A Sturdy Glass Ceiling: Representation of Women on Screen and Behind the Scenes of Hollywood’s Top 100 Films Throughout the Years

Student Author

Kara R. Miller is a senior in the at Indiana University in 1987 and at Dartmouth College Krannert School of Management in 2000. He has published widely on Italian and Italian studying marketing with a concen- American cinema. His translation of Pier Paolo Paso- tration in international business and lini’s Heretical Empiricism (Indiana University Press, a certificate of entrepreneurship. 1988, expanded, NAP, 2006) is widely cited. Most She is also pursuing minors in both recently he coedited Revisioning Terrorism: A Human- communication studies and film istic Perspective (Purdue University Press, 2016). He is and video studies. Immediately fol- a retired veteran of the Persian Gulf War. lowing her graduation from Purdue University in May 2018, she will attend the University of Southern Califor- nia’s Master of Science in Marketing program and will Gary Evans is a continuing lecturer also pursue a Graduate Certificate in the Business of in the Krannert School of Man- Entertainment. agement, where he is the course coordinator and primary instructor Mentors for MGMT 30500, Business Sta- tistics. He has a master’s degree in Ben Lawton (PhD UCLA, 1976) Mathematics from Indiana Univer- was cofounder and, for many years, sity and a PhD in Statistics from Director of the Interdisciplinary UCLA. At UCLA, he was a statistics instructor and Film/Video Studies Program at consultant and was named Most Outstanding Computa- Purdue. He is a recipient of the tional Statistician. His research interests are in survey campus-­wide Amoco (now Mur- sampling design and analysis, multivariate analysis, and phy) Best Undergraduate Teacher optimization algorithms. Award. He was visiting professor

24 Journal of Purdue Undergraduate Research: Volume 8, Fall 2018 http://dx.doi.org/https://doi.org/10.5703/1288284316736 Introduction Abstract In my research, I hope to expand awareness of My research analyzed the representation of the lack of representation of women across sev- women in the film industry, both on screen and eral decades within the film industry by providing behind the scenes. Specifically, I compared the new data that confirms it. This longitudinal look number of women on and off screen for the top at underrepresentation brings light to the barrier 100 films of 2017 (as of September) to the data women face when entering the industry. My research collected by Martha Lauzen for the top 100 follows up on research conducted by Martha Lauzen films of 1980,1990, 2000, 2010, and 2015. This (2015) on the representation of women in the roles of comparison graphically depicts the representation producer (associate and co-­producers; line, super- of women in film over the years. The positions vising, and consulting producers were not included), analyzed were producers, executive producers, executive producer, writer, editor, cinematographer, directors, cinematographers, writers, and editors. and director in 1980, 1990, 2000, 2010, and 2015. By comparing the data collected by Lauzen to the In addition to researching representation in these representation of women in 2017’s top 100 films, we roles, I examined what factors, if any, are more can determine any improvement or deterioration of likely to influence the presence of women in other female representation. roles. The strongest statistical factor in determin- ing the presence of women behind the scenes is In addition to researching the historical representa- the presence of a female lead. tion of women in the industry, I conducted statistical research to determine which roles are most influential I also compiled data on the top 30 films directed in determining the number of women present behind by men and compared the return on investment the scenes (BTS) on a film. For example, I analyzed (ROI), budget allocation, box office earnings, and the connection between having a female lead and the experience (quantified by number of films and likelihood of hiring a female director. The purpose television episodes they had directed prior to the of this test is to determine which roles have the most film listed) to the top 30 films directed by women. influence. If we can determine links between roles, Statistical analysis concluded that ROI was not we can determine the most effective ways to combat significantly different between men and women lack of representation at each level. directors. Interestingly, however, the two highest ROIs, by far, were from films directed by women. The second portion of my research concerns the directorial role specifically. I compiled a database The budget disparity for men and women direc- of the top 30 films directed by men and the films’ tors of Hollywood films is often noted. Statistical respective budget, gross box office earnings, return-­ tests also concluded that experience was not on-­investment (ROI), and the experience of the significant in determining budgets. director (determined by number of past directed films and television episodes) and compared it to the same Miller, K. R. (2018). A sturdy glass ceiling: categories for the top 30 films directed by women. Representation of women on screen and The purpose of this database was to determine if ­behind the scenes of Hollywood’s top 100 there was a significant difference in budget allocation, films throughout the years.Journal of Purdue ROI, experience, and box office earnings between the A Sturdy Glass Ceiling ­Undergraduate Research, 8, 24–31. https://doi genders. Various statistical analyses were undertaken .org/10.5703/1288284316736 to determine significance of differences. Specifically, ROI was tested to determine if female-­directed films Keywords result in less return than male-­directed films. Expe- rience was tested as a factor in budget allocation to gender, women’s studies, Hollywood, film determine if men received higher budgets based on ­studies, pay gap, directors, glass cliff, glass experience or on other factors. ­ceiling, sexism

Prior Research

The research collected and presented by Mar- tha Lauzen (2015) covers 1980, 1990, 2000, 2010, and 2015. This data looks at the top 100, 250, and

25 500 films from each of those years. Lauzen has continuously collected this data annually as “the longest-­running and most comprehensive study of women’s behind-the­ scenes employment in film avail- able”(Lauzen, 2015). The research shows that women have consistently been underrepresented across all fields in every year. Some numbers have improved slightly over time while others have remained stag- nant. Women are most underrepresented in the roles of cinematographer, writer, and director. The fields with the highest representation of women are producer and executive producer. Overall, not one role saw more than 22% women in these fields from 1980 to 2015. Furthermore, the percentage of women editors has declined over the years from its high of 20%, in 2010 and 2016, to 13%. This is the lowest percentage of women editors since 1980. Women face the biggest disparity in the role of cinematographer, accounting Figure 2. Number of male leads compared to female for only 3% of cinematographers at their peak. leads in 2017. These numbers are from the top 100 films of 2017 (as of September). In 2017, representation in editing declined from the 2010–2015 high of 20% to 13% (Lauzen, 2015). Rep- resentation in directing increased from 7% in 2015 to 14% (Lauzen, 2015). However, the peak percentage for women directors was 20% in 2010 (Lauzen, 2015). In 2017, no role surpassed 26% for female repre- sentation. In addition to representation behind the scenes, the percentage of female-­led movies was also calculated. In 2017, female-­led films accounted for just 30% of the top 100 films. A common response to the discrepancy of male and female leads is based on the assertion that men make up a higher percentage of the movie-­going population. However, the Motion Picture Association of America (2016) has concluded Figure 3. Representation of women in the role of director­ that females consistently make up a higher portion 1980–2017. of theater patrons. Although the difference between the numbers of male and female moviegoers is quite small, the representation of on-­screen leads ideally should be closer to 50/50 to reflect the actual popula- tion demographics of the United States and audiences.

Figure 4. Representation of women in the role of writer 1980–2017.

Figure 1. Disparity between men and women in the be- hind-the-scenes roles of the film industry. These numbers are from the top 100 films of 2017 (as of September).

26 Journal of Purdue Undergraduate Research: Volume 8, Fall 2018 A Sturdy Glass Ceiling - - 27 VIF 1.00

AIC 75.88 SE Coef 0.546 0.105 95% CI (1.0939, 1.6531) ole ’)) y R y Y y b y Deviance R-­Sq(adj) 9.68% Coef –2.927 0.296 Odds Ratio 1.3447 ’)/(1 + exp( Y Disparit Binary logistic regression: Female director versus versus director Female 1. Binary regression: logistic Table BTS. Women ’ = –2.927 + 0.296 Women BTS ’ Model Summary Deviance R-­Sq 10.92% Coefficients Term Constant BTS Women Odds Ratios for Continuous Predictors Continuous Odds Ratios for BTS Women Regression Equation Regression P(1) = exp( Y Gender Influence of Women in Different Roles Roles Women in Different Influence of on Other Roles With my research on 2017’s top 100 films, top 100 I sought research my With 2017’s on to analyze any links to between in role women one regres inthe Using presence women other roles. of the at analysis, connections I looked sion between directors,female leads, female and the total number behind women theof scenes. First, I generated a binary regression logistic directors female on versus behindwomen the scenes. A binary regres logistic is used tosion “predict the relationship between independent and dependent variables where the dependent variable is binary” Solutions, (Statistics The regression estimates thatn.d.). if there are 0 behindwomen the scenes, the probability that a if there is directorfemale will be is chosen 5.1%; 1 other behind woman the scenes, the probability Theseincreases results are to 6.7%. considered a However, statistically significant,slightly. only but binary Director Female regression logistic versus of that when shows BTS Lead andFemale Women controlling lead, a female the for impact women of behind the scenes the on assignment a female of director(regressionomitted). significant is not table thisTaking into account, was determined it that having lead a female is the significant most factor director. a female choosing when I conducted another binary regression logistic specifically to isolate relationshipthe that having a leadfemale has having on director. a female Using the regression equations, the having probability a of - - Representation of women in the role of cinema- role in the of women 5. Representation Figure 1980–2017. tographer - of execu in the role of women 7. Representation Figure 1980–2017. producer tive Representation of women in the role of editor of editor in the role of women 6. Representation Figure 1980–2017. Lauzen and Dozier present a hypothesis the lack for representation:of pursue perhaps do not women theseheavilyas fields menas do cannot and therefore reach equal representation. In order to evaluate this hypothesis, Lauzen requested the enrollment statis If women make up 50% of the popu United States’ of make upIf 50% women andlation are underrepresented not in film programs around the that country, is reasonable to conclude it berepresentation higher should than women of the representation presented in the data. tics for the toptics six for film Unitedin schools States. the studentsFemale underrepresented were only three at (Lauzen those schools of 1999). & Dozier, Model Summary Model Summary Deviance Deviance Deviance Deviance R-­Sq R-­Sq(adj) AIC R-­Sq R-­Sq(adj) AIC 27.52% 26.28% 62.49 9.13% 8.66% 455.32 Coefficients Coefficients Term Coef SE Coef VIF Term Coef SE Coef VIF Constant –3.512 0.718 Constant 1.0212 0.0651 Female Lead 3.106 0.809 1.00 Female Director 0.630 0.134 1.00 Odds Ratios for Continuous Predictors Regression Equation Odds Ratio 95% CI Women BTS = exp(Y’) Female Lead 22.3333 (4.5781, 108.9490) Y’ = 1.0212 + 0.630 Female Director Regression Equation Table 4. Poisson regression analysis: Women BTS versus P(1) = exp(Y’)/(1 + exp(Y’)) female director. Y’ = –3.512 + 3.106 Female Lead

Table 2. Binary logistic regression: Female director versus a female lead is predicted to increase women behind female lead. the scenes by 2.17. female director with a male lead is 0.03. However, the Next, female directors appear to have a fairly strong, probability of having a female director with a female noteworthy relationship to women behind the scenes; lead is 0.40. the model predicts that, with a female director, women behind the scenes increases by 1.88. In addition to the logistic regressions, I conducted Poisson regressions (Brilliant n.d.) to analyze the However, the third model shows that when both relationships in detail. Poisson regression is a type of female lead and female director are accounted count regression used when the dependent variable for, and female lead is controlled, the effect of is a whole number. I computed three Poisson regres- female directors is minor and likely not statistically sions: Women BTS versus Female Lead, Women significant for women behind the scenes. Thus I BTS versus Female Director, and Women BTS versus conclude that the apparent female director effect is Female Director and Female Lead. The relationships actually masking or working through association of these variables are comparable to the earlier logis- with the female lead effect and, therefore, having a tic regressions. The regression does show a notewor- female lead is by far the strongest factor in having thy relationship between women behind the scenes a female director and women in other behind-the-­ ­ and female leads. However, the fit for the model is scenes roles. not good and therefore determinate statements about statistical significance cannot be made, although the effect does appear to be strong. In particular, having Model Summary Deviance Deviance Model Summary R-­Sq R-­Sq(adj) AIC Deviance Deviance 21.60% 20.68% 430.41 R-­Sq R-­Sq(adj) AIC Coefficients 20.81% 20.35% 430.12 Term Coef SE Coef VIF Coefficients Constant 0.8283 0.0795 Term Coef SE Coef VIF Female Director 0.202 0.153 1.31 Constant 0.8348 0.0793 Female Lead 0.695 0.130 1.31 Female Lead 0.775 0.114 1.00 Regression Equation Regression Equation Women BTS = exp(Y’) Women BTS = exp(Y’) Y’ = 0.8283 + 0.202 Female Director + 0.695 Y’ = 0.8348 + 0.775 Female Lead ­Female Lead

Table 3. Poisson regression analysis: Women BTS versus Table 5. Poisson regression analysis: Women BTS versus female lead. female director and female lead.

28 Journal of Purdue Undergraduate Research: Volume 8, Fall 2018 A Sturdy Glass Ceiling ­ - 29 - - Fem Fem Value = ­Value Value ­Value SE Mean 0.67 0.67 SE Mean 1.5 0.80 μ (ROI-­ StDev 3.61 3.53 male versus ROI- male versus male (outliers StDev 8.12 4.37 Value = 1.42 P- ­Value Value = 0.06 P- ­Value Mean 6.50 5.16 N 29 28 0.555, 3.234) 3.29, 3.50) test and CI: ROI- ­ test and CI: ROI-­ test Mean 7.07 6.97 μ (ROI_1) Male (Outlier Del))Male -­ (Outlier ­ T- sample ­ T- sample N 30 30 Female (outliers del.). Female Male (Outlier Del)Male (Outlier Del.) (Outliers Fem (Outliers Del.)) (Outliers 0.162 DF = 54 = 0.951 DF = 44 Differences in the in male andROI between 9. Differences Figure directors. female ­ Two- 7. Table ­ Two- 8. Table female. ROI-­ del.), Test of difference = 0 (vs ≠): T- ≠): = 0 (vs of difference ­Test Test of difference = 0 (vs ≠): T- ≠): = 0 (vs of difference ­Test Difference = μ (ROI-­ Difference 1.340 difference: for Estimate (-­ difference: 95% CI for T- ROI-­ ROI-­ ROI ROI 1 = μ (ROI) -­ Difference 0.10 difference: for Estimate (-­ difference: 95% CI for T- ence statistically is not significant in determining budgetary allotment considering even inter possible action. The results that gender a strong is such show predictor budgetary of allotment that experience essentially has effect. no The last test conducted regression was a multiple analysis to determine if experience was a signifi cant factor in determining budgetary allotment. A common justification to budgetary the gap between male and directors female is that male directors experience more have andsimply deserve therefore to a higher be given The budget. regression analyses, found hadwhich fit, quiteexperithat model good - - ­sample Value ­Value

Wonder SE Mean 14 8.8 StDev 75.1 48.4 Between test to determine Value = 8.10 P- ­Value μ (Women Budget) Gap Mean 184.0 51.8 ­ t- ­sample test and CI: Male test budget versus Directors N 30 30 ­ T- sample y Allotment y Female ). Also, female directors Also,female ). and were, contin and = 0.000 DF = 49 Differences in the allocated budgets for male budgets in the allocated 8. Differences Figure an outlier. represents star point The directors. and female female budget. female ­ Two- 6. Table Test of difference = 0 (vs ≠): T- ≠): = 0 (vs of difference ­Test ale test. The results determined that in men were, fact, Estimate for difference: 132.2 difference: for Estimate (99.4, 165.0) difference: 95% CI for T- Male Budget Budget Women = μ (Male Budget) -­ Difference M Budgetar if ROI was significantly ifdifferent ROI betweengenders. If this the were case, argue could one that female directors are less due to paid their returns lack of the test determinedHowever, office. theat box that returnmen and women similar amounts when even controlling outliers. This for suggests that despite their budgets for receiving films,women lower they return similar amounts as men. gets, I conducted a two- uously are, consistentlyuously less than paid their male counterparts. As part research, my of I analyzed explanations behindseveral possible this discrepancy. In order to accurately represent the experiences of inmen and women directing, I created a database of the films top 30 directedFirst,women.menby and to determine if men actually were than more paid inwomen the directing I conducted a two- role, In 2017, 14 filmswere directed 14 women, by including In 2017, the second highest grossingyear filmof the ( After determining that higher men received bud Woman ­ T- significantlygiven higherbudgetary allotments. Model Summary S R-­sq R-­sq(adj) R-­sq(pred) 63.6488 53.24% 51.60% 48.43% Coefficients Term Coef SE Coef T-­Value P-­Value VIF Constant 47.8 14.5 3.30 0.002 Male Dir 131.4 16.5 7.95 0.000 1.01 Prior Exp 0.66 1.44 0.46 0.646 1.01 Regression Equation Budget ($Mil) = 47.8 + 131.4 Male Dir + 0.66 Prior Exp

Table 9. Model 1—no interaction effect.

Model Summary S R-­sq R-­sq(adj) R-­sq(pred) 63.7551 53.91% 51.44% 46.27% Coefficients Term Coef SE Coef T-­Value P-­Value VIF Constant 53.1 15.6 3.39 0.001 Male Dir 112.4 26.8 4.19 0.000 2.65 Prior Exp –0.22 1.74 –0.12 0.902 1.48 Male Dir:Exp 2.79 3.10 0.90 0.372 3.29 Regression Equation Budget ($Mil) = 53.1 + 112.4 Male Dir -­ 0.22 Prior Exp + 2.79 Male Dir:Exp

Table 10. Model 2—with interaction.

Overall Suggestions and Conclusion Finally, the research shows that although female directors do have some effect on the presence of This research shows that women are underrepre- other women behind the scenes, the strongest deter- sented in the roles of producer, executive producer, minant is having a female lead. When accounting cinematographer, writer, director, editor, and lead for a female lead, a female director has almost no for every year analyzed. Some roles have shown an effect. One could conclude that we simply need more increase in representation, albeit a miniscule one. female-­lead movies in order to solve the represen- There is also evidence that some roles have seen tation gap. However, that would be acceding to declining representation or stagnant representation standard sexist ideology. If we say that having more throughout the past 38 years. The roles of cinema- female-­lead movies is the solution, it puts us on a tographer, writer, and editor see the lowest percent- slippery slope to implying women should work only age of women. on female-­lead films and men should work only on male-­lead films. This study also demonstrates that female directors are statistically more likely to receive a lower Over the years, women have been fighting for their budget compared to their male counterparts. How- chance at creating, producing, and developing sto- ever, even when controlling for outliers, women ries to be shown on the big screen. Successes from and men have comparable ROIs. This shows that certain women in Hollywood were thought to have men and women make similar returns given their shattered any potential glass ceiling by the 2000s. different budgetary allotments. The research also However, with this data and the data collected shows that experience is not a statistically signif- continuously by Martha Lauzen (2015), we see icant factor in determining budget. This counters the same picture we have been seeing since 1980. the popular idea that budget is based on prior direc- Women simply are not given the same opportunities torial experience. in Hollywood as men are. Even when women are

30 Journal of Purdue Undergraduate Research: Volume 8, Fall 2018 A Sturdy Glass Ceiling ​ ​ ​ - ​

​ - ​ 31 ​ - ​ ​ ​ ​ ​.html ­the​ ​. e d u ​/tv - ​www ­up ​- ​www ​/ m a r k ­in ­scenes ​- ​/binary ­Theatrical ​journals ​- ​. s d s u ​/2018 ­times​- ­the- ​/ n e w s ​- ­a8153571 ​.com ​.com ​-­money ​/MPAA ​/ fi l m s ​. p d f ­the ​-­salary ​- ​- ­actors

​- ​variety ­all ​- ​. p d f ​w o m e n i n t v fi m l ­crown ​/03 ​/2018 ­1 ­pay ​- ​- ​_ R e p o r t ​/0196859999023004004 ​- ​-­screen Retrieved from https:// ­the ­distribution/ ​- ​- ​.1177 ​.statisticssolutions ­entertainment ​_ F i n a l ​- ​-­scott ​/uploads ​/10 ​_ C e i l i n g ­williams ​- ​-­smith ​/arts ​www ­2016 ​- ​/abs ​/poisson ­ridley ​.uk ­matt ­content ​- ​- ​/doi ​.co Retrieved from http:// ​/wiki ­michelle ​_ C e l l u l o i d ​- ​/wp ­regression/ ­Statistics ​.com ​- ​.org ​- ​- ­reshoots ​- ​.org ​/ 2 0 1 5 Variety. ­Market ­logistic ­world ­walhberg ­1202729883/ trieved from http:// - - screen and behind the scenes in the television and film industries: Review a program of research. of Journal of 23(4), 355–372. Inquiry, Communication http:// .sagepub #articleCitationDownloadContainer .mpaa - .independent rical market statistics Retrieved 2016. from https:// - / fi l e s Mark pay disparity Walberg investigated the by Screen Actors Guild. Independent. - employment of women on and the top 500 250, 100, films Retrieved fromof 2015. h t t p s : / / ​/petition /news brilliant donate parthis of salary from ‘The to Crown’ Time’s Up. References Statistics Binary Solutions. logistic (n.d.). regression. Re Lauzen, M., & Dozier, D. (1999). The women on of role Lauzen, (1999). & Dozier, M., D. Motion Picture theat Association MPAA America. of (2016). O’Connor, R. (2018, January 11). Michelle Williams January and R. 11). (2018, O’Connor, The celluloid ceiling: Behind- ceiling: The celluloid Lauzen, M. (2015). opportunity to conduct this research as faculty my advisorthrough WILKE Purdue Under University’s graduate also like I would Research Scholarship. to thank GaryDr. the Krannert of Evans of School Management his continued for assistance in thesta tistical the data analysis of presented and providing mentorship throughout the project. Brilliant. n.d. Poisson distribution. Retrieved from https:// Petition calls March on Matt 18). SmithClarke,to S. (2018, - - Up Up , despite actress (O’Connor, 2018), and petitions drafted were 2018), (O’Connor, cknowledgments A I would like toI would thank Benjamin Professor Lawton, Interim Director the Studies and Film Video of InterdisciplinaryProgram of the School of Studies Liberal Arts, of in the College giving for the me tion comes from the respective film pages listedon IMDB and Wikipedia. Claire Foy being the starClaire Foy character, to the Times-­ ence in salary made he The on Crown requesting that actor Matt Smith donate the differ This article cites information acquired the on top 100 (determined September) of (asby gross films2017 of office earnings)box and 30 filmsthe top directed by throughoutmen and women This history. informa In 2017 alone, several film alone, producers and television In 2017 castigatedwere their disparities for pay between andfemale male costars. reportedly Mark Wahlberg times” the salary costar “1,500 received of Michelle the MoneyWilliams in the All the reshoot of for World Going forward, further research studies be should conducted behind the the on presence women way of the scenes, lack thereof, or affects the portrayal of screen on women and vice versa. given thegiven opportunity, they are frequently underpaid compared to male counterparts. is through It 2018). Legal FundDefense (Clarke, continued research thatcan begin we to understand and analyze the prejudices that in affect women the film industry.