Attitudes Toward Energy Consumption: Segmenting the Gasoline Market

David J. Barnaby, University of Tennessee
Richard C. Reizenstein, University of Tennessee
ABSTRACT - This paper provides an empirical assessment of consumer attitudes and behavior relative to gasoline usage during the Arab oil embargo. The gasoline market is segmented into usage groups by AIO's and socioeconomic variables. Information included in segment descriptions suggests differential marketing strategies and tactics for each, directed toward reducing consumption and conserving energy.
[ to cite ]:
David J. Barnaby and Richard C. Reizenstein (1976) ,"Attitudes Toward Energy Consumption: Segmenting the Gasoline Market", in NA - Advances in Consumer Research Volume 03, eds. Beverlee B. Anderson, Cincinnati, OH : Association for Consumer Research, Pages: 246-251.

Advances in Consumer Research Volume 3, 1976      Pages 246-251


David J. Barnaby, University of Tennessee

Richard C. Reizenstein, University of Tennessee


This paper provides an empirical assessment of consumer attitudes and behavior relative to gasoline usage during the Arab oil embargo. The gasoline market is segmented into usage groups by AIO's and socioeconomic variables. Information included in segment descriptions suggests differential marketing strategies and tactics for each, directed toward reducing consumption and conserving energy.


For most of the twentieth century, American consumers have been accustomed to abundant supplies of inexpensive energy. In addition, a myriad of products and services provided by seemingly limitless sources of power have fostered an "energy affluent" style of life. This almost unconscious dependency on an unending energy supply came to an abrupt and disconcerting halt with the Arab oil embargo in late 1973. Suddenly, unexpectedly, the prospect of energy allocation, rapidly rising prices, and shortages became a reality, with gasoline rationing a very real possibility -- all within an extremely short time span. The public began to realize that those who had been forecasting just such conditions, those who had been labeled "doomsayers", were, in fact, forecasting a situation which could fundamentally alter the American life style: the energy crisis.

The sudden energy shortfall created a need for information on public awareness of, attitudes toward, and usage of world energy resources. The limited economic data available did not provide insights into the opinions and consumption patterns of various segments of the American populace. Thus, a people-oriented information need exists, in order to provide inputs to government, business, and consumer groups, and to assist in policy formulation, product and promotion development, and conservation planning.


Unfortunately, this need for people-oriented data on the energy crisis specifically related to attitudes toward, perceptions of, and usage of energy resources, has primarily remained unfulfilled. Major public opinion organizations such as Opinion Research Corporation have conducted weekly polls, but these are primarily longitudinal analyses of a limited number of key categories; they do not deal with in-depth exploration of consumer attitudes and perceptions [Opinion Research Corporation, 1974]. Talarzyk and Omura, on the other hand, have performed in-depth analyses of consumer attitudes, perceptions, and usage patterns related to energy; however, their data base is a cross-sectional study conducted in March, 1974, after the lifting of the Arab oil embargo [Omura and Talarzyk, 1974; Talarzyk and Omura, 1974]. The current study attempts to establish a baseline of relevant attitudinal and usage information anchored at a key point in time (February, 1974, at the peak of the oil embargo). Data is thus provided for the cross-sectional analysis to be discussed below; in addition, it yields a basis for future longitudinal analyses.

Although the intent of this research is primarily to describe varying attitudes and energy consumption patterns among users in the Southeastern United States, a conceptual framework specifying various stages in the decision process has been included to show the range and relationship of the variables measured. A general paradigm, presented in Figure 1, incorporates the types of variables used to characterize group behavior and defines the relationships between sets of variables. This description of the consumer decision process borrows heavily from and paraphrases the Engel, Kollat and Blackwell model [Engel, Kollat and Blackwell, 1971].

Briefly, the general model of the gasoline decision process formulated in Figure 1 suggests that several levels of variables are important in shaping gasoline consumption behavior. These influences include the constraints of environmental and social factors as determined by sources of information. Informational influences in turn may be filtered according to environmental factors which will affect perceptions. Within the individual's cognitive framework, attitudes shaped by perceptions and environmental factors may lead to behavioral intentions and ultimately to overt behavior which could provide feedback through various levels of the decision process.


For the purpose of this research only, portions of three of these sets of variables were included for analytical purposes. Within the environmental and social factor variable set, 15 demographic variables were used. In the attitudinal variable set, 31 AIO variables and 10 importance rankings of national issues were included in the analyses. Finally, a surrogate for behavior was measured by reported gasoline consumption. Given the interaction between variables shown in Figure 1, the framework for the analysis of the data was predicated on testing the relationship between behavior as measured by gasoline consumption, and demographic and energy/pollution related attitudinal items.


In February, 1974, at the peak of the Arab oil embargo, a mail questionnaire was sent to 2500 residents of three medium-sized (100,000-350,000 population) Southeastern cities. As the questionnaire dealt with the major issues of air pollution and air pollution-energy tradeoffs as well as the energy crisis, test cities were selected on the basis of the only universal measure of air pollution, the average number of suspended particulates of matter per cubic meter of ambient air. Thus, one low pollution city (Columbus, Georgia), one medium pollution city (Charlotte, North Carolina), and one high pollution city (Chattanooga, Tennessee) were chosen.

The total of 2500 questionnaires was distributed proportionate to the population of each of the three cities, based on 1970 census figures. Respondents were selected within cities from the telephone directory, using a systematic random sampling technique. Specially designed envelopes, colorful commemorative stamps, and a standard follow-up letter were all employed to maximize returns [Reizenstein and Barnaby, 1975]. A total of 2389 questionnaires were ultimately delivered (111 were undeliverable due to moves or insufficient address), resulting in 922 usable responses, a rate of 38.6%.


Importance of National Issues

To determine the relative importance of a set of national problems and issues, ranking of the importance of these issues were used to establish a framework of "consumer concern" against which specific energy data could be compared. Table 1 lists the various issues by order of mean ranking for a sample of 922 subjects. While the three most important national issues include "the energy crisis", it appears that this issue is perceived as somewhat less important than either "corruption in government" or "inflation". Given the timing of the survey, February, 1974, at the apex of the Arab oil embargo, it is interesting to note that two other issues are perceived to be of greater importance by the sample respondents. This fact is somewhat in contradiction to Gallup data [Gallup Poll, February 15-18, 1974 [Opinion Research Corporation, 1974].] which found the same three problems to be of greatest importance, but in inverse order; it may, however, be indicative of the lesser effect of the energy crisis on the Southeast as opposed to metropolitan areas across the nation.

Energy and Pollution AIO's

In addition to the global problem areas detailed in Table 1, a more specific set of 31 variables dealing directly with energy and pollution attitudes, interests and opinions was included in the analysis. The AIO's were scaled on a five-point interval from 1, strongly disagree, to 5, strongly agree. (AIO statements were phrased and ordered to attempt to minimize any halo effects.) Results of the responses to the 31 AIO's

are shown in Table 2. The energy related AIO's indicated various levels of disagreement with regard to the necessity of gasoline rationing, tax incentives for energy producers, and reduction of late evening radio and television programming. Various levels of agreement were shown for statements indicating the oil companies' deliberate creation of the energy crisis to enhance profitability, lessening of pollution from reduced automobile travel, negative effects on the economy from loss of jobs, and changes in our traditional life style.





In an effort to determine the extent of any overlap in the variety of AIO's and to provide a possible structure for the constructs-being analyzed, a factor analysis was performed. [The purpose of this factor analysis is considerably different from a similar analysis of energy related to AIO's by Talarzyk and Omura [Talarzyk and Omura, 1974]. Their purpose was to ascertain respondent logic and AIO item association.] Varimax rotation of the principal components solution yielded eight factors with eigenvalues greater than one. The eight factors explained more than 53% of the variance of the original 31 variables. (Therefore, the number of variables were reduced by a factor of 4 with a loss of less than one-half of the original information.) The varimax rotation successfully spread the variable loadings across the eight factors so no cross-loadings of variables were evident, and enhanced the interpretation of the factor. An arbitrary cut-off of .4 for variable loadings was used in preparing Table 3.

In Table 3 descriptive names are attached to the eight factors consistent with the variables having the highest loadings. Factor 1, "Pollution is of Secondary Importance", implies, once again, that energy issues outweigh pollution problems, given the existing "personally negative effects of pollution" (Factor 2). A "gloomy, pessimistic" dimension regarding energy and pollution is shown by Factor 3 as opposed to a "silver lining" dimension in Factor 7. Factors 4 and 6 underscore the primary importance of pollution and energy, respectively, while Factor 8 indicates a necessity for gasoline rationing in the immediate future.

Although the interpretation of the factors suggests a general structure on which the variety of AIO's may be grouped, further analysis using the factor scores generated from this program will provide descriptors of various segments of the gasoline market. The inclusion of factor scores along with demographic variables to attempt to segment the gasoline market via discriminant analysis is discussed below. [For a related analysis including sources of information as predictor variables relating to energy usage, see Barnaby and Reizenstein, 1975]



Gasoline Usage Segment Description

The dependent variable under investigation in this study is reported gasoline consumption during a critical period of gasoline supply. Weekly gasoline usage data, collected on a series of "fuel gauge" scales, provided a measure of the distribution of gasoline consumption. In Table 4, frequency of gasoline usage per week is shown for a range of five gallon increments. More than two-thirds of the 922 respondents report using between 6 and 20 gallons of gasoline weekly, with a non-uniform distribution over this range. Approximately 18% of the sample used more than 25 gallons per week, while over 7% reported zero consumption.



For purposes of discriminant analysis, three consumption groups were arbitrarily formed to represent "light", "medium" and "heavy" gasoline consumption segments of the market. The three groups included consumption ranges of less than 10 gallons per week, 10-19 gallons per week and 20 or more gallons per week. As shown in Table 5, the groups are of unequal sizes with group 1 the smallest group, and group 2 the largest. Approximately a two-thirds sample of the original 922 respondents was selected for analysis. The remaining 307 members of the total sample were withheld from analysis and used as a hold-out sample to test the predictive ability of the discriminant functions. All groups showed better than 50% correct classification despite greatly unequal group sizes.

The BMD and Cooley and Lohnes [Cooley and Lohnes, 1971] multiple discriminant analysis programs were used to test the ability of a set of 24 variables (8 AIO variables computed from factor scores and 16 demographic variables) to discriminate among the three levels of gasoline consumption. The results of this analysis are presented in Table 5. Since the Cooley and Lohnes program is not a stepwise routine, the BM]) program was used initially to determine the relative importance of those predictor variables that maximally separate usage groups. Of course, the extent of any correlation among the AIO variables is minimized because of the use of factor scores to represent orthogonal combinations of the original AIO variables. However, the demographic variables include a considerable overlap of information. Therefore, those predictor variables entering the analysis first may mask the effects of other correlated variables and may impair statistical significance. Table 5 includes the four AIO's and the seven demographic variables that are found to be statistically significant at the .05 level in discriminating among the three gasoline usage groups. The 13 non-significant variables are included as a footnote to the table.



Those factors that individually provide statistical separation among the groups include: (1) "Pollution is of secondary importance"; (3) "I am generally pessimistic and skeptical about energy-pollution issues"; (7) "The energy crisis has a silver lining"; and (8) "Gasoline rationing may be necessary". In addition, (12) family size; (14) families with children 6-14; (17) families with working spouses; (19) length of residence; (21) age; (22) education; and (23) income are descriptors that provide group separation.

Moreover, Table 5 gives a relative indication of differences across groups for each variable by presenting significantly different group means. For instance, group 3 is composed of individuals with relatively higher incomes than group 2, which in turn includes members with average income levels greater than group 1. Conversely, group 1 sees a greater necessity for gasoline rationing than group 2 or group 3. While a variable by variable comparison of group differences is provided in Table 5, an alternative method of displaying this same data which enhances the total description of each group is presented in Figure 2.

The Cooley and Lohnes program calculates N-1 discriminant functions among N groups. Hence, in our three group study, two discriminant functions are calculated. By plotting the scaled eigenvectors, and the three group centroids of the two functions, a spatial representation of descriptors and group centroids is possible in two-space [Johnson, 1971 and Massey, 1965]. This configuration of descriptor variables as vectors and group centroids shows a "picture" of the gasoline market and aids description of each group via the set of significant variables determined through the discriminant analysis. For example, in Figure 2, group 1, which is the low consumption group, is quite isolated from the other two groups; all descriptor vectors with the exception of the factor, "Gasoline rationing may be necessary", "load" less on this group than on groups 2 and 3. Group 2 gasoline consumers can be characterized by the proximate descriptor vectors as older, longer length of residence, greater likelihood of a working spouse, pessimistic and skeptical, and see the energy crisis as providing a silver lining. The high consumption group, 3, can be described as more educated with more children 6-14, a higher income level, and more working members of the family who generally see pollution as of secondary importance.



Such descriptions of consumer groups with various levels of gasoline consumption may provide insights into the appropriate marketing strategies required to reduce the energy consumption of the medium and heavy usage groups. Of particular importance is the selection of descriptor variables that provide not only a basis for segment identification, but suggest informational and attitudinal idiosyncrasies of each segment to aid in the formulation of appropriate marketing programs in an attempt to alter traditional consumption patterns. A mechanism for differentially reaching various segments with unique characteristics may be derived from the inclusion of informational descriptors as suggested in [Barnaby and Reizenstein, 1975]. Obviously, any number of additional socioeconomic variables, such as the gasoline consumption constraint inherent in ownership of a large domestic automobile, will provide useful information in discriminating among gasoline user groups.


The discriminant analysis of the gasoline consumption groups provides a vehicle for identifying distinct characteristics associated with such segments. These characteristics may be used as a basis for understanding group composition and the formulation of strategies directed toward reduction of excessive consumption where appropriate and feasible.

Group 1, the low and non-usage segment (less than 10 gallons per week), possesses the strongest positive attitude toward gasoline rationing. These individuals would appear to be younger, to have fewer children in the household, to have fewer working spouses, to have lower family incomes, to be less educated, and to be shorter term residents of the community. Therefore, it would seem highly likely that these individuals are young singles or married couples in the early stages of the family life cycle. This group might include owners of smaller imported and subcompact automobiles as well as those who use mass transit and car pooling to meet their transportation needs. (This hypothesis will be tested in future analyses.) Since this group is presently a low consumption segment, a continuation strategy is deemed advisable for them. Advertising should emphasize the continued existence of energy problems, highlighting solutions in use by individuals whose characteristics match those of users and nonusers within this group. A conservation orientation should be featured, targeted toward both owners and non-owners of automobiles. Ads directed toward the former might highlight gasoline saving tips, while those directed toward the latter could cite availability, convenience, and cost saving aspects of mass transit and car or van pooling.

Group 2, the medium usage segment (10 to 19 gallons per week), seems to have the attitude that the energy crisis has a silver lining. This attitude is reinforced by their generally skeptical outlook, another evidence of which is expressed in their opinion that pollution is of secondary importance. This group is older, more educated, long term community residents, and reasonably affluent (in part due to a working spouse). They seem to be skeptical about environmental issues which are heavily promoted in media as being of national importance; they require concrete, irrefutable evidence before their skepticism can be overcome. Therefore, a data-based communication strategy is recommended to promote decreased gasoline consumption within this group. Documented evidence from sources recognized by the public as having a high degree of credibility, [The New York Times, The Washington Post, recognized authorities and certain civic and public service organizations are examples of highly credible information sources.] could be used as a focal point of advertising and publicity directed toward this segment. The composition of this segment must thus be analyzed prior to any such campaign to determine which information sources would be most credible to them.

Group 3, the heavy usage segment (greater than 19 gallons per week), is least inclined toward gasoline rationing and think that pollution is of secondary importance. They are characterized by a large number of people in the household emphasizing the 6 to 14 age range. They, therefore, fall into a "full nest" stage of the family life cycle, where the parents spend a great deal of time (and gasoline) acting as chauffeurs for the children. This group, being the highest income segment of the three (as well as the highest education segment), can afford to buy the additional gasoline required for such activities. Like group 2, this segment is pessimistic and skeptical about energy and pollution issues.

Due to the similarity of attitudes between 2 and 3, a data-based communication strategy is also deemed appropriate for the latter. The tactical thrust, however, should be oriented toward the future in light of the family structure of group 3, emphasizing the effects of possible gasoline shortages on the future welfare of their children. This approach must stress the credibility of data sources, a procedure similar to that suggested for group 2. Naturally, there may be differences in source credibility between segments. Other tactics to reduce the group 3 consumption level, might be promotion geared toward car pooling and more efficient trip routing as a means of saving gasoline while not radically altering current life style.

Finally, it should be noted that segments 2 and 3 represent 73% of the respondents analyzed, thus indicating that higher gasoline consumption was the rule rather than the exception even at the peak of the Arab oil embargo. In terms of cost-benefit analysis, therefore, the data-based communication strategy and supporting tactics should receive paramount attention. It is this strategy which would appear to have the greatest potential for reducing consumption in medium and high usage segments of the gasoline market.


David J. Barnaby and Richard C. Reizenstein, "Profiling the Energy Consumer: A Discriminant Analysis Approach," a paper presented at the National ORSA/TIMS Conference (Chicago: May, 1975).

William W. Cooley and Paul R. Lohnes, Multivariate Data Analysis. New York: John Wiley & Sons, 1971.

James F. Engel, David R. Kollat, and Roger D. Blackwell, Consumer Behavior, Second Edition. New York: Holt, Rinehart, & Winston, 1973.

Richard M. Johnson, "Market Segmentation: A Strategic Management Tool," Journal of Marketing Research, 8 (February, 1971), 13-19.

William F. Massey, "Discriminant Analysis of Audience Characteristics," Journal of Advertising Research, 5 (February, 1965), 39-40.

Glenn S. Omura and W. Wayne Talarzyk, "Relationships Between Consumers, Shopping and Leisure Activities and their Attitudes Toward the Energy Crisis: A Cross Sectional Study," a paper presented at the Association for Consumer Research Annual Conference (Chicago: October, 1974).

Opinion Research Corporation, Trends in Energy Consumption and Attitudes Toward the Energy Shortage, Volume V, Princeton, N.J.: Opinion Research Corp., 1974.

Richard C. Reizenstein and David J. Barnaby, "Assessing Methods of Increasing Response Rates in Mail Questionnaires,'' a paper presented at the National ORSA/TIMS Conference (Chicago: May, 1975).

W. Wayne Talarzyk and Glenn S. Omura, "Consumer Attitudes Toward and Receptions of the Energy Crisis," a paper presented at the American Marketing Association Fall Conference (Portland: August, 1974).