The influence of electronic word-of-mouth volume on the length of box office firms' theatrical runs in the United States of Amercia : mediating role of sales

(1)

B a c h e l o r E c o n o m i c s a n d B u s i n e s s

S p e c i a l i z a t i o n : B u s i n e s s A d m i n i s t r a t i o n

The Influence of Electronic Word-of-Mouth Volume on

the Length of Box Office Films’ Theatrical Runs in the

United States of America: Mediating Role of Sales

Bachelor’s Thesis by:

Rosaria Cynthia Devi

10888225

Supervisor: dr. Frederik Situmeang Amsterdam, 26th_{June 2018}

(2)

Statement of Originality

This document is written by Student Rosaria Cynthia Devi who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are 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 completion of the work, not for the contents.

(3)

Abstract

Due to the raise of Internet, people tend to check on the Internet in advance before they do a purchase, especially in case of experience goods. This is where electronic word-of-mouth (eWOM) plays a big role on people’s purchase decision, and therefore the demand of the experience goods. This study investigated whether the volume of eWOM affect the decision of movie exhibitors on how long a movie should stay available in the movie theaters, and whether this decision is guided by the sales of the movie. This study made use of 1,757 movies, for which the data were collected from a combination of databases called Box Office Mojo and Metacritic, as its sample. The investigation of this study was carried out by examining four hypotheses, which were analyzed by using four separate Ordinary Least Squares (OLS) regressions in SPSS. Interestingly, the results of this study show that sales are not mediating the influence between eWOM volume and movies’ length of theatrical runs. Furthermore, this study also finds that the volume of expert reviews has a larger effect on sales than the volume of user reviews.

Keywords ⎯ Electronic Word-of-Mouth Volume, Expert Reviews, User Reviews, Theatrical Runs, Movie Industry, Box Office Films, Ordinary Least Squares Regressions, Data Mining.

(4)

Traditionally, when information regarding how good or bad a specific product was needed, many times we found ourselves asking people about it or looking for a better option from others (recommendation) (King, Racherla, & Bush, 2014). Clearly, Tony Hsieh is not the only one who has realized the importance of word-of-mouth. In fact, according to the findings of McKinsey & Co.’s study, up to 67% of United States’ consumer goods’ sales were gained because of word of mouth advertising (as cited in Liu, 2006). However, these days, we can just go to the Internet and find collections of reviews from all around the world easily (King, Racherla, & Bush, 2014). This raises the question whether these reviews affect the decisions made by organizations, specifically in the mass media entertainment industry such as movie theaters.

The fact that some box office films stayed longer than the others (or vice versa) shows that there are some considerations that were taken into account in order to make such decisions. To date, very little research has been conducted to find out what these considerations are. Yet, there is a substantial amount of studies that have found that there is a positive relationship between movies’ online reviews and their sales (Duan & Whinston, 2008; Duan & Whinston, 2008; Kim, Park, & Park, 2013; Liu, 2006), therefore, it is interesting to find out whether the movies’ online reviews actually have an effect on the decision of how long the movies stay available in the movie theaters, considering that sales are important not only for filmmakers but also movie exhibitors.

The research regarding online reviews in this paper is focused on the quantity of the online reviews that movies receive. The reason for this is because there are plenty of existing studies that have looked and proved the importance of

(6)

of them has directly connected it to the length of film screening in the movie theaters.

In this paper, the length of a movie stays in the movie theater is termed as “length of theatrical run” and the amount or quantity of online reviews is termed as “Electronic Word-of-Mouth (eWOM) Volume”. The proposed thesis aims to find out whether the volume of eWOM is one of factors of the length of theatrical runs of movies. The focus of the study is on those movies released in the United States of America. One reason for this is because film industry in the USA is highly demanded (Baumann, 2001) and data is widely accessible.

Thus, the following research question is built: “To what extent do sales

guide the influence of electronic word-of-mouth volume on the length of box office films stay in movie theaters in the United States of America?”

This paper is structured as follows; it begins with a theoretical framework, where related literatures and all variables used in this paper are given. Next is a methodology, which is a thorough explanation (including the steps) of how the research is conducted. Following this is the results part; this is where all results of this research are presented. Afterwards, a discussion is given, where the results in the previous section is explained. Lastly, a conclusion of the whole paper is given.

2 Theoretical Framework Length of theatrical run

One of the main differences of television shows and films that play in movie theaters is their cycle, for which films in movie theaters have way shorter cycle (Liu, 2006). It is very important to notice that dynamic is one of the main characteristics of box office films, some of them could only have a taste for a couple of weeks of screening while others could have a taste for a longer period (De Vany & Walls, 1999). In fact, the initial contract of screening for movies is typically between 4 to 8 weeks after the release date (De Vany & Walls, 1999). For that matter, the investigation regarding the importance of the length of theatrical runs of movies has come to the realization as one of the factors of movies’ success (Chang & Ki, 2005). This makes sense because when a movie runs longer in movie theaters, the filmmaker has a higher possibility of getting more revenues since the

(7)

availability to be watched by customers is longer and therefore, more profit can be gained from the movie.

eWOM Volume

The term eWOM has been mentioned many times so far, however, the clear definition regarding this term has not been explained yet. This term came from the traditional word-of-mouth (WOM) for which has multiple understandings, yet, all can agree on one thing, which is a form of interpersonal information exchange (Bae & Kim, 2013). Furthermore, as previously briefly mentioned, WOM has a big impact on customers’ purchase decision; this is because of the belief that WOM has a higher credibility in the eyes of (fellow) customers (Bae & Kim, 2013). Electronic Word-of-Mouth (eWOM) is basically the more advanced version of this traditional version, where it involves the invention of Internet, and by this, it means this act of information exchange occurs on the Internet (Hennig-Thurau, Walsh, & Walsh, 2003).

eWOM can be divided into two types based on the writer of the reviews namely user reviews and expert reviews, which according to Amblee and Bui (2007), user reviews refer to reviews that are written by customers who basically have experienced or used the product or service, and expert reviews refer to reviews that are written by professionals who often are hired by hosting portals with high integrity (as cited in Li, Huang, Tan, & Wei, 2013). The fact that experts are professionals who often are hired by an organization leads to a question mark concerning the credibility of the reviews that they write, for which questioning whether customers really put more faith on professionals’ comments or not (Purnawirawan, Eisend, De Pelsmacker, & Dens, 2015). Other than that, sometimes customers who are not very familiar with the nature of movie industry (complicated features of movies) may not be very influenced by experts’ preferences (Moon, Bergey, & Lacobucci, 2010). Yet, both types of reviews have effects on the success of movies (Kim, Park, & Park, 2013). Moreover, it can be said that based on the importance of eWOM itself; there are two attributes of eWOM that have been mostly researched, these are volume and valence. According to Purnawirawan, Eisend, De Pelsmacker, and Dens (2015), eWOM

(8)

negative” (p. 18). Meanwhile, eWOM volume represents “the total amount of WOM interaction” (Liu, 2006, p. 75).

Plenty of researches have been done in finding out the importance of eWOM, especially its effect on movies’ success, and most of them measure this success based on the revenues (Liu, 2006). Though, those who have looked into eWOM’s relationship with the length of theatrical run took ratings into consideration, which relates more to eWOM valence than eWOM volume, whereas, one study’s finding found that those 40-week theatrical-ran movies appeared to not ever being one of the top 10 movies (De Vany & Walls, 1997), these ranks were based on movie ratings given by the audiences. Yet, those who have done similar research, placing theatrical run as one of the dependent variables for measuring overall performance of the movies (Sochay, 1994), or with the audience rating as one of their independent variables, they found significant findings, which indicated a relationship between eWOM and the length of theatrical run (Chang & Ki, 2005). Moreover, Moul (2007) found consumer expectations toward some movies indeed had an effect on movies admissions, in which they took the length of theatrical run into consideration. However, as the author mentioned in the end of his conclusion part, his research was done in the beginning of Internet era, which clearly means that a new research is needed as the use of Internet is now has grown bigger than ever. Regardless, this shows that there are other factors than ratings (or maybe ratings are not even one of the components) that could be the factors of the decision of movies’ length of theatrical runs. Since none of these studies really focused on the volume aspect of these eWOM and its effect on the length of theatrical run, it is interesting to find out what their relationship is.

In this paper, it can be said that the relationship between eWOM and the length of theatrical run are associated with the signaling theory. Spence (1973) explains that signaling theory exists when information asymmetry existed in the first place, for which the existence of signals is to reduce this unequal condition. In case of this paper’s topic, the evaluations by both experts and users can be considered as signals for customers regarding how interesting the movie is (Situmeang, Leenders, and Wijnberg, 2014). Hence, just like how customers get the signal, eWOM could have been used to notify movie exhibitors on how long a

(9)

movie should be available in the theaters, in a sense that when a movie is actively being reviewed, this could attract more people to watch this movie and therefore the movie exhibitor should extend its length of theatrical run. Thus, the first hypothesis is as follows:

-H1: There is a positive relation between eWOM volume and the length of theatrical runs of box office films.

Sales

It is a common knowledge that sales is one of the most important factors for any kind of business in order to keep on running. Together with sales, that is how the company can profit.

Many researches have put eWOM volume as the predictor of movie sales, which their result appeared to be supported (Duan & Whinston, 2008; Duan & Whinston, 2008; Kim, Park, & Park, 2013; Liu, 2006; Rui, Liu, & Whinston, 2013). This is not a surprise because the more people talk about a movie, it means the greater the attention that will come to it, and therefore it will end up with more people wanting (and actually going) to see the movie and increase sales in the end (Liu, 2006). Furthermore, Kim, Park, and Park (2013) find that the frequency of eWOM plays a bigger role than the valence of eWOM especially for user written reviews. This is because the valence of eWOM do not necessarily influence the movies’ performance in the end (Duan & Whinston, 2008). Therefore, the second hypothesis is:

-H2: eWOM volume of box office films is positively related with their sales per week. Moreover, as previously mentioned, sales are not only important for filmmakers, but also for movie exhibitors and it appears that sales per week of a movie has an affect on the decision regarding the length of a movie runs in theaters in order to increase the overall returns (De Vany & Walls, 1997). This leads to the third hypothesis:

(10)

-H3: Sales per week of box office films is positively related with their length of theatrical runs.

As previously stated, a typical initial contract for screening is between 4 to 8 weeks, this number of weeks is a subject to change during the theatrical runs, such that the final decision of the length of theatrical run will be depended on the demand that occurs during the movies’ theatrical run (De Vany & Walls, 1999). In the previous two hypotheses, sales stand as either the result (H1) or predictor (H2). However, since there has been an indication that eWOM volume has a relationship with the length of theatrical run, and also an indication that sales have a relationship with both eWOM volume and the length of theatrical run, can sales actually be the mediator of this relationship between eWOM volume and the length of theatrical run? The answer has still remained unknown, and thus, is leading to the fourth hypothesis:

-H4: Sales per week is mediating the relationship between eWOM volume and the theatrical run of box office films.

Summarizing the hypotheses mentioned above, the proposed relationship between variables introduced is illustrated by the model in Figure 1.

H1

H2 H3

(11)

Figure 1. Conceptual model

3 Methodology

In this section, the report regarding how the study was conducted is give. This section is dived into four subsections. First, design and sample, where the thorough explanation on how the set of sample (dataset) used for this study was collected and prepared before the analyses is given. Second, measurement, this is where the explanations concerning how each variable was measured is given. Moreover, in this second subsection, explanations and premises of the chosen control variables are also included in this subsection. Third, analytical plan, this subsection contains the explanations on how the analyses were done for checking the correlations and regressions.

3.1 Design and sample

The data was collected by means of big data. The data set analyzed is composed of data collected by using a secondary data (data mining) with deductive approach that was mostly collected from the free data feeds named “Box Office Mojo” (www.boxofficemojo.com) and a reviews aggregator website named “Metacritic” (www.metacritic.com) by the thesis supervisor beforehand. Box Office Mojo was launched in 1999 and now is owned by IMDb which is an online data reporting website for box office movies, it contains many information regarding box office movies, including their sales (in US and worldwide), production budget, number of movie theaters, and many more. Meanwhile, Metacritic was launched back in 2001 and it collects data from approximately 40 different offline and online review sources and summarizes each review with a score (which they call it metascore) within the range of 0 to 100 (King, 2007). For more information regarding both websites, please find this in both websites.

However, as it is known that one of the limitations of secondary data is the different purpose of collecting the data at the first place (Cowton, 1998). This limitation will be avoided by using a careful analysis (data retrieval) of which inputs are appropriate to be used, as well as by controlling some variables. Other than that, the dataset collected by the supervisor does not have all the variables

(12)

needed for this specific research, and therefore, I also collected some extra data from the same free data feeds in order to be able to provide all relevant variables. The initial dataset covers box office films from the year 2000 until 2017, and it contains 5,231 films, with 203,512 user reviews and 185,368 critic reviews. However, for this research, dataset that was used is from the year of 2010 to 2017 only. Year 2010 to 2017 is chosen because of the Great Recession that started in the end of 2007. Great recession in 2007 had a dramatic economic impact in the USA, and it clearly had an impact on its citizens’ spending decisions, such that numerous citizens lost their jobs and led to those who still had their job scared of losing theirs too and therefore cut their spending as an act of anticipation (Perri & Steinberg, 2012). It is believed that the expenditure level was back to normal in 2010 (Perri & Steinberg, 2012), and that is why the data from 2010 onwards are chosen.

In the end, by using spreadsheet program named Microsoft Excel (Excel), the chosen dataset was being stored and cleaned. With this program, there were several actions taken such as store the initial data, sort, filter, and delete some data. During the data cleansing, it occurred that some movies have incomplete information, such as missing information regarding the number of release days and weeks, or missing information regarding one of the control variables. Thus, those movies that cannot be completed were ignored and deleted and the final amount of movies that were analyzed is 1,757 movies.

3.2 Measurement

3.2.1 Dependent variable ‘Length of Theatrical Run’

The length of theatrical run variable was measured by the number of weeks the movies stay in theaters. The numbers were collected from Box Office Mojo in shape of number of days, but after using Excel, the data was transformed into weeks by dividing the initial numbers by 7. Those movies that had incomplete number of days were being re-searched on Box Office Mojo and directly collected in terms of weeks, while those that did not have the information were being deleted. Moreover, movies that were released in 2009, but still available for screening in 2010 were not included in dataset for this study, because they were still considered to be 2009’s box office films.

(13)

3.2.2 Independent variable ‘eWOM Volume’

eWOM volume was measured in terms of two different categories namely ‘user reviews’ and ‘expert reviews’. Data for both categories were collected from Metacritic website. This way, other than using these two variables for its main function as the independent variables, it is also possible to find out whether one type of reviewers will have a stronger effect than another on the theatrical run or other variables. The volume of the eWOM is measured by the amount of reviews that both users and/or experts gave about the movies on different sources but had been collected by Metacritic and being shown on its website. Every review was being stored in one separate sheet on the same Excel document, and by using the “COUNTIF” formula, the amount of reviews that each movie has, could be recorded in terms of numbers.

3.2.3 Mediator ‘Sales per Week’

Initially, sales per week were collected as total sales of each movie. These total sales data for every movie were collected from Box Office Mojo. Furthermore, in order to keep the accuracy of the research, the sales amount of each movie has been corrected according to the US inflation conversion factor. The Inflation multipliers from different years were found on “Bureau of Labor Statistic” of the USA website (www.bls.gov). Every amount of total sales (grouped based on their released year) was converted into their equal amount in March 2018 after the inflation. The conversion was done by using the “VLOOKUP” formula on Excel. However, since the measurement of the length of theatrical run is per week, it was decided that the amount of sales is better to be measured per week as well. Thus, the total domestic sales collected were divided by the number of weeks that the movies were on the screening period. Thus, the amount of sales per week was measured in terms of average sales per week.

3.2.4 Control variables

In order to reduce bias, there were several variables that were kept constant (controlled). These variables are ‘beginning number of theaters’, ‘ending number

(14)

of theaters’, ‘age restriction’, ‘genres’, and ‘released year’. Reasons for choosing and how these variables were measured are explained below:

Number of theaters

Number of theaters were collected from Box Office Mojo website. In this paper, number of theaters is divided into two variables, the number in the beginning of the movie’s release date (which is the number of theaters in the first week) and the number in the end of its theatrical run (which is the number of theaters in the last week of the movie’s theatrical run). These two variables are considered to be relevant to be held constant because of the high variations of number of theaters where movies’ were played. Some movies were played in less than 10 theaters while some others were played in more than 2000 theaters in the beginning of theatrical run, and same story goes for the ending of theatrical run. Moreover, the number of theaters where it played in the beginning was mostly different from when it is the last week of the theatrical run. The possible reason for this is because of the different release strategies by film distribution, such that film distributor might decide to do exclusive booking for certain popular areas, where as they try to make an impression of limited screening in public eyes to increase sales (Litman, 1983).

Age restriction

Age restriction of movies was measured by MPAA (Motion Picture Association of America) Ratings (https://www.mpaa.org). This data regarding the MPAA ratings for all movies were collected from Box Office Mojo. This rating system is used to inform the content of the movie in advance to the audience, in order to keep the children stay on the appropriate shows only (Litman, 1983). According to the website of MPAA, there are 5 types of rating symbols by MPAA Ratings; these are G, PG, PG-13, R, and NC-17. The G rating means that the content in the movie has nothing inappropriate to be shown to children (all ages are permitted). PG rating means that there may be an inappropriate content in the movie for children (children should have a guidance by an adult). PG-13 rating means that there may be inappropriate material even for children under 13 (and therefore, parents needs to give a guidance). R rating means ‘restricted’, this means that the movie

(15)

contains some adult material, for which it urges parents to accompany their under 17 years old children. Last, NC-17 rating, this rating is for those movies that clearly permit only those who are above 17 years old (considered as adults). From all of these ratings, there is one symbol that was appeared to be more common than the rest, which is the R rating. Therefore, a 0-and-1 dummy variable was created for this variable by using an Excel sheet, where “IF” formula was used, with 1 means it is a R rating movie and 0 means the movie is classified into other types of ratings.

Genres

The data of genres is collected from Box Office Mojo. Data collected have various genres. Thus far, there has not been an official statement regarding what the main (official names for) genres are and different studies used different genre categories, there is a study which used 10 categories, there is also a study which used 5 categories (Kim, Park, & Park, 2013). Yet, for this study, 4 genres categories were chosen based on the 4 most common genres that the movies in the dataset have, these are Action, Comedy, Drama, and Thriller. A 0-and-1 dummy variable coding is used for each different type of genres. For each genre, 1 means one of the chosen genres and 0 is not one of the chosen genres. For example, for action, 1 means the movie has an action genre and 0 means the movie does not have an action genre. This categorization process used the same way as classifying the R rating movies, where “IF” formula was used on Excel to know whether the movie contains the specific genres or not.

Released year

According to Litman (1983), “Timing” is an important factor for the success of a movie; he believes that there is always the right timing to catch the most attention or being appealing in public eyes. For the same reason, I decided to take released year as one of the control variables as it is considered to be able to have an effect on the hypothesized relationships. The data of released years is collected from Box Office Mojo. Though, released year has been previously used to cut the initial data in order to reduced the bias that could be cause by Great Recession which

(16)

3.3 Analytical plan

In order to test the proposed hypotheses, the chosen and cleaned data were transferred to software named, Statistical Package for the Social Sciences (SPSS) for analyses. SPSS is one of the most used software to do regressions, especially in social science and psychology area of studies (Preacher & Hayes, 2004). The first thing that will be done is to check the mean (M) and standard deviation (SD) of all variables. Then, to check the correlation between variables, a bivariate correlation will be used; this way the level of correlations and whether one variable is positively or negatively correlated to another can be known. The correlation that will be used for this paper is Pearson Correlation coefficient. Table 1 shows the different correlation levels, which basically represents the rules of thumb for correlation.

Table 1

Interpretation of Pearson Correlation

Value Interpretation

-1 Perfect negative correlation -1 < r < -0.8 Very strong negative correlation -0.8 < r < -0.6 Strong negative correlation

-0.6 < r < -0.4 Moderately strong negative correlation -0.4 < r < -0.2 Weak negative correlation

-0.2 < r < 0 Very weak negative correlation

0 No correlation

0 < r < 0.2 Very weak positive correlation 0.2 < r < 0.4 Weak positive correlation

0.4 < r < 0.6 Moderately strong positive correlation 0.6 < r < 0.8 Strong positive correlation

0.8 < r < 1 Very strong positive correlation 1 Perfect positive correlation

Furthermore, for all regressions that will be performed, all control variables will be allocated to the independent variables section (or box) after the (main) independent variable(s) from each regression has been allocated.

(17)

For first hypothesis, to test whether there is a positive relation between eWOM volume and the length of theatrical run, an Ordinary Least Squares (OLS) regression will be used with eWOM volume (both user reviews and expert reviews variables) as the independent variable and length of theatrical run as the dependent variable. To test the second hypothesis, the positive relation between eWOM volume and sales per week, an OLS regression will be used with eWOM volume as the independent variable and sales per week as the dependent variable. For the third hypothesis, the positive relation between sales per week and the length of theatrical run, an OLS regression also will be used with sales per week as the independent variable and length of theatrical run as the dependent variable. Lastly to test the fourth hypothesis, the mediating role of sales per week on the relationship between eWOM volume and the length of theatrical run, another OLS regression will be used with eWOM volume and sales per week (the order of variables placement is eWOM volume and then sales per week) as the independent variables, and length of theatrical run as the dependent variable.

The confidence interval that is used for the regressions is 90%, meaning the significance level () for the research is set to 0.1. Any regressions’ result that has a significance level below 0.1 is supported and those that are higher than 0.1 are not supported.

4 Results

In this section, the results from SPSS are reported. This section is divided into three parts: first is the report regarding the descriptive statistics, second is the report regarding the correlation analyses, and the last one is the report regarding the regression analyses.

4.1 Descriptive statistics

The information regarding the descriptive statistics of all variables that were used for analyses are summarized in Table 2 and 3. In Table 2, information regarding the number of data (N), minimum statistic value of each variable (Min.), maximum statistic value of each variable (Max.), and the mean standard error (Std. error) are presented. In Table 3, the information regarding all the means and standard

(18)

standard deviations are represented in the table as SD. The amount of data for every variable (N) is 1,757; this is unarguably because only movies with complete information were taken into account.

All means and standard deviations were all above zero, where the larger the standard deviations, the larger the spread of the distribution (dispersion). Moreover, as it can be seen in Table 3, the mean of the number of user reviews was twice as many as the expert reviews, it means that it can be said that almost every movie had more user reviews than expert reviews (M = 46.440 for the volume of user reviews and M = 24.860 for the volume of expert reviews). In addition to that, from Table 2, it can be seen that the maximum number of user reviews was noteworthily higher than the number of expert reviews (Max. = 1,933 for the volume of user reviews and Max.= 55 for the volume of expert reviews).

Table 2

Descriptive statistics overview

Variable N Min. Max. Std. Error

Length of theatrical run (in weeks)

1,757 .286 273 .239

Number of user reviews 1,757 0 1,933 2.467

Number of expert reviews 1,757 0 55 .299

Sales per week 1,757 336.783 41759524.200 102660.760

Beginning theaters 1,757 1 4416 36.014 Ending theaters 1,757 1 2545 2.672 Released year 1,757 2010 2017 .048 R (dummy) 1,757 0 1 .011 Action (dummy) 1,757 0 1 .008 Comedy (dummy) 1,757 0 1 .009 Thriller (dummy) 1,757 0 1 .006 Drama (dummy) 1,757 0 1 .010 4.2 Correlation analyses

(19)

Other than the means and standard deviations, Table 3 also contains the relevant information regarding the correlations.

Whether a correlation is significant or not is determined by a p-value that less than 5%, and by using that determination, as it can be seen in Table 3, all correlations between the independent variable, dependent variable, and mediator were significant. Although the correlations are weak, the relation between eWOM volume and the length of theatrical run are still correlated positively, which means it is in line with the expectation, such that the volume of both user reviews and expert reviews are positively correlated with length of theatrical run, r (1,757) = .169, p < .01 for user reviews and r (1,757) = .234, p < .01 for expert reviews. Not only that, the relationship between eWOM volume and sales are also in line with the expectation, they are positively correlated, r (1,757) = .724, p < .01 for user reviews and r (1,757) = .530, p < .01 for expert reviews. It is good to note that the correlation between the volume of user reviews and sales per week is categorized as very high. Interestingly, it also shows that user reviews has a higher correlation with sales than expert reviews by approximately .2. Moreover, also as expected, the relation between sales and the length of theatrical run is positively correlated, r (1,757) = .141, p < .01.

Furthermore, regarding the correlations with the control variable, almost all control variables are negatively correlated with the length of theatrical run, such that only the number of beginning theaters is positively correlated with the length of theatrical run, r (1,757) = .094, p < .01. However, variables action and drama are not significant under 5% significance level (r (1,757) = -.014, p = .545 for action and r (1,757) = -.011, p = .644 for drama), meaning there is no sign of relationships for both drama and action with the length of theatrical run. Next, the correlations between control variables with eWOM volume. For the volume of user reviews, approximately half of the control variables indicated relationships with eWOM volume (r (1,757) = .499, p < .01 for number of beginning theaters, r (1,757) = .146, p < .01 for number of ending theaters, r (1,757) = .330, p < .01 for action, and r (1,757) = -.068, p < .01 for drama). The other half of the control variable were correlated insignificant (r (1,757) = -.044, p = .064 for released year, r (1,757) = .001, p = .980 for R rating, r (1,757) = -.040, p = .090 for comedy,

(20)

reviews, only one variable was insignificant, which is released year, r (1,757) = -.019, p = .432. The rest of the control variables were positively correlated with the volume of expert reviews. Moreover, the correlations between the control variables and the sales per week were mainly significant and positive, only comedy, r (1,757) = .035, p = .142, and thriller, r (1,757) = .006, p = .809, were not significant.

One may question the high correlation between number of beginning theaters and sales per week (r (1,757) = .769, p < .01). However, multicollinearity is not a big problem for this case, such that the value of VIF (Variance Inflation Factor) between these two variables was acceptable, meaning multicollinearity should not be an issue (see Appendix A). Same decision can be made from the correlation between sales per week and number of user reviews (r (1,757) = .724,

p < .01) (see Appendix A).

To conclude, all correlations between the main variables and mediator were significant and positive regardless the strength of the correlation, and therefore in line with the expectations.

(21)

Table 3

Descriptive statistics and correlations matrix.

M SD 1 2 3 4 5 6 7 8 9 10 11 1 Length of theatrical run 12.509 10.013 2 Number of user reviews 46.440 103.390 .169** 3 Number of expert reviews 24.860 12.532 .234** .498** 4 Sales per week 2467229.418 4303188.344 .141** .724** .530** 5 Beginning theaters 1172.240 1509.598 .094** .499** .530** .769** 6 Ending theaters 45.740 111.991 -.066** .146** .196** .315** .358** 7 Released year 2013.56 1.998 -.112** -.044 -.019 -.070** -.077** -.054* 8 R (dummy) .340 .475 -.128** .001 .251** -.056* -.054* .029 .007 9 Action (dummy) .110 .319 -.014 .330** .219** .358** .339** .110** -.005 .021 10 Comedy (dummy) .180 .387 -.056* -.040 .103** .035 .112** .078** -.053* .198** -.047 11 Thriller (dummy) .080 .271 -.066** .031 .094** .006 .058* .023 -.020 .154** -.027 -.123** 12 Drama (dummy) .22 .413 -.011 -.068** .118** -.109** -.132** .007 .132** .123** -.125** -.041 -.048*

(22)

4.3 Regression analyses

eWOM volume and length of theatrical run

To test the first hypothesis, an OLS regression for eWOM volume (user and expert reviews) on the length of theatrical run was used. Results are summarized in Table 4 and showed that an increase amount of eWOM results in a longer theatrical run,  = .071, t = 2.552, p < .05 for user reviews and  = .331, t = 10.974,

p < .001 for expert reviews. These results provided support for the first

hypothesis, which means that eWOM volume was indeed positively associated with the length of theatrical run.

eWOM volume and sales per week

To test the second hypothesis, I regressed eWOM volume (user and expert reviews) as the independent variable and sales per week as the dependent variable. The results of this regression are given in Table 4, and this showed an increase in eWOM volume results in an increase in sales per week,  = .434, t = 29.059, p < .001 for user reviews and  = .050, t = 3.075, p < .01 for expert reviews. These results reflected a support for the second hypothesis, meaning that eWOM volume was indeed associated with sales per week in a positive way.

Sales per week and length of theatrical run

To test the third hypothesis, another OLS regression was performed for sales per week on the length of theatrical run. The results of this regression are shown in Table 4, and these results showed that an increase in sales per week predicts more weeks of theatrical run,  = .179, t = 4.841, p < .001. These results provided a support for the third hypothesis, this means sales per week is indeed positively associated with the length of theatrical run.

Mediation test

In order to test whether sales per week mediates the influence of eWOM volume and the length of theatrical run, an OLS regression was used with eWOM volume as the predictor of length of theatrical run after controlling for the effects of sales per week. The results of this regression are shown in Table 4 and it showed that the association between eWOM volume and the length of theatrical run is still

(23)

there. However, this association has a lower significance level compare to the first single regression, at least for the user review,  = .061, t = 1.792, p < .1 for user reviews and  = .329, t = 10.903, p < .001 for expert reviews. Moreover, this regression shows an insignificant effect of sales per week,  = .024, t = .529, p = .597. In other words, this means that sales per week do not guide the influence of eWOM volume on length of theatrical run. These results do not provide enough evidence to support the fourth hypothesis.

(24)

Table 4

Regression Results for all 4 models

Model 1 Model 2 Model 3 Model 4

Dependent variable Length of theatrical run Sales per week Length of theatrical run Length of theatrical run Coefficient SE Beta Coefficient SE Beta Coefficient SE Beta Coefficient SE Beta Constant 1141.255*** 227.432 33480797.000 52290222.700 1129.835*** 236.305 1139.411*** 227.506 Number of user reviews .007* .003 .071 18077.993*** 622.104 .434 .006a _.003 _.061 Number of expert reviews .264*** .024 .331 17010.805** 5532.641 .050 .263*** .024 .329 Sales per

week 4.154E-7*** .000 .179 5.508E-8 .000 .024

Beginning theaters .000* .000 -.063 1402.662*** 47.380 .492 .000 .000 .024 .000 a _.000 _-.075 Ending theaters -.009*** .002 -.104 2513.921*** 492.093 .065 -.011*** .002 -.121 -.009*** .002 -.106 Released year -.562*** .113 -.112 -16818.907 25969.144 -.008 -.555*** .117 -.111 -.561*** .113 -.112 R (dummy) -3.816*** .519 -.181 -312112.670** 119353.015 -.034 -1.977*** .515 -.094 -3.799*** .520 -.180 Action (dummy) -2.613*** .766 -.083 362836.274* 176185.038 .027 -2.284** .791 -.073 -2.633** .767 -.084 Comedy (dummy) -1.474* .610 -.057 -115739.500 140236.492 -.010 1.365* .631 -.053 -1.467* .610 -.057 Thriller (dummy) -2.892* .853 -.078 611721.690** 196027.172 -.039 -2.233* .886 -.060 -2.859** .855 -.077 Drama (dummy) -.785 .571 -.032 154748.370* 131304.142 -.015 .589 .580 .024 -.777 .571 -.032 R2 _.136 _.753 _.067 _.136 Note 1. N = 1,757.

Note 2. The reported coefficients are the unstandardized coefficients and Beta represents the standardized coefficients.

* is significant at the 0.050 level (2-tailed).

(25)

5 Discussion

The discussion section is divided into four subsections. The first subsection is the

summary of the results, hypotheses, and unpredicted results. Second subsection is

the discussion points, where both supported and unpredicted results are explained with possible reasons. Third subsection is the explanation about the

contributions that this study gives in both theoretically and practically

perspectives. The last subsection is the limitations and suggestions for future

research.

5.1 Summary, hypotheses, and unpredicted results

This study investigated the interacting role of eWOM volume and sales per week in predicting the length of theatrical run. Thus, the following research question was built: “To what extent do sales guide the influence of electronic word-of-mouth volume on the length of box office films stay in movie theaters in the United States of America?” In order to answer the specified research question, it was tested whether sales per week mediates the influence of eWOM volume on the length of theatrical run.

The first hypothesis that expected a positive relation between eWOM volume and the length of theatrical run was supported along with the second hypothesis that stated that eWOM volume is positively correlated with sales per week. Not only that, the third hypothesis that expected a positive relation between sales per week and the length of theatrical run was also supported. Yet, the correlations between all variables in the first and third hypotheses were either very weak or weak. Only the correlation between eWOM volume and sales were quite strong, for which the amount of user reviews has a higher correlation with sales per week than the amount of expert reviews. Moreover, the unforeseen result was found for the mediating role of sales in the relationship between eWOM volume and the length of theatrical run, as it was not supported. Thus, the overall result was not in line with the prediction.

5.2 Discussion points and interpretation

To begin with, the difference between this paper and previous researches is that this research allocated sales per week as the mediator of the relationship

(26)

sales as their dependent variable, which they used as the representative of movies’ success (Duan & Whinston, 2008; Duan & Whinston, 2008; Kim, Park, & Park, 2013; Liu, 2006). However, the analyses done for this study shows no evidence for those sales per week guiding the influence of eWOM volume on the length of theatrical run. This means that the relationship between eWOM volume and the length of theatrical run is mediated by something else since they are still positively associated, which is the first discussion point for this study. Before explaining the possible reasons for this unpredicted result, one of the result points needs to be made clear; one may recall the fact that there is a high correlation between one of the control variables (number of beginning of theaters) with the one of the independent variables (sales per week), as previously explained, this is not a case of multicollinearity issue, such that a strong problem of multicollinearity occurs when VIF is larger than 10 (Lin, 2008), for which the tolerance value is when VIF is equal or under 5 (Friday & Emenonye, 2012). Therefore, using the same assumption, this study concludes that this is not the reason for this null finding.

One possible reason for the null finding regarding the failure to prove a mediating effect of sales could be because sales could not be the mediator for a relationship that involves eWOM as the predictor considering the amount of previous studies and the support for this paper’s second hypothesis that have proved that eWOM has a significant effect on sales, whereas sales stood as the dependent variable. Another possible reason for this null finding is the fact that this study only took volume into consideration. As previously explained, there are two attributes of eWOM; volume and valence. If the null finding is related to this fact, then it could be that sales are the mediator for the influence of eWOM valence on the length of theatrical run, or maybe when both attributes of eWOM are taken into the consideration to fulfill the definition of eWOM.

Next discussion point is regarding the relationship between eWOM volume and the length of theatrical run. Although the results show a very weak association, they do still have a relationship. However, in regard to the fact that the association is weak, this paper cannot confirm the present of signaling theory, which I expected to happen between the amount of eWOM and the decision made by movie exhibitors regarding the length of movies’ theatrical runs. One reason

(27)

for this doubt is the fact that eWOM volume does not literally explain whether the movie will excites customers or not. Meanwhile, eWOM valence could be a representative of the quality of the movie (Duan & Whinston, 2008). A movie can have hundreds of reviews, but if these reviews are mostly negative, it does not make sure that people will be willing to see this movie, such that negative reviews are perceived to be credible and even more credible than the positive ones (Pan & Chiou, 2011). Moreover, previous studies indeed found that the valence of eWOM does play some important roles in success of products, such that negative reviews have a negative impact on box office performance (Basuroy, Chatterjee, & Ravid, 2003), or even specifically on recommendation intension (Purnawirawan, Eisend, De Pelsmacker, and Dens, 2015). Thus, the amount of eWOM might not be a signal for movie exhibitors to decide how long a movie should stay in the movie theaters. There is a higher possibility that this signaling theory can be occurred in the case of the relationship between eWOM volume and sales considering the high correlation between the two variables. Furthermore, since movies are in entertainment industry and therefore categorized as experience goods, people tend to look for information in order to get the signals of how this experience good will be like and therefore another possible reason is that signals would have been found if the relationship was between eWOM valence and sales (Liu, 2006; Reinstein & Snyder, 2005). However, it did not appear that movie exhibitors get signal to extend the length of theatrical run of a movie.

Furthermore, one interesting result that should be pointed out is the fact that the amount of user reviews has a higher effect on sales than the amount of expert reviews. A reason for this could be because of the fact that expert reviews are written by professionals who were hired by (mostly) organizations to write reviews. In spite of the fact that they supposed to have a high integrity level, it could be that customers do not easily believe on what these professionals write about these movies. Therefore, I agree with the expectation from Purnawirawan, Eisend, De Pelsmacker, and Dens (2015), which stated that the credibility of expert reviews could be questioned which leads to some customers will probably put more faith on fellow (not being paid) customers, who they know do not get

(28)

5.3 Contributions and practical implication

The primary aim of this study is to contribute to the existing literatures about the effect of eWOM on the success of a movie, specifically the effect of one of the eWOM attributes, volume. So far, existing literatures are mainly measured ‘success’ by using the sales of the movie (Duan & Whinston, 2008; Duan & Whinston, 2008; Kim, Park, & Park, 2013; Liu, 2006; Rui, Liu, & Whinston, 2013). This study confirms the results of the previous studies regarding the positive relationship between eWOM volume and sales. However, in this study I have tried to shift the focus of movie’s success from the filmmakers’ perspective. This study focuses on the movie’s success from the movie exhibitors’ perspective. This was measured by the length of theatrical run, where the focus was on whether movie exhibitors will extend a movie’s length of theatrical run when this movie is being a hot topic in the internet (high volume of eWOM). Without ignoring the results from other studies regarding the influence of eWOM on sales, this study tested whether sales can actually be the mediator for the influence of eWOM on the length of theatrical run. However, as the results of this study shows, there is indeed an influence of eWOM volume on the length of theatrical run, but sales is not the mediator of this relationship. Therefore, the influence of eWOM volume on the length of theatrical run is guided by something else.

Another additional contribution to the existing literatures is the introduction of two new control variables, which are the number of theaters where the movies are screening in the beginning of the movie release and in the end of their theatrical run. As previously explained, these variables are considered to be important enough to be taken into account due to the variety of the amount of theaters for different movies. The result of this study also contributes to the stream of research on user reviews and expert reviews in terms of their volume (Purnawirawan, Eisend, De Pelsmacker, & Dens, 2015). This study confirms that expert reviews do not necessarily have a higher impact on customers’ decision (which presented in a shape of a lower effect on sales compared to effect of user reviews on sales) due to a possible threat of lower credibility.

(29)

As for the practical implication of this study, it is believed that this study can contribute to the making of marketing strategy for filmmakers, such that eWOM volume is indeed important for increasing sales. However, instead of spending so much money on convincing some organizations to make reviews on their movies, it may be more useful to motivate people who have watched their movies to make reviews about the quality of the movies. Moreover, as for the movie exhibitors, the results of this paper can be used to convince them on the accuracy of predicting the revenues that come from ticket selling based on the eWOM volume.

5.4 Limitations and future research

One of the limitations of this research is the fact that the sample of this study were taken from only two sources, namely Box Office Mojo and Metacritic. Even though that the size of the sample used for this study is actually good, the fact that Box Office Mojo website could not provide all the information needed to complete the data for every variables or some movies could not be found in one of the sources led to a smaller sample size. This is because only those movies with complete information for the variables were taken to be analyzed. Hence, the use of more sources of databases is recommended for future research in order to get a higher sample size.

Second limitation of this study is related with the fact that this study shows a lower effect of the volume of expert reviews on sales compare to the effect of the volume of user reviews on sales. Despite the explanation regarding the credibility of expert reviews in the eyes of customers, I believe that there should have been an extra control variable regarding the organization where these experts were hired from. This is because, there is a possibility that the type of organizations affects the credibility level perceived by customers. Therefore, future study may benefit by categorizing the type of organizations where the expert reviews are produced or maybe include a category of independent experts (they were not hired by any organizations).

Moreover, the unexpected result of this study gives an idea about the direction that the future research should be taking. First, as previously stated,

(30)

The fact that this study only considered volume into the analyses may have affected on the overall result in a way that the result of this paper contradicts with the expectations where the results of the regression models failed to give evidence for the mediating role of sales on the influence of eWOM volume on the length of theatrical run. With that being the case, combining eWOM volume and eWOM valence could affect the results.

Secondly, this study only made use of the software SPSS to analyze all its hypotheses. Meanwhile, there is a special tool for SPSS called “The PROCESS macro for SPSS and SAS”, where to test a mediating effect is one of its function. This tool was not officially used to check the possible mediating role of sales per week on the influence of eWOM volume on the length of theatrical run. Yet, there

was a sign of mediating role of sales when I used this tool to check this mediation

effect (see Appendix B), yet, this action did not take into account all variables, such that only the volume of user reviews (as the independent variable) and sales per week (dependent variable), with only 2 control variables (number of theaters in the beginning and ending of the theatrical run). Therefore, future research may use this tool to further analyze the effects and see whether it will really provide a different outcome than only using OLS regressions by SPSS.

6 Conclusion

In this study, a relationship between movies’ eWOM volume and their length of theatrical run was tested, not only that, this study took sales as the mediator of this relationship. This study used 1,757 movies that were released between the years of 2010 to 2017 in the USA as its sample. Thus, the following research question was built: “To what extent do sales guide the influence of electronic

word-of-mouth volume on the length of box office films stay in movie theaters in the United States of America?” Both length of theatrical run and sales were measured

in terms of weeks.

In order to answer the built research questions, four hypotheses were built. To find out whether those hypotheses are supported or not, four ordinary least squares (OLS) regressions were performed. The first three hypotheses were supported. These supported ones were regarding the positive relationship between eWOM volume and the length of theatrical run (H1), the positive

(31)

relationship between eWOM volume and sales (H2), the positive relationship between sales and the length of theatrical run (H3). The fourth hypothesis, which is about proving the mediating role of sales, was not supported. This means based on the analyses done in this study, sales were found not as the mediator of the influence of eWOM volume on the length of theatrical run. Therefore, this study provides a partly positive answer for the built research question, where sales do have positive independent relationships with eWOM volume and the length of theatrical run, yet, it cannot be confirmed that sales has a role of mediator within the relationship between eWOM volume and the length of theatrical run.

Furthermore, while looking for the answer of the research question, this study found some interesting results regarding the use of eWOM. First, the strong negative correlation between eWOM volume and the length of theatrical run that this study found has led to the doubt of the present of signals perceived by the movie exhibitors. Second, the low credibility perceived by customers on expert reviews causes the result of this study regarding the larger effect of the volume of expert reviews on sales compared to the effect of the volume of user reviews on sales. Finally, this study provides some recommendations for the direction of future research.

(32)

References

About Box Office Mojo. (n.d.). Retrieved from

http://www.boxofficemojo.com/about/

Bae, J., & Kim, B. D. (2013). Is the electronic word of mouth effect always positive on the movie? Academy of Marketing Studies Journal, 17(1), 61.

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.

Baumann, S. (2001). Intellectualization and art world development: Film in the United States. American Sociological Review, 66(3), 404-426.

Chang, B. H., & Ki, E. J. (2005). Devising a practical model for predicting theatrical movie success: Focusing on the experience good property. Journal of Media

Economics, 18(4), 247-269.

Cowton, C. J. (1998). The use of secondary data in business ethics research. Journal of Business Ethics, 17(4), 423-434.

De Vany, A. S., & Walls, W. D. (1997). The market for motion pictures: rank, revenue, and survival. Economic inquiry, 35(4), 783-797.

De Vany, A., & Walls, W. D. (1999). Uncertainty in the movie industry: Does star power reduce the terror of the box office? Journal of cultural

economics, 23(4), 285-318.

Duan, W., Gu, B., & Whinston, A. B. (2008). Do online reviews matter?—An empirical investigation of panel data. Decision support systems, 45(4), 1007-1016.

Duan, W., Gu, B., & Whinston, A. B. (2008). The dynamics of online word-of-mouth and product sales—An empirical investigation of the movie industry. Journal of retailing, 84(2), 233-242.

Film Ratings. (n.d.). Retrieved from https://www.mpaa.org/film-ratings/

Friday, O. R., & Emenonye, C. (2012). The detention and correction of multicollinearity effects in a multiple regression diagnostics. Elixir Stat, 49, 10108-10112.

Hall, R. E. (2010). Why does the economy fall to pieces after a financial crisis? Journal of Economic perspectives, 24(4), 3-20.

(33)

Hennig-Thurau, T., Walsh, G., & Walsh, G. (2003). Electronic word-of-mouth: Motives for and consequences of reading customer articulations on the Internet. International journal of electronic commerce, 8(2), 51-74.

Kim, S. H., Park, N., & Park, S. H. (2013). Exploring the effects of online word of mouth and expert reviews on theatrical movies' box office success. Journal

of Media Economics, 26(2), 98-114.

King, R. A., Racherla, P., & Bush, V. D. (2014). What we know and don't know about online word-of-mouth: A review and synthesis of the literature. Journal of Interactive Marketing, 28(3), 167-183.

King, T. (2007). Does film criticism affect box office earnings? Evidence from movies released in the US in 2003. Journal of Cultural Economics, 31(3), 171-186.

Li, M., Huang, L., Tan, C. H., & Wei, K. K. (2013). Helpfulness of online product reviews as seen by consumers: Source and content features. International

Journal of Electronic Commerce, 17(4), 101-136.

Lin, F. J. (2008). Solving multicollinearity in the process of fitting regression model using the nested estimate procedure. Quality & Quantity, 42(3), 417-426.

Litman, B. R. (1983). Predicting success of theatrical movies: An empirical study. The Journal of Popular Culture, 16(4), 159-175.

Liu, Y. (2006). Word of mouth for movies: Its dynamics and impact on box office revenue. Journal of marketing, 70(3), 74-89.

Metacritic: The History. (n.d.). Retrieved from http://www.metacritic.com/about-metacritic

Moon, S., Bergey, P. K., & Iacobucci, D. (2010). Dynamic effects among movie

ratings, movie revenues, and viewer satisfaction. Journal of

Marketing, 74(1), 108-121.

Moul, C. C. (2007). Measuring word of mouth's impact on theatrical movie admissions. Journal of Economics & Management Strategy, 16(4), 859-892.

Pan, Lee-Yun and Jyh-Shen Chiou (2011), “How Much Can You Trust Online Information? Cues for Perceived Trustworthiness of Consumer-generated Online Information,” Journal of Interactive Marketing, 25(2), 67–74.

(34)

Perri, F., & Steinberg, J. (2012). Inequality and redistribution during the Great Recession. Economic Policy Paper, 1.

Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behavior research methods,

instruments, & computers, 36(4), 717-731.

Purnawirawan, N., Eisend, M., De Pelsmacker, P., & Dens, N. (2015). A meta-analytic investigation of the role of valence in online reviews. Journal of

Interactive Marketing, 31, 17-27.

Reinstein, D. A., & Snyder, C. M. (2005). The influence of expert reviews on consumer demand for experience goods: A case study of movie critics. The

journal of industrial economics, 53(1), 27-51.

Reiss, R. (2010, July 1). Tony Hsieh On His Secrets Of Success. Forbes. Retrieved

from

https://www.forbes.com/2010/07/01/tony-hsieh-zappos-leadership-managing-interview.html#628870bd6d4e

Rui, H., Liu, Y., & Whinston, A. (2013). Whose and what chatter matters? The effect of tweets on movie sales. Decision Support Systems, 55(4), 863-870.

Situmeang, F. B. I, Leenders, M. A. A. M, & Wijnberg, N. M. (2014). The good, the bad and the variable: How evaluations of past editions influence the success of sequels. European Journal of Marketing, 48(7/8), 1466-1486.

Sochay, S. (1994). Predicting the performance of motion pictures. Journal of

Media Economics, 7(4), 1-20.

Spence, M. (1973). Job Market Signaling. The Quarterly Journal of Economics,

(35)

Appendix A VIF report.

Variable VIF

Number of user reviews 2.340

Number of expert reviews 1.843

Sales per week 4.043

Beginning theaters 2.929 Ending theaters 1.175 Released year 1.026 R (dummy) 1.230 Action (dummy) 1.208 Comedy (dummy) 1.126 Thriller (dummy) 1.081 Drama (dummy) 1.122

Note. N = 1,757, DV: Length of theatrical run.

All VIFs were below 5 and therefore, in accordance to the VIF < 10 and VIF < 5 assumptions, there should not have been a problem caused by multicollinearity.

(36)

Appendix B

Regression using PROCESS by Andrew F. Hayes

Note 1. N = 1,757.

Note 2. The reported coefficients are the unstandardized coefficients, with a number of bootstrap of 1,000.

* is significant at the 0.050 level (2-tailed)

** is significant at the 0.010 level (2-tailed) *** is significant at the level 0.001 level (2-tailed)

a_{is significant at the level 0.100 level (2-tailed)}

One of the differences between using only OLS regressions provided by SPSS and using PROCESS on SPSS is that PROCESS does not take into consideration the standardized coefficient. That is why there is not a beta reported on the table above. By using a 90% confidential level, it appeared that sales are mediating the relationship between the volume of user reviews and the length of theatrical run, t = 1.649, p < .1. The p-value of sales in this case was .0993, which is very close to .1. Thus, a future research with a deeper analysis regarding this mediation effect (using PROCESS) is recommended in the limitation and future research section of this study.

Model 1 Dependent

variable Length of theatrical run

Coefficient SE Constant 11.869*** .300 Number of user reviews .012** .003 Sales per week .000 a _.000 Beginning theaters .000 .000 Ending theaters -.0101*** .002 R2 _.040

The influence of electronic word-of-mouth volume on the length of box office firms' theatrical runs in the United States of Amercia : mediating role of sales

B a c h e l o r E c o n o m i c s a n d B u s i n e s s

S p e c i a l i z a t i o n : B u s i n e s s A d m i n i s t r a t i o n