Advances in Consumer Research Volume 2, 1975 Pages 455-464
A SITUATIONAL MULTI-ATTRIBUTE ATTITUDE MODEL
Kenneth E. Miller, University of Utah
[The research reported in this paper was supported in part by the College of Administrative Science, The Ohio State University. The author is indebted to Professor James L. Ginter for his constructive guidance during the research.]
[Kenneth E. Miller is Assistant Professor Marketing, College of Business, University of Utah.]
Several researchers have discussed the promise of inclusion of situational variables in consumer research. Wicker (1969) states that measurement of attitudes and behavior should be carried out under similar situational conditions. This research investigates the variability of attitude scores when measured for differing situations and the resultant efficacy in the prediction of consumer preference and choice. The situational multi-attribute attitude model is utilized to predict situational preference and situational choice for each individual in the large mail panel. For these same individuals the nonsituational (traditional) model is utilized to predict nonsituational (overall) preference and nonsituational choice. The results from these sets of analyses are contrasted. The situational multi-attribute model, when used to predict situational choice, outperforms the traditional model, when used to predict overall choice. The success of situational multi-attribute model is necessarily product-specific, in this case the fast food hamburger market in Columbus, Ohio. However, the success of situational variables in improving prediction of brand choice for a specific product category certainly improves the promise of inclusion of these variables in consumer behavior.
A potentially profitable area for improving the ability of attitude toward a brand to predict choice of that brand over competing brands by a consumer is consideration of the situational variables which affect consumer behavior. Wicker (1969) states "a general postulate regarding situational influence on attitude-behavior relationships is the following: the more similar the situations in which verbal and overt behavioral responses are obtained, the stronger will be the attitude-behavior relationship."
There has been limited research reported in consumer behavior which considers the benefits derived from measuring attitude-behavior relationships along situational dimensions. Research indicates that situational variables have a significant influence on consumer affect towards a product category. Belk (1974) demonstrated that situational variables accounted for nearly half of the explained variance in both meat and snack preferences. Respondents preferred differing types of meat products in differing situations. Sandell (1968) found beverage preferences obtained for different situations, such as "for breakfast," "with lunch," "at a party," explained a larger proportion of total variance in reported behavior, than did favored brand. Thus, evidence exists which suggests that explicit consideration of situational variables will improve the predictive ability of attitude and preference measures in the prediction of consumer choice.
The objective of this research is to compare the advantage of disaggregate v, aggregate situational analysis of attitude-behavior relationships for each individual in a specific product setting. The performance of a situational (disaggregate) model in predicting situational preference and choice will be compared with an aggregate model in predicting overall preference and choice for brands within a product category. The attitude model used is the multi-attribute model which has been widely used in marketing in beth predictive and diagnostic applications (see Wilkie and Pessemier for an extensive review).
The general form of the model used in these studies is:
Aj = attitude toward brand j
Vi = affective importance of attribute i
Bij = perceived amount of attribute i contained by brand j
Ii = ideal amount of attribute i (measured or assumed)
n = number of attributes considered
k = parameter of the weighted Minkowski space which determines the measure of distance calculations.
The axes of the perceptual space are the designated attributes for the product category and the attitude measure is represented as a measure of distance between the location of the brand and the product category ideal point in the multidimensional space. This model does not envisage situation-specific attitude measurement and is considered to be an aggregate model.
Situational measurement can be incorporated into the multi-attribute attitude model in several ways. It is hypothesized that perceived attribute importance will vary across differing situations. Depending on the specific product category chosen and salient attributes used, the performance of any brand along certain attributes also will vary across differing situations. The form of situational multi-attribute attitude model used is:
s = situation (1,2---p)
p = number of situations considered
Ajs = attitude toward brand j in situation s
Vis = importance of attribute i in situation s
Bijs = perceived amount of attribute i contained by brand j in a situation s
Iis = ideal amount of attribute i in situation s
The sensitivity of Vis, Bijs and Iis across situations is dependent on product category and specific attributes considered.
H1: The situational model does not perform better than the nonsituational model in the prediction of situational and overall preference, respectively.
H2: The situational model does not perform better than the nonsituational model in the-prediction of situational and overall choices respectively.
The data were collected from a mail panel residing in Columbus, Ohio, Each respondent completed five questionnaires, over a period of three months, concerning eight Columbus fast food restaurant chains. The initial panel sample of 744 was generated from a random list of names in the Columbus telephone directory. [The final panel size, reduced by experimental mortality, was 446.] During the three-month period an extensive advertising and couponing campaign was undertaken by one of these fast food restaurants.
Data collection was comprised of three stages:
1) Group interviews were conducted to isolate major situational usage patterns and salient product attributes;
2) A sample of 50 householders and 100 students were asked to indicate the frequency of visit(s) to fast food restaurants under each of the multitude of situations generated from (1);
3) Data were collected from the mail panel on the ratings of eight fast food restaurants on seven attributes (taste of food, speed of service, popularity with children, price, variety on menu, cleanliness, and convenience). The only brand rating obtained across situations was convenience. In addition, subjects were asked to rank preferences for the eight restaurants across situations and to rate the importance of each attribute for each situation. Choice information was also obtained from the panel. Thus, attribute importance, preference and choice as well as brand ratings on convenience, were obtained for all four situations considered.
The four situations chosen for detailed examination were selected with two criteria in mind. The situation should be encountered frequently so that adequate sample sizes could be obtained for each situation. Also, the situations should include different dimensions of situation. Belk (1974b) discusses five dimensions of a situation, being physical surroundings, time frame, interpersonal surroundings, mood and goal direction.
The four situations selected for detailed analysis were:
to eat at lunchtime on a weekday
to eat a snack after a shopping trip
to eat when rushed for time
to eat with the family when not rushed for time.
Lunch on a weekday evokes spatial and temporal dimensions for a major proportion of the panel. [E.g., those who were employed away from their place of residence.] Eating after a shopping trip evokes spatial and perhaps mood dimensions. Eating with the family when not rushed for time captures a consistent contrast in interpersonal surroundings. Eating when rushed for time encompasses time frame dimensions. Darley and Batson (1973) in their "Good Samaritan" experiment indicate that degree of haste in one's journey is a good situational predictor of helping behavior. The consumer's decision, as to which fast food restaurant to frequent, may be influenced by the amount of time he has available.
The respondents were also asked to complete an extensive AIO bank and media habit items. At the conclusion of the study, respondents were asked to give their opinions as to the nature of the study.
In order to ascertain the value of introducing the situational variables into the attitude-preference and attitude-behavior relationships, two sets of analyses were carried out for four time periods.
For each individual who indicated frequenting fast food restaurants in at least two situations, the traditional model (model I) is contrasted to the predictive power of the situational model (model II). For each model, the seven attributes [The popularity with children attribute was not applied to all respondents.] were applied to each subject to compute an attitude score. An assumed ideal was used in both models (more of any attribute is better). The ranked attitude scores for models I and II were correlated with rank order overall and situational preference, respectively, using Spearman's Rho to determine the degree of association. This analysis is presented in Figure 1.
The mean Spearman's Rho across individuals for each situation at time periods 2, 3, 4 and f are presented in Table 1. The Wilcoxin matched sign test indicated that the magnitude of correlations were significantly different in only one out of 16 cases.
A similar procedure was undertaken in the prediction of consumer choice. Following models I and II it is hypothesized that the brand closest to the product category ideal point will be the brand that is chosen. Situational and nonsituational attitude scores were computed and ranked. The brand closest to the assumed ideal point was ascertained and compared to the brand which was actually chosen by the respondent. The confusion matrix of brands predicted from the nonsituational attitude scores at period 1, given brand chosen from period 1 to period 2, is shown in Table 2.
The confusion matrix of brands predicted from the situational attitude scores at period 1, given brand chosen from period 1 to period 2, is shown in Table 3. This matrix is an aggregation of confusion matrices for all four situations.
AGGREGATE MODEL - PREFERENCE CORRELATION
SITUATIONAL MODEL - PREFERENCE CORRELATION
COMPARISON OF MODEL I AND MODEL II IN THE PREDICTION OF CONSUMER PREFERENCE
CONFUSION MATRIX OF BRAND PREDICTED FROM THE NONSITUATIONAL ATTITUDE SCORES AT PERIOD 1 GIVEN BRAND CHOSEN DURING PERIOD 1-2
CONFUSION MATRIX OF BRAND PREDICTED FROM THE SITUATIONAL ATTITUDE SCORES AT PERIOD 1 GIVEN BRAND CHOSEN DURING PERIOD 1-2
For each situational choice, the two models were used to predict this choice occasion, i.e., each choice was predicted separately. A matrix of brands predicted and brands chosen was constructed for each situation and the percentages of correct predictions are outlined in Table 4.
COMPARISON OF MODEL I AND MODEL II IN THE PREDICTION OF CONSUMER CHOICE
For the 20 comparisons of the model, the situational model outperformed the nonsituational model 19 out of 20 times. The Wilcoxin matched pairs test indicated that the magnitude of differences in choice prediction between the two-model forms was statistically significant (p < .001).
The results of the analysis illustrate that the disaggregate situational analysis is of value in improving the prediction of consumer choice, but is not superior to the aggregate model in the prediction of preference.
Several explanations may be advanced as to why the disaggregate analysis did not prove superior in the prediction of preference, Situational influence has been treated by certain consumer behavior theorists as mediating behavioral intention and choice. [E.g., Sheth (1969).] It could be hypothesized that, for a single product category, preference judgments are without situational variance but choice decisions were influenced by predictable situations. The use of the rank ordered preference measure leads to the use of less powerful nonparametric statistical tests. If situation has a relevant but small influence on preference more powerful parametric statistical tests may be required to isolate the effect. Use of more rich measurement of preference will alleviate the problem. [Pessemier, et al. (1971) present a dollar-metric preference measure which is interval scaled.] However, consistent with the theoretical proposals for situation-specific attitude and behavior measurement, situational attitude yields significantly improved prediction of consumer choice. The value of inclusion of situational variables is necessarily product-specific. Various factors favor the use of situation variables in the fast food hamburger restaurant market. The inclusion of attribute weights in the multi-attribute attitude model allow for variation of the number and nature of the salient attribute set across differing purchase situations. Also, the convenience of each restaurant varied with the spatial dimension of situation. These characteristics of the fast food consumer, in addition to the Columbus fast food restaurants exhibiting some degree of benefit segmentation, help explain why the inclusion of situation variables improves the relationship between attitude and behavior. Certainly the success of inclusion of situational variables in attitude measurement for brands in a product category, where attitudes, preferences and consumer choice are situation-specific, has ramifications for the consumer researcher.
Refinements to this research include examination of preference measurements to isolate situational variance. The power of situational rank order preference to predict choice should be contrasted to the predictive power of the overall preference measure. Analysis of variance should be conducted on attribute importance weights to ascertain situational variance. Examination of brand images along the specific situations considered in the research will give insight into the situational variation in attitudes for the brands. In addition to refinement and diagnosis of the research results presented here, the dimensions of situation need to be refined so that the study of consumers in predictable situational environments can be pursued.
Belk, R. W. An exploratory assessment of situational effects in buyer behavior. Journal of Marketing Research, 1974, 11, 156-163.
Belk, R. W. Situational influence in consumer behavior. Faculty Working Paper No. 195, College of Commerce and Business Administration. University of Illinois at Urbana-Champaign, 1974.
Darley, J. M. & Batson, C. D. From Jerusalem to Jericho, a study of situational and dispositional variables in helping behavior. Journal of Personality and Social Psychology, 19 73, 27, 100-108.
Fishbein, M. Attitude and prediction of behavior. In M. Fishbein, Readings in attitude theory and measurement. New York: John Wiley and Sons, Inc. 1967, 479-92.
Ginter, J. L. A study of attitude change and choice behavior during new product introduction in a laboratory setting. Unpublished doctoral dissertations Purdue University, 1972.
Ginter, J. L. An experimental investigation of attitude change and choice of a new brand. Journal of Marketing Research, 1974, 11, 30-40.
Pessemier, E. A., Burgers P., Teach, R. & Tigert, D. Using laboratory brand preference scales to predict consumer brand choice. Management Science, 1971, 17, B371-B385.
Sandell, R. G. Effects of attitudinal and situational factors on reported choice behavior. Journal of Marketing Research, 1968, 5, 405-08.
Sheth, J. N. Attitude as a function of evaluative beliefs. Paper presented at the American Marketing Association Workshop in Consumer Behavior, 1969.
Wicker, A. W. Attitudes versus actions: the relationships of verbal and overt behavioral responses to attitude objects. Journal of Social Issues, 1969, 25, 41-78.
Wilkie, W. L. & Pessemier, E. A. Issues in marketing's use of multi-attribute attitude models. Journal of Marketing Research, 1973, 10, 428-41.