THE REVIEWERS REVIEWED: HOW EXISTING REVIEWS
INFLUENCE THE REVIEW PROCESS OF SHORT MOVIES
by
EMILY TIMMERMAN
University of Groningen
TABLE OF CONTENT
1. INTRODUCTION ...4
2. LITERATURE REVIEW ...7
2.1 Characteristics of word of mouth ...7
2.2 Behavioral consumer responses ...8
2.3 Review posting behavior ...9
2.4 Sources of online reviews... 10
2.5 Social influence in review posting behavior ... 13
3. HYPOTHESIS AND CONCEPTUAL MODEL ... 15
3.1 Level of positivity of existing reviews ... 15
3.2 Source of the review ... 16
3.3 Consumers’ product involvement ... 17
3.4 Consumers’ degree of optimism... 18
3.5 Consumers’ degree of extraversion ... 18
3.6 Conceptual model ... 19 4. RESEARCH METHODS ... 21 4.1 Data collection ... 21 4.2 Data analysis ... 26 5. RESULTS... 27 5.1 Socio-demographic characteristics ... 27 5.2 Cronbach’s Alpha ... 29 5.3 Normality test ... 30
5.4 Analysis of variance - ANOVA ... 31
5.5 Multiple regression ... 34
6. DISCUSSION ... 38
6.1 Effects of independent variables ... 38
6.2 Effects of moderators ... 39
6.4 Managerial implications ... 41
6.5 Limitations and further research ... 42
REFERENCES ... 44
4
1. INTRODUCTION
In the past, marketing communication efforts were first picked up through opinion leaders, which in turn pass on the message to the less active consumers. In this system, word-of-mouth (WOM) played a very important role. Nowadays, the use of the Internet has changed a lot in this WOM concept (Bronner & de Hoog, 2012). Since consumers are able to share information online with a lot of people, friends and family but also people they don’t know and people from other countries, the range of WOM or electronic word-of-mouth (eWOM) became much wider (Burton & Khammash, 2010; Chen & Xie, 2008). Online reviews are closely aligned to eWOM.
In recent years, the amount of available online content increased dramatically due to the popularity of user generated content (Goldenberg, Oestreicher-Singer & Reichman, 2012). User generated content exists in many different forms. Consumers can create content among others through social websites as YouTube, Facebook, Twitter, and MySpace, and through blogs or discussion forums (Cheong & Morrison, 2008; Christodoulides, Jevons & Bonhomme, 2012). However, the main form of user generated content is online consumer reviews, in which consumers quantitatively assess their experience with the product or service and/or quantitatively rate the product or service online (Bronner & de Hoog, 2012; Shridhar & Srinivasan, 2012).
Online consumer reviews have become more important due to the still growing popularity of the Internet (Zhu & Zhang, 2010; Cheong & Morrison, 2008). Previous studies define consumer reviews in various ways. Mudambi and Schuff (2010, p. 186) define it as ‘peer-generated product evaluations
posted on company or third party websites’. Another definition, by Park, Lee and Han (2007, p. 125) is ‘an online consumer review is new information presented from the perspectives of consumers who have purchased and used the product. It includes their experiences, evaluations, and opinions.’
Benlian, Titah and Hess (2012, p. 238) define consumer reviews as ‘reviews written by consumers
about the quality of products based on personal experiences with the products’. Combining these
definitions results in the below mentioned definition of consumer reviews.
5 Consumer reviews often are supplemented by product ratings. This means that besides the qualitative assessments in the form of a short text, consumers also have the possibility to rate the product (Mudambi & Schuff, 2010; Sridhar & Srinivasan, 2012). Consumers can rate products usually on a scale of one to five, which provides a quick overview of the overall score of the product.
Online consumer reviews are out of control of companies (Shridhar & Srinivasan, 2012). Consumers write and share it on the Internet and firms cannot change or remove that content. Before a consumer reviews a product, usually there are already other reviewers who have evaluated the product, which might have an influence (Shridhar & Srinivasan, 2012). According to several studies, online reviews will have an influence on the online purchase decisions of consumers (Ludwig et. al, 2013; Shridhar & Srinivasan, 2012; Bronner & de Hoog, 2010). Depending on whether the reviewers are positive or negative about the product, the influences on consumers’ purchase decisions can also both be positive or negative. According to Cheong and Morrison (2008), negative reviews can also have harmful consequences for the brand equity. Online consumer reviews will also positively influence the trustworthiness (Ludwig et. al, 2013).
Consumers can have different reasons for writing online reviews, such as: (1) their desire for social interaction, (2) their desire for economic incentives, (3) consumers’ concern for other consumers, and (4) the opportunity to increase their self-worth (Hennig-Thurau et al., 2004). Since two of the four reasons are focused on a social aspect and helping others, it implies that the social factor is important in writing online reviews. Besides having social reasons to write a review, reviewers may also be socially influenced by others in writing their reviews. Sridhar and Srinivasan (2012) conclude that consumers weaken their extremity of reasoning due to other reviews; they rate less positive when other ratings are higher and less negative when other ratings are lower.
The degree in which consumers are influenced by existing reviews will also depend on the source of the existing reviews. Reviews provided by consumers or third-parties (experts) are for example often considered as being more credible and less biased than information that is provided by marketers of the company (Akdeniz, Calantone and Voorhees, 2013).
6 The central question in this research is:
To what extent are consumers in their review posting behavior influenced by the level of positivity of existing online reviews and do the source of the reviews, product involvement, degree of optimism and degree of extraversion have moderating effects on this relationship?
In marketing literature, a lot of studies have been done on online reviews and its effects. There are also a number of studies that researched the social aspect of online reviews. Which means the way in which reviewers are socially influenced by other reviewers in the reviews they write or the ratings they give. However, much less is known about the effects of different review sources in the social influencing of reviewers. Also product involvement, optimism and extraversion are, as far as I know, not yet examined in combination with the social influence of reviews and review sources. Therefore, this study will contribute to the academic literature on that area.
This study will also be relevant for retailers, since it is very useful to know in which ways consumers are socially influenced in their evaluations of products/services. It also provides insights for retailers on how to use or anticipate on online reviews (Sridhar & Srinivasan, 2012).
7
2. LITERATURE REVIEW
In this section existing research is used to bring together opinions, concepts and theories about online reviews. The first section describes characteristics of word of mouth. The following sub-section discusses existing research on behavioral consumer responses. The third sub-sub-section is about review posting behavior and the subsequent sub-section is about sources of online reviews. The last sub-section discusses the social influence in review posting behavior.
2.1 Characteristics of word of mouth
Some of the first studies on the effect of word of mouth (WOM) conclude that WOM is a very important source of information. In several situations, consumers are more influenced by WOM than by mass media or personal selling (Engel, Blackwell & Kegerreis, 1969; Katz and Lazarfeld, 1955). Correspondingly, Allsop, Bassett and Hoskins (2007) state that WOM is one of the communication channels that is most influential. According to them, results of a national U.S. survey show that consumers find WOM and recommendations from friends, family, colleagues or fellow students most important in their buying decisions for some everyday food products. Gruen, Osmonbekov and Czaplewski (2006) found that participants of electronic word of mouth (eWOM) have comparable motivations as WOM participants. This may mean that the effects of eWOM are very similar to the effects of WOM. eWOM is defined by Hennig-Thurau et al (2004, p. 39) as: ‘any positive or negative
statement made by potential, actual, or former customers about a product or company, which is made available to a multitude of people and institutions via the Internet’. eWOM is more permanent
than WOM and this longer life span and the wide range of opinions attracts readers (Burton & Khammash, 2010).
8
2.2 Behavioral consumer responses
Consumers can have different behavioral responses as a result of exposure to online reviews. This section will discuss existing research examining behavioral responses in relation with online reviews.
Several studies examined the effect of online reviews on purchasing. Chevalier and Mayzlin (2006) found a positive effect between online reviews and purchasing behavior for online bookstores Amazon.com and Barnesandnoble.com. They also found that both websites provide more positive than negative reviews, that the influence of negative reviews is bigger than the influence of positive reviews and that consumers prefer to read the review text rather than summary statistics. Another study of Dellarocas, Zhang and Awad (2007) investigated the movie market and also found a positive effect between online reviews and purchasing behavior. Also Riegner (2007) found a link between online reviews and the purchase decision. She also stated that online reviews are more likely to influence online purchases than offline purchases.
Besides the effects of online reviews on really purchase behavior, there are also found effects on consumers’ purchase intention. Positive reviews will have a positive effect on purchase intention, while negative reviews will reduce the purchase intention (Lin, Luarn & Huang, 2005; Park, Lee & Han, 2007). Xue and Zhou (2011) contribute to this by finding that consumers tend to develop stronger brand interest and purchase intentions after reading positive reviews.
Another behavioral consumer response, referral intention, is researched very limited in combination with online reviews. Referral intention means that consumers are likely to spread word of mouth about a certain product or service (Ryu & Feick, 2007; Biyalogorsky, Gerstner & Libai, 2001). In other words, it is about recommending the discussed product or service to others. What is only found about online reviews in combination with referral intention, is that consumers have a higher intention to refer negative reviews (Xue & Zhou, 2011). Since this research is very limited, there is a gap in the existing research with respect to the effects on referral intention.
9
2.3 Review posting behavior
Several studies have been done on the review posting behavior of consumers (Chen, Fay & Wang, 2011; Punj, 2013). For example, Chen, Fay and Wang (2011) examined the relationships between review posting behavior of consumers on social media and marketing variables such as product quality and price. They found that marketing variables have an effect on online consumer posting behavior and that this relationship differs for different levels of consumers’ Internet experience. Another study, of Punj (2013), investigated whether consumers who search online for product information will be more likely to post online reviews themselves. In particular, he examined the differences in characteristics of consumers who post online reviews after conducting online research and consumers who do not post online reviews. He found that older, higher educated women are more likely to post online reviews, while younger men with a moderate education are less likely to post online reviews. Consumers with higher incomes will search more for product information online than consumers with lower incomes, but it seems that consumers with higher incomes are less likely to post reviews themselves (Punj, 2013).
Research of Hennig-Thurau et al. (2004) found that the main reasons for eWOM behavior of consumers are: (1) their desire for social interaction, (2) their desire for economic incentives, (3) consumers’ concern for other consumers, and (4) the opportunity to increase their self-worth. eWOM behavior means in this case that consumers share their opinions and experiences with products and services on platforms on the internet. In other words, with eWOM behavior Hennig-Thurau et al. (2004) refer to review posting behavior. From the four reasons it becomes clear that the social aspect is very important for consumers in their review posting behavior, since two of the four reasons emphasize the social aspect. These two reasons focusing on the social aspect are consumers’ desire for social interaction and their concern for other consumers.
10 Sridhar and Srinivasan (2012) used online ratings and reviews of hotels in Boston and Honolulu on a third-party travel website to examine the effects of social influence in online ratings. They studied the effects of (1) positive product experience, (2) negative product experience, (3) product failure, and (4) product recovery as a conditional factor on product failure, on consumers’ online ratings. As a moderating effect, they used other consumers’ online ratings, to find out if social influence plays a role in online ratings. From their study it can be concluded that the positive effect of a positive product experience on a consumer’s online rating will become weaker when ratings of other consumers are higher. The same counts for the negative effect of a negative product experience on a consumer’s online rating, which will also become weaker when other consumers give higher ratings. In other words, when online ratings of other consumers increase, online consumer ratings will become less positive for a positive product experience, or less negative for a negative product experience. They also found that product failure has a larger negative effect on the online rating when ratings of other consumers increase and that this effect is weakened when the quality of product recovery increases.
Research of Schlosser (2005) investigated consumers’ publicly and private responses to product evaluations of others. She found that online consumer ratings will be less favorably after exposure to negative reviews of other consumers than after exposure to positive reviews. These online ratings will not be different when exposure to a positive review is compared with exposure to no review. Another finding of this study is that consumers who post reviews are more negatively influenced in their online rating by negative reviews of others than consumers who not post reviews.
Another research on the social part of online ratings examined the impact of social dynamics in the area of ratings on consumers’ rating behavior and product sales (Moe & Trusov, 2011). They looked at the effects of the valence and variance of reviews of other consumers on rating behavior. This research concludes that consumers are inclined to increase their online rating when the ratings of other consumers are lower. Although ratings of other consumers influence ratings behavior and ratings behavior can improve sales, Moe and Trusov (2011) state that these effects are of short duration due to indirect effects.
2.4 Sources of online reviews
11
Consumers. Online reviews provided by consumers belong to user generated content, which
is as the name says online content generated by consumers. In recent years, the amount of available online content increased dramatically due to the popularity of this user generated content (Goldenberg, Oestreicher-Singer & Reichman, 2012). Online consumer reviews is the main form of user generated content. Consumers quantitatively assess their experience with the product or service and/or quantitatively rate the product or service online (Bronner & de Hoog, 2012; Shridhar & Srinivasan, 2012). Reviews provided by consumers are based on the personal experience consumers had by using the product. This experience is very depending on preferences of consumers and the usage situation in which the consumer used the product (Chen & Xie, 2005). Online reviews are often affective and personally relevant and do not often include critical evaluations of technical aspects or something (Chakravarty, Liu & Mazumdar, 2010). For example for movie reviews, online consumer reviews will not include evaluations about the technical or artistic aspects of the movie. The tone of a consumer review is either positive or negative (Chakravarty, Liu & Mazumdar, 2010).
Companies. Online reviews provided by companies belong to company generated content.
Company generated content in online content, provided by the company through official sources such as the corporate website, adopted affiliates and press releases (Lee, Kim & Chan-Olmsted, 2011). Examples of company generated content are official product information on the corporate website or promotional materials as online advertisements (Cheong & Morrison, 2008; Ward & Ostrom, 2003).
Experts. Online reviews provided by experts are often called third-party reviews. These
12 expert reviews have become widely available. Therefore they can have a considerable effect on the success or failure of products (Chen & Xie, 2005).
Comparing sources. According to Akdeniz, Calantone and Voorhees ( 2013), reviews written
by consumers or third-parties are often considered as being more credible and less biased than information provided by (marketers of) the company. They state that consumers or third-parties, which are independent and reliable, do not have a reason to give a product a more positive review or higher rating than it really deserves. Companies do have (commercial) reasons for this. Woodruff (1972) contributed to this by stating that neutral sources and opinions of consumers are more likely to decrease uncertainty about quality perceptions than information provided by marketer-controlled sources. Lee, Kim and Chan-Olmsted (2011) just found that official brand websites and third-party sources were perceived as more credible than e-retailer sites and personal blogs. These three studies (Akdeniz, Calantone and Voorhees 2013; Woodruff 1972; Lee, Kim and Chan-Olmsted) belong to the few studies that mentioned both consumers, companies and experts. Reviews written by experts is the underlying factor in this: a lot of studies only focus on the comparison of reviews written by companies and consumers and there are a small number of studies discussing the comparison of expert and consumer reviews.
13 information is consulted more frequently than user generated information in the vacation decision process. A worldwide research of Nielsen (2007) conclude that consumers have the same level of trust towards online consumer opinions as towards brand websites.
As mentioned, about the comparison of consumer reviews and expert reviews, much less studies are available. Chakravarty, Liu and Mazumdar (2010) investigated the persuasive influence of consumer reviews and critical reviews (expert reviews) on the evaluations of to be released movies by moviegoers. They found that frequent moviegoers are more influenced by expert reviews, while for infrequent moviegoers consumer reviews are more relevant in their evaluation. This is primarily due to the fact that infrequent moviegoers attack more importance to the comments of regular moviegoers expressing the ‘usual’ taste, than experts who express ‘elite’ taste. Research of Kim, Park and Park (2013) investigates the impact of consumer reviews and expert reviews on the box office revenues of movies in the U.S. domestic market as well as in the international markets. They found that experts may focus more on artistic values than on values such as fun factor which are more important to the average consumer. Because of the better knowledge of experts, it may also be the case that their reviews are biased. The study of Kim, Park and Park (2013) conclude that consumer reviews have a larger impact on box office performances than expert reviews, both in the U.S. domestic market and in the international markets. The research of Woodruff (1972) found that neutral sources (to which expert reviews belong) are more likely to decrease consumers’ uncertainty regarding product quality than consumer-driven sources. Lee, Kim and Chan-Olmsted (2011) contribute to this by stating that third-party websites are perceived as a more credible product information source than personal blogs. They expect that this effect is due to the information driven focus of third-parties and the personality-driven focus of personal blogs.
2.5 Social influence in review posting behavior
14 Since consumers have different degrees of involvement for various products and high involvement products are of more personal relevance for consumers than low involvement products (Petty, Cacioppo & Schumann, 1983), product involvement may play a role in consumers’ review behavior. Therefore, consumers’ product involvement is also included in this study. Besides product involvement, consumers have also different levels of optimism and extraversion. These factors may also influence consumers’ review behavior and are therefore also included.
15
3. HYPOTHESIS AND CONCEPTUAL MODEL
In this section, the research of this study will be further clarified. First, hypothesis are formed with help of existing research and after that the conceptual model is displayed.
3.1 Level of positivity of existing reviews
The level of positivity of existing reviews is used as an independent variable in this research. The focus is on different levels of positivity of reviews (high, moderate and low). As mentioned in the literature review, existing research on the influence of reviews of others on the review posting behavior have different conclusions. Sridhar and Srinivasan (2012) conclude that higher ratings in reviews of other consumers weaken the reviews of consumers in their positivity or negativity. According to Schlosser (2005), consumers will be less positive in their reviews when they are confronted with negative reviews than with positive reviews. They also state that there is no difference when consumers are confronted with positive reviews compared with no reviews. Moe and Trusov (2011) conclude in their study that consumers write more positive reviews or give higher ratings when they are confronted with lower ratings or negative reviews.
The studies of Sridhar and Srinivasan (2012) and Moe and Trusov (2011) conclude a kind of weakening effect, they state that consumers are less extreme in there reviews after reading other reviews. This effect is referred to by Simonson (1989) as the compromise effect. It implies that consumers are more likely to choose middle options than extreme options in a selection set. In this case of online reviews, middle options are that consumers write less negative or less positive reviews. Extreme options in this case are the very positive or very negative reviews.
Thus, from existing studies mixed evidence is obtained. With help of the compromise effects, it can be expected that a weakening relationship exist between the level of positivity of existing reviews and consumers’ review posting behavior. In addition, also a majority of the studies conclude a weakening effect between the two variables. This results in the following hypotheses:
H1a1: There is a positive correlation between existing reviews and given reviews.
16
H1b: When existing reviews are high, given reviews are somewhat lower and when existing reviews
are low, given reviews are somewhat higher.
Existing reviews are existing online reviews and given reviews are the online reviews that consumers write. From now on the terms existing and given reviews will be used.
3.2 Source of the review
As described earlier, consumers react different on different sources of reviews. The sources of reviews that are taken into account in this study are consumers and experts, company reviews are not. Consumer and expert reviews are independent and can be unflattering about a product, while company reviews are obvious not independent and will always be complimentary about the product (Chakravarty, Liu & Mazumdar, 2010). Therefore, and also because there is less existing research on the comparison of consumer and expert reviews, only these two sources were taken into account.
As mentioned in the literature research, the study of Kim, Park and Park (2013) state that consumer reviews have a larger impact on movies’ box office performance than expert reviews. This means that consumers in this case are more influenced by other consumers, which suggests that consumers will also be more influenced by other consumers in their evaluation of movies. Chakravarty, Liu and Mazumdar (2010) found that frequent moviegoers are more influenced by expert reviews, while for infrequent moviegoers consumer reviews are more relevant in their evaluation. Lee, Kim and Chan-Olmsted (2011) found that consumers perceive third-party reviews as a credible product information source, while personal blogs (written by consumers) score much lower on credibility. Another study, of Woodruff (1972) contribute to this by stating that neutral third party sources (such as expert reviews) will significantly decrease consumers’ uncertainty about product quality. He also state that neutral sources are more likely to reduce uncertainty than consumer driven sources.
Taken these results together, it gives a bit of contradictory view. However, since most of the studies state that consumers are more influenced by experts than by consumers, it is expected that this will be true. This results in the following hypotheses:
H2a: The degree of positivity of reviews that consumers write is more influenced by reviews provided
17
H2b: There is a stronger correlation between the level of positivity of existing reviews and the degree
of positivity of given reviews if existing reviews are provided by experts than when they are provided by other consumers.
Figure 1 gives a visual representation of hypothesis H2b, to make the expected effect more clear.
FIGURE 1
Moderating Effect of Review Source
3.3 Consumers’ product involvement
Consumers have different degrees of involvement for various products. With product involvement is meant the personal relevance of a product for a consumer, which is based on needs, values and interests (Zaichkowsky, 1994). High involvement products are of greater personal relevance, have greater consequences and provoke more personal connections with consumers than low involvement products (Petty, Cacioppo & Schumann, 1983).
18 To conclude, it is expected that consumers with low product involvement will be more influenced in their review posting behavior by existing reviews than consumers with high product involvement, resulting in the following hypothesis:
H3: The relationship between the level of positivity of existing reviews and the degree of positivity of
given reviews is negatively moderated by consumers’ product involvement.
3.4 Consumers’ degree of optimism
Consumers differ in their degree of optimism. Optimism is a construct that refers to positive expectancies (Peter & Honea, 2012). It means that consumers that are optimistic tend to interpret a situation positively. They expect that things go their way and they anticipate good outcomes (Chan, Sengupta & Mukhopadhyay, 2013). People who are more optimistic act in a more adaptive, emotion-focused way, using strategies such as acceptance, humor, and they positively reframe the situation (Scheier, Carver & Bridges, 1994). Even if the situation is negative, optimism has the capacity to replace negative feelings with positive feelings (Peter & Honea, 2012).
Pessimists, on the other end of the continuum, expect things to go not as they want and believe that bad things will happen to them (Scheier, Carver & Bridges, 1994; Chan, Sengupta & Mukhopadhyay, 2013). The differences between optimists and pessimists lie in the favorability with which expectations are held, which means that optimists expect and perceive outcomes as more favorable than pessimists do (Chan, Sengupta & Mukhopadhyay, 2013).
Since optimistic people in general have a positive approach towards situations and things, it is expected that optimists are also more positive in their review posting behavior. They have positive expectations which may lead to higher ratings in comparison with pessimists, who have more negative expectations. This results in the following hypothesis:
H4: The relationship between level of positivity of existing reviews and the degree of positivity of
given reviews is positively moderated by consumers’ degree of optimism.
3.5 Consumers’ degree of extraversion
19 extravert is externally focused (Cuperman & Ickes, 2009). Characteristics of extraverts are impulsive, sociable, talkative, energetic, active, outgoing, excitement seeking, affectionate, friendly with others, and gregariousness (McCabe & Fleeson, 2012; Cuperman & Ickes, 2009; Turban, Stevens & Lee, 2009). Consumers’ degree of extraversion is especially relevant to their social behavior (Cuperman & Ickes, 2009).
According to Jung, Lee and Karsten (2012), extraverts have a wider attention span than introverts (who are internally focused). They state that people who are extravert seek more actively for external stimulation than introverts. Introvert persons have a more narrow focus and tend to avoid external stimulation. They also found differences in the working memory between extraverts and introverts. Due to their larger working memory, extraverts are able to store, search, retrieve and reproduce more information than introverts. Extraverts are looking for stimulation in the company of others and can process more cues in arousal conditions than introverts (Jung, Lee and Karsten, 2012).
Since extraverts are more externally focused, are actively seeking for external stimulation and can process more cues than introverts, it is expected that consumers with high levels of extraversion will be more influenced by existing reviews of others than introverts, who are more internally focused and avoid external stimulation. This leads to the following hypothesis:
H5: There is a stronger correlation between the level of positivity of existing reviews and the degree
of positivity of given reviews when consumers’ degree of extraversion is higher.
3.6 Conceptual model
20
FIGURE 2 Conceptual Model
Level of positivity of existing reviews
Consumers’ review posting behavior (degree of positivity of the review
21
4. RESEARCH METHODS
This section discusses the research methods. It starts with a sub-section about the data collection, which includes the method, the sample and measurement of the variables. The second sub-section is about data analysis, in which the used statistical techniques are explained.
4.1 Data collection
In this study it is investigated how the level of positivity of existing reviews influences the degree of positivity of the reviews that consumers write. Besides, also the moderating effects of review source, product involvement, optimism and extraversion are researched. The independent variable had three different levels, namely low level of positivity (rated with 2.5 stars), moderate level of positivity (rated with 3.5 stars) and high level of positivity (rated with 4.5 stars) The moderator review source had two categories, namely reviews provided by consumers and reviews provided by experts. The other moderators, consumers’ product involvement, optimism and extraversion are included in the research as continuous variables, since it cannot be determined on beforehand if consumers would have high or low product involvement and to what extent they are optimistic and extravert.
The product/service that is central in the questionnaire is a movie, specifically a short movie. Since most consumers are familiar with movies or have seen a movie before, they can quite easily evaluate the movie they get to see. Two different short movies are used, to be sure that the movie itself does not have too much influence. The setting of the questionnaire was the Cannes film festival. Respondents were told that since 2011 the Cannes film festival has a new category, namely short movies. For the first time this year a public jury is used to determine which short movie wins. Respondents are invited to take part of the public jury and judge one of the nominated short movies.
Method. The data for this study is collected through quantitative research, in the form of an
22
FIGURE 3
Experimental Conditions (3 x 2)
Independent variable Moderating variable
Different experimental methods can be used in performing an experiment. For this study a between-subjects design is used, which implies that there are two or more experimental groups and each group is exposed to one treatment (Malhotra, 2010; Charness, Gneezy & Kuhn, 2012). The six experimental groups of this study were confronted with different scenarios: (1) reviews with a low level of positivity provided by other consumers, (2) reviews with a low level of positivity provided by experts, (3) reviews with a moderate level of positivity provided by other consumers, (4) reviews with a moderate level of positivity provided by experts, (5) reviews with a high level of positivity provided by other consumers, and (6) reviews with a high level of positivity provided by experts. At the end the results of the six different groups were compared with each other, to find out if there are significant differences between the groups. Respondents were randomly assigned to one of the six experimental groups, which means that the technique of randomization is used (Malhotra, 2010).
Six structured questionnaires are used, another one for each experimental group. The questionnaires all started the same, with questions about demographics. The following part of the questionnaire consisted of questions about the moderators product involvement, optimism and extraversion. Subsequently, the actual experiment takes place, in which respondents got to see a short movie and existing ratings, followed by questions to test the relationship between the main independent and dependent variable and the influence of the review source. The questionnaire ended with questions to check the manipulations. The questionnaire can be found in appendix I. To verify if the questions are understood in the right manner, a pre-test of the questionnaire is done with five individuals. Later in this chapter, the measurement and manipulations of the variables are discussed.
Sample. As mentioned in the previous section, the sample consisted of six experimental
groups. Each group contained at least 50 respondents, which means that the total sample counted Low level of positivity
(average rating 2.5 stars)
Moderate level of (average rating 3.5 stars)
High level of positivity (average rating 4.5 stars)
Reviews provided by other consumers
23 303 individuals. 50 respondents per cell in an experiment is sufficient to get significant conclusions. The respondents were both men and women with a Dutch nationality. The sampling technique that was used to collect the data is snowball sampling. Through the Internet, via social media and email, an invitation to participate in this research was sent. In this way the initial group of respondents was selected randomly. Besides asking for participation in this research, it was also asked to share the invitation. Based on the referrals of initial respondents, subsequent respondents were selected (Malhotra, 2010).
Independent variables. The main independent variable, level of positivity of existing reviews,
is measured through manipulation. After seeing the short movie the experimental groups were confronted with reviews/ratings of the three different levels of positivity; two groups were confronted with an average rating of 2.5 stars (on a scale of 5 stars), two groups were confronted with an average rating of 3.5 stars (on a scale of 5 stars) and two groups were confronted with an average rating of 4.5 stars (on a scale of 5 stars), which are respectively low, moderate and high levels of positivity. In this way it could be investigated to what extent these different levels of positivity influence respondents.
The other independent variable, source of review, is also measured through manipulation. When the experimental groups were confronted with the three different levels of positivity after seeing the movie, they were also confronted with the two different sources of reviews, namely other consumers and experts. Three of the six groups that were confronted with different levels of positivity were told that other consumers have seen the movie and rated it on average with 2.5/3.5/4.5 stars. The other three groups were told that experts rated the movie on average with 2.5/3.5/4.5 stars. With help of this manipulation it could be investigated if there are differences in the influences of ratings provided by consumers and ratings provided by experts.
Dependent variable. The dependent variable, consumers’ review posting behavior (the
degree of positivity of the review consumers write), is measured by a 10-point Likert scale and two
24 that were used are: (1) I would like to recommend this movie to others, and (2) When iCandy
Productions has another new movie, I would also like to see it.
Moderators. The moderator source of review is measured through manipulation. For the
moderating effect of this variable, the same manipulation is used as used for source of review as an independent variable.
The moderator consumers’ product involvement was measured by a four-item five-point Likert scale question and an open question, asked in the beginning of the questionnaire when respondents were not yet confronted with the short movie. This moderator is not included as an experimental condition, but as a continuous variable. Beatty and Talpade (1994) also measured product involvement in their study, the same questions are used in this study. The four five-point Likert scale statements were: (1) In general I have a strong interest in movies, (2) Movies are very important to
me, (3) Movies matter a lot to me, and (4) I get bored when other people talk to me about movies.
The moderator consumers’ degree of optimism was measured by a ten-item five-point Likert scale question, asked in the beginning of the questionnaire after the questions about product involvement. This moderator is also included as a continuous variable. To measure optimism, the Revised Life Orientation Test (LOT-R) was used. Scheier, Carver and Bridges (1994) also used this test to measure optimism. The ten-item five-point Likert scale statements of the LOT-R are displayed in table 1. Of the ten items three are positively formulated (1, 4 and 10), three are negatively formulated (3, 7 and 9), and the other four are just filler items (2, 5, 6, 8) which are not used in scoring.
TABLE 1
Items measuring optimism Items
1. In uncertain times, I usually expect the best. 2. It is easy for me to relax.
3. If something can go wrong for me, it will. 4. I am always optimistic about my future. 5. I enjoy my friends a lot.
6. It is important for me to keep busy. 7. I hardly ever expect things to go my way. 8. I do not get upset too easily.
9. I rarely count on good things happening to me.
25 The moderator consumers’ degree of extraversion was measured by a twelve-item five-point Likert scale question, asked in the beginning of the questionnaire after the questions about optimism. This moderator is also included as a continuous variable. To measure extraversion, a part of the brief version of the Eysenck Personality Questionnaire (EPQ-BV) was used. Sato (2005) developed this revised version of the Eysenck Personality Questionnaire. The brief version measures only extraversion and neuroticism, two of the primary personality traits of Eysenck’s theory. For this study only the questions about extraversion are used. The items are displayed in this study as five-point Likert scales, instead of the original yes/no questions, to keep the manner of questioning constant and held it as easy as possible for the respondents. Table 2 shows twelve items that are used to measure extraversion.
TABLE 2
Items measuring extraversion
Manipulation check. The last questions of the questionnaire include manipulation checks, in
order to check whether the manipulations were successful. The respondents were asked what the average rating was with which the movie was rated by others. They were also asked by what kind of persons the existing ratings were provided. Respondents that gave a wrong answer to these questions were not successful manipulated and therefore seen as unreliable. The data of these respondents was excluded from this study.
Control variables. Extraneous variables may affect the dependent variable when they are not
controlled. Since this will lead to wrong conclusions, it is important to control for extraneous variables (Malhotra, 2010). In this study the volume of existing online ratings is included as a control variable, which is a possible source of social influence (Sridhar & Srinivasan, 2012). The number of existing ratings is held constant for all experimental groups, which means that they were all
Items
1. I am a talkative person. 2. I am a rather lively person. 3. I enjoy meeting new people.
4. I usually can let myself go and enjoy myself at a lively party. 5. I usually take the initiative in making new friends.
6. I can easily get some life into a rather dull party.
7. I tend to keep in the background on social occasions (i.e. on parties). 8. I like mixing with people.
26 confronted with an average of the last five ratings. This study also controls for online review characteristics, namely length of the review and the type of information disclosed. These variables may also affect consumers’ ratings (Sridhar & Srinivasan, 2012). To control for these variables, respondents were only confronted with an average rating, given as a grade. They were not exposed to long or short stories with different types of information, but just with the rating. This study also controls for the kind of movie. It may be the case that consumers react different to different movies. Therefore 2 movies are taken into account, to control for it. Another control variable is the knowledge of respondents about the displayed movie. They were asked if they already knew the movie, to control for this. The frequency of visiting the cinema or renting a movie is also a control variable. To control for this, respondents are asked how often they go to the cinema or rent a movie during a year.
4.2 Data analysis
To measure the moderator consumers’ product involvement, consumers’ optimism and consumers’ extraversion, multi-item scales were used. Cronbach’s Alpha is used to examine the internal consistency of these multi-item scales. In other words, it tested whether the different questions that are used to measure one variable are really measuring the same. According to Malhotra (2010), Cronbach’s Alpha is the right manner to establish if the internal consistency between the several items is sufficient.
27
5. RESULTS
In this part, the results of the research are discussed. First the socio-demographic characteristics of the respondents are discussed. Than the Cronbach’s Alpa of the multiple item scales are analyzed. The following sub-section contains the normality test. Further, the hypothesis are tested, by performing and describing the results of the analyses of variance and the multiple regression. This section ends with a description of the effects of demographic variables and control variables.
5.1 Socio-demographic characteristics
In total, 402 respondents filled in the questionnaire for this study. However, only 303 (75.4%) of these questionnaires could be used. There were several reasons for not including certain respondents. Four respondents were younger than 18 years old and therefore were too young for this research. 54 respondents filled in the questionnaire only half, most of them stopped when they should have to watch the short movie. The other 41 respondents were excluded from this study because they did not correctly fill in the manipulation check, which means that they were not successfully manipulated. So the total sample consisted of 303 respondents. To check whether the sample is representative for the Dutch population of 18 till 65 years, the sample is compared with population statistics of CBS and the government (Centraal Bureau voor de Statistiek, 2013; Ministerie van Onderwijs, Cultuur en Wetenschap, 2013). An overview of this comparison is included in table 3.
TABLE 3
Socio-demographic characteristics
Demographic variable CBS, 2013 Minocw, 2013 Sample
Gender Male Female 49.5 % 50.5 % 44.6 % 55.4 % Age 18 – 39 years 40 – 65 years 42.5 % 57.5 % 71.9 % 28.1 % Education VMBO, LBO Havo, vwo MBO HBO, WO 19.4 % 8.1 % 32.5 % 31.3 % 9.6 % 4.3 % 38.9 % 47.2 %
28 consists of way too many respondents of 18-39 years old (71.9%, while it should be around 42%) and way too less people of 40-65 years old (28.1%, while it should be around 58%). Furthermore, the sample contains a bit too much higher educated people (47.2% hbo/wo, which should be around 31%). The differences in age and education of the sample compared with the statistics of the Dutch population can be explained by the fact that a lot of students have filled in the questionnaire.
Since the gap in age between the Dutch population and the sample is rather big, the sample is weighted on age. A dummy is made, in which 0 refers to 18-39 years old, and 1 refers to 40-65 years old. There are too many respondents in the first group, so it is weighted 0.59 (42.5/71.9). The second group is too small and therefore it is weighted 2.05 (57.7/28.1). Taken these weights into account, the socio-demographic characteristics of the sample will change, as is shown in table 4.
TABLE 4
Socio-demographic characteristics weighted on age
Demographic variable Sample
Gender Male Female 41.2 % 58.8 % Age 18 – 39 years 40 – 65 years 42.5 % 57.5 % Education VMBO, LBO Havo, vwo MBO HBO, WO 16.3 % 5.4 % 42.8 % 35.6 %
From table 4 it can be concluded that the sample is perfectly distributed on age after the weighting. Due to the reweighting of age, also the distribution of gender and the distribution of education changed a bit. For the distribution of education it is positive. Since younger people (to which students belong) are less weighted, the surplus of higher educated people is also reduced. The distribution of gender has negatively changed due to the weighting. However, the sample still differs less than 10% in gender of the Dutch population and therefore it will be acceptable to work with. Thus, from now on the weighted data will be used in the analyses.
29 noticeable that the first group, which is exposed to a score of 2.5 stars, has the highest frequency of cinema visit or movie rental. This is accidental.
TABLE 5
Socio-demographic characteristics per experimental group
5.2 Cronbach’s Alpha
To measure product involvement, optimism and extraversion, multiple items are used to form a total score. Although these multiple items were derived from existing research, Cronbach’s Alpha is used to check if the internal consistency of the set of items forming the scale is high enough (Malhotra, 2010). Before the items could be checked by Cronbach’s Alpha, the reverse items should be reformulated.
The set of items of product involvement contains one reverse item: ‘I get bored easily when others
talk to me about movies’. Of the multiple items to measure optimism (LOT-R), three items were
negatively formulated: (1) ‘If something can go wrong for me, it will’, (2) ‘I hardly ever expect things
to go my way’ and (3) ‘I rarely count on good things happening to me’. Besides these reverse items,
which should be reformulated, the LOT-R also contains four filler items, which can be removed completely: (1) ‘It is easy for me to relax’, (2) ‘I enjoy my friends a lot’, (3) ‘It is important for me to
keep busy’, and (4) ‘I do not get upset too easily’ (Scheier, Carver and Bridges, 1994). The multiple
items to measure extraversion (EPQ-BV) contain two reverse items: (1) ‘I tend to keep in the
background on social occasions (i.e. on parties)’, and (2) ‘I am mostly quiet when I am with other people’ (Sato, 2005). The reformulation of the negatively formulated items should counteract
30 problems in the Cronbach’s Alpha analysis. Table 6 provides the Cronbach’s Alpha scores of the three variables.
TABLE 6
Cronbach’s Alpha scores
Cronbach’s Alpha Number of Items
Moderator Product involvement .778 4
Moderator Optimism .748 6
Moderator Extraversion .928 12
All Cronbach’s Alphas are higher than 0.6, which indicates that the multiple items that are used to measure the several variables have sufficient internal consistency reliability. In other words, the several items have measured the same. For the variables product involvement, optimism and extraversion it means that the average scores on the multiple items can be used in the analyses.
5.3 Normality test
To understand the nature of the distribution it is useful to use measures of shape, such as measuring skewness and kurtosis of the distribution (Malhotra, 2010). For the dependent variable degree of
positivity of given reviews, skewness and kurtosis are tested. Table 7 shows that the distribution is
moderate positively skewed (-.702 < -1), which means that the scores are a bit more grouped on the right and with a little longer tail on the left. The Kurtosis indicates that the distribution is a little more peaked than a normal distribution (.227 > 0), which is called leptokurtic distribution.
TABLE 7
Skewness and Kurtosis
Skewness Std. error Skewness Kurtosis Std. error Kurtosis Level of positivity of
given reviews
-.702 .140 .227 .279
31
5.4 Analysis of variance - ANOVA
To test the effects of the independent variables existing reviews and source of the existing reviews on given reviews a two-way analysis of variance (ANOVA) is used. Since the two-way ANOVA also measures the interaction between existing reviews and source of existing reviews, this analysis can also be used to test the moderating effect of source of the reviews.
Existing reviews. Table 8 shows for each experimental group the mean levels of positivity of
given reviews. From this table it is noticeable that the patterns of the means of the groups exposed to consumer reviews and expert reviews are some kind of the same. When existing reviews become more positive, given reviews first decrease and thereafter increase (consumers: 3.024, 2.992, 3.190; experts: 2.967, 2.883, 3.185). Further, it can be mentioned that the groups exposed to consumer reviews give higher reviews than the groups exposed to expert reviews. However, the differences are very small. Figure 4 gives a graphical representation of the mean given reviews.
TABLE 8
Mean given reviews per experimental group
Level of positivity existing reviews
Low (2.5 stars) Moderate (3.5 stars)
High (4.5 stars) Total
Source of reviews Consumers Experts Total 3.024 2.967 2.996 2.992 2.883 2.937 3.190 3.185 3.188 3.068 3.013 3.041 FIGURE 4
32 Table 9 shows the summary table of the two-way ANOVA. With an F-value of 2.211 and a p-value of .111 (>.05) it can be concluded that existing reviews do not have a significant effect on given reviews, since the p-value is higher than .05. Even on a marginal level (p-value < .10) there is no significant relationship between the existing reviews and given reviews. This means that hypotheses 1a1 and 1a2
cannot be supported.
TABLE 9
Two-Way ANOVA summary table
It is noticeable from the graph in figure 4 that the mean given reviews of the groups exposed to low and moderately positive existing reviews are close to each other and that the means of the groups exposed to highly positive existing reviews are higher. It might be the case that existing reviews of 2.5 and 3.5 stars (low and moderately positive) together are significantly different influencing given reviews than existing reviews of 4.5 stars (highly positive). Another two-way ANOVA, in which low and moderately positive reviews are grouped together, is used to test this. The results of this analysis are included in table 10.
TABLE 10
Two-Way ANOVA low and moderate reviews grouped together
Df N Mean given
reviews
F Sig.
Existing reviews (0 = 2.5 and 3.5 stars; 1 = 4.5 stars)
2.5 and 3.5 stars 4.5 stars
Source reviews (0 = consumers; 1 = experts)
Source reviews * Existing reviews
1 1 1 303 201 102 303 303 2.966 3.188 3.989 .183 .191 .047 .669 .663 R square Adjusted R square .038 .014
With a F-value of 3.989 and a p-value of .047 (<.05) the effect of low and moderately positive existing reviews on given reviews is significantly different from the effect of highly positive existing reviews. Respondents exposed to highly positive existing reviews give higher reviews themselves
Df N F Sig.
Independent variables
Existing reviews (1 = 2.5 stars; 2 = 3.5 stars; 3 = 4.5 stars)
Source of reviews (0 = consumers; 1 = experts)
Moderator
Existing reviews*source of reviews
33 than respondents exposed to low or moderately positive existing reviews. The mean of the given reviews of respondents exposed to highly positive existing reviews (4.5 stars) is 3.188 and the mean of the groups exposed to low and moderately positive existing reviews (2.5 and 3.5 stars) is 2.966. Thus, the given reviews of the respondents exposed to highly positive existing reviews are significantly higher than those of respondents exposed to low or moderate positive existing reviews and lower than the existing review they were themselves exposed to (3.188 is lower than 4.5). Therefore it can be stated that consumers exposed to highly positive existing reviews give somewhat less positive reviews, which means that hypothesis 1b can be partially supported.
It is also tested whether the short movie that respondents saw has an influence on the degree of positivity of given reviews. Two movies are used: One half of the respondents watched the movie ‘The Plan’, the other half watched ‘The Date’. Table 11 shows an overview of the ANOVA results which tested the influence of the two movies. The F-value is 1.235 and the p-value is .267, which means that there is no significant difference in the mean given reviews of the two movies. So, both movies have a same influence on the degree of positivity of given reviews. Therefore it does not have an influence on the further results whether respondents have seen ‘The Plan’ or ‘The Date’.
TABLE 11
ANOVA effect different movies
N Df Mean given reviews F Sig.
Movie (0 = The Date; 1 = The Plan)
The Date The Plan 303 145 158 1 3.07 3.00 3.12 1.23 5 .267
Source of reviews. The other independent variable, source of reviews, has a F-value of .354
and a p-value of .552, as mentioned in table 9. Therefore it can be concluded that there is no significant correlation between the source of the existing reviews and the positivity of given reviews. This was also suggested by the means in table 8. The means are very close to each other, suggesting already that the ratings of groups exposed to consumers reviews will not be very different of the ratings of groups exposed to expert reviews. Based on the non-significance, hypothesis 2a should be rejected.
34 with a F-value of .104. Since the p-value is larger than alpha (.901 > .05), the moderating effect of source of reviews is not significant. This means that the influence of existing reviews on given reviews is not influenced by the source of the reviews; whether they are provided by consumers or experts. Again this was already suggested by the means of table 8. Overall, it can be concluded that hypothesis 2b cannot be supported.
5.5 Multiple regression
To test the influences of the nominal moderators product involvement, optimism and extraversion, multiple regressions are used. First of all, a multicollinearity test is performed.
Multicollinearity test. Before multiple regressions can be done, the data should be checked
on multicollinearity. Multicollinearity means that the intercorrelations among the predictors or independent variables are very high (Malhotra, 2010). In other words, in case of multicollinearity the independent variables are correlated with one another. To test the data on multicollinearity, the various inflation factor (VIF) for each independent variable should be determined. Table 12 shows the VIFs of the variables of this study. According to Keller (2008), multicollinearity is too high when the VIF is 10 or higher. Looking to the VIF scores of this study, it is noticeable that the VIF scores of the variables source of reviews and the interaction SR * ER are larger than 10 (20.760 and 22.010 respectively). Since source of reviews is part of the interaction term SR * ER, it is logical that the VIF scores of these two variables are high, Therefore, there is nothing to worry about, multicollinearity is not an issue in this study.
TABLE 12
Multicollinearity between independent variables
Independent variables VIF
Main independent variables
Existing reviews (ER)
Source of reviews (SR) (0 = consumers; 1 = experts)
Moderating variables Interaction SR * ER
Interaction Product Involvement * ER Interaction Optimism * ER
Interaction Extraversion * ER
Consumer-related variables
Gender (0 = male; 1 = female)
Age (0 = 18 - 39 years; 1 = 40 – 65 years)
Education
Control variables
Movie (0 = The Date; 1 = The Plan)
35
Knowledge about movie (0 = no; 1 = yes)
Frequency cinema visit/movie rental
1.036 1.181
Multiple regression results. The results of the multiple regression analysis are presented in
table 13. The overall p-value of the model is .014 and the F-value is 2.022, which means that the overall model is significant (.014 < .05).
TABLE 13
Results multiple regression analysis.
Beta t Sig.
Constant 5.715 2.905 .004*
Independent variables
Existing Review
Source of Review (0 = consumers; 1 = experts)
Moderating variables
Product involvement Optimism
Extraversion
H3: Interaction Product Involvement * Existing Review H4: Interaction Optimism * Existing Review
H5: Interaction Extraversion * Existing Review
Demographic variables
Age (0 = 18-39 years; 1 = 40-65 years)
Gender (0 = male; 1 = female)
Education
Control variables
Movie (0 = The Date; 1 = The Plan)
Knowledge about movie (0 = no; 1 = yes)
Frequency cinema visit/movie rental
-.882 -.158 -.072 .191 -1.082 .018 -.068 .187 .245 -.046 .082 .149 .864 .007 -1.650 -.324 -.219 .397 -2.426 .389 -.980 2.998 1.924 -.402 1.328 1.341 1.222 1.057 .100 .746 .827 .692 .016** .697 .328 .003* .055*** .688 .185 .181 .223 .291 P-value of the model
F-value R square Adjusted R square .014 2.022 .098 .049 * Significant at 1% level ** Significant at 5% level *** Significant at 10% level
Product involvement. To test the influence of the moderator consumers’ product
36
Optimism. To test the influence of the moderator consumers’ optimism, there is also first
computed a new variable representing the interaction effect of optimism (optimism * existing review). In table 13 it is mentioned that the p-value of this interaction effect is .328, indicating that it does not have a significant influence (.328 > .05) on the effect of existing reviews on given reviews. These results lead to rejection of hypothesis 4.
Extraversion. Prior to performing the multiple regression to test the influence of consumers’
extraversion, a new variable for the interaction effect of extraversion is computed (extraversion x existing review). Table 13 shows that the p-value of this interaction term is .003, which is significant on a 1% level (.003 < .01). The Beta of the interaction effect of extraversion is .187. Since this is a positive Beta, a positive moderating effect of extraversion can be concluded, which results in supporting hypothesis 5.
The interaction effect of extraversion is plotted in figure 5. From the plot it can be concluded that extravert consumers give higher ratings when existing ratings are higher than when existing ratings are lower. Introvert consumers (low extraversion) are also positively influenced by existing reviews, but less strong than extravert consumers. This is corresponding to the expected effect that extraverts are more influenced by existing reviews than introverts.
Effects of control variables and demographic variables. The demographic variables age,
gender and education and the control variables movie (The Date or The Plan), knowledge of the movie and frequency of cinema visit or movie rental are included in the analyses to control for the
effects. Looking at the results in table 13, it is noticeable that gender and education have p-values larger than alpha (.688 and .185 respectively). This means that they do not have a significant effect on the level of positivity of given reviews. Age, on the other hand, has a p-value of .055 and is therefore significant on a marginal level (.055 < .10). The Beta of age is .245, which indicates that the older consumers are, the more positive their given reviews will be. The p-values of the control
FIGURE 5
37 variables are all larger than .10, namely .181 for movie, .223 for knowledge of the movie, and .291 for frequency of cinema visit or movie rental. Therefore, these three control variables do not have a significant effect on the level of positivity of given reviews. In other words, consumers are not significantly influenced in the level of positivity of their given reviews by the kind of movie, their knowledge of the movie and the frequency with which they visit cinemas or rent movies.
38
6. DISCUSSION
This section discusses the results of this study. It also describes the managerial implications of the results for retailers. This section end with limitations and suggestions for further research.
This study investigates whether the level of positivity of existing reviews has an influence on the level of positivity of given reviews. Further, it is examined if the source of the review has a direct influence on the level of positivity of given reviews and if review source has a moderating influence. The source of the review is either consumers or experts. Other moderators that are taken into account in this study are consumers’ product involvement, consumers’ optimism and consumers’ extraversion.
Table 14 gives a summary of the hypotheses and results. From this table it is noticeable that one of the eight hypotheses is completely supported and one of the hypotheses is partially supported. The other six hypotheses could not be supported.
TABLE 14
Summary of hypotheses and results
Variable Hypothesized relationship Result
Influence on positivity of given reviews:
H1a1: Positivity existing reviews
H1a2:Positivity existing reviews
H1b: Low positivity/high positivity existing reviews
H2a: Source of review
+ weakening + (low), - (high) present Not supported Not supported Partially supported Not supported Moderating effects: H2b: Source of review H3: Product involvement H4: Optimism H5: Extraversion present - + + Not supported Not supported Not supported Supported
6.1 Effects of independent variables
39 Although the main effect is not significant, there is found partial support for the hypothesis that given reviews are somewhat less positive when existing reviews are high and given reviews are somewhat less negative when existing reviews are low. This research could not support that given reviews are somewhat less negative when existing reviews are low, but it could support that given reviews are somewhat less positive when existing reviews are high. Customers exposed to highly positive existing reviews gave a significantly different review than customers that are exposed to low or moderately positive existing reviews. The reviews of customers exposed to highly positive existing reviews are more positive than the reviews of customers exposed to low or moderately positive existing reviews. However, the mean of their given reviews is lower than the level of positivity of existing reviews (4.5 stars), which implies that the effect of highly positive review is weakening. This conclusion corresponds to the conclusions of Sridhar and Srinivasan (2012) who found that higher ratings of other consumers weaken the reviews of consumers in their positivity or negativity.
Another independent variable used in this study is the source of the review. The two sources that were taken into account were experts and consumers. In accordance with existing studies of Chakravarty, Liu and Mazumdar (2010), Lee, Kim and Chan-Olmsted (2011) and Woodruff (1972) it was expected that existing reviews of experts will have a larger influence on given reviews than existing reviews of consumers. However, this study does not found a significant difference in the influence of either expert reviews or consumer reviews. This means that consumers do not attach more importance to and are not more influenced by existing reviews of experts than existing reviews of consumers.
6.2 Effects of moderators
The influences of four moderators, source of the reviews, consumers’ product involvement, consumers’ optimism and consumers’ extraversion, on the main relationship are examined in this research.
