Consumer Information By Objectives: an Approach For Defining Goals and Measuring Results

Klaus G. Grunert, University of Hohenheim
ABSTRACT - Starting from a general macro-goal of consumer policy, the theory of perceived risk is used to define micro-goals, to give directions on how useful consumer information material can be developed, and on how such usefulness can be measured. Results from a preliminary study are reported.
[ to cite ]:
Klaus G. Grunert (1980) ,"Consumer Information By Objectives: an Approach For Defining Goals and Measuring Results", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 664-668.

Advances in Consumer Research Volume 7, 1980     Pages 664-668

CONSUMER INFORMATION BY OBJECTIVES: AN APPROACH FOR DEFINING GOALS AND MEASURING RESULTS

Klaus G. Grunert, University of Hohenheim

[The paper draws on results from two research projects on consumer information, funded by the German Federal Departments of Commerce and Technology, respectively. These research projects are directed by Gerhard Scherhorn.]

ABSTRACT -

Starting from a general macro-goal of consumer policy, the theory of perceived risk is used to define micro-goals, to give directions on how useful consumer information material can be developed, and on how such usefulness can be measured. Results from a preliminary study are reported.

INTRODUCTION

Although a lot of progress has been made in consumer research in recent years, application of the insights gained has traditionally been confined mainly to the marketer side of the marketer-consumer dyad (Scherhorn 1978). While applications in the development of measures of consumer policy would seem to be an obvious supplement, this has only rarely occurred, firstly due to a lack of policy-oriented research, and secondly due to a lack of willingness, on the side of the policy makers, to use results from such research. While the latter problem seems to remain largely unsolved, there has been considerable progress in the former. Especially in the field of consumer information acquisition and processing, there have been studies with considerable policy implications. The work by Jacoby and his associates is just one example.

Consumer information programs as one of the main pillars of consumer policy have, however, not only suffered from the fact that their design was not based on behavioral science, but also from a lack of clearly and operationally defined goals to be attained by them, which in turn makes it impossible to develop means for measuring the effectiveness of such programs (Day 1976). It is the purpose of this paper to suggest a way in which a well known topic of consumer research, the theory of perceived risk, can be used to develop consumer information programs which have clearly defined goals and measurable results.

Policy Goals and Individual Goals in Consumer Information.

Explicit or implicit, the view most often advanced by practitioners and theoreticians alike is that the overall goal of consumer policy is to enhance the workability of a production/distribution system-usually the market system, but of course consumer policy can also be applied to public goods. "Enhance the workability" is held to mean that production and distribution should occur in a way which is in best possible accordance with the preferences of the system's members. Another stated goal is to further the interests of socially deprived people -e.g., the poor and the elderly. This will not be discussed here.

Consumer information programs, however, are directed at the individual consumer and not at the consumers as a whole. Thus, it becomes necessary to break down the macro-goal stated above into micro-goals which are hoped to be attained with the individual consumer and which are instrumental in reaching the macro-goal. Such a goal at the micro-level could be that the individual consumer should have the possibility to obtain the information he needs to select that product from the choice he has which best suits his personal needs. This is a view which seems to be widely held.

The problem with such goals is that consumers do not seem to be motivated to acquire all kinds of consumer reports, labeling information, etc. In many cases, the consumer does not even know which kinds of information could be useful in preparing a buy.

Numerous studies have shown the influence which socio-demographic variables, product characteristics, and product knowledge have on information acquisition behavior. The line of theoretical reasoning with the most explanatory power seems, however, to be the theory of perceived risk. Without going into any details, the main from of the theory suggests that the risk perceived by consumers before a purchase induces activities to reduce it, and that information acquisition might be one way to do this. However, risk reduction by means of brand loyalty, store loyalty, or other means more comfortable than information acquisition seems to be preferred by consumers (Roselius 1971). This suggests that the consumer's motivation to acquire information depends on:

- the strength of perceived risk

- the expected capacity of an information source to reduce risk

- the expected cost incurred in consulting that information source.

How does this relate to the policy goals stated above?

A consumer who has successfully reduced all risks he might possibly perceive in the purchase of some product, and hence has prevented any of the negative consequences associated with these risks to occur, obviously has attained the micro-goal stated above: he has bought the product which best suits his personal needs. Thus, in principle, we can expect consumers to be motivated at attain this goal. Goal attainment can, however, be impeded by a number of reasons:

- the consumer might not know about the risks which can be associated with the purchase of a particular product

- the consumer might not know which kinds of information are suitable to reduce these risks

- risk-reducing information may be available only at high cost, may be unavailable, unreliable, or false

Now it becomes possible to state in terms which are immediately accessible to empirical research what the functions of consumer information programs should be if the consumers are expected to use them, and if such programs are to attain the policy goal stated above. Consumer information should make sure that:

1. The consumer does not miss a risk before the purchase which afterwards might prove important for him

2. The consumer knows which kinds of information are suitable to reduce which kinds of risk

3. The consumer has at his disposal the information needed to reduce the kinds of risks he perceives

Obviously, developing consumer information programs in accordance with these subgoals requires some empirical investigations before actually setting up the information material. It has to be investigated which risks might be associated with the purchase of a particular product, how important they are to various consumers, and which kinds of information are. usually supplied by advertisers etc. For an example of how this can be done, cp. Grunert and Saile (1978). Our concern in the following will be how to measure the attainment of the three subgoals stated above for consumer information material.

MEASURING GOAL ATTAINMENT

Measures of consumer information influence the buying process - if they have any influence. Thus, their effectiveness can be measured by repeated measures of variables which characterize the buying process in a stage which lies further in the temporal sequence than the point at which the information exerted its influence. This means that in principle all effects could be measured by watching differences in the final purchase decisions. This, however, is not always the best solution, since by then the effect of the policy measure may have been confounded with other influences. We shall now discuss indicators of goal-attainment for the three sub-goals defined above.

Subgoal 1: Indicator Risk-consciousness

Since subgoal 1 deals with pointing out risks to consumers, risk-consciousness is the obvious indicator here. However, it would be foolish to consider some consumer information policy the more effective the more risk-consciousness it creates. The aim is to transform the consumer's actual risk-consciousness into a risk-consciousness which approximates as well as possible the one he would have in a situation of perfect information. Note that under perfect information a consumer might still experience very little risk - simply because many of what other consumers would call negative consequences of a buy are irrelevant to him, do not seem to be negative for him, cannot occur in his special situation etc. But it is desirable that a consumer does not suddenly experience after the purchase additional negative consequences which before the purchase he had not deemed possible.

Imagine that the purchase of some product is associated with a maximum of n possible items which can cause risk (or n risks, in short). Suppose further that the subjective importance of each such item for an individual consumer is measured on a fixed scale, yielding n scale values g. Then we can logically discern three vectors:

- a vector (g) of "ideal" scale values, i.e., the values the consumer would give under perfect information

- a vector (g^), giving the actually experienced relevance of the risks before the measure of consumer information is applied

- a vector (g~), giving the actually experienced relevance of the risks after the measure of consumer information is applied.

Using simple city-block-distances between g and g, we can compute a measure of distance between "ideal" risk-consciousness and actual risk-consciousness before the policy measure as

D1 = E|gi - g^|   (1)

              n

From this, we can compute the similarity-coefficient bounded by -1 and 1 (Schlosser 1976):

S1 = E (D1) - D1   (2)

             E (D1)

E(D1) is the expected value of D1, if there were no systematic relationship between the-two vectors. Specifying a scale on which to measure g and assuming a theoretical distribution over that scale, E(D1) can be obtained by computer simulation. Using a five-point scale, assuming that scale values are evenly distributed over the scale, and standardizing the scale values for a mean of zero and a variance of one, E(D1) turns out to be 1.13.

Similar to this, we can define D2 and S2 as measures of distance and similarity between g and g~.

From this, a rationale for an indicator of consumer information policy effectiveness in enhancing risk-consciousness can be developed as follows. If the measure were perfectly effective, we would have S2 + i. If it were completely ineffective, we would have S2 = S1. Thus, a measure of goal-attainment which is 0 for a completely ineffective and 1 for a completely effective program is

W = S2 - S1   (3)

         1 - S1

Subgoal 2: Indicator Knowledge

Subgoal 2 is concerned with telling consumers which kinds of information are suitable to reduce which kinds of risks - things like "if you want to make sure that your washing-machine spins dry, make sure it has a high r.p.m.". If anything, such statements increase the consumer's knowledge about the usage of certain informational items. This is therefore what we propose to measure. Here, too, there are individual differences in what a consumer ought to know, since a statement like the example above would hardly make sense for a consumer who never uses his washing machine for spinning. Still, this problem seems much less serious here, and measures of effectiveness can be constructed quite simply by administering items to respondents which have to be rated as true or false.

Subgoal 3: Indicator Purchasing Behavior

Subgoal 3 is concerned with the supply of factual information about products, information which can reduce risks. While such information can also enhance a consumer's knowledge, it is not very realistic to assume that a consumer memorizes all product information while buying some of the more complicates consumer durables. He might as well write them down, carry an issue of Consumer Reports with him etc. For this reason, knowledge is not a suitable indicator here. To assess the effectiveness of policy measures, we must use its impact on the purchase decision itself.

The informational items by which we describe the situation of the consumer before the purchase are n values gl' describing the subjective importance of the various risks, and j x n values rik, which are numerical representations of the risk-reducing information the effectiveness of which is to be tested, rik tells how well brand k does in avoiding risk i. Let us assume that all r are measured on the same scale (like ratings of a comparative-testing agency).

The basic problem is of course how to integrate these items to an overall judgement which in turn determines the purchase decision. While all kinds of integration models have been proposed (cp. Park 1976, 1978; Wright 1975), the one with the most explanatory power is still the simple linear model (Bettman 1971; Slovic and Lichtenstein 1973). Since we need an accurate predictor rather than a convincing psychological explanation, we use the linear model here.

So, under perfect information, the consumer will choose the brand k = a for which

f(k) = Ei girik -> Max!   (4)

Under imperfect information, of course, the consumer will probably choose a different brand b. The difference f(a)-f(b) can be used as a measure of the deficiency of the purchase decision due to lack of objective information - regardless of whether this lack of information is due to the fact that information is simply not available, that it is available but not acquired due to costs of acquisition, or that available information is deceptive hence not objective. If the information situation is improved by supplying risk-reducing information, the matrix of r-values changes, while the g-vector most probably remains constant. If therefore the consumer now, due to the additional information, chooses another brand c, we can measure the effectiveness of this information policy measure by

Z = f(c) - f(b)   (5)

      f(a) - f(b)

Z is bounded by 0 and 1.

A problem might arise here if the consumer's decision depends also on dimensions about which information is not given, e.g., psycho-social risks. While the consumer might learn from the additional information that brand c is in a functional way superior to brand b, he might still select b because b is superior in psycho-social aspects. In this case Z would show no effect. For how this problem can be remedied, see Grunert (1979).

Experimental Designs

In most cases, the indicators presented above can be measured only in a laboratory setting. Indicators for subgoals 2 and 2 call for measurement before and after treatment; the indicator for subgoal 1 calls even for three measurements to obtain values for the three vectors (g). The following section outlines a preliminary study in which the indicators tested above were used to test the effectiveness of information material designed to attain subgoals 1 and 2.

A PRELIMINARY STUDY

The purpose of the study was to investigate, on a preliminary basis, whether consumer information leaflets prepared according to the INVORMAT-method (Grunert and Saile 1978), designed to make consumers aware of possible risks and tell them which information can to used to reduce them, do have the desired effect, and whether differences in the presentation of the information exert an influence on the size of the effect. The latter question stems from the assertion that emotional "packaging" of information enhances information processing and hence neutral consumer information, too, should use emotional stimuli (Kroeber-Riel 1977).

Consumer information leaflets were prepared for three products: washing machines, bed linen, and car liability insurance. For each product, the leaflet was prepared in three versions, which were identical in content, but differed in their wording and graphical presentation. Version A was a plain sheet of paper presenting the information in a condensed form, version B had the text edited by a professional advertising man, version C had also the graphics done by a professional. Emotional "packaging" thus increased from version A to version C. Subjects were 116 students from Hohenheim University for the leaflet on car liability insurance, 80 females from the Stuttgart metropolitan area for the leaflet on bed linen, and 79 males and females from the Stuttgart metropolitan area for the leaflet on washing machines. The samples were convenience samples the selection of which was basically determined by the consideration that subjects should be expected to have some experience with the product informed about, thus ensuring a minimal level of motivation. The results for the three products are thus not strictly comparable. For the bed linen and washing machine samples, care was taken that various income and education groups were adequately represented.

The experimental design is depicted in figure 1. For each product, subjects were randomly assigned to one of the three treatments or to the control group. All subjects then were asked about risks associated with the buy of the product and rated the risks on a five-point scale according to their importance (0 = not important, 4-very important). Risks not mentioned were assigned the value 0. These values were taken for the vector (g). After the test-groups received and read the leaflet, the procedure was repeated to obtain (g). After that, each possible risk was presented to the subject on a card and was rated. This yielded (g). Finally, ten items were presented to the respondent which contained statements about which kinds of information are suitable to reduce which kinds of risks. These statements were a sample of the statements contained in the leaflet on this matter, though half of them were transformed to their contrary. The respondent had to rate them as either true or false. The number of correct responses was taken as a measure of attainment of subgoal 2. Attainment of subgoal 1 was measured by W and was computed from the three g-vectors.

Results

Subgoal 1.  Tables 1 to 3 show that the leaflets were indeed able to attain the desired effect as measured by the indicator W. Obviously, the means W for the control groups are much lower than for the test groups. Performing a t-test between the control group and the pooled test groups supports this observation. As for the differences in effect between treatments, the results are somewhat inconclusive. Although the size and nature of the sample do not allow generalizations, it seems that the relationship between amount of added emotional ingredients and effectiveness is not as simple as sometimes maintained. Also, the differential treatment effects seem to differ according to income and education; a consideration which cannot be born out here.

FIGURE 1

TABLE 1

MEANS W FOR TEST AND CONTROL GROUPS- LEAFLETS CAR LIABILITY INSURANCE

TABLE 2

MEANS W FOR TEST AND CONTROL GROUPS- LEAFLETS BED LINEN

TABLE 3

MEANSW FOR TEST AND CONTROL GROUPS-LEAFLETS WASHING MACHINE

An interesting observation, however, is the different size of the effects for the three products, which cannot be explained by the differences between the sample alone. Table 4 presents means, all subjects pooled, for the three products. Especially the difference between the washing machine and the other products is striking. The most plausible explanation for this is that while the number of possible risks associated with a product differs considerably, the absolute number of risks named by respondents in a free-recall situation seems to be more or less constant, as table 6 shows. Since the computation of W is, however, dependent on the size of the vector g, W can be expected to be lower for products with many risks. The consequences this has for the design of consumer information material have yet to be investigated.

TABLE 4

W FOR VARIOUS PRODUCTS - POOLED SUBJECTS

Subgoal 2.  Tables 5 and 6 show the means of Y for leaflets on bed linen and washing machines (no information concerning subgoal 2 was given in the leaflet on car liability insurance). While all of them are obviously higher than the expected value of 5 even respondents with no information could obtain, the means for control group and test groups do not differ very much, as the t-test with the pooled test-groups shows. A possible explanation for this is that subjects already had a considerable stock of knowledge before the treatment, which the leaflet could only moderately enhance. A before/after design, which was not used here in order to be able to combine the test of both subgoals, could give more clarity here. Also, a larger number of statements, thus giving Y a wider range, would hopefully result in more conclusive results.

TABLE 5

MEANS Y FOR TEST AND CONTROL GROUPS - LEAFLETS BED LINEN

TABLE 6

MEANS Y FOR TEST AND CONTROL GROUPS

CONCLUSION

An approach which allows a clear definition of goals and ways to measure their attainment is as important in consumer information policy as in any other area of public policy, if funds are to be allocated in an effective way. The approach outlined above has passed a first, though modest, test of application. However, much more research is needed to acquire a set of tools which is both reliable and valid. For example, different scales to measure the intensity of risk consciousness have to be tested, and an experiment using magnitude-scaling is underway. Also, follow-up interviews after some time has elapsed would yield data that captured aspects of information acquisition which have to be neglected in a single laboratory session. Self-administered questionnaires should be compared with personal interviewing to check for possible stage-effects. And, of course, the third indicator-purchasing behavior-still awaits its first test.

REFERENCES

Bettman, James R. (1971), "The Structure of Consumer Choice Processes," Journal of Marketing Research, 8, 465-71.

Day, George S. (1976), "Assessing the Effect of Information Disclosure Requirements," Journal of Marketing, 40(2), 42-52.

Grunert, Klaus G. (1979), Kriterien der Wirksamkeit von Verbraucherinformation: Zur Messung des Nutzens informationspolitischer Strategien. Working paper no. 4, University of Hohenheim (09200).

Grunert, Glaus G. and Saile, Heinz D. (1978), "Information Needs, Supply, and Demand as a Basis for the Development of Consumer Information Material: The INVORMAT Method," Journal of Consumer Policy, 2, 338-48.

Kroeber-Riel, Werner (1977), "Kritik and Neuformulierung der Verbraucherpolitik auf verhaltenswissen schaftlicher Grundlage", Die Betriebswirtschaft, 37, 89-103.

Park, C. Whan (1976), "The Effect of Individual and Situation-Related Factors on Consumer Selection of Judgmental Models," Journal of Marketing Research, 13, 144-51.

Park, C. Whan (1978), "A Few Questions on Consumer Information Processing: Two Product Cases, Coffee and Toothpaste," Journal of Marketing Research, 15, 243-49.

Roselius, Ted (1971), "Consumer Rankings of Risk Reduction Methods," Journal of Marketing, 35(1), 56-61.

Scherhorn, Gerhard (1977), "Ober die Bedeutung des Verbraucherverhaltens fnr die FunktionsfShigkeit des Marktes," Journal of Consumer Policy, 1, 20-31.

Schlosser, Otto (1976), Einfnhrung in die sozialwissenschaftliche Zusammenhangsanalyse, Reinbek: Rohwolt.

Slovic, Paul and Lichtenstein, Sarah (1973), "Comparison of Bayesian and Regression Approaches to the Study of Information Processing in Judgment," Human Judgment and Social Interaction, ed. L. Rapoport and D. A. Summers, New York: Wiley.

Wright, Peter (1975), "Consumer Choice Strategies: Simplifying Vs. Optimizing," Journal of Marketing Research, 12, 60-67.

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