Cognitive Personality Traits As Moderator Variables on the Depth of Search

ABSTRACT - Cognitive personality traits were shown to act as moderator variables on the relationships between depth of search in a set of information display board (IDB) tasks and such "product-specific psychological" variables as confidence, danger-risk, and perceived decision complexity. High cognitive-capacity individuals tended to secure more information, and low-capacity individuals less information, when they perceived low confidence, high danger-risk, or high complexity in the choice situation. These results were substantially weaker for IDB tasks which were more complex and novel.

Citation:

Charles M. Schaninger and W. Christian Buss (1984) ,"Cognitive Personality Traits As Moderator Variables on the Depth of Search", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 239-243.

Advances in Consumer Research Volume 11, 1984      Pages 239-243

COGNITIVE PERSONALITY TRAITS AS MODERATOR VARIABLES ON THE DEPTH OF SEARCH

Charles M. Schaninger, State University New York at Albany

W. Christian Buss, State University New York at Albany

ABSTRACT -

Cognitive personality traits were shown to act as moderator variables on the relationships between depth of search in a set of information display board (IDB) tasks and such "product-specific psychological" variables as confidence, danger-risk, and perceived decision complexity. High cognitive-capacity individuals tended to secure more information, and low-capacity individuals less information, when they perceived low confidence, high danger-risk, or high complexity in the choice situation. These results were substantially weaker for IDB tasks which were more complex and novel.

INTRODUCTION

Beginning with early work on perceived risk by Cox (1967a) and Bauer (1967), a number of researchers have investigated the influence of what we shall call product specific psychological variables on information search and processing. Among the more common have been perceived risk, prior knowledge, perceived complexity, and perceived differences among brands (Moore and Lehman, 1980). Recent studies have employed information display board (IDB) methodology to examine the effect of such variables on information search through either task-structure influences (Capon and Burke, 1977, 1980; Schaninger and Sciglimpaglia 1981) or individual differences (Green, Mitchell, and Staelin, 1977, Jacoby, Chestnut, and Fisher, 1978; Moore and Lehman 1980). This paper examines the influence of cognitive personality traits as moderator variables on the relationships between three of these choice-related variables and depth of search in IDB tasks

INFLUENCES ON DEPTH OF SEARCH

Three classes of variables appear to influence consumer information search and acquisition: Task structure variables (number of brands, number of attributes, the type of product and so on); general individual difference variables whose effects are stable across product classes; and product-specific individual differences whose effects are specific to each product class. Capon and Burke (1980) comment that mixed results were obtained from studies of these variables with search behavior apparently resulting from their interaction.

Task-Structure Influences

Employing an IDB task, Capon and Burke (1977, 1980) found that housewives examined more cues for the durable goods tested (microwaves, toaster ovens, and steam irons) than had previous studies of nondurables. More cues were examined for the riskier microwave decision. Significantly more cues and attributes were examined for the task which presented more attributes (27 vs. 14), a possible demand artifact. Schaninger and Sciglimpaglia (1981) also employed an IDB approach and found that housewives examined more cues for clothes dryers (perceived as more complex and higher in danger-risk presumably due to the higher cost) than for three instant beverage products. Number of cues and attributes examined was also a linear function of number of attributes available for the three instant beverage decisions. Thus, the degree of structural complexity (number of alternatives and attributes), and the degree of perceived risk in the product class influence depth of search. Budner (1962) recognized two other task structure influences: novelty (newness) and insolubility (internally conflicting values for an alternative on key attributes).

General Individual-Difference Influences

Schaninger and Sciglimpaglia (1981) found that demographics and cognitive style influenced depth of search in IDB tasks across products. Husband's occupational status and husband's and wife's education were positively related; while wife's age, state in family life cycle, and home-ownership were negatively related to depth of information search. Tolerance for Ambiguity and Self Esteem were positively related, and Cognitive Style (Simplifier) and Trait Anxiety were negatively related. Bariff and Lusk (1977) view Tolerance for Ambiguity as tapping the stress level at which individuals can effectively function in decision making tasks. Horton (1979) concluded that general anxiety and self confidence appeared to underlie the choice process for all products studied in a simulated shopping task. Rokeach (1960) concluded that Rigidity was negatively related to acceptance of new or discrepant information, and to ability to structure information in problem solving. Thus, individual factors which are not product specific affect search behavior.

Interactions Between General and Product-Specific Influences

Brody and Cunningham (1968) found measures of product-specific perceptions of performance risk and self-confidence enhanced the ability of personality tests to predict brand choice. Blake et al. (1973) found that tolerance for ambiguity (a general influence) interacted with perceived produce newness (a product-specific influence) in partially determining purchase intention. Thus, consumer personality traits may moderate the relationship between such product-specific measures as product newness and risk on purchase behavior and information search.

Individuals high in cognitive capacity (Low Trait Anxiety, High Self-Esteem, High Tolerance for Ambiguity, Clarifier Cognitive Style, and Low Rigidity) are likely to examine more information or choose riskier new products when they perceive less confidence, more danger-risk or higher complexity for a specific decision. Low capacity individuals are likely to act in an opposite manner.

Task structure influences are also likely to interact with the specific product and general influences. Complex and novel product classes are likely to introduce greater ambiguity (Budner 1962) and information overload (Bariff and Lusk 1977). An individual's perception of product specific influences should affect the overall level of decision-induced stress while cognitive-personality traits (general influences) affect the preferred and the maximum tolerable levels of stress. Thus, if task difficulty is too high, the moderator effects of general individual influences on product specific influences may break down. Therefore, task-structure variables must be varied across IDB tasks to examine interactive relationships.

Product-Specific, Individual-Difference Influences

Jacoby, Chestnut, and Fisher (1978) employed an IDB task on cold cereals (35 attributes, 16 brands) and found that depth-of-search was positively related to product-class importance, to being an optimizer rather than a satisficer, to past-purchase experience; negatively related to attitudinal brand loyalty; and not related to health or nutritional risk. Bettman and Park (1980) employed a protocol analysis and found that housewives of moderate knowledge and experience processed more available information than did high or low groups, suggesting an inverted U-shaped relationship. They argue that consumers low in prior knowledge and experience may be overwhelmed by the task of processing information in a complex and difficult task and hence process less information, and those high in prior knowledge and experience rely more on stored information and also process less information.

Moore and Lehman (1980) found that number of prior purchases of bread in a longitudinal IDB/purchase experiment was negatively correlated with information search on a given purchase occasion, and the repeat purchases of the same brand were negatively correlated with information acquisition. Thus, prior experience and knowledge may be negatively relatively to search only for less difficult tasks.

HYPOTHESES

Three sets of hypotheses are presented based upon the preceding sections. The first-order hypotheses are consistent with hypotheses presented in most of the prior literature and examine product-specific influences alone. The second-order hypotheses incorporate interactions between general and product-specific influences by treating cognitive-personality traits as moderator variables. The third-order hypotheses incorporate the interaction of task-structure influences with general and product-specific influences.

First-Order Hypotheses

H1: Confidence (reverse coded), danger-risk, and perceived complexity, (product-specific influences) will be positively correlated with number of cues, attributes, and alternatives (depth-of-search variables) examined in IDB tasks.

Second-Order Hypotheses

H2a: Positive correlations between confidence (reverse coded), danger-risk, and complexity and depth-of-search variables will be observed among individuals high in cognitive capacity -- e.g. Clarifiers, high Tolerance for Ambiguity, low Trait Anxiety, high self-esteem, and low rigidity (general influences).

H2b: Negative correlations between the above product-specific influences and depth of search will be observed among individuals low in cognitive capacity (general influences).

Third-Order Hypotheses

H3: For low-cost non-durable goods, the relationships in H2a and H2b will be strongest for a low novelty product with low task complexity (situation influences) and weakest for a high novelty product with high task complexity.

METHOD

The experiment consisted of a battery of questionnaire measures followed by four information display board shopping tasks. The experiment was specifically designed to include both convenience good tasks and durable good task. The four product classes chosen were instant coffees, non-dairy coffee creamers, instant lemonades, and electric clothes dryers.

Information Display Board Tasks

Instant coffee was chosen as a low-novelty, familiar, convenience good low in structural complexity (six attributes, fourteen alternatives). Coffee creamers were chosen to represent a less frequently purchased and less familiar (hence moderate novelty) convenience good higher in structural complexity (fourteen attributes, fourteen alternatives). Instant lemonades, then recently introduced to the market, were chosen to represent a convenience product higher in novelty and in structural complexity (sixteen attributes, thirteen alternatives). Dryers were chosen to represent a high-cost, infrequently purchased, high-risk (from high price) durable good of moderate to high complexity (ten attributes, thirteen alternatives). Attributes were generated from evaluations by Consumer Reports and by examination of information available to the consumer at the place of purchase.

The matrix boards themselves were constructed of hard pasteboard with information cues on cards held in paper coin envelopes glued to the boards. Multiple versions of the same brands (differing in size, product form or features) were used. Actual package information and price attributes were included for both store and national brand alternatives for the three low-cost convenience goods. For dryers, feature possession, capacity, and price attributes were utilized. Attributes were ordered alphabetically as rows in the matrix, and alternatives were labeled by brand name, with different sizes of the same brand name randomly ordered as columns. Based upon a pretest of sixteen subjects, the Bettman and Jacoby (1976) approach was modified in that a second "copy" of each information cue examined remained visible in its slot after selection rather than requiring subjects to remember or reaccess cues.

Questionnaire Measures

The questionnaire administered prior to the IDB task contained various demographic measures and seven-point sematic differential ratings for twelve nondurable and durable product classes [The twelve product classes for which these "product specific psychological" variables were determined were: instant lemonade mixes, clothes dryers, shampoos, instant coffees, microwave ovens, deodorants, soft drinks, carpeting, frozen vegetables, laundry soaps, refrigerators, and coffee cream substitutes. These measures for product classes other than four examined by the IDB task were included to allow response style bias or overall mean within individual differences to be measured and removed.] of: confidence in choosing among brands, degree of danger-risk present, and perceived decision complexity. Measures of the five cognitive-personality traits discussed previously were also obtained: Cognitive Style, as used by Cox (1967b) to measure the tendency to resolve cognitive uncertainty by simplifying (avoiding discrepant information) or by clarifying (seeking additional information to enhance understanding): Budner's (1962) measure of Intolerance for Ambiguity (with loadings reversed to measure tolerance for ambiguity); Spielberger's Trait Anxiety measure (Spielberger, Gorusch, and Lushene (1972); the Short Form of the Janis Test for Self Esteem, reported in Cox (1967a); and the Gough-Sanford Rigidity Scale as used by Rokeach (1960). Schaninger and Sciglimpaglia (1981) summarize the relationships between these measures. Respondents were divided into high and low groups based on their score on each cognitive-personality factor.

The Sample

One-hundred-twenty women were recruited from church-affiliated social groups in Lawrence, Kansas. These consumers agreed to participate in this study of "how consumers shop" in return for a small monetary contribution to their church group. The church groups were chosen from different areas of the city and varied with regard to the socioeconomic constituency of their membership. From 120 subjects 16 were used in a pretest and responses from two unsuitable subjects were dropped, resulting in usable data from 102 subjects participating in the revised experimental sessions. Although all age, family-life cycle, income, education, and occupation categories were represented, the sample was slightly skewed toward higher socioeconomic categories (as was the community) and toward older age categories.

RESULTS

In order to test the hypotheses, two possibilities were considered. The first, which we rejected, is to apply analysis of variance using cognitive-capacity scores to classify subjects into high and low factor levels. Since the primary purpose is to examine covariate relationships, we did not examine the main-effects of means, but for simplicity went directly to the covariance analysis by examining separate Pearson product-moment correlations within each set of subjects. To examine the overall relationships in a quasi-multivariate sense, each correlation was transformed to a Fischer Z value such that z = n-3 (1/2 in (l+r)/l 1-r)). The average z value for those in the high (and low) cognitive capacity groups is presented in each table. The t-test value is equal to the average of the five z values divided by their standard error, providing a test of whether the average of the set of five correlations is significantly different from zero.

First Order Hypotheses

In order to test H1, Pearson product-moment correlations were calculated between raw confidence, danger-risk, and complexity scores with depth of search for all four product classes using the entire sample. Only 4 of 48 correlations were significant (P < .05, two-sided) and three of these were opposite in direction to H1 When these raw scores were adjusted by "normalization" [Relative scores were calculated by dividing ar individual's score for a given product class by the average of the scores given to the 12 product classes by that individual on that dimension (e.g., confidence). This is directly equivalent to Bass and Wilkie's (1973) use of "normalization" on importance and evaluative belief ratings to remove response-style bias.] to remove response-style bias, weak results again occurred with 2 of 48 correlations significant in the expected direction and two opposite in direction to H1. Thus H. was not supported. These data are not presented because of the weak relationships.

The general failure of simple correlation analysis to support H1 is consistent with prior findings and suggests that the relationships between confidence, danger-risk, and complexity with information search are not the simple ones often postulated.

Second-Order Hypotheses

In order to test our second-order hypotheses, separate sets of Pearson product-moment correlations were computed for those high and low in cognitive capacity by splitting the sample at the median and to allow nearly equal-sized groups on each cognitive personality trait. These correlations were run on both raw and normalized scores on the three product specific influences. Only the correlations for the normalized scores are presented because they were consistently larger in magnitude and more consistent with each other and with H2a and H2b. These findings clearly indicate that response-style bias contaminates the raw rating scale scores and that raw scores should be adjusted to remove response-style effects and improve predictability

Tables 1, 2, and 3 present these correlations between depth of search and relative product-specific variables for those high and low in cognitive capacity. The average Fischer z-transformation and t-test values are also presented. Correlations for the creamer task are not presented because they were generally weak in magnitude and significance with the only significant pattern observed for relative confidence. H2a was most strongly supported for instant coffees and moderately supported for dryers when confidence is examined. H2a is supported only for instant coffees when relative danger-risk is examined, and only for dryers when relative complexity is examined. Surprisingly, the pattern of correlations for instant lemonades was significant in the opposite direction of H2a for relative confidence. (See the next section for further discussion). A comparison of the instant coffee correlations for confidence and danger-risk suggests that high cognitive-capacity individuals resolve low confidence situations by examining more alternatives, but resolve high danger-risk situations by examining more attributes.

Hypotheses 2b received limited support, with consistent patterns of significant correlations emerging primarily for the number of alternatives examined on both danger risk and complexity for instant coffees, and on confidence and danger-risk for instant lemonades. Although few other significant correlations emerged, the majority of the correlations for the low cognitive-capacity groups were negative (the expected direction), and 11 out of 27 significant t-tests on the average z transformation values were observed. Since the questionnaire measures were collected first, stronger results may have emerged in a post test.

Although the support for Hypotheses 2a and 2b is limited, the general pattern of findings clearly indicates that cognitive personality traits play a moderating role in influencing how individuals react to situations of low confidence, high danger-risk, or high complexity. Since the aggregation of high and low capacity individuals together for correlation analyses obscures the effect of product-specific variables, and results in near zero or "contradictory" correlations similar to those found by prior researchers.

Third-Order Hypotheses

Correlations in support of H2a and H2b were generally stronger for instant coffees than for coffee creamers, as suggested by H3. The patterns of correlations of H2a and H2b observed for instant lemonades were different than those observed for instant coffees and coffee creamers' not merely weaker as hypothesized. Patterns opposite of H2a were observed between confidence and all three depth-of-search criteria (in agreement with the first-order correlations reported previously). Although H2b was technically supported for confidence on instant lemonades, those negative correlations were generally lower in magnitude than the negative (contradictory) correlations found among high cognitive-capacity individuals. Thus, the instant lemonade task resulted in different information search processes than those hypothesized and demonstrated on the instant coffee task. The relationships specified by H2a and H2b tended to break down and be replaced by more complicated relationships, thus supporting a modified version of H3.

CONCLUSIONS

Bivariate correlation results show no meaningful or consistent pattern of relationships between depth of search and such variables as confidence, danger-risk and complexity. Incorporation of cognitive personality traits as moderator variables, however, does result in the discovery of weak but consistent and meaningful relationships in the directions hypothesized. High cognitive-capacity individuals evidence positive correlations between confidence, danger-risk, and complexity with depth of search, while low cognitive-capacity individuals show negative correlations. These findings were stronger and more consistent when raw scores were adjusted by normalization suggesting contamination by response-style bias. The degree of structural complexity and novelty present in an IDB task also appears to influence the form of these relationships. Support for hypothesized relationships is most definitive for instant coffee and dryers, with weaker results for creamers and contradictory results for instant lemonades. These "contradictory" lemonade results may be due to a high degree of ambiguity, stress or information overload being presented by this less familiar, more novel, and more structurally complex task. However, they may also result from gathering the ratings prior to the experimental task. Post-task ratings might well demonstrate stronger support for the hypothesized relationships.

TABLE 1

CORRELATIONS OF RELATIVE CONFIDENCE, LEVELS OF COGNITIVE CAPACITY WITH DEPTH OF SEARCH

TABLE 2

CORRELATIONS OF RELATIVE DANGER-RISK, LEVELS OF COGNITIVE COMPLEXITY AND DEPTH OF SEARCH

TABLE 3

CORRELATIONS OF RELATIVE COMPLEXITY, LEVELS OF COGNITIVE CAPACITY AND DEPTH OF SEARCH

Although many of the product-moment correlations were not significant the pattern of their signs were consistent with the hypotheses. Most of the average t-test values (which test the probability that the average of all five correlations is zero) were significant (26 of 54 in the direction hypothesized with 9 of 54 opposite in direction).

We conclude that there is a joint, moderating effect between general cognitive capacity influences and "product-specific influences" on search behavior. Much prior research and our weak bivariate correlation results highlight the difficulty of demonstrating this effect. The laboratory situation, possible respondent boredom, demand artifacts, and so on all may interact in IDB tasks to mask such effects. Thus, future research on the influence of product-specific individual influences (such as confidence, familiarity, danger-risk, or complexity) on information search and processing should not ignore the influence of cognitive capacity or cognitive personality traits. More theoretical and empirical work is needed to tie together the interactive effects of such cognitive personality and product-specific individual influences to task or structural variables (such as number and types of attributes, number of brands and alternative size or color/flavor variations within brands, the number of salient dimensions, the degree of actual interbrand differences on salient dimensions, the degree of newness of the product class, the amount of money at stake, and so on). Similarly, extensions of such research to more realistic settings such as laboratory test markets, or the use of survey or interview approaches to determine degree, depth, or type of search might allow us to expand or generalize our knowledge to a greater degree.

REFERENCES

Bariff, M.L. and Lusk, E. J. (1977) "Cognitive Personality Tests for the Design of Management Information Systems," Management Science, 23, 820-829.

Bauer, Raymond A. (1967) "Consumer Behavior as Risk Taking," in Donald F. Cox, (ed.), Risk Taking and Information Handling in Consumer Behavior, Boston: Harvard UniversitY Press, 23-33.

Bettman, James R. and Jacoby, Jacob (1976) "Patterns of Processing in Consumer Information Acquisition," B.B. Anderson, (ed.), Advances in Consumer Research: Volume III, Cincinnati: Association for Consumer Research, 7, 234-248.

Bettman, James R. and Park, C. Whan (1980) "Effects of Prior Knowledge and Experience and Phase of the Choice Process on Consumer Decision Processes: A Protocol Analysis," Journal of Consumer Research, 7, (December) 234-248.

Blake, Brian F., Perloff, Robert, Zenhausern, Robert, and Heslin, Richard (1973) "The Effect of Intolerance of Ambiguity Upon Product Perceptions," Journal of Applied Psychology, 58, 239-243.

Brody, Robert P. and Cunningham, Scott M. (1968) "Personality Variables and the Consumer Decision Process," Journal of Marketing Research (February), 50-57.

Budner, Stanley (1962) "Intolerance of Ambiguity as a Personality Variable," Journal of Personality, 30, 29-50.

Capon, Noel and Burke, Marion (1977) "Information Seeking in Consumer Durable Purchases," in Barnett A. Greenberg and Danny N. Bellinger (eds.), Contemporary Marketing Thought, 1977 Educator's Proceedings, Chicago: American Marketing Association 12, 110-115.

Capon, Noel (1980) "Individual, Product Class, and Task-Related Factors in Consumer Information Processing," Journal of Consumer Research, 7 (December), 314-326.

Cox, Donald F. (1967a) "Synthesis-Perceived Risk and Information Handling," in D.F. Cox (ed.), Risk laking and Information Handling in Consumer Behavior, Boston: Harvard University Press, 603-639.

Cox, Donald F. (1967b) "The Influence of Cognitive Needs and Styles on Information Handling in Making Product Evaluations," in D.F. Cox, (ed.), Risk Taking and Information Handling in Consumer Behavior, Boston: Harvard University Press, 370-392.

Green, Richard, Mitchell, Andrew A., and Staelin, Richard (1977) "Longitudinal Decision Studies Using a Process Approach: Some Results from a Preliminary Experiment," in Barnett A. Greenberg and Danny Bellinger (eds.), Contemporary Marketing Thought, Chicago: American Marketing Association, 401-66.

Horton, Raymond L., (1979) "Some Relationships Between Personality and Consumer Decision Making," Journal of , 16, 233-246.

Jacoby, Jacob, Chestnut, Robert W., and Fisher, William A. (1978) "A Behavioral Process Approach to Information Acquisition in Nondurable Purchasing," Journal of Marketing Research, XV, (November), 532-44.

Moore, William L., and Lehman, Donald R. (1980) "Individual Differences in Search Behavior for a Nondurable," Journal of Consumer Research, 7 (December), 296-307.

Rokeach, Milton (1960) The Open and Closed Hind, New York: Basic Books, Inc.

Schaninger, Charles M. and Sciglimpaglia, Donald, (1981) "The Influence of Cognitive Personality Traits and Demographics on Consumer Information Acquisition Behavior," Journal of Consumer Research, 9, (September).

Spielberger, C.D., Gorusch, D.L., and Lushene, R.E. (1970) Manual for the State-Trait Anxiety Inventory, Palo Alto, California: Consulting Psychologists Press.

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