comment: measuring consumer involvement in products: comment on traylor and joseph

6
Comment Measuring Consumer Involvement in Products: Comment on Traylor and Joseph Raj Arora Robert Baerl The concept of consumer involvement (and its related measurement problems) has been receiving the attention of more and more scholars (e.g., Arora 1982; Petty and Cacioppo 1983) who have recognized the complexity and the multidimen- sionality of the construct. However, the recent article by Traylor and Joseph (1984) suggests that consumer involvement in products can be assessed by a generalized scale. While the motivation to arrive at such a scale is commendable, questions may be raised concerning the usefulness of the Traylor and Joseph involvement scale for marketing management decisions for the following four reasons: 1) the scale seems to measure only an undimensional construct, 2) the validity of the 'The order of authorship was randomly determined and does not reflect the relative contributions of the authors. PSYCHOLOGY & MARKETINGNOL. 2 NO. 1 57

Upload: raj-arora

Post on 08-Aug-2016

218 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Comment: Measuring Consumer Involvement in Products: Comment on Traylor and Joseph

Comment Measuring Consumer Involvement in Products: Comment on Traylor and Joseph

Raj Arora Robert Baerl

The concept of consumer involvement (and its related measurement problems) has been receiving the attention of more and more scholars (e.g., Arora 1982; Petty and Cacioppo 1983) who have recognized the complexity and the multidimen- sionality of the construct. However, the recent article by Traylor and Joseph (1984) suggests that consumer involvement in products can be assessed by a generalized scale. While the motivation to arrive at such a scale is commendable, questions may be raised concerning the usefulness of the Traylor and Joseph involvement scale for marketing management decisions for the following four reasons: 1) the scale seems to measure only an undimensional construct, 2) the validity of the

'The order of authorship was randomly determined and does not reflect the relative contributions of the authors.

PSYCHOLOGY & MARKETINGNOL. 2 NO. 1 57

Page 2: Comment: Measuring Consumer Involvement in Products: Comment on Traylor and Joseph

CONSUMER INVOLVEMENT

scale has not been established, 3) the reliability of the test has not been adequately demonstrated and 4) problems exist pertaining to the factor analysis of the scale items. The present paper will address each of those issues in turn.

WHAT IS INVOLVEMENT?

There are four major streams of research in the area of involvement. The first one is based on social judgement theory (Sherif and Hovland, 1961). Sherif and his colleagues have focused on involvement from the perspective of attitudes towards the stimulus object and specifically an individual’s stand or commitment on a specific issue. In other words, involvement is referred to as some type of inflexibility with respect to a particular stand on an issue. The second stream of research findings can be attributed to Krugman (1965). While Krugman specifi- cally rules out interest or excitement toward the product in his conceptualization, he conceptualizes involvement as the number of conscious bridging experiences, connections, or personal references the viewer niakes between his or her own life and the advertisement.

The third stream of research findings focuses on involvement from the per- spective of information processing. Petty and Cacioppo’s (1983) work, based on an elaboration likelihood model, posits two routes to persuasion, i.e., a central route (compelling arguments for high involvement) and a peripheral route (low involvement-imagery). Petty and Cacioppo show that in the case of high in- volvement, one needs strong, convincing arguments (central route) and is therefore similar to the Sherif et a1 proposition, whereas in the low involvement case, per- suasion can be achieved by imagery and repetition (peripheral route) and is thus similar to Krugman’s proposition.

The fourth stream of research distinguishes between different types of in- volvements. Houston and Rothschild (1977) posit the existence of three types of involvements: situational, enduring, and response. Situational involvement is due to factors particular to a time and place of observation which do not follow from the knowledge of personal and stimulus attributes. An example of situational in- volvement may be the concern surrounding the purchase or consumption of a prod- uct due to the social-psychological environment. Enduring involvement is reflected by the relationship of an individual’s value system to the products. The importance attached to the products attributes reflects this dimension. Response involvement reflects the behavioral process characterizing the decision process. This may be search of information, purchase, and post purchase behavior.

The preceding review suggests that Traylor and Joseph’s literature review is inadequate. It is based on only Sherif‘s social judgement approach and Krug- man’s personal reference approach. While Sherif‘s and Krugman’s involvement is stimulus and individual specific, Traylor and Joseph appear to be assessing in- dividual involvement generalized across products. However, before one can de- velop a generalized scale, one must have a generalized definition and a generalized

58 SPRING 1985

Page 3: Comment: Measuring Consumer Involvement in Products: Comment on Traylor and Joseph

CONSUMER INVOLVEMENT

operational measure of this construct. This is precisely the problem that marketers are faced with today. In a work by Houston & Rothschild (1977) they posit three types of involvement and by Greenwald & Leavitt (1984) who, in terms of infor- mation processing, describe four levels of involvement, make it increasingly clear that involvement is not the simple undimensional construct that Traylor and Joseph would lead us to believe. However the six statements that comprise their involve- ment scale (# 8 , 9 , 12, 14, 16, & 19 in Table 1 of Traylor & Joseph) all “appear” to be assessing the degree to which the product is reflective of their self-concept.

A product may be used by some individuals as a means to express their values and as such may generate some concern and thus involvement. However, this type of concern is likely to be relevant to products that have an element of conspicuous consumption such as clothing, personal care, and so on. What about a vast mul- titude of products that are not used by individuals to exhibit their “self”?

VALIDITY

The validity of a test is defined by the extent to which the test measures the hypothesized underlying trait, construct, or factor. At least four basic approaches to assessing validity (content, construct, and criterion; predictive and concurrent) have been identified. Which type of validity is most appropriate must be deter- mined not by convenience, but with reference to the particular use for which the test is being considered (Anastasi, 1968). Traylor and Joseph present evidence that pertains only to their scale’s concurrent validity. It is argued here that, 1) con- current validity is not consistent with their stated purpose, and 2) the criterion chosen for their concurrent validity study is irrelevant and inappropriate.

Type of Validity

There are two types of criterion-related validity+oncurrent validity and predictive validity. Predictive validity refers to the extent to which a test (e.g., involvement scale) is capable of predicting future behavior (e.g., purchase). In this case, the test and criterion data (purchase) are collected at separate points in time. Concurrent validity is similar to predictive validity, but unlike predictive validity the test and criterion data are collected at the same points in time. Con- current validity is important when one intends to use an individual’s score on a test (involvement) as a substitute for some significant nontest variable; i.e., the criterion (Brown, 1970). What is of ultimate interest in the case of criterion-related validity is not the score on the test itself, but the individual’s status on the criterion. The test score is only important in that it predicts (predictive validity) or can serve as a substitute for (concurrent validity) the criterion. Since the test operates merely as a predictor or substitute rather than as a representation or sample, the content of the test is essentially, irrelevant (Brown, 1970). However, it is not Traylor and Joseph’s intent that their scale be used as a substitute for their criterion (purchase

PSYCHOLOGY & MARKETINGNOL. 2 NO. 1 59

Page 4: Comment: Measuring Consumer Involvement in Products: Comment on Traylor and Joseph

CONSUMER INVOLVEMENT

frequency) and thus the test of concurrent validity is irrelevant to their purpose. Since authors indicate that their aim is to develop a scale that would, “more ac- curately reflect the meaning of the involvement construct than do existing meas- ures” (p. 67), assessing the scale’s construct validity seems more germane to these goals.

Construct validity refers to the extent of which the test may be said to mea- sure the theoretical construct or trait (involvement). The central concern with con- struct validity is with the definition of the construct. Construct validity is important whenever a test is to be interpreted as a measure of some attribute or quality that people are presumed to possess (Brown, 1970). This would be consistent with Traylor and Joseph’s viewpoint that products are not inherently involving or un- involving, but that involvement is a characteristic of consumers. Although the scale’s convergent and discriminate validity and other quantitative information bear on construct validity, there is no single quantitative index that reflects the extent to which a scale measures the hypothesized underlying trait. In order to determine the construct validity of a scale, one must examine the entire body of evidence surrounding the test-the variables other than involvement with which the scores do and do not correlate. Since there is no such discussion in the Traylor and Joseph study, it is difficult to determine the nature of the construct they mea- sured.

Irrelevant Criterion

Traylor and Joseph selected purchase frequency over the past year as the criterion for their scale. The selection of a criterion, however, involves more than merely positing, on the basis of intuition and common sense, a relationship be- tween involvement and other variables. Rather, the selection of an appropriate criterion is the most important aspect of criterion-related validity. A criterion’s most important quality is relevance (Anastasi, 1968; Cascio, 1978; Thorndike and Hager, 1977). Relevance refers to the extent to which the criterion measure reflects and measures the important facets of the conceptual variable (involvement). Since a criterion is defined as a direct and independent measure of that which the test is designed to measure one’s standing on the criterion should exemplify the con- struct. Clearly purchase frequency does not reflect the multidimensional facets of consumer involvement, nor does it exemplify the construct. Since criterion-related validity evidence will be no better than the criterion measure used, the (weak) correlation between scores on the involvement scale and purchase frequency pro- vide a rather weak demonstration of the scale’s validity.

RELIABILITY

No test yet devised measures without error. Error represents the contribution of variables that are irrelevant to the purposes of the test and may result from a host of different reasons. Any test score actually obtained is a function of both the

60 SPRING 1985

Page 5: Comment: Measuring Consumer Involvement in Products: Comment on Traylor and Joseph

CONSUMER INVOLVEMENT

“true score” (i.e., the score one would obtain if the measuring instrument was perfect) and the random error involved in measurement. Reliability is an estimate of the degree for error free measurement. As in the case for validity, there are a variety of methods available for determining reliability (e.g., test-retest, split-half, alternative forms and internal consistency). Each method assess a different com- ponent of measurement error. Which particular method will be appropriate in any situation depends not upon convenience, but the purpose of the test (Anastasi, 1968; Brown, 1970; Guion, 1965).

Traylor and Joseph present evidence that pertains only to the internal con- sistency of their scale. Measures of internal consistency are widely used. Unfor- tunately, much of their popularity is probably due to convenience. These estimates can be obtained with only a single administration of the test. A better reason for their use is a need to know whether or not all items on the test measures one com- mon characteristic (homogeneous in content) or several characteristics (hetero- geneous in content). Homogeneous tests aid in the definition of the construct and help establish construct validity. However, by virtue of subjecting responses to items on the original questionnaire to factor analysis, we already know that the test is homogeneous. That is, items were selected for inclusion in the final ques- tionnaire because of their high degree of interrelatedness.

Before the Traylor and Joseph involvement scale can be said to be reliable, data pertaining to the scale’s retest reliability must be obtained. Retest reliability shows the extent to which scores on a test can be generalized over different oc- casions; the higher the reliability the less susceptible the scores are to random changes in the condition of the stimulus or the testing environment. Establishing retest reliability is important for traits assumed to be relatively stable over time and for variables used to make decisions involving long-term plans. Presumably, involvement as conceptualized by Traylor and Joseph falls into both of these cat- egories.

INCORRECT USE OF STATISTICAL ANALYSES

In the process of purifying the measure, Traylor and Joseph use factor anal- yses and retain six of the 10 statements that had loadings of 0.5 or more on the f ist unrotated factor for each of the products. They further state that the variance accounted for by the first unrotated factor ranged from a low of 67.3% to a high of 100%. There are two problems with the above analyses: (1) Almost all (ex- ploratory as opposed to confirmatory) factor analyses procedures use principle component solution to derive the unrotated loadings. Inherent in principle com- ponents procedure is the constraint that the first factor (principle component) has the largest variance. Thus, the selection of statements based on loadings of 0.5 or more on first unrotated factor yields a solution which is contaminated by procedural constraints since all items will appear to load heavily on the first factor. (2) The variance of a factor is the eigenvalue of the factor divided by the number of var-

PSYCHOLOGY & YARKETINGNOL. 2 NO. 1 61

Page 6: Comment: Measuring Consumer Involvement in Products: Comment on Traylor and Joseph

CONSUMER INVOLVEMENT

iables. The eigenvalue is the sum of squares of loadings. The variance of 100% for any factor means, (a) that the variable loads on one and only one variable and (b) that the loading for each variable is exactly one. This implies that all the va- riance of a variable is accounted for by one and only one common factor, or that there is no unique variance which may be contributed to either a unique factor or residual. Scholars in social and behavioral sciences recognize that measures in these areas are fallible or some degree of error no matter how small is always present in any assessment. Thus, the notion of 100% of variance is a virtual im- possibility. Besides the loadings on Table 2 are too far from unity (for all 10 var- iables on 12 products) to result in 100% variance extracted.

In sum, Traylor and Joseph have designed a homogeneous scale. But what this homogeneous scale measures, and whether it will do so consistently over time has yet to be determined.

REFERENCES

Anastasi, A. (1968) Psychological Testing (3rd ed.). London: The Macmillan Company. Arura, Raj (1982) Validation of an S-0-R model for situation, enduring and response components of

Brown, I.G. (1970) Principles ofEducationa1 and Psychological Testing. Hinsdale, ILL: The Dryden

Cascio, W.F. (1978) Applied Psychology In Personnel Management Reston, Virginia: Reston Pub-

Guion, R.M. (1965) Personnel Testing. New York: McGraw-Hill Book Company. HOUSIOII, Michael J and M.L. Rothschild (1977). A paradigm for research on consumer involvement,

working paper 11-77-46 Graduate School of Business, University of Wisconsin-Madison. Krugman, H.E. (1965). The impact of television advertising: Learning without involvement. Public

Opinion Quarterly, 29. 349-356. Petty, R.E. and J.T. Cacioppo (1983). Central and peripheral routes to persuasion: Application to

advertising. in Advertising and Consumer Psychology, eds., L. Percy and A. G. Woodside, Lexington, MA: Lexington Books, 3-23.

Sherif, M. and Carl I. Hovland (1Y81). Social Judgmenr Assimilation and Contrast Eflects in Com- municarion and Attitude Change, New Haven: Yale University Press.

Thomdike, R.L. and Hagen, E. P. (1977). Measurementand Evaluation in PsychologyandEducation (4th ed.). New York: John Wiley & Sons.

Traylor, M.B. and W.B. Joseph (1984). Measuring consumer involvement in products: Developing a general scale. Psychology and Marketing, 1, 65-71.

involvement, Journal of Marketing Research, (November) 505-5 16

Press, Inc .

lishing Company, Inc.

Raj Arora, and Robert Baer are both professors of Marketing in the Department of Mar- keting at Bradley University, Peoria, Illinois. 61625

62 SPRING 1985