The effect of cast diversity on the financial succes of movies

(1)

1

Sunaina Braam 10110364 22-06-2018

MSc. in Business Administration – Marketing track University of Amsterdam – Amsterdam Business School

Thesis supervisor : F. Situmeang

THE EFFECT OF CAST

DIVERSITY ON THE FINANCIAL

SUCCES OF MOVIES

(2)

2

Statement of originality

This document is written by student Sunaina Braam, who declares to take full responsibility for the contents of this document. I 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 completion of the work, not for the contents.

(4)

4

Acknowledgements

Writing my master thesis has been quite a bumpy road. While never encountering any

difficulties during my academic or high school journey before, finishing this master has been quite the challenge. I was faced with a personal tragedy during my first attempt for my thesis, I had to stop my studies and stop working. Trying to keep all those different balls in the air at that time resulted in all of them dropping on the floor all at once. This was a few years back though, and I came back stronger. However, each year I was hesitant to work on my thesis. I even though about giving up. In hindsight, I was afraid I would fail again and that believe held me back of actually doing it. Instead I focused my energy on what was going well for me, my professional career. Now that graduation is close, I feel privileged and thankful that I had the opportunity to pursue this degree with support from my family in a country that enables me to do so. I want to thank my boyfriend Joël for helping me with the enormous database during the thesis process, but more importantly his support every step of the way, during the good times but also the very bad. I came back stronger because you were there. I want to thank my thesis supervisor Frederik Situmeang for his continued support, flexibility, words of encouragement and trust. He never stopped believing in me. Last but not least, I want to thank my parents for the sacrifices they made to give me and my brother everything we ever wanted and their continued support in everything I do in life. I wouldn’t be close to where I am in life right now, if it weren’t for them. Thank you from the bottom of my heart.

(5)

5

Abstract

This research builds on existing literature on predicting box office revenues. The purpose of this study was to examine the relation between the level of ethnic diversity in a movie and box office success. This is the first study to research a sample of this size and incorporate diversity as one of the variables in a predictive regression model. A sample of 774 movies released between 2013 and 2017 was analyzed and the main actors were labelled by their ethnicity. The effect of diversity on customer and critic reviews was researched as being a possible mediator. Production costs, top distributor and the number of days a movie is shown in theatres, were used as control variables. The positive relationship between the number of non-white actors and box office success was not significant after controlling for the other variables. The data further shows that movie casts are not representative for the U.S population and that the relation between diversity and movie success should be further investigated. Results also indicate that there is a clear distinction to be made between the top grossing movies and the overall collection as they have different control variables and the effect of diversity is much larger in the top movies, though not significant enough in this study. This study debunks a long hold belief in Hollywood that diversity doesn’t sell overseas. The findings should encourage other scholars to examine the effects of diversity and may be used as input for prediction models with diversity as one of the independent variables.

(6)

6 “Our ability to reach unity in diversity will be the beauty and test of our civilization” –

Mahatma Gandhi

1 Introduction

Over the years Hollywood has received significant backlash for their lack of ethnic diversity. A telling example was discovered in the leaked Sony emails of 2014. A Sony producer contacted an executive complaining that The Equalizer, a movie starring Denzel Washington, failed to achieve success overseas because Denzel was black. The producer went further suggesting that Sony should avoid black actors in an effort to appease international markets. Hollywood was also scrutinized for its lack of diversity via the #oscarssowhite hashtag campaign on Twitter. Protesters from all over the world, including several famous actors, voiced their dismay over the lack of people of color among the Oscar nominees.

Deciding who should anchor a story is mostly imbued with financial considerations (Smith et al. 2016). In the film industry, hiring a non-white lead is deemed risky and

automatically considered as catering to a different audience. There is a common belief among filmmakers that movies with largely black casts will be seen as ‘‘Black movies’’, that white audiences will largely avoid (Weaver, 2011). For a long time, in Hollywood it has been often assumed that movies with non-white lead actors would not be commercially successful, especially overseas.

Is it financially risky to cast minority actors? Shattering records worldwide, some see Black Panther, a Marvel live-action movie with a 90% black cast, as evidence that diversity does sell. Recent studies seem to support this notion.

Over the past few years there have been a number of studies that researched diversity in Hollywood movies. The Creative Artists Agency (CAA) analyzed 413 movies that were released between January 2014 and December 2016, the study considered the ethnicities of

(7)

7 the top 10 billed actors of each film, 2,800 individuals total. For the 10 highest-grossing films in 2016, people of color comprised 47 percent of the opening weekend’s audience. Seven of the 10 top-grossing films boasted majority non-white audiences on their opening weekends. Truly diverse films, defined by CAA as a cast that is at least 30 percent non-white , earn more than non-diverse films in the opening weekend, even when controlled for the budget. As for audience members, “the average opening weekend for a film that has a ‘truly diverse’

audience, pegged at 38 percent to 70 percent non-white, is $31 million versus $12 million for films with non-diverse audiences.” The CAA study’s findings align with the Hollywood Diversity Report (2017), which states that movies with diverse casts boast the highest median global box office and the highest median return on investment. The report also noted that diverse films from 2011 to 2015 outperformed expectations at the box office.

Another study on this topic by Bunche Center was carried out in 2014 with the goal of determining the influence of diversity on the worldwide box office revenues for movies, discovered that the movies that had diverse actors performed better. The results showed that movies and TV shows with more than 40 percent minority actors in their cast performed twice better than those with less than 10 percent minority actors. According to Hoag (2016), the movies that combine actors from diverse group are more likely to perform better than the other movies. The author claims that films with actors from both minority and majority cultural groups perform better in the theatres than other movies.

The results from these recent studies suggest that diversity could actually be

financially rewarding for movie makers. The main goal of the studies mentioned above was to investigate the level of diversity and its evolution over the years, with moral viewpoints in mind. These studies cover only the well-known films and incorporated limited control factors. They focus solely on diversity and the methods of statistical analyses from the CAA and the Hollywood diversity report are not published in detail. Aside from the moral

(8)

8 considerations to represent a diverse group of people in movies, there are pressing financial considerations regarding diversity that can only be studied when incorporating other financial predictors. The demographics of the United States and other European countries have

changed significantly in ethnic composition (USA Census Report 2018, CBS 2017) and filmmakers might be leaving money on the table if they’re not serving the whole population. Given the fact that movie makers tend to avoid risk taking behavior such as casting a non-white lead actor, the trend of less risk-taking in casting to reduce financial risk could actually be hurting the performance and should be investigated

Because the movie industry is an industry that poses great financial risks, considering the big investments needed, researchers have been on a hunt for the Hollywood success formula, trying to discover the determinants of box office revenues from various academic disciplines including marketing, organizational science, economics, and mass communication (Kim et al. 2013). But researching diversity with the purpose of predicting box office success in a predictive model combined with other predictors hasn’t been done in the academic literature before.

The current study is the first to focus on diversity mainly as a determinant for financial success while controlling for known determinants like production budget, the number of in release days and the size of the production company behind the movie. It makes a distinction between the top grossing movies and the total collection of movies released between 2013 and 2017. With this set-up, it can help fill the gap of knowledge on how diversity in film works in the complex web of determinants for financial success.

In order to gain a comprehensive understanding of the role diversity plays, this study used a sample of movies that is larger than in previous studies. It covers 774 movies released between 2013 and 2017, and analyzes both the total collection of movies as well as a

(9)

9 above. In this research diversity in a movie is studied as a variable that can influence

customer reviews, critic reviews and box office success.

How did diversity in the lead cast influence the financial success of movies released between 2013-2017?

• 1. How diverse are the movies that have been released between 2013 and 2017? • 2. How does diversity effect box office results when controlled for common

determinants?

• 3. How does genre matter in the relation between the number of non-white actors and box office success?

• 4. How does diversity indirectly affect financial success as mediated customer and film critic reviews?

1.1 Thesis overview

This thesis is structured as follows: Chapter 2 provides the literature reviews where social identity theory, Uses and gratification theory and the effects of critic and online reviews for movies will be discussed along with the hypotheses. Chapter 3 explains the methodology in detail, describes the variables used in testing the hypotheses. The results of the analyses will subsequently be presented in Chapter 4 along with the output and descriptives from the statistical analysis performed. The empirical findings are discussed in Chapter 5 along with the managerial implication, the limitations and suggestions for further research and the final conclusions.

(10)

10

2 Literature review

By studying diversity as a possible predictor of box-office, this study adds to a growing body of research that is exploring the predictors of financial success of movies. Predicting weekly box-office demand is an important and challenging question. Because of the high uncertainty in the movie business and large investments needed to make a film, movie makers tend to avoid risk-taking behavior. Literature on forecasting success of new motion pictures can be grouped by the type of forecasting model employed.

The first category is the study of quantitative models that explore a combination of factors that influence box-office numbers of movies (Litman, 1983; Litman & Kohl, 1989; Litman & Ahn, 1998; Neelamegham & Chintagunta,1999; Ravid, 1999; Elberse &

Eliashberg, 2002; Sochay, 1994).

The second category consists of behavioral model focused research, that primarily examines the individual’s decision-making process when selecting a specific movie (De Silva, 1998; Eliashberg & Sawhney, 1994; Eliashberg et al., 2000; Sawhney & Eliashberg, 1996; Zufryden, 1996). The first category predicts either total revenues or total profits based on several elements (production cost, star power, etc.). The second category is more variable focused and wants to investigate how an individual’s decision-making processes can predict movie choice.

This study differentiates from previous studies by combining the two approaches. This research builds on behavioral theories from communication, marketing, and psychology to hypothesize on the effects of cast diversity on consumers’ movie preferences . At the same time it uses the existing literature on success predictors to propose a predictive model.

(11)

11

2.1 (Social) identity theory and Social cognitive theory

Identification occurs when an individual adopts behavior derived from another person or a group because this behavior is associated with a satisfying self-defining relationship to this person or group (Kelman, 1961). People do this all the time, consciously as well as

unconsciously. People assess their level of similarity with others and make similarity judgements about that source (Gregory et al., 2013). The more individuals perceive their appearance, personality, character, tastes or preference to be similar to that of another, the greater the identification. Race is an important feature that is easily recognized by physical features and is a common identifier, especially for those to whom racial identity is key to their identity (Phinney, 1989). In media context, this has been shown in studies where black audiences perceived themselves as more similar to black media characters than white media characters, regardless of social class (Appiah, 2004).

Social identity theory builds on identity theory in the sense that it underlines people’s tendency to identify with others who have similar characteristics, but also emphasizes the need of people to see themselves as part of a group and the important role of group

membership in one’s self-image (Tajfel, 1978). People are strongly motivated to maintain a positive self-concept and thus tend to favor groups they are a part of (ingroups) and

discriminate against outgroups (Tajfel, 1978). Even when group categorizations are made on a trivial basis (e.g., the flip of a coin), people will make judgments and take actions to

promote the status of their own group and devalue others (Billig & Tajfel, 1973). Through these favorable intergroup comparisons, one’s personal identity can be elevated. When applied to a media context, social identity theory has been used to predict that the need to create these intergroup comparisons can influence people’s viewing decisions (Blumler,

(12)

12 1979; Harwood, 1997; McQuail, 2000; Trepte, 2006). Viewers use media to gather

information and form impressions of social groups (Mastro, 2003). Because people are motivated to attain a positive social identity and elevate their ingroups, they are drawn to content that references their ingroup and preferably also depicts their ingroup in a positive way (Trepte, 2006). Moreover, audiences may selectively avoid movies and programs which they believe would cast their ingroup in a negative light (Abrams & Giles, 2007; Frutkin, 1998). Racial group categorization could influence selective exposure in this way. Race, because it is such a salient feature, is frequently used to make ingroup–outgroup

categorizations. For many people this is an important component of their social identity (Hewstone, Hantzi, & Johnston, 1991). This effect is much stronger for people from minority groups in a predominantly white society. This occurs because white people do not generally think of themselves as part of a specific ethnic group, as their skin color is the norm in their environment. It has been discovered that they consistently place significantly lower

importance on their racial and ethnic identity than for example black people (Phinney 1992). Given that people form ingroups based on race, social identity theory predicts that viewers would be drawn to media depictions that elevate the racial ingroup and would avoid depictions that would diminish the relative status of the racial ingroup. Abrams and Giles (2007) found evidence for selective avoidance among black audience members when their own ethnic group was perceived to be either underrepresented or misrepresented.

Social cognitive theory is related to social identity theory and has also been used to predict that individuals are drawn to media content featuring ingroup members, although in this case the attraction comes not from a need to elevate one’s ingroup but from a preference for observing similar behavioral models (Bandora, 2001; Knobloch, Callison, Chen, Fritzsche, &

(13)

13 Zillmann, 2005). According to social cognitive theory, people routinely observe other

people’s behavior to inform and evaluate their own actions (Bandora, 2001). Such social learning involves selective attention to behavioral models and as it turns out; people prefer to attend to, and learn more from others who are similar to themselves (Knobloch et al., 2005). For example, studies on gender have consistently shown that females are drawn to

programming featuring females and males are drawn to programming featuring males (Knobloch et al., 2005; Knobloch-Westerwick & Hastall, 2006; Oliver, 2000).

2.2 Uses and gratification theory

While the abovementioned theories focus on when and why consumers would prefer content with characters that resemble themselves, uses and gratification theory (UGT) complements this theory by explaining what motives people have to attend media content and what they receive in return from consuming it. UGT comes from a history of communication theories and research where audience members were largely seen as targets, or interpreters, rather than active seekers of content. This theory takes the viewpoint of an active audience that consumes mass media with a goal in mind. It is based in the socio-psychological

communication tradition, and focuses on communication at the mass media scale (West et al. 2007). The driving question of UGT is to find out why people use media and what do they use it for. According to scholars, uses and gratifications can be classified into five categories related to five groups of human needs (Katz, Gurevitch, Haas, 1973; Tan, 1985)

1. Cognitive needs – acquiring information, knowledge, understanding our social environment, curiosity, exploration.

(14)

14 3. Personal identity – self-confidence, personal stability, integrity, social status.

4. Integration and social interaction – family relations and friendship, connection with the outside world, the need for affiliation.

5. Escapism – the need to escape from everyday life.

Although this list has been expanded and rearranged by some, for the purpose of

entertainment media in this research this one offers the most suitable categorization. It shows that the experience of emotions can be gratifying for media users (Vorderer, Klimmt and Ritterfeld, 2004). Personal identity needs in the context of UGT means that viewers enjoy comparing their lives to a character in the film and assess the similarities and differences. the In order for entertainment media to fulfill these needs, especially the affective and the personal identity needs, the users need to be able to empathize and identify with the

characters displayed (Bartsch and Viehoff, 2010). Austin (1986) was one of the scholars to apply UGT to motives for movie goers. His study showed that especially frequent movie goers can identify their motives quite easily, “learning and information” and “learning about self” being reported as motives, which would fall under personal identity and cognitive needs. Pleasure and fun was also reported (integration and social interaction category). When examining social identity theory and UGT related to movie viewers, it could be hypothesized that a more diverse audience would prefer movies with actors that resemble themselves in order to identify themselves with them. Consumers selectively choose which movie to go to, based on the needs they are looking to fulfill. That is, audiences may be motivated to select content featuring same-race characters either because of a perception that such content will portray the ingroup (social identity theory) and this can satisfy the personal identification needs (UTG). Especially since race is an important part of the identity of minority ethnic groups in Western society but not so much for white people (Phinney, 1992),

(15)

15 diversity in movies could increase revenues as the ethnic composition of the western society has changed.

H1. A movie with a higher number of non-white actors will result in a higher global box office

2.3 Determinants of success

In this paragraph other variables that have a significant effect on box-office revenue will be discussed, so that the regression model can be corrected. Past success give an illusion of control, where there really is not (Langer, 1975; Presson & Benassi, 1996). Because of the high costs and risks of movie production, it is understandable that studios pursue successful franchises to control risk (Eliashberg et al. 2006).

The first multiple regression model used to predict the financial success of films was developed by Litman (1983). Factors which Litman (1983) viewed as important were the cast, director, production budget, and critical rating. Litman also stressed the contribution of major distributors for a film release. Top distributors all have their own niche and are a signal for movie goers of movies with significant quality. They can rely on a loyal target audience (Litman and Kohl, 1989). Besides that, they have greater bargaining power, more substantial financial resources, preferential access to theatres, and extensive distribution networks. Films being released by a major distributor significantly increases box office (Litman 1983;

Pangarker & Smit, 2013).

In a study by Kim (2013), main actor, actress and director were found to be the most likely determinants of box office success in their study, but the authors go on saying that many studies have had conflicting ideas and that there is no consensus. Star power was researched as determinant by De Vany and Walls (1999), together with movie budget. These

(16)

16 authors concluded that there is a significant relationship between production budget and box office sales, however neither a large budget, star power or past success guarantee box office success according to these authors. They deemed it impossible to attribute the success of a movie to individual causal factors, and go on saying that the audience decides the fate of a movie.

In line with that, Yahav (2014) proposed, a model for demand of a movie by utilizing information on a movie similarity network, the audience. This similarity network is defined as a network of similar movies based on appeal, genre, tone, and timeframe or release pattern. The author claims that movies that attract the same audience, should have similar demand structure. Sequels and trilogy’s are popular within the movie industry because these too have a lower perceived risk of failure. Studios produce sequels of successful movies, expecting that the sequel will be successful too. Yong et al. (2013) addressed this topic and discovered that the success of a parent movie enhances the first sequel attendance, but hardly transfers to movies after that. However, they also acknowledge that attendance rates of a parent movie are no guarantee for success of a sequel because the characteristics of the sequel itself are significant factors that influence its success. Numerous sequels that fail or have disappointing results prove this. On sequels, Basuro and Chatterjee (2006) note that weekly box office results also drop quicker for the sequel compared to the parent movie. The consensus is that the sequel of a movies does provide a signal of quality to the audience, but box office success is still influenced by a lot of other factors.

When it comes to genre and its effect on box-office success, there is no conclusive answer. Sawhey and Eliashberg (1996) find that movies in the drama genre take longer to show box office sales, whereas action usually exhibits revenues quickly. Neelamegham & Chintagunta (1999) found that the thriller genre is the most popular genre across countries, while romance genre is the least popular. This result contradicts the study from Ravid (1999).

(17)

17 He found that family films were very succesful. Other research from Collins et al. (2002) found that genre and its effect on box-office success remains less certain, partially because not all genres convert well in othe languages in culture. They concluded genre should be regarded as a control factor and not as a predictive variable. Walls (2005) found similar results, none of the genre classifications individually were significant. Even though some researchers find some significant results when it comes to genre, nothing conclusive can be drawn from this research.

Overall, the research on determinants of box office is growing and maturing but there is no single point of truth. In this study production budget and major distributor are taken as control variable as the author believes there is substantial evidence these variables influence box office. The number of release days in theatre is also incorporated as control variable as also suggested in previous studies (Walls 2005; Yahav 2015), the reasoning behind it being that the number of days a movie is showing in theatres allows for more ticket sales. On genre the research isn’t clear either. Since a genre is sensitive to cultural differences, the decision was made to explore the genre variable and test it in this model as a possible moderator on the relationship between the number of white actors and box office.

2.4 Movie reviews

One of the factors that has been discussed extensively by scholars in recent years is the movie evaluation of consumers and critics in (online) reviews. In the creative industries and for other experience goods, expert recognition can play an important role in consumer choices because it’s difficult for consumers to establish product quality before consumption.

(Wijnberg & Gemser, 2000; Caves, 2000). At the time Litman did his research in 1983, this effect was much smaller as internet was not as widely available. By now extensive research

(18)

18 has been done on the topic of online reviews from both critics and consumers, and its impact on box office revenue. Movies are popular topics when generating electronic word of mouth (Yeap, Ignatius & Ramayah, 2014). Especially the influence on performance of critic

reviews have been discussed extensively by scholars. A big portion of the scholars identified a positive relation between the critic’s reviews and financial success (e.g., Litman, 1983; Litman & Kohl, 1989; Sawhney & Eliasberg, 1996; Sochay, 1994). Nonetheless, other

scholars did not find a relationship (Delmestri, Montanari, & Usai, 2005; Reinstein & Snyder, 2005), while others have found a negative association (Hirschman & Pieros, 1985; Simonton, 2005a). These contradictory results reflect the complexity of the phenomenon. For instance, the precise relation between the critic review and box office returns varies across the weeks of the film’s theatrical run (Basuroy et al., 2003; Eliashberg & Shugan, 1997) and depends on whether the reviews are positive or negative in valence (Basuroy, Chatterjee, & Ravid, 2003). In general, there is consensus that word-of-mouth (WOM) influences people’s movie

selection (Austin, 1989; Bayus, 1985; Faber & O’Guinn, 1984; Neelamegham & Chintagunta 1999) and thus can influence the success of a movie (Basuroy et al., 2003; Gemser et al., 2007; Liu, 2006; Eliashberg & Shugan, 1997; Mckenzie, 2009). Ravid (1999) is another author who looked into the effect of movie reviews. From his study, he concluded that at times it does not matter whether a movie review is positive or negative. Irrespective of their nature, more reviews are likely to generate more revenues.

The difficulty of measuring the influence of expert reviews on demand is that products receiving positive reviews, logically tend to be of higher quality thus making it difficult to determine whether the review or the high quality is responsible for the high demand. Wyatt and Badger (1984) also nuanced the effect of critics by saying it’s not so much the rating but more so the amount of information a review contains about a movie, that encourages people watch a movie. Reinstein and Synder (2000) do support the notion that

(19)

19 positive critic reviews are correlated with the performance of a film, even when accounted for the assumed higher quality of movies with good reviews. In sum, empirical evidence for a relationship between box office revenues of a movie and critic reviews show positive but also insignificant relationships, depending on the time period of the study and the specification of the model. However, it seems the balance is skewed towards having a positive impact. Critics review movies for a living, they look at different cinematic aspects from a professional standpoint and don’t watch movies to fulfill identification of entertainment needs as described in UGT, like a regular consumer would. They will spend more time purposefully reviewing content and placing the item in the context of historical, political, social or theoretical context (Reinstein and Synder, 2000). Critics rate less on solely entertainment value and should therefore not favor a particular race/ethnicity in the movie, assuming quality is the same.

Based on this, the following was hypothesized

H2. The number of non-white actors in the lead cast has no effect on critic reviews

H3. Critic reviews have a positive effect on financial success, not influenced by diversity

Consumers however write their reviews based on their experience watching the movie. As hypothesized above, we expect that audiences, assuming the market is more diverse, are better able to fulfill their needs with diverse casts versus with a non-diverse casts. Therefore we expect that their reviews on these movies will be higher.

H4. Movies with a higher number of non-white actors will receives higher consumer ratings

H5. Higher consumer ratings for movies with a higher number of non-white actors will have a positive mediating role on the relation between non-white actors and financial

(20)

20

2.5 Conceptual model

3 Methodology

In order to answer the research question, a comprehensive data set was analyzed using excel and statistical analysis in SPSS. The data set contains public information of all movies from renowned movie websites; boxofficemojo.com and metacritic.com. Data from

boxofficemojo.com included movie titles, release date, director, (lead) actors, runtime, production budget, foreign sales, US domestic sales, genre and distributor. This was

combined with data from metacritic.com; customer review scores, expert review scores and the reviews itself.

Sample

The sample comprised 847 Hollywood movies released between 2013 and 2017. A sample of Hollywood movies was selected because in the western world this is the most influential movie industry with the highest revenues. Combining it with movies from other industry would lead to a sample that was not homogenous enough for regression analysis. Of the 847 movies in the initial sample, 774 remained after clearing the data of missing values for Userscore. The movies that didn’t have a Userscore on Metacritic.com were all lesser known

(21)

21 movies and as a result had no users post a review. The decision was made to leave these cases out of the analysis to ensure the N remained the same in the regression model for all variables. Most recent studies on diversity in movies only considered the top 100-200 movies. A sub-sample was created containing the top 119 grossing films to be able to compare the results of this study with other research. The data was enriched by manually coding actors to label their ethnicity. The variables used in the analyses are described on the following pages.

3.1 Independent variables

Diversity

To gain insight in the level of diversity in the cast, 4367 actors were coded with respective labels White (3450), Black (405) , Middle-Eastern and Indian (101), Asian (160) Native-American and Aboriginal (7) and Latin Native-American (244). The total number of non-white actors in lead roles was computed for each movie. Information used to code the actors were primarily their physical features (skin color, hair, bone structure, eyes, lips). In most cases this gave enough information to assign a label. When this was not the case, their personal information on public domains like IMDB or Wikipedia, their self-proclaimed ethnicity, birthplace or (type) casting for films were used to infer this data. This method led to 4794 actors labelled with their ethnicity, for 29 it was unable to retrieve their background via the internet because there was no information publicly available found on the internet. Given the time constraints of the master thesis and the minor effect on the sample size, the choice was made to not explore options for DVD or other footage of these actors but label them as Unknown. The actors labeled as Unkown were not used for analyses.

(22)

22

Reviews

For movie reviews the site Metacritic.com was used, because this is the only site that shows critic reviews and user reviews in full text next to each other for each film. It offers balanced aggregated scores and makes it easy to compare ratings side-by-side.

Critic reviews

To assess the critic review, the Metacritic score from Metacritic.com was taken. Metacritic is a website that aggregates reviews of media products, including movies. The company collects reviews from around the web and assigns them a score ranging from 0 to 100. In instances where a site uses a measurable metric, like a numerical rating system or a letter grade, Metacritic fills in a number that it most closely believes represents that figure. The site then takes a weighted average of all the reviews. The company doesn’t reveal how much weight it assigns to individual reviewers, but it does explain that certain reviewers are given more significance in overall score based on their stature.

Consumer reviews

Users can rate movies on a scale of 1 to 10 and Metacritic computes an average score based on these for each movie.

Box office

Box office is the number that represent the total cash value of all tickets sold for a movie. Even though a lot of the existing research focusses on domestic sales, this study used the combined number of domestic and foreign income because the hypotheses are partially drawn up based on changing global demand. The results were downloaded to the database from

(23)

23 boxofficemojo.com, owned by IMDB.com. It is the leading online box-office reporting service that is tracks all box office revenues in a systematic way and publishes these on their website. This measure for financial success is widely used to rank movies in order of success and in research. By choosing box office as measure for success the results can be compared to the majority of studies on movie success.

Genre

For each movie the main genre was taken from Metacritic.com, Genres were grouped into the following themes for analysis; Action (Action, Adventure) Suspense (Thriller, Horror,

Crime), Humor (Family, Comedy, Animation) , Sci-fi (Sci-Fi, Fantasy), Emotional (musical, drama, romantic) and Non-fiction (documentaries, foreign). Grouping the genres was necessary because for some genres the number of observations wasn’t large enough to execute a solid analysis.

3.2 Control variables

As discussed in the chapters before, a set of control variables was added to the model. In release days

The number of days a movie is shown in theatres, source boxofficemojo.com Top distributors

A film distributed by one of the 10 top distributors was assigned a score of ‘1’, the rest was assigned ‘0’. The list of top distributors is added in the appendix

Production budget

(24)

24 Table 1 Overview of all variables

3.3 Analysis

The current hypotheses were tested by running hierarchical regression analyses, two of which for hypotheses 1 to examine the different sample sizes. To test hypotheses a hierarchical regression was executed and to test mediation for hypotheses 2,3, 4 and 5 the SPSS PROCESS plugin by Hayes (2013) was used to effectively test for mediation.

4 Results

The following section reports correlations between all variables, descriptive data on diversity in movies and the results of the statistical analyses in order to test the hypotheses. Data

Variable Name Description

Userscore Average consumer rating from metacritic.com on a scale of 1- 10 ProductionBudget Movie production budget in $

Metacritic score Average expert review score from metacritic.com on scale 0 to 100 BoxOffice Worldwide box office sales in $

NWActors Number of non-white actors in the lead cast of a movie Genre

Genre classifications were clustered into 4 themes Suspense, Humor, Drama, Non-fiction, Fantasy

TopDistributor

Classification of the 10 biggest distributors worldwide. Dummy variable

TopMovies

Subsample of 119 movies with the highest worldwide box office sales films that comprise 70% of the total box office sales

(25)

25

Figure 1 Frequency number of non-white actors

In Figure 1 a visual distribution of the number of non-white actors is shown. The data is highly skewed to the left. 47% of the movies consisted of an all-white lead cast and 37% had 1 or 2 non-white actors, 85% had no more than 3 non-white actors in the lead cast.

The evolution over the years 2013 to 2016 is shown in Table 1Table 2. In all the years except for 2017, around 50% of the movies had 0 non-white actors. 2017 is an exception with only 17%, this can be explained because not all movies from 2017 were available in the data base at the time of this research (N = 63). When comparing the numbers on a yearly basis, 2017 can’t be compared with the other years because of this. The majority of the movies had 0 or 1 person of color in the lead cast. In 2013 this was 70%, in 2014 this was 81% of the total, in 2015 this percentage was 76% and in 2016 this was 75%. It tells us that the number of non-white actors in movies doesn’t show a lot of variance and in the time period studied in this sample, the number of diverse actors in movies does not approximates general population statistics. In the USA, 40% of the population is non- white (U.S Census, 2017) while Statistic Netherlands (CBS) doesn’t report ethnicity, it does report that 34% is a first, second or third generation immigrant (CBS, 2018), in the Randstad this number is even higher.

(26)

26

Table 2 Yearly evolution number of non-white actors

Table 3 Ethnic representation in movies

Years Percentage of movies per number of non-white actors

N 0 1 2 3 4 5 or more 2013 203 47% 23% 16% 5% 2% 7% 2014 196 50% 31% 6% 4% 4% 5% 2015 163 55% 21% 12 5% 3% 4% 2016 149 48% 27% 13% 4% 2% 6% 2017 63 17% 25% 11% 16% 19% 11% Ethnicities N N % U.S. Census information*

Aboriginal and native American 7 0% 1.3%

Asian 160 4% 6%

Black 405 9% 13%

Latin-American 244 6% 18%

Indian and Middle-Eastern 101 2% N/A

White 3450 79% 60%

Total 4367 100%

(27)

27 When the actors are split by ethnicity Table 3, it shows us that both in absolute numbers and in percentages, all minority groups are underrepresented and white people are

overrepresented. Latin American people are the most underrepresented group in this study. When the results are split by genre, there are no big differences between the genres except for Action, where most movies have 1 or 2 non-white actors (52%).

Table 4 Number of non-white actors per genre

Number of non-white actors

Genres N 0 1 2 3 4 5 6 or more Non Fiction 108 55% 19% 8% 4% 5% 4% 7% Suspense 130 49% 26% 15% 4% 3% 2% 1% Comedy 203 47% 30% 10% 4% 4% 3% 3% Fantasy 58 47% 24% 14% 9% 3% 3% 0% Drama 183 54% 21% 9% 6% 4% 1% 5% Action 92 25% 34% 19% 9% 5% 4% 4 %

(28)

28

4.1 Correlations

A correlational analyses was used to examine the relationship between all variables in the model. As can be seen in Table 5, there are several significant correlations between the control variables, as expected. InReleaseDays is significantly correlated with

ProductionBudget (p =.000) and TopDistributor (p = .000), ProductionBudget and

Topdistributor are also significantly correlated (p=.000). These variables were entered in the model because they are known predictors of box-office and are not subject of separate analyses.The correlation of 1 between Boxoffice and TopMovies is expected, since

TopMovies is a sub selection of Boxoffice. These two variables will not be analyzed in the same regression model so this does not pose a threat to the validity. The main independent variable, NWactors shows significant correlations only with ProductionBudget, Userscore and TopMovies. The correlation between NWactors and box office should not be completely ignored with p = .072, but when taking a statistical interval of 95% it is not significant. Which suggests that there is a relation between the rate of non-white actors and box office, but this relationship is only significant in the top grossing movies. Since all genres were dummy coded and mutually exclusive, the negative correlation between the genres is as expected. Fantasy and Action are positively correlated to ProductionBudget (p = .00) and to TopDistributor (p = .000). Action is also correlated to TopDistributor but only modest. There are only 2 genres that significantly correlate with NWActors, Suspense is negatively

(29)

29

Table 5 Means (M), Standard Deviations (SD) and Pearson correlations for all variables

Variables M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14

1. InreleaseDays* 83.01 55.31 -

2. ProductionBudget 3.24E+07 5.43E+07 .251** -

3. TopDistributor** 0.43 0.50 .191** .432** -

4. Userscore*** 6.43 1.38 .264** .035 -.010 -

5. MetacriticScore**** 61.03 15.84 .300** -.083* -.091* .679** -

6. NWActors 1.18 1.69 -.008 .075* 0.005 -.071* -.029 -

7. BoxOffice 1.32E+08 2.43E+08 .311** .770** .404** .116** .051 .065 -

8 TopMovies***** 5.95E+08 3.26E+08 0.149 .405** .197* .182* .306** .227* 1.000** - -

9. NonFiction .025 -.262** -.247** .173** .327** .038 -.231* N/A - 10. Suspense -.204** -.148** -.033 -.049 -.110** -.070* -.119** -0.173 -.200** - 11. Comedy .060 -.002 .098** -.174** -.191** -.031 .039 -0.154 -.267** -.244** - 12. Fantasy .055 .332** .124** .039 -.037 -.023 .242** .074 -.127** -.116** .055 - 13. Drama -.004 -.205** -.075* .033 .049 -.001 -.180** -.156 -.264** -.242** -.322** -.153** - 14. Action .084* .507** .209** .032 -.049 .096** .419** .232* -.164** -.150** -.200** -.095** -.198** - Note. N =774. * p < .05 ** p < .01 *** p < .001 (two-tailed).

(30)

30

4.2 Analysis

Prior to conducing two hierarchical regression analysis and two regression analysis with mediation in PROCESS, the relevant assumptions of statistical regressions were tested. First, the respective sample sizes of 774 and 119 were deemed adequate given the number of independent variables to be included in the analysis. An examination of correlations (see Table 5) revealed that no independent variables were highly correlated. The distribution of the outcome variable BoxOffice was skewed to the right with multiple outliers. Visual inspection learned that this variable violated the normality assumption and showed signs of heteroscedasticity. Removing outliers was an option, but since these outliers are in fact very highly grossing and influential movies, it was decided they should be included in the sample. Therefore to improve the validity of the regression model a natural logarithm (ln)

transformation was performed on the BoxOffice variable and BoxOffice(ln) was created. This resulted in a good fit for regression, with a normal distribution and no signs of

heteroscedasticity. Visualization of the normal distribution and heteroscedasticity was added in the appendix.

A three stage hierarchical multiple regression was conducted with BoxOffice(ln) as the dependent variable to test hypothesis 1. NWActors, MetacriticScore and Userescore were entered at stage 1 to examine if there was a main effect, comparable to other studies that didn’t include control variables. In step 2 the different Genre themes – 1 were (Action, Suspense, Fantasy and Drama) and their computed interaction variables NWAction x Genre were added in. In the last step the control variables ProductionBudget, InReleaseDays and TopDistributor were added to see if the main effects would still be signficant after controlling for these. The regression statistics for this model can be found in Table 6. The hierarchical multiple regression revealed that in model 1, NWActors, MetaCriticScore and Userscore all

(31)

31 contributed significantly to the regression model F (3, 768) = 13418, p < .00) and accounted for 5% of the variation in BoxOffice(ln). Introducing the Genre variables and the computed interaction variables explained an additional 25% of variation in BoxOffice(ln) and this change in R² was significant, F (13,758) = 24239, p < .00. Adding the control variables Inreleasedays, ProductionBudget and Topdistributor explained another 28% of BoxOffice(ln) (F(16,755)= 62627, p < 00). But when all control variables were included in stage 3 of the regression model only MetaCriticScore (p = .009), Drama (p =.000) and NonFiction (p = . 000) were modest though significant predictors of BoxOffice(ln). The three control variables ProductionBudget, InReleaseDays and TopDistributor predicted the majority of the model (28%). NwActors and Userscore were not significant anymore in model 3.

Hypothesis 1 was not supported with this analyses for the overall sample.

A second 3 stage hierarchical multiple regression was conducted with the same independent variables but with TopMovies as the dependent variable. This hierarchical multiple regression revealed that all 3 models are significant (p < .00). Model 1 predicts 19% of the variation of the model F (3, 115) = 9016, p < .00). Introducing the Genre variables and the computed interaction variables NWAction x Genre explained an additional 9% of variation in TopMovies and this change in R² was significant, F (6,112) = 7109, p < .00. Adding the control variables Inreleasedays, ProductionBudget and Topdistributor explained another 5% of TopMovies (F(14,104)= 3591, p < 00). When all independent variables were included in stage 3 of the regression model, MetaCriticScore remained a significant predictor (β =.406, p <.00) of TopMovies. The independent variable NWActors (β = .238 p = .052) seems to be a reasonable predictor but when adhering to a 95% confidence interval, the null hypothesis

(32)

32 cannot be rejected since p > .05. From the entered control variables, only ProductionBudget (p < .00) was a significant predictor in model 3.

This means hypotheses 1 was also not supported for the sub-sample of Top Movies

As Figure 2 illustrates, the regression coefficient between NWActors and UserScore was statistically significant, as was the regression coefficient between UserScore and BoxOffice, MetacriticScore and Boxoffice, and also for NWActors and BoxOffice. The standardized indirect effects were -.00 were tested using bootstrap procedures and not significant.

The same analysis was performed with TopMovies as dependent variable, NWActors as independent variable, Userscore and Metacritic score as mediating variables.

Figure 2 Regression analysis with mediation using PROCESS with dependent variable Boxoffice(ln)

(33)

33 The results are visualized in

Figure 3. The total standardized indirect effects were -.017 and tested for significance using bootstrap procedures, these indirect effects were not significant either.

Figure 3 Regression analysis with mediation using PROCESS with dependent variable TopMovies

Hypotheses 2, 4 and 5 were not supported by the results for either sample Hypotheses 3 was supported.

(34)

34 Table 6 Hierarchical regression with Boxoffice(ln) as dependent variable

Model 1 Model 2 Model 3

Estimates B SE β p B SE β p B SE β p NWActors .226 .064 .124*** .000 .094 .109 .052** .002 .099 .085 .055 .242 Userscore .293 .107 .132*** .006 .137 .094 .062* .040 -.017 .074 -.007 .822 MetaCriticScore -.048 .009 -.250*** .000 -.011 .009 -.059*** .000 -.018 .007 -.092** .009 NonFiction -3.905 .385 -.441 .386 -2.720 .308 -.307*** .000 Suspense -1.404 .361 -.171 .146 -.275 .288 -.033 .340 Fantasy .815 .493 .070 .183 -.243 .395 -.021 .539 Drama -1.716 .320 -.237*** .000 -.979 .252 -.135*** .000 Action 1.715 .434 .181*** .000 -.208 .365 -.022 .570 NWActors*Nonfiction .183 .167 .051 .099 .170 .131 .047 .195 NWActors*Suspense .088 .209 .018*** .000 .038 .164 .008 .818 NWActors*Fantasy .316 .277 .048*** .000 .003 .218 .000 .989 NWActors*Drama .124 .156 .039 .273 .137 .122 .043 .262 NWActors*Action -.032 .185 -.009 .672 .045 .145 .012 .755 ProductionBudget 1.968E-08 .000 .348*** .000 InreleaseDays .015 .002 .272*** .000 Top Distributor 1.435 .167 .231*** .000 R2 _.050 _.294 _.570 Adjusted R2 _.046 _.282 _.561 F-value 13.4** 24.2** 62.6** Note. N = 774. * p < .05 ** p < .01 *** p < .001 (two-tailed)

(35)

35

Table 7 Hierarchical regression with TopMovies as dependent variable

Model 1 Model 2 Model 3

Estimates B SE β p B SE β p B SE β p

NWActors 5.72E+07 2.03E+07 .244** .006 6.70E+07 2.86E+07 .286* .021 5.57E+07 2.84E+07 .238 .052

Userscore -1.95E+07 3.18E+07 -.070 .540 -3.91E+07 3.24E+07 -.139 .230 -4.47E+07 3.19E+07 -.159 .164

MetaCriticScore 9.15E+06 2.88E+06 .358** .002 1.10E+07 2.95E+06 .431*** .000 1.04E+07 2.94E+06 .406** .001

Suspense -3.06E+08 2.09E+08 -.189 .147 -1.27E+08 2.13E+08 -.078 .553

Comedy -6.01E+07 8.90E+07 -.084 .501 1.82E+07 9.06E+07 .025 .841

Fantasy -9.92E+06 1.02E+08 -.012 .923 8.66E+06 9.92E+07 .011 .931

Drama -2.46E+08 1.42E+08 -.190 .086 -1.54E+08 1.41E+08 -.119 .279

NWActors*Suspense 6.89E+06 1.51E+08 .006 .964 1.50E+07 1.52E+08 .013 .922

NWActors*Comedy -9.09E+07 5.29E+07 -.196 .088 -9.64E+07 5.25E+07 -.208 .069

NWActors*Fantasy -4.10E+07 5.16E+07 -.100 .429 -4.38E+07 5.03E+07 -.106 .386

NWActors*Drama -1.53E+08 2.42E+08 -.060 .530 -1.29E+08 2.36E+08 -.051 .586

ProductionBudget 1.35E+00 4.93E-01 .278** .007

InreleaseDays 6.19E+05 6.17E+05 .091 .319

Top Distributor 2.57E+07 8.33E+07 .028 .758

R2

.152 .263 .326

Adjusted R2

.130 .187 .235

(36)

36

5 Discussion

Literature on predicting movie success has yet to find a conclusive answer to the following question: what are the ingredients for top-selling movies? As a result, scholars have been examining different possible determinants and are putting an increased effort into researching questions that combine different variables. Additionally, a number of Hollywood

blockbusters in recent years have shown a growing number of non-white lead actors and heroes despite the fact that it was previously a common belief in Hollywood that non-white leads would not sell, especially overseas. In that vein, the current research set out to

investigate whether ethnic diversity in movies influences the success of a movie. This is the first study that assesses whether the number of non-white actors impacts both critic and customer reviews as well as box office results, when controlled for different variables. The answers to the research question and sub-questions will be answered in the following paragraphs

5.1 The amount of non-white actors in Hollywood movies is not representative for US population

The data shows that the number of non-white actors is highly skewed and that the lead cast in Hollywood movies does not resemble the ethnic composition of the USA or the Netherlands for that matter. The data doesn’t provide information on the type of role portrayed, so underrepresentation in absolute lead roles is likely even higher, as the main character is in most cases white (Park et. al 2006).

This is a concern to many institutions and people alike, who advocate for a more balanced representation of minority groups in (mass) media. The study of Park (2006) on representation also shows that in a lot of cases the actors of color are stereotyped. Ahmed

(37)

37 (2012) studied the effects of gender stereotypes in soap operas and pointed out that regular viewers often overestimate how often certain behavior portrayed in soaps (e.g. adultery, murder) occurs in the real world. The combination of underrepresentation and perpetual stereotyping of a group of people in Hollywood movies, could in this case also lead to overestimation of certain behavior portrayed by racial stereotypes. This in turn can have negative psychological and societal consequences especially for minority groups discussed by many scholars in communication and media studies (Sue & Morishima, 1982; Ponterotto & Pedersen, 1993).

5.2 There was no significant effect found of ethnic diversity on box office sales

The current hypotheses were drawn up based on the idea that an increasingly diverse (global) audience would prefer movies with a more diverse casts that resembles their own image. The data suggests the number of non-white actors in a movie does not influence box office results when controlled for other variables like production budget, the number of days a movie was in theatres, and the size of the film distributor.

This contradicts the suggestions made by a number of studies, namely, the Hollywood Diversity Report and the CAA study which were conducted in search of the effect diversity in movies. Each of these studies concludes that diversity pays off in terms of box office sales. When examining the hierarchical regression models, the number of non-white actors present in a film significantly influenced the number of box office sales before adding the control variables. This could explain why many of these studies that failed to control for these variables have encountered this result.

When performing the same analysis on the sub-sample with only the top 119 movies that were responsible of 70% of the total box office, a different pattern emerged. The effect of the number of non-white actors could not be proven with a 95% confidence interval, but with

(38)

38 a p value of .052 this result suggests that this relationship should be examined further as this provides a hint that there might be an effect even though, it wasn’t proven within this

particular model. Moreover, the effects of other predictors changed significantly compared to the larger sample, since only production budget had a significant effect on ticket sales in the smaller sub-sample. It can be reasoned why the top distributor variable was no longer a predictor, as 99% of the top movies were from top distributors. The same goes for in release days, there was not enough spread in this variable in the top movie category.

These results suggest that when looking at the big bulk of Hollywood movies, top distributors and in release days are predictors of success, however for the most successful movies there are other factors not included in this study, that predict a lot of the variance in box office. Diversity might be one of them, but that was not proven in this research.

The field of research specifically catered to the possible financial gains of diversity is relatively new. Diversity in movie hasn’t been studied that often, even more, diversity was considered a factor that could potentially cost revenues.

This study also proves is that the relationship between ethnic diversity and box office sales is not a simple equation that can be applied to the full spectrum of movies. It seems that there is not a one size fits all model to analyze Hollywood movies with diversity as subject of research. It could be inferred that a categorization based on box office sales should be made to draw conclusions that would give movie makers more decision making information and provide researchers with more insights into these dynamics.

Blockbusters like Black Panther, Star Wars and The Fast and Furious series are often named as proof that diversity does sell (Hollywood Diversity report, 2017). But these were either sequels or comic based movies or both, and had a lot of starpower in their cast. These factors have proven as significant contributors to box office in other studies and are often seen as safe choices for movie makers that limit the financial risks. Also news about movies

(39)

39 with non-white lead actors that exceeded their predictions by far like ‘Get Out’ and ‘Hidden Figures’ is covered in a lot of media, but these are more outliers than the norm. Another possible explanation that no significant relation was found, could be the lack of spread in the number of non-white actors in movies. 47% of the movies consisted of an all-white lead cast and 37% had 1 or 2 non-white actors. When there is such a low level of diversity overall, it’s hard to measure the effect compared to the big group of all-white casts.

The issue could also evolve around the types of roles that the non-white actors play. Social identity theory as an explanation why a diverse audience would prefer a more diverse cast, relies on an implicit assumption that the actors they are watching portray roles that they want to identify with. As mentioned before, racial stereotyping is still very common in movies and an easy way for movie makers to establish a character in a movie without spending to much time on developing the character. There are more negative racial

stereotypes in movies than there are positive ones (Park et al., 2006) If non-white characters portray roles that audiences don’t necessarily want to identify with, this could have the opposite effect.

It is also a possibility that our hypothesis on the race of an actor being a factor at all for audience members overestimates the role of race in movie choice. Race does play a role in our everyday lives, especially when one belongs to a minority group (Phinney, 1992). But it might not prevail over other features of a movie like the quality of the screenwriting, stardom or the plot in making the decision to go to the movies. These variables are less easy to quantify and the existing literature doesn’t offer a model which incorporates all of them.

5.3 Genre doesn’t influence the relation between diversity and box office

None of the genres interacted significantly with the relationship between non-white actors and box office results. It could be that the effect between the main independent variable

(40)

non-40 white actors and box office was too small in this study to show an interaction effect with the different genres. It is however, interesting to note that genres independently do influence the box office results. Drama and Nonfiction movies influence the box office sales negatively. The negative effect of drama is in line with the findings of Prag and Casavant (1994). This might be explained by the fact that these type of movies are less often very cinematic experiences in terms of visual effects, and can be enjoyed for its emotional or educational purposes outside the cinema’s through streaming, TV or DVD’s for example. In this study the nonfiction genre is a combination of documentaries and foreign movies. These movies

usually target a more niche consumer group, smaller in size and thus reasonably return lower box office sales.

5.4 The impact of diversity on reviews was not proven

It was also examined whether critics and consumer reviews were influenced by the number of non-white actors in a movie. There was only a modest negative though a significant effect found of the number of non-white actors on user reviews when considering the full sample of 774 movies. This effect should be further investigated, a possible explanation could be that some of the most diverse movies have in fact a lower budget and therefore maybe lack in other determinants for a good movie, such as starpower or visual effects.

A second explanation goes back to the theory. A positive effect was hypothesized based on social identity theory, however, it could be that the review population is not ethnically representative of the actual movie going population. The expected effect of diversity would not be seen in the consumer reviews in that case.

In the top movie sample there was no significant effect found on both consumer and critic reviews. When looked at the direct effect of reviews on box office success, a significant positive effect was found between critic reviews and user reviews on box office result when

(41)

41 looking at the total sample, this effect was not significant when examined in the sub-sample of the top performing movies.

Again this could be explained by the fact that the top movies other determinants they depend on for success.

5.5 Theoretical implications

The body of academic research on predicting box office sales is still maturing and this thesis adds to the literature by examining diversity as a determinant. It is the first study research this while controlling for some important predictors and looking at reviews as well it.

By incorporating social identity theory and uses and gratifications theory and economic forecasting models this research combined media psychology, communication theories and marketing theory.

Using a sample of this size is also a strength of this paper. It provides directions to other scholars to start examining diversity as one of the possible determinants and shows that the perceived effect in a lot of studies is probably due to other factors like production budget. The top movies had a higher diversity rate and the biggest blockbusters all showed high diversity rates. This was not the case for the majority of the movies however and these widely known successes skew the data as well as public opinion.

The research points out the number of non-white actors alone is not able to

significantly predict changes in box office results. Where recent studies have claimed that diversity does pay off, this study shows that diversity should always be studied in a model that controls for a couple of main predictors, otherwise it’s likely to produce results that give a distorted view on the nature of the link between diversity and sales.

(42)

42

5.6 Managerial implications

This study had both an academic goal as well as a managerial perspective to consider. Movie making will always be a financially risky business. Trying to create a perfect movie that will produce the biggest financial gains will be a never ending quest, as the perfect movie doesn’t exist. However, the western society is changing in its ethnic composition and populations from emerging markets are increasingly enjoying more money and leisure time to spend. The most important implication of this study is that even though this research hasn’t proven that diversity significantly increases box office, it does show that diversity doesn’t hurt sales either. All the correlations were positive and significant before adding the control variables. Decision makers in the move industry tend to be highly risk-averse, and these findings lower the perceived risk of hiring non-white actors.

This should strengthen decision makers in the movie industry to explore diversity in their movies and put the financial objections of not hiring non-white actors to rest for good.

5.7 Limitations and recommendation for further research

One of the limitations of this research was the fact that most data came from two major websites in the Hollywood movie industry. Ideally one would take data from multiple

websites and aggregate them in a manner that would give an average of the total industry and where one could control for consumer reviews that are representative for society.

The large sample size is a strength of this study. However, a caveat of using a large sample size in a time boxed period, is that it doesn’t allow for a lot of qualitative in depth research into the actors and the movies. Future research should look into the types of roles that non-white actors are portraying and the way this influences the audience and box office sales.

A third recommendation would be to examine how an intelligent categorization of movies of different magnitudes can be made to study the effects of diversity. The analysis of

(43)

43 the top movies suggests that the role of cast diversity is different for the most popular

mainstream movies when compared to the overall sample of Hollywood movies released. Another interesting perspective would be to examine determinants in relation to the total return on investment of a movie. This could be valuable insight as the highest grossing movies also have the highest production costs and because streaming has been a significant source of income in the movie industry that is not measured by box office sales.

The last recommendation for further scholars is to incorporate diversity behind the camera into their research as well. In this study the focus was solely on the actors on screen, but on screen portrayal of ethnically diverse actors is influenced by the composition of movie makers as well.

6 Conclusion

Despite the above-mentioned limitations, this study provides novel insight into diversity in Hollywood movies and its effects on box office. It advances the expanding body of

theoretical and empirical work by demonstrating that diversity is not directly affecting box office sales negatively, as some Hollywood decision makers believe. This complex

relationship should be explored more in depth. The results of these studies should encourage movie makers to be more explorative, as the perceived risks discussed above seem

unjustifiable in light of the data. The current research intends to promote extensive analysis of the financial impact of unconventionally diverse casts. Such efforts would deepen our understanding of the mechanisms at play.

Ultimately, the findings of this study can aid researchers in identifying these opportunities and providing more support for movie makers as well as agents of change within the industry. Ethically speaking, representation does matter. Three years after the

(44)

44 #oscarssowhitehashtag on twitter, a touching hashtag concerning the topic of diversity in movies emerged: #whatblackpanthermeanstome. Through use of this new hashtag, people (mostly black) shared pictures and quotes about the impact of witnessing black heroes in a movie fully capable of leading their own narratives. Every child, and adult for that matter, deserves the opportunity learn from heroes that look like them. This alone should be ample incentive for movie makers to employ culturally diverse actors in their movies. So even when it’s not clear how much diversity influences financial success, there is a lot to win. Maya Angelou once said “In diversity there is beauty and there is strength”. Who wouldn’t want that for their movie?