Research Contributions to Public Policy: Discussion Comments

Ivan Ross, University of Minnesota
[ to cite ]:
Ivan Ross (1979) ,"Research Contributions to Public Policy: Discussion Comments", in NA - Advances in Consumer Research Volume 06, eds. William L. Wilkie, Ann Abor, MI : Association for Consumer Research, Pages: 526-528.

Advances in Consumer Research Volume 6, 1979      Pages 526-528


Ivan Ross, University of Minnesota


The three papers composing this set share some things. First, each has a public policy implication (but, doesn't everything these days?). Second, each proposes or implies methodological procedures for estimating the effect of certain market actions (e.g., changes in attributes, repositioning of products) on consumer responses, perceptions, or behavior (Burger and Venkatesh and Golden, et. al.) or, vice-versa (Bourgeois). Indeed, the CB radio and transportation papers are very closely related, and were that it were possible for the authors of these two papers to have spent a day in seclusion before this Conference, we might have heard today one instead of two papers, and moreover, some of the confusion this discussant retains concerning the relative merits or demerits of alternative procedures for estimating attribute utilities (trade-off analysis, factor and discriminant analysis, etc.) might have been ameliorated as a result of such a coalition.

Third, each of the papers, in its own way, makes some advance in either methodological or conceptual thinking and application by extending our knowledge in three areas where, by this time, there is an accumulating "history" of research.


What is a Market?

There is no question but that procedures for determining or estimating the extent to which multiple "offerings" define a set within which high substitutability obtains (high cross-elasticities of demand) are useful. Moreover, as Bourgeois notes, there are noteworthy differences in approach to this question by various investigators. The troublesome thing about the conceptual positioning of this article, however, is that Bourgeois' proposed procedure for "defining" a market appears to be nothing more than a perceptual measure of substitutability which is offered as a "substitute" for a behavioral measure of substitutability. It is unclear how this advances the state of our knowledge or of our methodological repertoire.

Moreover, the author does not directly confront the question of the adequacy of a measure of substitutability as a measure of "market." Are not there many meanings of "market" which have relevance which do not directly depend upon any conception of cross-elastic demand? For example, in trademark infringement matters, is it not the case that confusion as to source or origin between two products has a "market definition" meaning, whether or not consumers perceive the products as substitutes?

Naturally, the author is free to employ any definition of "market" he wishes to demonstrate the value of the heuristic he offers to aid in its delineation, but some would consider it useful to read a considered evaluation of how this definition of market is most usefully applied or is more pertinent in terms of its applications.

Individual Versus Aggregate Analysis Levels

The terms "market" and "market segment" are being used in Figures 3 and 4 in a way that may lead to some confusion. This discussant reads the notation as defining the "market" as consisting of "all and only those" products seen in common by consumers as satisfying the same "need", and as defining "market segments" as those products seen by at least one, but not all consumers, as satisfying the same "need." But then, apparently, Figures 3 and 4 do not contain any products which are not part of either a market or a market segment (i.e., none is outside of a "C set"). Thus, when Bourgeois states that (with reference to Figure 3), "the present market is therefore defined as the set of those four available products...", he is restricting our attention to the "market" and not to the "market segments," and seems to be asserting that only these four products compose a "market" for these two consumers. This discussant sees fifteen products which compose "a market" for these two consumers, and he is 100% certain of that. That four of them "compose a market" for the same two consumers is undoubtedly of marketing interest but is not an entirely satisfactory basis for asserting we are "less certain" that the other eleven products are not "in" the market for "that" need. There may be a semantic problem here resulting from moving between "individual'' (consumer) and "aggregate" (market) levels of analysis.

The Value of the Proposed Market Delineation Heuristic

The approximation heuristic (equation 4) estimating the probability that a particular set of products are perceived as "substitutes" seems an efficient procedure for identifying "similar" product groupings. It may be of interest to inquire, however, whether or not consumers understand (perceive) the concept of "market" and the product set that composes it (them) to be determined simply and exclusively by whether or not they place the same two or more products in the same class/ usage situation cell. And how does the calculation of aggregate probability values handle obvious problems of segmentation with respect to product groupings/sets and interactions of such groupings with usage situations? Can the heuristic be used for investigating the existence of segments and the marketing implications of repositioning products (given some measurement on attributes which form the basis for a consumer's placement of a product in a class/usage situation)?

It would appear that the heuristic can be quite useful in various managerial applications, but it would be helpful if the author would provide examples of applications which would shed light on issues such as... what levels of confidence for "homogeneity" are relevant/required for what kinds of decisionsY, what if the same product belongs to several segmentsY, how do we bridge between probability estimates and product/ individual attributes which cause product sets to occur in the first place?



Linda L. Golden

John F. Betak

Mark I. Alpert

Individual Versus Aggregate Analysis Levels

Although the paper very usefully extends our methodological "tool kit" in looking at choice behavior in the mass transit situation, some will have a concern about the loss of individual level data as a result of calculating attribute utilities across subjects. Ultimately, if we want to predict the number of persons who might be moved from personal auto to mass transit via the change or addition of some determinant attribute or combination thereof, don't we have to retain data at the individual level and in some way simulate choice changes per each individual via manipulation (real or hypothetical) of attributes/levels of attributes?

Definitional Problems

Several issues deserve clarification. How did the authors begin with 27 attributes from which 11 were identified as (presumably), "determinant", and end up with using only nine in the trade-off analysis, and of these nine, two were "non-determinant"? Some of the "operational definitions" seem equivocal. For example, the number of hours per day and number of days per week that a transportation mode is available may well be a part of attributes such as "dependability", "flexibility'', and "convenience", but would not intuitively seem to exhaust the definitions of these attributes. Does not "dependable" suggest more than the number of hours/ days available? How "flexible" or "convenient" is a transportation mode if it is available eighteen hours a day but not at all between those hours in the morning or afternoon when a particular rider may require its use?

These issues would appear rather crucial since the information with which we are presented in Tables 2 and 3 presumably is sufficient to lead the authors (and readers) to decisions as to which attributes "need" changing most in order to increase the probabilities of mass transit patronage. Therefore, if there are problems in the way in which determinant attributes have been selected and defined, this information may be misleading.

Further, since the outcome of the conjoint analysis is so sensitive to the particular levels chosen to represent each attribute, the rationale for this procedure seems to deserve greater attention. How does one determine if the "levels" chosen to illustrate an attribute are reasonable, realistic, and "equivalent", so to speak, across attributes.

Not only does this issue receive scant attention in this research paper, it seems a common deficiency in most of the conjoint literature. To argue as do the authors that "care was taken to select levels which typified the relevant range for current transportation modes..." may not be a satisfactory assurance for many researchers. Consider, for example, the possible difference in part-worth utility attaching to a range of "never" to "sometimes but not often" re "encountering dangerous people" as compared to "never" versus "often" (as employed by the investigators).

Trade-offs, Segments, and Nonlinearity

Whereas ten of the eleven attributes appear to have a "favorable" (e.g., low cost per mile, high level of comfort), and unfavorable end (e.g., high cost per mile, low level of comfort), for most all S's, one attribute, "opportunity to socialize", may be nonlinear in preference levels which contributes to attribute interaction problems perhaps not well handled in the trade-off process, particularly where segments may exist. For example, if one wants to" often have an opportunity to socialize" then "longer total travel time" may be desirable, but only if "the possibility of encountering dangerous people" is "never." Intriguing interaction possibilities are endless. And although the authors recognize this nonlinearity issue they interpret it as a "golden mean" rather than as conflicting segments at the extremes of the dimension. Thus, as they note, not only is multiple regression questionable, but so is any linear assumption model.

It is equally important to the interpretation of these data to note that trade-offs were made for transportation modes in general, not for specific modes. Does this procedure result in different attribute importance inferences than if specific mode comparisons (e.g., auto-bus, auto-train, auto-carpool, etc.) had been employed? Again, attribute interaction combined with "situational" interaction (mode as situation, work versus school as situation, "in hurry" versus "not in hurry" as situation, etc.) seem relevant here.


Philip C. Burger

Alladi Venkatesh


The study as reported leads to the impression that it was initially undertaken to describe and "segment" the CB user market. The transition between these analyses and the simulation referred to is ambiguous. Were specific belief and attitude semantic differential items dealing with "telephonic needs" and "entertainment'' included, and how could a simulation of the effect of the proposed change(s) be conducted without any basis for estimating the nature and degree of perceptual (attitude-belief) change? Would not exposing a sample of respondents within each of the key segments to the proposed change at least in concept description form have provided helpful information for the simulation? And, exactly how does one compare the value of the increase in satisfaction with telephonic communication with the value of the decrease in satisfaction in entertainment value.., how was satisfaction measured?


A research contribution to public policy decisions is always laudable, even if the mechanics of the application are not always clear. Further, it seems particularly instructive that two rather different segments were identified here with respect to primary satisfactions sought in the use of the product; indeed, not only different, but antithetical in terms of the projected impact of the policy decision. This may reasonably be expected to be the case in various other policy decision contexts, and yet is rarely noted in discussion or research. For example, if younger children use advertising differently than older children, does the correct policy decision "trade-off" the sizes of these segments taking into consideration the costs and/or benefits associated with each proposed policy action? Or, is it even reasonable to talk about "tradeoffs" in comparing segments for policy decision purposes?


Historically, it may be true that when consumer researchers "discover" a new area of inquiry, in this case, public policy applications of consumer research, there is some tendency to let tool or technique considerations gain more attention than they should, and in so doing, sometimes to mask more fundamental issues. Each of use has our own favorite technique. But can't we let the problem determine the technique rather than vice-versa? Are utilities for CB and transportation mode utilities best determined via regression weights or trade-off analysis derived part-worths? Or doesn't it make any difference?

And, individual versus aggregate levels of analysis continue to pose interpretation problems. Certainly, it makes sense to retain individual level data if we are going to estimate the number of people who would shift this way or that given some change in marketing mix variables.

With respect to these issues, it would be desirable to have a "technical conference" co-sponsored by ACR and by all the Federal Agencies and other regulatory or quasi-regulatory groups so that standards regarding types of analyses, general design issues, etc., could be developed.

Two other suggestions seem pertinent:

(1) Most any research in public policy applications of consumer behavior seems unavoidably concerned with consumer satisfaction/dissatisfaction issues. Standardization in definition and measurement of CS/D dimensions would be very desirable so that "normative" or benchmark data could be developed in the various domains of public concern with consumer behavior. How "satisfied" (on a commonly employed scale) does a person have to be in order to ride a bus versus a car? Is there a comparable level of satisfaction required before a person will switch to a lower energy usage form of transportation? Should we be more concerned with the satisfaction enhancement or the dissatisfaction minimization side of the policy implication question?

(2) Just as the FTC and presumably other agencies have determined the worthiness of establishing priorities on issues to be investigated, so might ACR'ers be guided by such a schema. As consumer researchers, should the dollar impact (cost/savings) of a policy decision on consumers be considered as importantly as the increase in satisfaction or the years of additional health or X utilities of perceptions of increased personal competence in the marketplace? If we do not become entangled with "the criterion question" when we see regulators perhaps less prepared than we not fearing to plunge in, do we do a disservice to our public constituents?