Who Do Consumers Really Think Should Win Best Picture at the Oscars? Data Science Has the Answer

by Geoff Bakken with Bill Lattner, Gustavo Sanchez, and Maura Foley

With Oscar night fast approaching on Sunday, the question on the minds of movies fans is which of the nine nominated movies will walk away with the Best Picture award. As data scientists, we relish opportunities to test new ways to predict the outcomes of big events like the Academy Awards.

Conventional wisdom, which is largely based on the results of other awards shows leading up to the Oscars, has it that La La Land will win. While this conventional wisdom may be right, we wondered which movie would be most preferred by popular audiences and came up with a very different answer: Fences.

The case for La La Land rests on metrics like the number of nominations and wins accumulated, the proportion of nominations resulting in wins, or wins in particular contests. From these metrics, both La La Land and Moonlight perform well, as both have racked up over 200 nominations and won 43% of the time. But La La Land has the edge when we consider that it won Best Picture in both the Producers Guild and Directors Guild awards.

Instead of looking at the conventional metrics, we took a different approach, starting from the perspective of moviegoers. What do the predictions look like if we base them on an analysis of the aspects of movies that audiences care about most? While La La Land may be the favorite in the insider Hollywood horserace, I’ll break down our analysis to show why Fences is the movie that would be the most appealing to audiences.

Let’s start with an overview of our method.

We had to get around a huge problem with basing predictions on audience opinions: so few moviegoers (fewer than 1% of respondents in our survey) have seen all of this year’s Best Picture nominees and are able to express an informed opinion. The result is that when you ask people directly which movie should win Best Picture, they tend to choose the movies they have seen. In our survey, Hidden Figures and La La Land had been seen by the most people and were also the top two choices to win Best Picture.

To solve this issue, we went beyond asking people which movie they think should win. Instead, we conducted an experiment where we presented people with hypothetical movies that varied across four key dimensions that may influence people’s decisions about which movies to see:

Our hypothetical movies mixed and matched the actual features of the nominees for Best Picture. For instance, people who took our survey were presented with several questions like the one below and asked to choose which one they would be most interested in seeing.

Which of the following movies would you be most interested in seeing?

Using a Discrete Choice analysis, we identified which dimensions have the most impact on people’s choices and which specific features people find the most appealing. From there, we reconstructed which Best Picture nominees people would prefer, even if they haven’t seen them all.

Here are the highlights of what we found:

• Featured Actor has the most influence on which movies people want to see. While the importance of star power is to be expected, the fact that genre is nearly as important comes as more of a surprise. As media consultant Max Dawson of Penn Schoen Berland said in response to our findings, “The real revelation is the importance of genre. Genre quite literally has star power.” On the other hand, we found that a director’s Academy Award track record did not have much influence, with people demonstrating only a slight preference for Oscar-nominated and Oscar-winning directors relative to up-and-coming directors.

Dimension Importance Rating

• Better known actors are more likely to be more appealing. Actors like Denzel Washington, Emma Stone, Ryan Gosling, and Amy Adams were most appealing to respondents (everyone was indexed relative to Amy Adams, who served as our baseline). Lesser known actors like Trevante Rhodes and Mahershali Ali, who both appeared in Moonlight, were less likely to influence respondents’ choices. According to Dawson, this is a “concrete illustration of the important place that stars continue to occupy in the hearts and minds of moviegoers. It might not always translate into box office, but it does translate into viewer appeal and prestige.”

Actor Appeal Index

• Drama and Historical Drama were the most appealing genres. Musical and Western were the least appealing, while Science Fiction and War Drama fell in the middle.
• And the winner is…. Fences! Driven by the popularity of Denzel Washington and the Historical Drama genre, Fences clearly emerged as the movie audiences would prefer. Hell or High Water came in last, driven mostly by distaste for Westerns rather than a dislike for actors Jeff Bridges and Chris Pine. La La Land falls close to the bottom of the pack despite the appeal of its stars due to the strong dislike our respondents have for the Musical genre. If La La Land does end up the Best Picture winner, this indicates that exceptionally well-executed movies may outperform what we would expect on average from their characteristics.

Preference Share of Nominees

We’ll have to wait until Sunday night to see whether the Academy’s choice of Best Picture reflects the audience preferences we uncovered in our analysis.

Of course, this is just a fun experiment, but findings like these suggest that advanced analytics can “have immense implications for how movies are made and marketed,” according to Dawson. “For example, many awards contenders are marketed as the works of prestigious auteur directors. This research suggests that Hollywood has vastly overestimated the importance of directors’ Oscar track records to moviegoers.”

What if you could engineer a movie that would likely have high potential? Or from a broader business perspective, what if you could design offers based on the features of products or services consumers care about the most? There are many ways this type of analysis can be used — regardless of whether it will predict this year’s Best Picture.

Civis Analytics surveys approximately 1,200 American adults each week. Results used here come from an online survey fielded February 21st. The average margin of error for the survey is +/- 2.9 percentage points.

This post was published by Geoff Bakken along with contributions from Bill Lattner, Gustavo Sanchez, and Maura Foley.