A Theoretical Analysis of Two Recent Measures of Involvement

ABSTRACT - Two scales of involvement have appeared recently in major marketing/ consumer behavior journals. Of these, Laurent and Kapferer's (1985) scales assume multi-dimensionality in involvement; and Zaichkowsky's (1985) scale, while driven by a unidimensional view of involvement, is not unified. The sources of departure from unidimensionality are reviewed for each scale. Consistent with major, recent definitions, a unidimensional conception of involvement is utilized to develop a general model of involvement. The two scales are reconciled with this model, and subscales are identified in each which would measure involvement as a unified construct.

Citation:

Banwari Mittal (1989) ,"A Theoretical Analysis of Two Recent Measures of Involvement", in NA - Advances in Consumer Research Volume 16, eds. Thomas K. Srull, Provo, UT : Association for Consumer Research, Pages: 697-702.

Advances in Consumer Research Volume 16, 1989      Pages 697-702

A THEORETICAL ANALYSIS OF TWO RECENT MEASURES OF INVOLVEMENT

Banwari Mittal, Northern Kentucky University

[The theme of this paper emerged from my formative discussions with Brian Ratchford. Also, my subsequent discussions with him on an earlier draft benefitted the manuscript greatly. Kenneth Lord was also helpful with his suggestions. However, the author alone bears responsibility for the views expressed herein.]

ABSTRACT -

Two scales of involvement have appeared recently in major marketing/ consumer behavior journals. Of these, Laurent and Kapferer's (1985) scales assume multi-dimensionality in involvement; and Zaichkowsky's (1985) scale, while driven by a unidimensional view of involvement, is not unified. The sources of departure from unidimensionality are reviewed for each scale. Consistent with major, recent definitions, a unidimensional conception of involvement is utilized to develop a general model of involvement. The two scales are reconciled with this model, and subscales are identified in each which would measure involvement as a unified construct.

In 1985, Journal of Consumer Research published a scale of involvement (Zaichkowsky 1985). The same year, Journal of Marketing Research also published an article on the measurement of involvement (Laurent and Kapferer 1985). The two measurement approaches differ sharply on a key issue. Zaichkowsky (1985a) has conceptualized involvement to be unidimensional, and correspondingly has developed a measurement scale assumed to be unidimensional. Actually, however, the scale contains at least three dimensions as will be argued below. Laurent and Kapferer (1985) on the other hand adopt a multidimensional view and develop a 4-dimensional profile of involvement. This divergence in the assumed dimensionality goes far back to some early literature (e.g., Lastovicka and Gardner 1978). The issue of dimensionality in involvement, as indeed in any construct, is a fundamental one and must be resolved if the construct has to be employed in any theoretical network whatsoever. Because the above referred two scales are the only ones published in major journals in the field, it is important to resolve the dimensionality issue before they be employed uncritically. To employ them in future research (wherever involvement needs to be measured) is their raison d'etre, so by theoretical explication to cleanse them of their extant dimensional confounding is a service to their originators (at least that is how we see it) from the merit of whose work our essay does not in the least detract. Indeed, we later sort out the subscales from each which would measure involvement quite nicely, that is, more parsimoniously, and in a dimensionally unified manner.

Because several excellent reviews of the literature on involvement already exist (e.g., Zaichkowsky 1986a, Cohen 1983, Greenwald and Leavitt 1984, Antil 1984), this paper will not attempt a comprehensive review, but will instead be limited to delineating the dimensionality issue. Consistent with major, recent literature, the argument in this paper favors the unidimensional view. This view does not call for a per-se rejection of Laurent and Kapferer's empirical results, not does it imply an unqualified acceptance of Zaichkowsky's scale. Rather, the two works are reviewed here both from conceptual and empirical standpoints. Specifically, we dwell upon the sources of dimensionality in the two scales to exemplify as to why the dimensionality issue may have escaped resolution in prior empirical literature. Following the review of the two scales, a model of involvement is- presented which reconciles the multidimensional scales with the conceptually more elegant, unidimensional view of involvement.

Laurent and Kapferer's Involvement Profile

Laurent and Kapferer (1985) develop a consumer involvement profile comprising of four dimensions: importance/risk, risk probability, sign value, and pleasure value. Their work has several merits, including (a) organization of prior literature, (b) attention to diverse usage of the concept, (c) managerial focus, and (d) sifting of the various "sources" (i.e., antecedents) of involvement. It is the dimensionality issue, however, that seems to have been made murkier by their treatment.

Laurent and Kapferer (hereafter L & K) call their dimensions as "antecedents or facets of involvement" (p. 43) (emphasis added). This interchangeable usage of the two terms, antecedents and facets, obscures the issue of dimensionality. If we assume those four factors to be antecedents, then we have measured only antecedents and not involvement itself. On the other hand, if we are to consider the 4 factors as facets, then a multi-faceted view of involvement is implied. To be valid, this view must sit well with extant views of the concept in the literature, and departures be convincingly argued. The conceptual grounding of either "facet" or "antecedent" status would require that the first order of business be to define and conceptually explicate the construct.

Reading through the developmental portion of L&K's article, it is clear that in defining and delineating the concept, L&K encountered a difficulty that would confront us all: in prior literature, the concept has been used variously. Perhaps because L & K's work is driven by concern with aiding managerial decision making, they employ "current research and practice" (p. 43) as sources of their various antecedents, or facets, of involvement. Because "current market research and practice" is a pot-pourri of diverse and sometimes incompatible conceptions of involvement, L &c K's selection strategy could not have addressed the question of dimensionality in involvement, much less separate involvement proper from other seemingly related variables.

Turning to empirical analysis next, L & K present two sets of analyses: factor-analysis of items purported to measure involvement, and regressions of some behaviors on the emerging factors supposed to represent involvement. Consider the factor analysis first. In building their scales, L&K assumed four facets. Their factor analysis confirmed a four factor structure in their data. This at best demonstrated only that their item selection was well done (in that each facet's indicators showed convergent and discriminant validity). This demonstration does not imply that the involvement construct is four dimensional.

Next consider the regression analyses performed to show "predictive validity." Consumers' decision processes and communication receptivity are regressed on the four factors taken to represent involvement. Also included as independent variables are "perceived differentiation" and price. All six predictors have significant regression coefficients in one or the other regression. This result does not make price or perceived differentiation a component of involvement (and L dc K have not implied so). Analogously, the regression results cannot settle the question as to whether the four factors are ,involvement per se, or some or all of them are antecedents of involvement.

We will have been misunderstood if the present reader thinks that we consider L&K's analyses wrong. As we see it, L&K employed the right analytical methods (and with successful outcome) to show that (a) the scale items they assembled had a four factorial structure, and (b) these four factors influenced certain consumer behaviors (e.g., looking at advertising) non-uniformly. At the same time, we are concerned that an unsuspecting reader of L&K's work might conclude that L&K showed that the four factors emerging in their factor analyses are all dimensions of involvement (i.e., that it is involvement that they are dimensions of). Our point is that their empirical analyses were not designed to resolve the issue of dimensionality, and that their eclectic content selection strategy made it highly likely that involvement per se as well as involvement-related concepts could both have been included in their scales, and they actually were. This is an issue which may appropriately be resolved only through a priori concept explication, to which we shall return shortly. First, however, we need to examine the dimensionality question in Zaichkowsky's (1985a) scale of involvement.

Zaichkowsky's PII Scale

Zaichkowsky (1985a) presented a 20-item Personal-Involvement-Inventory (PII) to measure consumer involvement. Some particularly appealing aspects of her work are: (a) a dissertation-level effort to design a scale of involvement when none existed before; (b) the refreshing simplicity of the proposed scale and its applicability across products, brand-decisions, and advertisements as stimuli; and (c) attention to detail at the item screening stage and subsequent validation procedures. The objective of the present paper required, however, that our discussion be directed at a deficiency in PII. This deficiency concerns the dimensionality question.

Zaichkowsky (1985a) adopted a unidimensional conception of involvement defining it as "a person's perceived relevance of the object based on inherent needs, values and interests." However, the 20 items in her scale did not constitute a unidimensional construct, her rigorous item inclusion and screening procedures notwithstanding. This alleged absence of unidimensionality is apparent both on conceptual and empirical grounds.

Conceptually, content analysis of the PII items shows that it (the PII scale) contains at least 3 distinct constructs: (1) involvement proper, reflected in such items as unimportant/important, significant/insignificant, etc., (2) a hedonic factor, reflected in four items, namely, boring/interesting, mundane/fascinating, appealing/unappealing, and unexciting/exciting; and (3) an attitude-like construct, reflected in such items as valuable/worthless, beneficial/not beneficial. One may argue: "Wouldn't hedonic products, or purchases, be more involving?" Yes, but that establishes only that the hedonic factor is related to involvement, and not that it is involvement itself. (We shall later argue that it an antecedent of involvement.) To us the acid test of whether the two constructs are merely related or they are one and the same thing is whether or not it is possible to obtain one without the other. Because all involving purchases need not be hedonic (i.e., the high involvement purchase OS say, lawn mower, where the perceived risk rather than any hedonic factor would likely drive involvement), this implies that involvement is a separate construct from the hedonic factor.

What of the "attitude-like" construct? Zaichkowsky recognizes the need for keeping attitude items out, and for this purpose she reports avoiding bipolar semantic differentials. This we believe is an inadequate strategy because making the items unipolar (e.g., "not beneficial" in p!ace of "harmful") merely truncates the range of the attitude construct rather than eliminate it. (That is why our term for these items is "attitude-like.") Because attitude and involvement are distinct constructs -- involvement mediates attitude formation and change (e.g., Petty and Cacioppo 1979), the inclusion of attitude-like items in a scale of involvement confounds the latter (Bagozzi 1981).

Lending credence to these theoretical arguments are subsequent analyses by Zaichkowsky (1986). She analyzed the PII data for two products, namely, bubble bath and headache remedy. In each case, two factors emerged, but the item configuration also differed between the two products. For bubble bath, the first factor received high loadings from the items which we identified in the foregoing as reflecting the hedonic factor as well as the items reflecting the attitude-like construct. The second factor received high loadings from the rest of the items which can be construed to reflect involvement. For headache remedy, the first factor received high loadings from the involvement and attitude-like items, while the second factor received high loadings only from the items reflecting the hedonic factor. Thus, although only two factors were extracted in each solution, the item-factor loadings pattern reflected a three factor structure in the PII. Note that two conceptually distinct factors may in some empirical data come out as one if the two concepts happen to co-vary in the stimulus being rated. (Parenthetically, we might note that L&K hypothesized perceived importance and risk consequences as two separate factors, but in their data the two factors merged as one. They construed the emergent dimension as a new hybrid construct, termed "imporisk" and used it as such in their JMR analysis. They seem to have recognized the anti-theoretical nature of such use, since they keep these two constructs separate in their Journal of Advertising, 1986 article.)

Since Zaichkowsky (1986b) has herself noted the necessity of re-examining the original list of items, a general discussion of what may have contaminated her item-screening procedures is useful. Several factors may have caused the inclusion of items that were not unified. First, although Zaichkowsky adopted a definition that implicates a unidimensional- conception of involvement, she did not fully explicate what "relevance" meant in her definition. (For example, a pencil, a paper-clip, a briefcase, and a suit are all very relevant to a business executives, but they are not equally involving.) Secondly, her decision to generate a list of as many as 168 word pairs must have necessitated her outstepping the confines of involvement. She does not explain, in her JCR paper, the item generation step, but her earlier working paper (1985b) gives a clue that the included items had to be "either directly or tangentially related to the concept of involvement as it is defined throughout the marketing literature" (emphasis added). The use of the diverse marketing literature on involvement as source of item pool represents an eclectic strategy that can detract from the goal of measuring a unified construct. (Parenthetically note that such an eclectic strategy which L & K claim to have used, was precisely the reason for the emergence of multi-dimensionality in L & K's measures.) Furthermore, the decision to allow "tangentially related" items must have also added to the problem. Third, while the use of "expert" judges is a proper procedure, judges can not transcend the definitional framework supplied by the researcher. Unless the construct is carefully separated from its antecedents and consequences at the construct definition and explication stage, one runs the risk of burdening the construct with undesirable "excess baggage" (Cohen 1983). Zaichkowsky's scale appears to have been so burdened.

Summary. It is important to maintain clarity about how factor- analysis results in L & K's paper are cited here for not confirming the multi-dimensional view of involvement, but cited for disconfirming unidimensionality in Zaichkowsky's scale. Zaichkowsky's scale is argued to contain three conceptually distinct entities, and her subsequent empirical analyses tends to support this argument. So her empirical scale is being questioned, not her conceptual view of the construct. L & K's scales also yield multiple dimensions, but they adopt the multidimensional view right from the outset without, and this is important, carefully first defining involvement. That L & K's scale is multidimensional is not being questioned. Rather, the question being asked is, Is it involvement that their multidimensional scales are measuring? And, for Zaichkowsky's scale, the question is, Which items in her scale truly represent the unidimensional construct of involvement that she had set out to operationalize? To guide our decisions on this question, a conceptual model of involvement is presented next.

The Concept of Involvement: Establishing a Common Ground

Indeed, L & K are right in pointing out the diverse, vague, and confusing usage of the term involvement in the literature. Nevertheless, the field seems to have finally made considerable progress at converging toward a reasonably unified definition of involvement in the past five years or so. Some of these recent definitions are:

* Involvement is said to reflect the extent of personal relevance of the decision to the individual in terms of her basic values, goals, and self-concept (Engel and Blackwell 1982, p. 273; also adopted by Zaichkowsky 1985).

* Similarly, Greenwald and Leavitt (1984) conclude their literature review by stating that "there is a consensus that high involvement means (approximately) personal relevance or importance." (However, these authors appear to operationalize involvement in terms of attention.)

* Involvement is an internal state variable that indicates the amount of arousal, interest, or drive evoked by a particular stimulus or situation (Mitchell 1979, 1981); also adopted by Bloch (1982).

* It may be preferable to conceive of involvement as a person's activation level at a particular moment of time (Cohen 1983).

* Involvement is defined as a person's motivational state directed toward a goal object for accomplishing a specific goal. The goal object can be a product class, a purchase decision, a specific brand, or an advertisement. ... Also, involvement may be defined as a goal-directed arousal capacity (Park and Mittal 1985).

Despite differences in nuances, there does seem to be a common thread. This common thread may be construed as a "motivational state that has been activated" by a stimulus, situation, or a decision task. Engel and Blackwell's "personal relevance" implies that the decision bears on a person's goals, i.e., motivations. And, Mitchell's "internal states indicating the amount of arousal," and Cohen's momentary "activation level" may readily be paraphrased as "activated motivational state."

This concept of "activated motivational state" is unitary and unidimensional. The unidimensional view is clearly implied also in research by Petty and Cacioppo (1983). The diverse usage of the term in the literature with various prefixes (e.g., situational, enduring, and response involvement; ego involvement, personal involvement, solution involvement, task involvement, and emotional involvement) need not imply that those prefixes signify dimensions or aspects of involvement. Some of these are loose labels (e.g., emotional involvement), others differ in the corresponding goal-object (e.g., task vs solution involvement), and still others are either antecedents or consequences (e.g., the enduring/situational/response triology).

A General Model of Involvement

It is important that the construct of involvement be kept separate from its antecedents and consequences, otherwise, as Cohen (1982) has argued, an overly broad construct would result which would make any investigation of relationships (among involvement and other consumer behavior variables) necessarily imprecise. When involvement is construed as a unidimensional construct (defined, for example, as an "activated motivational state"), one may categorize all of its antecedents in two categories of goals: utilitarian and psycho-social. A stimulus becomes involving when it meets some important utilitarian or psychosocial goals, and were the person to not interact with this stimulus, he would incur the opportunity loss (or risk) of not satisfying those goals or not satisfying them as well. These relationships are shown in Figure 1. Ahtola (1985) uses a similar structure to model an "overall attitude construct." And the proposed model is also compatible with Bloch and Richins (1983) framework, although the focus of their work is not identical to ours.

Reconciling the Multidimensional Scales with the Unidimensional Conception of Involvement

In L & K's work, we suggest that only the "Importance" factor be deemed to represent involvement. This was a facet by itself at the outset, but was merged with risk consequences based on factor analytic results. On conceptual grounds, it should be kept distinct. Laurent and Kepferer's pleasure and sign facets should be modeled as antecedents and placed in the "psycho-social goals" box in Figure 1. The to-be-untangled risk-consequences and already separate risk probability should be two subfacets of risk; together, they should be placed in Figure 1 in the "Utilitarian goals" box. It must be noted that L&K's current importance facet taps product involvement; if purchase involvement is to be measured, then the importance facet would have to be modified suitably. For example, the current item "For me, _ does not matter," may be rephrased as "For me, it does not matter as to how well I make my selection of _."

(Like Zaichkowsky's scale, our model in Figure 1 is a general one. As such it can apply to product involvement or to purchase involvement. Involvement and its two categories of antecedents in Figure 1 should each be operationalized with the same target-stimulus (product or purchase).)

In Zaichkowsky's scale, involvement proper can be measured by 6 items: relevant, important, of concern, matters, means a lot, and significant. (The first five of these items were also identified by McQuarrie and Munson, 1986.) Many items in Zaichkowsky's scale (e.g., beneficial, useful, valuable, etc.) tap attitude and may not belong in a scale of involvement. Yet other items (e.g., the four items representing hedonic aspects) tap an antecedent of involvement rather than involvement itself. The six items suggested above appear to be conceptually unified, and prior factor analytic solutions (Zaichkowsky 1985a, 1986) have shown these items to always load on the same single factor.

Since McQuarrie and Munson (1986) present a revised PII, brief comments on it are in order. According to the arguments advanced above, they are right in discarding the attitude-like items in PII. But they borrow L&K's multi-dimensional view, and therefore all the arguments made here in-respect of L&K also apply to their work. Only the "Importance" factor in their revised, PII must be deemed to represent involvement. The other factors, Risk and Pleasure, must be deemed to be antecedents of involvement.

Empirical Validation.

The unidimensionality of a collection of items can be empirically tested. Zaichkowsky's subsequent work (1986) has shown lack of unidimensionality in her scale, so there was no need here for us to duplicate that demonstration. What about our assertion about L&K's scales? We argue that risk, hedonic, and sign facets must be treated as antecedents of involvement proper. Can this be demonstrated empirically? The answer is "No." LISREL analyses of L&K's four facets, treating them as four factors in a confirmatory factor analytic model, and then, alternatively, as a structural model with "importance" as an endogeneous construct and the three other facets as exogenous constructs -- both these model formulations will yield identical solutions both from overall model fit and parameter estimate standpoints. Nor would the incorporation of the aforementioned two alternative models in a more extended network of antecedents and consequences help resolve the issue. The causal influence on other variables (thought to be consequences of involvement) attributed to risk, hedonic, and sign facets would no doubt be different when these facets are modeled as antecedents of involvement than when they are modeled as facets of involvement itself. But these differences in parameter estimates are the outcome of the model relationships being specified differently on theoretical grounds. These differences can not serve as empirical proofs for the validity of one over the other model. Indeed, that the obtained estimates will be different is precisely why it is important to specify the models correctly so that the real-world interpretations of between-variable relationships are not misfounded. Some take Rothschild's (1984) admonition against the sometimes prevalent excessive flow-charting exercises in the literature to imply an abandonment of all a priori reasoning. We believe that such misreading will have the unintended effect of producing blind empiricism, an error this paper is purported to help avoid.

FIGURE 1

A LIMITED CAUSAL MODEL FOR INVOLVEMENT

Summary and Conclusion

Two scales of involvement have been published in recent literature, namely, Laurent and Kapferer (1985), and Zaichkowsky (1985a). In fact, these are the only scales that have been published in major marketing and/or consumer behavior journals. These scales differ in the assumed dimensionality of involvement, an issue that must be resolved before the proposed scales may be employed in any theoretical (or, for that matter, managerial) investigations of the effects of involvement. From this standpoint, both the scales were examined. The major, recent literature on involvement is shown to espouse a unidimensional view of involvement. The multidimensionality in L & K's scale is shown to stem from the* eclectic content-sampling strategy. Zaichkowsky's scale, on the other hand, is driven by the unidimensional conception of involvement, but the item inclusion procedures failed to ensure a unified construct. Both scales do contain within them the measures of involvement, and are therefore useful; both (especially L & K's) however also contain other related variables which our analysis has shown to be antecedents of involvement (rather than involvement itself). Once the antecedents and involvement-proper are separately identified, the two scales assume added utility, in that they furnish measures of involvement per-se as well as of its most proximate antecedents.

Our discussion of the two scales has led us to suggest a model which reconciles the multidimensional view with the theoretically more elegant unidimensional view of involvement. Our argument has been that the delineation of involvement-proper from its antecedents (or from other involvement-like constructs) is a theoretical task, not an empirical one. However the gains from such a theoretical exercise are both theoretical and pragmatic. On the theory side, contaminated constructs or scales introduce imprecision in theoretical relationships. On the pragmatic side, unified and conceptually valid scales enable proper estimation of magnitudes of causal influence of control variables.

REFERENCES

Ahtola, O. T. (1985), "Hedonic and Utilitarian Aspects of Consumption Behaviors: An Attitudinal Perspective," in E. C. Hirschman and M. B. Holbrook (eds.), Advances in Consumer Research, Provo, UT: Association for Consumer Research, XII, 7-10.

Bloch, Peter, and Marsha Richins (1983), "A Theoretical Model for the Study of Product Importance Perceptions," Journal of Marketing, 47, 69-81.

Cohen, J. B. (1982), "Separating the State from Its Causes and Effects," Paper presented at the Involvement Colloquium at NYU, June 34.

Cohen, J. B. (1983), "Involvement and You: 1,000 Great Ideas," in R. P. Bagozzi and A. M. Tybout (eds.), Advances in Consumer Research, Ann Arbor, MI: Association for Consumer Research, X, 325-328 .

Engel, J. F. and R. D. Blackwell (1982), Consumer Behavior, New York: Dryden Press.

Greenwald, A. G. and C. Leavitt (1984), "Audience Involvement in Advertising: Four Levels," Journal of Consumer Research, 11 (June), 581-92.

Houston, M. J. and M. L. Rothschild (1977), "A Paradigm for Research on Consumer Involvement," WP 11-7746, University of Wisconsin-Madison.

Krugman, H. E. (1966), "The Measurement of Advertising Involvement," Public Opinion Quarterly, 30 (Winter), 583-596.

Lastovicka, J. L. and D. M. Gardner (1978), "Components of Involvement,'' in Attitude Research Plays for High Stakes, eds., J. C. Maloney and B. Silverman, Chicago: American Marketing Association, 53-73.

Laurent, G. and Jean-Noel Kapferer (1985), "Measuring Consumer Involvement Profiles," Journal of Marketing Research, 22, 41-53.

McQuarrie, E. F. and J. M. Munson (1986), "The Zaichkowsky Personal Involvement Inventory: Modification and Extension," in Advances in Consumer Research, Toronto, October.

Mitchell, A. A. (1979), "Involvement: A Potentially Important Mediator of Consumer Behavior," in W. L. Wilkie (ed.), Advances in Consumer Research, 191-196.

Mitchell, A. A. (1981), "The Dimensions of Advertising Involvement," in K. B. Monroe (ed.), Advances in Consumer Research, 25-30.

Muncy, J. A. and S. D. Hunt (1984), "Consumer Involvement: Definitional Issues and Research Directions," in T. C. Kinnear (ed.), Advances in Consumer Research, Provo, UT: Assoc. for Cons Res, IX, 193-195.

Park, C. W. and Banwari Mittal (1985), "A Theory of Involvement in Consumer Behavior: Problems and Issues," in J. N. Sheth (ed.), Research in Consumer Behavior, I, JAI Press, Greenwich, CT, 201-232.

Park, C. W. and S. M. Young (1986), "Consumer Response to TV Commercials: The Impact of Involvement and Background Music on Brand Attitude Formation," Journal of Marketing Research, (February), 11-24.

Petty, R. E., J. T. Cacioppo and D. Schumann (1983), "Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement," Journal of Consumer Research, 10 (September), 135-146.

Rothschild, M. L. (1984), "Perspectives on Involvement: Current Problems and Future Directions," in T. C. Kinnear (ed.), Advances in Consumer Research, Provo, UT: Association for Consumer Research, XI, 196-198.

Zaichkowsky, J. L. (1985a), "Measuring the Involvement Construct," Journal of Consumer Research, 12, 341-352.

Zaichkowsky, J. L. (1985b), "Measuring the Involvement Construct," Working Paper, The American University.

Zaichkowsky, J. L. (1986a), "Conceptualizing Involvement," Journal of Advertising, 15 (2),4-14.

Zaichkowsky, J. L. (1986b), "The Emotional Aspect of Product Involvement," Advances in Consumer Research, Toronto, October.

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