Close Menu
    Screenwriting

    Frontiers | Women in directing and screenwriting: the capital allocation puzzle in commercial film

    By No Comments40 Mins Read

    Women in directing and screenwriting: the capital allocation puzzle in commercial film

    • Media Business Division, Department of Communication, Johannes Gutenberg University Mainz, Mainz, Germany

    Abstract

    Women remain underrepresented as directors and screenwriters in commercial film, yet the financial logic of this pattern is unclear. This study examines gender disparities in creative authorship using IMDb data on 199,656 feature-length narrative films released worldwide between 1994 and 2023. Across the sample, 73.8% of films credited no women in either role. Participation increased gradually, with no sustained acceleration in its rate of growth following the #MeToo movement. Using a financial subset of 4,245 commercially documented, predominantly U.S. theatrical films with verified budget and revenue data, the analysis compares production budgets, profit, and return on investment across four mutually exclusive groups. Films with a woman screenwriter and no woman director showed 25% higher median ROI and 34% higher median profits than male-only productions despite comparable budgets, with the ROI gap widening to 33% within the top 10% of production budgets in exploratory tier-level comparisons. These differences are descriptive; in a genre-adjusted robustness check, the woman-writer ROI advantage attenuates but remains significant. Films with women in both directing and screenwriting roles matched male-only returns while operating on approximately half the budget. Representation contracted sharply at higher investment levels. Across thirty years, no film with a woman director or a woman in both roles appeared in the top 1% of production budgets. Audience ratings were comparable across groups and visibility differences largely disappeared within the highest-budget tiers. These findings constitute an economically puzzling disconnect in which observed return distributions do not straightforwardly account for the allocation of production capital.

    1 Introduction

    Women remain substantially underrepresented as directors and screenwriters in commercial film production. Across a large-scale global sample examined in this study, women held directing roles without screenwriting credits in 4.5% of films and screenwriting roles without directing credits in 11.2%, with 10.5% featuring women in both roles and 73.8% crediting no women in either.

    This pattern has deep historical roots. Female representation in behind-the-camera roles declined sharply with the emergence of the Studio System in the early twentieth century and has never fully recovered (; ). That it persists today, however, is economically puzzling. Motion picture production is characterized by extreme uncertainty and highly skewed returns (; ; ). In such an environment, observable performance outcomes should inform subsequent capital allocation decisions. Yet disparities in directing and screenwriting have narrowed only slowly, despite growing evidence that gender-based performance differences diminish substantially once differences in project assignment, budgets, and distribution are taken into account (; ).

    The puzzle resembles other allocation patterns documented in the industry, such as the continued production of R-rated films despite profit distributions that are stochastically dominated by PG and PG-13 films (; ). In both cases, upstream investment decisions appear only weakly aligned with downstream performance distributions. Films involving women in directing or screenwriting roles similarly receive smaller production budgets and narrower distribution, constraining observed revenue potential and visibility independently of audience preferences (; ).

    Building on prior work that established this misalignment exists, this paper makes three contributions. First, it provides a unified analysis that jointly tracks participation in screenwriting and directing roles, audience engagement, and financial returns across the same sample – demonstrating that the allocation gap persists independently of both audience demand and observed financial performance. Second, it uses return on investment rather than gross revenue as the primary financial outcome, directly accounting for production cost and therefore more accurately reflecting capital efficiency. Third, it demonstrates that the allocation-performance gap is most pronounced at the highest budget levels – precisely where capital decisions carry the greatest financial consequences. The finding that films with a woman screenwriter and no woman director deliver 33% higher median return on investment than male-only films within the top 10% of production budgets – a pattern based on small cell sizes and most pronounced in these higher tiers – sharpens the puzzle considerably. Allocation patterns appear inconsistent with realized capital efficiency in a segment that generates comparatively strong financial returns. Why this pattern persists remains an open question that the evidence presented here is intended to motivate.

    Economic research on the motion picture industry has established that revenues are highly uncertain and that observable signals such as sequels, advertising expenditures, genre, star power, and critical information play central roles in shaping box-office outcomes (; ; ; ). Returns are heavily skewed, with a small number of blockbusters accounting for a disproportionate share of industry profits (). Within this environment of extreme uncertainty, recent work has examined whether gender influences economic outcomes.

    , using longitudinal data on feature film screenwriters from 1982 to 1992, find that the gender gap in earnings grows as men and women advance through their careers, with an initial gap of 4 to 6% at career entry widening to 40% or more by the fifteenth year. This pattern is consistent with cumulative disadvantage dynamics in which early-career differences compound over time. document a substantial unexplained earnings gap among Hollywood stars, driven primarily by action films. , using longitudinal data on French directors, find that the returns to box-office success are at least twice as large for women directors as for men in terms of both career survival and future earnings – a higher performance threshold that holds specifically in the commercial sector and is most pronounced for low and mid-achieving women. show that biases in project assignment, budgeting, and distribution explain apparent performance differences between male and female directors; once endogeneity and selection are accounted for, revenue gaps disappear. Parallel patterns appear in other markets. In Germany, women-directed films receive lower public funding, smaller budgets, and narrower releases despite strong artistic performance and efficient use of funding (; ). In the Russian market, male dominance in screenwriting and other creative positions is correlated with larger budgets and better distribution, while audiences showed no attendance penalty for films featuring central female protagonists in the period after 2015 ().

    On the demand side, evidence does not support a systematic consumer penalty for female representation. Films passing the Bechdel test or containing meaningful female dialogue generate higher domestic and international box-office revenues (; ; , ). Female-lead films elicit less consensus in audience evaluations, with higher dispersion and more positively skewed rating distributions, and independent studios generate higher box-office revenues from such films (). These findings suggest that observed disparities are more plausibly rooted in upstream allocation processes than in downstream consumer rejection.

    Earlier research demonstrates that women directors and writers are disproportionately concentrated in lower-budget projects, limiting revenue potential independently of audience preferences (; ). The present study extends this literature by providing a unified thirty-year analysis that jointly tracks labor supply (participation in directing and screenwriting), audience engagement (ratings and vote counts), and institutional investment (budgets and financial returns) across the full distribution of films and within the highest-budget tiers.

    2 Materials and methods

    2.1 Data

    The primary film data were drawn from the IMDb Non-Commercial Datasets (), covering global theatrical and non-theatrical releases. The sample was restricted to feature-length narrative films running at least 60 min and released between 1994 and 2023. The 1994 start point follows the period covered by the AFI Catalog’s data () on the first century of American cinema (1893–1993) and examined in prior work on the first century of women’s filmmaking participation versus recognition (). The present study extends that record forward. After removing duplicates, entries missing required fields (title, year, rating, votes, or credits), and films with unclassifiable gender credits (see Gender Classification below), the dataset contained 199,656 unique films. Metadata included title, release year, IMDb rating, vote count, and credited directors and screenwriters.

    To incorporate financial performance, production budgets and worldwide gross revenues were obtained from The Numbers (). Because IMDb and The Numbers maintain separate databases with different naming conventions, the two sources were merged using a fuzzy-matching procedure implemented with the RapidFuzz Python library (). Titles were aligned using a token-sort similarity threshold of at least 90 percent, with candidate pairs additionally required to share the same release year. Because matches were restricted to films from the same year, sequels and remakes released in different years could not be incorrectly paired despite similar titles.

    The financial subset of 4,245 films represents the complete set of The Numbers records that successfully matched to the IMDb dataset at this threshold, approximately 76% of the 5,611 The Numbers records within the study window that carried budget and revenue data. Match quality was high, with 98% of matched pairs being exact title matches within release year, and inspection of the lowest-scoring matches at the 90% threshold confirmed that the remaining differences reflected punctuation and subtitle formatting rather than incorrect pairings. All monetary values were converted to 2023 U.S. dollars using the Consumer Price Index for All Urban Consumers (). Because publicly reported financial data are concentrated in commercially prominent, predominantly U.S. theatrical releases, the financial subset represents this documented segment rather than global commercial filmmaking, and financial results should be interpreted accordingly.

    Two analytic tiers were used. The full sample (N = 199,656) was used to examine participation patterns and audience engagement. The financial subset (n = 4,245) was used to analyze budget allocation, profit, and return on investment.

    2.2 Gender classifications

    Gender classification used Wiki-Gendersort (), a freely available tool that assigns gender labels based on first names. The resulting classifications therefore reflect name-based patterns rather than independently verified information about the individuals credited. Prior evaluations show that the algorithm achieves accuracy above 95 percent for Western names () and above 90 percent in validation tests on large international datasets (). Only names confidently classified as male or female names were retained; ambiguous or unclassifiable entries were excluded during data cleaning. Across all credited directors and screenwriters in the data (327,456 unique individuals), Wiki-Gendersort classified 69.0% as having male names and 20.7% as having female names; the remaining 10.3% were excluded as unclassifiable (6.3% unknown, 3.8% unisex, and 0.2% initials only). This binary classification is consistent with the approach adopted in prior economic studies of gender representation in the film industry (; ).

    2.3 Variable construction

    Films were categorized into four mutually exclusive authorship groups based on name-based gender classification of credited directors and screenwriters:

    • –

      WOMAN_DIRECTOR: films with at least one director with a woman’s name and no screenwriter with a woman’s name.

    • –

      WOMAN_WRITER: films with at least one screenwriter with a woman’s name and no director with a woman’s name.

    • –

      DUAL_ROLE: films with at least one woman’s name in both directing and screenwriting roles.

    • –

      MALE_ONLY: films with no women’s names in either role (reference category).

    For readability, the remainder of the paper refers to these groups using “woman director,” “woman screenwriter,” and related shorthand. In all cases, these denote the name-based classifications defined here.

    Dependent variables captured audience engagement and financial outcomes. For audience engagement:

    • –

      RATING: the mean IMDb audience score (1–10).

    • –

      VISIBILITY: defined as ln(VOTES), where VOTES is the total number of IMDb user votes. The natural log transformation was applied because raw vote counts are highly right-skewed, with a small number of blockbuster titles attracting large audiences and most films receiving comparatively few votes. For the financial subset, raw vote counts are reported because this sample comprises commercially prominent releases with more uniform distributions.

    For financial outcomes, available in the subset of 4,245 films:

    • –

      BUDGET: CPI-adjusted production costs (2023 USD).

    • –

      PROFIT: gross revenue minus budget (2023 USD).

    • –

      ROI: return on investment, calculated as (Gross−Budget)/Budget. By normalizing returns against production cost, ROI captures capital efficiency more directly than gross revenue. This theatrical-gross measure excludes marketing and distribution costs as well as ancillary revenues such as home video, streaming, and merchandising.

    ) was used to examine temporal trends in participation rates.

    2.4 Analytical strategy

    The empirical analysis proceeded in five stages. First, participation patterns were examined by calculating annual proportions of films in each authorship category from 1994 to 2023. To summarize long-run trends, simple linear regressions of participation rates on release year were fitted to quantify trend slopes, with slope coefficients and R2 values reported to describe trend magnitude and stability. To assess whether participation trends shifted around the 2017 #MeToo movement, a Chow test and an interrupted time series specification (modeling a level shift and a slope change at 2017) were estimated.

    Second, audience reception was compared across authorship groups. One-way ANOVA tested for differences in mean RATING, followed by Tukey-Kramer post hoc comparisons for pairwise contrasts. VISIBILITY was compared using ln(VOTES) as defined in Variable Construction.

    Third, budget allocation and financial outcomes were assessed within the financial subset. Because BUDGET, PROFIT, and ROI exhibit non-normal distributions with heavy upper tails, medians are reported as the primary measure of central tendency. Ratings, by contrast, are bounded and approximately symmetric, and are therefore summarized using means. Kruskal–Wallis H tests were used to assess whether budget distributions differed significantly across authorship groups; ROI and PROFIT comparisons are reported descriptively given the focus on median differences. The group comparisons in this stage are unadjusted; a genre-adjusted robustness check is reported separately (see below).

    Fourth, to examine whether disparities intensify at higher investment levels, median BUDGET, ROI, PROFIT, VOTES, and RATING were reported separately for films in the top 10%, 5%, and 1% of production budgets.

    Fifth, to assess whether the ROI differences across authorship groups reflect genre composition, a genre-adjusted robustness check was conducted. ROI was regressed on authorship group (male-only as reference), one-hot encoded genre indicators (Drama, the most common genre, as reference), and release year. Production budget was not included, as it enters the ROI denominator and would create a mechanical dependence between outcome and predictor. Because ROI is heavily right-skewed, robust (Huber) regression was used (). A fuller treatment of genre as a structuring factor in capital allocation is beyond the present scope.

    All analyses were conducted in Python 3.11 using pandas (), NumPy (), and statsmodels ().

    2.5 Limitations

    Several limitations should be noted. The gender-classification process excluded names that could not be confidently assigned as male or female. A known limitation of Wiki-Gendersort is its reduced accuracy for romanized Chinese names, where gender cannot be reliably inferred from Pinyin spellings (). As a result, women filmmakers from certain linguistic backgrounds may be undercounted. Relatedly, the authorship groups are coded on an at-least-one-woman basis, which collapses variation in the proportion of women credited. A proportion-based specification would permit finer distinctions and is left to future work.

    A second set of limitations concerns sample composition. Publicly reported financial data skew heavily toward higher-budget theatrical releases, disproportionately excluding independent and non-theatrical productions where women creatives may be more active. The woman-director group (n = 83) may be particularly affected, given women’s disproportionate activity in lower-budget and independent production absent from this subset. The full sample, by contrast, encompasses the broad population of documented feature films rather than the commercial industry narrowly. It includes lower-visibility and non-theatrical productions, and country and production-type breakdowns were not retained in the analytic dataset, limiting finer characterization of its composition. Because incomplete records are also more common for lower-profile productions, where women are more active, this exclusion may undercount women’s participation.

    Finally, reception measures carry their own limitations. IMDb vote counts provide an imperfect proxy for visibility, as platform engagement skews toward English-language and U.S.-distributed films and toward male users, which may further depress engagement and ratings for films with women in creative leads. Reception is measured using IMDb audience ratings rather than professional critic scores. This distinction is informative for gender-comparative work. Professional critics tend to rate male- and female-fronted films similarly, whereas the gender gap widens under crowd-based user ratings, with female films receiving disproportionately many extremely low scores driven largely by male raters (; ). User ratings therefore represent the more conservative measure for the present comparison.

    3 Results

    This section presents findings from 199,656 films released between 1994 and 2023. Financial analyses are based on a subset of 4,245 films with verified budget and revenue data. Results are organized into five parts: participation trends, audience reception, budget allocation and financial returns, high-budget tier analysis, as well as a genre-adjusted robustness check.

    3.1 Participation trends (1994–2023)

    Analysis of 199,656 films revealed persistent underrepresentation of women in directing and screenwriting roles. Across the dataset, 73.8% of films (n = 147,373) credited no women in either role, while 10.5% (n = 20,893) involved at least one woman in both directing and screenwriting positions. Women held singular roles in 11.2% of films as screenwriters (n = 22,352) and 4.5% as directors (n = 9,038). Notably, among all films with a woman director, more than two-thirds (69.8%, or 20,893 of 29,931) also credit a woman screenwriter, indicating that women’s directorial participation rarely occurs independently of women’s screenwriting involvement.

    Growth patterns differed across categories. The most pronounced increase occurred in dual-role films, which nearly doubled from 6.7% to 12.6% over the study period (+0.21 percentage points annually, p < .001, R2 = .94). By 2019, dual-role films had surpassed woman-writer films to become the most common form of women’s creative involvement in the sample. The growth in dual-role films indicates increasing instances where women are represented across multiple creative positions within the same production. Woman-director participation rose from 3.4% to 5.1% (+0.05 percentage points annually, p < .001, R2 = .64). In contrast, woman-writer participation showed no meaningful change, remaining between 10% and 12% across the period (β = 0.01, p = .26, R2 = .04). Table 1 summarizes these trends.

    Category 1994 2023 Annual Change R2
    Women-inclusive 21.5% 29.4% +0.28%*** .94
    DUAL_ROLE 6.7% 12.6% +0.21%*** .94
    WOMAN_DIRECTOR 3.4% 5.1% +0.05%*** .64
    WOMAN_WRITER 11.1% 11.5% +0.01 .04
    MALE_ONLY 78.5% 70.6% −0.28%*** .94

    Trends in gender participation (1994–2023).

    N = 199,656. Annual change derived from linear regression of participation rates on release year. Women-inclusive=DUAL_ROLE+WOMAN_DIRECTOR+WOMAN_WRITER

    *** p < .001.

    Public attention to gender equity in film rose sharply around the 2017 #MeToo movement, raising the question of whether women’s participation accelerated in its wake. A Chow test indicated a break at 2017 (F = 7.90, p = .002), but an interrupted time series showed that this reflected a small one-time increase in participation (a level shift of roughly one percentage point, which did not reach significance, p = .055) rather than a change in the underlying trajectory. The annual rate of increase did not differ significantly before and after 2017 (slope change p = .16), and this held under a 2018 break specification (p = .32). Women’s participation thus continued rising at a similar pace through the late 2010s, with no sustained acceleration in its growth rate, though the short seven-year post-2017 window limits the power to detect gradual change. Figure 1 visualizes the annual share of films crediting at least one woman as director, screenwriter, or both. These trends indicate gradual but incomplete change. Even by 2023, fewer than one in three films credited women in either directing or screenwriting roles, and more than 70% of productions were exclusively authored and directed by men.

    3.2 Audience reception by gender group

    This subsection examines whether audience reception differs across gender authorship groups, using RATING as a measure of perceived quality and VISIBILITY as a measure of audience reach. Data are analyzed across the full dataset (N = 199,656) as well as the financial subset (n = 4,245).

    3.2.1 Full dataset

    In the full dataset, woman-director films received the highest average rating (6.66), followed by dual-role (6.43), male-only (6.19), and woman-writer (6.06). One-way ANOVA confirmed that differences in RATING across groups were statistically significant (p < .001), although the effect size was small (η2 = 0.008).

    Visibility patterns diverged from ratings. Despite achieving the highest ratings, woman-director films had the lowest log-transformed vote counts (4.12), compared to male-only (4.73), dual-role (4.58), and woman-writer (5.32). Woman-writer films received the lowest ratings yet garnered the highest visibility. Higher ratings did not correspond to greater visibility, indicating that factors beyond perceived quality shape audience reach (; ). Table 2 summarizes these findings.

    Category n Avg. Rating Avg. ln(Votes)
    WOMAN_DIRECTOR 9,038 6.66 4.12
    WOMAN_WRITER 22,352 6.06 5.32
    DUAL_ROLE 20,893 6.43 4.58
    MALE_ONLY 147,373 6.19 4.73

    Audience reception by gender group (1994-2023) – full dataset.

    N = 199,656. Rating is the mean IMDb audience score (1–10). ln(Votes) is the natural logarithm of IMDb vote counts

    3.2.2 Financial subset

    The financial subset revealed a different pattern (Table 3). Male-only films received the highest average rating (6.37) and the most votes (133,572). Dual-role films achieved comparable ratings (6.34, within 0.03 points) but 39% fewer votes (80,930). Woman-writer films received 23% fewer votes than male-only (102,207 vs. 133,572) with slightly lower ratings (6.28). Woman-director films had the lowest average rating (6.13), 0.24 points below male-only, with 34% fewer votes (88,043). Compared with the full dataset, the financial subset shows narrower rating differences across gender groups, with all groups within 0.24 points of male-only films, but persistent gaps in audience reach remain.

    Category Avg. Rating Avg. Votes
    WOMAN_DIRECTOR 6.13 88,043
    WOMAN_WRITER 6.28 102,207
    DUAL_ROLE 6.34 80,930
    MALE_ONLY 6.37 133,572

    Audience reception by gender group (1994–2023) – financial subset.

    n = 4,245. Subset includes films with verified budget and revenue data

    3.3 Budget allocation and financial returns

    This subsection examines production budgets and financial outcomes across gender authorship groups within the financial subset (n = 4,245).

    3.3.1 Visibility distribution

    To complement the mean visibility measures reported for the full dataset, Table 4 presents the distribution of vote counts across gender groups in the financial subset. Male-only films had the highest median votes (75,294). Woman-writer films achieved 60,678 median votes, 19% below male-only. Woman-director and dual-role films had the lowest median votes (40,735 and 37,818 respectively), approximately half that of male-only films. Similar patterns held across the distribution, with male-only films showing higher vote counts at both the 25th percentile (22,199 vs. 12,347–21,586) and 75th percentile (184,825 vs. 93,574–158,041). These visibility gaps mirror the budget disparities documented in the following section.

    Category 25th Percentile Median Votes 75th Percentile
    WOMAN_DIRECTOR 13,006 40,735 93,574
    WOMAN_WRITER 21,586 60,678 158,041
    DUAL_ROLE 12,347 37,818 102,793
    MALE_ONLY 22,199 75,294 184,825

    Distribution of IMDb vote counts by gender group (1994–2023).

    n = 4,245. Vote counts are raw

    3.3.2 Budget and financial outcomes

    Table 5 summarizes budget, profit, and return on investment (ROI) by gender group. Kruskal–Wallis tests confirmed that budget differences across groups were statistically significant (H = 46.91, p < .001).

    Category Median ROI Median Budget Median Profit n %
    WOMAN_DIRECTOR 0.44 $30.4M $13.8M 83 2.0
    WOMAN_WRITER 1.10 $41.4M $35.9M 752 17.7
    DUAL_ROLE 0.89 $19.9M $12.9M 217 5.1
    MALE_ONLY 0.88 $39.2M $26.8M 3,193 75.2

    Return on investment, budget, and profit by gender group (1994–2023).

    n = 4,245. Budgets and profits adjusted to 2023 USD using CPI-U ()

    Dual-role films operated on the lowest median budget ($19.9M), approximately half that of male-only films ($39.2M). Woman-director films received 22% less than male-only ($30.4M vs. $39.2M). Woman-writer films, with budgets comparable to male-only ($41.4M vs. $39.2M), generated the highest median profit ($35.9M), exceeding male-only by 34%. Woman-writer films also achieved the highest median ROI (1.10), 25% above male-only (0.88). Dual-role films achieved nearly identical ROI to male-only (0.89 vs. 0.88) despite receiving half the budget. Woman-director films had the lowest median ROI (0.44), the one group that departs from the broader pattern. This result should be interpreted cautiously given the small sample size (n = 83).

    3.4 High-budget tier analysis

    This section examines gender representation, financial outcomes, and audience perception among the highest-budget films in the financial subset: top 10% (n = 440), top 5% (n = 220), and top 1% (n = 44) of production budgets.

    3.4.1 Representation in high-budget tiers

    Gender disparities widened as production budgets increased. Male-only films dominated the highest-budget categories, accounting for 74.8% of the top 10%, 78.6% of the top 5%, and 88.6% of the top 1% (Table 6). Representation of women in directing roles contracted sharply with rising budgets. Woman-director films constituted only 1.4% of the top 10%, 0.5% of the top 5%, and none appeared in the top 1% across the entire study period. Dual-role films followed a similar pattern, present in the top 10% (3.4%) and top 5% (4.1%) but absent from the top 1%.

    Category Baseline Top 10% Top 5% Top 1%
    % of tier
     WOMAN_DIRECTOR 4.5% 1.4% 0.5% 0%
    WOMAN_WRITER 11.2% 20.5% 16.8% 11.4%
    DUAL_ROLE 10.5% 3.4% 4.1% 0%
    MALE_ONLY 73.8% 74.8% 78.6% 88.6%
    n
    WOMAN_DIRECTOR 6 1 –
    WOMAN_WRITER 90 37 5
    DUAL_ROLE 15 9 –
    MALE_ONLY 329 173 39

    Gender representation in top-budget tiers vs. overall averages (1994–2023).

    Baseline=average gender-group proportions in full dataset (N = 199,656). Tiers drawn from financial subset (n = 4,245). Top 10%: n = 440; Top 5%: n = 220; Top 1%: n = 44. Dashes (–) indicate no films in that category and tier

    3.4.2 Financial performance in high-budget tiers

    Financial outcomes diverged sharply across gender groups. Given the small cell sizes within gender groups at higher budget thresholds, formal significance testing was not conducted for tier-level comparisons. Thus, results should be interpreted as descriptive observations rather than inferential findings. Woman-writer films generated the strongest performance across all budget tiers. In the top 10%, they achieved a median ROI of 2.15 (compared with 1.62 for male-only films) and a median profit of $451.9M, nearly $133M higher than the male-only median (Table 7). These differences persisted at higher thresholds: in the top 5%, woman-writer films earned median profits of $556.2M (vs. $470.6M for male-only), and in the top 1%, they achieved both the highest median ROI (2.84) and the highest median profit ($930.4M). On the other hand, woman-director and dual-role films were less represented in these tiers, limiting direct comparisons. The single woman-director film in the top 5% posted median ROI of 0.90 and profit of $202.8M. Dual-role films displayed median ROI of 1.01 in both the top 10% and top 5%, with profits of $227.6M, values below woman-writer films but within the range of male-only productions.

    Category Top 10% Top 5% Top 1%
    n
    WOMAN_DIRECTOR 6 1 –
    WOMAN_WRITER 90 37 5
    DUAL_ROLE 15 9 –
    MALE_ONLY 329 173 39
    Median Budget
    WOMAN_DIRECTOR $189.0M $225.6M –
    WOMAN_WRITER $199.6M $239.0M $323.9M
    DUAL_ROLE $214.3M $236.2M –
    MALE_ONLY $209.4M $242.5M $337.7M
    Median ROI
    WOMAN_DIRECTOR 1.97 0.90 –
    WOMAN_WRITER 2.15 2.04 2.84
    DUAL_ROLE 1.01 1.01 –
    MALE_ONLY 1.62 1.83 2.32
    Median Profit
    WOMAN_DIRECTOR $359.5M $202.8M –
    WOMAN_WRITER $451.9M $556.2M $930.4M
    DUAL_ROLE $227.6M $227.6M –
    MALE_ONLY $318.4M $470.6M $817.6M

    Financial performance in top-budget tiers (1994–2023).

    Budgets and profits adjusted to 2023 USD using CPI-U (). Top 10%: n = 440; Top 5%: n = 220; Top 1%: n = 44. Dashes (–) indicate no films in that category and tier

    3.4.3 Audience reception in high-budget tiers

    Audience outcomes in the highest-budget films exhibited smaller differences than those observed in the broader dataset. However, woman-writer films consistently received the highest median ratings: 6.8 in the top 10%, 7.0 in the top 5%, and 7.4 in the top 1%. Overall, male-only films scored the same or marginally lower than the other female-led categories across all tiers, with the only exception being dual-role films in the top 5% (6.2 vs. 6.6). Woman-director and dual-role films performed competitively where sample sizes allowed, with median ratings of 6.9 (woman-director, top 10%) and 6.7 (dual-role, top 10%). Table 8 summarizes audience metrics across the top-budget tiers.

    Category Top 10% Rating Top 10% Votes Top 5% Rating Top 5% Votes Top 1% Rating Top 1% Votes
    WOMAN_DIRECTOR 6.9 270,698 6.6 449,108 – –
    WOMAN_WRITER 6.8 281,062 7.0 430,811 7.4 531,568
    DUAL_ROLE 6.7 301,458 6.2 406,094 – –
    MALE_ONLY 6.6 303,165 6.6 362,143 6.8 584,746

    Audience metrics in top-budget tiers (1994–2023).

    – indicates no films in category. Tiers drawn from financial subset (n = 4,245)

    Visibility also equalized in the high-budget context. Across the top tiers, vote counts were substantially higher than the financial subset overall, reflecting wider distribution and marketing reach. Within tiers, differences across gender groups were modest. In the top 10%, vote counts ranged from 270,698 to 303,165, and in the top 1% from 531,568 to 584,746 – suggesting that once a film receives a high production budget, audience reach no longer varies substantially by gender authorship.

    3.5 Genre-adjusted robustness check

    To test whether the woman-writer ROI advantage reflects genre composition, ROI was modeled as a function of authorship group, genre, and release year. After genre adjustment, the woman-writer advantage persists: films with a woman screenwriter and no woman director retain a significant ROI premium over male-only films under robust estimation (Huber β = 0.18, p = .033; Table 9). The woman-director and dual-role coefficients did not differ significantly from male-only. The advantage is therefore reduced relative to the unadjusted comparison in Section 3.3 but is not eliminated by genre.

    Predictor β [95% CI] p
    Intercept 0.91 [0.77, 1.06] <.001***
    Gender (vs. male-only)
    Director-only −0.33 [−0.79, 0.13] .157
    Writer-only 0.18 [0.01, 0.35] .033*
    Dual-role 0.02 [−0.27, 0.32] .876
    Control
    Year 0.015 [0.01, 0.02] <.001***
    Genre (vs. Drama)
    Documentary 0.73 [0.26, 1.21] .003**
    Adventure 0.56 [0.39, 0.74] <.001***
    Horror 0.94 [0.70, 1.17] <.001***
    Comedy 0.30 [0.16, 0.44] <.001***
    Romance 0.24 [0.05, 0.43] .012*
    Thriller 0.42 [0.22, 0.61] <.001***
    Mystery 0.42 [0.19, 0.65] <.001***
    Crime −0.34 [−0.52, −0.16] <.001***
    Action −0.11 [−0.27, 0.06] .208

    Genre-adjusted robust regression predicting return on investment (n = 4,245).

    n = 4,245. Robust (Huber) regression. Reference categories: male-only (gender), Drama (genre). Year is mean-centered.

    * p < .05, ** p < .01, *** p < .001.

    4 Discussion

    The financial evidence documented here reveals a consistent disconnect between capital allocation and realized returns. Films in the woman-writer category delivered 25% higher median ROI and 34% higher median profits than male-only productions despite receiving comparable budgets. Films with women in both directing and screenwriting roles matched male-only ROI while operating on approximately half the budget. In the highest-budget tiers, woman-writer films showed 33% higher median ROI, though small cell sizes in these tiers preclude formal testing and warrant a descriptive interpretation. Yet no woman-directed or dual-role film appeared in the top 1% of production budgets across the entire thirty-year period.

    Although the ROI measure omits marketing and distribution costs as well as ancillary revenues, the likely direction of these omissions is informative. Because documented distribution biases direct less promotional investment toward films involving women in creative roles (), a measure inclusive of these costs would likely widen rather than narrow the reported ROI advantages, leaving these estimates conservative in direction. Excluded ancillary revenues, by contrast, accrue disproportionately to franchise properties concentrated among male-only productions, an offsetting omission whose direction is less clear.

    Audience reception offers no clear explanation for these allocation patterns. Ratings were broadly comparable across all gender groups in the financial subset, with no group falling more than 0.24 points below male-only films. Vote count gaps, which were substantial in the broader sample, largely disappeared within the highest-budget tiers, indicating that distribution investment equalizes audience reach across gender groups when budgets are equivalent. This is consistent with evidence that visibility differences reflect marketing and distribution access rather than differential audience demand (; ).

    Prior work established allocation-performance differences primarily through shorter windows of U.S. theatrical releases and focused on revenue as the performance measure (; ). The present study extends this literature by examining return on investment across a thirty-year sample, with financial analyses drawn from the commercially documented, predominantly U.S. theatrical segment, demonstrating that the pattern persists when films of comparable budget are compared and is most pronounced at the investment levels where capital allocation decisions carry the greatest financial consequences. Despite heightened public and industry attention following the #MeToo movement in 2017, participation gains continued at a similar rate, with no sustained acceleration in their growth following 2017. This is notable given that find a positive effect of gender-inclusive content on box office returns specifically in the post-#MeToo period, suggesting that audience demand did respond to the cultural moment. offers a complementary explanation, arguing that while #MeToo generated visible activist momentum, structural change remained contingent on the very gatekeepers who benefit from existing arrangements. The present findings are consistent with that interpretation: cultural pressure and demonstrable financial performance appear insufficient to redirect capital allocation when institutional inertia is the binding constraint. That comparable allocation-performance patterns appear in Germany and Russia, markets with distinct funding structures and industry organizations, further suggests the disconnect reflects structural rather than market-specific dynamics (; ; ).

    Not all findings point in the same direction however. The group that most complicates the overall picture consists of films with at least one woman director and no woman screenwriter. This group posted the weakest financial returns in the sample, with a median ROI of 0.44 and median profit of $13.8 million, substantially below male-only productions. The genre-adjusted analysis (Section 3.5) did not find a significant difference for the woman-director group, though genre, project assignment, and directorial performance cannot be fully disentangled here. Previous research suggests several potential mechanisms that could shape these outcomes. show that genre concentration can substantially influence observed earnings patterns, while identify project assignment and budgeting as central drivers of apparent performance gaps. document higher performance thresholds for women directors in the French industry across all genres, suggesting that constrained budgets and elevated expectations may compound observed return differences independently of directorial quality. The present analysis cannot adjudicate among these explanations. The small sample size (n = 83) further limits definitive interpretation.

    From an economic perspective, the persistence of lower capital allocation to categories that deliver comparable or superior returns suggests a misalignment in an industry characterized by extreme uncertainty and highly skewed returns (, ; ). In such an environment, observable performance signals should, in principle, inform subsequent allocation decisions. That they appear not to do so resembles the R-rating puzzle, the continued production of R-rated films despite profit distributions stochastically dominated by lower-rated films (; ), and related allocation puzzles examined in downstream markets (). In both contexts, upstream investment patterns are not fully aligned with downstream performance distributions.

    Several theoretical frameworks offer candidate explanations for why such patterns persist. Status quo bias (), loss aversion under uncertainty (), and the reliance on conventions in creative industries described by may lead decision-makers to anchor on established production configurations rather than systematically update allocations in response to realized returns. Evidence that the gender of decision makers shapes the gender balance of other industry functions () suggests that relational and compositional dynamics within studios may compound the allocation patterns documented here. document how early disparities in screenwriting assignments compound into widening career earnings gaps over time. A similar path-dependent dynamic may operate at the project level, where initial allocation differences shape subsequent access to larger budgets and higher-visibility productions. provide direct empirical evidence of this dynamic in Hollywood, showing that distributors systematically over allocate scarce resources – such as favorable release dates and promotional effort – to production teams with whom they have prior exchange relationships, enacting a self-confirming pattern in which familiarity rather than projected quality drives resource allocation. That films embedded in these prior relationships actually perform worse at the box office after controlling for resource allocation suggests that relational inertia sustains investment patterns that are difficult to justify on purely financial grounds. While the mechanism differs, the broader pattern in which allocation responds to relational familiarity rather than projected performance, provides a parallel to the disconnect documented here. In the present context, this suggests that the concentration of capital among male-only production teams may reflect entrenched exchange relationships rather than differences in projected returns, echoing the self-confirming dynamic Sorenson and Waguespack document.

    Whether these mechanisms operate in this context cannot be determined from the present analysis alone. The genre-adjusted analysis reported above indicates that the woman-writer advantage is not explained by genre composition alone. A fuller account of how genre structures capital allocation across authorship groups, together with critical consensus, marketing expenditure, and risk-adjusted return measures, would further clarify the mechanisms underlying the patterns documented here.

    These results do not simply confirm that an allocation-performance disconnect exists in commercial filmmaking, a pattern prior work had already established. What thirty years of data reveal is the scale and concentration of the allocation pattern. Films in the woman-writer category reach equal or superior capital efficiency, an advantage that remains significant after adjustment for genre, though reduced in magnitude. Dual-role films achieve return parity at substantially lower budget levels, yet women remain sparsely represented in the highest investment tier. No woman-directed or dual-role film entered the top 1% of budgets during the study period. The pattern therefore cannot be explained solely by realized return distributions.

    What drives this pattern remains an open question. The present analysis does not identify the mechanisms, but it does establish that differences in realized financial performance are unlikely to account for the observed allocation gap.

    5 Conclusion

    This study documents a persistent puzzle in commercial film production, a disconnect between capital allocation and realized financial returns. Across thirty years and nearly 200,000 films, women’s participation in directing and screenwriting has grown only gradually, with no sustained acceleration following the #MeToo movement. Within the commercially documented segment, films with at least one woman screenwriter and no woman director delivered higher median ROI than male-only productions, an advantage that attenuates but persists after adjusting for genre, while films with women in both roles achieved parity on approximately half the median budget. Yet across the entire study period, no film with a woman director or a woman in both roles entered the top 1% of production budgets, and audience reception offers no clear justification for this pattern. Financial performance alone does not explain where production capital flows.

    For investors, studios, and production companies, these patterns suggest that how capital is allocated owes more to convention than to observed returns. Reviewing how projects are greenlit and budgeted is therefore both an equity question and a financial one. One possible explanation is relational inertia, a dynamic documented in this industry in which resources flow toward familiar exchange partners rather than toward projected performance (). Where such dynamics operate, established routines can persist even when they are costly. For those willing to examine their own allocation practices, this may represent an overlooked opportunity as much as a structural problem.

    For policymakers, public funding bodies, and industry organizations such as guilds and awards bodies, the findings caution against treating cultural attention or participation totals as substitutes for structural change. The absence of acceleration after 2017 suggests that visibility alone does not overcome institutional inertia. Attention is better directed at where women are positioned within the budget hierarchy. Monitoring representation across budget tiers and the composition of funding and slates could reveal a pattern that aggregate figures obscure.

    For scholarship, the study establishes the scale and concentration of the gap between allocation and performance and shows that it is not reducible to genre composition. The disconnect between where capital flows and where comparable or superior returns are realized is the central puzzle this study identifies, and it provides a foundation for further research on investment behavior in cultural markets.

    Statements

    Data availability statement

    The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

    AH: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing.

    Funding

    The author(s) declared that financial support was received for this work and/or its publication. The open access publication fee was covered by Johannes Gutenberg University Mainz.

    Conflict of interest

    The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

    Generative AI statement

    The author(s) declared that generative AI was not used in the creation of this manuscript.

    Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

    Publisher’s note

    All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

    References

    • AFI Catalog (n.d.). AFI catalog of feature films. American Film Institute. Available online at:https://aficatalog.afi.com(Accessed March 1, 2025).

    • AguiarL. (2024). Bad apples on Rotten Tomatoes: critics, crowds, and gender bias in product ratings. Mark. Sci45 (1), 63–79. 10.1287/mksc.2023.0236

    • AmaralL. A. N.MoreiraJ.DunandM. L.NavarroH. T.LeeH. A. (2020). Long-term patterns of gender imbalance in an industry without ability or level of interest differences. PLoS One15 (4), e0229662. 10.1371/journal.pone.0229662

    • BachmannM (2024). RapidFuzz (Version 3.8.1). Zenodo. 10.5281/zenodo.10938887

    • BasuroyS.DesaiK. K.TalukdarD. (2006). An empirical investigation of signaling in the motion picture industry. J. Mark. Res43 (2), 287–295. 10.1509/jmkr.43.2.287

    • BasuroyS.RavidS. A.GretzR. T.AllenB. J. (2020). Is everybody an expert? An investigation into the impact of professional versus user reviews on movie revenues. J. Cult. Econ44 (1), 57–96. 10.1007/s10824-019-09350-7

    • BérubéN (2020). Wiki-Gendersort. GitHub. Available online at:https://github.com/nicolasberube/Wiki-Gendersort(Accessed March 1, 2025).

    • BérubéNSainte-MarieMGhiasiGLarivièreV (2020). Wiki-Gendersort: Automatic gender detection using first names in Wikipedia. SocArXiv [Preprint]. Available online at:https://osf.io/preprints/socarxiv/ezw7p/(Accessed March 1, 2025).

    • BielbyD. D.BielbyW. T. (1996). Women and men in film: gender inequality among writers in a culture industry. Gend. Soc10 (3), 248–270. 10.1177/089124396010003004

    • BittmannS.MengerP.-M. (2026). The gendered returns of film success, women and men directors in French cinema. J. Cult. Econ50, 195–223. 10.1007/s10824-025-09569-7

    • Brannon DonoghueC. (2020). “Hollywood and gender equity debates in the #MeToo Time’s Up era,” in Women in the International Film Industry, ed. LiddyS. (Cham: Palgrave Macmillan), 235–252. 10.1007/978-3-030-39070-9_14

    • CavesR. E. (2000). Creative Industries: Contracts Between Art and Commerce. Cambridge, MA: Harvard University Press.

    • De VanyA. S.WallsW. D. (2004). Motion picture profit, the stable Paretian hypothesis, and the curse of the superstar. J. Econ. Dyn. Control28 (6), 1035–1057. 10.1016/S0165-1889(03)00065-4

    • De VanyA.WallsW. D. (1999). Uncertainty in the movie industry: does star power reduce the terror of the box office?J. Cult. Econ23, 285–318. 10.1023/A:1007608125988

    • De VanyA.WallsW. D. (2002). Does Hollywood make too many R-rated movies? Risk, stochastic dominance, and the illusion of expectation. J. Bus75 (3), 425–451. 10.1086/339890

    • EvansR. D.KarabasI.AndonovaY.Nochebuena-EvansL. (2024). Let’s not talk about men: when meaningful female-to-female interaction and dialogue drive higher box office sales. J. Glob. Scholars Mark. Sci34 (1), 57–70. 10.1080/21639159.2023.2248758

    • HarrisC. R.MillmanK. J.Van Der WaltS. J.GommersR.VirtanenP.CournapeauD.et al (2020). Array programming with NumPy. Nature585 (7825), 357–362. 10.1038/s41586-020-2649-2

    • Hennig-ThurauT.HoustonM. B.WalshG. (2007). Determinants of motion picture box office and profitability: an interrelationship approach. Rev. Manag. Sci1 (1), 65–92. 10.1007/s11846-007-0003-9

    • HeywoodJ. S.SanchezS. I.PaniaguaM. N. (2025). The Hollywood gender gap: the role of action films. J. Cult. Econ49, 459–485. 10.1007/s10824-024-09514-0

    • HuwilerA. G. (2026). 100 Years of women in filmmaking: behind-the-scenes labour and Oscar recognition in Hollywood. Hist. J. Film Radio Telev46 (1), 197–219. 10.1080/01439685.2025.2561971

    • IMDb (2024). IMDb non-commercial datasets. Available online at:https://developer.imdb.com/non-commercial-datasets/(Accessed March 1, 2025).

    • KahnemanD.TverskyA. (1979). Prospect theory: an analysis of decision under risk. Econometrica47 (2), 263–291. 10.2307/1914185

    • KarniouchinaE. V.CarsonS. J.TheokaryC.RiceL.ReillyS. (2023). Women and minority film directors in Hollywood: performance implications of product development and distribution biases. J. Mark. Res60 (1), 25–51. 10.1177/00222437221100217

    • LangD. M.SwitzerD. M.SwartzB. J. (2011). DVD Sales and the R-rating puzzle. J. Cult. Econ35 (4), 267–286. 10.1007/s10824-011-9149-z

    • LeontyevaX.KoltsovaO.VerhoevenD. (2024). Gender (im)balance in the Russian cinema: on the screen and behind the camera. J. Cult. Anal9 (1), 10.22148/001c.116142

    • LindnerA. M.LindquistM. A.ArnoldJ. R. (2015). Million dollar maybe? The effect of female presence in movies on box office returns. Sociol. Inq85 (3), 407–428. 10.1111/SOIN.12081

    • LoistS.PrommerE. (2019). Gendered production culture in the German film industry. Media Ind. J6 (1), 95–115. 10.3998/MIJ.15031809.0006.106

    • MattaM. (2025). Movies: ultimate metrics, features, and metadata. Kaggle. Available online at:https://www.kaggle.com/datasets/michaelmatta0/movies-ultimate-metrics-features-and-metadata(Accessed March 20, 2025).

    • McKinleyR. (2016). Celluloid and silicon ceilings: underinvestment in women directors and entrepreneurs. Duke J. Gend. Law Policy24 (1), 89–106. Available online at:https://scholarship.law.duke.edu/djglp/vol24/iss1/4(Accessed January 20, 2026).

    • McKinneyW. (2010). “Data structures for statistical computing in Python,” in Proc. 9th Python Sci. Conf., eds. van der WaltS.MillmanJ. (Austin, TX: SciPy), 51–56. 10.25080/Majora-92bf1922-00a

    • NguyenJ. (2023). The economic case for equality in screenplays: the Bechdel test, female dialogue and box office revenue. J. Screenwriting14 (2), 173–190. 10.1386/josc_00124_1

    • PrommerE.LoistS. (2020). “Where are the female creatives? The status quo of the German screen industry,” in Women in the International Film Industry, ed. LiddyS. (Cham: Palgrave Macmillan), 235–252. 10.1007/978-3-030-39070-9_3

    • RavidS. A.BasuroyS. (2004). Managerial objectives, the R-rating puzzle, and the production of violent films. J. Bus77 (2), 155–192. 10.1086/381638

    • SamuelsonW.ZeckhauserR. (1988). Status quo bias in decision making. J. Risk Uncertain1 (1), 7–59. 10.1007/BF00055564

    • SeaboldS.PerktoldJ. (2010). “Statsmodels: econometric and statistical modeling with Python,” in Proc. 9th Python Sci. Conf., eds. van der WaltS.MillmanJ. (Austin, TX: SciPy), 92–96. 10.25080/Majora-92bf1922-011

    • SeboP. (2021). Performance of gender detection tools: a comparative study of name-to-gender inference services. J. Med. Libr. Assoc109 (3), 414–421. 10.5195/jmla.2021.1185

    • SeboP. (2022). How accurate are gender detection tools in predicting the gender for Chinese names? A study with 20,000 given names in Pinyin format. J. Med. Libr. Assoc110 (2), 205–211. 10.5195/jmla.2022.1289

    • SorensonO.WaguespackD. M. (2006). Social structure and exchange: self-confirming dynamics in Hollywood. Adm. Sci. Q51 (4), 560–589. 10.2189/asqu.51.4.560

    • StroubeB. K.WaguespackD. M. (2024). Status and consensus: heterogeneity in audience evaluations of female- versus male-lead films. Strateg. Manag. J45 (5), 994–1024. 10.1002/smj.3575

    • U.S. Bureau of Labor Statistics (2024). Consumer price index for all urban consumers (CPI-U), Series ID: CUUR0000SA0. Available online at:https://www.bls.gov/cpi/(Accessed December 1, 2024).

    • ValentowitschJ. (2022a). Does female screen presence pay off at the box office? An exploratory analysis with special emphasis on the Bechdel test. J. Int. Bus. Econ22 (1), 40–69. 10.18374/jibe-22-1.4

    • ValentowitschJ. (2022b). Hollywood caught in two worlds? The impact of the Bechdel test on the international box office performance of cinematic films. Mark. Lett34, 293–308. 10.1007/s11002-022-09652-5

    • WallsW. (2005). Modeling movie success when ‘nobody knows anything’: conditional stable-distribution analysis of film returns. J. Cult. Econ29 (3), 177–190. 10.1007/s10824-005-1156-5

    Summary

    budgets, directors, filmmaker, gender gap, Hollywood, profits, screenwriters, women

    Huwiler AG (2026) Women in directing and screenwriting: the capital allocation puzzle in commercial film. Front. Commun. 11:1838492. doi: 10.3389/fcomm.2026.1838492

    Maria O’Brien, University of Galway, Ireland

    Ruba Saleh, Brussels Management School, Belgium

    Pete Jones, University of Alberta, Canada

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

    All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

    Share.

    Related Posts

    Add A Comment
    Leave A Reply