Spillover effects of ingredient branding
The importance of brand fit
Author: Lara Mochtar (11954612)
Supervisor: Tina Dudenhöffer
Date: 22-06-2018
University of Amsterdam - Amsterdam Business School
MSc Business Administration – Marketing track
2
Statement of originality
This document is written by Lara Mochtar who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Table of contents
ABSTRACT ... 4 1. INTRODUCTION ... 5 2. LITERATURE REVIEW ... 8 2.1 LEVERAGING BRAND EQUITY ... 8 2.1.1 Brand equity ... 8 2.1.2 Brand extensions ... 8 2.2 CO-BRANDING ... 9 2.2.1 Ingredient branding ... 10 2.3 SPILLOVER EFFECTS ... 11 2.3.1 Spillover effects in co-branding ... 12 2.3.2 Spillover effects in ingredient branding ... 13 2.4 THE ROLE OF FIT ... 14 2.4.1 Bases of fit ... 14 2.4.2 Brand fit in ingredient branding ... 16 2.4.3 Categorization theory ... 17 2.4.4 New-product-brand fit ... 17 2.4 THE ROLE OF PRIOR PARENT BRAND SATISFACTION ... 19 2.5 CONCEPTUAL MODEL ... 20 3. METHODOLOGY ... 21 3.1 RESEARCH DESIGN AND PROCEDURE ... 21 3.2 STIMULI DEVELOPMENT... 22 3.3 MEASURES ... 23 3.3.1 New-product-brand-fit (NPBF) ... 25 3.3.2 Brand satisfaction ... 25 3.3.3 Parent brand attitude change (PBAC) ... 26 3.3.4 Manipulation checks and control variables ... 26 3.4 DATA COLLECTION AND SAMPLE CHARACTERISTICS ... 27 4. RESULTS ... 28 4.1 RANDOMIZATION CHECKS... 28 4.2 MANIPULATION CHECKS ... 28 4.3 DESCRIPTIVES AND CORRELATIONS ... 30 4.4 NORMALITY ... 30 4.2 HYPOTHESES TESTING ... 31 5. DISCUSSION AND CONCLUSIONS... 35 5.1 GENERAL DISCUSSION ... 35 5.1.1 Main effects of NPBF on PBAC... 35 5.1.2 Moderating effect of NPBF host ... 36 5.1.3 Moderating effect of prior satisfaction with the host brand ... 36 5.3 THEORETICAL IMPLICATIONS... 37 5.3 MANAGERIAL IMPLICATIONS... 37 5.3 LIMITATIONS AND FURTHER RESEARCH ... 38 5.4 CONCLUSION ... 39 REFERENCE LIST ... 40 APPENDICES ... 49 A.1 PRE-TESTS FOR STIMULI DESIGN ... 49 A.1.1 Pretest 1 - qualitative ... 49 A.1.2 Pretest 2 – quantitative ... 49A.2 SURVEY PRE-TEST ... 55
A.3 CONTROL VARIABLES AND MANIPULATION CHECKS MAIN STUDY ... 59
A.4 SURVEY MAIN STUDY ... 60
A.5 HISTOGRAMS DATA DISTRIBUTION ... 68
A.6 PROCESS MODEL OUTPUTS ... 70
4
Abstract
Co-branding is an increasingly popular strategy for the launch of a successful new product, meaning that two brands decide to create a new product together. Prior research has examined the influence of brand fit between two brands in such an alliance but has overlooked the fit between the new product and the constituent brands. The purpose of this study is to examine the effects of this fit on the post-launch attitude change toward the parent brand, also known as spillover effects. Moderating effects of the fit between the main brand and the product are taken into account as well as prior satisfaction with that brand. An experiment is employed with eight manipulations randomized among 312 Dutch respondents. The focus of the study was in the product categories of Fast Moving Consumer Goods. Regression analysis are conducted to find main and interaction effects. The results show positive main effects of both fits between the new product and the two parent brands on the attitude change toward the main brand. No moderating effects were found. Overall, this research contributes to the literature on co-branding, brand fit and spillover effects by bridging the current conversations. Brand managers benefit from the findings as they can make a more informed decision when inviting another brand to be an ingredient in their new products.
5
1. Introduction
Companies continuously seek to grow and to increase financial benefits, which can be difficult in the saturated and competitive market of Fast Moving Consumer Goods (FMCG). They often launch new products to extend their brands beyond their original product offering, a strategy called brand extensions (Tauber, 1988). In 2015, it was found that 76% of all FMCG introduced in Europe over the past few years failed, meaning they were taken out of distribution before the end of the first year after their launch (Nielsen, 2015). To overcome this, a different strategy to create more successful products has been performed many times over the last years. Brands pair up with a different brand in order to increase market share (Desai & Keller, 2002) and competitiveness (Simonin & Ruth, 1998). This co-branding strategy is used by companies to leverage strengths of both brands (Washburn, Till, & Priluck, 2004) and to share the costs that come with introducing a new product or brand (Erevelles, Stevenson, Srinivasan, & Fukawa, 2008), as an alternative to brand extensions.
Looking at the current landscape of Dutch FMCG, a few recent examples can be named. An alliance between grocery store Albert Heijn and producer of vegetarian alternatives for meat De Vegetarische Slager who together created vegetarian microwave meals has just been announced. Also, in a collaboration between washing detergent producer Robijn and baby product brand Zwitsal a new detergent and fabric softener with the popular and characteristic smell of the latter were created. These products seem to be a great success, as the brands just launched an additional softener with an even more intense Zwitsal scent. Another example is the Broodje Unox sauce for use at home, a collaboration between sauce producer Calvé and the out of home sandwich brand Unox, that serves a typical and unique sauce on its sandwich. However, not all resulting products of co-branding strategies have led to positive consequences for the brands involved. Coffee brand Douwe Egberts decided to terminate the collaboration with milk producer FrieslandCampina and take the cold coffee they sell off the shelves, because it is not logical and confusing for the consumer (“DE neemt afscheid van Campina”, 2009). The former examples of co-branding are mainly of a certain type of co-branding, namely ingredient branding, which defined as a strategy in which an important attribute of a brand is used within a new product of another brand as an ingredient (Desai & Keller, 2002). In this strategy the first brand is more prominent, leading to interesting partnerships, asymmetrical effects for the allying brands and therefore important decisions to make.
6
There is consensus in the research field of brand extensions and co-branding that a good fit helps to influence whether the co-branding strategy will succeed or fail. A good fit effects both the co-branded product and the constituent brands in a positive way (Aaker & Keller, 1990; Simonin & Ruth, 1998). Effects on the brands that either deploy a brand extension or co-branding strategy are called spillover effects. Different bases of fit are explored, determining both the success for the new product and the brands involved. For brand extensions one of the most important types of fit seems to be the perceived fit between the original brand image of a brand and the extension. It leads to a better evaluation of the extension itself (Aaker & Keller, 1990) and positive spillover effects for the parent brand (Swaminathan et al., 2001). Although for branding multiple studies have found that a good perceived fit between the brands in a co-branding strategy create stronger positive spillover effects (James, 2005; Simonin & Ruth, 1998; Schnittka et al., 2017), it has not been empirically studied yet whether a perceived fit between the co-branding product and both constituent brands does the same. Bouten, Snelders, and Hultink (2011) and Thompson and Strutton (2012) did research the type of fit in a co-branding context, but only researched whether evaluations of the new product were influenced by it. To the author’s knowledge, until date, no one has conducted a study on the effects of spillover. Since the fit has been argued to be of huge importance for parent brands of brand extensions, it seems useful to explore any consequences for co-branding, as strategies are considered quite similar and can be used as alternatives.
Questions arise from these important determinants of spillover effects on brand extensions which seem to be overlooked in the field of co-branding. To answer these, the following research questions were formulated: (a) Does a high level of fit between a new co-branded product and it’s two parent brands lead to a more positive attitude toward the host brand, and if so (b) is the choice of the ingredient as important if the host brand has a good fit? Furthermore, (c) is the strength of this effect different for different levels of prior brand satisfaction with the parent brands? Contributions to theory will be made as current knowledge can be extended by the results of this study. There is not yet an integration of spillover effects with the product-brand fit in the field of co-branding. This research will provide understanding of whether the fit between a new product and its constituent brands will induce a positive attitude change toward the main brand in the alliance. A different addition to the literature will be the incorporation of prior brand satisfaction, as no study has yet related this to spillover effects. For brand managers this paper will contribute to the understanding of spillover effects due to a good brand fit between
7 constituent brands and the co-branded product. It might be that although one of the brands has a perfect fit, the ally has a lower perceived fit. It will be interesting for practitioners to see whether it makes a different for the attitude change toward their brand when they take the one brand or another. As the competition is fierce and product introductions are risky, they can benefit from the extra knowledge that will be derived from this research and make more informed decisions. The following chapter of this paper contains a literature review, in which an overview is given of the current state of relevant literature on co-branding and spillover effects. After positioning the research in light of existing literature, a conceptual framework is proposed at the end of the chapter. The chapter thereafter describes the sample that is used and outlines in what way the research is designed. The results section explains which analyses were done and whether the hypotheses are supported based on the collected data. Subsequently, this paper provides a thorough discussion of the results. Conclusions are drawn in the last chapter as well, including limitations of the study, directions for future research and implications for scholars and
8
2. Literature review
This chapter provides an overview of the current literature on ingredient branding strategies and their spillover effects on the parent brands. Based on the review of the literature, several hypotheses are established, that lead to a conceptual model for this study.
2.1 Leveraging brand equity
2.1.1 Brand equity
Brands are intended to identify the product of a seller and to differentiate it from a competitor (Kotler & Turner, 1979). Customer-based brand equity is defined as “the differential effect of brand knowledge on consumer response to the marketing of the brand” (Keller, 1993), and in the remainder of this paper referred to as brand equity. This brand equity is considered as positive (negative) when people respond more (less) favorably to a brand’s marketing than they would respond to similar marketing from an unbranded version of that product or service (Keller, 1993). Brand knowledge consists of two components, namely brand image and brand awareness. The brand image of a brand is defined as “perceptions about a brand as reflected but the brand associations held in consumer memory” (Keller, 1993). Those associations can be of different types, like attributes of the products or services a brand offers, benefits of those products or services, and attitudes toward a brand. Brand attitudes are defined as an overall evaluation of a brand (Wilkie, 1986) and are a function of the evaluative judgement of most prominent associated attributes and benefits of a brand (Ajzen & Fishbein, 2000). Brand associations also include any additional meaning of a brand to the consumer, like the typical user of a brand or the fact that people can use it to express their personality with (Keller, 1993). The brand associations making up the brand image can be of varying favorability, strength and uniqueness. A positive brand image is one with favorable, strong and unique associations, enabling brands to set higher prices, making marketing communication more effective and stimulating consumers to put more effort into finding distribution channels for the product (Keller, 1993). 2.1.2 Brand extensions With fierce competition in the FMCG industry, a substantial amount of risk comes with the launch of a product. To overcome the risk that a new product fails and to reduce high costs that are associated with the introduction, many companies try to leverage the equity of their already well-established brands (Reddy, Holak, & Bhat, 1994). When entering new markets, the use of a brand extension strategy is attractive because there is the advantage of the recognition of a brand
9
image and name. The strategy entails the introduction of new products in a category similar or different to the category of a parent brand, under the same brand name (Tauber, 1988). Consumers may base their decisions to adopt the new product on secondary associations coming from the brand they already know (Keller, 2005). Consumers tend to use existing value perceptions related to the original branded product to evaluate a new offering from that same brand (Aaker & Keller, 1990; Boush & Loken, 1991). If this is the case, the brand affect from the parent product is transferred to the extension product (Liu, Hu, & Grimm, 2010). Affect refers in this context to a consumer’s liking of a brand. The affect transfer from a branded product to a brand extension can be explained with the theory of schema-triggered affect transfer (Andersen & Baum, 1994). A schema is “a cognitive structure that represents knowledge about a concept or type of stimulus, including its attributes and the relations among these attributes”(Fiske & Taylor, 1991, p.93). The theory proposes that affect transfer happens when a new item matches an existing schema (Fiske & Pavelchak, 1986). Fiske & Taylor (1991) argue that affect is carried by a schema that is based on prior experiences and knowledge. In the context of brand extensions, a consumer’s schema about an existing product include both product knowledge and affect. Consequently, when a stimulus in the form of a brand extension is congruent with existing schema, the prior knowledge and affect will transfer to the brand extension (Liu et al., 2010).
2.2 Co-branding
An alternative strategy to brand extensions that has become more popular recently is co-branding. Co-branding can be defined as two brands combining their forces to create a single, new product (Besharat & Langan, 2014). It should not be confused with different types of brand alliances, that do not always lead to co-development of a totally new product but can also be a type of strategic alliance meant to incorporate the associations of two brands and improve their image (Cooke & Ryan, 2000). Co-advertising is also often mistaken for co-branding, but this is a communication strategy in which two brand both promote their own products in one single advertisement (Leuthesser, Kohli, & Suri, 2003). To be able to make a clear distinction between co-branding and other brand partnerships, Besharat and Langan (2014) suggest the conceptualization of the strategy to have some key characteristics. First, it should be a long-term relationship between two (or more) brands to launch a new service or product. This can be both in a new or already existing market. Also, both brands should be present and visible to consumers on the logo, product or package. Furthermore, complementary tangible (i.e. physical) attributes or intangible (i.e. brand image) attributes need to be present. Lastly, both brands in the co-branding strategy must maintain their own identity, so they can be marketed independently (Besharat & Langan, 2014).
10
Several benefits of co-branding arise from the field of literature. In co-branding, the strengths of two brands can be leveraged (Washburn et al., 2004) and the high costs that are associated with introducing new products and brands can be shared among the constituent brands (Erevelles et al., 2008). Chances of adoption of a newly introduced product are greater if two known brands are involved (Blackett & Russell, 1999). Moreover, when one of the brands is unknown and partners with a highly reputable ally, it will signal marketplace credibility as consumers base their assumption of overall quality on the second brand name that has high levels of brand equity (Rao & Ruekert, 1994). Rao, Qu, and Ruekert (1999) showed that co-branded products provide an improved signal of quality in comparison to mono-branded products, when a brand has unobservable attributes. Consumers can better evaluate the product when a brand is allied with another brand that is perceived as more vulnerable to sanctions by consumers. These are reasons that contribute to opinions that co-branding can be in many cases a superior alternative for brand extension (Helmig, Huber, & Leeflang, 2008) 2.2.1 Ingredient branding One particular type of co-branding is ingredient branding. It is defined as a strategy in which an important attribute of a brand (the ingredient brand) is used within a new product of another brand (the host brand) as an ingredient (Desai & Keller, 2002). Whereas in other co-branding strategies the product belongs to both brands in the alliance, in ingredient branding the product belongs only to the host brand, as the ingredient cannot be marketed separately for this particular product category (Norris, 1992). For this reason, the roles are clearly divided between the host brand and the ingredient brand. The alliance is defined as being asymmetrical with the host brand being more dominant and thus the primary brand that is associated with the co-branded product (Uggla, 2004). Uggla & Åsberg (2010) also explain that an ingredient brand can add both functional and symbolic associations to the host brand. A functional association would be relating to an attribute, for example the great taste of a certain ingredient and a symbolic association would be a luxurious feel that the ingredient brand adds to the product. Ingredient branding is a relatively new branding strategy that originated in the 1980s (Norris, 1992), but has been popular ever since. It is used in multiple product categories, like electronics (e.g. Dell computers as host with Intell processor as ingredient), foods (e.g. McFlurry as host with M&M’s as ingredient) and soft drinks (Coca Cola as host with Stevia sweetener as ingredient in Coca Cola life). In the FMCG industry this type of co-branding is used as a predominant alternative to brand extensions (Baumgarth, 2004), as illustrated by the recent Dutch examples mentioned in the
11
introduction of this paper. The popularity is the reason for the focus on ingredient branding in this study.
In response to the numerous real-life cases, various studies have been conducted on application of the strategy and its effects (Desai & Keller, 2002; McCarthy & Norris, 1999a; Park, Jun, & Shocker, 1996). Several benefits have emerged from the literature and scholars seem to generally agree that ingredient branding is a very beneficial way to introduce a new product. Desai and Keller (2002) show that a co-branded ingredient evokes better quality perceptions and more favorable evaluations of the extension than an ingredient from the host brand itself, when the ingredient is an entirely new attribute or characteristic to the product. This favorable attitude toward the co-branded extension even lays a foundation for subsequent category extensions introduced by the host brand alone. Park et al. (1996) also argue that the weaknesses that a host brand has in a product extension category can be overcome by allying with a strong ingredient brand in this category. They combined a host brand and an ingredient brand with complementary attribute levels to a new extension that had a better attribute profile than a similar direct extension of the host brand. The brand that benefits the most from ingredient branding with a well-known brand is a moderate-quality host brand, as purchase intentions, taste perceptions and product evaluations are more influenced compared to for a high-quality host brand.
2.3
Spillover effects
As discovered from existent literature, a product that is introduced through either a brand extension or co-branding strategy can be both positively and negatively affected as a result of that strategy. The theory on spillover effects suggests that this evaluation of the new product is only one of two processes that make up the existing framework of feedback. The second process involves re-evaluation of the parent brand after the product is launched (Dwivedi, Merrilees, & Sweeney, 2010), also referred to as spillover effects. Gürhan-Canli and Maheswaran (1998) suggested schema change theory is involved in this process as well. Each piece of new brand information, in this case an extension of that brand, causes small modifications in the brand knowledge structure of a consumer. This change in knowledge and resulting attitude changes are spillover effects from the extension on the parent brand (Gürhan-Canli & Maheswaran, 1998). Both positive and negative spillover effects can occur as a result of extending a brand. Balachander and Ghose (2003) found that positive exposure to brand extensions can positively influence the image of the parent brand, which then has an impact on sales in other categories of that brand12 (Balachander & Ghose, 2003). On the contrary, unsuccessful brand extensions can hurt the brand because of negative perceptions generated by such a failure (Aaker, 1996). 2.3.1 Spillover effects in co-branding
Researchers in the field of co-branding have found that from a co-branding strategy changes in beliefs and attitudes toward the constituent brands can result (Simonin & Ruth, 1998). Brand values can be influenced because both brands in a co-branding product look similar in quality. This is relevant especially for lesser-known brands that cannot effectively signal high quality (Rao e.a., 1999). Levin & Levin (2000) have also demonstrated that for lesser known brands that are not well-defined yet in the mind of a consumer assimilation effects exist in dual brand evaluations. When there is one new brand that has ambiguous attributes, affect transfers easily from the well-known brand to the unknown brand. An alliance has the ability to improve the attitude towards the weaker brand of the two, as also shown by (Washburn, Till, & Priluck, 2000). They indicate that consumers can form positive impressions about an unknown brand when they see it together with a high-equity partnering brand in a co-branding relationship. All these studies agree on that the weaker brand of the two will be perceived of higher quality in a co-branding strategy with a well-known and strong brand. The associations of the well-known brand are extended to the lesser-known brand (Besharat, 2010). Simonin & Ruth (1998) find that the attitude toward a co-branding case can consequently have a positive effect on the post-attitude toward both of the brands involved, but overall, co-branding tends to help an unfamiliar brand more than an already familiar brand. Aforementioned effects are all positive effects, as most research on the topic of spillover effects is on positive effects and rule out negative effects. It is argued that a brand with low brand equity can benefit from associating with a high-equity brand without harming the brand value of the latter (Washburn et al., 2004). Similarly, Votola and Unnava (2006) argued that the main brand is not weakened by any negative associations from another brand, besides when consumers actually think the failing of the relationship is related to this strong brand. However, Keller (2005) argues that leveraging a brand through another brand can be risky, because some control of the brand image is given up. The other brand will have a whole lot of own associations, only some of which a brand manager wants to be transferred. According to Keller (2005) it is difficult to manage the transfer process in a way that no negative spillover effects occur. There are a few studies that do cover negative spillover effects. Researchers have identified potential negative effects of co-branding for both the constituent brands, like the losing of brand esteem (Lebar et
13
al., 2006). When a high esteem brand partners with a brand of lower esteem, the overall esteem perceptions would decrease, leading to dilution of the high esteem parent brand (Lebar et al., 2006). Schnittka et al. (2017) showed that when brands are presented jointly to consumers, evaluations of those brands are contrasted and evaluations of one brand improve at the expense of the partner brand. 2.3.2 Spillover effects in ingredient branding Similar findings on spillover effects have been found particularly for ingredient branding. Many scholars have illustrated only positive spillover effects of ingredient branding. An ingredient by a well-known brand incorporated into a lesser-known host brand improves quality perceptions about the latter (Vaidyanathan & Aggarwal, 2000). Desai & Keller (2002) find a different kind of positive effect of an ingredient brand on a host brand, which is that the expansion helps the extendibility of the host brand. Lebar et al. (2006) argue that this is because ingredient branding builds more points of differentiation, for the host brand in particular. In contrast, a few others say that risks are also at play in ingredient branding. A study has shown that poor selection of an ingredient brand may not only lead to a failure of the new product but can potentially harm the host brand through dilution of the brand equity (Janiszewski & Osselaer, 2000) or they can overpower it (McCarthy & Norris, 1999). Ponnam and Balaji (2015) suggest that for the equity of the host brand ingredient branding should be preferred to an incremental product improvement, but only when the category is perceived as a low involvement (like food) and brand equity of the host brand is low. In a high involvement context and when brand equity is already high, it should be preferred to improve the current product, in order not to dilute the host brand. An advantage for the ingredient brand is that it has been argued to be relatively immune to negative spillover, especially when it is a very well-known and strong national brand (Leuthesser, Kohli, & Suri, 2003) However, Radighieri, Mariadoss, Grégoire, and Johnson (2014) found that when an ingredient branded product is a success it indeed positively affects both of the parent brands involved, but when a product fails, only a strong ingredient brand is somewhat safe from negative spillover effects (Radighieri et al., 2014).
Considered the above, there is clearly ambiguity in the field of spillover effects of co-branding and ingredient branding in particular. As co-branding researchers did find both positive and negative feedback on brands due to an ingredient branding alliance, this paper proposes that spillover effects can most likely occur. However, the ambiguity among researchers asks for further
14 investigation to gather understanding of what those potential effects are for attitudes toward the parent brand and what may influence those effects. As in ingredient branding there is a clear role division between the host and ingredient brands and the host brand has ownership over the new product, the host is perceived as the leader and will be held most responsible for any success or failure (Radighieri et al., 2014). As the brand manager of a host brand is most likely to select a brand to use as an ingredient in a product, potential spillover effects on the host brand are probably considered to be most important during this selection. Together with the findings by various researchers that the risk at spillover is bigger for host brands (Leuthesser e.a., 2003; Radighieri e.a., 2014) this is the reason for a research focus on spillover effects for the host brand and not the ingredient brand. In this study, the conceptualization of spillover effects is based on research by Dwivedi et al. (2010). Because feedback effects have been conceptualized as changes in the schema of a parent brand (Gürhan-Canli & Maheswaran, 1998), they focus on the investigation of impact on the change in parent brand attitudes. This study adopts the dependent variable parent brand attitude change (PBAC) toward the host for the conceptual model and used the same measure.
2.4
The role of fit
The consensus in the field of brand extension and co-branding research is that fit plays a major role in determining the success and spillover effects of both strategies (McCarthy & Norris, 1999). A good fit can have enhancement effects (Gürhan-Canli & Maheswaran, 1998; Zimmer & Bhat, 2004) and a lack of fit can have dilution effects on the parent brand (Martínez & Pina, 2003; Thorbjørnsen, 2005). The reason is that a good perceived fit strengthens associations, which leads to strong positioning of a brand and increased brand attitude (Sheinin, 1998) A bad perceived fit leads to negative associations and attitudes toward a parent brand (Keller & Aaker, 1992). Dwivedi et al. (2010) even argue that the perceived fit between the parent brand and the extension is the most important factor in determining spillover effects in a brand extension strategy, overpowering the importance of attitude toward the product itself. 2.4.1 Bases of fit No consistent conceptualization of fit has been used, and various bases of fit have been identified in the literature, including product fit, concept consistency and brand fit (Lanseng & Olsen, 2012). In a brand extension context there is only one brand involved. The fit in this context is naturally based on the fit between the parent brand and the extension product. Aaker and Keller (1990) found that when a new product fits the current products of a brand on a basis of physical15 attribute similarity, it will be evaluated more positively than when it does not, as a good fit ensures easy transferability of the experience of a brand from the original to an extension product. This similarity is also related to as product fit. In the case of ingredient branding for food products, there is by definition a good product fit between the constituent brands and the new product, as the host brand delivers a majority of the ingredients of the product and the ingredient brand adds an essential ingredient to the co-branded product (Vaidyanathan & Aggarwal, 2000). Due to the nature of an ingredient branding strategy, this study only considers brand fit. Park et al. (1991) reveal that for brand extensions not only a fit at the product-level is important, but likewise consistency between the concept of the brand and the extension, also referred to as brand fit. The product-level fit is defined as a more concrete type of fit, while brand concept consistency has a high level of abstraction (Park et al., 1991). Park, Jaworski, and Maclnnis (1986) proposed these concepts in an earlier paper. A distinction is made between functional and symbolic brand concepts. Functional brands satisfy practical and immediate needs, while the practical usage of symbolic brand is only incidental and they are used more to satisfy symbolic needs as those for prestige, status and expression of one’s personality (Park et al., 1986). An often used example of the difference between the two brand concepts is one in the watch category (Bhat & Reddy, 1998; Monga & John, 2010). In that category the brand Timex would be considered a functional brand as it is primarily useful to tell the time. Rolex on the other hand would be seen as a symbolic brand as it is bought and used primarily as a status symbol. Bhat and Reddy (2001) argue in their paper that brand image fit is broader in scope than Park et al.'s (1991) consistency with a symbolic or functional parent brand concept. Their concept of brand image fit is concerned with similarity of the extension product with specific associations of the parent brand, which for example also include the quality of a brand (Bhat & Reddy, 2001). This is supported by Keller (2003), who stated that any brand association in the consumer memory may serve as a basis of fit. The flexible perspective of brand fit as referring to the extent to which consumers see the new product as a logical and expected extension from the brand, was also adopted by some other scholars (Dwivedi et al., 2010; Tauber, 1988) and in the study (Bouten et al., 2011) that was partly extended by this research. For that reason, this study uses that perspective as well.
16 2.4.2 Brand fit in ingredient branding An extra dimension of co-branding in comparison with brand extensions is the second brand that is involved. In comparison to the strategy of brand extensions, for practitioners co-branding comes with the extra challenge of making a choice to collaborate with a brand that has good perceived fit with the own brand. The literature on the role of brand fit is therefore mainly about the fit between the two parent brands (Baumgarth, 2004; Moon & Sprott, 2016; Simonin & Ruth, 1998) Simonin & Ruth (1998) showed that a good perceived fit between the two brands in a co-branding alliance leads to positive co-Simonin & Ruth (1998) showed that a good perceived fit between the two brands in a co-branding evaluations. The share of brand- and product specific associations results in positive attitudes toward the co-branding product and the stronger the degree of fit is, the more positive consumers are about the resulting product (James, 2005). Results of a different study indicate that a good fit between partner brands, based on both brand image and product category, positively influences intentions to purchase an ingredient branded luxury product (Moon & Sprott, 2016). Another example was illustrated by Decker and Baade (2016), who found that partner dissimilarity in terms of firm size and country-of-origin image can negatively affect brand fit perceptions and consequently product evaluations. In a study that researched the relative importance of parent brands attitudes and brand fit for the purchase likelihood of an ingredient branded product in the FMCG category, it was found that brand fit is more important (Dalman & Puranam, 2017) On the contrary, Monga and Lau-Gesk (2007) argued that co-branding between brands with different brand personalities leads to better product perceptions that a single personality co-brand, as consumers are self-referencing and consequently prefer varied cognitive structures. When it comes to spillover effects, Simonin & Ruth (1998) also argue that perceived fit between the brands has a significant positive influence on them. In their often cited and replicated research they studied the car and consumer electronics industries and found that a good brand fit is related positively to the brand alliance attitudes, including spillover effects. Spillover effects were found to be much stronger for the lesser-known brand in an alliance, but as long as perception of brand fit between the two allies is not detrimental, the better-known brand might experience at least some positive spillover effects. (Simonin & Ruth, 1998). To find whether those results hold for FMCG, Baumgarth (2004) replicated the study and discovered that in this industry the perceived brand and product fit between allies have significant positive effects on spillover effects as well. More recently, research showed that negative spillover effects can occur when
17 there is a misfit between the two constituent brands, meaning they carry incongruent associations in the mind of consumers (Schnittka et al., 2017). This leads to a contrast of evaluations so that the evaluation of one brand improves while the other brand is diluted. 2.4.3 Categorization theory Many scholars have explained the transfer of associations in the brand extension and co-branding literature with the categorization theory (Aaker & Keller, 1990; Bhat & Reddy, 2001; Dalman & Puranam, 2017; Swaminathan, Reddy, & Dommer, 2012). The theory of schema-triggered affect transfer and categorization theory can be combined to understand why a fit is important in determining spillover effects. According to categorization theory, consumers use their schemas as categorical representations and try to assign any new piece of information into an existing category (Loken, Barsalou, & Joiner, 2008). In a brand extension context it is argued that brand extensions that are congruent with and processed in the same category as the parent brand are preferred (Aaker & Keller, 1990). If there is perceived consistency between the new information and attributes of the category schema, new information is evaluated more easily and therefore positively (Goodstein, 1993). If consumers can easily categorize a new product based on a shared associations that are also sensed in the brand image of the parent brand, there is a good perceived fit (Broniarczyk & Alba, 1994; Park et al., 1991). This fit leads to easy activation of the schema of the parent brand, which facilitates affect transfer of the brand to the new extension (Aaker & Keller, 1990). Because in a co-branding alliance two brands are involved, consumers also form a perception of whether the two brand images are consistent and belong to the same schema. Little consistency between the new information and the activated category schema would lead to more difficulty of categorization, negative thoughts and more negative spillover (Simonin & Ruth, 1998). 2.4.4 New-product-brand fit
As the previous paragraphs illustrated, in the current research fit in co-branding and ingredient branding is mainly based on the fit between the two parent brands. While co-branding and specifically ingredient branding is an important and rich field of research, surprisingly there are not many studies that research the effects of variation in fit between the partner brand images and the new product (Bouten et al., 2011; Thompson & Strutton, 2012). Helmig et al. (2008) already called for an investigation in this topic. This is reasonable, as such effects were found to be very important in a brand extension context, which is similar to co-branding and in particular ingredient branding. Furthermore, the research that is done only considered consequences for
18
the co-branded product evaluation and did not include any spillover effects on the parent brands (Bouten et al., 2011; Thompson & Strutton, 2012). Thompson & Strutton (2012) proposed, based on the categorization theory, that a fit between the parent brands and the new product is an important determinant for evaluations of the product. This was confirmed in their study and they even found that this is a more important factor for success than the fit between the constituent brands. Bouten et al. (2011) found similar results and conceptualized the fit between the image of the parent brands and the new product as new-product-brand fit (NPBF). In this study that conceptualization is adopted to extend their model. Both studies that researched this fit in the co-branding context only considered the impact on the evaluation of the new product itself. (Bouten et al., 2011; Thompson & Strutton, 2012) They did not include spillover effects. This study focusses on the more long-term effects of spillover compared to effects on the product. Also, the NPBF has not been studied in an ingredient branding context yet.
Dwivedi et al., (2010) presented that perceived fit between the parent brand and the extension product was found to be even more important to determine spillover effects then the attitude toward the brand extension itself. As brand extensions are so closely related to ingredient branding strategies, the considered importance of the brand fit variable calls for more attention to it in a co-branding context. Therefore, this study aims to extend the research by (Bouten et al., 2011) and will empirically research whether the by them introduced new-product-brand fit (NPBF) in co-branding also has an effect on the change in parent brand attitudes, also known as spillover effects. This type of fit is defined as the fit between the image of a brand and a new product (Bouten et al., 2011). NPBF is determined based on a shared concept formed by brand-unique associations rather than through comparing features of existing products with those of a new product (Bouten et al., 2011). The discovered significant positive effects of high levels of NPBF on co-branding product evaluations (Bouten et al., 2011) and the finding that a fit between a co-branding product and its parent brands is more important than a fit between the two brands lead to the following hypotheses:
H1: New-product-host-brand fit (NPBF host) positively affects host brand attitude change (PBAC host).
H2: New-product-ingredient-brand fit (NPBF ingredient) positively affects host brand attitude change (PBAC host).
19
As mentioned earlier, in ingredient branding the host brand has ownership over the new product, is perceived as the leader and is most prominently present in the alliance (Radighieri et al., 2014). It is proposed that because of that the fit of the host brand is more important and moderates the effect of NPBF ingredient on the spillover effects, so that when a host brand has a good brand fit with the product already, the NPBF ingredient matters less to determine spillover effects. This reasoning leads to the third hypothesis:
H3. The positive effect of new-product-ingredient-brand fit (NPBF ingredient) on host brand attitude change (PBAC host) is moderated by new-product-host-brand fit (NPBF host) such that higher levels of NPBF host weaken the positive effect on host brand attitude change.
2.4
The role of prior parent brand satisfaction
It is argued that a brand extension strategy can be used to optimize relationships with existing customers (Davis & Halligan, 2002). A favorable initial relationship enables consumers to anticipate other favorable interactions and derive benefits from them (Dwyer, Schurr, & Oh, 1987). Dwivedi and Merrilees (2012) propose therefore that a favorable initial relationship based on prior satisfaction with a brand positively influences changes in attitudes toward the brand as a result of brand extension. This was supported empirically in their research in the retail industry. Consistent with that finding, (Simonin & Ruth, 1998) found that prior attitudes through earlier satisfaction with the brand are predictive for spillover effects, because attitudes are rather stable psychological concepts (Fishbein & Ajzen, 1975). That negative spillover can be reduced through prior satisfaction has been shown as well (Connelly, 2001). Prior consumer profiles of a brand have shown to influence people’s opinions of that brand after a brand scandal. After Firestone recalled tires on Ford cars in 2000 because of more than 150 deaths and many injuries, 86 per cent of Ford satisfied consumers reported same or better opinions (Connelly, 2001).No studies on spillover effects have included the consumer characteristic of prior satisfaction with the brand yet, while investigation might find a difference in the strength of spillover effects among different consumer profiles. Besharat and Langan (2014) pointed out this need for further research earlier. This study does include prior satisfaction with the host brand as a moderator and proposes that the positive effects of NPBF host and NPBF ingredient are stronger for people who have had positive experiences with the brand and are thus satisfied than for consumers who have had a less positive experience. The following hypotheses were developed:
20
H4a. The positive effect of new-product-host-brand fit (NPBF host) on host brand attitude change (PBAC host) is moderated by prior satisfaction with the host brand such that higher levels of satisfaction strengthen the positive effect. H4b. The positive effect of new-product-ingredient-brand fit (NPBF ingredient) on host brand attitude change (PBAC host) is moderated by prior satisfaction with the host brand such that higher levels of satisfaction strengthen the positive effect.
2.5
Conceptual model
In Figure 1 a visualization of the hypotheses can be found. In order to empirically answer the research question that was put forward in the introduction of this paper, the model was created. In the next chapter it is described how the research is conducted. Figure 1 Pre-test constructs and their items New-product-host-brand fit (NPBF host) New-product-ingredient-brand fit (NPBF ingredient) Host brand attitude change (PBAC host) Prior host brand satisfaction H1 + H2 + H4a + H4b + H3 -21
3. Methodology
3.1 Research design and procedure
The aim of this study was to examine spillover effects of co-branding alliances between brands that fit or don’t fit with the newly created product and whether those effects are influenced by the role division of those brands. To do this, an experiment was performed. Two different fictional co-branded products were developed, using the same four brands in which the new-product-brand-fit (NPBF) and the role division between the host and partner brands were manipulated. Brands of two product categories and two different brand image concepts were used to create the new co-branded products. Consequently, the experimental design was a between-subjects 2 (host brand-new-product fit: high or low) x 2 (ingredient brand-new-product-fit: high or low) x 2 (co-branded product: A or B) factorial design and a total of eight manipulations was created, as shown in the grids below. Product A Ingredient NPBF: high Category B, brand 3 Ingredient NPBF: low Category B, brand 4 Product B Ingredient NPBF: high Category A, brand 1 Ingredient NPBF: low Category A, brand 2 Host NPBF: high Category A, brand 1High + high High + low Host NPBF: high
Category B, brand 3
High + high High + low
Host NPBF: low Category A, brand 2
Low + high Low + low Host NPBF: low
Category B, brand 4
Low + high Low + low
For this study, an online questionnaire was conducted to generate numerical data. To create and conduct this questionnaire, the software survey program Qualtrics was used. 312 respondents were randomly assigned to one of eight manipulations, with varying degrees of NPBF for the brands involved. The survey first showed an introduction to the study followed by items about the two brands that were involved in the assigned manipulation, including items about brand attitude, prior brand usage and brand satisfaction. Then, the new product was introduced, but yet without brands. Respondents were asked to indicate their level of agreement about the NPBF-items for the two brands separately. Next, items for the control variables about the product were shown and participants had to answer the questions about their post-attitudes toward the two brands. Finally, questions were asked about demographics. The procedure was designed this
22 way because influential scholars in the field have used similar methods (Simonin & Ruth, 1998; Bouten, Snelders, & Hultink, 2011). Exact measures of the study can be found in the measures section (3.3) of this chapter.
3.2 Stimuli development
Two pre-tests served to identify suitable brands and co-branded products for this study. Consistent with other, important studies in the field (Simonin & Ruth, 1998; Desai & Keller, 2002; Baumgarth, 2004; Bouten, Snelders, & Hultink, 2011), existing brands were used as stimuli for this research to increase external validity (Desai & Keller, 2002). For people to evaluate whether a co-branded product fits a certain brand, they need to have genuine associations with that brand, which is more likely if it is real and they are familiar with it (Simonin & Ruth, 1998).A few requirements needed to be met in order for the stimuli to be suitable. For the fictional co-branded products, they were: they should be plausible, realistic and appealing. Also, the host and ingredient brand should both be perceived as making an important contribution, as co-branding is defined as two brands creating a new product together. The host brand should clearly be the main brand and the ingredient brand should unmistakably supply the ingredient. Furthermore, this research required two co-branded products of which the role division of the host and partner brand could be interchanged, while keeping the same set of brands. This way the possibility of any effects appearing that are due to the use of different brand names could be excluded or at least highly limited. For the brands that were tested it was required that they were familiar so that people would have genuine associations with the them. Also, it was necessary that the two brands in the same category were clearly of different brand concepts, as this would likely influence the perceived fit with the symbolic co-branded products. The procedures and analyses of both pre-tests can be found in Appendix 1. Based on the pre-test results, two products and four brands were considered suitable for the main study. Both products met the criteria of being plausible, realistic and appealing and the possibility of interchanging host and ingredient brand. The brands were all familiar and held a clear brand concept. The variation in NPBF was not fully provided by the pre-test, but with alterations based on the results it was considered sufficient to go on with the main study. The manipulations were formed with the product and brands as shown in the grids below.
23 Creamy Iced Coffee (CIC) Ingredient NPBF: high Ice cream, Magnum Ingredient NPBF: low Ice cream, Hertog Espresso Ice Cream (EIC) Ingredient NPBF: high Coffee, Nespresso Ingredient NPBF: low Coffee, AH Perla Host NPBF: high Coffee, Nespresso High + high Nespresso + Magnum High + low Nespresso + Hertog Host NPBF: high Ice cream, Magnum High + high Magnum + Nespresso High + low Magnum + AH Perla Host NPBF: low Coffee, AH Perla Low + high AH Perla + Magnum Low + low AH Perla + Hertog Host NPBF: low Ice cream, Hertog Low + high Hertog + Nespresso Low + low Hertog + AH Perla
3.3
Measures
During the main study the items were measured with statements and 7-point Likert scales. Table 1 contains the scales and their items used in the research. The measures are discussed in further detail below. Table 1 Main study constructs and their items Items References + answer options New-product-brand-fit (NPBF) 1. The new product fits with my idea and image of the brand [brand name]. 2. I think the brand [brand name] and the new product complement each other. 3. I think the brand [brand name] fits the product. 4. I think the new product adds to the brand [brand name]. 5. I think this is a very appropriate product for the brand [brand name]. (Bhat & Reddy, 2001) (Bouten, Snelders, & Hultink, 2011) 1 = Strongly disagree … 7 = Strongly agree Prior usage experience Have you used the brand [brand name] before? Yes / No Brand satisfaction 1. I am satisfied with my decision to buy the brand [brand name]. 2. Using the brand [brand name] has been a good experience. 3. The brand [brand name] has not worked out as well as I thought it would. * 4. I have truly enjoyed the brand [brand name]. (Lau & Lee, 1999) 1 = Strongly disagree … 7 = Strongly agree24 Parent brand attitude change (PBAC) 1. My attitude toward [brand name] would become ... 2. My disposition toward [brand name] and its products would become ... 3. My admiration toward [brand name] would become ... 4. My opinion regarding [brand name] as having a great reputation would become... (Dwivedi et al., 2010) 1 = strongly negative as compared with before … 7 = strongly positive as compared with before Note: * indicates that the item was reverse coded because it was negatively worded. To test convergent validity in this study, two confirmatory factor analyses were conducted with all main constructs: one for the host brand and one for the ingredient brand. The items for NPBF, brand satisfaction and PBAC were included in a principal axis factoring analysis. For both host and ingredient brands, the Kaiser-Meyer-Olkin measure verified the sampling adequacy, as KMO = .880 and .878, being bigger than the necessary .6. Bartlett’s test of sphericity was significant for both analyses (χ2(78) = 2661.59, p = .00 and χ2(78) = 3344.84, p = .00), indicating that it was appropriate to use the factor analytic model on the data. The analyses both showed 3 factors with Eigenvalues over Kaiser’s criterion of 1 and in total they explained 73,42% and 79.22% of variance. Also, the two scree plots levelled off after the third factor. For measure purification, items that load below .30 should be suppressed from the matrix (Nunnally, 1978), but none of the items were excluded in these cases. Table 2 shows the factor loadings in a pattern matrix for the host brand and Table 3 for the ingredient brand. The items that cluster on the same factors suggest that factor 1 represents NPBF, factor 2 brand satisfaction, and factor 3 PBAC. None of the items show cross-loadings, so it was confirmed that the scales indeed measured different and intended constructs. Table 2 Table 3 Factor loadings for host brand Factor loadings for ingredient brand
Item Factor loadings Item Factor loadings
NPBF BrSat PBAC NPBF BrSat PBAC
Brand fits product Appropriate product for brand Complement each other Product adds to brand Product fits with idea and image Good experience Truly enjoyed .961 .934 .914 .825 .822 .919 .847 Appropriate product for brand Brand fits product Complement each other Product fits with idea and image Product adds to brand Good experience Satisfied with decision to buy .941 .933 .916 .903 .883 .963 .928
25 Note: * indicates that the item was reverse coded because it was negatively worded. 3.3.1 New-product-brand-fit (NPBF) To assess the independent measure NPBF, four items that were previously used by Bouten et al. (2011) were adopted. One item from a similar scale was added (Bhat & Reddy, 2001), because of the limited variation in NPBF in the pre-test results. Also, because this fit is regarding brand image, it was considered appropriate to add an item to the scale that mentioned that image. The factor analysis confirmed this consideration. The original scale was reported by Bouten et al. (2011) to have a Cronbach’s alpha of .95 and thus being reliable. In their study, a confirmatory factor analysis confirmed validity as all factor loadings were significant (p < .05) and the item loadings were all above .50. An assessment of the expanded scale reliability in this study shows a Cronbach’s alpha of .96, which is much above the reliability threshold (.70)(Nunnally, 1978). Final scores for the variable were created by averaging the five items. 3.3.2 Brand satisfaction To assess respondents’ satisfaction with a brand if they had used it before, a scale created by Lau & Lee (1999) was adapted. Their scale consisted of seven items, which was reduced to four items for this study to limit respondent fatigue. This was considered appropriate, because many items in the original scale asked the exact same question but were reverse scaled. It was made sure that with the four questions in the used survey, all aspects of brand satisfaction from the original scale were included. That original scale was reported to be reliable with a Cronbach’s alpha of .95 and no problems with validity were revealed (Lau & Lee, 1999). The reduced scale in this study showed to be reliable as well with a Cronbach’s alpha of .84. Final scores for the variable were created by averaging the four items. The questions about brand satisfaction were not asked to respondents if they indicated that they never used the brand before. Satisfied with decision to buy Has not worked out as well* Admiration would become Attitude would become Great reputation would become Disposition would become .798 .498 .916 .905 .864 .776 Truly enjoyed Has not worked out as well* Admiration would become Great reputation would become Disposition would become Attitude would become .910 .495 .922 .900 .886 .858
26
3.3.3 Parent brand attitude change (PBAC)
The dependent variable PBAC was measured with a scale originally from (Dwivedi et al., 2010). The items measured the change in attitude toward the parent brands as a consequence of the co-branding alliance and they were anchored as 1 = “strongly negative as compared with before” and 7 = “strongly positive as compared with before” with a mid-point of 4 = “same as before”. This dependent variable was chosen to avoid noise that would arise in measuring pre- and post-attitudes toward the brands and then measuring the difference in levels (Dwivedi et al., 2010). An exploratory factor analysis ensured validity for this scale and reliability was confirmed with a Cronbach’s alpha of .77. For the current study a Cronbach’s alpha of .93 was reported, showing sufficient reliability. Final scores for the variable were created by averaging the items. 3.3.4 Manipulation checks and control variables The manipulation checks in the main study were very similar to the items used for the pre-test. In Appendix 3 an overview can be found of all manipulation checks and control variables in the main study. A change was made in the familiarity scale, extending it to the three items that were also used by (Bouten et al., 2011). This was done to be absolutely sure that people know the brands involved and that they have genuine associations and opinions for them. The scale was considered reliable with a Cronbach’s alpha of .90. Final scores for familiarity were created by averaging the three items. To control for any differences in familiarity that still might arise, the variable was inserted in the analyses when hypotheses testing, because (Simonin & Ruth, 1998) found that familiarity can have an effect on spillover.
Brand quality (Aaker & Keller, 1990) was assessed with a one-item scale, just as plausibility, realism and appeal of the co-branded products and overall liking of the categories coffee and ice cream. It is argued that for doubly concrete constructs for which both the object and attribute of measurement are clear and unambiguous (e.g. brand attitude) single-item measures can perform equally well as multiple-item measures (Drolet & Morrison, 2001; Bergkvist & Rossiter, 2007). It has even been suggested that multiple-item scales potentially cause lower validity (Drolet & Morrison, 2001). Brand quality would be controlled for if significant differences were shown by results of the brand manipulation checks, to see whether the changes in attitudes were actually influenced by the NPBF or that other components of brand equity played a role. The control variables regarding the co-branded products were added to be sure that PBAC host was not influenced by the products or categories of the products themselves. For that reason, a question was added to the main study that assessed whether respondents liked the parent brands
27
categories at all. Aaker & Keller (1990) argued that whether people like the category of constituent brands may have an effect on evaluation of a branding strategy.
3.4
Data collection and sample characteristics
Before distribution, five people tested the survey to check whether all questions were clear and how much time it took to complete it. As a result of this test, it was decided to not show the co-branded product combined with brands before asking whether it fit with the partner brands, but only the product with no brands. People would answer that question with the combination of brands in mind and did not just answer whether they thought the product itself would fit the brand, regardless of the co-branding alliance. Other than this change, no changes were made, and the survey was distributed among friends and family through WhatsApp and social media like Facebook, LinkedIn and Instagram. This convenience sampling method was combined with snowball sampling, as many respondents recruited new respondents by forwarding the survey link and sharing the social media message. People that participated in the pre-test were excluded from the main study. People who filled out the survey had the chance to win a €30 gift voucher of choice. The whole survey can be found in Appendix 4. Because the tested co-branding alliance was about food products, everyone who buys food products could fill in the survey. The questions controlled for people who don’t like one of the product categories coffee or ice cream. This meant everyone above the age of around ten could participate. As two of the brands used in the study are only sold in The Netherlands (AH Perla and Hertog), the survey also contained questions to find out whether respondents are Dutch or have at least lived in the Netherlands in the past year. Hogg, Tanis and Zimmerman (Bhat & Reddy, 2001) argue that a sample size larger than 25 is sufficiently large. As the experiment design consists of eight different manipulations, the total sample size should be at least 200 respondents. A total of 333 respondents filled out the survey completely. After screening the data with boxplot analyses, univariate outlier analysis and a case-by-case analysis of respondents who answered the same on every question, some responses were removed, so the final sample size was N = 312. Of all respondents, 24.4% was male, 75% was female and two preferred not to disclose their gender. Almost a half of all participants was in the age group of 15-24 (44.1%), a somewhat smaller group of 25-34 (40.9%),
28
4. Results
4.1 Randomization checks
Some preliminary analyses were conducted to check for biases in the allocation of respondents to different conditions. First, a Chi-square test of independence was conducted to check for significant differences in gender between all manipulations. The result showed that the number of males and females did not significantly differ between the conditions (χ2 (14) = 17.82, p = .22). A one-way ANOVA was used to check whether any significant differences in age were present among the manipulations. The result indicated this was not the case F(7, 304) = .1.16, p= .38. Based on the results of the two tests, it can be concluded that randomization was successfully achieved for all the manipulations.
4.2
Manipulation checks
Consistent with the findings of the pre-test, both co-branded products showed similar scores on all variables. Results from the manipulation checks of the two products are shown in Table 4. They are discussed below.
Table 4
Descriptives of co-branded product criteria
Plausibility Realism Appeal Important contribution Category
Host Ingredient Host Ingredient
CIC – M (SD) (N = 158) 5.35 (1.21) 5.52 (1.13) 4.85 (1.69) 5.32 (1.32) 4.89 (1.42) 5.15 (1.56) 3.45 (1.66) EIC – M (SD) (N = 154) 5.58 (1.34) 5.75 (1.22) 4.85 (1.71) 5.51 (1.16) 5.03 (1.47) 6.05 (1.37) 2.37 (.97) One sample t-tests showed that both the CIC and EIC were perceived plausible, realistic and appealing, as the mean scores were significantly above the midpoint of the scale (all p-values < .05). One sample t-tests also confirmed that the host and ingredient brands scored all above the midpoint of the scale for making an important contribution to the new products (all p-values < .05), considered as important enough. Both products were considered as belonging to the
29 category of the host brand. One-sample t-tests showed that the CIC scored significantly above the midpoint of the scale for being perceived as coffee and significantly below that midpoint for being perceived as ice cream (both p-values < .05). For the EIC scores it was the other way around (both p-values < .05). Analysis of the control variables of the brands used in the study showed similar results as in the pre-test, except for the NPBF. Results of the manipulation checks are shown in Table 5 and discussed below. Table 5 Descriptives of brand criteria
Familiarity Quality Attitude NPBF
CIC EIC Nespresso – M (SD) N = 162 6.43 (.82) 5.79 (1.15) 5.80 (1.14) 4.63 (1.30) 4.60 (1.40) AH Perla – M (SD) N = 161 4.99 (1.86) 4.33 (1.18) 4.39 (1.08) 3.65 (1.15) 3.55 (1.29) Magnum – M (SD) N = 156 6.60 (.56) 5.86 (1.00) 5.85 (1.16) 4.42 (1.30) 4.84 (1.44) Hertog – M (SD) N = 167 6.47 (.72) 5.59 (1.08) 5.57 (1.13) 3.59 (1.25) 3.62 (1.03) Familiarity scores were all significantly above the mid-point (4) of the scale (all p-values < .05). These results confirmed that the four brands derived from the pre-test were indeed familiar enough to use in the manipulations. All brands scored significantly above the mid-point of the scales for quality and attitude toward the brand as well (p-values < .05). Though, as the brand AH Perla scored significantly lower on these variables than the other three brands, both variables were included as controls during hypotheses testing to make sure any effects were not caused by differences in brand perceptions. Results of the pre-test did not show the expected variance in NPBF for the different brand-product combinations. Measuring with a more comprehensive scale and stressing the symbolic concept of the product more has improved the manipulations. The means of the NPBFs are all still around the mid-point of the scale, as can be seen in Table 5. In this main study the scores were all significantly above the mid-point for the symbolic brands and significantly below that point for the functional brands though, so they were considered variable enough. Based on all manipulation checks, the manipulations were successful.
30
4.3
Descriptives and correlations
In Table 6 an overview of the descriptive statistics, correlations and reliabilities of the scales is presented. When observing the correlations, it was found that some variables are interrelated. However, many of these relations showed correlation coefficients so small they can be considered as little if any correlation (r < .30) (Hinkle, Wiersma, & Jurs, 2003). Just one correlation was bigger. The liking of the host brand category correlated significantly with satisfaction with the host brand (r = .327). Yet, these correlation coefficients are still not big (r < .50) and considered as low positive correlation (Hinkle et al., 2003). Table 6 Means (M), standard deviations (SD) and correlations M SD 1 2 3 4 5 6 7 8 1. NPBF host 4.19 1.35 (.96) 2. NPBF ingredient 4.00 1.38 .269 (.96) 3. Brand satisfaction host a 5.53 .95 .063 .078 (.84) 4. PBAC host 4.39 .84 .196 .209 .018 .257 5. Familiarity host 6.22 1.11 .115* .020 .203 -.056 (.93) 6. Familiarity ingredient 6.05 1.40 -.109 .088 -.044 .067 -.004 (.90) 7. Liking host category 5.92 1.43 .037 .034 .327 .162 .198 -.032 (-) 8. Liking ingredient category 5.50 1.80 -.033 .015 -.099 .090 -.083 .174 -.101 (-) Notes. N = 320. a N = 245. Significant correlations (p < .01) are presented in bold, except for one correlation indicated with * which is significant at .05 level.
4.4
Normality
The independent and dependent variables of the study have been tested for a normal distribution of the data. A Shapiro-Wilk test was conducted which resulted in significantly non-normal distributions for all variables (all p-values <.05). However, this test is very powerful and for large samples of 200 or more the assumption of normality is often rejected (Field, 2013). For that reason, scores for skewness and kurtosis were calculated. The values for skewness and kurtosis for the different variables were as following. For NPBF host the skewness was -.055 (SE = .138) and kurtosis -.789 (SE = .275). For NPBF ingredient brand the skewness was -.080 (SE = .138) and kurtosis -.645 (SE = .275). For satisfaction with the host brand the skewness was -1.271 (SE = .156) and kurtosis 2.176 (SE = .310). For PBAC of the host brand the skewness was .855 (SE = .138) and kurtosis 1.172 (SE = .275).
