The Importance of a General Measure of Brand Engagement on Market

Behavior: Development and Validation of a Scale

DAVID SPROTT, SANDOR CZELLAR, and ERIC SPANGENBERG

WEB APPENDIX:
SCALE DEVELOPMENT AND NOMOLOGICAL VALIDITY

Initial Item Generation and Selection

An initial pool of 36 items was developed on the basis of extant branding and self-concept

literature streams (e.g., Belk 1988; Fournier 1998) and on previously validated scales including

self-brand connections (Escalas 2004), object attachment (Ball and Tasaki 1992), and RISC

(Cross et al. 2000). This item pool was given to nine consumer behavior research experts for

review; all experts possessed psychometric expertise (cf. Hardesty and Bearden 2004). Experts

were provided with a definition of BESC and were asked to carefully read each item and rate it

with regard to how well they believed it represented the construct. Experts rated each item on a

nine-point scale, anchored with “Does not tap construct” (1) to “Taps construct” (9). Space was

provided for experts to comment further about particular items as they felt appropriate.

Results of expert review led to eleven items being dropped from the initial pool, including

five items with a mean and median below the scale midpoint and six items identified by written

comment as inappropriate to the measure (e.g., bad wording). Remaining items included three

reverse-scored items, which was raised as a minor concern by one reviewer. Thus, seven

additional reverse-scored items were created (based on rewording of the most highly ranked

items from the expert review), resulting in 32 items used for initial data collection (see Table

A1).
 Table A1

ORIGINAL ITEMS IN INITIAL DATA COLLECTION

1. My favorite brands feel like a part of me.

2. Brands are not that important to who I am.
3. Brands that I most prefer reveal little about me.
4. Overall, the brands I prefer most are those that are part of how I view myself.
5. I have a special bond with the brands that I like.
6. If a friend criticized a brand that I like, I would not feel offended.
7. I consider my favorite brands to be part of myself.
8. The brands that I most prefer clearly indicate something about me.
9. I often feel a personal connection between me and my brands.
10. Sometimes I view the success of my brands as my own personal success.
11. I am offended when someone mocks one of my favorite brands.
12. Part of me is defined by important brands in my life.
13. I feel as if I have a close personal connection with the brands I most like.
14. I can identify with the important brands in my life.
15. There are links between the brands that I prefer and how I view myself.
16. I would feel like I had lost part of my identity if one of my favorite brands were no longer
available to me.
17. I would be a very different person without my favorite brands.
18. When friends criticize brands that I like, I feel offended.
19. I rarely feel a personal connection with any brands.
20. Brands that I prefer have nothing to do with how I view myself.
21. I feel a personal sense of pride when my favorite brands introduce new successful products.
22. My brands are an important indication of who I am.
23. Brands do not indicate much about who I am.
24. Many brands really reflect who I am.
25. I do not feel that my favorite brands are part of me.
26. No brands reflect who I am.
27. Very little of me is defined by important brands in my life.
28. Most brands have nothing to do with how I view myself.
29. When someone criticizes a brand that is important to me, I often feel personally attacked.
30. When a brand is important to me, it feels as if the brand defines who I am.
31. You can learn a lot about me by looking at the important brands in my life.
32. I dislike brands that do not reflect who I am.
Scale Reduction

The 32 initial scale items were then administered to a sample of 430 undergraduate

students in exchange for extra credit. Participants were instructed:

“We are interested in learning more about the role that brands play in the lives of

consumers. Below are a series of statements about brands and your life. Please read each

statement carefully and indicate your level of agreement with the statement by circling the

appropriate number. As you read the statements below, think about the brands that you

prefer in your own personal life.”

Participants then evaluated each of the scale items utilizing seven-point scales anchored by 1

(“strongly disagree”) to 7 (“strongly agree”).

Following established scale development procedures, the sample was randomly split into

two equal groups (Noar 2003). Exploratory factor analysis was performed on one half to reduce

the number of items to a more parsimonious set. Confirmatory factor analyses (CFA) were then

performed on the other half in order to identify final scale items and establish unidimensionality.

Exploratory factor analysis. A principal components analysis on the 32 items highlighted

the presence of a single dominant factor explaining 48.72 % of total variance. Seventeen items

were then eliminated due to factor loadings and item-to-total correlations of less than .70. A

principal components analysis was conducted on the remaining 15 items (items 1, 5, 7, 8, 9, 12,

13, 14 15, 23, 25, 26, 27, 28 and 30), resulting in a single dominant factor explaining 62.32 % of

total variance (eigenvalue = 9.35; all the other factors < 1).

Confirmatory factor analysis. An initial CFA was then conducted on the 15 items with

AMOS 5.0 using the covariance matrix as input, one latent variable, uncorrelated measurement

errors and maximum likelihood estimation. Indices for this first model suggested considerable
room for improvement in terms of model fit (CMIN = 396.15, CMIN/DF = 4.40, GFI = .76,

AGFI = .68, NFI = .84, CFI = .87, RMR = .19, RMSEA = .13, pclose = .00). To improve model

fit resulting in a more parsimonious scale, the number of items was further reduced by following

the procedure implemented by Voss, Spangenberg, and Grohmann (2003). This procedure

improves the psychometric characteristics of a scale while shortening it by iterative deletion of

the poorest fitting items. One first selects and drops the item with the lowest item-to-total

correlation. After that, two CFA models are estimated—one model based on the original set of

items, and the other based on the reduced set. These two models are then compared on two

criteria: (1) a χ2 difference test; and (2) AGFI indexes. If the χ2 test proves significant and if the

AGFI index increases, the reduced scale is used for the next iteration wherein the item with the

lowest item-to-total correlation is eliminated and the χ2 test and AGFI are considered for the next

iteration.

Following Voss et al. (2003), iterations are stopped when one of two possible results is

obtained: The χ2 test shows no significant difference between the models and/or the AGFI

decreases. For our analysis, the AGFI reached a plateau at eight items (AGFI = .92). At this

point, other fit indices also met conventional criteria (CMIN = 42.36, CMIN/DF = 2.12, CFI =

.98, GFI = .96, NFI = .97, RMR = .06) and the hypothesis of close fit was not rejected (RMSEA

= .07, pclose = .11). The remaining eight items were therefore accepted as the final BESC scale

(see Table A2 for factor loadings and item-to-total correlations). Reliability of the final scale was

α = .94. Respondents’ average scores on the final scale were not affected by either their age (r =

.04 n.s.) or gender (Mmales = 3.79, Mfemales = 3.59, F(1, 213) = 1.37, n.s.).
 Table A2

ASSESSING BRAND ENGAGEMENT IN SELF-CONCEPT:

PSYCHOMETRIC CHARACTERISTICS OF FINAL SCALE ITEMS

Factor Item-Total
Items
Loadingsa Correlation
1. I have a special bond with the brands that I like. .787 .726
2. I consider my favorite brands to be a part of myself. .839 .787
3. I often feel a personal connection between my brands
.870 .824
and me.
4. Part of me is defined by important brands in my life. .858 .808
5. I feel as if I have a close personal connection with the
.895 .855
brands I most prefer.
6. I can identify with important brands in my life. .859 .809
7. There are links between the brands that I prefer and
.848 .798
how I view myself.
8. My favorite brands are an important indication of who
.820 .764
I am.
a
Unrotated factor loadings in principal components analysis. Test-Retest Reliability Studies

Following conventional practice, we performed inter-temporal reliability checks with a

two-month delay between two administrations of the BESC measure to two new independent

samples (e.g., Aaker 1997). The first sample was composed of 118 university staff members

participating on a voluntary basis while the second sample included 83 undergraduate students

fulfilling a course requirement. For the staff sample, the BESC scale demonstrated acceptable

test-retest reliability with the two administrations being significantly correlated after a two-month

delay, r = .62 (p < .001). The mean of the individual-item correlations was r = .51 (all ps < .001).

For the student sample, the correlation between the two measures was r = .78, (p < .001) and the

mean of the individual-item correlations was r = .59 (all ps < .001). These test-retest reliability
coefficients are comparable to those obtained by Cross et al. (2000) with the structurally similar

RISC scale, ranging between .63 and .73.

Nomological Validity

Our next step was to position BESC within a broader network of self-relevant constructs

in the literature and to demonstrate that it captures a unique dimension of the self. We also

investigated the relationship between BESC and Material Values, expecting BESC to be related

to but distinct from the concept of materialism. Additionally, we examined the relationship

between BESC and the most routinely assessed form of response bias, social desirable

responding (Paulhus 1998). Further, we sought to ascertain whether the BESC scale was free of

potential gender biases arising from the possibility that some branded goods may reflect strong

masculine or feminine gender portrayals (Lindsay and Widiger 1995).

Samples and scales. Four independent undergraduate samples participated in this study,

whereby each sample completed a survey including the eight-item BESC scale and some of the

other scales under scrutiny. Thus, the first sample (N = 398) completed the following: Self-

Esteem (Rosenberg 1965), Satisfaction with Life (Diener et al. 1985), and Material Values

(Richins 2004). The second sample (N = 279) completed Relational-Interdependent Self-

Construal (Cross et al. 2000). The third sample (N = 107) completed Self-Concept Clarity

(Campbell et al. 1996) as well as the Masculinity and Femininity sub-scales of the Interpersonal

Bem Sex Role Inventory (Brems and Johnson 1990). Finally, the fourth sample (N = 199)

completed Independent Self-Construal (Singelis 1994), Interdependent Self-Construal (Singelis

1994), Collective Self-Esteem (Luhtanen and Crocker 1992), as well as the Self-Deception and

Impression Management sub-scales of the Balanced Inventory of Desirable Responding (Paulhus

1998).
 Correlational results. Correlation coefficients support the expected relationships between

BESC and the aforementioned constructs. BESC was not related to measures of social aspects of

the self, however, there was a significant positive correlation between BESC and RISC (r = .15, p

< .05). Further, BESC was unrelated to general self-esteem, self-concept clarity and satisfaction

with life. There is no evidence that BESC is subject to social desirable responding or gender bias.

BESC was significantly and positively correlated to MATVALUES (r = .42, p < .01).

Discriminant validity between BESC and Material Values. Specifying Material Values

and BESC items to load on two separate factors, two tests of discriminant validity were

performed. In the first, two models were estimated wherein the correlation between factors was

either constrained to correlate perfectly (i.e., a single factor model was forced), or unconstrained

and estimated (Bagozzi, Yi, and Phillips 1991). A χ2 difference test indicated that the second

model provided a significantly better representation of the data than the first (χ2 = 32.63, p < .01).

A second test focused on the unconstrained model; using Fornell and Larcker’s (1981) criteria,

the average variance extracted of .54 for this model was greater than the squared correlation

between the two factors of .22.

 REFERENCES

Aaker, Jennifer L. (1997), "Dimensions of Brand Personality," Journal of Marketing Research,

34 (August), 347–56.

Bagozzi, Richard P., Youjae Yi, and Lynn W. Phillips (1991), “Assessing Construct Validity in

Organizational Research,” Administrative Science Quarterly, 36 (September), 421–58.

Ball, Dwayne A. and Lori Tasaki (1992), “The Role and Measurement of Attachment in

Consumer Behavior,” Journal of Consumer Psychology, 1 (2), 155–72.

Belk, Russell W. (1988), “Possessions and the Extended Self,” Journal of Consumer Research,

15 (September), 139–68.

Brems, Christiane and Mark E. Johnson (1990), “Reexamination of the Bem Sex-Role Inventory:

The Interpersonal BSRI,” Journal of Personality Assessment, 55 (3&4), 484–98.

Campbell, Jennifer, Paul D. Trapnell, Steven J. Heine, Ilana M. Katz, Loraine Lavallee, and

Darrin R. Lehman (1996), “Self-Concept Clarity: Measurement, Personality Correlates,

and Cultural Boundaries,” Journal of Personality and Social Psychology, 70 (1), 141–56.

Cross, Susan E., Pamela L. Bacon, and Michael L. Morris (2000), “The Relational-

Interdependent Self-Construal and Relationships,” Journal of Personality and Social

Psychology, 78 (4), 791–808.

Diener, E., R. A. Emmons, R. J. Larsen, and S. Griffin (1985), “The Satisfaction with Life

Scale,” Journal of Personality Assessment, 49 (1), 71–75.

Escalas, Jennifer Edson (2004), “Narrative Processing: Building Consumer Connections to

Brands,” Journal of Consumer Psychology, 14 (1&2), 168–80.

Fornell, Claes and David F. Larcker (1981), “Evaluating Structural Equation Models with

Unobservable Variables and Measurement Error,” Journal of Marketing Research, 28

(February), 39–50.
Fournier, Susan (1998), “Consumers and Their Brands: Developing Relationship Theory in

Consumer Research,” Journal of Consumer Research, 24 (March), 343–73.

Hardesty, David M. and William O. Bearden (2004), “The Use of Expert Judges in Scale

Development: Implications for Improving Face Validity of Measures of Unobservable

Constructs,” Journal of Business Research, 57 (2), 98–107.

Lindsay, Kathleen A. and Thomas A. Widiger (1995), “Sex and Gender Bias in Self-Report

Personality Disorder Inventories: Item Analyses of the MCMI-II, MMPI, and PDQ-R,”

Journal of Personality Assessment, 65 (1), 1–20.

Luhtanen, Riia and Jennifer Crocker (1992), “A Collective Self-Esteem Scale: Self-Evaluation of

One's Collective Identity,” Personality & Social Psychology Bulletin, 18 (3), 302–18.

Noar, Seth M. (2003), “The Role of Structural Equation Modeling in Scale Development,”

Structural Equation Modeling, 10 (4), 622–47.

Paulhus, Delroy L. (1998), Manual for the Balanced Inventory of Desirable Responding.

Toronto, Canada: Multi-Health Systems.

Richins, Marsha L. (2004), “The Material Values Scale: Measurement Properties and

Development of a Short Form,” Journal of Consumer Research, 31 (June), 209–219.

Rosenberg, M. (1965), Society and the Adolescent Self-Image. Princeton, NJ: Princeton

University Press.

Singelis, Theodore M. (1994), “The Measurement of Independent and Interdependent Self-

Construals,” Personality & Social Psychology Bulletin, 20 (5), 580–91.

Voss, Kevin E., Eric R. Spangenberg, and Bianca Grohmann (2003), “Measuring the Hedonic

and Utilitarian Dimensions of Consumer Attitude,” Journal of Marketing Research, 40

(August), 310–20.