Trade-Offs in Attribute Levels Made By Ecologically Concerned Unconcerned Consumers When Buying Detergents

Karl E. Henion, University of Texas at Austin
Russell Gregory, University of Texas at Austin
Mona A. Clee, University of Texas at Austin
ABSTRACT - Environmental concern is less intense than it was a decade ago. Today when selecting products, consumers nay be giving much greater weight to price than to ecological attributes of product. Whether this suspected trend holds for ecologically as well as non-ecologically concerned buyer segments is the principal focus of this paper.
[ to cite ]:
Karl E. Henion, Russell Gregory, and Mona A. Clee (1981) ,"Trade-Offs in Attribute Levels Made By Ecologically Concerned Unconcerned Consumers When Buying Detergents", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 624-629.

Advances in Consumer Research Volume 8, 1981      Pages 624-629


Karl E. Henion, University of Texas at Austin

Russell Gregory, University of Texas at Austin

Mona A. Clee, University of Texas at Austin

[The authors gratefully acknowledge the financial assistance of the Sperry and Hutchinson Company whose grant in ecological marketing helped make this study possible. The also express their appreciation to Professor Mark Alpert for many ideas for improving the analysis and discussion sections; also to three blind reviewers for numerous helpful suggestions.]

[Karl E. Henion is Professor of Marketing; Russell Gregory is a doctoral student in the Department of Marketing; and Mona A. Clee is Assistant Instructor of Marketing.]


Environmental concern is less intense than it was a decade ago. Today when selecting products, consumers nay be giving much greater weight to price than to ecological attributes of product. Whether this suspected trend holds for ecologically as well as non-ecologically concerned buyer segments is the principal focus of this paper.


"During the past fourteen months we have seen a virtual storm of public protest and journalistic attention devoted to the nation's ecology. One of the prime targets of the environmental hue and cry has obviously been the business community and the role that it plays in our deteriorating environment." Times change, Written almost ten years ago, that opening paragraph of an article on ecological factors and buying behavior (Herberger and Buchanan 1971, p. 644) seems strangely out-of-step with public opinion today. No longer does environmental concern rank particularly high in national priorities (Jaroslarosky 1980). In the early and middle 1970's it did, and national policy favored a strong, regulatory approach to problems like environmental protection (Nixon 1973). As that decade drew to a close, national sentiment seemed to reflect greater sympathy for the problems of business, especially the regulatory burden it carries, which is widely considered to be an important cause of inflation (DeMuth 1980). And inflation seems to be on everybody's mind. More and more, consumers are shopping price and they seem to be less interested in buying products that are ecologically benign than they were at the start of the decade (Marketing News 1979).

This turn of events is critical to the credibility of a concept that has been referred to in the marketing literature as ecological marketing (Henion 1976). Both a suggested national policy alternative to increased regulation (Henion 1978) and itself a sub-discipline of social marketing, this concept has one central idea. It is that environmental protection and resource conservation can be better advanced through less regulation by the public sector and more enterprise in the private sector. This idea, in turn, is based on the premise that the ecologically concerned consumer is a legitimate but largely untapped market segment -- one that is identifiable. accessible and measurable (Henion 1976). Such a segment is a potential market for products that are environmentally benign (EN-products, for short).

But, clearly, no customer -- whatever his or her degree of ecological consciousness -- will buy EN-products on the basis of their ecological impact alone. The consumer obviously buys a collection of attributes. In the case of a detergent, the attributes would probably include price, cleaning ability and -- for some shoppers, such as the ecologically concerned consumer -- ecological impact. Just as for other products, a bundle of attributes is sought; and the EN-product bought will depend on trade-offs the consumer is willing to make among various levels of such attributes.


What trade-offs in attribute levels then do consumers make today when choosing products that are particularly relevant from an ecological perspective? Only two studies in the marketing literature have addressed this question.

The earliest study, which took up the issue in indirect fashion, was based on a simulated shopping experiment whose data today are ten years old (Herberger and Buchanan 1971). In that study price and ecological information were experimentally manipulated for four brands of detergents. Ecological information was found to influence brand preferences across all three price levels studied. The effect was in favor of low-phosphate brands at the expense of high-phosphate brands. Yet, in the context of the extent of the change in preference, change was not as great when phosphate-free detergents were priced disadvantageously. Less conclusive results were reported for soft drinks.

In the other study, Kerin and Paterson (1974) considered the subject of trade-offs more directly. Responding to mailed questionnaires, housewives rank-ordered detergent and soft drink product configurations, each made up one of two levels of price, brand and ecological impact. Regardless of high or low level of environmental consciousness, respondents compromised ecological compatibility first, price second and most favorite brand third.

The aim of the present study was to make a current assessment of attribute level trade-offs for an EN-product. Former assessments reflected in the studies mentioned are dated and limited. For example, neither study reported results based on observed behavior. Also, in the earlier study ecologically unconcerned consumers were not distinguished from concerned consumers -- an important distinction. Although such a distinction was made in the later study, no information was supplied about either the scale used to measure ecological consciousness or its reliability and validity. Moreover, in the 1974 study, the only price levels studied were those below and above prevailing prices. Consequently, overlooked was the realistic and important situation in which products are priced the same.

In the present study the expectation was that ecologically concerned consumers would value attribute levels of an EN-product differently from unconcerned consumers. This prediction is consistent with key findings in much of the ecological marketing literature. Numerous studies (e. g. 's, Anderson, Henion and Cox 1974, Kinnear, Taylor and Ahmed 1974, Webster 1975) have shown that these two segments exist and that consumers comprising them are inclined to believe and behave in different ways when buying products that are particularly ecologically relevant.


As in the previous two studies, laundry detergents were also the product focus of the present study. The field setting was five stores of a supermarket chain situated in a medium-size city in the Southwest. They served shoppers living in middle to upper-middle class neighborhoods. Each store carried the same line of about 25 brands of powdered and liquid detergents. Several of them were phosphate-free: the powdered detergents Arm & Hammer, Purex, Trend, Topco Orange and Ivory Snow; and the liquid detergents ERA and Dynamo.

Trained interviewers, stating that they were conducting a survey on consumer preferences using a technique called conjoint analysis, administered a pre-tested five-minute questionnaire to a total of 188 in-store shoppers, mostly as they filed by the detergent aisles of the stores. Of this number 110 were observed and recorded, prior to the interview, as actually taking a particular brand of detergent from the shelf and selecting it for purchase. The three-part questionnaire was administered to the respondents in the following order: (a) trade-off tables of paired attribute levels; (b) an index of ecological concern (IEC) developed by Kinnear and Taylor (1973); and (c) several demographic questions, which are of no direct concern to the present study. Evidence of construct validity of the IEC has been reported by the above investigators (1973, pp. 192-193); additional evidence is provided elsewhere (Henion and Wilson 1976).

Data were collected by the trade-off method instead of the full-profile method (Green and Srinivasan 1978) for several reasons. Knowledge of the relative importance of the attributes was necessary for testing one of the hypotheses of the study; and this can be easily done by comparing utility ranges based on the information that respondents provide directly in the trade-off method (Johnson 1974). Also, the experience of one of the authors gained from other field experiments using both data collection methods resulted in a decided preference for the trade-off method because it was easier to apply and explain to busy shoppers in supermarkets. The issue of interactions, which cannot be estimated under this method, was not considered a potential problem; the attributes of the present study were viewed as independent and non-redundant and hence orthogonal.

Another issue that should be addressed arises from self-perception theory. According to that theory, some persons who bought low-phosphate detergents conceivably might have labeled themselves as ECCs and developed and registered an attitude on the IEC consistent with such behavior, both of which may have had low reliability if measured on a later occasion. However, disguiser steps were taken so that subjects would not be self-conscious about the fact that the ecological aspect of their buying behavior was being monitored. Moreover, the limited time shoppers were willing to take for the interviews did not warrant embedding the IEC in a lengthy battery of other measures.

The interviewer first explained to the respondent how to fill out a trade-off table in which pairs of attribute levels were to be rank-ordered; then the interviewer had the respondent practice on two tables containing attribute pairs for unrelated products. These tables helped disguise the ecological interest of the study. Next, three 3X3 trade-off tables for detergents were separately presented. (Since only one of the three attributes was ecologically related, the disguise was further protected.) The three detergent attributes (and their three levels) were price (154 more than, 154 less than and same price as usual); cleaning power (somewhat cleaner, equally clean, and somewhat less clean); and phosphate content (high, average, and low). Choice of these attributes served the main purpose of the study, namely, to clarify their previous ambiguous role in the literature of ecological marketing. Other detergent attributes, such as softness or odor, while relevant for different purposes such as new product development, were not relevant for this study.


Motivated by the studies of Herberger and Buchanan (1971) and Kerin and Paterson (1974), the following predictions were made:

H1:  A low phosphate detergent has mere utility for ECCs than for non-ECCs; and, conversely, a high phosphate detergent has more disutility for ECCs than for non-ECCs;

H2:  A detergent's phosphate content will be considered by ECCs to be more important, and by non-ECCs to he less important, than other attributes.

H3:  An EN-Product that is higher priced and/or less functional than its environmentally less desirable substitute will have less overall disutility for ECCs than for non-ECCs, e.g., a low-phosphate detergent priced higher and with less cleaning power than a high-phosphate detergent.


From the sample of 110 respondents who were observed buying a detergent, 18 ecologically concerned and 29 unconcerned consumers were identified, as well as 63 neutral or inconsistent consumers. Identification of the first two groups (ECC and non-ECC) was based on two criteria: observed purchase of a detergent and score on the IEC, whose range was from 0 to 25 points. If shoppers bought a phosphate-free detergent and also scored above the third quartile (15) on the IEC, they were classified as ECCs. If they bought a phosphate detergent and scored below the first quartile (7), they were classified as non-ECCs. Interquartile scorers were considered neutral whichever detergent type was bought; there were only six respondents whose scores and purchase behavior were inconsistent.

From the rankings of attribute level pairs made by each of the 110 identified respondents, a parametric utility for each attribute level was estimated using conjoint analysis. The analysis was based on an additive model and was carried out using $FTROFF (Nehls, Seaman and Montgomery 1976). This algorithm rests on the assumption that the dependent variable is, at most, ordinally scaled.

$FTROFF prints out a badness-of-fit measure, 0, for each respondent. It is based on the extent to which the sums of the estimated utilities for each pair of intersecting attribute levels in a trade-off table have rank orders dissimilar to the input ranking that the respondent made of the sample pairs. This measure, which accounts for both number and size of errors arising from the revealed dissimilarities, accumulates such errors across all of the tables that the respondent fills out, The average f for the ECC group was .054; for the non-ECC group, .085. The average proportion of correct comparisons of utility sums in each cell for the ECC group was 87.5%; for the non-ECC group, 84.3%.

For each attribute level, an average utility was calculated from the utilities derived from the conjoint analysis of ranks at the individual level. The results for the consumers in the ECC and non-ECC groups are presented in Table 1. The algorithm, forces the algebraic sum of each respondent's utilities for the three levels of each attribute to equal zero; hence the sum of the utility averages for each attribute at the group level approximates zero.

The utilities for the nine attribute levels (3X3) were cast as independent variables first in a multiple discriminant analysts, with the ECC vs. non-ECC vs. neutral and inconsistent group membership (utilities not shown) as the criterion variable. The results were not significant using the whole spectrum of subjects. However, the theoretical interest was in the ECC and non-ECC groups. Hence, the indifferent and inconsistent group was dropped from the model and a simple discriminant analysis was next run using the other two groups. Forty percent of the variance in the resulting discriminant function was explained by the group variable. The canonical correlation coefficient was .63; the Wilks Lambda of .60 was significant at .014 (X2 = 20.71 with 9 d.f.).

The jackknife technique was used in cross-validating the results of a parallel stepwise discriminant analysis, owing to the small size of the sample. Correct classification and a-probabilities for the two groups were: ECC (78%, p <.02) and non-ECC (55%, p<.28). The rest of the paper focuses on findings for these two groups.



Within each attribute for each consumer group in Table 1, the pairs of correlated means (i.e., those in the vertical direction) were significantly different for each of the three possible pairs of attribute levels (t's had p<.001 or better in all cases but two, which had p<.02 instead: one in the ECC group, d.f. = 17, and one in the non-ECC group, d.f. = 28).

The direction and size of the utility means show that persons in both groups place higher utility on lower price and on higher cleaning power. This result one would expect to find. Of special interest, however, is the finding that not only ECCs but also non-ECCs place higher utility on low phosphate content than on average or high levels.

We turn next to a comparison of uncorrelated means across the two groups (i.e., those in the horizontal direction of Table 1). While there are 9 possible comparisons, only 6 are particularly relevant since values of the utilities for the 3 middle levels for the attributes will tend toward zero, owing to the logic of the $FTROFF algorithm. Of the six comparisons, three revealed differences in mean pairs that were significantly different. The other three were not significant, although they were in the expected direction; the probability that three pairs would randomly be arrayed in the expected directions, given that each of them is equal likely to have been reversed, is .125, which approaches some statistical significance. Overall, strong support was found for H1.

The disutility of high phosphate content was significantly higher for ECCs than for non-ECCs (p<.001, d.f. = 65); and the utility of low phosphate content was significantly higher for ECCs than for non-ECCs (p<.003, d.f. = 45). In a similar vein, low cleaning power had less disutility for ECCs than for non-ECCs (p<.05, d.f. = 45); and, while not statistically significant (p<.27) high cleaning power had less utility for ECCs than for non-ECCs. Unfortunately, none of the price comparisons produced significant differences. It seems that price qua price (i.e., when not compared with other attributes) is about equally important to both groups of consumers. Nevertheless, numerically, the values of the means are in the expected direction. High price had slightly less disutility for ECCs than for non-ECCs; low price, slightly less utility,


The relative importance of a set of attributes can be indexed by computing the range of utility values across the levels of an attribute and comparing it with the corresponding ranges for the other attributes. The greater the range the greater the importance. For each respondent the range of utilities for each attribute was calculated and the means of the ranges for the three attributes for the two groups are presented in Table 2.

The rank order of the attributes according to mean utility ranges is quite different for the two consumer groups. As mentioned earlier, low phosphate content had positive utility for both groups. However, ECCs considered this attribute to be of first importance (with the highest range of .9021), cleaning power next and price last. But non-ECCs considered phosphate content to be of least importance (with the lowest range of .5628), placing price and cleaning power ahead of it.



The one-way ANOVA of the mean ranges for each consumer group produced significant F-values. For the ECC group, the results of t-tests for the three pairs of correlated means were: no significance between price and cleaning power (p<.20); approaching significance between cleaning power and phosphate content (p<.10); and significance between price and phosphate content (p<.02). For the non-ECC group, the results for the sane pairs were, respectively; approaching significance (p<.10); significance (p<.05); and significance (p<.005). Consequently, good support was found for H2.

Amplifying these findings were the results of t-tests for two of the three pairs of uncorrelated means (i.e., those in the horizontal direction). For two attributes --phosphate content and cleaning power -- significance (p<.005 and p<.056, respectively) was found between the ECC and non-ECC groups, but not for price.

Product Simulations

For each respondent, a composite utility was calculated for each product simulation or configuration. A configuration was made up of 3 attribute levels -- one from each of the 3 attributes. The utilities of the 3 levels were summed to form a respondent's composite utility for the configuration. The means of the respondents' composite utilities (uc) for the 27 possible product configurations are shown in Table 3 for the ECCs and the non-ECCs.

Direction signs, - 0 +, in Table 3 indicate whether a level for an attribute is desirable relative to the attribute's other two levels. These signs were employed in the following manner. Product configurations have been listed in Table 3 in descending order according to the sum total of the advantages or benefits that various attribute's levels in a configuration are Judged to have. Thus, the first entry in the list is a detergent that has a low (L) price (i.e., 15 cents leas), high (H) cleaning power and low (L) phosphate content. This configuration, symbolized by the ordered triple (LHL), is also represented by an ordered triple of signs (+ + +). Here a + indicates the most desirable level of an attribute; a 0, a middle level; and a -, the least desirable level. Let a + and a - be interpreted as +1 and -1, respectively. If equal weights (1 : 1 : 1) are then arbitrarily used as attribute weights for each sign in the triple, then the algebraic sum of the signs indexes the utility of a configuration. Hence, the first triple yields an index of +3. Expectedly, for configuration 1 the uc obtained for each consumer group has the highest value; for configuration 27, the lowest value.



Abstracted from Table 3 and shown separately in Tables 4 and 5 are the mean composite utilities, respectively, for the nine detergent configurations with low phosphate content and the nine with high phosphate -- results on which H3 principally depend.

Based on the composite utilities, support was found for this hypothesis. Considered first will be the "and" part of the "and/or" statement of H3.  For ECCs the mean composite utility (uc) for configuration 20 was -.299901 but only -.330226 for item 21, although the difference in the correlated means was not significant; whereas, for non-ECCs the corresponding uc's were -.661557 and -.038328 (p<.001). Thus, a low-phosphate detergent priced higher (by one level) and with less (by one level) cleaning power than a detergent with a high level of phosphate content was considered by ECCs to have more utility (right direction): by non-ECCs, to have significantly less. However, these reversals did not hold if both the disadvantages of higher price and less cleaning power were at the extreme levels, i.e., by two levels: for ECCs the utilities were -.299901 (again configuration 20) vs. .106660 (configuration 8) with p<.01; for non-ECCs, -0.661557 vs. .419263 with p<.001. These last results however, are based on extremely strict criteria. The ECC is being asked to sacrifice for ecology not only price and cleaning power (function) but also to the extent of the most disadvantageous levels of these attributes.





It is noteworthy that in Table 4 the composite utility mean for the ECC group is higher than that for the non-ECC group for each of the nine product configurations with low phosphate content. The reverse is true in Table 5: the uc is lower for the nine with high phosphate content. The probability of either of these outcomes is less than .002 (binomial test). Obviously the ECC places greater value on phosphate content being at a lower level than does the non-ECC.

The utilities were next examined under criteria that were lees strict, namely, the "or" part of the "and/or" in H3. Rather than both disadvantages, only one disadvantage or the other were considered -- again at one level and two levels. The results, which are broken out from Table 3 and summarized in Table 6, provide strong support for H3. In the case of these somewhat less unattractive configurations -- namely, a low-phosphate detergent with either a higher price (whether 1 or 2 levels) or less cleaning power (also whether 1 or 2 levels) -- the EN-product was still considered by ECCs to have higher utility than its environmentally less desirable substitute; by non-ECCs, lower utility. In six of the eight possible pairings of the uc's of the EN-product with those of the substitutes, significant differences in correlated means were found (p<.01 in four cues and p<.05 in two); in the other two pairings -- both in the non-ECC group -- the one with configuration 25 approached significance (p<.08) while the other with 24 was not significant.




These results, based on concomitant measures of observed behavior and attitude, help buttress the case of ecological marketing. Today inflation is driving consumers more and more to shop price, and yet concerned consumers in this study are at least implying -- by the trade-offs they registered -- that they would be willing to subordinate price to ecology. Unconcerned consumers, however, place price and function ahead of ecology. Therefore, our findings contradict those of Kerin and Peterson (1974) who reported both consumer groups as placing price ahead of ecology. The strength of the present findings is our identification of the ecologically concerned consumer is more valid than theirs. One caveat is that these findings are applicable to consumers with middle to upper-middle socio-cultural background.

Given this caveat, these findings have relevance for a long-standing, unresolved issue in ecological marketing research. If consumers show greater concern for price, as is expected, will sales of an EN-product be adversely affected (Henion 1976)? That depends. Is a particular EN-product less or more expensive than the product for which it is a more ecologically desirable substitute? (Obviously the critical condition is when it is more expensive.) What is the cross-elasticity of demand for the two products? How do patterns of price inelasticity of an EN-product differ for ecologically concerned and unconcerned consumers? Based on the earlier studies, the presumption is that the patterns are the same for both groups of consumers, namely, little if any elasticity (and further that substantial cross-elasticity exists for an EN-product and its functional counterpart which is ecologically insulting). However, the present assessment of attribute levels trade-offs -- which were made by one sample of respondents for one EN-product and analyzed using a decompositional model of consumer preference --suggest a different conclusion. Concerned consumers might possibly be willing to pay a premium for an EN-product, even during times of high inflation, although unconcerned consumers would probably not -- based on the results of the range analysis.

However, the external validity of the conjoint measurements taken is still an open question, making predictions of the degree to which consumers actually would trade-off higher prices against lower phosphate content quite another matter. As with previous studies, the design of this study may have increased the salience of the area. For we have presumed that buyers perceive phosphate levels as salient; further that such salience was probably augmented by labeling the levels for respondents in the trade-off tables. In the real world attribute levels compete for buyer recognition with other cues that may be more dominant, e.g., price; and in the case of detergents usually, phosphate content is merely registered rather than highlighted (although it was highlighted on the package of one of the six phosphate-free detergents in the present study). External validity would be more assured if phosphate content had been labeled prominently as it was in one study (Henion 1972); surveying instruments would then not differ much from the information available to buyers when shopping. At any rate, results from the present study can justifiably be compared with those of prior studies, since whatever cue bias exists in the present study was present in the others, too.

Accordingly, field experiments need to be conducted in which prices of EN-products are experimentally manipulated with phosphate content highlighted. If inelasticity of EN-products were found for the ecologically concerned consumer, then the suspicion of ecological lip service would be largely groundless for this group. Meanwhile, confirmation that concerned consumers value attribute levels of an EN-product differently from unconcerned consumers are encouraging from the policy perspective of ecological marketing, coming as it does five years after the 1974 study with its negative results.


Anderson, W. Thomas, Henion, Karl E, and Cox, Eli P. (1974), "Socially vs. Ecologically Responsible Consumers," American Marketing Association Combined Conference Proceedings, 36, 304-311.

Demuth, Christopher C. (1980), "Constraining Regulatory Costs, Part One: The White House Review Programming," Regulation, (January/February), 13-26.

Green, Paul E. and Srinivasan, V. (1978), "Conjoint Analysis in Consumer Research: Issues and Outlook," Journal of Consumer Research, 5, 103-123.

Henion, Karl E. (1972), "The Effect of Ecologically Relevant Information on Detergent Sales," Journal of Marketing Research, 9, 10-14.

Henion, Karl E. (1976), Ecological Marketing, Columbus, Ohio: Grid.

Henion, Karl E. (1978), "Ecological Marketing: From Normative to Descriptive Model?" in Future Directions in Marketing, Johan Arndt, George Fisk, and Kjell Gr°nhaug, eds., Boston: Marketing Science Institute, 301-311.

Henion, Karl E., and Wilson, William H. (1976), "The Ecologically Concerned Consumer and Locus of Control," in Ecological Marketing, Karl E. Henion and Thomas C. Kinnear, eds., Chicago: American Marketing Association.

Herberger, Roy A. and Buchanan Dodds I. (1971), "The Impact of Concern for Ecological Factors on Consumer Attitudes and Buying Behavior," American Marketing Association Combined Conference Proceedings, 33, 644-646.

Jaroslovsky, Rich (1980), "Douglas Castle's Balancing Act," The Wall Street Journal, (April 11), 18.

Johnson, Richard M. (1974), "Tradeoff Analysis of Consumer Values," Journal of Marketing Research, 11, 121-214.

Kerin, Roger A. and Peterson, Robert A. (1974), "Selected Insights of Ecologically Responsible Behavior," American Institute for Decision Sciences Proceedings, 6, 33.

Kinnear, Thomas C. and Taylor, James R. (1973), "The Effect of Ecological Concern on Brand Perceptions," Journal of Marketing Research, 10, 191-197.

Kinnear, Thomas C., Taylor, James R. and Ahmed, Sadrudin A. (1974), "Ecologically Concerned Consumers: Who Are They?", Journal of Marketing, 38, 20-24.

Marketing News (1979), "Permanent Inflation Is Key Factor in Future of Marketing," (September 7), 3-4.

Nehls, Lyle, Seamen, Bruce, and Montgomery, David B. (1976), "A PLI Program for Trade-off Analysis," Cambridge, MA: Marketing Science Institute.

Nixon, Richard M, (1973), "The President's Massage," Environmental Quality: The Fourth Annual Report of the Council on Environmental Quality, Washington, D.C.: U.S. Govt. Printing Office, vii.

Webster, Fred E., Jr., (1975), "Determining the Characteristics of the Socially Conscious Consumer," Journal of Consumer Research, 2, 188-196.