Industrial Search Behavior and Perceived Risk



Citation:

Thomas P. Copley and Frank L. Callom (1971) ,"Industrial Search Behavior and Perceived Risk", in SV - Proceedings of the Second Annual Conference of the Association for Consumer Research, eds. David M. Gardner, College Park, MD : Association for Consumer Research, Pages: 208-231.

Proceedings of the Second Annual Conference of the Association for Consumer Research, 1971     Pages 208-231

INDUSTRIAL SEARCH BEHAVIOR AND PERCEIVED RISK

Thomas P. Copley, Wilkes College, Wilkes-Barre, Pa.

Frank L. Callom, GTE Sylvania Inc., Towanda Pa.

[Assistant Professor of Business Administration. Washington State University, Pullman, Washington. Effective August 1971.]

INTRODUCTION

One of the many fertile hypotheses suggested by Howard and Sheth's Theory of Buyer Behavior (1969) is the notion that the perceived risk of a purchase situation is strongly related to the search for information engaged in by buyers. The main purpose of this study is to test this hypothesis in an industrial marketing setting.

Anyone generally familiar with the pioneering work of Bauer, Cox, Cunningham, and others at Harvard (Cox, 1967) on the manifestations of perceived risk and information handling will note the importance of the theoretical foundations of Howard and Sheth's contention. Indeed, the latter researchers suggest that perceived risk may be an analog of a more generic construct, such as that of stimulus ambiguity, developed by Berlyne, K. C. Montgomery, and others (Howard and Sheth. p. 259). The hypothetical relation is represented in Figure 1.

The Berlyne Function

Concerning the X1 to X2 segment of the curve in Figure 1, the buyer is motivated to obtain more information, thereby reducing the stimulus ambiguity (perceived risk) to the left. The X, to X3 portion of the curve is called 'supramaximal inhibition'. From an industrial buying viewpoint, this might be regarded as a zone in which the risk of a particular situation appears so great that the purchasing agent tends to freeze up rather than to seek information about new sources of supply that would reduce the perceived risk. ln such cases the agent's information seeking would drop markedly, thus moving from right to left on the curve. To the left of X1 the ambiguity (perceived risk) is so small that the industrial buyer increases his search behavior due to curiosity and a need to relieve the monotony of continually using the same sources of supply. The present study tests the existence of this 'two bump' (two local maximums)relationship between the perceived risk and industrial search described in Figure 1.

Styles of Risk Perception and Search Patterns

A secondary objective of this study is to determine whether or not there are distinct styles of risk perception. and whether or not these styles, if they exist, in any way affect purchasing agents' search for information.

Several writers have suggested conflicting theories in an effort to explain how industrial buyers search for information. For example, Webster (1965) says, "The search process starts with an evaluation of goals and if the present state of goal attainment is satisfactory, there is no need to search." From this one can see that if goals are raised the search process will be activated. However. any search effort has high cost and time constraints. Webster suggests that one stops searching when the value is less than the cost of gathering information.

Ammer (1968) disagrees that the buyer is limited in his search process at all. He feels that good buyers go to great pains to develop new and untried suppliers. even when there is nothing new or revolutionary about the product. if for no other reason than to stimulate competition. He says. "Buyers should always be in the process of developing new suppliers in an effort to lower prices. assure continuity of supply. consistency of quality, and to gain favorable relations with suppliers."

However, the work of Bauer (1960) disagrees somewhat with both Webster and Ammer. Citing the work of Katz. Menzel, and Coleman on the adoption of new drugs by physicians. he says that people concentrate on ways to reduce risk after a decision is made and seek out information that confirms the wisdom of their decision. People change their own responses to bolster their perception of the desirability of their actions.

The work of Wilson (1970) suggests a reconciliation of the divergent views on the search process. He would say that there are distinct 'styles' of search behavior which are related to personality variables. such as the need for certainty. the need to achieve, and generalized self confidence. In his study three decision styles were evident. There was (I ) a normative group which generally tended to make choices congruent with a normative decision model (analogous to Webster's view) and. (') a conservative group which tended to avoid uncertainty and large negative outcomes (the Bauer argument). Wilson also found (3) a group which appeared to have more vague behavior which could neither be categorized as 'conservative' nor 'normative'. We suggest that this group represents one which tends to have a wider range of search strategies that would closely approximate the Berlyne curve of Figure 1.

Under some conditions (e.g. low risk) the switchers would undertake considerable search for 'curiosity' type motives. as cited above by Ammer. In conditions of medium risk they would undertake search patterns more consistent with 'rational' or economic considerations - that is. they would respond to price and product characteristics as opposed to perceived risk. [Cardoza and Cagley (1971) found that generalized risk was associated with selection of vendors. but that risk specifically associated with price or product differences was not influential on vendor selection. This indicates that there are not strong interaction effects between the buyer s risk perception and his price/product perception.] With higher risk perception, the switcher would act to reduce risk by increasing his information up to a point of overload. At the overload or 'supramaximal inhibition' point, information seeking would decline. In fact. if one accepts the Berlyne relation as a valid explanation of the relation between perceived risk and search, that of a "switcher" would be the most normal style of decision-making. Conservative or normative styles would be deviant from the norm. (A style which is normal. of course, is not necessarily the most effective. )

This study will attempt to show that there are distinct patterns of search behavior. and that these styles are correlated with styles of risk perception.

Organization of the Paper

This paper is organized into six major parts: (1) the data base, (2) a description of the instruments. (3) methodology, (4) hypotheses and results, (5) alternative explanations for data. and (6) implications for further consumer research.

DATA BASE

The data for this study were obtained from questionnaires returned through the mail by 71 purchasing agents belonging to two active chapters of the National Association of Purchasing Management. both located in the northeastern United States. Of 409 listed members in these two chapters. 200 were mailed questionnaires. Seventy-one (71), or 35%, of the 200 members contacted returned usable questionnaires.

A secondary questionnaire was necessary to obtain independent judgments by a second, but presumably similar, group of purchasing agents. One hundred (100) members of the same two northeastern V.S. chapters, who had not received the main questionnaire, were distributed copies of the secondary questionnaire. Thirty (30) of the 100 questionnaires were returned and usable. In both cases the cover letter accompanying the questionnaire was made purposely vague in regard to the objectives of the study in order to avoid engendering a positive response set or bias. Given the meager information that the cover letter supplied, and the lengthiness of the forms, the relatively high rate of return, that is, 35%, and 30%, is surprising.

THE INSTRUMENTS

The instruments were the main questionnaire and the secondary questionnaire. The former was used for measuring risk and search, and the latter was used only for scaling, that is, to determine the scale values of the search strategies.

The Case Used as the Basis for Responses

Both questionnaires used the same hypothetical purchasing case, citing the "DeLar Industries" It read as follows:

"DeLar Industries produces metals and chemicals for the electrical and electronic industry. Recently the company developed a new process for manufacturing a copper metal powder more efficiently. The plant producing this powder has now been in operation for six months.

"Ted Tallon, the purchasing manager for the division, established the sources for the raw materials used in the process. The major chemical used in the process is soda ash. It would have been possible to use caustic soda: however, at the time the availability and price were more favorable to the use of soda ash. During the first six months of operation there were no major difficulties with either the price or availability of soda ash. "

Directions accompanying the main questionnaire instructed the purchasing agent to play the role of Ted Tallon in answering twelve situational variants of the main case. (See Figure 2.) Since the secondary questionnaire involved the scaling of industrial search, slightly different instructions were given. For each situation area, the respondents to the secondary questionnaire were asked to indicate their assessment of the degree of 'search' represented by each strategy that DeLar could adopt. The instructions stated that search could be formally defined as:

"The systematic investigation and development of alternative sources of supply, selection, and new products consistent with purchasing objectives and company goals. "

Each of the twelve situational variants used in the main questionnaire included two five-point rating scales. one for certainty and one for consequences, respectively. Four or five search strategy alternatives were provided for each variant. Each situational variant in the main questionnaire had its counterpart in the secondary questionnaire. However, the rating scales for scaling search, which are described in the next section, were quite different. The similarities and dissimilarities between the main and secondary questionnaires will become clear after examination of Figures ' and 3.

Scaling of Perceived Risk and Industrial Search

In order to maintain continuity with the earlier Harvard work, a technique almost identical with Cunningham s operational measure of perceived risk was used (Cox. p. 84-86). Measures of certainty and consequences (danger) were obtained on five category rating scales. The number one was arbitrarily assigned to the first category, two to the second, etc. The product of the certainty rating and that of the consequences rating was taken as a measure of risk. In spite of some methodological problems with the scaling procedure, as noted by Cunningham, this method has been assumed in the past to be reasonably indicative of the perceived risk variable. (See Figure 9. )

The scaling of industrial search called for a different approach. The purchasing agents were given a set of 4 or 5 search strategies per question and were asked to choose the one which they were most likely to adopt given the purchase situation. (Also see Figure 2.)

The method adopted for obtaining scale values of search for each strategy was the method of successive intervals first introduced by Thurstone (see Saffir, 1937), and later refined by Edwards and Thurstone (1952) and Torgerson (1958). This method involves the use of a separate set of "judges" who do not fill out the main questionnaire; rather, their only contribution is to rate each strategy according to how much "search" they believe it represents. To some extent. then, the validity of the scale values is dependent upon the judges having been drawn from exactly the same population as the respondents to the main questionnaire. Scale values obtained in this manner have been found to be highly valid, even when as few as 15 judges were used per group (Rosander, 1936).

Independent judgments are obtained for the values of items, such as, in our case, the (a), (b), (c). and (d) strategies. In this study these strategies were sorted into a series of eleven categories. Cumulative frequencies were obtained for the number of times a strategy was placed in each of the eleven categories. The cumulative frequencies were converted to their normal deviates. and then the strategies were scaled according to the formula:

EQUATION

where:

Si = the value of the ith search strategy

l = the lower limit of the interval on the 'psychological continuum' on which the median falls

SPb = the sum of the proportions below the interval in which the median falls

Pw = the proportion within the interval in which the median falls

wj = the width of the interval on the 'psychological continuum

Despite the criticism that there are confounding effects associated with the judges being presented with a different task than the subjects, the method of successive intervals seems to be a reliable scaling procedure. (See Scott. 1968).

METHODOLOGY

A polynomial regression was run. treating all values of risk and search across subjects as separate cases. In order to demonstrate a two bump curve similar to the one in Figure 1, a polynomial regression equation of the form

Y = b0 + b1X + b2X2 + b3X3

would have to be statistically significant. (In general. a curve with one "bump" has the equation Y = b0 + b1X = b2X2. For each successive term of higher degree added, such as b3X3 or b4X4, the curve acquires one more "bump".) However, the particular routine which was used (Dixon, 1968) proceeds in a step-wise fashion. First it fits a linear relation, then a quadratic, then a cubic, until the user specifies that it stop. At each stage an analysis of variance is performed and an F ratio is calculated.

In order to isolate distinct styles of risk perception, a Q-type factor analysis with orthogonal rotation was performed on the risk data. In other words, each subject was considered to be a variable, and each of the twelve situational areas to which the subjects responded was regarded as an observation on the variables. The effect of this type of factor analysis is to group subjects into distinct groups according to their loadings on each factor. Four factors were rotated. An arbitrary minimum cut-off point was set for deciding whether or not to include a subject in a given factor. Only if his loading was equal to or greater than 0.50 with, say, factor 1, but lower than 0.40 with any other factor was he included.

Polynomial regressions were run for each significant risk factor. In addition, an R factor analysis was run. The risk measure for each of the twelve situational areas was taken to be a variable, and the responses of each of the subjects was taken to be an observation. Regressions were run on each factor of any significance.

RESULTS AND HYPOTHESES

Hypotheses

The basic hypotheses of this study were:

1) That a Berlyne shaped curve would fit the relationship between perceived risk and search behavior: and

2) That there would be specific types, that is, groups, of risk perceivers, and that each group would demonstrate a characteristic search behavior different from, yet consistent with, the Berlyne curve.

Design Considerations

The polynomial regression program adopted in this study was taken from the University of California Biomedical Program Library (BMD 05R) edited by Dixon (1968). The program in question allows only a maximum of 500 observations on each variable. Since 71 purchasing agents returned questionnaires, and since there were 12 responses per questionnaire, there was a total of 12 x 71 = 852 responses - 352 more than could be analyzed. It would. of course. be possible to eliminate 352 responses on a random basis, thus bringing the total down to 500. However. if certain of the twelve questions evoked a radically different set of responses from the norm, a random selection would not ensure representativeness.

It was decided to factor-analyze both the 12 risk questions and the 12 search questions. If any distinct factors were determined on either variable, a random selection of responses would be precluded. The results of the two factor analyses are presented in Table 1.

The factor analyses indicate that it is not possible to group the questions together on the basis of search. The first factor on the search variable accounted for only 10% of the variance and only had one question which loaded on it. On the other hand, the first factor in the risk factor analysis accounted for 32% variance and had three questions which loaded on it. The mean rating on it for risk was 15.74, which was considerably higher than the overall average of 11.48. (Range was 1 to 25.)

As a result, the data were divided into two groups of questions - (I) all of the responses for variables loading high on the first risk factor, which will be called the "high risk " group of questions, and (r) all other responses, which will be called the "normal risk" group.

The curve fit to the "high risk" factor appears in Figure 4. The means, correlation co-efficients. ANOVA tables, et cetera, are given in Table 2. Similarly, the curve fit to the "normal risk" questions appears in Figure 57 and Table 3 presents the data on it.

Figure 4, representing the "high risk" questions, does exhibit some of the characteristics of the Berlyne curve, such as two local maximums. However, the curve is dissimilar in some important respects. No overload point is observed beyond which search declines. Also, at low ranges of risk, search actually declines. The most disturbing quality of this group, though, is the almost zero correlation co-efficient between risk and search (R / = 0.001 ).

The second set of data, representing the "normal risk" questions, exhibits a higher correlation co-efficient (R2 = 0.289); however. the curve fit to the data is nowhere close to the Berlyne curve. The curve fit to the data (Figure 5) is almost linear with a slight bow to it. Searching only gradually increases with the risk.

At this point it could not be argued that the data generally supports the Berlyne formulation of hypothesis 1. However, the authors believe that the low correlation co-efficient and the unexpected shape of the curves can be explained as the result of an artifact of composition. that is, the result of regressing several heterogeneous groups of risk perceivers. The effect of this would be to confound the actual relationship between risk and search.

The next section is devoted to analyzing the results of segmenting the respondents into distinct groups of risk perceivers. It will be seen that the effect was to remove some of the 'noise from the data. and to provide results more clearly consistent with the hypotheses.

The Q-Factor Analysis of Subjects

A factor analysis of perceived risk was conducted with varimax rotation of four factors. Each subject was considered to be a variable, and each of the twelve questions was considered as a separate observation. Only 64 subjects were included, because several were eliminated for not answering a complete set of questions. The results of the factor analysis are presented in Table 4.

The fact that the first four factors account for such a large percentage of the total variance on the risk variable strongly supports the division of subjects into groups of distinctly different risk perception. In fact. the first ten (10) factors in the analysis accounted for a total of 98% of the variance. Additionally. the pattern of variable (subject) loadings was very unambiguous - subjects tended to exhibit loadings of 0.70-0.99 on one of the rotated factors, and 0.35 or less on all other factors.

The relationship of these risk perception groups to search behavior will be the subject of the next section

Polynomial Regressions for Each Group of Risk Perceivers

On the basis of their loadings on the Q-factor analysis, the subjects were divided into five groups. The first four groups were composed of those subjects, respectively, who unambiguously loaded on the first four factors. The fifth group, the largest, was composed of those subjects who did not clearly load 011 any of the first four factors. Table 5 indicates the composition of each of the five groups.

For each of the five groups, referred to as Groups I - 5, search was regressed onto risk by the polynomial regression routine. The analysis of variance for each of these regressions are presented in Tables 6a through 6e. Figures 6A through 6E present the line of regression fit by the routine to each group of data.

The results of the regressions for each of the five risk groups were consistent with both the Berlyne curve hypothesis and the notion that style of risk perception affects the search pattern. Although the means of risk and search for each of the five groups were fairly similar, the frequency distribution of risk, that is, the skew towards the low risk (left)end of the frequency distribution. (See Figures 6a and 6b.) For reasons which will be explained in the following pages, these two groups were dubbed the "search simplifiers." (2) Group number 3 had a risk frequency distribution which approximated the normal curve. (See Figure 6c.) This group was called the "search norms". (3) Finally, groups 2 and 5 had a skew to the high end of the risk scale. (See Figures 6d and 6e.) These two groups were called the "searchers: mirror image groups". The results of the regression analyses follows.

The Search Simplifiers

The search simplifiers, who are made up of groups I and 4, tend to be low risk perceivers. Their means are lower than the mean for all groups, and their modes are considerably lower than the average. The mode in each case was 4, versus the overall average for all groups of 11.48. The type of search curve was very similar for both groups I and 4. (See Figures 6A and 6B.) The simplifiers' search curves increased very gradually as the risk was increased, and at the higher values of risk the amount of searching leveled off and declined. The most obvious thing about these curves is the very slow rate of change of search and the absence of inflection points similar to the Berlyne curve. One might argue that this group is similar to what Wilson (1970) called the "conservative" group. They seem to have no curiosity 'bump' at the low levels of risk. Instead, they seem to be motivated to seek more information up to a certain level of risk. At that point of risk overload, the simplifiers reduce their information intake

Conservative often means a disposition to do things in traditional ways. The search process, as Bauer (1961) notes. often fulfills the need to justify previous decisions, that is, in the case of a conservative purchasing agent, decisions made according to traditional criteria. In this case the agent is likely to be examining and reexercising his well worn criteria and searching to find more support for them. At levels of high risk the criteria he is using are likely to be challenged the most. To avoid the possibility of discovering new information which might enlarge the challenge to his traditional criteria, he reduces his search for new information. In other words. he simplifies the situation by avoiding new information. The fact that these persons tend to perceive risk as low could be accounted for by the fact that they refuse to look at the whole range of consequences which might result from their purchase decisions. We might further speculate that these individuals are likely to believe that observance of "tried and true" decision rules. or "rules of thumb ", will always minimize the possibility of harmful consequence.

The Search Norms

"Search norms" is the term used to designate group 3. (See Figure 6C.) The mean (10.94) of this group on risk was only 54/100 of a unit of risk different from the overall mean of 11.48. The mode of the group was 12, which was even closer to the overall mean. The type of search curve observed was very similar to the Berlyne curve illustrated in Figure 1. It had a characteristic bump at the low end of the search scale, representing the curiosity motive; at the higher levels of risk search increased and then eventually dropped off at the point of supramaximal inhibition. The curve appears to be a good deal flatter than the theoretical curve of Figure 1, as indeed do all of the observed curves. The reason for this flatness may be largely attributed to the fact that search is scaled on a ranee of 0 to 3.2, while risk is scaled on a ranee of 1 to 25.

The search norms are probably the same group to which Wilson (1970) refers as the "switcher" group. This group tends to seek information for its own sake at lower levels of rick, but at the higher levels search behavior is consistent with that of the search simplifiers, or conservatives. The search activity at low levels of risk is similar to what Ammer (1968) meant when he said that buyers go to great pains to develop new and untried suppliers, even when there is nothing new or revolutionary about the product, if for no other reason than to stimulate competition.

The fact that this group accounted for only 8% of the total variance indicates that in the purchasing profession most agents do not fall into the switcher category. On the contrary, they tend either to be search simplifiers or searchers.

The Searchers: Mirror Image Groups

The searchers were called "mirror image" groups because their search curves (see Figures 6D and 6E) were mirror images of the familiar Berlyne curve. This fact suggests that these individuals tend to regard as risk just the opposite of what would be expected. It will be recalled that the original justification for the hypotheses stated that perceived risk should be considered an analog of "stimulus ambiguity". Thus the higher the perceived risk, the higher the stimulus ambiguity. For the searchers, however, it appears that a low risk situation represents a greater ambiguity than a high risk situation.

The risk means for each of the searcher groups were higher than the overall mean: 12.51 and 11.25 versus 11.48. The modes in each case were 16. (See Figures 6d and 6e.) The fact that the searchers see all situations as more risky may be the key to why they exhibit a mirror image search curve. These purchasing agents may fit the model suggested by Webster (1965): that is, "the search process starts with an evaluation of goals, and if the present state of goal attainment is satisfactory, there is no need for search." These persons' curiosity motive would be activated only when goals are higher than goal attainment. This would certainly be a situation in which there was very high risk. However, to the searcher, such a clear gap between goals and goal attainment would be a very unambiguous situation. Hence, their curiosity would be aroused.

As the risk becomes smaller, goals and goal attainment would be closer together Ambiguity would increase because the searchers would no longer have a clear goal at which to aim. Paradoxically, as the risk declines even more, the response level increases. This might occur when the searcher begins to feel too comfortable. At the very lowest level of risk the searcher's search would fall off completely, as he would seek to reduce the information overload of being concerned with a situation in which his goals and goal attainment are in perfect harmony. The searcher then is primarily an active person in his outlook. His idea of ambiguity is a situation in which he can't act. In contrast. the search simplifier, or conservative, is basically passive in his outlook. To him an ambiguous situation is one in which he must act.

Summary of Results

The results of this study were generally consistent with the Berlyne relationship between perceived risk and search suggested by Howard and Sheth (1969). Through a Q-factor analysis of perceived risk 5 distinct groups of risk perceivers were isolated These 5 groups exhibited distinct stvies of search behavior. On the basis of their search behaviors the five groups were reduced to three groups of distinct risk search inter-relationships. These three grouPS were labeled (1) the "search simplifiers", (2) the "search norms", and (3) the "searchers: mirror image groups". The search simplifiers exhibited an almost linear curve with slight positive slope; it had a slight bow to it, making it concave to the x-axis. The search norms had a curve identical to the Berlyne relationship of Figure 1. The searchers had a curve that was the mirror image of the Berlyne curve. The frequency distributions of risk for each of these three groups was consistent with an explanation of their shape.

ALTERNATIVE EXPLANATIONS FOR DATA

The rule of parsimony for use with polynomial regression is to drop from the equation those terms of higher degree which do not enlarge the overall significance of the equations. Examination of Tables 6a through 6f will demonstrate that. with one exception, the equations were significant at the 0.01 level or below for each stage (degree) of the equation for each of the five groups. This implies that the quadratic and cubic terms should be dropped from every equation considered. because they do not contribute meaningfully to the explanation of the data In other words, a strong case could be made for saying that all relationships observed were linear. Acceptance of the linear point of view would necessitate rejection of the Berlyne curve explanation.

The best counterargument for the linear claim is a consideration of the presence of a considerable "noise", that is, random, unexplained variance, in the data. The direct mail questionnaire. when filled out by busy executives in their offices. is an inherently fallible tool. Interruptions to the executive by his secretary and others, insufficient time to study each question, and the distractions of work pressure are likely to create such random error. In such a case the "noise' is likely to take the form of considerable scatter in the data points.

Polynomial regression, like linear multiple regression, uses the least square technique of fitting the line of regression to the data points. When the data points are widely scattered the polynomial regression losses much of its sensitivity to patterns in the data. In such cases a simple linear equation is likely to be as significant, statistically speaking, as a polynomial one. Nevertheless, it is not then possible to accept a null hypothesis of no polynomial equation. One can only say that no polynomial was observed.

The strongest support for the hypothetical Berlyne relationship between risk and search is the curves themselves (Figure 6A through 6E). They represent least square cubic fits of lines to the data. In other words, each curve represents a cubic polynomial which is drawn such that no other cubic curve better fits the data. This being the case, it would be an amazing coincidence to obtain 5 curves which so closely approximate our a priori expectations for each of the five data groups.

IMPLICATIONS FOR FURTHER CONSUMER RESEARCH

From the viewpoint of purchasing management, certainly the most important implication of this research has to do with the three styles of risk-search interaction. The question should be asked as to which of the three styles, the simplifier, the norm, or the searcher, tends to be the most effective in specific situations. For example, would it be better to have a searcher or a simplifier if the company is rapidly expanding its facilities? Each style would be likely to have its own peculiar weaknesses and strengths.

Marketers would, of course, be concerned with risk perception styles, particularly from a segmentation viewpoint. The Berlyne curve relationship also suggests points to be considered in personal selling to purchasing agents. For example, a searcher is more likely to respond to a moderate or high pressure sales approach than to one which is extremely low key. An industrial salesman using a low key approach on a searcher should be careful that "low key" does not become synonymous with low risk. On the other hand, a search norm is likely to respond well to a low risk - low key sales approach.

From the standpoint of consumer research, the fact that perceived risk can be tucked into the stimulus ambiguity paradigm suggests that other constructs with which we have been working might fit the mold too. Some possible candidates are source credibility, persuasive fear, divided attention, and cognitive clarity. Perhaps a more generic understanding of these constructs is a building.

TABLE 1

R-FACTOR ANALYSES OF RISK AND SEARCH: PROPORTION OF VARIANCE ACCOUNTED FOR AND FACTOR LOADINGS

TABLE 2

ANALYSIS OF VARIANCE FOR "HIGH RISK" FACTOR

TABLE 3

ANALYSIS OF VARIANCE FOR "NORMAL RISK" FACTOR

TABLE 4

Q-FACTOR ANALYSIS OF RISK ACROSS SUBJECTS: PROPORTION OF VARIANCE ACCOUNTED FOR AND FACTOR LOADINGS

TABLE 5

COMPOSITION OF THE FIVE GROUPS OF SUBJECTS

TABLE 6a

ANALYSIS OF VARIANCE FOR GROUP 1 - THE "SEARCH SIMPLIFIERS"

TABLE 6b

ANALYSIS OF VARIANCE FOR GROUP 4 - THE "SEARCH SIMPLIFIERS"

TABLE 6c

ANALYSIS OF VARIANCE FOR GROUP 3 - THE "SEARCH NORMS"

TABLE 6d

ANALYSIS OF VARIANCE FOR GROUP 2 - THE "SEARCHERS: MIRROR IMAGE GROUP"

TABLE 6e

ANALYSIS OF VARIANCE FOR GROUP 5 - THE "SEARCHERS: MIRROR IMAGE GROUP"

FIGURE 1

THE BERLYNE FUNCTION

FIGURE 2

TYPICAL FORMAT USED FOR THE 12 SITUATIONAL AREAS OF THE MAIN QUESTIONNAIRE

FIGURE 3

TYPICAL FORMAT USED FOR THE 12 SITUATIONAL AREAS OF THE SECONDARY QUESTIONNAIRE

FIGURE 4

CURVE REPRESENTING THE "HIGH RISK" QUESTIONS

FIGURE 5

CURVE REPRESENTING THE "NORMAL RISK" QUESTIONS

F1GURE 6A

LEAST SQUARES CUBIC FIT OF GROUP 1. THE FREQUENCY CURVE FOR RISK IS SKEWED TO THE LEFT.

FIGURE 6B

LEAST SQUARES CUBIC FIT OF GROUP 4 DATA. THE FREQUENCY DISTRIBUTION APPEARS TO BE BIASED TO THE LOWER VALUES.

FIGURE 6C

LEAST SQUARES CUBIC OF GROUP 3 DATA. THE FREQUENCY CURVE FOR RISK APPEARS TO APPROXIMATE A NORMAL DISTRIBUTION.

FIGURE 6D

LEAST SQUARES CUBIC FIT OF GROUP 2 DATA. THE FREQUENCY CURVE FOR RISK APPEARS TO BE VERY CLOSE TO A NORMAL DISTRIBUTION.

FIGURE 6E

LEAST SQUARES CUBIC FIT OF GROUP 5 DATA. THE FREQUENCY CURVE FOR RISK IS SKEWED RIGHT.

BIBLIOGRAPHY

Ammer, Dean S., Materials Management, Homewood: Richard D. Irwin, 1968.

Bauer, R. A., "Consumer Behavior as Risk Taking", in Dynamic Marketing for a Changing World, R. S. Hancock, editor, Proceedings of the 43rd Conference of the A.M.A., 1960, p. 389-400.

Bauer, Raymond A., " Risk Handling in Drug Adoption: The Role of Company Preference ", POQ, Vol. 25. D. 546-559.

Berlyne. D. E., Conflict, Arousal and Curiosity, McGraw-Hill, New York, 1963.

Berlyne, D. E., "Motivational Problems Raised by Exploratory and Epistemic Behavior", in Psychology: The Study of Science. McGraw-Hill. New York. 1963, Vol. 5, p. 284-364.

Berlyne, D. E., "Curiosity and Exploration", Science. Vol. 153, No. 3731, p. 25-33.

Cattell. Raymond B., (ed.) Handbook of Multivariate Experimental Psychology. Rand McNally, Chicago, 1966.

Coombs, Clyde H., A Theory of Data, John Wiley & Sons. Inc., New York, 1964.

Cox, Donald F., (ed.) Risk Taking and Information Handling in Consumer Behavior, Division of Research, Graduate School of Business Administration, Harvard University, Boston, 1967.

Dember, W. N., Psychology of Perception. Holt, Rinehart, & Winston, New York, 1965.

Dixon, W. J., Biomedical Computer Programs, University of California Press, Los Angeles. 1968.

Edwards, Allen L., Technique of Attitude Scale Construction, Appleton-Century-Crofts, Inc., New York. 1957.

Edwards, A. L. and Thurstone, L. L., "An Internal Consistency Check for Scale Values Determined by the Method of Successive Intervals ", Psychometrika, 1952, 17, p. 169-180.

Fowler, H., (ed.) Curiosity and Exploratory Behavior, The MacMillan Co., New York, 1965.

Howard, John A. and Sheth, J. N., The Theory of Buyer Behavior,, John Wiley and Sons, Inc., New York, 1969.

Jones, Lyle V., "Some Invariant Findings Under the Method of Successive Intervals". in Psychological Scaling: Theory and Application, by H. Gulliksen and Samuel Messick. John Wiley and Sons, Inc., New York, 1960.

Montgomery, K. C., "The Role of the Exploratory Drive in Learning ", Journal of Comparative Physiological Psychology, 1954, Vol. 47, p. 60-64.

Purchasing Management Association of Buffalo, New York, "Membership Directory". September 1, 1970.

Rosander, A. C., "The Spearman-Brown Formula in Attitude Scale Construction ", Journal of Experimental Psychology, 1936, 19, p. 486-495.

Saffir, M. A., "A Comparative Study of Scales Constructed by Three Psychophysical Methods", Psychometrika, 1955 20, p. 307-318.

Scott, William A., "Attitude Measurement ", in G. Lindzey and E. Aronson, The Handbook of Social Psychology (2nd ed.), Addison-Wesley, Reading, Mass. 1968.

Sheth, J. N., "Perceived Risk and Diffusion of Innovations", in Insights into Consumer Behavior, Johan Arndt, editor, Allyn and Bacon, Inc.. Boston. 1968.

Sterling, T. D. and Pollack, Seymour V., Introduction to Statistical Data Processing, Prentice-Hall, Englewood Cliffs, New Jersey, 1968.

Torgerson, W. S., Theory and Methods of Scaling. John Wiley and Sons, Inc., New York, 1958.

Webster, Fredrick E., "Modeling the Industrial Buying Process", Journal of Marketing Research, Vol. II, November 1965, p. 370-376.

Westing, J. H., Fine. I. V., and Zenz, Gary J., Purchasing Management, New York: J. Wiley and Sons, 1969.

Wilson. David T., "An Exploratory Study of Purchasing Agent Decision Style," Unpublished Paper. Penn State University. 1970.

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

Authors

Thomas P. Copley, Wilkes College, Wilkes-Barre, Pa.
Frank L. Callom, GTE Sylvania Inc., Towanda Pa.



Volume

SV - Proceedings of the Second Annual Conference of the Association for Consumer Research | 1971



Share Proceeding

Featured papers

See More

Featured

When Implementation Intentions Backfire: Illusion of Goal Progress in Financial Decisions

Linda Court Salisbury, Boston College, USA
Gergana Y. Nenkov, Boston College, USA
Min Zhao, Boston College, USA

Read More

Featured

Consuming Products with Experiences: Why and When Consumers Want Mementos

Charlene Chu, Chapman University
Suzanne Shu, University of California Los Angeles, USA

Read More

Featured

L7. The Joy of Shopping: Reconciling Mixed Effects of Positive Emotions on Shopping Behavior

Kelley Gullo, Duke University, USA
Duncan Simester, Massachusetts Institute of Technology, USA
Gavan Fitzsimons, Duke University, USA

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.