Dem
50
GOP
50
FiveThirtyEight has some advice about predicting the winner of the presidential election: Flip a coin. In four of the swing states, the margin is less than 1 point; in the other three it is less than 2 points. The margin of errors on these things is typically ±4 points or so, not counting methodological errors. With numbers like that, all seven could go one way or the other way and it could end up being a landslide in the Electoral College in either direction.
This raises the question of: How good are the polls? Actually, that question can be answered. Sort of. After all, there is a lot of historical data. In particular, 538 took a look at historical polling data and election results for president, governor, and senator in the swing states. Here are the results since 2016 (they also looked at earlier data, but they might be less relevant than the data produced since Donald Trump appeared on the scene).
| State | # Polls | Error |
| Nevada | 86 | 3.3% |
| Georgia | 188 | 3.4% |
| Arizona | 162 | 3.7% |
| North Carolina | 157 | 4.4% |
| Pennsylvania | 147 | 4.5% |
| Michigan | 133 | 5.4% |
| Wisconsin | 81 | 5.6% |
Clearly, the error is substantial in all cases, with polling in the Sun Belt states being better than in the Rust Belt states. This suggests we could be in for big surprises next week.
The article compares polling in the seven swing states to polling in other states. It is actually much better in the swing states. In Idaho, Oklahoma, South Dakota, and Wyoming, the average error is over 10 points. There is far more polling in the swing states than in these four, so maybe practice makes perfect—well, all right, less bad. (V)