Separating Brand-Choice Involvement From Product Involvement Via Consumer Involvement Profiles

Banwari Mittal, Northern Kentucky University
Myung-Soo Lee, State University of New York at Buffalo
ABSTRACT - Laurent and Kapferer (1985) proposed a four-faceted Consumer Involvement Profile as a way of operationalizing consumers' involvement in products. Two gaps in that profile are addressed in this research: (a) expansion of Laurent and Kapferer's "illustrative" scale items into full scales, and (b) operationalization of the four facets separately at the product- and brand-choice levels. Using confirmatory factor analytic procedures, the convergent and discriminant validity for the proposed scales is shown to be supported, and directions for future research are suggested.
[ to cite ]:
Banwari Mittal and Myung-Soo Lee (1988) ,"Separating Brand-Choice Involvement From Product Involvement Via Consumer Involvement Profiles", in NA - Advances in Consumer Research Volume 15, eds. Micheal J. Houston, Provo, UT : Association for Consumer Research, Pages: 43-49.

Advances in Consumer Research Volume 15, 1988      Pages 43-49

SEPARATING BRAND-CHOICE INVOLVEMENT FROM PRODUCT INVOLVEMENT VIA CONSUMER INVOLVEMENT PROFILES

Banwari Mittal, Northern Kentucky University

Myung-Soo Lee, State University of New York at Buffalo

ABSTRACT -

Laurent and Kapferer (1985) proposed a four-faceted Consumer Involvement Profile as a way of operationalizing consumers' involvement in products. Two gaps in that profile are addressed in this research: (a) expansion of Laurent and Kapferer's "illustrative" scale items into full scales, and (b) operationalization of the four facets separately at the product- and brand-choice levels.

Using confirmatory factor analytic procedures, the convergent and discriminant validity for the proposed scales is shown to be supported, and directions for future research are suggested.

INTRODUCTION

One of the welcome, recent developments in research on consumer involvement (which for a long time was an elusive concept) is the appearance of empirical scales to measure that concept. One such scale was developed by Laurent and Kapferer (1985). These authors suggest a four-faceted involvement scale and present data which show that different facets have differing implications for specific consumer behaviors. While we find this scale appealing, we believe that Laurent and Kapferer do not address the question of whether their facets reflect product-class involvement or brand-choice involvement or both. Our analysis will show that two of the facets reflect product-involvement, and two reflect brand-decision involvement. The purpose of this research is to propose (and empirically test) the measures of each of the four facets separately for product and brand-choice involvement. Because these two "forms" of involvement are distinct but not unrelated (Bloch 1983), our data will allow examination of relationships among the facets both within and across the two forms.

This paper is organized as follows. First, the contents of Laurent and Kapferer's scale are discussed with a view to identifying the form (product- or brand-choice) of involvement each facet measures. Next, parallel measures for each facet are developed for each of the two forms. Third, empirical data, collected for beer, are analyzed to demonstrate the structure of the measures, and also the relationship of each measured facet to selected criterion variables. Finally, implications of our findings for future theoretical and applied research are outlined.

THEORY

Product- and Brand-Choice Involvement

A number of involvement researchers have distinguished product involvement from brand choice involvement (Bloch 1983, Bloch and Richins 1983, Zaichkowsky 1985, 1986). Product involvement is the degree of interest of a consumer in a product category on an on-going basis. Brand-choice involvement is the motivation of a consumer to make the right choice. As Bloch and Richins (1983) point out the two are not identical. For example, a consumer is seldom involved in the washing machine on an enduring basis, but he/she is likely to be very involved in making the brand selection.

Houston and Rothschild (1977) make a distinction between enduring- and situational-involvement and view the former as a mean level of involvement across situations, while the latter (i.e., situational involvement) provides temporary, situation-bound deviations from the mean level. It is possible, then, that for a consumer, and for a given product, the enduring involvement is low, but brand-choice involvement (which is a form of situational involvement) is high. Thus it would seem useful to distinguish *e two forms of involvement and operationalize Laurent and Kapferer's profile separately at the two levels.

Laurent and Kapferer's Involvement Profile

Laurent and Kapferer (1985) propose 4 facets, namely, (1) the importance of the product, (2) perceived risk associated with the product purchase, which in turn has two subfacets: (a) perceived importance of negative consequences from a poor choice, and (b) the perceived probability of making such a mistake; (3) the symbolic or"sign" value, and (4) the hedonic value of the product. In the factor analysis of their data, they found that the perceived product importance and perceived importance of negative consequences of a mispurchase loaded on a single factor which they termed "imporisk." Risk probability formed a separate factor, while sign value and -pleasure value each provided one additional factor. These four factors are then deemed by Laurent and Kapferer as constituting the consumer involvement profile.

The question is, which involvement does this profile tapCproduct-category involvement or brand-choice involvement? Because Laurent and Kapferer base their framework, in large part, on Houston and Rothschild's paradigm (a paradigm that incorporates both forms of involvement), Laurent and Kapferer's consumer involvement profile (CIP) contains measures both of product category and brand-choice involvement. Houston and Rothschild separately identify the two forms, but Laurent and Kapferer do not. This is a task addressed below.

To discern the form of involvement (product- or brand-choice-) being tapped by the four facets, both the description and the corresponding measurement items for each are reviewed below. Laurent and Kapferer present only illustrative items; therefore, an assumption was made that other items were conceptually identical. That is, if the illustrative item reflects product-category(brand-choice) involvement, the omitted items also reflect product-category- (brand-choice-) involvement.

To recapitulate, two of the four facets (perceived importance, and hedonic value) are conceived by Laurent and Kapferer at the product level, and one facet, namely, perceived risk is conceived at the brand-choice level. The fourth facetCsign valueCis defined at an all inclusive level so that it can refer to product- or to brand-choice involvement; however, it seems to have been measured only at the brand-choice level.

Developing Separate Measures for Product- and Brand-Choice Involvement

Of interest here is the development of measures of each of the four facets for product involvement and, separately, for brand-choice involvement, based on Laurent and Kapferer's illustrative measures. Because Laurent and Kapferer have not presented the complete measurement instrument, the proposed operationalizations may also serve to fill this gap.

Hypotheses

To establish the separation of facets across the two forms of involvement, we propose to conduct tests of trait validity and discriminant validity just as Laurent and Kapferer (1985) did. For trait validity, items purported to measure a construct should measure a single dimension, and for discriminant validity the two scales should not correlate highly (Campbell 1960). Our hypothesis is: Each of the four facets are separate constructs across product involvement and brand choice involvement. This hypothesis entails four subhypotheses:

H1. Perceived importance of the product is a separate construct from perceived importance of brand choice.

H2. Perceived sign value of the product is a separate construct from the perceived sign value of the brand

H3. Hedonic value of the product is a separate construct from the hedonic value of she brand.

H4. Perceived risk of the product is a separate construct from the perceived risk in brand - choice.

Because Laurent and Kapferer found perceived importance and perceived risk to constitute a single factor, we propose to also test the discriminant validity of these two facets. As hypothesized by Laurent and Kapferer, these two facets have "face validity" as two constructs rather than one. It is in the nature of exploratory factor analysis that scale items measuring two correlated factors might sometimes load on a single factor (more discussion later). We shall instead be using confirmatory factor analytic procedures to test the following hypotheses:

H5. Perceived importance of the product is a separate construct from perceived risk from the product.

H6. Perceived importance of the brand choice is a separate construct from perceived risk in brand choice.

H7. Perceived importance of the product is a separate construct from perceived risk of brand choice.

The reason for testing the last hypothesis is that Laurent and Kapferer had operational: zed perceived importance at the product level but risk at brand choice level and it is these operationalizations which in their analysis yielded a single factor.

In addition, our empirical study will also examine the relationship of the various facets with selected consumer behavior variables. Laurent and Kapferer (1985) have shown that such consumer behavior variables as 'looking at advertising' or 'the decision making process' are differentially related to their four facets. Our interest here is in testing whether the selected behaviors correlate with a facet differentially across the brand level versus the product level.

METHOD

We measured the "Consumer Involvement Profiles" for beer for student consumers. Beer was chosen due to its relevance to our respondents. It was also intuitively judged to provide potential for the varied occurrence of each of the facets. For example, not all consumers would see beer as providing a sign value but many would; Similarly, the pleasure of drinking beer would be realized for some consumers from any beer whatsoever, and by others only from specific brands of beer.

Of the 100 undergraduate and graduate business students who answered the questionnaire, 22 were "never' drinkers. Consistent with Laurent and Kapferer's respondent screening strategy, all reported results are based on the 78 beer drinkers. Readers may note, however, that the support to be reported for the hypotheses was not materially affected when the analyses were rerun with all 100 respondents.

Measures. Measures of each of the 4 facets operationalized differently for product and brand-choice levels are shown in Table 1. These operationalizations resulted from an extension of Laurent and Kapferer's illustrative items, and our own intuitive translation of the facets. In addition, 6 selected aspects of consumer behavior were also measured: (1) use frequency: I drink beer (never - 1, occasionally - 2, often - 3, regularly - 4, very frequently - 5); (2) perceived brand differences: Beer brands are all very similar - 1 / all very different- 7; (3) Brand comparison: I have done extensive (1) / I have not done any (7) brand comparisons; (4) Brand commitment: If your favorite brand of beer is not available in the store, you will: 1 - go to another store, 2- buy another favorite brand, 3 - buy whatever is available; (5) Interest in product article: If there is an article about beer, I would be interested in reading it: Strongly disagree -1, Strongly agree - 7, and (6) Attention to advertising: I often pay attention to advertising: Strongly disagree - 1, Strongly agree - 7. (Brand comparison and Brand commitment scales were reverse-scored.)

ANALYSIS

Exploratory factor analysis The elicited measures of the consumer involvement profile were subjected to an exploratory factor analytical procedure. As Bagozzi (1983) and Burnkrant and Page (1984) have pointed out, the exploratory factor analysis can mislead with respect to dimensionality. High measurement errors can inflate item correlation estimates, which in turn can lead to merging of items that measure distinct though correlated constructs (Bagozzi 1983). Therefore, the exploratory factor analysis results are presented merely as a first step. Later, we will use confirmatory factor analysis to test the hypotheses.

TABLE 1

OPERATIONALIZATION OF THE FOUR FACETS OF INVOLVEMENT AT THE PRODUCT- AND BRAND CHOICE LEVELS

An oblique solution is presented in Table 1. Six factors with eigenvalues exceeding 1.0 were extracted with a total explained variance of 70.1 %. The "cleanest" factors were those for sign value at product and brand-choice levels (Factors 5 and 4, respectively), and for product risk (Factor 3). Among the other, not-so-clean factors, Factor 2 can be interpreted to represent Brand-risk. Factor 1 is a product-importance factor but it also received significant loadings from brand- importance and brand hedonic scales. One could view it as a "general" product factor. The last factor, Factor 6, can be viewed as a "general" brand-level factor. As might be expected, there were many cross-over factor-item loadings.

Tests of the Hypotheses via Confirmatory Factor Analysis

For each of the four facets, two models (A and B) are constructed. Model A treats a given facet across the two forms of involvement as one and the same construct. Model B treats that facet as two separate constructs, one for each form of involvement. See Figure 1 for the "importance" facet. Two statistics are examined. First, the chi-square values and their significances are examined for absolute fit of the model to the data. Second, improvement in chi-square value is examined across Models A and B. The difference in the chi-square values between the two models also has a chi-square distribution and it should be significant for the difference in the degrees of freedom between the two models.

Hypothesis 1. Perceived importance differs across the product and brand choice. The results of the two factor and one-factor model are presented in Figure 1. The chi-square value for the one factor model is 5251 (degree of freedom=9, P,.0011, indicating a poor fit. This statistic for the two-factor model is 9.55 (d.f.=10, p=.481), indicating a good fit. Adjusted goodness of fit index (AGFI) and Root mean square (RMS) values also indicate the two-factor model fits much better than the one-factor model. Chi-square difference of 42.96 for a difference of 1 in degrees of freedom is highly significant at p <.05. Thus, Hypothesis 1 is supported. The inter-factor correlation of .691 does imply that the two constructs are well-correlated.

At this point, we must note that the two models similar to those shown in Figure l for the "importance" facet were also constructed for all other facets. Due to space limitations, the schematic diagrams (such as those in Figure 1) are not included here; only the chi-square fit statistics are presented below. An earlier, longer version of the paper, obtainable from authors, furnishes complete details.

Hypothesis 2. Perceived sign value differs across product and brand choice. The-one factor motel does no fit (chi square=63.65, d.f.=20, p=.001), ant the two-factor model barely approaches significance (chi-square 32.34, t.f.21, p=.054). However, the chi-square difference of 31.31 between the two models is significant (d.f.+l, p,.05), indication the superiority of the two-factor model. The correlation between the two factors is a modest.395. Thus, Hypothesis 2 is supported.

TABLE 2

ITEM CORRELATIONS

FIGURE 1

MODELING PRODUCT- AND BRAND-CHOICE IMPORTANCE AS ONE VERSUS TWO CONSTRUCTS

Hypothesis 3. Hedonic value differs across the product and brand choice. The one factor model is significant although not highly: chi-square=9.8 1, d.f.-5, p=.081 (Note: "p" values above .05 indicate significant). Two-factor models has a good a solute fit (p=.980) Also, the chi square difference is significant (difference_8.68, d.f. 1, p,.05) indicating the superiority of the two-factor model. Thus Hypothesis 3 is supported.

Hypothesis 4. Perceived risk in product is separate from Perceived risk in Brand Choice. In Table 1, under the "perceived risk" facet for the "brand-choice level" items 2 and 5 were meant to measure "risk consequences." (Item 12 and 20 measure risk-probability, and will not be further dealt with here.) These two items with the two items (items 23 and 25) of product level perceived risk were employed in the two factor versus one factor models.

Neither model fitted the data well. (The fit was bad enough that most of the statistics could not be estimated.) The correlation matrix was examined to detect the possible causes. items 2 and 5 are correlated (r = -.496) and items 23 and 25 are correlated (r = -0504); however, neither of the two correlations is very high. The cross factor correlations of items 2 and 5 on the one hand and, of 23 and 25 on the other are extremely low. For discriminate validity, these latter correlations must be and are lower than those in the convergence triangles (see Table 2). This shows that brand-choice-risk and product-choice-risk are discriminated (Campbell and Fiske 1959). Lack of any relation whatsoever (as opposed to low correlation) might have blocked the model estimate in the LISREL Procedure. The failure to obtain very high correlations between the two-items of either factor is also a cause for concern, but, in Table 2 the separation of the two factors itself is reasonably supported. Thus, the support for Hypothesis 4 is a qualified one in the LISREL estimation gave a poor fit, but the cross-factor item correlations were very low.

Hypotheses 5 thru 7. To conserve space, only the salient aspects of the analyses are briefly summarized below for these 3 hypotheses.

H5: Product Importance and Product Risk. The two-factor model fitted (chi-square 6.53, d.f.=6, p=.366) whereas the one- factor model did not (chi-square=29.11, d.f.=5, p < .0001).

H6: Brand-choice Importance and Brand-choice Risk. The one factor model was not significant (chi-square=16.86, d.f.=5, p=.005). Although the two-factor. model also did not have a significant chi-square value (p=.027), the chi-square difference statistic showed the two-factor model to be superior to the one-factor model. However, the two factors were highly correlated (r=.883).

H7: Product Importance and Brand Risk. The two factor model was significant (though not highly) with chi-square value of 11.18 (d.f.=6, p=.083), while the one-factor model was not significant (chi-square=28.12, d.f.=5, p=.0001). The chi-square difference test showed the two-factor model to be superior to the one-factor model.

For both the product and the brand level, the other two facets, namely, "hedonic" and "sign" were proven separate constructs, and each was also proven to be separate from importance and risk constructs. That is, for each pair, a two factor rather than a one factor model was supported.

Reliabilities

The coefficient alpha of internal reliability for each facet was as follows: perceived product importance, .839; perceived brand-choice importance, .880; product level sign value, .856; brand level sign value, 0.711; product level hedonic value, .779; These are considered good reliabilities. The three remaining facets comprised of two items each correlated modestly: brand level hedonic value (items 17 and 22), .561; product risk (items 23 and 25) 0.677; and perceived brand risk (items 2 and 5), .661.

Relationship of the 8 Facets with Other Consumer Behavior Variables

Six variables related to the respondents' behaviors with respect to beer were also measured (See Method section for measures) and it was of interest to examine their correlations with each of the 8 facets of the involvement profile. Regression analyses were not employed for two reasons. One, significant multicollinearity among facets can render (and in our data it did) otherwise significant predictors insignificant. Secondly, because respondents have been engaging (to a greater or lesser degree) in these behaviors already, predictor/criterion distinction between these behaviors on the one hand and the facets of involvement on the other may no longer hold. Illustratively, perceived brand differences could have once led to a feeling of brand-choice involvement; that involvement could have in turn led to paying attention to advertising, which may have fed back to an enhancement in perceived brand differences. Therefore, examining simple correlations was considered more appropriate than regression analyses.

Table 3 presents the correlations. It is immediately apparent that the brand-choice level and product level facets differ in their correlations with the 6 behavioral elements. Notable patterns are:

i) Product level risk is unrelated to any of the variables. This is because regardless of frequency of drinking, everyone seemed to at least moderately agree that drinking beer could be harmful and disagree that nothing could go wrong by drinking beer (Mean=4.80 S.D.=1.43, on a 7-point scale). We read this to mean not that people suspected harm, but rather that they could not deny the possibility of some harm.

ii) Frequency of drinking beer is related more to product level involvement than to brand level involvement. This seems reasonable.

iii) Perceived brand differences is related more to brand choice level involvement. This result seems logical.

iv) Brand commitment is related more to the brand level than to the product level of the first and fourth facets. This is logical. However, the pattern of correlations with the other two facets is not explicable.

v) Brand comparison seemed related about equally well to both levels of involvement. Although we would have expected brand level involvement to show higher influence, the equally strong influence of product level involvement is not difficult to accept (even if on a posthoc basis).

vi) Interest in reading an article about beer was related more with product level involvement than with brand level involvement. This would be expected if an article about beer can be assumed to contain information about the product per-se than about brands.

vii) Attention to advertising was only moderately related to any of the facets, but it was related about equally well with brand level and product level involvement facets.

SUMMARY AND DISCUSSION

Laurent and Kapferer's consumer involvement profile is seen here as an important device to map consumers' relationships with products. However, those authors did not distinguish between product-class- and brand-choice involvement. Rather inadvertently, they conceived and measured two of the four facets at product level and the other two at the brand level.

We sought to make this distinction for each of the four facets. Moreover, because Laurent and Kapferer did not present the entire scale, to expand their illustrative item list towards a full scale was also considered a benefit of this research. The items we developed had good internal reliabilities for perceived importance of brand choice, sign value of the product, sign value of the brand, and hedonic value from the product. The three other facets (perceived product risk, perceived brand risk, and hedonic value at the brand level) had modest reliabilities.

TABLE 3

CORRELATLONS OF SELECTED CONSUMER BEHAVIOR VARIABLES WITH VARIOUS FACETS OF THE CONSUMER INVOLVEMENT PROFILE (N=78)

While the exploratory factor analysis provided unclear factor solutions (which may be expected), the more elegant confirmatory factor procedures confirmed the separation of each facet at the product- and brand-choice levels. For perceived importance, and likewise for sign value, the support was unequivocal for the two-factor rather than one-factor model (brand-choice level and product-level involvements being modeled respectively as two constructs or as a single construct). Similar was the case for the hedonic aspect. For perceived risk, the support for the hypothesized two factor model was a qualified one. The one-factor and two-factor models were comparable (i.e., neither was superior to the other), but the cross-factor item correlations supported the discrimination between the two factors. The less than desired levels of model fit may have been due to: (a) less than perfect measures for risk at both the levels; (b) the product beer not perhaps being seen as risky, and (c) possible vagueness about the meaning of product risk. Nonetheless, there was support for each facet being different across the product- and brand-choice levels.

Some consumer behaviors were about equally related to a given facet at either level of involvement, while some other facets were differentially related. For example, perceived brand differences and brand commitment were related more with brand- level than with product-level involvement And, interest in reading about the product was related more with product level than with brand level involvement. These results are logical, and they would have remained undiscovered if the facets of the profile were not separately measured at brand-choice and product levels.

Our results are obviously limited by the use of only a single product and convenience sampling and also due to the smallness of the sample size. In addition to overcoming these tactical limitations, an important task for future research is the further theoretical and empirical explication of each facet. The distinction between the brand level and product level involvement is intuitively apparent for three of the four facets, namely, perceived importance, perceived risk, and sign value. For example, all essential items (e.g., salt, facial tissue) could be important at the product level but not necessarily at the brand level. Also, some not-so-essential products such as airline travel could be important at the product level but not at the brand choice level. Then, some products could be risky at the product level (e.g., some medical procedures) but not any more risky at the brand-level (choice of surgeon). And for many established products (e.g., appliances) brand-choice is risky but products themselves are not perceived to be risky. Finally, sign value can be associated with the product itself rather than with the brand when the product is new, or a luxury or both (e.g., cellular phone, video cameras, or diamonds).

The case of the hedonic facet is less clear. If one finds a brand hedonic, the product would seem to become hedonic inevitably. If the product is hedonic, repeated use may result into the preferred or usual brand being perceived as more hedonic than other brands. The two concepts may be difficult to separate empirically, or the relationship may be asymmetrical. The relationships among all facets (within and across product/brand-choice levels), and of each facet with other consumer behavior variables definitely need more a priori hypothesis development and empirical testing. The present research is a step in that direction.

REFERENCES

Bagozzi, R. P. (1983), "A holistic methodology for modeling consumer response to innovation," Operations Research, 31, 128- 176.

Bloch, P. H. (1983), "Involvement Beyond the Purchase Process: Conceptual Issues and Empirical Investigation," in A. A. Mitchell (ed.), Advances in Consumer Research, IX, 413-417.

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

Burnkrant, R. E. and Page, T. 1. (1984), "A Modification of the Fenigstein, Scheier and Buss self-consciousness scales," Journal of Personality Assessment, 48, 629-637.

Campbell, D. T. (1960), "Recommendations for APA Test Standards Regarding Construct, Trait and Discriminant Validity," American Psychologist, 15, 546-53.

Campbell, D. T. and D. W. Fiske (1959), "Convergent and Discriminant Validation by the Multitrait-Multimethod Matrix," Psychological Bulletin, 56 (March), 81-105.

Houston, M. I. and M. L. Rothschild (1977), "A Paradigm for Research on Consumer Involvement," WP No. 11-7746, University of Wisconsin-Madison.

Joreskog, K. G. and D. Sorbom (1986) LISREL VI, User's Guide, Chicago: National Educational Resources.

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

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

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

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