A New Twist on Polling Averages

The Upshot has an interesting twist on polling averages, showing the key battleground states along with what would happen with a polling error similar to those in 2016 or 2012.

But Jonathan Bernstein has two caveats: “One is that the chart assumes any error would be in the same direction, which it may not be; I’d rather they demonstrated the size of the error, which could favor either candidate. The other thing is that the ‘polling error’ here includes the final three weeks of polls in each of those previous elections, but there’s good reason to believe that the race shifted in Donald Trump’s direction in the final days in 2016, so part of what’s being called an error here is really polls taken before voters changed their minds.”