Women and Minority Film Directors in Hollywood: Performance Implications of Product Development and Distribution Biases

This article examines the performance of 2,386 motion pictures released in the United States between 1994 and 2016, for which domestic box office revenue, production budget, consumer review, and distribution data are available. Women directed only 4.95% of these films, and people of color (POC) directed 10.52% of them. A more comprehensive examination of the IMDb data set of more than 4,000 films released domestically and internationally over the same period reveals that women directed only 4.99% of all widely released U.S. films, while POC directed 12.35% of all movies and 11.8% of all widely released films.

Perhaps more surprisingly, while we might expect an ongoing improvement in gender- and race-related imbalances over time, evidence indicates the opposite. As we demonstrate, the proportion of widely released U.S. movies directed by women and POC actually decreased during the period considered in this study (Figure 1). This lack of improvement is noteworthy because gender and race participation in the creative arts is robust. Thus, highly qualified female and POC candidates for leadership positions in the movie industry should not be rare. Yet industry sources often reflect common views that female and minority directors pose certain risks. For example, Bielby and Bielby (1996) cite industry views that women cannot “open” a movie and that studios are uneasy about trusting a woman with an action movie or a big-budget project. Moreover, studios falsely believe that women are not interested in directing large-scale films (Anderson 2015). Similar biases against minority directors manifest in a lack of industry support, isolation, and even blatant discrimination (Ugwu 2019).

OPEN IN VIEWER

Figure 1.

Percentage of Movies Directed by Women and POC over Time.

On the surface, there appears to be some justification for these biases against female and minority directors. The average female-directed movie in our sample achieved an inflation-adjusted gross domestic box office revenue of $31.3 million, whereas the average male-directed film realized $43 million, a difference of 37%. Similarly, the average POC-directed movie had an inflation-adjusted gross domestic revenue of $30.9 million compared with $43.9 million for nonminority directors, a difference of 42%. However, there is ample reason to believe that these differences in performance may be due in large part to selection bias on the part of Hollywood studios. To illustrate, Figure 2 provides data on average domestic box office revenue as a function of movie genre, Motion Picture Association (MPA) rating category, and whether the movie is a sequel/prequel or remake, with categories arranged in descending order of box office performance. It also shows the percentage of female (Panel A) and minority (Panel B) directors associated with each type of film. As is evident, female and minority directors are selected disproportionately for movies in specific lower-grossing categories and are chosen far less often for high-grossing categories, including sequels, animation, adventure, and action films.

OPEN IN VIEWER

Figure 2.

Movie Versus Director Characteristics.

What Figure 2 does not show is the potential bias within each category in the way high- and low-potential projects are assigned and in the way decisions related to production budgets and distribution are made for movies from directors of different genders and races. Our goal in this study is to characterize the extent of these biases and their effect on the performance of female-directed and POC-directed movies in the marketplace. To do so, we model (1) the assignment of women and minority directors to projects with different revenue potential, (2) the production budgets given to films directed by these directors, and (3) the number of screens secured for movies from these directors. We also examine the extent to which the degree of bias these directors experience changes as a function of (1) the gender of the lead actor in the film, (2) the race of the lead actor in the film, (3) how much clout the director has in the industry, as indicated by their box office revenue history, and (4) whether the director is international. We define minority directors as directors of Black/African American, Asian, Indian, Hispanic/Latinx, Native American, and Pacific Islander ancestry.

The literature on motion picture performance in marketing has focused primarily on improving the accuracy of box office forecasts (Eliashberg et al. 2000; Neelamegham and Chintagunta 1999; Sawhney and Eliashberg 1996). It has also examined specific issues related to movie performance and distribution, including the importance of star buzz (Karniouchina 2011), critical reviews and word of mouth (Basuroy, Chatterjee, and Ravid 2003; Basuroy, Desai, and Talukdar 2006; Eliashberg and Shugan 1997; Liu 2006), box office premiums associated with sequels (Basuroy and Chatterjee 2008; Hennig-Thurau, Houston, and Heitjans 2009), lead–lag effects and endogeneity between distribution intensity and revenues (Elberse and Eliashberg 2003; Krider et al. 2005), movie screen optimization (Swami, Eliashberg, and Weinberg 1999), and cross-cultural differences (Song et al. 2018).

Our study adds to this literature by examining two factors that appear on their surface to strongly influence consumer response to a particular film: director gender and race. The central issue is whether the observed shortfall in revenue performance of female- and minority-led projects is due to consumer preferences for the offerings of male, nonminority directors (or perhaps even for the directors themselves) or supply-side biases in project assignment and support on the part of Hollywood studios and exhibitors. This support includes two categories of special relevance to marketing: budgeting to support product development/production and distribution of the film.

While the practices of Hollywood studios seem to indicate that director gender and race are important determinants of movie performance, the impact of director identity on box office revenue has received relatively little attention in empirical research. Directors are frequently grouped with actors to capture the overall star power of the film or are omitted from the analysis altogether. Often, the subsample of women and minorities is so small that researchers face difficulty studying the gender and race of the director as relevant factors. For example, in the case of female directors, Elliott and Simmons (2008) analyze a sample of 527 movies and code actors and directors into star and nonstar categories. Using industry lists, Julia Roberts was the only woman categorized as a star in the actor category. Citing a lack of industry lists in the director category, the authors imposed their own categorization based on the director's ability to attract publicity, which resulted in no women or POC in the star director category (Robert Altman, Stanley Kubrick, George Lucas, Michael Moore, Guy Ritchie, Martin Scorsese, Steven Soderbergh, and Steven Spielberg were identified as star directors). The resulting variable proved insignificant and was later removed from the analysis. In this study, we use a broad sample of movies to ensure that we have a sizable group of women and minority directors. We first present our analysis with women and minority directors combined to maximize sample size and then go on to show that the results hold for each group.

Interest in gender and racial imbalances among directors in the motion picture industry reflect a more general interest in the acceptance and performance of women and POC as leaders (Helfat, Harris, and Wolfson 2006; Smith et al. 2019). A relatively small body of literature has examined the relationship between female leadership and organizational performance. These include studies of the performance of women-owned small businesses (e.g., Du Rietz and Henrekson 2000; Fasci and Valdez 1998; Loscocco et al. 1991; Robb and Watson 2012; Rosa, Carter, and Hamilton 1996; Swinney, Runyan, and Huddleston 2006), the effects of female board members (e.g., Adams and Ferreira 2009; Carter, Simkins, and Simpson 2003; Erhardt, Werbel, and Shrader 2003; Farrell and Hersch 2005; Krishnan and Parsons 2008; Rose 2007; Smith, Smith, and Verner 2006), and the impact of female chief executive officers (CEOs) (e.g., Gondhalekar and Dalmia 2007; Khan and Vieito 2013; Lam, McGuinness, and Vieito 2013; Lee and James 2007; Martin, Nishikawa, and Williams 2009; Wolfers 2006). Surprisingly, studies of minority leaders are limited and often focus on entrepreneurship and small to medium-sized businesses. They usually point to structural barriers and the need to create support ecosystems to mitigate impediments (Avery and Rendall 2002; Bates 1997; Bates, Jackson, and Johnson 2007; Boston and Boston 2007; Fairlie and Robb 2007). As a whole, these studies often highlight a lack of resource parity but yield little agreement on the overall effect of female or minority leadership.

The present study contributes to this broader literature in several ways. First, we examine vital factors that explain observed differences in performance between female and male leaders and between minority and nonminority leaders that cannot be observed as easily in other contexts. In particular, we examine biases in project assignment and support in the form of production budgets and distribution. While extant research has modeled the selection of leader gender and race to varying degrees, no study has offered insight into the role of these crucial factors in generating the ecologically observed effects of women and minority leaders.

Second, motion picture performance ultimately rests on acceptance and support from studio executives, channel partners, and consumers. Unlike most board members, CEOs, or entrepreneurs, it is possible that audiences are attuned to director identity. This adds an additional dimension to the analysis in that biases may reflect both studio expectations about the capabilities of female/minority directors as well as their assumptions about audience acceptance of female/minority directors. To the extent that audiences are less accepting of female- and minority-directed movies, this would provide a basis to rationalize studio biases under the premise that such biases reflect externally imposed constraints on female/minority director performance. We provide some exploratory evidence that suggests that consumers do not pay much attention to director identity. Therefore, studio decisions appear to reflect their own biases rather than mirror those of audiences.

Third, we showcase important sources of heterogeneity in the biases that female and minority leaders face. In particular, we examine how the degree of bias changes as a function of the gender of other key members of the project, the race of other key members of the project, the extent to which the leader possesses clout in the industry, and whether the leader is international or from the United States. In the first two instances, we focus on the gender and race of the lead actor, arguably the most important individual to the success or failure of the production. In the case of clout, we investigate whether high-clout female and minority directors are subject to less discrimination or whether they are subject to more bias compared with high-clout male, nonminority counterparts.

Our estimation consists of a two-stage least squares model in which movie revenue, budget, and screens are predicted by director gender, director race, other endogenous variables, and several control variables in the second stage. In the first stage, we model the selection of director gender and race and introduce instrumental variables to address the endogeneity of revenue, budget, screens, and audience reviews. A simple representation of the conceptual model appears in Figure 3.

OPEN IN VIEWER

Figure 3.

Primary Conceptual Relationships.

The results provide evidence of substantial biases in project assignment, budgeting, and distribution in favor of male, nonminority directors. Once the analysis accounts for endogeneity, the remaining underperformance associated with female and minority directors is small and statistically insignificant. Rather than female and POC director underperformance, we find clear evidence of a bias toward male, nonminority directors such that the revenue potential of the project, the production budget, and screens all increase as the probability of a male, nonminority director increases. Our examination of heterogeneity reveals that biases against female and minority directors are generally stronger when the lead actor is female, when the lead actor is a minority, when the director is from the United States, and among low-clout directors. The last effect suggests that biases dissipate to a degree as female and minority directors gain experience in the industry. Interestingly, however, this does not occur for budgeting decisions, for which the lack of trust appears to be more pervasive.

Our conclusions from the econometric model are supported with matched-sample analyses for both women and minority directors. For female directors, the closest movie in the male-directed sample released within two calendar years of the focal film had a significantly higher budget. Yet the differences in revenues between these samples are not significant. For minority directors, we find no significant difference in budget favoring minority directors, but we do find a moderately significant difference in revenue (p-value = .078).

We organize the rest of this article as follows: First, we present our research questions. Second, we review background literature on the impact of female and minority leadership and summarize current industry trends pertaining to women and minority directors. We also present findings from a simple survey and an in-depth text analysis of audience reviews, which suggest that only a small proportion of respondents pay attention to characteristics of the movie director. Third, we examine results from our two-stage model and the matched-sample analyses. Finally, we discuss the findings, managerial implications, and directions for further research. Our results indicate that current imbalances are linked to industry player conduct rather than factors inherent to either female or minority directors.

Research Contributions

In this article, we examine the following specific research questions:

RQ₁: Do female and minority directors generate lower revenues than male, nonminority directors, and how do biases in project assignment, budgeting, and distribution affect the performance of female- and minority-directed movies?

Surface evidence pointing to the underperformance of female- and minority-directed movies is far from subtle. However, to isolate the causal effect of female and minority directors, we must control for various biases. We show that female and minority director underperformance is due to the inherently lower revenue potential of the projects to which they are assigned as well as less support in the form of production budgets and distribution. After we control for selection on the part of studios and exhibitors in these areas and other sources of endogeneity affecting revenue, budgets, distribution, and audience ratings, there is no significant revenue gap. Because revenue performance is not a statistical function of gender or race, the magnitude of the ecological underperformance of female- (37%) and minority- (42%) directed movies points to a high degree of bias on the part of studios and exhibitors in the industry. We find clear statistical evidence of biases in favor of male, nonminority directors in project assignment, budgeting, and distribution.

RQ₂: How does the presence of a female or minority lead actor affect the biases facing female and minority directors?

Research suggests that women and minority leaders perform best when other women and minorities occupy prominent positions within the enterprise. In the context of motion pictures, female and minority directors may better understand source material that calls for a female or minority lead actor and may be better able to relate to other women/minorities in the lead actor role. This would suggest a better fit between female/minority directors and female/minority lead actors, thus increasing studios’ confidence in such pairings. The result would be stronger budget and distribution support as well as reduced bias in project assignment (i.e., women and minority directors will not be assigned to less promising projects among films featuring female or minority lead actors). This is consistent with broader organizational research showing that the allocation of resources is more equitable along gender and racial lines when teams are more diverse (Joshi, Liao, and Jackson 2006).

However, the strong anecdotally reported biases in the movie industry suggest that a less favorable outcome is also possible; that is, studios may be more reticent to employ a female or minority director for a project that is already “risky” due to a female or minority lead actor. Our results suggest that women and minority directors are subject to stronger project selection and distribution biases when the film also features a female or minority lead actor. This does not carry over to budgetary decisions, where the bias against female and minority directors is equally strong regardless of the lead actor’s gender or race.

RQ₃: Does the clout of the director mitigate the level of bias that female and minority directors encounter relative to their male, nonminority counterparts with similar clout?

The clout of a director should result in better projects and stronger budget and distribution support compared with what the same director received earlier in their career. The more interesting question is whether clout helps reduce the degree of bias experienced by female and minority directors compared with similarly high-clout male, nonminority peers. An alternative way to phrase this question is whether and to what degree the bias against female and minority directors persists or diminishes as their experience within the industry grows. We find that high-clout female and minority directors are subject to weaker biases. This provides some evidence that biases do not persist at the same level as a director's career progresses. Of course, it also indicates more extreme biases against female and minority directors with less clout and experience at the outset of their careers.

RQ₄: Are the biases facing female and minority directors more pronounced if they are also international directors?

Biases against female and minority directors exist in part because they are atypical and, thus, part of an outgroup within the industry. On the one hand, female and minority directors who are international are even more atypical than the usual Hollywood director. On the other hand, there is a higher proportion of female and especially minority directors in the international subsample. Thus, the degree of bias experienced by female and minority directors may be lower among international directors than among U.S. directors. We find that the biases experienced by female and minority directors are indeed smaller if they are international, though distribution biases do not change across domestic and international directors.

Related Literature and Background

Several distinct streams of literature are relevant to this study: minority entrepreneurship, diversity in organizations, the performance of women-owned small businesses, the impact of female board members and CEOs, and leadership in the movie industry. In the following subsections, we summarize the main insights from these various streams.

Minority Entrepreneurship

While entrepreneurship is often considered a route to economic freedom for racial/ethnic minorities (Bates 1993), research on minority entrepreneurship is somewhat sparse. The evidence suggests that Black-owned and other minority businesses underperform their white-owned counterparts in sales, profits, employment, and survival (Bates 1997; Fairlie and Robb 2007) due to structural barriers facing entrepreneurs of color (Bates 1997; Bates, Jackson, and Johnson 2007). In addition to drastic differences in wealth (Avery and Rendall 2002) and access to outside capital (Bates, Jackson, and Johnson 2007), research has observed the lack of an “intergenerational link” in business ownership in Black communities. To reverse some of these trends, additional support in the areas of education, access to capital, and access to markets is often required. For example, extant research has found that access to government contracts and national markets are defining characteristics of fast-growing small to medium-sized businesses with African American CEOs (Boston and Boston 2007).

Evidence on Female Leadership

In the managerially oriented literature, Robinson and Dechant (1997) present evidence from a survey of Fortune 100 companies that workforce diversity leads to superior performance. Diversity helps companies better utilize their workforce, enhance their understanding of the marketplace, improve breadth in leadership positions, and boost creativity. Following these and similar claims, many industry studies have explored the importance of increasing female participation in the workplace. For example, a large number of nonprofit organizations have presented evidence from large corporations suggesting that both gender diversity within the workforce (Catalyst 2004) and increased female participation on corporate boards (Joy et al. 2007) lead to superior financial outcomes. Management consulting practitioners have also come to embrace this premise (Desvaux, Devillard-Hoellinger, and Meaney 2008).

However, within the academic literature, the debate on whether female leadership is associated with superior business outcomes is far from settled. While the premise that diversity leads to better team performance is widely accepted (e.g., Krishnan and Parsons 2008), the finding that women leaders deliver superior (or equivalent) organizational performance seems to depend on the unit of analysis and the specific measures used to gauge performance. Indeed, several studies find that companies led by women underperform those led by men (see Web Appendix A). However, these studies tend to focus on small entrepreneurial firms or samples of companies of mixed sizes that include many small firms. They also suggest that differences are industry- and country-specific (Swinney, Runyan, and Huddleston 2006). For example, differences in education are an important consideration in developing nations where substantial gaps in access to education are observed between men and women. In addition, several of these studies find that differences in performance are attenuated when controls linked to access to funds or human capital resources are incorporated into the analysis (Fairlie and Robb 2009).

In the entrepreneurial context, Coleman and Robb (2009) find that women not only raise substantially less external capital to launch their businesses but also encounter less funding support in subsequent fundraising rounds, even after controlling for initial startup capital and sales. In other words, even with a proven track record, women face barriers to obtaining external financing, which impedes their ability to scale their businesses. Previous literature also indicates that women entrepreneurs tend to enter less profitable sectors of the economy (Du Rietz and Henrekson 2000; Watson 2003).

Most studies that focus on larger companies find that women CEOs are responsible for equivalent or superior returns compared with their male counterparts (e.g., Gondhalekar and Dalmia 2007; Khan and Vieito 2013; Wolfers 2006), especially for innovation-driven firms (Dezsö and Ross 2012). Khan and Vieito (2013) find that, ceteris paribus, profitability increases if a female CEO manages the firm. They also point out that women are more careful when spending company funds and extract fewer personal benefits from the companies they lead. Other studies show that female CEOs tend to have relatively long tenures and are associated with more risk-averse and ethical decisions (Ford and Richardson 1994). Research also suggests that the influence of female CEOs may unfold over a longer-term time horizon. For example, female CEO gender has been found to indirectly affect long-term financial success by leading to a stronger market orientation (Davis et al. 2010). These findings differ from the entrepreneurial literature, which tends to document the lower performance of firms led by women. However, it is worth noting that even the entrepreneurial literature finds that underperformance tends to be attenuated or nonexistent in larger ventures (Du Rietz and Henrekson 2000; Watson 2003).

The relatively few studies that find adverse effects of female leadership in large organizations (e.g., Lee and James 2007) are usually based on financial measures such as cumulative abnormal returns or Tobin's q, while studies that focus on return on assets and other accounting measures tend to find a premium associated with women leaders. This pattern suggests that there may be biases in how equity markets react following earnings announcements from companies led by female CEOs. In addition to absolute performance, many studies note a reduction in risk associated with female leadership that is not necessarily reflected in point estimates of performance (Krishnan and Parsons 2008; Martin, Nishikawa, and Williams 2009). Some research also points out that women appear to need a critical mass or the presence of other women to succeed. For example, Konrad, Kramer, and Erkut (2008) find that women benefit corporate boards in a positive way only when there are three or more women on the board.

Female and Minority Leadership in the Movie Industry

On the whole, studies that focus on larger firms find either no significant differences in performance or significant advantages associated with female leadership. While the aforementioned research on female CEOs is informative, the insights gleaned from extant research are not easily transferable to the modular structure of the movie industry, where many studios are formed to produce a single feature, and the work is largely entrepreneurial in nature. Even within the corporate structure of the major studios, individual films are produced in a fairly modular way, with movie directors having substantial influence on the creative thrust and quality of the final product. Thus, it is unclear if any of the previously identified benefits associated with female leadership persist in this space or if the more negative effects found in the entrepreneurial literature are more applicable.

Despite the open nature of the fundamental question, the degree of favoritism of male and nonminority directors evidenced in the motion picture industry is unquestionably surprising. Consider industry recognition of female directors. Since the Oscars were established in 1929 through 2022, only three women—Kathryn Bigelow, Chloe Zhao, and Jane Campion—have won an Academy Award for Best Directing, in 2010, 2021, and 2022, respectively. Lina Wertmüller, Sofia Coppola, Greta Gerwig, and Emerald Fennell are the only other female directors who have been nominated for the award. Only six Black directors were ever nominated for this award (John Singleton, Lee Daniels, Steve McQueen, Barry Jenkins, Jordan Peele, and Spike Lee), but to date no wins have been recorded. At the same time, the pressure to hire female and minority directors may be increasing, and firms may be paying more attention to what other industry players are doing in this respect. In 2022, Jane Campion, a female director from New Zealand, and Ryusuke Hamaguchi, a male director from Japan, were nominated in the Best Director category. In Figure 1, we present data from the IMDb data set of films released in the United States and internationally during the 1994–2016 time period. While the overall proportion of movies directed by women is slowly increasing for all movies (both in the United States and internationally), the proportion of films directed by women among U.S. wide releases is actually declining (Panel A). For POC, the proportion remains stable for all movies worldwide but is decreasing in the United States (Panel B).

A possible explanation is related to opportunities along the career path that leads to leadership positions in the film industry. While there is limited peer-reviewed literature on the experience of women directors in Hollywood, there is research on women screenwriters. Bielby and Bielby (1996) highlight how the gender gap in earnings grows as writers advance in their careers. They also show that women are underrepresented in the Writers Guild of America, in which more than 80% of members are men. It is especially difficult for Black women to break into positions of power in Hollywood (Burton 2010). “White men have almost exclusively created the narratives and myths that comprise Hollywood cultural production, while narratives by women and racial/ethnic minorities are fewer and less prominent” (Erigha 2015, p. 78). The situation is similar to the general problem with management roles across sectors. Only 13% of African American women graduating from the Harvard Business School have achieved the highest-ranking executive positions, compared with 19% of African American men and 40% of a matched sample of non–African American Harvard Business School alumni (Washington and Roberts 2019).

Alternative Explanations

An alternative explanation for differences in director hiring revolves around audience acceptance of female and minority directors. As we have alluded to, if audiences do not accept films from female/minority directors as readily, this could explain studios’ reticence to use women and POC as directors, as well as the lower levels of support such directors receive in financing and distribution once they are in charge of a project. Previous research indicates that when it comes to minority actors, audiences actually reward rather than penalize diversity on the screen (Kuppuswamy and Younkin 2020). Because there is limited evidence regarding consumer sentiment toward directors and their identity, we conducted an exploratory empirical study of online movie reviews to determine how much attention people pay to movie directors, which we present as background.

We analyzed more than 970,000 reviews for films from more than 1,000 directors on IMDb. We identified 166,938 mentions of directors’ first/last names, indicating that only about 17% of user reviews mention directors. Thus, audiences do not seem to be especially attuned to director identity. The most commonly mentioned directors were the “big names” with long successful careers, while the overall intensity of mentions was the highest for those who appeared in front of the camera and directed simultaneously. The leaders in the overall number of mentions were Christopher Nolan, Peter Jackson, Ridley Scott, Steven Spielberg, and George Lucas. For reviews that mentioned individual directors relative to the aggregate number of reviews for their projects, the top spots belong to Chris Rock (80%), Woody Allen (75%), Tom Green (71%), Albert Brooks (68%), Tyler Perry (67%), Tom Hanks (who only has a couple of director credits [66%]), and Drew Barrymore (65%). Mentions clearly concentrate on a select group of directors. The top 5% of directors account for 56% of all mentions, and the top 20% account for 83% of all mentions.

In a separate Amazon Mechanical Turk survey, we asked respondents if they paid attention to various salient factors, including director credits, when deciding to watch a movie. The importance of director credits was second from the bottom (ahead of production studio). At the top of the list were the story, cast, genre, critical reviews, and feedback from friends and family. Again, there is little evidence that director identity matters. ¹ We also asked an open-ended question about respondents’ favorite director. Of the 507 respondents, 428 provided a verifiable director name. Steven Spielberg was the most frequently mentioned director (62 respondents), followed by Christopher Nolan (40 respondents), with Quentin Tarantino and James Cameron sharing the third spot (29 respondents each). Our respondents listed 148 unique, verifiable director names (106 were mentioned only once), with only nine of them being women (i.e., Maren Ade, Janicza Bravo, Barbra Streisand, Ava DuVernay, Susanne Bier, Kathryn Bigelow, Lana Wachowski, Nicole Kassell, and Shandra McDonald). Ava DuVernay was the only woman with more than one mention; she received two mentions, which put her in the same category as Oliver Stone, Guy Ritchie, Michael Bay, and Spike Lee. POC were relatively more represented, including the aforementioned Janicza Bravo, Ava DuVernay, Spike Lee, and Tyler Perry, and adding Denzel Washington, Hayao Miyazaki, Guillermo del Toro, Zhang Yimou, Park Chan-wook, Mostofa Sarwar Farooki, Alejandro González Iñárritu, Akira Kurosawa, Jackie Long, James Wan, Justin Lin, Mohammed Ehteshamuddin, Stephen Chow, and many others, including some prominent Bollywood directors.

Six percent of all mentioned directors were women, and 10.75% were POC. These figures are consistent with women making just under 5% of all widely released movies and minority directors accounting for 11.8%. It seems that the popularity of these directors is proportional to their exposure.

Data

The analysis is based on data obtained from BoxOfficeMojo.com, with 29,547 weekly revenue observations and 28,451 weekly screening observations pertaining to 2,386 movies widely distributed in the United States. We use all available observations in each model to utilize the information as completely as possible. Reestimating the models to include only complete observations (where screening information is available) does not materially change the results. We follow Chen, Chen, and Weinberg (2013) in defining a wide or national release as a movie that opened on at least 200 screens. We also use IMDb data to capture director identity, production budgets, cast identity, genre, MPA ratings, and so on. The Appendix summarizes the variables we used in this study and their sources, and Web Appendix B presents descriptive statistics and a correlation table.

Models and Results

We estimate a system of three equations predicting revenue, screens, and budget in a two-stage least squares framework. In the first stage, we instrument the endogenous variables and also predict the selection of a female or minority director. The estimation is complicated by the fact that movie performance is driven by both time-varying and time-invariant factors. We therefore estimate a weekly model that includes time-varying dependent variables, time-varying independent variables, and movie-specific fixed effects. We then use the movie-specific fixed effects as dependent variables, along with the film's production budget, in a set of movie-level equations that contain the time-invariant independent variables, including director gender and race.

At the weekly level, we model revenue as a function of screens, the cumulative average audience rating, the number of weeks since the movie was released, the cumulative Los Angeles Times mentions of the director, the cumulative Los Angeles Times mentions of the lead actor, the box office rank of the film, and a movie-specific fixed effect. We find some evidence that the number of weeks since release acts in a nonlinear manner and therefore include a quadratic term for this variable. We then model the fixed effect at the movie level as a function of the time-invariant characteristics: director gender and race; the production budget of the film; the duration of the film; director age and revenue history; lead actor gender, race, age, and revenue history; whether the director and lead actor are international; whether the film was written by a female writer; the number of Academy Award nominations the film received; the film's MPA ratings category and genre; and whether the film is a sequel/prequel, remake, or original picture. Among these variables, the director and actor revenue history capture clout within the industry. The number of Academy Award nominations serves as a measure of critical acclaim. We include an overall time trend at the movie level as well as additional fixed effects to capture seasonality based on the calendar week in which the movie was first released. We include these effects in the movie-level model to reflect that major differences due to time are likely to show up across movies with different release dates rather than within the (typically short) run of an individual film.

We take the natural log of all variables (plus one) except for dummy variables, as recommended in the previous literature (Elberse and Eliashberg 2003; Karniouchina 2011; Zufryden 1996). Because the movie-specific fixed effects are generated using the logged dependent variable in the weekly estimation, these fixed effects are not log-transformed in the movie-level estimation.

The second-stage equations for revenue at the weekly and movie levels are as follows:

ln _ revenu e it = ω 0 + ω 1 ln _ screen s it ^ + ω 2 ln _ audience _ ratin g it ^ + ω 3 ln _ weeknu m it + ω 4 ln _ weeknum _ square d it + ω 5 ln _ ran k it + ω 6 ln _ director _ la _ times _ mention s it + ω 7 ln _ lead _ actor _ la _ times _ mention s it + v i + e it ,

(1)

v i = θ 0 + θ 1 female _ or _ minorit y i + θ 2 female _ or _ minority _ λ i + θ 3 ln _ budge t i ^ + θ 4 M P A i + θ 5 sequel _ preque l i + θ 6 remak e i + θ 7 G e n r e i + θ 8 ln _ duratio n i + θ 9 lead _ actor _ femal e i + θ 10 lead _ actor _ minorit y i + θ 11 lead _ actor _ internationa l i + θ 12 ln _ lead _ actor _ ag e i + θ 13 ln _ lead _ actor _ revenue _ his t i + θ 14 director _ internationa l i + θ 15 ln _ director _ ag e i + θ 16 ln _ director _ revenue _ his t i + θ 17 female _ writte n i + θ 18 ln _ academy _ nomination s i + θ 19 ln _ opening _ perio d i + ∑ j = 1 51 ( γ j × calendar _ wee k j ) + u i .

(2)

Here, v_i is the movie-specific fixed effect from Equation 1; opening_period indexes the week the movie first opened within the overall sampling period to capture the time trend, and calendar_week indexes the calendar week in which the movie first opened (1–51) to reflect seasonality. In addition, MPA_i is a vector of dummy variables reflecting MPA classification obtained from IMDb (PG, PG-13, and R), and Genre_i is a vector of dummy variables corresponding to IMDb genre coding. The calendar week variable also captures some of the competitive dynamics along with the rank variable in the weekly models because high-quality movies tend to cluster around desirable release dates (Einav 2007). The excluded categories in the movie-level model are G-rated movies, action films, original pictures, and films that opened in week 52 of the year.

The model for screens is similar. The main difference is that we need to control for the revenue potential of the film during a given week because studios and exhibitors anticipate revenue when selecting the number of screens. We do this in the weekly model by specifying the screens secured during a given week as a function of the revenue for that week. The revenue a movie achieves is the best estimate of the studios’ and exhibitors’ projections. The revenue variable is instrumented to remove supply-side effects, as we describe subsequently. Including revenue contemporaneously reflects that screening decisions are forward-looking and anticipate the full range of observable and unobservable factors that influence box office performance during a given week rather than only the past performance of the film. An alternative is to model screens as a function of the previous week's revenue. However, this approach ignores unobservable information available to decision makers that is not reflected in the previous week's revenue. Other variables are identical to the weekly model for revenue. The model also includes a movie-specific fixed effect, τ_i. At the movie level, we model the movie-specific fixed effect as a function of the same variables used to predict revenue. The equations for screens are as follows:

ln _ screen s it = δ 0 + δ 1 ln _ revenu e it ^ + δ 2 ln _ audience _ ratin g it ^ + δ 3 ln _ weeknu m it + δ 4 ln _ weeknum _ square d it + δ 5 ln _ ran k it + δ 6 ln _ director _ la _ times _ mention s it + δ 7 ln _ lead _ actor _ la _ times _ mention s it + τ i + ς it ,

(3)

τ i = ρ 0 + ρ 1 female _ or _ minorit y i + ρ 2 female _ or _ minority _ λ i + ρ 3 ln _ budge t i ^ + ρ 4 M P A i + ρ 5 sequel _ preque l i + ρ 6 remak e i + ρ 7 G e n r e i + ρ 8 ln _ duratio n i + ρ 9 lead _ actor _ femal e i + ρ 10 lead _ actor _ minorit y i + ρ 11 lead _ actor _ internationa l i + ρ 12 ln _ lead _ actor _ ag e i + ρ 13 ln _ lead _ actor _ revenue _ his t i + ρ 14 director _ internationa l i + ρ 15 ln _ director _ ag e i + ρ 16 ln _ director _ revenue _ his t i + ρ 17 female _ writte n i + ρ 18 ln _ academy _ nomination s i + ρ 19 ln _ opening _ perio d i + ∑ j = 1 51 ( φ i × calendar _ wee k j ) + ψ i ,

(4)

Budget differs from revenue and screens in that it is time invariant. Therefore, we model it only at the movie level. We predict budget with the same set of variables as Equations 2 and 4, with the exception of Academy Award nominations, which are unknown and largely unforeseeable at the time budget is selected:

ln _ budge t i = η 0 + η 1 female _ or _ minorit y i + η 2 female _ or _ minority _ λ i + η 3 M P A i + η 4 sequel _ preque l i + η 5 remak e i + η 6 G e n r e i + η 7 ln _ duratio n i + η 8 lead _ actor _ femal e i + η 9 lead _ actor _ minorit y i + η 10 lead _ actor _ internationa l i + η 11 ln _ lead _ actor _ ag e i + η 12 ln _ lead _ actor _ revenue _ his t i + η 13 director _ internationa l i + η 14 ln _ director _ ag e i + η 15 ln _ director _ revenue _ his t i + η 16 female _ writte n i + η 17 ln _ opening _ perio d i + ∑ j = 1 51 ( π i × calendar _ wee k j ) + ζ i .

(5)

Endogeneity and First-Stage Estimates

Endogeneity enters our estimation due to the nonrandom selection of female and minority directors to head certain projects as well as unobserved factors that influence revenues, budgets, distribution, and audience reviews. We devise appropriate instruments and properly excluded variables to identify the selection models for director gender and race as follows.

We base our instrument for revenue on the weather during a given week, using precipitation data obtained from the National Oceanic and Atmospheric Administration's National Centers for Environmental Information. Specifically, we use a measure of the population-weighted precipitation observed during a given week, adjusted for the average amount of precipitation observed during the same calendar week over the period of the study. To create the instrument, we first calculate the mean precipitation across all weather stations within a given state in a given week. Then, we weight these amounts by state population and aggregate them to obtain a population-weighted nationwide precipitation measure to match our national box office data. Finally, we adjust this measure using the average population-weighted precipitation observed during the same calendar week over the period of the study. Thus, the final instrument reflects the amount of precipitation over/under the typical amount for a given week of the year. The logic behind the instrument is that people are more likely to engage in indoor activities, such as attending movies, rather than outdoor activities when the weather is poor. Our identifying assumption is that the amount of precipitation influences audience demand to see a given movie during a given week (relevance) in a way that is unrelated to the appeal of the focal movie, the supply of films available to watch, the budget of these films, or the gender/race of the director (exogeneity). A similar instrument has been used for blog posts (Gopinath, Chintagunta, and Venkataraman 2013) and for product reviews (Yazdani, Gopinath, and Carson 2018).

We base our instrument for screens on Gopinath, Chintagunta, and Venkataraman (2013). As they observe, theaters often close for maintenance and refurbishment, which shifts the total supply of screens available for all movies from week to week in a manner that is not correlated with the demand for any particular movie. Specifically, we use as our instrument the net reduction in total screens during a given week, adjusted for the average net reduction in screens observed during that same calendar week over the period of the study. The adjustment reflects the greater likelihood of theaters closing screens during historically lower box office periods, so we remove this potential demand-side factor. Our identifying assumption is that the closure of screens directly affects only the supply of screens available (relevance) but not the demand for a specific focal movie, its budget, or the gender/race of its director (exogeneity).

As our instrument for budget, we use the average budget allocated to all movies released within a one-year (52-week) period by other production studios before the release of the focal movie. This naturally excludes the focal movie. By “release,” we mean the first week in which the film was shown. Thus, the one-year period is unique for each week in the data. Furthermore, because we exclude the focal film and the focal studio, the instrument for each movie will generally be based on a different period and a different set of other movies. Our identifying assumption is that while it is possible that an individual studio could spend a large portion of its resources on a given film, thereby reducing the budget for other films it produces, the average budget across all movies from other studios is likely to reflect funding practices in the industry (relevance) rather than the anticipated performance of the specific focal movie, the number of screens on which it will be shown, or the gender/race of its director (exogeneity).

For audience ratings, we base our instrument on the tendency of particular reviewers to give either high or low ratings to movies in general. To calculate the instrument, we created a review-level data set consisting of more than 500,000 observations. For each review for the focal movie, we calculate the average rating provided by that reviewer for all other movies. Next, for each week in the focal movie's theatrical run, we calculate the mean of these average review scores across all reviewers who have reviewed the focal movie up to that week. This enables us to capture how the specific group of reviewers who reviewed a given movie by a certain week rated other films, on average. We then use this variable as the instrument for the cumulative audience rating of the focal movie in a given week. Both the set of reviewers and the other movies they have reviewed will change from week to week for the focal movie; thus, each movie will have a unique instrument. Our identifying assumption is that the way reviewers rate other movies will be correlated with their rating for the focal film (relevance) but will not directly affect people’s decisions to see or not see the focal movie (exogeneity). The same instrument has been used for user ratings by Li, Gopinath, and Carson (2022).

For director gender and race, we also require unique variables in the first-stage probit models for identification that are properly excluded from the second-stage estimation. To find these excluded variables, we consider supply-side constraints on the availability of female and minority directors. Specifically, we observe a strong tendency for studios to employ female directors for films that call for a female lead actor. Using this observation and the fact that there are relatively fewer female directors in Hollywood, we anticipate that the supply of female directors will decrease to the extent that a larger number of other movies released within a given window feature a female lead actor. Therefore, we take as our excluded variable in the female director selection model the percentage of movies released within a one-year (52-week) period before the release of the focal movie, not including the focal movie, with a female lead actor. Again, “release” refers to the very first week the movie was shown. Our identifying assumption is that while the gender of the lead actor in other movies will influence the availability of directors for the focal movie (relevance), this factor will not otherwise affect the performance of the focal movie, once the gender of its own lead actor and director are controlled for (exogeneity). Thus, the variable is properly excluded from the second-stage estimation.

For the model of minority director selection, we find the strongest empirical association with a slightly different and more direct measure of supply-side constraints on director availability. In particular, we use the percentage of movies released within a one-year period before the release of the focal movie, not including the focal movie, with a minority director. As with gender, our identifying assumption is that while the race of the director chosen for other movies will influence the availability of directors for the focal movie (relevance), this factor will not otherwise affect the performance of the focal movie, once the race of its own director is controlled for (exogeneity). For both minority directors and female directors, we find that these excluded variables have the strongest predictive effect in the first-stage probit models. The slight difference is due to the fact that whereas female directors are closely tied to the gender of the lead actor, the same is not true of minority directors, mainly because so many nonminority directors are selected for projects involving minority stars.

Table 1 provides first-stage estimates at the weekly level for the regressions of revenue, screens, and audience ratings on the instrumental variables, other independent variables, and movie-specific fixed effects. Each instrument predicts its corresponding endogenous variable with strong statistical significance and with the expected sign. Net closures also negatively predict revenue, as we would expect. The F-test in each regression for the joint significance of the instruments is highly significant (all p < .0000), with F-values at or above 10, as suggested by Stock and Yogo (2005). Thus, there is evidence that the instruments are not weak. Subsequently, we discuss relationships between the dependent variables and the other independent variables in the second-stage estimates.

OPEN IN VIEWER

Table 1.

First-Stage Weekly Estimates.

	(1)	(2)	(3)
	DV: ln_revenue	DV: ln_screens	DV: ln_audience_rating
ln_precipitation	.0383*** (.0102)	.0074 5 (.00938)	.00125 (.00197)
ln_netclosures	−.0258*** (.00775)	−.0340*** (.00706)	.00173 (.00150)
ln_audience_rating_reviewer_other	.0607 (.0385)	.123*** (.0352)	.428*** (.00746)
ln_weeknumber	−1.996*** (.0309)	−.342*** (.0285)	.0102* (.00599)
ln_weeknumber_squared	.0446*** (.00788)	−.135*** (.00732)	−.00692***(.00153)
ln_rank	−.0517*** (.000268)	−.0448*** (.000254)	.000127** (.0000519)
ln_director_la_times_mentions	.0436*** (.00558)	.0325*** (.00517)	.000538 (.00108)
ln_lead_actor_la_times_mentions	.0617*** (.00844)	.0474*** (.00782)	.00311* (.00164)
Constant	17.47*** (.115)	7.783*** (.105)	1.155*** (.0222)
R2	.84	.86	.07

*p < .10.

**p < .05.

***p < .01.

Notes: DV = dependent variable. Standard errors are in parentheses. We do not report movie-specific fixed effects.

Table 2 presents first-stage estimates at the movie level for the regression of budget on its instrumental variable as well as the other exogenous variables, the overall time trend, and seasonality fixed effects. We find a strong statistical effect of the instrument on budget with the expected positive sign, providing evidence that the instrument is not weak. We discuss relationships between budget and the other independent variables subsequently.

OPEN IN VIEWER

Table 2.

First-Stage Production Budget Estimates.

	DV: ln_budget
ln_average_budget_other_studio	.854***	(.128)
PG	.0811	(.102)
PG-13	.291**	(.120)
R	.0268	(.122)
Sequel_prequel	.0619	(.0392)
Remake	.214***	(.0536)
Adventure	.324***	(.0435)
Fantasy	.264***	(.0456)
Sci-fi	.278***	(.0451)
Thriller	.196***	(.0418)
Drama	−.245***	(.0373)
Comedy	.0164	(.0422)
War	.312***	(.0948)
Biography	−.184**	(.0786)
Documentary	−1.482***	(.166)
Music	−.0301	(.0700)
Romance	−.0167	(.0405)
History	.0623	(.102)
Horror	−.336***	(.0564)
Crime	.0560	(.0448)
Animation	.562***	(.0774)
Family	.321***	(.0808)
Sport	.0894	(.0758)
ln_duration	1.754***	(.116)
lead_actor_female	−.0722**	(.0328)
lead_actor_minority	−.127***	(.0440)
lead_actor_international	.0794**	(.0349)
ln_lead_actor_age	.145***	(.0481)
ln_lead_actor_revenue_hist	.0194***	(.00236)
Director_international	.127***	(.0326)
ln_director_age	.385***	(.0734)
ln_director_revenue_hist	.0195***	(.00202)
female_written	.00160	(.141)
ln_opening_period	−.157***	(.0235)
Constant	−8.302***	(2.290)
R2	.55

*p < .10.

**p < .05.

***p < .01.

Notes: DV = dependent variable. Standard errors are in parentheses. We do not report seasonality fixed effects.

Director Selection Estimates

We model endogeneity in the choice of director gender and race at the movie level, following Heckman’s (1979) recommended approach. Specifically, we estimate a selection equation that predicts whether the studio will employ a female or a minority director for a given picture. We then include the nonselection hazard (inverse Mills ratio) from this probit model in the movie-level revenue, screens, and budget equations.

In specifying the equation predicting director identity, we must again recognize that the studios are forward-looking and model the effect of the anticipated revenue potential and anticipated screening potential of the film. To do so, we take the predicted values for weekly revenue and weekly screens from the first-stage instrumental variables estimation and aggregate them up to the movie level as the predicted total revenue for the film and the predicted total number of screens (screen-weeks) on which the film will be shown. We emphasize that these variables are instrumented and included to reflect the anticipated revenue potential and anticipated screening potential of the film. We control for the overall time trend and seasonality in the selection equation, as we do in all the movie-level models. The model that predicts the use of a female or a minority director is as follows:

female or minorit y i = β 0 + β 1 ln _ percent _ female _ lead _ othe r i + β 2 ln _ percent _ min _ dir _ othe r i + β 3 ln _ budge t i ^ + β 4 ln _ total _ screen s i ^ + β 5 ln _ total _ revenu e i ^ + β 6 M P A i + β 7 sequel _ preque l i + β 8 remak e i + β 9 G e n r e i + β 10 ln _ duratio n i + β 11 lead _ actor _ femal e i + β 12 lead _ actor _ minorit y i + β 13 lead _ actor _ internationa l i + β 14 ln _ lead _ actor _ ag e i + β 15 ln _ lead _ actor _ revenue _ his t i + β 16 female _ writte n i + β 17 ln _ opening _ perio d i + ∑ j = 1 51 ( κ i × calendar _ wee k j ) + o i .

(6)

The results of the probit estimation predicting the use of a female or minority director appear in Table 3. Beginning with the excluded variables, we observe that female/minority directors are less likely to be selected when other films released within a one-year prior period featured either a female lead actor or a minority director. The strong significance of these variables suggests that the model is identified by more than just the nonlinearity of the probit. Turning to the other variables, the most important result is that female or minority directors are significantly less likely to be chosen for films with larger budgets. As we will show, this lack of budgetary support is a key driver of the ecological underperformance of films from female and minority directors. We also observe that female or minority directors are more likely to be chosen for films featuring minority lead actors and for longer films. Marginally significant results (p < .10) indicate that female and minority directors are also chosen more often for films featuring female leads, films with younger lead actors, and films in the war and romance genres, while they are chosen less often for horror movies.

OPEN IN VIEWER

Table 3.

First-Stage Female/Minority Selection Probit Model Estimates.

	Female_or_minority
ln_pct_fem_lead_­other	−3.058***	(1.016)
ln_pct_min_dir_other	−1.741**	(.869)
ln_budget_predicted	−.582***	(.161)
ln_total_screens_predicted	−.0530	(.0590)
ln_total_revenue_predicted	.0213	(.0331)
PG	.108	(.268)
PG-13	.413	(.308)
R	.151	(.306)
Sequel_prequel	.151	(.0951)
Remake	.0355	(.135)
Adventure	−.142	(.121)
Fantasy	.124	(.118)
Sci-fi	.102	(.124)
Thriller	.0831	(.105)
Drama	.0290	(.0947)
Comedy	−.0723	(.0973)
War	.380*	(.217)
Biography	−.0739	(.179)
Documentary	−.617	(.408)
Music	.155	(.149)
Romance	.174*	(.0907)
History	−.219	(.220)
Horror	−.273*	(.145)
Crime	.0293	(.103)
Animation	.308	(.226)
Family	.271	(.204)
Sport	−.132	(.176)
ln_duration	1.055**	(.428)
lead_actor_female	.129*	(.0772)
lead_actor_minority	.992***	(.0894)
lead_actor_international	.109	(.0830)
ln_lead_actor_age	−.233*	(.121)
ln_lead_actor_revenue_hist	.000894	(.00630)
female_written	−.153	(.376)
ln_opening_period	−.00753	(.0655)
Constant	4.873***	(1.712)
Pseudo R2	.17

*p < .10.

**p < .05.

***p < .01.

Notes: Standard errors are in parentheses. We do not report seasonality fixed effects.

The probit includes a large number of independent variables such that relatively few are statistically significant. This in itself is not problematic, as we are ultimately interested in the nonselection hazard from the model rather than the significance of individual predictors. The accuracy of the hazard will increase with the inclusion of more variables (information); therefore, we preserve all variables in the final model. To evaluate overall model fit, we compare the predicted probabilities and hazard from the probit model with the actual percentage of female or minority directors observed in the different categories in Figure 4. There is close correspondence, with an almost perfect correlation for the predicted probabilities as expected and a correlation of .993 for the hazard. ² Thus, the fitted probit model appears to reproduce the effects of selection well. We present the results for the probit models predicting the use of female directors and minority directors as separate groups in Web Appendix C.

OPEN IN VIEWER

Figure 4.

Estimated female hazard from probit model versus observed percent female director.

Revenue Estimates

We begin by analyzing the effect of director gender and race on revenue using the weekly and movie-level models in Equations 1 and 2. The results from the weekly model appear in Table 4. The estimates in Column 1 reveal that revenue is significantly related to the number of screens on which a picture is shown, with an elasticity of .800, reflecting diminishing returns as more screens are added. We find a limited effect of audience ratings on revenue but a significant effect on screens in Column 2. Thus, ratings appear to affect revenues indirectly through their impact on screen allocation, with highly rated movies enjoying broader exhibition. Films engaging more prominent directors and stars (as indicated by Los Angeles Times mentions) generate higher revenues. In addition, revenue decreases with the number of weeks since the film's release, but with a slower rate of decay for movies that have “legs” (i.e., enjoy longer theatrical runs). Finally, the control variable for competition has the expected effect, with higher numerically ranked movies exhibiting lower revenues (p < .10).

OPEN IN VIEWER

Table 4.

Weekly Revenue Estimates.

	(1)	(2)
	ln_revenue	ln_screens
ln_revenue_hat		.703*** (.177)
ln_screens_hat	.800*** (.209)
ln_audience_rating_hat	−.0647 (.104)	.181** (.0862)
ln_weeknumber	−1.711*** (.0784)	1.058*** (.356)
ln_weeknumber_squared	.139*** (.0287)	−.165*** (.0110)
ln_rank	−.0162* (.00938)	−.00858 (.00917)
ln_director_la_times _mentions	.0204** (.00864)	.00165 (.00932)
ln_lead_actor_ la_times_mentions	.0308** (.0128)	.00354 (.0135)
Constant	11.34*** (1.637)	−4.578 (3.128)
R2	.84	.86

*p < .10.

**p < .05.

***p < .01.

Notes: Standard errors are in parentheses. We do not report movie-specific fixed effects.

Turning to the movie level, the results for revenue appear in Column 1 of Table 5. The critical variable in this estimation is the female or minority lambda term, which is the nonselection hazard from the first-stage probit model predicting the use of a female or minority director. This coefficient captures the effect of selection on the revenue associated with a given film (e.g., Heckman 1979). Because the outcome in the probit is coded with a value of one for female or minority directors, the nonselection hazard is monotonically increasing in the probability that a film is directed by a male, nonminority director (Heckman 1976, p. 479). The hazard is a transformation of this probability, but the correlation between the hazard and the probability is high (.992) for our model.

OPEN IN VIEWER

Table 5.

Movie-Level Revenue Estimates.

	(1)		(2)		(3)
	DV: fe_revenuea		DV: fe_screens		DV: ln_budget
female_or_minority	−.0580	(.0402)	.00179	(.0231)	−.111**	(.0473)
female_or_minority_lambda	.837***	(.153)	.178**	(.0888)	.676***	(.131)
ln_budget_predicted	.498**	(.243)	.540***	(.142)
PG	−.0361	(.0793)	−.0972**	(.0410)	.0920	(.104)
PG-13	−.0318	(.138)	−.181**	(.0722)	.410***	(.145)
R	−.197*	(.113)	−.185***	(.0565)	.114	(.148)
sequel_prequel	.413***	(.0411)	.144***	(.0223)	.108**	(.0433)
Remake	−.163**	(.0700)	−.0949**	(.0377)	.155***	(.0443)
Adventure	−.335***	(.0885)	−.149***	(.0497)	.131**	(.0536)
Fantasy	−.152**	(.0751)	−.138***	(.0426)	.263***	(.0458)
Sci-fi	−.194**	(.0805)	−.164***	(.0449)	.260***	(.0448)
Thriller	−.107*	(.0610)	−.110***	(.0349)	.177***	(.0428)
Drama	.132**	(.0673)	.0734*	(.0397)	−.145***	(.0419)
Comedy	−.0541	(.0425)	−.0229	(.0227)	−.0371	(.0453)
War	−.0745	(.122)	−.0723	(.0706)	.399***	(.0842)
Biography	.0333	(.0906)	.0217	(.0570)	−.171**	(.0853)
Documentary	.662	(.406)	.593**	(.236)	−1.353***	(.332)
Music	−.0150	(.0736)	−.00716	(.0387)	.0525	(.0749)
Romance	.116***	(.0438)	.0307	(.0253)	.0931**	(.0437)
History	−.373***	(.101)	−.194***	(.0720)	−.0900	(.112)
Horror	.0703	(.0962)	.193***	(.0558)	−.385***	(.0641)
Crime	−.0649	(.0451)	−.0291	(.0251)	.0665	(.0434)
Animation	.122	(.153)	−.0930	(.0875)	.525***	(.0770)
Family	.119	(.119)	−.00726	(.0630)	.375***	(.0980)
Sport	−.204***	(.0755)	−.0220	(.0419)	−.0353	(.0831)
ln_duration	.156	(.445)	−.595**	(.257)	1.766***	(.108)
lead_actor_female	.142***	(.0389)	.0578***	(.0219)	.0282	(.0370)
lead_actor_minority	.726***	(.129)	.117	(.0738)	.463***	(.116)
lead_actor_international	−.0602	(.0403)	−.0963***	(.0233)	.111***	(.0349)
ln_lead_actor_age	−.289***	(.0626)	−.142***	(.0366)	−.0184	(.0588)
ln_lead_actor_revenue_hist	−.0122**	(.00521)	−.0116***	(.00302)	.0131***	(.00285)
director_international	−.179***	(.0430)	−.105***	(.0250)	.0871***	(.0321)
ln_director_age	−.823***	(.116)	−.396***	(.0657)	.247***	(.0834)
ln_director_revenue_hist	−.0100**	(.00490)	−.0102***	(.00283)	.0123***	(.00243)
female_written	.113	(.110)	.00666	(.0643)	−.0740	(.116)
ln_academy_nominations	.117***	(.0130)	.0356***	(.00592)
ln_opening_period	−.145**	(.0705)	−.0639	(.0439)	−.127***	(.0323)
Constant	−5.226***	(1.742)	−3.571***	(1.040)	5.953***	(.689)
R2	.48		.40		.55

*p < .10.

**p < .05.

***p < .01.

^aHere, and in other exhibits, fe_revenue and fe_screens are the fixed effects for revenue and screens from the weekly model.

Notes: DV = dependent variable. Standard errors are in parentheses. We do not report seasonality fixed effects.

The coefficient on the nonselection hazard thus reflects the effect of an increase in the probability that a given film will be assigned to a male, nonminority director. A positive coefficient reflects a bias in favor of male, nonminority directors in that they have a higher probability of being assigned to better-performing films; therefore, uncorrected estimates will overstate the negative effect of a female or minority director (e.g., Certo et al. 2016, p. 2643; Greene 1993, p. 713). The statistical significance of the coefficient serves as a direct test of the presence of selection bias (Heckman 1976; Wooldridge 2010). In the revenue model, such bias is largely due to the nonrandom matching of directors of different genders and races with projects of different revenue potential. As is evident, a statistically significant selection effect is present, indicating a strong bias in favor of male, nonminority directors. After we control for the selection of director gender and race, the budget of the film, and screens (in the weekly estimation), the revenue earned by movies from female and minority directors is statistically indistinguishable from that achieved by movies from all other directors, as shown by the insignificant coefficient on the female_or_minority dummy. ³

Turning to the other coefficients, we observe that higher revenue is associated with larger production budgets as well as with sequels and prequels. Conversely, remakes realize lower revenue, on average. Interestingly, films with female and minority lead actors perform better at the box office. We attribute this to their relative rarity, as well as higher hurdles that must be met for such films to be greenlighted. As Column 3 shows, they also receive lower budgets from the studios. Older lead actors and older directors are associated with lower revenues, as are international directors. Interestingly, both actors and directors with more significant revenue histories are associated with lower box office performance, suggesting mean reversion. However, both groups receive significantly higher budgets due to their clout, as shown in Column 3. Not surprisingly, films that receive more Academy Award nominations generate higher revenues. Coefficients on the genre categories must be interpreted carefully because they are relative to the excluded category (action movies) and are influenced by the inclusion of budget in the model. After we partial out the effect of their large budgets, adventure, fantasy, and science fiction films all perform worse, on average. History and sport films also perform worse, although this is less related to budget than their inherent appeal. Dramas and romances perform better after accounting for their low budgets. R-rated films also have moderately lower revenues (p < .10). Finally, the time trend indicates that inflation-adjusted revenue declines over time, which is consistent with an examination of the raw data. ⁴

Distribution Estimates

Recall that a major contribution of this study is its ability to examine factors that explain the observed underperformance of films from female and minority directors at the box office. Chief among these factors is the level of support such films receive in the form of distribution and budget. We present the results of the weekly model predicting the number of screens in Column 2 of Table 4. The results indicate that the number of screens allocated to a given picture in a given week is a significant function of revenue, with an elasticity of .703. Revenue in this equation has been instrumented in an attempt to remove the influence of supply-side effects; thus, the finding that elasticity approaches one suggests that studios and exhibitors are forward-looking and reasonably accurate in matching screens to revenue potential for a given week.

As we have alluded to, we also find a positive effect of viewer ratings on the number of screens allocated during a given week. This factor seems to have a stronger effect on exhibitors’ decision making than media coverage, as evidenced by the insignificant effects of Los Angeles Times mentions for both the star and the director. In the only other significant effects, we observe a quadratic relationship for the number of weeks since the film was released, with screens first growing briefly and then contracting rapidly.

Movie-level results appear in Column 2 of Table 5. We begin with the female or minority lambda term. Its sign indicates that the number of screens secured for a given picture increases with the hazard that the film is directed by a male, nonminority director. Thus, we again find evidence of a strong and significant bias favoring male, nonminority directors. After we control for this and other endogenous selection factors, the number of screens allocated to films from female or minority directors is statistically indistinguishable from those allocated to movies from all other directors. This does not mean that such differences in distribution intensity do not exist. Rather, they are explained predominantly by bias favoring male, nonminority directors in the choices made by studios and exhibitors. Because screens are closely related to revenue in the weekly estimation, we can conclude that this bias has a strong effect on the performance of films from women and minority directors.

Turning to the other coefficients, we find similar results to those in the revenue model. Signs and significances for the genre variables are virtually all the same, with the exceptions of romance movies, which do not appear on a significantly larger number of screens despite their superior box office performance, and sports movies, which appear on the same number of screens as other movies despite their lower revenues. Sequels and prequels receive significantly greater distribution, whereas remakes receive significantly less. Longer movies are more difficult to exhibit, as are those with an international lead actor or international director. As was the case with revenues, lead actor age, lead actor revenue history, director age, and director revenue history are all negatively related to screens, whereas movies with female lead actors (but not minority lead actors) and those with more Academy Award nominations are shown on more screens. Finally, we detect significant effects in the MPA rating categories, with PG, PG-13, and R-rated movies appearing on fewer screens than the excluded G-rated category.

Budget Estimates

The results from the movie-level model predicting production budget appear in Column 3 of Table 5. Again, we find a significant bias in favor of male, nonminority directors, as reflected in the positive coefficient on the lambda term. Unlike revenue and screens, we still find a significant, negative effect of a female or minority director on budget, even after we control for selection and endogeneity. The negative coefficient may be due to effects of selection/endogeneity that we are unable to capture in the model. Alternatively, it could reflect that female and minority directors tend to operate on lower budgets than their male, nonminority counterparts in a way that is not conditional on the factors included in the model (recall that a similar result has been found for female CEOs). Unconditionally lower spending may be viewed as a virtue, but given the strong effect budget has on revenue, the tendency to spend less could also explain some degree of underperformance associated with female- and minority-directed films that is not directly attributable to studio or exhibitor bias. These directors may be requesting lower budgets or spending too conservatively, perhaps due to greater scrutiny of their financial performance.

Examining the remaining coefficients, we note several significant effects. Sequels/prequels and remakes receive higher budgets, as do PG-13, adventure, fantasy, science fiction, thriller, war, romance, animation, family, and longer movies. Dramas, biographies, documentaries, and horror films tend to have smaller budgets. Considering the characteristics of the lead actor and director, larger budgets tend to be allocated to films with minority and international lead actors (but not female lead actors), international and older directors, and directors and actors with more significant revenue histories. Recalling the results for revenue, more established directors receive higher budgets, but do less with them. Finally, the time trend shows that budgets, like revenues, have been declining over time on an inflation-adjusted basis, as is evident in the raw data.

Matched-Sample Analysis

To provide a robustness check for our results, we compiled a matched sample of movies by taking all 123 movies directed by women in our sample and finding the closest match in terms of release date, genre combination, MPA ratings categories, and budget. Following the same approach, we compiled another matched sample of movies that includes all 257 movies directed by minorities and their closest match. We then examined the pairwise differences in revenues, screening intensity, and movie quality (see Table 6).

Despite our attempt to select female-directed movies with similar characteristics and comparable budgets, the closest movie in the male-directed sample released within two calendar years of the focal film had a significantly higher budget. Yet the difference in revenue between the male-directed and female-directed samples was not significant. We also failed to detect differences in viewer ratings or critical acclaim, which is in line with the idea that women directors may be better at delivering quality outcomes under conditions of resource scarcity (Karniouchina et al. 2020).

OPEN IN VIEWER

Table 6.

Matched-Samples Results.

	Male-Directed	Female-Directed	Minority-Directed	Non-Minority-Directed
Average budget ($)	62,911,531.97**	51,728,217.95**	48,916,405.47	49,416,776.05
Average revenue ($)	76,161,836.12	65,224,135.60	67,399,513.91*	56,900,114.04*
Average duration	107.57	107.2	107.62	105.50
Average screens	2,367.70	2,208.89	2,272.50	2,332.18
Average Academy nominations	.56	.37	.20	.16
Average lead actor nominations	2.70	2.61	1.74	2.11
Average IMDb score (audience)	6.15	6.07	5.88***	6.16***

*p < .10.

**p < .05.

***p < .01.

Considering the pairwise differences between the minority-directed and non-minority-directed samples, we find that minority-directed movies produced marginally significantly higher revenues (p < .10) but had no significant difference in budget. Minority-directed films also received significantly lower IMDb audience ratings. Despite the lower audience rating, the difference in the average number of Academy Awards nominations was not significant, suggesting lower exposure and a weaker consumer following, despite comparable critical acclaim.

Moderator Analysis

Finally, we examine how several key moderating variables affect the extent of bias evidenced with respect to revenue, screens, and budget. We do so by examining interactions between four moderating variables and the female or minority lambda term, which reflects the magnitude of selection effects. The results of the moderator analysis appear in Table 7. Stronger biases against women and minority directors exist with respect to both revenue and screens when the movie has either a female lead actor or a minority lead actor. This confirms the view that studios are less likely to trust a female or minority director with a movie when the picture is already risky (from the perspective of the studio) due to a female or minority lead. However, a weaker bias is evidenced for movies in which the female or minority director has a higher revenue history. Thus, the bias against female and minority directors appears to decline with accumulated success in the industry. Finally, female and minority directors who are also international experience weaker biases in terms of revenue and budget, although this moderating effect is not statistically significant for screens. This indicates that international directors face somewhat different biases than their U.S. counterparts. It may also reflect the greater opportunity for women and POC to amass experience in the international film industry before coming to Hollywood.

OPEN IN VIEWER

Table 7.

Interaction Model Estimates: Female and Minority Directors.

	(1)		(2)		(3)
	DV: fe_revenue		DV: fe_screens		DV: ln_budget
female_or_minority	−.0468	(.0402)	.00627	(.0232)	−.0993**	(.0471)
female_or_minority_lambda	1.055***	(.162)	.295***	(.0931)	.788***	(.140)
female_actor_interaction	.346**	(.142)	.156*	(.0863)	.170	(.167)
minority_actor_interaction	.538***	(.173)	.235**	(.112)	.340	(.223)
director_revenue_interaction	−.193**	(.0828)	−.130***	(.0499)	−.0667	(.104)
director_intl_interaction	−.298**	(.142)	−.133	(.0897)	−.493**	(.193)
ln_budget_predicted	.412*	(.244)	.505***	(.142)
PG	−.00248	(.0794)	−.0794*	(.0409)	.109	(.105)
PG-13	.0523	(.139)	−.141*	(.0729)	.440***	(.146)
R	−.147	(.114)	−.159***	(.0568)	.140	(.149)
Sequel_prequel	.434***	(.0418)	.152***	(.0226)	.114***	(.0434)
Remake	−.158**	(.0705)	−.0943**	(.0376)	.155***	(.0445)
Adventure	−.348***	(.0880)	−.157***	(.0494)	.121**	(.0539)
Fantasy	−.135*	(.0748)	−.131***	(.0424)	.259***	(.0457)
Sci-fi	−.180**	(.0803)	−.159***	(.0448)	.250***	(.0447)
Thriller	−.104*	(.0608)	−.110***	(.0347)	.173***	(.0429)
Drama	.137**	(.0669)	.0757*	(.0396)	−.132***	(.0422)
Comedy	−.0703*	(.0426)	−.0310	(.0227)	−.0423	(.0456)
War	−.0435	(.123)	−.0599	(.0712)	.401***	(.0843)
Biography	.0232	(.0904)	.0176	(.0569)	−.169**	(.0851)
Documentary	.559	(.406)	.547**	(.236)	−1.330***	(.333)
Music	.0269	(.0727)	.0123	(.0388)	.0752	(.0742)
Romance	.141***	(.0443)	.0420	(.0256)	.102**	(.0442)
History	−.391***	(.103)	−.201***	(.0727)	−.0975	(.111)
Horror	.0187	(.0971)	.169***	(.0562)	−.396***	(.0641)
Crime	−.0614	(.0452)	−.0273	(.0252)	.0623	(.0430)
Animation	.146	(.153)	−.0865	(.0874)	.513***	(.0773)
Family	.165	(.119)	.0128	(.0633)	.380***	(.0986)
Sport	−.232***	(.0761)	−.0359	(.0423)	−.0484	(.0847)
ln_duration	.306	(.446)	−.534**	(.257)	1.767***	(.108)
lead_actor_female	−.168	(.139)	−.0812	(.0847)	−.121	(.166)
lead_actor_minority	.448***	(.154)	−.00410	(.100)	.270	(.187)
lead_actor_international	−.0441	(.0407)	−.0894***	(.0234)	.116***	(.0349)
ln_lead_actor_age	−.309***	(.0623)	−.153***	(.0365)	−.0251	(.0586)
ln_lead_actor_revenue_hist	−.0127**	(.00517)	−.0119***	(.00300)	.0121***	(.00292)
director_international	.117	(.145)	.0269	(.0917)	.572***	(.199)
ln_director_age	−.820***	(.116)	−.394***	(.0656)	.218***	(.0837)
ln_director_revenue_hist	.188**	(.0849)	.123**	(.0514)	.0801	(.107)
female_written	.0727	(.110)	−.0128	(.0650)	−.0935	(.119)
ln_academy_nominations	.116***	(.0130)	.0352***	(.00592)
ln_opening_period	−.140**	(.0700)	−.0609	(.0440)	−.126***	(.0325)
constant	−4.936***	(1.747)	−3.489***	(1.041)	5.864***	(.688)
R2	.69		.33		.58

*p < .10.

**p < .05.

***p < .01.

Notes: DV = dependent variable. Standard errors are in parentheses. We do not report seasonality fixed effects.

In the foregoing analyses, we combined female and minority directors into a single category to maximize the sample size of underrepresented directors. In Web Appendix C, we present analyses for female and minority directors as separate groups. The results show that both groups experience similar biases in terms of project assignment, distribution, and budgetary support.

Discussion

Our results indicate that women and POC direct films that generate similar revenues to films directed by male, nonminority directors when they are given the same opportunities. However, they are not currently being given the same opportunities. Biases against female and minority directors begin with project assignment and “build” throughout the process to include decisions about budgeting and exhibition. Importantly, these biases exist independently; for example, limited exhibition is not merely reflective of an “earlier” bias in project assignment, but reflects a bias that exists in addition to the bias in project assignment. Thus, women and POC encounter new biases at each stage in the process.

Biases against female and minority directors appear to be especially strong when the lead actor is female or a POC. This is consistent with the concern that studios are more reluctant to entrust women and POC with resources for projects that already fall outside the “norm.” This is unfortunate, as such directors have lived experiences that allow for deeper insight into projects centered on women and POC. It is also unfortunate given evidence suggesting that women and minorities are able to contribute more when surrounded by other women and minorities. In contrast, international directors and more established directors with higher clout who are also women or POC experience less discrimination. Successful women and minority directors are thus able to overcome some of the biases they face as their careers progress, but they should never have to face such obstacles in the first place.

The disparities observed in this study result in both inefficiency and inequity in the labor market. Talented filmmakers miss out on directorial projects that could further their career, the public misses out on creative output that will never come to fruition, and the studios and exhibitors miss out on profits that could have been made from successful movies. Unfortunately for those trying to break into the business, hiring in Hollywood depends on previous success (John, Ravid, and Sunder 2017). With the first big-budget film being a major hurdle, it is critical for major studios to support women and POC early in their careers and to seek out and elevate directing talent from the independent movie sector where women and minorities have a broader representation. Using data on 1,200 films from 2007 to 2018, Smith et al. (2019) report that only 4.9% of top-grossing films have female directors compared with 34.5% of smaller independent features. Unfortunately, this somewhat more balanced representation has not crossed over to films slated for major release.

What, then, is the solution? Our analysis suggests three ways to support women and minority directors: increasing the number of women and POC selected to direct high-potential wide-release films, increasing the budgets allocated to female- and POC-directed films, and increasing the distribution of films directed by women and minorities. This conclusion echoes industry insights, in this case regarding female directors:

This imbalance is particularly egregious at the lower end of the budget spectrum where female driven films are seen on only a third as many movie screens as male driven films. Implicit bias may be at play here across the marketplace. The fact that the average budget for women-directed films is also substantially less than the average budgets for films directed by men, implies an institutionalized lack of financial commitment that starts from the developmental stages of a film's life and extends all the way through to its theatrical exhibition (Slated 2016).

Our findings add appropriate rigor to the industry study by Slated (2016), which is largely based on an examination of descriptive statistics across different categories of films. Both suggest that women and POC need greater access to directing projects along with sufficient budgets and adequate distribution. The distribution side, in particular, is an area in which the gender and racial makeup of buyers has received little attention. In Web Appendix D, we present a sample of various organizations focused on getting aspiring women and minority directors started in filmmaking through mentoring or small grants for independent films. While these efforts undoubtedly help build a diverse talent pipeline, they do not adequately assist midcareer directors in getting their films into wide distribution. Focusing on women in film distribution, Follows (2018) reports that as of 2018, 44% of professionals attending the major film markets (movie industry trade shows) were women. However, the percentage was lower among those in high-status jobs attending these markets (32%) and was also lower when considering attendees at the American Film Market (35% for 2017, the latest year of available data). If 30%–40% of the buyers in the market are women but only 4% of top-grossing films are directed by women, increasing the number of women making distribution decisions seems unlikely to solve the problem. Additional research is needed to determine whether women distributors are more likely to bet on films directed by women and to explore racial representation in distribution. A greater understanding of return on investment for female- and minority-directed films needs to be developed and shared with the distributor and exhibitor communities to make greater progress.

Managerial Implications

The results of this study have several implications for marketing. First, the matched-sample analyses suggest that financial gains are likely to accompany an increase in the employment of female and minority directors by Hollywood studios. Female directors produce statistically equivalent revenues with smaller budgets than their male peers, whereas minority directors produce higher revenues with the same budget as nonminority directors. In both cases, the result is a higher return on investment. While it is possible that such differences reflect the inherent nature of female and minority directors, it is more likely that (1) female and minority directors are under greater financial pressure to perform because of the biases they face, resulting in more efficient use of resources, and (2) the creative output from diverse directors adds a degree of novelty to their work that audiences welcome, especially in a Hollywood culture plagued by formulaic rehashes of well-known tropes. This, in turn, increases differentiation and may make the job of promoting the film and breaking through the clutter somewhat easier.

Second, by releasing more projects led by female and minority directors, Hollywood studios can expand their audience base. Films such as Black Panther and Wonder Woman illustrate how Hollywood can connect with minority and female audiences by expanding minority and female participation in the creative process. Black Panther involved minority writers, directors, and actors and clearly resonated with both minority and nonminority audiences, while Wonder Woman proved that a well-executed superhero action movie does not require a male director or protagonist. Furthermore, these movies demonstrated that big-budget superhero films delivered by minority and women directors and leading cast members can compete for the top box office spots. Black Panther had the top domestic box office result in 2018, and Wonder Woman finished at number three in 2017, trailing only the two Disney movies released in the same year.

Third, while we do not have promotional data available in our study, we suspect that similar decisions are being made with respect to the promotion of films from female and minority directors as are observed for production budgets and exhibition. Indeed, a common assumption in movie research is that promotional budgets are a fixed proportion of production budgets. Returning to the matched sample for minority directors, the performance of these films, despite the relatively poor reception from mainstream audience reviewers, may also indicate insufficient marketing of the films outside targeted minority audiences. Box office performance suggests that the appeal of these films may be more universal if properly marketed.

Fourth, the extant literature highlights the benefits associated with diversity (in leadership and team composition), such as greater long-term orientation, innovation, and customer focus. Given the evidence of biases that reinforce Hollywood's lack of diversity, we may reasonably assume that the movie industry would similarly benefit from more female and minority involvement. Previous research also points out the positive impact of the expansion of nontraditional leadership. For example, following the introduction of gender quotas on corporate board seats in Norway, the companies affected by the new rules had higher payrolls and more stable employment (Matsa and Miller 2013).

Limitations and Future Research Directions

Our inquiry is limited to financial measures of performance. We do not examine the social implications of female or minority leadership in the movie business, which may be especially pronounced in this highly visible sector of the economy. Our study examines data starting in 1994, when detailed weekly screening, user review, and budget information became available for a large sample of films. This later period coincides with increased globalization and heightened focus on women and minority participation across many industries. We are unable to generalize our findings to previous periods. Future studies should draw connections between broader societal trends and biases against women and minority leaders.

For empirical simplicity, we equate gender with biological sex and assume an obviously flawed gender dichotomy. While we admittedly fail to surface a more nuanced picture of gender identity, this approach still allows us to shed light on female-directed movie performance, as previous research points out that findings related to feminine gender identity are typically replicated when using biological sex (e.g., Winterich, Mittal, and Ross 2009; Zhang, Feick, and Mittal 2014). While the state of the industry is disappointing when it comes to the representation of women, the representation of nonbinary people also deserves urgent attention. We hope that future work will offer a more detailed examination of gender dynamics as an increasing number of people identify as gender fluid, gender nonconforming, genderqueer, and so forth.

There are several other important avenues for further research. First, our results indicate that women and minority directors make do with lower budgets, even after controlling for selection bias. Understanding this result in more detail is an important issue that deserves further investigation—in particular, whether it is the result of enhanced scrutiny of these directors, evidence of greater efficiency, or symptomatic of depressed women and minority director compensation. More broadly, the support of female and minority directors after they have been assigned to a picture deserves further investigation, as this is an potential source of ongoing bias. We have highlighted budget and distribution decisions, but promotion, the participation of big-name stars of different genders/races, and supporting talent on the production side (e.g., cinemaphotographers) could also be examined. Future studies might also examine how women and minority directors perceive their own roles in the movie industry in a more systematic way. In particular, do they behave in a way that reflects industry biases even after having been brought onboard? For example, do they request lower budgets or exhibit more conservative spending patterns and underspend the budgets assigned to their films?

Another contribution could be made by further exploring audience reactions to female and minority directors. Future work should conduct more detailed analyses of the language used in consumer and critical reviews of projects from various actors and directors. Using advanced natural-language-processing tools, researchers could determine how consumers and critics talk about different types of actors and directors and possibly link uncovered differences to movie ratings and box office performance. There is ample reason to suspect that the gender/race of the reviewer would also play a role, which can be inferred using machine learning techniques. Marketing scholars are especially well-positioned to conduct such inquiries.

Future work might also examine more subtle genre-level effects of female and minority directors; for example, certain directors may perform better in specific genres or be influenced by a particular script, ensemble cast, or studio characteristics. On a related note, anecdotal evidence suggests that many female and minority directors direct their own scripts. While this level of creative involvement is typically perceived as a virtue in Hollywood, it may also reflect that women and POC are only given opportunities when they take extraordinary measures to succeed. An investigation of the role of professional networks would complement such inquiries.

Finally, the movie industry is currently implementing diversity guidelines—for example, only movies with diverse casts and crews can be nominated for some awards (e.g., the “50/50 by 2020” pledge). Further research needs to guide the optimal design of such programs. In promoting equity, one must look beyond simple representation, as gaps in authority and wages often remain even when the representation gap narrows (Jacobs 1992).

Supplemental Material

sj-pdf-1-mrj-10.1177_00222437221100217 - Supplemental material for Women and Minority Film Directors in Hollywood: Performance Implications of Product Development and Distribution Biases

Supplemental material, sj-pdf-1-mrj-10.1177_00222437221100217 for Women and Minority Film Directors in Hollywood: Performance Implications of Product Development and Distribution Biases by Ekaterina V. Karniouchina, Stephen J. Carson, Carol Theokary, Lorien Rice and Siobhan Reilly in Journal of Marketing Research