The Effect of Product Expertise on Decision Making and Search For Written and Sensory Information

Fred Selnes, Norwegian School of Management BI
Row Howell, Texas Tech University
ABSTRACT - Product expertise is generally thought to be an important variable in consumer behavior theory. Much attention has been devoted to how consumer expertise affects the amount of information acquired in consumer information search. In this article we suggest that not only the amount, but also the type of information sought is relevant in understanding the differences between experts and novices. A distinction is made between written and sensory information, where written refers to information acquired through reading, and sensory refers to information acquired (or sensed) through hearing, feeling, tasting, and seeing.
[ to cite ]:
Fred Selnes and Row Howell (1999) ,"The Effect of Product Expertise on Decision Making and Search For Written and Sensory Information", in NA - Advances in Consumer Research Volume 26, eds. Eric J. Arnould and Linda M. Scott, Provo, UT : Association for Consumer Research, Pages: 80-89.

Advances in Consumer Research Volume 26, 1999      Pages 80-89

THE EFFECT OF PRODUCT EXPERTISE ON DECISION MAKING AND SEARCH FOR WRITTEN AND SENSORY INFORMATION

Fred Selnes, Norwegian School of Management BI

Row Howell, Texas Tech University

ABSTRACT -

Product expertise is generally thought to be an important variable in consumer behavior theory. Much attention has been devoted to how consumer expertise affects the amount of information acquired in consumer information search. In this article we suggest that not only the amount, but also the type of information sought is relevant in understanding the differences between experts and novices. A distinction is made between written and sensory information, where written refers to information acquired through reading, and sensory refers to information acquired (or sensed) through hearing, feeling, tasting, and seeing.

We develop a process model that integrates product expertise, information search, and cognitive processes, which allows us to test both direct and indirect effects of expertise on information search in decision making. First, we find, as predicted, that product experts seek more information than novices do. Second, we find that the degree of product expertise reduces reliance on written cues and increases the amount of inference drawn from sensory cues. Thus, product experts rely more on sensory information in decision making as compared to product novices. The preference and motivation to apply sensory information cues in consumer decision making will affect, among other things, how electronic marketing should be organized in order to succeed.

INTRODUCTION

The increased demand for high-quality products and the recent developments in new technology for information search (i.e. Internet) accentuates our understanding of how product expertise affects choices and evaluations. In order to appreciate the intrinsic value of a high-quality product, consumers may either trust the reputation of the brand, make his or her own inferences, or a combination thereof (Rao and Sieben 1992). As the consumer in general appears to be becoming increasingly more knowledgeable about products, marketers are now addressing product expertise as an important variable in order to understand current and future needs. In particular, expertise may affect how consumers acquire and utilize information (Brucks 1985; Johnson and Russo 1984). It has been argued that not only the amount, but also type of information is relevant in order to understand differences between product experts and novices (e.g., Park and Lessig 1981; Rao and Sieben 1992). We suggest that physical inspection of a product provides a different type of information as compared to verbal or oral descriptions, and that experts and novices differ with respect to their ability and thus inclination to acquire the former type. We will later refer to physical inspection as acquisition of sensory information. Although information type has been addressed in the literature, the empirical investigations has to our knowledge been limited to studying acquisitions of only different forms of descriptive information.

Another major limitation with the information search literature addressing product class knowledge is the lack of explication of the cognitive processes underlying the search behavior. Expertise has a direct affect upon the problem solving process itself. An important observation was made by De Groot (1965), who found that experts in chess did not seek out more alternative moves than novices do. The major difference was that experts were better at making good moves, that is, finding good solutions. Thus, understanding and framing the problem are important traits of expertise. The most common approach is to measure overt search behavior and infer the underlying cognitive processes. Product class knowledge and overt search behavior have been analyzed both in laboratory simulations (e.g., Brucks 1985; Jacoby, Chestnut and Fisher 1978; Sheth and Vanketesan 1968; Moore and Lehman 1980; Swan, 1969; Tyebjee 1979) and through retrospect analysis with surveys (e.g., Anderson, Engeldow, and Becker 1979; Bennett and Mandel 1969; Bucklin 1966; Formisano, Olshansky and Trapp 1982; Katona and Mueller 1955; Kiel and Layton 1981; Locander and Herman 1980; Newman and Staelin 1972; Punj and Staelin 1983; Urbany, Dickson, and Wilkie 1989). These studies have generally not integrated overt search measures with cognitive process measures.

One interesting exception is Bettman and Park (1980). In a laboratory simulation using information display board (IDB) methods they examined the cognitive process. The subjects were asked to think-aloud, and their verbalizations were tape-recorded. They studied the interrelationship between prior knowledge, phase of the choice process, and consumer decision processes, finding that consumers with moderate knowledge and experience did more processing of available information than did the high or low groups. More knowledgeable consumers tended to process by brand. Consumers tended to use attribute-based evaluations in early phases and brand-based evaluations in later phases of choice. One limitation with this study is that only those search activities that were verbalized were registered, and thus analyzed.

The second, and major limitation with this and other IDB simulations is the limitation of the information available. In a real setting, consumers have several other types of information available. Information related to design, functionality, real size, smell, sound, and similar attributes are never available in the traditional display boards. A distinction is commonly made between extrinsc and intrinsic attributes or product cues. Extrinsic cues are easily available from either the product or descriptions of it. Thus, the IDB provides extrinsic cues. Intrinsic cues, on the other hand, are not directly observable but must be inferred. Such inference can be made through physical inspection or use experience with the product. This type of information is not available in IBD, as the real products are not provided. It is quite reasonable to believe that this latter type of information will affect information search in the decision process.

Similarly, it is important to distinguish between product class knowledge and familiarity with the choice set. Familiarity will reduce search as information has already been processed and is stored in memory. For frequently purchased products, like most consumer non-durables, the decision process becomes automated. Familiarity with the choice set, therefore, works as a moderating variable on search. Experience with a setting or familiarity with a choice set has been found to have a negative effect upon amount of information acquisition (Lehman and Moore 1980; Sheth and Vanketesan 1968; Swan 1969). The problem with empirical studies (and surveys in particular) is that product expertise and familiarity with the choice set are highly correlated, and thus confound the relationship between expertise and search.

This article examines the effect of product expertise on acquisition of written and sensory information. The first type of information is written information available in brochures, advertisements and other descriptions of the product. (This is thus the type of information available in the traditional information display boards.) The second type (sensory) is information that requires some sort of inference and domain specific product knowledge in order to utilize it. We examine the relationship between product expertise, the cognitive processes underlying the problem solving task, and acquisition of written and sensory information.

HYPOTHESIZED MODEL

Most models of decision making include four steps that the decision-maker goes through: problem definition, information search, evaluation of the alternatives, and choice (Hansen 1972). An implicit assumption is that the decision process (related to a purchase) is a purposively goal directed behavior (Zaltman and Wallendorf 1979), and a type of "productive problem solving" (March and Simon 1958). It is further assumed that the decision process is iterative. A subject may, for example, redefine the problem after information search or after a number of evaluations.

We can perceive the problem definition phase as a process of generating hypotheses to be tested. In this stage, the decision-maker will address his or her memory for relevant information. Thoughts generated at this stage reflect the importance of attributes, relationships among attributes, definition of product use, and search or decision strategies. The problem framing thoughts define what to search for, and thus initiate overt search for written or sensory information cues.

Overt search behavior stimulates thoughts related to evaluations. These are thoughts reflecting actual characteristics of an alternative, evaluations of these properties, and overall evaluations of one or more alternatives. Evaluations may drive more thoughts about the problem definition, or more conclusive thoughts related to decisions about the choice set. This includes thoughts like acceptance or rejection of one or more alternatives, ranking of alternatives, intention, or final decision. Decision thoughts may thus induce further problem definition thoughts. The system of hypothesized relationships is provided in Figure 1 below.

The relationship between product expertise and information processing has received considerable attention in the area of consumer behavior theory. The vast majority of the contributions have focused on the amount of information acquired. The consumer informatio search literature indicates that product class knowledge encourages information search by facilitating search. That is, knowledge facilitates retrieval of relevant information from memory (e.g., Chase and Simon 1973; Tversky and Kahneman 1973; Voss, Vesonder, and Spilich 1980), and it facilitates integration of new information (e.g., Fiske, Kinder, and Larter 1983; Hayes-Roth 1977). A positive relationship between expertise and the amount of search has also been confirmed in several studies (e.g., Brucks 1985; Formisano et al. 1982; Jacoby, et al. 1978; Locander et al. 1980; Selnes and Troye 1989).

The literature further indicates that product expertise will reduce search because experts are better able to address information that is relevant for the problem to be solved (Jacoby, Troutman, Kuss, and Mazursky 1986; Taylor and Crocker 1980). One implication is that experts often base their evaluations on a different set of attributes (Alba 1983; Beattie 1983; Park and Lessig 1981). The facilitating and efficiency effects of expertise should act differently on search for written and sensory cues. Since the cognitive structure of experts is more complex and integrated, fewer written cues are needed in order to form an impression when compared to novices. Also, experts are better able to detect the most relevant cues. Thus, as shown in Figure 1, we hypothesize a negative effect of expertise on search for written cues.

The facilitating and efficiency effect of knowledge further suggests that experts are better able to infer intrinsic cues from a product. First, experts have the necessary competence to evaluate intrinsic cues from a physical inspection or trial experience. Second, as their cognitive structures are more integrated, experts are probably more motivated to seek out intrinsic information in order to validate hypotheses generated from extrinsic cues. Thus, as shown in Figure 1, we hypothesize a positive effect of expertise on search for sensory cues.

A number of studies have found that experts and novices differ with respect to how a problem is solved (De Groot 1965). Chi, Glaser and Rees (1981) found that novices were more or less unable to formulate the problem in a physics test, and suggested that this was the main reason for their poor performance. In think aloud studies, experts have been found to produce more statements related to problem definition and to have different objectives with the different steps in the decision process (Johnson 1981; Selnes and Troye 1989). Bouwman (1982) also found that non-experts were far less analytical in their search. They appeared to employ a strategy similar to induction, whereas the experts appeared to use a more deductive approach. Thus, product expertise is proposed to increase the number of problem definition thoughts.

It follows that expertise has an indirect effect on search through the decision process. Subjects who perceive a choice situation as cognitively complex (a large number of problem definition thoughts) is more likely to experience more uncertainty than others (Hansen 1972) are. In addition, more uncertainty will increase the motivation to search and thus increase overt search behavior. Thus, the indirect effect of product expertise on information search appears because experts generate more problem definition thoughts, which next generates a need for more information. Or stated differently, novices usually do not perceive a decision as difficult or complex because their ignorance prevents them from perceiving uncertainty, and thus they are not motivated to search out as much information as their comparative experts are.

METHODOLOGY

Subjects

The subjects were 87 business school students who were randomly selected from the student directory. Each individual was called, informed about the study and asked to participate. The respondents were motivated by a small payment for partcipation. Four students were excluded from the data analysis because they had recently bought the product, and thus had very clear preferences for the available alternatives. In the debriefing after the task, the respondents were asked not to discuss the task with fellow students. A total of 83 respondents successfully completed the task.

Selecting a focal product

To investigate the different types (i.e., written and sensory cues) of information searched during a decision process, it was important to use real products. As product class expertise may lead to automation (Alba and Hutchinson 1978), it was also necessary to use a product category with a high degree of innovation in terms of new brands, models, and features. A product satisfying this requirement was a portable stereo. We had no guarantee that the subjects would be unfamiliar with all models in the choice set, but the important point was that the choice set as a whole was perceived as unfamiliar (Swan 1969).

A manipulation check was conducted in order to test the variability in product class knowledge in the student population for the focal product. A sample was selected from a subgroup of students other than those who participated in the main study. This subgroup was asked to rate their perceived degree of product class knowledge for eight different products. We told the respondents that the results were to be used in a market survey of new products. Degree of perceived expertise was measured on a four point scale including no, little, some, and very high confidence. The distribution was approximately 10%, 20%, 50% and 20%. Thus, there was sufficient variation in the population. In addition, this variation was centered around medium product class knowledge with tails in both directions.

FIGURE 1

HYPOTHESIZED MODEL

Procedure and tasks

A laboratory setting with real products was constructed. Twelve different models representing seven different brands were used as product stimuli. These were arranged on a "wall," similar to the product display found in stores. To avoid possible choice-set effects, we randomly moved the positions of the models. Next to each model we placed a pamphlet containing the type of information usually found in brochures and advertisements for this type of product. In addition to brand and model name, there was information about price and technical features like frequency, power, and wow & flutter. There was also information about aspects such as whether the speakers could be detached, if a pick-up could be connected, if there was a built-in microphone, if there were plugs for earphones, if both batteries and net could be used, and weight. The respondents were free to listen to the radio and to three different cassettes with pop music.

In the laboratory the subjects were told to simulate a purchase where the chosen product should cover their own needs and fit their own budget. They were also asked to think aloud, that is, to tell the administrator what they were doing and thinking during the simulation. Search activities and think aloud statements, were recorded with a hidden video camera. On average the respondents spent 14.85 minutes to complete the choice-task. The standard deviation is 7.23 minutes, and thus a 95% confidence interval is from 13.3 to 16.4 minutes.

Information search

We wanted to measure the extent of search activities in a choice process. That is, the observable action consumers engage in to collect information from the environment. Search activities were observed and recorded during the simulation. A coding scheme with eight codes was developed (see Table 1). The information search activities were broken down into sequences corresponding to the activities I1 to I8. The means and standard deviations for the I-codes are reported in Table 1. As can be seen, I2 (look at information pamphlet) is the most frequent search activity.

A factor analysis of the I-codes gave two factors with eigenvalues larger than 1.00. The loading matrix is reported in Table 2. As expected, search for written information (I1 and I8) load on a different factor than the other search activities. Four of the six other search activities load high (above 0.7) on a common factor characterized by search for sensory information. Two items, I6 and I7, did not load high on the sensory search factor. We therefore decided to remove these two items from the rest of the analysis. In the analysis we also combined I4 and I5, adjustments of volume and tone. Sensory search was thus measured with three indicants, including the number of times the subject looked at one model (I1), listened to one model (I3), or adjusted the tone or volume (I4+I5). Written search in the case of this product was measured with two indicants, I2, the number of time a subject looked at a pamphlet, and I8, the number of comparisons of pamphlets made by a subject.

TABLE 1

CODES FOR INFORMATION SEARCH ACTIVITIES AND THEIR MEANS AND STANDARD DEVIATIONS.

TABLE 2

FACTOR LOADINGS FOR SEARCH-ACTIVITIES.

Product expertise

The traditional methods for measurement of cognitive complexity like Sentence completion tests (Schroeder and Streufert 1962; Streufert and Streufert 1978), Impression formation tests (Schroeder et al. 1967), Role concept repertory test (Kelly 1955), Listing and comparing (Scott, Osgood, and Peterson 1979), Multidimensional scaling (Torgerson 1958), Ratio estimation (Torgerson 1958), and Rating in test retest (Torgerson 1958) were all evaluated but found not to be applicable in the context of the cognitive complexity related to products. In terms of measurement we chose to focus on cognitive structure. In a complex cognitive structure there will be a rich number of dimensions, the level of each dimension will be advanced or abstract, and the dimensions will be highly interrelated (Streufert and Streufert 1978).

Product class knowledge or expertise was measured using three indicants. Two of these indicants were based on two questions that were administered before the simulation task. The first question asked the respondents to list attributes and/or features a person would use to evaluate a portable stereo in a purchase situation. The second question asked the respondents to list the attributes and/or features that are typically different between a low priced and a high priced product. The time limit was one minute on both questions.

Two raters were instructed to give each respondent a score on questions, #RAT1’ and #RAT2’. In the coding process, the first question for all respondents was evaluated first, and then the second question was evaluated. This was done in order to eliminate the effect of influence between the two evaluations. Inter-rater reliability on RAT1 was 0.80 and 0.82 on RAT2. RAT1 and RAT2 are thus represented as the average of the two raters’ evaluations. The instructions for the coders are provided in Table 3. The correlation between RAT1 and RAT2 was 0.57. The two measures were summed to form RAT, which is one of the three indicants of expertise. This score ranged from 2-9, with a median of 5.

The second indicant of expertise reflected the average complexity or level (LEV) of the attributes and features the respondents produced on the two questions. In order to assess the level of complexity, we first ordered the complexity of the attributes/features within a given dimension. The nine dimensions were appearance, functionality (design), price, amplifier, radio, cassette deck, speakers, sound quality, and others. The dimension related to amplifier will serve as an example. Attributes/features with a low level of complexity were reflected in answers such as "power," "effect," and "amplifier". More complex items might be "bass control," "loudness," "equalizer," "frequency range," or "sinus watt". All the items within each dimension were categorized into four degrees of complexity. LEV1 is the respondents average level of complexity of the items listed on the first question, that is, the complexity of attributes the subject would use to evaluate a portable stereo in a purchase situaton. LEV2 is the average on the second question, that is, the complexity of the attributes that are typically different between a low and high priced product. The correlation between LEV1 and LEV2 is 0.51. The two measures were summed and labeled LEV. The range of this variable was 1-3, with a median of 1.65.

TABLE 3

INSTRUCTIONS FOR CODERS

The third indicant of expertise was the understanding of general product terminology (TER). To avoid any queuing effect, this measure was obtained after the simulation. Care was taken not to provide any information during the simulation that could be used by the subjects to answer these questions. Brucks (1985), who found high correlation between this measure and other objective tests, used a similar measure. The subjects were asked to explain the meaning of ten terms or expressions such as "stereo," "tuner," "booster," "equalizer," etc. Answers were coded (2=correct, 1=partly correct, 0=not correct) and analyzed using principal component analysis. Using the scree-test, one factor was retained (eigenvalue 4.34 for the first and 1.22 for the second factor). The TER index is the respondent’s average score on the ten items. This score ranged from 0.1 to 2 with a median of 1.3.

Decision process

The respondents were asked to think aloud while completing the task. Their thoughts were recorded and written down in protocols. The thoughts produced are believed to reflect short-term memory processing and thus also internal deliberation (Bettman 1979; Ericson and Simon 1980). The protocols were then broken down into smaller blocks representing different aspects of the problem-solving task. We tried to break them down into as small blocks as possible (Newell and Simon 1972). Such procedures have been employed in consumer choice settings, so these were used to aid the development of the coding procedure (Bettman and Park 1980; Lussier and Olshavsky 1979; Payne and Ragesdale 1978).

Thought blocks were coded PROBACT if they were related to the definition of the problem. This not only included statements related to the objective of the decision or purchase, but also criteria used to make the decision and the strategy or planned procedure to reach the decision. Such thoughts were: importance of an attribute, preference for a specific level of an attribute, explanations of why an attribute is important, description of relationship between attributes, reference to long term memory, definition of product use, and decision strategies.

Thought blocks were coded EVALACT if they were related to evaluations of any of the models. Such thoughts were description of actual properties of a model, evaluation of a model’s properties, evaluation of properties, several models’ and overall evaluations of one or more models. Thought blocks were coded DECIACT if they were related to acceptance or rejection of the alternatives, that is, during the task several decisions were made. For some subjects, the models had to meet certain criteria to be further evaluated. For example, some subjects first screened out alternatives outside a given price range. Such thoughts were acceptance of one or more models, rejection of one or more models, ranking of models, intention or final decision.

The protocols were first divided into smaller blocks. Two coders cooperated on this task. The protocols were coded independently by the same two coders, and they initially agreed on about 80% of the codes. They discussed the thought blocks on which they disagreed until agreement was reached. The average number of problem definition thought blocks (PROBACT) is 19.45, with a standard deviation of 11.04. The average number of evaluation thoughts (EVALACT) is 20.88, with a standard deviation of 14.47. Number of decision type of thoughts is low compared to the other two types, with an average of 5.18 and a standard deviation of 2.98. Thus, we observe that on average, consumers spend about equally much thought activities in structuring and framing the problem to be solved (i.e. choose a portable stereo), as the amount of thoughts spent on evaluating information acquied.

RESULTS

The correlation matrix of the eleven variables is reported in Table 4. We see that all three expertise variables (RAT, LEV and TER) correlate with amount of problem framing thoughts. This is consistent with Spence and Brucks (1997) who argue that experts outperform novices in ill structured, but structurable problem situations because they can use existing knowledge to identify viable solution strategies. We further observe that the relationship between expertise and the other two thought categories, evaluations and decisions (EVALACT and DECIACT) are close to zero. An observation is that all thought categories are quite highly correlated (around 0.5). This indicates that thought activities of one kind tend to generate more activities of other kinds. Thus, if the consumer has many thoughts related to defining the problem, this will generate more evaluations and decisions.

TABLE 4

CORRELATION MATRIX OF THE LELVEN VARIABLES MEASURED.

FIGURE 2

ESTIMATED MODEL

The full model tested appears in Figure 2. The estimated coefficients (loadings) of the measurement model are reported in Table 5. This Latent Variable Structural Equations model was estimated for the covariance matrix using LISREL 7 (Joreskog and Sorbom 1988). The scale of measurement for each of the latent constructs with multiple indicants was established by fixing the loading of one of the indicants to 1.0. Three of the constructs in the model were measured by one indicant. For each of these, the loading was fixed at .9, with the associated measurement error term fixed at .19. This corresponds to an assumption of .81 reliability, and is equivalent to the initial percentage of agreement scores obtained by the two coders. It was felt that some moderate amount of measurement error more accurately reflected the true reliability of these measures than would the assumption of zero measurement error often used with single-indicant constructs (Howell 1987).

Following the two step procedure recommended by Anderson and Gerbing (1988), the adequacy of the measurement model was first assessed. The measurement model is simply a confirmatory factor analysis model, with all inter-construct correlation free to be estimated (zero degrees of freedom in the structural model). As noted by Anderson and Gerbing, this allows the measurement model to be adjusted as necessary while preserving the integrity of the test of the restrictions imposed by the structural model (of the relationships among the constructs), and establishes a baseline lack of fit against which the lack of fit resulting from constraints on the structural model can be compared.

TABLE 5

ESTIMATED COEFFICIENTS (LOADINGS) OF THE MEASUREMENT MODEL

The resulting measurement model (loading reported in Table 5) fit the data reasonably well, with a chi-square value of 42.54 with 35 degrees of freedom (p=.178), suggesting that the difference between the observed covariance matrix and the covariance matrix reproduced as a function of the parameters of the model can be attributed to sampling error. Further, there were no large modification indices associated with the fixed parameters, and all factor loadings were significantly different from zero. Thus, the model as depicted in Figure 2 was estimated. This model has four restrictions as compared to the measurement model: the effect of Expertise on Evaluations and Decisions is hypothesized to be indirect only, through Written and Sensory Search, which are hypothesized to have an indirect effect on Decisions, through Evaluations. With these restrictions, the fit of the model has a chi-square value of 44.07, with 39 degrees of freedom. The Chi-square difference test is 1.53 with 4 degrees of freedom, which is not significant at the .10 level, suggesting that the four restrictions imposed by the structural theory do not significantly alter the fit of the model. [Since the model is scale invariant, using the correlation matrix or the covariance matrix as input, results in proportional estimates of the parameters of the model and equal chi-square values. However, the estimates of the standard errors of the parameters can be biased when using the correlation matrix as input since they are based on the sampling distribution of the covariance matrix. Thus, the t-values reported are based on the unstandardized values computed from the analysis of the covariance matrix, although the parameter estimates reported are from the standardized solution (wherein all construct standard deviations are set equal to one), as estimated on the correlation matrix, for ease of interpretaion.]

As can be noted in Figure 2, two of the hypothesized parameters in the model are less than twice their respective standard errors. That is, they are not significantly different from zero at the .05 level. These are the paths frm Problem Definition to Sensory Search, and the covariance of the structural error terms between Written and Sensory Search (indicating that the partial correlation between Written and Sensory Search, given Expertise and Problem Definition, is within sampling error, zero.).

Again following Anderson and Gerbing (1989), these paths were fixed to zero and the model in Figure 2 was re-estimated. The reduced model has a Chi-square value of 47.63 with 41 degrees of freedom, which is again not significant at the .10 level. Further, as compared with the measurement model, the reduced model increases Chi-square by 5.09, with an additional 6 degrees of freedom, indicating strong support for the adequacy of the reduced model. The goodness of fit index for the reduced model is .93, and the Tucker-Lewis Index is .94. Thus, even though with the relatively small sample size employed here, sampling error could account for a substantial amount of lack of fit, leading to acceptance of a poor model The other fit indices suggest that the fit of the model is in fact adequate. [Using results from McDonald and Marsh (1990), the estimated model noncentrality is .08, and the index of centrality is .96. Both the index of centrality and the Tucker-Lewis index reported in the text (.94) are independent of sample size (McDonald et al. 1990; Gerbing and Anderson 1992). The Root Mean Square Error of Approximation (RMSEA) proposed by Steiger (1990) is .086, for which the hypothesis of   'close fit' cannot be rejected at the .10 level (Browne and Cudeck 1992).]

Summary

As predicted, experts seek more sensory (or intrinsic) information. Search for this type of information appears, however, to be driven by expertise only and not the decision process (Problem definition). A post-hoc interpretation is that this type of information can only be acquired if the subject has the necessary cognitive skills. Such skills are not developed during a decision process.

Experts seek less written (or extrinsic) information as compared to the novices. Search for written information is, as expected, driven by the amount of internal deliberation in terms of problem defining thoughts. As experts have more internal deliberation during their decision process, expertise will also affect degree of written search. Thus, a major finding is that the direct effect of expertise upon written search is negative, but the indirect effect (through complexity of the decision process) is positive. If we do not control for the effect of internal deliberation, an observed relationship between product expertise and written search may take any form. This can reconcile several of the conflicting observations in the search literature.

DISCUSSION AND CONCLUSION

The results provide evidence that product experts and novices differ in terms of decision making and information acquisition. Experts have cognitive skills that enable them to better search for sensory (or intrinsic) information cues. As novices lack these skills they are more likely to base their evaluations and subsequent choice on more written (or extrinsic) descriptions. In the case of portable stereos, some of the novices did not even listen to the product they selected. One implication for marketing is that product class knowledge may be important in terms of selecting the appropriate type of communication. If the degree of expertise varies considerably among the buyers, marketers should be careful in their market communication. If a company is selling a high-technology product, the prospective buyer must have considerable expertise in order to understand the benefits of the product, and thus be willing to pay a higher price. Electronic trade companies should probably focus more on products/services where decisions are made based on descriptive information (for example, financial services or travel reservations), and be more careful in selecting products/services where decisions usually are made on the basis of physical inspection and sensory search (for example, designer clothes and fresh food). The results should also have implications for personal sales or customer service. The salesperson should try to establish the customer’s degree of expertise, and based on this, employ different communication procedures.

A major finding in this study is that experts are beter at evaluating sensory information than novices, and thus search for more of this type of information. In contrast they search for less written information than novices do because they are more efficient. As suggested by Spence and Brucks (1997), the relationship between expertise and information search in problem solving is likely to be contingent on the decision environment. In the case of portable stereos, sensory information is difficult to judge and thus knowledge or expertise is required. In other situations (e.g. hotel cleanliness), sensory information is easy to judge. Similarly, sometimes written information takes expertise to decipher (e.g. computer features) and sometimes it doesn’t (e.g. calculator features). Thus, the relationship between expertise and search for sensory vs. written information may depend on the ambiguity of information. Thus, in situations where sensory information is ambiguous and difficult to judge, we expect a positive relationship between expertise and search for sensory information. This relates to a number of product categories in addition to stereo equipment, like for example food and beverages (e.g. find a good wine), sports equipment (e.g. find good down-hill skies), entertainment (e.g. find a good movie), clothing (e.g. find a comfortable sweater) and similar. Also, a context-specific finding is the relationship between expertise and problem framing activity. It is likely that our choice situation is similar to what Spence and Brucks (1997) would label ill structured, but structurable. This type of choice situation is likely to produce the strongest effect between expertise and problem defining thoughts as compared to well-structured or inherently unstructured environments. We believe future research should explore in more depth what makes sensory information ambiguous, and how different types of sensory-dependent decision environment affect search behavior.

The interrelationship between expertise, decision-making, and information search suggests that future research in this area should include the cognitive processes in their analysis. The methodology employed in this study suggests that the underlying decision process can be attached and analyzed. We believe that research in the future should focus on developing more advanced methods for analyzing cognitive processes, thus focusing more attention on the sequence of the elements in the decision process.

This study has only focused on the decision to acquire a product. Future research should also address the post-purchase differences between experts and novices, that is the "outcome" variable of the decision process. It is very likely that an expert and a novice will develop different expectations toward the product that they acquire. This can later affect their total satisfaction and subsequent loyalty. Another implication is that experts and novices will differ in their ability to evaluate a product’s performance, which will also affect the buyers’ satisfaction, and thus future loyalty.

REFERENCES

Alba, Joseph W. (1983): "The Effect of Product Knowledge on the Comprehension, Retention and Evaluation of Product Information," in Advances in Consumer Research, , Vol. 10, eds. R. Bagozzi and A. Tybout, Chicago, IL: Association for Consumer Research, 577-580.

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

Anderson, Ronald D, Jack L. Engeldow and Helmut Becker (1979), "Evaluating the Relationships Among Attitude toward business, Product Satisfaction, Experience and Search Effort," Journal of Marketing Research,Vol 16 (Aug), 394-400.

Anderson James C. and David W. Gerbrin (1988), " Structural Equation Modeling in Practice: A Review and Recommended Two-Step Approach," Psychological Bulletin, Vol. 103, No. 3, 411-423.

Beattie, Ann E. (1983), "Product Expertise and Advertising Persuasiveness," in Advances in Consumer Research, Vol. 10, eds. R. Bagozzi and A. Tybout, Chicago, IL: Association for Consumer Research, 336-341.

Bennet, Peter D. and Robert M. Mandel (1969), "Prepurchase Information Seeking Behavior of New Car PurchaseBThe Learning Hypothesis," Journal of Marketing Research, Vol 6, 430-433.

Bettman, James R. (1979), An Information Processing Theory of Consumer Choice, Reading, MA: Addison-Wesley.

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.

Bouwman, B.J. (1982), "Expert versus Novice Decision Making in Accounting; A Process Analysis," Working Paper, University of Oregon.

Browne, Michael W. and Robert Cudeck (1992), "Alternative Ways of Assessing Model Fit," Sociological Methods & Research, 21, 12 (Nov) 230-258.

Brucks, Merrie (1985), "The Effects of Product Class Knowledge on Information Search Behavior," Journal of Consumer Research, Vol. 12, (June), 1-16.

Bucklin, Louis P. (1966), "Testing Propensities to Shop," Journal of Marketing, 30, 22-27.

Chase, William G. and Herbert A. Simon (1973), "Perception in Chess," Cognitive Psychology, 4 (January), 55-81.

Chi, Michelene T.H, Robert Glaser, and Ernest Rees (1981), "Expertise in Problem Solving," in Advances in the Psychology, of Human Intelligence, ed. Sternberg, Hillsdale, NJ: Lawrence Earlblum.

De Groot, Adriaan D. (1965), Thought and choice in chess, The Hague: Mouton.

Ericson, K. Anders and Herbert A. Simon (1980), "Verbal Reports as Data," Psychological Review, Vol. 87, No. 3, 215-251.

Fiske, Susan, Donald R. Kinder, and Michael W. Larter (1983), "The Novice and the Expert: Knowledge Based Strategies in Political Cognition," Journal of Experimental Psychology, 19, 381-400.

Formisano, Roger A., Richard W. Olshavsky, and Shelley Tapp (1982), "Choice of Strategy in Difficult Task Environment," Journal of Consumer Research, Vol. 8 (March), 474-479.

Gerbing, David W. and James C. Anderson (1992), "Monte Carlo Evaluations of Goodness of Fit Indices for Structural Equation Models," Sociological Methods & Research, 21, 12 (Nov) 132-160.

Hansen, Flemming (1972), Consumer Choice Behavior, New York: Free press.

Hayes-Roth, Barbara (1977), "Evolution of Cognitive Stuctures and Processes," Psychological Review, Vol. 84, No. 3, 260-278.

Howell, Roy D. (1987), "Covariance Structure Modeling and Measurement Issues: A Note on "Interrelations Among a Channel Entity’s Power Sources," Journal of Marketin g Research, Vol. 24 (Feb), 119-126.

Jacoby, Jacob, Robert W. Chestnut, and William A. Fisher (1978), "A Behavioral Approach in Non-Durable Purchases," Journal of Marketing Research, 15, 532-544.

Jacoby, Jacob, Tracy Troutman, Alfred Kuss, and David Mazursky (1986), "Experience and Expertise in Complex Decision Making," Advances in Consumer Research, Vol. XIII, ed. J. Lutz, 469-472.

Johnson, Eric J. (1981), "Expertise in Admission Judgement," Working Paper, Carnegie-Mellon University.

Johnson, Eric J. and J. Edward Russo (1984), "Product Familiarity and Learning New Information," ournal of Consumer Research, 1 (June), 542-550.

Joreskog, Karl and Dag S├Ěrbom (1988), "LISREL 7: A Guide to the Program and Application," Chicago, SPSS, Inc.

Katona, George and Eva Mueller (1955), "A Study of Purchase Decisions in Consumer Behavior," in Consumer behavior: The dynamics of consumer reaction, Vol. 1, ed. L.Clark, New York: New York University Press, 30-87.

Kelly, George A. (1955), The Psychology of Personal Constructs, New York: Norton.

Kiel, Geoffrey C. and Roger A. Layton (1981), "Dimensions of Consumer Information Seeking Behavior," Journal of Marketing Research, May, 233-239.

Lehman, Donald R. and William Moore (1980), "Validity of IDB: An Assessment Using Longitudinal Data," Journal of Marketing Research, November.

Locander, William B. and Peter W. Herman (1980), "The Effect of Self-Confidence and Anxiety on Information Seeking in Consumer Risk Reduction," Journal of Marketing Research, Vol. 16 (May), 268-274.

Lussier, Denis N. and Richard W. Olshavsky (1979), "Task Complexity and Contingent processing in Brand Choice," Journal of Consumer Research, Vol.6, 154-165.

March, James G. and Herbert A. Simon (1958), Organizations, New York: Wiley.

McDonald, Roderick P. and Herbert W. Marsh (1990), "Choosing a Multivariate Model. Noncentrality and Goodness of Fit," Psychological Bulletin, 107, 247-255.

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

Newell, Allan and Herbert Simon (1972), Human Problem Solving, Engelwoods Cliff, NJ: Prentice Hall.

Newman, Joseph and Richard Staelin (1972), "Prepurchase Information Seeking for New Cars and Major Household Appliances," Journal of Marketing Research, Vol.9, 240-257.

Park, C. Whan and V.Park Lessig (1981), "Familiarity and Its Impact on Consumer Decision Biases and Heuristics," Journal of Consumer Research, Vol. 8 (Sep), 223-230.

Payne, John W. and Easton Ragesdal (1978), "Verbal Protocols and Direct Observation of Supermarket Shopping Behavior: Some findings and a discussion of methods," Advances in Consumer Research, Vol. V, ed. K. Hunt, 571-577.

Punj, Girish and Richard Staelin (1983), "A Model of Consumer Information Search Behavior for New Automobiles," Journal of Consumer Research, 9 (March), 366-380.

Rao, Akshay R. and Wanda A. Sieben (1992), "The Effect of Prior Knowledge on Price Acceptability and the Type of Information Examined," Journal of Consumer Research, Vol. 19 (Sep), 256-270.

Schroeder, Harold M., Michael J. Driver and Siegfried Streufert (1967), Human Information Processing, New York: Holt, Rineheart and Winston.

Schroeder, Harold M. and Siegfried Streufert (1962), " The Measurement of Four Systems of Personality Structure Varying in Level of Abstractness: Sentence Completion Method," Princeton University: ONR Technical Report.

Scott, William A., D.Wayne Osgood, and Christopher Peterson (1979), Cognitive Structure: Theory and Measurement of Individual Differences, New York: John Wiley.

Selnes, Fred and Sigurd V. Troye (1989), "Buying Expertise, Information Search and Problem Solving," Journal of Economic Psychology, Vol. 10, No.3, 411-428.

Sheth, Jagdish and M. Vanketesan (1968), "Risk Reduction Processes in Repetitive Consumer Behavior," Journal of Marketing Research, Vol. V (Aug), 307-310.

Spence, Mark T. and Merrie Brucks (1997), "The Moderating Effects of Problem Characteristics on Experts and Novices’ Judgements," Journal of Marketing Research, Vol. XXXIV, May, pp.33-247.

Steiger, J.H. (1990), "Structural Model Evaluation and Modification: An Internal Estimation Approach," Multivariate Behavioral Research, 25, 173-180.

Streufert, Siegfried and Susan C. Streufert (1978), Behavior in the Complex Environment, New York: John Wiley.

Swan John E. (1969), "Experimental Analysis of Predecision Information Seeking," Journal of Marketing Research, 6, 62-65.

Taylor, S.E. and J. Crocker (1980), "Schematic Bases of Social Information Processing," in Social Cognition: The Ontario Symposium, Vol. 1, eds. E.Tory Higgins, C.Peter Herman and Mark P. Zanna, Hillsdale, NJ: Earlbaum.

Torgerson, Warren S. (1958), Theory and Methods of Scaling, New York: Wiley.

Tversky, Amos and Daniel Kahneman (1973), "Availability: A Heuristic for Judging Frequency and Probability," Cognitive Psychology, 5, 207-232.

Tyebjee, Tyzoon T. (1979), "Response time, Conflict and Involvement in Brand Choice," Journal of Consumer Research,, Vol. 6 (Dec), 295-304.

Urbany, Joel, Peter R. Dickson, and William L. Wilkie (1989), "Buyer Uncertainty and Information Search," Journal of Consumer Research, Vol.16, No. 2, (Sep), 208-215.

Voss, James, Greg Vesonder, and George J. Spilich (1980), "Text Generation and Recall by High-knowledge and Low-knowledge Individuals," Journal of Verbal Learning Verbal Behavior, 19, 651-667.

Zaltman, Gerald and Melanie Wallendorf (1979), Consumer Behavior; Basic Findings and Management Implications, New York: John Wiley.

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