The Effects of Sampling and Information on Brand Choice When Beliefs in Quality Differences Are Ambiguous

George Brooker, University of Puget Sound
John J. Wheatley, University of Washington
John S. Y. Chiu, University of Washington
ABSTRACT - Extant research has suggested or demonstrated that when consumers believe large quality differences exist among brands in a product class there is a tendency for consumers to choose the higher priced Alternative. These studier have failed to answer a number of important questions. For example, what is the relative effectiveness of information and sampling in affecting preference and can monetarily constrained choice be influenced when beliefs about quality differences among brands in a product category are ambiguous.
[ to cite ]:
George Brooker, John J. Wheatley, and John S. Y. Chiu (1986) ,"The Effects of Sampling and Information on Brand Choice When Beliefs in Quality Differences Are Ambiguous", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 272-276.

Advances in Consumer Research Volume 13, 1986      Pages 272-276

THE EFFECTS OF SAMPLING AND INFORMATION ON BRAND CHOICE WHEN BELIEFS IN QUALITY DIFFERENCES ARE AMBIGUOUS

George Brooker, University of Puget Sound

John J. Wheatley, University of Washington

John S. Y. Chiu, University of Washington

ABSTRACT -

Extant research has suggested or demonstrated that when consumers believe large quality differences exist among brands in a product class there is a tendency for consumers to choose the higher priced Alternative. These studier have failed to answer a number of important questions. For example, what is the relative effectiveness of information and sampling in affecting preference and can monetarily constrained choice be influenced when beliefs about quality differences among brands in a product category are ambiguous.

INTRODUCTION

Prior beliefs about large quality differences among brands in a product category with a tendency for consumers to prefer the higher priced alternative in such circumstances has been suggested or demonstrated by a number of researchers. (Leavitt 1954, Tull Boring and Gonsior 1964, Lambert 1972, Obermiller and Wheatley 1984). However, the earlier studies did not attempt to measure preference when money was involved and the more recent study, while imposing a monetary constraint on choice behavior failed to answer a number of important questions.

First, the work by Obermiller and Wheatley (1984) on price effects indicated that individuals changed beliefs and behaviors consistent with self-perceptions of evidence given then about products. The nature of the evidence given was a taste test of (identical) brands and information about the brands. The information given was either "perfect" (indicating no essential differences existed between brands) or irrelevant (sales figures, usage data, user demographics). All participants in the study got both a taste of the brands and one of the information treatments in a before-after between-subjects design. From the perspective of the marketing change-agent, there is so e uncertainty regarding which of the treatments (taste and information) was influential in causing the changes that were reported. Although there is little reported research on the relative effectiveness of sampling versus advertising, possibly because both are typically used simultaneously in promotional efforts, the former is generally regarded as the more effective method of persuading consumers to choose a product (Kotler 1984, p. 663). Product trial is thought to generate stronger attitudes than advertising at least in part because, unlike advertising, there is little risk of source derogation, discounting and counterarguing (Smith and Swinyard 1983). Preferences based on sensory experience with products are viewed by consumers as valid and usually unquestioned (Bem 1970). Therefore, we hypothesize that sampling will be more influential than information in determinant choice behavior.

Pre-measures and post-measures of beliefs about brand differences and strengths of those beliefs were taken by Obermiller and Wheatley. These beliefs along with changes in perceptions of quality differences were used to assess switching patterns between low and high priced brands. While switching patterns were consistent with beliefs, some of the patterns were inconsistent (e.g., people initially choosing a low priced brand, subsequently changed to a high priced brand). A second question which arises from these results deals with the nature of the beliefs that are changed (leading to changes in preference). Some beliefs are stronger than others, and are difficult to change. Rokeach (1969) has discussed the sources of beliefs as one reason for difficulty in changing them. Generally, beliefs emanating from personal experience are most difficult to change; beliefs arising from an authority figure will be somewhat less resistant to change, changing when the authority changes; inconsequential beliefs, such as those developed by advertising are easiest to change. From this perspective, it would be expected that individuals whose beliefs about products are based on personal experience are less likely to change those beliefs (and consequent behavior) in response to marketer activities than would those whose beliefs are less resistant to change as, for example, advertising based beliefs.

The third question, and perhaps the most important, has to do with the types of products chosen by Obermiller and Wheatley to demonstrate the effect of prior beliefs on brand preference. The product categories, colas and pop corn, were ones that consumers believed contained either large (cola) or negligible (popcorn) quality differences, among brands. In reality, a large number, if not most product categories are believed to include brands that are quite similar in terns of quality. The issue of how consumers utilize price as an indicator of quality in such circumstances was not addressed. If the hypothesis that individual differences, such as prior beliefs about the presence or absence of quality differences among brands in a product category, determines whether consumers utilize price as a quality cue is true, we should anticipate ambiguous results in the case of product groups that are viewed as consisting of brands that, while not alike in terms of quality, are not very different from one another. An appropriate choice of products should offer an opportunity to demonstrate discriminant validity.

Consequently, this research has four major objectives: (1) to determine the relative effects of sensory (taste) and cognitive (informational) experience on quality assessments and behavior, (2) to ascertain the relationship between the source and strength of beliefs and subsequent effects on perceptions of quality and behavior and (3) to infer the effects of price on quality perceptions and choice when neither very large nor very small quality differences are thought to exist among brands. A fourth objective was to verify the generalizability of previous results to adult consumers, since the work by Obermiller and Wheatley was done primarily on student populations.

METHODOLOGY - EXPLORATORY RESEARCH

A pretest was performed on twenty-two products to select those appropriate for the subsequent experiment. The products studied were food products that could be used in a taste-test procedure.

Products were separated into four groups for the pretest. Convenience samples of 32-35 people rated the products (on a seven point scale) in each group for quality differences among brands in the product class. Selection for the experiment was based on a reasonably uniform distribution of ratings across the entire scale and a mean close to the scale midpoint. The products chosen for the research from the pretest were potato chips and frozen orange juice.

METHODOLOGY - CAUSAL RESEARCH

Two separate 2x2x2 split plot factorial designs were used. For each of the two products, two information conditions (relevant or irrelevant) and two sample conditions (taste or no taste) were created between subjects. The two price conditions were within subjects, that is, all subjects were exposed to the high and low prices then prevailing in the marketplace for potato chips and frozen orange juice.

Data collection was performed at two different sites (grocery stores in a medium-sized Northwestern city). One product was studied at each research site to minimize the potential for cross-product contamination by reusing respondents. The research sites were about ten miles apart. Only one person was run in the experiment at one time to minimize respondent suspicion regarding the sample condition. (Having one person sample the product while a second was not allowed to sample at the same time has the potential to create questions regarding the procedure in the minds of respondents.) Individuals were randomly assigned to experimental conditions at each site.

In the procedure, individuals were screened to be sure they used the products. Then they were asked about their beliefs regarding differences among brands of the product and about the strength of these beliefs. This was followed by three questions on possible sources for their beliefs: personal experience, knowledge gained from advertising, and expert opinion. the remainder of the experiment was completed by the respondent, with the experimental treatment and all remaining measures contained in a booklet.

Experimental treatments were couched as information drawn from articles in Consumer Reports. In the relevant information conditions, it was stated that there were no essential differences between brands of the product and a blind taste test panel could not tell the brands apart. In the irrelevant information conditions, no mention was made of brand differences. The various treatments are presented in Appendix A.

Following the presentation of the information condition, individuals were shown samples of the (identical) product with "brand names" Xb and Wt attached. (Xb and Wt have been demonstrated to be neutral "brands.") Retail prices of higher priced and lower priced brands were included as part of the "brand" information. (The higher priced brand was Xb.) In the taste condition, people were given samples of each "brand" at this time. Individuals then rated the quality of both brands on a five-point scale, after which they made a choice between the brands.

By way of introducing a monetary constraint, the choice was presented as the prize the individual would win if selected in a drawing. If the lover priced brand were chosen, the individual would receive that brand plus the cash difference between the lower and higher priced brand. If the higher priced brand were chosen, the individual would receive only the product.

After the choice was made, individuals were asked about the reasons for making their choice; views on Consumer Reports and manipulation checks were taken; demographics were collected and individuals were asked their beliefs regarding the purpose of the experiment. Upon completing the booklet, the people were thanked for their participation and were given a folded and stapled sheet of paper which they were asked to open and read when they got home. This sheet debriefed the subjects; it explained that some of the information given was not correct, and related it to the general purpose of the experiment.

RESULTS

[We wish to acknowledge the assistance of Ms. J.M. Morris in the computer analysis of the data generated in this experiment.]

The profiles of the groups participating in the two studies were quite similar except for income. In the orange juice experiment, 74 people participated: 23 males, 50 females, and one was unreported. Median age was 34.1 years; 69% had more than a high school education and the median income was $19,615. In the potato chips study, 77 people participated: 23 males, 53 females and one unreported. Median age was 34.8; 75% were educated beyond high school and median income was $29,999.

Manipulation checks focused on the relevant and irrelevant information treatments. The checks measured on a five point scale the strength of agreement by respondents with factual statements about the information given to then "from Consumer Reports." The groups compared, on an individual product basis, were those receiving the relevant and those receiving the irrelevant information treatments. T-tests were used to sake the comparisons. For potato chips, those receiving the relevant information agreed ore strongly (p < .05) that there were "no essential differences between brands," and those receiving the irrelevant information agreed were strongly (p < .05) with a factual statement about the consumption of the product in the U.S. for orange juice, the irrelevant treatment check (about consumption) produced a significant (p < .05) and parallel difference between groups. The relevant treatment check was not significant (p < .30), but the group means showed that the relevant information group was in greater agreement with the "no difference" statement. Therefore, it was concluded that the treatments were effective.

T-tests between information treatment groups on their views of the factualness of Consumer Reports did not yield significant differences for either product. Similarly a question about the usefulness of the information from Consumer Reports in helping respondents choose between brands produced no significant (or substantial) differences.

Responses to the two product studies were analyzed separately. Table One presents the results of the treatment effects on choice behavior in the orange juice experiment. As hypothesized, the effect of sampling the product had a significant effect on the choice made, with those receiving the sample tending to choose the lower-priced (Wt) brand. (Table Two shows the proportions of those in the sampling conditions choosing each brand.) Information did not affect choice, and the interaction effect was insignificant. [This method of analysis say be unfamiliar to the reader. Briefly, the vertical axis in the ANOVA, which is cast in a regression ode, ranges from O to 1 and represent the probability of purchase of the expensive brand. Non-random choice behavior would differ significantly from .5 in either direction. The intercept in this case was .536 with a standard error of .055. Thus the inferred lack of significance of price.] The inferred price effect was also weak, as anticipated.

TABLE 1

ORANGE JUICE: ANOVA ON CHOICE USING UNIQUE SUMS OF SQUARES, SAMPLING AND INFORMATION AS DUMMY VARIABLES

[The response function involving a binary dependent variable is essentially linear in the middle range. Therefore, the logit transformation was not necessary here. (See graph, page 329, Neter and Wasserman 1974.)]

TABLE 2

ORANGE JUICE: PROPORTIONS CHOOSING BRANDS BASED ON SAMPLING CONDITIONS

Analysis of Covariance using initial beliefs in brand differences, strength of those initial beliefs, belief difference x belief strength and confidence in the factualness of Consumer Reports as covariates did not change the pattern of significance shown in Table One. None of these covariates was significantly associated with brand choice.

In the belief that absolute quality ratings of each of the brands were not meaningful in themselves, a variable was created to reflect the differences in relative quality ratings given to the brands by each person in the study. This was done by subtracting the quality rating given to it from the quality rating given to Xb. An ANOVA performed on this variable indicated that differences in relative quality ratings were not affected by the information and sampling treatments in the orange juice experiment. The inferred effect of price was, however, marginally significant ( p < .068) as indicated in Table 3. The relative quality ratings apparently were the result of the price cue being utilized by the subjects to perceive quality differences where, objectively speaking, none existed. (In this ANOVA, we infer the price effect from the intercept; the ordinal axis is scaled to reflect differences in the quality ratings of the expensive and inexpensive "brands.") The same concomitant variables used to examine the choice process also yielded nothing of statistical significance in connection with the relative quality ratings of the two brands.

TABLE 3

ORANGE JUICE

In the potato chip experiment, ANOVA revealed no significant effects of sampling or information on brand choice or on differences in quality ratings. The price variable vas, however, significant for quality ratings as indicated in Table 4. Once again, none of the concomitant variables were significant nor did they alter the results of ANOVA.

TABLE 4

POTATO CHIPS

DISCUSSION

Sources of beliefs had negligible effects in both experiments. Unfortunately, responses to the questions on sources of beliefs were highly skewed, with a preponderance of people reporting personal experience as the basis for their beliefs and only a few reporting advertising or reliance on other knowledgeable people as the sources of their beliefs. Thus, we were unable to deal with the research hypotheses having to do with sources and strength of belief and choice. In retrospect, it seems plausible to suggest that the products chosen for this experiment say have contributed to this outcome. If orange juice and potato chips were regarded as relatively low involvement - low risk products, that alone could account for the large number of experimental subjects who said that personal experience was the source of their beliefs.

Nevertheless, this research was partially successful in clarifying the results of the previous research by Obermiller and Wheatley. As anticipated, sampling is more powerful than "limited" advertising in affecting preference. In the orange juice study at least, the effects of sampling resulted in choice differences; information effects were non-significant. However, neither of these treatment effects were significant in the potato chips study. Price, on the other hand, was significant in affecting quality ratings for both potato chips and orange juice. In assessing quality, adult consumers apparently still use price as a cue even when they believe that quality differences among brands in a product category are not large. This is a significant finding because it strongly suggests that the quality of a large proportion of the products available in the marketplace may be judged on the basis of price information, i.e., the price-quality relationship may not be restricted to the smaller number of product categories in which large differences among brands are believed to exist. As long as product differences are not negligible, e.g. salt, popcorn or sugar, prices seem to matter; they create a price-quality perceptual bias. Additional and different types of products should be tested to confirm this conjecture. On the other hand, consumers may not feel strongly enough about this difference in quality to make a price constrained choice of the more expensive brand. This is consistent with the earlier observation by Wheatley et. al. (1980) "a price-quality perceptual bias my not necessarily affect choice behavior."

These research results may also be product specific. The pattern of results is somewhat different between the two products. However, aside from product form, there is no obvious difference between the products that would cause dissimilar results and there is no obvious reason why product form should affect the way beliefs influence quality assessments and brand choice. It may be that there is a difference in the decision process used to choose among brands in these product classes, but this observation is merely speculative.

One of the interesting, though understandable, outcomes of this project is the lack of impact of the information treatment. The manipulation checks indicated the information treatments were perceived accurately and retained. Few, if any, students of the communication process would suggest supplying information to change beliefs, attitudes, or behaviors is a hopeless task. It is more likely that the effects of information require wore than a single exposure to be reflected in the types of measurement used here Counter arguments and negative thoughts about the message may also have played a role. In addition, as mentioned before, in this experiment most of the participants beliefs were found to be based on personal experience Such beliefs are strongly held and difficult to change. They will limit what will be accepted and remembered about a message because a message recipient relies more heavily on his or her evaluative response to the content of a message than on the content of the message itself (Wright 1973).

The effects of sampling are equally intriguing Only in the orange juice experiment was an impact found on choice behavior; no such finding was observed in the potato chips experiment. Here we see a marketer behavior (sampling) affecting sensory experience (taste) in turn affecting choice behavior by demonstrating that the samples were indistinguishable. This is similar to an effect reported by Steinberg and Yalch (1976), who found sampling had an effect on the purchase behavior of obese people This suggests that sensations may beget behaviors in some circumstances This is a notion which ought to be explored to define the conditions under which it holds true.

While we have inferred that quality difference ratings were attributable to price differences, it is possible that they were the result of demand characteristics inherent in the experiment Most consumers like to feel that they can notice differences between brands of a product Advertisers certainly do their best to encourage such sentiments. The two samples presented to the respondents were identified as being different brands having different prices. Thus, the shoppers were encouraged to believe that differences existed. Fully half the sample, however, was led to believe that the samples were identical. The emergence of an inferred price effect, in such circumstances, tends to lend credence to the interpretations we have given, and attenuates the alternative explanation of a demand artifact.

The failure to find effects based on the sources of beliefs probably is due, in part, to the lack of control over the belief sources in the research. The skewed responses, problems with recall, and confounding of multiple sources could be overcome if source of belief were manipulable. Although it was not the focus of the study, manipulation of sources has been tried with undergraduate students in a study by Obermiller and Wheatley (1985). That approach could be adapted for use with different populations. The Rokeach typology of beliefs is relatively untested in consumer research in spite of the fact that beliefs add belief change have been studied for years. The typology may represent an opportunity to develop greater understanding of the problems of changing or maintaining beliefs with a research design adapted to test it.

Another interesting step would be to determine what proportion of consumer products are in each of the three product categories discussed in this paper, i.e., product groups in which consumers believe that there are large, intermediate or very small differences among brands. It would also be worthwhile to learn what fraction of consumers believe strongly in these characterizations. This work would have obvious implications for the development of pricing strategies on the part of sellers.

REFERENCES

Bem, D.3. (1970), Beliefs. Attitudes and Human Affairs, Belmont: Brooks/Cole.

Kotler, P. (1984), Marketing Management, Englewood Cliffs: Prentice Hall.

Lambert, Z.V. (1972), "Price and Choice Behavior," Journal of Marketing Research 9 (February), 35-40.

Leavitt, H.J. (1954), "A Note on Some Experimental Findings About the Meaning of Price," Journal of Business 27 (July), 201-10.

Neter, J. and W. Wasserman (1974), Applied Linear Statistical Models, Homewood, Ill.: Richard D. Irvin.

Obermiller, C. and J.J. Wheatley (1984), "Price Effects on Perceptions Under Varying Conditions of Experience, Information and Beliefs in Quality Differences," Advances in Consumer Research, Vol. 11, 453-58.

Obermiller, C. and J.J. Wheatley (1985), "Beliefs in Quality Differences and Brand Choice, n Advances in Consumer Research, Vol. 12, 75-78.

Rokeach, H. (1969), Beliefs. Attitudes and Values, San Francisco: Josses Bass.

Smith, R.E. and W.R. Swinyard (1983), "Attitude-Behavior Consistency: The Impact of Product Trial Versus Advertising," Journal of Marketing Research 20 (August), 257-267.

Steinberg, S. and R. Yalch (1978), "When Eating Begets Buying: The Effects of Food Samples on Obese and Nonobese Shoppers," Journal of Consumer Research 4 (March) 243-246.

Tull, D.S., R.A. Boring and H.H. Gonsior (1964), "A Note on the Relationship of Price and Imputed Quality, n Journal of Business 37 (April) 186-91.

Wheatley, J., R. Yalch and J. Chiu (1980), "In Search of the Economists Consumer: The Effects of Product Information, Money and Prices on Choice Behavior," Advances in Consumer Research, Vol. VII, 533-537.

Wright, P.L. (1973), "The Cognitive Processes Mediating Acceptance of Advertising, n Journal of Marketing Research 10 (February), 53-62.

APPENDIX A - THE INFORMATION TREATMENTS

A study by Consumer Reports states:

Most Americans start their day with a glass of some sort of juice. By far, the most popular juice consumed is orange juice. Orange juice basically contributes vitamin C and sugar to the diet. The sugar in orange juice - a combination of sucrose, fructose, and glucose contributes most of the 70 or so calories per glass. Orange juice is available in several forms - freshly squeezed, frozen, refrigerated, and canned. More recent variations are the imitation juices made from powder which is mixed with water. All forms are available in local supermarkets. (Orange Juice -Irrelevant)

A study by Consumer Reports states:

Orange juice basically contributes vitamin C and sugar to the diet. Different kinds of orange juice don't differ significantly in vitamin C or in calories. The big difference is in what processing does to the flavor. Frozen orange juice is the least processed of the supermarket juices. It's basically fresh juice with most of the water removed. If the juice isn't defrosted and refrozen, all frozen orange juices will taste pretty close to freshly squeezed juice after they are reconstituted. A scientifically conducted experiment with a large sample of frozen orange juice users found it was impossible to tell the brands apart in a blind taste test. (Orange Juice - relevant)

A study by Consumer Reports states

The potato is an important staple in the American diet. On average, every American eats 123 pounds of potatoes a year - 58 pounds fresh and 65 pounds processed. Processed potatoes include frozen french fries, instant dehydrated potatoes, and potato chips (Potato Chips Irrelevant)

A study by Consumer Report states:

The content of all commercially available potato chips in this country is the same, except for flavorings such as barbecue, hickory smoke, or cheese, that may be added. Differences in color, fat content, moisture content, preservatives and salt were found to be very small. Since potato chips are all made with similar raw materials and utilize almost identical manufacturing processes, our researchers were not surprised to discover that a scientifically conducted experiment with a large sample of potato chip users revealed that it was impossible for consumers to tell any of the various brands of unflavored chips apart in a blind taste test. (Potato Chips - Relevant)

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