Journal of Business Research 57 (2004) 209 – 224

Developing and validating measures of facets of

customer-based brand equity
Richard G. Netemeyera,*, Balaji Krishnanb, Chris Pulliga, Guangping Wangc,
Mehmet Yagcid, Dwane Deane, Joe Ricksf, Ferdinand Wirthg
a
Department of Marketing, McIntire School of Commerce, University of Virginia, Charlottesville, VA 22904-4173, USA
b
Fogelman College of Business, University of Memphis, Memphis, TN 38152, USA
c
Dept. of Business, Pennsylvania State University-Hazleton, Hazleton, PA 18201, USA
d
College of Business, Loyola University-New Orleans, New Orleans, LA 70118, USA
e
School of Business, East Carolina University, Greenville, NC 27858, USA
f
College of Business, Xavier University-New Orleans, New Orleans, LA 70125, USA
g
Agricultural Economics Dept., University of Florida, Gainesville, FL 32611, USA

Abstract

This article presents four studies that develop measures of ‘‘core/primary’’ facets of customer-based brand equity (CBBE). Drawing from
various CBBE frameworks, the facets chosen are perceived quality (PQ), perceived value for the cost (PVC), uniqueness, and the willingness
to pay a price premium for a brand. Using numerous advocated scale developmental procedures, the measures of these facets showed
evidence of internal consistency and validity over 16 different brands in six product categories. Results also suggest that PQ, PVC, and brand
uniqueness are potential direct antecedents of the willingness to pay a price premium for a brand, and that willingness to pay a price premium
is a potential direct antecedent of brand purchase behavior.
D 2003 Elsevier Science Inc. All rights reserved.

Keywords: Customer-based brand equity; Validation; Brand choice behavior

1. Introduction
In recent years, customer-based brand equity (CBBE) has
garnered considerable attention. Several conceptualizations
of CBBE exist, and these conceptualizations have offered
valuable insight into the processes that consumers evaluate
and choose brands within a given product category. How-
ever, many CBBE facets have not been systematically
measured or validated within a nomological framework.
The purpose of this research is to measure ‘‘core/primary’’
facets of CBBE and examine their relationships with related
brand associations and brand response variables (i.e., a
‘‘nomological net’’). As further validation of the CBBE
measures, we examine their relationships with actual brand
purchase behavior in three product categories.
This article will proceed as follows. We first briefly
review conceptualizations of CBBE and offer our rationale
* Corresponding author. Tel.: +1-434-924-3388; fax: +1-434-924-
7074.
E-mail address: rgn3p@forbes2.comm.virginia.edu (R.G. Netemeyer).
for the facets we have chosen to measure. Two studies are
used to derive and initially validate measures of the CBBE
facets. Two more studies test their ability to predict brand
choice behavior. Finally, a discussion with implications for
future research is presented.
2. Core/primary facets and related brand associations
Though the terms ‘‘CBBE’’ and ‘‘brand equity’’ have
been used interchangeably, the present research will focus
on ‘‘CBBE.’’ Two frameworks that encompass the facets
espoused in most CBBE conceptualizations, are those of
Aaker (1996a) and Keller (1993).1 Aaker views CBBE as a
1
As noted by Keller (1998), the concept of ‘‘brand equity’’ has been
defined and operationalized in disparate ways. These include brand equity
as financial strength, the profit and sales margins due to a brand relative to
comparable brands, and conceptualizations focusing on consumer percep-
tions. Keller offers a concise review of the various conceptualizations of
brand equity, as well as their similarities and differences. In this research,
we limit our focus to the more consumer perception-based frameworks of
‘‘CBBE.’’

0148-2963/$ – see front matter D 2003 Elsevier Science Inc. All rights reserved.
doi:10.1016/S0148-2963(01)00303-4
210 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224
set of assets (liabilities) linked to a brand’s name and
symbol that adds to (or subtracts from) the value provided
by a product/service to the customer. A consumer perceives
brand equity as the ‘‘value added’’ to the product by
associating it with a brand name. Though this ‘‘value
added’’ is a function of several facets, the ‘‘core’’ facets
are the primary predictors of brand purchase intent and
behavior. Core CBBE facets espoused by Aaker include
‘‘perceived quality’’ (PQ), ‘‘perceived value for the cost’’
(PVC), ‘‘uniqueness,’’ and the ‘‘willingness to pay a price
premium’’ of a given brand.
Keller (1993, p. 2) views CBBE as ‘‘the differential
effect of brand knowledge on consumer response to the
marketing of the brand.’’ He also views CBBE as a process
whereby CBBE occurs ‘‘when the consumer is familiar with
the brand and holds some favorable, strong, and unique
brand associations in memory’’ (Keller, 1993, p. 2). The
favorable, strong, and unique associations are termed
‘‘primary’’ associations that include brand beliefs and
attitudes encompassing the perceived benefits of a given
brand (Keller, 1993). These beliefs and attitudes can be
functional and experiential (i.e., PQ and value relative to
other brands) or more symbolic (i.e., its ‘‘uniqueness’’). As
with Aaker’s framework, it is the ‘‘primary’’ brand associ-
ations of PQ, PVC, uniqueness, and the willingness to pay a
price premium that are the strongest predictors of purchase
intent and purchase behavior in Keller’s framework. As
such, the focus of our research will be on those ‘‘core’’ or
‘‘primary’’ CBBE facets common to the Aaker, Keller, and
other frameworks.
Given this article’s focus on measuring and validating the
core/primary CBBE facets, it is useful to delineate related
brand associations and concepts. Fig. 1 shows a nomolog-
ical network where the core/primary CBBE facets are the
most salient predictors of the brand response variables of
purchase intent and brand choice behavior (as indicated by
the solid line). Also note that the willingness to pay a price
premium is posited as key linkage between the other core/
primary CBBE facets of PQ, PVC, and brand uniqueness
and the brand response variables. Five related brand asso-
Fig. 1. Potential relationships with CBBE facets.
ciations are also included in Fig. 1. Brand awareness,
familiarity, popularity, organizational associations, and
brand image consistency are brand associations that are
viewed as related to the core/primary CBBE facets (as
indicted by the curved line) in several brand equity frame-
works (Aaker, 1996a; Blackston, 1995; Farquhar, 1989;
Keller, 1993, 1998).
Consistent with these frameworks, brand awareness is
viewed as the degree to which consumers automatically
think of a brand when a given product category is men-
tioned (i.e., a top-of-the-mind awareness). Brand familiarity
is viewed as the degree to which consumers are familiar
with the brand name, and brand popularity reflects the
degree to which consumers feel the brand is popular with
and used by others. Organizational associations are those
beliefs held by the consumer that the company that markets
the brand is honest, trustworthy, and cares about its cus-
tomers. Brand image consistency is viewed as the degree to
which consumers feel the brand has a rich heritage/history
and a consistent and positive image. Given that most of
these brand associations will not be as predictive of the
brand response variables as core/primary facets (as indicted
by the dotted line in Fig. 1), we term them ‘‘related’’ brand
associations. Still, these brand associations are important as
they should be nomological correlates of the core/primary
facets. As such, their relationships with the core/primary
facets are examined in the studies that follow. We now offer
the conceptual rationale for each of the core/primary facets.
2.1. Perceived quality
PQ is considered a ‘‘core/primary’’ facet across CBBE
frameworks (Aaker, 1996b; Dyson et al., 1996; Farquhar,
1989; Keller, 1993). A definition that has gained some level
of acceptance views PQ as the customer’s judgment of the
overall excellence, esteem, or superiority of a brand (with
respect to its intended purposes) relative to alternative
brand(s). PQ is at a higher level of abstraction than any
specific attribute, and differs from objective quality as PQ is
more akin to an attitudinal assessment of a brand—a global
affective assessment of a brand’s performance relative to
other brands (Aaker, 1996b; Keller, 1993; Zeithaml, 1988).
PQ is considered a core/primary CBBE construct because it
has been associated with the willingness to pay a price
premium, brand purchase intent, and brand choice. PQ may
also be a surrogate for other elements of CBBE (i.e., PVC),
and it is applicable across product classes (Aaker, 1996a;
Keller, 1993, 1998).
Theories based in consumer memory, particularly the
means – end chain model and expectancy value theory,
offer useful frameworks for explaining how PQ judgments
are formed. The means – end chain approach suggests that
a consumer’s cognitive structure holds brand-related
information in memory at different levels of abstraction
(Zeithaml, 1988). At the simpler levels are brand attributes
suggesting ‘‘quality’’ benefits (i.e., functional/practical)
 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224
that lead to an overall ‘‘value’’ or payoff from using the
brand. Information relevant to PQ can be obtained via
promotions where the general quality of the brand is
stressed or where quality is inferred from providing
information about intrinsic or extrinsic brand attributes
(Kirmani and Zeithaml, 1993). Quality judgments may
also be inferred via direct experience with a brand, and
judgments from direct experience are stronger and are
more easily ‘‘accessed’’ from memory (Fazio and Zanna,
1981). Consistent with expectancy value theory, Keller
(1993) also theorizes that brand associations are at differ-
ent levels of abstraction where brand attributes, benefits,
and an overall affective brand attitude represent the levels
hierarchically. The overall affective judgment can be
represented by the ‘‘primary’’ CBBE facet PQ, where
PQ is a multiplicative function of the attributes and
benefits espoused in expectancy value theory (Ajzen and
Fishbein, 1980).
What is implicit in both theoretical approaches to PQ is
the notion of ‘‘value.’’ Both the means – end chain and
expectancy value models include PVC as a part of, or a
consequence of, PQ judgments. Though conceptualized as
distinct, disentangling PQ from PVC judgments, as well as
their effects on brand-related response variables in the
minds of consumers, may be problematic. As the next
section suggests, the manner in which PVC judgments are
formed are theoretically similar to the manner in which PQ
judgments are formed. Further, many suggest that PQ
reflects an overall value judgment (Holbrook and Corfman,
1985; Zeithaml, 1988), or, that the two facets can be
combined to form one overall summary construct of brand
attitude (Aaker, 1996a,b).
2.2. Perceived value for the cost
PVC is a ‘‘core/primary’’ facet and is considered a
cornerstone of most CBBE frameworks (Aaker, 1996a;
Farquhar, 1989; Keller, 1993). PVC is defined as the
customer’s overall assessment of the utility of the brand
based on perceptions of what is received (e.g., quality,
satisfaction) and what is given (e.g., price and nonmonetary
costs) relative to other brands. PVC involves the trade-off of
‘‘what I get’’ (i.e., functional and emotional benefits) for
‘‘what I give’’ (i.e., time, money, and effort) (Kirmani and
Zeithaml, 1993).
Means – end chain and expectancy value theories again
are useful for explaining PVC and its links to other CBBE
constructs. In the means –end chain theory, PVC is at a
higher level of abstraction than any attribute or benefit of a
brand. The attributes and benefits (i.e., ‘‘what I get’’) can be
functional, experiential, or symbolic, and include PQ.
Though PQ has been viewed at a higher level of abstraction,
some feel it is incorporated into PVC judgments (Holbrook
and Corfman, 1985) and should be considered the primary
‘‘what I get’’ component of the PVC equation. Similarly,
expectancy value models suggest that combining attributes
211
and benefits, including PQ, result in an overall PVC
judgment about the brand (Keller, 1993).
As noted by several theorists, consumers are not likely to
draw a distinction between a brand’s PQ and its PVC (Aaker,
1996a; Holbrook and Corfman, 1985). In addition, the recent
Equitrend study found that PQ explained 80% of the vari-
ance in PVC (Aaker, 1996b). Other evidence, though, does
suggest that the two constructs may be separate. PQ may
have a higher prestige aspect associated with the brand,
whereas PVC may be associated more with the functional
utility of the brand. A bottom line is that the same conclusion
may occur whether PQ, PVC, or both are used to predict
brand-related response variables (Aaker, 1996b).
2.3. Uniqueness
Uniqueness is defined as the degree to which customers
feel the brand is different from competing brands—how
distinct it is relative to competitors. If the brand is not
perceived as unique from competitors, it will have a difficult
time in supporting a higher price relative to other brands. As
such, brand uniqueness is considered a ‘‘core/primary’’
CBBE facet (Aaker, 1996b; Agarwal and Rao, 1996).
Judgments of a brand’s uniqueness can be inferred via
differentiating advertising claims or from direct experience
with a brand. Regardless of how it is formed, if a brand is
considered unique, it can command a price premium in the
marketplace (Aaker, 1996b).
Choice theory offers an explanation as to the effective-
ness of uniqueness as a core/primary CBBE facet. When
faced with a choice among brands, features common to
alternative brands may cancel each other out because they
offer little diagnostic information toward preference (Tver-
sky, 1972; Dhar and Sherman, 1996). In contrast, unique
features do offer diagnostic information by differentiating
the brand from other brands. Given that consumers tend to
be cognitive misers, the unique features offer a simplifying
‘‘heuristic’’ for choosing among alternatives. Recent evid-
ence supports this view as unique aspects of a brand affected
both preferences and the willingness to pay a higher price for
a brand (Carpenter et al., 1994; Kalra and Goodstein, 1998).
Further, uniqueness is likely related to PQ and PVC judg-
ments in that consumers may infer that unique aspects have
‘‘value’’ or quality. As such, a strongly held unique asso-
ciation implies that PQ, PVC, and uniqueness are related.
2.4. Willingness to pay a price premium
The willingness to pay a price premium is defined as
the amount a customer is willing to pay for his/her
preferred brand over comparable/lesser brands of the same
package size/quantity. It is one of the strongest indicators
of brand loyalty and may be the most reasonable summary
measure of overall brand equity (Aaker, 1996a). Though
this ‘‘price premium’’ construct is conceptualized as a
‘‘core/primary’’ CBBE facet, it has also been considered
212 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224
a differential brand response construct in other CBBE
frameworks. For example, willingness to pay a price
premium is viewed as a result of managing other CBBE
facets well where PQ, PVC, and uniqueness are important
reasons for the willingness to pay a price premium (Black-
ston, 1995; Keller, 1993). Consistent with Fig. 1 then, the
other core/primary CBBE facets should be predictive of
this price premium facet.
Several consumer theories offer rationale for drawing
associations between price premium and other CBBE facets.
According to memory theory, once information is stored, an
associative network forms that connects or links the asso-
ciations in some way (Alba et al., 1990). Brand associations
formed from direct experience (i.e., PQ and PVC) tend to be
stronger and are more quickly retrieved from memory than
those formed via other means (Fazio and Zanna, 1981).
Such accessible associations (PQ and PVC) guide responses
to a brand and brand choice (Farquhar, 1989). Further, the
favorability and strength of the brand associations are also
affected by brand congruence. Congruence refers to the
‘‘extent to which a brand association shares content and
meaning with another brand association’’ (Keller, 1993, p.
7). As such, favorable PQ, PVC, and uniqueness are
congruent with a greater willingness to pay a price premium
for a brand.
Pricing theories also suggest strong relations between the
core/primary CBBE facets of PQ, PVC, and willingness to
pay a price premium. For example, Monroe (1990) offers a
model that posits the willingness to pay a particular price for
a brand/product as a function of the total perceived value
and quality of the brand/product. Other researchers share
similar views where PQ and PVC affect the willingness to
pay a price premium for a brand and brand purchase
(Kirmani and Zeithaml, 1993; Sethuraman and Cole,
1997). Finally, choice theory again suggests a uniqueness
to price premium relationship. Unique features offer dia-
gnostic information by differentiating a brand from other
brands, thus offering a simplifying ‘‘heuristic’’ for choosing
among alternatives (Tversky, 1972). Further, unique aspects
of a brand can affect the willingness to pay a higher price for
a brand (Kalra and Goodstein, 1998).
2.5. Summary
Our review suggests the following commonalities among
CBBE frameworks. First, CBBE is multifaceted consisting
of the ‘‘primary’’ or ‘‘core’’ associations of PQ, PVC,
uniqueness, and the willingness to pay a price premium
for a brand. Second, nomological correlates of the core/
primary facets include brand awareness, familiarity, popu-
larity, organizational associations, and image consistency.
Third, response variables relevant to the core/primary
CBBE facets are brand purchase intent and brand purchase
behavior. Fourth, valid measures of core/primary CBBE
facets applicable across product categories are needed to
help managers track CBBE and academicians interested in
CBBE research (Aaker, 1996b; Keller, 1993). It is a CBBE
measurement approach focusing on the four core/primary
facets that is the emphasis of this article. Once these
measures are developed, the strength of the relationships
between the CBBE measures and response variables can
provide a basis for specifying the relations among CBBE
facets and response variables (Aaker, 1996a). We now offer
several pretests and studies encompassing numerous advo-
cated psychometric procedures that derive and validate
scales for the core/primary CBBE facets.
3. Measurement pretests
3.1. Focus groups and expert judging pretests
Prior to the main studies, we conducted two focus
groups, expert item judging, and one small pretest study.
The purposes of these procedures were to choose product
categories and brands for the main studies, to determine if
our author/literature generated definitions of the core/prim-
ary CBBE facets concurred with the public’s view, and
generate items for the CBBE measures. Such qualitative
procedures are being increasingly advocated as useful scale
development tools (Bearden and Netemeyer, 1998; Haynes
et al., 1999).
Ten product categories were examined by two consumer
focus groups of various ages and ethnic backgrounds. The
categories were colas, toothpaste, jeans, athletic shoes, fast-
foods, camera film, greeting cards, canned soups, pain
relievers, and breakfast cereals. These categories chosen
because they have a high rate of purchase, and existing
surveys of brand strength (i.e., the Young and Rubicam
Asset Valuator and Total Research’s Equitrend) show that
these products have strong and weak brands (Aaker, 1996b).
Focus group participants were asked numerous scaled and
open-ended questions about their perception of brands with
these categories, and their perceptions of quality, value, and
uniqueness associated with brands.
From the focus groups, literature review, and our own
judgment, 65 items were generated to tap the four core/
primary facets of CBBE. Many of the items were adapted
from studies that had examined aspects of brand equity,
brand loyalty, PVC and PQ, etc. (cf., Aaker, 1996a; Zei-
thaml, 1988). Two marketing professors with backgrounds
in both measurement and brand choice then judged the
items for representativeness, resulting in 37 items retained
with a minimum of 8 and a maximum of 11 items per
CBBE facet.
3.2. Pretest study
To further trim this item pool to a more reasonable
number, a pretest with a sample of 44 MBA students was
conducted. The pretest entailed a long take-home survey
that asked the students to respond to the 37 items across
 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224
four brands in four product categories. The product cat-
egories were cola, toothpaste, jeans, and athletic shoes.
These categories reflect relatively inexpensive and fre-
quently purchased products (i.e., cola and toothpaste), as
well as two (i.e., jeans and athletic shoes) that are more
costly and less frequently purchased. Given that a focus of
this research was to develop measures applicable across
product categories, we felt that these product categories
provide for a stronger initial test of the internal structure and
validity of the measures. The brands were Coca-Cola (i.e.,
all types including Coke Classic, Diet Coke, etc.), Nike
athletic shoes (i.e., all types such as running shoes, cross-
trainers, basketball, etc.), Levi’s jeans, and Crest toothpaste
(i.e., all types including, tartar control, plaque removing,
etc.). Within their product classes, these brands had the
highest brand equity-related ratings from the previous
sources listed (e.g., Landor Associates) and the focus group
results.
The responses were analyzed via principal components
and item analyses. Given the small sample size and the
exploratory nature of this step, items that comprised each
CBBE facet were looked at separately within each of the
four CBBE facets. Consistent with recent psychometric
literature, we used a combination of statistical heuristics
and content validity judgments to delete or retain items
(Haynes et al., 1999). For each CBBE facet, if an item
consistently had low ( < .50) or very high factor loadings
( > .95) (Bagozzi and Yi, 1988), consistently had low or
very high item-to-total correlations, and was consistently
highly correlated with another item within its facet (>.80),
it was considered for deletion. Items with low loadings and
low item-to-total correlations may not be tapping the same
construct (Nunnally and Bernstein, 1994), and items with
very high loadings, very high item-to-total correlations, and
high intercorrelations with other items tend to reflect
empirical redundancy as they are worded too similarly
with other items. Such redundancy can result in an
‘‘attenuation paradox’’ where an item does not contribute
substantively to content validity or reliability (Clark and
Watson, 1995). Overall, these statistical heuristics and
judgment procedures resulted in the retention of 23 items
(a minimum of five and a maximum of seven per CBBE
facet).
4. Study 1
4.1. Sample and procedures
The purpose of this first study was to develop and refine
the CBBE measures and obtain initial estimates of their
psychometric properties. Four samples of adults from a
southeastern city participated in the study. These samples
were collected by undergraduate marketing students as a part
of their ‘‘subjects pool’’ requirement. Students were in-
structed to have one nonstudent adult complete the survey.
213
The person who completed the survey had to provide his/her
address and phone number as it was instructed to the student
who delivered the survey that respondents to several surveys
would be randomly called to ensure that the person whose
name was on the survey actually had responded. Two
hundred surveys were drafted per product category and
given to subjects pool students. Usable responses were
obtained from the four samples ranging in size (n) from
138 to 154 participants. As such, return rates ranged from
69% (138/200) to 77% (154/200). The average age of
respondents was 34.58 years, 49% were college educated,
53% were female, and the median household income was in
the US$40,000 – 49,999 range.
Four product categories were chosen (one for each
sample) where three brands were evaluated within each
category. The product categories and brands were: (1) colas:
Coca-Cola, Pepsi, and RC Cola; (2) toothpaste: Crest,
Colgate, and Close-up; (3) athletic shoes: Nike, Reebok,
and Fila; and (4) jeans: Levi’s, Lee, and Wrangler. These
product categories each have two strong brands with a third
brand having lesser market share, but a highly recognizable
brand name (Aaker, 1996b). We felt that having two strong
brands and one weaker brand within commonly used
product categories could enhance the generalizability of
the CBBE measures.
Surveys corresponding to the four products categories
were drafted. CBBE items pertaining to the brands were
counterbalanced within each survey such that three versions
of each survey were distributed. For example, approxi-
mately an equal number of respondents who were distrib-
uted the ‘‘cola’’ survey responded to Coca-Cola items first,
Pepsi Cola items first, and RC Cola items first. In addition
to the 23 CBBE items, measures relevant to nomological
validity were included. Single-item measures of brand
awareness, familiarity, and popularity were randomly dis-
persed throughout the survey (see Table 1). Consistent with
Fig. 1, most CBBE frameworks include these brand associ-
ations as related to the core/primary CBBE facets (Aaker,
1996a; Keller, 1993; Lassar et al., 1995). Single-item
measures of purchase intent and a past percentage of brand
purchases were also included for nomological validity test-
ing. These measures are displayed in Table 1 and were
culled or adapted form several sources that examined
aspects of brand strength or brand equity (Aaker, 1996b;
Agarwal and Rao, 1996; Cobb-Walgreen et al., 1995). The
past percentage measure included competing brands. For
colas, the brands were Coke, Pepsi, RC, and a ‘‘store brand’’
(SB); for toothpaste the brands were Crest, Colgate, Close-
up, and an SB (a brief description of what an ‘‘SB’’ is was
given to the respondents). Given that SBs are increasingly
capturing market share for many grocery store products,
they were used as an option for colas and toothpaste (Burton
et al., 1998). For jeans, the brands were Levi’s, Lee,
Wrangler, and ‘‘Other Brand (please specify)’’; and for
shoes the brands were Nike, Reebok, Fila, and ‘‘Other
Brand (please specify).’’ Study participants were instructed
214 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224
Table 1
Nomological validity measures for Studies 1 and 2
Study 1
Brand awareness: When I think of (product category), (brand name) is the
brand that first comes to mind.
Brand familiarity: (Brand name) is a brand of (product category) I am very
familiar with.
Brand popularity: (Brand name) is s a very popular brand of (product
category).
Brand purchase intent: The next time I buy (product category), I intend to
buy a (brand name) brand.
Past percentage of brand purchases: Below is a list of four brands of
______. On average, when you have bought _______, what percent of
your purchases were devoted to each brand? The sum of the purchases
across the four brands should add up to 100%.
Study 2
Organizational associations
1. The firm that makes (brand name) is a good corporate citizen.
2. The (brand name) company is honest with its customers.
3. The company that makes (brand name) of (product category) is socially
responsible.
4. The company that markets (brand name) really cares about its
customers.
Image consistency
1. (Brand name) brand of (product category) has a rich history.
2. (Brand name) brand of (product category) has a strong brand image.
3. (Brand name) brand of (product category) has a consistent brand image.
4. Over the years, (brand name) of (product category) has a maintained a
strong image.
5. Over time, (brand name) brand has been very consistent in what it
stands for.
Purchase intent
1 – 3. For my next purchase of (product category), I intend to buy a (brand
name) brand. (With scale endpoints of unlikely – likely, probably will –
probably will not, disagree – agree.)
4. The next time I buy (product category), I intend to buy a (brand name)
brand.
Past percentage of brand purchases
Below is a list of four brands of ______. On average, when you have
bought _______, what percent of your purchases were devoted to each
brand? The sum of the purchases across the four brands should add up to
100%.
to respond to all measures as pertinent just to their own
personal views/uses of all brands.
4.2. Measurement purification and results
To purify the measures, an iterative approach was adop-
ted that included both quantitative and qualitative judgments
(Haynes et al., 1999). For each brand, a factor model
corresponding to an a priori theoretical structure was esti-
mated via LISREL8 covariance structure modeling (Jore-
skog and Sorbom, 1996). As such, 12 four-factor models
were estimated (i.e., one for each brand over the four core/
primary CBBE facets where the factors were freely corre-
lated). These models constituted the first iteration. Items
were deleted that consistently resulted in high within- and/or
across-factor correlated measurement errors and low ( < .50)
or extremely high (>.95) completely standardized within-
factor loadings. However, if an item showed high ‘‘face
validity’’ to its definition, it was retained (Haynes et al.,
1999). This iteration resulted in deletion of four items. In the
next iteration, the remaining 19 items were again specified
to a four-factor model across the 12 brands. Here, we
focused on items that had high cross-loadings to a factor
other than the intended factor and the correlation among the
latent factors themselves. We deleted two items with cross-
loadings of .75 or greater and then estimated another four-
factor model.
At this point, we examined internal consistency and
discriminant validity. All reliability estimates were above
.75 (Nunnally and Bernstein, 1994), and all but one
average variance extracted estimates (AVE) were .50 or
above, a level advocated as indicative of high internal
validity (Fornell and Larcker, 1981). Still, the ‘‘disattenu-
ated’’ correlations among the latent factors of PQ and PVC
(the j correlations) were extremely high. In fact, for 11 of
the 12 brands this correlation ranged from .93 to .99. On a
most stringent test of discriminant validity (i.e., the mean
of the average variance extracted estimates of two latent
factors being greater than the square of the j estimate
between the factors), evidence of discriminant validity was
not found between PQ and PVC. For uniqueness, price
premium, and PQ and PVC though, all but two j estimates
showed evidence of discriminant validity. That is, for a
total of 70 possible j comparisons over the 12 brands, only
the price premium – PVC correlation for Crest and the
price premium – PQ correlation for Levi’s did not show
discriminant validity via the rigorous Fornell and Larcker
(1981) criterion.
At the end of Study 1, 17 CBBE items were retained:
four for PVC, five for PQ, three for uniqueness, and five for
willingness to pay a price premium. However, the extremely
high correlations among PQ and PVC suggest a severe lack
of discrimination between these constructs. These results
lead us to conduct Study 2 to further examine and establish
the dimensionality of PQ and PVC and gain estimates of
reliability and validity.
5. Study 2
5.1. Sample and procedures
A sample of 186 nonstudent adults from a southeastern
city was used in Study 2. As with Study 1, undergraduate
subjects pool participants delivered the survey to the adult
respondents. Two hundred fifty surveys were drafted and
186 were completed for a return rate of 74.4%. The average
age of respondents was 33.73 years, 43% were college
educated, 58% were female, and the median income was in
the US$30,000 – 39,999 range. Three brands were chosen:
(1) Coca-Cola, (2) Reebok athletic shoes, and (3) Levi’s
jeans. Again, these brands were chosen for their high degree
of familiarity and use and their rankings across various brand
rating studies (cf., Aaker, 1996b). The survey contained 18
 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224 215

CBBE items (four for PVC, five for PQ, five for price factor structure where PQ and PVC were modeled as
premium, and four for uniqueness retained from Study 1). separate, but correlated factors. Across brands, no evidence
We added one more uniqueness item.2 of discriminant validity was found for the two-factor model
For nomological validity purposes, two multi-item meas- (Fornell and Larcker, 1981). As such, the decision was made
ures that should correlate with the core/primary CBBE to form an eight-item PQ/PVC factor. This decision is
facets were included: a four-item measure of organizational consistent with the theoretical and empirical findings (par-
associations and a five-item measure of brand image con- ticularly those of Study 1) related to these constructs. PVC
sistency. Consistent with Fig. 1, organizational associations is a customer’s overall assessment of the utility of the brand
of honesty, trustworthiness, and caring for customers may be based on perceptions of what is received (e.g., quality,
considered ‘‘secondary’’ brand associations that are related satisfaction, liking) and what is given (e.g., price, non-
to the core/primary CBBE facets via an inferencing process monetary costs) relative to other brands. This suggests that
(Keller, 1993; Brown and Dacin, 1997). That is, a brand PQ judgments are likely incorporated into PVC judgments
perceived to have high PQ, PVC, and uniqueness may also (or vice versa). Also, as noted by several theorists, in the
be viewed as having strong organizational associations of minds of consumers there is likely little or no distinction
company reputation via the belief inferencing process (Park between a brand’s PVC and PQ (Aaker, 1996b; Holbrook
and Srinivasan, 1994). Brand image consistency is also a and Corfman, 1985). With the exception of the PQ/PVC –
brand association that has been hypothesized as related to price premium correlation for Levi’s, correlations among the
the core/primary facets of PQ, PVC, and willingness to pay other facets ranged from .41 to .79 with evidence of
a price premium (Aaker, 1996b; Farquhar, 1989). discriminant validity.
Two other validity measures were also included. The first Based on these results, a three-factor model that corre-
was a four-item brand purchase intent measure and the sponded to an eight-item PQ/PVC factor (i.e., four PVC
second was the past percentage of brand purchase measure items and four PQ items), a four-item uniqueness factor, and
used in Study 1. (Table 1 displays all nomological validity a four-item price premium factor was estimated. Table 2
measures.) As not to fatigue the respondents, the survey was presents three fit indices viewed as robust to sampling
split into two parts of equal length (about four pages) and characteristics to evaluate model fit: the non-normed fit
responded to over a 10-day interval. Items for all CBBE index (NNFI), the comparative fit index (CFI), and the root
facets across the brands were completely randomized to mean square error of approximation (RMSEA). Values in
minimize the potential of ‘‘yea-saying’’ response set bias. the .90 range and above have been noted as designating
adequate fit for CFI and NNFI (Hu and Bentler, 1995), and
5.2. Measurement purification and results values ranging from .05 to .10 have been deemed acceptable
for the RMSEA (Browne and Cudeck, 1993). As Table 2
The responses were again subjected to confirmatory shows, all NNFI and CFI estimates were above .90 and all
factor analyses using LISREL8. For each brand, a four- RMSEA estimates were at or below .10.
factor model corresponding to a four-item PVC factor, a Evidence of discriminant validity for the three constructs
five-item PQ factor, a four-item uniqueness factor, and a was examined and found. The square of the parameter
five-item willingness to pay a price premium factor was
estimated. For the PVC and PQ factors correlations were
Table 2
again at or above .90 across the three brands (.88 to .95). Study 2 measurement model fit and internal consistency estimates
One item in the PQ facet and one item in the price premium Fit statistics
facet still showed extremely high cross-loadings with other
c2 df CFI NNFI RMSEA
facets or were judged to be somewhat redundant in item
wording. These two items were deleted from all further Coke 257.30 101 .95 .94 .10
Reebok 190.30 101 .97 .96 .07
analyses. Levi’s 186.75 101 .97 .96 .07
At this stage, two more analyses were conducted. First,
we combined the four remaining items for PQ and the four Internal consistency
items for PVC into one factor and compared it to a two- PQ/PVC Uniqueness Price premium
Coke
2
We added another uniqueness for the following reasons. We had five Coefficient a .95 .94 .90
items for uniqueness going into Study 1. Two of the items were reversed Average .78 .79 .70
worded and just did not load well (i.e., well below .50) for any of the brands
of Study 1. Thus, we added one more positively worded item: ‘‘(Brand Reebok
name) is distinct from other brands of (product category)’’ for Study 2. Coefficient a .96 .91 .87
Given that our scales could be used in a variety of studies, we wanted to Average .78 .72 .66
have at least four items per construct as a three-item scale will be ‘‘just
identified’’ as a single-factor model via confirmatory factor analysis— Levi’s
precluding any measures of fit. A four-item scale will be ‘‘over- Coefficient a .96 .91 .91
identified’’—giving estimates of fit. Average .79 .71 .74
216 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224

Table 3
Study 1 validity-related correlations
Colas Intent Past % Awareness Familiarity Popular
Coke Pepsi RC SB
Coke
PQ/PVC .84 .41
.27
.31
.30 .71 .62 .60
Uniqueness .65 .37
.28
.20
.24 .56 .43 .41
Price premium .64 .43
.33
.31
.29 .44 .43 .36

Pepsi
PQ/PVC .66
.36 .48 .07 ns .15 ns .63 .35 .33
Uniqueness .30
.08 ns .12 ns
.01 ns
.04 ns .33 .13 ns .21
Price premium .56
.19 .22 .12 ns
.03 ns .48 .09 ns .10 ns

RC
PQ/PVC .53
.16 .08 ns .23 .12 ns .45 .36 .37
Uniqueness .19
.01 ns
.04 ns .13 ns
.07 ns .21 .22 .38
Price premium .60
.07 ns
.03 ns .25
.03 ns .64 .19 .34

Pastes Intent Past % Awareness Familiarity Popular

Crest Colgate Close-up SB
Crest
PQ/PVC .75 .64
.35
.29
.26 .80 .73 .70
Uniqueness .63 .49
.23
.23
.17 .62 .47 .35
Price premium .66 .65
.29
.38
.22 .66 .50 .49

Colgate
PQ/PVC .77
.48 .42 .04 ns
.03 ns .76 .58 .55
Uniqueness .57
.32 .30 .04 ns
.04 ns .60 .47 .43
Price premium .73
.35 .31
.14 ns
.02 ns .68 .38 .38

Close-up
PQ/PVC .70
.38 .04 ns .67 .01 ns .71 .47 .52
Uniqueness .46
.33 .11 ns .39 .01 ns .48 .45 .42
Price premium .73
.35 .06 ns .49
.13 ns .69 .40 .37

Jeans Intent Past % Awareness Familiarity Popular

Levi’s Lee Wrangler Other
Levi’s
PQ/PVC .81 .48
.15 ns
.12 ns
.27 .73 .68 .64
Uniqueness .54 .29
.04 ns
.12 ns
.16 .55 .50 .35
Price premium .79 .44
.21
.14 ns
.17 .60 .49 .51

Lee
PQ/PVC .78
.12 ns .64
.01 ns
.29 .78 .56 .71
Uniqueness .47
.03 ns .38
.08 ns
.17 .48 .28 .26
Price premium .75
.02 ns .66 .03 ns
.40 .75 .42 .58

Wrangler
PQ/PVC .68
.14 ns
.02 ns .57
.16 .70 .52 .64
Uniqueness .29
.06 ns
.15 ns .25
.01 ns .31 .21 .28
Price premium .61 .05 ns .11 ns .48
.36 .58 .37 .43

Shoes Intent Past % Awareness Familiarity Popular

Nike Reebok Fila Other
Nike
PQ/PVC .76 .48
.21 .06 ns
.22 .69 .55 .47
Uniqueness .56 .33
.21 .04 ns
.19 .59 .54 .40
Price premium .76 .48
.43 .10 ns
.17 .54 .48 .26
(continued on next page)
 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224 217

Table 3 (continued)
Shoes Intent Past % Awareness Familiarity Popular
Nike Reebok Fila Other
Reebok
PQ/PVC .80
.35 .55 .05 ns
.12 ns .81 .44 .63
Uniqueness .63
.22 .35 .07 ns
.11 ns .66 .44 .53
Price premium .71
.22 .42 .16 ns
.18 .70 .29 .45

Fila
PQ/PVC .55
.18 .13 ns .12 ns
.04 ns .57 .34 .55
Uniqueness .48
.11 ns
.01 ns .07 ns .09 ns .40 .42 .41
Price premium .68
.13 ns .06 ns .27 .00 ns .68 .30 .38
‘‘ns’’ denotes a nonsignificant correlation ( P>.05). All other correlations are significant at the .05 level or better.

estimate between pairs of constructs (j2) was less than the
average variance extracted estimates of the two constructs in
all but one instance (Fornell and Larcker, 1981). (Only the j
correlation between PQ/PVC and willingness to pay a price
premium for Levi’s did not meet this rigorous criterion.)
Evidence of internal consistency is provided by coefficient
a and average variance extracted estimates. As Table 2
shows, a ranged from .87 to .96 across CBBE facets and the
average variance extracted estimates were all above .50.
Appendix A shows the final version of the CBBE facets
retained from Studies 1 and 2.
5.3. Evidence of validity: Studies 1 and 2
In Studies 1 and 2, measures related to nomological
validity of the CBBE facets were gathered. Table 3 shows
these results for Study 1. It was expected that PQ/PVC,
uniqueness, and the willingness to pay a price premium
would be correlated with brand purchase intent (‘‘intent’’)
and past percentage of brand purchases (i.e., ‘‘past %
Coke’’). The data confirmed these expectations. Of the 72
correlations of each CBBE facet with ‘‘intent’’ and
‘‘past %,’’ for their respective brands, 68 were significant.
It was also felt that the correlations of the CBBE facets
would be negative or nonsignificant with past percentage of
brand purchases for the other brands examined within a
given product category. For example, the Coke CBBE
facets were expected to be negatively correlated or uncor-
related with ‘‘past %’’ for Pepsi, RC, and SB. This
expectation was largely confirmed for the brands of Study
1. Of the 144 correlations of the CBBE facets with past
percentage of brand purchases for other brands, all were
either negative or nonsignificant. We also correlated the
core/primary CBBE facets with brand awareness, brand
familiarity, and brand popularity. Several frameworks sug-
gest that these constructs are related to the core/primary
CBBE facets (Aaker, 1996a; Keller, 1993, 1998). Of the
144 correlations pertaining to these constructs, 141 were
significant in the expected direction.
Table 4 displays the validity results for Study 2. The
correlations of PQ/PVC, uniqueness, and willingness to pay
a price premium with brand purchase intent (‘‘intent’’) and
past percentage of brand purchase (‘‘past %’’) were again all
positive and significant across brands. PQ/PVC and will-

Table 4
Study 2 validity-related correlations
Coke Intent Past % Image consistency Organizational associations
Coke Pepsi RC SB
PQ/PVC .62 .56
.20
.29
.29 .53 .45
Uniqueness .60 .44
.16
.23
.30 .76 .63
Price premium .68 .43
.13 ns
.17
.34 .59 .57

Levi’s Intent Past % Image consistency Organizational associations

Levi’s Lee Wrangler Other
PQ/PVC .80 .53
.12 ns
.04 ns
.42 .57 .52
Uniqueness .48 .40
.04 ns
.14 ns
.31 .62 .61
Price premium .68 .50
.09 ns
.06 ns
.36 .32 .33

Reebok Intent Past % Image consistency Organizational associations

Reebok Nike Fila Other
PQ/PVC .64 .53
.22 .04 ns
.25 .62 .56
Uniqueness .69 .45
.19 .11 ns
.25 .82 .72
Price premium .73 .49
.22 .01 ns
.22 .62 .58
‘‘ns’’ denotes a nonsignificant correlation. All other correlations are significant at the .05 level or better.
218 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224
ingness to pay a price premium were also more strongly
correlated with ‘‘intent’’ and ‘‘past %’’ than was uniqueness,
and the correlations of the CBBE facets with past percent-
age of purchase for other brands (including ‘‘SB’’) were all
negative or nonsignificant. The correlations of the core/
primary CBBE facets with brand image consistency and
organizational associations were significant as well.
5.4. Summary of Studies 1 and 2
Two studies were used to derive the final form of the
core/primary CBBE measures, test their dimensionality and
internal consistency, and gain estimates of nomological
validity. These studies resulted in retaining an eight-item
‘‘PQ/PVC’’ measure, a four-item ‘‘uniqueness’’ measure,
and a four-item ‘‘willingness to pay a price premium’’ for
the brand measure. Across 12 brands, the measures
showed evidence of dimensionality and internal consist-
ency. These measures showed evidence of validity as they
were correlated with measures of brand purchase intent,
past percentage of brand purchases within a given product
category, and other brand associations (i.e., awareness,
familiarity, popularity, brand image consistency, and organ-
izational associations).
6. Study 3
6.1. Sample and procedures
Though Studies 1 and 2 provide evidence of distinct and
internally consistent core/primary CBBE measures, and that
these measures are related to past behavior and intent, little
was offered about the measures relative to future behavior.
This third study represents a field test that extends the scale
validation process by examining the CBBE measures rela-
tions with actual brand purchase behavior. We contacted a
local Independent Grocers Association supermarket in a
southeastern city. The manager of the supermarket agreed
to allow us to contact shoppers and supplied us with
US$5.00 gift certificates redeemable for any merchandise
at the store as an incentive for shoppers to participate.
During a 4-day period (Wednesday to Saturday), shoppers
were contacted as they entered/exited the store. They were
first shown the incentive to participate and screened to
determine if they (or household members) used coffee and
cola—the target product categories. If they agreed to par-
ticipate and satisfied the screening question, the study
procedures were explained to them. Shoppers were handed
a packet containing a four-page survey, the US$5.00 gift
certificate, three postage-paid return envelopes, and a page
of instructions that fully reminded them of the study
procedures. They were told that the survey was to be
completed at home and mailed back to the researchers
within a 2-week period in one of the postage-paid enve-
lopes. They were also told that we needed their grocery store
receipt(s) for their next purchases of any coffee and cola
from stores that printed the brand names of the products on
the receipts. They were to mail the receipts back to us (with
their name printed on the receipts) in the other postage-paid
envelopes. (Given the limited time in the store, these
procedures were repeated on a separate page at the end of
the survey as a reminder of what the study entailed.) A total
of 250 shoppers were given the study packet.
Two weeks after the last day, we distributed the study
packets, 101 shoppers had returned the survey for an
effective response rate of 41% for the survey. Of these, 63
returned grocery store receipts with the brand names of cola
and/or coffee printed on the receipt. As such, we mailed
reminder letters to those who had not yet sent in their receipts
for cola, coffee, or both products. From this procedure, 14
more behavioral responses were gathered such that of the
original 101 completed surveys, 77 respondents provided
behavioral data (i.e., grocery receipts) for cola purchases and
66 provided behavioral data for coffee purchases. Average
age of the respondents was 50.37 years, 50% were college
educated, 71% were female, and 86% classified themselves
as the primary grocery shopper for their household.
6.2. Measures
The survey contained measures specific to two brands: (1)
Coca-Cola and (2) Community Coffee. Coca-Cola was again
used as a focal brand due to its ratings as a national brand of
high equity across numerous studies. Community Coffee is a
regional brand sold in only a few southeastern states that has
developed a strong customer base. In general, on a per unit
basis, it is priced well above competing national brands such
as Folger’s, Maxwell House, etc. Thus, as a further test of the
validity of the CBBE facets, a strong regional brand was
used as a focal brand in the product category. For both Coca-
Cola and Community Coffee, participants responded to the
final versions of the CBBE measures (see Appendix A). We
used the returned grocery receipts as the behavior measure.
(All grocery receipts were checked for a date to ensure that
the focal product categories were purchased after the ques-
tionnaire was completed and returned.) If the receipt indi-
cated the respondent had purchased a Coke brand (i.e., Coke
Classic, Diet Coke, Caffeine-Free Coke, etc.), and no other
brand (including Dr. Pepper and root beer), this response was
coded a ‘‘1,’’ i.e., did purchase. If the receipt indicated any
soda brand other than a Coke brand, this response was coded
a ‘‘0,’’ i.e., did not purchase. This same coding scheme was
used for coffee brands and Community Coffee.
6.3. Results
Confirmatory factor models of the eight-item PQ/PVC,
four-item willing to pay a price premium, and four-item
uniqueness facets fit the data well for Coke (CFI=.93,
NNFI=.92, RMSEA=.11), but marginally for Community
Coffee (CFI=.89, NNFI=.87, RMSEA=.14). Still, other
 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224 219

Table 5
Study 3 MANOVA/ANOVA results
Coca-Cola
CBBE construct Purchased (n = 59) Did not purchase (n = 18) Multivariate F value Univariate F value h2
Mean S.D. Mean S.D.
4.81 .18
PQ/PVC 47.81 8.27 39.11 12.16 11.65 .16
Uniqueness 23.69 4.74 19.06 6.60 10.50 .17
Price premium 17.40 6.34 11.87 6.62 8.90 .12

Community Coffee
CBBE construct Purchased (n = 51) Did not purchase (n = 15) Multivariate F value Univariate F value h2
Mean S.D. Mean S.D.
6.45 .26
PQ/PVC 47.67 8.27 35.93 12.55 17.41 .24
Uniqueness 23.96 4.82 19.33 5.69 9.64 .17
Price premium 18.62 6.60 10.92 7.41 13.94 .23
All F values are significant at the .05 level or better.

aspects of measurement model acceptability were strong as 7. Study 4

coefficient a estimates ranged from .89 to .95, and AVE
estimates ranged from .67 to .77 across measures for both
brands. Correlations among these four constructs ranged
from .32 to .89 ( P < .01) for Coke and .34 to .81 ( P < .01)
for Community Coffee.
Given these correlations, MANOVA with follow-up
univariate ANOVAs was used to assess mean differences
on the CBBE measures between the ‘‘purchased’’ and
‘‘nonpurchased’’ groups. Such an approach reflects
‘‘known-group’’ validity testing where those who did pur-
chase should show higher mean scores on the CBBE facets
than those who did not purchase (Bearden and Netemeyer,
1998). Each CBBE facet was used as a dependent variable
with the independent variable being purchase behavior
coded from the grocery receipt. As Table 5 shows, both
multivariate and univariate effects were significant and
effect sizes (i.e., h2) ranged from .12 to .26. All mean
scores for the CBBE facets were higher for those who ‘‘did
purchase’’ the brand compared to those who ‘‘did not
purchase’’ the brand. In sum, the CBBE measures were
related to actual purchase behavior.3
Prior literature suggests that once CBBE measures are
developed, models should be used to determine which facets
predict the willingness to pay a price premium and brand
purchase behavior (Aaker, 1996a; Keller, 1993). The
strength of the relationships between CBBE measures and
criterion variables can provide a basis for prioritizing con-
structs such that a model relevant to several brands in a
product category can be specified. In this study, we use our
literature review and findings from the previous three
studies to specify and test such a predictive model.
7.1. Proposed model
Consistent with the nomological framework presented
in Fig. 1, Fig. 2 offers a model that specifies the core/
primary CBBE facets as potential antecedents of brand
purchase behavior. The model posits PQ/PVC and unique-
ness as correlated constructs that are direct antecedents of
the willingness to pay a price premium for a brand.
Furthermore, willingness to pay a price premium directly

3
Though a more prediction-based technique suitable for a dichotomous
dependent variable is logistic regression, we wanted to be able to show that
the all the CBBE measures were related to actual purchase behavior via
MANOVA. The MANOVA approach reflects ‘‘known-group’’ validity
testing in scale development where behaviors theoretically consistent with
attitude/belief-based scaled variables are empirically consistent with
attitude/belief-based variables, i.e., show known-group differences (Beard-
en and Netemeyer, 1998). Still, and consistent with the model estimated in
Study 4, we ran hierarchical logistic regressions for Study 3 data for both
brands. Specifically, we entered willingness to pay a price premium as the
direct predictor of brand purchase on Step 1 and PQ/PVC and uniqueness
on Step 2. Consistent with theory (Aaker, 1996a), willingness to pay a price
premium was a significant predictor on the first step, but when PQ/PVC and
uniqueness were entered on the second step model fit and explained
variance in purchase behavior did not improve appreciably. Fig. 2. Model of core/primary CBBE facets and brand purchase.
220 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224
affects brand purchase behavior. Each of these effects
(i.e., nondirectional and directional paths) is elaborated
upon below.
PQ/PVC and uniqueness are modeled as correlated
exogenous constructs. Strong PQ/PVC judgments are most
likely formed via brand experience (Fazio and Zanna, 1981;
Monroe, 1990; Zeithaml, 1988), whereas uniqueness judg-
ments are more likely affected via promotional activities
(Aaker and Biel, 1993; Cobb-Walgren et al., 1995). This
suggests that the processes of how PQ/PVC and uniqueness
judgments form may not directly affect one another. In
sum, consumers may form judgments about these CBBE
facets simultaneously or through different processes, sug-
gesting no direct paths among PQ/PVC and uniqueness.
Fig. 2 shows a nondirectional relationship among these
constructs (j12).
Recall that the ‘‘willingness to pay a price premium’’
represents the amount a customer is willing to pay for his/
her preferred brand over comparable/lesser brands of the
same package size/quantity. There is an abundance of
theoretical literature suggesting that PQ/PVC directly affects
a consumer’s differential response to a brand in the form of
willingness to pay a price premium. Many feel that PQ/PVC
is the primary antecedent of willingness to pay a price
premium (Blackston, 1995; Kirmani and Zeithaml, 1993).
Further, the results of our previous three studies consistently
show a strong correlation of PQ/PVC with the willingness to
pay a price premium, and this correlation was generally
stronger than the correlation between uniqueness and will-
ingness to pay a price premium. As such, we hypothesize a
direct path from PQ/PVC to willingness to pay a price
premium (g11 in Fig. 2).
Uniqueness is the degree to which customers feel the
brand is different from competing brands. Several theorists
argue the uniqueness to price premium link. Unique features
offer diagnostic information by differentiating the brand
from other brands, offer a simplifying ‘‘heuristic’’ that
positively affects brand preference and the willingness to
pay a higher price for a brand (Carpenter et al., 1994; Kalra
and Goodstein, 1998). Further, it has been suggested that
uniqueness is a key driver of the willingness to pay a price
premium for a brand (Aaker, 1996a; Blackston, 1995; Kalra
and Goodstein, 1998). Our Studies 1 and 2 results also show
a consistent, and moderate to strong, correlation between
uniqueness and the willingness to pay a price premium.
Given this evidence, we hypothesize a direct path from
uniqueness to the willingness to pay a price premium (g12
in Fig. 2).
The most important brand-related response variable is
brand purchase. Willingness to pay a price premium is
theorized as a primary predictor of brand purchase (Aaker,
1996b; Blackston, 1995; Dyson et al., 1996), and all three of
our previous studies show a strong relation between will-
ingness to pay a price premium and brand purchase. Thus,
we hypothesize a direct path from willingness to pay a price
premium to brand purchase (b21 in Fig. 2).
7.2. Sample, procedures, and measures
To test the hypothesized model for brands within a
product category, a multiple-time period study was con-
ducted. Two hundred two undergraduate business students
at a major state university in the southeast participated in
the study as part of an ‘‘extra credit’’ subjects pool. In a
central location, all participants responded to the main
survey containing the final form of CBBE measures toward
three fast-foods restaurants—McDonald’s, Burger King,
and Wendy’s. Fast-foods and the McDonald’s, Burger King,
and Wendy’s chains were chosen as the focal product and
brands for the following reasons. First, across several brand
equity-related ratings (i.e., the Young and Rubicam Asset
Valuator, Total Research’s Equitrend, and the Landor Asso-
ciates Power Rating), these three brands were consistently
ordered as first (McDonald’s), second (Burger King), and
third (Wendy’s) in terms of overall brand strength within
their product category (cf., Aaker, 1996b). Second, fast
food purchases are relatively inexpensive and frequent for
students, and on/near the campus where the study was
conducted, there are McDonald’s, Burger King, and
Wendy’s outlets.
At the end of each week for the next 5 weeks following
the initial survey administration, participants provided self-
report behavioral data on the frequency of which they
patronized the three fast-food restaurants. If respondents
indicated that they had patronized a given restaurant,
responses were coded a ‘‘1’’; if they had not patronized,
responses were coded as ‘‘0.’’ As such, the behavioral
measure became a continuous scale that could range
from ‘‘0’’ to ‘‘5.’’ The reasoning for gauging behavior
over a 5-week time period was threefold. First, it has been
shown that belief/attitudinal variables (such as the core/
primary CBBE facets) may not predict single incidences of
behavior all that well as there are a myriad of situational
factors that could prohibit a strong relation between such
predictors and a single-shot behavior (e.g., Ajzen and
Fishbein, 1980; Epstein, 1980). Conversely, it has been
shown that belief/attitudinal predictors do predict aggrega-
tions of thematically related behaviors across situations
and time intervals.
By aggregating behavioral measures over time, greater
reliability (and validity) of the behavioral measure is
obtained giving a more accurate estimate of the relation
between predictor and criterion (Ajzen and Fishbein, 1980;
Epstein, 1980). Second, and consistent with the brand
loyalty and satisfaction literatures, aggregating brand behav-
iors over time better taps ‘‘affective’’ loyalty to the brand as
it indicates a cumulative satisfaction over purchase occa-
sions (Oliver, 1999). Given the many idiosyncratic influen-
ces affecting single purchase occasions (i.e., situational
factors, variety seeking), aggregating brand purchases offers
a better representation of consumer loyalty to a brand.
Third, a continuous behavioral variable was required to
operationalize the model via a structural equations meth-
 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224 221

odology (SEM). Mixing categorical and continuous data in
such an analytical framework may violate the statistical
assumptions of SEM and/or require extremely large sample
sizes (Schumacker and Lomax, 1996). Of the 202 that
completed the main survey, 167 provided behavioral data
for all 5 weeks (i.e., 83%). These 167 responses were used
in all analyses.
7.3. Modeling procedures and results
Prior to estimating the model in Fig. 2, a three-factor
measurement model was estimated that corresponded to an
eight-item PQ/PVC factor, a four-item uniqueness factor,
and a four-item price premium factor using covariance
structure modeling via LISREL8. The top half of Table 6
presents the fit indices and internal consistency estimates for
this model. Both the fit and internal consistency estimates
were acceptable. Correlations among these three factors
ranged from .39 to .63 for McDonald’s, .37 to .79 for
Wendy’s, and .46 to .73 for Burger King. Evidence of
discriminant validity was examined via the same test used
in Study 2, and across all brands and possible pairs of
factors, discriminant validity among the CBBE facets was
found.
To test the model in Fig. 2, we again used LISREL8. The
behavioral construct in the model was operationalized via
the single continuous variable derived above. However, we
did incorporate the potential effect of random measurement
error by setting the measurement loading of the construct to
the square root of its reliability estimate and its measure-
ment error term to the product of one minus the internal
consistency estimate and the variance of the construct.
(KR-20 estimates of internal consistency for the behavioral
measure ranged from .57 to .70 across the three brands.) The

Table 6
Study 4
Measurement model fit and internal consistency estimates
Fit statistics
c2 df CFI NNFI RMSEA
McDonald’s 285.59 101 .91 .89 .10
Wendy’s 248.69 101 .93 .92 .09
Burger King 223.62 101 .94 .93 .09

Internal consistency
PQ/PVC Uniqueness Price premium
McDonald’s
Coefficient a .92 .90 .85
Average .67 .70 .61

Wendy’s
Coefficient a .93 .94 .91
Average .73 .79 .73

Burger King
Coefficient a .91 .88 .86
Average .64 .68 .64

Structural model fit, path, and explained variance estimates

Fit statistics
c2 df CFI NNFI RMSEA
McDonald’s 305.49 116 .91 .89 .10
Wendy’s 260.30 116 .94 .93 .09
Burger King 251.16 116 .93 .91 .09

Path and explained variance estimates

McDonald’s Wendy’s Burger King
PQ/PVC ! Price premium: g11 .36 (4.29) .71 (9.05) .20 (2.57)
Uniqueness ! Price premium: g12 .51 (5.91) .17 (2.66) .69 (7.89)
Price premium ! Behavior: b21 .36 (3.52) .51 (5.41) .44 (4.05)
PQ/PVC ! Uniqueness: j12 .58 (5.21) .53 (5.24) .63 (5.62)

R2—Price premium .59 .67 .68

R2—Behavior .13 .26 .19
Path estimates are completely standardized and numbers in parentheses represent t values. All paths are significant at the .05 level or better.
222 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224
bottom half of Table 6 presents the results. In general, the fit
indices suggest an acceptable level of fit across the three
brands. More importantly, though, are the path and
explained variance estimates (i.e., R2) for the price pre-
mium, and behavioral constructs. Of the nine directional
paths over the three brands, all were significant in the
prediction direction. The exogenous variables of PQ/PVC
and uniqueness explained between 59% and 68% of the
variance in willingness to pay a price premium, and will-
ingness to pay a price premium explained between 13% and
26% of the variance in purchase behavior. In sum, evidence
of the model’s ability to predict the brand-related response
variable of behavior over time was found.
8. Discussion
8.1. Summary
The primary objective of this research was to measure
core/primary facets of CBBE. Several consumer-based the-
ories (i.e., memory theory, choice theory, and pricing
theory) were used to delineate the CBBE facets and specify
their interrelationships, and, numerous advocated psycho-
metric procedures were used to derive their measures. These
procedures included focus groups of consumers to help
derive definitions and generate items, item judging by
marketing professors with expertise in measurement and
branding, exploratory and confirmatory factor analyses,
dimensionality and assessment of internal consistency, and
nomological, known-group, and predictive validity testing.
Importantly, the validity testing including relating our
CBBE facets to actual brand purchases. Across pretests
and studies that covered over 1000 respondents, and exam-
ining 16 different brands over six product categories,
consistent evidence for the internal consistency and validity
of CBBE measures was found.
8.2. Implications and future research
Though several facets of CBBE have been identified,
those most predictive of key brand-related response vari-
ables derived and measured in this research are the core/
primary facets of: (1) a PQ/PVC facet, (2) a uniqueness of
the brand facet, and (3) a willingness to pay a price
premium facet. Throughout the early stages of our
research other non-core/primary CBBE facets were meas-
ured, but were not as predictive of brand-related response
variables as the facets retained. As such, for researchers
interested in predicting and explaining brand purchase
intent and brand purchase behavior, the CBBE measures
derived should prove very useful. The measures are brief,
easy to administer, and applicable across product categor-
ies and brands.
Note specifically that the PQ/PVC measure has eight
items: four reflecting PQ and four reflecting PVC. We
purposefully retained more items for this facet due to the
nature of PQ and PVC. Though some theoretical literature
suggests that consumers are not likely to draw a distinction
when making PQ and PVC judgments (e.g., Aaker, 1996b;
Holbrook and Corfman, 1985; Kirmani and Zeithaml,
1993), and our empirical results bear this out, there are
occasions where the two may be separate constructs. For
some goods, it is possible that consumers could judge the
‘‘quality’’ of a brand to be high without judging its ‘‘value
for the cost’’ to be high (e.g., BMW is perceived to be the
best car in its class (quality), but its price might not justify a
high ‘‘value’’ judgment in the minds of many consumers).
In these instances, our PQ and PVC measures could be used
as separate four-item scales, depending on the specific
research objective.
Also, we are not suggesting that non-core/primary (i.e.,
related) brand associations are of little utility. Future
research should examine these brand associations as ante-
cedents or consequences of the core/primary CBBE facets.
For example, experimental studies manipulating brand
image consistency, familiarity, and popularity to assess their
effects on PQ/PVC and willingness to pay a price premium
are of interest. Studies could also be conducted that examine
how advertising and promotion affect consumer judgments
of not only core CBBE facets, but organizational associa-
tions as well. Given the recent interest on the effects of
corporate citizenship on brand judgments (Brown and
Dacin, 1997), research examining how ads promote the
corporate citizenship, and in turn their effect on brand-
related response variables, is needed.
Also of interest are the relations among the CBBE facets
and Aaker’s (1997) brand personality dimensions. It is
unclear if brand personality affects some CBBE facets (or
vice versa), or if personality and CBBE judgments are made
independently. It is possible that secondary belief inferenc-
ing affects these correlations. Thus, studies examining how
CBBE and brand personality relationships form are of
interest (Fournier, 1998).4
Finally, given the brevity of our measures, managers and
practitioners may find them useful as predictors or corre-
4
As a further check of validity, Study 2 included the brand personality
scales of Aaker (1997). These scales measure the human-like characteristics
that consumers may attach to a brand and have been posited to be related to
facets of brand equity and brand choice. Five 5-item semantic differential
scales assessing the brand personality traits of ‘‘competence,’’ ‘‘sincerity,’’
‘‘ruggedness,’’ ‘‘sophistication,’’ and ‘‘excitement’’ for all three brands
were included on the survey. All five of the brand personality dimensions
were significantly correlated with our CBBE measures for Levi’s and
Reebok. (Across the CBBE facets and personality dimensions these
correlations ranged from .22 to .70 for Levi’s and Reebok.) For Coke, only
one significant correlation was found (i.e., uniqueness and sincerity had a
correlation of .19). However, the core/primary CBBE facets were more
highly correlated with the brand response variables of intent and past
purchase behavior than any of Aaker’s (1997) brand personality
dimensions.
 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224 223

lates of other measures of brand strength, consumer sat-
isfaction, or financial strength. For example, ongoing rank-
ings of brand strength (e.g., Total Research’s Equitrend and
Young and Rubicam’s Brand Asset Valuator) and company
strength (e.g., Fortune’s annual best and worst list) may be
potential outcome variables of the core/primary CBBE
scales. Similarly, overall customer satisfaction with a variety
of brands within a given product category may be function
of the CBBE facets we have measured (Fornell et al., 1996).
Practitioners may also examine the CBBE scales vis-à-vis
financial measures to determine if consumer views of a
brand’s equity are related to financial indices of brand and
firm strength (e.g., Simon and Sullivan, 1993). A question
such as ‘‘do high ratings on core/primary CBBE facets
predict market share and financial returns?’’ could be
addressed by using our measures. Lastly, connecting
responses to our CBBE measures with scanner data is of
interest. Given the rise in ‘‘club-card’’ grocery store mem-
berships, managers can assess the degree to which a brand’s
PQ/PVC and uniqueness affect brand response variables. In
particular, how do PQ/PVC and uniqueness affect the will-
ingness to pay a price premium for a brand (and brand sales)
in the face of a price increase or the introduction of a similar
competitive offering? The use of our scales with scanner
panel data could be helpful in providing answers to such
questions.
8.3. Limitations
Like most studies, ours are not without limitations. First,
our samples were all convenience-based. Whether our
results will generalize to randomly drawn samples repres-
entative of the population is not known. Second, though we
looked at 16 different brands, all brands were either fre-
quently purchased ‘‘grocery’’ goods (i.e., fast-foods, coffee)
or infrequently purchased nondurables (i.e., athletic shoes,
jeans). Whether our CBBE measures hold as well for
‘‘durable’’ goods has yet to be tested. Third, while the
model we proposed and tested in Study 4 suggests a
temporal ordering of constructs, only the behavioral meas-
ure was collected at a time other than the cross-sectional
data collection of the CBBE facets. As such, all we can
conclude is that PQ/PVC, uniqueness, and the willingness to
pay a price premium were related at one point in time.
Finally, we used single-item measures of awareness, famili-
arity, and popularity as nomological correlates of our CBBE
scales in Study 1. Though the results of correlating these
measures were very consistent across the 12 brands and four
product categories of Study 1, single-item measures are of
an unknown level of reliability.
Acknowledgements
The authors would like to thank Scot Burton for helpful
comments on this article.
Appendix A. Items in the CBBE scales
PQ/PVC
1. Compared to other brands of (product), (brand name) is
of very high quality.
2. (Brand name) is the best brand in its product class.
3. (Brand name) consistently performs better than all
other brands of (product).
4. I can always count on (brand name) brand of (product)
for consistent high quality.
5. What I get from (brand name) brand of (product) is
worth the cost.
6. All things considered (price, time, and effort), (brand)
brand of (product) is a good buy.
7. Compared to other brands of (product), (brand name)
is a good value for the money.
8. When I use a (brand name) brand of (product), I feel I
am getting my money’s worth.
Note: Items 1, 2, 3, and 4 represent PQ and Items 5, 6, 7,
and 8 represent PVC.
Uniqueness
1. (Brand name) is ‘‘distinct’’ from other brands of
(product).
2. (Brand name) really ‘‘stands out’’ from other brands
of (product).
3. (Brand name) is very different from other brands
of (product).
4. (Brand name) is ‘‘unique’’ from other brands of
(product).
Willingness to pay a price premium
1. The price of (brand name) would have to go up quite a
bit before I would switch to another brand of (product).
2. I am willing to pay a higher price for (brand name)
brand of (product) than for other brands of (product).
3. I am willing to pay ___% more for (brand name) brand
over other brands of (product): 0% 5% 10% 15% 20%
25% 30%, or more.
4. I am willing to pay a lot more for (brand name) than
other brands of (product category).
Note: With the exception of Item 3 of Willingness to Pay
a Price Premium, all items are measured on seven-
point strongly disagree to strongly agree scales.
References
Aaker DA. Measuring brand equity across products and markets. Calif
Manage Rev 1996a;38(3):102 – 20.
Aaker DA. Building strong brands. New York: Free Press, 1996b.
Aaker JL. Dimensions of brand personality. J Mark Res 1997;34:347 – 56
(August).
Aaker DA, Biel A. Brand equity and advertising: an overview. In: Aaker
DA, Biel A, editors. Brand equity and advertising. Hillsdale (NJ): Law-
rence Erlbaum Associates, 1993. pp. 1 – 8.
Agarwal MK, Rao V. An empirical comparison of consumer-based meas-
ures of brand equity. Mark Lett 1996;7(3):237 – 47.
224 R.G. Netemeyer et al. / Journal of Business Research 57 (2004) 209–224
Ajzen I, Fishbein M. Understanding attitudes and predicting social behavior.
Englewood Cliffs (NJ): Prentice-Hall, 1980.
Alba JW, Hutchinson JW, Lynch JG. Memory and decision making. In:
Kassarjian HH, Robertson TS, editors. Handbook of consumer theory
and research. Englewood Cliffs (NJ): Prentice-Hall, 1990. pp. 1 – 49.
Bagozzi RP, Yi Y. On the evaluation of structural equation models. J Acad
Mark Sci 1988;74 – 94 (Spring).
Bearden WO, Netemeyer RG. Chapter one: introduction. In: Bearden WO,
Netemeyer RG, editors. Handbook of marketing scales. 2nd ed. Thou-
sand Oaks (CA): Sage Publications, 1998. pp. 1 – 14.
Blackston M. The qualitative dimension of brand equity. J Adver Res
1995;2 – 7 (July/August).
Brown T, Dacin P. The company and the product: corporate associations
and consumer product responses. J Mark 1997;61:68 – 84 (January).
Browne MW, Cudeck R. Alternative ways of assessing model fit. In: Bollen
KA, Long JS, editors. Testing structural equation models. Newbury
Park (CA): Sage Publications, 1993. pp. 136 – 62.
Burton S, Lichtenstein DR, Netemeyer RG, Garretson JA. A scale for
measuring attitude toward private label products and an examination
of its psychological and behavioral correlates. J Acad Mark Sci 1998;
26:293 – 306 (Fall).
Carpenter GS, Glazer R, Nakamoto K. Meaningful brands from meaning-
less differentiation: the dependence of irrelevant attributes. J Mark Res
1994;31:339 – 50 (August).
Clark LA, Watson D. Constructing validity: basic issues in scale develop-
ment. Psychol Assess 1995;7(3):309 – 19.
Cobb-Walgren KJ, Ruble CA, Donthu N. Brand equity, brand preference,
and purchase intent. J Adver 1995;24:25 – 41 (Fall).
Dhar R, Sherman SJ. The effect of common and unique features in con-
sumer choice. J Consum Res 1996;23:193 – 203 (December).
Dyson P, Farr A, Hollis NS. Measuring and using brand equity. J Adver Res
1996;36:9 – 21 (November/December).
Epstein S. The stability of behavior: II. Implications for psychological
research. Am Psychol 1980;35:790 – 806.
Farquhar PH. Managing brand equity. Mark Res 1989;1:24 – 33 (September).
Fazio RH, Zanna MP. Direct experiences and attitude – behavior consis-
tency. In: Berkowitz L, editor. Advances in experimental social psychol-
ogy, vol. 14. New York: Academic Press, 1981. pp. 161 – 202.
Fornell C, Larcker DF. Evaluating structural equation models with unob-
servable variables and measurement error. J Mark Res 1981;18:39 – 50
(February).
Fornell C, Johnson MD, Anderson EW, Cha J, Bryant BE. The American
consumer satisfaction index: nature, purpose, and findings. J Mark
1996;60:7 – 18 (October).
Fournier S. Consumers and their brands: developing relationship theory in
consumer research. J Consum Res 1998;24:343 – 73 (March).
Haynes SN, Nelson K, Blaine D. Psychometric issues in assessment re-
search. In: Kendall PC, Butcher JN, Holmbeck G, editors. Handbook of
research methods in clinical psychology. New York: Wiley, 1999. pp.
125 – 54.
Holbrook MB, Corfman KP. Quality and value in the consumption experi-
ence: Phaedrus rides again. In: Jacoby J, Olsen J, editors. Perceived
quality. Lexington (MA): Lexington Books, 1985. pp. 31 – 57.
Hu L-T, Bentler PM. Evaluating model fit. In: Hoyle RH, editor. Structural
equation modeling: concepts, issues, and applications. Thousand Oaks
(CA): Sage Publications, 1995. pp. 76 – 99.
Joreskog K, Sorbom D. LISREL8: user’s reference guide. Chicago (IL):
Scientific Software International, 1996.
Kalra A, Goodstein RC. The impact of advertising positioning strategies on
consumer price sensitivity. J Mark Res 1998;35:210 – 24 (May).
Keller KL. Conceptualizing, measuring, and managing customer-based
brand equity. J Mark 1993;57:1 – 22 (January).
Keller KL. Strategic brand management. Englewood Cliffs (NJ): Prentice-
Hall, 1998.
Kirmani A, Zeithaml VA. Advertising, perceived quality, and brand image.
In: Aaker DA, Biel A, editors. Brand equity and advertising. Hillsdale
(NJ): Lawrence Erlbaum Associates, 1993. pp. 143 – 61.
Lassar W, Mittal B, Sharma A. Measuring customer-based brand equity. J
Consum Mark 1995;12(4):11 – 9.
Monroe KB. Pricing: making profitable decisions. 2nd ed. New York:
McGraw-Hill, 1990.
Nunnally JC, Bernstein IH. Psychometric theory. 3rd ed. New York:
McGraw-Hill, 1994.
Oliver R. Whence consumer loyalty? J Mark 1999;63:33 – 44 (special issue).
Park CS, Srinivasan V. A survey-based method for measuring and under-
standing brand equity and its extendibility. J Mark Res 1994;31:271 – 88
(May).
Schumacker RE, Lomax RG. A beginner’s guide to structural equation
modeling. Mahwah (NJ): Lawrence Erlbaum Associates, 1996.
Sethuraman R, Cole C. Why do consumers pay more for national brands
than for store brands? MSI Working Paper No. 97-126. Cambridge
(MA): Marketing Science Institute, 1997.
Simon CJ, Sullivan MW. The measurement and determinants of brand
equity: a financial approach. Mark Sci 1993;12:28 – 52 (Winter).
Tversky A. Elimination by aspects: a theory of choice. Psychol Rev
1972;79:281 – 99 (July).
Zeithaml VA. Consumer perceptions of price, quality, and value: a means –
end model and synthesis of evidence. J Mark 1988;52:2 – 22 (July).