1
Amsterdam, June 14th 2015
Thesis Seminar Business Studies 2014/2015 BSc Economics and Business
Supervisor Dr. Frederik Situmeang written by Pablo Aguirre 10.272.674
IS FALLING SHORT OF EXPECTATIONS
2
Table of Contents
Statement of originality ... 3 Foreword ... 3 Abstract ... 4 1 Introduction ... 4 2 Literature review ... 62.1 Electronic Word of Mouth (eWOM) ... 6
2.2 Online consumer reviews (OCR) ... 7
2.3 Sequel movies as brands ... 8
2.4 OCR valence related to movie performance ... 10
2.5 OCR valence related to sales for sequels ... 11
3 Methodology ... 13 3.1 Research design ... 13 3.2 Data collection ... 14 3.3 Sample ... 16 3.4 Measurements ... 17 4 Results ... 19 4.1 Descriptive statistics ... 19 4.2 Test of hypothesis 1 ... 20 4.3 Test of hypothesis 2 ... 21 4.4 Additional tests ... 21 5 Discussion ... 22 5.1 Summary of findings ... 22 5.2 Theoretical implications ... 23 5.3 Managerial implications ... 25
5.4 Limitations and suggestions for future research ... 25
6 Conclusions ... 27
3
Statement of originality
I, Pablo Aguirre (student number 10.272.674), declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of comple-tion of the work, not for the contents.
Foreword
This thesis was written for my Bachelor degree in Business studies at the University of Am-sterdam. While writing this thesis I received help from several people who I would like to thank here.
First of all, I thank my supervisor Dr. Frederik Situmeang for assisting me in this challenging process of writing my first bachelor thesis. His expertise provided me with unmatchable guid-ance for this thesis process.
Secondly, I would like to thank my fellow students of the Economics and Business program that have accompanied me over the past few years. Particularly, I would like to express my gratitude to both Christoph van Balen and Gwennis Keja, as well as to Sebastiaan Hersmis and Anniek Schepers.
Lastly, I would like to show my special gratitude to my mentors during the years in Amster-dam: Marcel Edwin van den Heuvel and Bas Robert Wooldrik. They have not only offered me innumerable resources, but also stood aside me as my second family here in Amsterdam. I hope that whoever may be reading this, he or she enjoys it as much as I did while writing it.
4
Abstract
The aim of this study was to gain more understanding of the (dis)satisfaction of consumers recorded online and the broader implication regarding the consumption goods. Therefore, online consumer reviews and box office sales data of 195 prequel and sequel movies from The-Numbers.com were collected. This study found that dissatisfaction, that is the propor-tional difference between subsequent prequel and sequel movies, have no significant impact on the repurchase of a subsequent product as measured in box office sales, even when control-ling for star power and production budget. Moreover this study found that evaluations of one product of the series are positive related to evaluations of the next product of the same series. Theoretical and managerial implications are discussed.
1 Introduction
“If curiosity killed the cat, it was satisfaction that brought it back” was a famous phrase of Holly Black, in the book published this year called Tithe. This phrase underlines a wide spread belief: satisfaction plays a major role in modifying behavior. This idea could originate on the fact that satisfaction is based on past experiences. Past experiences, in turn, have been shown to have a role among regular people in shaping product performance. However, it is unclear what the role of (dis)satisfaction of consumers is on modifying their behavior. There-fore, this study will focus on clarifying this role by analyzing the effect of word of mouth (WOM) on product performance, specifically investigating the impact of consumer (dis)satisfaction.
To know the consumers’ evaluations of an experience product, marketers can use con-sumers’ Word of mouth (WOM). WOM behavior “involves informal communication among consumers about products and services” (Liu, 2006, p.74). Before the Internet became widely available, WOM records ‘could disappear into thin air’ (Dellarocas, Zhang & Awad, 2007). With the advent of a world more internet-mediated, reviews posted by consumers became available to the public, including marketers and other consumers (Kostyra, Reiner, Natter & Klapper, 2015). This in turn, created expectations from the consumer side, that possibly need to be met by the firm in order to increase product performance of its next product.
The literature shows that consumers usually perceive WOM as a credible and trustwor-thy source of information (Banerjee, 1992; Brown and Reingen, 1987; Murray, 1991) and that it can influence general product performance (Liu, 2006). A widely used proxy for WOM is electronic WOM (Cheng & Xie, 2008). Furthermore, online consumer reviews (OCR) are the most common type of electronic WOM (Chatterje, 2001). Such reviews can influence
deci-5
sions about a variety of products and services, e.g. which movie to watch (Dellarocas, 2003). In particular, numerous studies found that valence of online consumer reviews (OCR) are related to product performance (i.e. Moldovan, Goldenberg & Chattopadhyay, 2011, Ludgwig et al., 2013; Lee, Park, & Han, 2008), and future product performance (Anderson & Sullivan, 1993).
Consumers’ valence of reviews has been expressed, in the literature of the motion pic-ture industry, as the expression of consumer (dis)satisfaction (i.e. Chintagunta, Gopinath, & Venkataraman, 2010). Consumer (dis)satisfaction can be defined as “the consumer's response to the evaluation of the perceived discrepancy between prior expectations and the actual per-formance of the product as perceived after its consumption” (Day, 1984 in Tse and Wilton, 1988). To analyze past and present expectations, this paper focuses on the prequels and se-quels of the movie industry. Particularly, this paper addresses sese-quels which are considered as brand extensions of movies. This means that consumer movie performance of a sequel movie could be influenced by the (dis)satisfaction of consumers of the corresponding prequel. With recorded OCR, people can express satisfaction or dissatisfaction through OCR valence. Therefore, this paper analyzes the moderation effects of consumer (dis)satisfaction on the relation between OCR valence and repurchase intentions in experience goods.
Anderson and Sullivan (1993) analyzed the effects of OCR on repurchase intentions. In contrast with the latter authors, this study tries to add a third sequel examination to test the findings of Anderson and Sullivan (1993), establishing a relationship between consumer (dis)satisfaction and repurchase intentions for consumption products. The following text is organized as follows: first, a literature review in which the concepts of WOM, (dis)satisfaction and product performance are developed, and two hypotheses are drawn; then, the methodology section describes the methods to answer the hypotheses; subsequently fol-lows the results section, the discussion of these results. The paper concludes with an overview of this study’s limitations and a final conclusion.
6
2 Literature review
In this section, a review of the current literature regarding the outlined concepts in the intro-duction is presented. Moreover, the hypotheses testing the theory of the research question are established. In terms of the literature, first, a summary of the concepts of WOM, electronic WOM, and online consumer reviews is given. Then, the concept of product performance and its relationship with OCR valence are explained, followed by the first hypothesis. Further-more, the concept of (dis)satisfaction related to the relationship with product performance is investigated, concluding with a corresponding hypothesis.
2.1 Electronic Word of Mouth (eWOM)
Word of mouth (WOM) "involves informal communication among consumers about products and services” (Liu, 2006, p. 74). Post-purchase product perceptions (Bone, 1995) as well as consumer choice have been shown to be significantly influenced by WOM (Katz & Lazarfeld, 1955; Engel et al., 1969; Arndt, 1967; Richins, 1983; Gruen, Osmonbekov, & Czaplewski, 2006). Mahajan, Muller and Kerin (1984) found ‘friends’ and ‘movie reviews’ as the main sources of awareness of a new movie using primary data. As a proxy for analyzing consumers’ informal communications, research has used inferences (i.e. Bass, 1969), surveys (i.e. Bow-man & Narayandas, 2001), and laboratory experiments (i.e. Borgida & Nisbett, 1977). How-ever, all these methods have their own specific errors: causality, self-reporting behavior, and generalizability, respectively (Dellarocas, Zhang, & Awad, 2007).
The advent of the Internet brought new information to test the WOM properties directly through user-generated online content (Dellarocas, Zhang, & Awad, 2007), i.e. market data. People often make offline decisions on the basis of online information (Lee, Park, & Han, 2008), such as which movie to watch (Dellarocas, 2003). As for other products, this phenom-enon was called electronic word-of-mouth (eWOM). Similarly to WOM, eWOM is found to be higher in credibility, empathy, and relevance to other customers as opposed to information coming from marketers (Bickart and Schindler, 2001).
7 2.2 Online consumer reviews (OCR)
Online consumer reviews (OCR) are different to other eWOM with regard to measurability, source, volume, and reachability (Chatterjee, 2001; Cheng & Xie, 2008). Nevertheless, Lee, Park, and Han (2008) argued that OCR are highly effective, can reach far beyond the local community through the Internet, are easy to observe, and the number of people who recom-mend a product can easily be counted. Chatterjee and Patrali (2001) contradict Lee, Park, and Han (2008) by stating that wide availability does not necessarily mean that consumers actual-ly access this information. Chatterjee and Patrali (2001) also posed that prior knowledge of their existence and conscious effort by the consumer are a condition sine-qanoon for its con-sumption. Despite its limitations, OCR is the most accessible and prevalent form of eWOM (Chatterje, 2001).
The vast existing literature on this topic has divided OCR into two groups: qualitative and quantitative OCR (i.e. Sridhar & Srinivasan, 2012). Qualitative OCR consists of a con-sumer review in written form (Kostyra et al., 2015). In those ‘written reviews’, the customer can assist other consumers, vent negative feelings about the product, express concerns about the product, exert positive self-enhancement, achieve social benefits, reach economic incen-tives, help the company, and seek for advice (Hennig-Thurau, Gwinner, Walsh & Gremler, 2004). Moreover, Situmeang, Leenders and Wijnberg (2014) found that for qualitative OCR, length of reviews positively relates to sales performance. Nonetheless, although the latter au-thors established a relationship with sales, the effects of consumer reviews develop through their volume and the generated offline WOM (Liu, 2006; Gu & Whinston, 2008; Karniouchina, 2011). This last finding is in line with quantitative OCR, where the reviewer is forced to summarize the evaluation in a single rating or grade. The single ratings from cus-tomers are commonly pooled together into a summary statistic (Lee, Park and Han, 2008). Thus, quantitative OCR might make it simpler and faster for the customer to accurately pro-cess large amount of information without the need to read each review and analyze its content (Lee, Park and Han, 2008). In this paper, it is the latter measure – quantitative reviews – that is used for measuring OCR.
Nonetheless, another important factor that cannot be addressed with the summary of sta-tistics for OCR is trust, which is based on how the consumer relates to the other consumer writing reviews (Chatterjee and Patrali, 2001). The same interaction in an online platform can have different consequences for consumers who trust compared to those who do not trust each other (Friedman, Khan & Howe, 2000). Yet, a way to overcome this misbalance in the
conse-8
quences related to trust is proposed by Senecal and Nantel (2004), who argue that customers’ trust in the recommender system is increased over time as customer satisfaction increases with previously recommended products by that system. In other words, if a consumer uses one review system repeatedly then this could possibly indicate the increased trust he/she has with respect to that system.
Chintagunta, Gopinath, and Venkataraman (2010) decomposes a quantitative OCR into valence, volume and variance. These metrics have been widely used in the experience goods literature for analyzing the OCR phenomena (i.e. Dellarocas & Narayan, 2006; Dellarocas, Zhang & Awad, 2007; Moe & Trusov, 2011). According to Chintagunta, Gopinath, and Venkataraman (2010), valence is the average rating in a single cipher, and it “represents aver-age customer satisfaction” (p.6). Volume is defined as the total number of ratings (Chintagunta, Gopinath, & Venkataraman, 2010).
Finally, variance is the variation of the number of customer ratings for each valence level (Chintagunta, Gopinath, & Venkataraman, 2010). Kostyra et al. (2015) add that variance represents the degree of disagreement or heterogeneity among customers’ evaluations. Vari-ance has been shown to have different levels of influence over performVari-ance (if at all), and therefore, they are not part of this analysis. This is consistent with current studies in which variance of OCR related to product performance was not analyzed (see Amblee and Bui, 2011; Chen Wan, & Xie, 2011; Chevalier and Mayzlin, 2006; Cui, Lui, & Guo., 2012; Dellarocas et al., 2007; Dhar and Chang, 2009; Duan & Whinston, 2008; Ho-Dac Carson & Moore, 2013).
To conclude, OCR are a valid proxy for eWOM as they help to easily process reviews information and have a more extensive reachability compared to other forms of eWOM. Im-portant aspects of OCR are its valence, representing consumer satisfaction evaluation, and its volume, representing the total number of ratings.
2.3 Sequel movies as brands
Experience goods are products whose quality cannot be determined by mere inspection. In-stead, consumers have to actually consume the product in order to evaluate it (Wernerfelt, 1988). Experience goods, such as movies, are an example for the consumer’s inability to as-sess the good before consumption. It is not possible to asas-sess the value of a movie until the movie has been watched (Chang & Ki, 2005). However, if compared to other common con-sumption goods such as music, movies seem to be a rather unique case. That is, they have a
9
higher cost of consumption compared to other goods like music (Adermon & Liang, 2010), since it is more time-consuming (Caresa, 2015). Therefore, it follows that since a movie is a higher level ‘experience good’, consumers are exposed to higher risks when deciding to con-sume a movie compared to music.
The consumption of movies demands a higher time risk from the consumers. A way to reduce consumer risk for the producer is to apply the ‘signaling theory’ (Connelly, Certo, Ire-land & Reutzel, 2011). It proposes that when two parties have asymmetric information, one side can choose to send signals (or communicate) to the other side (Connelly, Certo, Ireland & Reutzel, 2011). For the signaling side it is often costly to communicate these signals to the receiver (Spence, 1973). In this context, signaling explores a rational consumer who expects a firm to honor the implicit commitment transmitted through a signal because not honoring the commitment is economically unwise (Kirmani & Rao, 2000). The signaling approach consid-ers the firm’s stimulus and it is most beneficial for products whose quality is unknown before purchase (i.e. experience goods). For instance, signals from the supply side such as star power and movie budget have been positively related to movie performance (Basuroy, Chatterjee, & Ravid 2003; Elberse & Eliashberg 2003; Litman 1983).
Besides signals with information about the brand originating from the supply side (i.e. producer), the production of a sequel in itself could be a signal. According to brand theories, sequels can be conceptualized as a brand extension (Sood & Drèze, 2006; Chang & Ki, 2005). A known family brand is often the most important factor for predicting the trial of a new product (Claycamp and Liddy, 1969) because the known brand name is an important factor for reducing consumer risk (Milewicz and Herbig, 1994). Hence, if a movie succeeds, the studio producing the movie could try to duplicate the formula as closely as possible (Ravid, 1999) by plotting the same characters evolving in a new situation -a sequel movie (Sood & Drèze, 2006; Moon et al., 2010). This is consistent with the signaling role of brand names that is illustrated in the context of umbrella branding (Wernerfelt, 1988). Umbrella branding oc-curs when a firm uses an established brand name in promoting a new good (Wernerfelt, 1988). The new product is inviting consumers to join their experience with the two products allowing the consumer to infer the quality of both (Wernerfelt, 1988). The results, however, are contra-dictory regarding the success of this strategy. If success is unpredictable, sequels performance will be no better than any other movie performance (Ravid 1999; Basuroy et al., 2006). How-ever, Dhar, Sun, and Weinberg (2012) demonstrated that a movie sequel on its own is a posi-tive determinant of revenue. Additionally, Saburoy and Chatterjee (2006) and Brewer et al.
10
(2009) showed that sequels perform better than non-sequels in monetary values. The problem of these studies lies in the fact that the originating prequel of the sequel are also shown to per-form better than non-sequel movies (Dhar, Sun, & Weinberg, 2012). This might bias the re-sult in favor of sequel movies.
To summarize, the quality of experience goods cannot be assessed by the consumers be-fore its consumption and movies have higher cost of consumption compared to other experi-ence goods. Rational consumers, in turn, can reduce their consumption risk by looking for certain cues emitted by the producer before acquiring those experience goods, such as star power and movie budget. However, the literature is contradictory to whether a sequel is a pos-itive determinant of revenue or the higher revenue is due to the fact that a sequel movie suc-cess is tightly related to its corresponding prequel movie.
2.4 OCR valence related to movie performance
Consumers can look at quality signals emitted by the supply side by considering the costly signals from the producer regarding to a movie (i.e. budget and stars). Consumers can also look at signals emitted by other consumers. For example, if a movie is appreciated by its customers, the release of a sequel movie can create expectations. Situmeang et al., (2013) wrote: [movie prequels] “wakes awareness, excitement and anticipation, which will help sub-sequent sales”. This in turn, could imply that the signal from other consumers (i.e. OCR ) could have an impact on future economical performance.
Sales could be seen as a straightforward way to quantify economical performance of a movie and might be of a high interest for producers who are profit seekers. On the other hand, previous research in the motion picture industry, regarding product performance, has investi-gated the customers who attend the cinema (i.e. Austin, 1984). When analyzing real market data, however, it is difficult to assess an accurate number of people who watched a movie, since there might be people not using money in exchange for watching a movie. This is espe-cially the case with pirated movies, which typically do not involve the use of money. In addi-tion, this phenomenon also occurs in the case of the motion pictures reproduced in other me-diums, where the money paid for the movie is different across different platforms (i.e. DVD vs. Videos). Nonetheless, performance measures containing monetary values are prevalent in the studies of OCR, such as sales (i.e. Moe & Trusov, 2011), box office (i.e. Chintagunta,
11
Gopinath, & Venkataraman, 2010), and revenues (i.e Kindem, 1982). In this study, sales are used as a proxy for performance for its practical implications.
The positive effect of valence on product performance on the cultural industry has been reported as significant for most of the studies (Chevalier & Mayzlin, 2006; Chintagunta et al., 2010; Clemons, Gao, & Hitt, 2006; Cui et al., 2012; Dellarocas, Zhang, & Awad, 2007; Dhar & Chang, 2009; Sun, 2012). However, some scholars have not found a significant effect when using individual movies OCR data (Amblee & Bui, 2011; Duan et al., 2008). Kostyra, Reiner, Natter and Klapper (2015) state that a limitation of previous studies is that products with a low valence have generally less chances of being reviewed frequently and therefore, do not reach statistical significance.
In summary it can be said that generally a movie relates to the next movie because it generates expectation and thereby could have an influence on movie performance. Further-more, sales is a feasible metric to measure movie performance. Finally, OCR valence has been positively related to product performance. Therefore, the following hypothesis is pro-posed:
Hypothesis 1: The higher (lower) the valence of a prequel movie, the higher (lower) the sales will be of its corresponding sequel movie, even when controlling for appearance of stars and movie budget
2.5 OCR valence related to sales for sequels
Consumer’s satisfaction has been defined as “the consumer’s response to the evaluation of the perceived discrepancy between prior expectations (or some norm of performance) and the actual performance of the product as perceived after its consumption” (Tse and Wilton, 1988, p. 204). Consumers satisfaction regarding to movies means the consumers expectation the quality of a prequel to be similar to the quality of a sequel.
Although the direct connection between prequels and sequels has not been investigated with the use of market data in the motion picture industry, the literature connects the OCR of prequels with sequels related to performance. For instance, Duan, Gu, and Whinston (2008), indicate that a unique aspect of the WOM effect is the presence of a positive feedback mecha-nism between WOM and sales. Situmeang, Leenders and Wijnberg (2014) also highlight a
12
latent relationship between sequels OCR valence and past editions in video games. Moreover, people relate sequels with their predecessors. Before watching a movie Anderson and Sullivan (1993) found that an evaluation (i.e. OCR valence) falling short of expectations has a greater impact on satisfaction and repurchase intentions (i.e. paying for watching the next sequel) than an evaluation that exceeds expectations.
Besides the established connection between prequels and sequels related to performance, Chen, Fay & Wang state that “WOM can be used to express satisfaction or dissatisfaction” (2011, p.86). For example, consumers engage in negative word-of-mouth in order to vent hos-tility (Jung 1959) and to seek revenge (Allport and Postman 1947; Richins 1984). Anderson (1998) finds that extremely satisfied and extremely dissatisfied customers are more likely to initiate information flows (i.e. OCR) than consumers with more moderate experiences (Hennig-Thurau, Gwinner, Walsh & Gremler, 2004). However in this case, the literature does not distinguish between dissatisfaction with the product and the dissatisfaction with the norm of evaluation of the product in itself. Meaning that it is not clear whether the consumers are (dis)satisfied with the prequel movie in relation to the sequel movie regardless of the prequels’ rating or the consumers are (dis)satisfied with the prequels rating in relation to the sequel movie.
To conclude, even though the evaluation of a product could contain other underlining factors, the dissatisfaction in evaluation, as in the difference between evaluations, is a factor to be analyzed since most of the sales can be explained by the valence of OCR.
Hypothesis 2: The higher the negative (positive) proportional difference between the OCR valence of the prequel and the OCR valence of the second sequel, the lower (higher) the in-teraction effect between valence of the second sequel and box office sales of the third sequel.
13
3 Methodology
In the previous section, the theoretical concepts and the operationalization of the concepts that guide the explanatory research have been outlined. The research design and methodology sec-tions are oriented at the research question, that is, what is the effect of recorded consumer (dis)satisfaction on future product performance for experience products. As mentioned before, the aim of the study is to advance the existing literature, as well as to give managerial recom-mendations. In order to answer the research question, this chapter presents the measures to test the two aforementioned hypotheses. First, the overall research design, which provides the framework for the hypotheses will be tested are discussed. Subsequently follows a discussion of the sample, the data collection process, and finally, the measures.
3.1 Research design
The research question reflects the positivist nature of this thesis, in which large amount of data from the cultural industry is used to test theory (deduction). Furthermore, a case study was chosen as an appropriate research strategy. By doing so the phenomena can be explored in a real-life context (Saunders et al., 2012). Real market data analysis has the advantage that the data collected is considered part of reality (Saunders et al., 2012). Moreover, the presence of endogeneity issues is mitigated by the analysis of the control variables. Further, although experiment data may have the advantage of high internal validity (since the variables to be researched can be isolated and highly controlled) (Saunders et al., 2012), the real market data used in this study might provide an unbiased image of the effects of OCR valence on Box Office sales, and are therefore preferred for this study. However, the amount of data to be collected is limited by its availability. Nonetheless, the fact that previous studies analyze OCR using real market data makes it more feasible to reach more robust conclusions and allow for a deeper understanding of the phenomenon at hand.
This research considers all the movies indexed by The-Numbers.com (www.the-numbers.com) that have had at least three sequels. This assures that the movies are the sequels of the whole population. However, to address the fact that not all experience goods are se-quels, and therefore do not behave the same way, other control variables specific of the mo-tion picture industry are used. This is in line with variables that have been shown in the past to have an effect on performance. Such variables are star power (‘StarPower’) and production
14
budget (‘ProductionBudget’). Both of them have been found to have an influence on movie performance (Basuroy, Chatterjee, & Ravid, 2003; Elberse & Eliashberg, 2003; Litman, 1983). Lastly, studies using data from the website The-numbers.com have shown to be in line with previously drawn theory.
Because of the limited time and money, this research is cross-sectional. Saunders et al. (2012) argue that a cross-sectional research design can save money and time. Smith and Smith (1986) warn about the difficulty of analyzing a cyclical (cultural) industry with a cross-sectional study, e.g. there is a higher number of OCR when the movie is just launched. Mov-ies are unique in that they have product life cycles that are both short and steered by predicta-ble, exponential patterns (Moe & Fader 2001; Sawhney & Eliashberg 1996). Moe and Trusov (2011) propose that results can be sensitive to the moment in the product life cycle when the researcher collects the data. For that reason, a cross-sectional analysis could bias the result for analyzing the industry. To reduce such bias, this research incorporates a unique sampling strategy that is discussed in the next section.
3.2 Data collection
Movies OCR valence and volume have been analyzed by regional roll outs (Chintagunta, Gopinath, & Venkataraman, 2010), by aggregated country data (i.e. Moe and Trusov, 2011), and by all regions using the English language (Chevalier & Mayzlin, 2006). Because of the widely available data in the English language and extensive literature in the topic, most of the research that has been done in this field was for the United States Motion Picture industry (Lei-cowa, 2012). Moreover, according to Redfern (2013), Hollywood is the largest market in terms of box office and revenues. Thus, this research is also based on the United States Mo-tion Picture industry.
According to Saunders et al. (2012), three major issues of data quality have to be con-sidered before collecting the data: reliability, generalizability and validity. These are dis-cussed in the following paragraphs. Reliability is of particular concern for studies with large amount of data, since it is easier to find significant effects. This issue is addressed by using public-available information, by building up on existent data collection methods, and by de-signing a detailed methodology section.
Next to reliability, the next issue to address for assuring data quality is generalizability. This means that this research aims to establish that the effects of word of mouth on sales is
15
moderated by (dis)satisfaction in the cultural industries and that this phenomenon does not only exist in the motion picture industry for sequels or in the particular sample used, but can also be used for other experience goods. Information about the OCR was previously collected from Numbers.com and was originally generated by Rotten-Tomatoes.com. The-Numbers.com claim to have no affiliation with Rotten-Tomatoes.com. However, a mere statement of that does not prove that there is no affiliation with the aforementioned third party. Although we cannot assess the reliability of the data in this respect, we can increase the sam-ple size by using multisam-ple websites, and weighing them according to the amount of reviews. In general, increasing the sample size and the use of multiple sources is a used way to reduce bias reliability (Saunders et al., 2012).
Finally, the last issue of data quality to be addressed is validity. To be more precise, there is an explanation of each variable and its relation to each other. Firstly, the data for the dependent as well as independent variable (OCR valence and volume) were collected from the website The-Numbers.com. It contains detailed information about the Box Office sales per-formance, as well as a summary of the OCR of past movies. The former aspect originates from sales data presented by BoxOfficeMojo (http://www.boxofficemojo.com/) while the latter aspect (OCR information) is taken from Rotten-Tomatoes.com. There are four main advantages of using this kind of secondary data: saving large amount of resources (Ghauri and Gronhau, 2010), it is considered an unobtrusive method because the data has already been collected and thus cannot be influenced (Cowton, 1998), it is often of the higher-quality than the one that oneself could collect (Smith, 2006), and finally, it is widely available for testing. The main disadvantages of using secondary data is that there is no real control over data quality and that the data might have been collected for other purposes, thus increasing bias (Saunders et al., 2012). As mentioned before, the increase of the sample size is used.
Moreover, Chevalier and Mayzlin (2006) have found that reviews on one website can be higher than on another website. The websites from where the OCR valence and OCR vol-ume is collected are consistent with the websites accessed by van der Elst (2012): the Internet Movie Database (IMDB.com), Rotten-Tomatoes.com, and Yahoo! Movies (mov-ies.yahoo.com). They are in the English language and, according to van der Elst (2012), are the most accessed websites for OCR.
To analyze the moderator variable – (dis)satisfaction – this research contrasts OCR Va-lence (OCRVaVa-lence) of the first movie with OCRVaVa-lence of the second movie, and compares it with the Box Office sales of the third movie. For that reason only three-sequel movies are
16
selected, including in the analysis only the population of sequel movies that have been re-leased in the past 15 years with at least two sequels. The number of years is consistent with the point of time in which Internet started to become more widely available, as proposed by Dellarocas, Zhang and Awad (2007). In this study, the selected sequel movies were acquired from an online available list generated by online users IMDb. The list contains the names of the sequel movies that have been reviewed by thousands of people during the past 15 years. That list was triangulated with sequels proposed by a know site with over a million monthly visits known in the industry: Movie Insider (http://www.movieinsider.com/). This triangula-tion showed no major differences between the two lists.
Lastly, in order to collect the information regarding the control variables star power and production budget, this research also used compiled data from the webpage The-Numbers.com.
3.3 Sample
The total of prequels and sequels movies released over the past 15 years was 596, however, only 432 of those sequels have at least a third sequel. Because of incomplete data, this study bases its analysis on 195 sequels. The use of non-probability sample of sequels could jeopard-ize the generalization to the whole population of experience goods. In order to avoid this problem, this study strived to match the distribution of age rating and movie genre with the population of movies. However, it was not possible to match the distribution because by do-ing that, the overall data sample to be analyzed could shrink considerably, creatdo-ing the prob-lem of no statistical significance.
Information regarding the BoxOffice sales performance, ProductionBudget, and star power data was collected from the The-Numbers.com retrieved on the 22nd of June 2015. The values with respect to OCR valence and OCR volume were retrieved on the 29th of May of 2015 from IMDb, Rotten-Tomatoes.com, and Yahoo! Movies. Moreover, there was a positive weight towards the reviews posted on IMDb, in comparison with the other two sites, driving the OCRValence towards the values of OCR valence collected from the IMDb website. How-ever, the values corresponding to the valence of a movie for the three sites were not signifi-cantly different (average OCR Valence: IMDb=59.60, Rotten Tomatoes=60.46 , Yahoo! Movies=60.38), meaning that the possible bias might not be large.
17 3.4 Measurements
The dependent variable, OCR valence, is the average rating in a single cipher. This is con-sistent with the study of Chintagunta, Gopinath, and Venkataraman (2010) among others. However, this research takes the OCR valence from each site and weighs it with its corre-sponding OCR volume. Previous studies have focused their sample on relating the various OCR valence values of one movie to product performance. In contrast with those studies, this research takes all the OCR valence values of one movie together as a singular value of OCR valence for that one movie. Consequently, it considers the OCR valence ratings of the movie as a whole, and thereby, reduces the OCR valence of the movie to a single convergent digit. For this statement to hold true, the marginal valence of a review should tend to 0. This is mostly true for products with several reviews, where one extra review is not likely to alter the single digit OCR valence of the movie as a whole. This research also incorporates the ratings of other main websites, weighing the OCR valence of each rating according to its correspond-ing OCR volume. This new variable is then called OCRValence
To measure (dis)satisfaction, this research follows the method of Churchill and Suprenant (1982) who state that “operationally, satisfaction is similar to attitude in that it can be assessed as the sum of the satisfactions with various attributes of the product or service” (p.493). Moreover, to test the moderation effect of (dis)satisfaction on the basis of sequels, the only necessary condition to be met is that there are sequels with higher and lower valence than the previous sequel. Thus, a simple subtraction of the OCR valence of the first (second) sequel by the OCR valence of the second (third) sequel and divided by the valence of the first (second) sequel was performed and named S/DS.
To create the moderation variable, this study creates a ‘central mean’ variable for the (dis)satisfaction variable and also for the OCRValence of movie 2 variable. It then multiplies both together. Next, the new variable S/DS was included in the regression to the Box Office 3 value.
Moreover, the cultural industry has its own characteristics, and therefore the authors proposed to check for those peculiarities in the data (i.e. Basuroy & Chatterjee, 2008). This is in line with a conventional quantitative research analysis, as Saunders et al. (2012) propose to
18
incorporate controls to ensure validity of the data. As suggested in the literature, this study controls for star power and production budget.
Star power is related to movie revenues and it is divided in peers, experts and consum-ers (Ravid, 1999). However, Limsuval (2015) suggests that when regressing OCR measures, consumers rating are not needed to measure consumer star power, and peers and experts awards are enough to measure star power. In line with Lumsuval’s (2015) findings, this paper uses Academy Awards (as a proxy for peers) and Golden Globes (as a proxy for experts) be-cause both have a significant effect on movie revenues when using OCR measures. This is also in line with the work of Elberse (2007), who demonstrated a direct relationship between movies financial performance and actors’ and actresses’ recognition through awards. There-fore, this research uses Limsuval’s (2015) findings to control for appearance of stars of peers and experts.
With respect to star power, this research considered the nominations in the Oscar and Golden Globes in the categories of best actor/actress in a leading role, best actor/actress in a supporting role, and best director. The occurrence of award winners of each movie was count-ed, added together, and each nomination of actors or actresses was included in the analysis as a simple number. For instance, if the movie had two actors that have been nominated and one actor or actress who won an award, the movie would receive a value of two for nomination, and a value of one for winning award, resulting in a total value of three.
For collecting data regarding the movie budget, the corresponding information taken from The-numbers.com matched with other sites that report movie budgets, and are preveni-ent from the studios themselves. However, in the case of some sites, the analysis depended on the used currency, therefore, all the numbers were converted to dollars. Important to note is that the collected data about movie budgets do not include information about marketing ex-penditures after the roll out.
19
4 Results
In the first part of this chapter the descriptive statistics resulting from the analysis are shown. Thereafter, a description of the multiple regression analyses used to establish the relationship between box office sales, OCRValence, and the control variables of appearance of star and budget.
4.1 Descriptive statistics
Correlation between Box Office sales of movies 1 and 2, and OCR measures and Movie revenue determinants
Mean SD 1 2 3 4 5
1 Box Office of Movie 3 295918426.11 337406527.65
2 OCRValence of Movie2 60.45 12.78 .424**
3 OCRValence of Movie3 58.99 13.54 .514** .749**
4 Proportional dif. Movie 1 & 2 -0.06 0.19 .177* .470** .262** 5 Production Budget of Movie 3 76512598.43 70314934.39 .786** .403** .412** .144
6 Star Power of Movie 3 3.31 5.52 -.039 .150 .020 .016 -.041
**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).
The table outlined at the begining of this section presents the correlation matrix, which shows the means, standard deviations, and correlations of the variables to be analyzed according to the theoretical framework. In total this sample consists of 195 data observation points for the BoxOffice sales variable, 127 for production budget (‘ProductionBudget’) of movie three, 332 and 316 for the OCRValence of movie two and movie three respectively, 103 for the appear-ance of stars (‘StarPower‘), and lastly 320 data points for the proportional difference between movie two and movie three. In the same correlation matrix below, the values of the Person correlations are shown. They can be used as a preliminary evidence of a linear relation be-tween variables. A correlation higher than .5 means a high positive relation, whereas a corre-lation between .5 and .2 means a moderate positive correcorre-lation, and values between .2 and 0 imply a weak positive correlation.
Based on the presented data, there is a strong positive relation between OCRValence of movie 2 (’OCRValence2’) and Box Office of movie 3 (‘BoxOffice3’) (r(195)=.514, p.<0.01). This finding offers preliminary support to hypothesis 1, that establishes a relation between the va-lence ratings of a prequel, with its immediate sequel box office performance. This is in line
20
with the previously discussed theory in which two sequels were analyzed. In addition, as ex-pected, production budget of movie 3 (‘ProductionBudget3’) is strongly positive related to the ‘BoxOffice3’ (r(112)=.786, p.<0.01). Unexpectedly, Star power of movie 3 (‘StarPower3’) is not significantly related to ‘BoxOffice3’ of its next sequel, as measured by the nomination and winning awards given by peers and experts. In fact it was on the negative side, (r(99)=-.039, ns). This might be due to the small sample taken regarding to the stars, not allowing to reach significant effects (n=103).
Moreover, there is a positive correlation between the proportional difference of movie two and movie three with respect to ‘BoxOffice3’ (r(195)=.177, p.<0.05) given a preliminary support to hypothesis 2. ‘OCRValence2’ is also significantly positively related to the fore-mentioned proportional difference of OCRValence values (r(320)=.470, p.<0.01).
Also unexpectedly, the proportional difference is also related to OCRValence3 (r(316)=.262, p<0.01). Moreover, OCRValence2 and OCRValence3 are highly positively related (r(316)=.749). This last finding could be use as preliminary evidence for the that the release of subsequent products influence each-other through the valence. This is in line with previous theory that proposes that success of one sequel is related to success of the next. This finding needs further exploration.
4.2 Test of hypothesis 1
After performing a Pearson’s correlation test, a linear hierarchical regression between ‘OCRValence2’ and ‘BoxOffice3’ was performed. The objective was to investigate the ability of ‘OCRValence of the previous sequel to predict Box Office sales performance of the next sequel movie (hypothesis 1).
In line with current research (i.e. Trusov, 2011), this study modeled the relationship be-tween product performance and ratings by linearly regressing product performance against measures of ratings. Nonetheless, before performing a linear regression, this relation is con-trolled for non-linearity, as urged by Chevalier and Mayzlin (2006) who suggested that the relation between valence of OCR and product performance might not be linear, and that an incremental negative review is more powerful in decreasing books sales than a positive one.
First this study controlled for Star power and Production budget, this model was statisti-cally significant F(2,70)=37.981 p<0.000, and explained 52.0% of the variance of Box Office
21
performance in movie three. Only ProductionBudget had a significant effect on Box office, β=3.714 (sd=.426, p<0.000), and Star Power had a β=1655770.7 (ns). This means that an in-crease of one percent of the movie budget could magnify the Box Office performance by al-most four times while accounting for Star power. Although statistically not significant, the last result regarding Star Power was expected due to the low correlation of Star Power of movie 3 with Box Office of the same movie in a series, albeit being in the positive side. After accounting for valence of movie 2, as proposed in hypothesis 1, the total variance ex-plained by the model as a whole increased to only 52.1% F(3,69)=25.036 (p<0.000). The model increased its explanatory power by 0.1%, F(1,69)=.111 (ns). To conclude, hypothesis 1 is rejected, in support of the simpler model. The higher (lower) valence of a sequel does not significantly increase the box office sales on the next sequel movie when accounting for StarPower and ProductionBudget of the movie of the series.
4.3 Test of hypothesis 2
Regarding hypothesis 2, another hierarchical multiple regression was performed to in-vestigate the ability of perceived (dis)satisfaction to predict box office performance, after con-trolling for appearance of stars and production budget. In the first step of the hierarchical, multiple regression, two predictors were entered: ProductionBudget, and star power. This model was statistically significant F(2,35)=39.753, p<.0001 and explained 69.4% of variance in box office sales. After entry of the (dis)satisfaction variable and valence of the second movie variable in Step 2, the total variance explained by the model as a whole was 69.6% F(33,4) =18.868; p < .0001. In step 3, the moderator variable was introduced. The third model explained 69.7% of the variance in F(5,32)=14.755 (p<0.000). However, contrary to the pre-dictions, the introduction of the (dis)satisfaction did not increase the explained the variance in box office sales (r2 change=.002, F(1,32)=.180 (ns). Therefore, hypothesis 2 is also rejected.
4.4 Additional tests
In light of the inconclusive results, it was decided to perform another linear regression in order to test whether the OCRValence of the third movie was related to the box office per-formance when controlled for movie budget and star power. For that reason, another regres-sion to control whether the OCRValence of the second movie has an effect on the
22
OCRValence of the third movie, since there was an unusually high correlation between the two. Moreover, as mentioned in the theoretical framework, there might be a relation between the prequel and the sequel. However, this link could be related to the valence of the next mov-ie, and not to the box office as hypothesized.
To test the relationship between OCRValence3 and BoxOffice3, a linear regression was performed. However, no significant effects were found. Therefore, we turned to test the rela-tionship between OCRValence2 and OCRValence3. To test this relarela-tionship between valenc-es of different products of the same serivalenc-es, a hierarchical linear regrvalenc-ession was performed be-tween Box office of movie OCRValence3 and OCRValence2, StarPower and ProductionBudget. When testing for a direct effect of StarPower and ProductionBudget both of movie 3, on the ‘OCRValence3’, the model was significant F(2,72)=3.419 (p=.38), and predicted 8.7% of the change in the OCRValence3. Production budget of movie 3 was an al-most significant driver of the change β=0.0000 (t=2.592, p=0.012), whereas Star Power was not significant β=0.08 (t=.448, ns).
When adding the variable OCRValence2, the model significantly improved the predic-tion power (F(2,71)=10.255, p=0.002). to 20.3% F(3,71)=5.99 (p<0.000). However, only OCRValence2 had an effect on OCRValence3, β=0.352 (t=3.202, sig=0.002), Production budget movie 3 β=0.000 (ns), and Star Power β=-0.51 (ns).
5 Discussion
In this section the results from the previous sections are interpreted and linked back to the literature. In the first part, the effects word of mouth (WOM) of consumption goods are dis-cussed in relation to product performance and disdis-cussed (section 5.1). In the following section its relation to the current theory (section 5.2) is described, followed by the discussion of the effects of consumer (dis)satisfaction on product performance. In the subsequent section (sec-tion 5.3), managerial implica(sec-tions are discussed. Lastly, the limita(sec-tions and how future studies can extend this research are outlined (section 5.4).
5.1 Summary of findings
There is an increasing importance of the word of mouth phenomena in a more Internet medi-ated world. However, little has been researched on the mediating effect of consumer
23
(dis)satisfaction on repurchase behavior for consumption goods. There were two big gaps in the literature that needed to be addressed. One was the fact that researchers have provided inconsistent results regarding the impact of OCR reviews in relation to the performance of a product in the movie industry, particularly the impact of WOM on repurchase intentions. The second is the mediating role of (dis)satisfaction. This paper tried to assess the effect of word of mouth (WOM) on product performance, specifically investigating the impact of consumer (dis)satisfaction on repurchase behavior. Among other forms of electronic WOM, online con-sumer reviews (OCR) of movies were chosen to assess its impact on future box office sales of these consumption goods. Moreover, this study analyzed OCR valence of different movies of the same series in relation with its box office sales at an aggregated level. The (dis)satisfaction of consumers was a proportional difference between the evaluation of previ-ous products of the same series.
This study predicted for high WOM behavior to result in high repurchase behavior. This was not the case in this sample, meaning that there are no conclusive results indicating that online consumer reviews with a high valence result in higher purchase behavior of differ-ent product of the same series. This was contradicting the predictions made in the theoretical framework.
The inclusion of the role of online consumer reviews was of great importance, since vast literature in the consumption goods found a high correlation of WOM and current pur-chase intention. It was found that a high online consumer review of one movie would be high-ly related to the previous movie of the same series.
The experience with only one product before of the same series might have not been enough. Therefore the effect of a mismatch in the past experiences of two products of the same series - (dis)satisfaction was investigated. The results were not in line with the second hypothesis. In this particular sample, (dis)satisfaction was found to be not significant as a moderator to the relationship between WOM and product performance.
5.2 Theoretical implications
No significant support was found for the first hypothesis, suggesting that WOM has no signif-icant positive effect on product performance. This study tested the consumers’ evaluations as assessment of quality (i.e. OCR valence) in relation to the customers’ repurchase intentions. The first hypothesis was build on the foundation of signaling theory considering the signals
24
from the supply side (i.e. production budget and star power), together with expectations from the consumer side (i.e. OCR valence). Signals from the supply side were found to have a sig-nificant effect on box office sales of the next movie in this study, strengthening the lines of support for signaling theory. However, the effects of the producer signals outweighed the ef-fects emitted by the consumers toward the box office sales of the next movie.
A deeper insight into this phenomena revealed that the expectation of the consumers transmitted through OCR might have affected the OCR of the subsequent movie, and not its box office sales. These findings give support to brand theories, which propose that sequels can be conceptualized as a brand extension (Sood & Drèze, 2006; Chang & Ki, 2005). These contradicting findings could also be due to the fact that OCR might not be related to box of-fice sales at all.
An alternative explanation could be that movies OCR act as an independent indicator for the people accessing the information posted online, and having no more reach than the online readers. This alternative theory suggests that the mass accessing to online information is not sufficiently large to create network externalities. This was suggested by Chatterjee and Patrali (2001) who posed that prior knowledge of the online community existence and con-scious effort by the consumer are a pre-condition for the consumption of the product. Moreo-ver, this could contradict the theory of Ravid (1999) and Basuroy et al. (2006) who state that success in unpredictable and success of one movie is independent from the previous movie. Further investigations on the amount of network externalities are needed to shed light on this issue.
Thirdly, no support for the second hypotheses was found, implying that (dis)satisfaction might not moderate the relationship between WOM and product performance. One explana-tion could be the sample size. This study tested the moderaexplana-tion effects of consumer (dis)satisfaction on repurchase behavior, therefore only movies with at least three sequels were selected from all the movies of the past 15 years in the United States motion movie in-dustry. Unfortunately, the lack of enough available data made it difficult to find significant results. People with access to a bigger data base could gain a deeper inside in this issue. How-ever, another plausible explanation might be that a moderation effect of (dis)satisfaction does not exist, meaning that consumers do not actually take rational decisions. A view that current-ly is gaining support is the use of heuristics to predict consumer behavior (see Tversky & Kahneman, 1974).
25
To conclude, the current study adds several new ideas to the field of online consumer reviews and electronic word of mouth. It is, to our knowledge, the first study that provides an analysis of movies as experience goods using a third sequel. The present research also seems to be the first one that elaborates further on the (dis)confirmation of consumers related to the performance of experience goods such as movies. Furthermore, this study suggests a link be-tween consumer word of mouth of two products of the same series.
5.3 Managerial implications
Besides the theoretical implications of this research about WOM, (dis)satisfaction, and product performance, managers could also profit from this research. This study demonstrated that information regarding the evaluation of experience goods could be informative about the following evaluations of another product in the same line. This study shows that although consumers consider evaluation of the previous product when making purchases, a mismatch between those evaluations has not been shown to have explanatory power as a driver for not purchasing products. Moreover, signals coming from the supply side can offset the signals from the consumers. If the reviews of a product were not high enough, managers could con-sider increasing the expenditures on other activities such as increasing the production budget of the product.
Managers who would like to increase the sales of consumption goods should not be constrained by previous product performance (in terms of sales). However, managers who seek to have positive reviews on their products should take into account that the evaluation of existing products of a series can be influential for the evaluation of a new product in that se-ries. If a product had very low reviews, those managers could initiate another line of product that is not associated with the original product.
5.4 Limitations and suggestions for future research
In this section some limitations of this research are described which should be kept in mind before applying the results to real market situations. Moreover, some light is shed on potential avenues for further research.
Several limitations were encountered for the present study. First, a limitation to take in-to account is the sample. Although two different lists containing sequels names were used for
26
this study, they were both used to generate convenience samples. For lack of data points, this research could not match the distribution of genres and ratings as intended. Thus, as with most online studies, due to the possible self-selection bias it is not possible to confirm that the se-lected movies are representative of the population of all movies, and therefore, for all experi-ence goods. Future studies could try to match the distribution of genre and rating, to get more conclusive results.
Second, this research was based primarily on our analysis of the demand side of the mo-tion picture industry and was limited in integrating the supply side to a similar extent. Only signals from the supply side, such as production budget and star power, were taken into ac-count. Since supply side factors influence movie profits as much as demand side factors, fur-ther research could take a more balanced approach towards demand and supply aspects of the industry. For instance, analyzing trailers as determinants for deciding on watching a movie or not (Smeaton et. al, 2006). In the same line of thought, Feng and Papatla (2011) studied how increased advertising can be associated with reduction in online WOM. The increase of adver-tising might have consequences on the valence and volume of the recorded OCR for certain product. More studies on this area could be needed to advance in the understanding of the behavior of OCRs and its relationship with advertising.
Third, qualitative ratings were used, yet their content was not analyzed. The authors (Chevalier & Mayzlin, 2006; Wyatt & Badger, 1990; Situmeang, Leenders, & Wijnberg, 2014) warned about the use of quantitative reviews in the analysis of OCR. Future research could use a combined method of quantitative and qualitative reviews to account for an possi-ble overlapping impact of both quantitative and qualitative reviews.
Fourth, individual viewers’ post-consumption evaluations were captured but not neces-sarily their actual behavior of going to the movies per se. In fact, the choice issue was only examined at the aggregate movie level using Box Office sales data. Research on individual choices for movies could be conducted with proper data acquisition, such as individual re-views instead of averages of all the rere-views. In this study, however, single digits with equal weights were used. In addition, in order to yield more accurate results, data could have been aggregated by the researcher instead of sampling pre-aggregated data. Note that as warned by Moon, Bergey, and Iacobucci (2010), such research would require a sophisticated choice model development because the set of available movies is large and changes across viewers and time.
27
Last, star power was measured by the appearance of stars, and the movies that had win-ning awards were given more weight than the movies that were just nominated. It might be the case that winning awards do not have an effect on the next movie. As proposed by Chiriaki (2015), star power should only be measured by nominations and not winning awards. This is because star power has an effect as a signal only before releasing the movie and not after releasing the movie. The author adds that the information related to OCR has more weight on driving consumers sales.
6 Conclusions
The aim of this study was to gain more understanding of the WOM phenomena regard-ing to product performance of consumption goods. Moreover, the present study went into depth on the mediation effect of (dis)satisfaction on product performance as measured in re-purchase behavior. Therefore, online consumer reviews and box office sales data were col-lected with respect to prequel and sequel movies. This study found that dissatisfaction regard-ing to products of the same series have no significant impact on the repurchase of a subse-quent product in that series when controlling for star power and production budget. Moreover this study found that evaluations of one product of the series are positive related to evalua-tions of the next series. In light with this discovery, more studies need to extend the satisfac-tion literature to see whether this phenomena happen in other than consumpsatisfac-tion products.
28
7 Bibliography
Adermon, A., & Liang, C. Y. (2010), Piracy, music, and movies: A natural experiment (No. 2010, 18), Working Paper, Department of Economics, Uppsala University.
Amblee, N. F., & Bui, T. (2011), Harnessing the influence of social proof in online shopping: The effect of electronic word of mouth on sales of digital microproducts. International Journal of Electronic Commerce, 16(2), 91-113.
Amblee, N. F., & Bui, T. (2011), Harnessing the influence of social proof in online shopping: The effect of electronic word of mouth on sales of digital microproducts. International Journal of Electronic Commerce, 16(2), 91-113.
Anderson, E. W., & Sullivan, M. W. (1993), The antecedents and consequences of customer satisfaction for firms. Marketing science, 12(2), 125-143.
Austin, B. A. (1981), "Film Attendance: Why College Students Chose to See Their Most Re-cent Film," Journal of Popular Film and Television, (3), 43-49.
Banerjee, A. V. (1992), A simple model of herd behavior. The Quarterly Journal of
Econom-ics, 797-817.
Bass, F. (1969), A new product growth model for consumer durables. Management Science, 15, 215-227.
Basuroy, S., & Chatterjee, S. (2008), Fast and frequent: Investigating box office revenues of motion picture sequels. Journal of Business Research, 61(7), 798-803.
Basuroy, S., Chatterjee, S., & Ravid, S. A. (2003), How critical are critical reviews? The box office effects of film critics, star power, and budgets. Journal of Marketing, 67(4), 103-117.
Bowman, D., & Narayandas, D. (2001), Managing Customer-Initiated Contacts with Manu-facturers: The Impact on Share of Category Requirements and Word- Of-Mouth Be-havior. Journal of Marketing Research, 38, 291-297.
C. Dellarocas (2003), The digitalization of word of mouth: promise and challenges of online feedback mechanisms, Management Science, 49(10), 1407-1424.
Caresa, A (2015), Thesis Seminar Business Studies, Universiteit van Amsterdam, 1-59
Chang, B. & E. Ki. (2005), Devising a Practical Model for Predicting Theatrical Movies Suc-cess: Focusing on the Experience Good Property. Journal of Media Economics, 18, (4), 247-269.
Chatterjee & Patrali (2001), Online Reviews: Do Consumers Use Them?, Association for Consumer Research, ACR 2001 proceedings, M. C. Gilly, J. Myers-Levy, EDS.: 129-134
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.
Chen, Y., Fay, S., Wang, Q. (2011), The Role of Marketing in Social Media: How Online Consumer Reviews Evolve, Journal of Interactive Marketing, 25 (2), 85-94.
Chen, Y., Wan, Q., & Xie, J. (2011), Online social interactions: A natural experiment on word of mouth versus observational learning. Journal of Marketing Research, 48 (2), 238-254.
Chevalier, J. A. & Mayzlin, D. (2006) The Effect of Word of Mouth on Sales: Online Book Reviews. Journal of Marketing Research, 43 (3), 345-354.
Chintagunta, P.K., Gopinath, S., & Venkataraman, S. (2010), The effects of online user re-views on movie box office performance: Accounting for sequential rollout and aggre-gation across local markets. Marketing Science, 29 (5), 944-957.
Claycamp, H. J., & Liddy, L. E. (1969), Prediction of new product performance: An analyti-cal approach. Journal of Marketing Research, 414-420.
29
Clemons, E. K., Gao, G., & Hitt, L. M. (2006),When online reviews meet hyper-differentiation: A study of the craft beer industry. Journal of Management Information
System, 23(2), 149-171.
Connelly, B.L, Certo, S.T., Ireland, R.D. & Reutzel, C.R. (2011), Signaling Theory: A Re-view and Assessment. Journal of Management, 37 (1), 39-67
Cowton, C. J. (1998), The use of secondary data in business ethics research. Journal of
Busi-ness Ethics, 17 (4), 423-434.
Cui, G., Lui, H. K., & Guo, X. (2012), The effect of online consumer reviews on new product sales. International Journal of Electronic Commerce, 17(1), 39-57.
Dellarocas, C., Zhang, X. M., & Awad, N. F. (2007), Exploring the value of online product reviews in forecasting sales: The case of motion pictures. Journal of Interactive
mar-keting, 21(4), 23-45.
Dhar, T., Sun, G., & Weinberg, C. B. (2012), The long-term box office performance of sequel movies. Marketing Letters, 23(1), 13-29.
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.
Duan,W., Gu, B., & Whinston, A. B. (2008), Do online reviews matter? — An empirical in-vestigation of panel data. Decision Support Systems, 45(4), 1007-1016.
Elberse, A. (2008), Should you invest in the long tail?. Harvard business review, 86(7/8), 88. Elberse, A., & Eliashberg, J. (2003), Demand and supply dynamics for sequentially released
products in international markets: The case of motion pictures. Marketing
Sci-ence, 22(3), 329-354.
Eliashberg, J., & Shugan, S. M. (1997), Film critics: Influencers or predictors? The Journal of
Marketing: 68-78.
Feng, J. & Papatla, P. (2011), “Advertising: Stimulant or Suppressant of Online Word of Mouth,” Quarterly Journal of Economics, 25, 2: 67-76.
Friedman, B., Khan Jr, P. H., & Howe, D. C. (2000), Trust online. Communications of the
ACM, 43(12), 34-40.
Ghauri, P., & Gronhaug, K. (2010), Research Methods in Business Studies 4th Edition
Gruen, T. W., Osmonbekov, T., & Czaplewski, A. J. (2006), eWOM: The impact of custom-er-to-customer online know-how exchange on customer value and loyalty. Journal of
Business research, 59(4), 449-456.
Hardoon D. R. & Shmueli, G (2013) “Getting Started with Business Analytics: Insightful De-cision Making”, Chapman and Hall/CRC.
Ho-Dac, N. N., Carson, S. J., & Moore,W. L. (2013), The effects of positive and negative online customer reviews: Do brand strength and category maturity matter? Journal of
Marketing, 77(6), 37-53.
Jiménez, F., & Mendoza, M. (2013), Too popular to ignore: The influence of online reviews on purchase intentions of search and experience products. Journal of Interactive
Mar-keting, 27(3), 226-235.
Johnson, B. & Reingen, P. (1987), "Social Ties and Wordof-Mouth Referral Behavior".
Jour-nal of Consumer Research, 350-62.
Kindem, G. (1982), "Hollywood's Movie Star System: A Historical Overview," in The Amer-ican Movie Industry: The Business of Motion Pictures, Gorham Kindem, ed. Carbon-dale, IL: Southern Illinois. University Press: 79-94.
Koenker, R., & Hallock, K. (2001), Quantile regression. Journal of Economic Perspectives, 15(4), 143-156.
Kostyra, D. S., Reiner, J., Natter, M. & Klapper, D., (2015) Decomposing the effects of online customer reviews on brand, price, and product attributes. International Journal of
