The effect of word-of-mouth, sentiment, and disappointment on product sales of videogames.

University of Amsterdam

Faculty of Economics and Business Master in Business Studies

The Effect of Word-of-Mouth, Sentiment, and

Disappointment on Product Sales of Videogames.

Author: Renzo Hijman (10003553)

University of Amsterdam, Faculty of Economics and Business MSc Business Studies, Marketing Track

Supervisor: dr. Frederik Situmeang Second Supervisor: dr. Umut Konus

Table of Contents 1. Introduction……….4 2. Theory………..6 3. Method………...20 4. Results……….24 4.1 Test of Hypotheses………...34

5. Discussion and Conclusion………40

5.1 Unexpected Results………..40

5.2 Interpretation of Results……….43

5.3 implications, Limitations, and Future Research………...45

5.4 Implications to Theory……….46

5.5 Conclusion……….48

Abstract

The purpose of this study is to contribute to the marketing literature and marketing practice by examining the relationship between consumer and expert reviews (opinion mining and Word of Mouth (WOM)) of videogames and the global product sales of these video games. A set of hypotheses was developed, based on earlier research of Miller (2005) and MacInnis, Rao and Weiss (2002), and a dataset was obtained from websites, such as Metacritic.com and vgchartz.com, to test the hypotheses by

performing regression analyses. Online WOM in consumer- and expert product evaluations/scores, of videogames, has a positive effect on the product sales of videogames. Also sentiment displayed in consumer- and expert product reviews, of videogames, has a positive effect on the sales of videogames. Disappointment displayed in consumer product reviews, of videogames, has a negative effect on the sales of videogames. The results have several implications for marketing practice and marketing research. This study is the first attempt to apply opinion mining to the discussion of videogames. Opinion mining offers a new way for studying the impact of consumer and experts reviews of videogames on the sales of videogames. The results of this study contribute to the ongoing discussion about the impact of consumer and expert reviews on the global product sales, which has obtained mixed findings. The results have important managerial implications, and are of commercial importance. It is important for managers to understand how their product sales are affected by the sentiment and disappointment displayed in consumer- and expert reviews.

Keywords: Consumer Evaluation, Expert Evaluation, Word of Mouth, Sentiment Mining, Global Product Sales

1. Introduction

More and more people make offline decisions that are based on online information or online conversations. Textual online reviews are used more often as information sources than traditional information sources (Godes & Mayzlin, 2004). According to a market study of Channel Advisor (2010) 92% of online consumers read product reviews when considering buying a product. Also online shoppers have more trust in opinions of other shoppers than in marketer-initiated sources (eMarketer, 2010).

It is clear that textual online reviews can have a tremendous effect on product sales of retailers (Chevalier & Mayzlin, 2006). However, a market study by

Econsultancy (2011) shows that 81% of (online) retailers have no understanding why customers leave without buying a product. Thus, there is a managerial need to come up with insights into the influence of text-based online reviews on product sales to improve the understanding of retailers.

Current research that is focused on textual online reviews offers almost no guidance. Most of the studies that are done used only quantitative measures to measure the content of the review (Mudambi & Schuff, 2010). The current studies focus on the influence of star ratings, review volume and word of mouth (WOM) on sales, but the results of these studies are mixed (Chen, Wu & Yoon, 2004; Chevalier & Mayzlin, 2006; Dellarocas Xiaoquan & Awad, 2007; Duan, Gu &Whinston, 2008).

Currently, there are now studies that focus on the textual properties of reviews and their impact on retail performance (Chevalier & Mayzlin, 2006). It might be that positive and/or negative emotion (for example, “I love the new Grand Theft Auto” or “I hate the new Call of Duty”) provided in texts (for example textual online reviews) has an influence on respondents’ attitudes (Cohen et al. 2008). However it is not clear if affective words serve as predictors of the impact of customer reviews on retail

success, because there is limited evidence of a relationship between affective

activation and product evaluations (Andrade, 2005). Beyond the review content, there are also researchers that looked to the sentiment in online blogs and the impact on movie success (Mishne & Glance, 2006).

In light of this gaps and concerns this research aims to provide theoretical and managerial guidance on the influence of online word of mouth (WOM), sentiment (positive emotion) and disappointment (negative emotion) of textual online reviews on product sales of videogames from stores. First, this research looks at the impact of online WOM on global product sales. Previous work on online WOM measured the construct by looking to review ratings and postdate (Duan, Gu &Whinston, 2008). This research wants to measure WOM by also looking at the review ratings of

consumers and experts and their impact on global product sales. Second, this research looks at the impact of sentiment in online textual reviews on global product sales. Previous work on sentiment (positive emotion) measured the construct by looking at weblogs about movies both before and after the movie was released (Mishne, Glance, 2006). This research wants to focus mainly on textual reviews from Metacritic.com of videogames and their impact on global product sales of this videogames from stores. Third, this research looks at the impact of disappointment (negative emotion) in online textual reviews on product sales of videogames from stores. From previous work it is not clear if affective cues (positive/negative) serve as predictors of the impact of textual online reviews on retail success, because there is limited evidence of a relationship between the two constructs (Andrade, 2005), so this research wants to try to find evidence for this relationship. This study is guided by the following research question: what is the effect of online Word of Mouth, sentiment and

disappointment in online textual reviews of videogames on global product sales from stores?

This study makes use of a secondary dataset that is based on online textual reviews on metacritic.com. The researcher will perform text analysis with the software program MAXQDA version 11 for Mac. Ghose and Ipeirotis (2011) used opinion mining in their research. They found that the information, subjectivity and readability of reviews matter in influencing product sales. Ghose and Ipeirotis (2011) focused on the market of digital cameras and audio and video players, so opinion mining (analysis) is a useful tool to use in this current research.

The remaining parts of this research are organized as follows. A review of the related literature is presented in the next section. The section after that introduces the methodology and the dataset used in this study. Section 4 reports the results of this study and section 5 concludes and discusses the implications of the study and suggestions for future research.

2. Theory

In the past a lot of researchers have looked at the influence of online

reviews/evaluations on product sales (Litman, 1983; Reddy, Swaminathan and Motley 1998; Basuroy, Chatterjee and Ravid, 2003; Elberse, Eliashberg, 2003; Reinstein, Snyder, 2005; Zhang, Dellarocas, 2006; Boatwright, Basuroy and Kamakura, 2007). In table 1 there is an overview of their findings.

From research of Gemser, van Oostrum and Leenders (2006) it was argued that product reviews/evaluations play two roles. First, Gemser, van Oostrum and Leenders (2006) talk about the influence effect of reviews. This type of effect actively influences consumers in their selection process. Second, Gemser, van Oostrum, and Leenders (2006) talk about prediction effect of reviews. This type of effect shows if

the product will be a success or a failure. Therefore, product reviews/evaluations are an important information source for consumers.

Miller (2005) argued that expert evaluations/scores could serve as an input for the formation of individual attitudes. For example, the scores of game experts or film reviewers can affect an individual attitude about a videogame or movie. The game industry is an example of a market where consumers can access the expert

evaluations/scores of products. Furthermore, there are now websites that gather all these expert evaluations/scores as well as consumer evaluations. As a result, one can expect that, if a videogame receives a positive review score, these positive review scores may contribute to shaping the purchase intentions of consumers positively. This will have a positive influence on the product sales of that particular video game.

H1a: Positive WOM, of experts, has a positive impact on product sales of videogames from stores (see figure 1).

More and more people are using the Internet to communicate. This is possible, because of the introduction of Web 2.0. O’Reilly (2005) argues that interaction is one of the core characteristics of Web 2.0. There are different platforms on the Internet

can communicate one-on-one or in groups. Opinions and information are exchanged on the Internet. This is possible with another distinguishing mark of Web 2.0 namely: user generated content (Shao, 2009). People who use the Internet are now able to create their own content. People are also able to influence other people through the information they provide or the opinions they put on the Internet.

When this is done in an offline environment researchers call this Word-of-Mouth. Word-of-mouth is the informal communication between two private parties concerning evaluations of goods and services (Dichter, 1966; Fornell & Bookstein, 1982; Singh, 1988; Westbrook, 1987). Word of Mouth (WOM) may be positive, negative or neutral (Anderson, 1998). In the literature there is theoretical support that WOM has an impact on the actions of consumers. In the article of Banerjee (1992, 1993) there are two models that suggest that people are influenced by the information and opinions of other people.

When people communicate in an online environment researchers call it online Word-of- Mouth or electronic Word-of-Mouth (eWOM). Davis and Khazanchi (2008) came up with a workable definition of online word of mouth. According to Davis and Khazanchi (2008) online word of mouth is when people have the ability to

communicate and share experiences with each other online. The most important difference between WOM and online WOM is the distance between sender and receiver (Steffes & Burgee, 2009). Offline WOM is very intimate. On the other hand, with online WOM or communication the moment of interaction is separated by time and space. A second difference between offline WOM and online WOM is the size of the audience (Steffes & Burgee, 2009). Online WOM has a larger scope.

On the Internet there are lot of platforms where people can communicate online. The platforms differ in their scope and the degree of online interactivity

(Litvin, Goldsmith & Pan, 2008). Blogs and online communities have a great scope, but they face difficulties with interactivity. Compared to chat rooms where people face no difficulties with interactivity.

Another important online platform are review sites. Like metacritic.com. Research of Porter Novelli (2011) indicates that 78% of the Dutch respondents read product reviews online on websites. Furthermore, 49% of the Dutch respondents find online reviews more important if there are many of them, so online reviews and conversations offer a cheap and easy opportunity to measure WOM according to Godes and Mayzlin (2004).

Another important difference between online and offline one-on-one

communication is the trust the receiver have in the sender of the message (Steffes & Burgee, 2009). When someone communicates offline most of the time he or she knows the one he or she speaks with. The receiver is most of the time able to form a judgment about if he trusts the sender or not. With online one-on-one communication this is not possible, because most of the time a receiver does not know the sender. The social influence should therefore be different between online and offline one-on-one communication. With eWOM the message also plays an important role. According to Sussman and Schneier Siegal (2003) there are three important rules for online

messages: (1) quality of the argument, (2) reliability of the source, and (3) the degree of usability of the information.

Another research article of Mayzlin (2004) looks at online WOM and the potential it has for firms. Mayzlin (2004) argues that firm can create firm-to-consumer communication that looks like consumer-to-consumer communication, when firms pretend that they are consumers themselves. Mayzlin (2004) concludes that even if this is possible for the firm, consumers will still look at anonymous online posts.

Therefore, it can be a good strategy for a firm to pretend that they are a consumer in an online environment. Therefore textual online posts are an important source of information for firms.

Over the years researchers have noticed the importance of WOM (Coleman, 1966). When people started to use the Internet to write recommendations and reviews the reach of WOM became bigger. This started to attract researchers to study WOM and the effects of WOM in this digital age (Chen & Xie, 2004; Dellarocas, 2003).

There have been a few empirical attempts to provide support for the effect of WOM on product sales. Chevalier and Mayzlin (2003) found that an improvement in book reviews leads to an increase in relative sales at Amazon.com. Similar research has focused on music sales (Dhar & Chang, 2009). Dhar and Chang (2009) found that future music sales have a positive correlation with the volume of blogs about a music album and if a major music label released the album. However, Dhar and Chang (2009) where not able to conclude, that blog chatter really was the only cause of an increase in music sales. On the other hand, Duan, Gu and Whinston (2008) found that WOM is important to generate and sustain retail revenue. That’s why in this research the expectation is that positive WOM has a positive impact on product sales of videogames from stores.

Miller (2005) argued that expert evaluations/scores could serve as an input for the formation of individual attitudes. A similar argument can be made with regard to consumer evaluations/ scores of videogames as these evaluations too can contribute to shaping the purchase intentions of other consumers positively. However, consumer evaluations/scores of videogames are also a reflection of how much the consumer loves the product (Liu, 2006). Consumers directly signal experiences and satisfaction with the product (Chen, Fey and Wang, 2011). So, positive WOM/ positive review

scores, of consumers, have a positive impact on product sales of videogames from stores.

H1b: Positive WOM, of consumers, has a positive impact on product sales of videogames from stores (see figure 1).

However, there are possible other influences on product sales. Mishne and Glance (2006) concludes that sentiment is an important predictor of the success of movies. Sentiment analysis, which is also known as opinion mining, is related to online WOM (Jansen, Zhang, Sobel and Chowdury, 2009). Sentiment is the next possible predictor of product sales, so this is the focus of the next section.

Table 1: Previous Research Related to Professional and Consumer Reviews.

Study Method Data Review Findings

Litman (1983) Multiple Regression Movies (1972-1978)

Critics Critics‘ ratings are significant factors to explain box office revenues. Basuroy, Boatwright and Kamakura (2003)

Diffusion model Movies

(1992-2001)

Critics Critics’ reviews are significantly correlated with box office revenues Elberse and Eliashberg (2003)

Demand/Supply model Movies (1999)

Critics Positive reviews

mean more opening

revenue, but less positive reviews correspond to a higher number of opening screens. Godes and Mayzlin (2004)

Multiple Regression TV shows (1999-2000)

Amateur Online

Word-of-Mouth (WOM)

Reinstein and Snyder (2005)

Difference in difference Movies (1999)

Critics The influence of

critics’ reviews is smaller than previous studies but still detectable. Dellarocas,

Awad and Zhang (2005)

Diffusion/forecasting model

Movies (2002)

Amateur Online amateur

ratings can be used as a proxy for WOM Chevalier and

Mayzlin (2006)

Difference in difference Books (2003-2004)

Amateur Online amateur

book ratings affect consumer purchasing behavior Zhang and Dellarocas (2006)

Diffusion model Movies

(2003-2004)

Amateur/ Critics Significant influence of star ratings of online reviews on demand.

The effect of sentiment (positive emotion) in online textual product reviews on product sales of videogames from stores.

Blogs or online reviews have become popular types of online media to express opinions (Liu, Huang, An & Yu, 2007). Blogs and online reviews are some kind of online diaries and they can show the opinions of people on particular subjects, such as on mainstream topics (for example music or movies) or subjects in which people have a high personal interest (Kumar, Novak, Raghavan & Tomkins, 2004). Many bloggers or consumers express their opinions online, so blogs or online reviews can serve as a good indicator of sentiments and opinions.

Sentiment is a thought, idea or opinion based on a feeling about a situation. Or a way someone thinks about something (Cambridge Dictionaries Online, 2013).

This research focuses on the predictive power of sentiments, which are

reviews that contain a review on a product, because what the general consumers thinks of a product or what experts think about a product can have an influence on how good a product sells. The researcher thinks a good understanding of the opinions and sentiments, which are expressed in online textual reviews, is of high importance, because it can be a very good indicator of the product’s sales performance.

Earlier work on the predictive power of blogs and online reviews has used the volume of blogs and online reviews or link structures to predict the trend of product sales (Gruhl, Guha, Kumar, Novak & Tomkins, 2005; Gruhl, Guha, Liben-Nowell & Tomkins, 2004). They were not able to consider the effect of the sentiments present in the blogs and online reviews, but in the literature there is evidence that sentiment has an impact on the actions of consumers. There seems to exist a strong correlation between blog mentions and sales (Gruhl, Guha, Kumar, Novak & Tomkins, 2005; Gruhl, Guha, Liben-Nowell & Tomkins, 2004), but only using the volume or link structures as predictor of sales performance is not good enough. This is illustrated by Liu, Huang, An, and Yu (2007) in their research paper. They show with their example that sentiments expressed in online reviews or blogs are more predictive than using volumes. Mining sentiments and opinions from online reviews is necessary for predicting product sales.

Most of the existing work on opinion mining, also called sentiment mining, focuses on determining the sentiment in different documents. Some of the existing work tries to come up with positive and negative classifiers at document level. In earlier work researchers used different methods to measure sentiment in all different kind of documents. First, were Pang, Lee, and Vaithyanathan (2002), which used three machine learning approaches to label the polarity of IMDB movie reviews. Second, there are researchers who work on a deeper level and they used words to

classify subjects. They classify the words in two groups. The words are good or bad. Then they use a special function to estimate how good or bad the scores for the

documents are. Kamps and Marx (2002) tried to evaluate the semantic distance from a word to good/bad with the use of WordNet.

Another research of Turney (2001) measures the strength of sentiment by looking at the difference of the mutual information between the given phrase and “excellent” and the mutual information between the given phrase and “poor”. Two other researches of Pang and Lee (2005) and Zhang and Varadarajan (2006) tried to determine opinions in reviews with different rating scales (for example the number of stars). For this research, the researcher is convinced by the earlier work, to stick with the kind of analysis mentioned above.

According to Gruhl, Guha, Kumar, Novak and Tomkins (2005) online blogs/ reviews can predict the sales of books, but in this research the researchers looked at the raw number of blogs, which was a good predictor.

However, in reviews or blogs people have a lot of opinions: positive, negative, mixed or neutral. Sentiment analysis makes it possible to look at the level of positive or negative mentions in blogs or reviews (Mishne & Glance, 2006).

There is a lot of work of sentiment analysis on reviews/blogs. Earlier work showed that when there are more references to a book in online reviews this will be followed by increasing sales of that particular book (Gruhl et al., 2005). Tong (2001) had similar results in his earlier work. Tong (2001) concludes that references, that people make, of movies in newsgroups has a positive impact on sales.

However in the research of Mishne and Glance (2006) the correlation between sentiment and sales was not high enough to come up with a model for sales that was only based on sentiment.

On the other hand, Liu, Huang, An & Yu (2007) came up with two models in their research. First model is the Sentiment-PLSA (S-PLSA) model. In this model a blog is viewed and the researchers are able to obtain a summary of the sentiment information that is hidden in the blogs. With the second model (the ARSA model) the researchers are able to predict product sales by looking at the blog information from the S-PLSA model. Experiments conform the superiority of the proposed models. That is why in this research the expectation is that sentiment has a positive impact on product sales.

H2a: Sentiment (positive emotion), of consumers, has a positive impact on product sales of videogames from stores (see figure 2).

H2b: Sentiment (positive emotion), of experts, has a positive impact on product sales of videogames from stores (see figure 2).

However in this research there will be looked at another possible influence on product sales, namely disappointment (negative emotion). According to research of Ladhari

(2007) WOM is related to emotion. Therefore, disappointment (negative emotion) is the focus of the next section

The effect of disappointment (negative emotion) in online textual product reviews on product sales of videogames from stores.

Existing theories of product choice mainly look at consumers as people who analyze the tangible features of the products before making a purchase decisions (Bettman, Johnson and Payne, 1991; Meyer and Kahn, 1991). According to these existing theories, people make decisions through heuristic decision rules or either on the basis of a cost-benefit analysis. However, these theories don’t take in account the role affective experience might play in the choices consumers make (Holbrook and Hirschman, 1982). On the other hand, research that has the focus on attitude judgment recognizes the importance of affect (emotion) in judgment (Darke, Chattopadhyay and Ashworth, 2002). Darke, Chattopadhyay & Ashworth (2002) expanded the role of affect (emotion). They did this by examining choices under high elaboration

conditions and whether affective cues (emotion) are likely to be used when making these choices. Before this research of Darke, Chattopadhyay & Ashworth (2002) the view in other researches was that affect (emotion) influences choice only under low elaboration conditions. Low elaboration conditions mean that consumers don’t have the motivation to make accurate decisions or don’t have the capacity to engage in information processing.

Russell and Carroll (1999) describe affect as the subjective moods and

feelings (emotion) we have in our minds, rather than thoughts about specific events or objects. A more comprehensive description, for affect, is given by Darke,

describe affect as general positive or negative feelings (emotion). Also, they make a distinction between mood and affective cue, which is important for this current research. Isen (1984) and Schwarz (1990) describe mood as a broad negative or positive state that arises over time. On the other hand, affective cue is described as more discrete positive or negative feelings (emotion) that can be attached to specific products (Zajonc, 1990). Most of the existing studies, about the impact of affect in judgment, used mood state manipulations (Darke, Chattopadhyay & Ashworth, 2002). In contrast, this current research uses affective cues (emotional cues).

Past research argues that emotion (positive/negative) can have a significant impact on evaluation (Lench, Flores and Bench, 2011) decision-making (Shiv and Fedorikhin, 1999) and choice (Loewenstein and Lerner, 2003).

More and more researchers are focusing on textual properties in reviews and their influence on retail performance. Affective cues (emotion) might have an impact on consumers’ attitudes (Cohen et al., 2008). However it is not clear if affective words serve as predictors of the impact of customer reviews on retail success, because there is limited evidence of a relationship between affective activation and product evaluations (Andrade, 2005).

In the past there have been some research about emotion and the effect on sales. For example Sutton and Rafaeli (1988) found a negative relationship between the emotion displayed by employees to consumers and the product sales, but the sample where only stores from America and Canada. Tsai (2001) came up with opposite conclusions. Tsai (2001) concluded that positive emotions displayed by employees have a positive effect on consumer willingness to return to the store.

MacInnis, Rao and Weiss (2002) found similar results, but focused on

of MacInnis, Rao and Weiss (2002) it is shown that advertisements for frequently purchased brands in mature categories were likely to create greater sales when they used affectively based cues. From research of Gemser, van Oostrum and Leenders (2006) it was argued that product reviews play two roles. First, Gemser, van Oostrum and Leenders (2006) talk about the influence effect of reviews. This type of effect actively influences consumers in their selection process. Second, Gemser, van Oostrum, and Leenders (2006) talk about prediction effect of reviews. This type of effect shows if the product will be a success or a failure. Therefore, product reviews are an important information source for consumers.

Miller (2005) argues that expert reviews can serve as an input for the formation of individual attitudes. For example, an individual attitude about a videogame can be affected by the opinion of game experts. The game industry is an example of a market where consumers can access the expert reviews of products. Furthermore, there are now websites that gather all these expert reviews as well as consumer reviews. As a result, one can expect that, if a videogame receives a negative expert review, these negative reviews may contribute to shaping the purchase

intentions of consumers negatively. This will have a negative influence on the product sales of that particular video game. That is why in this research the expectation is that disappointment of experts, in product reviews has a negative impact on product sales.

H3a: Disappointment (negative emotion), of experts, has a negative impact on product sales of videogames from stores (see figure 3).

Miller (2005) argued that expert reviews could serve as an input for the formation of individual attitudes. A similar argument can be made with regard to consumer reviews

of videogames as these reviews too can contribute to shaping the purchase intentions of other consumers negatively. However, consumer reviews of videogames are also a reflection of how much the consumer loves or hates the product (Liu, 2006).

Consumers directly signal experiences and satisfaction or dissatisfaction with the product (Chen, Fey and Wang, 2011). So, disappointment (negative emotion), of consumers, in product reviews has a negative impact on product sales of videogames from stores.

H3b: Disappointment (negative emotion), of consumers, has a negative impact on product sales of videogames from stores (see figure 3).

In the next section the research design will be explained, which will be used to answer the research question.

3. Method

In order to analyze the relationship between online word of mouth,

disappointment (negative emotion), sentiment (positive emotion) and their effect on the product sales of videogames the researcher will use a dataset that consist of consumer and expert reviews on videogames from metacritic.com. To analyze the dataset from metacritic.com the researcher will perform text analysis with the software program MAXQDA version 11 for Mac. MAXQDA is one of the best qualitative data analysis software in the world. There is no better software, the researcher can code the data in different ways, MAXQDA is user friendly, and the researcher can use the mixed methods that MAXQDA provides (MAXQDA, 2014).

To gather information about the product sales of the videogames and movies the website vgchartz.com is used. On this website the researcher can find the global sales of all the videogames of all the video platforms (for example PS4 or Xbox one).

In MAXQDA version 11 for Mac the researcher can import certain positive words (for example fun, must buy or great) and than can run an analyses to look if people use these kind of words in their reviews. However, it is not sufficient enough to only use 3 words, so the researcher used oxforddictionaries.com to come up with synonyms of the words fun and great. Synonyms for the word ‘fun’ that are used in this research are: enjoyable, amusing, diverting, pleasurable, pleasing, agreeable, and interesting.

Synonyms for the word ‘great’ that are used in this research are: enjoyable, amusing, delightful, lovely, love, pleasant, congenial, diverting, exciting, thrilling, excellent, marvelous, wonderful, superb, first-class, first-rate, admirable, fine, splendid, very good, and good (Oxford Dictionaries, 2014). In MAXQDA version 11 for Mac the researcher can also import certain negative words and than can run an analysis to look if people use these kind of words in their reviews. The negative words that are used in this research are: underwhelming, disappointment, failure, disaster, flop, debacle, fiasco, disappoints, hate, dislike, aversion, distaste, disfavor, disgust, bad, poor, unsatisfactory, awful, worthless, miserable, shit, boring, dull, worst, trash, and sucks (Oxford Dictionaries, 2014). The Oxford English Dictionary (OED) is considered to be the authority on the English Language. It is a guide to the meaning, and

pronunciations of 600.000 words from across all the countries in the world were the people speak English. In the OED people are able to learn about the history of individual words (Oxford Dictionaries, 2014).

Table 2. Number of Reviews, Number of Games and Global Product Sales

Platform Number of Reviews Number of Games Global Product Sales

Playstation 1 3147 289 730.660.000 Playstation 3 44.710 2450 789.400.000 Playstation 4 371 10 16.780.000 PSP 4039 1005 285.930.000 PS Vita 2969 426 22.730.000 Xbox 8277 861 258.410.000 Xbox 360 59.184 2478 856.200.000 Xbox One 1150 41 12.810.000 Nintendo Wii 10.184 1990 875.370.000 Nintendo 3DS 2840 508 132.860.000

Note: Global Product Sales is in millions. Global Product Sales from vgchartz.com

In table 2 first there is some information about the dataset. In the dataset there were reviews about a lot of games from various platforms (for example Playstation 1). Second, there is some information about the number of games of each platform (for example 289 videogames from Playstation 1). Finally, there is some information about the global videogames sales of the different video platforms (for example 730.660.000 games for Playstation 1).

3.1 Model Specification

The hypotheses are tested according to the following statistical model. A detailed overview of the variables is presented in Table III. The model takes the reviews from consumers and experts into account and the model takes also the scores from

consumers and experts into account.

SALESi = β0 + β1.CEi + β2.RACINGi + β3.ACTION ADVENTUREi + β4.FPSi +

β5.MATUREi + β6.TEENi + β7.EVERYONEi + β8.EEi + β9.METASCOREi +

Table 3 Overview of the Variables Variable name Description

SALESi The worldwide unit sales (million copies) of videogames. RACINGi Genre classification of videogame is Racing. Dummy variable. ACTION

ADVENTUREi Genre classification of videogame is Action Adventure. Dummy variable.

FPSi Genre classification of videogame is FPS (First Person Shooter). Dummy variable. MATUREi ESRB rating of videogame is Mature. Dummy variable.

TEENi ESRB rating of videogame is Teen. Dummy variable.

EVERYONEi ESRB rating of videogames is for Everyone. Dummy variable. CE i Consumer review of video game i (Sentiment).

EE i Expert review of video game i (Sentiment). METASCOREi Expert scores of video game i (Word of Mouth). USERSCOREi Consumer scores of video game i (Word of Mouth). NCEi Consumer review of video game i (Disappointment) NEEi Expert review of video game i (Disappointment)

The expert scores (WOM) and consumer scores (WOM) are obtained from Metacritic.com video game database. Metacritic.com is a widely used site for evaluation scores and information about video games (Luan & Sudhir, 2010). Also the positive expert reviews (sentiment) and positive consumer reviews (sentiment) are obtained from Metacritic.com. The negative expert reviews (disappointment) and negative consumer reviews (disappointment) are obtained from Metacritic.com. The information about the sales is obtained from vgchartz.com. The other variables such as age rating dummies (TEENi, MATUREi, and EVERYONEi) and the genre dummies (FPSi, ACTION ADVENTUREi, and RACINGi) are obtained from Metacritic.com and vgchartz.com (see table 3).

4. Results

Table 4 to Table 4.8 presents the descriptive statistics and the correlation matrix of the key variables of the study. The correlation analysis provides preliminary evidence that positive consumer reviews of videogames have a positive and significant correlation with global product sales of videogames (3DS: r = .663, p < .01; PS1: r = .548, p < .01; PS3: r = .645, p < .01; PSP: r = .588, p < .01; Vita: r = .289, p < .01; Xbox: r = .326, p < .01; Xbox One: r = .535, p< .01; Xbox360: r = .681, p < .01), and also the analysis provides preliminary evidence that positive expert reviews of videogames have a positive and significant correlation with global product sales of videogames (3DS: r = .487, p< .01; PS1: r= .276, p< .01; PSP: r = .396, p< .01; Vita: r = 244, p< .05). There is a significant correlation between global product sales and the user score (consumer score) from Metacritic.com (PS1: r = .323, p< .01; PS3: r = .062, p< .05; PSP: r = .123, p< .05; Xbox: r = .254, p< .01; Xbox360: r = .145, p< .01), and also between global product sales and the expert score (meta score) from Metacritic.com (PS1: r = .373, p< .01; PSP: r = .260, p< .01; Wii: r = .153, p< .01; Xbox: r = .298, p< .01; Xbox360: r = .213, p< .01). The analysis provides preliminary evidence that negative consumer reviews of videogames have a negative and significant correlation with global product sales of videogames (3DS: r = -.631, p< .01; PS1: r = -.459, p< .01; PS3: r = .703, p< .01, PSP: r = .555, p< .01; Vita: r = .355, p< .01; Xbox: r = -.428, p< .01; Xbox360: r = -.603, p< .01, Xbox One: r = -.747, p< .01), and also the analysis provides preliminary evidence that negative expert reviews of videogames have a negative and significant correlation with global product sales of videogames (3DS: r = -.169, p< .01).

Table 4 Descriptive Statistics and correlations 3DS

Variable name Mean SD 1 2 3 4 5 6 7 8 9 10 11 12

1 SALESi .354 1.234 2 CE i 4.080 12.633 .663** 3 ACTION ADVENTUREi .016 .125 -.007 -.023 4 RACINGi .022 .146 -.031 -.047 -.019 5 MATUREi .028 .164 -.010 .144** -.021 -.025 6 EVERYONEi .500 .500 .046 -.072 -.063 .095* -.168** 7 USERSCOREi 6.291 2.062 .015 .308** -.038 -.102 .125 .155* 8 METASCOREi 65.683 13.289 .347** .490** -.082 -.056 .092 .013 .747** 9 TEENi .165 .372 -.035 -.006 -.056 .007 -.075 -.445** .061 -.025 10 EE i 2.136 4.371 .487** .781** -.004 -.073 .201** -.101* .334** .500** .019 11 FPSi .004 .063 -.020 -.008 -.009 -.011 .063 .141* -.031 12 NCEi -1.634 4.701 -.631** -.783** .027 .049 -.187** .059 -.163* -.259** .016 -.692** .008 13 NEEi -.921 1.760 -.169** -.276** -.087 .055 -.137** .128** -.027 .220** -.020 -.479** .033 .464** Note: * Significant at p < .05, ** Significant at p < .01, Sales in million units

Table 4.1 Descriptive Statistics and correlations PS1

Variable name Mean SD 1 2 3 4 5 6 7 8 9 10 11 12

1 SALESi .976 1.756 2 CE i 3.640 10.780 .548** 3 ACTION ADVENTUREi .083 .276 .180** .065 4 RACINGi .132 .338 .066 -.090 -.117* 5 MATUREi .076 .266 .059 .047 .623** -.112 6 EVERYONEi .561 .497 -.209** -.180** -.214** .221** -.324** 7 USERSCOREi 49.087 40.702 .323** .310** .174** -.122* .156** -.269** 8 METASCOREi 60.242 28.370 .373** .227** .077 .064 .046 -.167** .361** 9 EE i 1.069 1.521 .276** .217** .085 -.011 .081 -.189** .330** .421** 10 TEENi .318 .467 .205** .192** -.098 -.200** -.196** -.772** .226** .217** .155** 11 FPSi .014 .117 .018 .012 -.036 -.046 .078 -.134* .086 .062 .014 .110 12 NCEi -1.910 7.175 -.459** -.894** -.007 .068 .006 .134* -.222** -.134* -.138* -.166** -.010 13 NEEi -.782 1.192 -.028 .014 -.076 -.037 -.031 .039 .125* .144* .122* -.007 .003 .040

Table 4.2 Descriptive Statistics and correlations PS3

Variable name Mean SD 1 2 3 4 5 6 7 8 9 10

1 SALESi .539 1.265 2 CE i 10.260 40.988 .645** 3 ACTION ADVENTUREi .071 .256 .064* .065* 4 RACINGi .039 .194 .028 -.022 -.056** 5 MATUREi .247 .431 .200** .167** .194** -.111** 6 EVERYONEi .204 .403 -.061* -.077** -.128** .174** -.290** 7 USERSCOREi 35.980 34.862 .062* .084** .046* -.033 .078** .030 8 METASCOREi 36.718 36.324 .019 .051* .010 .011 .027 .087** .764** 9 TEENi .292 .455 -.039 -.037 -.012 -.088** -.367** -.325** -.043* -.068** 10 FPSi .049 .216 .064* .057** -.063* -.046* .199** -.115** .040* .019 -.025 11 NECi -8.575 41.873 -.703** -.765** -.067** .018 -.163** .073** -.071** -.041* .048* -.046* Note: * Significant at p < .05, ** Significant at p < .01, Sales in million units

Table 4.3 Descriptive Statistics and correlations PSP

Variable name Mean SD 1 2 3 4 5 6 7 8 9 10 11 12

1 SALESi .361 .654 2 CE i 1.97 6.776 .588** 3 ACTION ADVENTUREi .047 .211 .161** .120** 4 RACINGi .049 .215 .019 .000 -.050 5 MATUREi .059 .235 .139** .198** .265** -.057 6 EVERYONEi .415 .492 -.096* -.138** -.167** .138** -.210** 7 USERSCOREi 6.976 2.417 .123* .154** .062 .083 .096* .121** 8 METASCOREi 39.226 34.519 .260** .289** .067* .022 .144** -.140** .630** 9 EE i 1.882 3.207 .396** .479** .054 .043 .187** -.206** .180** .495** 10 TEENi .265 .441 .020 .113** -.015 -.063* -.150** -.505** .181** .261** .172** 11 FPSi .003 .055 .006 .070* -.012 -.012 .064* -.046 .036 .034 -.009 .050 12 NCEi -1.005 3.428 -.555** -.866** -.155** .014 -.237** .174** -.068 -.240** -.401** -.099** -.048 13 NEEi -1.010 1.781 -.077 -.128** -.060 -.053 -.208** .226** -.063 -.296** -.272** -.194** -.041 .209** Note: * Significant at p < .05, ** Significant at p < .01, Sales in million units

Table 4.4 Descriptive Statistics and correlations Vita

Variable name Mean SD 1 2 3 4 5 6 7 8 9 10 11

1 SALESi .135 .208 2 CE i 5.00 20.665 .289** 3 ACTION ADVENTUREi .002 .048 -.002 4 RACINGi .002 .048 -.012 -.002 5 MATUREi .080 .271 .268** .145** -.014 -.014 6 EVERYONEi .382 .486 -.060 -.051 -.038 -.038 -.231** 7 USERSCOREi 6.611 2.540 -.203 .037 -.003 .035 .101 .021 8 METASCOREi 70.922 11.393 .006 -.091 -.081 .022 .722** 9 TEENi .173 .379 -.010 .052 -.022 -.022 -.135** -.360** .220** .143 10 EE i 1.803 4.053 .244* .591** -.022 -.022 .130** -.067 .138* .068 .050 11 NCEi -2.304 12.731 -.355** -.811** .005 .009 -.146** .025 .041 .234** -.025 -.340** 12 NEEi -.726 2.154 -.165 -.448** -.006 .016 -.078 -.026 .022 .349** -.044 -.383** .636**

Table 4.5 Descriptive Statistics and correlations Wii

Variable name Mean SD 1 2 3 4 5 6 7 8 9 10

1 SALESi .467 2.747 2 CE i 2.57 16.054 .019 3 ACTION ADVENTUREi .045 .208 .000 .084** 4 RACINGi .048 .214 .008 -.012 -.049* 5 MATUREi .019 .135 -.002 .008 .185** -.031 6 EVERYONEi .584 .493 .022 .018 -.140** .142** -.163** 7 USERSCOREi 6.668 2.327 .064 .086* .051 -.046 .053 .101** 8 METASCOREi 62.524 15.224 .153** .087* -.029 -.117 .021 -.020 .690** 9 FPSi .014 .118 -.005 .065** -.026 -.027 .078** -.133** .039 .059 10 TEENi .150 .357 -.012 .016 .017 -.075** -.058** -.498** .096** .033 .177** 11 NCEi -1.616 7.952 -.031 -.779** -.030 -.007 -.019 .002 -.070* -.057 -.104** -.033

Table 4.6 Descriptive Statistics and correlations Xbox

Variable name Mean SD 1 2 3 4 5 6 7 8 9 10

1 SALESi .286 .523 2 CE i 3.47 8.273 .326** 3 ACTION ADVENTUREi .142 .349 -.028 .069* 4 RACINGi .122 .327 -.039 -.028 -.151** 5 MATUREi .201 .401 .097** .093** .220** -.178** 6 EVERYONEi .342 .474 -.006 -.094** -.215** .233** -.361** 7 USERSCOREi 53.760 36.462 .254** .106** .080* -.086* .178** -.075* 8 METASCOREi 66.039 20.008 .298** .107** .058 -.017 .070 .064 .510** 9 FPSI .071 .257 .153** .133** -.112** -.103** .268** -.199** .091** .055 10 TEENI .388 .488 -.057 .020 .032 -.064 -.399** -.573** .023 -.002 0.12 11 NCEi -2.317 10.907 -.428** -.790** -.023 .034 -.077* .063 -.049 -.057 -.133** .004

Table 4.7 Descriptive Statistics and correlations Xbox One

Variable name Mean SD 1 2 3 4 5 6 7 8

1 SALESi .344 .516 2 CE i 11.800 25.504 .535** 3 EVERYONEi .195 .401 .056 .165 4 USERSCOREi 37.781 29.240 .254 .327* .343* 5 METASCOREi 34.805 36.354 .223 .345* .491** .693** 6 TEENI .049 .218 -.174 -.093 -.111 .045 .030 7 EE i 3.195 5.188 .215 .659** .209 .364* .513** .013 8 NCEi -10.585 25.370 -.747** -.509** -.062 -.235 -.333* .046 -.737** 9 NEEi -2.439 5.000 .215 .020 .056 -.030 -.018 -.782** -.234 .138

Table 4.8 Descriptive Statistics and correlations Xbox360

Variable name Mean SD 1 2 3 4 5 6 7 8 9 10

1 SALESi .553 1.437 2 CE i 12.435 47.685 .681** 3 ACTION ADVENTUREi .068 .251 .027 .043* 4 RACINGi .049 .216 -.002 -.030 -.061** 5 MATUREi .262 .440 .213** .228** .183** -.131** 6 EVERYONEi .239 .427 -.077** -.107** -.140** .158** -.334** 7 USERSCOREi 45.001 33.980 .145** .179** .067** -.019 .157** -.026 8 METASCOREi 45.759 34.051 .213** .233** .036 -.005 .030 .105** .757** 9 TEENi .258 .438 -.069** -.060** -.009 -.074** -.352** -.331** .084** .068** 10 FPSi .051 .221 .047 .049* -.063** -.053** .207** -.130** .073** .039 -.020 11 NCEi -11.613 62.845 -.603** -.846** -.045* .026 -.161** .081** -.076** -.129** .057** -.036 Note: * Significant at p < .05, ** Significant at p < .01, Sales in million units

4.1. Test of the Hypotheses

The results of the regression analysis are presented in Table 5 to Table 5.4. The relationship between user score of videogames and global product sales of

videogames is significant and positive (Xbox: β= .131, p< .001), this is also the case for experts scores from Metacritic.com (Meta score) and product sales of videogames (PS1: β= .187, p< .01; Wii: β= .164, p< .01; Xbox: β= .219, p< .001; Xbox360: β= .110, p< .001). These results partially support H1a, H1b, because it is not the case with all the video game platforms. The relationship between positive consumer review (sentiment) of videogames and global product sales of videogames is

significant and positive (3DS: β= .425, p< .001; PS1: β= .504, p< .001; PS3: β= .249, p< .001; PSP: β= .291, p < .001; Xbox360: β= .575, p< .001), this is also the case for positive expert reviews (sentiment) of videogames and product sales of videogames (PSP: β= .148, p< .01). These results partially support H2a and H2b, because it is not the case with all the video game platforms. The relationship between negative

consumer reviews (disappointment) of videogames and global product sales of videogames is significant and negative (PS3: β= -.508, p< .001; PSP: β= -.233, p< .01; Vita: β= -.501, p< .05; Xbox: β= -.469, p< .001; Xbox One: β= -.978, p< .01; Xbox360: β= -.103, p< .01). These results partially support H3b, because it is not the case with all the video game platforms. The relationship between negative expert reviews (disappointment) of videogames and global product sales of videogames is not significant and negative (3DS: β= -.051, ns; PSP: β= .000, ns; Vita: β= -.076, ns; Xbox One: β= .346, ns). These results not support H3a.

Table 5 Regression Results 3DS and PS1

Variable Name 3DS PS1

Coefficient SE Beta Coefficient SE Beta

Constant -.341 .717 -.337 .453 CE i .028** .010 .425 .076*** .018 .504 EE i .004 .027 .020 .076 .063 .069 USERSCOREi -.149 .101 -.160 .003 .003 .076 METASCOREi .017 .012 .168 .012** .004 .187 NCEi -.038 .022 -.219 .011 .026 .047 NEEi -.027 .050 -.051 -.083 .077 -.056 RACINGi .645* .281 .123 ACTIONADVENTUREi .259 .530 .036 1.106** .401 .181 FPSi .264 .816 .017 TEENi -.101 .265 -.031 -.014 .465 -.004 MATUREi -.842* .405 -.164 -.741 .588 -.119 EVERYONEi .392 .217 .153 -.253 .436 -.072 N 508 289 R2 .667 .647 Adj R2 .445 .418 * Significant at p< .05 ** Significant at p< .01 *** Significant at p< .001

Table 5.1 Regression Results PS3 and PS4

Variable Name PS3 PS4

Coefficient SE Beta Coefficient SE Beta

Constant .084 .059 2.492 .056 CE i .006*** .001 .249 EE i USERSCOREi .001 .001 .019 -.012* .001 -.997 METASCOREi .000 .001 .012 -.016 .013 -.786 NCEi -.012*** .001 -.508 -.061 .030 -.895 NEEi RACINGi .308* .126 .047 ACTIONADVENTUREi .079 .095 .017 FPSi .126 .111 .023 TEENi .119 .067 .042 MATUREi .134 .075 .044 EVERYONEi .167* .069 .058 N 2450 10 R2 .726 .994 Adj R2 .526 .989 * Significant at p< .05 ** Significant at p< .01 *** Significant at p< .001

Table 5.2 Regression Results PSP and PS Vita

Variable Name PSP Vita

Coefficient SE Beta Coefficient SE Beta

Constant .051 .116 -.011 .151 CE i .022** .007 .291 -.001 .002 -.163 EE i .028** .009 .148 .003 .006 .082 USERSCOREi .010 .017 .028 -.004 .025 -.024 METASCOREi .002 .002 .066 .002 .003 .095 NCEi -.035** .013 -.233 -.010* .004 -.501 NEEi .000 .015 .000 -.007 .011 -.076 RACINGi .189 .144 .053 ACTIONADVENTUREi .257* .124 .088 FPSi .137 .612 .009 TEENi -.108 .083 -.070 -.007 .049 -.018 MATUREi -.164 .122 -.067 .201* .077 .307 EVERYONEi -.029 .093 -.017 .029 .049 .070 N 1005 427 R2 .612 .605 Adj R2 .375 .366 * Significant at p< .05 ** Significant at p< .01 *** Significant at p< .001

Table 5.3 Regression Results Wii and Xbox

Variable Name Wii Xbox

Coefficient SE Beta Coefficient SE Beta

Constant -1.932 1.034 -.152* .075 CE i .020 .044 .041 -.006 .003 -.091 EE i USERSCOREi -.098 .170 -.034 .002*** .001 .131 METASCOREi .051** .018 .164 .006*** .001 .219 NCEi -.021 .069 -.027 -.022*** .002 -.469 NEEi RACINGi .636 1.044 .027 -.007 .051 -.004 ACTIONADVENTUREi .134 .751 .008 -.064 .049 -.043 FPSi -.242 1.474 -.007 .154* .068 .075 TEENi -.251 .566 -.023 -.123 .068 -.114 MATUREi -.247 .999 -.012 -.073 .075 -.056 EVERYONEi .940 .510 .099 -.086 .070 -.078 N 1993 861 R2 .196 .533 Adj R2 .039 .284 * Significant at p< .05 ** Significant at p< .01 *** Significant at p< .001

Table 5.4 Regression Results Xbox One

Variable Name Xbox One Xbox360

Coefficient SE Beta Coefficient SE Beta

Constant .190 .107 -.030 .078 CE i -.002 .006 -.096 .014*** .001 .575 EE i -.006 .036 -.057 USERSCOREi .005 .004 .295 -.001 .001 -.013 METASCOREi -.006 .005 -.455 .006*** .001 .110 NCEi -.027** .007 -.978 -.002** .001 -.103 NEEi .031 .024 .346 RACINGi .120 .142 .017 ACTIONADVENTUREi .010 .110 .002 FPSi .106 .123 .017 TEENi .345 .491 .175 -.057 .082 -.018 MATUREi -.050 .091 -.015 EVERYONEi .097 .209 .079 .007 .082 .002 N 41 2478 R2 .830 .690 Adj R2 .689 .476 * Significant at p< .05 ** Significant at p< .01 *** Significant at p< .001

5. Discussion and Conclusion

The starting point of this research was that most of the studies that are done used only quantitative measures to measure the content of the consumer and expert reviews (Mudambi & Schuff, 2010). The current studies focus on the influence of star ratings, review volume and word of mouth (WOM) on sales, but the results of these studies are mixed (Chen, Wu & Yoon, 2004; Chevalier & Mayzlin, 2006; Dellarocas Xiaoquan & Awad, 2007; Duan, Gu &Whinston, 2008). This study adds to the discussion, by showing that sentiment (positive emotion) of consumers and experts in reviews have a positive impact on the product sales and disappointment (negative emotion) of consumers in reviews have a negative impact on the product sales.

The findings offer an in-depth perspective on the role of expert and consumer evaluations of videogames on global product sales of videogames. Not only WOM is important predictor of global product sales. More importantly, the research shows that sentiment of consumers and experts displayed in reviews is a good predictor of the global product sales and disappointment of consumers displayed in reviews is a good predictor of the global product sales.

This research showed that WOM and sentiment displayed in consumer and expert reviews of video games, has a positive effect on the global product sales of videogames. Also this research showed that disappointment displayed in consumer reviews of video games, has a negative effect on the global product sales of

videogames.

5.1 Unexpected results

There were some unexpected results in this research. First, unexpected result of this research was that there was no relationship between consumer evaluations, in the Vita and Wii dataset, and global product sales of videogames. From previous research of

and sustain retail revenue, so the expectation was that consumer evaluations, in the Vita and Wii dataset, had a positive effect on global product sales of videogames from stores. This was not the case, mainly because there was no correlation between

consumer evaluations, in the Wii dataset, and product sales. In the Vita dataset, the number of games was not so high and it is one of the newest game devices, so maybe not so much people own one.

Second, unexpected result of this research was that there was no relationship between positive expert reviews, in the Vita, 3DS, and PS1 dataset, and global product sales of videogames. From previous research of Miller (2005) it turned out that expert reviews could serve as an input for the formation of individual attitudes. As a result, if a videogame receives a positive expert review, these positive expert reviews may contribute to shaping the purchase intentions of consumers positively. This was not the case, mainly because, in the PS1 and Vita dataset, the number of games was not so high. In the Vita dataset there was also no correlation between expert evaluations and Sales, this was also the case in the 3DS dataset.

Third, unexpected result of this research was the negative significant

relationship between UserScore, in the PS4 dataset, and global product sales of PS4 videogames. From previous research of Miller (2005) it turned out that consumer evaluations of videogames could serve as an input for the formation of individual attitudes. As a result, if a videogame receives a positive consumer review, these positive expert reviews may contribute to shaping the purchase intentions of other consumers positively. This was not the case in the PS4 dataset, mainly because the dataset only consisted of 10 videogames and because it is one of the newest game devices there was no global product sales information about all the PS4 games.

Fourth, unexpected result of this research was that there was no relationship between UserScore, in the 3DS, PS1, PSP, PS3, Vita, Wii, XboxOne, and Xbox360 dataset, and global product sales of videogames. From previous research of Miller (2005) it turned out that consumer evaluations of videogames could serve as an input for the formation of individual attitudes. As a result, if a videogame receives a positive consumer evaluation, these positive consumer evaluations may contribute to shaping the purchase intentions of other consumers positively. This was not the case in the 3DS, PS1, PSP, PS3, Vita, Wii, Xbox360 and XboxOne dataset, mainly

because in the 3DS, Vita, XboxOne, and Wii dataset there was no correlation between UserScore and Sales. In the other cases (Xbox360, PSP, PS3 and PS1) not all the games had a UserScore, so there were a lot of missing values. This can influence the results.

Fifth, unexpected result of this research was that there was no relationship between MetaScore, in the 3DS, PS3, PSP, Vita, and XboxOne dataset, and global product sales of videogames. From previous research of Miller (2005) it turned out that expert evaluations could serve as an input for the formation of individual attitudes. As a result, if a videogame receives a positive expert evaluation, these positive expert evaluations may contribute to shaping the purchase intentions of consumers positively. This was not the case in the 3DS, PS3, PSP, Vita and XboxOne dataset, mainly because in the 3DS, PS3, XboxOne, and Vita dataset there was no correlation between MetaScore and Sales. Also, not all the games had a MetaScore, so there were a lot of missing values. This can probably influence the results.

Sixth, unexpected result of this research was that there was no relationship between disappointment in the expert review, in the 3DS, PS1, PSP, Vita, and Xbox One dataset, and global product sales of videogames. From previous research of

Miller (2005) it turned out that expert reviews could serve as an input for the formation of individual attitudes. As a result, if a videogame receives a negative expert review, these negative expert reviews may contribute to shaping the purchase intentions of consumers negatively. This was not the case in the 3DS, PS1, PSP, Vita, and Xbox One dataset, mainly because in the PS1, PSP, Vita, and Xbox One dataset there was no correlation between disappointment and global product sales of

videogames. In the case of the 3DS dataset there weren’t enough negative words used in the expert reviews (only 469 words) in 6000 reviews.

5.2 Interpretation of Results

Duan, Gu and Whinston (2008) found that WOM is important to generate and sustain retail revenue. Davis and Khazanchi (2008) came up with a workable definition of online word of mouth. According to Davis and Khazanchi (2008) online word of mouth is when people have the ability to communicate and share experiences with each other online. This research is consistent with the research of Duan, Gu and Whinston (2008) with the finding that online WOM in consumer- and expert product evaluations has a positive effect on the product sales. This means that when

consumers or experts give a positive evaluation/score about a product, this evaluation/score has a positive influence on the sales of that particular product.

Second, there is a lot of work on sentiment analyses on reviews/blogs. Earlier work showed that when there are more references to a book, in online reviews, this would be followed by increasing sales of that particular book (Gruhl et al., 2005). Sentiment (positive emotion) might have an impact on consumers’ attitudes (Cohen et al., 2008). Sentiment is a thought, idea or opinion based on a feeling about a situation. Or a way someone thinks about something (Cambridge Dictionaries Online, 2013) and Darke, Chattopadhyay & Ashworth (2002) describe affect as general positive or

negative feelings. This research is consistent with the research of Gruhl et al., 2005) with the finding that sentiment displayed in consumer- and expert product reviews has a positive effect on the product sales. This means that when consumers or experts write product reviews with a lot of positive words, this review will have a positive influence on the sales of that particular product.

Third, from earlier research of Gemser, van Oostrum and Leenders (2006) it was argued that product reviews play two important roles. First, there is the influence effect of reviews. This type of effect actively influences consumers in their selection process. Second. There is the prediction effect of reviews. This type of effect shows if the product will be a success or failure (Gemser, van Oostrum & Leenders, 2006), so product reviews are an important information source for consumers.

Miller (2005) argued that consumer reviews could serve as an input for the formation of individual attitudes. For example, an individual attitude about a

videogame could be affected by the opinion of other consumers. The game industry is an market where consumers can access the product reviews of other consumers, because there are now websites that gather all these consumer reviews. As a result, one can expect that, if a videogame receives a negative consumer review, these negative reviews may contribute to shaping the purchase intentions of other

consumers negatively. This will have a negative influence on the product sales of that particular video game. This research is consistent with the argumentation of Miller (2005) that disappointment (negative emotion) displayed in consumer reviews has a negative effect on the product sales. This means that when consumers write product reviews with a lot of negative words, these reviews will have a negative influence on the sales of that particular product.

5.3 Implications, Limitations and Future Research

The results have several implications for marketing practice and marketing research. First, this study is the first attempt to apply sentiment/opinion mining to the

discussion of videogames. Sentiment/ opinion mining offers a new way for studying the impact of consumer and experts evaluations of videogames on the global product sales of videogames.

Second, the results of this study contributes to the ongoing discussion about the impact of consumer and expert evaluations on the global product sales, which has obtained mixed findings so far (Gruhl, Guha, Kumar, Novak & Tomkins, 2005; Mishne and Glance, 2006).

Third, the results have important managerial implications, and are of

commercial importance. It is important for managers to understand how their product sales are affected by the opinions and emotions displayed in consumer- and expert reviews. The significant influence of opinions and emotion displayed both in

consumer and expert reviews and the significant influence of WOM further illustrates the necessity for intensive management of these external elements, therefore it is important for companies that their products are positively regarded by consumers and experts. Overall, the insight generated within this research can serve as a basis for future decision-making regarding the possible release of videogames on different platforms, as this research provides an overview of consumer- and expert

sentiments/opinions, emotions and scores of videogames from different videogame platforms.

However, this research has several limitations. First, the empirical

relationships found in this study come from the video game industry. More research is needed to validate the findings of this research and analyze other types of products,

especially non-creative products, such as hotels, restaurants, and automobiles where past consumer and expert reviews may also influence sales of products and services. Second, the high anonymity of online consumer evaluations means that with the current approach, the researcher cannot determine the actual purchasing behavior of individual consumers who post online reviews. For future research it is important to use data that reflect the purchase behavior of individual consumers. Third, the data sources, especially Metacritic.com also limits the repeatability of this study, as both MetaScores and UserScores are constantly subject due change to the removing and adding of existing reviews. Finally, the data source, especially vgchartz.com also limits the repeatability of this study, because the global product sales is constantly subject due change to the adding of new sales data.

5.4 Implications to theory

Miller (2005) argued that expert evaluations/scores could serve as an input for the formation of individual attitudes. A similar argument can be made with regard to positive consumer evaluations/ scores as these evaluations too can contribute to shaping the purchase intentions of other consumers positively (Miller, 2005).

This research confirms the research of Miller (2005). It is noted from this current study that positive expert- and consumer evaluations/scores contribute to shaping the purchase intentions of consumers. This finding is consistent with that presented by Zhang and Dellarocas (2006), Chevalier and Mayzlin (2006), and Litman (1983) who also showed that expert-and consumer evaluations/scores are important for explaining revenue, sales and purchasing behavior, but contradicts that of Dhar and Chang (2009). However, Dhar and Chang (2009) did research in the music industry and not in the videogame industry.

Miller (2005) also argued that expert reviews could serve as an input for the formation of individual attitudes. For example, an individual attitude about a videogame can be affected by the opinion of game experts. The game industry is an example of a market where consumers can access the expert reviews of products. Furthermore, there are now websites that gather all these expert reviews as well as consumer reviews. As a result, one can expect that, if a videogame receives a negative expert review, these negative reviews may contribute to shaping the purchase

intentions of consumers negatively. This will have a negative influence on the product sales of that particular video game. Miller (2005) argued that expert reviews could serve as an input for the formation of individual attitudes. A similar argument can be made with regard to consumer reviews of videogames as these reviews too can contribute to shaping the purchase intentions of other consumers negatively This research confirms the research of Miller (2005). It is noted from this current study that negative consumer reviews contribute to shaping the purchase intentions of

consumers. When consumers use disappointment (negative emotion) in their reviews this has a negative impact on the product sales of videogames. This study was as far as the researcher knows the first attempt to apply text mining in the videogame industry. This research expands the previous studies by using text mining to measure disappointment (negative emotion) in expert- and consumer reviews.

Earlier work on the predictive power of blogs and online reviews has used the volume of blogs and online reviews or link structures to predict the trend of product sales (Gruhl, Guha, Kumar, Novak & Tomkins, 2005; Gruhl, Guha, Liben-Nowell & Tomkins, 2004). They were not able to consider the effect of the sentiments present in the blogs and online reviews. This research expands the previous studies by using opinion mining to measure sentiment in expert- and consumer reviews. It is noted

from this current study that positive expert- and consumer reviews contribute by shaping the purchase intentions of consumers. When experts or consumers use

sentiment (positive emotion) in their reviews this has a positive impact on the product sales of videogames. This study was as far as the researcher knows the first attempt to apply opinion mining in the videogame industry. This research expands the previous studies by using opinion mining to measure sentiment (positive emotion) in expert-and consumer reviews.

5.5 Conclusion

The research question of this research was: what is the effect of online Word of Mouth, sentiment and disappointment in online textual reviews of videogames on product sales from stores? WOM and sentiment in online textual reviews from experts and consumers have a positive impact on the product sales from stores.

Disappointment in online textual reviews of videogames from consumers has a negative impact on the product sales from stores.

6. Bibliography

Anderson, E.W. (1998). Customer Satisfaction and Word of Mouth. Journal of Service Research, 5-17.

Andrade, E.B. (2005). Behavioral Consequences of Affect: Combining Evaluative and Regulatory Mechanisms. Journal of Consumer Research, 32 (3), 355-62. Banerjee, A. (1992). A Simple Model of Herd Behavior. Quarterly Journal of

Economics, 797-817.

Banerjee, A. (1993). The Economics of Rumors. The Review of Economic Studies, 309-327.

Basuroy, S., Chatterjee, S., Ravid, S.A. (2003). How Critical Are Critical Reviews? The Box Office Effects of Film Critics, Star Power, and Budgets. Journal of Marketing, 103 – 117.

Bettman, J.R., Johnson, E.J., Payne, J.W. (1991). Consumer decision making. Handbook of consumer behavior, 50-84.

Boatwright, P., Basuroy, S., Kamakura, W. (2007). Reviewing the reviewers: The Impact of individual film critics on box office performance, Quantitative Marketing and Economics, 401-425

Cambridge Dictionaries Online (2014). Sentiment.

(http://dictionary.cambridge.org/dictionary/british/sentiment_1), 15 January.

ChannelAdvisor (2010), Through the Eyes of the Consumer: Consumer Shopping Habits Survey. Morrisville, NC: Channel Advisor Corporation.

Chen, P., Wu, S., Yoon, J. (2004), The Impact of Online Recommendations and Consumer Feedback on Sales. Proceedings of the International Conference on Information Systems, ICIS Washington, DC: Association for Information Systems. Chen, Y., Fay, S., Wang, Q. (2011). The Role of Marketing in Social Media: How

Chen, Y., Xie, J. (2004). Online Consumer Review: A New Element of Marketing Communications Mix. Working Paper, Department of Marketing, University of Florida.

Chevalier, J.A., Mayzlin, D. (2006). The Effect of

Word of Mouth on Sales: Online Book Reviews. Journal of Marketing Research, 43 (August), 345-54.

Cohen, J.B., Pham, M.T., Andrade, E.B., Haugtvedt, C.P., Herr, P., Kardes, F. (2008). The Nature and Role of Affect in Consumer Behavior. Handbook of Consumer Psychology, New York: Taylor & Francis Group, 297-348.

Coleman, J.S. (1966). Medical Innovation: ADiffusion Study. Bobbs-Merrill.

Darke, P., Chattopadhyay, A., Ashworth, L. (2002). Going With Your “Gut Feeling”: The Importance and Functional Significance of Affective Cues in Consumer Judgement and Choice, Insead Working papers, 1-68.

Davis, A., Khazanchi, D. (2008). An Empirical Study of Online Word of Mouth as Predictor for Multi- product Category e-Commerce Sales. Electronic Markets- Taylor & Francis.

Dellarocas, C. (2003). The Digitization of Word-of-Mouth:

Promise and Challenges of Online Feedback Mechanisms. Management Science, 49 (10), 1407–1427

Dellarocas, C., Xiaoquan, Z., Awad, N.F. (2007). Exploring the Value of Online Product Reviews in Forecasting Sales: The Case of Motion Pictures. Journal of Interactive Marketing, 21 (4), 23-45.

Dhar, V., Chang, E.A. (2009). Does Chatter Matter? The Impact of User-Generated Content on Music Sales. Journal of Interactive Marketing, 300-307