“Numbers trigger weird things in people’s brains”

7/19/2018
Nate Silver at OMR18
Nate Silver speaking at OMR18 in Hamburg, Germany

After forcing the political mainstream to take notice in 2008 and one-upping his success four years later in the 2012 US Presidential Elections, Nate Silver and his predictive data models took a beating in the press in the run-up to the 2016 elections. While the surprising (for most) election result vindicated Silver and his models, the misconception of how forecasting works remains. In this episode of The OMR Podcast, Hamburg-area serial entrepreneur Erik Siekmann sits down with Silver, founder and Editor-in-Chief of data-analytics site FiveThirtyEight to talk about the power and powerful misconceptions of numbers, the difference between prediction and forecasting, the similarities between politics and sports and why if you understand forecasting, neither the Trump victory nor the Brexit result were all that surprising.

“Prediction is bullshit”

After accurately calling 49 of 50 states in the 2008 elections, Silver and FiveThirtyEight seemed to improve on that performance in 2012, by getting every single state right. But Silver is quick to point out that such a perception mischaracterizes what it is they do. “We make forecasts that are probabilistic. We think the right way to measure something is, if you say something has an 80% likelihood of occurring, then over the long run, does it occur about 80% of the time?” He is much more concerned with his models landing within the margin of error—and not with being results oriented. “What we are doing is not prediction. Prediction is bullshit. Forecasting is what we are doing; forecasting is probabilistic. Forecasting is about preparing one for risk and uncertainty.”

Watch Nate Silver’s keynote at OMR18.

“Finding needles in haystacks”

Although 2008 marked a breakthrough for Silver personally with FiveThrityEight, to say that data first gained in importance during the 2008 campaign is not entirely true. “2008 was primarily a change in media coverage. Data-driven campaigns hit the mainstream in 2008—but campaigns have always been fairly data driven.” Even George W. Bush and his campaign manager, Karl Rove, whose presidency preceded Barack Obama’s, understood the importance of data, according to Silver. And due to the nature of the US political system and landscape, with its electoral college, two-party system and polarized voter bases, it necessitates a more data-driven approach because “you are trying to reach a fairly small segment of swing voters in the United States; you’re trying to find needles in haystacks.”

In 2016: “nothing went wrong”

Although many touted the 2016 as evidence of shortcomings in predictive data analysis when applied to politics, Silver sees it as vindication. “Our forecasts said Trump had a 30% chance of winning. Other forecasts said that it was almost impossible for Trump to win. If you understand what we’re doing, then that is highly valuable information.” Silver also pointed out that there was a lot of crap published regarding the 2016 elections, although “the polls were about as accurate in 2016 as they have been historically.”

“Defining a solvable problem”

Sports and games have played a role in Silver’s career to date—from getting his start from winnings as a professional poker player to working as a writer for analytics site Baseball Prospectus. And when the opportunity presented itself in 2012 to join US sports behemoth ESPN, Silver jumped at the opportunity to “do something bigger” and meld two analytically predictive, yet fundamentally disparate sectors. “Sports and politics are more different than they are alike.” Because of its defined rule structure, “sports are much easier to predict. They are closed systems. They don’t change that much over time.”

Politics, on the other hand, are an entirely different kettle of fish. Although the electoral college in the US make politics “sportslike,” predicting broader political trends is not a part of forecasting. “If we have polling data, then at short time horizons, we can build a forecast into the probability of each party winning. It doesn’t mean that politics is an inherently predictable enterprise.”

“Campaigns don’t make much of a difference”

When asked about Cambridge Analytica and their impact on the campaign, Silver was reluctant to give the controversial, and now defunct, British company too much credit. “The Clinton campaign invested four or five times more money than the Trump campaign and it didn’t do them much good—both in terms of how they allocated resources and also they didn’t persuade many swing voters.” Furthermore, Silver sees the efficacy of Cambridge’s microtargeting approach in a national, presidential election as limited at best on only at the margins, where “it might swing 1% of the vote.” That plays into a larger point that “campaigns don’t have that much of an ability to influence relevant to the messages that people are seeing and hearing.”

OMR Podcast with FiveThirtyEight founder Nate Silver at a glance:

  • On the difference between US and European audiences (1:40)
  • On how Nate Silver got his start as an entrepreneur (02:30)
  • On what went into his decision to found FiveThirtyEight in 2008 (03:23)
  • On 2008 marking a paradigm shift in the importance of data in political campaigns (04:28)
  • On the emergence of data scientists into the mainstream (06:07)
  • Why Nate Silver doesn’t like to view the 2012 elections as more successful than those in 2008 (07:28)
  • On the important difference between predicting and forecasting (08:24)
  • How did the 2012 elections impact Silver from an entrepreneurial perspective? (09:50)
  • On the commonalities between sports and politics (11:50)
  • Why politics are harder to forecast (12:41)
  • On how sports is ahead of politics with regard to analytics (14:20)
  • On what went wrong in the 2016 elections (16:40)
  • On the concept of margin of error in polling (19:15)
  • What are some tactics Silver relies on to communicate his approach to readers? (23:10)
  • On why Cambridge Analytica’s efforts to influence the election had at most a negligible impact (26:40)
  • On the extent to which microtargeting and campaigns impact an election (27:59)
  • On why Cambridge Analytica not only doesn’t deserve the scrutiny, but also the publicity (31:50)
  • On whether or not data science will have an expanded role in upcoming elections (35:40)
  • What does Nate Silver see as the “next big thing” in data science? (38:34)
  • On finding qualified persons to fill the hybrid role of data scientist and journalist (40:12)
  • On creating effective data visualizations and FiveThrityEight’s signature style? (42:31)

Be sure to leave us a like and subscribe to our Soundcloud and Spotify channels.

Current stories and the most important news for marketers straight to your inbox!
Show me an example