Master thesis
Expert reviews, online comments and ticket sales of theatres.
Is there a relation?
Marije Lüschen Studentnr: 6063624
Supervisor: dr. F.B. Situmeang
Second reader: prof. dr. N.M. Wijnberg
MSc Business Studies: Entrepreneurship and Management in the Creative Industries. 30 June 2014
2
§1. Table of contents
2. Abstract P. 3
3. Introduction P.4
4. Literature review and theoretical framework P.7
4.1 Selection systems P.7
4.2 Influence of traditional reviews P.8
4.3 Influence of online reviews and comments P.10
4.4 Influence of text analysis P.16
5. Research design and methodology P.20
5.1 Data collection P.20 5.2 Variables P.21 5.3 Methodology P.21 6. Results P.23 6.1 Descriptive statistics P.23 6.2 Correlations P.25 6.3 Panel regression in R P.28 7. Discussion P.31
7.1 Theoretical foundations and hypotheses P.31
7.2 Limitations and directions for further research P.34
7.3 Implications P.37
8. Conclusion P.39
9. References P.40
3
§2. Abstract
In this thesis the influence of affective cues in expert reviews and online comments on ticket sales is tested. This relation was explored in the Dutch theatre industry. In contrast to previous studies, this thesis made use of text analysis tool LIWC. Therefore, focus was not only on the quantitative aspects of reviews and comments, but also on the contents and emotions expressed in these reviews and comments. Based on previous literature, there was an expectancy to find a positive relation between positive affective cues in expert reviews and online comments on theatre ticket sales, and a negative relation between negative affective cues in expert reviews and online comments on theatre ticket sales. In support of the first hypothesis was found that positive expert reviews have an immediate positive influence on theatre ticket sales. Negative expert reviews have a negative effect on theatre ticket sales, this effect occurs immediately and one week later. Opposite results were found for the second hypothesis. Positive online comments have an immediate negative effect on theatre ticket sales. Negative online comments have a positive effect on theatre ticket sales, this effect occurs after one week. In addition was found that increase in the number of online comments, the length of the comments and the theatre ticket sales of the previous week had a positive effect on theatre ticket sales. The number of expert reviews had an immediate negative effect on theatre ticket sales. These results contribute to existing theories since there was no research done in the theatre industry while using text analysis tools. Furthermore, this research opens a lot of doors for further research since these results are not in line with results of previous studies.
Keywords: Dutch theatre industry, Text analysis tools, affective cues, expert reviews, online comments, theatre ticket sales.
4
§3. Introduction
Consider this quote made by actor James Franco:
“Sadly Ben Brantley and the NYT have embarrassed themselves. Brantley is such a little bitch he should be working for Gawker.com instead of the paper of record. The theatre community hates him and for good reason, he’s an idiot.”
This quote is a response to a bad review posted in the New York Times for the theatre show where Franco plays one of the leading roles. Considering the reaction of Franco, Franco might think this bad review harms his reputation and maybe even the reputation of the show, and therefore choses to defend himself. Franco is not the only one who attaches value to reviews. The influence of reviews has been studied extensively. Reviews are especially important in the creative industry, since cultural products are sold in a different way than non-cultural products. Consumers buy cultural products without knowing the quality of the product upfront (Caves, 2003). Consumers try to get information about the quality through reviews and other information like for example commercials. According to Cameron (1995), criticism provides information which can be used by consumers in assembling the hedonic price for the cultural demand. It also serves to distinguish in terms of quality and originality. Reputation is often established and built by critics (Cameron, 1995), this explains the enraged reaction of Franco.
Since the rise of Web 2.0, there was a rise in information sources. The technological part of internet stayed quite consistent but the content and use did not. The development of Social Network Sites (SNS) made it possible for consumers to share more information with each other. The definition of SNS by Boyd and Ellinson (2008) will be used in this thesis: “Social network sites are web-based services that allow individuals to (1) construct a public or semi-public profile within a bounded system, (2) articulate a list of other users with whom they share a connection, and (3) view and traverse their list of connections and those made by
5 others within the system.” SNS like Facebook or public forums are places where consumers can share information about products and services. Besides the traditional reviews in for example newspapers, these sources of information on SNS can help consumers to make purchase decisions. This might be especially helpful for purchases of cultural products, since consumers have limited information about these products (Elberse, 2007; Hadida, 2009). The influence of SNS on performance in the creative industries has been studied a lot. Dhar and Chang (2009) researched whether online chatter has an influence on sales in the music industry. Chevalier and Mayzlin (2003) explored whether online consumer reviews had an influence on book sales. Duan, Gu and Whinston (2008) looked at the relation between online content and movie sales.
The Dutch theatre industry will be the main focus of this thesis. The Dutch theatre industry is big. Theatre visits in the Netherlands are number one of the European rankings (Raad van Cultuur, 2011). The industry is partly subsidized by the government to enable all different theatre genres to be present in the Dutch cultural landscape. Around 47% of the Dutch population older than twelve visits a theatre once or more than once a year (Langeveld, 2006). This seems like a lot, but while the average number of theatre visits stayed the same, there is an increase in the number of shows, theatres and seats (Commissie d’Ancona, 2006). Creativity is needed to get more people in the theatres (NRC Handelsblad, 2014). This thesis may provide more information about factors, like reviews and online comments, that might have an influence on their sales. In this research there will be focused on the influence of traditional reviews and online comments on theatre ticket sales. Theatre has received little attention regarding the relation between online comments, reviews and sales. Therefore research in this industry is needed. This research might reveal other relations between reviews, online comments and sales than already revealed in studies about other cultural industries. Theatre is, for example, in comparison with books, movies and music, not easy to
6 duplicate online. Besides posting criticism on social media forums, consumers of books, movies and music can also share the actual products, like movies or music singles, online (Dewan & Ramaprasad, Forthcoming). This in contrast to theatre, which has little ability to be distributed or shared online. Another contribution of this research are the methods used. Previous research focused on the number of reviews, the number of online comments or online star rankings (Shrum, 1991; Duan et al., 2008). This study will make use of the text analysis program linguistic inquiry and word count (LIWC). Via this program the affective content exposed in traditional reviews and online comments will be looked at. Affective content words reveal the content of a text. People can, for example, express anger, happiness or sadness (Ludwig et al., 2013, Pang & Lee, 2008). Furthermore, a match in linguistic style can increase credibility and shared perceptions, which might lead to purchase (Ireland & Pennebaker, 2010). The type of relation between traditional reviews, the affective content on SNS and theatre ticket sales is being explored in this thesis. It is expected that the affective words expressed in expert reviews and online comments about theatre shows influences the ticket sales of theatre shows.
In short, this research is needed because it makes use of new text analyses methods. Perhaps focusing on affective content reveals other relations or a stronger relation between reviews, online comments and sales. Furthermore, by focusing on the Dutch theatre industry, this research might give more insight in factors that can have an influence on their ticket sales. This is needed since the theatre population stayed consistent while the number of seats, shows and theatres has increased (Commissie d’Ancona, 2006).
In the next sections existing literature will be reviewed, thereafter the methodology will be explained, followed by results of the research and it will finish with the final conclusion and discussion.
7
§4. Literature review and theoretical framework
In this section existing literature about this thesis topic is reviewed. First a short explanation on selection system theory will be presented, thereafter the influence of traditional reviews on sales, third the influence of online comments on sales and thereafter research that made use of text analyses will be covered. This section shall finish with the hypothesis and the theoretical framework.
§4.1 Selection systems.
Value for creative goods is created through selectors. Mol and Wijnberg (2011) describe this process as follows: “Expounding how product value is socially constructed, selection system theory describes competitive processes in terms of the ‘selected’, actors who are being selected, and ‘selectors’, actors who attribute value of the products of the selected” (Wijnberg & Gemser, 2000; Gemser & Wijnberg, 2001; Wijnberg, 2004; Priem, 2007; in Mol & Wijnberg, 2011). Wijnberg (1995) distinguishes three types of selection systems: market-selection, peer-selection and expert selection. “These selection systems provide a short description of the relation between the selected and selectors and more generally of competitive process” (Wijnberg, 1995).
Market selection is compared with natural selection, a term which is often used in biology (Wijnberg, 1995). The consumer decides what they like and will consume. The producers are the selected and the selectors are the consumers. When a group of selectors and the group of those who are being selected are the same, this could be considered as peer-selection (Wijnberg, 1995). An example of peer-peer-selection might be when a painter is also reviewing paintings of his colleagues. Selectors who are not members of the group in which the selection takes place or are not consumers, are experts (Wijnberg, 1995). In this thesis, reviews written by experts will be taken into account. Thus, emphasis will be on the expert
8 selection system. Market selection will also play a role since the online comments of potential consumers also will be taken into account. Peer selection is not considered in this research.
§4.2 Influence of traditional reviews.
In this section, earlier research about the influence of traditional reviews will be discussed. Traditional reviews are, in this thesis, considered as reviews published in traditional media like for example newspapers or magazines.
Shrum (1991) was one of the first to examine the relationship between reviews and audience participation in the performing arts. He wondered whether there was a difference in influence of reviews in different performance genres and if there was an association between the kind of reviews and the size of the audience who attended to the researched shows. Shrum (1991) conducted his research on the Fringe festival, a festival for performing arts in Edinburgh. He took the different kind of shows, reviews for these shows and audience attendance numbers into account. He made a distinction between the visibility of a show through reviews, - whether a show was reviewed and if so, how many times it was reviewed- and modality – the kind of attention it received in reviews, either positive or negative-. Shrum (1991) found a mediating effect of reviews between performing arts and audience size. More positive reviews are in relation with larger audience, but reviewers do not have the power to make or break a show. The visibility of a show is more important than the modality of the reviews. “Reading or seeing reviews in a number of places highlights the importance of a show and maintains its salience for taste publics in ways that may be more effective, but serve much the same function, as advertising” (Shrum, 1991).
Eliashberg and Shugan (1997) researched the relation between market performance of entertainment services and the potential role of critics. Before their research few scholars have focused on this topic. Eliashberg and Shugan (1997) questioned whether critics play a role in
9 predicting and/or influencing box office performance. They focused on the movie industry to research this potential relation. They found no relation between critical reviews and early box office results, while there was a correlation with the total cumulative box office results. Eliashberg and Shugan (1997) concluded that film critics are not influencers but predictors of a movies success.
Gemser, van Oostrum and Leenders (2007) used the research of Eliashberg and Shugan (1997) as an inspiration. They focused on the influence of reviews on movie sales and wondered which factors determine whether film reviews are predictors or influencers. They predicted a difference in influence of reviews between mainstream and art house movies. Gemser et al. (2007) argued that art house moviegoers were led by film reviews when making a movie choice, so reviews are influencers in this case. While on the other hand, mainstream moviegoers rely mainly on other information sources when they make a choice, like movie stars or advertising. Gemser et al. (2007) looked at reviews written by experts in Dutch newspapers, they only included reviews which were published one day before the movie première or on the première day itself. They looked at the effects of these reviews on the opening weekend and on the cumulative box office revenue. Gemser et al. (2007) found that the number and size of a review had a direct influence on the behavior of the art house moviegoers. Reviews are an influencer in this case. The number and size of reviews about mainstream movies does not have this influence, they only function as a predictor for movie performance.
These three articles show that reviews can have an influence on performance and sales. Although the type of influence may depend on the cultural goods where the research focuses on. However, these researchers decided to focus on the quantitative aspects of the reviews. This thesis focuses on the content of the reviews. It is expected that the effect of reviews on ticket sales stays the same while looking at the influence of the content of the
10 reviews. So expectation is to find a positive relation between the positive affective words in expert reviews and theatre ticket sales, and a negative relation between negative affective words in expert reviews and theatre ticket sales. This leads to hypothesis 1 (figure 1):
H1A: Positive affective content in expert reviews about a theatre show lead to an increase in the ticket sales for this show.
H1B: Negative affective content in expert reviews about a theatre show lead to a decrease in the ticket sales for this show.
Negative affective content in expert reviews Positive affective content in expert reviews Ticket sales theatres
-+
(Figure 1, Hypothesis 1).§4.3 Influence of online reviews and comments
The influence of online reviews and word-of-mouth (WOM) plays an important role in consumers’ purchase decisions (Chen & Xie, 2008, Park, Lee & Han, 2007, Zhang et al., 2010). Chen and Xie (2008) even argue that online consumer reviews can serve as a new
11 element in the marketing communications mix. Articles discussed hereafter will show the influence of online reviews and WOM.
During the rise of web 2.0, Chevalier and Mayzlin (2003) decided to study the effect of consumer reviews on firms’ sales patterns. During this rise of web 2.0 there was a widespread belief that websites must provide community content to build brand loyalty. However, to the knowledge of Chevalier and Mayzlin (2003), there was not any literature about the influence of online community content in consumer decision making. The authors compared consumer reviews and sales on book websites Amazon.com and Barnesandnoble.com. They chose to compare two websites because in this way they could research what happened with sales if someone posted a negative review on, for example, Amazon.com but not on Barnesandnoble.com. They were wondering whether this would have an impact on the sales of Amazon.com and if the sales of Barnesandnoble.com would stay the same in this case. Chevalier and Mayzlin (2003) found that reviews are extremely positive on both sites, but Amazon.com has a larger amount of reviews and those are longer. They found that customers rely more on the reviews text than on summary statistics. However, one star reviews and rankings have a bigger influence than five star reviews. Amazon.com has a larger number of reviews and greater relative sales. Thus, an increase in the number of reviews on Amazon.com relative to Barnesandnoble.com continues to improve sales at Amazon.com relative to Barnesandnoble.com (Chevalier & Mayzlin, 2003). In conclusion, the more reviews the more sales.
Chen, Wu and Yoon (2004) were, after Chevalier and Mayzlin (2003), the first to start exploring the impact of online reviews on sales. They saw that firms began to invest in creating a user friendly virtual community where consumers could share opinions and experiences. This seemed necessary since the ocean of information, made possible by the Web 2.0, caused uncertainty in quality and high search costs for identifying relevant
12 information (Chen et al., 2004). Chen et al. (2004) explored the effectiveness of the effort the firms made. Since earlier research focused primarily on the quality of information provided on the internet and trust issues between buyers and sellers (Houser & Wooders 2000; Lee et al., 2000; Lucking-Reiley et al., 2000; Resnick, et al., 2002; Ba & Pavlou 2002 in Chen et al., 2004). So Chen et al. (2004) switched from focusing on the service to focusing on the product itself. They investigated the impact of recommendations and consumer feedback on sales based on data from Amazon.com. This study might look the same as the study of Chevalier and Mayzlin (2003), which was mentioned earlier. However, Chevalier and Mayzlin (2003) only focused on consumer recommendations while the retail recommendation might have an influence too. Chen et al. (2004) take these retailer recommendations into account, allowing them to distinguish between the two effects. The results indicate that a larger number of retailer recommendations improve sales of Amazon. The recommendations work better for less-popular books than for more-popular books. This might be explained by the so called search cost argument. When a consumer has to search longer for less-popular books, search costs are higher and thus they may rely more on recommendations to locate a product of interest (Chen et al., 2004). Consumer ratings are not found to be related to Amazon’s sales but the number of consumer reviews do have a positive effect on sales. So this research shows that the effort firms put in their online mediums pays off.
Duan et al. (2008) looked at the relation between online content and products sales. They chose to focus on online word-of-mouth and product sales in the movie industry. They describe the positive feedback mechanism between word-of-mouth (WOM) and product sales. WOM leads to more product sales, and more product sales lead to more word-of-mouth, which in turn generates even more product sales. “Word-of-mouth is not only a driving force in consumer purchase but also an outcome of retail sales” (Duan et al., 2008). This positive feedback mechanism is an extension of the earlier research from Shrum (1991) on traditional
13 reviews. His findings indicate the importance of attention for a show, it doesn’t even matter a lot if this attention is positive or negative. This positive feedback mechanism shows the importance of the attention given by reviews or WOM. They chose to look at this positive feedback mechanism in the movie industry, because experts agree word-of-mouth is one of the most important factors underlying the length of stay of a movie and has a certain influence on sales. Duan et al. (2008) looked at online reviews and rankings of 71 movies. From this data they developed the variables word-of-mouth valence and word-of-mouth volume. They showed that a movies box office revenue and mouth valence influence word-of-mouth volume. In turn, word-of-word-of-mouth volume leads to a higher box office performance. “This positive feedback mechanism highlights the importance of word-of-mouth in generating and sustaining retail revenue” (Duan et al., 2008).
Research did not only focus on online reviews but also on online chatter and blogs about products/services. Gopinath, Chintagunta and Venkataraman (2013) studied the effects of blogs and advertising on the local-market movie box office performance. They looked at the pre- and post-release blog volume, blog valence and advertising of 75 movies in 208 geographic markets in the United Stated. They measured the effects of the pre-release factors on opening day box office performance and the pre- and post-release factors on box office performance one month after the release. They found that release day performance is most effected by pre-release blog volume and advertising. Post-release box office performance is influenced by advertising and post-release blog valence.
Stephen and Galak (2012) studied the effects of traditional and social earned media on sales. The authors looked at the effects of earned media, so not paid media like advertising, on sales. The setting is the non-profit organization Kiva. An organization that operates on online marketplace for microloans for entrepreneurs in developing countries. They analyzed the daily sales of Kiva from new customers, the daily sales from repeat customers, the
14 advertisement and articles in traditional media, blogs posts and the number of posts on the community forum. They found that social and traditional earned media both influenced the sales. Social earned media seemed to play an important role in driving traditional media activity. Per-event sales impact of traditional earned media is larger than social earned media. But after adjusting for event frequency, social earned media activity is more frequent, social earned media sales elasticity is greater than traditional earned media’s. The study of Stephen and Galak (2012) confirms the importance of both traditional and online media. Therefore both traditional reviews and online comments are used in this thesis.
Dhar and Chang (2009) and Dewan and Ramparasad (2013 and Forthcoming) looked at the music industry. Dhar and Chang (2009) examined whether online chatter on blogs and social networking sites could predict sales in the music industry. They looked at the online chatter about 108 music albums, four weeks before and four weeks after their release dates. With these data they tried to predict the album sales one, two and three weeks ahead. They found that future sales are positively correlated with the number of blog posts about an album. Also traditional factors influenced the album sales. For example the status of the record label and reviews in traditional mainstream media had an effect on future album sales. Dewan and Ramaprasad (2012) studied how online blogs could influence the consumption decisions of consumers. They found that the intensity of music sampling was positively related with the popularity of the blog among previous consumers. This effect is stronger for niche music than mainstream music. This is in line with the later research from Dewan and Ramaprasad (Forthcoming) on the music industry. Based on their 2012 research, Dewan and Ramaprasad (Forthcoming) decided to look at the effects of social media and traditional media on music sales. They compared the influence of blog buzz and radio play and looked what kind of influence they had on the music sales of albums and songs. They expected that the traditional media would have a positive effect on music sales, because songs played on the radio get a
15 tremendous exposure. Effects of social media could either be negative or positive for music sales. On one hand social media creates word-of-mouth, whereby online blogging and sharing can persuade readers to buy the music. While on the other hand, social media also creates a big platform not only for sharing opinions but also about the real music. This would have a negative effect on music sales. Dewan and Ramaprasad (Forthcoming) also took the kind of music into consideration. They made a distinction between niche and mainstream music. As expected, Dewan and Ramaprasad (Forthcoming) found a positive relationship between traditional media and music sales on both album and singles. However, social media are not related with album sales and negatively related to single sales. So sharing of music has a stronger and negative influence on music sales than the potential positive word-of-mouth.
Influence of online reviews and online WOM has been studied extensively. The articles discussed above have different methods to research the relation between online content and sales. Most of them found a positive relation between the amount of reviews and WOM on sales (Chen et al., 2004; Chevalier & Mayzlin, 2003; Duan et al., 2008; Gopinath et al., 2013). Furthermore, Stephen and Galak (2013) showed that both traditional reviews and online chatter have an influence on sales. Volume of reviews might not be the only indicators of increase of sales and performance. Dewan and Ramaprasad (2012) found a negative relation between online WOM and single sales. Chen et al. (2004) did find a relation between the amount of reviews and sales but not between ratings and sales. Different methods might lead to different results. Counting the number of reviews is one method, but this thesis uses another method. In this research the affective content in online comments is measured. Therefore a choice is made for a text analyzing approach. The influence of positive and negative affective words in reviews and comments will be measured. There are several researches which describe the importance of affect, Ludwig et al. (2013) explored whether affective content has an influence on sales. Their article will be discussed in the next section.
16
§4.4 The influence of expert reviews and online comments while making use of text analysis
This last section of the literature review is about a research that looked at relations between online WOM and sales/performance while making use of text analyzing tools. “Text mining, also known as text data analyzing or knowledge discovery from textual databases, refers to the process of extracting interesting and non-trivial patterns or knowledge from text documents” (Tan, 1999). Several studies used text mining programs to explore what kind of influence words can have (Liu, 2010, Ludwig et al., 2013, Pang & Lee, 2008). This thesis will also use a text analyzing program which focusses on affective analyses, namely LIWC. The term affect describes the internal feeling state (Cohen et al., 2008). Affective content words are words such as ‘perfect’, ‘terrible’, they describe emotions like anger or happiness (Ludwig et al., 2013). Affective stimuli can influence negative or positive assessments (Strahan, Spencer & Zanna, 2002). Ludwig et al. (2013) did study the effects of affective content in the creative industry, they focused on book sales. This research is an important guideline for this thesis.
Ludwig et al. (2013) gave a good example in what way a study, while using LIWC which focuses on affective cues, should be conducted. Ludwig et al. (2013) argued that current research on online reviews offered little guidance. Since most studies focus merely on quantitative surrogates of reviews, such as review volume and the influence of star ratings, the effect of these reviews on sales remains ambiguous (Ludwig et al., 2013). It is certain reviews contribute in helping with consumers purchase decisions, but maybe affective cues provided in these reviews – e.g. I love it or I hate it- might influence consumers attitudes (Cohen et al., 2008 in Ludwig et al., 2013). Earlier research on affective cues in text showed that it is likely that affective content in reviews drives behavior (Das, Martinez-Jerez, and Tufano 2005; Jones, Ravid, and Rafaeli 2004 in Ludwig et al., 2013). However, it is not certain whether
17 these affective words are immediate predictors of the influence of customer reviews on retail success. There is limited evidence of nonlinear relationships between these two (Ludwig et al., 2013). Ludwig et al. (2013) aimed to provide more clarity on the influence of textual properties of consumer reviews on online retailers’ conversion rates. They examined the impact of affective content from a dynamic perspective, they noted how changes in the affective content influence changes in conversion rates over time. Furthermore, they want to add to recent research by noting the impact of linguistic style of customer reviews on online conversion rates. Ludwig et al. (2013) distinguish between linguistic content and style elements. Contents are nouns, regular verbs and many adjectives and adverbs. Content always has style words, also known as function words. These are pronouns, prepositions, conjunctions, auxiliary verbs and other esoteric categories (Tausczik & Pennebakker, 2010). Affective content words which convey for example happiness, reveal the intent of a text. A match in linguistic styles between a reviewer and reader of the review might lead to a more positive evaluation of the review. Ludwig et al. (2013) tested the linear and quadratic –when the reviews on one certain site are too positive compared to other sites- relationships between affective content in reviews and conversion rates. They collected reviews and sales data on Amazon.com. Amazon.com made conversion behavior of their customers publicly available during data collection, so Ludwig et al. (2013) could make a direct link between customer reviews and retail performance. They analyzed the collected reviews automatically with the linguistic inquiry and word count (LIWC) program. There was a strong positive and significant effect found between increasing levels of positive affective content on subsequent conversion rates (Ludwig et al. 2013). Furthermore, extreme positive changes in affective content of reviews had a smaller impact on conversion rates than moderate changes in the affective content of reviews. This is not the case for negative changes. An increasing degree of linguistic style matches (LSM) also increases conversion rates, and a combination of
18 positive changes in affective content and increasing degrees of LSM significantly predicts increases in conversion rates. In conclusion, affective words do have an influence on conversion rates.
The research of Ludwig et al. (2013) is a good example on how to research the effects of affective content, therefore this thesis will further elaborate on the methods used in their research. Based on the results of all earlier reviewed articles in this thesis and in line with hypothesis 1, it is expected to find a positive relation between the positive affective content of online comments and theatre ticket sales, and a negative relation between the negative affective content of online comments and theatre ticket sales. This leads to the following hypothesis (figure 2):
H2A Positive affective content in online comments about a theatre show lead to an increase in theatre ticket sales.
H2B Negative affective content in online comments about a theatre show lead to a decrease in theatre ticket sales.
Negative affective content in online comments Positive affective content in online comments Ticket sales theatres
-+
(Figure 2, Hypothesis 2).19 These two hypotheses lead to the following conceptual model (Figure 3). In the next section the methods of research will be explained, starting with the data collection, second explanation about the variables and last about the methodology.
Negative affective content in online comments Positive affective content in online comments Ticket sales theatres
-+
Positive affective content in expert reviews Negative affective content in expert reviews+
20
§5. Research design and methodology
In this section data collection is described, thereafter the variables will be named and the method of this research will be described.
§5.1 Data collection.
This study is a quantitative study which focuses on three types of data: 1) ticket sales data from theatres, 2) expert reviews about the theatre shows and 3) online comments written about the theatre shows.
Theatre and theatre production companies in the Netherlands have been identified via keyword search on Google. There were 52 theatres and theatre companies selected and approached through email. In this email an explanation was given about the thesis and asked to get access to their data about ticket sales. Theatres and theatre production companies who chose to cooperate, provided data about ticket sales. Ticket sales data was gathered from three theatre shows and varied from ticket sales from four up to five months. The ticket sales were measured one week before the première of the show up to respectively four or five months after.
One day after the première of the show, most expert reviews were published in the Dutch newspapers. Copies of these reviews were found on the websites of the theatre shows. Furthermore, a search was conducted via search engines to verify that all the reviews were placed on the websites. User generated content was gathered from the Facebook page of the shows involved and forum website Musicalworld. Since this thesis looks at online comments and not specifically online consumer reviews, a public forum with topics about the shows was a good place for online chatter. These webpages were selected, because they were among the most popular pages for theatre consumers to visit. Content was gathered from the première of the show until four or five months after the première. This process was done manually. There
21 were on average around 45 comments posted in a week on the researched webpages per production. In total around 1000 comments were gathered per production.
§5.2 Variables.
The dependent variable was the number of tickets sold per week for the theatre shows. The independent variables was the affective content in expert reviews and the online comments. Furthermore control variables were conducted for the length of the online comments and expert reviews and the number of comments and reviews was also taken into account. These were measured since it is expected that these variables are factors which have an influence on the researched relations. Earlier research already found this relation (Shrum, 1991; Chevalier & Mayzlin, 2003; Chen et al., 2004).
§5.3 Methodology
In this research the influence of expert reviews and online comments on ticket sales of theatres was measured. Previous research focused only on the quantitative factors reviews and comments which could have an influence on sales. This thesis also focuses on the content of the reviews and comments. To measure the qualitative content and change it into quantitative content, a content analysis was needed. Content analysis is an increasingly popular method to study user-generated comments online (Singh, Hillmer and Ze, 2011). Content analysis uses automated, systematic procedures that ensure the objectivity, reproducibility and reliability of the data analysis (Chung, Pennebaker and Fiedler, 2007).
The user generated content from Facebook and Musicalworld was automatically analyzed with the linguistic inquiry and word count program (LIWC). The expert reviews were also analyzed in LIWC. The program is developed in order to provide an efficient and effective method for studying the various emotional, cognitive, and structural components
22 present in individuals’ verbal and written speech samples (Pennebaker et al., 2007). LIWC dictionaries offer strong, reliable convergence between the dimensions they extract and content ratings performed by human coders (Ludwig et al., 2013). Content from the reviews and comments will be coded on emotions expressed. For example, the comment “Fantastic!’ will be coded as positive. Via LIWC the expert reviews and online comments about the shows were analyzed. The results of the positive and negative emotions expressed by the expert and potential consumers were used for further analysis.
An Excel sheet was created with the shows, number of weeks and the ticket sales data. The positive and negative emotions values gathered in LIWC were added to this. The length and the number of the comments and reviews was also measured in Excel. These values were added to the existing Excel sheet. The Excel sheet with the panel data was further analyzed via statistical program R. Panel data are cross-sectional and time series. There are multiple entities, each of which has repeated measurements at different time periods. An OLS regression analysis processed in the more commonly used SPSS is therefore not useful. The regression analysis does not take the different entities and time periods into account which would result in an incorrect analysis. R was able to cope with the panel data set used in this thesis. The Breusch-Pagan Lagrange multiplier was used to analyze the data.
In the next section of this thesis the results of this analysis will be presented, thereafter these results will be discussed.
23
§6. Results
In this section the results will be presented. First the descriptive statistics are presented, second the correlations and third the results of the regression analysis.
§6.1 Descriptive statistics.
Table 1 shows a summary of the descriptive statistics of the data. For the three shows which were being researched, respectively sixteen, twenty-two and twenty-two weeks of ticket sales data, expert reviews and comments were collected. The expert reviews were published in
Table 1. Descriptive statistics.
N = 60
Variables Frequency in weeks Frequency in %
Show 1 16 26,70%
2 22 36,70%
3 22 36,70%
Minimum Maximum Mean Std. Deviation
Number of expert reviews 7 9 7,9 0,796
Average Length Expert Reviews 395,5 524,56 468,28 56,68
Variance Length Expert Reviews 9873,53 18582,48 13190,35 4133,72
Positive Emotions Expert Reviews 1,76 2,55 2,17 0,32
Negative Emotions Expert Reviews 0,61 1,04 0,87 0,2
Number of online comments 11 152 48,17 31,94
Average Length online comments 9,11 54,27 25,45 9,36
Variance Length online comments 61,18 15876,75 2061,57 3100,06
Positive emotions online comments 1,31 7,43 3,98 1,32
24 week three of the dataset since on the last day of week two, the première took place. The number of expert reviews varied from nine till seven reviews. The average length of these reviews was around 470 words. The minimum value of the positive emotions in the expert reviews was 1,76. The maximum was 2,55. The standard deviation for those positive emotions expressed in expert reviews was 0,32. The reviewers were all quite consistent in their reviews. This also holds while looking at the negative emotions expressed by experts in their reviews. The minimum value was 0,61 and the maximum 1,04. The standard deviation was 0,2. Notable is the difference between the positive and negative emotions expressed in the reviews. Overall more positive emotions are expressed by reviewers in all three shows than negative emotions.
The number of online comments varies from 11 comments up to 152 comments per week. The average length of the online comments is around 25 words. The minimum value of positive emotions expressed online was 1.31, the maximum was 7,43. The standard deviation was 1,32. The negative emotions expressed online , had a minimum value of 0,21 and a maximum of 2,01. The standard deviation was 0,43.
There were a lot more online comments than expert reviews. Reviews were only published in the third week of the data set while the online comments were published online every week. The expert reviews were longer than the online comments. The differences in length were greater for online comments in comparison to the expert reviews. Experts might follow a certain standard when it comes to writing a review. Both expert reviews and online comments were overall more positive than negative (resp. 2,17 and 3,9 compared with 0,87 and 1,01). Potential consumers were overall more outspoken about their negative and positive emotions than experts. The average values of both negative and positive emotions were higher than those of experts.
25
§6.2 Correlations.
Table 2 shows the means, standard deviations and correlations of the dataset. On average, there were around 8 expert reviews published per show and around 45 online comments published per show per week. The variable positive expert emotions had an average of 2,17 with a standard deviation of 0,32 and negative expert emotions had an average of 0,87 with the standard deviation of 0,20. Positive emotions expressed online had a mean value of 3,98 with a standard deviation of 1,32, negative emotions had a mean value of 1,01 and a standard deviation of 0,43. Mean ticket sales was 0,05 with a standard deviation of 0,03. The table shows all the correlations between the variables. The significant correlations that have an influence on the relation that is being researched will be discussed.
Both the number of online comments and the total ticket sales had significant negative correlations with the week of the shows (resp. r = -,507, P < 0,1; r = -,507, P < 0,1). This negative correlation suggests that, when the number of weeks a show plays increases, the number of online comments and ticket sales decreases. The relation between sales and the number of weeks a show plays is shown in figure 4.Ticket sales and the number of online comments, however, have a significant positive correlation with each other (r = 0,5 P < 0,1). This suggests that, when the number of online comments increases, the amount of tickets sold will also increase. Positive emotions expressed by potential consumers have a significant negative correlation with ticket sales (r = -,299, p < 0,5) . This could suggest that increase of positive online comments leads to decrease in ticket sales. However, since panel data is not taken into account in this exploration of the data, the next analysis might show other results. Correlation offers an indication of the relation between variables but do not take the different shows and weeks in this panel data set into account. Therefore a panel regression is performed in the next section.
26
27
Table 2.
Means, standard deviations and correlations
Variable Mean SD 1 2 3 4 5 6 7 8 9 10 11 12 13
1 Show 2,10 0,80 -
2 Week 10,70 6,13 ,183 -
3 Ticket sales 0,05 0,03 -,234 -,507** -
4 Number of expert reviews 7,90 0,80 -1,000** -,183 ,234 -
5 Average length exp. review 468,28 56,68 -,075 -,132 ,176 ,075 -
6 Variance length exp. r. 13198,35 4133,72 ,884** ,107 -,132 -,884** ,400** -
7 Positive emotions exp. r. 2,17 0,32 ,996** ,173 -,219 -,996** ,009 ,920** -
8 Negative emotions exp. r. 0,87 0,20 ,017 -,115 ,155 -,017 ,996** ,482** ,100 -
9 Number of online comment 49,17 31,94 ,177 -,507** ,500** -,177 ,108 ,214 ,187 ,125 -
10 Average length comments. 25,45 9,36 -,027 -,085 ,111 ,027 -,059 -,053 -,032 -,062 ,301* -
11 Variance length comments. 2061,57 3100,06 ,120 ,057 -,042 -,120 -,172 ,029 ,105 -,161 ,253 ,776** -
12 Positive emotions com. 3,98 1,32 ,177 ,123 -,299* -,177 -,082 ,124 ,170 -,066 -,285* -,486** -,177 -
13 Negative emotions com. 1,01 0,43 ,058 -,009 ,024 -,058 -,075 ,018 ,052 -,070 ,009 ,151 ,010 -,297* - **. Correlation is significant at the 0.01 level (2-tailed).
28
§6.3 Panel regression in R
The previous two sections showed more descriptives about the data and showed correlations. The following dataset consists of panel data. A regression analysis cannot be performed, since a regression analysis would not consider heterogeneity across groups or time (Torres-Reyna, N.D). Therefore a long panel regression was conducted. After performing the Breusch-Pagan test, presence of heteroskedasticity was found (P = 8.677e-08). Therefore the results in table 3 are controlled for heteroskedasticity. Besides the control for heteroskedasticity, variables were lagged. Since, for example, comments of the previous week can have an influence of comments on the week after. With the results in table 3, hypotheses can be tested. The R-squared has a value of 0.855, meaning that around 85% of these variables explain the relation. 15% is caused by other variables which are not taken into consideration in this thesis. Possible variables that are not being researched in this thesis, will be discussed in the next section.
H1A predicted that positive affective written expert reviews will result in an increase in ticket sales. This hypothesis is confirmed (t=5.81, Pr=3.02e-06, P < 0). This positive effect is found in lag 0. Meaning that positive expert reviews have an immediate positive effect on theatre ticket sales. Positive expert reviews do not have a significant effect in lag 1 or 2. H1B predicted that negative affective written expert reviews will result in a decrease in ticket sales. This effect was confirmed in lag 0 (t=-3.56, Pr= 0.001, P < 0.001) and lag 1 (t=-2,08, Pr=0.046, P < 0,001). No significant effect was found in lag 2. Meaning that negative affective written expert reviews lead to a decrease in theatre ticket sales immediately and up to one week later.
H2A predicted that positive affective written comments lead to an increase in ticket sales. However, the results show the opposite effect (t=-2.12, Pr= 0.040. p < 0.01) indicating that more positive online comments lead to a decrease in ticket sales. This result was found in lag 0, meaning that this effect is immediate. This effect was not found in lag 1 and 2, so there
29 was no effect of positive comments on ticket sales found after one or two weeks. Hypothesis H2A is not supported. H2B predicted that negative affective written comments lead to a decrease in ticket sales. Lag 0 and 1 showed no significant effects. Lag 2 does show a significant effect (t=5.45, Pr=3,513e-06, P<0). This indicates that an increase in negative written online comments lead to an increase of ticket sales after a while. Negative comments had a slower effect on ticket sales. The consequences of these results will be discussed in the paragraph seven.
The T test for coefficients also shows significant results for the control variables. A direct positive, significant effect was found between the number of online comments and ticket sales (t=2.53, Pr=0.016, p<0.01). Meaning that an increase in online comments leads to an increase of ticket sales. A direct, negative effect was found for the number of expert reviews on theatre ticket sales (t=-5.41, Pr=9.118e06, P<000). Meaning that an increase of expert reviews lead to a decrease in theatre ticket sales. Another positive, significant effect is found between the length of comments and ticket sales (t=2.12, Pr=0.041, P<0.01). This effect was found in lag 1. The length of the comments had an effect on ticket sales after one week. These effects were not found in lag 0 and 2. Last, a positive, significant effect was found between sales and ticket sales in lag 1 (t=4.31, Pr=0.001, P<0). Meaning, ticket sales of the previous week had an effect on ticket sales of the current week. There was no significant effect found between ticket sales and ticket sales in lag 2.
In the next section will be discussed what these results mean for the theoretical contributions of this thesis. Furthermore, the implications and directions for further research will be discussed.
30
Table 3. T test for coefficients
Positive emotions expert week 0 Positive emotions expert week 1 Positive emotions expert week 2 Negative emotions expert week 0 Negative emotions expert week 1 Negative emotions expert week 2 Number of expert reviews week 0 Number of expert reviews week 1 Number of expert reviews week 2 Positive emotions comments week 0 Positive emotions comments week 1 Positive emotions comments week 2 Negative emotions comments week 0 Negative emotions comments week 1 Negative emotions comments week 2 Number online comments week 0 Number online comments week 1 Number online comments week 2 Average length online comment week 0 Average length online comment week 1 Average length online comment week 2 Ticket sales week 1
Ticket sales week 2
Estimate 8.4159e-02 -1.3558e-03 -8.3991e-03 -6.4769e-02 -2.4551e-02 3.5632e-02 -2.2177e-02 5.0861e-03 -2.6949e-03 -1.6910e-03 -9.1592e-04 -9.2562e-04 -1.2579e-03 -6.2289e-03 3.6171e-03 5.6729e-04 1.3055-e04 -7580e-05 1.2771e-04 3.3374e-04 2.8713e-04 4.6021e-01 -5.6595e-02 Std. Error 1.4475e-02 2.0153e-02 1.3459e-02 1.8202e-02 1.1779e-02 1.3939e-02 4.1012e-03 5.2553e-03 3.7777e-03 7.9597e-04 9.3811e-04 8.1511e-04 2.4902e-03 4.1446e-03 6.6394e-04 2.2397e-04 1.2367e-04 4.7053e-05 2.7798e-04 1.5726e-04 3.6225e-04 1.0681e-01 1.1063e-01 t-value 5.8140 -0.0673 -0.6241 -3.5583 -2.0843 2.5563 -5.4073 0.9678 -0.7134 -2.1244 -0.9763 -1.1356 -0.5052 -1.5029 5.4478 2.5329 -1.0556 1.0112 0.4594 2.1222 0.7926 4.3088 -0.5115 Pr(>|t|) 3.022e-06*** 0.946842 0.537639 0.001354** 0.046379* 0.016929 9.118e-06*** 0.341427 0.481514 0.0403810 * 0.3352330 0.2634374 0.6164424 0.1413535 3.513e-06 *** 0.0156849* 0.2980025 0.3184902 0.6486241 0.0405776* 0.4330512 0.0001163*** 0.6120063 Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 R-squared : 0.855 Adj. R-squared: 0.444
31
§7. Discussion
In this section the results of the analysis will be discussed, second the limitations of this study will be covered, directions for further research are provided, contributions of this research will be discussed and last, practical implications are presented.
§7.1 Theoretical foundations and hypotheses.
This thesis explored the effects of affective words in expert reviews and online comments on ticket sales. The setting was in the Dutch theatre industry. Previous studies did research effects of reviews and online WOM on sales (Chen et al., 2004; Dewan & Ramaprasad, 2012; Shrum, 1991; Mayzlin & Chevalier, 2003; Ludwig et al., 2013; Gopinath et al., 2013). But, to my knowledge, no study investigated this effect with the use of text analysis tool LIWC in the theatre industry. LIWC makes it possible to look not only at the quantitative numbers but also focus on the qualitative aspect of reviews and comments, namely the content. Based on earlier research, there was an expectation to find a positive relation between the positive affect words expressed in reviews and online comments and ticket sales of theatres. A negative relation was expected between negative affective words expressed in reviews and online comments and the ticket sales of theatres. This relation was confirmed for expert reviews. However, the analysis of the online comments, showed counter-intuitive results. There was a negative relation found between positive online comments and ticket sales, while a positive relation was found between negative online comments and ticket sales. Other significant relations were found between the number of online comments, the length of the online comments and the sales of the previous week on the theatre ticket sales.
Hypothesis 1A suggested that positive affective cues in expert reviews have a positive influence on ticket sales. An immediate positive effect of positive affective cues in expert reviews on theatre ticket sales was found. More positive reviews lead to an increase in theatre
32 ticket sales. This is in line with the results found in the research of Ludwig et al. (2013), who found that an increase in positive affective content leads to higher conversion rates. Hypothesis 1B suggested that negative affective cues in expert reviews will lead to a decrease of theatre ticket sales. This is confirmed in lag 0 and lag 1. There was no significant effect found in lag 2. The negative influence of negative affective cues in expert reviews had not been found in research from Ludwig et al. (2013), who made use of the same method. Ludwig et al. (2013) did not found any effects of negative affective content. The effects found in the current research might be explained by the different industry were the research is done. Reviews might be more important in the theatre industry since people can find less information about the shows compared to other cultural products. Dewan and Ramaprasad (Forthcoming) found a negative relation between online WOM and music single sales. Books, music and movies can be shared online and give people more information about the products. Theatre shows cannot be shared online, perhaps potential consumers rely more on the reviews since they have less information. Negative reviews might have a bigger impact in this case.
Hypothesis 2A suggested that positive online comments about a theatre show lead to an increase in theatre ticket sales. Hypothesis 2B suggested that negative online comments about a theatre show lead to a decrease in theatre ticket sales. The results show the opposite. Positive online comments have an immediate negative effect on theatre ticket sales. Negative online comments have a positive effect on theatre ticket sales, but this effect occurs later on. None of the previous discussed research has found these relations. Ludwig et al. (2013) found that positive affective content had a positive influence on sales, this was also expected, but not found in this thesis. Other previous discussed research found that the number of reviews and comments has an influence on sales (Chevalier & Mayzlin, 2003; Dhar & Chang, 2009; Duan et al., 2008). None of these researches found relations in line with these results. An explanation of these results might be found in the non-scientific literature like newspapers.
33 Around two years ago, several big newspapers and news sites revealed the existence of creation of fake online WOM and reviews (NY times, 2012; Forbes, 2012). An example is the website ‘gettingbookreviews.com’, were authors can literally buy positive online comments and reviews. This phenomenon was discussed on a lot of media platforms. Consumers might have more knowledge about the creation of these reviews and online comments. Perhaps negative comments are seen as more truthful and realistic, while too much positive reviews can lead to skepticism. This skepticism and perceptions on negative and positive comments and reviews might explain the results found in this thesis. Skepticism might lead to a decrease in purchase. While comments and reviews about shows which are being received as more realistic and truthful might lead to curiosity and an increase in purchase. These possible explanations are starting points for further research. Possibilities for further research are discussed in the next chapter, but first the other significant results will be discussed.
Besides the counter-intuitive results found for hypothesis two, other significant effects were found for the control variables. The number of online comments had a positive significant effect on theatre ticket sales, meaning that an increase in the number of these comments would lead to an increase in theatre ticket sales. This relation is in line with previous research from Shrum (1991), Chen et al. (2004) and Dhar and Chang (2009). However, an opposite effect was found for the number of expert reviews. The number of expert reviews had an immediate negative effect on theatre ticket sales. Meaning that an increase in expert reviews will lead to a decrease in theatre ticket sales. This is not in line with results in previous research (Shrum, 1991; Chen et al.; 2004; Chevalier & Mayzlin, 2003) but may be explained by another relation. The correlation table on page 26 shows a negative significant correlation between positive written expert reviews and the number of expert reviews (-0.996, P<0.01). The more reviews the less positive the content is. While positive content has a positive effect on theatre ticket sales. Another explanation might be that more
34 reviews lead to less consensus in all the different reviews, what might make potential consumers doubt. Another positive, significant effect was found between the length of the comments and theatre ticket sales. This effect was found in lag one, so it was not an immediate effect. Chevalier and Mayzlin (2003) found that consumers rely on the text of reviews and comments itself and not on summaries or statistics like star rankings around them. A longer comment gives a reader more information about the product or service they want to purchase. This could lead to a more positive attitude towards the comments and the product. Which in turn, might lead to an increase in purchase. This will be further discussed in the next section.
The last positive, significant effect found in this thesis was between theatre ticket sales of the previous week and theatre ticket sales of the current week. Indicating that an increase in theatre ticket sales in the previous week, leads to an increase of theatre ticket sales in the current week. Or, vice versa, a decrease in theatre ticket sales in the previous week leads to a decrease in theatre ticket sales in the current week. Both possibilities are shown in figure 4 of this thesis. Before the première, theatre ticket sales are increasing each week. After the première, theatre ticket sales decrease every week a bit more. This relation could be explained by the other variables researched. For example, the number of comments decreased after the première, this could be one of the explanation of the decrease in theatre ticket sales, while the number of comments increased from the week before the première until the première.
§7.2 Limitations and directions for future research.
This research has several limitations. These limitations are at the same time new starting points for future research. First, data selection seemed harder than expected. A lot of theatres were not willing to share their ticket sales data. This resulted in a data set which included three theatre shows which provided data of 60 weeks. Future research could replicate this
35 study, while using a larger dataset. It is suggested that data collection for this thesis was hard for several reasons. First, data collection started around January and lasted until April. Theatres showed their interest in the research topic, but indicated that they were very busy with conducting the program for the next theatre season. Future research might start data collection in a more calm period. Data gathering might start around the end of June until the end of August. Another reason for the theatres not to share their ticket sales data might be caused by the economic crisis. The Dutch theatre industry has suffered severely from the economic crisis (NRC Handelsblad, 2014). Perhaps the employees who were approached, did not want to reveal any losses or show that their theatre seats were not filled. A thesis student asking for these private, important data might lead to some restraints. Future research cannot solve all the economic problems, of course, but it might be possible that theatres are more willing to share data with a more prestigious researcher. For example when a PhD specialized in this topic or a professor or someone who has more contacts in the Dutch theatre industry asks for these private data. Since those might be taken more serious or more trusted as they already know the person who is going to use it.
Another limitation of the current research is the end product of the dataset. The dataset consists of 3 different theatre shows. This dataset is very homogeneous when it comes to the type of theatre show. Star quality is present, the theatre company has a high prestige and they are all from the same theatre genre. Therefore further research could use this master thesis as a starting point for more. A more heterogeneous dataset gives a better picture of how the Dutch theatre industry is constructed and perhaps leads to other results. When someone else decides to expand the current data set, taking into consideration, differences in genre, star quality and the prestige of the theatre company more specific results could be found. Perhaps there is a difference of influence of affective word cues in reviews and comments when it comes to the type of show. Gemser, van Oostrum and Leenders (1997) found a different
36 influence effect of reviews when it comes to arthouse and mainstream movies. Arthouse moviegoers were influenced by reviews, while mainstream moviegoers rely mainly on other resources such as advertising. Chen et al. (2004) found the same effect and explain this by their called cost argument. People have to look harder for, in their case, less popular books. Since search costs are higher, potential consumers tend to rely more heavily on reviews and online comments. This effect might also hold in the theatre industry. It would be very interesting to investigate this in future research.
Another limitation of the current study is the way in which the online WOM was gathered. This data was gathered manually. Further research can make use of automatic web crawling tools. This might result in more online comments and therefore a more reliable data set. Furthermore, in this study was not controlled for several variables. Variables like advertising and discounts have not been taken into account in this study, while they can also have an important influence on the consumers purchase decision. Future research might take this into consideration. It is expected that the larger, more mainstream theatre shows make more use of advertising. This can have an effect on the relation between the influence of reviews/online comments and sales.
Based on the results of this study, it would be very interesting to investigate why people decide to purchase when the number of comments is high and when they read negative online comments. It would be very interesting to investigate what kind of – unconscious- considerations consumers have during the purchase process. A research in an experimental setting might give more insights into those aspects. For example a situation where participants are being exposed to either a lot of negative reviews and/or comments, a lot of positive reviews and/or comments, a few negative reviews and/or comments and a few positive reviews and/or comments. When they are being asked to purchase, for example, tickets for the show of which they read the reviews and comments, there might be differences between the
37 various experimental groups. It would also be interesting to explore whether this consideration is conscious or unconscious. Last, it is important to explore the further research possibilities while making use of text analysis tools like LIWC.
§7.3 Implications.
This study has several practical and theoretical implications. Based on earlier research, there was an expectation to find a positive relation between the positive affective words expressed in reviews and online comments and the ticket sales of theatres. A negative relation was expected between negative affective words expressed in reviews and online comments and the ticket sales of theatres. This hypothesis was confirmed for expert reviews. However, the analysis showed counter-intuitive results for online comments. Positive results were found for the number of online comments, length of the comments and theatre ticket sales of the previous week on the current week.
Since both negative comments and the number of comments have a positive influence on theatre ticket sales, theatres could invest in motivating consumers to write online comments. It might seem hard for theatres to influence this, but there are some options to stimulate online WOM. For example, several theatre companies send their visitors an email after they have seen the show. In this email they ask to write about their experience and opinion on the website. Since negative comments do not have an influence on ticket sales they do not take that many risks when asking for this. An even extremer example is the musical Sister Act. At the end of the show the actors encourage the visitors to take photos and post them on the SNS of the show. This study confirms that these efforts of theatres are being rewarded.
This study also has theoretical implications. The results found in this study were counter-intuitive when it came to online comments. The current results have major impact on
38 the results of previous studies who investigated this relation. Their conclusions about the influence of negative and positive reviews and comments do not hold in this study. It seems that the previous conclusion, which led to more consensus, on this topic is partly in contradiction when there is made use of text analysis and the Dutch theatre industry as research area. A lot of research in the creative industry has been done on the relation between reviews, online WOM and sales. This study focused on the theatre industry. To my knowledge, no research had been done on this relation while making use of text analysis tool LIWC. Focusing on the content of reviews and online comments, instead of only on the quantitative aspects, might have led to other possible relations between these variables. This research brought the use of text analysis tools and the theatre industry under attention. This research is, hopefully, the beginning of more research with this method in this industry since creative industries are still suffering from the economic crisis. According to the NRC Handelsblad (2014), creativity is needed to fill the chairs of the Dutch theatres. More research can help to get a better understanding of the factors that have an positive effect on theatres ticket sales.
39
§8. Conclusion
The aim of this thesis was to test the influence of affective cues in expert reviews and online comments on ticket sales. This was tested in the Dutch theatre industry. Based on previous literature, expectation was to find a positive relation between positive affective cues in expert reviews and online comments on theatre ticket sales and a negative relation between negative affective cues in expert reviews and online comments on theatre ticket sales.
In support of the first hypothesis was found that positive expert reviews have an immediate positive influence on theatre ticket sales. Negative expert reviews have a negative effect on theatre ticket sales. Opposite results were found for the second hypothesis. Positive online comments have an immediate negative effect on theatre ticket sales. Negative online comments have a positive effect on theatre ticket sales, this effect occurs after one week. In addition was found that an increase in the number of online comments, the length of the comments and the theatre ticket sales of the previous week had a positive effect on theatre ticket sales. The number of expert reviews have an immediate negative effect on theatre ticket sales.
In conclusion, this study contributes to the literature by showing the urgency of more research in the theatre industry. It made a first step in exploring what kind of effect the affective aspects in reviews and online comments has on sales. Furthermore it confirmed that theatres are doing a good job by motivating consumers to write on their SNS websites. The findings of this study indicate a need for future research to further investigate this topic since some results are the opposite of results found in previous studies.
40
§9. References
Boyd, D. M., & Ellison, N. B. (2008). Social network sites: Definition, history, and scholarship. Journal of Computer‐Mediated Communication, 13(1), 210-230. Cameron, S. (1995). On the role of critics in the culture industry. Journal of Cultural
Economics, 19(4), 321-331.
Caves, R. E. (2003). Contracts between art and commerce. Journal of economic Perspectives, 73-84.
Chen, P. Y., Wu, S. Y., & Yoon, J. (2004). The impact of online recommendations and consumer feedback on sales. ICIS 2004 Proceedings, 58.
Chen, Y., & Xie, J. (2008). Online consumer review: Word-of-mouth as a new element of marketing communication mix. Management Science, 54(3), 477-491.
Chung, C., Pennebaker, J., and Fiedler, K. (2007). The Psychological Functions of Function Words. Social Communication, K. Fiedler, ed. New York: Psychology Press, 343-59.
Cohen, J. B., Pham, M. T., & Andrade, E. B. (2008). The nature and role of affect in consumer behavior. Handbook of consumer psychology, 297-348.
Dewan, S., & Ramaprasad, J. (2012). Research Note—Music Blogging, Online Sampling, and the Long Tail. Information Systems Research, 23(3-Part-2), 1056-1067.
Dewan, S., & Ramaprasad, J. (Forthcoming). Social media, traditional media and music sales.
Management Information Systems Quarterly, Forthcoming.
DiMaggio, P., & Useem, M. (1987). Cultural democracy in a period of cultural expansion: The social composition of arts audiences in the United States. Art and Society. Readings
in the Sociology of the Arts. Albany, NY: State University of New York, 141-176.
Dhar, V., & Chang, E. A. (2009). Does chatter matter? The impact of user-generated content on music sales. Journal of Interactive Marketing, 23(4), 300-307.
