Why the Academy Awardsâ may fail to pick the "best picture" -- again

by Steven J. Brams and Paul Hager

[Note: There is no objective standard by which movies or other works of art can be judged.  The Latin phrase that describes this is "De gustibus non est disputandum" - in matters of taste there can be no dispute.  However, in researching this subject for Professor Brams, I looked at two different sources of "informed taste": The American Film Institute list of 100 top movies and the Internet Movie Database top 250 movies (as of 2002).  The AFI represents top critics and the IMDB represents movie aficionados.  A comparison was made between the available Best Movie choices for each Academy Awards year and the ranking of the choices in the two lists.  The window was from 1952 to 1996 in order to include the Golden Globes as a possible biasing effect on the Academy Awards choice.  We ultimately excluded discussion of the Golden Globes because it was tangential to the main argument, which is the effect of the voting system.  The result showed that AFI and IMDB agreed with each other more frequently than they did with the actual Best Movie choices. (In some cases both lists picked movies that weren't nominated in a given year.)  Further, both AFI and IMDB disagree with the actual Best Picture choice 12 times and, of those, agreed with each other 10 times on what the Best Picture choice should have been.  (The choice of window understated the number of "bad choices" by the Academy.  For example, both AFI and IMDB would have picked Citizen Kane as the best picture for 1941.) This article selected two examples from the 10 choices where both movie buffs and critics think the Academy picked wrong. - Paul Hager, 2005.]

Spring is approaching and, with it, the 74th Academy Awardsâ ceremony. Both professional handicappers and garden variety cinephiles will be spending the coming weeks trying to divine the collective mood of Academy voters in order to guess the winner.

Experience has shown that picking the artistically "best" movie for Best Picture is not an optimal strategy. Movies with a challenging theme or innovative style indicative of a cinematic masterpiece often fail to win, particularly if there is an accessible "crowd pleaser" among the nominees. But how can this be? The voting members of the Academy of Motion Picture Arts and Sciences (AMPAS) are professionals with a greater knowledge and appreciation of the blend of art and technology involved in movie making than those outside their fraternity. One would expect their collective judgment to be superior to that of the rest of us.

It turns out that mathematics and the little known science of voting explain why AMPAS so often fails to select the "best" picture. Although AMPAS refuses to divulge any information on the voting method it uses, it is an open secret that Academy members vote for only one nominee in a given category. This voting method, known as plurality voting, is the same method used throughout the United States for electing people to public office.

Plurality voting is a perfectly acceptable voting method whenever voters are choosing between only two candidates (or nominees). As soon as there are three or more options, the method breaks down.

When there are two (or more) artistically similar candidates that split the votes of their supporters, a minority candidate can win. Split voting may well have had a decisive impact a number of times in Academy voting, with the most infamous incident occurring in 1950. That year, both Bette Davis and Anne Baxter received Best Actress nominations for All About Eve. When Judy Holliday became the surprise winner in Born Yesterday, it was widely believed that Davis and Baxter had split the vote.

Vote splitting is a special case of the more general problem voting science has demonstrated with plurality voting: it is prone to selecting something other than the true favorite or "majority winner" of a group of voters. A majority winner is the candidate who would defeat all others in a series of one-on-one match-ups.

Following are two of many examples in which the Best Picture winner selected was probably not the majority winner.

In 1976, the five nominees for Best Picture were Rocky, All the Presidentís Men, Bound for Glory, Network, and Taxi Driver. Rocky was the winner. Certainly Rocky was a good movie, but if it had been matched only with Taxi Driver would it have won? Most movie buffs would probably say no. Network would arguably have won a head-to-head match-up with Rocky as well.

The winner in 1979 was Kramer v. Kramer. The other nominees were All That Jazz, Apocalypse Now, Breaking Away, and Norma Rae. Apocalypse Now is generally considered a great movie; it is hard to imagine that Academy voters would have favored Kramer v. Kramer over it in a direct comparison.

Since mathematicians first demonstrated the defects of plurality voting in the late 18th century, many theorists have contributed to what became the discipline of voting science. Better alternatives to plurality voting have been proposed and analyzed. One of the best is known as approval voting. Approval voting allows a voter to choose more than one candidate (or nominee) in a given race. All of the votes cast are totaled. Whichever selection gets the most votes wins. Approval voting eliminates the split-vote problem (in 1950, members could have voted for both Davis and Baxter) and almost always selects the majority winner -- the true favorite. If AMPAS were using it, the winners in the various categories would have a more legitimate claim to the title of "best".

Approval voting is currently used by eight professional engineering and scientific organizations with, collectively, 600,000 members. AMPAS should become the ninth. Until it does, there will be no way of knowing whether the Oscarâ winners reflect the artistic judgment of the Academy voters or the vagaries of a seriously flawed voting method.

[696 words]

[Steven J. Brams is a professor of politics at New York University and co-author with Peter C. Fishburn of the 1983 book Approval Voting. Paul Hager is a software engineer and voting reform activist based in Bloomington, Indiana.]

February, 2002.