Consumer evaluations of brand extensions: evidence from India.

ABSTRACT

Aaker & Keller's (1990) article is a seminal study in the area of systematic brand extension research. More than eight replications of original study have been conducted in the developed countries. We replicate and extend the original study in India to examine the generalizability and robustness of the Aaker & Keller's model in the context of developing economies. Our study provides support for three out of four hypotheses of Aaker and Keller's model. QUALITY of original brand and three fit variables, TRANSFER, COMPLEMENT and SUBSTITUTE, have strong positive effect on consumers' attitude towards the brand extension. Only the interaction of QUALITY and TRANSFER is positive and significant and hence hypothesis 2 is partially supported. The relationship of difficulty in making the extension with customers' attitude towards the brand extension could not be supported. We find support that consumers' attitudes towards brand extensions are driven primarily by the main effects.

1. INTRODUCTION

Brand extension is a popular brand strategy to attach an existing brand name to a new product introduced in a different product category (Swaminathan, Fox and Reddy, 2001). Brand extension strategies are widely employed because of the beliefs that they build and communicate strong brand positioning, enhance awareness and quality associations, and increase the probability of trial by lessening new product risk for consumers (Taylor and Bearden, 2002). In 1990, 63 percent of products introduced were line extensions, and another 18 percent were category extensions (Keller, 1999). Extending brands both within and beyond the original product category enables firms to tap into their substantial brand equity investment (Lane, 2000). Through brand extension, new offerings are stamped with instant credibility and a built-in endorsement from the underlying brand (Davis and Halligan, 2002). The brand extensions help in increasing the chance of new product success and lower its launch costs (Kapferer, 1997). Through the history of brand extensions, multiple successes have been observed. However, not all brand extensions have been successes (Gronhaug, Hem and Lines, 2002). Brand extension can generate enormous profits but it can be dangerous, too. An ill-conceived brand extension may seriously damage the original product and preclude the establishment of another brand with its unique associations and growth potential (Aaker, 1990).

We have structured the article as follows. Section 2 discusses the Aaker and Keller's (1990) original study (henceforth A&K) and different replications done around the world. In section 3, we discuss the research design of the study, describing the model being tested and the data collection method. In section 4 and section 5, we present the results. The article concludes with a number of theoretical and managerial implications of replication of A&K in Indian context.

2. THE AAKER AND KELLER MODEL AND DIFFERENT REPLICATIONS

The area of brand extension has received considerable interest of marketing academic researchers and practitioners during the last decade (Barwise 1993; Chen and Paliwoda, 2002; Grime, Diamantopoulos and Smith, 2002; Zhang and Sood, 2002). The past 15 years have witnessed the development of an important body of empirical evidence on consumer attitude vis-a-vis brand extensions (Czellar, 2003). A&K's seminal study was the first systematic research on consumer behavior towards brand extension. More than eight replications of A&K original study have been conducted in different parts of world. Sunde and Brodie (1993) in New Zealand (henceforth S&B), Nijssen and Hartman (1994) in Holland, Alexandre-Bourhis's (1994)in France, Bottomley and Doyle's (1996)in UK, Holden and Barwise's (1995, 1996)in three different places UK, France and US, replicated A&K study. Riel, Lemmink and Ouwersloot (2001) replicated and extended A&K in the services context. Almost all the replications of A&K study were conducted in the developed countries. Our replication of this study in India allows us to examine the generalizability and robustness of the A&K study in the context of developing economies. To be applicable to brand managers in different countries, the findings must possess external validity (Gronhaug, Hem and Lines, 2002). Barwise (1993) first stressed the need for further replication of the A&K study, study using identical stimuli. Such replication studies are needed to examine the robustness of A&K's results as well as their sensitivity to cross-cultural differences with other experimental conditions held constant (Barwise, 1993). Replications are needed in order to develop empirical generalizations (Bottomley and Doyle, 1996). A & K's study shows that

* the "Fit" between the original and extension product classes has a direct positive association with the attitude toward the extension. The three dimension of fit are the perceived applicability of the skills and assets of a competent manufacturer in the original product class for making the product extension (TRANSFER), the perceived product class complementarity (COMPLEMENT), and the perceived product class substitutability (SUBSTITUTE).

* the complementarity and substitutability fit measures and perceived quality of the original brand has an interactive effect on evaluation of an extension.

* perceived difficulty (DIFFICULT) in making the extension has a positive relationship with evaluation of an extension.

However A&K's study shows that there is no direct link from perceived quality of the brand to the attitude towards the extension.

Despite the wide acceptance and diffusion of A&K's findings, almost all the replications gavn varying results and thus questioning the empirical generalizability of A&K's original findings. The first lack of agreement between A&K and replication emerged in Sunde and Brodie (1993) New Zealand study. In contrasts to A&K, S&B replication study provided strong evidence for a direct relationship of QUALITY of the original brand and the attitude toward the extension, without the need for some dimension of "fit". Another difference between A&K's and S&B's was New Zealand study failed to support relationship between difficulty in making the extension and consumer evaluation of extension. One possible reason for this, as given by S&B, was that there might be a cultural difference between New Zealanders and Americans--with the latter being more "informed" and experienced consumers and a lot more wary in terms of paying higher prices for branded products. On S&B findings, Barwise (1993) commented "this paper exemplifies why we need more replications in marketing: if a result has not been shown to generalize, it is not--or not yet--a scientific finding". Aaker and Keller (1993) argued that the cross-cultural differences in S&B's replication--New Zealand versus US business students and mostly different brands, could be the possible explanation for the differences in the results. Only one of the six parent brands (stimuli) was common to both studies. The results of interactive effect of QUALITY and fit measures on extension evaluations in S&B were inconclusive perhaps because of presence of multicollinearity. To deal with the problem of multicollineaity, Nijssen and Hartman (1994) used "Residual centering technique" in their Dutch replication of A&K study. Contrary to A&K's finding, Nijssen and Hartman (1994) also found strong support for individual effect of parent brand quality on customer's attitude towards brand extension. Similarly Alexander-Bourhis (1994) and Bottomley and Doyle (1996) reported different results than A&K's findings. Correcting for multicollinearity, Bottomley and Doyle (1996) found that consumers' attitudes towards brand extensions were driven primarily by the main effects and moderated via the interaction terms. Moreover they didn't find support for Aaker and Keller (1993) contention that different stimuli (both parent brands and extensions) might have caused the discrepancies between A&K's and S&B's findings.

Bottomley and Holden (2001) did a secondary analysis using a comprehensive data set containing the data from the original study and seven replications, to investigate the empirical generalizability of A&K's model. They found support for the A&K's full model despite published results, including Aaker and Keller's own that support only some of the hypotheses. Their finding also supported that the level of contribution of each of components in the model varies by brand and culture. Gronhaug, Hem and Lines (2002) investigated the impact of similarity and brand reputation in Scandinavian setting and found that evaluations of brand extensions co-vary positively with perceived similarity to, and reputation of, the parent brand. Their results corroborated previous findings, and indicated that they possessed external validity and thus could be generalized--at least to some extent--across national borders. Riel, Lemmink, and Ouwersloot (2001) replicated A&K's study in the services domain and found the support for the direct effect of perceived quality of the original brand, perceived fit on the evaluations of extensions and moderating role of quality with respect to fit dimension. However, they didn't find support for relationship between difficulty in making extension and evaluation of extension.

3. RESEARCH DESIGN

We essentially test the following model of A&K (1990) and subsequent replicated by other researchers:

Aext = [[beta].sub.0] + [[beta].sub.1]QUALITY + [[beta].sub.2]TRANSFER + [[beta].sub.3]COMPLEMENT + [[beta].sub.4]SUBSTITUTE + [[beta].sub.5]QUALITY*TRANSFER + [[beta].sub.6]QUALITY*COMPLEMENT + [[beta].sub.7]QUALITY*SUBSTITUTE + [[beta].sub.8]DIFFICULT

Based upon above model, we test the following four hypotheses:

1. Higher quality perceptions towards the original brand (i.e. higher quality) are associated with more favorable attitudes toward the extension.

2. The transfer of a brand's perceived quality is enhanced when the two product classes in some way fit together. When the fit is weak, the transfer is inhibited.

3. The fit between the two involved product classes has a direct positive association with the attitude towards the extension.

4. The relationship between the difficulty of making the product class of the extension, 'DIFFICULT' and the attitude towards the extension is positive.

Five Indian brands were chosen using Aaker and Keller's criteria (i.e., relevance to the respondents, high quality, having a strong brand image, and not having been extended broadly before). We attempted to select identical product class categories for parent brand and the extension. Based on response of two focus groups consisting of 20 students, similar to subjects in actual data collection, we didn't take beer product class category for parent brand. In India, consumption of beer or any other alcoholic beverages is not seen as a part of culture and consumption is not very prevalent among the subjects of study. Out of five brand chosen, four (Colgate, Clinic Plus, Ray-Ban and Baskin Robbins) are well known brands by multi-national corporations operating in India. The fifth brand "Barista" is essentially Indian brand associated with premium coffee retailing.

A questionnaire was developed using similar rating scales to those used by A&K, S&B and Bottomley and Doyle (1996). The questionnaire was administered to 106 Indian students. All respondents were enrolled for MBA program at the XLRI Jamshedpur, India. Subjects varied in age between 18 and 26 years old, 80% were male, and 20% were female. In the A&K original study and most of the replication studies, the subjects were the students, with the exception of the Nijssen and Hartman (1994) study, which used a stratified sample of the Dutch population. The dependent variable for the models was the attitude towards the brand extension, operationalized as the average of the perceived quality of the extension and the likelihood of trying the extension. The correlation between two indicators of Attitude is found to be 0.63. Total 106 subjects evaluated Colgate, Ray-Ban, Barista and Clinic Plus brands. Only 38 subjects could evaluate Baskin Robbins, because of simple omission. Hence we got total 924 (108*8 plus 38*2) brand extension observation. Five (in some case four because of omission) brands at a time didn't place heavy measurement burden for each respondent.

We estimate regression models at both the aggregate level (by treating each subject's response to each brand extension as a separate observation), and also at the individual brand extension level. A&K, S&B and Bottomley and Doyle (1996) reported the presence of high degrees of multicollinearity between the main effects and interaction terms, which made the interpretation of results difficult. For example, in S&B study, correlations between TRANSFER, COMPLEMENT and SUBSTITUTE and their corresponding interactions with QUALITY were 0.88, 0.92, and 0.93 respectively. In our study also we find very high correlation of 0.94, 0.95, and 0.96 between TRANSFER, COMPLEMENT and SUBSTITUTE and their corresponding interactions with QUALITY respectively (Table 1).

Following Riel, Lemmink, and Ouwersloot (2001), we find Variance Inflation Factors (VIFs) as high as 48.6. VIF indicates the extent to which each independent variable is explained by the other independent variables and large VIF values (a usual threshold is 10.0) indicate a high degree of multicollinearity among the independent variables (Hair et al., 1998). Nijssen and Hartman (1994) in Dutch replication first used Lance (1988) "residual centering method" to deal with the effect of multicollinearity.

Residual centering approach is a two step regression procedure. In step 1, each interaction term is regressed on its component parts.

QUALITY * FIT= [??] + [[??].sub.1] QUALITY + [[??].sub.2] FIT.

In second step, residual term retrieved from step 1 is used in place of the respective original interaction term and then full effect model is estimated using ordinary least square.

ATTITUDE = [[??].sub.0] + [[??].sub1] fit QUALITY + [[??].sub.2] FIT + [[??].sub.3] QUALITY * FIT (Residual).

After using residual centering method, multicollinearity has almost disappeared, VIF values in all cases are lower than 1.7. Beta coefficients for full effect model after applying residual centering method are given in Table 2. Table 2 also includes the beta coefficients of secondary analysis by Bottomley and Hoiden's (2001). We find strong support for the positive relationship of QUALITY of original brand and three fit measures with the customer's attitude towards extension. All the standardized regression coefficients related to QUALITY, TRANSFER, COMPLEMENT and SUBSTITUTE are significant. Hence A&K hypothesis 1 and 3 are also supported in Indian context. Except in A&K's study, in all replications hypothesis 1 was supported. In A&K study the coefficient for QUALITY was nonsignificant. In Bottomley and Holden's (2001) secondary analysis of original study and seven replications hypothesis 1 was strongly supported in all data sets including for A&K study. Riel, Lemmink, and Ouwersloot (2001) also found direct positive relationship between perceived quality of the original brand and extension evaluation for both services and goods. Similarly Bottomley and Holden (2001)found standardized coefficients related with three fit variables significant virtually in all studies. Only in Nijssen and Hartman (1994) data set, coefficient for TRANSFER was nonsignificant. We also find strong support for positive effect of all three fit variables on extension evaluation. Beta coefficients of TRANSFER and COMPLEMENT were higher than SUBSTITUTE, showing their relatively more important effect on extension evaluation than SUBSTITUTE. Another possible reason for this may be the limited number of brand extensions of a substitute nature contained in our study as hypothetical extensions.

We find mixed support for the interaction of QUALITY with the three fit measures. Interaction of QUALITY and TRANSFER is positive and significant. However interaction of QUALITY and SUBSTITUTE is negative and significant. The possible explanation for negative beta coefficient is that consumers find brand extensions launched in substitute product categories confusing. In other word, brand extensions that compete with the parent brand may appear some what more irrational to the consumer for higher quality brands than for comparatively weaker brands (Bottomley and Holden, 2001). Hence hypothesis 2 is only partially supported. Bottomley and Holden (2001) found mixed support for the interaction of QUALITY with each of the three fit variables in their secondary analysis.

The interaction of QUALITY and TRANSFER was significant in five of the eight studies, the interaction of QUALITY and COMPLEMENT was significant in four studies, and the interaction of QUALITY and SUBSTITUTE was significant in two studies, but in one case, the relationship was negative. Thus in our study also it comes out that interaction effect is secondary to the main effects in determining consumers' overall attitude towards the brand extension. This is converse to results of A&K study which didn't correct for multicollinearity. Consumers' attitudes towards brand extensions are driven primarily by the main effects and moderated by the interaction effects (Bottomley and Doyle, 1996; and Bottomley and Holden 2001).

In A&K study fourth hypothesis postulating positive relationship between difficulty of making extension and extension evaluation was supported. The possible reasons given for this were that customer may feel an extremely easy-to-make extension incongruous and even exploitative. We find coefficient related with DIFFICULT nonsignificant and hence hypothesis 4 could not be supported. In Bottomley and Holden (2001) secondary analysis, in four out of seven studies, beta coefficients were significant. One possible reason for this, as speculated by S&B also, there may be a cultural difference between respondents from developing countries (India in present study) and developed countries--with the former being less "informed" and experienced consumers and less price sensitive for branded products.

In main effect model also, we find support for positive relationship of QUALITY and three fit variables with the extension evaluation. Beta coefficients related with QUALITY, TRANSFER, COMPLEMENT and SUBSTITUTE are 0.23, 0.36, 0.31 and 0.11 respectively (all significant at the 0.01 level) are relatively higher for TRANSFER and COMPLEMENT than SUBSTITUTE. Beta coefficient related with DIFFICULT is nonsignificant and hence hypothesis 4 could not be supported in main effect model also. We find R2 value is (0.54) much higher than the original study (0.26). In subsequent replication R2 value had increased. A possible explanation could be that brand extensions have become much more prevalent over the years, Successful extensions could thereby have set or reinforced the standards for evaluation in consumers, at the same time improving the subsequent explanatory power of the independent variables (Riel, Lemmink, and Ouwersloot, 2001).

In order to see the effect of brand knowledge and familiarity on extension evaluation, we introduced another variable brand usage in main effect model. Broniarczyk and Alba (1994) found brand knowledge role as a moderating variable in extension evaluation. Brand association represents as aspect of brand knowledge and strongly influence how consumers evaluate brand extension. Brand usage as another independent variable in main effect model explained only an additional variance of 2.5 percent Nevertheless beta coefficient for usage is significant at the 0.01 level. This indicates the subject-specific effects such as usage also act as moderating variables in extension evaluation.

5. EXTENSION LEVEL ANALYSIS

In brand level analysis, we disaggregate the data to the individual brand extension level. This allows us to examine whether the same results are found across all or most extensions. Bottomley and Doyle (1996) argued "although the principle of statistical generalization to other subjects is well understood researchers often neglect the principle of generalization to other materials (in this case brand extension)" Brand extension level results have managerial implications also, because it is at this level that managers must make decisions about future branding strategies (Bottomley and Holden, 2001).

In the first stage we examined the effect of interaction terms on consumers' attitude towards the brand extension. Similar to residual centering approach of aggregate level analysis, we regressed the three sets of residuals of interaction terms are regressed on consumers' attitude towards the brand extension (Table 3). As we have fewer observations for Baskin Robbins extension, we dropped its two extensions at this level of analysis. Evidence from the past studies indicated that the chance of making a Type II error was in excess of .50 under similar conditions with relatively few observations (Mason and Perreault, 1991 Bottomley and Holden, 2001). In Table 3, the coefficient for interaction term is significant only in one case (Q*T residual in case of Colgate Mouthwash) out of three interaction terms for eight brand extensions. The three interaction terms on average explain 2.2 percent of variance. Hence full effects model is rejected in favour of the main effects models. The main effect model is presented in Table 4. In terms of statistical significance, 4 out of 8, 7 out of 8, 6 out of 8, 3 out of 8 and 2 out of 8 beta coefficient is significant for QUALITY, TRANSFER, COMPLEMENT and SUBSTITUTE respectively. It seems that the aggregate main effect model is applicable in most of the individual level brand extension.

TRANSFER and COMPLEMENT seem to more important variable of fit than SUBSTITUTE. Bottomley and Holden 2001, in their secondary analysis at brand extension level, also found strong support for the contribution of QUALITY and the three fit variables. Of 131 brand extensions, coefficients for QUALITY, TRANSFER, COMPLEMENT, SUBSTITUTE and DIFFICULT were significant for 68, 56, 57, 36 and 14 extensions respectively. Brand extension-level analysis supports the findings of the aggregate level analysis discussed in previous section. Main effects primarily influence consumers' evaluations of the majority of brand extensions but interaction terms have weaker role in extension evaluation.

6. CONCLUSION

Since all the replications of A&K study were conducted in the developed countries, our Indian replication allows us to examine the generalizability and robustness of the original study in developing economies context. Our study provides strong support for three out of four hypothesis of A&K model. QUALITY of original brand and three fit variables, TRANSFER, COMPLEMENT and SUBSTITUTE have strong positive effect on customers' attitude towards the brand extension. Only the interaction of QUALITY and TRANSFER is positive and significant and hence hypothesis 2 is partially supported. The relationship of difficulty in making the extension with customers' attitude towards the brand extension could not be supported. In brand level analysis, the full effects model couldn't be generalized across brand extensions while mail effects model was present in majority of extensions. We find support that consumers' attitudes towards brand extensions are driven primarily by the main effects.

7. LIMITATIONS AND DIRECTION FOR FUTURE RESEARCH

One of the limitations of the present study is, like original study and subsequent replication; we have used single item scales to measure the Attitude and the perceived fit dimension. As suggested by Bottomley and Doyle (1996), multiple item scale to measure the customers' attitude towards extension should be developed. The three fit variables are strongly correlated among themselves, and hence further improvement is possible in terms of development of sufficiently distinct constructs (Riel, Lemmink, and Ouwersloot, 2001). Further better conceptualization for perceived fit should be developed. The degree of perceived fit is a function of both product-feature similarity perceptions and brand-concept-consistency perceptions (Park, Milberg and Lawson, 1991). Whereas A&K focused only on product-feature relationship, concept-consistency perceptions rely on the extension product's ability to accommodate the brand concept. The original study and almost all replication investigated consumers' attitude toward extensions in controlled conditions in a marketplace vacuum. Limited information environment restrict generalization of results (Taylor and Bearden, 2002). However consumers' attitudes toward the extension are sensitive to competitor activity, retailer-level decisions as well as other information sources like press, consumer reports and word-of-mouth (Czellar, 2003). In-store factors such as the spatial proximity of a brand and its extension may also influence (or reduce) the effect of similarity (Broniarczyk and Alba, 1994). Researcher should investigate this and other local environmental factors. Further research should be done in more real-life-like environment. Longitudinal research to investigate the stability of consumers' attitude toward extension over time is needed.

TABLE: 1 CORRELATION COEFFICIENT MATTRIX

             ATTIT-    QUALI-    TRAN-    SUBS-    COMP-    DIFFI-
             UDE       TY        SFER     TITU.    LEM.     CUL.

ATTITUDE      1
QUALITY       0.34     1
TRANSFER      0.61     0.1        1
SUBSTITU.     0.41     0.08       0.45     1
COMPLEM.      0.6      0.22       0.52     0.38     1
DIFFICUL.    -0.24     0.05      -0.3     -0.13    -0.3      1
Q*T           0.65     0.4        0.94     0.42     0.54    -0.27
Q*S           0.46     0.32       0.44     0.95     0.41    -0.11
Q*C           0.61     0.47       0.48     0.34     0.95    -0.27

             Q*T      Q*S     Q*C

ATTITUDE
QUALITY
TRANSFER
SUBSTITU.
COMPLEM.
DIFFICUL.
Q*T          1
Q*S          0.51     1
Q*C          0.6      0.45    1

TABLE 2: FULL EFFECT MODELS AFTER RESIDUAL
CENTERING: BETA COEFFICIENT

                A&K        N&H        B&D        S&B        LBS

QUALITY         0.22 **    0.25 **    0.22 **    0.25 **    0.39 **
                (11.50)    (9.00)     (11.40)    (13.10)    (21.20)
TRANSFER        0.26 **    0.60 **    0.31 **    0.27 **    0.14 **
                (12.00)    (19.20)    (13.80)    (12.60)    (6.80)
COMPLEMENT      0.17 **    0.00       0.31 **    0.30 **    0.09 **
                (8.20)     0.00       (13.80)    (14.50)    (4.80)
SUBSTITUTE      0.07 **    0.07 *     0.18 **    0.18 **    0.18 **
QUALITY         (3.40)     (2.30)     (7.60)     (8.40)     (8.70)
TRANSFER        0.03       0.08 **    0.08 **    0.08 **    0.06 **
                (1.40)     (2.70)     (3.40)     (4.10)     (3.00)
QUALITY         0.06 **    -0.02      0.05 *     0.05 *     0.05 *
COMPLEMENT      (3.10)     (0.5)      (2.30)     (2.30)     (2.50)
QUALITY         0.05 *     -0.07 *    0.03       -0.1       0.03
SUBSTITUTE      (2.20)     (2.3)      (1.20)     (-0.6)     (1.60)
DIFFICULT       0.11 **    OFR        0.01       0.03       0.12 **
                (5.40)                (0.60)     (1.60)     (6.20)
Adjusted R 2    0.25       0.49       0.49       0.44       0.28
Sample size     2101       693.00     1358.00    1559       2204

                Illinois   Essec      A-B        India

QUALITY         0.21 **    0.46 **    0.09 **    0.23 **
                (7.50)     (23.70)    (3.80)     (10.00)
TRANSFER        0.07 *     0.19 **    0.27 **    0.36 **
                (2.30)     (8.60)     (10.80)    (12.80)
COMPLEMENT      0.10 **    0.15 **    0.29 **    0.31 **
                (3.50)     (7.90)     (11.50)    (11.20)
SUBSTITUTE      0.22 **    0.15 **    0.29 **    0.11 **
QUALITY         (6.80)     (7.10)     (11.90)    (4.10)
TRANSFER        0.03       0.05 *     0.03       0.05 **
                (0.80)     (2.20)     (1.20)     (1.90)
QUALITY         (0.02)     0.02       0.02       -0.05
COMPLEMENT      (-0.8)     (0.80)     (0.90)     0.00
QUALITY         -0.01      0.02       0.01       -0.06 *
SUBSTITUTE      (-0.4)     (1.00)     (0.20)     (-2.1)
DIFFICULT       0.10 **    0.07 **    0.03       -0.03
                (3.50)     (3.60)     (1.10)     (-1.2)
Adjusted R 2    0.17       0.36       0.38       0.54
Sample size     1086       1793       1151        924

Figures in parentheses are student t-values. OFR = omitted from
research. A&K = Aaker and Keller, 1990; N&H = Nijssen and Hartman,
1994; B&D = Bottomley and Doyle, 1996; S&B = Sunde and Brodie, 1993;
A-B = Alexander-Bourhis, 1994. Holden and Barwise's three data sets are
referred to by the university where the data were collected (i.e., LBS,
ESSEC, and Illinois). * p < 0.05, ** p < 0.01

TABLE 3: BRAND LEVEL INTERACTION TERMS: BETA COEFFICIENTS

                                          Q * T         Q * C
Brand extension              [R.sup.2]    (Residual)    (Residual)

Barista Fast Food            0.010        -0.01          0.042
Barista Theme park           0.020         0.04         -0.117
Ray-Ban Wallets              0.012        -0.083        -0.03
Ray-Ban Sportswear           0.020        -0.024        -0.104
Clinic Plus Suntan lotion    0.030         0.126         0.037
Clinic Plus Perfume          0.012        -0.051         0.126
Colgate Mouthwash            0.060         0.289 **     -0.096
Colgate Chewing gum          0.008        -0.141         0.071

                             Q * S
Brand extension              (Residual)

Barista Fast Food            -0.101
Barista Theme park            0.117
Ray-Ban Wallets              -0.013
Ray-Ban Sportswear           -0.042
Clinic Plus Suntan lotion     0.077
Clinic Plus Perfume          -0.012
Colgate Mouthwash            -0.167
Colgate Chewing gum           0.038

* p<.1 & ** p<.05

TABLE 4: BRAND LEVEL MAIN EFFECTS MODELS: BETA COEFFICIENTS

Brand                        QUAL-       TRANS-      COMPL-
extension       [R.sup.2]    ITY         FER         EMENT

Barista Fast    0.62          0.404 *    0.206 *     0.3 **
Food
Barista         0.35         -0.023      0.395 **    0.218 *
Theme park
Ray-Ban         0.3           0.205 *    0.363 **    0.197 *
Wallets
Ray-Ban         0.33          0.084      0.359 **    0.161
Sportswear
Clinic Plus     0.48          0.219 *    0.303 **    0.36 **
Suntan
lotion
Clinic Plus     0.39          0.236 *    0.076       0.452 **
Perfume
Colgate         0.3          0.112       0.469 **    0.093
Mouthwash
Colgate         0.55         -0.008      0.558 **    0.241 *
Chewing
gum

Brand           SUBST-       DIFFI-
extension       ITUTE        CULT

Barista Fast     0.144 *     -0.1
Food                         *
Barista          0.254 **    0.241 *
Theme park
Ray-Ban          0.033       0.063
Wallets
Ray-Ban          0.1         -0.2 *
Sportswear
Clinic Plus      0.144       0.02
Suntan
lotion
Clinic Plus      0.042       -0.12
Perfume
Colgate         -0.187 *     0.105
Mouthwash
Colgate         -0.002       0.05
Chewing
gum

* p <.05 & ** p <.01

8. REFERENCE

Aaker, D. A., "Brand extensions: The good, the bad, and the ugly", Sloan Management Review, Vol. 31, 1990, 47-56.

Aaker, D. A. and Keller, K. L., "Consumer evaluations of brand extensions", Journal ofMarketin.q, Vol. 54 (January), 1990, 27- 41.

Aaker, D.A. and Keller, K. L., "Interpreting cross-cultural replications of brand extension research", International Journal of Research in Marketing, Vol. 10 (1), 1993, 55-59.

Alexandre-Bourhis, Nathalie, "L'Evaluation des Extensions de Marque par les Consommateurs: Une Etude Empirique", Workinq Paper No. 49, IAE-Caen, France, 1994.

Barwise, Patrick, "Introduction to the Special Issue on Brand Equity" International Journal of Research in Marketing, Vol. 10 (1), 1993, 3-8.

Bottomley, Paul A. and John R. Doyle, "The Formation of Attitudes Towards Brand Extensions: Testing and Generalising Aaker and Keller's Model", International Journal of Research in Marketing, Vol. 13 (4), 1996, 365-77.

Bottomley, P. A. and Holden, Stephen J.S., "Do we really know how consumers evaluate brand extensions? Empirical generalizations based on secondary analysis of eight studies", Journal of Marketing Research, Vol. 38 (November), 2001, 424-500.

Broniarczyk, Susan M. and Joseph W. Alba, 'q-he Importance of the Brand in Brand Extension," Journal of Marketing Research, Vol. 31 (May), 1994, 214-28.

Chen, Junsong and Paliwoda, Stanley, "Adoption of new brands from multi-branding firms by Chinese consumers", Journal of Euro--Marketing, Vol. 12 (1), 2002, 63-77.

Czellar, Sandor, "Consumer attitude toward brand extensions: an integrative model and research propositions", International Journal of Research in Marketing, Vol. 20 (1), 2003, 97-115.

Davis and Halligan, "Extending your brand by optimizing your customer relationship", The Journal of Consumer Marketing, Vol. 19, 2002, 7-11.

Grime, Ian., Diamantopoulos, Adamantios and Smith, Gareth, "Consumer evaluations of extensions and their effects on the core brand: Key issues and research propositions". European Journal of Marketing, Vol. 36 (11/12), 2002, 1415-1438.

Gronhaug, Hem and Lines, "Exploring the impact of product category risk and consumer knowledge in brand extensions", Journal of Brand Management, Vol. 9 (6), 2002, 463-476.

Hair, Joseph F., Rolph E. Anderson, Ronald L. Tatham, and William C. Black, Multivariate Data Analysis, 5th ed., Prentice Hall, New Jersey, 1998.

Holden, Stephen J.S. and Patrick Barwise, "An Empirical Investigation of What It Means to Generalise", in Holden, Stephen J.S. and Patrick Barwise, "Generalizing from Replication Studies: An Exploratory Case Study", Workinq Paper 96-004, Australian Graduate School of Management, 1996.

Kapferer, J. N., Strategic brand management, 2nd ed., Kogan Page, London, 1997.

Keller, K. L., Strategic brand management: building, measuring, and managing brand equity, Prentice Hall, New Jersey, 1998.

Lane, Vicki R., "The impact of ad repetition and ad content on consumer perceptions of incongruent extensions", Journal of Marketing, Vol. 64 (April), 2000, 80-91.

Mason, Charlotte and William D. Perreault, "Collinearity, Power, and Interpretation of Multiple Regression Analysis", Journal of Marketing Research, Vol. 28 (August), 1991,268-80.

Nijssen, E.J. and D. Hartman, "Consumer evaluations of brand extensions: An integration of previous research", In: J. Bloemer, J. Lemmink and H. Kasper (eds.), Proceedings of 23rd. European Marketing Academy Conference, Maastricht, 1994, 673-683.

Park, Whan, Sandra Milberg, and Robert Lawson, "Evaluation of Brand Extensions: The Role of Product Feature Similarity and Brand Concept Consistency", Journal of Consumer Research, Vol. 18 (September), 1991, 185-93.

Riel, A. C. R. V., Lemmink, J and Ouwersloot, H., "Consumer evaluations of service brand extensions", Journal of Service Research, Vol. 3 (3), 2001,220-231.

Sunde, L. and Brodie, R. J., "Consumer evaluations of brand extensions: Further empirical results", International Journal of Research in Marketing. Vol. 10(1), 1993, 47-53.

Swaminathan, Fox and Reddy, "The impact of brand extension introduction on choice", Journal of Marketing, Vol. 65 (October), 2001, 1-15.

Taylor, Valerie A and Bearden, William O., "The effects of price on brand extension evaluations: The moderating role of extension similarity", Journal of the Academy of Marketing Science, Vol. 30 (2), 2002, 131-140.

Zhang, Shi and Sood, Sanjay, ""Deep" and "surface" cues: Brand extension evaluations by children and adults", Journal of Consumer Research, Vol. 29 (June), 2002, 129-141.

Sanjay K. Patro is a professor of marketing at the XLRI, Jamshedpur, India. His research interests include strategic marketing, brand management and consumer behavior.

Anand K. Jaiswal is a doctoral candidate in marketing at the XLRI, Jamshedpur, India. His research interests include service quality, customer satisfaction and electronic business.