In this mini analysis we work with the data used in the FiveThirtyEight story titled “The Dollar-And-Cents Case Against Hollywood’s Exclusion of Women”. Your task is to fill in the blanks denoted by ___.

Bechdel Article Summary

The Bechdel test is a test invented by cartoonist Alison Bechdel intended to analyze whether or not the movie has two women that have a meaningful conversation that doesn’t involve a man. Movies that fail this test either dont feature a conversation between two women that isn’t about a man or don’t have women in the movie at all. The author of the article found that the budgets of movies that pass the Bechdel test is substantially lower than those that don’t, however movies that do pass the test may have a higher return on investment than those that don’t.

Data and packages

We start with loading the packages we’ll use.

library(fivethirtyeight)
library(tidyverse)

The dataset contains information on 1794 movies released between 1970 and 2013. However we’ll focus our analysis on movies released between 1990 and 2013.

bechdel90_13 <- bechdel %>% 
  filter(between(year, 1990, 2013))
print(bechdel90_13)
## # A tibble: 1,615 × 15
##     year imdb    title test  clean…¹ binary budget domgr…² intgr…³ code  budge…⁴
##    <int> <chr>   <chr> <chr> <ord>   <chr>   <int>   <dbl>   <dbl> <chr>   <int>
##  1  2013 tt1711… 21 &… nota… notalk  FAIL   1.3 e7  2.57e7  4.22e7 2013…  1.3 e7
##  2  2012 tt1343… Dred… ok-d… ok      PASS   4.5 e7  1.34e7  4.09e7 2012…  4.57e7
##  3  2013 tt2024… 12 Y… nota… notalk  FAIL   2   e7  5.31e7  1.59e8 2013…  2   e7
##  4  2013 tt1272… 2 Gu… nota… notalk  FAIL   6.1 e7  7.56e7  1.32e8 2013…  6.1 e7
##  5  2013 tt0453… 42    men   men     FAIL   4   e7  9.50e7  9.50e7 2013…  4   e7
##  6  2013 tt1335… 47 R… men   men     FAIL   2.25e8  3.84e7  1.46e8 2013…  2.25e8
##  7  2013 tt1606… A Go… nota… notalk  FAIL   9.2 e7  6.73e7  3.04e8 2013…  9.2 e7
##  8  2013 tt2194… Abou… ok-d… ok      PASS   1.2 e7  1.53e7  8.73e7 2013…  1.2 e7
##  9  2013 tt1814… Admi… ok    ok      PASS   1.3 e7  1.80e7  1.80e7 2013…  1.3 e7
## 10  2013 tt1815… Afte… nota… notalk  FAIL   1.3 e8  6.05e7  2.44e8 2013…  1.3 e8
## # … with 1,605 more rows, 4 more variables: domgross_2013 <dbl>,
## #   intgross_2013 <dbl>, period_code <int>, decade_code <int>, and abbreviated
## #   variable names ¹​clean_test, ²​domgross, ³​intgross, ⁴​budget_2013
## # ℹ Use `print(n = ...)` to see more rows, and `colnames()` to see all variable names

Summary of Abraham Lincoln: Vampire Hunter

It’s budget was 67.5 million dollars. It’s domestic gross was 37.5 million dollars. It’s international gross was 115.1 million dollars. Abraham Lincoln: Vampire Hunter fails the bechdel test, meaning it has fewer than two women and the women don’t talk at all or the women only talk about other men. The result “dubious” means that contributors were skeptical whether or not the movie passed the test.

bechdel2004 <- bechdel %>% 
  filter(between(year, 2004, 2004))
print(bechdel2004)
## # A tibble: 81 × 15
##     year imdb    title test  clean…¹ binary budget domgr…² intgr…³ code  budge…⁴
##    <int> <chr>   <chr> <chr> <ord>   <chr>   <int>   <dbl>   <dbl> <chr>   <int>
##  1  2004 tt0337… 13 G… ok    ok      PASS    3  e7  5.71e7  9.64e7 2004…  3.70e7
##  2  2004 tt0212… 2046  ok    ok      PASS    1.2e7  1.44e6  1.92e7 2004…  1.48e7
##  3  2004 tt0343… 50 F… men-… men     FAIL    7.5e7  1.21e8  1.96e8 2004…  9.25e7
##  4  2004 tt0356… A Ci… ok    ok      PASS    1.9e7  5.14e7  7.01e7 2004…  2.34e7
##  5  2004 tt0359… A Ho… dubi… dubious FAIL    6.5e6  1.03e6  1.03e6 2004…  8.02e6
##  6  2004 tt0344… A Ve… ok    ok      PASS    5.5e7  6.17e6  5.96e7 2004…  6.78e7
##  7  2004 tt0357… Anch… men-… men     FAIL    2.5e7  8.41e7  8.94e7 2004…  3.08e7
##  8  2004 tt0327… Arou… nota… notalk  FAIL    1.1e8  2.40e7  7.20e7 2004…  1.36e8
##  9  2004 tt0370… AVP:… dubi… dubious FAIL    7  e7  8.03e7  1.73e8 2004…  8.63e7
## 10  2004 tt0414… Banl… nowo… nowomen FAIL    1.5e7  1.20e6  1.12e7 2004…  1.85e7
## # … with 71 more rows, 4 more variables: domgross_2013 <dbl>,
## #   intgross_2013 <dbl>, period_code <int>, decade_code <int>, and abbreviated
## #   variable names ¹​clean_test, ²​domgross, ³​intgross, ⁴​budget_2013
## # ℹ Use `print(n = ...)` to see more rows, and `colnames()` to see all variable names

There are 1615 such movies.

The financial variables we’ll focus on are the following:

  • budget_2013: Budget in 2013 inflation adjusted dollars
  • domgross_2013: Domestic gross (US) in 2013 inflation adjusted dollars
  • intgross_2013: Total International (i.e., worldwide) gross in 2013 inflation adjusted dollars

And we’ll also use the binary and clean_test variables for grouping.

Analysis

Let’s take a look at how median budget and gross vary by whether the movie passed the Bechdel test, which is stored in the binary variable.

bechdel90_13 %>%
  group_by(binary) %>%
  summarise(
    med_budget = median(budget_2013),
    med_domgross = median(domgross_2013, na.rm = TRUE),
    med_intgross = median(intgross_2013, na.rm = TRUE)
    )
## # A tibble: 2 × 4
##   binary med_budget med_domgross med_intgross
##   <chr>       <dbl>        <dbl>        <dbl>
## 1 FAIL    48385984.    57318606.    104475669
## 2 PASS    31070724     45330446.     80124349

Next, let’s take a look at how median budget and gross vary by a more detailed indicator of the Bechdel test result. This information is stored in the clean_test variable, which takes on the following values:

bechdel90_13 %>%
  group_by(clean_test) %>%
  summarise(
    med_budget = median(budget_2013),
    med_domgross = median(domgross_2013, na.rm = TRUE),
    med_intgross = median(intgross_2013, na.rm = TRUE)
    )
## # A tibble: 5 × 4
##   clean_test med_budget med_domgross med_intgross
##   <ord>           <dbl>        <dbl>        <dbl>
## 1 nowomen     43373066     44891296.    89509349 
## 2 notalk      56570084.    63890455    123102194 
## 3 men         39737690.    56392786     99578022.
## 4 dubious     35790994     49173429     89883201 
## 5 ok          31070724     45330446.    80124349

In order to evaluate how return on investment varies among movies that pass and fail the Bechdel test, we’ll first create a new variable called roi as the ratio of the gross to budget.

bechdel90_13 <- bechdel90_13 %>%
  mutate(roi = (intgross_2013 + domgross_2013) / budget_2013)

Let’s see which movies have the highest return on investment.

bechdel90_13 %>%
  arrange(desc(roi)) %>% 
  select(title, roi, year)
## # A tibble: 1,615 × 3
##    title                     roi  year
##    <chr>                   <dbl> <int>
##  1 Paranormal Activity      671.  2007
##  2 The Blair Witch Project  648.  1999
##  3 El Mariachi              583.  1992
##  4 Clerks.                  258.  1994
##  5 In the Company of Men    231.  1997
##  6 Napoleon Dynamite        227.  2004
##  7 Once                     190.  2006
##  8 The Devil Inside         155.  2012
##  9 Primer                   142.  2004
## 10 Fireproof                134.  2008
## # … with 1,605 more rows
## # ℹ Use `print(n = ...)` to see more rows

Below is a visualization of the return on investment by test result, however it’s difficult to see the distributions due to a few extreme observations.

ggplot(data = bechdel90_13, 
       mapping = aes(x = clean_test, y = roi, color = binary)) +
  geom_boxplot() +
  labs(
    title = "Return on investment vs. Bechdel test result",
    x = "Detailed Bechdel result",
    y = "Return on Investment",
    color = "Binary Bechdel result"
    )

What are those movies with very high returns on investment?

bechdel90_13 %>%
  filter(roi > 400) %>%
  select(title, budget_2013, domgross_2013, year)
## # A tibble: 3 × 4
##   title                   budget_2013 domgross_2013  year
##   <chr>                         <int>         <dbl> <int>
## 1 Paranormal Activity          505595     121251476  2007
## 2 The Blair Witch Project      839077     196538593  1999
## 3 El Mariachi                   11622       3388636  1992

Zooming in on the movies with roi < 16 provides a better view of how the medians across the categories compare:

ggplot(data = bechdel90_13, mapping = aes(x = clean_test, y = roi, color = binary)) +
  geom_boxplot() +
  labs(
    title = "Return on investment vs. Bechdel test result",
    subtitle = "Zoomed in to results lower than 16", # Something about zooming in to a certain level
    x = "Detailed Bechdel result",
    y = "Return on investment",
    color = "Binary Bechdel result"
    ) +
  coord_cartesian(ylim = c(0, 15))