Consumer Evaluations of Brand Extensions: Empirical Evidence From China

ABSTRACT - This paper examines the applicability of Aaker and Keller’s (1990) US developed model of consumers’ evaluation of brand extensions in China, a developing economy where consumers are less familiar with brands. Residual Centering regression is used to correct the collinearity between main and interactive variables in a regression model (Lance 1988). The results re-confirmed Bottomley and Doyle’s (1996) conclusion that consumers’ attitudes towards brand extension are primarily driven by the main effects that are moderated by interaction terms. In contrast to the earlier results from developing economies, this study found that extending established brand to a product that the consumer sees as being easy to make can unfavorably affect consumers’ attitudes towards the extension.


Fu Guoqun and John Saunders (2002) ,"Consumer Evaluations of Brand Extensions: Empirical Evidence From China", in AP - Asia Pacific Advances in Consumer Research Volume 5, eds. Ramizwick and Tu Ping, Valdosta, GA : Association for Consumer Research, Pages: 395-399.

Asia Pacific Advances in Consumer Research Volume 5, 2002      Pages 395-399


Fu Guoqun, Peking University, China

John Saunders, Aston Business School, England


This paper examines the applicability of Aaker and Keller’s (1990) US developed model of consumers’ evaluation of brand extensions in China, a developing economy where consumers are less familiar with brands. Residual Centering regression is used to correct the collinearity between main and interactive variables in a regression model (Lance 1988). The results re-confirmed Bottomley and Doyle’s (1996) conclusion that consumers’ attitudes towards brand extension are primarily driven by the main effects that are moderated by interaction terms. In contrast to the earlier results from developing economies, this study found that extending established brand to a product that the consumer sees as being easy to make can unfavorably affect consumers’ attitudes towards the extension.


A large volume of academic research has followed the increasing importance of branding and brand equity that has occurred in the early 1980’s (Shocker and Weitz, 1988; Farquhar, 1989; Aaker and Keller, 1990). The very high cost of launching new brands has led most leading firms to use brand extensions as a means of launching new products (Smith and Park, 1992). Nike’s extension of its sports shoes name into sports clothing is a successful example. Cadbury’s extension of their chocolate brand to Cadbury’s Smash instant mashed potato was less successful. Although the academic contribution to understanding branding has been modest, the study of brand extension is an exception (Barwise, 1993). Brand extension studies have dominated branding research, with two thirds of branding literature in the area (Laforet, 1995). Much recent research has followed Aaker and Keller’s (1990) seminal work that developed and tested a theoretical model of brand extensions. This study examines whether Aaker and Keller’s model applies as well in a developing economy, where brands are little understood, as it does in the developed economies where the model was developed and tested.


There are several rationales for extending existing brands rather than launching new ones:

1. Brand extensions can reduce the introductory cost of the new product by trading the recognition and reputation of the parent brand (Smith and Park, 1992).

2. Brand extension could increase the chance of new product success. According to Bragg (1986) only two out of ten new brands are successful. Faced with such risks of failure, firms opt for brand extension as a safe means to growth.

3. Retailers are more likely to give shelf space to a brand they already know.

4. Retailers’ own labels increasingly challenge manufacturers’ brands. Some retailers use the threat of producing their own brands "to keep manufacturers in line" (Low and Fullerton, 1994). In an increasingly busy market place, brand extensions allow manufacturers’ brands to hold more shelf space and retain a higher profile in the consumer’s mind (Farquhar, Herr and Fazio, 1990).

In developing markets there are other reasons why brand extensions may be preferred to launching a series of independent brands. In developing markets, brand identities are less well established than in mature markets so could be more malleable. In addition, brands, such as Coca-Cola, that are mature in the developed world can be exciting and prestigious in emerging markets and so more able to add value to brand extensions. Finally, manufacturers may need to launch a huge range of products in a developing market in order to gain scale economies. Brand extensions also offer the chance of sharing that cost across several products. Against these advantages is the lack of brands with clearly established reputations in developing markets.

Aaker and Keller (1990) theorized that the consumer acceptance (A) of the brand extension is more likely to be positive if the following qualities exist.

Quality (Q): if the parent brand has a high perceived quality it can lend its aura of quality to an extension. For example, the Sony brand lending its reputation to its new line of Palm Top Organizers.

Fit (F): if there is a good perceived fit between the original product class and the proposed product extension, the values of the donor brand can flow to the extension. For example, Colgate markeing sugar free dental chewing gum.

Difficult (D): if the extension is regarded as a significant development that is not too easy to make. An incongruity occurs if consumers perceive the extended product to be exploited by a "trivial" extension. The incongruity of using an established brand on a "trivial" extension may itself trigger rejection or could lead to the belief that the brand name implies a higher than is justified price. One such trivial example was the BBC’s launch of blank audiocassette tapes.

Fit has three components:

Transfer (T) reflects the perceived ability of any firm operating in the first product class to make a product in the second product class. If consumers feel that the people, facilities and skills a firm uses to make the original product is helpful, the favorable attitude or associations about the original product may transfer to the extension. Disney realize this as a potential problem so do not use their name on their adventure movies aimed at teenagers.

Complement (C) reflects the extent to which consumers view two product classes as complementary. Products are considered complementary if they are consumed jointly to satisfy some particular need. For example, pen and ink are more complementary.

Substitute (S) reflects the consumers’ view of two product classes as alternatives to be used to perform the same function, for example, a pen and ballpoint are substitutes.

Aaker and Keller add interaction effects where fit moderates the transfer of perceived quality to extension. Even if the donor brand is perceived to have high quality, the transfer of the quality to the extension is diminished if transfer, complement or substitute is not strong. Rolls Royce could be an example. The brand name is synonymous with quality but the ability of that quality to transfer to fine wines is moderated by lack of fit. The skills involved in making cars do not transfer to wines, cars and wine are not substitutes and, unless the car is chauffeur driven, they are not complementary.

An algebraic representation of Aaker and Keller’s full effects model is:

A=Q + D + T + C + S + Q*T + Q*C + Q*S (1)

This embodies the hypotheses:

H1: A higher quality perception toward the parent brand is associated with more favorable attitude toward the extension.

H2: The transfer of a brand’s perceived quality is enhanced when the two product classes in some way fit together.

H3: The fit between the two product classes has a direct positive association with the attitude toward the extension.

H4: The relationship between the difficulty of making the product class of the extension, and the attitude toward the extension is positive.

Aaker and Keller’s ordinary least squares (OLS) estimation of the full effects model (1) using a US sample rovided some support for hypotheses 2, 3 and 4. Three years later Sunde and Brodie’s (1993) replication of the study using New Zealand data provided quite different conclusions. The inconsistency resulted in calls to refine and replicate the estimation of the brand extension work internationally (Barwise, 1993; Aaker and Keller, 1993). In response, Nijssen and Hartman (1994) and Bottomley and Doyle (1996) performed further replications in different countries. In doing so, Bottomley and Doyle’s UK study found that most of the incongruities between the earlier studies stemmed from collinearity between variables. Using residual centering (RC) regression (Lance, 1988) they reduced the multicollinearity problem to obtained results that strongly supported hypotheses 1, 2 and 3. The aim of this study is to test the brand extension model that has been designed and tested in a series of developed economies to see if it applies to China, a developing economy.


Other than in estimation, where both OLS and RC regression are used, the method follows that in Aaker and Keller’s original study six parent brands were each associated with three hypothetical extensions. The brands chosen were relevant to the respondents, were of high quality, had strong brand images and had not been widely extended. Because of differences in brand availability and the different development stages of the US and China markets, the brands chosen differed (Table 1).

Seven-point Likert scales were used to gather consumers’ responses. The dependent variable, the overall attitude toward the extension, was an average of two dimensions: the overall perceived quality of the extension quality (1=inferior, 7=superior) and the purchasing probability of buying the extension (1=extremely unlikely, 7=extremely likely). The independent variables were measured as quality (1=inferior, 7=superior) and difficult (1=extremely easy, 7=extremely difficult),

To measure the fit, subjects were asked to assess the extent to which the products were substitutes in certain usage situations or complements in that they would be likely to be used together (1=extremely low, 7=extremely high). To assess transfer, respondents were asked if the people, facilities, and skills used in developing, refining, and making the original product would be helpful if the manufacturer were to make the extension product (1=extremely unhelpful, 7=extremely helpful).

Three questionnaires were developed and each contained two parent brands and six extensions. To enable the questions put in the questionnaires to capture the true meaning of the terms and constructs used in Aaker and Keller’s study, very brief explanations were given for some terms that might be ambiguous under different language contexts.

The respondents were second and third year undergraduate business students. The questionnaires were distributed to six classes by the researcher along with a colleague who taught the class. Respondents were briefed before the questionnaire was distributed and encouraged to ask for clarification where needed. To minimize the possible interactions between different questionnaires and different classes, each questionnaire was assigned to around one third of the students in a class. Preliminary checks found 17 questionnaires were unusable leaving a usable sample of 2814 cases.

As in the original Aaker and Keller’s study full effects (1) and a main effects models (2), the model was first estimated using OLS regression:

A=Q + D + T + C + S (2)

The RC regression is a two-stage process. In stage 1, the interaction term is regressed on the individual variables from which it is composed using OLS to estimate the regression coefficients a, q and t in (3), for example,

Q.T=a + q .Q + t .T (3)

Values of Q and T are then substituted in the estimated equation (3) to provide QTp, predicted values of Q.T. Next, subtracting the predicted value, QTp, from the observed value, Q.T, gives a residual QTr that captures the variance associated with the interaction term that is not explained by the two component variables Q and T.

In stage 2, the residual term QTr is substituted for the original interaction term into the full effects model (1) to give an equation that is estimated using OLS:

A=Q + D + T + C + S + QTr + QCr + Qsr (4)


Table 2 presents the results of main effects models. The four estimates are similar in their support for two of the main effects hypotheses: consumers’ attitudes towards a brand extension depend on the quality of parent brand (H1) and the fit between the two product categories measured in terms of transfer, complement and substitute (H3). However, the Chinese results contrast with the others in having all their coefficients significant. The Chinese main effects estimation is therefore unique in supporting the hypothesis that the difficulty of making the extension is important (H4).







The results for the full effects model (Table 3) are more complex and vary with the regression technique used. Using OLS regression, the results from the four data sets show an almost random distribution of significant coefficients. This variation suggests that multicollinearity is a problem. In contrast both sets of RC results are similar and reveal a wider range of significant main and interaction effects. These are discussed with the overall evidence for each of the hypotheses.

Hypothesis 1: Higher quality perceptions toward the parent brand are associated with more favorable attitudes toward the extension.

OLS regression analysis of the Chinese and the original Aaker and Keller full effects models produced similar insignificant results for the main effect of quality. This suggests hypothesis 1 should be rejected. However, when using the RC to correct for collinearity between the main effects and the interaction variables, the Chinese and RC estimated UK results provided support for Hypothesis 1.

Hypothesis 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.

Using OLS, the Chinese results are similar results to Aaker and Keller’s original findings. Of the three interaction terms, only quality x complement is significant. However, Aaker and Keller’s results find the interaction effect quality x substitute significant while the Chinese results do not.

Using RC regression on the Chinese data gives the same results as OLS, with quality x complement being significant. Bottomley and Doyle’s RC analysis also exposes quality x complement as significant but in this case alongside a significant result for quality x transfer. The Chinese results therefore support Hypothesis 2 but only on the basis of a single interaction term.

Hypothesis 3: The "fit" between the two product classes has a direct positive association with the attitude toward the extension.

Regardless of the regression analysis techniques used, the Chinese study supports Hypothesis 3. Using the OLS approach, the supporting evidence from the Chinese data and Aaker and Keller’s original study is based only on the statistical significance of transfer, a single fit variable.

Hypothesis 4: There is positive relationship between the difficulty of making the extension and the attitude toward the extension.

The evidence in support of this hypothesis is inconsistent. While Aaker and Keller’s OLS analysis supports the hypothesis, the subsequent replications in developed countries found no supporting evidence. Neither of Bottomley and Doyle’s OLS or RC analyses nor Sunde and Brodie’s analysis found significant beta coefficients for difficult. In contrast, the OLS and RC results from the Chinese analysis of the full effects model found difficult to be significant. Given the known multicollinearity in Aaker and Keller’s OLS analysis, this suggests that Hypothesis 4 is supported for China, the developing economy, but rejected for developed economies.


The agreement between the findings of this study, and those reported in earlier research, suggests that there is some international heterogeneity in the way that consumers evaluate brand extensions. The success of Japanese companies, such as Sony, Panasonic and Mitsubishi, in extending their brands in both developing and developed countries provides further supporting evidence for this conclusion.

Opportunities exist for companies to use brand extension strategies in both developed and developing economies. This does not mean that brand extensions will always work in the same way from country to country. In the Chinese analysis two fit variables, transfer and complement, along with difficult, are more important than the perceived quality of the parent brand in determining the success of a brand extension. The positive interactive effects between fit of the brand and its extension and quality of the parent brand reinforce the importance of fit. Thus, any decision to extend a brand should be preceded by evaluating consumers’ perception of the fit between the two product categories involved.

This China study and Aaker and Keller’s original results differ from other replications in finding that the difficulty of making the extension had a role in determining consumers’ attitude toward the brand extension. There may be two possible explanations for this variation. From a methodological perspective, the stretch from original products to extensin products in this study could be greater than in other replications and so more likely to draw respondents’ attention to the dimension. Alternatively, and more likely, the different economic and legal environments in developed and developing countries have shaped consumers’ thinking. An abundance of low quality products and relatively weak protection of intellectual property or consumer rights characterize developing countries. Under these conditions, new offerings are more likely to elicit suspicions or negative associations when evaluating extensions that are too easy to make. For example, in China consumers may be more confident in buying a Sony TV than buying a pair of Sony batteries, because the more complex item is less likely to be a fake. It is also easier to assess the quality of a complex item, such as a camera, than a product, such as a drug, whose appearance is easily copied.

We finally conclude that RC regression is a more robust way of testing Aaker and Keller’s brand extension. The RC estimates from the UK, a developed economy, and China, a developing economy, is similar. This suggests that brand owners can adopt a similar approach to extending their brands in developed and developing economies. However, there is one very significant difference between the UK and China RC results. In the developing economy, the chance of transferring the positive values of a brand to an extension is greatest when consumers see the extension as difficult to make. Trivial extensions are less likely to be accepted by consumers than extensions that embody more skill in their making.


Aaker, D. A. And K. L. Keller (1990), "Consumer evaluations of brand extensions," Journal of Marketing, 54, 27-40.

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

Barwise, P. (1993), "Brand equity: Snark or boojum?" International Journal of Research in Marketing, 10, 93-104.

Bottomley, P. A. and J. R. Doyle (1996), "The formation of attitudes towards brand extensions: Testing and generalizing Aaker and Keller’s model," International Journal of Research in Marketing, 13, 365-377.

Bragg, A. (1986), "Back to the future," Sales and Marketing Management (November), 61-62.

Farquhar, P. H. (1989), "Managing brand equity," Marketing Research (September), 24-33.

Farquhar, P. H., P. M. Herr and R. H. Fazio (1990), "Relational Model for Category Extensions of Brands," In: M. E. Goldberg, G. J. Gorn and R. W. Pollay (eds.), Advances in Consumer Research, Vol. 17, 856-860, UT: Association for Consumer Research.

Laforet, S. and Saunders, J.A., (1994), "Managing Brand Portfolios: How the Leaders Do It", Journal of Advertising Research, 34(5), pp. 64-76

Lance, C. E. (1988), "Residual centering, exploratory and confirmatory moderator analysis, and decomposition of effects in path models containing interactions," Applied Psychological Measurement 12, 163-175.

Low, G. S. and R. A. Fullerton (1994), "Brands, brand management and the brand manager system: A critical-historical evaluation," Journal of Marketing Research (May), 173-190.

Nijssen, E. J. and D. Hartman (1994), "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, 673-683.

Shocker, A. D. and B. Weitz (1988), "A perspetive on brand equity principles and issues: Summary of Marketing Science Institute Conference," Report No. 88-104, Cambridge, MA.

Smith, D. C. and C. W. Park (1992), "The effect of brand extension on market share and advertising efficiency," Journal of Marketing Research 29 (August), 296-313.

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



Fu Guoqun, Peking University, China
John Saunders, Aston Business School, England


AP - Asia Pacific Advances in Consumer Research Volume 5 | 2002

Share Proceeding

Featured papers

See More


Born to Shop? A Genetic Component of Deal Proneness

Robert M Schindler, Rutgers University, USA
Vishal Lala, Pace University
Jeanette Taylor, Florida State University

Read More


Turning the Titanic: Creating Consumer-Centric Cultures and Improved Consumer Experience in Large, Established Health Care Systems

Gregory Carpenter, Northwestern University, USA
Beth Leavenworth DuFault, University at Albany
Ashlee Humphreys, Northwestern University - Medill, USA
Lez Ecima Trujillo Torres, University of Illinois at Chicago, USA

Read More


E8. Perceptions of Out-Group Members: The Effects of Language Abstraction

Afra Koulaei, University of South-Eastern Norway
Daniela Cristian, City University of London, UK

Read More

Engage with Us

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