Oscar Bait

A Scientific Investigation

That tedious and never-ending period drama. A complete chronicle of Will Smith's inner demons. Two and a half hours of Leonardo DiCaprio stumbling through the snow. Whatever the specific flavor, we've all suffered through the Oscar Bait film.

For the uninitiated, "Oscar Bait" is a pejorative term used to describe a film released for the sole purpose of tallying up as many Oscar nominations and wins as possible, often at the expense of the film's entertainment value. These films often have specific "Bait-y" characteristics:

  • Dramas, particularly period and historical dramas
  • Lavish production budgets
  • Lengthy runtime (i.e. "epic" length films)
  • A late in the year release date (so as to be "fresh" in the minds of the academy)
  • Better critical than audience reception

But do these "Oscar Bait" films actually tally up more Oscar nominations overall, or does it just seem that way? Is there a quantifiable relationship between the "Bait-y" characteristics listed above and nominations? For my second Metis project, I set out to investigate.

What Types of Movies are Nominated?

For my analysis, I web-scraped data on movie genres, runtimes, release date, production budget, and (of course) Oscar nominations for approximately 7,000 movies released in the last ten years (i.e. 2005-2015) from Box Office Mojo. I subsequently matched this movie data to Rotten Tomatoes critical and audience ratings provided on the Open Movie Database (OMDb) API.

After collecting and cleaning the data, I began to investigate the relationship between "Oscar Bait" movie characteristics and Oscar nominations. The bar graphs below show the the percentage films nominated for at least one Oscar by film characteristics (e.g. Genre, runtime, etc.).


These descriptive  seem to, on the whole, support our "Oscar Bait" theory. Drama ranks as the third most likely genre to receive nominations, only trailing family/ animation and "other" genres. Films over two hours are much more likely to get Oscar nominations than their counterparts, as are films released during the last few months of the year. Finally, films where the critic TomatoMeter exceeds the audience meter by 1 to 25 percentage points are the most likely to get nominations, suggesting a relationship between better critical than audience reception and nominations as hypothesized.

Testing The Oscar Bait Theory

Now that we've explored our data and have an idea as to which characteristics might affect the number of Oscar nominations, let's throw them into a model and see what sticks.

The below table shows the results from a dummy-adjusted OLS linear regression model. The "dummy adjustment" refers to the treatment of missing values - since production budget and critical/ audience ratings were missing for over half of all movies, I decided to set them =0 when they were null. This allowed me to retain movies with missing budget and ratings data in the model while still including these important explanatory variables. I also created two missing data flags (imdb_null and budget_null) that =1 if the IMDb ratings or budget data was missing, =0 otherwise to control for any non-randomness in budget or ratings missingness.

Regression Coefficients and Significance

Feature Coefficient
Intercept -1.235***
Genre - Comedy 0.064
Genre - Documentary -0.109*
Genre - Drama 0.052*
Genre - Family/ Animation 0.077*
Genre - Foreign -0.032
Genre - Horror/ Thriller/ Sci-Fi 0.116**
Genre - Other 0.185***
Release - Q2 0.021
Release - Q3 -0.027
Release - Q4 0.251***
Runtime (in minutes) 0.005***
TomatoMeter Critical Rating 0.018***
Tomato Meter Audience-Critical Rating Difference 0.010***
Not in IMDb Database 0.971***
Production Budget (per $100,000) 2.593***
Production Budget Missing -0.255***
R2 17.9%
Adjusted R2 17.7%

*Significant at the 10% level; **Significant a the 5% level; ***Significant a the 1% level.

The parameters of our model generally support our Oscar Bait theory. The statistically significant positive coefficient on dramas, 4th quarter releases, runtime, and critical reception indicate that these factors are related to increased Oscar nominations. Unexpectedly, horror, thriller and science fiction films also predict a higher number of nominations, as does the audience vs. critical reception difference and missing reviews. I would have predicted that the horror genre, better audience than critical reception, and the lack of an Rotten Tomatoes review (indicating a lesser-known film) would predict fewer Oscars, not more.

But some funkiness is to be expected given the low R-Squared and potential reverse causality in the model (it is conceivable that the promise of Oscar nominations drove movie executive decisions on genres, release dates, etc. instead of the other way around). As such, the results of our model should be interpreted with caution.

So while our descriptive statistics highlight potential connections between specific "Oscar Bait" movie characteristics and Oscar nominations, I'm unable (so far) to draw clear explanatory links between film characteristics and Oscar nominations using regression. The scientific investigation will have to continue another time.

Interested in checking out the code? Take a look at the GitHub repo.