2pm
Sunday, March 15th
Description: Fivethirtyeight.com was one of the biggest winners in the 2008 November election, successfully predicting the Obama landslide. Veteran BusinessWeek columnist Stephen Baker talks with Nate Silver, the man who's statistical analysis powered the site.
(Note these are all paraphrased responses. Questions from Baker are posted at http://thenumerati.net/)
Baker: What were you doing during Iowa last year?
Silver: Not a big fan of polls, there's just room for improvement. Narrative was simplistic, for example, Obama can't win among working whites. It wasn't looking at the big picture. From a data geeks perspective it was a big experience.
Baker: What did you learn about race in America?
Silver: It's complicated. Can't put any one racial group together. People are quick to assume something is about race when it isn't.
Baker: A lot of research I do, people say demographics is dead. There's lots of ways to crunch data. Are there new "tribes?"
Silver: Sometimes we have micro-targeted groups that make us short sited. At some point it makes sense but we can go to far about being too overly specific.
Baker: Most Americans don't care about specific issues-would you agree?
Silver: In some ways the issues in a campaign are superficial.. I think people are looking at things outside the issues. Would be interesting to have qualitative look at people, say 1,000 one-on-one interviews with voters. That could be fascinating.
Baker: How does Pekota relate to politics?
Silver: They're pretty different. I believe in being really meticulous. If you really want to solve a problem the most interesting things you learn are when you are at the part of the curve with little difference. Decisions made on margins. But both are long seasons, you don't find all that much info at once,. One poll in June means almost nothing. People are trained to overreact. A lot of what I was doing was urging patience.
Baker: Was there an instance that swing a state for you?
Silver: There's a variable in ancestry. Its a badge of pride in certain areas. is is what leads to the forecast?
Baker: How much is Manny Ramerez worth this year?
Silver: Baseball puts too much emphasis on what just happened. He's being paid twice what he is worth/ Teams underrate how rapidly players decline.
Baker: What effect does the economic crash have on voters?
Silver: We've never seen this in the modern era, this level of not being able to see the light. People are almost fearful. It's almost a 9-11 type of event. At some point maybe in 2010 there will be an intersection between Obama's approval rating and the fearfulness. There's about a 3-6 month grace period but after that, after about 18 months people will assign him much blame as much as Bush. he needs things to turn around soon. Although people think the recession will last a long time, they are more pessimistic than economists so he could actually beat expectations.,
Baker: Is anyone polling zip codes for high foreclosure states?
Silver: Not really. Rate of migration is lowest its been on record except during wartime. There are a few areas of higher unemployment, That is the key area to look at. Employment may actually lag behind recovery an you will still get blamed if that is the case.
Baker: You talked ab out recency and how people have ignored past recessions. How can you compare different recessions when you on;t have the internet an there are so many changes?
Silver: Over the long run the volatility of the economy has decreased.. Still people dismiss recessions of the past. System like the economy are complex. We're good at making up explanations when we see random data.
We try to put probabilities around everything we do. If we're off were we wrong or unlucky? It's hard to say. For example, a baseball player playing badly might be due to him breaking up with his girlfriend and drinking too much.
Baker: Project looks at workers behaviors and networks and predict what a worker will be worth in a few years, do we provide training, etc.?
Silver: A lot of what Gates is f=doing is how to we evaluate teachers? I;m a little suspicious that a poorly executed objective option is worse that a well executed subjective one. At a 30k foot level it might be helpful in deciding general things, But there is a give and take there.
Baker: If you could get a masters in something what would it be?
Silver: Computer Science. In some fields a master's doesn't help you. But if I could write some of my own code that would be helpful.
Baker: Would you put your genome on the Internet?
Silver: No, I am relatively private. I'm surprised at how much people are willing to share every detail of their life right now. It's GenY'rs. Average 15 years old girl sends 2500 text messages a month, constant desire to be sharing. I think there's a balance there.
Trying to branch out a bit. Tried to do Oscar predictions but didn't go that well. Looked at 30 years to find variables that are predictable. Got best actor wrong. But some are ok to get wrong.
If you know you're gonna be wrong keep working on your f* model.
Baker: Are you going to get into predicting who is going to be a terrorist?
Silver: In baseball you have 100 years of stats plus no cost if you are wrong. Terrorism you're looking for a needle in a pile of needles. It's hard. No one has great answers. A lot of people working hard on it but collecting data that is reliable is a challenge.
Baker: People on twitter inters ted in predicting whether flights would be cancelled.
Silver: I like JetBlue because they have TV but if you look at on-time for airlines, JetBlue is punished because it flies out of JFK so you would like something relative to the airports.
Baseball is the perfect data set. You can see a 4th and 5th order correlation but it spoils you in a way.
Questions from the audience?
Q: How did you face the challenge of collecting the polls?
Silver: Not too challenging, steal off the polling sites an they steal off us; folks would email us etc. But no way to validate whether it's real or even good.
Baker: A challenge for media is to acquire the skill of collecting data.
Silver: I'm not a huge fan of prediction. Stock market is biggest example and hasn't been doing well lately. I think crowds are generally pretty wise but they can also be every wrong sometimes. This is the big question: how do people form their opinions?
0 comments:
Post a Comment