Evolving Gender Representation in Cinema: Analyzing the Bechdel Test’s Trends and Relationships with Movie Performance throughout History

ageng

jiwooyou

mcheong

rpv

Motivation

Due to historic biases, women are seldom given equal representation in media. Movies are no exception. Thus, in this analysis, we seek to understand the varied portrayal of women in movies over time, as well as the relationship between the portrayal of women and various other characteristics of movies. As our measure of representation, we use the Bechdel Test, a litmus test developed by cartoonist Allison Bechdel. In order to past the test, a movie must fulfill all of the following three criteria:

1. The movie must have at least 2 women characters
2. Who talk to each other
3. About something other than a man

While not a foolproof way of assessing the complexity given to female characters in film, the Bechdel gives insight into if female characters are given the opportunity to exist beyond their relationships to male characters. Understanding these trends in gender representation over time is crucial because it reflects broader societal changes and attitudes towards gender equality. By studying these trends, we can identify progress and areas that still need improvement, which can inform future efforts to promote gender equality in media and beyond. Additionally, analyzing these trends/relationships helps us understand the potential cultural/economic impacts of gender representation in films.

By studying a dataset of movies that pass or fail the Bechdel Test, we can gain valuable insights into several key areas:

1. Trends Over Time: Analyzing how the proportion of movies passing the Bechdel Test has changed over the years can reveal broader societal changes and shifts in gender representation in cinema.
2. Financial Characteristics: Investigating the financial aspects, such as budgets, of movies that pass or fail the Bechdel Test can uncover trends and differences across different decades, shedding light on the economic dynamics of gender representation.
3. Commercial Success: Exploring the relationship between passing the Bechdel Test and a film’s commercial success can help us understand the potential impact of gender representation on a movie’s financial performance and its reception in the market.

---

Dataset Description

Our dataset was obtained from the TidyTuesday package, which sourced its data from FiveThirtyEight. It includes data on 1,794 movies from 1973 to 2013. It contains 1,794 observations (movies) and 34 variables. The variables include basic features of the movie (title, release year and date, actors, plot, rating, country, writer, director, genre, runtime, etc.), the results of the Bechdel Test (a binary pass-fail and the highest criterion it follows), budget (including actual budget and budget normalized to 2013) and measures of success (domestic and international gross, metascore, IMDB rating, awards, etc.).

Research Questions

1. How have trends in passing the Bechdel Test in cinema evolved over time/place?

2. What are the financial characteristics and trends of movies that pass/fail the Bechdel Test respectively, and how have these evolved over different decades?

3. What is the relationship between pass/failing the Bechdel Test and a film’s commercial success?

---

RQ1: How have trends in passing the Bechdel Test in cinema evolved over time/place?

To explore the trends in gender representation in cinema over time, we start by examining the distribution of Bechdel test outcomes across multiple time periods throughout history.

The above stacked bar plot shows the proportion of movies that pass and fail the Bechdel test in each year group. The various shades of red denote the 3 ways a movie can fail the test. The graph shows that generally, the proportion of movies that pass the Bechdel test has increased over time, indicated by the increasing height of the blue portions of the bars. Just over 25% of movies in the 1970-1974 group pass the test. The next group jumps up to a 50% passage rate, followed by a slight decline in passage, then a steady increase until the passage rate just about levels off. Also, when movies fail, the majority of the time, the reason is because while there are at least 2 women characters, there is not an instance of two or more women talking to each other.

The increasing proportion of movies passing the Bechdel test suggests a gradual improvement in gender representation in cinema. However, the fact that most movies fail the test simply because there weren’t any conversations between women in the first place still highlights room for improvement.

To examine how trends in passing the Bechdel Test in cinema are different across regions, we should focus on the differences in Bechdel Test passage rates between domestic (US-made) and international movies. The stacked bar chart illustrates the difference in Bechdel Test passage by region. Specifically, by domestic movies (movies made only in the US) and international movies (movies made in countries outside of the US, or by multiple countries that may include the US). The bar chart displays the proportion of movies that pass the Bechdel Test (following all three criteria) compared to the movies that do not pass the Bechdel Test. The graph suggests that there is no significant difference for passing the Bechdel Test by region, as both domestic movies and international movies have similar proportions of passing compared to not passing. Both domestic and international movies have more observations that fail the Bechdel Test, which suggests that movies regardless of their country of origin have a tendency to fail the Bechdel Test.

---

RQ2: What are the financial characteristics and trends of movies that pass/fail the Bechdel Test respectively, and how have these evolved over different decades?

One of the financial characteristics that we can analyze about movies that pass/fail the Bechdel test could be their budgets. This time series graph shows the budgets of movies that pass and fail the Bechdel test over time. In order to make the budgets comparable, they have been adjusted for inflation to 2013 money, as 2013 is the year the most recent film in the dataset was released. The plot shows that for the most part, movies that pass the Bechdel test have consistently lower budgets than those that fail. Both categories have relatively similar patterns of peaks and troughs, but failing movies are consistently higher in budget. This graph provides evidence that movies with more robust and complex portrayals of women are valued less than more male-centric movies.

In order to formally test the budget differences that were observed above in the plot, we also conducted a two sample one sided t-test to observe whether the average log-budget (the distribution of budgets were skewed to the right so we applied a log transformation) for movies that failed the Bechdel test was statistically higher compared to those that passed. We received a t-value of 5.00, giving us a p-value of 3.242e-07, which is less than alpha=0.05. Hence, we reject the null hypothesis in favor of the alternative which is that the average log-budget for movies that failed the test is higher. Overall, this suggests that there may be disparities in funding based on women representation in film.

Another key financial metric are the movies’ domestic/international grosses. The above scatterplot compares the domestic and international gross of the movies (adjusted to 2013), colored by whether the movie passes (blue) or fails (red) the Bechdel Test. A regression line is fitted for each result of the Bechdel Test. We find that most movies (regardless of Bechdel Test passage) fall within a domestic and international gross of $500 million, which is shown by the high density of points towards the lower left corner of the graph. From the linear regression lines, we find that movies that fail the Bechdel Test tends to have higher domestic gross compared to international gross, while movies that pass the Bechdel test tend to have higher international gross compared to domestic gross. This may suggest that international audiences have cultural preferences that value gender representation more heavily than domestic audiences. Another possibility is that movies that pass the Bechdel Test are promoted more aggressively in international markets compared to domestic markets.

To further understand how a movie’s budget is associated with the probability of passing the Bechdel Test, we fitted a logistic regression model with pass/fail as the outcome (regressing on budget, year, and language). This plot visualizes probability of passing across different decades, with normalized log-budgets in 2013 in the x-axis. In these plots, we are able to see that higher budgets consistently lower the probability of passing the Bechdel Test in most decades from 1970 to 2010, but the relationship is stronger in recent decades (2000’s and 2010’s). We can also see that lower-budget movies in recent decades have a higher likelihood of passing, possibly indicating a shift towards more diverse gender inclusion in recent decades. Thus, we are able to see the inverse relationship between budget and passing the Bechdel Test.

Finally, the violin plot shows log-transformed budgets (normalized to 2013) for movies that passed the Bechdel Test across different decades. We can see that movies that failed the Bechdel test have slightly higher median budgets and larger budget variability. In addition, the distribution of movies that fail the test appear to be wider and is skewed towards higher budget values, compared to those that passed. However, in years past 2000, the difference in budgets appears to be smaller, possibly suggesting that there is a shift towards diverse representation over time. Overall, this graph shows evidence that movies that failed the Bechdel Test have higher variability and larger budgets.

---

RQ3: What is the relationship between pass/failing the Bechdel Test and a film’s commercial success?

For our last research question, we wanted to investigate how the outcome of the Bechdel test was associated with a movie’s commercial success. One measure of a movie’s “commercial success” could be how positively it was received by audience reviews. Therefore, we started by examining the differences in the distribution of IMDB ratings for movies that passed the Bechdel test vs. movies that failed to pass. The above plot shows the probability density distribution of IMDB ratings for movies that passed the Bechdel test (blue) and for movies that didn’t (red). We divided this further into two different time periods: before the year 2000 (on the left), and after the year 2000 (on the right), which created a total of four different distributions. This would allow us to observe whether the relationship between the test’s outcome and imdb ratings changed between the two periods. All four densities in the plot appeared to follow a normal distribution. For both time periods, it appears that the distribution of IMDB ratings for movies that failed the Bechdel test is slightly higher compared to those that failed it. Initially, this might suggest that movies that fail the Bechdel test have better audience reception.

However, it is not obvious whether the aforementioned differences in IMDB ratings are statistically significant. To determine this, for each time period, a two-sample K-S test was conducted for the distributions of IMDB ratings that passed and failed the Bechdel test. Both of these tests were significant, with p-values of 0.022 and 0.002, respectively. Hence, for both time periods, we reject the null hypothesis that the distribution of IMDB ratings between those that fail the test versus those that pass the test come from the same underlying distribution. This gives us reason to believe that the outcome of the test is associated with different levels of public reception. From the graphs, we could also hypothesize that movies that fail the Bechdel Test have a slightly more positive audience reception, but the K-S test does not tell us anything about the nature of this difference.

Because the K-S test cannot tell us anything about the nature of the difference in distributions, we also conducted a one-sample T-test to compare the mean IMDB ratings of movies that passed the Bechdel Test with those that failed. This analysis was performed separately for movies released before 2000 and after 2000. The null hypothesis states that the mean IMDB rating for movies that failed the Bechdel Test is less than or equal to the mean rating for those that passed. For movies released before 2000, the test yielded a test statistic of 3.06 and a p-value of 0.001. For movies released after 2000, the test yielded a test statistic of 4.58 and a p-value of 2.62e-06. In both cases, we rejected the null hypothesis in favor of the alternative hypothesis that the mean rating for movies that failed the Bechdel Test is greater than that for movies that passed. Thus, we confirm our hypothesis that movies failing the Bechdel Test are associated with greater positive reviews (before and after the year 2000).

Finally, the most obvious way of measuring the commercial success of a movie would be to examine its gross revenue. In order to compare the commerical success for movies that passed the test vs those that didn’t, we decided to plot how the gross revenue of movies changed over time for each binary outcome of the test. This leads to the time-series graph displayed above. The x-axis shows the movies’ release date, and the y-axis shows the rolling average of the 30 previously released movies’ domestic gross. Keep in mind that the gross revenues for each date has been normalized to “2013 money” in order to adjust for inflation. The green line plots movies that have passed the Bechdel test while the red line plots movies that haven’t. We can see that before 1995, domestic gross of movies that failed the Bechdel test was relatively higher by a fair margin. This was especially true between the years 1970-1980. After 1995, the difference in gross starts to close. Past 2010, there no longer an obvious margin of difference between the two lines. This shows how movies that failed the Bechdel test were associated with a relatively higher gross revenue in the past, but that same statement does not hold true for more recent time periods.

Conclusion

From the data, we have observed a notable progression in gender representation within cinema. The proportion of movies passing the Bechdel Test has steadily increased, suggesting a growing effort towards creating more gender-inclusive films.

However, when we shift our focus to the financial characteristics of these films, the picture becomes more complex. Despite the increase in gender representation, movies that pass the Bechdel Test have consistently lower budgets compared to those that fail. This trend has persisted across different decades. It seems that while the industry is making strides in gender inclusivity, this progress is not yet fully reflected in the financial backing these films receive.

We also found from the data that historically, movies which failed the Bechdel Test achieved higher gross revenues. However, differences in revenues have started to close in more recent years, which may or may not suggest a change in consumer preferences. Audiences today may be more appreciative of films that offer diverse and inclusive narratives, even if the industry has been slower to adapt financially.

In summary, while there has been significant progress in the representation of women in cinema, as evidenced by the increasing number of films passing the Bechdel Test, this has not yet translated into equal financial support.

Future Work

While our analysis provides insights into gender representation in movies, using the Bechdel Test as a benchmark, there are several ways that future research can build upon these findings.

First, future work can expand the metric for gender representation. While the Bechdel Test is a useful starting point, it has its limitations for capturing nuances of female portrayal in movies. Future studies could incorporate other measures, such as the proportion of screen time a female character has, the complexity of their character, and their impact on the storyline. Obtaining these metrics were out of the scope of our research, but would be very interesting to observe.

Second, future work could expand the time period of movies used in their analysis. While our dataset only extended from 1970 to 2013, future studies could include older movies and more recent movies to better understand the passage of Bechdel Test over time. More recent movies can also be used to observe how movements like #MeToo have influenced the film industry and their portrayal of women.

Lastly, future work can explore how different audience segments view movies that pass or don’t pass the Bechdel Test. There may be a difference in popularity for movies that pass the Bechdel Test with younger audiences compared to older audiences, women compared to men, or people in one region compared to another.

These recommendations are motivated by the limitations encountered during our analysis. Addressing these questions would require access to more detailed datasets and additional time. These future directions will help us better understand gender representation in the film industry, as well as provide insight for filmmakers seeking to promote inclusivity and diversity into their work.