Supplier Information in the German Automobile Market: an Analysis of Information Gaps

ABSTRACT - Information gaps in consumer markets are introduced as a problem of economic performance, and defined by quantitative and qualitative criteria. Results from a computerized quantitative content analysis of a sample of sales brochures and advertisements of German automobile manufacturers are presented, indicating that the information-supplied is not in all cases the information consumers want.

Citation:

Ingrid Gottschalk and Klaus G. Grunert (1983) ,"Supplier Information in the German Automobile Market: an Analysis of Information Gaps", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 427-433.

Advances in Consumer Research Volume 10, 1983      Pages 427-433

SUPPLIER INFORMATION IN THE GERMAN AUTOMOBILE MARKET: AN ANALYSIS OF INFORMATION GAPS

Ingrid Gottschalk, University of Hohenheim

Klaus G. Grunert, University of Hohenheim

ABSTRACT -

Information gaps in consumer markets are introduced as a problem of economic performance, and defined by quantitative and qualitative criteria. Results from a computerized quantitative content analysis of a sample of sales brochures and advertisements of German automobile manufacturers are presented, indicating that the information-supplied is not in all cases the information consumers want.

THE PROBLEM: INFORMATION GAPS AND ECONOMIC PERFORMANCE

Any economy based on a division of labor, and any market economy in particular, can function properly only if the flow of information between suppliers and consumers is continuous, contains the right information, and works in both directions. If the flow of information is inhibited, consumers cannot make the optimal choices among the goods offered, and hence incur a welfare loss (Dedler, Gottschalk and Grunert 1981), and producers cannot adapt their products to consumers' needs and wants. Whenever this is the case, we can say that there is an information gap in this market.

There is widespread agreement among consumer economists that information gaps abound in reality. However, there has been hardly any attempt to ascertain empirically whether information gaps exist, what kinds of information are lacking, and how they can be filled by measures of information policy.

This paper reports about an attempt to do just this: Measure empirically the information gap in a consumer market.

WHAT IS AN "INFORMATION GAP"?

Basically, an information gap can be said to exist when there is a discrepancy between information needs and information supply. We distinguish two kinds of such discrepancies: Quantitative and qualitative information gaps.

Quantitative Information Caps

Information about things consumers would like to know about particular products can simply be missing, i.e. it is not supplied by those communicators forming the consumers' information system (Thorelli and Thorelli 1977). More precisely, we say that a quantitative information gap exists if information about product attributes of interest to consumers is not available for at least n% of the brands in the market.

In application, this poses two problems: It has to be determined which attributes are of interest to consumers, and a specific n%-threshold has to be decided upon. The most obvious way to find out attributes of interest to consumers might be to ask them directly which kinds of information they would like to see. However, this approach fails for two reasons. Firstly, in many cases consumers feel well-informed, even if they are grossly underinformed by any conceivable objective standard. Secondly, consumers do not usually know all the hazards associated with a particular purchase.

The solution proposed here is to ask consumers not about their information wants, but rather about the risks they associate with a particular purchase. Consumers experience risk if they fear that due to a suboptimal buying decision some negative consequence of the purchase might occur, like malfunctioning of the product, physical damage, financial loss etc. (Bauer 1960). Of course, consumers try to reduce risk before the purchase, and one way to do this is to acquire information (Roselius 1971). Thus, if we know which kinds of risks are associated with a particular product by the consumers collectively, we can define consumer information needs as the ensemble of product attribute information that would be suitable to reduce these risks (Grunert 1981). If we find that some of this information is not supplied, a quantitative information gap has been found. Such an application of the well-known concept of perceived risk entails a more differentiated concept of risk than is usually used in the literature: it is not sufficient to assess inasmuch consumers perceive risks of a financial, functional, or psycho-social kind, but rather we need to know about which attributes of the product the consumer is ignorant and fears negative consequences (Dedler, Gottschalk and Grunert 1981). Thus, we develop a multi-attribute view of perceived risk.

Risks associated by consumers with the purchase of a new car were ascertained by interviewing a sample of 60 drivers using an open-ended questionnaire. The aim was to obtain a list of risks as comprehensive as possible. While representativeness of the sample was not deemed necessary, care was taken to assure that all types of users were included. For this reason, interviews were conducted in both rural and urban areas. and all age groups were included.

The list of potential risks compiled from the questionnaire was then shown to automobile experts. The task of the experts was to make judgments about the occurrence of these risks, i.e. whether they occur at all, and whether they are suitable to discriminate between brands. Also, the wording of the risks was checked, and, where necessary, corrected. However, it was attempted to adhere to respondents' original wordings as completely as possible, since the aim was to measure consumer information needs, not expert information needs. The list of risks emerging from this consultation process contained 82 items.

This list can be used to check whether the information supply contains informations suitable to reduce the various risks. However, it seems plausible that not all of the 82 risks are or equal importance to Consumers, i.e. subjective disutility experienced when one of the risks materializes can be expected to differ over the 82 items . Since a lack of information found concerning a very important item seems more serious than a lack of information concerning a less important item, a survey to ascertain relative importance of the 82 items was conducted. A multistage, stratified random sample was drawn to obtain 500 respondents representative for the south-west German state Baden-Wurttemberg. All respondents had purchased a new car during the past 4 years. In order to obtain data that allow for maximal differentiation of the weights given to the various risks, the method of magnitude scaling was used (Lodge 1981; Wegener 1973). This method, adapted from psychophysics, does not employ prespecified scales, but specifies only a modality in which the respondent can express the intensity of the response to a stimulus. Numbers and lines are the modalities used most often: Respondents first assign an arbitrary number or a line to a standard stimulus, and then scale all .other stimuli by specifying numbers of different magnitude or lines of different length, relative to the standard stimulus. Line drawing was the modality used in this study.

Because of the unusually high number of items, the total set of 82 risks had to be split into two parts, each of which was administered to one half of the sample. Four items were included in both questionnaires, including the standard stimulus, ensuring that the scale weights obtained can be assembled to a coherent list of 82 values later . For each respondent, the raw value for each item was obtained by measuring line length in millimeters. The values were then standardized, assigning an arbitrary value of 10 to the standard stimulus. Overall scales values for each item were obtained by computing geometric means over all respondents. The 82 items, the geometric means of their standardized scale values, and the resulting ranks are displayed in table 1. Risks concerning physical dangers, safety aspects, durability, and financial losses are ranked as very important; performance characteristics of the car like top speed etc. are ranked relatively low.

TABLE 1

GEOMETRIC MEANS OF STANDARDIZED SCALE VALUE AND RANKS FOR 82 RISKS ASSOCIATED WITH BUYING AN AUTOMOBILE

This list can be used to delineate quantitative information gaps: A quantitative information gap is said to exist if for one of the risks listed there is not enough information supplied suitable to reduce-it. If such an information gap is found, the scale value of the risk can be used to characterize its seriousness. The "not enough" criterion is defined as referring to the number of brands in a market about which there is no information. The rationale for this is that consumers use product information to compare brands; however, if there is a lot of information about some product attribute for one brand, but none for all others, this is not particularly helpful for him. Thus, the percentage of brands in a market about which information about a certain purchase risk is available can serve as an indicator whether there is "enough" information. Choosing a certain threshold value is of course always arbitrary. We decided to speak about a quantitative information deficit if information suitable to reduce a certain risk is available for less than 25% of the brands in a market. Obviously, this is not a very strict criterion.

Qualitative Information Gaps

Given that information about a certain purchase risk is found to be available for a sizeable number of brands. its usefulness can still be questionable.

The quality of the information can be wanting. For these cases, where information is available quantitatively, but the quality is considered unsatisfactory, we use the term "qualitative information gap".

If information gaps are analyzed from the viewpoint of the consumer, as is done here, it follows that the quality of the information also has to be determined from the consumers' viewpoint. We want to introduce three criteria for the quality of information suitable to reduce risks perceived by consumers.

First, information can be verifiable or non- verifiable. A sentence of the type "This car consumes 8.5 liters/ 100 km- contains verifiable information, while a sentence like "this car has very low fuel consumption does not. "Verifiable" is understood in a very general sense, i.e. the information has to be verifiable by someone, not necessarily by the consumer himself. For example, the information that the car body has been galvanized twice to prevent rust cannot be verified by the average consumer, but can be verified by product experts. We assume that verifiable information is more helpful to the consumer than non-verifiable information, and hence is a better quality of information.

Secondly, information can refer to the risk itself or can be related to a risk by means of an indicator. The sentence just mentioned, "this car consumes 8.5 liters/100 km", directly refers to the risk that the fuel consumption of the car is too high. A sentence "this car has a low air resistance value of 0.4 also says something about fuel consumption, since a low air resistance value means low fuel consumption. Thus, it is related to a risk, but it does not refer to a risk directly, since none of our respondents expressed a fear that the car's air resistance value may be too low. We assume that information dealing with a risk directly is more helpful for the consumer. Information about an indicator for a risk not only requires more cognitive processing of the consumer, because he has to relate the indicator to the risk. It may also be completely useless to him, because he may not know that the information given can be used as an indicator for one of the risks perceived by him.

It has to be mentioned, however, that for several of the risks listed in table 1, information has to be about indicators in order to be verifiable. Take the example proneness to rust". A sentence like 'our car won't rust' refers to the risk itself, but is non-verifiable. But in order to be verifiable, indicators, like rust-prevention measures taken, have to be used.

Thirdly, the information concerning a certain risk given by various communicators can be more or less comparable. We say that information about a risk is comparable across brands if all manufacturers use the same verifiable information or the same indicator. If all manufacturers give information about fuel consumption in liters per 100 km, this is comparable information. If each manufacturer uses a different vocabulary to describe the rust prevention measures he has taken, the information is very difficult to compare for the consumer. Of course, different degrees of comparableness are conceivable between these two extremes.

These three quality dimensions can be used to delineate qualitative information gaps. Gaps of various seriousness can be distinguished according to which of the three quality criteria employed are not met by the information supply.

ASSESSING INFORMATION GAPS BY CONTENT ANALYSIS: AN EXAMPLE

The Object of Investigation: The German Automobile Market

The German market for automobiles was used as an example to ascertain information gaps empirically. There are two main reasons for this choice: First, automobiles are a very complex product involving all kinds of physical, financial and performance risks, thus making the task of information search for the consumer especially difficult. Secondly, the information system in this market is quite elaborate. making it a very interesting object of study.

Neglecting personal communication, the main communicators in the market are manufacturers, retailers, and neutral sources. Manufacturers use sales brochures and advertisements as their main media; retailers use advertisements and personal selling. Neutral sources comprise mainly tests in automotive magazines. All sources and media have been analyzed (Dedler, Gottschalk, Grunert, Heiderich, Hoffmann and Scherhorn 1983), but only the part of the study dealing with sales brochures and advertisements will be presented here. These are the two most important elements of car suppliers' communication policy.

In order to be manageable, the market and media to be analyzed had to be defined more precisely. The following criteria were used:

- The analysis referred to the information supply in 1978;

- only information from manufacturers having a market share of more than 0.9: was included, and only about models not exceeding or falling short of certain thresholds for price (8000 - 25000 DM) and cylinder capacity (850 - 3000 ccm). This served to delimit the market;

- within these limits, all sales brochures available from manufacturers (totalling 38 documents) were included in the analysis;

- as for advertisements, 5 magazines and 4 newspapers were selected by systematic criteria. From these, issues were selected randomly, resulting in a total of 395 advertisements. Discounting duplicates, a sample size of 261 was obtained.

The sample of documents thus obtained was to be submitted to a content analysis to find information gaps.

The Method: Computerized Quantitative Content Analysis

In applying content analysis, an important restriction has to be kept in mind. Starting with the classic paradigm of communication research 'who says what to whom, how and with what effect?', content analysis itself can give answers only to the questions "what" and "how". Answers to all other questions have to be inferred and are by necessity hypothetical. Thus, by means of a content analysis we cannot measure whether information can actually reduce risks. We can only apply certain criteria which we think delimit the types of information that are suitable to reduce risks. Still, we thought that a content analysis is a more useful instrument here than an experimental analysis of information effect, for two reasons: First, experimental analysis can measure only subjective risk reduction. This is unsatisfactory from a policy viewpoint, because we want to know whether the information supplied is actually suitable to indicate whether a certain risk will occur, not whether the consumer has been made believe that it will not occur. Secondly, a content analysis is more apt to give insights as to where and how the information supply can be improved.

A content analysis involves splitting the documents into units of analysis and sorting these into categories. In this case, the sentence was used as the unit of analysis. The category system reflects the dimensions referred to above: the risk, if any, alluded to in the sentence, and the manner the risk is dealt with, i.e. verifiable/non-verifiable and direct/indirect. Comparableness was analyzed separately.

A pretest with manual coding for some selected ads and risks showed severe reliability problems, so that we decided to attempt a computer-assisted_ content analysis using the program TEXTPACK (Hohe, Klingemann, Radermacher and Schickle 1980). Computer-assisted content analysis requires that

- all documents be transcripted to machine-readable format:

- all categories are defined in an extensive way, i.e. by a comprehensive list of words or word-combinations which, when they appear in a sentence, constitute evidence that this sentence belongs to a certain category.

Construction of the category system thus becomes a complex interactive process in which word functions are defined, tested with the data, and redefined until the validity of the category system seems satisfactory. The possibility of recoding the data as often as seems necessary to construct good computer-assisted content analysis, along with the fact that problems of reliability are completely eliminated. However, the large amounts of computer time needed to process data sets of the size obtained here sets a definitive limit to the number of possible recoding runs.

Photos and drawings presented a special problem. In order to be amenable to the content analysis, they had to be translated to certain key words. Three categories of photos and drawings were distinguished and assigned to corresponding key words:

- Photos/drawings illustrating some assertion made in the text, which may be suitable to reduce a risk;

- photos/drawings suitable to reduce a risk by themselves, without reference to the text;

- photos/drawings not suitable to reduce any risks, like photos where persons or landscape dominate.

Results Concerning the Quantitative Information Gap

The documents analyzed comprise a total of 23,745 sentences. An average of 530 sentences are used in a sales brochure, 13.8 sentences in an advertisement. On the average, 58.4% of all sentences in sales brochures were coded as dealing somehow with any of the 82 risks listed in table 1, compared to 73.9% in advertisements. At first sight, especially the figure for advertisements appears quite high. However, treatment differs considerably among the 82 risks. Of course, these differences could reflect the differences in importance assigned to the various risks by consumers. However, table 2 shows that this is not the-case. Table 2 gives correlation coefficients between the weights assigned to the risks as documented in table 1 and the number of sentences dealing with these risks in one of the ways distinguished here, as well as the total number of sentences. Most correlations are strikingly low or even negative: There seems to be no positive relationship between the subjective importance assigned to certain product attributes by consumers, and the space assigned to describing them in sales brochures and advertisements. Only for the number of sentences dealing with indirect information, i.e. with indicators relevant to certain risks, there is a slight positive relationship.

TABLE 2

CORRELATION BETWEEN CONSUMER-SUPPLIED SCALE VALUES FOR RISKS AND COVERAGE OF RISKS IN SUPPLIER MEDIA

Table 3 shows the results of applying the criterion for quantitative information gaps to sales brochures: It lists all those risks for which there was no information found in at least 75% of the sales brochures. The mean number of sentences referring to it in those sales brochures that did contain some information is also given. Twenty-three risks are listed in this table, i.e. more than one fourth of the total list of 82. This would be less serious if these were the risks ranked as least important by consumers. But the ranks from table 1, repeated in table 3, show that this is not the case. Thus, the quantitative information gap found can be considered as serious.

TABLE 3

RISKS NOT COVERED IN AT LEAST 75% OF THE SALES BROCHURES

A similar analysis for advertisements is not shown, since it was found that information there concentrated on very few risks, which is not surprising given the limited information capacity of advertisements. Attributes addressed most often in advertisements were purchase price, engine power, fuel consumption. and outward appearanCe.

Results Concerning the Qualitative Information Gap

Information quality was characterized in three dimensions: Verifiability, directness, and comparability. Table 4 shows results concerning the first two dimensions. Those risks not forming part of the quantitative information gap are grouped according to whether information concerning them is predominantly verifiable and/or predominantly direct. The risks listed in the lower left cell indicate where information is least helpful for the consumer. However, information about risks grouped in the lower right or upper left cells can be improved as well. This is especially important for the lower right cell, because the four items scaled as most important by consumers can be found there. Information about these risks is available and mostly verifiable, but it is mostly about indicators and thus limited in its usefulness to those consumers who have enough technical expertise to be able to link the indicator to the risk in question. We know from another study (Gottschalk and Schneider 1982) that few consumers have that expertise.

TABLE 4

QUALITATIVE ANALYSIS OF INFORMATION IN SALES BROCHURES

This problem is emphasized if comparableness is low, because in this case the consumer has to know all the different indicators used by the various manufacturers. We have analyzed comparability for five selected risks (brake mechanism needs repairs frequently, proneness to rust, susceptibility to aquaplaning, high fuel consumption, steering characteristics in curves are not neutral). For none of these five items there was at least one indicator or other verifiable information used by all manufacturers in the sample. For two items, at least 75% of the manufacturers used at least one common information, for the other three items between 50% and 75% of the manufacturers used a common information. Thus, for none of the five risks used as examples the information is strictly uncomparable, but comparability could certainly be improved.

DISCUSSION AND IMPLICATIONS

Substantive Implications: Supplier Information and Public Policy

The result that 23 of 82 risks are characterized by a quantitative information gap, and that for many of the others the quality of the information could he improved, does not imply, of course, that policy measures should be called for to provide additional information about all these risks. Information is costly, both for the supplier and the user. But the analysis also gives hints about where additional information might be most urgent, and how it might be supplied most favorably.

Two types of information can guide such decisions: The criterion by which a risk was classified as part of the information gap, and the rank assigned to it by consumers. Thus, the most obvious candidates for measures to supply additional information are those risks which have high ranks and which failed because of the quantitative criterion. 'Shivering steering and 'expensive spare parts" are examples for this. The most obvious candidates for improvements of existing information are those in the upper left or lower right cells of table 4 having a high ran's, like 'proneness to rust' and "aquaplaning". While additional information could be supplied by neutral sources, improvements of existing information can come about by legal requirements on supplier information. For example, there seems to be quite a considerable amount of information concerning rust in sales brochures, but it is mainly about indicators. Revising some standardized way or relating this information to the risk itseLf could bridge this part of the information gap. On the other hand, for risks where there is absolutely no information available, supplying neutral consumer information is a viable alternative to improving supplier information.

Methodological Implications: Computerized Content Analysis in Consumer Research

Some insights about the feasibility of computerized content analysis in consumer research emerged in the course of the study that might be useful for potential users of this method.

First, transcribing all documents to machine readable format is quite a task, and is bound to prevent large scale use of this instrument in the immediate future. However, with electronic typesetting equipment becoming widely used, it may become easier to sample documents in a format that is machine readable from the start.

Secondly, even though 38 sales brochures and 261 advertisements does not seem like a very large sample, a data base 23,745 sentences is certainly quite large for the average university computer, resulting in very high computing costs. However, a manual content analysis would have resulted in the same number of cases, so that this seems to be a general problem of content analysis. Of course, a larger unit of analysis than sentences could be used. But this would not only lead to less detailed results, but also to a decrease in reliability.

Thirdly, the task of building word functions to categorize sentences according to 89 purchase risks proved to be extremely tedious and time consuming. Also, some of the word functions became so complex that it was at times difficult to trace their logic, so that the high reliability achieved by the computerized procedure may have been gained at the expense of high validity in some cases. A simple and straightforward research question would be the best strategy to circumvent this problem. Unfortunately, information gaps are not a simple Problem.

REFERENCES

Bauer, Raymond A. (1960), "Consumer Behavior as Risk Taking", in Dynamic Marketing for a Changing World, R. Hancock (ed.), Chicago: American Marketing Association, 389-398.

Dedler, Konrad, Gottschalk, Ingrid, and Grunert, Klaus G. (1981), "Perceived Risk as a Hint for Better Consumer Information and Better Products: Some New Application of an Old Concept", in Advances in Consumer Research, Vol. 8, Rent B. Monroe ted.), Chicago: Association for Consumer Research, 391-397

Dedler, Konrad, Gottschalk, Ingrid, Grunert, Klaus G., Heiderich, Margot, Hoffmann, Annemarie L., and Scherhorn, Gerhard (1983), Das Informationsdefizit der Verbraucher, Frankfurt: Campus.

Gottschalk, Ingrid, and Schneider, Iris (1982), "Die Verstandlichkeit von Anbieterinformationen-', Working Paper No. 20, Institut fur Haushalts und Konsumokonomik, University of Hohenheim.

Grunert, Klaus G. (1981), 'Consumer Information Programs and the Concept of Perceived Risk", in Advances in Economic Psychology, W. Molt, H.A. Hartmann, and P. Stringer (eds.), Heidelberg: Heyn, 161-174.

Hohe, Jurgen, Klingemann, Hans Dieter, Radermacher, Klaus, and Schickle, Cornelia (1980), TEXTPACK Version V Release 2, Mannheim: ZUMA.

Lodge, Milton (1981), Magnitude Scaling, Beverly Hills: Sage.

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

Thorelli, Hans B., and Thorelli, Sarah V. (1977), Consumer Information Systems and Consumer Policy, Cambridge: Ballinger.

Wegener, Bernd (1978), "Einstellungsmessuna in Umfragen: Kategoriale vs. Magnitude-Skalen", umanachrichten, 3, 3-27.

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