Research, Experimentation, and Development of a Local Consumer Information System

James N. Morgan, University of Michigan
Greg J. Duncan, University of Michigan
ABSTRACT - The development and testing of a local consumer information service allows research not only on the main question of whether such a service can be supported by charging its users, or deserves public subsidy, or both, but also on the optimal amounts of different kinds of information to collect, on the marketing of information, and on consumer and market response to improved information.
[ to cite ]:
James N. Morgan and Greg J. Duncan (1980) ,"Research, Experimentation, and Development of a Local Consumer Information System", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 241-245.

Advances in Consumer Research Volume 7, 1980     Pages 241-245


James N. Morgan, University of Michigan

Greg J. Duncan, University of Michigan

E. Scott Maynes, Cornell University


The development and testing of a local consumer information service allows research not only on the main question of whether such a service can be supported by charging its users, or deserves public subsidy, or both, but also on the optimal amounts of different kinds of information to collect, on the marketing of information, and on consumer and market response to improved information.

It is embarrassing to be still talking about a research design that has never been funded, but perhaps all this time to reconsider will improve our research-and-action project if and when it is funded. We have been proposing a project combining action and research which is unique in the sense that:

1) The action part promises substantial benefits, particularly through the design of an entire information system exportable to other communities.

2) There are four different kinds of research imbedded into the project.

3) The pieces of the project fit together into an efficient, integral whole which make it inadvisable to chop up, but this intricate design combined with the infant-industry aspects of the project (start-up costs), makes it difficult to get underway.

In this paper, we discuss some research related elements of the design of a local consumer information system. Thought must be given to several crucial parts of such a system. We hope that by explaining what we feel will be the potential problems and our proposed solutions we will stand a better chance of success in this project once it is in progress.


The basic proposal is to develop, test, document, and evaluate a system for the collection, processing, distribution, and financing of local market information. The consumer user of such a system could obtain point-of-purchase answers to the following kinds of questions:

1) Where should I take my disabled TV set (or car, or hi-fi, or other consumer durable) to have it repaired cheaply and effectively. Call this type of information experience rating of vendor services.

2) Where is the cheapest place in my shopping area to buy "regular" gasoline (or film, prescription drugs or other products with relatively little quality variation)? Call this local price information.

3) What are the top-rated models of (say) single-lens reflex cameras? Where can they be purchased in my shopping area and at what price? Call the latter information on local accessibility.

4) What have been consumers' reactions to their purchase experiences with local retailers (vendors)? Call this vendor purchase ratings.

The proposed research would place greatest emphasis on experience ratings of vendor services, but it would also test the feasibility of collecting and distributing the other three types of information cited. While the consumer product testing organizations provide reproducible quality ratings of nationally marketed tangible goods and some services, there exists no information regarding the quality of local consumer repair services. This deficiency would be made good if our proposed research is successful.

The information to be collected on repair services--satisfaction ratings from continuing surveys with a representative sample of customers--would formalize, quantify, and systematize the kind of information which now passes informally among consumers, namely, how satisfied they have been with the service rendered to them by particular vendors--automobile repair services, TV servicemen, etc.--relative to the price of the repair.

By relying on simple ratings of satisfaction, probability sample surveys, and statistical tests, we would reduce random "noise" and avoid treacherous decisions which might bedevil a more sophisticated assessment of service quality. It is an assumption of this research that a sufficient sharing of standards exists so that the reported satisfaction ratings of an entire population of consumers will prove useful to other consumers. It might be added that persons exposed to experience ratings on automobile repairs, TV repairs, and hi-fi repairs in a pretest survey found them meaningful and stood ready to use them! The same pretest found statistically significant differences in reports of satisfaction among the vendors even with a small sample. In addition to an overall experience rating of a service, information will be collected regarding sources of satisfaction or dissatisfaction: promptness, the quality of work performed, and reasonableness of price.

To avoid unfair injury to either local vendors or local consumers, various checks and balances will be built into the system design. These include such devices as reporting only statistically significant differences, reporting changes in ratings over time, allowing affected vendors to feed their comments and explanations to system users, and adjusting ratings for the average level of satisfaction of each vendor's customers on all the other things they rate--to allow for differences in the types of customers (the curmudgeon factor).

To maintain an up-to-date bank of local consumer information, it will be necessary to contact large samples of consumers recurrently. This need becomes a virtue since it facilitates feedback from users of the data regarding the usefulness of the data collected, the appropriateness of alternative information delivery systems, felt needs for new types of data, and possible changes in the ratings of particular vendors. In addition, it will become possible to determine the extent to which vendor ratings vary among different kinds of customers (old vs. young, poor vs. rich, mobile vs. immobile, etc.).

Methods of disseminating the information would also be assessed. We propose to try a variety of methods ranging from cable TV to mail. Direct personal telephone contacts are too costly as a permanent dissemination mode but will be used initially to improve the more mechanical and impersonal processes. Since consumer acceptance and use of the information is crucial, methods of communicating with them are am essential link in the whole process. Some mix of promotion and actual information will be incorporated into the information getting process.


Before describing the four research aspects, let us look at the practical benefits that could come out of the project. If local markets are not performing their resource allocation function efficiently, and one cause of the problem is insufficient information among potential users of locally provided goods and services, the potential gains from the system are those generally attributed to the gains of perfect competition. [Consumer and producer surpluses, to use the old coinage, are enlarged by expanding the business of the "low cost" producers (meaning here high quality relative to cost) and providing services to consumers at lower prices (meaning here higher quality relative to price).]

This does not mean that there have to be excessive profits somewhere in order for some benefit to be possible. But it also does not necessarily mean that any vendor is driven out of business. It is entirely possible that a substantial unmet need for repair services is inhibited by lack of information and fear of high-cost-low quality work. The main result of better information may easily be a great deal more total business.

In addition, the same information that consumers would be getting goes to the managers of organizations providing services and goods, who may lack information about how their own mechanics are treating customers, or even which of their mechanics are doing the best work. Improvements in quality may easily result. There may be fewer new appliances purchased, and more existing ones repaired, with a saving of resources and a substitution of (repair) labor for resources (steel and energy to produce new things).

There are two possible outcomes of the experiment in terms of the functioning of local markets. If the information is widely used and adjustments of vendors rapid (e.g., bad vendors improve or go out of business, good vendors are overloaded with customers) the correlation between price and quality may so improve that consumers as individuals hardly need any information about the trivial remaining differences in quality/price. In this case, it is clearly appropriate that the system be publicly subsidized in order to continue and maintain its efficiency and to provide like systems in every community. Alternatively, the use of the information may be incomplete, the market sluggish, and the changes sufficient so that there is always some value to an individual in having and using up-to-date information. In that case, user-funded local consumer information systems would be economically viable and the system exportable to other communities with little or no subsidy.

Differences between various services in the response to better informed consumers, and differences in the adaptation of consumers to new information might well lead to a conclusion that some kinds of services deserve subsidized information, while others combine such slow adaptation with such rapid changes that it would always be worthwhile for consumers to pay for information. So as an experiment in the need for local market information and possible ways of providing it, the design is neat in that there are alternative outcomes with real social significance.

Why, if information is potentially so useful, is it nowhere provided? Largely, we argue, because there is an infant-industry problem. An information system must become visible, well-known and trusted, and used by enough people to be viable if not to produce total rationalization of the local market. Such a system requires a broad range of continuously updated information, computerized editing, adjusting, and summarizing, and efficient methods of dissemination. It also requires protection against manipulation and against undependable findings.

But once a system is developed and tested, the procedures can be easily reproduced in other communities. Some of the alternative dissemination methods such as cable television may not be available everywhere, but the main output format is likely to be selective computer-printout of subsets of information on demand, perhaps by vending machines, perhaps by mail.

How can any such system overcome the basic economic fact that information is a non-marketable "social good" that can be given away without losing it--failing the principle of exclusivity? Basically ours does it by selling tailored, up-to-date information only optimal for those seeking a particular service just now, and individuals ordinarily do not know who else seeks the same information. Even if the information were flashed on cable television channels, the chance that many other individuals who happened to be watching that channel just then wanted just that information, is small. A little free-riding would not be disastrous in any case because the information has social benefits even to non-users if it improves the local markets correlations between quality and price.

Even such fixed publications as Consumer Reports manage to secure some revenue by selling convenience, a sense of membership, and ease of retrieval of information that does not date too rapidly, such as quality differences of appliances that are only periodically redesigned. But the kind of local information we talk about has a much greater rate of obsolescence, it is as though the information were in disappearing ink. The highest rated repair services may become overloaded or hire other mechanics and the lowest-rated ones may clean up their act.

It is in that intermediate period when information is still valid, but relatively recent (so it isn't available in libraries of older reports) that a consumer information system can find an economically viable existence.


Our proposal involves more than social improvement; it involves and indeed requires, research on four topics:

1) optimal methods of data collection and summarization, combining information theory and sampling statistics

2) optimal methods of dissemination and marketing the information, combining computer and TV engineering and business administration

3) research on human behavior and behavioral change, combining social psychology and sociology

4) research on the functioning of local markets and the effects of improved information and competition on that functioning, combining microeconomic production theory and consumption theory

We take up these four research areas in order.


Starting with some measure of the value of various kinds of information, we can ask what is the most efficient way of getting information--the most information per dollar spent. Optimization has to be a sequential adaptive process, since the reliability of information depends not just on the number of independent bits of information on any one vendor, but the variability of that information across reports and over time. Satisfaction ratings may vary because survey respondents differ (the curmudgeon factor to which we return) or because quality varies even for one vendor depending on the type of service, the make and model of what is repaired, or which individual mechanic did the work.

There is also a possibility that any one satisfaction rating may be more reliable if it is done by experiment rather than sampling experiences, as when a car or appliance with a known defect is taken to many repairers. Our proposal is based on the assumption that there is such a high variability in those seemingly precise measures of quality that they are next to worthless in drawing inferences about the vendors. They constitute a one-degree-of-freedom estimate of quality that may unfairly penalize those repairs shops with all but one good mechanic or a good mechanic who is "tested" on a bad day. We argue that the much less precise reports of samples of those who have had repairs done, with all their potential unreliability, are safer when averaged over many customers and corrected for potential biases.

Our continuing surveys will provide a time series of data on the average satisfaction with each vendor at different points in time. The most useful information for consumers is whether the recent reports of satisfaction with a particular vendor are significantly higher than average or, alternatively, which vendor or group of vendors are significantly better than the rest. But with the information or ratings at different points in time, we can also test for significant trends in satisfaction ratings (e.g., is the slope of a regression line of satisfaction on time significantly positive or negative?) System users may find the trend information as important as the average ratings.

There remain interesting problems of optimal sample size. Some experiences are sufficiently frequent and unimportant that only a fraction of the sample need be asked to report them, or the time period covered could be reduced asking only about such experiences in the last month rather than half-year or year, or longer period. And the desired number of reports on any one type of service (and the emphasis on recency) would also depend upon how much they tended to vary over time. The extent to which there were substantial differences among vendors would also help determine how much information was required to establish those differences. It would be wasteful to have "ton much" information in one area at the expense of not being able to be sure of differences in others.

Clearly a dynamic optimization of the information system could maximize the value of the information relative to its cost. We would have to start with some Bayesian guesses and improve the system as it went. Fortunately we have some notions on how frequently people purchase things, and how much is at stake in various services.

One particular bias in need of adjustment is differences in affect, tone, or personality of individuals reporting their experience. A vendor would be unfairly penalized if he tended to attract hypercritical customers. Since each individual reports or more than one experience, a statistical comparison of their reports with averages of others reporting on the same vendors, allows estimation of a "curmudgeon factor" by which each individual's ratings can be adjusted.


We do not want to go from the present situation in which we have relatively good systems for disseminating relatively inadequate consumer information, to the reverse. A local market is limited, so how can we entice enough customers not only to justify the system, but to assure that it can affect the market so we can study its effects? Again, systems have to be developed that are adaptive. Experiments with rigid alternatives in time-honored design-of-experiment sense can only prove what will not work (namely any system that does not pay attention to problems and change to solve them). We would propose trying a number of alternatives at the same time, with methods of adapting them. And we would try a number of ways of getting attention, including free widespread dissemination of some widely usable findings along with the announcement of the availability of the information service.

Many years ago we discovered that vending machines have relatively low break-even points because they require little labor. Since then the engineering of remote consoles to central computers has advanced substantially. So one method of making the data available would be a vending machine printing directly from the computer a summary of a selected subset of up-to-date information. The information would include only statistically reliable differences, and an indication of the speed with which ratings have been changing over time.

Another possibility would be by cooperation with cable television, flashing the information on the screen and adding the small service charge to the next cable-TV bill.

A more versatile and flexible method would be telephone-order and mail delivery of printed reports. Monthly or annual subscription fees with extra charges only for excess use might be tried.

And to ease collection of fees-for-service, telephone orders and use of some cooperative drive-in pick-up windows, perhaps even banks, would serve a population that still drives a lot. One could order by 4 PM and pick up the information on the way home from work.


The same probability sample telephone interview that collected the information on vendors could collect information on the demographic characteristics and behavior of the respondents, including their use of and response to the information service itself. Indeed, the widespread telephone interviewing itself becomes a way of informing people of the existence of the service, in a realistic way since they know on just what questions the summarized are based.

There may well be wide difference in the effects of information on behavior, not only as between individuals with different levels of income and education, but also among the different kinds of services. People may have more loyalty to certain vendors, such as an "authorized dealer." Those with very low income may make so many emergency decisions as to make it difficult to use information, and those with very high incomes may not be concerned with saving money. Recent immigrants to an area may have less access to background information and local informants, and hence more need of organized sources. (Note that our information is essentially a summarization of the kind of information on gets from asking friends and acquaintances, "Where is a good place to __________?" without the possible bias in sample or instability from small numbers of reports.)

We might well imbed some reinterviews into the system to compare actual changes in behavior with estimates based on retrospective questions about changes in behavior.


By counting the number of survey respondents who consume the services of each vendor, we can calculate an unbiased estimate of the market share of each vendor over time. Furthermore, by asking whether respondents use the system's information we obtain an unbiased estimate of the amount of information in the market over time. When combined with the time series information on the satisfaction scores themselves, we will have a data base which will allow us to estimate the interaction between satisfaction ratings and market information level as they affect market shares. The extent to which the market is being rationalized by the system can also be monitored with an estimate of the variability of satisfaction ratings over time and how responsive that variability is to the level of information in the system.

Figure 1 shows the expected changes over time in quality/price and in share of the business, first (solid lines) with no improvement in information but simply from regression toward average, and second (dotted lines) with a combination of improved information and regression. The improved information effects must be examined in comparison with some norm that takes account of the regression phenomenon, as with time series multiple regression.

In contrast to the time series analysis of market function, it is also possible to argue that there should be a control area in order to make this a real experiment. Such a control area would be a similar city where information on market efficiency and changes in market shares is collected, but no improvements made in the information flow to consumers (or to vendors). It is doubtful that this is worth the cost, or that the statistical matching of areas would be effective, but it could be considered if funds were available. Even without the control area, there would be a period while sufficient information was being collected when data would accumulate on variations in satisfaction and on whether there was movement toward the mean--poor performers dropping our or getting better, and good performers backsliding.

In cases where both the prices and quality of goods or services could be obtained, the economic value of the information could be estimated by noting the extent to which individual price/quality pairs move toward the "efficiency frontier." One might even use the slope of that frontier isoquant as a rate of substitution between quality and price, in order to convert the departures from it into dollars. Then the economic value of information, and of the improvements in the market, would be crudely quantified. This would be particularly important if the market were so improved that it would seem difficult to continue to sell the information. Justifying public subsidy would be simpler if an economic measure of the benefit were available. This would be done by translating composite changes in both price and quality (satisfaction) into an equivalent change in price alone thereby quantifying the benefits of changes in dollar terms. In Figure 2, a movement from a to b would then be evaluated by the vertical distance between the two straight lines, which have a slope equal to the efficiency frontier at some compromise slope. The gain would then be $AA1 or $BB1.





Since repairs differ a great deal, an average or actual price would not mean much, yet would be better than nothing. Indeed, for repairs made in the Nome, so much of the cost is a fixed "overhead" charge that the actual cost might be usable. Quality can be crudely quantified by a five-point "satisfaction" scale. If the information system proves to be fundable and viable by user charges, such quantification is needed only to determine in borderline cases whether a particular area should continue to be covered, and even then one might want to cover a wide range of areas simply to provide a "full service." However, if it became clear that information improved markets so much that almost all the vendors ended up along the efficiency frontier (price and quality correlated almost perfectly), then in order to justify an expenditure of public money to continue the information service, one would have to provide some quantitative estimate of the benefits in dollars to compare with the dollar cost of the system. Each vendor's improvement would be weighted by the estimated number of units (of repair) sold by that vendor. Where shares change, one would average the before- and after-information sales like any other consumer surplus estimation.


It is easy to demonstrate the immediate individual benefit to users of the system's information since they are directed toward vendors that offer the highest ratios of quality to price or, perhaps equivalently, the greatest satisfaction to their customers. The social benefits of improved price/ quality correlation and more uniform satisfaction ratings are more difficult to estimate. Fortunately, the same survey that collects the basic satisfaction information also provides data on market shares and system use that can be used to test whether markets are indeed functioning more efficiently.

But before the potential benefits of the system are forthcoming, a number of research-related questions must be addressed. The most important of these concern the form of the information itself, the marketing of that information and the use of that information by consumers. Ail must be resolved to at least some extent before a consumer information-system is given a fair trial.