David R. Henderson  

Barack Beane Obama

PRINT
A Question for Steve Sailer's ... We're Going Too Far...

One of my favorite books of the last decade was Michael Lewis's Moneyball. It's about how Oakland A's general manager Billy Beane used statistical analysis to offset the advantage of more heavily funded teams and make his small-market team into a contender. I've written about it here and here.

I was reminded of that book when I read Michael Scherer's article, "Inside the Secret World of the Data Crunchers Who Helped Obama Win." Obama's campaign had an incredibly sophisticated data-crunching operation for raising money and targeting ads.

An excerpt on fund raising:

For the general public, there was no way to know that the idea for the [Sara Jessica] Parker contest had come from a data-mining discovery about some supporters: affection for contests, small dinners and celebrity. But from the beginning, campaign manager Jim Messina had promised a totally different, metric-driven kind of campaign in which politics was the goal but political instincts might not be the means. "We are going to measure every single thing in this campaign," he said after taking the job. He hired an analytics department five times as large as that of the 2008 operation, with an official "chief scientist" for the Chicago headquarters named Rayid Ghani, who in a previous life crunched huge data sets to, among other things, maximize the efficiency of supermarket sales promotions
.
An excerpt on getting out the vote:
The magic tricks that opened wallets were then repurposed to turn out votes. The analytics team used four streams of polling data to build a detailed picture of voters in key states. In the past month, said one official, the analytics team had polling data from about 29,000 people in Ohio alone -- a whopping sample that composed nearly half of 1% of all voters there -- allowing for deep dives into exactly where each demographic and regional group was trending at any given moment. This was a huge advantage: when polls started to slip after the first debate, they could check to see which voters were changing sides and which were not.

It was this database that helped steady campaign aides in October's choppy waters, assuring them that most of the Ohioans in motion were not Obama backers but likely Romney supporters whom Romney had lost because of his September blunders. "We were much calmer than others," said one of the officials. The polling and voter-contact data were processed and reprocessed nightly to account for every imaginable scenario. "We ran the election 66,000 times every night," said a senior official, describing the computer simulations the campaign ran to figure out Obama's odds of winning each swing state. "And every morning we got the spit-out -- here are your chances of winning these states. And that is how we allocated resources."


It probably goes without saying, but I'll say it anyway: Dick Morris and his ilk are dinosaurs. Like co-blogger Bryan, I bet on various issues to discipline my thinking and to learn. I've learned. Losing the bets helped, and the Scherer article helps.

Having credited the sophistication of the Obama campaign, though, I want to point out that the overall message still matters and that Romney had essentially 90 minutes of good campaigning in the midst of a few-months campaign. I'm reminded of a campaign I was involved in against a sales tax increase in Monterey County, one that my band of brothers and I won. After it was over, I found out just how sophisticated the losers were. Of course, it helped that our side needed only one third of the vote to stop the tax increase. Here's an excerpt from an article I wrote after the campaign about my discovery of the losing side's sophistication:

We then walked in the other direction and John showed us the headquarters of the Yes on Q campaign, which had already been vacated and was for rent. The For Rent sign stated that the building had 20 phone lines. We peered into the building and saw that it was almost the size of a high-school gymnasium. John explained that during the campaign the building was a hive of activity. They had had maps of the voting area, block by block, and had obtained data on who had voted and who hadn't, so that they could target their efforts. So this is part of how they had spent their $450,000, I said to myself. I felt a quiet awe. We had taken them on and beat them. It was as if we had fought our way across no-man's land between the two enemy trenches on the assumption that although there were more of the enemy, they had the same kind of weapons. Then when we got there, we discovered that while we had M-16s, they had machine guns. Of course, there are two problems with the analogy. First, our weapons were our words and ideas and, compared to their words and ideas, ours were machine guns and theirs were M-16s. Second, I never regarded them as the enemy. But you get the point.


Comments and Sharing





COMMENTS (6 to date)
Ian B writes:

I'm curious how analytical Romney's campaign was. Several authors have criticized the conservative media channels for getting the election so wrong. But my (naive?) assumption is that with Romney's skin on the line, he would have had a much clearer picture than the media had.

Speaking of skin on the line, I made some bets on the election too: I bet that Romney wouldn't reach 290 electoral votes and that Obama wouldn't reach 300. I put most of the money into the second bet and got killed when Obama easily reached 303 (and maybe 332!)

I made the bet using the intrade state data (e.g. % probability that Obama will win Florida). I downloaded all of the data using intrade's API, then ran a monte carlo simulation that suggested that Obama would win 70-75% of the time with 281-290 electoral votes. I gave Obama a 25% chance of reaching 300 and a 5% chance of reaching 330. I bet on the 300 threshold because the odds were "so good" thanks to all of the Nate Silver fans flooding that particular bet.

My lessons learned were:

1. You're only competitive in forecasting if you have as much data as your competitors: my intrade state-level data was no match for Nate Silver's poll-level data (and maybe even district-level data).

2. Backtest your model!! I built my model sitting in a hotel room during Hurricane Sandy. Didn't have time to backtest it. If I had, I may have found that intrade's state-level data wasn't accurate enough.

3. Build less risky bets: my computer program I wrote spat out which intrade contracts had the highest rate of return based on price vs. my probability distributions. I could easily have hedged my bets better and taken a slightly lower rate of return. Instead I maximized return and got killed when all the money I sunk into Obama at 300 was lost (I sold most of my contracts for pennies on the dollar late election night).

jeremylapin writes:

I guess the lesson for this is, if you are in the minority or the underfunded side, lie to every pollster. Better yet, make a 5 dollar donation to the Democrats, get on their phone and mailing list and skew their data and resources as much as possible.
Still I don't see enough people would follow this strategery to make much difference.

jure writes:

out of topic question but: i've been reading right now about limitations on gas consumption that mayor Bloomberg is making on our Slovenian news pages. They are comparing those policies with policies of a socialist commisar Milka Planinc in Yugoslavia when we had to drive on odd and even registration numbers. And of course they are trying to show how great are those policies by Bloomberg. I really am wondering how is it possible that such a prominent figures as mayors don't have enough advice and knowledge or both for just letting the price to jump and give the incentive for more supply? How is it possible that New Jersey and NY are adopting policies that were proved to be wrong? I dont understand where is mayor Bloomberg making here a political point for himself, why is this in his self-interest? he just makes people more angry because of gas shortage.

drycreekboy writes:

RCP's current reckoning of the popular vote gives Obama %50.5. So all that micro-targeting and data-mining may have made the difference in what was bound to be a close election. However:

The most telling metric coming out of the exit polls was how many people thought George W. Bush was to blame more than Barack Obama for the economy: 60 to 40. Given that ratio roughly reflects the public as a whole, what if those numbers were reversed? All those whiz kid quants working for Jim Messina could run 66 million simulations a night and Obama would've still lost the election.

George W. Bush is the republican Jimmy Carter because of a failed war and a giant economic meltdown in the last six months of his administration. Republicans need to come to terms with this, and accept that, barring a meltdown of Obama's own, the electorate will need enough time to forget Bush. Until then it's a steep uphill battle -- however many data analysts they hire.

Steve Sailer writes:

Funny isn't it, how Mr. Uniter wound up presiding over the most scientifically divisive campaign in history?

magilson writes:

So you are telling me the campaign of a politician in favor of centralized planning won his campaign by collecting, analyzing, and acting on data. That's just the ego boost he needs to eliminate the last shred of humility he has.

That's just stupendous for all of us, I'm sure.

Comments for this entry have been closed
Return to top