The power law of the Oscars

This was an unusual year for the Oscars: No single film dominated the field and swept the board. Granted, two movies — Million Dollar Baby and The Aviator — did very well, and Million Dollar Baby grabbed all the “important” awards.

But as a couple of scientists found when they studied the data, in most years the Oscars are dominated by a single powerful film that stands out, much the way The Lord of the Rings did in 2003, or Titanic in its famously award-studded season. In 2002, Alan Collins — an economist at the University of Portsmouth — and Chris Hand, a media arts scholar at the Royal Hollaway University of London, gathered stats on the winners of Oscars and the Golden Globe awards from 1983 to 2000. They discovered that the winners follow a power-law distribution — or a Zipf-curve or Yule distribution, depending on what terminology you prefer. That means that a relatively small number of films snared the majority of awards: Sweeping the board is indeed the norm, not the exception.

But why? That’s an interesting question, because most often power-law distributions take hold in systems where the winners have a first-mover advantage — and can thus avail themselves of the “rich get richer” phenomenon. If you’re the first blog in your field, you’ll get linked to by every follower, ensuring you have the highest traffic; if you’re the first city to build a major airport, any new airports will connect to you, ensuring you’re the biggest hub; if your stock gets a little bump on NASDAQ, you might just find yourself benefitting from a herd-mentality stampede as everyone jumps on board.

But voting for the Oscars happens, theoretically, in secret. There’s no information flow, and thus no way for a power law to take hold, right?

Well, sure. Except of course it’s not really a secretive, traditionally democratic process — the members of the Academy talk all the time amongst themselves about their preference. As Collins and Hands drily note in the paper (which you can download as a PDF here):

It is the spread of opinion from colleagues which may result in the clustering of voters’ opinions. Information cascade models generally require decisions to be made sequentially and for the decision of the n+1th consumer to be influenced by the nth consumer. However … information cascade models based on local interactions can also produce heavy tailed / power law distributions.

In this case, for “local interactions” read “Academy members trading gossip while doing blow in the bathroom during a party”. That’s the beauty of network science: According to the theory, even Tara Reid is a potential source of data.


blog comments powered by Disqus

Search This Site


Bio:

I'm Clive Thompson, the author of Smarter Than You Think: How Technology is Changing Our Minds for the Better (Penguin Press). You can order the book now at Amazon, Barnes and Noble, Powells, Indiebound, or through your local bookstore! I'm also a contributing writer for the New York Times Magazine and a columnist for Wired magazine. Email is here or ping me via the antiquated form of AOL IM (pomeranian99).

More of Me

Twitter
Tumblr

Recent Comments

Collision Detection: A Blog by Clive Thompson