attribute (i.e., pair of stores in a given item) independently. In order to cancel out such "profile effects," three forms of the questionnaire were constructed. The second form presented the last one third of the items at the top, and the left pole descriptions were interchanged with the ones on the right pole. The third form moved the middle one third of the items to the top of the list and similar interchanges of the poles were made.

A total number of 270 questionnaires (i.e., 90 of each fora) were mailed to potential respondents in four adjacent zip code areas in a Midwestern city. A city directory was used as the sample frame and the addresses were selected through a systematic random sampling method. The primary grocery shopper was asked to fill out the questionnaire. Thirty-five questionnaires could not be delivered. Eighty-nine questionnaires were returned making the response rate 37.9 percent. Four of the returns were not useable. The final number of useable responses for the three forms of the questionnaire were 27, 33 and 29.


Perceived Correlates of Price Image

The data regarding the perceptions of correlation between the store attributes and the store price image are summarized in Table 3 and Table 4. While the t-tests on means in Table 3 test whether the data confirm the expected direction of the perceived correlations, Table 4 summarizes the x2 tests on variances which attempt to assess the consistency of the perceived correlation across the sample.

As it is presented in Table 3, consumers expect extra operating expenses and investments in the store to be reflected in higher prices (rarely waiting a long time in the check-out line because a sufficient number of lines are open, ready availability of salespeople, being located at a major shopping center and therefore paying higher rent, [In Brown and Oxenfeldt's (1972) study, being "located at a major shopping center' was perceived to be related to higher volume, and, therefore, associated with lower prices. During the questionnaire construction stage, the housewives who were interviewed associated it with higher rents, therefore, higher prices. The respondents of the survey seem to believe along the same lines.] being open 24 hours, being tidy and neat, displaying products on the shelves rather than carton boxes, providing extra services, having elegant store and lighting). Similarly, small volume of business at the store and company level is believed to indicate higher prices (few shoppers on Saturday, one of the smaller food retailers in the city, family owned, independent store rather than being part of a chain organization which owns a large number of stores). Also, if the assortment is extended to include gourmet food, deli and bakery, and non-food products, price image is adversely affected. [This finding regarding extension of the assortment should be viewed with caution since the scales in Table 2 ask the respondents to indicate the store for which they expect their grocery bill will be higher. It is possible for some consumers to think of some additional items that they can purchase when the assortment is richer and therefore except their grocery expenses to increase for the store with the extended assortment rather than thinking about possible differences in the prices of a given basket of items in the two stores described as the voles of the scale.] Finally, high frequency of printed retail advertising and large number of featured items per advertising are perceived to be associated with lower prices. Overall, these results parallel Brown and Oxenfeldt's (1972) findings.

The above perceived correlations seem to be consistent across the sample as suggested by the x2 tests on variances of the perceived correlation scores for each store attribute. It was judged to have a variance less than unity for each attribute so that only a relatively small percentile o the distribution of the scores would extend over to either side of the central category. Store attributes numbered 1, 2, 3, 4, 5, 7, 8, 11, 14, 15 and 16 met this criterion.



Prediction of store price image from the above store attributes was confirmed by two follow-on studies. In the first study, two store profiles were constructed by retaining the subset of the above attributes for which the average score was significantly different from zero (implying positive or negative perceived correlation) and variance less than or equal to one (implying relatively consistent perceived correlation across the sample). While the combination of those verbal descriptions which defined the negative poles of the scales constituted the "high" priced store profile, the combination of the verbal descriptions at the positive poles made up the "low" priced store profile (Buyukkurt 1985). The two store profiles were tested in a drop-and-pick-up survey using an independent sample and found to lead to significantly different perceptions of "high" versus "low" priced stores (t=12.30, d.f.=62, p=.000). ConFidence in these judgments was high for both store (x > 7.4 given a 10-point rating scale) but not significantly different across the two stores (t=1.35, d.f.=62, p=.163). In this survey, field workers contacted 75 residences and completed 69 interviews, out of which 64 were usable. 29 and 35 respondents were exposed to the "high" and "low"-priced store profiles respectively.

The same two profiles were later used in an experimental setting as a manipulation of store profile effects where 240 grocery shoppers judged the perceived value of a self selected basket of grocery items. Again, price judgments were affected by store profile (F=5.886, d.f.= 1,190, a = 0.02)(Buyukkurt 1985),.

Regression Results

The hypotheses in Table 1 regarding the relationship between the individual related variables (independent variables) and the propensity to use the perceived correlates to store image as predictors of store image (dependent variable) were tested by conducting a step-wise multiple regression analysis.

The dependent variable was operationalized by summing the absolute values of the perceived correlation score across all store attributes in Table 2 for a given subject. This sum increases (1) as the degree of perceived covariation between an attribute and price image increases and (2) as the number of such perceived covariations increases. As such, the sum, as an indicator of the propensity to use the perceived correlates as predictors of price image, is based on the assumption that the probability that a cue will be used in predicting a criterion increases as the perceived predictive validity of that cue increases.

Also, the higher the number of such predictors, the more likely it is that the prediction will be affected by such perceived correlates. [Naturally, the number of predictors that can be used does not increase indefinitely because of information processing capacities of human beings.] If the degree of perceived correlation between an attribute and price image is conceptualized as the strength of a belief, then a theoretical justification can be provided for the above operationalization, for example, from the attitude research (Fishbein and Azjen 1975) and research on information search (Duncan and Olshavsky 1982) where beliefs and tendency to act are theorized and shown to be related.

Stepwise regression was preferred to ordinary least squares mainly because of the exploratory nature of the study. Selection of the variables into the equation was controlled through "tolerance" and on t-value-to-enter. Technically, tolerance is equal to one minus the squared multiple correlation between an independent variable which has not yet been included in the equation and those that are already in the equation. In this study, tolerance was set equal to 0.9 F-value-to-enter was 2.8 which assures statistical significance at a=.10 for the regression coefficients of the entered variables.

A stepwise inclusion history of the independent variables is presented in Table 5. Four independent variables were included in the equation with final adjusted R2 = .38. In their order of inclusion, these variables are perceived difficulty of making price image comparisons by sampling prices (X7), not being able to pay enough attention to grocery prices because of time pressures (X4), belief that no significant differences exist in grocery prices from one store to another (X3), and self ratings of experience as a grocery shopper (X2). [As far as multicollinearity is concerned, the null hypothesis of zero population correlation coefficient could be rejected at a = .05 only for the sample correlation between sc and Y7 (r = .475). Because of the stringent tolerance value, it is apparent that X5 was kept out of the equation after the entry of X7.] Except for the last variable, the signs of the regression coefficients are expected. The positive coefficient for X2 indicates that the propensity to use the perceived covariates as indicators of price image increases as self perception of experience as a shopper increases. A posteriori, this finding is not extremely surprising because (1) the perceived covariates may in fact be valid predictors of price image in the market for this sample and their beliefs may have been confirmed over time or (2) since people often see what they expect to see (Hogarth 1980, Nisbett and Ross 1980), confirmation of their hypotheses may have been mainly illusory since accurate judgments based on self acquired price samples will involve substantial uncertainty.

The Bootstrap and Cross-Validation

One immediate question about the regression findings is the predictive ability of the final model since the regression weights are based on a relatively small sample size. To eliminate this concern the bootstrap was used to estimate the expected excess error of prediction (Efron 1979) by randomly drawing samples of a fixed size with replacement from the original sample. A parallel estimate of expected excess error was obtained from another method of sample reuse widely known in the marketing literature as cross-validation (Appendix ).

The expected error of prediction is 2.237 and 2.120 for cross-validation and the boostrap respectively which are not substantially large given a dependent variable with a mean of 14.31, a minimum value of 4.0 and a maximum value of 31.


More than a decade after Brown and Oxenfeldt's (1972 study, the data from a geographic area which was not included in their study confirmed that consumers still predict store price image through store attributes which are perceived to be indicators of expense and volume. Furthermore, these perceived correlations seem to be rather consistent as suggested by the sample studied here.



The propensity to use such perceived correlations in predictions of price image seem to be positively related to (1) perceived difficulty of basing the price image judgment on self-acquired samples, (2) time pressures, and (3) self perception of experience as a shopper, and negatively related to whether the consumer believes that there are significant differences in terms of price across stores. the regression coefficients on which the above conclusions are based seem to be stable and the expected prediction error seem not to be excessive as indicated by the nonparametric resampling methods. These methods, however, should not be regarded as perfect substitutes for sufficiently large sample sizes, and therefore, this study should be replicated by taking larger samples, preferably in different geographic areas, to reflect possible variations at the retail level. Such an effort is already under way with replications in the northeast and the northwest.

Further research in this area seems to be warranted. Knowledge about the cognitive integration of the perceived correlates in a judgment context and the relative subjective importance assigned to each correlate should be valuable to retail strategists. Also, perceived correlates of store price image can he conceptualized as beliefs (Duncan and Olshavsky 1982) or hypotheses Bettman, John and Scott 1984), and their development and revision as new information is gathered can be studied.








Anderson, Norman H. (1981), Foundations of Information Integration Theory, New York: Academic Press.

Azjen, I. (1977), "Intuitive Theories of Events and the Effects of Base-Rate Information on Prediction," Journal of Personality and Social Psychology, 35, 303-314.

Berry, L.L. (1969), "The Components of Department Store Image: A Theoretical and Empirical Analysis," Journal of Retailing, 45 (Spring), p. 3-20.

Bettman, James R., Deborah Roedder John and Carol A. Scott (1984), 'Consumers' Assessment of Covariation," in Thomas C. Kinnear (ed.), Advances in Consumer Research, Vol. XI, Provo, UT, Association for Consumer Research.

Brown, F.F. and A.R. Oxenfeldt (1972), Misperceptions of Economic Phenomena, New York: Sperr and Douth.

Buyukkurt, B.K . (1985), "Integration of Serially Sampled Information: Modeling and Some Findings," Working Paper Series, Faculty of Commerce and Administration, Concordia University, Montreal, Quebec, Canada.

Castellan, N.J. (1973), "Comments on the 'Sen's Model" Equation and the Analysis of Multiple Cue Judgment Tasks," Psychometrica, (March), 87-100.

Christensen-Szalanski, Jay J.J. (1978), "Problem Solving Strategies: A Selection Mechanism, Some Implications, and Some Data," Organizations Behavior and Human Performance, 22, p. 307-323.

Crocker, Jennifer (1981), "Judgment of Covariation by Social Perceivers," Psychological Bulletin 90, 2, p. 272-292.

Duncan, Calvin P. and Richard W. Olshavsky (1982), "External Search: The Role of Consumer Beliefs," Journal of Marketing Research, XIX (February),

Efron, Bradley (1979), "Bootstrap Methods: Another Look at the Jacknife," Annals of Statistics, 7, 1-26.

Fishbein, Martin, and Icek Azjen (1975), Belief Attitude, Intention and Behavior, Reading, Mass : Addison-Wesley.

Hansen, R. and T. Deutscher (1977-78), "An Empirical Investigation of Attribute and Importance in Retail Store Selection," Journal of Retailing, (Winter), p. 59-73.

Hogarth, R.M. (1970, Judgment and Choice, Chichester: Wiley.

Huber, Joel and John McCann (1972, "The Impact of Interential Beliefs on Product Evaluations," Journal of Marketing Research, XIX (August) 324-333.

Kelley, H.H. (1973), "The Process of Causal Attribution," American Psychologist, 28, 107-128

Kelley and R. Stephenson (1967), "The Semantic Differential: An Information Source for Designing Retail Patronage Appeals," Journal of Marketing, (October), p. 43-47.

Levin, Irwin P. and Richard D. Hohnson (1984), Estimating Price-Quality Trade-offs Using Comparative Judgments," Journal of Consumer Research, 11 (June), p. 593-600.

Monroe, Kent (1973), "'Buyers' Subjective Perceptions of Price," Journal of Marketing Research, 10 (February), p. 70-80.

Newell, A. and Simon, H.A. (1972), Human Problem Solving, Englewood Cliffs, N.J.: Prentice-Hall.

Nisbett, Richard and Lee Ross (1980), Human Inference: Strategies and Short-comings of Social Judgment, Englewood Cliffs, N.J.: Prentice-Hall.

Payne, John W. (1976), "Task Complexity and Contingent Processing in Decision Making: An Information Search and Protocol Analysis," Organizational Behavior and Human Performance, 16, p. 366-387.

Progressive Grocer (1981). July. p. 90-91.

Progressive Grocer (1982), July, p. 74-75.

Svenson, Ola (1979), "Process Descriptions of Decision Making," Organizational Behavior and Human Performance, 23, p. 86-112.

Wheatley, John S. and John S.Y. Chiu (1977), "The Effects of Price Store Image, and Product and Respondent Characteristics on Perceptions of Quality," Journal of Marketing Research, XIV (May) 91-186.