A Study of Heuristic Cues affecting the perceived
Credibility of Online Consumer Review Sources
from an ELM Perspective
Sven Giesen
A Study of Heuristic Cues affecting the perceived
Credibility of Online Consumer Review Sources
from an ELM Perspective
Master ThesisJune 27, 2016
Sven Giesen University of Groningen Faculty of Economics and Business
MSc Marketing
Specialization Marketing Management
Abstract
In a traditional face-to-face communication, people usually knew their interaction partner and could otherwise stick to the required skills to judge the credibility of the information given by their counterpart. The computer-mediated communication significantly changes the rules of this game. Due to the complexity of the current media landscape, where information is uncer-tain and often manipulated, the present paper suggests that – with regard to the ELM and HSM – people increasingly rely on profile characteristic cues in judging the credibility of an online consumer review source. The main objective of this study is to investigate the effect of the reviewer’s profile picture and reputation as stated by other users. Firstly, this study essen-tially contributes to the current credibility research literature by providing a tangible and robust conceptual delineation of credibility. Thereby, the present paper provides empirical evidence that the essential factors driving credibility are defined through the three-component model consisting out of trustworthiness, expertise and homophily. Concluding the findings of the following analysis, the profile picture as well as the reviewer’s reputation positively im-pacts the three-component model and credibility itself. Reasoned mainly by the principle of social presence and the identity heuristic, a profile picture showing a real person turns out to positively impact those perceptions. Furthermore, not only a picture of a human being, but also a computer-based entity such as an avatar shows such a positive effect compared to no picture being present, strongly providing evidence for the machine heuristic. Next to the profile picture, the reputation of the reviewer as a cue generated by fellow consumers finally has an even higher impact on the perceptions of credibility, trustworthiness, expertise and homophily.
Acknowledgements
This thesis is submitted in fulfilment of the requirements for the Master's Degree in Marketing Management of the University of Groningen. It is the result of work done in the period from February to June 2016. On this occasion, I want to express my sincere gratitude to those people who essentially supported me during this time and who are responsible for even hav-ing the possibility to be in this situation. Firstly, I would like to thank my supervisor Dr. Liane Voerman for all her support. Not only as my mentor, but also as a very good and caring per-son, she was always available with her advice and understanding. Next to her excellent guid-ance and her honest feedbacks during our meetings, she also supported me in terms of private issues within the time when writing my Master Thesis.
I also wish to thank my mother Annette and my brother Lars, who were essential for the mental support I needed in this time.
Unfortunately, during the time of writing my thesis, I had to accompany the most adorable person of my life, my dad Jörg Giesen, on his debilitating fight against cancer. I will be ever grateful for his support and assistance and I am enormously sad that he cannot see me gradu-ate. Knowing that he is incredibly proud of me, I dedicate him my master thesis.
Finally, I want to thank my girlfriend Radostina Zlatanova for always being there for me and incredibly supporting me with all her love, kindness and backing.
Sven Giesen
I
Table of Content
CHAPTER 1: INTRODUCTION ... 1
1.1 The transition from traditional WOM to eWOM ... 1
1.2 The consequences of the eWOM phenomenon on consumers’ credibility evaluations ... 1
1.3 The role of heuristic cues within an OCR ... 2
1.3.1 The prominence-interpretation theory by Fogg (2003) ... 3
1.3.2 Introduction of the Elaboration-Likelihood Model (ELM) and Heuristic-Systematic Model (HSM) ... 4
1.3.3 The usage of peripheral cues in OCRs ... 4
1.4 Introduction of the independent variables profile picture and stated reputation... 4
1.5 Definition of credibility and its determinants – The three-component model ... 6
1.6 Problem statement and research questions ... 7
1.7 Academic and managerial relevance ... 8
1.8 Structure of the following chapters ... 8
CHAPTER 2: THEORETICAL FRAMEWORK ... 10
2.1 The three-component model of perceived source credibility ... 10
2.1.1 The effect of trustworthiness on the credibility perception of an OCR source ... 10
2.1.2 The effect of expertise on the credibility perception of an OCR source ... 10
2.1.3 The effect of homophily on the credibility perception of an OCR source ... 11
2.2 The effect of the profile picture on the three-component model of source credibility ... 12
2.2.1 The effect of a real person profile picture on the three-component model of source credibility... 12
2.2.1.1 The uncertainty reduction theory ... 13
2.2.1.2 The social presence heuristic ... 13
2.2.1.3 The identity heuristic ... 14
2.2.1.4 Hypotheses – Real picture condition ... 14
2.2.2 The effect of an avatar as a profile picture on the three-component model of source credibility ... 15
2.2.2.1 The machine heuristic ... 15
2.2.2.2 The anthropomorphisation of avatars and the social presence heuristic ... 15
2.2.2.3 The anthropomorphic extent of an avatar ... 16
2.2.2.4 Hypotheses – Avatar picture condition ... 16
2.2.3 The comparison of the effects of a real person picture and an avatar ... 17
2.3 The effect of the reputation of an OCR source on the three-component model of source credibility .... 17
2.3.1 The reputation and endorsement heuristic... 18
2.3.2 The authority heuristic ... 19
2.3.3 The bandwagon heuristic and the principle of social proof... 19
2.3.4 Hypotheses – Reputation of the OCR source ... 19
2.4 The interaction effect between the profile picture cue and reputation cue ... 20
II
2.6 Summary and Conceptual Model ... 21
CHAPTER 3: RESEARCH DESIGN ... 23
3.1 Experimental design ... 23
3.2 Setup of the Survey ... 23
3.2.1 Introduction and Priming Questions ... 24
3.2.2 The Manipulation ... 25
3.2.3 Dependent Variables ... 26
3.2.4 Manipulation Check ... 26
3.2.5 Background and Control Variables ... 27
3.2.6 Population and Distribution ... 28
3.2.7 The Sample ... 28
3.3 Operationalization ... 29
3.3.1 Measurement of the DVs, moderators and covariates ... 29
3.3.2 Validation of the Scales ... 32
3.3.3 Measurements – Manipulation Check ... 34
3.3.4 Validation of the Manipulation Success ... 35
3.3.5 Preliminary Analysis of the Priming Questions and Manipulation Checks ... 39
3.4 Plan of Analysis ... 39
3.4.1 Testing Disturbance Term Assumptions ... 41
CHAPTER 4: RESULTS ... 43
4.1 Two-Way ANOVAs ... 43
4.2 Multiple regressions ... 45
4.2.1 Testing the antecedents of credibility – the three-component model ... 46
4.2.2 The effects on perceived credibility ... 47
4.2.3 The effects on perceived trustworthiness ... 49
4.2.4 The effects on perceived expertise ... 50
4.2.5 The effects on perceived homophily ... 52
4.2.6 Summary of main findings of the linear regressions ... 53
4.3 Mediation analyses ... 55
4.3.1 Mediation as supposed in the initial conceptual model ... 56
4.3.2 Mediation – Likability as an essential mediator ... 57
4.3.3 Mediation – Initial model adjusted by Likability as a mediator ... 60
4.4 Summary of main findings ... 63
4.5 Validation of hypotheses ... 64
CHAPTER 5: CONCLUSIONS AND RECOMMENDATIONS ... 67
5.1 Discussion ... 68
5.2 Managerial implications ... 69
III
1
Chapter 1: Introduction
1.1 The transition from traditional WOM to eWOM
With the rise of online channels, there is a great transition in the retail market. This allows consumers not only to make purchases from their homes, but also enables them to interact and communicate with each other online (Verhoef et al., 2015). More precisely, this rising trans-parency allows consumers to share their thoughts, opinions and experiences with products and services through so called electronic Word-of-Mouth (eWOM) (Bickart & Schindler, 2001; Duan et al., 2008; Hennig-Thurau et al., 2004; Xu, 2014). This new form of Word-of-Mouth (WOM) entails “all informal communications directed at consumers through Internet-based technology related to the usage or characteristics of particular goods and services, or their sellers” (Litvin et al., 2008, p. 461).
While traditional WOM limits the distribution of experiences and information to the inner circle of the message source identified as friends or relatives and consists out of spoken words in face-to-face situations (Bickart & Schindler, 2001), eWOM significantly changes the “rules of the game”. The eWOM phenomenon allows people to interact with an unlimited broader range of unknown customers in form of the written word (Sun et al., 2006). As a conclusion, Phelps et al. (2004) argue that eWOM might be even more influential than traditional WOM because of its convenience, one-to-many reach, rapidity of its distribution and finally because it does not entail the human pressure of face-to-face communication. (Cisco Internet Business Solutions Group 2013; Sun et al., 2006)
2
and not physically present (Fogg et al., 2007; Riedl et al., 2011). More precisely, computer channels make it impossible to conveniently interpret one another, as self-presentation cues such as e.g. nonverbal behavior like physical appearance, emotional expression and body language are not as easy to detect as in a traditional WOM setting (Ma & Argwal, 2007). As a conclusion, the judgement of credibility is way more complicated within the cyberspace of eWOM, as those traditional indicators of credibility are not applicable anymore (Metzger et al., 2010; Dou et al., 2012). The steady increase of the complexity of the current media land-scape even hampers this by a huge variety of eWOM being present (Metzger et al., 2010). Therefore, this paper contributes to “a supremely key concern in the new media environment” by studying drivers of credibility online (Sundar, 2008, p.73).
As those recent developments are “virtually absent in the credibility research literature” (Metzger et al., 2010, p.415), this paper closes the research gap in the credibility perception in eWOM by examining several cues within an online consumer review (OCR), as it is one of the most important ways where eWOM can be present (Duan et al., 2008). In this regard, Park and Park (2008, p.744) define OCRs as “any positive or negative statements made by poten-tial, actual, or former customers about their experiences, evaluations, and opinions on prod-ucts and services”. Finally, it is the object of this paper to shed light into the credibility judgement of OCRs by examining various cues triggering the use of heuristics (Wathen & Burkell, 2002).
1.3 The role of heuristic cues within an OCR
As stated above, the communicationmediated by technology entails various challenges and barriers regarding the effectiveness of social interactions.
As a conclusion, an OCR setting forces the reader to judge credibility based on altered and reduced cues, as the availability of personal information is limited, which makes an appropri-ate judgement very difficult (Cheung et al., 2008; Gotlieb & Sarel, 1991; Schindler & Bickart, 2005). Those heuristic cues are somewhat hidden in OCRs as they appear as so-called credi-bility markers, mental shortcuts, rules-of-thumb or guidelines that enable the reader to finally make a credibility judgement (Sundar, 2008; Metzger et al., 2010; Dou et al., 2012).
3
specific heuristic cues is sorely needed to understand how people judge credibility in digital media. As especially cues concerning the source characteristics are not adequately researched yet (Lis, 2013), the present paper examines the impact of the source’s characteristics on the credibility perception of the reader. Source characteristics are especially interesting regarding the natural anonymity of an eWOM setting and hence, the difficulty to determine the proper-ties of the sender (Lis, 2013). For instance, the proper-ties between the author (in this study called “source”) and the reader are typically lacking in strength and consumers have difficulties determining the credibility of information given by that stranger (Park & Lee, 2009; Bronner & De Hoog, 2010; Ehsaei, 2012). In this regard, Dou et al. (2012, p.1555) describe investigat-ing source effects as “a crucial step to understandinvestigat-ing the psychological process of people’s online review evaluation”. The present paper therefore focuses on these source-related cues. For this, the prominence-interpretation theory, the Elaboration Likelihood Model (ELM) and the Heuristic-Systematic Model (HSM) are introduced.
1.3.1 The prominence-interpretation theory by Fogg (2003)
4
1.3.2 Introduction of the Elaboration-Likelihood Model (ELM) and Heuristic-Systematic Model (HSM)
Petty and Cacioppo (2012) define the ELM as consisting out of two routes of processing information leading to persuasion. The cognitively effortful central route is characterized by a deep evaluation of the content of a message. On the other hand, the peripheral route describes the processing of information as an automatic and low-effort form of processing information relying on peripheral cues. Similarly, the HSM differentiates systematic processing as “refer-ring to a detailed analytical consideration of judgment-relevant information” from heuristic processing as “relying on mental shortcuts to judgmental rules (or heuristics) that are already stored in memory” (Chaiken, 1980; Sundar, 2008, p. 74).
1.3.3 The usage of peripheral cues in OCRs
5
include a representative picture. This trend dismisses the anonymity assumption of previous research and facilitates forming an impression of the user (Utz, 2010). Since privacy issues may restrict the publication of a picture of oneself (Pesce et al., 2012) it might be of high managerial importance to see whether an avatar picture might have a similar effect on the perceptions of credibility, trustworthiness, expertise and homophily. More precisely, this paper will investigate whether an avatar might compensate for not uploading a real picture. For instance, Nowak and Rauh (2005) report an increasing usage of avatars as an instance for communication and interaction not only in online communication contexts like OCRs, but also in E-commerce.
Conclusively, with the reviewer’s profile picture as a source characteristic cue, this study distinguishes between having either no picture, a real picture of the reviewer or an avatar as a form of visual representation of oneself (Nowak & Rauh, 2005).
According to Metzger et al. (2010), people traditionally referred to experts in a top-down manner in contexts of information scarcity. In the new media landscape, a shift towards a bottom-up evaluation process, where people rather rely on community or collective efforts, is observable. With the development of new IT features in the online context, this trend of a peer-to-peer credibility judgement is nowadays often present in the form of user ranking systems (Ma & Argawal, 2007; Metzger et al., 2010). However, the majority of studies con-sider credibility as an individual judgement made in isolation, thereby neglecting the social influence that other users could have. Hence, there is a lack of research investigating credibil-ity perceptions of user-generated and shared content by considering group-based tools (Metz-ger et al., 2010). For example, users on Amazon or eBay credential other users profile based on their experiences with them (Metzger et al., 2010). For instance, the feedback score pro-vided at eBay is one of the most important claims of a user profile (Ehsaei, 2012). Additional-ly, the website of the New York Times lists the most e-mailed stories of the day and a high amount of e-commerce websites use the phenomenon of cross-selling, indicating what other users with same interests have bought (Sundar, 2008). Conclusively, Ehsaei (2012) argues that an enhancement of feedback systems in OCRs is needed.
6
1.5 Definition of credibility and its determinants – The three-component model
Literature defines source credibility as the perception of the believability of a source by the reader of the message” (Hovland & Weiss, 1951). The definition of credibility as such is broadly discussed in research as previous studies treat it either as the sum of components and determinants (in this study trustworthiness, expertise and homophily) that yield to an overall credibility evaluation, while other studies investigate credibility at an aggregate level without specifying sub dimensions (Lis, 2013; McKnight & Kacmar, 2006; Metzger et al., 2010). Similarly to Pornpitakpan (2004), this study will cover both perspectives.
Hovland et al. (1953) define credibility in the scope of their source-credibility model as a function entailing the dimensions “expertise” and “trustworthiness”. The majority in the field of credibility research supports this view of credibility as mainly consisting out of these two sub dimensions (Dholakia, 1977; Wathen & Burkell, 2002). Another approach of conceptual-izing credibility is done by the source-attractiveness model of McGuire (1985), stating that the similarity between the source and the receiver is a further central dimension of credibility, which is stated as homophily in the present paper (von Wangenheim & Bayón, 2004; Lis, 2013). Finally, similarly to Ohanian (1990, 1991), this paper integrates the two models of Hovland et al. (1953) and McGuire (1985) into a three-component model consisting out of expertise, trustworthiness and homophily, which are declared as the essential determinants of message credibility.
Trustworthiness can be described as the belief of the audience that the information provided
by a source is objective and honest (Ohanian, 1991). As already mentioned, past interactions are missing and the reader of an OCR has to rely only on non-verbal cues in judging the credibility of the source. In this regard, cue-based trust can be defined as “the extent to which a person is confident in another based on information cues provided by or associated with the one to be trusted” (Xu, 2014, p.137).
As a further dimension of credibility, cues of an OCR might also drive the perceived expertise of the source. The concept of expertise refers to an assessment of the capability of the source, that assertions are correct and valid (Hovland et al., 1953). In terms of one’s expertise, the reviewer is perceived as credible under the condition that he or she endows specific knowledge of a product or has made experiences with it (Jain & Posavac, 1999).
Homophily finally describes the similarity among individuals (Ehsaei, 2012). Accordingly,
7
2012). Social homophily is defined by the structure of every type of network ties not only including traditional WOM ties like family and friends, but also relationships like information transfer and comembership (McPherson et al., 2001). Bringing this definition to an online context, the relationship strength can be hidden as nodal profile similarities (Xiang et al., 2010). As it “limits people’s social worlds in a way that has powerful implications for the information they receive, the attitudes they form, and the interactions they experience” (McPherson et al., 2001, p.415), it is crucial in the credibility judgement process.
1.6 Problem statement and research questions
In the digital media century, fellow consumers help each other by providing their experiences and opinions with products and services through so called eWOM. Those recommendations are a highly valuable source of information next to seller-generated information, where objec-tivity is usually lacking. Actually, this seemingly win-win situation to get an objective im-pression of products and services without being manipulated by marketers has an essential downside. Even this kind of information is uncertain in terms of credibility, as eWOM blows up the boarders of traditional WOM by connecting people unknown to each other online. As a solution, the present paper suggests that consumers judge the credibility of an OCR source based on different properties of the user profile of the OCR source, namely the profile picture and the reputation of the source.
Conclusively, the following research questions will be addressed:
RQ1a: How do the presence and type of the profile picture affect the perceived trustworthi-ness and expertise of the source and the homophily perception with the source?
RQ1b: How does the stated reputation of the source affect the perceived trustworthiness and expertise of the source and the homophily perception with the source?
RQ1c: How do the profile picture and stated reputation of the reviewer impact the perceived source credibility mediated by trustworthiness, expertise, and homophily?
RQ2: How do the perceptions of trustworthiness, expertise and homophily impact the per-ceived source credibility?
RQ2a: Which of the above mentioned determinants (see RQ2) has the highest impact on the perceived source credibility?
8
1.7 Academic and managerial relevance
At first, this study contributes to the conventional credibility literature by combining the source-credibility model (Hovland & Weiß, 1951) with the source-attractiveness model of McGuire (1985). Furthermore, considering the increasing complexity of the web, this study provides empirical evidence that people tend to facilitate their credibility judgement solely based on available heuristic cues, i.e. the profile picture and stated reputation of the reviewer. In this regard, the present study examines the assumptions of the ELM and HSM concerning the relationship between involvement and the peripheral, heuristic style of information pro-cessing.
Regarding the managerial relevance, the consequences of eWOM are diverse and broadly discussed in literature. For instance, Weber and Shandwick (2012) highlight in their study that 65% of potential consumers decided to buy a product that was not in their initial consideration set after being exposed to a favorable consumer review. Additionally, Leskovec et al. (2007) already find an influence of eWOM on the consumer’s decision making process leading to an increased likelihood of finally buying a product. Furthermore, according to Chen and Xie (2008), especially novice consumers benefit from reviews written by pioneer experts. Conclu-sively, the phenomenon of eWOM enables the seller to easily create “free sales assistants” (Chen & Xie, 2008, p.6). Bearing in mind the extent and consequences of the rising transpar-ency between consumers, companies are forced to take this transition into account (Hennig-Thurau et al., 2004). Therefore, the present paper contributes to investigate drivers of con-sumers’ credibility perceptions online as a consequence of eWOM, in order to come up with suggestions to implement these findings into the marketing strategy.
1.8 Structure of the following chapters
9
involvement on the above mentioned effects of source characteristic cues on the perceived credibility and its determinants.
Within Chapter 3, the methodology used in this paper is described. Therefore, as a first step, the composition of the manipulations of the picture and reputation is explained. Furthermore, the scales measuring the dependent variables, the moderator and control variables, as well as the manipulation checks are introduced. Afterwards, the scales are validated with a factor analysis and the manipulation success is controlled by a descriptive analysis and several One-sample T-tests.
In Chapter 4, the hypotheses are tested firstly by conducting several ANOVAs and AN-COVAs, followed by multiple regressions. The first multiple regression evaluates the impact of trustworthiness, expertise and homophily on credibility as suggested in 1.5. This is fol-lowed by four linear regressions analyzing the impact of the IVs on those four DVs. As a last step of analysis, possible mediating effects are analyzed.
10
Chapter 2: Theoretical Framework
The theoretical framework mainly provides relevant theories in current literature about the effects of the IVs as mentioned in 1.4 on the DVs as mentioned in 1.5. As a first step, the assumptions of the three-component model introduced in 1.5 are tested. Afterwards, the ef-fects of the profile picture and reputation of the source on the four DVs are explained by several psychological theories, followed by the derivation of interaction effects of the IVs. Lastly, the moderating role of product involvement is elaborated leading to a final overview of all hypotheses and the conceptual framework.
2.1 The three-component model of perceived source credibility
2.1.1 The effect of trustworthiness on the credibility perception of an OCR source
Research has found that highly trustworthy and expert sources are more likely to immediately affect attitude change than sources lacking in those attributes (Dholakia, 1977). More precise-ly, literature states that a highly credible source is perceived to be more trustworthy than a low credible source (Harmon & Coney, 1982). Furthermore, Xu (2014) finds that cognitive and affective trust both have a significant positive effect on the perceived review credibility. Similarly, Brown and Broderick (2007) define trustworthiness of the source as the source bias and state that a source “less prone to bias” (Brown & Broderick, 2007, p.6) is perceived as more credible. Therefore, the following hypothesis is formulated:
H1: The perception of trustworthiness positively impacts the perceived credibility of the OCR source.
2.1.2 The effect of expertise on the credibility perception of an OCR source
11
Broderick (2007) identify source expertise as one of the elements positively affecting the credibility perception.
H2: The perception of expertise positively impacts the perceived credibility of the OCR source.
2.1.3 The effect of homophily on the credibility perception of an OCR source
The perceived similarity between the sender and the recipient might play a crucial role in OCRs, as two more or less foreign users are communicating online. For instance, already Aristotle stated that humans “love those who are like themselves” (Aristotle, 1934, p.1371). Conclusively, in order to trust the judgement of the author of the review, there has to be some kind of fit between the characters of those individuals in order to consider a specific review in the decision making process. In this regard, Bronner and de Hoog (2013) argue that perceived similarity between the source and the reader is determined by the user profile of the source and might have an impact on the perceived credibility of an OCR. Additionally, Risselada et al. (2014) find that tie strength and homophily are important weighting factors in models of social influence (Risselada et al, 2014). For instance, Kossinets and Watts (2009) find a posi-tive effect of homophily (e.g. by having the same gender and age) on the formation of ties, which might positively impact credibility evaluations.
The research of Bickart and Schindler (2001) has further shown that the perceived similarity between the source and the reader might increase the persuasive influence of an OCR. Even if no similarities regarding demographics and lifestyles are present, the source and the recipient have in common that they are at least both interested in the same product and therefore fellow consumers. This degree of similarity might therefore determine not only the credibility of an OCR, but also its relevance to the reader. The simplest similarity of being fellow customers might be stronger than any try of a marketer to present its product. Linked to the similarity, also the ability to feel empathy with the source might have an impact on the persuasiveness of the OCR (Bickart & Schindler, 2001).
12
argue that a feeling of similarity can reduce uncertainty between interaction partners and therefore, this reduced uncertainty might lead to a more positive credibility evaluation. There-fore, the following hypothesis is formulated:
H3: The perception of homophily between the source and the reader positively impacts the perceived credibility of the OCR source.
2.2 The effect of the profile picture on the three-component model of source credibility 2.2.1 The effect of a real person profile picture on the three-component model of source credibility
As already mentioned, the information about the person behind an OCR is limited and it is difficult for the reader to judge the credibility of the source as easy as in a face-to-face situa-tion (Bickart & Schindler, 2001; Lis, 2013; Xu, 2014). In this regard, attribusitua-tion theory states that people use the social information available to make inferences about the identity and personality of others (Kelley, 1973). Therefore, the reader needs to judge the credibility of the review based on available alternative indicators of credibility (Metzger et al., 2010; Dou et al., 2012) like in this case, the profile picture of the reviewer.
For instance, Xie et al. (2011) suggest that readers may use personal identifying information (PII) of the reviewer as a credibility indicator of an OCR. They argue that competent and genuine reviewers are generally more likely to disclose personal information and therefore, the presence of these specific PII cues improves the reader’s judgement of the perceived credibility of OCRs. They indeed find a positive impact of the presence of PII in form of the name, day of stay and state of residence on the perceived OCR credibility. As an extension of their research, this paper suggests to investigate the effect of more informative descriptive PIIs, like in this case the reviewer’s profile picture on the perceived OCR’s credibility.
13
2.2.1.1 The uncertainty reduction theory
A first explanation for the possibly positive effect of the presence of a profile picture is given by the uncertainty reduction theory. The assumption behind this theory is the primary aim of people interacting with strangers to reduce uncertainty and increase predictability of their own and the other’s behavior (Berger & Calabrese, 1975). More precisely, the uncertainty reduc-tion theory states that people strive to reduce uncertainty about their counterpart in an interac-tion (Nowak & Rauh, 2005). According to Berger & Calabrese (1975), the uncertainty level is determined by the communication environment. Therefore, this paper suggests that especially in a computer-mediated communication environment such as an OCR, uncertainty plays a crucial role (Tidwell & Walther, 2002). Conclusively, people tend to try to avoid this discom-fort by i.e. interpreting the cue given by a profile picture with the help of the social presence and identity heuristic (Berger & Calabrese, 1975).
2.2.1.2 The social presence heuristic
According to Ma and Agarwal (2007), a profile picture is one way of self-presentation online. The suggestion that the mere presence of a picture has an effect on credibility perceptions of the reader can be explained by the psychological construct of perceived social presence through a picture, which shifts the previously distant relationship between the source and the reader to a more emotional dimension affecting a perception of trust (Steinbrück et al., 2002). More precisely, with respect to Sundar (2008), the theory of social presence states that specif-ic cues are able to trigger individuals towards the impression of communspecif-icating with a real person. Tidwell and Walther (2002) argue that especially in a computer-mediated communi-cation (CMC) setting, the ability to associate the communicator with a kind of social presence is important in developing an impression of the source.
investi-14
gate the effect of social presence with either a picture or without a picture in a social media setting, in this case LinkedIn. They find that users who own a profile picture are perceived as more competent and socially attractive than users without a picture, indicating an effect on the perception of similarity. Similarly to attractiveness, Xiang et al. (2010) found sources of homophily based on hidden effects of nodal profile similarities in their latent variable model in an online social networks setting. However, other studies argue that it is difficult or even impossible to detect similarity in an eWOM setting (Chatterjee, 2001; Park & Lee, 2009). On the other hand, the presence of a reviewer’s profile picture facilitates assessing similarity between a reviewer and a reader and thereby influences credibility perceptions (Bronner & de Hoog, 2013). Conclusively, this study suggests that, based on attribution theory and the pres-ence of heuristic cues, it is indeed possible. Finally,the mere presence of a profile picture is expected to be sufficient to positively impact credibility evaluations without cues indicating authority or other forms of power being present (Edwards et al., 2015).
2.2.1.3 The identity heuristic
Additionally, not only the social presence heuristic, but also the identity heuristic, stating that a reviewer tries to assert himself via technology, might be triggered through a profile picture and thereby increase the perceived credibility (Sunder, 2008). More precisely, as a high amount of digital media engagement is explained by user’s self-presentation, users might see themselves identified by a review source or, in this case, by his profile picture. For instance, Ma and Agarwal (2007) argue that recognizing the identity of knowledge contributors, in this case reviewers, helps readers determining source credibility and similarity.
2.2.1.4 Hypotheses – Real picture condition
Based on the previous findings and theories, this paper suggests that sorely the presence of a profile picture significantly impacts the perceived trustworthiness (Xu, 2014), expertise (Sun-der, 2008) and homophily (Edwards et al., 2015) of the OCR source. Additionally, with re-spect to 2.1, this paper suggests that the effect of a real profile picture is to some extent medi-ated by these constructs, which leads to the following hypotheses:
H4a: Compared to the no picture condition, a profile picture containing a real person posi-tively impacts the perceived trustworthiness of the OCR source.
15 H4c: Compared to the no picture condition, a profile picture containing a real person posi-tively impacts the perceived homophily between the reader and the OCR source.
H4d: Compared to the no picture condition, a profile picture containing a real person posi-tively impacts the perceived credibility of the OCR source through trustworthiness, expertise and homophily.
2.2.2 The effect of an avatar as a profile picture on the three-component model of source credibility
Avatars are defined as “computer generated visual representations of people or bots” (Nowak & Rauh, 2005, p.153) and are increasingly present in social virtual environments or ecom-merce. Hence, in the online context, an individual chooses an avatar to express his or her social attitude or personality. This facilitates the readers understanding of the social and per-sonal identity of the source (Golder & Donath, 2004; Smith et al., 2000). In the following, the machine heuristic and the social presence heuristic paired with a discussion of the antro-morphisation of avatars are introduced.
2.2.2.1 The machine heuristic
People got nowadays used to say that they got information “of the computer”. Thereby, they attribute information obviously originating from a human being to the computer itself. This process is based on the machine heuristic, stating that a machine-like interface is automatical-ly perceived as more objective (Sundar, 2008). Similarautomatical-ly, the so-called “computer=true”- or the “if it comes from a computer it must be true” heuristic states that people increasingly tend to believe information originating from a computer (Fennis & Stroebe, 2015). This might thereby trigger more positive credibility and especially expertise judgements (Fennis & Stroe-be, 2008; Sundar, 2008).
2.2.2.2 The anthropomorphisation of avatars and the social presence heuristic
16
perceive social presence when either human, but also computer-controlled entities are present through a picture. The perceived social presence in their study was higher in the avatar condi-tion, than in the no picture condition. For instance, in an advertising context, Rodgers and Thorson (2012) argue that a computer system like an avatar might no longer be treated as impersonal, but as a ”personified recommendation agent” (Rodgers & Thorson, 2012, p.363) causing a perception of social presence. More precisely, the avatar thereby becomes a social entity instead of being perceived as an inanimate object. Furthermore, Sundar (2008) supports that view of anthropomorphic cues triggering the social presence heuristic. Moreover, Nowak and Rauh (2005) find that avatars of an anthropomorphic nature are perceived as more attrac-tive and credible compared to other forms of avatars. Further, they find that feminine avatars are perceived as more attractive than masculine ones and that people generally tend to prefer a match of one’s gender and the gender of the avatar. The attractiveness perception and gender matching preference again relates back to the construct of homophily in this study.
2.2.2.3 The anthropomorphic extent of an avatar
The effect of the anthropomorphic extent of avatars is widely discussed in research. In this regard, anthropomorphism is defined as the extent of a character endowing human appearance or behavioral attitudes (Nowak & Biocca, 2003; Nowak, 2004; Nowak & Rauh, 2005). While some studies report that more anthropomorphic avatars are more likeable, engaging and cred-ible (Koda, 1996; Wexelblat, 1997), Nowak (2004) found a positive credibility effect of less anthropomorphic pictures. An extensive study of different kinds of avatars by Nowak and Rauh (2005) finally shows that the five highest scores on credibility and homophily are all introduced by human-like images. Additionally, they find that animal- and object-like avatars even reduce the credibility perception.
2.2.2.4 Hypotheses – Avatar picture condition
With regard to the stated effects above, this paper focuses on an anthropomorphic avatar, on which the following hypotheses will be tested:
H5a: Compared to the no picture condition, a profile picture containing an avatar positively impacts the perceived trustworthiness of the OCR source.
17 H5c: Compared to the no picture condition, a profile picture containing an avatar positively impacts the perceived homophily between the reader and the OCR source.
H5d: Compared to the no picture condition, a profile picture containing an avatar positively impacts the perceived credibility of the OCR source through trustworthiness, expertise and homophily.
2.2.3 The comparison of the effects of a real person picture and an avatar
As the present paper suggests that a profile picture including either an avatar or a real person each causes a perception of social presence and positively affects the perceived source’s credibility and its antecedents, it is of great interest to compare the extent of these effects. In this regard, there are two opposing views in literature (De Visser et al., 2012). The contrast hypothesis states that human-machine trust is different from human-human trust. On the other hand, the similarity hypothesis states that machine-like entities as in this case the avatar is able to trigger a similar perception of trust (de Visser et al., 2012). In a multi-round trust game, Riedl et al. (2011) support the latter, as people trusted avatars to a similar degree as humans. More precisely, the neuroscientific findings showed a differential activation of brain networks, but a similar outcome of trust perception. Additionally, the learning process of trustworthiness of an interaction partner is stated as similar. Furthermore, Nass et al. (1994) find that avatars are perceived as social actors and thereby cause social responses similar to a human-human interaction.
Based on the discussion about the anthropomorphic perception of avatars above, the follow-ing hypothesis is formulated:
H6: Compared to the real picture condition, an avatar similarly affects the perceived credi-bility, trustworthiness, expertise and homophily.
2.3 The effect of the reputation of an OCR source on the three-component model of source credibility
18
aggregate user representations (AURs). Within a reviewer’s profile, they are usually present through either ratios or counts (e.g. amount of people rating the reviewer as helpful) or as visuals (e.g. stars). In their paper, they specifically distinguish between deliberate and inci-dental AURs. While deliberate AURs represent the collective evaluation of user-generated content (UGC) such as OCRs, incidental AURs reflect the trace or record of a user, generated by a computer and not intended to have a signaling purpose (Walther & Jang, 2012). Similar-ly, Xu (2014) defines a reputation cue as aggregated opinions or namely the collective en-dorsement of others, ranking the source in terms of trustworthiness and helpfulness.
As internet users tend to use the low effort peripheral route of processing, the reviewer’s reputation might not affect the perceived credibility caused by only one, but a variety of cog-nitive heuristics, which are introduced in the following (Walther & Jang, 2012).
2.3.1 The reputation and endorsement heuristic
The stated reputation of a reviewer facilitates inferences and validations about the source’s expertise and identity and therefore relates back to the above discussed construct of the identi-ty heuristic (Ma & Argawal, 2007). For instance, previous research indicates that people tend to rely on others to make judgements about credibility and that they therefore make use of group-based tools (Metzger et al., 2010). Metzger et al. (2010) label this tendency as the
reputation heuristic. The relationship between reputation and trust is the object of a variety of
different settings and research areas (Ba & Pavlou, 2002; Fuller et al., 2007; Zucker, 1986). Metzger et al. (2010) label a way of processing such reputational cues as the endorsement
heuristic, stating that people tend to perceive sources as credible if others do so. For instance,
19
2.3.2 The authority heuristic
Previous studies have already shown that a reviewer’s self-claim concerning expertise leads to a more useful perception of the OCR and that it positively impacts the perceived credibility (Eastin, 2001). Additionally, Sparks and Browning (2011) find that positively framed infor-mation in terms of positive reputation leads to an increase in trustworthiness. This effect can be explained by the authority heuristic (Sundar, 2008). The authority heuristic describes “a cognitive decision rule established through prior experience that teaches that experts are a valid source of information due to their authority on a subject” (Willemsen et al., 2011, p.22). In this context, it is important to highlight that this effect can be present irrespective of the content of information (Willemsen et al., 2011). As consumers tend to judge the credibility of a review based on more than only one review characteristic (Xu, 2014), this paper suggests that expertise might not just appear in terms of a self-claim in the content of a review but might be reflected in the reputation cue, representing the rating by other users.
2.3.3 The bandwagon heuristic and the principle of social proof
Research suggests that the information provided by fellow users is the favorite source to build one’s judgement on, which is best explained by an increasing usage of the “bandwagon heu-ristic” (Sundar, 2008, p. 83). In this regard, the bandwagon heuristic describes the tendency of adopting the majority’s opinion (Moe & Schweidel, 2012). For instance, Sundar et al. (2013) find that user ratings significantly impact the perception of the reader of an OCR. Similarly, the principle of social validation or as labeled by Cialdini (2009) the principle of social proof describes “the tendency to look at other people to infer the ‘correct’ way of behaving in a given situation” (Fennis & Stroebe, 2015, p. 272). More precisely, people tend to use the actions of others as an indicator for the appropriate action (Cialdini, 1993). As this principle or cue is usually used when alternative sources of information are not available or accessible, this paper suggests that social proof might be an important factor in the eWOM setting, where source information is rare (Fennis & Stroebe, 2015). Finally, as the principle of social valida-tion is already extensively used in advertising, it might as well play an important role in the online review context (Fennis & Stroebe, 2015).
2.3.4 Hypotheses – Reputation of the OCR source
20 H7a: Compared to the low reputation condition, a high reputation positively impacts the perceived trustworthiness of the OCR source.
H7b: Compared to the low reputation condition, a high reputation positively impacts the perceived expertise of the OCR source.
H7c: Compared to the low reputation condition, a high reputation positively impacts the perceived homophily between the reader and the OCR source.
H7d: Compared to the low reputation, a high reputation positively impacts the perceived credibility of the OCR source through trustworthiness, expertise and homophily.
2.4 The interaction effect between the profile picture cue and reputation cue
Regarding the peripheral nature of information processing, the presence of various cues is expected to impact each other (Walther & Jang, 2012). For instance, Riegelsberger et al. (2003) find a significant interaction effect between a profile picture and the vendor’s reputa-tion. Their results indicate that the effect of a positive reputation dissipates when adding a profile picture and vice versa that the presence of a profile picture might compensate a nega-tive reputation. On the other hand, Walther and Parks (2002) introduce the warranting princi-ple suggesting that peoprinci-ple increasingly rely on information with a higher warranting value. In this regard, the warranting value defines the reader’s perception of the extent to which the given information is “immune to manipulation by the person to whom it refers” (Walther & Parks, 2002, p. 552). Bearing in mind that the reputation is an other-generated cue, and there-by resistant to manipulation, Walther et al. (2009) argue that this kind of information might count more for the credibility perception of the reader. Therefore, the impact of a picture (being either a real picture or an avatar) might be lower, if a reputation cue is present. That leads to the following hypothesis:
H8: The presence of a reputation cue lowers the impact of a profile picture on the perceived source credibility and its determinants.
2.5 The moderating effect of the recipient’s involvement on the effect of source characteristics on perceived OCR source credibility and its determinants
21
involvement. Their results indicate that highly involved consumers tend to use the high effort central route of processing, sticking to quality relevant cues, whereas low involved consumers tend to use the peripheral route of processing, judging a product by external, non-product related cues (Petty et al., 1983). Therefore, it is the aim of this paper to investigate a possible moderating impact of involvement on the effect of the different source credibility cues (reputation and profile picture cue) on the perception of credibility and its antecedents (Xie et al., 2011). More precisely, regarding the suggestions of 1.3.3 that the peripheral route is the most commonly used route to judge OCRs, a low involvement level is expected to positively impact the effect of peripheral cues as in this case the profile picture and reputation of the source.
Petty and Cacioppo (1990), as well as Lee, Park and Han (2008) find that within the credibility judgement process, effects driven by the characteristics of the source tend to interact with receiver-related factors, such as in this case, the level of involvement. In this regard, previous research shows that source characteristics seem to have a higher impact on the perceived credibility under the low involvement condition (Wathen & Burkell, 2002). Additionally, as already mentioned, the source characteristics chosen in this paper are expected to represent heuristic cues, thereby being especially relevant for low involved consumers. Contrarily, non-heuristic cues like the quality of the content of the OCR are known to have a higher impact, when the involvement is at a high level (Lee et al., 2008). Following the ELM and the findings of Petty et al. (1983), it can be expected that low involved consumers will strongly process the picture and reputation cue, while highly involved customers might not even recognize them. This leads to the following hypotheses:
H9: The positive effect of the picture and stated reputation cues on the coutcome variables perceived credibility, trustworthiness, expertise and homophily will increase when involvement is low.
2.6 Summary and Conceptual Model
22
visualizes these expected relationships, followed by table 1, which summarizes the hypothe-ses, which are going to be tested in chapter 4:
Figure 1: Conceptual Model
H1 The perception of trustworthiness positively impacts the perceived credibility of the OCR source.
H2 The perception of expertise positively impacts the perceived credibility of the OCR source.
H3 The perception of homophily between the source and the reader positively impacts the perceived credibility of the OCR source.
H4a Compared to the no picture condition, a profile picture containing a real person positively impacts the perceived trustworthiness of the OCR source.
H4b Compared to the no picture condition, a profile picture containing a real person positively impacts the perceived expertise of the OCR source.
H4c Compared to the no picture condition, a profile picture containing a real person positively impacts the perceived homophily between the reader and the OCR source.
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, expertise and homophily.
H5a Compared to the no picture condition, a profile picture containing an avatar positively impacts the perceived trustworthiness of the OCR source.
H5b Compared to the no picture condition, a profile picture containing an avatar positively impacts the perceived expertise of the OCR source.
H5c Compared to the no picture condition, a profile picture containing an avatar positively impacts the perceived homophily between the reader and the OCR source.
H5d Compared to the no picture condition, a profile picture containing an avatar positively impacts the perceived credibility of the OCR source through trustworthiness, expertise and homophily.
H6 Compared to the real picture condition, an avatar similarly affects the perceived credibility, trustwor-thiness, expertise and homophily.
H7a Compared to the low reputation condition, a high reputation positively impacts the perceived trustwor-thiness of the OCR source.
H7b Compared to the low reputation condition, a high reputation positively impacts the perceived expertise of the OCR source.
H7c Compared to the low reputation condition, a high reputation positively impacts the perceived homophi-ly between the reader and the OCR source.
H7d Compared to the low reputation, a high reputation positively impacts the perceived credibility of the OCR source through trustworthiness, expertise and homophily.
H8 The presence of a reputation cue lowers the impact of a profile picture on the perceived source credi-bility and its determinants.
H9 The positive effect of the picture and stated reputation cues on the coutcome variables perceived credibility, trustworthiness, expertise and homophily will increase when involvement is low.
Table 1: Overview of Hypotheses
23
Chapter 3: Research Design 3.1 Experimental design
The present paper uses an experimental design by manipulating the independent variables
presence and type of a profile picture and level of stated reputation. The study investigates a
different set of dependent variables. The main dependent variable is the perceived credibility of the review source. Three other dependent variables are expected to mediate the effect of the IVs on the main DV, i.e. the perceived trustworthiness, expertise and homophily. Therefore, a 3 (profile picture: no, real, avatar picture) x 2 (reputation of review: positive, negative) be-tween-subjects design (table 2) is chosen to explore how the source characteristics and their interaction with each other affect users’ perceived review credibility (Field, 2013).
No Real picture Avatar
High Condition 1 Condition 3 Condition 5
Low Condition 2 Condition 4 Condition 6
Table 2: Overview of conditions
As the content of the review and the visual appearance are exactly the same across conditions, it is not appropriate to choose a within- or mixture-design for the following reasons:
1. Participants might be bored, as the text is all the same and there is no variation
2. The manipulation becomes too obvious and gets too much conscious attention, leading to the danger that people easily compare the manipulations and the choice is pre-determined (e.g. everybody judges the one with a profile picture the highest as default) In a between-subjects factorial design, participants are randomly assigned to different levels of the independent variables, called conditions (Field, 2013). Hence, in this study, there are six different conditions, where the participant is either confronted with no profile picture, with a picture showing a real person or with a picture of an avatar and, at the same time, either a high or low reputation. As for each condition, at least an amount of 30 respondents is needed to be perceived as reliable, a minimum of 180 respondents is needed for this research.
3.2 Setup of the Survey
For gathering the data, a survey hosted by Qualtrics is conducted. The following paragraphs guide through the structure and content of the questions the respondent will be asked to an-swer.
24
3.2.1 Introduction and Priming Questions
Instead of choosing an introduction that triggers the participant towards carefully and con-sciously looking at the user profile (e.g. picture and reputation), the survey starts with a quite general and brief introduction, only including a salutation and general explanation of the overall OCR topic in order to avoid an unintended manipulation of the results. More precise-ly, the reason behind that is that participants should spend the attention they normally spend when scanning an OCR in order to eliminate or at least limit the drawbacks an experimental study has compared to a field study by assuring a realistic and true-to-life setting. However, to ensure that people recognize and notice the stimuli in the manipulation, the study makes use of the priming effect (Bargh et al., 1996) in the first two questions. The questions will be asked as follows:
1. Imagine you want to write a review on tripadvisor.com. While creating your account you are asked, whether you want to upload a profile picture. Which of the following options would you choose?
a) I would upload a picture of myself
b) I would choose an avatar (a virtual proxy of a real person) c) I wouldn’t upload a picture at all
2. Usually, review sites provide ratings, where other users rate the reviewers in terms of positive or negative user ratings. Do you consider these ratings when reading reviews? (measured with a 5-point scale ranging from always to never)
Asking the first question serves this study in two ways. Firstly, according to the priming effect, it triggers participants to unconsciously watch out for a profile picture cue. Secondly, this question can be used to make inferences about hypothesis H1d and H2d in further analy-sis, as the perceived homophily might be triggered by the same beliefs and attitudes of choos-ing a profile picture. The second question is again essential for this study in two ways. Firstly, it again makes use of the priming effect which by now triggers the attention of the participant to look for reputation cues in the following. Secondly, it might be useful to make inferences about hypothesis H2d in further analysis, as the perceived homophily might be triggered by the perception of having a similar or non-similar reputation as the reviewer in the respective condition.
25
3.2.2 The Manipulation
The survey consist out of one OCR, showing on the right a review text concerning a specific service and on the left a user profile showing the respective manipulation of the IV. Below the text and the picture, the second manipulation of the stated reputation of the reviewer is shown. Figures 2-4 show the exemplary conditions avatar/high reputation, no picture/low reputation and real picture/high reputation to illustrate the structure of the chosen OCR style.
Figure 2: Examplary OCR condition (Avatar/High Reputation)
Figure 3: Examplary OCR condition (No Picture/Low Reputation)
26
The placement of the picture and text is optimal as the principle of hemispheric lateralization and the principle of matching activation state that the right brain hemisphere is responsible for processing pictures and is activated by a stimulus on the right and the left hemisphere for text vice versa (Fennis & Stroebe, 2015). The profile picture will have a reasonable size, which is consistent across conditions, in order to assure that the picture is recognized and to avoid a confounding effect of various sizes across conditions. Below the review text and the profile picture, the stated reputation manipulation is shown, introduced by the phrase “how other users rated this reviewer”. The positive reputation condition is signaled by 314 positive user ratings and a green thumb up and 2 negative user ratings with a red thumb down. The nega-tive reputation condition is built up vice versa. The wording of the review text is kept very simple, straight-forward and neutral, trying to avoid any indicators of credibility, homophily, trustworthiness or expertise within the text. The product or better service evaluated on in the review is a hotel. This choice is supported by the findings of Jain and Posavac (1999), as in case of search attribute information, source credibility matters less in the evaluation process. Contrastingly, for experience attribute information, consumers are forced to verify the experi-ence claim and therefore rely on cues. Similarly, Park and Lee (2009) support that eWOM effects are greater for experience goods. Each condition includes the same questions and is only manipulated on the independent variables. More precisely, by showing exactly the same review text to the respondents and by manipulating only the reviewer’s profile, it can be assured that different results are due to the manipulation.
3.2.3 Dependent Variables
After being exposed to the manipulation, respondents are asked to answer questions concern-ing the source on the perception of credibility, trustworthiness, expertise and homophily. Furthermore, the involvement of the respondent is measured. This set of questions is followed by a manipulation check, testing the recognition of several components of the manipulation, which is explained in the following paragraph 3.2.4.
3.2.4 Manipulation Check
27
present instead. Finally, they are asked to remember the gender and hair color of the person at the profile picture. The first manipulation check concerning recognition is essential for further analysis, as this condition necessarily needs to be fulfilled in order to draw causal relation-ships from the results of analysis in chapter 4. The more specific questions serve this study in a way that it allows drawing inferences about how deeply the participant processed the picture and that in turn might serve as an additional indicator regarding the involvement of the re-spondent. Finally, it might also indicate, whether respondents are successfully primed as expected in 3.2.1.
Furthermore, participants are asked to state whether they actually noticed an indicator of the review source’s reputation. If yes – respondents should indicate whether the reputation was positive or negative and in which way it was present. Furthermore, they are asked to remem-ber how many people actually rated the reviewer as positive and negative. Similarly to the procedure explained above, the first question serves to ensure that the manipulation is actually noticed and the more specific question allow inferences about the attention, involvement, sensitivity and priming success.
The last two manipulation check items control for the perception of the content of the review. The respondent is asked to judge the strength of the arguments given in the OCR. This serves the study to control for the drawbacks of the ELM as identified by the Unimodel. According to the ELM, a low involvement should increase the effect of heuristic cues (Petty et al., 1983; Sundar, 2008). Following the suggestion of Fennis and Stroebe (2015), the Unimodel ques-tions the generalization of the effect of the involvement, suggesting that even highly involved consumers might switch from assessing the arguments to scanning the heuristic cues in case arguments are perceived as low. Vice versa, if heuristic cues are perceived as weak, low involved consumers might switch to processing arguments. Hence, the effect of the profile picture and reputation should be stronger, when arguments are perceived as low. Additionally, asking the respondent to remember what the reviewer said about the service personnel of the hotel allows controlling for the depth of processing of the – within this paper – non-relevant content-part of the review. Hence, it is expected that the higher involved people are, the better (more correctly) they remember the content of the review.
3.2.5 Background and Control Variables
28
OCRs, experience with OCRs, one’s opinion-leadership on hotels and usage frequency of the internet and finally the age of the respondent.
3.2.6 Population and Distribution
The population of respondents will be between the age of 16 and 65. Using the convenient sampling method, the sample frame is built up of family, friends, fellow students, sports and work colleagues (Malhotra, 2009). Those respondents are randomly assigned to the different conditions. As the present paper will - as an extra - control for a differential effect of the participant’s gender, the randomization considers filling the conditions with an equal amount of men and women. The survey is going to be distributed online (via social media, e-mail and WhatsApp) to reach as many respondents as possible and necessary. Additionally, the partici-pant is neither distracted, nor forced to answer the questions at a set time. The time needed to finish the survey will be limited to not more than 10 minutes and participants remain anony-mous and can leave the survey at any time. In the following paragraph 3.2.7, the actual sam-ple is described.
3.2.7 The Sample
29
3.3 Operationalization
The following paragraph 3.3.1 firstly guides through the scales used from literature to meas-ure the constructs stated above. Furthermore, the scales are checked on their validity by doing a factor analysis and reliability analysis in 3.3.2. This is followed by the validation of the manipulation success with the help of several One-Way ANOVAs and descriptive analyses in 3.3.3. In order to validate the significance of comparing the descriptive findings, several chi squares are calculated manually using the following formula: 𝜒² = ∑(𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑−𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑)²
𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 Finally, a preliminary analysis of the priming questions is conducted.
3.3.1 Measurement of the DVs, moderators and covariates
Firstly, the construct of credibility is measured by a subset of items as suggested by Putrevu and Lord (1994) including the following items on a 7-point Likert-scale ranging from “totally agree” to “totally disagree” expanded by an item directly asking for the credibility perception:
I1 = The reviewer is credible.
I2 = The claims of the reviewer are true.
I3 = I believe the claims of the reviewer.
In the following, the participant will be asked to indicate his perceived trustworthiness, using a 7-point bipolar scale of Ohanian (1990) including the following items:
I1 = Dependable – Undependable
I2 = Honest – Dishonest
I3 = Reliable – Unreliable
I4 = Sincere – Insincere
I5 = Trustworthy – Untrustworthy
Afterwards, the construct of perceived expertise will be measured by applying the 7-point bipolar scale of Ohanian (1990) asking the following items:
I1 = Expert – Not an expert
I2 = Experienced – Unexperienced
I3 = Knowledgeable – Unknowledgeable
I4 = Qualified – Unqualified
30
As a last step of measuring the set of dependent variables, the construct of similarity, i.e. homophily, will be measured using a 7-point Likert-scale including the items suggested by Feick and Higie (1992) as follows:
I1 = The reviewer probably has similar values and beliefs.
I2 = The reviewer is quite a bit like me.
I3 = It is likely that the reviewer and I have similar tastes and preferences.
As already stated in 1.5, previous research struggles differentiating the sub constructs exper-tise and trustworthiness from the overall credibility constructs, this paper considers it appro-priate to rely on the scales stemming from only one study, as Ohanian (1990) successfully and explicitly distinguishes between those sub dimensions.
As the present paper strives to investigate the moderating role of involvement, the study makes use of a 7-point scale as suggested by the Modified Personal Involvement Inventory (PII) of Mittal (1995) with the following items:
I1= important – unimportant
I2= means a lot to me – means nothing to me
I3= matters to me – does not matter
I4= significant – insignificant
I5= of no concern – of concern to me (reversed)
In the scope of this paper, involvement might be even better measured in terms of the amount of time spent, when looking for a suitable hotel. Therefore, this item is also included into the survey with possible choices of 1 day, 2-4 days, 5-7 days, 1-2 weeks, or more than 2 weeks. The last part of the survey will include the following set of control variables, which might as covariates directly impact the different DVs.
Attractiveness (one item of the scale ass suggested by Feick and Higie, 1992)
Likability (Feick & Higie, 1992)
Skepticism (Holbrook & Batra, 1988)
Usage/Channel experience (a subset of items of Murray & Schlacter, 1990)
Personal attitude towards OCRs (Lee, Park & Han, 2008)
