Attitudes and Values As Predictors of Energy Information Behavior Patterns

ABSTRACT - The focus of research lies on the investigation of the impact of informational marketing measures for energy conservation. In a reanalysis of representative consumer data two causal models are tested integrating values. demographics, attitudinal and behavioral variables. The analysis is carried out by applying the Lisrel approach.


Lutz Hildebrandt (1984) ,"Attitudes and Values As Predictors of Energy Information Behavior Patterns", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 574-578.

Advances in Consumer Research Volume 11, 1984      Pages 574-578


Lutz Hildebrandt, Science Center Berlin, International Institute for Environment and Society


The focus of research lies on the investigation of the impact of informational marketing measures for energy conservation. In a reanalysis of representative consumer data two causal models are tested integrating values. demographics, attitudinal and behavioral variables. The analysis is carried out by applying the Lisrel approach.


During the last decade, energy conservation has been a major issue in economic and environmental policy. - After the second energy crisis in the late seventies, the German government put forward a comprehensive program intended to stimulate energy conservation in the residential sector. Roughly 40 percent of the total amount of energy consumption can be attributed to this sector (see Dritte Fortschreibung des Energieprogramms der Bundesregierung 1981). An important part of the program is a set of social and informational marketing measures for stimulating energy conservation through insulation and the substitution of oil. The government uses three channels to communicate with the consumers, i.e., government brochures, product-testing organizations (e.g. Stiftung Warentest) and a consumer advisory service via consumer organizations.

Little is known about the effects of these information programs. Recent evaluations (see Ifo Institute 1982) estimate a total energy conservation effect of .9 percent as a result of information marketing. However, there are scarcely reliable data on impacts and household acceptance of the program.

This paper attempts to clarify the determinants and their causal effects on energy related information behavior, assuming that the use of different information sources will lead to different energy behaviors. A former content analytic study of energy conservation programs (Hildebrandt/Joerges 1982) shows that there is a strong relation between the information sources and the conservation actions relevant to households. Consumer organizations generally concentrate on conservation, whereas the primary aim of business and utilities is the substitution of oil by other energy sources,e.g. electricity. Based on literature studies and other information sources hypotheses are derived and integrated into a causal model


The first general hypothesis is founded on information from the 'test' Institute about the impacts of the conservation-related articles and reports in the periodical 'test'. It is assumed that the reception of 'test' will result in a favorable attitude towards energy conservation as well as r energy conserving actions, which on the other hand causes an increase in the demand of energy conserving equipment and an additional search for information for pre-and post-decisional situations.

This assumption is borne out by the normative appeal of 'test' magazine and the high credibility of 'test' information (Raffee/Silberer 1981). There are a number of theoretical studies which support the hypotheses, although some of .ne relations are rather controversial. Normative appeal, for instance, has been shown to have little effect or energy conservation, if the financial effect remains unclear (Heberlein 1975). There is no direct financial effect of the 'test' reception, but cost-benefit and quality considerations of energy conserving investigations are usually the major theme of articles.

Up till now, attitudinal variables have contributed little to the explanation of energy conservation behavior. Exceptions in this respect are the investigations by Seligman, Darley and Becker (1978) and Seligman, Kriss, Darley, Fazio, Becker and Pryart (1979) who found attitudinal factors regarding energy related problems (in summer) which explain more than 50 percent of the variance in energy use. But in a replicatory study carried out in winter, only 18 percent of energy use could be explained (see Becker, Seligman, Fazio and Darley (1980). Verhallen and van Raaij (1981) also found weak relations between energy attitudes and actual behavior.

The failure of predictive power is, among other reasons, attributed to methodological problems. The most criticized points are a lack of specificity between attitudinal and behavioral variables (Heberlein/Black 1976; Fishbein/Ajzen 1977), and the use of self-report measures instead of actual observed behavior (McDougall, Claxton, Ritchie and Anderson 1981). A general agreement exists in regarding attitude as an important mediator variable between contextual variables such as family income, family size, and climate which have been shown to be significant predictors of energy related behavior (Black 1978; Fritzsche 1981; Olsen 1981). Following the study of Stern, Black and Ellworth (1982) for the second hypothesis, it is assumed that a causal chain exists between demographic status and income and energy-related behavior, which is mediated by attitudinal variables. The causation works from general contextual variables through behavior-specific cognitive variables to conservation-related behavior.

The third set of hypotheses concerns personal variables which are also assumed to be relevant for energy-related behavior patterns, and are working on the same level as other structural contextual variables. Different studies have shown that there is a kind of general association between personal values and conservation, although the strength of the relationship varies across the studies,e.g. Rankin (1981); Fietkau, Kessel, Tischler (1980). Generally it is assumed that the value system effects beliefs and attitudes (Howard 1977, p. 99; Rokeach 1973). which on the other hand determines the behavior. Therefore, the causal impact on a behavioral variable will usually be low, but much stronger on an attitudinal variable.

Testing the effect of contextual variables on the value-behavioral-relationship, Newman (1982) found that while contextual attributes exert a clear impact on behavioral efforts to conserve, there was no proof of the systematic influence of personal values mediated by contextual variables. The impact of one value orientation will be assessed in relation to a contrasting value orientation - in this cas consumption orientation versus ecological orientation.

The hypothesized sets of causal impacts and their signs can now be summarized:

H1 : The more intense to reception of the 'test' magazine, the more favorable are the conservation attitude and the conservation related behavior.

H2 : The impact of structural context variables on energy. related variables decreases according to the level of specificity of these variables.

H3 : Ecological value orientations have a significant impact on energy related variables, whereas the consumption value has none



Past investigations have shown that readers of 'test' can be characterized as well educated members of the upper middle class who are also particularly busy consumers. It is, therefore, expected that there is an effect of the "consumption value" on the reception of 'test'. The model of causal hypothesized relations is shown in Figure 1 above.

The Data

The data were collected at random in a representative survey of 1000 individuals from West Germany and West Berlin in 1980. To test our hypotheses, a subsample of heads of households was drawn (N=558). Original aim of the survey was an evaluation of the policy of the "Stiftung Warentest" and to provide empirical information for future product-test planning. A large section of the questionnaire was concerned with media behavior and interest in energy saving products and services.

Most of the questions referring to energy related behavior were simple binaries or three point scales, although underlying variables were continuous, some questions have been recoded to obtain ordinal or quasi-interval data. For the exogenous variables occupation and education are indices of the social status, reflecting the position of the household' heads. Income, on the other hand, represents an estimation of the total household income after tax. The scorings of the value items were collected on rating scales which have proved to be an alternative to the originally proposed ranking method (see Munson/McIntyre 1979; Rankin/ Grube 1980). The items a "comfortable life" were selected for the consumption orientation, and a"world of peace" and a "world of beauty" for ecological concern. The relation of these values to consumption and ecological orientation has been Proved (Hildebrandt 1983; Newman 1982).

Past 'test' reception behavior serves to measure the regularity of the reception, i.e. whether the periodical is purchased on a monthly or sporadic basis. A general energy conservation attitude variable is represented by a question on the importance of energy conserving equipment and services in one's own home, 'product interest' in consumer research will usually be regarded as an indicator for communication arising from a relevant product (see Kroeber-Riel 1982, p. 480).

Indices of the preferred information sources are formed by adding binaries about the actual use of an information source before. or coinciding with an energy relevant purchasing decision. Three variables measure the preference of (a) consumer information sources (neutral product testing, consumer advisory boards) (b) business information sources (shopping advice, business leaflets, folders) and (c) other information sources, e.g. radio and TV reports. Private communication with friends and relatives is not included. an overview on the scales of the variables is given in Table 1, next page.

In a reanalysis of data collected with another aim in view, the semantic content of the variables must be examined very carefully. Neither group of variables bears a direct relation to energy conservation, all are indirectly related. Demographics for instance have been shown to have an impact But the same also can be said of the variable representing past reception of 'test' . This variable, however, us related to the object 'test' and not to energy conservation. More specific for energy conservation is the general attitude towards it; it seems to be on the same level as product interest. The use of information source is related to a special medium as to the content of the "energy conservation" information too.




To test the hypotheses the LISREL V program is used (Joreskog/Sorbom 1981). Marketing and consumer researchers paid special attention to this approach. The LISREL model consists of an integration of confirmatory factor models and econometric (path) models. The model structure explicitly distinguishes between hypothetical constructs and measured variables, causal and measurement relations. Under the assumption of linearity and multinormality of the data, a maximum likelihood procedure for estimation is available (Bagozzi 1980; Fornell/Larcker 1981; Hildebrandt 1983).

The equations of the factor model and causal relations model in LISREL V are:

x = Lx x + d  (1) 

y = Lyh + e  (2)

for the measurement part, and

h = Bh + G x + z   (3)

for the causal relations part, where B and G are coefficient matrices for the construct relations, z is a vector of residuals in the equations. The vector of constructs h and x is related to the vectors of the observed variables by two factor models (1), (2) with Lx, Ly the matrices of factor leadings and d, e the errors of measurement, (for further details see Bagozzi 1980). f, qd, qe, Y are covariance matrices of x, d, e, z, respectively.

In the case of multiple measurements for the construct,the observed variables can be factorized and the observed variances separated in variance due to the construct and variance due to measurement error. The confirmation of a model by a Chi2-test can be interpreted as a proof for validity.

In the LISREL V software some new features are integrated which are appropriate for this data analysis. Most of the questions and indices in our study measure the underlying construct on a few scale steps. Although regarded as being normally distributed, some of the observed variables exhibited a moderate degree of skewness. In this case Joreskog/ Sorbom (1981, p.IV:1) propose not to use standard errors and Chi2-tests to evaluate the model, but instead of that to use available fit indices. If this is unsatisfying, it is also possible to calculate polychoric correlations and to use an Unweighted Least Square Estimation (ULS). In our case, we still use the maximum likelihood procedure, having in mind that the T values are to be interpreted cautiously.

Three kinds of fit-indices are used in addition to evaluate the models: (1) the Goodness of Fit-Index (GFI),which is a measure of the relative amount of variances and covariances jointly accounted for by the model. (2) this measure adjusted by degrees of freedom (AGFI), and (3) the Root Mean Square Residual (RMR), which is a measure of the average residual variances and covariances. GFI and AGFI should be close to one, RMR close to zero. The coefficient of determination (R2) is also used for evaluation of the models.

Specification and Test of the Model

The model testing is carried out in two stages. In a submodel, an attempt will be made to confirm hypothesis one,concerning the impacts of the 'test' reception on energy conservation related constructs.Subsequently the personal and structural background variables will be integrated to test the sequence of effects and the impact of value structures. Because of the hypothesized complex impact of the 'test' reception on energy related attitudes and behavior in the model, it is assumed that there are direct and indirect effects of h1, h2 and h3 on energy related information behavior. The model is represented in Figure 2.



h1, h2, h4, h5, h6 are each measured by one observable.The measurement paths are set to 1. The variable is assumed to have been measured without error. Because the model is fully recursive and correlated errors are not specified, no identification problem arises. For purposes of interpretation,only causal paths larger than 0.10 are marked with a coefficient,which means that at least one percent of variance is explained in the determinated variable. Examining the standard errors all interpreted path coefficients were also significant on the 5% level.

The model does not fit using the Chi -test in evaluation, but all other fit-indices are sufficient in size. Especially the RMR measure indicates a good fit,there is virtually no residual variance. On the other hand the predictive power is low.The model explains about 14 percent of the total variance in the endogenous variables. But 10 percent of the variance is explained solely in the product interest and the use of consumer information (.12; .19,respectively) The measurement paths are considerably high in the interest variables, all loadings are larger than .8. Table 2 presents all fit-indices.



The complete model is tested step by step, first omitting paths which are close to zero (.005) in the original model. The full model of Figure 2 is integrated in the structure to ascertain whether there are changes in the coefficients due to effects of the intervening context variables. The exogenous variables are assumed to be correlated. Education/occupation and the values "a world of beauty" and "a world of peace" are each treated as measures of one construct respectively.



The errors in the relations to predict the information construct are allowed to be correlated too.This was also the case in Model 2' Inspecting the fit-indices of the complete model, the Chi is insufficient, but the goodness of fit-indices GFI and AGFI show borderline coefficients. AGFI is close below .90, and GFI is close below .95. Both values are mentioned in the literature as lower limits. Because RMR .024 was still very low and 22 has a large value (.34), the model is regarded as fitting.

Discussion of the Results

The first model provides some evidence for the hypothesis that the reception of 'test' (n1) leads to a favorable attitude towards energy conservation (n2) B21 = .19. Although the relation is not strong, 'test' reception is one of the reasons for this attitude. Attitude itself has a much higher causal impact on the interest in energy conservation (n3) equipment B32 = 44 The direct effect of 'test' reception on pro duct interest is not worth mentioning. Also the indirect effect, mediated by the attitude is very low (B21 B32 = .08). Therefore, we can state that there is an impact but it is hard to quantify. On the other hand, the interest in energy conservation equipment leads to a negative effect as regards using business information. The negative sign means that the higher the product interest is, the lower is the use of business information in an energy related purchasing decision.

The strong causal effect of 'test' reception (n1) on the use of energy related consumer information (n5) does not appear to be a particularly interesting result. However, it implies that the reception of 'test' leads to a search for and use of information from other consumer information sources related to energy conservation.

The result of the test is not spectacular, but it is somewhat surprising that a negative effect of product interest on business information can be observed.

All paths in the total model explained explicitly show sufficient T values > 2.0 In an overall evaluation it can be stated that the sequence of effects caused by structural and personal background variables is mediated by the attitudinal construct (n2). This holds for income (E2) Y22 = .16 and for the ecological value orientation (E4) which has a considerable effect on attitude Y24 = .25 . These two effects led to a causal path which was below .10 in the submodel, but now has a causal impact on the use of other information (n6) with B62 = .15.

The model also shows that the impact of 'test' reception on energy related attitudes and behavior is strongly influenced by the educational status (E1) of the consumers as a background variable Y11 = .34 . This result is in accordance with former empirical results of the demographic profile of the 'test' readers (see Silberer 1979, p. 180). The direct impact of status on the use of consumer energy information is also confirmed by Y51 = .18, the coefficient for the direct effect of past 'test' reception is B51 = .24 on the use of energy related consumer information.

As an indicator for the validity of the value constructs, the causal path of (E3) the consumption value and (E4) the ecological values are to be evaluated. Consumers with a strong consumption orientation measured by the Rokeach value, have a negative influence on the interest in energy products Y33 = -.18 . On the other hand, the ecological value orientation has a negative impact on the use of business information (Y44) which means that people who are ecologists do not like using business information for energy investigations in the household.

The assumption that consumption value is also related to 'test' reception, could not be confirmed by the model.


The relations between context variables, attitudinal variables and concrete behavior, especially in the field of energy, are extremely difficult to analyses due to observation problems of actual behavior. Therefore, a re-analysis of cross-sectional data can provide only weak results. Using our results, the activities of the "Stiftung Warentest" are to be regarded as valuable. From a public policies perspective the informational marketing of the "Stiftung Warentest" influences an important mediator variable which in turn is the cause for the relevant behavior: energy conservation. The total causal models show that attitude mediates for instance the impact of the effect of income. In contrast to structural context variables attitudes are changeable.


Ajzen, Icek and Martin Fishbein (1977), "Attitude Behavior Relations: A Theoretical Analysis and Review of Empirical Research", Psychological Bulletin 84, 888-918.

Bagozzi, Richard P. (1980), Causal Models in Marketing, New York, Wiley & Sons.

Becker, Lawrence, Clive Seligman, Russell H. Fazio and John M. Darley (1980), "Relationship between Attitudes and Residential Energy Use", Environment and Behavior.

Black, J. Stanley (1978), Attitudinal, Normative, and Economic Factors in Early Response to an Energy Use Field Experiment, University of Wisconsin, Department of Sociology, Unpublished Doctoral Dissertation.

Dougall, Gordon H.G., John D. Claxton, Bent J.R. Ritchie, and Dennis C. Anderson, "Consumer Energy Research: A Review", Journal of Consumer Research, 8, 343-354.

Fietkau, Hans J., Hans Kessel and Wolfgang Tischler (1982), Unwelt im Spiegel der offentlichen Meinung, New York, Frankfurt, Camsus.

Fornell, Claes and David F. Larcker (1981), "Evaluating Structural Equation Models with Unobservable Variables and Measurement Error", Journal of Marketing Research, XVIII, 35-50.

Fritzsche, David J. (1981), "An Analysis of Energy Consumption Patterns by Stage of Family Life Cycle", Journal of Marketing Research, XVIII, 227-232.

Heberlein, Thomas A. (1975), "Conservation Information: The Energy Crisis and Electricity Consumption in an Apartment Complex", Energy Systems and Policy, 1, 2, 105-118.

Hildebrandt, Lutz (1983), Konfirmatorische Analysen von Modellen des Konsumentenverhaltens, Berlin, Duncker & Humblot.

Hildebrandt, Lutz and Bernward Joerges (1983), "Consumer Energy Policies in the Federal Republic of Germany", Working Paper, Internationales Institut fur Umwelt und Gesellschaft. Wissenschaftszentrum Berlin.

Howard, John A. (1977), Consumer Behavior: Application of Theory, New York.

Joreskog, Karl G. and Dag Sorbom (1981), "LISREL V - Analysis of Linear Structural Relationships by Maximum Likelihood and Least Squares Methods", Research Report 81-8, University of Uppsala, Department of Statistics.

Joreskog, Karl G. and Dag Sorbom (1982), "Recent Developments in Structural Equation Modelling", Journal of Marketing Research, XIX, 404-416.

Kroeber-Riel, Werner (1980), Konsumentenverhalten, Munchen, Vahlen

Newman, Keith A. (1982), "Human Values: Do they make a Difference in Individuals' Commitment to Energy Conservation?" University of California (USA). Paper presented at the International Conference on Consumer Behavior and Energy Policy, Noordwijkerhout (NL).

Olsen, Marvin A. (1981), "Consumer Attitudes toward Energy Conservation", Journal of Social Issues, 37, 2, 108-131.

Raffee, Hans and Gunter Silberer (eds.) (1981), Informationsverhalten des Konsumenten - Ergebnisse empirischer Studien -. Wiesbaden.

Rankin, W. and J. Grube (1980), "A Comparison of Ranking and Rating Procedures for Value System Management", Journal of Social Psychology, 10, 233-246.

Rokeach, Milton (1973), The Nature of Human Values, New York, Free-Press.

Scholz, Lothar and Peter Ramner (1982), "Quantitative Wirkungen der Energiesparpolitik in der Bundesrepublik Deutschland", Ifo-Studien zur Energiewirtschaft, 3/2.

Seligman, Clive, John M. Darley and Lawrence J. Beeker (19?8), "Behavioral Approaches to Residential Energy Conservation", in: R. H. Sokolow (ed. ), Saving Energy in the Home: Princeton's Experiments at Twin Rivers, 231-254.

Seligman,Clive, Mitchell Kriss, John M. Darley, Russell H. Fazio, Lawrence J. Becker and John B. Pryart (1979), "Predicting Residential Energy Consumption from Home Owners' Attitudes", Journal of Applied Psychology, 9, 70-90.

Silberer, Gunter (1979), Warentest, Informationsmarketing, Verbraucherverhalten. Berlin.

Stern, Paul C., J. Stanley Black and Julie T. Ellworth (1982), "Personal and Contextual Influences on Household Energy Adaptions", National Academy of Sciences, University of Illinois (USA). Paper presented at the International Conference on Consumer Behavior and Energy Policy, Noordwiikerhout (NL).

Verhallen, Theo M. M. and W. Fred van Raaij (1981), "Household Behavior and the Use of Natural Gas for Home Heating", Journal of Consumer Research, 8, 253-257.

Dritte Fortschreibung des Energieprogramms der Bundesreaieruna (1981. Drucksache 9/983 Bonn.



Lutz Hildebrandt, Science Center Berlin, International Institute for Environment and Society


NA - Advances in Consumer Research Volume 11 | 1984

Share Proceeding

Featured papers

See More


F13. A Story of Waste: Trust, Symbolic Adoption & Sustainable Disposal

Marwa Gad Mohsen, Babson College, USA

Read More


D6. How to Boast Appropriately When Word of Mouth Flows Internationally?

Xingyu Wang, Huazhong University of Science and Technology, China
Yaping Chang, Huazhong University of Science and Technology, China
Jun Yan, Huazhong University of Science and Technology, China

Read More


A Salience Theory of Three Novel Exposure Effects

Kellen Mrkva, Columbia University, USA
Leaf Van Boven, University of Colorado, USA

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.