The role of affective content in online reviews : how evaluations of past editions influence the performance of video game sequels

(1)

Faculty of Economics and Business

The role of affective content in online reviews: how

evaluations of past editions influence the performance of

video game sequels

Master Thesis 30-06-2014

Author: David Bastiaan Jakob Schimmel (5881285) Supervisor: dr. Frederik Situmeang

Second Supervisor: dr. Umut Konus Msc Business Studies, - Marketing Track

(2)

2

Abstract

This study examines the role of affective content in expert reviews and consumer reviews of earlier editions on the performance of sequels. A set of hypotheses was developed based on earlier research of Ludwig et al. (2013) and Situmeang, Leenders and Wijnberg (2013). The empirical setting is the video game industry and the dataset was obtained from the websites metacritic.com and vgchartz.com. The hypotheses were tested by performing linear regression analysis. Affective content in consumer- and expert reviews has a positive effect on the global product sales of video games. The effect of affective content in consumer- and expert reviews of earlier editions on the global product sales of sequels has however not been found. But the sales of earlier editions have a positive effect on the performance of sequels. The results have several implications for marketing purposes. For instance, this study is the first to examine the influence of affective content in reviews of earlier editions on the performance of sequels. It provides new insights into the impact of product evaluations. The results have some important managerial implications too. It is important for managers to understand how sales are affected by opinions and emotions displayed in reviews. This study shows that affective content in reviews play a significant role in predicting performance. With this knowledge, companies can act on that by strategically manage reviews and their communication with consumers. But this study has also shown that the affective content in expert reviews of earlier editions does not predict the performance of the sequels. The sales of the earlier editions are good enough to predict the success of the sequels.

Keywords: Sequels, Consumer reviews, Expert reviews, Affective Content, Global Product Sales

(4)

4

1. Introduction

Experience goods are products or services where the quality and utility for a consumer can only be determined upon consumption (Nelson, 1970). Because the quality is difficult to ascertain prior to consumption, there are several sources where people can attain information about the qualities of experience goods. Examples of sources are for instance advertising, information from peers and branding (Reinstein and Snyder, 2005).

One of the most influential resources of information transmission since the beginning of society, especially for experience goods, is word-of-mouth (Duan, Gu and Whinston, 2008). In this digital era people are using online word-of-mouth, expert reviews, user reviews and evaluations of product information very often. Online customer review systems function as one of the most powerful channels to generate online word-of-mouth (Dellacoras, 2003). There are several websites where people can share their thoughts about the products with other internet users and influence decisions. For instance, the most popular source that people use to attain knowledge about experience goods is the Internet Movie Database (IMDb). This database has more than 4 million unique visitors per day (data from websitelooker.com).

The influence of online reviews has been examined several times. Earlier research did not give a consentient outcome of the main online WOM driver of product sales. It has been shown that various aspects of online WOM influence the performance (Duan, Gu and Whinston, 2008). Also several studies have been conducted to examine the impact of expert evaluations on sales (Gemers, van Oostrum and Leenders, 2006; Basuroy, Chatterjee and Ravid, 2003). All these earlier studies nearly exclusively focused on ‘’quantitative surrogates’’ of review contents as review volume, grading or star rating (Ludwig et al, 2013)

With these mixed results it is interesting to explore whether the qualitative content of user reviews has an influence on performance. Research from Korfiatis et al. (2011) found that the qualitative content of reviews matters. It was found that word length and readability

(5)

5 scores were indicators for a review to be considered highly helpful by a consumer. Therefore this study will examine the qualitative content of reviews, with the focus on the supposed influence of affective content in reviews. Earlier research of Isen (2001) states that it is widely accepted that affect regularly plays a role in cognitive processes as decision-making and evaluating. Furthermore Cohen et al. (2008) states that affective cues provided in text can influence respondent attitudes. Therefore this study is going to examine whether the affective content of reviews influences the product performance.

The aim of this study is to determine whether the affective content in evaluations of past editions has an influence on the performance of their sequel. The main research question of this research is: does the positive affective content in reviews of past editions affect the performance of the sequel? In earlier research sequels are examined as extensions of a brand (Henning-Thurau et al., 2009; Sood and Dreze, 2006; Bassuroy and Chatterjee, 2008). Sequel series are most of the best-selling products in experience goods industry. On average, sequels produce even higher revenues than non-sequels (Hennig-Thurau et al., 2009). A reason for the good performance of sequels is given in earlier research. Sunde and Brodie (1993) stated that higher perceptions toward the original brand are associated with more favourable attitudes towards the extension. Furthermore Situmeang (2013) states that the original brand has established a positive image which makes sequels more attractive, and also create awareness, excitement and anticipation for the sequel.

This study will focus on the video game industry. As mentioned before, this study is about experience goods and will therefore focus on a creative industry. The video game industry has experienced an enormous growth over the past decades. The past 25 year, the industry has grown between 9% and 15% per year (Zackariasson and Wilson, 2010).

Nowadays it even is the largest entertainment industry in the world (VG sales, 2013). The enormous growth of the industry, the huge amount of sequels in this industry, and

(6)

6 availability of large online databases are making it a very suitable industry to perform this research. This research is organized in the following manner: in the next section, relevant literature is reviewed which provides detailed empirical background for this study. The following section will describe the methodology and dataset which will be followed with the section in which the results of the study will be reported. Subsequently a section will be included in which the results of the study will be discussed. The last section will contain a conclusion with limitations of the research and possible suggestions for future research.

(7)

7

2. Literature review

2.1 The impact of evaluations on performance

Product evaluations are an important source for information about products in the creative industries. These products are experience goods and the quality of the product is difficult to estimate before consumption. There has been done much research to the impact of evaluations on the performance of products. According to literature of Gemser, van Oostrum and Leenders (2006) reviews can play two basic roles. There is the influence effect of reviews, which actively influence consumers in their selection process. And there is the

prediction effect of reviews, which forecasts whether the product will be a success or not.

Both the influence and the prediction effect on the demand have been established in earlier research in the movie industry (Gemser, van Oostrum and Leenders, 2006; Bassuroy, Chatterjee and Ravid, 2003). And these effects of reviews are applicable to other industries in which it is difficult to ascertain the quality prior to consumption (e.g., book publishing, theatre, financial markets, recorded music and the video game industry).

With the digitalization of this era, reviews on the Internet have become very important to consumers. Reviews are available all throughout the web and guide decisions of consumers all over the world. Research has even shown that users tend to trust peer reviews more than advertising and other content created by marketing departments and advertising agencies (Kardon, 2007). Throughout the internet there are user reviews and expert reviews. Earlier research has looked into the influence of the online reviews. For instance Duan, Gu and Whinston (2008) have focused on the impact of online Word-of-mouth (WOM), user reviews, in the movie industry. Word-of-mouth (WOM) has been recognized as one of the most influential resources of information transmission since the beginning of human society (Godes and Mayzlin 2004; Maxham and Netemeyer 2002; Reynolds and Beatty, 1999).

(8)

8 Traditional Word-of-Mouth (WOM) only took place within small social circles and the lifetime of this information was short. This has however changed with the rise of the Internet. Laroche et al. (2005) mention that the advances of information technology and the emergence of online social network sites have profoundly changed the way information is transmitted and has transcended the traditional limitations of WOM.

Online Word-of-Mouth (WOM) takes many forms, as blogs, discussion boards or chat-rooms. This study however focuses on user reviews. Because statistics suggest that user reviews are more prevalent than other forms of WOM communication (Duan, Gu and Whinston, 2008). One important difference between online user reviews and other types of WOM is that user reviews usually reflect user experience and consumer satisfaction, which are mainly viewed as a source of product information (Chen and Xie, 2004; Li and Hitt, 2008). Prior studies had shown that WOM influences retail sales. Duan, Gu and Whinston (2008) found that higher ratings do not lead to higher sales, but the number of posts is significantly associated with sales. Other studies found that WOM dispersion (Godes and Mayzlin, 2004) and valence (Chevalier and Mayzlin, 2006) have effect on product sales, while research from Chen, Wu and Yoon (2004) and Ludwig et al. (2013) found that the WOM volume is a key driver of product sales. The association of the number of posts and WOM volume with product sales can be contributed to the informative component of reviews. A higher WOM volume enhances the brand awareness of a product and awareness is a key variable in consumer decision-making (Vermeulen and Seegers, 2009).

The role of expert reviews has also been examined in earlier research. Expert reviews are posted by paid evaluators, who provide in-depth and unbiased evaluations of a product. These reviewers are expected and perceived to have high standard of integrity (Amblee and Bui, 2007). Eliashberg and Shugan (1997) studied the influence of expert reviews on box office success and found that experts can have two possible effects. Experts have the ability

(9)

9 to influence consumers’ decisions and the ability to predict consumers’ decisions. Their study found a significant correlation between critical reviews and eventual box office performance. Miller (2005) also argued that evaluations of experts can serve as input for the formation of individual attitudes. This implies that the individual attitude about a video game can be affected by the opinion of reviews of game experts. Also Basuroy, Boatwright and Kamakura (2003) have found that critics’ reviews are significantly correlated with box office revenues in the movie industry. And Litman (1983) found that the rating of movies by critics play a significant role in explaining box office revenues.

So according to this literature the relationship between online WOM and product sales has been established. Therefore one can expect that if a video game receives positive reviews this will have a positive effect on the product sales of that particular video game.

2.2 Affective content in product reviews

Research has shown that affect can have a significant impact on judgement, choice (Loewenstein and Lerner, 2003; Darke, Chattopadhyay and Ashworth, 2002) decision-making (Shiv and Fedorikhin, 1999) attitude (Morris et al. 2002) and evaluation (Lench, Flores and Bench, 2011). The vast majority in earlier research has found that affect can influence judgment under low elaboration conditions, where the motivation to make accurate judgments is low or the consumer lacks the capacity to engage in information processing. Research from Darke, Chattopdhyay and Ashworth (2002) however provided evidence that affect can influence judgment under high elaboration conditions, where judgment is determined through effortful and purposeful thought, as well.

Emotions can exert a direct impact and an indirect impact on behaviour via their impact on judgements of expected consequences and emotional reactions (Loewenstein and Lerner, 2003). Research of Morris et al. (2002) has shown that affect is a powerful predictor

(10)

10 for intention. They have even shown that affect, compared to cognition, accounts for almost twice the variance towards intention. So measuring emotions can help to determine the intention of consumers.

Cohen, Pham and Andrade (2008) mention to the term ‘affect’ to describe an internal feeling of state. Whereas Russel (1981) states that affect refers to genuine subjective feelings and moods. Earlier research has made a distinction between affect and affective cues, which are important for this study. The term affective cue refers to more discrete feelings of positivity or negativity that become attached to, or paired with specific products (Leventhal, 1980; Zajonc, 1980). Research of Lau-Gesk and Meyers-Levy (2009) has shown that more positive affective cues lead to more positive evaluations and judgements. Furthermore Lau-Gesk and Meyers-Levy (2009) stated that affective content has a strong impact on behaviour. This can indicate that the affective content of product evaluations has a strong impact on the behaviour of the consumer.

The research to affective content in reviews is very relevant because earlier research has established many reasons to suggest that affective content plays a significant role in people judgement and behaviour. For instance, Ludwig et al. (2013) stated that the affective content of words reveal the intent of the entire text. Where Strahan, Spencer and Zanna (2002) found that even without producing a measurable effect on people’s affective experience, affective stimuli, like words, can influence people’s positive or negative assessment. And applicable to product evaluations, Zajonc (1980) states that the affective cues in evaluative judgements are even more accessible than the factual or descriptive information. According to Lau-Gesk and Meyers-Levy (2009) and Lench, Flores and Bench (2011) just reading a text with affective content is enough to influence thoughts and behaviours of people. Furthermore Ortony, Clore and Foss (1987) stated that text based affective content words provide rapidly accessible and diagnostic signal about targets. This

(11)

11 earlier literature proves the relevance of research to the relationship between affective content in reviews with product performance.

And there has been done research to relationship between affective content and product evaluations. According to this earlier literature a linear relationship between the two variables was expected. But research from Andrade (2005), Roehm and Roehm (2005) and Ludwig et al. (2013) found evidence of a non-linear relationship between affect and the thoughts and behaviour of consumers. Andrade (2005) found that people are more willing to try a product when experiencing positive affect. However when people were already experiencing positive affect; the impact of the affective cue was less present, resulting in an inverted u-shape relation. This non-linear relationship is because of the effect of extreme (positive and negative) affect.

For instance, Ludwig et al. (2013) found a strong positive effect of changes in affective content on subsequent conversion rates changes. But when the extreme intensity of affective content changed, it exhibited in a quadratic relationship. It was also found that negative affective cues were more powerful than positive affective cues for driving judgement and behaviour. So it was found that the relationship between affective content and performance is non-linear due to the effect of extreme (positive and negative) content in reviews.

So there has been done research to role of affect in judgements and decision-making. Earlier research has shown that affect has a significant impact on evaluations and judgements. Also the relationship between affective content and product evaluations has been examined. Earlier research has found a non-linear effect of affective content. Affective content led to more willingness of trying a product but extreme affect did not lead to more willingness of trying a product. Despite the results of earlier research, this research expects that positive affect will lead to better performance.

(12)

12 Hypothesis 1a: Positive affective content in expert reviews has a positive effect on the

performance of video games.

Hypothesis 1b: Positive affective content in consumer reviews has a positive effect on the

performance of video games.

2.3 The relationship between evaluations of past editions and sequels

In recent literature, sequels have been examined as brand extensions of experiential goods (Sood and Dreze, 2006). Earlier research from Bassuroy and Chatterjee (2008) to the success of sequels in the movie industry has shown that sequels perform worse than their parent films. But Hennig-Thurau et al. (2009) and Bassuroy and Chatterjee (2008) both found that sequels perform better than contemporaneous non-sequels. The success of sequels can be explained by the transfer of positive parent brand affect to their sequel. The acceptance and success of an extension will be higher if the perceived quality and the hold of positive beliefs by the consumers of the parent brand are high (Aaker and Keller, 1990; Sunde and Brodie, 1993). Research of Hennig-Thurau et al. (2009) showed that parent brand awareness and

(13)

13 brand image are the strongest predictors of overall success of sequels. The affect from the parent brand is transferred to the extension, when the perceived familiarity is high (Sood and Dreze, 2006) and also acts as a risk reducer for consumers (Milewicz and Herbig, 1994). Sequels have a high familiarity because they are somewhat similar to their past edition and carry-over the awareness and perceived quality of the past edition. The familiarity and increased knowledge with a series of sequels or editions of a product can be considered to be present at consumers (Rotschild and Gaidis, 1981). Although familiarity seems to be a key driver for the success of sequels, research from Sood and Dreze (1984) has showed that consumers tend to prefer sequels that differ in a substantive way from their earlier edition.

Earlier research of Hennig-Thurau et al. (2009) and Situmeang (2013) has shown that positive evaluations of past editions are associated with sequel performance success. Situmeang (2013) found that both expert and consumer evaluations in past edition of series have a significant positive relationship with sales performance of the sequels. Therefore it is to assume that evaluations of past editions have a significant positive impact on the performance of sequels.

Hypothesis 2a: Positive affective content in expert reviews of past editions has a positive

effect on the performance of sequels.

Hypothesis 2b: Positive affective content in consumer reviews of past editions has a positive

(14)

14 2.4 Past sales performance and the performance of sequels

Another issue that is going to be examined is the relationship between past sales and the performance of sequels. Past sales can serve as a reflection, to some extent, on how well the series have been appreciated over time by consumers in general across past editions in the series (Situmeang, Leenders and Wijnberg, 2014). Sales reflect the performance of a product, and indicate whether the product was a success or not.

As for most things in life, we seek to repeat in the future what we have liked or enjoyed in the past, and avoid or dread further experiences with what we have disliked or found aversive (Fredrickson, 2000). Therefore it is plausible to assume that sequels of highly sold video games shall perform well too. As Sunde and Brodie (1992) concluded that consumer acceptance of brand extensions will tend to be higher if the perceived quality of the brand is high.

As mentioned before, evaluations of products are narrowly linked to the performance of products. Research from Ho-Dac, Carson and Moore (2013) found that higher sales lead to a larger number of positive online customer reviews. Successful products generate more WOM than unsuccessful products, which could affect subsequent performance. As it was found by research of Chen, Wu and Yoon (2004) that popular products tend to have more

(15)

15 reviews and an increase in information could lead to more trust. So according to the literature, popular products tend to have more reviews and thus more product information. And this leads to higher trustworthiness of reviews which ultimately leads to more influential reviews of popular products.

Experts within the movie industry appear to agree that WOM is a critical factor of its ultimate financial success (Duan, Gu and Whinston, 2008). But sales can have an influence on the likelihood of WOM too. One unique aspect of the WOM effect is that it distinguishes from more traditional marketing effects by the positive feedback mechanism. That means that WOM leads to more sales, but also indicates that WOM is not only a driver of sales but also a result of sales (Duan, Gu and Whinston, 2008). Research from Situmeang, Leenders and Wijnberg (2014) did found evidence for the positive relationship between the sales performance of past editions in the series for both expert and consumer evaluations of the new edition of the product. The results of these earlier studies have led to the expectation that better sales of past editions will lead to better performance of their sequels.

Hypothesis 3: Better sales of past editions lead to better performance of sequels.

Summarizing the earlier literature, both expert and user reviews influence the performance of experience goods. Also there has been done research to the affective content of reviews and their influence on the performance of the products. There was found that there

(16)

16 is a non-linear relationship between those variables due to the effect of extreme positive and negative reviews. But there has not been done any research into the relationship of affective content in reviews of past editions and the performance of the sequels.

According to the literature, there is a strong relationship between those two because of the carry-over of awareness and perceived quality of past editions. Furthermore earlier literature suggests that there is a relation between past sales and the performance of new editions. The main gap that has been found in the earlier literature to evaluations and brand extensions is that the combination of affective content in reviews of past editions and their influence on the performance of sequels has not been researched before.

(17)

17

3. Methodology

This research will use a dataset that consists of data from metacritic.com. The dataset consists of evaluations of video games both for experts and consumer who are registered on the website. Separate text files for all reviews had to be created to analyze the linguistic properties per review. To analyse the linguistic properties of the reviews, text analysis with the software program Linguistic Inquiry and Word Count 2007 (LIWC) program (Pennebaker et al.2007) was performed. The LIWC program can analyze hundreds of standard ASCII text files or Microsoft Word documents in seconds (LIWC, 2007). The validity of LIWC has been confirmed in more than 100 studies that applied this methodology to various texts (Ludwig et al. 2013) and research suggests that LIWC accurately identifies emotion in language use (Tausczik and Pennebaker, 2010). Using word counts for a given text, LIWC calculates the proportion of words that match predefined dictionaries.

In this research the Internal Pennebakker LIWC 2001 Dictionary of LIWC was used. This research is about the positive- and negative emotions in reviews. There are 406 words of positive emotion and 499 words of negative emotion in the dictionary. Words of positive emotion (nice, sweet) and negative emotion (hurt, hate) in the review texts determined the measure of affective content in the reviews. For example, ‘’love’’ would appear in the positive emotion dictionary and be counted as 1 in the total amount of positive affective content words in the review text. Whereas ‘’ugly’’’ would appear in the negative emotion dictionary and be counted as 1 in the total amount of negative affective content words in the review text. LIWC calculates the total of appearances of negative and positive words in a review, divided by the total number of words in the review, to determine the percentage of the text that falls into a particular linguistic category. The output of LIWC resulted in two main files per video game platform. One that consisted of the output for consumer reviews and one that consisted of the output of expert reviews. Subsequently, one main database per

(18)

18 video game platform was made in Microsoft Access. The sales data, product data and review data and all control variables were combined to create one main file with all data per video game platform.

The equation used in this research is based on the equation that was used in earlier research of Ludwig et al. (2013):

ACit represents the overall intensity of affective content in reviews for game I. This research focuses on the body of the review text. The subscript (B) denotes the body of the review text. PAiB and NAiB stand for the sum of positive affective- and negative affective content words across all reviews. So the score of the equation will be the sum of positive affective content words in the review minus the sum of the percentage of negative affective content words in the reviews. The website vgchartz.com was used to obtain information about the product sales of the video games. This website records sale data for all the video games of all the platforms (for example Nintendo Wii or Playstation 3).

(19)

19 In Table 1 there is information about the dataset.

Table.1 Number of Games, Number of Expert- and Consumer Reviews, and number of Sales Platform Nr. Of Games Nr. of Expert Reviews Nr. of Consumer Reviews Global Sales

Nintendo 3DS 508 5867 2840 135.790.000 Nintendo Wii 1990 16903 10175 877.070.000 Playstation 1 289 2855 3145 730.660.000 Playstation 3 2450 36912 44678 795.820.000 Playstation 4 10 159 370 20.830.000 Playstation Vita 426 3713 2967 23.860.000 PSP 1005 13324 4035 286.360.000 Xbox 861 21976 8273 258.410.000 Xbox 360 2478 57338 59158 861.450.000 Xbox One 41 668 1150 14.910.000 PC 10092 55481 74779 261.300.000

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

Table 1 shows the dataset that consists of reviews of games of various platforms. With the PC having the largest amount of games (10.092) and the biggest amount of expert- (55.481) and consumer reviews (74.779). Nintendo Wii has the largest amount of global video game sales (877.070.000 games).

Model specification

The hypotheses in this research are tested with the following statistical model. The model takes the affective content in expert- and consumer reviews into account, as well as the affective content in expert- and consumer reviews of sequels and the global product sales of earlier editions. A detailed overview of the variables is presented in Table II.

SALESi = β0 + β1.C_EMOi+ β2.E_EMOi + β3.C_EMOt-1i + β4.E_EMOt-1i +

β5.SALESt-1i + β6.FPSi + β7.RACINGi + β8.ACTIONADVENTURE+ β9.MATUREi +

(20)

20 Table.2 Overview of the variables

Variable Name Description

SALESᵢ Worldwide unit sales (million copies) of video games.

C_EMOᵢ Positive affective content in consumer reviews of video game i. E_EMOᵢ Positive affective content in expert reviews of video game i. C_EMOt-1ᵢ Positive affective content in consumer reviews of previous

edition of video game i.

E_EMOt-1ᵢ Positive Affective Content in expert reviews of previous edition of video game i.

SALESt-1ᵢ Worldwide unit sales (million copies) of previous edition of video game i.

FPSᵢ Genre classification of video game is FPS (First Person Shooter). Dummy variable.

RACINGᵢ Genre classification of video game is RACING. Dummy variable.

ACTIONADVENTUREᵢ Genre classification of video game is ACTION ADVENTURE. Dummy variable.

MATUREᵢ ESRB rating of video game is Mature. Dummy variable. TEENᵢ ESRB rating of video game is Teen. Dummy variable. EVERYONEᵢ ESRB rating of video game is Everyone. Dummy variable. METASCOREᵢ Expert scores of video game i.

USER_SCOREᵢ Consumer scores of video game i.

The information about the sales is obtained from the database of vgchartz.com. This website contains all legal copies of video games sold worldwide. The consumer reviews and the expert reviews are obtained from metacritic.com. The expert scores (METASCOREᵢ) and consumer scores (USER_SCOREᵢ) are also obtained from Metacritic.com as well as genre (FPS, RACINGᵢ, ACTION ADVENTUREᵢ) and age rating (MATUREᵢ, TEENᵢ, EVERYONEᵢ) dummies. The database of metacritic has been widely used before (Situmeang, Leenders and Wijnberg, 2014).

The operationalization of the variables

Dependent variable

The dependent variable in this research is the performance of video games. The global product sales of video games have been chosen to measure the performance of video games. The number of legal copies sold is a good indicator for the performance of the video games.

(21)

21 Although it is inevitable to recognize that people are downloading games digitally, especially for the PC platform. The number of legal copies sold is however still a good indicator of the performance of video games.

Independent variables

The independent variables in this research are the affective content in reviews of consumer and experts. The consumer reviews and expert reviews have been quantified and analyzed with LIWC. This resulted in a sum of both positive- and negative emotions of reviews per video game. The sum of the positive emotions per video game minus the sum of negative emotions per video game resulted in the positive affective content in the reviews per video game. This have been done for all reviews of all video games of all the video game platforms and resulted in the positive affective content of consumer reviews of a video game (C_EMOᵢ) and the positive affective content of expert reviews of a video game (E_EMOᵢ).

For examining the effect of reviews of previous editions to the performance of their sequels all non-sequel games were deleted from the data, and the expert- and consumer reviews of previous editions of sequels were established. This resulted in the independent variables of positive affective content of consumer reviews of previous editions (C_EMOt-1ᵢ) and positive affective content of expert reviews of previous editions (E_EMOt-1ᵢ). The same has been done for the sales of previous edition of sequels of video games (SALESt-1ᵢ).

For examining the effect of reviews of previous editions to the performance of their sequels, not all the video game platforms have been analyzed because some of the platforms did not have enough sequels in their database.

(22)

22 Table 3 shows an overview of the video game platforms and number of games that have been analyzed for the examination of the effect on of reviews on sequels.

Table 3. Platforms and number of games with sequels Platform Number of Games

Nintendo Wii 342 Playstation 1 84 Playstation 3 757 PSP 234 Xbox 360 Xbox 360 823 PC 1771 Control variables

Age rating (Teen, Mature and Everyone) and Genre (FPS, Racing and Action Adventure) are dummy variables. These variables are the control variables in this research. They control for possible attribute change that may influence consumer of expert reviews. The expert scores (METASCOREᵢ) and consumer scores (USER SCOREᵢ) are also control variables. All control variables are obtained from metacritic.com.

(23)

23

4. Results

4.1 Descriptive statistics and correlations

First, the descriptive statistics and the correlations between the variables will be presented. The results are presented in Table 4 to Table 4.10. The results of the correlation analysis can be used as preliminary evidence to support the hypotheses. The correlation analysis provides preliminary evidence that positive affective content in consumer reviews has a positive and significant correlation with the performance of the video games for almost all platforms (3DS: r = .590, p < .01; PS1: r = .413, p <.01; PS3: r = .334, p < .01; PSP: r = .492, p < .01; Vita: r = .415, p < .01; Xbox: r =.580, p < .01; Xbox360: r = .262, p < .01; PC: r = .140, p < .01; Wii: r = .540, p < .01).

Furthermore the correlation analysis provides evidence that positive affective content in expert reviews has a positive and significant correlation with the performance of video games for almost all platforms (3DS: r = .472, p < .01; PS1: r = .301, p < .01; PS3: r = .174, p < .01; PSP: r = .282, p < .01; Vita: r = .195, p < .01; Xbox: r = .355, p < .01; Xbox360: r = .144, p < .01; PC: r = .141, p <.01; Wii: r = .325, p < .01).

The correlation analysis also provides evidence that positive affective content in consumer reviews of earlier editions has a positive and significant correlation with the performance of their sequel (PS3: r = .123, p < .01; Wii: r = .172, p <.01; Xbox: r = .267, p <.01; Xbox360: r = .157, p < .01; PSP: r = .196, p < .01; PC: r =.150, p < .01). This positive and significant correlation has again been found for almost all the platforms.

The positive affective content in expert reviews of earlier editions also has a positive significant correlation with the performance of their sequel for several platforms (Wii: r = .165, p < .01; Xbox: r = .130, p < .05; PSP: r = .139, p < .05; PC: r = .119, p < .01).

Finally, the correlation analysis provides evidence that the performance of earlier editions has a positive and significant correlation with the performance of their sequel (PS1: r = .488, p

(24)

24 <.01; PS3: r =.122, p < .01; Wii: r = .334, p < .01; Xbox: r = .248, p < .01; Xbox360: r = .733, p < .01; PSP: r = .453, p < .01; PC: r = .533, p < .01). This correlation has been found significant for every video game platform.

(25)

25 T a b le 4 . D es cr ip tiv e s ta tis tic s a n d c o rr el atio n ma tr ix PS1 V ar iab le N am e M e an S. D . 1 2 3 4 5 6 7 8 9 10 11 12 13 1 S al es 1. 070 1.830 2 C _ Em o 74. 023 130. 469 . 413** 3 E _ em o 32. 294 33. 291 .301** .253** 4 C _e m ot -1 7. 470 133. 978 .099 .261* .113 5 E _e m ot -1 38. 543 35. 956 .011 . 009 .234** . 235** 6 S al es t-1 1. 484 2. 249 .488** . 183 .228** .317** . 181 7 FPS .017 .132 .039 -.021 .003 -. 031 .0 51 . 008 8 R aci n g .106 . 308 .127 -. 110 .047 -.137 -. 244* .020 -.046 9 A ctio n .0929 . 290 .217** .007 -.001 -.022 -. 065 .203 -.043 -. 11 10 A dve nt ur e 10 M at u re .885 . 284 .052 -.018 -.030 -,018 . 048 . 256* . 076 -.107 . 598** 11 T een .354 8 .480 .180* .143* .107 .111 .236* . 212 .110 -.198* -.112 -. 233** 12 E ve ryone .508 8 .501 -. 186** -.093 -.079 -. 072 -.270* -.332** -.137* -.2 24** -.204** -.317** -. 761** 13 M et as co re 70. 97 17. 196 .438** .288** .553** . 0 05 . 227 .0 83 .033 . 031 .091 -. 034 . 093 -. 091 14 U ser _ sco re 7. 88 1. 792 .183** . 3 02** .215** .143 -.010 . 063 .010 .055 .095 -. 006 .148 .072 .575** * S igni fi ca nt a t p < .05. ** S igni fi ca nt a t p < .01 . S ale s in millio n u n its

(26)

26 T ab le 4.1 D es cr ip tiv e s ta tis tic s a n d c o rr ela tio n ma tr ix PS3 V ar iab le N am e M e an S .D . 1 2 3 4 5 6 7 8 9 10 11 12 13 1 S al es .767 1.634 2 C _ Em o 67.480 109.508 .334** 3 E _ em o 89.565 98.620 .174** .586** 4 C _e m ot -1 79.726 120.966 .123** .106** .016 5 E _e m ot -1 98.312 102.559 .016 .069 .188** .591** 6 S al es t-1 .8003 1.724 .122** .012 .003 .028 .028 7 FPS .0593 .236 .135** .082** -.044 .036 -.06 3 .0 60 8 R aci n g .0420 .200 -.002 -.043 .021 -.025 -.007 .000 -.053 9 A ctio n . 0871 .282 .038 .062* .013 -.002 -.008 -.015 -.078** -.0 65* A dve nt ur e 2980 10 M at u re .2875 .452 .260** .180** -.090 .10 7** -.11 1** .034 .227** -.125** .192** 11 T een ..457 -.106** -.050 .029 -.041 .026 .070 -.018 -.071** -. 009 -.414** 12 E ve ryone .2110 .408 -.071** -.101 ** -.044 - .075 -.04 3 -.130 .203** -. 153** -.328** -.337 13 M et as co re 70.89 13.545 .331** .447** .533** .123** .148** .035 -.002 -.0 27 -.002 -.050 -.086** .037 14 U ser s co re 4.96 3.492 -.040 .010 -.069 .0 55 .114 .080 .064 -.019 .075 .249** .186** .0 90 .719** * S igni fi ca nt a t p < .05. ** S igni fi ca nt a t p < .01 . S al es i n m il li on u n its

(27)

27 T ab le 4.2 D es cr ip tiv e s ta tis tic s a n d c o rr ela tio n ma tr ix N int endo W ii V ar iab le N am e M e an S .D . 1 2 3 4 5 6 7 8 9 10 11 12 13 1 S al es .92 3.928 2 C _ Em o 45.067 97.383 .540** 3 E _ em o 75.395 101 .20 .325** .678** 4 C _e m ot -1 59.260 120.996 .172** .285** .384** 5 E _e m ot -1 81.608 105.168 .165** .242** .352** .732** 6 S al es t-1 1.410 5.516 .334** .111* .205** .6 16** .378** 7 FPS .224 .148 -.016 .140** .071* .137* .056 -.013 8 R aci n g .061 .239 .001 -.038 -.087* .049 .038 -.016 -.039 9 A ctio n .071 .257 -.021 .027 -.059 .060 -.070 -.017 -.042 -.07 1* A dve nt ur e 10 M at u re .424 .201 -.0 23 .081 -.009 .047 -.052 -.032 .093** -.05 4 .231** 11 T een .213 .409 .044 .034 .020 -.022 -.086 -.070 .168** -.12 0** .034 -.110** 12 E ve ryone .468 .499 .081* .004 .034 .016 .096 .026 -.142** .240** -.202** .198** -.4 89** 13 M et as co re 64.86 15.008 .143** .399** .586** .127** .209** .077 .062 -.1 33** -.055 -.012 . 032 -.011 14 U ser _ sco re 6.78 2.139 .066 .224 ** .288** .091 .064 .022 .038 -.052 .036 .052 .086* .012 .688** * S igni fi ca nt a t p < .05. ** S igni fi ca nt a t p < .01 . S al es i n m il li on u n its

(28)

28 T ab le 4.3 D es cr ip tiv e s ta tis tic s a n d c o rr ela tio n ma tr ix XB OX Va ri ab le N am e M e an S .D . 1 2 3 4 5 6 7 8 9 10 11 12 13 1 S al es .3176 .5683 2 C_ Em o 51.637 75.291 .580** 3 E _ em o 80.553 79.247 .355** .542** 4 C _e m ot -1 64.133 86.930 .267** .220** .206** 5 E _e m ot -1 87.755 79.247 .130* .083 .144** .497** 6 Sa les t-1 .440 .729 .248** .152** .168** .617** .422** 7 FPS .0756 .264 .139** .121** .045 .096 .093 .097 8 R aci n g .1167 .321 -.033 -.0 4 4 -.018 -.037 .0 18 -.027 -.104** 9 A ctio n .1397 .346 .013 .041 .079* .168** .124* .121* -.115** -.146* A dve nt ur e 10 M at u re .2197 .413 .090* .093** . 127** .133* .122* .109* .272** -.182** .226** 11 T een .3923 .488 -.040 .026 -.053 -.048 -.051 -.001 -.001 -.071 .032 .424** 12 E ve ryone .3321 .471 -.028 -.08 5 -.043 -.155 -. 055 -.081 -.202** -.244** -.221** -.372 ** -.567** 13 M et as co re 69.87 14.283 .374** .470** .615** .154** .129** .161** .038** -.018 .002 -.001 -.104** 1.06** 14 U ser s co re 7.51 1.723 .133** .289** .289** .098 .027 -.014 -.041 .016 .000 -.047 .001 .118** .592** * S ig n if ic an t a t p < .05. ** S igni fi ca nt a t p < .01 . S al es i n m il li on u n its

(29)

29 T ab le 4.4 D es cr ip tiv e s ta tis tic s a n d c o rr ela tio n ma tr ix XB OX3 6 0 V ar iab le N am e M e an S .D . 1 2 3 4 5 6 7 8 9 10 11 12 13 1 S al es .84 2.172 2 C _ Em o 69.733 105.888 .262** 3 E _ em o 114.1877 118.405 .144** .575** 4 C _e m ot -1 82.1057 116.843 .157** .170** .130** 5 E _e m ot -1 126.1689 127.499 .065 .148** .287** .581** 6 S al es t-1 1.2843 2.665 .733** .018 -.029 .230** .111** 7 FPS .0666 .24945 .104** .058 -. 027 .060 -. 065 .161** 8 R aci n g .0473 .2123 -.044 -.00 4 .068* -.047 .028 -.032 .060* 9 A ctio n .0809 .2728 .003 .048 .061* .058 .089* .001 -.079** -.066* A dve nt ur e 10 M at u re .2801 .4492 .294** .224** -.035 .167** .017 .285** .243** -.131** .195** 11 T een .2607 .4392 -.140** -.03 2 .037 .014 .021 -.145** -.008 -.071** -.021 -.370* * 12 E ve ryone .2457 .4306 -.103** -.131 ** -.044 -.101** -. 036 -.102** -.152** -210** -.151** - .356 ** -.339* * 13 M et as co re 69.03 14.349 .270** .401** .620** .148** .219** .109** .020 -.011 .004 . 040 -.054 .006 14 U ser s co re 6.5 2.135 -.057* .194** .223** .064 .092 -.172** .017 .020 .027 .039 .051 .015 .624** * S igni fi ca nt a t p < .05. ** S igni fi ca nt a t p < .01 . S al es i n m il li on u n its

(30)

30 T ab le 4.5 D es cr ip tiv e s ta tis tic s a n d c o rr ela tio n ma tr ix PSP V ar iab le N am e M e an S .D . 1 2 3 4 5 6 7 8 9 10 11 12 13 1 S al es .31 .755 2 C _ Em o 37.744 62.82 .492** 3 E _ em o 76.759 83.69 .282** .543** 4 C _e m ot -1 45.696 72.33 .196** .119 .258** 5 E _e m ot -1 85.04 88.22 .139* .185** .469** .507** 6 S al es t-1 .3974 .8339 .453** .123 .146* .434** .170** 7 FPS .0058 .0758 -.032 .0 43 -.061 .028 -.046 .126 8 R aci n g .0481 .2141 .026 .039 .029 -.034 -.054 .051 -.01 7 9 A ctio n .0635 .244 .153** .063 .023 .223** .133** .163** -.020 -.05 9 A dve nt ur e 10 M at u re .0923 .2897 .096* .160** .074 .207** .187** .132* .063 -.072 .244** 11 T een .3635 .4841 .020 -.015 -.033 -.067 -.139* * .065 .048 -.076 -.033 -.241** 12 E ve ryone .3154 .4651 -.106* -.067 -.016 -.072 .020 -.078 -.052 .157 ** -.160 - .216** -.513** 13 M et as co re 68.60 11.820 .245** .414** .575** .176 * .280** . 111 -.055 -.01 0 -.041 .005 -.078 .132** 14 U ser s co re 7.25 2.219 .103* .198* * .167** .167* .148* .092 .033 .092 . 020 .069 .123** .098* .523** * S igni fi ca nt a t p < .05. ** S igni fi ca nt a t p < .01 . S al es i n m il li on u n its

(31)

31 T ab le 4.6 D es cr ip ti v e sta tis tic s a n d c o rr ela tio n ma tr ix PC V ar iab le N am e M e an S. D. 1 2 3 4 5 6 7 8 9 10 11 12 13 1 S al es .0822 .44283 2 C _ Em o 80.9525 123.856 .140** 3 E _ em o 4109.815 5308.45 .141** .517** 4 C _e m ot -1 96.4628 133.686 .150** .286** .203** 5 E _e m ot -1 4949.353 5704.61 .119** .168 ** .322 ** .523** 6 S al es t-1 .1095 .5024 .533** .104** .098** .173** .141** 7 FPS .0804 .2719 .013 .057** -.002 .078** -.0 10 .041 8 R aci n g .0364 .1873 -.033 -.03 0* -.031 -.044 -.05 1 .041 -.057** 9 A ctio n .0583 .2342 -.015 .010 -.017 .004 -.01 2 .001 -.074** -.04 8** A dve nt ur e 10 M at u re 1864 .38 95 .082* * .173 ** .056** .160** .088 ** .107* * .307** -.089** .159** 11 T een .2939 .4556 -.010 .027 .084** -.003 -.0 16 -.017 -.010 -.0 71** -.013 -.309** 12 E ve ryone .1603 .3669 -.035* -.10 2** -.059** -.115** -.0 97** -.062** -.129** .2 50 .068** -.209** -.2 82** 13 M et as co re 70.21 13.75 0 .127** .399** .476** .151* * .146 ** . 101** -.051** .021 -.068* .017 -. 028 .034 14 U ser s co re 6.80 1.959 .0 02 .245** .186** .060* .072* * -.024 -.004 .01 8 -.009 .040** .123** .061** .614* * . * S igni fi ca nt a t p < .05. ** S igni fi ca nt a t p < .01 . S al es i n m il li on u n its

(32)

32 Ta b le 4 .7 D es cr ip tiv e s ta tis tic s a n d c o rr ela tio n ma tr ix N int endo 3ds V ar iab le N am e M ean S .D . 1 2 3 4 5 6 7 8 9 10 1 S al es .43 1.524 2 C _ Em o 55.564 114.079 .590** 3 E _ em o 97.457 116.741 .472** .750** 4 FPS .00 84 .0917 -.026 - .045 -.066 5 R aci n g .0042 .0649 .-.019 -.036 -.053 -.006 6 A ctio n .0211 .1446 -.0 03 -.039 -.0 57 -.014 -.010 A dve nt ur e 7 M at u re .046 4 .2108 -.050 .060 .007 -.020 -.014 -.032 8 T een .16 88 .3753 -.092 -.025 -.021 -.205** -.029 -.06 6 -.099 9 E ve ryone .41 35 .4935 .109 .009 .022 -.077 .078 -.12 3 -.185** -.378** 10 M et as co re 67.2 3 12.912 .170* .515** .568** -.112 .078 -.041 .062 11 U ser s co re 6.5 1 1.927 .073 .267** .241** -.062 .119 .088 .141 .740** * S igni fi ca nt a t p < .05. ** S igni fi ca nt a t p < .01 . S al es i n m il li on u n its

(33)

(34)

34 T a b le 4 .8 D es cr ip tiv e s ta tis tic s a n d c o rr ela tio n ma tr ix XB OX ONE V ar iab le N am e M ean S .D . 1 2 3 4 5 6 7 8 1 S al es .52 .643 2 C _ Em o 97.407 130.89 .246 3 E _ em o 66.000 99.59 .049 .657** 4. R ac ing .0370 .192245 .150 .323 .423* 5 M at u re .2593 .44658 .586** .411* .077 -.116 6 T een .074 1 .26688 -.233 -.211 -.360 -.055 -.167 7 E ve ryone .259 3 .44658 -.084 -.039 .216 .331 -.350 -.167 8 M et as co re 68.7 16.746 .143 .253 .554* .159 .206 -.596* .200 9 U ser s co re 5.81 3.498 -.093 .030 .241 -.047 -.017 -.108 -.042 .688** * S igni fi ca nt a t p < .05. ** S igni fi ca nt a t p < .01 . S al es i n m il li on u n its

(35)

35 T a b le 4 .9 D es cr ip tiv e s ta tis tic s a n d c o rr ela tio n ma tr ix PS4 V ar iab le N am e M ean S .D . 1 2 3 4 5 6 7 1 S al es 1.00 1.069 2 C _ Em o 148.625 138.53 .373 3 E _ em o 67.6250 67.0285 .373 .613 4 M at u re .500 0 .53452 1.00** .373 .373 5 T een .250 0 .46291 -.577 .138 .128 -.577 6 E ve ryone .125 0 .35355 -.378 -.398 -.510 -.378 -.218 7 M et as co re 68.56 14.842 .918** .581 .418 .918** -.510 -.641 8 U ser s co re 5.57 2.225 .100 .527 .187 .100 .439 -.708 .327 * S igni fi ca nt a t p < .05. ** S igni fi ca nt a t p < .01 . S al es i n m il li on u n its

(36)

36 4.10 D es cr ip tiv e s ta tis tic s a n d c o rr ela tio n ma tr ix P S V ita V ar iab le N am e M e an S. D . 1 2 3 4 5 6 7 8 9 1 S al es .03 .175 2 C _ Em o 59.361 103.805 .415** 3 E _ em o 69.062 94.92 .195** .726** 4 R aci n g .0052 .07236 -.013 -.030 -.050 5 A ctio n .0052 .07236 -.013 -.041 -.050 -.005 A dve nt ur e 6 M at u re .094 2 .29293 .147* .277** .108 -.023 -.023 7 T een .26 18 .44076 -.107 -.010 -.049 -.043 -.043 -.192 ** 8 E ve ryone .293 2 .4564 2 -.050 -.131 .026 -. 047 -.047 -.208** -.384** 9 M et as co re 71.3 1 11.455 -.139 .251** .420** -.097 .106 .074 10 U ser s co re 6.84 2.338 -.213** -.002 .060 .034 -.011 .067 .200** .008 .726** * S igni fi ca nt a t p < .05. ** S igni fi ca nt a t p < .01 . S al es i n m il li on u n its

(37)

37 4.2 Hypotheses testing

A linear regression was conducted to test the hypotheses. The results of the regression analysis are presented in Table 4 to Table 4.5. The results suggest that the relationship between affective content in consumer reviews and the performance of video games is significant for (3DS: β= .533, p < .001; PS1: β= .241, p = .026; PSVita: β= .472, p < .001;

PSP: β= .419, p < .001; Xbox360: β= .222, p < .001; Xbox: β= .523, p < .001; Wii: β= .637, p

<.001; PC: β: .085, p = .003). And the relationship between affective content in expert reviews and the performance of video games is significant for (3DS: β= .186, p = .041). Also there has been found a significant relationship between affective content in expert reviews and the performance of video games, this effect is however negative (PSVita: β= -.276, p = .018). These results partially support H1a and H1b because the relationship between affective content in the reviews is not significant for all the video game platforms.

To examine the relationship between affective content in reviews of earlier editions and the performance of their sequels a linear regression has been performed. The relationship between the affective content in consumer reviews of earlier editions and the performance of their sequels is significant but negative for (Wii: β: -.320, P = .001).

The relationship between affective content in expert reviews of earlier editions and the performance of their sequels is not significant for any of the video game platforms. H2a and H2b have not been supported by the results. To establish the relationship between past sales of earlier editions and the performance of their sequels a linear regression has been performed again. The relationship between the sales of earlier editions of the video games and the performance of their sequels is significant for (PS1: β: .272, p = .018; PS3: β: .386, p = .001;

Wii: β: .451, p < .001; PSP: β: .413, p < .001; PC: β: .438, p < .001; Xbox360: β: .582, p <

(38)

38 Table 4. Regression Results 3DS Regression Results Vita

Variable Name B SE β Sig. B SE β Sig. Constant .279 .510 .585 .131 .084 .122 C_EMOᵢ .006 .001 .533 .000 .001 .000 .472 .000 E_EMOᵢ .002 .001 .186 .041 .000 .000 -.276 .018 ACTION .415 .488 .049 .397 ADVENTUREᵢ MATUREᵢ -.347 .375 -.054 .356 .037 .042 .083 .383 TEENᵢ -.036 .224 -.010 .871 -.029 .031 -.092 .356 EVERYONEᵢ .497 .181 .177 .007 -.006 .032 -.018 .855 METASCOREᵢ -.143 .010 .041 .642 -.002 .002 -.130 .285 User SCOREᵢ .415 .085 -.141 .092 .001 .014 .009 .941

Table 4.1. Regression Results Xbox One Regression Results PS4

Variable Name B SE β Sig. B SE β Sig. Constant 1.576 1.044 .160 2.000 C_EMOᵢ -.001 .002 -.118 .741 -1,676E-16 0.000 E_EMOᵢ .003 .063 .841 .418 -1,794E-16 0.000 RACINGᵢ -.092 .967 -.095 .926 3.493-16 0.000 MATUREᵢ .796 .426 1.806 .098 4.495E-16 0.000 TEENᵢ -.153 .625 -.069 .811 -2.000 0.000 -.913 EVERYONEᵢ -.072 .402 -.051 .861 -2.000 0.000 -.707 METASCOREᵢ .011 .016 .271 .513 USER SCOREᵢ -.378 .253 -.1493 .163

(39)

39 Table 4.3. Regression Results Xbox360 Regression Results Xbox

Variable Name B SE β Sig. B SE β Sig.

Constant -.135 .484 .781 -.299 .389 .443 C_EMOᵢ .005 .001 .222 .000 .005 .000 .523 .000 E_EMOᵢ .001 .001 .032 .388 .000 .001 .047 .391 C_EMOt-1ᵢ -.001 .001 -.031 .357 .001 .001 .078 .207 E_EMOt-1ᵢ -.001 .001 -.050 .136 .000 .001 -.035 .507 SALESt-1ᵢ .597 .030 .582 .000 .094 .061 .090 .125 FPSᵢ -.262 .262 -.028 .318 .209 .141 .073 .139 RACINGᵢ -.499 .313 -.044 .111 -.047 .124 -.018 .704 ACTION -.242 .236 -.028 .306 -.010 .122 -.004 .932 ADVENTUREᵢ MATUREᵢ .380 .216 .073 .079 .186 .195 .090 .340 TEENᵢ -.268 .210 -.047 .204 .028 .186 .016 .879 EVERYONEᵢ -.180 .218 -.031 .408 .044 .181 .026 .807 METASCOREᵢ .031 .008 .160 .000 .011 .005 .141 .024 USER SCOREᵢ -.272 .061 -.150 .000 -.069 .036 -.101 .052

Table 4.2. Regression Results PS3 1 Regression Results PS3 3 Variable Name B SE β Sig. B SE β Sig Constant -.3869 3.887 .324 -3.133 2.165 .156 C_EMOᵢ .004 .002 .241 .026 .005 .044 .222 .219 E_EMOᵢ -.017 .009 -.192 .065 -.005 .004 -.234 .199 C_EMOt-1ᵢ .001 .002 .030 .778 -.004 .005 -.132 .387 E_EMOt-1ᵢ -.013 .008 -.173 .097 .003 .003 .150 .328 SALESt-1ᵢ .296 .121 .272 .018 .537 .151 .386 .001 FPSᵢ -.927 1.474 .505 .000 .953 .793 .432 .236 RACINGᵢ 1.418 .959 -.166 .140 .587 1.862 -.281 .754 ACTION 1.611 .819 -.060 .535 -.354 1.010 .151 .727 ADVENTUREᵢ MATUREᵢ 3.012 2.351 .174 .146 1.410 .951 .035 .146 TEENᵢ 4.668 2.201 .217 .055 .713 .970 -.041 .467 EVERYONEᵢ 2.704 2.131 .370 .260 .317 1.042 .315 .762 METASCOREᵢ .097 .021 .860 .039 .078 .027 .148 .005 USER SCOREᵢ -.617 .411 .485 .210 -.408 .197 .053 .044

(40)

40 Table 4.4. Regression Results PSP Regression Results Wii

Variable Name B SE β Sig. B SE β Sig.

Constant -.317 .501 .528 2.158 2.040 .291 C_EMOᵢ .005 .001 .419 .000 .030 .003 .637 .000 E_EMOᵢ -.002 .001 -.161 .066 -.006 .004 -.112 .161 C_EMOt-1ᵢ -.001 .001 -.078 .308 -.016 .005 -.320 .001 E_EMOt-1ᵢ .001 .001 .047 .528 .007 .004 .133 .094 SALESt-1ᵢ .463 .075 .413 .000 .467 .066 .451 .000 FPSᵢ -.927 1.474 -.141 .020 -1.486 1.965 -.039 .450 RACINGᵢ 1.418 .959 .065 .282 -1.65 2.080 -.004 .937 ACTION 1.611 .819 .136 .036 -.054 1.112 -.003 .961 ADVENTUREᵢ MATUREᵢ 3.012 2.351 -.012 .872 -.717 1.385 -.028 .605 TEENᵢ 4.668 2.201 -.025 .782 -.035 .838 -.003 .967 EVERYONEᵢ 2.704 2.131 -.080 .369 .826 .776 .067 .288 METASCOREᵢ .097 .021 .101 .234 -.038 .032 -.081 .239 USER SCOREᵢ -.617 .411 -.015 .826 .003 .287 .001 .993

Table 4.5. Regression Results PC Variable Name B SE β Sig. Constant -.189 .086 .028 C_EMOᵢ .000 .000 .085 .003 E_EMOᵢ 4.0867E-06 .000 .046 .108 C_EMOt-1ᵢ .000 .000 .026 .341 E_EMOt-1ᵢ 7.218E-07 .000 .008 .775 SALESt-1ᵢ .453 .024 .483 .000 FPSᵢ -.043 .041 -.025 .297 RACINGᵢ -.043 .070 -.014 .535 ACTION -.063 .048 -.030 .194 ADVENTUREᵢ MATUREᵢ .020 .038 .016 .608 TEENᵢ -.017 .034 -.015 .623 EVERYONEᵢ -.047 .041 -.034 .234 METASCOREᵢ .005 .001 .120 .000 USER SCOREᵢ -.026 .010 -.072 .010

(41)

41

5. Discussion

The influence of evaluations of experience goods and their effect on the performance of those products have been discussed earlier in literature (e.g. Gemser, van Oostrum and Leenders, 2006; Bassuroy, Chatterjee and Ravid, 2003). Earlier literature has found that reviews play a significant role on the performance of products in different ways. According to Gemser, van Oostrum and Leenders (2006) reviews can play a role due both their influence effect and their prediction effect. Duan, Gu and Whinston (2008) found that the number of posts is associated with sales while Godes and Mayzlin (2004) found that WOM dispersion is an important predictor of the performance of products. Chevalier and Mayzlin (2004) mentioned the importance of WOM valence and research from Chen, Wuu and Yoon (2004) and Ludwig et al. (2013) suggested that WOM volume is highly associated with sales.

This research focused on the combination of affect in evaluations and the performance of products. Earlier literature has shown that affect influences judgement, choice (Loewenstein and Lerner, 2003; Darke, Chattopadhyay and Ashworth, 2002) decision-making (Shiv and Fedorikhin, 1999) attitude (Morris, Woo, Geason and Kim, 2002) and evaluation (Lench, Flores and Bench, 2011). Regarding this literature, it was expected that affect in reviews would play a significant role on the performance of video games.

Next to the research to the supposed influence of affect on behavior, earlier literature of Ludwig et al. (2013) has yet explored if the affective content in evaluations has effect on the performance. The study Ludwig et al. (2013) found an asymmetrical relationship between affective content in evaluations and the performance, meaning that greater increases in positive affective content in customer reviews have a smaller effect on subsequent increases in conversion rate. Andrade (2005) and Roehm and Roehm (2005) also found a non-linear relationship between affect and thoughts and behaviors, due to the effect of extreme affect.

(42)

42 The results of this study suggest that affective content in evaluations has a positive effect on performance of experience goods. User reviews and expert reviews both had a positive effect on the global product sales of video games. Especially affective content in user reviews seem to have a strong positive effect on performance, considering that 8 of the 11 video game platforms showed a strong positive relationship between affective content in reviews and the global product sales.

Not for all the video game platforms there have been found relationships between affective content in reviews and the performance of the video games. There was no relationship between affective content in consumer reviews, in the Xbox One and PS4 dataset, and global product sales. This however can be explained due to the fact that the number of games in the dataset was low for both Xbox One (N= 41) and PS4 (N = 10).

The relationship between affective content in expert reviews has only been found for the video game platform 3DS. Although the correlation analysis found that there were many significant relationships between the affective content in expert reviews and the performance of video games. There has even been found a significant negative effect for the influence of expert reviews on the performance of video games for video game platform Vita. This contradicts the findings in earlier research of Litman (1983) and Basuroy, Boatwright and Kamakura (2003) that found that expert reviews play a role in predicting the performance of products. It seems that affective content in expert reviews plays a less significant role in the performance of video games than affective content in user reviews. This is consistent with earlier research of Dellacoras et al. (2003). In this study, the impact of both movie-critics and user reviews was examined, and it was found that user reviews were twice as powerful in predicting box-office revenues as reviews by movie-critics. A possible explanation for the difference between the influence of expert reviews and the influence of consumer reviews can be that expert reviews are independent evaluators. Expert reviewers often make up their

The role of affective content in online reviews : how evaluations of past editions influence the performance of video game sequels