Response from Sam Wang, Princeton Election Consortium

September 17, 2014 1:19pm


First, a small correction. In the FiveThirtyEight article, there may be confusion between snapshots and predictions. The Princeton Election Consortium provides snapshots as a precise measure of where the race stands on any given day. In contrast, predictions are less certain. This is why, for the question of whether Democrats+Independents reach 50 votes, the snapshot is at 80%, and the prediction for Election Day is at 70%. As a validation of PEC's this poll-based predictive approach: when applied to Presidential races, it gives a cliffhanger in summer of 2004 and likely Obama wins in summer of 2008 and 2012.


The current argument distracts from the real point: PEC is doing a polls-only calculation, as do HuffPost,, RealClearPolitics, DailyKos Poll Explorer, and TalkingPointsMemo. All of these show a tight race, and in some cases show Democrats+Independents slightly favored. If one adds assumptions about where the Senate race "ought" to be, the outcome looks a little bit different. Several major models (NYT, 538) have said that at the start of 2014, conditions favored the GOP. However, for most of the year, polls have shown that Republicans are slightly underperforming, relative to those expectations. That is the real political story.


In my view, it is helpful for readers to be able to separate what the polls say, and where data pundits say the polls ought to be ("fundamentals"). The NYT's Upshot does that here - it's a good explorational tool for anyone who wants to understand this important disagreement.


Finally, regarding Senate polls: in the aggregate, they are pretty good. When poll medians are applied to Senate races, they perform slightly better than polls-plus-fundamentals. In 2010, both approaches worked -  though they both got the Nevada Senate race wrong. And then, of course, in 2012, there were two differences:


North Dakota: FiveThirtyEight, Berg 92% win probability. PEC, Heitkamp >90%. Winner: Heitkamp.


Montana: FiveThirtyEight, Rehberg 66% win probability. PEC, Tester 71-80%. Winner: Tester.