1
THE EFFECT OF HOMOPHILY ON PERCEIVED
RISK AND THE RELATIVE EFFECT OF
DEMOGRAPHIC INDICATORS ON PERCEIVED
DEMOGRAPHIC HOMOPHILY IN OCRS
By
2
THE EFFECT OF HOMOPHILY ON PERCEIVED
RISK AND THE RELATIVE EFFECT OF
DEMOGRAPHIC INDICATORS ON PERCEIVED
DEMOGRAPHIC HOMOPHILY IN OCRS
Master thesis
By
T. van den Berg University of Groningen Faculty of Economics and Business First supervisor: Dr. J. A. Voerman Second supervisor: Prof. Dr. L. M. Sloot
3
TABLE OF CONTENT
1. INTRODUCTION ... 7
1.1 The degree of perceived risk ... 7
1.2 The degree of perceived risk and (e)WOM ... 8
1.3 The degree of perceived risk and OCRs ... 8
1.4 The concept of homophily ... 9
1.4.1 Demographic and perceptual homophily ... 9
1.4.2 The use of different demographic indicators to measure demographic homophily ... 10
1.5 Contribution of this study ... 11
1.6 Problem statement and research question ... 11
1.7 Academic and managerial relevance ... 12
1.8 Structure of the thesis ... 13
2. THEORETICAL FRAMEWORK... 14
2.1 Theory of homophily... 14
2.1.1 The theory of cognitive dissonance... 14
2.1.2 The cognitive balance theory, liking principle and attractiveness model ... 15
2.1.3 The hedonic fluency model ... 15
2.1.4 The social comparison theory, social identity theory and social network theory... 16
2.1.5 The uncertainty reduction theory ... 16
2.2 Study one: effect of demographic and perceptual homophily on perceived risk ... 17
2.2.1 Current findings on demographic homophily in (e)WOM communication ... 17
2.2.2 The effect of demographic homophily on perceived risk ... 18
2.2.3 Current findings on perceptual homophily (e)WOM communication ... 19
2.2.4 The effect of perceptual homophily on the degree of perceived risk ... 20
2.2.5 The joint effect of perceptual and demographic homophily on perceived risk ... 21
2.2.6 Other variables ... 21
2.2.7 Hypotheses and conceptual model study one ... 22
2.3 Study two: the relative effects of demographic indicators on demographic homophily ... 23
2.3.1 Gender homophily... 23
2.3.2 Occupational homophily ... 24
2.3.3 Age homophily ... 24
2.3.4 Location homophily ... 25
2.3.5 Name homophily ... 26
2.3.6 Relative strength of demographic indicators to demographic homophily... 27
3. RESEARCH DESIGN ... 28
3.1 Choice of research ... 28
4
3.1.2 Conjoint analysis study two ... 28
3.2 Population and sample ... 30
3.3 Data collection ... 31
3.4 Manipulation ... 31
3.4.1 Manipulation of demographic homophily ... 32
3.4.2 Manipulation of perceptual homophily ... 32
3.5 Operationalization of study 1 ... 34
3.5.1 Operationalization of the degree of perceived risk ... 34
3.5.2 Manipulation check ... 34
3.5.3 Control variables and Covariates ... 35
3.6 Operationalization of study 2 ... 35
3.7 Factor analyses ... 36
3.8 Plan of analysis ... 39
3.8.1 Plan of analysis study one ... 39
3.8.2 Plan of analysis study two ... 41
4. RESULTS ... 43
4.1 Two-way ANOVA study 1 ... 43
4.2 Multiple regression study 1 ... 44
4.4 Conjoint analysis study 2 ... 46
4.4.1 Estimation of results... 46
4.4.2 Goodness of fit: assessment of model fit ... 47
4.4.3 Preferences per choice set ... 47
4.4.4 Relative attribute importance ... 48
4.5 Validation of hypotheses ... 48
4.5.1 Validation of hypotheses study 1 ... 49
4.5.2 Validation of hypothesis study 2 ... 50
5. CONCLUSION AND DISCUSSION ... 51
5.1 Conclusion study 1 ... 51
5.2 Conclusion study 2 ... 52
5.3 Discussion, limitations and future research study 1 ... 52
5.4 Discussion, limitations and future research study 2 ... 53
5.5 Scientific and managerial relevance ... 55
REFERENCE ... 56
APPENDIX ... 64
5
ABSTRACT
Customers often experience some kind of risk during their purchase decisions. This is especially true for online purchase decisions. Customers have one thing in common when it comes to perceived risk: they want to abate it. When perceived risk is reduced during a purchase decision, customers are more likely to buy a particular product. The literature suggests that homophily is one way to reduce perceived risk. Homophily refers to the degree to which individuals are similar regarding certain attributes. The literature distinguished two types of homophily: demographic and perceptual homophily. Demographic homophily refers to the degree to which individuals are similar regarding demographic indicators such as gender, age, occupation, location and education whereas perceptual homophily refers to the degree to which individuals are similar regarding values, attitudes, beliefs and lifestyles. Only little is known about the effect of perceptual homophily in OCRs. The objective of this paper is twofold. The first study examines the effect of perceptual homophily on the degree of perceived risk and its joint effect with demographic homophily in an OCR setting. A regression analysis shows that homophily does not have a significant effect on the degree of perceived risk. That is, demographic, perceptual and the joint effect of both dimensions of homophily do not have a significant effect on the degree of perceived risk in OCR. Finally, to study demographic homophily, the literature made use of different demographic indicators. For example, one study used the demographic indicators age, location, photo and occupation to measure demographic homophily whereas another used location, age, gender and name. The aim of the second study is to measure the relative strength of five demographic indicators – location, gender, occupation, name and age - to perceived demographic homophily in an OCR setting. The results of the conjoint-analysis show that all demographic indicators but location are significantly important to perceive demographic homophily. Thus, individuals do not gain any utility when the reviewer of an OCR is from the same residence. The results show that the demographic indicator ‘occupation’ has the relative strongest effect to perceive demographic homophily. The effect of occupation homophily to perceived demographic homophily is that strong, that the indicator alone, is more important than the four other demographic indicators combined. The relative strength of occupation on perceived demographic homophily is followed by gender, age, name and location respectively.
6
PREFACE
This thesis has been carried out in order to complete the master Marketing Management at the University of Groningen in the period from January to June 2017.
Through this way, I would like to thank my supervisor Dr. Liane Voerman for all the feedback and support she gave me over the last six months. She did not only gave me feedback and support during consulting hours, but she was always available for any additional questions, even at night. Her excellent way of coaching definitely contributed to quality improvements of this thesis.
I also want to thank Dr. Felix Eggers for the feedback and support he gave me regarding specific questions for the conjoint-analysis. In his leisure time, Felix Eggers gave me several guidelines so I could properly conduct my second study.
Moreover, I want to thank Dr. Hans Risselada for his feedback and guidelines during the conceptual part of my thesis.
Finally, I want to thank my girlfriend, Sanne Bolks and my parents, Dick van den Berg and Mineke van den Berg, for their understanding, encouragement and mental support over the last six months.
I am deeply grateful to all persons mentioned above.
“I can do all this through Him who gives me strength.” - Philippians 4:13
7
1. INTRODUCTION
1.1 The degree of perceived riskCustomers often experience some kind of perceived risk when buying a product (Jacoby & Kaplan, 1974; Park & Park, 2008). Consumer research refers to perceived risk in terms of uncertainty and consequences, whereas perceived risk increases with higher levels of uncertainty and with the chance of greater negative consequences (Ogiethorpe & Monroe 1987; Cox & Rich, 1964; Campbell & Goldstein, 2001; Jacoby & Kaplan, 1974). For example, if a customer wants to book a hotel, perceived risk associated with the purchase could emerge because the customer does not know whether he will like his stay in the hotel (uncertainty), and is therefore worried whether he will have a comfortable holiday (consequence). According to Jacoby & Kaplan (1974), Rosario et al. (2016), Campbell & Goldstein (2001) and Park & Park (2008), customers can experience several types of perceived risk during their purchase decisions, such as performance risk (“does this hotel perform as I expect?”), financial risk (“is this hotel my money worth?”), privacy and security risk (“is this hotel respecting my privacy?”), social risk (“what do my friends and relatives think when I book this hotel?”) and psychological risk (“does this hotel fits well by my image?”). The combination of these several types of perceived risk (e.g. performance and financial risk) can be defined in an overall level, or, degree of perceived risk. The degree of perceived risk is determined by two general factors – the amount at stake in the purchase decision and the individual feeling of subjective certainty (Cox & Rich, 1964). The amount at stake is determined by the value attached to attaining some set of goals (e.g. degree of importance of having a comfortable holiday), and by the costs involved in attempting to achieve these goals (financial, time, psychological). Subjective certainty is when a consumer can win or lose all or some of the amount at stake (Cox & Rich, 1964). Although customers perceive risk in different ways during their purchase decisions, they have one thing in common: they want to abate it (Campbell & Goldstein, 2001). It is thus not surprising that customers are more likely to buy a particular product or service when perceived risk is reduced during their purchase decision (Moe & Trusov, 2011).
8
product meets certain expectancies. To reduce the perceived risk during purchase decisions, customers therefore rely on two main strategies: they consult others and they search for, and rely on, existing information (Cox & Rich, 1964). One way to search for and rely on information is through word-of-mouth (WOM) communication (Campbell & Goldstein, 2001).
1.2 The degree of perceived risk and (e)WOM
In WOM communication, customers share information, opinions and news with their friends, relatives and other social ties (Chen & Berger, 2016). For example, when a friend, relative or other social tie has a positive opinion with a certain hotel and shares this information (e.g. “I had a really comfortable holiday due to this hotel”), the subjective certainty of the receiver of this information could increase due to that the ‘loss’ of the amount at stake is minimized (e.g. “if he had a comfortable holiday, the chance is high that I will also have it”). It is not surprising that (WOM) communication reduces the degree of perceived risk (Roselius 1971; Dichter 1966; Campbell & Goldstein, 2001). Since the rise of consumer internet usage, electronic word-of-mouth (eWOM) has become the most important source of product information for customers (Fennis & Stroebe, 2016). In comparison with traditional WOM, customers find eWOM a more powerful, effective communication device, since it is easier to decipher given its written form and given its accessibility (Floyd et al., 2014).
1.3 The degree of perceived risk and OCRs
9
that OCRs are one of the most effective information sources for reducing perceived risk. Since OCRs are such effective for decreasing perceived risk, several techniques have been studied to further abate the degree of perceived risk in OCRs. For example, making OCRs more trustworthy by adding a membership duration of the reviewer, or the possibility to send the reviewer an email, decreases perceived risk (Racherla, Mandviwalla & Connolly, 2012; Rosario et al., 2016).
1.4 The concept of homophily
WOM theory suggests that another technique to reduce the degree of perceived risk among customers, is when the sender and receiver have similarity traits such as similar age or a similar lifestyle (Brown & Reingen, 1987). Rogers (1983) argue that individuals tend to perceive a stronger influence from people who are similar to themselves. This view is supported by Brown & Reingen (1987), which stated that similar individuals are more likely to interact with each other than dissimilar individuals. WOM theory uses the term “homophily” to refer to this perceived similarity between individuals (Lazarsfeld & Merton, 1954). Homophily refers to the degree to which people are similar regarding certain attributes (Lazarsfeld & Merton, 1954; Rogers, 1983; Brown & Reignen, 1987; McPherson, Lovin, & Cook, 2001; Gilly et al., 1998). Although the existence of homophilous behaviour was first noted by Tarde (1903), it were Lazarsfeld & Merton (1954) who first studied the effect of homophily and who distinguish two types of homophily: status homophily and value homophily. Status homophily, or demographic homophily, refers to the degree to which people are similar regarding demographic indicators such as gender, age, occupation, location and education (Lazarsfeld & Merton, 1954; Rogers, 1983; Brown & Reignen, 1987; McPherson, Lovin, & Cook, 2001; Gilly et al., 1998). Value homophily, or perceptual homophily, refers to the degree to which people are similar regarding values, attitudes, beliefs and lifestyles (Lazarsfeld & Merton, 1954; Rogers, 1983; Brown & Reignen, 1987; McPherson, Lovin, & Cook, 2001; Gilly et al., 1998).
1.4.1 Demographic and perceptual homophily
10
remarkable, since theory suggests that perceptual homophily also influences WOM (Rogers, 1983; Brown & Reingen, 1987; Gilly, Graham, Wolfinbarger & Yale, 1998; McPherson, Smith-Lovin & Cook, 2001). More specific, Brown & Reingen (1987) argue that future research could benefit by incorporating perceptual homophily to examine the effects and degrees of homophily on WOM behavior. It is thus questionable whether the generalizations made by research based on demographic homophily (e.g. Forman, 2008; Naylor et al., 2011; Rosario et al., 2016), are valid for homophily as a whole. Another interesting finding is that research which measured both dimensions of homophily, found stronger effects of perceptual homophily compared to demographic homophily (Bruyn & Lilien, 2008; Shen & Zhang, 2016). Therefore, this study examines both dimensions of homophily and the interaction effect on the perceived risk of customers.
11
demographic (e.g. age, income, gender). The increase in targeting options have led to an increase in personalized online content and websites (Goldfarb & Tucker, 2011; eMarketer, 2016). Interestingly, this trend is something what online customers expect and want (eMarketer, 2016). A recent survey found that the most important element in the online shopping journey is to make sure that the content is relevant, i.e. personal, for the customer (eMarketer, 2016). This personalization (e.g. targeting) can be further increased with homophily between the sender(s) and the reader(s). However, lacking the knowledge of the relative strength of each demographic indicator could diminish the effect. In addition, online review sites make use of machine learning algorithms to predict visitor profiles (Booking.com, 2016). These profiles can be classified based on visitor’s demographics such as gender, education level, age, occupation and location (Bock & Poel, 2009). Thus, using algorithms allows websites to personalize their offers (Bock & Poel, 2009; Booking.com, 2016). Over the years, these algorithms are constantly improved and offer more precise predictions of visitor demographics (Booking.com, 2016). For example, when a visitor is predicted to be 30 years old, the algorithm can automatically show OCRs of other reviewers who are 30 years old, while a visitor, predicted with age 50, could mainly see OCRs of reviewers which are 50 years old, which increases the effectiveness of OCRs.
1.5 Contribution of this study
The contribution of this paper is twofold. The first part of this paper contributes to homophily and OCR research by measuring both dimensions of homophily and their joint effect on perceived risk in OCR. The second part of this study contributes to homophily and OCR research by measuring the relative impact of different demographic indicators on perceived demographic homophily.
1.6 Problem statement and research question
The literature argues that future research is required to measure circumstances in which eWOM is a more powerful risk-reducing tool (Rosario et al., 2016). This paper posits that homophily is such a circumstance. This paper suggests that the degree of perceived risk is reduced by increasing both perceptual and demographic homophily between the sender and receiver in OCRs. Furthermore, this paper studies the relative strength of each demographic indicator to perceived demographic homophily as discussed in 1.4.21. This paper asks the following main research question:
1 The current literature found that homophily exerts a strong influence on hedonic goods but not on utilitarian
12
“What is the effect of homophily on perceived risk in OCRs and what is the relative effect of each demographic indicator towards perceived demographic homophily?”
1. What is the effect of demographic homophily between sender and perceiver on perceived risk in OCR?
2. What is the effect of perceptual homophily between sender and perceiver on perceived risk in OCR?
3. What is the effect of demographic and perceptual homophily on perceived risk?
4. What is the relative strength of each demographic indicator to perceived demographic homophily?
To answer the formulated questions above, this paper will conduct two studies. Study one measures the effect of demographic and perceptual homophily on the degree of perceived risk. Study two measures the relative strength of five demographic indicators: age, residence, name, gender and occupation, towards demographic homophily.
1.7 Academic and managerial relevance
As far as we know, this is the first study which measures the individual relative effect of each demographic indicators to demographic homophily and the joint effect of perceptual and demographic homophily on perceived risk in OCRs.
We contribute to the literature by enriching the understanding of the influence of online reviews by measuring the joint effect of homophily on perceived risk. Furthermore, for practitioners, it is interesting to gain insights in which demographic indicator have the highest relative effect to demographic homophily in OCRs. For academics, it is interesting to know which demographic indicators have the highest relative strength on perceived demographic homophily such that it can be measured properly. Knowing the relative effects of each indicator on demographic homophily makes it possible to optimize the effect of OCRs. For instance, next to the opportunity for algorithms (see 1.4.2), Tripadviser.com and Booking.com both use a location indicator in their OCRs, but do not use any age indicators. However, the relative strength of the latter could be higher on demographic homophily between the sender(s) and reader(s) than the former. Thus, knowing the relative strength of each indicator allows practitioners to optimize their OCR effects.
13
1.8 Structure of the thesis
14
2. THEORETICAL FRAMEWORK
As discussed in 1.6, this paper consists of two studies. The dependent variable of study one is the degree of perceived risk whereas the dependent variable of study two is the perceived demographic homophily. To better understand the effects of the two dependent variables and their independent variables, this chapter will discuss relevant findings and theories in the current literature. First, the theories of homophily will be discussed and it will be related to perceived risk in 2.1. Next, specific empirical support of demographic homophily and perceptual homophily is discussed in paragraph 2.2. Finally, the relative strength of each demographic indicator on the perceived demographic homophily will be discussed in 2.3.
2.1 Theory of homophily
Before discussing the underlying mechanisms of homophily (see 2.1) and the empirical findings of homophily in the eWOM literature (see 2.2), it is interesting to discuss the findings of the effects of homophily in the WOM literature. WOM Research found that communication between homophilous individuals, is more influential, more efficient, more effective, more likely to occur and more trustworthy compared to communication between heterophilous individuals (Rogers, 1983; Gilly et al., 1998; McCroskey, Richmond, & Daly 1975; McCroskey, McCroskey & Richmond, 2006; McPherson et al., 2001; Szmigin & Piacentini, 2015). To understand these findings, several underlying mechanisms are discussed in this paragraph. The (e)WOM literature makes no distinction in the underlying mechanisms of how perceptual and demographic homophily work. Therefore, this paragraph will discuss the mechanisms of the general impact of homophily on perceived risk. There are several underlying mechanisms which explain how homophily between sender and receiver may positively effects communication and why it works to reduce the degree of perceived risk.
2.1.1 The theory of cognitive dissonance
15
effective communication increase subjective uncertainty since less effective communication could increase the chance of losing some or all of the amount at stake which subsequently increases the degree of perceived risk (Reimer & Benkenstein, 2016).
2.1.2 The cognitive balance theory, liking principle and attractiveness model
Based on the theory of cognitive dissonance (see 2.1.1), it is not surprising that individuals like cognitive consistency. The cognitive balance theory (Heider, 1958) suggests that individuals have a certain tendency to seek balance with interpersonal relations. The cognitive balance theory of Heider (1958), as well as the like-me principle, or liking principle (Laumann, 1966) and the attractiveness model (Feick & Higie, 1992) show that the attitude of individuals to a particular object is influenced through interpersonal relations in a positive or negative way. To explain this mechanisms, Heider (1958) introduced the P-O-X model, where P, O and X interact with each other. In this model, P is a person, O is another person and X is an element. When person P has a positive attitude towards element X and person P has a positive relation with person O (e.g. they are similar), the theory argues that it is reasonable to infer that person O also has a positive attitude towards element X.
Thus, translated to an OCR setting, when the reviewer (person P) has a positive attitude towards a product (element), and the reviewer and the recipient (person O) are similar (positive relation), it is likely that the latter forms a positive attitude towards the product (element) and consequently reduces perceived risk.
2.1.3 The hedonic fluency model
16
Thus, familiarity (i.e. similarity) among two individuals may increase the subjective ease of information processing which positively affects the evaluation of the receiver of an OCR which subsequently decreases the degree of perceived risk.
2.1.4 The social comparison theory, social identity theory and social network theory In addition, the social comparison theory (cf. Festinger, 1954), the social identity theory (cf. Tajfel & Turner, 1986) and the social network theory (cf. Granovetter, 1973) explain the mechanisms how individuals interact with other individuals, groups or networks. These theories make a distinction between the interaction with similar and dissimilar individuals. Similar individuals, groups or networks are referred to as ‘convergent’, ‘in-group’ and ‘strong ties’ respectively. Dissimilar individuals, groups or networks are referred to as ‘divergent’, ‘out-group’ and ‘weak tie’ respectively. All three theories argue that individuals tend to interact and compare themselves with other individuals when uncertainty is present. Furthermore, these theories argue that individuals are more likely to interact with similar others to create trust and reduce uncertainty (Festinger, 1954; Brown & Reignen, 1987; Rogers, 1983; Racherla et al., 2012). Thus, according to these theories, homophily could increase the likelihood of interaction between individuals and could further reduce (subjective) uncertainty. It is noteworthyto mention that Festinger (1954) argues in his social comparison theory that individuals compare themselves with similar other individuals to ensure that their evaluation is more accurate. More accurate evaluations may decrease subjective uncertainty, and consequently, perceived risk (see 1.1). Moreover, Festinger (1954) argue that it is not possible to compare and evaluate the opinion of another person accurately, if this person is dissimilar to oneself.
In an OCR context, this implies that a receiver of a positive OCR could not accurately evaluate the opinion of the reviewer, when this latter is dissimilar to the receiver. Due to the lack of the accurate evaluation, a perceived existence, or even increase, of subjective uncertainty, could occur. This means that homophily between a sender and receiver should decrease perceived risk in OCRs whereas perceived risk occurs or increases when the sender and receiver are dissimilar (heterophilous).
2.1.5 The uncertainty reduction theory
17
Calabrese, 1975). Moreover, the URT argues that individuals seek any available information to decrease uncertainty when cues are limited (Ramirez, Walther, Burgoon, & Sunnafrank, 2002). In an OCR setting, these cues are indeed relatively limited compared to offline communications. Thus, the few social cues that are present in an OCR setting, become more salient and important in predicting one´s behaviour (Racherla et al., 2012). Moreover, the URT argues that the level of liking negatively effects uncertainty in communication between two or more individuals (Berger and Calabrese, 1975). This view is consistent with the liking principle (see 2.1.2). Note that similarity among communicators positively affect the level of liking.
Thus, when two individuals like each other, i.e. are similar, the predictability between two individuals increases and (subjective) uncertainty decreases which may decrease the degree of perceived risk in OCRs.
2.2 Study one: effect of demographic and perceptual homophily on perceived risk
This paragraph discusses the current findings in the literature of both demographic and perceptual homophily on the degree of perceived risk. Paragraph 2.2.1 discusses the current findings of demographic homophily in (e)WOM communication, followed by the discussion of the effect of demographic homophily on perceived risk in paragraph 2.2.2. Next, paragraph 2.2.3 discusses the current findings of perceptual homophily in (e)WOM communication, followed by the discussion of the effect of perceptual homophily on perceived risk in paragraph 2.2.4. In paragraph 2.2.5, the interaction effect of demographic and perceptual effect on the degree of perceived risk is discussed. In addition, paragraph 2.2.6 shortly discusses some other variables which may impact the degree of perceived risk and the relation between homophily and perceived risk. Finally, in paragraph 2.2.7, the hypotheses of study one are formulated and a conceptual model is depicted.
2.2.1 Current findings on demographic homophily in (e)WOM communication
18
effect on the helpfulness rating and on sales when reviews included identity descriptive information such as geographical location, nickname and real name positively (Forman, 2008). Furthermore, Shen & Zhang (2016) found that recipients were more likely to discount their own information and imitate others when they perceived demographic homophily. Moreover, Naylor et al., (2011) studied the persuasion effect of similar reviews, dissimilar reviews and ambiguous reviews, whereas the latter refers to reviewers without identifying information. They found that recipients are naturally making inferences about ambiguous reviewers and interpret them more like similarity reviews than dissimilar reviews because recipients tend to define ambiguous reviews in their minds as similar to the self. Moreover, they found that the persuasive impact of ambiguous reviews is significantly greater than that of dissimilar reviews but lower than that of similar reviews.
Although the study of Bruyn & Lilien (2008) was not conducted in an OCR but in a viral marketing setting, their findings are remarkable. They found that demographic homophily has a negatively effect on all three stages of the decision making process (i.e. awareness, interest and final decision). This finding is remarkable since previous discussed studies find positive effects. In the study, acquaintances send an unsolicited e-mail to an individual asking to participate in a survey and spread the word about it. This approach is quite impersonal (i.e. respondents do not really care since it is unsolicited). Furthermore, it offers respondents new information where they did not asked for (e.g. sending a unsolicited mail). Thus their involvement should be quite low and their responses quite functional rather than entertaining (e.g. there is nothing exiting about sending an e-mail). This could indicate that their responses were more utilitarian rather than hedonic. These factors contribute to the finding of the social network theory, which argue that weak ties (e.g. dissimilarity) are more effective in situations where new information is communicated (Granovetter, 1973). Furthermore, as we highlighted in the introduction (see 1.6), homophily tends to be less effective in an utilitarian contexts which may explain their findings.
2.2.2 The effect of demographic homophily on perceived risk
19
usefulness may further reduce subjective uncertainty and subsequently perceived risk. This findings is in line with the uncertainty reduction theory (Berger & Calabrese, 1975). Furthermore, based upon the liking principle (see 2.1.2), when someone’s demographic identity is perceived as similar, liking increases and uncertainty is likely to decrease. However, Gilly et al., (1998) found in their WOM study that messages from sources with similar demographic characteristics will generate more interest and are more influential in situations where a high degree of trust, intimacy and confidence are required. Higher perceived trust rates increases the integrity and intentions of the reviewer (Racherla et al., 2012; Mayer & Davis, 1999). More specifically, Mayer & Davis (1999) argue that trust and uncertainty are two ends of the same continuum in the sense that the higher the uncertainty, the lower the trust and vice versa. Therefore, it is plausible that demographic homophily reduces subjective uncertainty, and consequently reduce perceived risk. Or, in other words, demographic homophily has a negative effect on the degree of perceived risk.
2.2.3 Current findings on perceptual homophily (e)WOM communication
Perceptual homophily is, next to demographic homophily, the second dimension of the homophily concept. Paragraph 1.4 defined perceptual homophily as the degree to which people are similar regarding values, attitudes, beliefs and lifestyles (Lazarsfeld & Merton, 1954; Rogers, 1983; Brown & Reignen, 1987; McPherson, Lovin, & Cook, 2001; Gilly et al., 1998). Surprisingly, only few eWOM studies measured the effect of perceptual homophily. However, the OCR studies which took perceptual homophily into account, found positive effects. The study of Shen and Zhang (2016) found that recipients of OCRs discounted their own information and were more likely to imitate others when perceptual homophily was perceived. Furthermore, Bruyn and Lilien (2008) found in their viral marketing study that perceptual homophily positively affected recipients interest. Moreover, as discussed before, Racherla et al. (2012) found that the trust scores of OCRs increases in presence of perceptual homophily.
20
that individuals with similar values are more likely to trust each other which is consistent with the finding of Racherla et al. (2012).
2.2.4 The effect of perceptual homophily on the degree of perceived risk
21
2.2.5 The joint effect of perceptual and demographic homophily on perceived risk The interaction effect of perceptual homophily on the relation of demographic homophily on perceived risk should have the strongest negative effect. The homophily construct is defined as the degree to which people are similar (Rogers, 1983). Based on this definition it is arguable to state that the more similarities among individuals (e.g. demographic and perceptual homophily), the higher degree of perceived homophily and, consequently, the stronger the effect on perceived risk. This suggestion is supported by Rogers & Shoemaker (1971), who argue that homophily is most effective in the presence of both dimensions. Moreover, Fischer (1982) argue that the pattern of homophily is getting stronger when more similarities exists between two individuals. Furthermore, Rogers (1983 p.147) stated that the more homophilous individuals are, the more likely it is that their communication will be effective. Thus, based on previous discussed mechanisms, in the presence of both demographic homophily and perceptual homophily, individuals should perceive a lower degree of perceived risk than when individuals only perceive one of the two dimensions.
2.2.6 Other variables
To ensure high validity during the measurement of the effects of the independent variables on the dependent variables in both studies, some other variables are discussed which could covariate or clutter with the measured effects. First, a distinction can be made between individuals with a individualistic orientation and individuals with a collectivist orientation (Jain et al., 2007). Individualists define themselves as independent from others and focus on personal goals, values and motives whereas collectivists, define themselves as interdependent, rather than independent. Thus, individuals with an collectivist orientation could be more effected by homophily than individualists.
Second, and according to regulatory focus theory (C.f. Higgins, 1997), individuals with a promotion focus have the chronic or acute goal to focus on achieving gains and acquire positive outcomes. In contrast, individuals with a prevention focus are not so much motivated by achieving gains, but more by preventing losses and negative outcomes (Haws et al., 2010). Thus, this paper suggests that individuals with a preventing focus are more effected by risk reducing cues (e.g. homophily) than individuals with a promotion focus, since the former is actively trying to prevent losses. This paper controls for this variable.
22
individuals who are confident and often use reviews are accustomed to reviews, meaning that their trust in reviews is higher than individuals who are not often using reviews. Thus, individuals with a high OCR usage experience may perceive lower perceived risk rates. Finally, personal attitude towards OCRs may clutter the internal validity of the predicted effects between the independent and dependent variables (Lee, Park & Han, 2008). Individuals with a negative attitude towards OCRs may naturally have a higher degree of perceived risk, whereas individuals with a positive attitude towards OCRs may naturally have a lower degree of perceived risk. In addition, this paper controls for this variable.
2.2.7 Hypotheses and conceptual model study one
Based on the discussing in the preceding paragraphs, eight hypotheses are formulated. These hypotheses are depicted in figure 1:
H1. Perceived demographic homophily has a negative effect on the degree of perceived risk
H2. Perceived perceptual homophily has a negative effect on the degree of perceived risk
H3. The negative effect of perceived perceptual homophily is stronger on the degree of perceived risk than the negative effect of demographic homophily.
H4. The negative effect of demographic homophily on perceived risk becomes more negative in the presence of perceived perceptual homophily.
H5 The negative effect of homophily on the degree of perceived risk is weaker when individuals have an individualistic rather than an collectivist orientation
H6 The negative effect of homophily on the degree of perceived risk is stronger when controlled for individuals with a prevention and promotion focus
H7 The negative effect of homophily on the degree of perceived risk is stronger when controlled for individuals experience with OCRs
23
Figure 1: Conceptual model study one
2.3 Study two: the relative effects of demographic indicators on demographic homophily
Thus far we have discussed the effects of demographic homophily and perceptual homophily on perceived risk. However, as indicated in 1.4.2, the eWOM literature uses different indicators to measure demographic homophily. This paragraph is devoted to explain the relative effects, or strength, of each demographic indicator to demographic homophily. Within the research field of OCRs, five different demographic indicators are used to measure the construct of demographic homophily: Gender, age, occupation, location and name2 (Shen & Zhang, 2016; Rosario et al., 2016; Forman, 2008; Naylor et al., 2011; Racherla et al., 2012; Reichelt et al., 2014). We will discuss each characteristic separately and discuss their relative strength on demographic homophily.
2.3.1 Gender homophily
We define gender homophily as the similarity between people regarding their sex. The networks (e.g. friendships, families) of people are relatively gender-integrated compared to dimensions like race, age and education (McPherson et al., 2001). This is due to that men and women are roughly equal in numbers and are linked together in networks such as households and families. This causes similarities in, social class, residence and other characteristics (McPherson et al., 2001). While 22% of people does not have any cross-gender confidants,
2
24
37% of people have networks that are nearly perfectly mixed by gender. McPherson et al. (2001) argue that this pattern is misleading since these confidant networks contain family links, containing both genders. The study of Marsden (1987) controlled for family links and found that among family links, the heterogeneity of networks is very close. He also found that gender homophily is much stronger among non-family networks. After controlling for family, Marsden (1988) found that gender homophily is positive, but substantially weaker compared to other social dimensions such as race and education. This finding is consistent with the finding of Smit et al., (2014) who found that, although gender homophily is positive on demographic homophily, its effect is weaker than other demographic indicators.
Within the context of OCRs, online reviews of other consumers could be seen as a network which is consistent with the social network theory (see 2.1.4). When recipients read an OCR that includes gender homophily, it is likely that this positively affects recipients attitude and behaviour towards a product or service (McPherson et al, 2001).
2.3.2 Occupational homophily
We define occupational homophily as the degree of similarity between people regarding their work situation. Smit et al (2014) argue that occupational homophily affects perceived demographic homophily. Research found strong occupational homophily in strong ties like marriage (McPherson et al., 2001). However, research suggests that occupational homophily could be more important for relatively superficial ties. Verbrugge (1977) found that occupational homophily has a stronger effect on superficial ties compared to kin-ties.
2.3.3 Age homophily
25
strong differences in individuals interests. This strong effect could be due to that age homophilous ties are more close and perceived as more personal (Fischer,1982).
2.3.4 Location homophily
We define location homophily as the degree of similarity between people regarding their place of residence. Surprisingly, as far as we know, the literature did not measured the relative strength of location homophily compared to other demographic homophily indicators. However, the location indicator may have the strongest effect on perceived demographic homophily. In paragraph 2.2 the findings of the effect of demographic homophily were discussed. Recall that all studies found positive effects of demographic homophily but the study of Bruyn & Lilien (2008) found negative effects. When the demographic homophily constructs of these studies are further examined, it is remarkable to conclude that all studies inserted location homophily but the study of Bruyn & Lilien (2008) did not. The strong effect of geographical location is further supported by the OCR study of Forman (2008), who highlighted the important role of geographical location, as shared geographical location increases the relationship between disclosure and online product sales. Although correlation is no causation, and other differences between the study of Bruyn & Lilien (2008) and the other studies are present, like a different study context (i.e. viral marketing and OCRs) more findings in the WOM literature support the strength of location homophily on perceived demographic homophily. For instance, Smith et al., (2014) argue that location homophily will strongly affect the rate of homophily. Moreover, McPherson et al. (2001) argue that individuals are more likely to have contact with individuals who live closer in geographical location than individuals who live further away. This phenomenon can be explained due to that it takes more energy to connect and interact with individuals who live further away than those who already live nearby (Zipf, 1949). Moreover, Verbrugge (1983) found that location homophily is the strongest indicator of how often friends socialize.
26
Thus, sharing the same geographical location makes individuals feel closer with another which could increase tie strength and, consequently, have a greater effect on perceived demographic homophily. For example, unlike other demographic indicators, one can have thoughts and experiences about a geographical location. When the residence of individuals is similar, the experiences and thoughts can bond the individuals since it is likely that both have experienced the same. Based on previous discussing, we argue that location homophily has the strongest effect on perceived homophily.
2.3.5 Name homophily
Another interesting similarity-liking relationship for OCRs is the name-letter effect. The name-letter effect is defined as: the tendency to have a positive predisposition to the letters in one’s own name, especially the first and last initials (Nuttin, 1987; Fennis & Stroebe, 2016; Coulter & Grewal, 2014). When individuals perceive the same letter, individuals are reminded of themselves and tend to express greater liking towards those same letter related objects. The underlying mechanism of this effect is implicit egotism which refers to the concept that people have positive associations about themselves, and these associations can non-consciously transfer to another individual or object that they associate with themselves (Coulter & Grewal, 2014; Nuttin, 1987). Having positive associations with oneself can lead to perceived similarity (Nuttin, 1987). However, this transfer can only occur if implicit egotism cues are provided that enable the individual to make the self-object connection (Nuttin, 1987). Furthermore, Koole & Pelham (2003) found that personal names defines a individuals self-concept, and positively affect it. For example, an individual named ‘Liane’ should like and identify herself more with an individual named ‘Laurens’ than ‘Sanne’, since the name-letter effect is especially caused by the first (and last) initials. It is thus plausible that when individuals share the same letter in OCRs, they perceive themselves as more similar, which increases demographic homophily.
27
of communication and information processing (Price & Feick, 1984). When only the first letter of a name is the same, it is harder to process that information and perceive similarity with another individual compared to demographic indicators which are plenary similar.
2.3.6 Relative strength of demographic indicators to demographic homophily
So far, the discussed literature in 2.3 argues that all five demographic indicators affects demographic homophily. But their relative strength differs. Based on the findings of Marsden (1987) and Smit et al (2014), it is plausible that the effect of gender homophily to demographic homophily is weaker compared to other demographic indicators. Research showed that the effect of occupational homophily tends to be slightly stronger than the effect of gender homophily but weaker than other demographic indicators (Verbrugge, 1977; Smit et al., 2014). Age homophily tends to have a stronger effect to perceived demographic homophily than occupational and gender homophily (Smith et al., 2014; Hsu, Hackett & Hinkson, 2014; McPherson et al., 2001). Moreover, based on the discussion in 2.3.4 and 2.3.5, location homophily tends to have the strongest effect to perceived demographic homophily whereas name homophily tends to have the lowest effect respectively.
9.Location homophily has the strongest relative effect on perceived demographic homophily, followed by age, occupation, gender, and name homophily respectively.
This hypothesis is graphically depicted in figure 2.
28
3. RESEARCH DESIGN
This chapter discusses the choice of research, the method for data collection, the manipulation, the operationalization of the constructs and the plan for analysing the data respectively.
3.1 Choice of research
As discussed in 1.6, this paper conducts two studies using the same set of respondents. Study one measures the effect of perceptual and demographic homophily on the degree of perceived risk. Study two measures the relative strength of demographic indicators on perceived demographic homophily. Both studies have their own dependent variable and their own independent variables as discussed in paragraph 2.2 and 2.3. Moreover, each dependent variable requires a different analysis to measure the stated hypotheses stated in chapter two. Hence, in the following paragraphs, both studies are discussed separately. Study one is a non-disclosed study, meaning that respondents are not aware of the purpose of the study (Aronson, Wilson & Brewer, 1998). In study two, respondents may be aware of the purpose of the study. Respondents are first treated to study one and afterwards to study two.
3.1.1 Experimental design study one
An experimental design is chosen to determine the causal relationship between the variables and, consequently, to measure the hypotheses. This experiment uses a 2 (demographic homophily: dissimilar, similar) x 2 (perceptual homophily: dissimilar, similar) between-subject design (see table 1) to measure the effects of demographic and perceptual homophily on the degree of perceived risk (Aronson et al., 1998). Each respondent is treated to one of the four conditions of perceived risk in study.
Table 1: Overview of conditions study 1 3.1.2 Conjoint analysis study two
To measure the relative strength of each demographic indicator on perceived demographic homophily, a conjoint analysis is conducted. According to Louviere & Woodworth (1983), in a conjoint analysis, the utility of a product (perceived demographic homophily in OCR) equals the sum of the utilities of the attribute levels (demographic indicators). Thus, the level
Demographic homophily
Perceptual homophily
Similar Dissimilar
Similar Condition 1 Condition 2
29
of utility in the construct depends on the level of utility in the attributes. This indicates a formative scale. This study uses a decompositional choice-based conjoint-analysis which consists of five attributes with each two levels (Louviere & Woodworth, 1983). These five attributes are: gender, name, age, location and occupation (see 2.3). In a full factorial design, this implies that each respondent has to view and rate 32 different choice alternatives, where a choice alternative refers to a unique setting of similar or dissimilar demographic indicators. For example, one choice alternative could be `age similar, occupation dissimilar, residence similar, gender similar, name dissimilar´. However, rating 32 choice alternatives is quite exhausting. Therefore, the choice alternatives are reduced to eight alternatives, with a fractional factorial design depicted in table 2 (Louviere & Woodworth, 1983).
Choice Alternative Age Name Gender Occupation Location
1 Dissimilar Dissimilar Similar Similar Dissimilar
2 Similar Similar Dissimilar Dissimilar Dissimilar
3 Dissimilar Similar Dissimilar Similar Similar
4 Dissimilar Dissimilar Dissimilar Dissimilar Similar
5 Similar Dissimilar Dissimilar Similar Dissimilar
6 Similar Similar Similar Similar Similar
7 Dissimilar Similar Similar Dissimilar Dissimilar
8 Similar Dissimilar Similar Dissimilar Similar
Table 2: Overview fractional factorial design study 2
Note that each choice alternative is compared to another choice alternative in a choice set. When presented with a choice set, respondents will select one of the two choice alternatives with which they feel most similar. Thus, to fully test the conjoint-model, each respondent is presented with a total of four choice sets each consisting of two choice alternatives. For an efficient design, each attribute level is displayed an equal number of times and each attribute level combination appears an equal number of time. Furthermore, each choice alternative is matched with its opposite choice alternative for minimal overlap between the attribute levels in a given choice set (Louviere & Woodworth, 1983). This way the conjoint-analysis will have a balanced and orthogonal design (Huber & Zweringa, 1996). The choice sets, each including two choice alternatives are depicted in table 3.
Choice set Choice alternative Choice alternative
1 1 3
2 2 4
3 5 7
4 6 8
30
3.2 Population and sample
Since the measurement of the both dependent variables are independent of each other, each respondent is treated to both studies. Hence, in this paper, there are four different conditions. Each respondent is treated to one of the four conditions of study one, as well as to the same choice sets in study two. Respondent are randomly assigned to one of the conditions in study one. To make sure that the effects in this study are reliable, at least 30 respondents are needed for every condition (Malhorta, 2009). Thus, a minimum of 120 respondents is needed. This number of respondents is also sufficient for the conjoint analysis (Malhorta, 2009).
Within 12 days, 178 respondents were collected. 146 respondents remained after deleting those respondents who did not finished the survey. The sample consists out of 55.5% females and 44.5% males. The average age of all respondents was 33 years with a standard deviation (S.D.) of 13,13. Whereas the age of the youngest participant was 15 and the age of the oldest respondents was 69. Most respondents have a fulltime job (38,5%), followed by a parttime job (31,5%), student (26%) and no job (4.0%). 58,9% of the respondents preferred an active vacation, while 41,1% preferred a relax vacation. 68,5% of the respondents attaches more value to breakfast compared to dinner (31,5%). Moreover, 54,1% prefers a far travel as a holiday, followed by a sun vacation (30,8%), city trip (10,3%) and weekend away (4,8%). Finally, 67,8% of the respondents is most concerned when the hotel is not clean, followed by not having a comfortable bed (27,4%), not having a the hotel service as expected (3,4%) and lastly, not having a luxurious bathroom (1,3%). The results are depicted in table 4.
Variable Classification Mean (S.D.) Sample (%)
Age In years 33 (13,13)
Gender Male 44.5
Female 55.5
Occupation Fulltime job 38.5
Parttime job 26.0 Student 31.5 No job 4.0 Vacation Active 58.9 Relax 41.1 Food Dinner 31.5 Breakfast 68.5
Trip Far travel 54.1
31
City trip 10.3
Weekend away 4.8
Most concerned Hotel not clean 67.8
Comfortable bed 27.4
Hotel service 3.4
Bathroom 1.3
Table 4: Summary sample descriptive
A One-way ANOVA shows that gender does not significantly differs across the four conditions. In other words, the variation between male and female does not significantly differ between conditions of study one.
When analysing the distribution of age, an ANOVA shows that the mean of age significantly (p=,028) differs between the four conditions of study one. Subsequently, a post-hoc test reveals that the age of respondents which were treated to the demographic homophily condition (condition 3, table 1) is significantly higher compared to respondents treated to conditions one (p=,049), two (p=,004) and four (p=,026). To test whether age has a significant effect on the dependent variable, the degree of perceived risk, a regression analysis is conducted. The results show that age has a significant positive (p=,052) effect of ,013 on perceived risk. Thus, when age is increased with one year, the degree of perceived risk is increased with ,013 points3.
3.3 Data collection
Due to the quantitative data approach, this study used an online survey to collect the data. This survey is digitalized using Qualtrics Survey Software. The software allows respondents to conduct the survey on their smartphones, tablets or computer. Through a hyperlink, respondents can get access to the survey, which is distributed on social platforms like Facebook and Linkedin. The data collection took place in week 17,18 and 19 (2017). The survey is conducted in The Netherlands. Thus, there is a high probability that most respondents are Dutch. Therefore, the survey is in the Dutch language.
3.4 Manipulation
This paragraph discusses the manipulation of demographic homophily and the manipulation of perceptual homophily.
3 The effect of age on perceived risk is not significant anymore when it is added as a control variable to the
32
3.4.1 Manipulation of demographic homophily
To manipulate demographic and perceptual homophily such that every respondent is matched with dissimilar or similar demographic and/or perceptual traits, depending on the treated condition (see table 1), this study makes use of the ‘piped text insertion’ function in Qualtrics. In sum, this function offers the opportunity to make the stimuli dynamic. Thus, specific text parts of the stimuli (e.g. demographic indicators) are automatically matched (e.g. dissimilar, similar) in accordance with the respondent and the condition the respondent is randomly treated to. Thus, a respondent of 24 years who is a male, lives in Groningen and is a student, will be matched in condition one and three (see table 1), with a stimulus which is similar to these traits due to the use of ‘piped text insertion’. Similarly, a female who lives in Hardenberg, has a fulltime job and is 54 years will automatically be matched with these similar demographic traits when she is treated to either condition one or condition three.
In order to show dissimilar demographic indicators, respondents will see the opposite gender to control for the gender indicator. To manipulate dissimilar occupation, respondents who indicated to be a student or indicate to have no job or a parttime job, will see an OCR of an individual who has a fulltime job. In contrary, respondents who indicate to have a fulltime, parttime or no job will see an OCR of an individual who is a student. Furthermore, a ´fake´ residence name is used to show a dissimilar residence. In addition, to manipulate for the first-letter effect, respondents will see a name with a different first first-letter than that of the respondent. Note that this study uses unisex names (i.e. names which can be used for both female and male) in order to conceal gender in the name indicator. To control for dissimilar age, the same manipulation as Naylor et al. (2011) is used which means that there is a difference of 18 years between the respondent and the OCR.
3.4.2 Manipulation of perceptual homophily
33
perceptual homophily, the perceptual traits are similar. To control for dissimilar perceptual traits, the opposite of 1 and 2 is displayed and the least preferred option is displayed for 3 and 4. In order to match respondents with a similar or dissimilar stimuli using the piped text insertion, respondents first indicate their gender, age, residence, occupation, their first letter of their names and their likes, attitudes and lifestyles. Consider the following example for study one. A respondent who is treated to condition 1 (see table 1), is 30 year old male, lives in Groningen, has a fulltime job, the first letter of his first name starts with a ‘R’, prefers breakfast, an active vacation, a clean hotel and a city trip, will see the stimulus presented in figure 34.
Figure 3: Example condition one study one
4 The stimulus is in the Dutch language and is translated here. Note that the hotel name is fictitious. 20-4-2017 – The review of Robin, 30 years old, male, occupation: has a fulltime job, residence: Groningen. During my holiday I have been a week in the Crescent Park hotel and I am very satisfied about it!
34
Furthermore, the same respondent could be, based on the same demographic traits as in study one, treated in study two to a choice set presented in figure 45.
Figure 4: Example choice sets study two 3.5 Operationalization of study 1
This paragraph will discuss the questions that will be used in this paper for study one. More specifically, the paragraph will discuss the operationalization of the degree of perceived risk, the manipulation checks and the control and covariate variables.
3.5.1 Operationalization of the degree of perceived risk
Several studies have measured the construct of perceived risk, using different items (Stone & Gronhaug, 1993; Inmann, 2001; Erem, Tulin & Swait, 2004; Kim, Ferrin & Rao, 2009). This study uses the construct of Stone & Gronhaug (1993), given the manipulation and the wide adoption of this construct by other studies which measure perceived risk in an online setting. The construct of perceived risk includes five items on a seven-point Likert scale (1 = strongly disagree, 7 = strongly agree) which are stated in table 56.
3.5.2 Manipulation check
After the exposure to one of the four conditions (see table 1), several questions will be asked to check whether the respondent perceives demographic homophily and/or perceptual homophily. Given the wide use in the homophily literature, the homophily scales of
5 The choice set are in the Dutch language and are translated here.
Left review: The review of Xin, 58 years, male, occupation: has a fulltime job, residence: Langeroord.
I have had a great experience with this hotel! It is a little bit decadent for someone who lives in Langeroord, but it is quite affordable. I have a fulltime job in my private live, and as a 58 year old male it is delightful to relax and to enjoy the huge hotel rooms and excellent hotel service. I should really recommend it!
Right review: The review of Robin, 58 years, female, occupation: has a fulltime job, residence: Groningen.
I have had a great experience with this hotel! It is a little bit decadent for someone who lives in Groningen, but it is quite affordable. I have a fulltime job in my private live, and as a 58 year old female it is delightful to relax and to enjoy the huge hotel rooms and excellent hotel service. I should really recommend it!
6
35
McCroskey & Richmond (1975) and that of McCroskey et al. (2006) are used for the manipulation check of the manipulation of perceptual and demographic homophily. The demographic homophily construct consists of four items which are measured on a seven-point Likert scale (1= strongly disagree, 7 = strongly agree) which are stated in table 5. However, since this paper consists out of two studies and given the large amount of questions already presented in the survey, this paper has combined the four items into one item to check for manipulation of demographic homophily (see table 5).
The construct of McCroskey & Richmond (1975) is used to measure perceptual homophily given its wide use in the literature. This construct includes four items on a seven-point Likert scale (1 = strongly disagree, 7 = strongly agree) which are stated in table 5. This paper uses all four items but IT4, since this item does not fits well in the context of this paper and due to the large extent of questions. This can be done since the construct of perceptual homophily is formative (McCroskey & Richmond, 1975).
3.5.3 Control variables and Covariates
In the survey, after the randomized stimulus of study one, questions will be include to measure the collectivist and individualist orientation of the respondent, which might covariate with the perceived demographic and perceptual homophily and their joint effect on perceived risk (see 2.2.6). In addition, questions are asked to measure the prevention and promotion focus (Higgins, 1997; Haws et al. 2010), the personal attitude of respondents towards OCRs (Lee, et al.,2008) and the experience of the respondents with OCRs (Murray & Schlacter, 1990) such that they can be controlled for.
3.6 Operationalization of study 2
In study two, all choice alternatives are exactly similar to each other but for demographic traits (see figure 4). Therefore, the dependent variable, perceived demographic homophily, is operationalised by asking respondents which of the two reviewers is most similar to them, when presented with a choice set.
Construct (source) Scale Items α
Degree of perceived risk (Stone & Gronhaug,1993) Dependent variable study one 7-point Likert scale
IT1. The thoughts of purchasing this item causes me to be concerned with experiencing some kind of loss if I went ahead with the purchase.
IT2. All things considered, I think I would be making a mistake if I purchased this item.
IT3. I really feel that purchasing this item poses problems for me that I just don´t need.
IT4. If I purchase this item for myself, I would be concerned
.854
36
that I really would not get my money’s worth from this item.
IT5. If I were to purchase this item, I become concerned that this item will not provide the
level of benefits that I would be expecting.
.845 Demographic homophily McCroskey & Richmond (1975), McCroskey, et al., (2006) Dependent variable study two 7-point Likert scale
Original IT1. The person who wrote the review shares a similar background, IT2. The person who wrote the review is culturally different from mine IT3. The person who wrote the review shares the same social class IT4. The person who wrote the review has a different economic situation like mine
Used in this paper for manipulation check study one: IT1. The person who wrote this review has a similar background (residence, age, occupation and gender) as me.
Used in this paper for dependent variable study two: “which of these two reviewers is the most similar to you?”
n/a n/a n/a Perceptual homophily McCroskey & Richmond (1975) 7-point Likert scale
IT1 The person who wrote the review thinks like me IT2 The person who wrote the review does not share my values
IT3 The person who wrote this review has similar likes and dislikes
IT4 This person expresses attitudes different from mine
.453 .360 .902 n/a Collectivist/ individualist (Jain et al., 2007) 7-point Likert scale
IT1. I often do ‘my own things'.
IT2. I am a unique person, separate from others.
IT3. I often sacrifice my self-interest for the benefit of my group.
IT4. It is important to me that I respect the decisions made by my groups. n/a n/a n/a n/a Promotion/prevention focus (Haws et al., 2010) 7-point Likert scale
IT1. When I get something I want, I feel excited and energized.
IT2. I worry about making mistakes.
n/a n/a Attitude towards
OCRs
(Lee, Park & Han, 2008)
7-point Likert scale
IT1. The reviews presented on the website are helpful for my decision-making.
IT2. The reviews presented on the website make me confident in purchasing the product.
IT3. I always read reviews presented on the website.
.870 .804 .865 Usage experience (Murray & Schlacter, 1990) 7-point Likert scale
IT1. I have a great deal of experience with reviews. IT2. I am very confident in using reviews.
IT3. I have used or been exposed to reviews in the past.
.810 .837 .842 Table 5: Overview of items per construct
3.7 Factor analyses
