Online Customer Reviews:

Online Customer Reviews:

The Effect of Humor and Emotion Words on the Review

Helpfulness

By

Stephanie Nanninga

University of Groningen

Faculty of Economics and Business

MSc Marketing Management

26th of January, 2015

Lindelaan 10,

2243 BS Wassenaar

Tel: +31 6 30336935

E-mail: s.n.nanninga@gmail.com

Student number: 1885111

2

Management Summary

This thesis focusses on online consumer reviews on shopping websites. Online consumer reviews play a major role in consumers’ decisions about whether or not to purchase products or services (Duan et al., 2008). Identifying the most helpful product reviews represents an important task in order to reduce an information overload for the consumer and therefore relevance of this topic has increased over the years. Review helpfulness is of particular importance since they constitute a focal point for examining consumer decision making during the purchase process (Korfiatis et al., 2012). Research into factors that affect this review helpfulness is the aim of this study; the style factors “humor” and “emotion words” are researched. A gender perspective is taking into account by looking at the moderating effects of the “style of processing” and the level of “emotional awareness”.

A 2x3 between subject factorial design is conducted with 170 respondents who filled in an online questionnaire. The results do not show significant differences between the six different online reviews regarding review helpfulness. Additionally, no significant moderating effects are indicated from this study.

From this research it can be concluded that there are no significant differences between humorous and non-humorous reviews, and no significant differences between using positive emotion words, negative emotion words and neutral words regarding review helpfulness. Therefore, it cannot yet be demonstrated through this study that managers should or should not have to intervene when someone posts for example a humorous review without emotion words for a product that is aimed for women.

It can only be said that women and men indeed differ in their style of processing and level of emotional awareness.

4

Acknowledgements

After graduating my Bachelor in International Business and Management, I started in September 2013 my Master Marketing Management. After a few months of the start of this master, the master thesis was already introduced. The idea of writing your master thesis sounded scary to me, as this is ‘your final master piece of all your years of studying’. However, within no time I was referred to the thesis topic I wanted and before I knew I was in the middle of writing it. As my supervisor, Dr. Liane Voerman, told us at the very beginning, writing your thesis can be fun but will also give you a moment that you will dislike it really much. Indeed, it was challenging, and I have seen both sides of this continuum that Dr. Liane Voerman was talking about. Consequently, this makes me even more proud that I completed writing this ‘master piece’. This thesis marks the end of my great time as a student at the University of Groningen and my great time in Groningen as well. Even though I moved out of Groningen in September already, being there every two weeks for the thesis meetings gave me lots of fun. Not being a student anymore means a new part of your life will start. A new adventure is waiting for me as I will start in January my Trade Marketing internship at Nestlé, which I really look forward to.

I would like to take this opportunity to thank some people that have helped me writing my thesis and who supported me in the tough times during the last few months. First of all, I would like to thank Dr. Liane Voerman for her thesis meetings, that were really helpful and fun, and for her productive feedback by e-mail or phone. Also, thanks to my fellow students who helped me out with their constructive feedback during the group meetings and the group communication on Facebook. Additionally, thanks to everyone who filled in my questionnaire. Last but not least, thanks to my family and friends who have unconditionally supported me during the whole writing process of this thesis and all the other months of this difficult year for me.

6

Table of Content

1.2 Rely on online customer reviews ...8

1.3 Factors influencing the helpfulness of online customer reviews ...9

1.4 Style factors ...9

1.5 A Gender Perspective ...11

1.6 Problem statement ...11

1.7 Academic and Managerial Relevance ...12

1.8 Structure of the thesis ...12

2. Literature Review...14

2.1 Style factor - Humor ...14

2.2. Style factor: Emotion Words ...15

2.2.1 Positive and Negative Emotion Words ...17

2.2.2 Interaction Humor and Emotion Words ...18

2.3 Gender and Style of Processing ...18

2.3.1 Humor and Style of Processing ...20

2.3.2 Emotion Words and Style of Processing ...22

2.4 Gender and Emotional Awareness ...23

2.4.1 Emotion words and Emotional Awareness ...24

2.5 Conceptual Model ...25

3. Methodology ...26

3.1 Research Design and Participants ...26

3.2 Procedure ...27

3.3 Operationalization and reliability of scales ...30

3.3.1 Operationalization of the IV’s Humor and Emotion Words ...30

3.3.2 Operationalization of the DV Review Helpfulness ...31

3.3.3 Operationalization of the Moderator Style of Processing ...32

3.3.4 Operationalization of the Moderator Emotional Awareness ...32

3.4 Control Variables ...34

3.5 Plan of Analysis ...35

7

4. Results ...38

4.1 IV’s on Review Helpfulness (Q1 and Q2) ...38

4.1.1 Conclusion ...40

4.2 IV’s and Moderators on Review Helpfulness (Q1 & Q2) ...40

4.2.1 Conclusion ...43

4.3 Moderatos and gender ...43

4.3.1 Conclusion ...43 4.4 Alternative Model ...44 4.5 Overall Results ...44 5. Conclusion ...46 5.1 Discussion ...46 5.1.1 Humor ...46 5.1.2 Emotion words ...47 5.1.3 Moderators ...48

5.2 Limitations and further research ...48

8

1. Introduction

1.1 General Introduction

Imagine buying a new smartphone. Most likely one will do some pre-research online. Research indicates that online ratings and reviews on retailer websites (52%) were included among the top three sources of information most frequently used by respondents, ahead of advice from friends and family members (49%) and advice from store employees (12%) (Cisco Internet Business Solutions Group 2013).When reading several online reviews, there might be some reviews perceived more helpful for you than other ones. Consequently, you would rely (more) on the reviews that you perceived as helpful and thus not on all of the reviews presented. Consumers’ incorporation of online product reviews into their decision-making has not escaped the notice of retailers, who actively try to harness electronic word-of-mouth (eWOM) as a new marketing tool (Floyd et al., 2014). Retailers invite their consumers to post personal product evaluations on seller websites or availing consumers to information provided about their products by other third-party sources (Dellarocas 2003). Identifying the most helpful product reviews represents an important task in order to reduce an information overload for the consumer (Siering & Muntermann, 2013). Especially because customer reviews are more and more available online, which makes it difficult for consumers to choose reliable reviews. In the recent years, more and more researchers are investigating what online reviews make them helpful to customers and how to classify helpful reviews versus unhelpful reviews (Zhiming et al., 2011). Therefore, the focus in this thesis is to better understand the factors that make online reviews appealing to consumers.

1.2 Rely on online customer reviews

9 helpfulness as the extent to which consumers perceive the product review as being capable of facilitating judgment or purchase decisions. How helpful an online review is (review helpfulness) is of particular importance, since they constitute a focal point for examining consumer decision making during the purchase process (Korfiatis et al., 2012). However, not every online review has a similar effect on someone’s purchase decisions. Reviews that are considered more helpful by consumers have stronger effects on consumer purchase decisions than other reviews (Chen et al., 2008; Chevalier & Mayzlin, 2006).

1.3 Factors influencing the helpfulness of online customer reviews

In order to optimize message adoption from online customer reviews, it is crucial to explore which specific factors in the online customer review do influence the decision making process. Previous studies have already researched several determinants of review helpfulness, such as characteristics of the reviewer, the receiver of the review, the review itself or the product type (Zhiming et al., 2011; Baek et al., 2012; Mudambi & Schuff, 2010). The focus in this study will be on the review characteristics, specifically the style factors of the review. Some studies have already measured several different factors of the review itself, such as valence (positive, negative or neutral), volume, content quality, argument quality, style, credibility, the rating of the review and accuracy (Chan & Ngai, 2011; Dellarocas et al., 2007; Park et al., 2007; Sussan et al., 2006; Lee 2008; Baumeister, 2001; Duan et al, 2008; Cheung et al, 2007). To sum up, many studies have indicated that various antecedents in an online review influence a purchase decision, with an important role for the review itself.

1.4 Style factors

10 that even when the content of the message is the same, individuals can express themselves verbally with their own distinctive styles (Pennebaker & King, 1999).

11

1.5 A Gender Perspective

As stated earlier, not only the characteristics of the review can affect the outcome on helpfulness of the review; consumer characteristics affect the way the consumer processes a message. Among demographic variables, gender is the main indicator of consumer behavior (Lin, 2014) and is therefore one of the most common forms of segmentation and the basis of marketing strategy used by marketers in particular (Kim, Lehto, & Morrison, 2007). Besides, according to sociolinguistic theory, cultural factors, in particular gender, affect communication (Cyr & Bonanni, 2005). For instance, women write and respond to writing differently than men (Heermann, 2010) and women communicate differently than men (Gefen & Ridings, 2005). In addition, prior research suggests that men and women, due to differences in socialization (Bern 1974), may exhibit different patterns of judgments and behaviors (Zhang et al. 2013).

Marketing studies have also shown significant differences in the way product information is processed by men and women. For instance, also in an online setting, Awad and Ragowsky (2008) indicate that the effect of eWOM is stronger for women than for men when it comes to online shopping and those men and women value different factors of eWOM. Additionally, Fan and Miao (2012) find that gender differences affect perceived eWOM credibility, use and acceptance of eWOM, and purchasing decisions. Therefore, it will be interesting to explore the influence of online customer reviews when read by males versus females and what the possible underlying mechanisms are. In other words, gender in particular is taken into account in this study. This study aims to examine whether men and women act differently to the use of humor and presence of emotion words in an online review for a specific product.

1.6 Problem statement

Given the above, the problem statement for this study is as follows:

“How do the online customer reviews’ style characteristics, more particularly humor and emotion words, influence the review helpfulness from a gender perspective?”

In order to be able to answer this problem statement, the following research questions are derived and will be investigated in this study.

1. What is the influence of humor on the helpfulness of online customer reviews? 2. What is the influence of emotion words on the helpfulness of online customer

12 3. Which characteristics from a gender perspective should be taken into account as

moderators?

1.7 Academic and Managerial Relevance

This research will add information to the existing pool of research in the area of online customer reviews. It will be of practical use for manufacturers to get more insight in the way women and men find online customer reviews helpful. For example, when companies aim to target women for their product it would be helpful what type of online customer reviews do support their objective. Online retailers could filter the most relevant reviews and show the most helpful product reviews first (Martin & Pu, 2014), in order to offer greater potential value to customers (Mudambi & Schuff, 2010). Investigating online customer reviews components is thus important since it offers business potential for hosts of websites specialized to consumers-generated content (Stauss, 1997; Stauss, 2000; Mariussen et al., 2010), as helpful reviews gain a strategic advantage in consumer attention and “stickiness” (Connors et al. 2011). To sum up, a better understanding of perceived review helpfulness offers clear benefits to online retailers and review providers.

1.8 Structure of the thesis

14

2. Literature Review

As stated above in the introduction, two style factors are chosen to be examined in this study; “humor” and “emotion words”. The aim is to evaluate whether these two factors might affect the helpfulness of a review. According to Li et al. (2013) review helpfulness is a formative construct existing of the components “perceived source credibility”, “perceived content diagnosticity”, and “perceived vicarious expression”. Yet, in this research, the review text itself will be the main basis to evaluate the helpfulness of a specific review, and other factors, like the source identity and review rating, will be eliminated. Therefore, one of the components of review helpfulness, namely “perceived source credibility”, is not applicable and replaced by the so-called “review credibility” component, which does tap into the credibility, and thus helpfulness, of the review text. The two style factors will be firstly described, followed by the theoretical part of the gender perspective that is taken into account in this study. It is important to note that humor and emotion words in a message can both evoke emotional feelings by the receiver of the message; however, they are not the same. Emotion words are often used to emphasize on the authors opinion, and do not have the intension to be perceived as amusement, while humor is used to make readers laugh or smile and in this way making a link with the product (review). As the gender perspective is taken into account, there will be looked at the style of processing information of individuals and their emotional awareness level. Hereafter, these two moderators and their interaction with the style factors humor and emotion words will be described.

2.1 Style factor - Humor

15 comprehension, and offer support for the information processing theory, which assumes a consumer who makes decisions following acquisitions, integration and evaluation of information (Bettman, 1979; Zaltman and Wallendorf, 1983). Additionally, in other studies it was indicated that humor increases intentions to purchase the product (Perry et al., 1997). Besides, humor may evoke an informal tone that would help the reader feel a connection with the reviewer (Fraley & Aron, 2004). To sum up, regarding the authors above, the use of humor will positively influence the processing of the ad and or the product.

This is contradictory to the study of Wu, Crocker and Rogers (1989) where no effect of humor on ad responses (recall, attitude toward the brand, and purchase intention) was found and contradictory to the study of Madden and Weinberger (1984) where comprehension is decreased by humor. Besides, mixed effects are found in the study of Chattopadhyay and Basu (1990) and in de study of Weinberger and Gulas (1992), who state that it is inappropriate to generalize the persuasive effect of humor. To sum up, abundant research with mixed evidence on the impact of humor on advertising exists.

To conclude, on the one hand humor decreases comprehension and does not increase recall, attitude toward the brand, and purchase intention, while on the other hand humor does enhances attention, consumer’s mood, message comprehension, and intention to purchase.

By analogy with the above, that humor has a positive influence in advertising, and based on Eisend’s (2009) meta-analysis review, that indicates that humor significantly enhances positive affect, brand attitude and purchase intention, the first hypothesis is formulated regarding online reviews:

H1: An online customer review that contains humor, compared to an online customer review without humor, will positively affect the product review helpfulness.

2.2. Style factor: Emotion Words

16 frequently described as overwhelming forces that put forth a wide influence on behavior. Martin and Pu (2014) believe that emotions are compelling instruments for communication as they will almost certainly arouse the feelings of others and engage their reactions. Emotional messages contain information about how strongly someone feels about an issue, or even whether or not to believe verbal statements (Gilbert, 2001). The messages that contains emotion words can evoke emotions that drive evaluation and decision making (Lench et al., 2011) and recognize emotions in text and analyze them can serve as an interesting way to classify documents (Mohammed, 2011; Martin et al., 2014). Simply reading a text with emotional words may be sufficient to influence thoughts and behaviors as emotion words drive people’s action and regulate their decision process (Lau-Gesk & Meyers-Levy, 2009; Lench, Flores, & Bench, 2011). Besides, people often let their emotional feelings outplay their logical thoughts when it comes to making purchases.

As emotions are good to trigger reactions, it is suspected to help review readers putting themselves in the writer’s place. These emotions can be induced by emotion words; these words can bring out positive (or negative) emotions and feelings in people that might be uncorrelated to the facts. By using emotion words, advertisers and propagandists can bring across positive or negative feelings about their product to consumers. Reviewers can add emotions to their reviews by using verbal cues such as emotion words and by using nonverbal cues such as emoticons. Verbal cues are emotion words or phrases (Planalp, 1998) in spoken or written language that show emotions like happy, love, sad, and angry.

Previous theory and research suggest that emotion words evoke automatic affective responses, which require few processing resources, emerge rapidly, and guide attitudes and actions (Baumeister et al., 2007; Cohen et al., 2008). Stieglitz and Dang-Xuan (2013) find that Twitter messages with positive or negative emotions are more likely to be shared than neutral messages. Kanske and Kotz (2007) also find facilitation for both positive and negative emotion words over neutral words. Additionally, Maddock et al. (2003) confirmed that emotion words were consistently better recalled than neutral words. Consequently, the presence of emotion words in online reviews assumes to evoke affective responses and therefore impact someone’s decisions based on this online review.

17 increase the emotional value of the product and thus the review can be more helpful as it signalizes that the product is worth to consider (Schindler & Bickart, 2012). However, other research suggests that sensing strong emotions in another person’s opinion based on the emotion words used negatively affect source trustworthiness and competence (Bowers, 1963; Buller et al., 1998), and, by that, the value, and thus, the helpfulness of the review. Additionally, contradictory results come forth from the studies done by e.g. Mudambi and Schuff (2010), Cao et al. (2011), Pan and Zhang (2011), and Hu et al. (2014) who studied emotion words (in their studies called sentiment) on review helpfulness. Hence, there is not a lot of clarity about the effects of the use of emotion words in written WOM, e.g. online reviews, due to these mixed results.

Based on the above, we assume that emotion-laden words create excitement and enthusiasm (or disdain and dissatisfaction) about a product, and, by that, simplify the decision process and thus being helpful to the reader of reviews (Schindler, 2012). Hence, the second hypothesis is formulated as follows:

H2: An online customer review that contains emotion words, compared to an online customer reviews without emotion words, will positively affect product review helpfulness.

2.2.1 Positive and Negative Emotion Words

The valence of emotion words can be either positive or negative. Therefore, in this study there will be made a distinction between neutral words, positive emotion laden words and negative emotion laden words.

Some research is done into the difference in impact of negative and positive words, leading to the conclusion that negative words have more impact on someone than positive ones. For example, Nass (2010) states that negative information generally involves more thinking and the information is processed more thoroughly than positive ones. People automatically devote more attention to negative information than to positive information (Smith et al., 2006) and negative words have longer color-naming latencies than positive words, suggesting that negative words were automatically drawing more attention than positive words (Pratto and John’s (1991). Besides, negative information contributes more strongly to the final impression than positive information (Baumeister, 2001).

18 review emotion increases the helpfulness of reviews about experience goods. Additionally, Cao (2011) suggests that reviews with extreme opinion receive more helpfulness votes than those with neutral opinions, as the brain handles positive and negative information in different hemispheres. Furthermore, Schindler and Bickart (2012) find that negative sentiment is detrimental, but positive sentiment is beneficial to review helpfulness.

The above findings can partly be related to the so-called negativity bias (Yin et al. (2014), whereby negative reviews tend to be more influential. Bad things will produce larger, more consistent, and more multifaceted or more lasting effects than good things (Baumeister et al., 2001). A frequently quoted reason is that because negative information is rare or unexpected, it is perceived as more useful for decisions (Fiske, 1980). According to Resnick and Zeckhauser (2002), this logic is particularly relevant to eWOM, where negative feedback tends to be much rarer than positive feedback.

Translating the above to the current research, the third hypothesis, is formulated: H3: The use of negative emotion words in a review will have a bigger effect on review helpfulness than the use of positive emotion words.

2.2.2 Interaction Humor and Emotion Words

When the two independent style variables humor and emotion words both exist in an online review, there might be an interaction effect. According to Wegener et al. (1995), humor may influence the persuasiveness of a message by inducing positive moods in listeners, causing them to attend to peripheral, heuristic cues rather than to the strength of the argument via central processing. In other words, due to the presence of humor in a review, causing a positive mood in readers, the negative emotion words might be perceived less negative and thus the review will become less helpful for the reader compared to a review with negative emotions words and without humor. This would similarly indicate that the positive effect of humor on review helpfulness will increase when positive emotion words are also used in the review. Based on this, the fourth hypothesis is formulated:

H4: There is an interaction effect between humor and emotion words on review helpfulness: the presence of humor will lead to a decrease (increase) of the effect of negative (positive) emotion words on review helpfulness.

2.3 Gender and Style of Processing

19 underlying processes that can explain these differences. The style of processing might be a gender specific variable that impacts the way an individual acts on an online review, and thus can be used to draw conclusions about the gender perspective in this study.

Individuals have diverse processing capabilities, motivations and prior experiences (Bettman, 1979). The method to acquiring information, the strategy during this information acquisition and the way individuals use this information are three areas where individuals vary in when creating judgments (Henry, 1980). Segmenting the individuals based on style of processing can be done by looking at gender. Perhaps the best known theory explaining gender differences in the way of processing information is the Meyers-Levy's (1989) Selectivity Hypothesis. This theory states that male and female differences in information processing emerge as women tend to process information in a more comprehensive, effortful manner than men, and may have a higher tendency to perform systematic processing than male consumers. Therefore, women are the so called ‘comprehensive processors’. They consider all available information before making a decision or judgment. Men are the so called ‘selective processors’ and tend to focus on one, or a small number of cues that are highly available and salient, rather than try to process all the information available in a communication instead of detailed message elaboration. Hence, males are expected to engage in peripheral processing. The Selectivity Hypothesis proposes that ads directed at men should be simple and focus on a single theme and ads directed at women should contain a lot of product information. The model also suggests that women have a lower threshold for elaborative processing than men. According to Chan (2011), males often do not even engage in comprehensive processing of all available information and incline to employ various heuristic devices. Consequently, the interpretations above suggests that gender differences regarding information are due to differences in depth of processing.

20 easier to be persuaded (Chan, 2011). The HSM (Chaiken, 1980), a model to explain how people process persuasive messages, also states that individuals rely on two different styles of information processing. One style is called systematic processing, which is more thoughtful, deliberate, and analytical (comparable to central route), while the other style is called heuristic processing, which is more reflexive or automatic and is based on heuristic cues (Cohen, 2010; Seiter & Gass, 2004). This latter style is comparable to the peripheral route of the ELM. The HSM focuses on different persuasion processes that can operate in different situations. Although these two models are quite similar, they differ in the fact that the ELM suggests that as people increase scrutiny through the central processing route, peripheral cues have less impact, and vice versa, while in the HSM, systematic processing and heuristic processing operate independently and may occur simultaneously (Larson, 2010).

When linking these two dual process models to the Selectivity Hypothesis Model, the comprehensive processors of the HSM and the central processing route of the ELM are corresponding to the systematic processors of the Selectivity Hypothesis theory. The selective processors of the Selectivity Hypothesis Model are in line with the heuristic processors of the HSM and the peripheral processing route of the ELM. For this study the focus is on the Selectivity Hypothesis Model, which has been widely attributed to explain the detection of gender differences regarding advertising response (e.g. Carsky & Zuckerman 1991; Darley & Smith, 1995), as well as online consumer behavior (Grassmann & Brettel, 2009).

As the Selectivity Hypothesis Model theory states, it needs to be highlighted that men and women obtain and process online information differently. For online reviews this would mean that comprehensive processors, and thus women, will take all the information in the online review into account, while selective processors, and thus men, will process only a small number of cues in the online review. Hypothesis 5 is formulated as follows:

H5: Women are significantly more comprehensive versus selective in their processing than men.

2.3.1 Humor and Style of Processing

21 Sternthal and Craig (1973) believe that humorous messages attract attention due to enhanced processing of advertising claims, leading to more rather than less comprehension.

Several factors may moderate the effect of humor on persuasion (Weinberger & Gulas, 1992). Humorous effects vary by the target audience’s gender and ethnicity (Madden & Weinberger 1982). According to Li et al. (2009) females indicate a preference for friends who makes them laugh, whereas males prefer a friend who laughs at their humor. Provine, 2000; Kothoff, 2006) find that women laughing more and men making people laugh more, which complies with the theory that men are funnier than women (see e.g. Lewis, 2000). Additionally, women and men vary in how humor is used and appreciated states (Reiss, 2011). Lammers and his colleagues (1983) found a positive effect for humor on persuasion, which was only moderated by males, not females. On the other hand, Brandt (2005) has examined sex-specific differences in the brain's response to humor (Brandt, 2005) and found that some brain regions were activated more in women when exposed to humor. These included the left prefrontal cortex, suggesting a greater emphasis on language and executive processing in women. So, women, instead of men, seem according to this study more prone to the appreciation of humor.

Besides linking humor to gender, which is done above, the effect of humor can be further linked to someone’s style of information processing. Humor can make a person feel warm and comfortable therefore kicking into gear the heuristic theory (Lyttle, 2001). A recent study states that males often do not engage in comprehensive processing of all available information and incline to employ various heuristic devices, and that males are thus expected to be persuaded by humor advertisements more readily than females (Chan, 2011). The use of humor is a good attention-getting device, and can be perceived as a heuristic cue that can catch the audience’s eye and convince individuals (Maibach & Parrott, 1995). Consequently, as humor can be seen as a heuristic, and selective processors rely on a small numbers of salient heuristic cues, they would be more influenced by humor in online customer reviews compared to comprehensive processors. As in contrast to males, females tend to use a ‘comprehensive strategy’ and usually attempt to engage in effortful elaboration of message content. Therefore, it is expected that humor has a less positive effect on review helpfulness when individuals are comprehensive processors. The seventh hypothesis is formulated:

22 2.3.2 Emotion Words and Style of Processing

In the previous part humor and the style of processing is discussed. Here, the second independent variable, emotion words, and the interaction with the moderator style of processing will be looked at. Men and women do respond differently to the specific emotional contents of particular stories (Precht, 2008). However, one could argue whether males or females are more influenced by emotion words when used in online reviews, and how this relate to their different style of processing.

Several theories posit that emotionally salient stimuli, like emotion words, have privileged access during information processing (LeDoux, 2000; Vuilleumier & Huang, 2009), however, it is not clear yet whether these emotion words will be picked up and used by males or females when reading these words in an online review. One could argue whether emotion words will be processed systematically as women will be more triggered by these words, or that these emotion words can be perceived as heuristic cues, and thus picked up by systematic processors.

Males acquire information in a heuristic fashion, therefore missing subtle cues like written emotion words throughout a text, whereas females tend to engage in an effortful, comprehensive and itemized analysis of all possible information. According to Voss and Van Dyke (2001), women are indeed more influenced by the emotional content than men. They studied the emotional contents of statements, as well as the gender aspect, and find that scenarios constructed by men are more skeletal and heuristic and based on evidence, whereas female representations include more scenario components that took the emotional statements into account. This suggestion is consistent with the Meyers-Levy (1989) Selectivity Hypothesis, women processing more available information while men focusing on just the salient cues from evidence. Subsequently, previous work demonstrates that women engage in less risky behaviors than men and those women’s lower risk preferences and less risky behavior is robust across a variety of contexts (Byrnes et al., 1999; Powell & Ansic, 1997). As emotion words can emphasize the author’s opinion, women systematically process all the words in a review, including the emotion words in order to reduce risky behavior. Comprehensive information processors allow a number of interrelated factors to influence their decision making.

23 situation. The cue is usually regarded to be the most important aspect of the task. The Selectivity Hypothesis maintains that decision making will be guided primarily by this cue selected. This leads to the selective processors being more influenced by emotion words, when perceived as heuristic cues, compared to the comprehensive processors. However, to the author’s knowledge, no literature exists about this perspective. Taking both perspectives into consideration it is not clear whether the review helpfulness will be more influenced by emotion words by one style of processing or the other, or whether both styles of processing will have a moderation effect.

Based on all the above, two plausible competing hypotheses can be formulated;

H7: Emotion words have a more positive effect on review helpfulness when individuals are comprehensive or systematic processors versus selective or heuristic processors. The competing hypothesis is the other way around.

2.4 Gender and Emotional Awareness

The second moderator for this study is emotional awareness. Lane and Schwartz (1987) have defined emotional awareness as the ability to identify and describe one’s own emotions, and those of other people. Emotional awareness is part of the Emotional Intelligence (EI) concept, which is a set of interrelated skills that allows people to process emotionally relevant information efficiently and accurately (Mayer, Caruso, & Salovey, 1999). Creating a higher emotional awareness is perhaps the first step towards furthering the development of this person’s EI (Hein, 2010). Barret et al. (2000) and Bajgar et al. (2005) find females to score higher on levels of emotional awareness than males. Additionally, from the study done by Wang and Hsieh (2007) and the study done by Chentsova-Dutton and Tsai (2007), it is indicated that women are more emotional than men are. Additionally, Schmidt and Cohen (2008) find that females tend to be better at sensing emotional messages in conversations compared to men. O’Kearney (2004) suggests that females more frequently report or express emotion terms referring to inner-directed emotions and more intense positive and negative feelings. In addition, Wang & Hsieh (2007) also stated in their article that women carry out emotional and linguistic tasks more precisely, as they are more sensitive to that kind of stimuli.

24 2.4.1 Emotion words and Emotional Awareness

One could argue how emotion laden words will affect the helpfulness of the review differently in accordance with the emotional awareness level of an individual. Emotion words can be seen as a critical factor to communicate and influence the reader of an online customer review, though only when these words are picked up by the reader and thus influence him or her. The higher the number of emotion cues contained in the message and correctly picked up by the receiver, the stronger the senders’ emotions perceived by the receiver (Harris and Paradice, 2007). However, the extent of influence depends on the level of emotional

awareness of the receiver. The concept of emotional awareness suggests that some more than others pick up, understand, and appreciate emotional messages. Individuals vary in their ability to process information of an emotional nature and in their ability to relate emotional processing to a wider cognition (Connolly, 2013). In this respect, individuals high in emotional awareness use emotional information to guide thinking and behavior (Mayer, 2008). Emotional information can be observed in emotion words, and this information can be a useful source of information by helping one to make sense of and navigate the social

environment (Mayer & Salovey, 1997; Salovey & Grewal, 2005). This means in turn that the more aware someone is of his or her own emotions, the easier it will be for him or her to pick up on what others are feeling and accurately read their wants and needs (Segal, Smith, and Robinson, 2006).

The Emotion as Information Model (EASI-Model) states that emotional expressions affect observers' behavior by triggering inferential processes and/or affective reactions in them (Kleef, 2009). Differences in level of emotional awareness will affect the information processing of these emotional expressions made by others, like emotion words, and thus affecting the reaction of the observer. The understanding of the sender’s emotions is thus necessary to trigger the processing of emotion (words), which is only the case when someone is emotional aware (Chanel et al., 2013). Consequently, a higher emotional awareness leads to someone being more influenced by emotional expression made by others (Mayer

&Salovey, 1997).

Based on the above, when emotion words are included in the text of online reviews it can be assumed that this would influence individuals with a higher level of emotional

25 not have any emotional awareness at all (Hein, 2010). These findings lead for this study, to the assumption that a higher emotional awareness will increase the impact of emotion words on review helpfulness, compared to low emotional awareness.

H9: High emotional awareness will positively moderate the effect from emotion words on review helpfulness.

2.5 Conceptual Model

By summarizing the literature above, the conceptual model is designed with its hypotheses. The aim is to investigate whether the presence of humor and emotion words affect the online review helpfulness and for what kind of information processor and level of emotional awareness. The conceptual model can be seen as existing out of 2 models; the first model are the two independent variables (humor and emotion words), the two moderators (style of processing and emotional awareness) and the dependent variable review helpfulness. The second model consist of the effect of gender on the two variables style of processing and emotional awareness.

26

3. Methodology

The previous chapter focused on the literature of this study and several hypotheses are formulated and drawn into a conceptual framework. In order to test these hypotheses and to answer the research question, an empirical research is conducted. This section, the methodology of this research, elaborates on the research design (data method) followed by the method of data collection. Subsequently, a plan of analysis is described.

3.1 Research Design and Participants

The research question of this study aims to investigate the influence of humor and emotion words on the review helpfulness. The conceptual model, consisting of nine hypotheses, will be empirically tested through quantitative research. An online survey will be send out, which contains self-report measures, to collect data.

In order to test the hypotheses a between-subject factorial experiment is designed, which is often called an independent measures design (Gravetter & Forzano, 2011, p. 230). An experimental design is used in order to determine a causal relation, where one or more of the variables are manipulated and control variables can be included (Malhotra, 2007). A between-subject factorial design is used, in order to avoid carryover effects; an effect that can arise because of participating in several experimental conditions (Greenwald, 1976). This kind of design decreases the chances of participants suffering boredom, as every participant is only subjected to a single scenario instead of many scenarios, and skewing the results, as they become will not become experienced due to the only one test instead of many tests. These effects can exist within-subjects designs where respondents perform in more than one condition. Besides, a within-subject design is not suitable for this study as subject variables (e.g. processing style) impose restrictions on the research design as well as analysis because they cannot be used as within-subject or repeated-measures factors (Judd, Smith & Kidder, 1991).

27

Table 3.1 Between-Subject Design

The research was conducted by gathering data from 181 respondents that filled out the online survey. For this study, approximately 180 respondents were at least needed, as 30 respondents per cell should lead to about 80% power, which is the minimum suggested power for an ordinary study, according to Cohen (1988). However, due to incomplete data of some of the respondents, 11 questionnaires needed to be excluded. The net sample of respondents per condition is listed in Table 3.2. Due to these unfinished and/or incomplete questionnaires, some of the conditions received fewer respondents than 30. The sample consists of 68 (39%) men and 105 (61%) women, with an average age of 39,77, ranging from 21 to 88 years old. The highest frequency for educational level is shown at the level of WO masters or higher.

Condition Humor Emotion Words Number of

respondents (N) 1 Humor Neutral 31 2 Humor Positive 24 3 Humor Negative 25 4 No humor Neutral 28 5 No humor Positive 31 6 No humor Negative 31

Table 3.2 Respondents per condition 3.2 Procedure

Before the questionnaire was sent out to the respondents, a pre-test was done regarding to what will be defined as a humorous product review. The reason for this is because humor is hard to define and not everyone will perceive the same humor as amusement. To assure that the online review that will be used in the questionnaire will indeed be perceived as humorous (funny), and thus causing amusement, a short questionnaire is prepared containing four online

Positive emotion laden review Neutral emotion laden review Negative emotion laden review Humorous review Condition 1 Condition 2 Condition 3

Non-humorous review

28 product reviews. A print screen of the frame of the established website Amazon.com is used in order to give the respondents a more real experience when actually reading the online customer review instead of designing an unknown web template or only presenting a few sentences of text as the online review. However, the name Amazon is removed to minimize the potential bias a respondent has to Amazon and thus to the given reviews in this study. Besides, additional features like star ratings and source information are excluded. This is done in order to prevent respondents being influenced by these kind of features instead of influencing them only with the review text, which is the purpose of this study. One of the four online reviews that are used for the pre-test is shown in Figure 3.1. The four online reviews in the pre-test are exactly the same, except for the last sentence in the review text, where the author is trying to make a humorous impression. Ten respondents for this pre-test are confronted with all four online reviews and are asked after to which extent they think each review is funny, on a seven point Likert scale (1 = totally not funny, 7 =totally funny). The most humorous review, according to these respondents, will be used in the questionnaire for the main experiment of this study. The ten respondents for this pre-test will be gathered online by e-mail, containing the link to the questionnaire in Qualtrics, and will all be Dutch respondents as the questionnaire is in Dutch.

The product used in the reviews is a tablet. The reason, for using a tablet as the product in the online review, is that electronic products are frequently purchased in online shopping websites (Park & Lee 2009) and that consumers tend to rely on comments and reviews from previous users due to the fact that electronic products are generally complicated (Park & Lee 2009). The review text is self-composed and added into the existing online review web template. Besides, the text contains a two-sided message to eliminate the chance of respondents finding the review not helpful and credible like with one-sided messages (Mudambi & Schuff, 2010).

29

Figure 3.1 Print screen of online review

The questionnaire will start off with an online consumer review for the tablet, where the review text will be designed as for one of the six experimental conditions. The most humorous review is selected from the pre-test and used for the three humorous conditions. For the three non-humorous conditions, the humorous sentence in the review text will be left out. Two conditions will include positive emotion words in the review text, two other conditions will include negative emotion words, and the last two conditions will include neutral words.

30 In total, the questionnaire exceeds the suggested amount of items to ask a respondent, however, in order to reliably measure the concepts, items could not be left out (Brent, 2010). In order to make sure the questionnaire will not take respondents too long, in a quasi-pre-test the questionnaire was completed by three random respondents who were time recorded. It took them on average four minutes to fill in the questionnaire, and is therefore considered as an acceptable questionnaire (Brent, 2010).

3.3 Operationalization and reliability of scales

3.3.1 Operationalization of the IV’s Humor and Emotion Words

The independent variables (IV’s) humor and emotion words are manipulated for the six different conditions. These six different reviews are shown in Appendix B. The respondent might be faced with the online review that contains humor in the last sentence of the review text, or a review that does not contains this last humorous sentences. The rest of the review text includes neutral words, positive emotion words or negative emotion words. These positive and negative emotion words are derived from a study done by Li and Zhan (2011). To check these manipulations, two control questions, concerning these two IV’s are placed after the items that measures the dependent variable (DV) review helpfulness. These two questions are asked on a semantic differential scale using polar adjectives, measuring the opinion of the respondents with regard to the positive versus negative and the not funny versus funny concepts on a 5 point scale. The results of the ANOVA tests for the manipulation checks, to determine whether humor and emotion words in the different reviews were correctly manipulated, show significant effects for humor (t (160)=2.504, p = 0.013). Respondents exposed to the humorous review did rate the review as significantly more humorous (M= 2.7) than respondents who were exposed to the non-humorous reviews (M =2.18). However, it should be marked that although the reviews with humor score higher on the scale of finding the review funny versus not funny, the average is not that high on the scale as expected1. Also significant effects are found for emotion words (F=152.523, p < 0.001). Respondents exposed to positive emotion laden review did rate the review as significantly more positive (M=4.15) than respondents exposed to neutral reviews (M=3.42) or negative emotion laden reviews (M=1.61). So, overall the manipulation checks were successful.

1

31 3.3.2 Operationalization of the DV Review Helpfulness

In order to test the hypotheses, the concept review helpfulness should be turned into measurable instruments. Product review helpfulness is a formative construct consisting of three dimensions according to Li et al. (2013): (1) perceived source credibility, (2) perceived content diagnosticity, and (3) perceived vicarious expression. As source characteristics and thus source credibility is not taken into account, it is not suitable for this study to include the perceived source credibility dimension. However, this dimension can be ‘replaced’ by the perceived eWOM review credibility (Cheung et al., 2013), which does make sense instead.A reader who thinks the received review is credible will have more confidence in adopting the eWOM comments and using the review for making purchase decisions (Nabi & Hendriks, 2003). EWOM credibility can be defined as to which extent someone perceives a review as true (Nabi & Hendriks, 2003), as factual (Tseng & Fogg, 1999), and as believable (Fogg et al., 2002). All three are used to determine review credibility for this study so three items are asked for this dimension. The second dimension, perceived content diagnosticity, is measured according to the scale of Jiang and Benbasat (2007), while the items for the third dimension, perceived vicarious expression, are conformed from Manz and Sims (1981) their scale.

Besides Li et al. (2013), a study of Sen and Lerman (2007) investigate the review helpfulness, named under the heading usefulness in their article. Both studies have overlap in measuring review helpfulness. Sen and Lerman their study was designed to have three dependent measures: (i) Attitude toward the Review, (ii) the Attributions about the Reviewer, and (iii) Attitude toward the Product. Only the items to measure attitude toward the review are relevant for this study. According to Sen and Lerman (2007), measuring this dimension of review helpfulness is done by exploring the attitude toward the review by analyzing usefulness, accurateness and informative. As these items are already covered by the items of Li et al. (2013), these will not be measured again. However, the additional (general) question in the study of Sen and Lerman (2007), “Assuming that you were thinking of buying this product, how likely would you be to use the above consumer review in your decision-making?” is added to the items of Li et al. (2013) for this study. This because the attitude towards the reviews and this additional question are of significant value (coefficient α=.85) in their study and has the purpose of being a control question. All the items of the review helpfulness in this study are asked on a 7 point Likert scale (1= totally disagree, 7= totally agree).

32 is conducted for the variables of review helpfulness to verify the internal consistency. As the results show in Table 3.1, all the items will be used to measure this dependent variable. 3.3.3 Operationalization of the Moderator Style of Processing

Measuring the respondent’s style of processing is done by items that are tapping into systematic processing and heuristic processing. As no scale items are designed yet to measure the style of processing according the Selectivity Hypothesis Model (see footnote, an e-mail from Meyers-Levy to the author of this thesis2), the author of this present study searched for a model that covers the concepts of the comprehensive and selective processors of the Selectivity Hypothesis Model. The items used in this study are derived from from Chaiken (1980;1989) and Griffin et al. (2002) (α=.69), who researched the Heuristic Systematic Model, and are measured on a 7 point Likert scale (1= totally disagree, 7= totally agree). The items and reliability analyses can be found in Table 3.1 on page 30. For systematic processing a Cronbach’s Alpha of .759 was found. As this is >.6 the items for this style of processing can be considered as acceptable (Gliem & Gliem, 2003). For heuristic processing a Cronbach’s Alpha of .562 was found, which is <.6 and therefore does not meet the condition for internal consistency (Gliem & Gliem, 2003). However, the Cronbach’s Alpha could not be improved by deleting any of the items. Consequently, for heuristic processing all the items are used for further analysis.

3.3.4 Operationalization of the Moderator Emotional Awareness

With the aim of investigating whether someone scores high or low on EA, an emotion awareness questionnaire is used. The focus will be on the emotional awareness scale that is used in the study of Monti and Rudolph (2014). In their study, an Emotional Awareness scale of 15 items is used existing of three subscales; clarity, description, and expression. The items in this measure scale were drawn from several established measurements of emotional awareness (Salovey et al., 1995; Bagby, Parker, & Taylor, 1994; Kring, Smith, & Neale, 1994). Respondents are asked to rate the degree to which extent they agree up on each item on a 7 point Likert scale (1= totally disagree, 7= totally agree). The items measuring emotional awareness and its reliability analysis can be found as well in Table 3.1 on page 30.

33 All the three constructs, clarity, description and expression, have a Cronbach’s Alpha above .6 and therefore all the items for each construct will be used for further analyses.

Variable Items Cronbach’s Alpha

(α) Review

Helpfulness

(Li et al., 2013)

(Sen & Lerman, 2007)

Perceived eWOM review credibility

1) I think the review is factual. 2) I think the review is accurate. 3) I think the review is credible.

Perceived Content Diagnosticity

1) The review helped me familiarize myself with the product.

2) The review helped me evaluate the product.

3) The review helped me understand the performance of the product.

Perceived Vicarious Expression

1)By reading this product review, I can feel what the author is trying to say about the product and his/her usage

experience.

2) By reading this product review, I can imagine what the author is trying to say about the product and his/her usage experience.

3) By reading this product review, I can envision what the author is trying to say about the product and his/her usage experience.

General question

Assuming that you were thinking of buying this product, how likely would you be to use the above consumer review in your decision-making?

.830

(α does not increase when an item will be deleted)

.827

(α does not increase when an item will be deleted)

.945

(α does not increase when an item will be deleted)

Style of Processing

(Chaiken, 1980; Griffin et al., 2002)

Systematic processing

1) After I encounter information about a review, I am likely to stop and think about it

2) If I need to make a purchase decision after reading a review, the more reviews I read, the better.

3) After thinking about the information in the review, I have a broader understanding.

4) When I encounter information in a review, I read or listen to most of it, even though I may not agree with its

perspective.

5) It is important for me to interpret information in a review in a way that applies directly to my life.

Heuristic processing

1) When I encounter information in a review, I focus on only a few key points.

2) There is far more information in reviews than I personally need.

3) When I read a review, I rarely spend much time thinking about it.

4) If I need to make a purchase decision after reading a review, the advice of one expert is enough for me.

.759

(α does not increase when an item will be deleted)

.562

34 Emotional Awareness (Monti and Rudolph, 2014) Clarity

1) I am usually confused about how I feel 2) I can’t make sense out of my feelings 3) I am usually very clear about my feelings 4) I usually know my feelings about a matter 5) I almost always know exactly how I am feeling

Description

1) It is difficult for me to find the right words for my feelings 2) I am able to describe my feelings easily

3) People tell me to describe my feelings more 4) I find it hard to describe how I feel about people

5) It is difficult for me to reveal my innermost feelings, even to close friends

Expression

1) I keep my feelings to myself

2) I display my emotions to other people

3) I don’t like to let other people see how I am feeling 4) People can read my emotion

5) I hold my feelings in

.792

(α does not increase when an item will be deleted)

.859

(α does not increase when an item will be deleted)

.879

(α does not increase when an item will be deleted)

Table 3.1 Operationalization of the Variables (all measured on a 7-point Likert scale)

3.4 Control Variables

At the end of the questionnaire several control variables were asked. These variables can describe the participants in the research and give insights in the representativeness of the sample compared to the population in the Netherlands. One of the control questions is about what the gender of the respondents is, which is needed in order to test hypotheses 5 and 8. The other control questions can provide extra information about the effects in the conceptual model. The demographic control questions in the questionnaire looked as follows:

1) What is your gender?

This is a nominal variable with the answer possibilities Male and Female. 2) What is your age?

This is an open question resulting in ratio data. 3) What is your highest-rounded educational level?

This is an ordinal variable with a 7-point scale. The answer possibilities are: basic education or lower, Vmbo/mbo 1/ avo lower primary school, Mbo 2/3, Mbo 4, Havo/Vwo, Hbo/Wo bachelor, or Wo masters/higher.

35 These questions were asked as follows:

4) How many hours a day do you spend online?

This is an ordinal variable on a 5-point scale. Answer possibilities are: less than 1 hour, 1-2 hours, 2-4 hours, 4-6 hours and more than 6 hours.

5) To what extent do you agree with the following statement? “I often use online product reviews when buying a product online”.

This is an ordinal (quasi interval) variable on a 7-point Likert scale ranging from totally disagree to totally agree.

3.5 Plan of Analysis

The conceptual model is tested via the statistical software program SPSS, where analyses are done to test the directs effects and the interaction effects. Firstly, there will be looked at whether there are any significant differences between the six different reviews on the review helpfulness, with the independent t-tests and ANOVA tests. As discussed in part 3.3.2, the review helpfulness consists of three concepts; review credibility, content diagnosticity and vicarious expression. These are measured in the first question of the questionnaire and will therefore be called the review helpfulness from question 1 (Q1). The added general question from Sen and Lerman (2007), which is question 2 in the questionnaire, is the fourth dependent variable in this study; the review helpfulness from question 2 (Q2).

As the conceptual model in this study also contains interactions variables, the second step of the analysis is a hierarchical multiple regression analyses to test which variables and or interaction effects influence the review helpfulness and to what extent they influence the review helpfulness. For each of the four DVs a multiple regression analysis is performed, so four regression analyses will be done. The first step of the hierarchical regression analysis is the “base model”, only modeling the independent variables (the experimental variables) humor and emotion words and their interaction. Hereafter, in the second step, the first moderator, style of processing, and its interaction with both the IV’s is added. Subsequently, the second moderator, emotional awareness, is with its interaction with the IV emotion words is added to the base model in the third step. The final step is testing the IV’s, the moderators and their interaction effects in one.

36 gender as a moderator into a multiple regression while leaving out the previous moderators style of processing and emotional awareness.

3.6 Validity Model Check

A validity check is developed in order to test the dependent variable review helpfulness. In this model there is critically looked at whether the 3 constructs from Q1 (credibility, content diagnosticity and vicarious expression) indeed explain the general question (Q2) measuring review helpfulness as well. A correlation analyses is carried out to establish whether these 3 constructs correlate with review helpfulness. Next to this, a regression analyses is performed to test to what extent each construct influences the dependent variable Q2. An important assumption for carrying out regression analysis is the absence of multicollinearity. In order to make sure there is no multicollinearity present in this model, the VIF values are checked and they are all below 3 (see Table 3.2), so no multicollinearity exists.

The results of the regression analysis show that the 3 constructs explain 37% of the variance of review helpfulness (Adjusted R2 = 0.369). This means that the 3 constructs from Q1 indeed explain Q2 in this study, so it is in line with the literature from Li et al. (2007). However, 63% of the variance of review helpfulness cannot be explained by these three variables; there must exist other variables (constructs) that have influence on review helpfulness. Unstandardized Coefficients (B) p-value VIF Credibility 0.257 0.0320 2.731 Content Diagnosticity 0.235 0.0530 2.503 Vicarious Expression 0.346 0.0000 1.67

38

4. Results

This chapter presents the results of the analyses done with the data of the online questionnaire that were filled in by the respondents. Descriptive statistics will be discussed firstly. Hereafter, the results of the t-tests and ANOVA’s of the IV’s on the DV’s are described. Subsequently, the multiple regression analyses results will be described. A regression analysis is performed in order to establish whether review helpfulness can be predicted with the IV’s and the moderators used in this study. The interaction variables are included in this analysis from each independent variable with each moderator. Subsequently, the results are described of gender on the style of processing and the level of emotional awareness. Finally, an alternative model is tested.

4.1 IV’s on Review Helpfulness (Q1 and Q2)

In Table 6.1 in Appendix C the frequency and mean results of the dependent variable review helpfulness (Q1 & Q2) per independent variable is shown. The first question (Q1) in the questionnaire measures review helpfulness by the three constructs review credibility, content diagnosticity and vicarious expression. Question number two (Q2) in the questionnaire represents the fourth dependent variable (the general question) measuring review helpfulness. These descriptive statistics (in Table 6.1) of humor and emotion words on these three constructs (Q1) and on the general question (Q2) show that the means of both IV’s lie very close to each other for each dependent variable. In the table 4.1, on the next page, the mean of each condition per dependent variable and the overall means per dependent variable are shown.

In order to test whether the mean differences were significant, an ANOVA and independent samples t-test are carried out. The Levene’s test shows that there is homogeneity of variance amongst all groups. This means that the assumption of homoscedasticity for carrying out a t-test or ANOVA is met. The independent samples t-test for humor on review helpfulness (Q1) is for none of the three constructs of review helpfulness significant (see Table 6.2a in Appendix D). The same counts for the means of emotion words on review helpfulness (Q1) (see ANOVA in Table 6.2a in Appendix D ). Therefore, overall there are no significant differences between the means of humor and emotion words on the 3 constructs of review helpfulness (Q1)3.

39 Positive emotion words Neutral emotion words Negative emotion words Overall means (µ) Humor M DV1 = 3.639 M DV2 = 3.319 M DV3 = 4.681 M DV4 = 3.833 M DV1 = 3.860 M DV2 = 3.763 M DV3 = 4.247 M DV4 = 4.387 M DV1 = 3.293 M DV2 = 3.027 M DV3 = 4.493 M DV4 = 4.320 M DV1 = 3.614 M DV2 = 3.404 M DV3 =4.512 M DV4 = 4.296 Non-humor M DV1 = 3.527 M DV2 = 3.194 M DV3 = 4.462 M DV4 = 4.065 M DV1 = 3.607 M DV2 = 3.417 M DV3 = 4.679 M DV4 = 4.571 M DV1 = 3.516 M DV2 = 3.312 M DV3 = 4.839 M DV4 = 4.581 M DV1 = 3.522 M DV2 = 3.253 M DV3 = 4.651 M DV4 = 4.323 Overall means (µ) M DV1= 3.576 M DV2 = 3.249 M DV3 = 4.558 M DV4 = 3.964 M DV1= 3.860 M DV2 = 3.763 M DV3 = 4.247 M DV4 = 4.387 M DV1= 3.480 M DV2 = 3.262 M DV3 = 4.683 M DV4 = 4.500 M DV1= 3.5804 M DV2 = 3.3490 M DV3 = 4.562 M DV4 = 4.3060

M DV1 = Mean of Dependent variable Review Credibility M DV2 = Mean of Dependent variable Content Diagnosticity M DV3 = Mean of Dependent variable Vicarious Expression M DV4 = Mean of Dependent variable General Question (Q2)

Table 4.1 Overview of Means per condition

Testing the main effects of humor and emotion words on review helpfulness (Q2) is done in the same way as for the three constructs of review helpfulness from Q1; performing an ANOVA and independent samples t-test (see Table 6.2b in Appendix D). Regarding humor, the results show that there are no significant differences between the reviews with or without humor (t (168) =0.09, p =0.928. The ANOVA (see also Table 6.2b in Appendix D) shows that that there are no significant differences between reviews that contain positive, neutral or negative emotion words on review helpfulness (Q2) (F(2.167)=1.483, p=.230).