Predictions play a central role in our financial markets, natural-disaster planning and most human decisions that involve risk — but despite growing mountains of data, too often predictions are dead wrong when it matters.
Journalist Nate Silver, founder of news site FiveThirtyEight.com and author of “The Signal and the Noise,” cautioned against blind faith in data models during a presentation last week to 1,800 financial analysts attending the CFA Institute’s annual conference in Seattle.
Silver, who became well-known for his forecasts during the 2012 election, is a news entrepreneur who brings averages, correlations and other statistics to bear on themes of the day.
He and his staff often write about how current events square (or don’t) with long-term trends, complete with a geek-fest of charts, such as a recent blog post on why no team can beat the NFL draft.
- Mount St. Helens, still steaming, holds the world’s newest glacier
- Whitest big county in the U.S.? It’s us
- Seattle sets heat record for July 4
- For escapee, prison now will mean 23 hours a day in a cell
- Sound Transit planning heats up for light-rail expansion and public vote
Most Read Stories
From the housing bubble to the 2012 elections to Japan’s devastating 2011 earthquake, Silver warned, even experts’ predictions often fail because they tend to focus on signals that reinforce their view of reality and ignore potentially fatal risks that are hard to measure.
“The gap between what we know and what we think we know is widening,” Silver told the crowd at the Washington State Convention Center.
Compounding the problem, he said, is that forecasters rarely disclose how uncertain they really are — and the public rarely questions the human assumptions behind these models.
Silver, 36, grew up in East Lansing, Mich., graduated with an economics degree from the University of Chicago and launched FiveThirtyEight.com in 2008. The New York Times picked up his blog in 2010.
Journalists have been using data and statistics to inform their stories for decades. The idea of presenting the news through a statistical lens isn’t new either.
But perhaps better than anyone, Silver captured the zeitgeist after he correctly predicted the winner in all 50 states in the 2012 presidential election — prompting a writer on the website Mashable to opine, “Statistics, big data, neutral mathematical models — this, it turns out, is what people want. Who knew?”
Last summer, after an intense courtship, ESPN poached Silver from The New York Times to expand his brand of data-driven journalism to economics, science, life and sports. The broadened FiveThirtyEight site, which relaunched in March, has a staff of 20 and is growing.
The site offers a statistical take on current buzz: After the uproar over L.A. Clippers’ owner Donald Sterling’s racist comments, Silver posted an article on how the Clippers, like most NBA teams, have a majority minority fan base.
But he’s having some fun with it too: He told the analysts at the CFA Institute to keep an eye out for an interactive data project on the best burrito in America, based on data developed in a partnership with Yelp.
“There are 65,000 places in America serving burritos,” Silver said.
He sat down with The Seattle Times for a brief chat. Here’s an edited transcript:
Q: In your book you talk about our ability to predict events. Here in Seattle there’s been concern that we’re overdue for a big earthquake. What are your thoughts on the quality of our ability to predict that?
A: The fact that you have smaller earthquakes from time to time gives you a sense there could be a larger one once every couple hundred years. The fact that you haven’t had one recently is not relevant to the prediction exactly … Definitely sometimes those areas that haven’t had an earthquake in (the) generational memory are less prepared — in terms of building codes, in terms of what people do when there’s a disaster. That’s a risk that if I were mayor of Seattle or governor of Washington, I’d be sensitive to.
In the book, you write about economist Robert Shiller’s bubble-detection method using the price-earnings ratio. If you were a betting man, which you are, how do you feel about the belief the S&P 500 is heading for a correction?
A: I think Shiller’s hypothesis carries some weight that would suggest the market’s a little overvalued right now. The fact that there’s discourse and people are aware it might be overvalued is better than if the reverse were true: “Oh it’s a new era, a new regime and everything is different now.” That’s when you really might worry … I don’t try to make aggressive bets on financial markets. To the extent I invest myself, it’s very diversified, index-fund type of stuff.
Q: What are the lessons you’ve learned so far with FiveThirtyEight.com?
A: There are a lot of lessons. Until you’re getting feedback from customers, then you don’t really have a functional product. I think we underestimated how challenging it would be in the first couple of weeks as people’s expectations are changing. We’re not a politics site anymore. We cover politics, but it’s about 20 percent of our content and about 20 percent of our traffic.
Q: How much traffic is the relaunched site getting?
A: I can’t give you numbers without consulting my boss … but it’s a multiple higher than it was in a couple periods at The New York Times.
Q: Do you believe the business model for online media is sustainable?
A: We’re trying to produce good content. I’m not saying we’re doing anything revolutionary from an economics point of view. … Between ESPN, ABC News and other properties in the Disney family, there are a lot of natural synergies here. I hate that term, but it’s a case where it does apply. ESPN has always been good about letting people build products that have a longer time horizon, and they’ve also been good about monetizing content, getting eyeballs on things, selling deals to advertisers, so it’s a really well-run business. My concern is just to figure out how to put out a great product every day.
Q: Do you think data journalism is a fad or here to stay?
A: No, I don’t think it’s a fad. All the people we’ve hired from organizations like The Times, The Wall Street Journal and The Guardian, they are replacing those people. So they’re not saying it was just kind of a one-off thing. They’re saying it’s an essential function of what we’re trying to do.
Q: How did you like math in high school and what would you say to parents who are trying to get their kids interested in statistics and math?
A: I got to a point in calculus classes where I got a little burned out; I didn’t find it as useful. I like applied math. … I would love telling students, here’s a big data set. Go play around with it and make some mistakes. Do experiments in class. Things that are more hands on. For high-school students, probability statistics, personal finance, foundational skills, frankly, those things could be emphasized more than calculus. If you need to know calculus, you’re going to have to relearn it in college and then probably again in graduate school.
Q: Any regrets about leaving The New York Times?
A: It was not a critique of The New York Times’ journalism. It was a business decision. We wanted to do something that’s entrepreneurial, where we have very strong leadership, and that’s what we have at ESPN. We’ve been focused on written stories, but we also have film projects launching, interactive features, podcasts we’re hoping to do. We have a pretty big new team we’ve hired, and they’ve been supportive of that.
Q: And how about that burrito feature?
A: The burrito feature is coming soon. There might be one detour to Seattle there. The Pacific Northwest is not highly known for burritos, but yeah, we’re going to have fun and make it part of a summer package we’re doing.
Sanjay Bhatt: 206-464-3103 or firstname.lastname@example.org