The Impact of Race and Gender in Film Casting on Box Office Revenue. Will Burchard. University of Oregon. Economics 525 Research Proposal.

Transcription

1 The Impact of Race and Gender in Film Casting on Box Office Revenue Will Burchard University of Oregon Economics 525 Research Proposal Spring 2012 Abstract: We test whether or not race and gender affect box office revenue of wide-release films in the United States. We explain box office revenue using data on production budget, critical rating, number of theaters at which a film is playing, genre, award nominations, MPAA rating, and whether a film is a sequel. While these measures have been used to study box office revenue in previous research, we include race and gender of film cast as explanatory variables. We consider the order in which a cast is credited and analyze the race and gender of the first through fourth billed actors. Our aim is to encourage discussion and share knowledge on how media spending is related to race and gender and how cultural norms are related to media spending. 1

2 Project Summary: The film industry in the United States is influential both economically and culturally. A natural metric that is tracked closely is that of box office revenue. That is, the money consumers spend to see a particular film in theaters in the United States. In this project, we will study the relationship between box office revenue for films in the United States and the race and gender of the cast in a film. In addition, we aim to encourage discussion and thought on how cultural norms are related to media and how consumers decide to spend money on media-related goods. This project is motivated by the history of bias in the United States film industry in terms of race and gender. While biases were much more prominent in the early days of film, to detect anything of the sort in modern films requires more subtle techniques and considerations. Race and gender have been analyzed in fields of academia such as film studies and gender studies, but rarely with a concrete quantitative backing. We will provide this quantitative backing that will allow the plausible theories already set forth in academia to be tested and applied to modern film. To analyze race and gender in the film industry, we consider the measure of box office revenue. Our research will be confined to box office revenue earned in the United States and will consider wide-release films only (films screened in 600 theaters or more at peak). We consider films screened between the years of 2004 and 2011, which requires the inclusion of about 1000 observations. We find data are readily available on common explanatory variables associated with box office revenue. Included measures are production budget, critical rating, number of theaters at which a film is playing, genre, award nominations, MPAA rating, and whether a film is a sequel. In addition to these common explanatory variables, we add dummy categories for race and gender. We consider analysis of the first four credited actors of the cast. This allows for analysis on demographic make-up of cast through various interaction terms. That is, we can analyze how casting just a nonwhite lead actor in a film influences box office revenue, or we can analyze how a cast that is mostly nonwhite is related to box office revenue. As shown in (Brewer et al. 2009), these data can be analyzed using OLS methods. Our classical assumptions hold, which allows for simple, intuitive analysis. We correct for nonlinearities in box office revenue and production budget, and expect no problems with endogeneity. Inflation is adjusted for using a base year of 2004, and time fixed effects are implemented based on year. This research has the potential to influence academic fields including economics, film studies, and gender and race studies. It will contribute to the existing economics literature on race and gender bias, and will test theories already set forth in other academic fields. In addition to benefiting those in academia, this research will encourage discussion and thought on the relationship between cultural norms and the media. 2

3 The United States Film Industry: The history of the film industry in the United States is marked by certain gender and race roles that are possibly still present today. Many of these biases date back to the development of the United States film industry in the 1910s and 1920s. As (Dixon and Foster) present in A Short History of Film, an extreme example of racial bias in film form is D.W. Griffith s The Birth of a Nation (1915). The Birth of a Nation was, and still is, surrounded by much controversy in that the film revolves around the Ku Klux Klan rescuing a woman from a gang of African-Americans. This film sparked a revival of the Ku Klux Klan in said time period, and created a further divide between blacks and whites of the time. Another early and prevalent example of prejudice is that of blackface, seen in early films like The Jazz Singer (1927), in which white actors painted their faces black to imitate black actors. While we do not see such extreme racial prejudices on film today, as (Benshoff and Griffin) note in America on Film, culture has been shaped such that certain perceptions are triggered each time we watch a film. An example highlighted in America on Film is that of modern romantic comedies in which male and female characters are involved in some sort of comedic romantic relationship. One such film, You ve Got Mail (1998), involves a relationship between two white actors, Tom Hanks and Meg Ryan. Another film that has a very similar premise, Two Can Play That Game (2001), stars two black actors, Morris Chestnut and Vivica A. Fox. While both films follow the same plot line, Two Can Play That Game is considered a black film, while You ve Got Mail is considered a romantic comedy (as opposed to a white film ). 3

4 Just as nonwhites were misrepresented in early American film, so were females. Much of this stemmed from the types of jobs women and men held in the film industry. It was the norm that in the 1930s-1950s, women were not considered for financial, studiohead, director, or producer positions in films. It was also the case that actresses in the early film era played characters that were non-dynamic, and were actually filmed in different styles than men. These styles include the use of certain camera shots that force the viewer to focus on the actress entire body, rather than the conversation or dialogue in consideration. Often, shots of men were filmed such that men were looking at something, while shots of women were filmed such that women were being looked at. Notable examples of films, as analyzed in America on Film, include Footlight Parade (1933), in which a group of chorus girls are objectified in a choreographed scene as part of a male fantasy. This film was directed by Busby Berkley, who was famous for these choreographed scenes in which women simply moved around stage and were put on display for the audience. An example that combines both race and gender bias is in Gilda (1946), starring Rita Hayworth. Hayworth was an actress of Latina heritage. In Gilda, the filmmakers dyed her hair and raised her hairline to hide her Latina heritage. Whenever Hayworth is seen in a close-up shot in Gilda, it is interjected between shots of men looking at her. This makes her appear as the object of attention. Gilda demonstrates the early film industry s tendency to objectify women and also suppress racial diversity. While these issues in the American Film Industry have become less prominent over time, it is certainly possible that there are still certain issues regarding race and gender in the United States film industry. These issues will be the focus of our research. Box office revenue, a readily available metric that tracks how much money a film makes at the box 4

5 office, will be our tool of study. Our hypothesis is that if race or gender plays a factor in film production and cultural norms, then this will be reflected in box office revenue. We will analyze this hypothesis by examining the demographics of film casts. It is also possible that certain films that target a specific audience earn more box office revenue out of loyalty from said audience. That is, niche audiences influence the success of a film. It is possible that certain racial demographics are more likely to watch films that star certain actors or actresses that they can relate to. Existing Literature: While there is research and literature available on film roles and how they have evolved throughout history, there is little in the realm of quantitative research involving gender and race. Existing research on race and gender in films is limited to academic fields like film studies and gender studies. However, there have been economic studies conducted on the general make-up of the film industry and the correct usage of explanatory variables that can be expected to influence box office revenue. (Brewer et al. 2009) find that an ex post OLS regression model on box office revenue yields significant explanatory variables that include budget, peak number of screens for a film, sequels, critical reviews, seasonal releases, word-of-mouth, award nominations, and star power. This is relevant research in that the OLS methods used are similar to those we shall use. Brewer et al. also make a distinction between ex ante and ex post regression models. That is, information used to measure revenue in the ex ante model includes that available to the public prior to film release, and information used to measure revenue in 5

6 the ex post model includes information available to the public before and after film release. For the purposes of our study, we will only make use of the ex post model. (Reinstent and Snyder 2005) find that positive film reviews have a positive impact on box office revenue, most notably in the genres of drama and narrowly-released movies. While we do not consider narrow-release films in this study, it is expected that positive film reviews have a positive impact on box office revenue for drama and other categories considered in our research. (Eliashberg and Shugan 1997) also show that critic reviews positively affect box office revenue, but that they do so mostly for the later box office receipts. That is, initially, the reviews do not greatly affect box office receipts. This is expected as it takes time for critical reviews to circulate and impact an audience. (Einav 2009) finds that movie releases are largely clustered around holidays, and that this provides incentive for distributors to alter the release of their films by a few weeks either before or after the holiday. Although we are not concerned with exact timing of film release in our study, this result does provide evidence that it is necessary to include a control for seasonal releases. (Fowdur et al. 2009) find that ratings for a film with a black lead actor are about 6% lower than all other ratings, controlling for various measures like film production costs, genre, MPAA rating, etc. This study hints that there is some sort of racial bias in the film industry, and provides further motivation for the analysis we propose. (Ravid 1999) studies the impact of star power on box office revenue. Simple means data show that star power appears to be higher in films that have higher box office revenue. However, Ravid also finds that any big investment increases film revenue. For this reason, we are confident in our hypothesis (as noted in the data section) that star 6

7 power is already accounted for in the production budget. In addition to work on star power, Ravid also finds that sequels and film accessibility (how easy it is to see a film) positively affect box office revenue. Although no research exists on race and gender in the film industry, studies exist that analyze racial bias in the pricing of baseball and basketball cards. (Stone and Warren 1999) find that there is no evidence of racial bias in the market for basketball cards. That is, customers are willing to pay the same amount for basketball cards regardless of player race, controlling elsewhere. (McGarrity et al. 1999) correct prior findings that consumers are willing to pay different prices for baseball cards based on the race of players. Previous studies find that race is a factor in the price of baseball cards, but after making model adjustments, McGarrity et al. find that this is not an acceptable conclusion, and that race does not, in fact, play a factor in baseball card pricing. While there is no direct parallel between the price of trading cards and box office revenue, it is noteworthy that literature exists on the impact of race on pricing. Ultimately, research conducted on the film industry thus far has dealt with the prediction of film releases and how they will perform at the box office. These analyses use a common set of explanatory variables like production budget, critical rating, MPAA rating, awards nominations, and other film characteristics. None of these studies, however, attempt to analyze what impact race and gender may have on box office revenue. Our research will add measures of race and gender to the already standard set of explanatory variables for box office revenue. These measures will be used to analyze how race and gender affect box office revenue, and how cultural norms influence what movies consumers choose to attend. 7

8 Available data: We focus data analysis on the metric of box office revenue in the United States between the years 2004 and Box office revenue is tracked by (The Numbers). Such an analysis can pose difficulties if unrefined in that lesser-known films are less likely to have data available for box office revenue analysis. In addition, data on film-specific traits that can be used as natural explanatory variables are not available for lesser-known films. The apparent issue here is that data are available for well-known films but not lesserknown films. To address this issue, we use a well-defined film industry measure to target a specific group of films. As defined by the film industry, a wide-release film is one that, at its peak, screened in at least 600 theaters across the United States. We embrace this definition and consider only wide-release films in this research. This allows for collection of data on nearly all films and allows for the consideration of roughly 1000 films. It is also a natural data set to study for the proposed research, given that higher production budgets, and thus advertising budgets, are associated with wide-release films. A limitation of United States box office revenue data is that it is not available on a disaggregated scale. That is, it cannot be separated by city, county or state. The reason for this is that revenue is reported to tracking companies through film studios (Box Office Mojo). Thus, obtaining disaggregated data would mean the film studios would have to report disaggregated data to the tracking companies, which, for unknown reasons, does not happen. Box office revenue can be explained by traits also represented by available data. The most natural explanation of box office revenue is the production budget associate with the film. Production budget encompasses special effects associated with the film, star 8

9 power, director notoriety, and possibly studio predictions on popularity of the film. As long as we are not concerned with these individual measures, production budget acts as a neat and concise measure of how much a studio is willing to invest in a film. Given that our concern is in the race and gender aspects of film, it is acceptable to use production budget as an explanatory variable in this way. Considering only wide-release films requires a measure of maximum theaters screened at a film s peak screening. These data are also available from (The Numbers, Box Office Mojo). This measure allows for the separation of wide-release films from non-widerelease, but also acts as another natural control variable for box office revenue. The range of this measure is from 625 to 4393 theaters. The number 600 may seem arbitrary, but the data suggest this is a natural cutoff for the films we consider. Critical rating as an explanation for box office revenue is included as an average of available ratings compiled from magazine and website ratings. (Rotten Tomatoes) provides these data in a percentage format. Lower percentage ratings are less favorable and higher percentage ratings represent films that critics reviewed positively. It is expected that films with higher critical rating earn more box office revenue, controlling elsewhere. MPAA ratings, another readily available metric, reflect the accessibility of films, and are thus expected to impact box office revenues for films. MPAA ratings enter the study as a group of dummy variables representing ratings G, PG, PG-13, and R-rated films. G-rated films are most accessible and R-rated films are least accessible, based on the maturity level of the content in the film. 9

10 Certain genres can be expected to influence expected box office revenue as well. The major genres that we use to explain box office revenue in this research are: action, adventure, drama, horror, comedy, thriller, and other. Other is a sort of novelty category including musicals and films that do not fit into the main categories already mentioned. There are few films that fall into the other category. This is an ambiguous measure in that it is difficult to predict which categories will influence box office revenue and in which direction. Many modern films are either adapted from a book, television series, or previous story. Also, many films are simply sequels to films previously released. Data on these traits are easily found. Whether a film is a sequel or adaptation can be expected to positively influence box office revenue, since a previous fan base has been built. Data on award nominations are available and are included as a dummy category. The award nominations in consideration are those for the Academy Award for Best Picture. The correlation between award nomination and box office revenue is expected to be positive. The explanatory variables mentioned thus far have been considered in previous studies, but the main focus of this study is to analyze the effects of race and gender on box office revenue. This analysis involves data on the race and gender of the casts of films in consideration. The actors credited on each film are done in billings. That is, the first billed actor (also the lead ), is the actor most highly associated with the film. His or her name is the first listed in the cast credits. After the first billed actor come the second billed, third billed, and so on. To conduct a thorough analysis, we include the first through fourth billed actors. This data is obtained at (IMDB, The Numbers). For each billed actor, we include a 10

11 dummy variable for gender (fem) and race (nonwhite). For example, we can include dummy variables for gender of first billed, second billed, third billed, fourth billed, and race of first billed, second billed, third billed, and fourth billed. All explanatory variables noted have available associated data. However, there a large amount of pre-processing is necessary to compile a complete data set. The need for pre-processing comes from the data being scattered across various websites. Also, actor billings must be done individually. This is tedious in that each actor must be thoroughly researched in order to determine race and gender. Model and methodology: As (Brewer et al. 2009) find OLS to be an appropriate method for conducting their analysis on ex post factors affecting box office revenue, we also find this to be an appropriate method. Brewer et al. find that the data in question does not suffer from heteroskedasticity and that all models pass the normality tests necessary to make use of the classical ordinary least squares method. Since our data are similar to those used in (Brewer et al. 2009), we expect to obtain the same results regarding model specification. This leads us to analysis of the following model: (1) ln(rev)=δ1(fem)+δ2(nonw)+β1ln(prod)+β2(thea)+β3(crit)+δ3(mpaa) +δ4(genre)+δ5(award)+δ6(seq)+δ7(adap)+δ8(year)+e Note that this model is simplified from the data described above. The female and nonwhite variables here encompass the first through fourth billed actors, the MPAA and 11

12 genre variables represent the entirety of their categories, and the year variable represents all years in question. This model is specified such that revenue and production budget are logged. This is appropriate given the trend of the revenue and production budget data. That is, revenue and production budget appear to increase at an increasing rate. Fixed effects are included for year, which account for variance in prices from year to year. In addition, revenue and production budget are adjusted for inflation using 2006 as a base year. Dummy variables include MPAA rating, genre, award, sequel, adaptation, year, female, and nonwhite. Thus, we can include many intuitive interactions in our analysis, such as interactions between the gender and race categories, gender and genre categories, race and genre categories, race and critical ratings categories, and gender and critical ratings categories. Our model also allows for a noteworthy interaction in that we can interact the first billed, second billed, third billed, and fourth billed actor. This allows for rigorous analysis of the gender and racial make-up of a cast. For example, we can analyze the effect on box office revenue of a film having a first and second billed nonwhite actor versus films that do not have such actors. This allows for analysis on how a film with just a first billed nonwhite or female actor differs from a film with a cast composed mainly of female and nonwhite actors. In addition, we can analyze how a cast made up entirely of nonwhite actors affects box office revenue. It should be noted that we can include even more billings in our model, but in general, the cast composition is reflected in the first few billed actors. 12

13 Potential Reach: This research is geared first towards those who have interactions with the film industry and study of the film industry. Our study provides a benchmark for analysis on the relation of race and gender to box office revenue of films in the United States. This will allow economists to study the film industry further, and will also benefit those in other academic disciplines like film studies, gender studies, and racial studies. Often it seems that these areas of study have excellent theory and literature that would be greatly enhanced by quantitative analysis. Beyond this, research on film and culture has a broader impact on the way people absorb media. Spreading knowledge on how we perceive certain representations in media can prepare us for scrutinizing any misrepresentations that may take place. While obvious misrepresentations can be made by a glancing eye, it takes quantitative analysis to uncover the more subtle impacts that race and gender may have on parallels between media and culture. This research will best fit in a journal geared towards media and culture. A few journals that have published research on film and other media include The Journal of Cultural Economics, Applied Economics (Taylor and Francis), the The Journal of Media Economics, and The Journal of Industrial Economics. Many of the references in this proposal were published in these journals (as noted in reference section). The film industry is integrated in the United States both economically and culturally. Any impact of gender and race on the industry s revenue, or any impact of the industry s production methods on the shaping of cultural norms warrants an in-depth analysis. While literature exists on the historical development of the film industry and how race and 13

14 gender impacted this history, in-depth quantitative analysis has not been done on this matter. Using box office revenue to model this impact, it is possible to shed light on an area that has not been previously explored. This new information will encourage debate and further study on how media shape our culture and how cultural norms are reflected in media spending. 14

15 References Benshoff, Harry M., and Sean Griffin. America on Film: Representing Race, Class, Gender, and Sexuality at the Movies. Malden, MA, USA: Wiley-Blackwell, Print. "Box Office Mojo." Box Office Mojo. Web. 28 May < Brewer, Stephanie, Jason Kelley, and James Jozefowicz. "A Blueprint for Success in the US Film Industry." Applied Economics 41.5 (2009): Print. Dixon, Wheeler W., and Gwendolyn A. Foster. A Short History of Film. New Brunswick: Rutgers UP, Web. Eliashberg, Jehoshua, and Steven M. Shugan. "Film Critics: Influencers or Predictors." Journal of Marketing 61.2 (1997): Web. 28 May Einav, Liran. "Not All Rivals Look Alike: Estimating An Equilibrium Model Of The Release Date Timing Game." Economic Inquiry 48.2 (2009): Print. Fowdur, Lona, Vrinda Kadiyali, and Jeffrey Prince. Racial Bias in Expert Quality Assessment: A Study of Newspaper Movie Reviews. Diss. Cornell University, Web. 28 May IMDb. IMDb.com. Web. 28 May < McGarrity, Joseph, Harvey D. Palmer, and Marc Poitras. "Consumer Racial Discrimination: A Reassessment Of The Market For Baseball Cards." Journal Of Labor Research 20.2 (1999): EconLit with Full Text. Web. 28 May The Numbers. Web. 28 May < Ravid, S. Abraham. "Information, Blockbusters, and Stars: A Study of the Film Industry." The Journal of Business 72.4 (1999): Web. 28 May Reinstein, David A., and Christopher M. Snyder. "The Influence Of Expert Reviews On Consumer Demand For Experience Goods: A Case Study Of Movie Critics*." Journal of Industrial Economics 53.1 (2005): Web. 28 May Rotten Tomatoes. Web. 31 May < Stone, Eric W., and Ronald S. Warren. "Customer Discrimination in Professional Basketball: Evidence from the Trading-card Market." Applied Economics 31.6 (1999): Web. 28 May