Data journalist and election predictor Nate Silver came to campus for Saturday's Zeidman Memorial Colloquium on Politics and the Press. As founder and Editor-in-Chief of FiveThirtyEight, Silver publishes data journalism and developed models to forecast the outcomes of the 2014 midterm elections. The Chronicle's Gautam Hathi spoke with Silver about the elections, FiveThirtyEight and choosing the career paths.
The Chronicle: FiveThirtyEight, as well as other modeling organizations, seemed to miss the mark in several key races on election night. A few races had outcomes which fell outside of what FiveThirtyEight said was the 90 percent probability range of outcomes. What happened?
Nate Silver: I would dispute a number of things in that characterization. Our job is to describe the range of outcomes that can happen in an uncertain world. I think we did a really good job of that. We had the single most likely outcome as being 53 Republican seats [in the Senate]. It’ll wind up being 53 or 54, depending on Louisiana. But that had a range around it where, if Republicans did even a little bit better than their polls, then they could easily win 54 or 55 seats. You’re supposed to also have 10 percent of the cases where you do have an error outside of the 90 percent confidence interval.… You’re supposed to have, if you’re making 70 calls between governor and senate races, you’re supposed to have seven of those [be incorrect] if you’re describing the uncertainty properly. You’re wrong if you don’t have seven of those…. I don’t know if other folks have spent as much time trying to understand and describe the uncertainty.
TC: In general, what have you learned from the midterms for your forecasting methods and how does that change the way you operate?
NS: I don’t think it would change very much. The reason being is that we have already collected data going back to 1990 for Senate elections and 1968 for presidential elections, and there have been a lot of years like this in the past. In 1994 the polls really underestimated how much of a Republican wave you would have, in '98 the polls underestimated how well Democrats would do, in 1980 Ronald Reagan did much better than his polls. So if you’re doing the modeling properly to begin with, then an outcome like this is not a huge outlier—it’s fairly common. That’s why we have this big range of possibilities…. I’ve kind of lost patience with the folks who are like, “It’s FiveThirtyEight’s job to call every state.” If that were the case, then we wouldn’t invest as much time in the both in the modeling and in the description and in the visualization about understanding the uncertainty and the probability in the problem.
TC: You’ve had your own site for a few months now. You’ve been putting out these messages about how people need to look at uncertainty, about the fact that they need to accept errors in models, but it seems like people have a hard time accepting those kind of messages. So what do you think that your version of journalism has contributed so far, and how can people overcome the resistance to accept messages that aren’t often easy to hear?
NS: It’s easier when you have a continuing dialogue and discourse with people. The people who are reading our site every day, they do understand that stuff over time. What’s challenging is that it goes against a lot of instincts that some journalists might have, where you want to put a ribbon on something and say, “This candidate is going to win” or if you don’t do that, then you say, “Too close to call. We have no idea. It’s 50-50.” We don’t exist in a 100-0 or 50-50 world. Most things in the world are 75-25 outcomes. There’s more evidence on one side, but not so much that you can be clairvoyant about it.… But look, FiveThirtyEight has inspired a lot of competitors. You see The New York Times and the Washington Post and the Huffington Post and all these other folks all trying to do this, so that strikes me as pretty significant. It’s great to exist in a world where you have the monopoly profits of being the only person doing something, but it’s not how the market works and it’s not how ideas spread over time. I hope that—probably not by next election cycle, but by 2024 or 2028—it’s just kind of incorporated into the DNA of political coverage and encourages people to think more about uncertainty.
TC: What’s next for FiveThirtyEight? Are you comfortable with the model you have set up in terms of coverage and what you are reporting on? If not, what are you changing?
NS: We always want to get better.… I think we’ve gotten really good at how to cover big, major events like the election and the World Cup and the NCAA tournament. That we do really well with. Figuring out what’s the range of coverage on a slow news day, that’s a challenge for any organization.… It’s very important to us that we’re not just sports and politics, but economics, science, both highbrow and lowbrow lifestyle topics. We’re also figuring out how to do everything from the short 200-word blog posts written off the news to a longer-form piece that involves more investigative reporting, just having a more ambitious range of things we’re doing. We also have a film series with ESPN. We’re really excited to do more with those guys.
I think we have a philosophy about what we think our brand of journalism should be, and I think at first that makes it harder because it means you can’t do just anything. You can’t just say, “Well, I’d love to write a nice essay about the racial politics of Ferguson,” as opposed to what we did, which is have a reporter go out there and report on the economic context…. But I think in the long run having constraints is really important for the creative process. It gives you more focus and allows you to differentiate yourself in a crowded market. We’re finding ways to expand people’s understanding of what data journalism means, I think to show that it’s not incompatible with traditional reporting. Sometimes it’s very much in line with it…. The thing about this is that it’s an enterprise where you just get better with practice.
TC: The stuff that you’ve done—data journalism and quantitative approaches to politics—existed but wasn’t really in the mainstream before you came along and pushed it there. What have you learned from that process of taking something that’s on the fringes and bringing it into prominence.
NS: I think it’s like any other technological innovation. At first people are resistant to change and then it finds an audience and then it proves to have a use case and actually informs people. So you have a big ramp-up in expectations. Now we’re at the next phase where the disruption has occurred and there’s maturation that has to take place. We would assert at FiveThirtyEight that there’s still an undersupply of this content. It’s definitely not the only way to cover the news, we don’t think it should be the only way to cover the news, but maybe now only 2 percent of coverage is in this vein. as opposed to the dozens of outlets who do it the same way as they’ve always done it.
TC: You’ve done a whole bunch of different things that are different but oddly similar in your career. What general advice do you have for people, like many on this campus, who don’t know exactly what they want to do but have some idea?
NS: Get specific. This is cliche advice, but find something you’re passionate about and start doing it. Be a little obsessive about something. I think one of the curses of the millennial generation is that there is not as much economic opportunity, but in terms of having choices about how to spend your time and what media to consume and who to interact with, there’s so much choice that I think that people can get stuck where they’re not specializing in anything very much. So avoiding that mushy middle ground where on the one hand you’re spending time deeply learning about a problem and on the other hand you’re spending time stepping back and trying to take a more global vantage point.
TC: Any other advice you have for students?
NS: It gets into cliches very fast, but do what you love to do. I had, for my first four years, a consulting job, which wasn’t where my interests were really, and it was pretty miserable. The quality of life impact between having a job that you’re passionate about and one that you’re not is just enormous. For better or worse, most of us are spending the majority of our waking hours at work. Every job is hard, but if you have the choice between taking a job that fulfilling and one that’s not, then take the job that’s fulfilling.
The second piece of advice is don’t always follow the crowd. We’re so interconnected now that people come to conclusions more quickly than they should. Keep your distance. Don’t always assume the herd is right.
Get The Chronicle straight to your inbox
Signup for our weekly newsletter. Cancel at any time.