If I take one PhD in science and ask them to guess the number of pennies in a jar, it's not going to be close, but if I ask 1,000 regular people to guess, the mean average of their answers is going to be eerily accurate.
This is the concept behind averaging polls too. In 2008, a baseball statistical analyst named Nate Silver made a splash by correctly getting a lot of the electoral college results right. It wasn't that he got the Obama victory right - so did I, so did most people with a clue. Getting the states right is usually a little trickier because some states are 'in play' and polls are tricky business. The dozen major polling agencies - and the fact that there are a dozen tells you this is big money - get bragging rights for the next four years.
2012 was a different animal in numerous ways. The economy was stuck in reverse, the president had engaged in controversial policy, the electorate had thrown his party out of control of the House in 2010 - all offset by the fact that Americans hadn't tossed out a sitting president in 20 years. It was controversial when his opponent, Gov. Mitt Romney (in case you forgot him already, like most people did) declared that 47% of people were never going to vote for him because they were getting something from the bloated government Pres. Obama represented - it was also true. In two candidate races it is almost unthinkable to get less than 47% of the vote. 47% of people are voting by party so there are 6% of people determining a president. But where those 6% of people are in any given year is the magic sauce of punditry.
Given the political variables above, and the fact that polling was primarily being done among demographics that were dwindling, like people who have a landline phone, I didn't think Silver could duplicate his feat of 2008. Americans are just not that predictable and polarized, I believed. It wasn't the election that was in doubt, I called that for Obama in June, I believed the battleground states were not being sampled properly. I even bet against him improving his performance.
Well, I was wrong, Americans were absolutely as predictable and polarized as we feared. Obama carried the super-rich and people making under $50,000 a year. Romney carried the group in between, just like everyone said for months would happen. There were a lot more poor people this time around and that meant there were no surprises or even any drama. Silver got 50 out of 50 state results correct, as did everyone who updated their predictions the day of the election. All of the people who updated the day before got 49 out of 50 correct - just like Silver did. Even Intrade got it right, and those are just Europeans placing bets. Why did Silver get the glory when everyone was right? It helps to write in the New York Times.
But what else can be predicted? Why not the Oscars? Of course, no one is going to pick all the Academy Awards correctly but we don't need to - Silver only got one race correct and he had dozens of polls - he didn't predict every Congressional race and so I don't need to predict a winner in Cinematography (1).
We don't have dozens of polling companies tackling the Academy Awards, so that is a limitation, but we have informal polls, done by every media company in the entertainment business. And since we only need to be right the day of (Silver was off by dozens of electoral votes in the weeks before the election) I have access to the latest data and can do some science - if you call averaging statistics science and, judging by my Twitter feed and all of the science media types genuflecting before Silver in November, 2008, they do believe computing statistics is science.
So I am going to poll average the following sources, create a measure of 'lean' based on how they predicted one month and two months ago, and then calculate a winner for Best Picture.
I got data from Betfair, Intrade, Hollywood Stock Exchange, The Atlantic, IndieWire, Huffington Post and the New York Times.
In the chart below, you can see a real trend. Through most of January one film was doing quite poorly. What was instead leading in polls was "Zero Dark Thirty", a movie about the hunt for Osama Bin Laden. Hollywood loves when a Democrat engages in some imperialism and neo-con thinking became cool in January of 2009. But then they actually watched the movie it reminded them that Guantanamo Bay has not actually been closed and that torture is still going on - so they defaulted to a movie set in the days of another Democratic president, Jimmy Carter. Even better, this is a Hollywood movie where a Hollywood movie saves the hostages. Once oscar voters watched the movie, this really took off in polls.
Thus, we have a clear winner, using the awesome power of science. If statistics are science, that is.
My prediction: "Argo" has an 87.3% chance of winning in the "Best Picture" category unless entertainment publications are statistically biased.
Dear New York Times: Sure, you can hire me to write for you based on my numerical wizardry. I can do this all day on any topic you want, as long as I don't have to conduct the polls.
(1) And isn't this one a waste of time in the digital age? I get their is artistic creativity - when I watched the "Thor" movie I was amazed at how the artists had gotten Asgard to look exactly as it should look - but that is not cinematography. It would be better to have a fun replacement category, like 'Worst Accent In A Movie'. Who wouldn't want to picked Kevin Costner for his "Robin Hood" (sorry, Mr. Costner - love your baseball movies, though!) or Sean Conney in "The Hunt for Red October"?
- How Accurate Are Those Political Polls?
- Do Biases And Analytical Improvements Of Polls Help Predict Elections?
- Are Americans Polarized Or Is Polling More Scientific?
- Poll Averaging Was No More Accurate In 2012 Than It Is Now
- Thousands Of Papers Using Social Media Data For Behavior Could Be Full Of Flaws