The Effects of Expertise on Preference and Typicality in Investment Decision Making

ABSTRACT - Preference is positively correlated with typicality in many product categories. As expertise increases, however, preference and typicality may change as well as the relationship between them. In this study of investment decision making, the strength of the correlation between preference and typicality decreased as subjects' expertise increased. In addition their ratings of preference and typicality differed by level of expertise. Two approaches were used to measure expertise--one cognitive measure and one experiential. This article presents an explanation of the results and a discussion of their implications.

Citation:

W. Steven Perkins and Valerie F. Reyna (1990) ,"The Effects of Expertise on Preference and Typicality in Investment Decision Making", in NA - Advances in Consumer Research Volume 17, eds. Marvin E. Goldberg, Gerald Gorn, and Richard W. Pollay, Provo, UT : Association for Consumer Research, Pages: 355-360.

Advances in Consumer Research Volume 17, 1990      Pages 355-360

THE EFFECTS OF EXPERTISE ON PREFERENCE AND TYPICALITY IN INVESTMENT DECISION MAKING

W. Steven Perkins, Penn State University

Valerie F. Reyna, University of Arizona

ABSTRACT -

Preference is positively correlated with typicality in many product categories. As expertise increases, however, preference and typicality may change as well as the relationship between them. In this study of investment decision making, the strength of the correlation between preference and typicality decreased as subjects' expertise increased. In addition their ratings of preference and typicality differed by level of expertise. Two approaches were used to measure expertise--one cognitive measure and one experiential. This article presents an explanation of the results and a discussion of their implications.

INTRODUCTION

Alba and Hutchinson (1987) have hypothesized that, "Brand choices are less influenced by prototypicality for experts than for novices, except when prototypicality or atypicality is valued for its own sake (p. 418, hypothesis 2.9)." They suggest that as consumers gain experience the relationship between product preferences and product typicality changes. Novices' choices may be based upon whether a product is a good example of its product category while experts discriminate between products on more functional attributes.

This research builds upon two major areas in cognitive psychology that relate to decision making. The first pertains to typicality--how well an object represents its cognitive category (Rosch 1975, 1983). The second area concerns the effects of expertise. Experts think differently not only because they know more, but also because they organize and use their knowledge better (Chi 1983).

After reviewing the relevant literature, three hypotheses are proposed. An exploratory study is used to examine these hypotheses. The remainder of this article describes its results and discusses the implications, limitations, and goals for future research.

Typicality and Preference

Categorization is the process of forming classifications by which nonidentical stimuli can be treated as equivalent (Rosch and Mervis 1975). According to Rosch, objects in the same cognitive category may differ in terms of typicality (or prototypicality), meaning that some objects represent the grouping better than others. Objects in a category lie on a gradient of typicality, ranging from extremely good examples to very poor examples of the category (Barsalou 1985). Prototypical members of a category serve as reference points to which other category members can be compared (Rosch 1975, 1983) and prototypes are usually learned first (Mervis and Rosch 1981). Differences in typicality have been noted in both natural object categories, such as fruit (Rosch and Mervis 1975), and product categories, such as snack food (Ward and Loken 1986). A competing but similar theory of categorization argues that category membership is based upon comparison to specific, retrieved instances, or exemplars, rather than abstracted prototypes (Medin 1983).

Typicality can be an important factor in product preference and choice. Nedungadi and Hutchinson (1985) asked students to rate the prototypicality of several magazines and beverages, then rate how much they liked each product. After aggregating the responses across all subjects, they found a strong positive relationship (r>.61)) between liking and prototypicality. Loken and Ward (1987) also obtained a high positive correlation (r=.88) between typicality ratings and attitudes toward brands of shampoo. These results are in contrast to Rosch (1973) who found a weak relationship in natural object categories. One explanation for these differences could be that consumers construct product categories "ad hoc" in order to satisfy the goal of evaluating and purchasing a product (Barsalou 1983). Thus, affect may play a much more important role in structuring functional, product categories than natural object categories used in much of the original categorization research (Cohen 1982). Natural categories structure our perceptual experience without regard to preferences.

Effects of Expertise on Preference and Typicality

The second theoretical underpinning for this research concerns expertise. After extensive research (see Alba and Hutchinson 1987), it appears that experts organize their knowledge hierarchically by purpose as compared to novices who organize data linearly in discrete bits by attributes. Expert and novices appear to use the same types of problem solving heuristics and search methods (Chi 1983). Instead, expert performance is due to organizing their knowledge on the basis of domain relevant functionality. Novices, on the other hand, tend to organize their knowledge about a domain at a perceptual level.

As the organization of knowledge changes with increasing expertise, the relative typicality of products as well as the use of typicality in decision making may also change. Prototypical objects are learned first, but the level of perceived prototypicality for an object differs by an individual's level of familiarity with the domain (Hermann and Kay 1977). One experiment (Murphy and Wright 1984) found that as people know more about an area (in this case psychiatric diagnosis), the simple, black and white distinctions and definitions blur into fuzzier, generalities. Framed in terms of typicality, the experts based their decisions upon factors other than stereotyped, prototypical knowledge.

Also as expertise increases, experts not only know more products they know more about efficiently and effectively choosing products. Bettman and Park (1980), for example, found that novice microwave buyers based their decisions on a brand they "had heard of," rather then evaluating brands on their attributes. Preference depended solely upon availability in memory. Novices represent objects at a superficial level as compared to a deeper structure level (Chi 1983). As noted by Alba and Hutchinson (1987), mere familiarity, the number of experiences accumulated by a consumer, is not the same as expertise, the successful performance of tasks. Expertise requires familiarity, but familiarity does not guarantee expertise.

In summary, the objects in cognitive categories differ in terms of typicality. Preference correlates positively with typicality in many product categories. As expertise increases, the organization of knowledge in memory becomes more functional, and processing becomes less superficial. Also as experience increases, atypical examples and instances are encountered and stored in memory. Thus, the influence of typicality diminishes in decision making and choice, resulting in a weaker link between preference and typicality.

THE STUDY

As expertise increases, consumer preferences are less influenced by product typicality (Alba and Hutchinson 1987). The objective of this research is to examine three aspects of that statement. Specifically, it is hypothesized that:

H1: Product preference will differ by level of expertise.

H2: Product typicality will differ by level of expertise.

H3: Preference and typicality will be positively related, but the strength of the relationship will decrease with expertise.

To test these hypotheses, data were gathered from a domain in which subjects could be expected to differ in knowledge and experience: personal investments. A category such as investments is actively, even voluntarily learned, as compared to natural object categories or even common product categories. Most previous empirical research on investment decision making (Cooley 1977; Soutar and Ascui 1980) has not specifically considered the effects of expertise, though Slovic (1972) compared the decision strategies of one novice and one expert stock broker.

An overview of the tasks and derived measures follows

Task:

1. Unaided recall of investments

2. Sort 18 investments

3. Rate 18 investments

4. Check off ones owned

Measure:

1. # recalled

2. typicality

3. preference

4. # owned

The first task asked subjects to write down all the personal investments they could recall. The remaining tasks involved a set of 18 investments given to the subject. They sorted the investments into groups of similar items, then rated their preference for each and noted whether they had ever owned each one. These 18 investments were selected based upon a previous experiment in which 60 MBA students listed all the investments they could recall. The 18 most frequent examples were used in the present experiment.

Subjects and Administration

A total of 40 people completed usable questionnaires. One half were undergraduate students, mostly seniors, in an upper division math class. They were expected. to be relatively naive about investments. The other respondents had completed at least a bachelor's degree. This second group included both part-time MBA students and other non-students, many of whom worked on a day to day basis in the investment field. These financial employees included a commodities broker, international money trader, vice president of finance and others considered experienced in investments. The questionnaire was administered individually, or in small groups, to these subjects because of the difficulty of gaining access to them. Although the administration of the questionnaire varied somewhat across the groups, the study required gathering data from "real world" experts as well as novices.

Measures and Procedure

On the first page of an eight page questionnaire, subjects wrote down all the personal investments they could recall in a 2 minute time period. Subjects had no previous exposure to the stimuli and no warning as to the task, thus it required them to search their knowledge base to recall examples. This page was then returned to the experimenter. For the first measure of expertise, the total number of investments recalled was calculated, regardless of whether the example recalled was included in the target list of 18 investments. The average subject recalled 10.5 investments, with a range from 4 to 21. In previous research, the total number of factors recalled by a subject has been employed as a measure of expertise, under the assumption that it corresponds to the "richness of the cognitive process undertaken by the individual in this specific problem area (Larreche and Moinpour 1983)." Also the examples recalled were later matched with the 18 investments used in the latter part of the study to ensure that the stimuli covered the majority of the investments recalled by subjects. The 18 prespecified investments captured almost 75% of the investments mentioned during free recall. Of the 25% mentioned but not subsequently encountered in other tasks, many were redundant or unusual items.

The remaining pages of the questionnaire were then completed at each subject's own pace with most finishing in less than 15 minutes. The second page of the questionnaire gave an overview and instructions for the remaining tasks. For the second task, subjects received an envelope containing 18 pieces of paper with the name of an investment on each one. Subjects were told to sort them into separate piles based upon what they considered similar types of investments. Then within each sort group, they arranged them from the best example of that type of investment to the poorest example. Subjects recorded their sorting results on another page. The measure of typicality derived from the sorting results considered the relative position of an investment in its sort group, specifically:

                       total # of items in the sort group

Typicality  =                - order in the group           

                      total # of items in the sort group - 1

With this formula, typicality is scaled so that the first investment in a group equals 1 and the last equals 0. For example, in a group of four items the typicality scores would be 1, .66, .33, and 0. If a sort group consisted of only one investment, it was given a typicality score of 1. Sorting provides a less direct, but possibly more valid, approach to measuring typicality than the rating methods used in previous work. Subjects may find it difficult to separate product preference and typicality when asked to rate both on the same type of scale (e.g., Loken and Ward 1987; Nedungadi and Hutchinson 1985)

In the third task, subjects rated their preference for each of the same 18 investments on a scale from 0 (not preferred at all, would not invest in it) to 10 (totally preferred, my ideal investment). The question sought to measure their likelihood of investing rather than the amount of money they would invest. Ratings were later converted into z scores by individual in order to make them more comparable across subjects.

For the last task, respondents checked off which of the 18 investments they had owned or anyone in their household had owned. For the second measure of expertise, the total number of investments owned was calculated. The average subject owned (or their household owned) 7.8 of the investments with a range from 0 to 16. This measure of expertise was based on the same assumption used in previous research that expertise increases with greater frequency of purchase and use (e.g., Bettman and Park 1980).

Finally, subjects answered several demographic questions concerning their age, number of children, and annual household income. Two additional questions asked for their self-rated level of knowledge about investments and the amount of time they spend working on investments. These two questions used a 5 point scale increasing from very little (1) to a great deal (5).

RESULTS

The four major measures are summarized in Table 1: the mean preference rating and typicality score for each investment, and the percentage recalling and owning each investment. Considering "shares of stock," for example, 88% of the subjects recalled it and 70% owned it, while it received an above average (positive) preference rating and a relatively high typicality score. "Oil and gas partnership," for instance, was much lower on every measure. At this aggregated level, mean preference and mean typicality correlated positively (r=.57, p=.01) as expected.

The first two hypotheses state that preference and typicality will differ by level of expertise. Using the total number recalled and the total number owned as measures of expertise, a multiple analysis of variance (MANOVA) was performed over the 18 preference ratings and over the 18 typicality scores. Table 1 indicates that both did differ by the number of investments recalled by a subject, typicality differing to a somewhat greater extent (p<.01) than preference (p<.04). The number of investments owned by a subject also significantly differentiated between typicality scores (p<.04) but not preferences (p<.22). Both H1 and H2 are confirmed, though the number of investments recalled was a more fruitful measure of expertise than the number owned. Because the questionnaire was administered differently to undergrads and others, the same MANOVA analyses were performed using method of administration as the explanatory variable. As expected, neither preferences (F = 0.85, p<.63) nor typicality (F = 1.23, p<.32) differed by method.

The third hypothesis suggested that preference and typicality will be positively related but that the size of that correlation will decrease as expertise increases. By individual, preference and typicality were correlated, resulting in a median Pearson coefficient of .19 indicating a positive but weak relationship. A positive coefficient occurred for 75% of the subjects, but only 23% had a positive, significant coefficient (r>.31, p<.05, n=18). Thus, for the majority of subjects preference and typicality correlated positively, though the coefficients ranged widely from -.49 to +.62.

To detect changes in this relationship between preference and typicality with changes in expertise, the Pearson correlation was regressed on each of the measures of expertise. As shown in Table 2, the sign of the regression coefficient for both measures of expertise is negative and significant, indicating that the size of the correlation between preference and typicality decreases as expertise increases. By both measures of expertise, H3 is confirmed. (A Fisher r to z transformation was made on each Pearson correlation coefficient before regressing on expertise).

Finally, because much of this research hinges on the concept of expertise, the validity of the two measures of expertise should be considered. The number of investments recalled and the number owned correlated .57, signifying that the two are relatively similar. In addition, each can be correlated with the demographic and self-rating data from the end of the questionnaire as shown in Table 3. (These data were considered ordinal, necessitating a Spearman correlation.) The results have strong face validity, especially for the number of investments owned. As the number owned increased, subjects were older, had more children, had a higher income, were less likely to be undergraduates, rated themselves as more knowledgeable, and spent more time on investments. Of course, many of these factors are highly related themselves. The number recalled differentiates less well, though the results were in the same direction.

TABLE 1

SUMMARY DATA ON INVESTMENTS

TABLE 2

REGRESSION EQUATIONS CORRELATION BETWEEN PREFERENCE AND TYPICALITY REGRESSED ON EXPERTISE

TABLE 3

SPEARMAN CORRELATIONS BETWEEN EXPERTISE AND DEMOGRAPHIC DATA

CONCLUSIONS AND IMPLICATIONS

Based upon this research in the domain of investments, it appears that both preference and typicality do change with expertise. In addition, preference and typicality were found to be positively related in the majority of subjects. This is in accord with previous research which has found a positive relationship between preference and typicality (Nedungadi and Hutchinson 1985). One aspect this current research contributes is the finding that the strength of this relationship decreases with expertise. Thus, it seems that novices' preferences are tied more closely to judgments of typicality (and/or vice versa), while experts' preferences are based upon other factors, possibly risk and return.

Limitations to this research should be addressed. First, the method for measuring typicality assumes a linear, equal-interval relationship between the objects in a sort group. Instead, the graded category structure might be nonlinear, with objects clustered together according to their similarity (Barsalou 1983). The measure employed in this paper should also be compared with the results derived from a more direct rating of typicality. Second, the measures of expertise, though employed in previous research, did not perform consistently across the analyses. For example, the number of investments owned by the subject or their household did not differentiate between preference ratings but it did correlate as expected with the demographic data. The number of investments originally recalled measures the individual's level of cognitive complexity and thus probably more closely approximates expertise. Third, though the primary purpose of the research was to investigate the changing relationship between typicality and preference, measures of investment risk propensity would also have been an interesting parallel question. Risk influences preferences for different investments, but also may influence or even be a part of typicality.

The research has implications for both marketing and decision making. Very typical objects in a category may serve as cognitive reference points in decision making. Rosch (1983) argues that judgment can be based on formal logical reasoning or on reference point reasoning. The latter type of reasoning, though relatively primitive, occurs quite often and can be valid. In a risky class of decisions, such as investments, logical thinking may play a more important role, especially for experts. In many product classes the difference between a novice and an expert is inconsequential. But in a complex product class the degree of familiarity can greatly influence an individual's ability to make reasonable decisions. First time buyers of cars, computers, and homes face an enormous amount of information. With little cognitive structure developed to hold the new information, they may rely on intuitive heuristics such as reference points in making decisions.

Future research could investigate methods for improving decision making and the resulting effects on the relationship between preference and typicality. Can we teach novices to make decisions like experts? Most research to date has concentrated on judgmental mistakes (Kahneman, Slovic, and Tversky 1982) with little said about training decision makers. The next phase of research on expertise should explore the means to improve decision making abilities.

REFERENCES

Alba, Joseph W. and J. Wesley Hutchinson (1987), "Dimensions of Consumer Expertise," Journal of Consumer Research, 13 (March), 411-454.

Barsalou, Lawrence W. (1983), "Ad Hoc Categories," Memory and Cognition, 11 (May), 211-227.

Barsalou, Lawrence W. (1985), "Ideals, Central Tendency, and Frequency of Instantiation as Determinants of Graded Structure," Journal of Experimental Psychology: Learning, Memory, and Cognition, 11 (October), 629-654.

Bettman, James R. and C. Whan Park (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.

Chi, Michelene T. H. (1983), "The Role of Knowledge on Problem Solving and Consumer Choice Behavior," in Advances in Consumer Research, Vol. 10, Cincinnati,OH: Association for Consumer Research, 569-571.

Cohen, Joel B. (1982), "The Role of Affect in Categorization: Toward a Reconsideration of the Concept of Attitude," in Advances in Consumer Research, Vol 9, e 1. Andrew A. Mitchell, Provo, UT: Association for Consumer Research, 94-100.

Cooley, Philip L. (1977), "A Multidimensional Analysis of Institutional Investor Perception of Risk," Journal of Finance, 32 (March), 67-78.

Hermann, Douglas J. and Barbara E. Kay (1977), "Familiarity and Organization of Category Terms in Semantic Memory," Memory and Cognition, 5, 139-145.

Kahneman, Daniel, Paul Slovic, and Amos Tversky (Eds.) (1982), Judgment Under Uncertainty: Heuristics and Biases. New York: Cambridge University Press.

Larreche, Jean-Claude, and Reza Moinpour (1983), "Managerial Judgment in Marketing: The Concept of Expertise," Journal of Marketing Research, 20 (May), 110-121.

Loken, Barbara, and James Ward (1987), "Measures of the Attribute Structure Underlying Product Typicality," in Advances in Consumer Research, Vol 14, eds. Melanie Wallendorf and Paul Anderson, Provo, UT: Association for Consumer Research, 22-26.

Medin, Douglas L. (1983), "Structural Principles in Categorization," in Perception, Cognition, and Development: Interactional Analyses, Thomas J. Tighe and Bryan E. Shepp (eds.), Hillsdale, NJ: Lawrence Erlbaum Associates, 203-230.

Mervis, Carolyn B. and Eleanor Rosch (1981), "Categorization of Natural Objects," in Annual Review of Psychology, Vol 32, eds Mark R. Rozenzweig and Lyman W. Porter, Palo Alto, CA: Annual Reviews, Inc., 89-115.

Murphy, George L. and Jack C. Wright (1984), "Changes in Conceptual Structure with Expertise: Differences Between Real-World Experts and Novices," Journal of Experimental Psychology: Learning Memory and Cognition, 10, 144-155.

Nedungadi, Prakash, and J. Wesley Hutchinson (1985), 'The Prototypicality of Brands: Relationships with Brand Awareness, Preference and Usage," in Advances in Consumer Research, Vol 12, eds. Elizabeth Hirschman and Morris Holbrook Provo, UT: Association for Consumer Research, 498-503.

Rosch, Eleanor (1973), "On the Internal Structure of Perceptual and Semantic Categories," in T.E. Moore (ed.) Cognitive Development and the Acquisition of Language, 111-114, New York: Academic Press.

Rosch, Eleanor (1975), "Cognitive Reference Points," Cognitive Psychology 7, 532-547.

Rosch, Eleanor (1983), "Prototype Classification and Logical Classification: The Two Systems," in New 'Trends in Conceptual Representation, Ellin Kofsky Scholnick (ed.), Hillsdale, NJ: Lawrence Erlbaum, 73-85.

Rosch, Eleanor and Carolyn B. Mervis (1975), "Family Resemblances: Studies in the Internal Structure of Categories," Cognitive Psychology, 7, 573-603.

Slovic, Paul (1972), "Psychological Study of Human Judgment: Implications for Investment Decision Making," Journal of Finance, 27 (Sept), 779-799.

Soutar, Geoffrey N. and Alvaro F. Ascui (1980), "Analyzing Preferences for Investment Projects: A Multidimensional Scaling Approach," Decision Science, 11 (July), 471-481.

Ward, James and Barbara Loken (1986), 'The Quintessential Snack Food: Measurement of Product Prototypes," in Advances in Consumer Research, Vol 13, ed. Richard J. Lutz, Provo, UT: Association for Consumer Research, 126-131.

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