After our archaic electoral college system gave Donald Trump the presidency on November 8th, the public and the media were all shocked. And one of the first questions asked was -- how could the presidential polls have been so wrong?
I'll admit that I was even asking that on election night. But now that I've had some time to think, I'm starting to realize that most polls really weren't far off from the actual result. There were a few outliers, but most polls were showing a race that was closing and was close on election day -- many of them even having the race within their margin of error.
As I write this, Hillary Clinton has a popular vote margin of over 2 million votes, and it is expected that as the remaining absentee and provisional ballots are counted, that margin will continue to grow. She will probably finish with about a 2.5 million vote margin over Trump.
Isn't that about what most of the national polls predicted? Didn't they at least come close to predicting the popular vote outcome? They did. Where they fell down was in electoral college predictions -- and that was mostly in three states (Wisconsin, Michigan, and Pennsylvania). Even then, many polls had them within the margin of error. Most other states were correctly assessed.
What went wrong in the electoral college predictions was not that the polls got the feeling of voters wrong (remember Clinton won the popular vote by a pretty good margin), but that they misjudged who would show up to vote. They expected about the same number of people to vote that voted in 2012 (and that was a rational assumption), but that didn't happen.
While Trump got about the same number of voters that Romney did in 2012, Clinton got about 5 million fewer votes than Obama did in 2012. Did those people stay at home thinking Clinton had it won, and didn't need their votes? Maybe. Anyway, it was the sample that was flawed, and not the polling numbers.
Actually, the polls didn't do that bad a job. I need to remind readers of two things about polls:
1. At their best, even the scientific polls are an educated guess at what will happen.
2. Always look at the margin of error of a poll. If the poll prediction is within the margin of error, then the outcome could go either way. If a poll predicts X will win by 2 points over Y, but the margin of error is 3 or 4 points, then Y could win without the poll being wrong.
(NOTE -- The caricatures of Hillary Clinton and Donald Trump above are by DonkeyHotey.)