• No results found

A cross-cultural study on the consumer perceived credibility of an online recipe

N/A
N/A
Protected

Academic year: 2021

Share "A cross-cultural study on the consumer perceived credibility of an online recipe"

Copied!
69
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

0 | P a g e

A cross-cultural study on

the consumer perceived

credibility of an online

recipe

Master Thesis – IBM

Rijksuniversiteit Groningen

Student: Arnte Willem van Dam

Program: MSc International Business & Management Student nr: 2040131

Date: 18-2-13

(2)

1 | P a g e

Table of content

1 INTRODUCTION 3 2 LITERATURE REVIEW 6 2.1 DEFINING CREDIBILITY 6 2.2 CONSTRUCT OF CREDIBILITY 7 2.2.1 SOURCE CHARACTERISTICS 7 2.2.2 MESSAGE CHARACTERISTICS 10 2.2.3 RECEIVERS CHARACTERISTICS 12 3 THEORETICAL FRAMEWORK 16

3.1 ELABORATION LIKELIHOOD MODEL 17

3.2 DUAL-PROCESS THEORY &HOVLAND-YALE MODEL 18

3.2.1 CULTURE 19

3.3 CONCEPTUAL FRAMEWORK 20

4 HYPOTHESES 20

4.1 EFFECTS ON PERCEIVED MESSAGE CREDIBILITY OF AN ONLINE RECIPE 20

4.1.1 EFFECT OF DIFFERENT TYPES OF RECIPE SOURCES 20

4.1.2 EFFECT OF RECIPE MESSAGE RATING ON PERCEIVED MESSAGE CREDIBILITY OF RECIPE 23

4.2 MODERATING EFFECT OF CULTURE 24

4.2.1 SOURCE-ENDORSEMENT 24

4.2.2 RECIPE RATING 26

4.3 EFFECT OF PERCEIVED MESSAGE CREDIBILITY ON LIKELIHOOD OF RECIPE ADOPTION 27

5 METHODS 27 5.1 DATA 27 5.2 SURVEY DESIGN 28 5.3 MEASUREMENTS 30 5.3.1 INDEPENDENT VARIABLES 32 5.3.2 MODERATING VARIABLE 33 5.3.3 DEPENDENT VARIABLE 33 5.4 CONTROL VARIABLES 34 5.5 RELIABILITY OF MEASUREMENT 35

(3)

2 | P a g e

5.5.2 LIKELIHOOD OF RECIPE ADOPTION 36

5.5.3 BRAND-EQUITY SCALE 37

6 RESULTS 38

6.1 CONTROL VARIABLES 39

6.2 EFFECT OF SOURCE AND RATING 41

6.3 INFERENCE ABOUT CULTURAL DIFFERENCE 44

6.4 INFERENCE ABOUT LIKELIHOOD OF RECIPE ADOPTION 45

7 THEORETICAL IMPLICATIONS 46 8 MANAGERIAL IMPLICATIONS 48 9 LIMITATIONS 49 10 FURTHER RESEARCH 50 11 CONCLUSION 51 12 BIBLIOGRAFIE 51 13 APPENDICES 62

13.1 APPENDIX 1–DUTCH RECIPE DESIGNS 62

(4)

3 | P a g e

1 INTRODUCTION

The evaluation of information credibility, also defined as believability, on online consumer discussion forums has increasingly been studied in the academic marketing and communication literature (Cheung, Lue, Sia, & Chen, 2009; Doh & Hwang, 2009; Metzger, Flanagin, & Medders, 2010). A recent phenomenon is the popularity of online consumer recipe forums (Park, Lee, & Han, 2007). Such online consumer forums have over thousands of active members whom interactively and instantaneously exchange their cooking and baking knowledge, recipes, opinions and recommendation with other consumers (Brown, Broderick, & Lee, 2007). For instance, a popular Dutch recipe forum, Smulweb.nl, has 2.400.000 visitors per month seeking for recipes, or for example the Polish Znajdzprzepisy.pl with an average of 1.500.00 visitors per month. As an increasing number of consumers consult online recipe forums to decide what will be on the dinner table, the process by which the consumer evaluates the credibility of an online recipe becomes interesting for marketers and practitioners. Particularity interesting because online recipe forums provide great potential for local and multinational enterprises (MNE’s) to influence the buying-behaviour of consumers through online recipes. In addition, comprehending the implication of cultural differences in the credibility evaluation of an online recipe may steer future international marketing strategy on online discussion forums.

(5)

4 | P a g e

judgement of an online message (Hovland, Janis, & Kelley, 1953; Chueng & Thadani, 2010). The source, the correspondent of who originated the message (Hovland & Weiss, 1951), was found to be one of the most crucial determinants on online consumer product forums (Cheung, Lue, Sia, & Chen, 2009; Dou, Walden, Lee, & Lee, 2012). Message factors identified to significantly influence the credibility were for instance, aggregated ratings (e.g. product recommendation rating) and argument strength (Flanagin A. , Metzger, Pure, & Markov, 2011). Concerning receivers’ characteristics, mainly involvement and prior knowledge of the recipient were investigated and found to significantly influence the credibility evaluation of online information.

(6)

5 | P a g e

respondents from Poland and from the Netherlands. The choice for these two cultures follows from the fundamental cultural differences according to the dimensions of Hofstede (Hofstede & Hofstede, 2005). Lastly and for completeness sake, this study will address the relationship between the message credibility of an online recipe and the recipient’s likelihood of adopting that recipe, as such contributing to studies on the role of credibility in the persuasiveness of online information.

By applying the theoretical lens of Deutsch and Gerrard’s dual-process theory (Deutsche & Gerrard, 1955) in combination with the Yale-Hovland model (Hovland, Janis, & Kelley, 1953) and guided by the elaboration likelihood model (Petty & Cacioppo, 1986), this study explores how a different source of an online recipe and a recipe’s rating affect the perceived message credibility and likelihood of recipe adoption. The Deutsche and Gerrards dual-process theory is used as a theoretical framework to determine the influences of normative (recipe rating) and informational (source) variables on the credibility evaluation. Furthermore, the Hovland-Yale model is integrated into the dual-process model of Deutsche and Gerrards as a refinement to determine if cultural differences (Poland vs. the Netherlands) affect the strengths of the latter mentioned effects on the perceived message credibility of an online recipe. Moreover, the ELM provided guidance to comprehend the persuasion process of an online recipe. Hence, the research questions are formulated as followed:

1. How do source and recipe rating affect consumer online recipe credibility evaluation, and does this differ between Dutch and Polish consumers?

2. How does perceived message credibility affect the likelihood of adoption an online recipe?

(7)

6 | P a g e

of the online recipe. Finally, this study confirmed that message credibility of an online recipe is an important determinant in the adoption of the recipe.

The structure of this research paper is as follows. In the subsequent section, a brief literature synopsis on the main factors that have found to influence the credibility evaluation in the online context will be shared. Next, the theoretical framework will include the factors of focus. Subsequently, based on the discussed literature and the theories elaborated on in the theoretical framework, five hypotheses are presented. Hereafter, the methodology section will provide the steps taken to test hypothesis. Then, the results and its implications will be discussed. Lastly, this study will elaborate on its limitations, future research directions and managerial implications.

2 LITERATURE REVIEW

This literature review will consider past research touching upon how credibility of an online recipe may be influenced by recipe source, rating and culture. It aims to offer insights into the theoretical frameworks and perspectives on how users judge the credibility of an online message. First the term ‘credibility’ is defined according to the focus of this paper. Subsequently a literature synopsis on factors influencing credibility in the online context (the Internet and Web) is provided, hence elaborating on past and present knowledge on the information credibility assessment in the online context.

2.1 Defining credibility

(8)

7 | P a g e

It is important to comprehend that credibility is a subjective perception of believability or quality by the information recipient (Fogg, 2003; Metzger, Flanagin, & Medders, 2010). As it is well “understood that the credibility of a source or message is a recipient-based judgment” (Metzger, 2007, p. 2078). Hence, when discussing credibility it is emphasized from a recipient’s perception rather than an objective perspective.

2.2 Construct of credibility

So what marks or influences the credibility of online information? This question has been investigated within multiple academic disciplines, including communication, information science, and psychology and marketing science (Pornpitakpan, 2004). Since the earliest discussion on credibility, “message credibility is generally agreed to result from an interaction of source characteristics (e.g., expertise, trustworthiness), message characteristics (related to message content, encompassing factors such as plausibility, internal consistency, and quality), and receivers’ characteristics (e.g., cultural background, previous beliefs)” (Walthen & Burkell, 2002, p. 135). Source characteristics are defined as perceived characteristics of the communicator who transmits the message (Chueng & Thadani, 2010). In addition, message characteristics refer to the message features pertaining to the characteristics of a message (Hung T. , 2006), while receivers’ characteristics concern characteristics of the individual who receives the message (Hovland, Janis, & Kelley, 1953).

This study examines the source, message and receivers’ characteristics (namely, culture) on the credibility assessment of an online recipe. Hence, the following subchapters are devoted to a literature synopsis on the source, message and receivers characteristics that found to affect the credibility evaluation of a message.

2.2.1 Source characteristics

(9)

8 | P a g e

information, and therefore impacts the recipient’s credibility judgement of a message and its acceptance thereof (Hovland, Janis, & Kelley, 1953; Ohanian, 1990; Pornpitakpan, 2004).

Most scholars agree that, similar to the offline world, critical source factors affecting the credibility of the message are the perceived expertise, trustworthiness and attractiveness of the message source (Hovland, Janis, & Kelley, 1953; Ohanian, 1990; O'Keefe , 2002; Rieh & Danielson, 2007; Pornpitakpan, 2004). Table 1 summarizes the source factors found to associate the credibility judgement of an online message in relation to online consumer forums. However, online forums make it hard to judge certain source factors such as physical appearance, due to the nature of the virtual discussion, which may not always permit to convey such cues (Cheung, Lue, Sia, & Chen, 2009).

Constructs Definitions Authors

Source credibility

- Expertise - Trustworthiness

Expertise refers to an “intrinsic

characteristic ascribed to the message sender, who is supposed to have a profound knowledge of a specific theme” (Guido, Prete, & Sammarco, 2010, p. 4).

Trustworthiness concerns the

morality or goodness of the source and is often defined with terms such as truthful, unbiased and well intentioned (Hovland, Janis, & Kelley, 1953)

(Cheung, Lue, Sia, & Chen, 2009; Dou, Walden, Lee, & Lee, 2012)

Attractiveness

- Similarity - Familiarity - Likability

Similarity refers to the resemblance

between the information source and the recipient (Bickart & Schindler, 2001)

Familiarity concerns “the

knowledge of the source through exposure or past association whereby a level of comfort with the source is established for the

receiver” (Kiecker & Cowles, 2001, p.

(10)

9 | P a g e

351)

Likability is defined as the

recipient’s affection for the information source “as a result of physical appearance, behaviour or other personal traits” (Chueng & Thadani, 2010, p. 334)

Disclosure of identity Disclosure of the source identity to others.

(Forman, Ghose, & Wiesenfeld, 2008)

Table 1 Source factors associated with the credibility of an online message

Various scholars advocate that especially Internet users tend to rely more on the source than on the merits of the message to judge the credibility of an online message (Doh & Hwang, 2009; Dou, Walden, Lee, & Lee, 2012; Hilligoss & Reih, 2008; Metzger, Flanagin, & Medders, 2010). This is argued to be a result from Internet users being bombarded with online information from various sources; e.g. online forums where various authors create and exchange information. To save mental effort and avoid confusion, Internet users need to evaluate information swiftly, thereby focusing more on environmental cues of the message (e.g. the source). The study of Chueng et al. (2009), that investigated the message credibility of product recommendations on an online consumer product forum, provides an empirical example. Their research demonstrated that the message source (e.g. the author of the product recommendation) was the most influential factor affecting the credibility of interpersonal exchange and messages created on online consumer forums. Similar findings were found by Doh & Hwang (2009) that examined the effect of the source on the credibility of a video review.

(11)

10 | P a g e

hand, argue that Internet users find the website and the particular information on the website to be distinct. Hence, this should be investigated separately “because credibility assessments may vary depending upon which source attributes are salient at the time of evaluation” (Flanagin & Metzger, 2007, p. 322). In line with earlier empirical studies of Doh & Hwang (2009) and Chueng et al (2009) on product recommendations, this study views the source not as being an attribute to the authorship of the website, but an attribute to the particular source of the information on the particular website.

In this area research is scarce and has solely explored consumer product or video review forums, leaving other popular types of online forums unexplored (Cheung, Lue, Sia, & Chen, 2009; Doh & Hwang, 2009). Moreover, these studies have exclusively focused on consumer-endorsed messages, also referred to as electronic Word-of-Mouth (eWOM). Nevertheless, typical other message sources such as advertisers or anonymous sources are also prevalent on online forums. For instance, a submitted message on an online forum may be branded, anonymous or even be submitted by stealth marketers that pose as a consumer (Guido, Prete, & Sammarco, 2010). Comprehending how these distinct types of sources (consumer, branded or anonymous) influence the credibility of a message is of vital importance in the online interpersonal setting, as it can aid marketing strategists on the new social media and contributes to the discussion on the concept of the online source.

2.2.2 Message characteristics

(12)

11 | P a g e

message presented on a website, such as the message framing or the strength of the message arguments. Indeed, in any given online message, numerous message features may influence the credibility judgement. Often these message features are categorized to message structure (i.e. message organization), message content (i.e. argument quality), or message delivery (i.e. assertive language) (Metzger M. , Flanagin, Eyal, Lemus, & McCann, 2003).

The area of interest in this study is the ‘message content’ feature of an online message. Certain message content features found to influence the judgement of credibility are the use of relevant evidence (i.e. statistic, quotes or testimonials), argument strength, message framing, message or argument quality, and product review ratings. The current study focuses on the effects of aggregate ratings, namely aggregated recipe rating which construct is similar to product review ratings. This contribution is valuable as aggregated ratings of messages are thus far solely explored on online consumer discussion forums, leaving other type of popular forums unexplored (e.g. recipe forums). Table 2 summarizes the factors that have found to impact credibility of an online message.

Constructs Definitions Authors

Message/argument quality

- Accuracy

- Comprehensiveness - Currency

Argument quality “refers to the

persuasive strength of arguments embedded in an informational message” (Cheung, Lee, & Rabjohn, 2008, p. 232)

Accuracy refers to the reliability of

the message or arguments. This also includes the perceived correctness of the information (Reih & Belkin, 1998)

Comprehensiveness of the message

concerns its completeness. (Chueng & Thadani, 2010)

Currency refers to whether the

messages are up-to-date and timely. (Reih & Belkin, 1998)

(13)

12 | P a g e Argument strength

“The extent to which the message receiver views the argument as convincing or valid in supporting its position” (Cheung, Lue, Sia, & Chen, 2009, p. 15)

(Cheung, Lue, Sia, & Chen, 2009)

Use of evidence

The extent to which statistic, quotes or testimonials support the message.

(O'Keefe , 2002; Hung T. , 2006)

Product review rating

Overall rating in stars (i.e. 1 = not recommended, to 5 = highly recommended) given by other readers, due to peer-to-peer evaluation systems that allows the reader to review a message by providing a star from (1 lowest to 5 highest)

(Cheung, Lue, Sia, & Chen, 2009; Flanagin A. , Metzger, Pure, & Markov, 2011)

Message framing

Whether the valance of a message is positive or negative (Doh & Hwang, 2009)

(Cheung, Lue, Sia, & Chen, 2009; Doh & Hwang, 2009)

Table 2 Message content factors found to affect the credibility of an online message

2.2.3 Receivers characteristics

Literature often defines the receiver as the recipient who receives the information (Janis & Hovland, An overview of persuasibility research, 1959). The actual effect of the information received by the recipient may vary from individual to individual. As Chaiken (1980) advocates, identical information can engender a variety of responses in different recipients. This has encouraged scholars to gain interest in the influence of receivers’ characteristics in credibility evaluation of information and the actual information adoption.

(14)

13 | P a g e

may influence the credibility evaluation of the message are summarized in table 3. These receivers’ characteristics are often used as control (Bickart & Schindler, 2001; Briggs, Burford, Angeli, & Lynch, 2002; Eysenback & Kohler, 2002; Flanagin & Metzger, 2007; Hung T. , 2006) or moderating variables in credibility studies (Cheung, Lue, Sia, & Chen, 2009; Dou, Walden, Lee, & Lee, 2012).

Constructs Definitions Authors

Prior Knowledge

Prior knowledge that the recipient has on the specific topic discussed (Dou, Walden, Lee, & Lee, 2012)

(Bickart & Schindler, 2001; Briggs, Burford, Angeli, & Lynch, 2002; Cheung, Lue, Sia, & Chen, 2009; Doh & Hwang, 2009; Dou, Walden, Lee, & Lee, 2012; Eysenback & Kohler, 2002; Pornpitakpan, 2004)

Involvement

“The degree of psychological identification and affective, emotional ties” (Chueng & Thadani, 2010, p. 337)

(Bickart & Schindler, 2001; Briggs, Burford, Angeli, & Lynch, 2002; Cheung, Lue, Sia, & Chen, 2009; Doh & Hwang, 2009; Dou, Walden, Lee, & Lee, 2012; Eysenback & Kohler, 2002)

Gender

The gender of the recipient (male or female)

(Flanagin & Metzger, 2007; Pornpitakpan, 2004)

Age

The duration of the recipients life in years

(Flanagin & Metzger, 2007; Pornpitakpan, 2004)

Table 3 Receivers factors found to affect the credibility of an online message

2.2.3.1 Culture

When considering the impact of receivers’ characteristics on the credibility evaluation of a message, Fogg (2003) suggests that culture may also play a crucial role. In addition, Liu & Huang (2005, p. 109) noted that, “credibility perception is deeply embedded in and heavily influenced by social and cultural contexts”. However, a credibility assessment of online messages by Internet users in different cultures remains largely unexplored.

(15)

14 | P a g e

people do assess credibility different that Americans. Hence, Fogg (2003) included cultural background into its Prominence-Interpretation Theory that explains how people assess credibility of web-based information. In addition, Liu & Huang (2005) did a cross-cultural study on how Chinese and American students evaluate the credibility of scholarly information present on the web. Their results indicated Chinese and American students show similarities and differences in their credibility evaluation. As such, Liu & Huang (2005) concluded that culture may affect the credibility evaluation of online information. Lastly, based on the elaboration likelihood model by Petty & Cacioppo (1986), Pornpitakpan & Francis (2008) compared Thais and Canadians in terms of source expertise and argument strength. Their findings indicated that culture alters the effect of both the source and the argument strength in persuasion. Hence, concluding that culture may affect the evaluation and persuasiveness of information.

Neither of the aforementioned studies has specifically touched upon the role of culture differences in the evaluation of different sources (i.e. brand, consumer and anonymous) nor on the aggregated rating (i.e. low and high recipe rating) when judging the credibility of an online message. In addition, all prior studies have been particularly keen on comparing individualistic versus collectivistic cultures, ignoring to compare countries with equal individualistic cultures or collectivistic cultures. Such comparison seems a vital theoretical contribution to comprehend which specific dimensions of a culture may influence the credibility evaluation of information. Furthermore, comprehending the effect of cultural differences on credibility judgement of a message is of vital importance to international/global advertising and marketing research and practitioners. As noted by Pornpitakpan & Francis (2008), cultural differences often dictate the MNE’s decision to “standardize international advertising and the degree and kind of adaptation to the advertising message that is needed” (Pornpitakpan & Francis, 2008, p. 80).

(16)

15 | P a g e

In brief, the Hofstede model (Hofstede, 1986; Hofstede & Hofstede, 2005) identified five dimensions on which cultures differ: Power Distance (PDI), Individualism vs. Collectivism (IDV), Masculinity vs. Femininity (MAS), Uncertainty Avoidance (UAI), and Long-term vs. Short-term orientation (LTO). This study will only focus on power distance, uncertainty avoidance and masculinity. The next paragraph elaborates on this choice and the dimensions specified to the Netherlands and Poland. Building on findings of 22 countries, the model offers a scale for various countries in which each country has a position comparative to other countries. In figure 1 the Hofstede’s dimensions are depicted concerning the countries investigated in the current study: Poland and the Netherlands.

Figure 1 Comparison of Dutch and Polish culture according to Hofstede & Hofstede (2005)

(17)

16 | P a g e

consumers in Poland are more concerned with the security of life as oppose to taking risk. The last dimension is masculinity (MAS), which represents a preference in a culture for the stress on achievement or nurture (Hofstede & Hofstede, 2005). For instance, in high MAS countries, like Poland, social gender roles are perceived to be clearly distinct and society stresses the focuses on material success while failing is a disaster. In low MAS countries, like the Netherlands, social gender roles overlap and society is more concerned with the quality of life in which failing is just a minor accident.

Overall, the difference between the Netherlands and Poland in the degree to which diverse members of society wield influence and command respect (PDI), how the reader copes with uncertainty (UAI) and the difference in stress on achievement or nurture (MAS), makes Poland and the Netherlands suitable candidates to investigate the moderating role in the credibility assessment of an online message.

Second, comparing two countries with low IDV is an interesting contribution as past cross-culture credibility studies have exclusively focused on countries with high difference on IDV. For instance, comparing America and China (Liu & Huang, 2005) or Canada and Thailand (Pornpitakpan & Francis, 2008). Utilizing respondents from Poland and the Netherlands countries may allow more specific empirical testing of the moderating effect on credibility assessment of the other cultural dimensions identified by Hofstede (1986).

Lastly, marketing practitioners often view Poland as a suitable representation of the Eastern-European culture, besides being the fastest growing economy in Europe, and the Netherlands as a well representative for the Western-European culture (Unilever, 2012).

3 THEORETICAL FRAMEWORK

(18)

17 | P a g e

factors (Deutsche & Gerrard, 1955) and the Hovland-Yale model (Hovland, Janis, & Kelley, 1953). These theories will be elaborated on next.

3.1 Elaboration Likelihood Model

To comprehend to what extent perceived credibility changes attitude or behaviour, the process of persuasion should be investigated. A dual-process theory widely applied in marketing literature to investigate how attitudes are formed or altered, is the Elaboration Likelihood Model (ELM) by Petty and Cacioppo (1986) (Cheung, Lue, Sia, & Chen, 2009; Sussman & Siegal, 2003). The ELM model posits that there is a central and a peripheral route via which persuasiveness of a message is processed. The choice for either of the two routes is dependent on the recipients’ motivation and ability to process the information.

The central route entails careful scrutiny and cognitive evaluation of the message by the recipient to determine the merits of the message, hence requiring considerable effort of the recipient. For instance, by evaluating the message by carefully examining the message arguments. In this central route, the recipients’ cognitive responses to the merits of the message, determine its persuasiveness. If approving thoughts are the outcome of the elaboration-process, the message will likely be accepted. If disapproving thoughts are formed, it is likely that the message will be rejected. In order for a recipient to process a message via the central route, it is required that the recipient has the ability to do so and the recipient attributes high relevance to the message (i.e. is involved).

In the peripheral route, on the other hand, attitudes are shaped by less cognitively demanding means. Rather, the recipient relies on environmental cues of the message, also referred to as peripheral cues (i.e. source or quality of presentation). Hence, this process uses a mental shortcut to decide whether to accept or reject the message and outside stimuli considerably affect the decision-making process. In addition, this route is taken when the recipient is unable to process the merits of a message due to lack of motivation or ability to do so. Depending on the degree of the recipients’ motivation and ability, either of the two routes to persuasion might dominate.

(19)

18 | P a g e

cognitive efforts required to cope with the abundant amount of information that is likely to be presented to them online (Flanaging & Metzger, 2008; Fogg, Soohoo, Danielson, Marable, Stanford, & Tauber, 2002; Metzger, Flanagin, & Medders, 2010; Walthen & Burkell, 2002). This may especially be applicable to online forums were readers are confronted with a nearly boundless amount of information. Important peripheral cues identified in the online environment are the source and social aggregated feedback systems available (i.e. aggregated recipe rating) on present online forums (Metzger, Flanagin, & Medders, 2010; Petty & Cacioppo, 1986; Flanagin A. , Metzger, Pure, & Markov, 2011). As the current study focuses on credibility evaluation of online messages (recipes), these prominent peripheral cues of an online message are included: source and social aggregated rating systems. This study justifies the use of these variables by indication that Internet users tend to utilize the peripheral route as indicated by prior studies that adopted the ELM. In fact, the ELM will also serve as groundwork to enable to make predictions about the relative impact of effects of the perceived message credibility on the likelihood of adopting an online recipe.

3.2 Dual-process theory & Hovland-Yale model

In order to understand how the source and aggregated recipe rating as well as culture, may influence the perceived message credibility of an online recipe, this study will combine the dual-process theory of normative and informational influences (Deutsche & Gerrard, 1955) and Hovland-Yale model (Hovland, Janis, & Kelley, 1953).

(20)

19 | P a g e

in today’s online forums as a normative factor. The theory posits also that both normative and informational factors work together to shape the recipient’s credibility judgement (Cheung, Lue, Sia, & Chen, 2009; Deutsche & Gerrard, 1955).

The Hovland-Yale model (Hovland, Janis, & Kelley, 1953) is according to scholars a refinement on the informational influences as categorized by Deutsche en Gerrards (1955) (Cheung, Lue, Sia, & Chen, 2009). This model suggests there are three key informational components affecting the message credibility evaluation that can ultimately influence the recipients’ perceptions, opinions and actions (Hovland, Janis, & Kelley, 1953). As elaborated in the literature review, these three key informational factors are source-, message- and receiver characteristics. The integrating of the Hovland-Yale model into the information influences of the dual-process model provides a theoretical foundation to explore both the influences of the recipe source (informational influence) and the recipe rating (normative influence) on the credibility judgement of an online recipe. In addition, the Hovland-Yale model suggests that receiver characteristics, such as culture may influence the evaluation of the message. This will be further elaborated on next.

3.2.1 Culture

(21)

20 | P a g e

support this conceptual relation is the study of Bond and Smith (1996) that examined 134 studies that utilized the Asch paradigm of conformity in a meta-analysis. Hence, it is argued that culture may have the tendency to alter the effective strength of both informational and normative factors in shaping the credibility judgement.

3.3 Conceptual framework

Based on the previous section, it is summarised that the different type of recipe sources investigated in this study are perceived as informational factors that may affect the perceived message credibility. Rating of a recipe, on the other hand, is depicted as a normative factor that may also influence the perceived message credibility. In addition, as discussed in previous culture section, culture is argued to play a moderating role between the effects on perceived message credibility of the recipes. Finally, based on the Dual-process model of normative and informational influences and the ELM, it is argued that perceived message credibility will affect the likelihood of adopting an online recipe. Hence, the proposed theoretical framework is depicted as follows, see figure 2.

Figure 2 Theoretical framework

4 HYPOTHESES

4.1 Effects on perceived message credibility of an online recipe 4.1.1 Effect of different types of recipe sources

(22)

21 | P a g e

1955). As identified by the Hovland-Yale model, the source is one of the key informational components in the message evaluation and can alter the credibility of the message. As pointed out by Hovland & Weis, messages with positive source attributes are assumed more credible than message with less positive source attributes. In addition, the source cue tends to have the strongest impact on the message evaluation when assessed via the peripheral route (Petty & Cacioppo, 1986). Argued by scholars, especially in the online environment, Internet users tend to follow the peripheral route when evaluating the message evaluation (Hilligoss & Reih, 2008; Metzger, Flanagin, & Medders, 2010). Hence, it is argued that the source of an online recipe plays an important role in credibility assessment of an online message.

On online recipe forums, unless the recipe source is anonymous (undisclosed), the source name, credentials and or graphical representation (i.e. avatar picture) are disclosed with the recipe message. This enables the recipient to see who has submitted the recipe on the online recipe forum. This is particularly interesting as anonymous sources, as well as fellow consumers, which are often total strangers to the recipient and even advertisers, can reach and exchange recipes and baking and cooking knowledge with a “vast and geographically dispersed group of strangers“ (Chueng & Thadani, 2010, p. 334). Hence, the effect of anonymous-, consumer- or brand-endorsed messages will be further elaborated on.

(23)

22 | P a g e

finds support in the statement by Buhalis and Law (2008, p. 612): “[..] consumers increasingly trust their peers, rather than marketing messages”.

Another source attribute that may influence the credibility of the recipe is the perceived expertise of the source. It can be challenging for the recipient to evaluate to what degree a source is an “expert” on a specific topic. For a brand-endorsed recipe, the authority behind the recipe is known (i.e. Blue Band or Kasia in this study) which can be verified on its qualifications. In addition, these recipes often include official credentials such as “master chef” that expresses occupational qualifications. Consumers tend to value a source more credible with official credentials than in absence of official credentials (Rieh, 2002). Consumer-endorsed recipes often solely include the name of the author. In the virtual environment, the evaluation of the individuals’ knowledge and background is very limited. Hence, when relying on expertise to judge the credible of a recipe, it is argued that a consumer will likely perceive brand-endorsed message to have greater knowledge about cooking and baking than a consumer-endorsed recipe. Noted however, the perceived brand equity may positively or negatively influence the credibility perception of the endorsed recipe (La Ferle & Choi, 2005).

In summary, while consumers will likely perceive a consumer-endorsed recipe superior in trustworthiness, less prone to bias and more similar in attractiveness, they on the other hand will probably perceive a brand-endorsed recipe to have greater knowledge on the topic, thus more expertise. Based on Hovland and Weis (1951), as well as derivable from the ELM model (Eagly & Chaiken, 1993; Petty & Cacioppo, 1986), a message with more positive source attributes is assumed to be more credible than message with less positive source attributes. Hence, it is postulated that:

Hypothesis 1a: If the recipe is consumer-endorsed, then the consumers will

perceive this recipe more credible than a brand-endorsed recipe.

(24)

23 | P a g e Hypothesis 1b: If the recipe is brand-endorsed, then the consumers will perceive

this recipe more credible than an anonymous-endorsed recipe.

4.1.2 Effect of recipe message rating on perceived message credibility of recipe

Derived from the dual-process theory by Deutsche and Gerrards (1955), it is argued that aggregated recipe rating as normative influence may exert a salient and vital effect on the credibility judgement of the message. According to scholars that adopted the ELM, Internet users frequently trend to rely on social aggregate feedback systems, like social aggregated ratings, when evaluating the online credibility of a message (Metzger, Flanagin, & Medders, 2010). Aggregated recipe rating present on an online recipe message indicates the overall ratings by other readers on a specific recipe message. Thereby allowing the reader to evaluate the message; that is, providing the information with a low or a high rating according to how they perceive the information (1= not recommended to 5= highly recommended). Since aggregated recipe rating provides a position favoured by others community members during the exchange of information, it is argued to be a normative influence (Kaplan & Miller, 1987).

If the aggregate rating of a recipe is high, this may imply that the majority of the previous readers perceived the overall recipe to be credible or believable and therefore highly recommend it. In contrast, when most readers provide a very low rating to the message, this may indicate that the majority of the consumers perceived the overall recipe to be unreliable or unbelievable and would not recommend it. This could provoke suspicion with the recipient who would consequently doubt the credibility of the message.

(25)

24 | P a g e

Flanagin A. , Metzger, Pure, & Markov, 2011). Yet, previous online studies have exclusively focused on online consumer product forums that include aggregated product ratings. The question remains whether these results are also to be found in other types of online community forums, like online recipe forums. Hence, the following hypothesis is developed:

Hypothesis 2: If the recipe has a high rating, then consumers will perceive this

recipe more credible than when a recipe with a low rating.

4.2 Moderating effect of culture 4.2.1 Source-endorsement

(26)

25 | P a g e

The masculinity dimension will not be elaborated on although the countries highly differ on this dimension. This because the masculinity dimension is argued to be less applicable, since the credibility evaluation of message source is argued to be less related to which a society stress nurture or achievement (Hofstede & Hofstede, 2005). The final dimension of importance is power distance (PD). Power distance is a characteristic of a culture that to a certain extent regulates which diverse members in a society command respect and yield influence. In large power distance cultures, such as Poland “individuals in positions of influence hold the power over what is right and good” (Pornpitakpan & Francis, 2008, p. 80). Polish consumers therefore tend to find it important that one’s social status is clear so that others can show appropriate respect, in contrast to Dutch consumers being from a low PD culture. As pointed out by Dawar, Parker, & Price (1996), high PD affects opinion leadership behaviour that causes Polish consumers to distrust fellow consumers. In contrast, low PD cultures such as the Netherlands tend to minimize inequality.

Based on the above discussion and derived from the Hovland-Yale model, it is argued that culture as a receivers’ characteristic may affect the strength of the relationship between the source and the credibility judgement of the message. This finds support in the study of Pornpitakpan & Francis (2008) that adopted the ELM. They found that culture plays a moderating role between the source and the persuasiveness of the message. As credibility is a key stage in the persuasion process, this study argues also that cultural has tendency to moderate the effect of the source on the message evaluation process (Petty & Cacioppo, 1986). Hence, the following is postulated:

Hypothesis 3a: If the recipe is brand-endorsed, then Polish consumers will perceive

this recipe more credible than Dutch consumers.

Hypothesis 3b: If the recipe is consumer-endorsed, then Dutch consumers will

perceive this recipe more credible than Polish consumers.

(27)

26 | P a g e Hypothesis 3c: If the recipe is anonymous-endorsed, then Polish consumers will

perceive this recipe less credible than Dutch consumers.

4.2.2 Recipe rating

Based on the dual-process theory of Deutsche and Gerrard’s (1955) it is argued that the recipe rating as a normative cue could exert a noticeable and significant effect in the evaluation of the message credibility. According to the social impact theory, the strength of a normative effect lays amongst other factors in the importance of a group opinion to the individual. Scholars suggest that such conformity may differ per culture (Bond & Smith, 1996).

The Netherlands and Poland are both individualistic cultures, but it is argued that uncertainty avoidance of a culture is believed to moderate the effect of rating on the credibility assessment of a recipe. Power distance is not directly applicable because of the missing knowledge on the social position of the audience that expressed their opinion. In addition, masculinity is neither directly applicable as it is characterised by the stress of nurture and achievement in a culture (Hofstede & Hofstede, 2005). Consumers in strong UA cultures are likely to be more risk averse than consumers in weak UA cultures (Hofstede & Hofstede, 2005). Hence, it can be argued that consumers from high UA cultures may tend to rely more on the opinions of others to decrease the perceived risk of adopting the information.

Considering that Poland has a strong UA and the Netherlands a low UA, it is postulated that consumers from Poland will have a higher tendency to conformity than consumers from the Netherlands. As such, it can be argued that culture may moderate the reliance on rating in the credibility assessment of an online recipe. Hence, the following is postulated:

Hypothesis 4a: If the recipe has a high rating, then Polish participants will perceive

this recipe more credible than Dutch participants will.

Hypothesis 4b: If the recipe has a low rating, then Polish participants will perceive

(28)

27 | P a g e 4.3 Effect of perceived message credibility on likelihood of recipe adoption

The elaboration likelihood model (ELM) by Petty and Cacippo (1986) postulates that persuasion is a conjoin function of the involvement of the recipient and the credibility of the message (Brown, Broderick, & Lee, 2007). It is also argued that the message credibility is an initial key stage in the persuasion process of a message (Petty & Cacioppo, 1986). If the online recipe message lacks credibility, the message will be discounted and therefore not very persuasive. Since scholars suggest that online messages are frequently processed via the peripheral route, according to the ELM, the keys determined in this persuasion process are the peripheral cues: the source and the recipe rating. In line with the ELM, the Hovland-Yale model also postulates that factors (source, message and receiver) influence the persuasiveness of the message. From both models can be derived that a message with positive attributes will be more persuasive than a message will less positive attributes (Eagly & Chaiken, 1993; Hovland, Janis, & Kelley, 1953; Petty & Cacioppo, 1986). As such, behavioural influences are expected to be higher when the credibility of the message is high compared to when the credibility of the message is perceived to be low. Hence, it is argued that the credibility of a message can determine how much the reader adopts and learns from the received message. A recipient who perceived an online recipe as credible is likely to adopt the recipe and make purchasing decisions on it. On the other hand, if the recipient perceives the online recipe to be less credible, the effect will be discounted and the recipient will seek to avoid potential risk, hence is unlikely to follow or adopt the online recipe. This finds direct support in the studies of McKnight & Kacmar (2006) and Cheung, Lue, Sia, & Chen (2009). Their studies show the positive effect of the recipients perceived message credibility on the willingness to accept the eWOM recommendation. Consequently, the following is postulated:

Hypothesis 5: Perceived message credibility of an online recipe has a positive effect

on the likelihood of recipe adoption.

5 METHODS

5.1 Data

(29)

28 | P a g e

was distributed amongst Dutch and Polish consumers that were willing to fill in the online questionnaire. In this study, the Dutch sample consists of 164 subjects (59 men and 105 female, mean age = 37 years). The Polish sample utilized in this study consists of 185 subjects (82 men and 103 female, mean age = 35 years). Thus, the entire sample consists of 349 subjects. Because the respondents were obliged to answer all questions, the sample did not contain missing data. However, 3 cases were excluded out of the dataset, due to unreliable answers indicated by answering to all questions the lowest option. The dataset was also screened on univariate outliers. Using the rule of thumb by Tabachnick & Fidell (2007), if the dataset includes more than 80 respondents, univariate outliers are to be excluded if exceeding a Z-score of 3.29. This rule of thumb exposed no outliers.

5.2 Survey design

Nearly all instrument items used in the self-administered online questionnaire were adapted from prior studies, though with some alterations to fit the context of this study (see table 4). Since the official instruments were in English, the questions were first translated into both Dutch (in consultation with the graduation supervisor) and Polish (by a native Polish speaker who has a major in cross-cultural studies and psychology). Disagreements in the meaning and wording of the questions were resolved through profound discussions with both the thesis supervisor and the interpreter. In both Poland and the Netherlands, a pilot study was done before the actual collection of the data. This included 5 randomly selected subjects both in Poland and in the Netherlands. After filling in the questionnaire, the subjects were asked if there had any difficulties in understanding the questions and were invited to give suggestions for improvement. The results indicated no major issues in the comprehension of constructs measured in the questionnaire.

(30)

29 | P a g e

recipe and related control questions. To permit this study to examine credibility evaluation of an online recipe, the subjects were asked to evaluate an image that depicted an invented online consumer recipe forum with a Dark Chocolate & Raspberry Ripple muffins recipe (see figure 3). This image was manipulated on two points, namely by recipe source (anonymous, consumer or brand) and recipe rating (low or high). Hence, each subject was exposed to one out of six versions with each a combination of the source (consumer, brand or anonymous) and combination of the rating (high or low). For instance, a brand-endorsed recipe with low rating was one of the options presented, see figure 3. When the subject entered stage four, they were randomly appointed exclusively to one of the six variations. The Dark Chocolate & Raspberry Ripple muffins recipe was chosen in consultation with the regional Unilever Chef Jan-Peter Versteegen. This recipe followed the criteria of being simple, having a few ingredients and is a recognizable recipe for both Polish and Dutch subjects. Thereby trying to minimize the effect of the actual recipe on the credibility evaluation by the subject. In the final stage, additional control questions were asked and the subject was debriefed. While completing the questionnaire, subjects were restricted to go back to the previous stage. This was necessary to safeguard against the possibility of altering the answers in the previous stage, to insure that the answers were not influenced by obtained knowledge in the process of completion.

(31)

30 | P a g e 5.3 Measurements

Construct Items Mean / SD Instruments Cronback

alpha Authors Dependent variable Perceived message credibility 7 point Likert scale

I find the information about the muffin recipe on this website:

8.9

(Flanagin & Metzger, 2000; Flanagin & Metzger, 2007; Lowery, 2004; Roberts, 2010) 5.92 (1.16) - Believable 5.65 (1.21) - Accurate 5.64 (1.25) - Trustworthy 5.48 (1.50) - Bias 5.72 (1.27) - Complete Likelihood of recipe adoption 7 point Likert scale

4.67 (1.63) I would bake this recipe for myself and/or for others.

8.4

(Cheung, Lue, Sia, & Chen, 2009; Park, Lee, & Han, 2007)

4.32 (1.56) I would recommend this recipe to my friends and/or family.

Independent variables

Recipe source Dummy Primed/ stimulated via an avatar figure of an anonymous, consumer or branded source Author

Recipe rating Dummy Primed/ stimulated via a star rating system (1=not recommend to 5=highly recommended) (Cheung, Lue, Sia, & Chen, 2009)

Moderating variable

Culture Nominal What is your nationality? (Pornpitakpan & Francis,

2008) Control variable Brand-equity 7 point Likert scale

I find Blue Band or Kasia,

7.9 (Vogel, Evanschitzky, & Ramaseshan , 2008) 5.95 (1.18) - an strong brand

5.46 (1.46) - an attractive brand 4.57 (1.62) - an unique brand

(32)

31 | P a g e Ordinal 5.16 (1.08) How frequently do you bake muffin recipes? Flanagin & Metzger,

2000) Involvement Ordinal 3.03 (1.53) When you would like to bake sweet baking recipes (e.g. cakes, cookies, muffins and pies),

do you search for recipes on the Internet?

(Park, Lee, & Han, 2007)

Internet usage Open 27.72 (20.13) How often do you use the Internet? Number of hours per week: (Flanagin & Metzger, 2000)

Allergy Nominal Are you allergic to any of the ingredients used in this recipe? Author Find recipe

delicious Nominal

Do you think that the dark chocolate raspberry muffins, prepared according to this recipe,

will be delicious? Author

Feasibility of recipe

7 point Likert scale

6.09 (1.15) I find the muffin recipe on this website:

- Infeasible – feasible Author

Age Open 36 (12.38) What is your age?

(Fogg & et al., 2001; Flanagin & Metzger, 2000)

Gender Nominal What is your gender? Author

Income Ordinal 2.23 (.776) What is your gross annual income? Scale adopted to Dutch or Polish (below, around or

above average gross annual income) based on Eurostat. Author Degree of

education Ordinal 5.02 (.1.144) What is your highest degree or level of education that you have completed? Author

(33)

32 | P a g e 5.3.1 Independent variables

As shown in figure 3, for this study an image was designed of a fictitious online consumer recipe that presented a baking recipe. The image included all the familiar aspect of an online consumer recipe forum, based on recipe forums such as http://www.smulweb.nl and http://www.znajdzprzepisy.pl. In order to explore whether the source of a recipe and/or rating affected the perceived message credibility of an online recipe, the recipe message was manipulated on two-points: recipe source and recipe rating.

The recipe source variable consisted of three different type of sources, namely anonymous, consumer, and brand source. It was chosen to take a real brand name as opposed to a fictitious brand to increase reality of the source. With the authorisation of Unilever, the margarine brand Blue Band was utilized for The Netherlands and Kasia for Poland. According to Unilever (2013), both brands are popular baking margarine brands, which share a similar family focus, marketing approach and brand attributes. In addition, according to market research by Unilever (2013) Blue Band and Kasia are considered by consumer to be equivalent brand. As such, the confounding effect of brand familiarity will be minimized on the dependent variable (Yoon, Kim, & Kim, 1998, p. 160). The consumer, brand and anonymous source of the recipe were articulated by an avatar figure as shown in the figure 4, 5, and 6. In addition, the credentials of the source were presented below the avatar picture. By providing a visual as well as the credentials, it is believed that the subjects have a clear understanding that the recipe is coming from a marketing, consumer, or anonymous driven source.

Figure 4 Dutch Consumer Figure 5 Dutch Brand Figure 6 Dutch anonymous source

Rating of recipe variable is the overall rating of the recipe that is provided by other

(34)

33 | P a g e

stars. By showing extremes, it is argued that the effect of the rating may be better measured. See for example figure 7 for the extremely low rating and figure 8 for the extremely high rating, as depicted for the Dutch respondents.

Figure 7 Low rating of recipe 1/5 stars Figure 8 High rating of recipe 4.5/5

5.3.2 Moderating variable

Following the study of Pornpitakpan & Francis (2008), cultural influences were operationalized via the culture of the respondents. The Polish culture was assumed to present a large power distance, strong uncertainty avoidance and high masculinity culture, whereas the Dutch culture was assumed to represent a culture with small power distance, weak uncertainty avoidance and small masculinity. While these cultures differ on all the previous mentioned dimensions, they are equal on the individualism-collectivism dimension. In this study, the Hofstede dimensions are not tested or measured. In order to make inference about the culture, it will be included as a nominal variable.

5.3.3 Dependent variable

With the focus of this study to investigate the consumers’ credibility evaluation of an online recipe and the adoption thereof, this section will discuss the operationalization of two dependent variables, namely the perceived recipe credibility and the likelihood of recipe adoption.

Perceived message credibility of an online recipe as dependent variable was

(35)

34 | P a g e

used scale of Meyer (1988) to measure messenger credibility, as both include trust, bias and accuracy.

Likelihood of recipe adoption refers to the recipients’ intention to actually bake

the recipe or suggest it to friend and/or family. As such, the variable measures the message acceptance of a recipe. In order to gauge the likelihood of recipe adoption, this study implemented the 7 point-Likert scale of purchase intention by Park, Lee, & Han (2007). Amendments were made to the questions to fit in the context of this study. As such formulating it to the following measurements of construct:

• I would bake this recipe for myself and/or for others.

• I would recommend this recipe to my friends and/or family.

5.4 Control variables

(36)

35 | P a g e

brand equity scale by Vogel, Evanschitzky, & Ramaseshan (2008) were adopted. “I find a brand likable” was excluded based on translation and interpretation issues. Hence the measurement based on a 7 point Likert scale, existed out of I find a brand; (1) attractive, (2) strong, and (3) unique.

5.5 Reliability of measurement

5.5.1 Perceived message credibility scale

In order to make inference about the perceived message credibility of the online recipe, the bipolar adjectives (believability, accuracy, trustworthiness, bias, completeness) were compounded into one variable. Before merging the variables into one artificial new variable, 3 steps were performed. First, the correlation between these variables was explored using a Pearson product-moment correlation coefficient (see table 5). This indicated that all variables, except for Bias were strongly positive correlated, at a p< 0.01 (Pallant, 2007). The variables Bias was medium positive correlated.

Pearson Correlations 1 2 3 4 (1) Believability 1 (2) Accuracy .681** 1 (3) Trustworthiness .783** .752** 1 (4) Bias .445** .433** .494** 1 (5) Completeness .685** .805** .751** .525** **. P< 0.01 level (2-tailed). N=349

Table 5 Correlation of perceive message credibility

(37)

36 | P a g e Component Matrixa Component 1 Believability .825 Accuracy .860 Trustworthiness .883 Bias .588 Completeness .875 Extraction Method: Principal Component Analysis.

Table 6 Factor loadings Table 7 Eigenvalues

Lastly, the Cronbach’s alpha coefficient was explored to analyse the internal consistency of the scale. This test showed that variables had an internal consistency of 8.9. Above 8 is perceived as good (Pallant, 2007). The above statistical evidence provided enough prove to merge the 5 bipolar adjectives of the perceived message credibility scale into one new variable. However, this study chose to exclude the Bias variable when forming a new variable that formed the message credibility. This was partly based on face validity of the variable and statistical evidence. The investigation of the dataset suggested that subjects found a brand-endorsed recipe to be mostly unbiased, while in some cases the subjects found a consumer-endorsed recipe to be extremely biased. Thereby, indicating that this item may not have measured what it was intended to measure. Furthermore, the loading of Bias into the component was not high, solely 5.88. In addition, by excluding bias, the component accounted for 80% of the variances between the variables. Also, the Cronbach’s alpha coefficient increased from 8.9 to 9.3. Based on these findings and prior researches that have applied the scale (Flanagin & Metzger, 2000; Flanagin & Metzger, 2007; Roberts, 2010), it was chosen to merge believability, accuracy, trustworthiness and completeness into one new variable. Hence, based on the statistical evidence and prior research, this study adopted and treated this new variable as the perceived message credibility of an online recipe.

5.5.2 Likelihood of recipe adoption

In order to make inference about the likelihood of recipe adoption, the two adjectives by which the scale was measured were merged into one new variable. Hence, the same

Total Variance Explained

Component Initial Eigenvalues Total % of Variance Cumulative % 1 3.312 66.247 66.247 2 .742 14.845 81.092 3 .463 9.269 90.361 4 .254 5.086 95.447

(38)

37 | P a g e

steps were followed as elaborated on in the previous section. First, the correlation amongst the variables was explored by preforming a Pearson product-moment coefficient analysis. This demonstrated a strong positive correlation between the two variables, (r=.713, n=349, p<0.01). Second, a Principal Component Analysis was executed. This analysis showed that both variables loaded into 1 component, as illustrated in table 8 Furthermore, as shown in table 9 this one component accounted for 86% of the variance of the two variables. This component also had an eigenvalue of 1.731, which is acceptable as above 1 can be retained according to Hatcher (1994).

Component Matrixa

Component

1

(1) I would bake this recipe

for myself and/or for others. .930

(2) I would recommend this recipe to my friends and/or

family. .930

Extraction Method: Principal Component Analysis.

Table 8 Factor loadings Table 9 Eigenvalues

Thirdly, the internal consistency of the scale was checked with the Cronbach’s alpha coefficient. The Cronback’s alpha coefficient was .844, which indicated a good internal consistency of the scale (Pallant, 2007).

The above statistical evidence justified the merging of these variables together into one new variable. Finally, taken into account that past literature demonstrated that the two variables capture the construct of likelihood of adoption (Park, Lee, & Han, 2007), the above results confirmed this. Hence, this new variable was adopted and treated as the likelihood of recipe adoption, and was used for further testing.

5.5.3 Brand-equity scale

In order to make inference about the perceived brand-equity of either Blue Band or Kasia, the bipolar adjectives (a strong brand, an attractive brand, a unique brand), as discussed in the control variable section, was merged into one variable. To justify the merge of the 3 variables into one variable, the identical steps will be followed as outlined in the previous section. The Pearson product-moment correlation coefficient

Total Variance Explained

Component Initial Eigenvalues Total % of Variance Cumulative % 1 1.731 86.574 86.574 2 .269 13.426 100.000 Extraction Method: Principal Component

(39)

38 | P a g e

that was explored between the three variables as depicted in table 10. This table shows that there was a strong positive correlation between the 3 variables (Pallant, 2007).

Pearson Correlations 1 2 (1) Strong brand 1 (2) Attractive brand .648** 1 (3) Unique brand .546** .560** **. P< 0.01 level (2-tailed). N=349

Table 10 Correlation of perceive brand-equity

Component Matrixa Component 1 Strong brand .863 Attractive brand .870 Unique brand .818

Extraction Method: Principal Component Analysis.

Table 11 Factor loadings Table 12 Eigenvalues

Second, by the Principal Component Analysis it was showed that all three variables loaded in one component as is indicated in table 11. Table 12 indicates that this component accounts for 72.4% of the variance in the three variables, which is relatively high. Lastly, the Cronbach’s alpha coefficient test was preformed that indicated a scale internal consistency of .796. This is in the limits of acceptance, as above 7 is acceptable though above 8 is preferred (Pallant, 2007). In line with prior research and statistical evidence provided by this study, a new variable was created from merging the three brand-equity items into one new variable.

6 RESULTS

This study has a 3 x 2 between subjects factorial design, in which the recipe source was varied (brand-, consumer, or anonymous-endorsed) and the recipe rating (low rating or high rating) of an online recipe. For this experiment subjects were randomly divided into six groups (Group 1: Anonymous-endorsed recipe with low rating, Group 2:

Total Variance Explained

Component Initial Eigenvalues Total % of Variance Cumulative % 1 2.171 72.371 72.371 2 .478 15.934 88.305 3 .351 11.695 100.000

(40)

39 | P a g e

Anonymous-endorsed recipe with high rating, Group 3: Brand -endorsed recipe with low rating, Group 4: Brand-endorsed recipe with high rating, Group 5: Consumer-endorsed recipe with low rating, Group 6: Consumer-Consumer-endorsed recipe with high rating). In order to make an inference on whether these groups are statistically significant differed in perceived message credibility and the moderating role of culture, an analysis of variance or covariance (ANOVA or ANCOVA) was conducted. In addition, the relationship between perceived message credibility of an online recipe and the likelihood of recipe adoption is explored by a multi-regression analysis. Before elaborating on the results of the analysis, first the effects of the control variables will be discussed. In table 13 the correlations presented of the continuous and ordinal variables are presented. Results were found to be significant if the significant level was at p<0.10.

Correlations matrix

Variables Mean (SD) 1 2 3 4 5 6 7 8 9

(1) Perceived message

credibility 5.73 (1.1) 1

(2) Age 36 (12.38) -0,011 1

(3) How many hours a week do you use internet?

27.72 (20.13) -0,058 -,270** 1 (4) Brand equity 5.32 (1.22) ,133* 0,004 -,113* 1 (5) Education 5.02 (.1.144) -0,005 -,278** ,188** -0,091 1 (5) Income 2.23 (.776) 0,072 ,316** -,167** ,267** -0,027 1 (6) Baking frequency 4.13(1.08) 0,034 0,004 0,078 -,135* 0,022 -,141** 1

(7) Bake muffin frequency 5.16 (1.08) 0,068 ,182** -0,03 -0,1

-0,002 -0,033 ,606** 1

(8) Search recipe via

Internet 3.03 (1.53) -0,051 ,213** -,119* -,138*

-0,052 0,035 ,406** ,284** 1

(9) Feasibility of recipe 6.09 (1.15) ,707** 0,03 -0,038 ,166** 0,044 0,086 -0,087 -0,025

-0,05

** Correlation is significant at the 0.05 level (2-tailed).

*** Correlation is significant at the 0.01 level (2-tailed).

Table 13 Correlation matrix of continuous and ordinal variable

6.1 Control variables

(41)

40 | P a g e

credibility of an online recipe. This was done for two reasons. Firstly, though prior research may have identified the control variable, it is possible that in this study the control variable does not have a significant effect on the dependent variable. Hence, it may not control for what it was intended to control. Secondly, ANOVA and ANCOVA analyses are sensitive for the number of control variables included, so the less control variables included the better (Pallant, 2007). To statistically test the effect of the control variables on the dependent variable, this study executed a one-way ANOVA for ordinal and nominal control variable and for continues control variables in a one-way ANCOVA. Table 14 summaries all control variables as well as the main interaction effects. From the results shown in the table it can be derived that solely (1) Search for recipe via Internet, (2) feasibility of recipe, (3) Perceived delicious, (4) brand-equity were found to significantly (p<.05) effect the perceived message credibility. Hence, these variable were identified as significant control variables. As brand-equity is a continuous variable, it is viewed as a covariate in the analysis of variance. This explains why in certain tests a two-way ANCOVA is preformed or a two-way ANOVA

Independent: Dependent: Perceived message credibility

Sig. Main variables F df1 df2 P Sig. Interaction effects F df1 df2 P

Source .710 2 346 .492 Source * rating .006 2 343 .994

Rating .015 1 347 .903 Source * culture .131 2 349 .878

*** Culture 2.841 1 347 .093 Source * search internet .710 2 344 .279

Control variable F df1 df2 P Source * feasibility of recipe 2.992 10 330 .001

Income .871 2 347 .491 Source* perceived delicious .228 2 343 .796

Sex .045 1 347 .831 Rating * culture 1.060 1 345 .304

Age .042 1 347 .838 Rating * search internet 1.210 1 345 .272

Education .380 6 342 .891 Rating * feasibility of recipe 2.938 5 336 .013

Income .871 2 346 .419 Rating * perceived delicious .114 1 345 .736

Baking frequency .770 5 343 .572

Bake muffin frequency 1.050 5 343 .388

**

Search recipe via

Internet 3.091 4 344 .044 Internet usage 1.184 1 346 .227 * Feasibility of recipe 73.454 6 342 .000 Allergy to ingredients 2.253 1 347 .134 * Perceived delicious 33.412 1 347 .000 ** Brand equity 5.937 1 332 .015

Referenties

GERELATEERDE DOCUMENTEN

In episode three, the editor/author utilises bodies and spaces such as the king, the Babylonians, Daniel, the lions’ den, the prophet Habakkuk and food to demonstrate the

[r]

 Positioning based on the trader’s economic benefit: this occurs if the business providing the ranking places a product higher the more it benefits from the sale of this

A commercial practice is also misleading if the marketing of the product causes confusion, for example concerning products, trademarks, trade names or other characteristics by which a

In the current study it is hypothesized that the effect of the independent variables (the presence of demographic/ psychographic characteristics attached to an OCR)

The loop assured that the new created datasets report information at the level of consumers’ individual purchase journeys and only include the touchpoints related

H4d Compared to the no picture condition, a profile picture containing a real person positively impacts the perceived credibility of the OCR source through trustworthiness,

The aim of this empirical research is to analyze the relationship between the sender’s expertise with the product and the quality of the arguments presented in an online