A Multivariate Analysis of Prepurchase Deliberation and External Search Behavior

Citation:

Stephen J. Miller and William G. Zikmund (1975) ,"A Multivariate Analysis of Prepurchase Deliberation and External Search Behavior", in NA - Advances in Consumer Research Volume 02, eds. Mary Jane Schlinger, Ann Abor, MI : Association for Consumer Research, Pages: 187-196.

Advances in Consumer Research Volume 2, 1975      Pages 187-196

A MULTIVARIATE ANALYSIS OF PREPURCHASE DELIBERATION AND EXTERNAL SEARCH BEHAVIOR

Stephen J. Miller, Oklahoma State University

William G. Zikmund, Oklahoma State University

[The authors wish to thank the Oklahoma Publishing Company for providing the survey data for this study.]

[Stephen J. Miller is an Associate Professor of Marketing and William G. Zikmund is an Assistant Professor of Marketing at Oklahoma State University.]

Several studies have indicated that economic and demographic characteristics such as age, income, education and family life cycle, relate to household shopping behavior. Based on a major survey of household grocery shopping behavior, this study offers further support to such evidence. The study indicates that deliberation and search are not simple univariate phenomena but should be examined in a multivariate framework. Canonical analysis was utilized to test this multivariate framework. The analysis yielded profiles of households associated with a variety of deliberation and search activities.

Information evaluation and search activities in grocery shopping behavior is a topic of considerable interest to the public policy maker as well as the marketing manager. For example, consumer advocates recommend that disadvantaged consumers need to receive more and better product information. Marketing managers need to know the best information medium for each market segment. While many psychological factors, such as perceived risk and personality, are considered to be determinants of search behavior, research efforts show socioeconomic variables are also determinant factors.

Thorelli (1971) has observed, "There is no reason to assume that product information is either homogeneously or normally distributed among the population." Several research studies indicate that individuals are more responsive to information sources they perceive to be compatible with their own personal characteristics, such as family life cycle or social class (e.g., Engel, Kollat, and Blackwell, 1973; Katona and Mueller, 1955; Levy, 1969; Thorelli, 1971). One might expect upper middle class women, married less than ten years, to organize shopping more purposefully (less impulsively) and more efficiently than lower class women (e.g., see Kollat and Willet, 1967; Levy, 1969). Or increases in multi-store food shopping might be expected to correlate with multiple-automobile availability or education (Prasad, 1972).

While the evidence suggests that socioeconomic factors or personal characteristic factors do appear to influence the propensity to search, it has been pointed out that there is a lack of studies investigating a combination of variables (Engel, Kollat, and Blackwell, 1973, p. 383). For instance, the influence of income or education on search behavior might vary at different stages of the family life cycle. Furthermore, utilization of one source of information may be correlated with the use of other sources of information (Katona and Mueller, 1955). Thus, researchers investigating personal characteristics, and their relationship to shopping deliberation and information utilization, should consider the interaction effects within both sets of variables.

This research extends the inquiry into the nature of the relationship between prepurchase deliberation and search behavior and personal characteristics, by utilizing multivariate analysis. Prepurchase deliberation and search behavior might include a number of activities such as frequent shopping trips, the reading of newspaper advertising, shopping list preparations and multi-store shopping. The view is taken in this study that the activities represent a total process of behavior rather than simply a number of discrete and independent variables. Likewise, socioeconomic factors or personal characteristics that identify segments differing by shopping process need not be examined separately or used to form a global index of social class, life cycle or the like. Investigating these variables, with their potential interactions as process and descriptor groupings in mind, should add to the understanding of the determinants of various search behavior.

The research reported in this study sought to answer the following question:

Are there profiles of personal characteristics of households that significantly relate to various prepurchase deliberation and search behavior processes?

The area of household grocery shopping was chosen as the focal point for investigation. This shopping activity has a variety of characteristics that made it amenable to the study. First, it is an activity in which all households participate. Second, the food budget represents a sizeable percent of the household expenditures which makes extensive shopping behavior an economically rewarding task. Finally, the food industry provides a large amount of marketing-dominated information.

The complete list of variables identified for the study is given in Table I. A variety of planning and search process variables reflect the criterion set. The predictor set of variables are those frequently examined individually as determinants of shopping behavior.

The data for the study was gathered by The Daily Oklahoman and the Oklahoma City Times newspapers in their 1972 consumer audit of the Oklahoma City SMSA. The sample included-2020 households utilizing an area-probability sample for the metropolitan area. Within each sample household, an in-home personal interview using professional interviewers was conducted with the "chief marketer," usually the housewife. Questions included brand purchase behavior, shopping behavior, media exposure characteristics, and household personal characteristics

In the research, bivariate correlation is first used to examine simple relationships between the criterion and predictor variables. Then canonical analysis is employed to isolate profiles of personal characteristics and shopping behavior patterns. A canonical correlation maximally correlates two linear additive sets of variables measured across a group. For a more complete discussion of this statistical technique, see Alpert and Peterson (1972).

THE BIVARIATE RESULTS

A variety of studies have been cited earlier in the paper that dealt with an explanation of prepurchase deliberation and search behavior components (shopping behavior) based on personal characteristics. A bivariate correlation matrix for the shopping and personal characteristic variables is given in Table II. Many of the correlations between shopping and personal characteristic variables were statistically significant. However, the correlations were quite low with the highest being .20 for number of shopping trips and number of cars in family. Thus, only four percent of the variation in the shopping variable can be explained by transportation availability.

TABLE I

RESEARCH VARIABLES

TABLE II

CORRELATION MATRIX  -  SHOPPING - PERSONAL CHARACTERISTICS

Another striking feature of the correlation matrix is the apparent low intercorrelation among the shopping variables. The highest correlation is .21 for a variety of stores shopped and number of trips. This moderate interaction indicates that a number of separate dimensions of the planning-search process of shopping are represented by the data.

A canonical analysis of the data should cast light on the various shopping patterns exhibited by households and the characteristics descriptive of such households. While bivariate correlations revealed few shopping-personal characteristic relationships, canonical analysis might indicate combinations of variables of the criterion and predictor sets that interrelate. Market segments based on shopping process variables will have been defined and identified.

THE CANONICAL ANALYSIS

The canonical analysis included the seven criterion variables and fourteen predictor variables as listed in Table I. Of the twenty-one variables, five "point" measures are included (list usage, home ownership, household head status, marital status, and race). The assumptions of canonical analysis include interval scaled data. However, as with similar multivariate techniques, the assumption can be relaxed with little damage to the analysis (Peterson and Alpert, 1972).

Seven canonical roots were generated by the analysis, one for each criterion variable. A summary of these are given in Table III along with relevant statistical tests of significances. For example, the first canonical R of .41 indicated that 16.8 percent of the variation in the linear combination of shopping variables could be explained by the linear combination of personal variables in the predictor set. Bartlett's "chi-square test" was used to test the statistical significance of each canonical correlation coefficient in the table. In all, five of the seven coefficients were significant at the .05 level. Thus, it appears that shopping patterns do relate to personal characteristic variables.

TABLE III

CANONICAL CORRELATION ROOTS

The next step in the analysis was an examination of each statistically significant root to ascertain shopping patterns that emerged and the various personal characteristic profiles that relate to each pattern. As with factor analysis, the significant relationships are not always readily interpretable However, potentially useful inferences can be drawn from the interpretation although they must be treated as exploratory. One could examine the standardized canonical coefficients for both the criterion and predictor variables to identify salient variables from each set. However, this approach to analysis can be misleading if intercorrelations exist within either set of variables. Such is the case for the predictor set where some moderately high bivariate correlations exist. An alternative method of interpretation is to examine the correlations between each variable of the criterion and predictor sets and its respective canonical variate (linear combination of all the variables in the set). The relative size of the loadings within a set indicate the general influence each variable exerts in defining the dominant group of variables displayed by the root being investigated. This form of interpretation is analogous to an examination of factor loadings in factor analysis.

The analysis of variate-variable correlations for each canonical relationship began with identification of the highest loadings (correlation). Variables were included in the interpretation of the equation in the order of their strength, stopping either when the loading was less than .25 or the new variable no longer "made sense" intuitively. Obviously, care is taken to avoid a posteriori rationalization to fit prior notions. The correlations of each variable with its respective criterion or predictor variate are provided in Table IV for the five statistically significant roots. Each root is explained separately below as patterns of shopping process and related profiles of personal characteristics are identified.

Canonical Root One

Although the criterion variables differ in the magnitude of their correlations, every shopping variable could conceivably be included in defining this form of shopping. Preplanning of shopping is conducted through grocery ad readership in a variety of papers and a shopping list is developed to reify the planning. Then, many trips are made to a variety of stores, some a distance away from the home, as the planned purchases are secured and comparison occurs. In summary, this shopping pattern epitomizes the "ideal shopper" in planning and search for market needs. An examination of the predictor set of variables provides a personal characteristic profile. The highly correlated variables are reflected in a marriage household with both the husband and wife having higher than average education, high income, and the husband being somewhat older. Multiple-car ownership emerges as provider of the mobility for comparative shopping. The personal characteristics description in this case reflects the general findings from previous studies. Better educated, higher income households (upper middle class) in moderate stages of the family life cycle purposefully organize shopping efforts.

TABLE IV

CANONICAL VARIATES - VARIABLE CORRELATIONS

Canonical Root Two

A quite different form of shopping appears in the second root. Here is seen the process involving virtually no preplanning through ad readership or list preparation. Instead, frequent trips occur, perhaps for some distance, to one or two stores. Little comparative shopping is done among stores and the lack of planning precipitates the need for frequent shopping. This pattern might be labeled as "unplanned shopper" in the market place. The household profile drawn from the loadings indicated a different family life cycle and educational level than in root one. Here is found the renter, new to the neighborhood, a young wife of lower education level, and children in the house of all ages. One sees the lower education and age combining to diminish the preplanning. This, combined with a larger family of children, necessitates frequent shopping.

Canonical Root Three

The third shopping pattern to be considered introduces preplanning of one form but little comparative shopping. List preparation loads highly although food ad readership does not. Few trips are made to the shopping site a long distance from the home. In-house planning occurs while the shopping task is a minimum-effort process. The "one-stop shopper" or, perhaps the "in-store shopper" is illustrated in this case.

The personal characteristic profile that emerges for this root resembles the immediately previous root. Key variables reflect that the woman of the house if young, a renter new to the area, and has small children. However, in contrast to the previous root, there are no school-age children and the housewife's education appears higher. This profile reflects preplanning as suggested by the restrictive burden of young children hampering ability to comparatively shop and the housewife's better education.

Canonical Root Four

Root four illustrates the type of shopping pattern where there specifically is no preplanning through ad readership in Paper C (inferring the newspaper's personality). Multiple trips are made to a variety of stores, yet a short distance traveled infers that any comparative shopping is in close proximity to the home. The personal characteristics indicative of this pattern are a white household with a young wife and no children. An automobile is available for mobility but does not enter the shopping process. A "neighborhood shopper" image appears as would be common in ethnic neighborhoods of many large cities. As an alternative interpretation, the behavior may be indicative of residents of a high retail density areas.

Canonical Root Five

Here the variables reflect preplanned shopping through food ad readership limited strictly to Paper C and combined with list preparation although specifically not consulting the other newspapers. The only other variable that loads heavily is the variety of stores visited. Neither number of trips nor distance are salient characteristics.

The characteristics that correlate highly with this pattern indicate the single woman with no children, probably living alone, in rental housing. This cluster of individuals spans a number of age levels since it didn't enter the description. A label is somewhat difficult for this household although it appears to represent a "single-moderate planner" type.

CONCLUSIONS

The findings from several studies unidimensionally investigating the role of economic and demographic characteristics on shopping behavior have indicated age, income, education, family life cycle, and other such variables as relating to shopping behavior. The results of the bivariate analysis in this study offers further support to such evidence. Our study indicated middle-aged female grocery shoppers are more prone to deliberate and to utilize shopping information. Individuals in earlier stages of the life cycle "shopped" less. Income was correlated with external search activity (distance traveled, number and variety of trips). Education in this study, and in all the studies reviewed, appeared to be a consistent factor related to information utilization.

The study suggested that deliberation and search are not simple univariate phenomena. Further, the combination of variables generated for respective sets of criterion and predictor variables adds clarity to our understanding of their interacting influence. Certain households read food ads in many different papers while preparing a shopping list. Others read only specific papers or just prepared a list. In the search activity, some households shopped a variety of stores over a broad area. Others restricted shopping to a variety of stores in close proximity. The variety of personal characteristics that described the criterion variable appear to be related to prepurchase deliberation and search behavior. Variate one portrays a deliberate shopper, characterized by a mobile family higher in income, and education and in one of the later stages of family life cycle. Other combinations (e.g., variate 2 vs. variate 3) suggest that factors influencing the perceived cost of search (e.g., children and babysitter expense) are moderated by other demographic factors such as education.

This study offers evidence that the total search process should be viewed as a "system" of activities and that market segmentation, perhaps, may be more fruitful when two or more variables are utilized. Some groups of people emphasize in-store activities in their search process while other differing on socioeconomic factors use a much different process of gathering information. From a public policy point of view, it seems that better educated consumers as well as truthful product labeling information, is the most fruitful area for improving disadvantaged consumers' positions.

REFERENCES

Alexis, M., Simon, L. S., and Smith, K. M. Some determinants of food buying behavior. In M. Alexis, R. S. Holloway and R. S. Hancock (Eds.), Empirical foundations of marketing. Chicago: Markham, 1969.

Alpert, M. I., and Peterson, R. A. On the interpretation of canonical analysis. Journal of Marketing Research, 1972, 2, 187-192.

Bucklin, L. Consumer search, role enactment and market efficiency. Journal of Business, 1969, 42, 416-438.

Darden, W. R. and Reynolds, F. D. Shopping orientations and product usage rates. Journal of Marketing Research, 1971, 8 s 505-508.

Engel, J. F., Kollat, D. F., and Blackwell, R. D. Consumer behavior. New York: Holt, Rinehart and Winston, 1973, 156-157 and 373-392.

Frank, R. E., Massy, W. F., and Wind, Y. Market segmentation. New York: Prentice-Hall, 1972.

Katona, G., and Mueller, E. A study of purchasing decisions. In L. H. Clark (Ed.), Consumer behavior: The dynamics of the consumer reaction. New York: New York University Press, 1955, 30-87.

Kollat, D. T. and Willett, R. P. Consumer impulse purchasing. Journal of Marketing Research, 1967, 4, 21-31.

Levy, S. J. Social class and consumer behavior. In S. W. Newman (Eds.), On knowing the consumer. New York: John Wiley and Sons, 1968.

Martineau, S. P. Social class and spending behavior. Journal of Marketing, 1958, 23, 121-130.

Peterson, R. A., and Alpert, M. I. Consumer choice patterns and psychographics: A test of predictability. Proceedings of the Fourth Annual Meeting of the American Institute for Decision Sciences, 173-180.

Prasad, V. K. Correlates of multi-store food shopping. Journal of Retailing, 1972, 48, 74-81.

Rich, S. U., and Jain, S. C. Social class and life cycle as predictors of shopping behavior. Journal of Marketing Research, 1968, 5, 41-49.

Schapker, B. L Behavior patterns of supermarket shoppers. Journal of Marketing, 1966, 30, 46-49.

Sparks, D. L., and Tucker, W. T. A multivariate analysis of personality and product use. Journal of Marketing Research, 1971, 8, 67-70.

Thorelli, X. B. Concentration of information power among consumers. Journal of Marketing Research, 1971, 8, 427-432.

----------------------------------------