EFFECTS OF BRAND EXTENSION FIT

EFFECTS OF BRAND EXTENSION FIT

The Moderating Role of Social Approval, and the Mediating Role of Brand Attitude: Evidence from the Dutch Charitable Market

NICOLE VAN BEEK

University of Groningen Faculty of Economics and Business

Master Thesis MSc Marketing 22 August 2016

First supervisor: Dr. J.C. Hoekstra Second supervisor: Dr. J. Berger

Palfrenier 12 1188DP Amstelveen 06-41112502 n.a.van.beek@student.rug.nl

Student number: S2793083

ABSTRACT

Brand extensions (the use of established brand names to launch new products) form an important part of strategic business growth, but are not without risks. Low-fit brand extensions (low similarity to the parent brand) can cause the formation of negative associations in consumers’ minds. This study examines if social approval can reduce the negative effects of a low-fit brand extension. It is hypothesised that social approval negatively moderates the effect of brand extension fit on brand attitude and behavioural intention and that brand attitude will mediate the relationship between fit and behavioural intention. 847 donors of a Dutch charitable organisation participated in an experimental study using a 2 (high/low brand extension fit) x 2 (social approval majority/minority) research design.

Against expectations, social approval does not moderate the relationship, no results have been found for the direct effect of fit on brand attitude and behavioural intention, and no results have been found for a mediating effect of brand attitude. Implications of the results and limitations are discussed.

Keywords: charity, brand extension fit, brand extension, brand attitude, social approval

TABLE OF CONTENT

1 INTRODUCTION 4

2 LITERATURE REVIEW AND HYPOTHESES 6

2.1 Conceptual model 6

2.2 Theory and hypotheses 7

3 METHODOLOGY 11

3.1 Selection of stimuli pretest 11

3.2 Data collection 12

3.3 Sample 14

3.4 Construct measurement 14

3.5 Manipulation check 17

3.6 Method of analysis 18

4 RESULTS 19

4.1 Descriptive statistics 19

4.2 Hypotheses testing 20

5 CONCLUSIONS AND RECOMMENDATIONS 26

REFERENCES 29

APPENDIX I 34

Scales 34

APPENDIX II 36

Pretest (in Dutch) 36

APPENDIX III 39

Survey (in Dutch) 39

1 INTRODUCTION

Introducing new products can be a smart strategy for a business to grow. Brand extensions, -

‘the use of established brand names to launch new products’ - (Völckner & Sattler 2006) and line extensions - the use of the current brand name to enter a new market segment in its product class - (Aaker & Keller 1990) are strategies frequently used by companies to expand their portfolio, to lower the risk of new product failure, and to foster growth (Ahluwalia 2008;

Lau & Phau 2007). The added value of a brand extension comes from the belief that the reputation of the parent brand will reduce the expenses of the introduction of the new product (Kim & Yoon 2013).

While brand extensions might seem like a good way to expand the business portfolio, it is not without risks. Brand extensions often fail in achieving their goals (Lau & Phau 2007;

Völckner & Sattler 2006). Research suggests that only two out of ten brand extensions succeed (Hem, de Chernatony, & Iversen 2003). For example, in 1999, woman’s magazine Cosmopolitan extended its brand after noticing a link between readership and dairy products.

Cosmopolitan launched its own brand of yoghurt, but it was quickly withdrawn from the market due to disappointing results (Schutte 2014). Disappointing sales results are not the only concern for companies. When brand extensions fail, they can seriously damage the brand image, for example by creating negative associations in the mind of the consumer, and linking these to the parent brand (Aaker 1990). An example of creating negative associations is the

‘Harley Davidson cake decorating kit’ that was launched by motorcycle brand Harley Davidson. The associations of ‘freedom’ and ‘macho’ that consumers held of Harley Davidson were now mixed with ‘disneyfied’ associations (Schutte 2014). It is therefore not surprising that a serious amount of research has focused on the factors concerning the creation of successful brand extensions (Ahluwalia 2008).

Völckner and Sattler (2006) reviewed 45 brand extension studies over the past 15 years and summarized 10 significantly proven factors that influence brand extensions success.

They found that the most important factor for a brand extension to succeed is the perceived fit, “the extension’s perceived similarity to the parent brand”, (Ahluwalia 2008). A large number of studies found the same results (e.g. Aaker & Keller 1990; Boush & Loken 1991;

Park, Milberg, & Lawson 1991; Sunde & Brodie 1993).

A brand extension that fits the parent brand benefits from carry-over effects of

perceptions and preferences of the parent brand (Batra, Lenk, & Wedel 2010). A fitting

intention (Swaminathan, Fox, & Reddy 2001). This is the case because a fitting brand extension induces shared associations with the parent brand and will therefore foster the flow of associations between them. Besides a direct positive effect of a fitted extension on behavioural intention, Roberts, and Bacon (1997) show that attitude partly mediates the effect of environmental concern on purchase intentions of green products. It could therefore be that perceptions and preferences of the parent brand influences attitude towards the brand and that this change in attitude thereby affects behavioural intentions.

When a brand extension is created, mixed associations may derive in consumers’

minds (Aaker 1990). These mixed associations might create confusion in consumers’ minds (Martinez & Pina 2003), which may make people uncertain about their attitude towards the extension. The degree of uncertainty will especially be high in case of an ill-fitted brand extension because the number of shared associations with the parent brand will be lower.

When people are uncertain, they rely on others’ opinions to form their own (Cialdini &

Goldstein 2004). Therefore social approval could be an important aspect to help overcome this uncertainty and should have a larger effect in higher uncertainty situations. For example, positive online customer reviews have a greater effect for weak brands than for strong brands (Nga, Carson, & Moore 2013) and these positive reviews are especially important in the introduction phase of new products (Cui, Lui, & Guo 2012). This signals that social approval may have a stronger effect in higher uncertainty situations, in this case weak brands and new products. In other words, social approval might play a moderating role in the relationship between an extension and its success. This paper therefore studies the effect of brand extension fit on brand attitude and behavioural intention, the mediating role of brand attitude, and the moderating effect of social approval.

A Dutch charitable organisation cooperated in this study. This charitable organisation is in the animal welfare business and has been helping monkeys in need for over 40 years.

Over the years, the charitable organisation has also rescued other animals every now and then.

The charitable organisation always profiled itself as a monkey rescue and sheltering charity and did not actively brand itself with other animals. Due to changing circumstances, this organisation is now extending its business from only helping monkeys in need to also helping other animals in need and is considering changing their communication accordingly. Because of the charitable context of this study, the dependent variable of behavioural intention is measured in terms of donation intention.

This study contributes to the brand extension literature by empirically examining the

Brand extension fit

Social approval

Brand attitude

Donation intention outcomes. An ill-fitted brand extension could have negative effects on consumers’ brand attitude and behavioural intention. Therefore this study investigates the moderating role of social approval and examines if social approval is a factor that could reduce the negative effects of an ill-fitted brand extension. The moderating role of social approval in this relationship has, to our knowledge, never been investigated before and therefore adds new insights in the brand extension literature. Next to this, this study is conducted under real market conditions, which is important for the external validity of the results.

The structure of this paper is as follows: chapter 1 introduces the aim of the study, chapter 2 defines the conceptual model, the constructs and hypotheses, chapter 3 describes the methodology including the research design, sample and data analysis, chapter 4 provides the results, and chapter 5 describes the conclusions and recommendations.

2 LITERATURE REVIEW AND HYPOTHESES

This chapter will describe the literature review and hypotheses. In section 2.1 the conceptual model is explained, and in section 2.2 the theory and hypotheses are described.

2.1 Conceptual model

Figure 1 shows the conceptual model where fit is hypothesised to have a positive influence on brand attitude and donation intention. A low-fit extension will have less shared associations

FIGURE 1:

Conceptual model

with the parent brand than a high-fit extension. With less shared associations, uncertainty

about the extension and parent brand may be larger and social approval (the knowledge that

others endorse the brand extension) could have a stronger effect. Therefore social approval is

assumed to increase the level of brand attitude and donation intention in the low-fit condition,

but not in the high-fit condition, and thereby negatively moderates the relationships. Next to the direct effect of fit on donation intentions, the mediating role of attitude is investigated.

Ample literature shows that brand attitude is a predictor of purchase intentions (e.g. Hee Yeon

& Jae-Eun 2011; Dall’Olmo Riley, Pina, & Bravo 2015; Webb, Green & Brashear 2000; Kim

& Yoon 2013). This implies that, in order to observe behavioural intentions, a change in attitude should have taken place. Therefore it is hypothesised that brand attitude mediates the relationship between brand extension fit with donation intention.

2.2 Theory and hypotheses

Defining fit. Fit is defined as the “extension’s perceived similarity to the parent brand”

(Ahluwalia 2008). There is no unequivocality in literature on how perceived similarity is determined (Mao & Krishnan 2006) or how it should be measured (Bhat & Reddy 2001).

Aaker and Keller (1990) state that consumers evaluate extension fit based on three components, namely: complementarity (how it complements the existing product), substitutability (if it is substituting the existing products), and transferability (if the product has the same manufacturing resources). Park, Milberg, and Lawson (1991) found that extension fit is perceived on both concrete functional features and abstract symbolic features such as prestige. In a more recent study, Huifang and Krishnan (2006) argue that fit is based on brand prototype fit (consistency between the image of the parent brand and the extended brand) and product exemplar fit (consistency between the extension product and existing products). Bhat and Reddy (2001) specified similar components in earlier research: product category fit (how similar the product is to the brands’ current products) and brand image fit (how similar the image is of both the brand and the extension). Evangeline and Ragel (2016) recently conducted a study in which they investigated in which way consumers evaluate brand extensions and find that indeed product category fit and brand level fit are determinants of consumers’ judgements of perceived fit. Most evidence is related to the conception of fit based on product category fit and brand image fit, therefore perceived fit will be determined by both product category fit and brand image fit in one construct.

Brand attitude. Brand attitude is defined as “consumers’ overall evaluations of a brand”

(Keller 1993). Low-fit extensions can result in evoking negative associations towards the

parent brand (Lau & Phau 2007) because brand extensions can create new associations and

thereby confuse the current associations consumers have of the brand (Martínez Salinas &

Pina Pérez 2009). Confusion can weaken what the brand means to consumers (John, Loken,

& Joiner 1998) and could thereby have a negative effect on brand attitude.

High-fit extensions on the other hand lead to more positive evaluations of the brand extension (Boush & Loken 1991), to more positive product evaluations (Park, Milberg, &

Lawson 1991), and to more positive attitudes towards the brand (Völckner & Sattler 2006).

This is because high-fit extensions cause the transfer of associations and emotions from the parent brand to the extension (Pina, Iversen, & Martinez 2010).

When charities support a cause (in this case a type of animal that was not supported before) that fits with the current causes (the current animals the charity supports), donors are assumed to hold more positive attitude towards the brand than when charities support a cause that does not fit with their current activities. Therefore, the following is hypothesised.

H1: Brand extension fit positively influences brand attitude.

Donation intention. Behavioural intention is measured in terms of donation intention, where donation intention is specified as the ‘intention to donate money to a specific charitable organisation’. Donations are very important for charitable organisations because they form more than half of their income (Charity Aid Foundation 2014). Intentions have shown to predict behaviour (Randall & Wolff 1994; Ajzen et al., 2011) and will therefore be incorporated in the model.

Research concerning brand extensions and purchase intentions found that a high-fit extension resulted in a higher willingness to pay a premium, and a low-fit extension resulted in less willingness to pay a premium (DeiVecchio & Smith 2005). This was also found in the study of Pracejus and Olsen (2004), consumers were willing to pay significantly more for a adventure park ticket when $5 was donated to a charitable organisation that fitted with this specific adventure park than when it did not. The strongest support for the belief that brand extension fit has a positive influence on donation intention comes from Swaminathan, Fox, and Reddy (2001), who investigated successful and unsuccessful brand extensions and found that fit positively influenced purchase behaviour.

Donors have usually no control over what is done with their donation once it is

received by the charitable organisation. It can be imagined that if a charitable organisation

uses money to support a cause that fits with the parent brand a donor is more willing to donate

than when the charitable organisation uses money to support a cause that does not fit with the

than before. This could give the donor the feeling that his donation matters more. Based on this reasoning, the following hypothesis is developed.

H2: Brand extension fit positively influences donation intentions.

Attitude as mediator. There are a lot of studies that report a positive relationship between consumers’ attitude and behavioural intentions (e.g. Hee Yeon & Jae-Eun 2011; Dall’Olmo Riley, Pina, & Bravo 2015; Webb, Green & Brashear 2000; Kim & Yoon 2013). For example, Webb, Green and Brashear (2000) found that consumers’ attitude towards charitable organisations has a positive influence on their donation intention. Kim and Yoon (2013) found that fit has an influence on attitude, which in turn influences purchase intentions. These studies imply that in order for a behavioural intention to occur, a change in attitude precedes.

Therefore the following is hypothesised:

H3a: Brand attitude positively influences donation intentions

Evidence for a mediated relationship comes from Roberts and Bacon (1997) who show that environmental concern positively influences purchase intentions of green energy and that this relationship is partly mediated by attitude. This study can be linked to the charitable organisation context because both environmental concern and donating to charities require consumer behaviour directed at a societal problem.

The transfer of emotions and associations from the parent brand to the extension ensures a fitting extension to have a positive influence on attitude towards the brand (Pina, Iversen, & Martinez 2010; Völckner & Sattler 2006). When fit increases donors’ attitude towards the brand, consistency in responses is ought to explain the effect of attitude on behaviour (Ajzen & Fishbein 1977). When donors’ attitude has changed it is logical for donors to be consistent in further responses and their positive attitude therefore positively influences donation intentions. Hence, the following is hypothesised.

H3b: The relationship brand extension fit with donation intention is partly mediated by brand attitude.

Social approval. It is long known that people rely on other people’s opinions to form their

conforming and agreeing with a visible majority (Jahoda 1959). In this respect, social approval is a persuasive message highlighting the fact that other people are satisfactorily using or endorsing the brand (Myers & Sar 2013). When consumers endorse a product, other consumers are likely to evaluate that product more positive accordingly (Myers & Sar 2013).

This could be the case because people use the information of others as reference level for their own subsequent evaluations (Mcferran et al. 2013).

Many brands use the effect of social approval cues in their marketing activities to increase brand attitude and purchase intentions, for example McDonalds that communicates

“billions and billions served” and a TV provider that states: “join the 50 million Americans that already enjoy [TV provider]” (Myers & Sar 2013). A good illustration comes from Erb et al. (1998) who studied the influence of social approval on participants’ attitudes towards a construction project. Participants received a message that a majority of people (85%) versus a minority of people (15%) approved of a construction project. In case of the majority manipulation, participants held significantly more favourable attitudes towards the construction project compared to the minority manipulation. Bischoff and Egbert (2013) give another example in which they describe a television show in which donations to charities are collected and publicly announced. Giving this social information increases donation amounts up to 12% (Shang & Croson 2009). Research therefore shows that social approval influences both attitude and behaviour.

When consumers are confronted with mixed associations a brand extension induces, confusion may arise (Martínez Salinas & Pina Pérez 2009; Kim & Yoon 2013). Confusion could lead consumers to be uncertain about their attitude towards the brand. People are motivated to reduce uncertainty and in times of uncertainty, people look for leadership and guidance and rely on the opinion of others to form their own (Cialdini, & Goldstein, 2004;

Hogg, Kruglanski, & van den Bos, 2013; Smith, Hogg, Martin, & Terry 2007). In case of brand extensions, donors could therefore look for cues to help reduce uncertainty created by mixed associations. Textual social approval cues lead to more positive brand attitudes and stronger purchase intentions (Myers & Sar 2013). The approval of others could facilitate the reduction in uncertainty and lead to more positive brand attitudes and donation intentions.

Because a low-fit extension is assumed to create a larger level of uncertainty, social

approval will have a greater effect on donors when they are confronted with a low-fit

extension than when donors are confronted with a high-fit extension. Hence, the following

has been hypothesised:

H4a: Social approval will negatively moderate the effect of brand extension fit on brand attitude.

H4b: Social approval will negatively moderate the effect of brand extension fit on donation intention.

3 METHODOLOGY

A Dutch charitable organisation specialised in the rescue of monkeys provided the possibility to do an experiment among their donors. This organisation is extending their brand to other types of animals. Therefore an experimental field study was conducted using a 2 (high fit/low fit extension) x 2 (social approval majority/minority) between-subjects design. Participants (donors) were confronted with either a high-fit or a low-fit animal to simulate a high and low fit brand extension. Due to restrictions imposed by the charitable organisation, the experimental study was executed among lapsed donors (donors that had not been donating for at least 18 months).

In section 3.1 the selection of stimuli is described that was done in the first study in order to determine which animal was perceived as low-fit and high-fit brand extension. In section 3.2 the data collection is discussed followed by the sample in section 3.3, in section 3.4 the construct measurement is discussed and in section 3.5 the manipulation check results are discussed. In the final section, 3.6, the method of analysis is discussed.

3.1 Selection of stimuli pretest

213 donors of which 28% male and 72% female with an average age of 58 (SD=13.5) participated in the pretest (see appendix II). Six animals were included in the survey:

raccoons, squirrels, pumas, silver foxes, porcupines, and ferrets. These animals were chosen because the charitable organisation wanted to use animals they at least had sheltered once before. All animals were shown in random order to participants to avoid respondents using the first animal as reference level. A picture of the animal was shown before each set of questions. Although it is known that visually presenting an animal can narrow its meaning (Phillips 1996), it is chosen to do so in order to limit the degree of confusion. To minimize the influence of the use of pictures, all animals are portrayed from the side, showing the full body, not directly looking into the camera and in a passive mode.

The Dall’Olmo Riley, Hand, & Guido (2014) perceived fit scale (see Table 2) was

used to measure the degree of fit for each animal using five statements on a Likert scale

(Likert, 1932) ranging from 1 “totally disagree” to 7 “totally agree”. A factor analysis and reliability analysis were conducted to reveal if the scale items measure one construct and are internally consistent. Both KMO analysis (Kaiser 1981) (.907) and Barlett’s test of sphericity (Bartlett 1937) (p < .001) showed that factor analysis was appropriate to use. The analysis shows that the five items to measure fit belong to one construct (eigenvalue 4.35, variance 87.04% on one dimension) and were highly internally consistent (α .963) (Cronbach 1951).

TABLE 1:

Fit scores pretest (N=213)

Animal Raccoon Squirrel Puma Silver fox Porcupine Ferret

Mean

SD

The mean fit scores can be found in Table 1. As indicated in the Table, the raccoon is judged as the best fitting animal (M=5.48, SD=1.16) and the puma as least fitting (M=3.8, SD=1.74). A t-test shows that the means of both animals are significantly different from each other (p < .001). These animals will therefore form the high (raccoon) and low (puma) fit manipulation conditions for the experiment.

3.2 Data collection

An online survey was developed to collect data. The data collection was held in the period April – June 2016 with Qualtrics (a research software tool). The online survey was distributed in Dutch because the participants (donors) are Dutch. The population existed of lapsed donors (donors that had not donated for 18 months or more). Next to this, donors had to be told up front that the news message (manipulation), that they were going to see, was fictive. To limit the possible effect of this message to not be taken serious, donors were told that they should read the message as if it was real.

Respondents were shown a news article with the text: “For over more than 40 years,

stichting AAP supports monkeys in need. Soon stichting AAP will shelter raccoons/puma’s

next to monkeys. In this way, stichting AAP will commit itself to long term sheltering of

raccoons/puma’s. A poll has shown that 85%/15% of the donors agree with sheltering

raccoons/pumas.” To manipulate fit, the raccoon (high fit) and puma (low fit) were used. In

order to manipulate social approval, following (Erb, et al., 1998) the donors were told that

either a majority of respondents (85%) or a minority of respondents (15%) agreed with the decision of the charitable organisation to extend their brand towards raccoons/pumas. The stimulus material can be seen in Figure 2, and the full survey can be found in appendix III.

FIGURE 2:

Four manipulations used in the survey

A 2 x 2 between subjects design was used. Respondents were randomly assigned to one of the four treatments. First, two opening questions were asked. The first one to get an idea of the knowledge level of participants (donors were asked which animals the charitable organisation supports), the second was the manipulation check of fit (see section 3.5). The manipulation check was asked before the start of the experiment in order to get unbiased answers with respect to the judgment of fit of the animal to the parent brand. Hereafter, the manipulation was shown (see Figure 2).

After this message, respondents answered statements about donation intention and brand attitude. On the next page, a manipulation check of social approval was incorporated to examine if participants had seen the information in the news article. Respondents were asked:

“to what extent did donors agree with sheltering raccoons/pumas?” A seven point Likert scale

(Likert 1932) using various percentages measured their response. This question was asked after respondents answered the attitude and donation intention items to avoid revealing the purpose of this experiment. If respondents would have seen this question before answering the attitude and donation intention items, they could have known that the social information given was of interest in this study and therefore give desired answers and bias the results. Finally, the survey ended with control variable questions gender, age, income, education level, and donation behaviour last year.

3.3 Sample

865 donors participated in the survey of which 847 responses were useful. 18 participants that indicated the same number on every Likert scale statement (so-called straigtliners, as checked by a reversed asked statement) were removed from the analysis. The dataset yields 22.9%

male and 76.4% female with an average age of 57.83 years old (SD=13.01). 9.2% had a yearly income of less than €15.000, 18.8% had an income between €15.000 and €29.999, 22.4% had an income between €30.000 and €44.999, 9.9% had an income between €45.000 and €59.999, 7.7% had an income of €60.000 or more, and 32% of the respondents did not want to answer this question. Further, .2% of the respondents had no education, 1.8% had a primary school education, 17.7% was lower educated (MAVO, VMBO), 31.6% was middle educated (HAVO, VWO, MBO) and 48.6 was higher educated (HBO, WO). 91.4% of the participants were aware that the charitable organisation also sheltered other animals besides monkeys. About 20% of the donors were aware that the charitable organisation sheltered puma’s and about 76.4% were aware of sheltering raccoons and 95.6% of the participants had donated one year or less ago.

3.4 Construct measurement

The scale of Dall’Olmo Riley et al. (2014) was used to measure fit, the scales of Mathwick

and Rigdon (2004), Webb, Green, and Brashear (2000) were used to measure brand attitude

and the scales of Putrevu and Lord (1994), Posavac et al. (2004) and Grohmann (2009) were

used to measure donation intention (see Table 2). These scales are validated in earlier

research and have a high internal consistency (α= >.8). The scale items were comprised of

different scales used by various researchers because the existing scales were not specific

enough for this charitable organisation context. A professional translator translated the scale

items to Dutch. Some items were adapted to suit the current research (appendix A shows an overview of the adapted items).

Because the scales used in this study are comprised of different existing scales, a factor analysis was conducted to assess the structure of the scales. In all cases both KMO analysis (Kaiser 1981) (> .7) and Barlett’s test of sphericity (Bartlett 1937) (p < .05) showed that factor analysis was appropriate to use (Malhotra 2010 p.638). Next to this, all factor loadings, corresponding to the factor they belong to, were well above the cut-off value (.4) (Ford, Maccallum & Tait 1986). After the factor analysis, a reliability analysis was conducted to check for the internal consistency of the scales (Cronbach 1951). The results of these analyses can be found in Table 2.

First, the fit scale was examined using a factor analysis to reveal its structure. The fit scale showed that one component was extracted (eigenvalue 4.01, variance explained 80.2%

on one dimension). The reliability analysis showed that the scale was internally consistent (α

= .937). The five items measuring fit are combined into one factor by taking its average to conduct further analysis.

Next the attitude scale was examined by a factor analysis to reveal the structure of the scale. It was found that four items loaded on one factor and five items loaded on a second factor. A reverse coded item loaded on both factors and was therefore excluded and the factor analysis was performed again. The results show that two factors were loaded. Five items on factor one and three on factor two. A separate factor analysis for the five items of factor one yielded an eigenvalue of 3.46, a variance of 69.21% on one dimension with a Cronbach alpha of .878 and the factor analysis for the three items loaded on factor two yielded an eigenvalue of 1.78, a variance of 59.38% on one dimension with a Cronbach alpha of .603. This shows that the five items have a considerably higher variance explained and are significantly greater internally consistent. Next to this, Osborne and Costello (2009) state that a factor analysis with five or more strongly loading items (> .50) is desirable and indicates a strong factor.

When examining the external validity, the three item attitude measure show more donation

intention related questions (attitude towards donating and best charity to donate to) and

therefore do not give the impression of a correct attitude measure. The last five items signal

the measurement of brand attitude better (the image of the charity is discussed, the

performance of the organisation is assessed and the usefulness of the organisation within

society is examined). Therefore these five items (averaged) will be used as attitude measure to

conduct further analysis with.

TABLE 2:

Scales, varimax rotated factor loadings, and Cronbach’s alpha reliability results

Factor loadings

α

Perceived fit 1

Dall’Olmo Riley, Hand, & Guido,

• This animal is similar to the animals stichting AAP is

currently sheltering .829 .94

2014 • Stichting AAP is capable of sheltering this animal .849

• Sheltering this animal fits with the brand image I have of

stichting AAP .919

• Sheltering this animal is logical for stichting AAP .942

• Sheltering this animal is appropriate for stichting AAP .933

Brand attitude 1 2

Webb, Green, & • I say positive things about stichting AAP to other people. .216 .770 .60 Brashear (2000) • I have a favorable attitude towards donating to stichting

AAP over the next few years. .264 .770

Mathwick, &

Rigdon, (2004)

• To me, stichting AAP is clearly the best charity

organisation to donate to. .106 .678

• Much of my donation to stichting AAP is wasted.

(Reverse item) - -

• I believe stichting AAP is a good charity. .744 .227 .88

• The money given to stichting AAP is used for good

causes. .790 .047

• My image of stichting AAP is positive. .842 .321

• Stichting AAP has been quite successful in helping

animals in need. .825 .268

• Stichting AAP performs a useful function for society. .800 .239

Donation intention 1 .79

Putrevu, & • It is very likely that I will donate to stichting AAP. .750 Lord (1994);

Grohmann, (2009)

• I will donate to stichting AAP the next time I donate to a

charity. .803

• I will definitely donate to stichting AAP. .788 Posavac et al.

(2004)

• How likely are you to donate to stichting AAP in the near

future? .747

• If you were to donate today, how likely is it that you

would donate to stichting AAP? .641

* For the brand attitude scale, Cronbach alpha was conducted for the bold items of factor one and two separate.

Donation intention was measured with five items that loaded on one factor with an eigenvalue of 2.8 and a variance of 55.95%. The five items were internally consistent (α = .79). Again, these items were averaged to conduct further analysis.

3.5 Manipulation check

The perceived fit scale was used as manipulation check to confirm that the two animals from the pretest were indeed judged as a high and a low fit animals. To determine the effectiveness of the manipulation of fit, a t-test was conducted. Both the means of the high and low fit manipulation were significantly different from each other with an average fit score of 5.37 for the raccoon and an average fit score of 4.17 for the puma (t(777.5) = -12.6, p < .001).

To check for the social approval manipulation respondents were asked to indicate the extent to which donors in the news article agreed with the brand extension. This question was asked on a 7-point Likert scale (Likert, 1932) using the following options in increasing order:

no one agreed, 15% agreed, 35% agreed, 50% agreed, 70% agreed, 85% agreed, and everyone agreed. A t-test showed a significant difference (t(682.7) = -20.28, p < .001) between the means of the 15% manipulation (M=3.32) and 85% manipulation (M=5.52). The results of both manipulations can be found in Table 3.

TABLE 3

Manipulation check results using a t-test

N mean sd df t Sig.

Fit High 419 5.37 1.59 777.5 -12.6 < .001

Low 428 4.17 1.15

Social approval Majority 446 5.52 1.22 682.7 -20.28 < .001

Minority 401 3.32 1.84

To assess whether the four groups differed from each other on background

characteristics, an ANOVA test was computed for age (in years), education level (from lowest

level 1 to highest level 5), and last donated to the charitable organisation (in years) a chi-

square test was computed for income (1 lowest, 5 highest, 6 do not want to tell), and gender

(woman, man). The ANOVA test showed that the four groups are statistically significantly

different from each other with respect to age (p = .039) and education (p = .001) but are not

statistically significantly different based on last donated (p = .902). The chi-square test showed that the four groups do not differ significantly on income (p = .94) and gender (p = .085). A regression analysis of all control variables on brand attitude and donation intention revealed that at a 5% confidence level only education was found to influence donation intention (b = -.150, t(836) = -.138, p < .001). Therefore education will be controlled for in subsequent analysis where donation intention is the dependent variable. Age was different between the four manipulation groups but had no significant influence on attitude or donation intention and will therefore not be incorporated as control variable in the subsequent models.

3.6 Method of analysis

Regression analyses were performed for all hypotheses. Because education was the only control variable that had influence on donation intention, this variable will be included in analyses in which donation intention functions as dependent variable. Fit is in each model dummy coded with 1 high fit and 0 low fit, social approval is dummy coded with 1 majority and 0 minority, education is categorical measured on a scale from 1 low educated to 5 high educated, and brand attitude and donation intention are both interval scale measured.

Model 1 was used to test hypothesis 1. Models 2 and 4 were used to examine hypothesis 2. To examine hypothesis 3a and 3b a mediation analysis was conducted using Baron and Kenny’s (1986) approach to mediation testing. To test for mediation the following regression equations must be estimated:

1: Fit must influence the mediator variable (attitude, assessed by model 1) 2: Fit should have a direct effect on donation intention (assessed by model 2)

3: The mediator variable (attitude) must affect the dependent variable (donation intention), which is measured by model 3.

4: The full model has to be examined including fit, attitude and donation intention (assessed by model 4).

If partial mediation occurs, the regression coefficient (β

) of the direct effect of fit on donation intention (model 2) should be smaller than regression coefficient (β

) in model 4.

When the relationship of fit with donation intention has become non significant in the fourth equation, the relationship of fit with donation intention is fully mediated by brand attitude.

Models 5 and 6 will examine the moderating effect of social approval. A significant

interaction coefficient indicates moderation (Dawson 2014).

Model 1 Model 2 Model 3

BA = β

+ β

FIT + ε

DI = β

+ β

FIT + β

ED + ε

DI = β

+ β

ED + β

BA + ε

Where:

BA = Brand attitude FIT = Fit (1 high/ 0 low) ε

= Error term

Where:

DI = Donation intention FIT = Fit (1 high/ 0 low) ED = Education

ε

= Error term

Where:

DI = Donation intention ED = Education

BA = Brand attitude ε

= Error term

Model 4 Model 5 Model 6

DI = β

+ β

ED + β

FIT + β

BA + ε

BA = β

+ β

FIT + β

SA + β

FIT*SA + ε

DI = β

+ β

ED + β

BA + β

FIT + β

SA + β

FIT*SA + ε

Where:

DI = Donation intention ED = Education

FIT = Fit (1 high/ 0 low) BA = Brand attitude ε

= Error term

Where:

BA = Brand attitude FIT = Fit (1 high/ 0 low) SA = Social approval (1 majority/ 0 minority) FIT*SA = Interaction term ε

= Error term

Where:

DI = Donation intention ED = Education

BA = Brand attitude FIT = Fit (1 high/ 0 low) SA = Social approval (1 majority/ 0 minority) FIT*SA = Interaction term ε

= Error term

4 RESULTS 4.1 Descriptive statistics

Table 4 shows the mean values for attitude and donation intention for the different levels of

fit. As can be seen from the table, there is hardly any difference between attitude in the low fit

condition (M=6.47) and the high fit condition (M=6.52). This lack of difference is also

present in the low fit condition (M=5.99) and the high fit condition (M=5.95) of donation

intentions. The t-test confirms the lack of statistically significant differences for both attitude

(t(845) = 1.01, p = .315 and donation intention t(845) = -.706, p = .480).

TABLE 4

Descriptive statistics and t-test for attitude and donation intention grouped by fit.

Fit N mean sd df t Sig.

Attitude High 419 6.52 .564 845 1 .315

Low 428 6.47 .715

Donation intention High 419 5.95 .811 845 -.71 .480

Low 428 5.99 .815

4.2 Hypotheses testing

The effects on brand attitude. Table 5 shows the results of models 1 and 5. Both models were not significant (p >.05). Model 1 is examined to test hypothesis 1 (brand extension fit positively influences brand attitude). The results show that fit has no influence on respondents attitude towards the brand (β = .035, t(846) = 1.01, p = .315). H1 is therefore not supported.

Model 5 in Table 5 was used to test the moderating effect of social approval in the relationship of fit with brand attitude. The model did not significantly improve by adding the interaction term (R

change .004, p >.05) and the results show that there is no moderating effect of social approval (β = .006, t(846) = .105, p = .916). H4a is not supported. Figure 3 visually presents the (insignificant) interaction result.

The effects on donation intention. Table 5 shows the results of model 2, 4 and 6. All three models fit the data well (p <.05). To test hypothesis 2 (brand extension fit positively influences donation intention) model 2 and 4 are assessed. The results show that fit has no significant influence on donation intention (β = -.066, t(846) = -.863, p = .388). Although the change in R

was significant (p <.05) for model 4, the addition of attitude did not change the effect of fit on donation intention (β = -.045, t(846) = -1.501, p = .134). H2 is not supported.

Model 6 examines if social approval negatively moderates the effect of brand extension fit on

donation intention. Adding the interaction term does not improve the model (R

change .002,

p >.05). Social approval does not moderate the effect of fit on donation intention (β = .061,

t(846) = 1.02, p = .308). This concludes that H4b is not supported. Figure 3 visually presents

the (insignificant) interaction result.

TABLE 5

Regression results (standardized betas) for brand attitude and donation intention

Brand attitude Donation intention

Hypothesis (effect)

Model 1

Model 5

Hypothesis (effect)

Model 2

Model 4

Model 6

Model 3

Constant 6.474^a 6.434^a 6.542^a 2.644^a 2.678^a 2.613^a

Main variable

Fit 1 (+) .035^n.s. .032^n.s. 2 (+) -.029^n.s. -.045^n.s. -.078^c

Social approval .059^n.s. -.066^n.s.

Attitude 3a (+) .471^a .469^a .469^a

Control variable

Education -.132^a -.117^a -.121^a -.116^a

Interaction effect Fit*Social

approval 4a (-) .006^n.s. 4b (-) .055^n.s.

R² .001 .005 .018 .239 .241 .237

(Adjusted R²) (.000) (.002) (.016) (.236) (.236) (.235)

R² change .004^n.s. .221^a .002^n.s.

F-value 1.01^n.s. 1.43^n.s. 7.69^a 88.25^a 53.34^a 131.05^a

Note: ^a p-value <.01; ^b p-value <.05; ^c p-value <.10; ^n.s. p-value >.10

FIGURE 3

Interaction (fit*social approval) results for brand attitude and donation intention.

Brand attitude Donation intention

Note: y axis for brand attitude ranges from 6 to 7 and the y axis for donation intention ranges from 5.5 to 6.5.

Mediation. Model 3 examines hypothesis 3 (brand attitude positively influences donation intention) and fits the data well (p <.05). Brand attitude positively influences donation intention (β = .469, t(846) = 15.6, p <.001). This supports H3a, but because of the lack of significant relationships of fit with attitude and fit with donation intention, there is no mediation effect possible. H3b is not supported.

TABLE 6 Hypotheses results

Hypothesis Supported

H1 Brand extension fit positively influenced brand attitude. No

H2 Brand extension fit positively influences donation intention. No

H3a Brand attitude positively influences donation intention. Yes

H3b The relationship brand extension fit with donation intention is mediated by brand attitude. No H4a Social approval will negatively moderate the effect of brand extension fit on brand attitude. No H4b Social approval will negatively moderate the effect of brand extension fit on donation

intention.

No

Table 6 shows the final results of the analyses. Only hypothesis 3a is supported. The other hypotheses are not supported. In the next section, additional analysis are performed and discussed.

4.3 Additional analyses

The raccoon (high-fit) and puma (low-fit) were chosen to function as the high and low fit

manipulation conditions. From the analyses (see Table 3) emerged that participants judged the

fit of the puma extension as moderately fitting with an average fit score of 4.17 on a seven-

point scale. Therefore the puma did not function entirely as a low-fit brand extension. To

follow up on the results obtained in section 4.4, additional analyses are performed to examine

if the relationships still would be non-significant under conditions where only extreme fit

observations are retained. 573 respondents remained, 374 in the high-fit condition, and 199 in

the low-fit condition. All hypotheses will be tested again under conditions where respondents

that gave an average fit score on the puma (low fit condition) of more than four and an

average fit score on the raccoon (high fit condition) of less than four are omitted.

Descriptive statistics. Descriptive statistics were calculated as can be seen in Table 7. As expected, the fit scores and social approval (manipulation check) scores are statistically significantly different (p < .001) between the two fit and two social approval groups.

TABLE 7

T-test of the fit and social approval manipulation check scores in the additional analyses

N mean sd df t Sig.

Fit High 374 5.66 .792 353.68 -38.01 < .001

Low 199 2.72 .926

Social approval Majority 296 5.46 1.32 495.37 -2.13 < .001

Minority 277 3.33 1.86

A t-test is performed to check for significant differences between groups with respect to the dependent variables attitude and donation intention (see Table 8). Respondents in the high fit (M = 6.54) condition significantly held higher attitudes towards the brand in comparison to the low fit (M = 6.38) condition (t(297.98) = -2.65, p = .008). There are no significant results found for donation intention.

TABLE 8

T-test for the scores of brand attitude and donation intention in the additional analyses

Fit N mean sd df t Sig.

Brand attitude High 374 6.54 .749 297.98 -2.653 .008

Low 199 6.38 .510

Donation intention High 374 5.99 .768 571 -1.191 .234

Low 199 5.90 .818

Hypotheses testing. Table 9 shows the results of the hypotheses tests performed for the additional analysis. All models are significant and therefore indicate to fit the data well (p

<.05). Model 1 examines the relationship of fit with brand attitude. The results show that fit

has an influence on attitude towards the brand (β = .123, t(572) = 2.97, p = .003). H1 is

therefore supported. When testing the moderating effect of social approval with model 5, no

significant results were found and the model did not change significantly by adding the

interaction term (R

change .008, p >.05). Social approval does not moderate the relationship

between fit with brand attitude (β = -.072, t(572) = -.871, p = .384) and H4a is therefore not supported.

TABLE 9

Regression results (standardized betas) for brand attitude and donation intention in the additional analyses

Brand attitude Donation intention

Hypothesis (effect)

Model 1

Model 5

Hypothesis (effect)

Model 2

Model 4

Model 6

Model 3

Constant 6.385^a 6.304^a 6.382^a 2.243^a 2.257^a 2.239^a

Main variable

Fit 1 (+) .123^a .158^a 2 (+) .043^n.s. -.016^n.s. -.026^n.s.

Social approval .134^c .001^n.s.

Attitude 3a / 3b (+) .489^a .488^a .487^a

Control variable

Education -.121^a -.095^b -.096^a -.094^b

Interaction effect

Fit*Social approval 4a (-) -.072^n.s. 4b (-) .019^n.s.

R² .015 .024 .017 .251 .252 .251

(Adjusted R²) (.013) (.018) (.014) (.248) (.245) (.249)

R² change .008^c .235^a .000^n.s.

F-value 8.82^a 4.59^a 4.93^a 63.73^a 38.16^a 95.63^a

Note: ^a p-value <.01; ^b p-value <.05; ^c p-value <.10; ^n.s. p-value >.10

Model 2 and 4 examine the effect of fit on donation intention. The regression analysis

shows no significant coefficients for the relationship of fit with donation intention (β = .043,

t(572) = 1.03, p = .304). This relationship is also not significant when in model 4, attitude is

added (β = -.016, t(572) = -.436, p = .663). Therefore H2 is not supported. Adding the

interaction term in model 6 does not significantly improve the model (R

change .000, p >.05)

and no moderating role of social approval is found (β = .019, t(572) = .255, p = .799). Figure

4 presents the (insignificant) interaction results.

FIGURE 4

Interaction (fit*social approval) results for brand attitude and donation intention in the additional analyses

Brand attitude Donation intention

Note: y axis for brand attitude ranges from 6 to 7 and the y axis for donation intention ranges from 5.5 to 6.5.

Mediation. To be able to check for mediation, fit must have an effect on brand attitude. Table 9, model 1 shows that fit effects attitude with a positive significant beta coefficient (β = .123, t(572) = 2.97, p = .003). The second requirement is that fit has to have an effect on donation intention. Table 9, model 2 shows that there is no significant relationship of fit with donation intention (β = .043, t(572) = 1.03, p = .304). The third requirement is that attitude must affect donation intention. Model 3 examines the effect of attitude on donation intention and yields a highly significant positive beta coefficient (β = .487, t(572) = 13.41, p < .001). The final requirement is that in the full model (model 4), the fit coefficient should be smaller than in model 2. Although the beta coefficient for fit is indeed smaller in model 4 (-.016) than in model 2 (.043), they are both not significant. The lack of the relationship of fit on donation intention implies there is no relationship to be mediated and therefore attitude does not mediate the relationship between fit with donation intention.

This concludes that when extreme fit scores are retained only hypothesis 1 (brand

extension fit positively influences brand attitude) and hypothesis 3 (brand attitude positively

influences donation intention) are supported.

5 CONCLUSIONS AND RECOMMENDATIONS

This study investigates the moderating role of social approval on the relationship between fit and brand attitude and fit and donation intention. Expected is that brand extension fit will positively influence brand attitude (hypothesis 1) and donation intention (hypothesis 2) and that attitude positively influences donation intention (hypothesis 3a) and functions as a partial mediator in the relationship of brand extension fit with donation intention (hypothesis 3b).

Because ill-fitted brand extensions are shown to have negative effects with respect to consumers’ brand attitude and behavioural intention, the moderating role of social approval is investigated. Predicted is that social approval will lower the negative effects of an ill-fitted brand extension and thereby negatively moderates the effect of brand extension fit on brand attitude (hypothesis 4a) and donation intention (hypothesis 4b).

Against predictions, fit has no influence on brand attitude or on donation intention, and social approval does not moderate these relationships. The only significant finding is that brand attitude has a positive influence on donation intention.

The lack of results in the current study could have been caused by the fact that donors had recently donated (95.6%) and were aware that the charitable organisation also sheltered other animals besides monkeys. Therefore brand associations might have already been diluted and could have caused the low-fit extension associations to be more lined up with the parent brand associations than if donors had not been aware that the charitable organisation is shifting its focus towards rescuing other animals. Kim and Yoon (2013) found that association ambiguity is negatively related to consumers’ evaluation of perceived fit, which could explain the high perceived fit score for the low-fit manipulation (M=4.17 on a seven-point scale).

This may have caused the lack of results for the relationships of fit with brand attitude and behavioural intention. Also, if extension associations had not confused donors, uncertainty may not have occurred and social approval might not have had the desired effect. Hogg, Kruglanski, and van den Bos (2013) state that only when people feel uncertain about their perceptions, beliefs, and attitudes, they look for the guidance of others. This indicates that, in absence of uncertainty, social approval should have had no effect.

To test if the high association similarity in the low-fit condition could have caused the

results to be non-significant, additional analyses were performed where only respondents with

extreme fit scores (less than 4 on a 7 point scale and more than 4 on a seven point scale) were

retained. When respondents were omitted, results are only found for the relation of fit with

brand attitude (with fit explaining 1.3% of the variance in brand attitude) and for the effect of

An important outcome of this research is the valuable insights that were gained for the charitable organisation that cooperated in this study. The results suggest that the charitable organisation does not have to consider the amount of fit when extending their brand to a new animal. Extending the brand to a similar or dissimilar animal has no influence on donors’

brand attitude and donation intention.

This study has been conducted in the charitable organisation market, while brand extension fit research has mainly been conducted in the (fast moving) consumer goods market (Völckner & Sattler 2006). The results found in this study could be valid for the charitable organisation context but not for others. Although this sounds promising, certain limitations have to be taken into account.

This study was conducted among a relatively old age group (average 57.83 years old).

This limits the generalizability of the results. This also goes for the fact that the study was conducted in the Dutch market, which limits the geographical generalizability. Research in the field of brand extensions has already shown that there are differences between countries in respondents’ responses to fit (Pina, Iversen, & Martinez 2010). Future research could investigate if different results are obtained in other age groups and other countries.

Another limitation is the fact that donors were told up front that the news message (the manipulation) was composed of fictive information. Donors might have processed the message superficially, which could have caused the effectiveness of the fit manipulation and social approval cue (A poll has shown that 85% of the donors agree with the extension) to diminish completely. Myers and Sar (2013) state that third party social approval cues are evaluated based on source expertise (accurate knowledge), trustworthiness (willingness to communicate accurate knowledge), and perceived social value (similarity to others). In this experiment donors were told up front that the message (manipulation) was fictive. Donors therefore knew that the approval of others was spurious and hence was not trustworthy, nor a source of accurate knowledge.

A recommendation for further research is to provide clarity in the conception of fit.

Brand extension fit is measured in many different manners (e.g. Aaker & Keller 1990; Park,

Milberg, & Lawson 1991). Because there is a lack of coherence in how fit should be

determined or measured (Bhat & Reddy 2001; Mao & Krishnan 2006), it is difficult to

interpret fit results from different academic research. Future research would benefit from a

clear conception of fit. Also, no results for the moderating role of social approval in the

charitable organisation context have been found. This does not imply that no results will be

contexts such as the (fast moving) consumer goods market, where academic research has shown that low-fit extensions have negative effects on the outcomes.

Acknowledgements. I would like to thank the people who were involved during the creation

of this research paper, especially Dr. J.C. Hoekstra for her superior supervision and support

during the whole process. I would also like to thank the charitable organisation for the

opportunity to do a field experiment.

REFERENCES

Aaker, D. (1990). Brand Extensions: The Good, the Bad, and the Ugly. Sloan Management Review, 31(4), 47-56.

Aaker, D. A., & Keller, K. L. (1990). Consumer Evaluations of Brand Extensions. Journal of Marketing, 54(1), 27-41.

Ahluwalia, R. (2008). How Far Can a Brand Stretch? Understanding the Role of Self- Construal. Journal of Marketing Research (JMR), 45(3), 337-350.

Ajzen, I., & Fishbein, M. (1977). Attitude-behavior relations: A theoretical analysis and review of empirical research. Psychological Bulletin, 84(5), 888-918.

Ajzen, I., Joyce, N., Sheikh, S., & Cote, N. G. (2011). Knowledge and the Prediction of Behavior: The Role of Information Accuracy in the Theory of Planned Behavior. Basic &

Applied Social Psychology, 33(2), 101-117.

Bartlett, M.S., 1937. Properties of sufficiency and statistical tests. In: Proceedings of the Royal Statistical Society of London A pp.160 268-282.

Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51(6), 1173-1182.

Batra, R., Lenk, P., & Wedel, M. (2010). Brand Extension Strategy Planning: Empirical Estimation of Brand–Category Personality Fit and Atypicality. Journal of Marketing Research (JMR), 47(2), 335-347.

Bhat, S., & Reddy, S. K. (2001). The impact of parent brand attribute associations and affect on brand extension evaluation. Journal of Business Research, 53(3), 111-122

Bischoff, I., & Egbert, H. (2013). Social information and bandwagon behavior in voting: An economic experiment. Journal of Economic Psychology, 34270-284.

Boush, D. M., & Loken, B. (1991). A Process-Tracing Study of Brand Extension Evaluation.

Journal of Marketing Research (JMR), 28(1), 16-28.

Burnkrant, R. E., & Cousineau, A. (1975). Informational and Normative Social Influence on Buyer Behavior. Journal of Consumer Research, 2(3), 206-215.

Cialdini, R. B., & Goldstein, N. J. (2004). Social influence: Compliance and Conformity.

Annual Review of Psychology, 55(1), 591-621.

Charity Aid Foundation 2014. Accessed 5-2-2016.

https://www.cafonline.org/docs/default-source/about-us-

publications/caf_wgi2014_report_1555awebfinal.pdf

Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16297-334.

Cui, G., Lui, H., & Guo, X. (2012). The Effect of Online Consumer Reviews on New Product Sales. International Journal of Electronic Commerce, 17(1), 39-58.

Dall’Olmo Riley, F., Hand, C., & Guido, F. (2014). Evaluating brand extensions, fit perceptions and post-extension brand image: does size matter?. Journal of Marketing Management, 30(9/10), 904-924.

Dall’Olmo Riley, F., Pina, J. M., & Bravo, R. (2015). The role of perceived value in vertical brand extensions of luxury and premium brands. Journal of Marketing Management, 31(7- 8), 881-913

Dawson, J. F. (2014). Moderation in Management Research: What, Why, When, and How.

Journal of Business & Psychology, 29(1), 1-19.

DeiVecchio, D., & Smith, D. C. (2005). Brand-Extension Price Premiums: The Effects of Perceived Fit and Extension Product Category Risk. Journal of The Academy of Marketing Science, 33(2), 184-196.

Erb, H., Bohner, G., Schmälzle, K., & Rank, S. (1998). Beyond conflict and discrepancy:

Cognitive bias in minority and majority influence. Personality And Social Psychology Bulletin, 24(6), 620-633.

Evangeline, S. J., & Ragel, V. R. (2016). The Role of Consumer Perceived Fit in Brand Extension Acceptability. IUP Journal of Brand Management, 13(1), 57-72.

Ford, J. K., MacCallum, R. C., & Tait, M. (1986). The application exploratory factor analysis in applied psychology: a critical review and analysis. Personnel Psychology, 39(2), 291- 314.

Grohmann, B. (2009). Gender Dimensions of Brand Personality. Journal of Marketing Research (JMR), 46(1), 105-119.

Hee Yeon, K., & Jae-Eun, C. (2011). Consumer purchase intention for organic personal care products. Journal of Consumer Marketing, 28(1), 40-47

Hem, L. E., de Chernatony, L., & Iversen, N. M. (2003). Factors Influencing Successful Brand Extensions. Journal of Marketing Management, 19(7/8), 781-806.

Hogg, M. A., Kruglanski, A., & van den Bos, K. (2013). Uncertainty and the roots of extremism. Journal of Social Issues, 69(3), 407-418.

Huifang, M., & Krishnan, H. S. (2006). Effects of Prototype and Exemplar Fit on Brand

Extension Evaluations: A Two-Process Contingency Model. Journal of Consumer