Republican candidate Donald Trump surprised pollsters and pundits last Tuesday by delivering a decisive victory in the presidential election on November 8th. Trump’s success is largely due to a dominating performance in the rust belt in which he flipped several key states that were predicted for Democratic candidate, Hillary Clinton.
Play with the interactive map below to explore the differences between the election outcomes and several popular political forecasts and polls.
Several models reveal the probabilities of each electoral vote outcome. Play with the chart below to explore these distributions and how they predicted the actual election outcome.
The Elections Forecaster primarily uses data from TheUpshot — which gathers predictions from The New York Times, FiveThirtyEight, Huffington Post, YouGov, PredictWise, Princeton Election Consortium, Daily Kos, The Cook Political Report, The Rothenberg & Gonzales Political Report, and Sabato’s Crystal Ball — and YouGov. The data was scraped from TheUpshot on November 7th and YouGov on November 6th and is binned into 7 categories:
- 50%-65%: Tossup (the outcome could reliably go either way)
- 66%-84%: Lean Dem., Lean Rep.
- 85%-95%: Likely Dem., Likely Rep.
- 96%->99%: Solid Dem., Solid Rep.
Sources which do not have percentage forecasts — namely YouGov, Cook, Roth, and Sabato — follow the same 7 categories and are matched into their corresponding bin. YouGov notably does not have any “Likely Dem.” or “Likely Rep.” categories.
The reported aggregate score “All models combined” is an average of each forecast result, treating each prediction on a 7-point scale.
Daily Kos and YouGov do not have additional data for the sub-districts of Maine and Nebraska, the two states which give a portion of their electoral votes to Congressional districts. For these two sources, it is assumed sub-districts have the same prediction as the entire state.
This model assumes that Trump won 306 electoral votes, as this corresponds to the outcome of the election if no electors had been faithless. No models took faithless electors into account.
The Electoral Vote Distribution graphic uses data from The New York Times, 538, Huffington Post, and PredictWise, as these sources were the only to make their simulation data public on their website or available upon request.