Election Fraud Nonsense

Everyone who thought the election mishegas would be over after the election, raise your hand. C'mon, don't be shy. I certainly thought things would calm down a day or two after the election.

So I was wrong. Starting with election night, when Hillary was apparently too crushed, or -- dare I say it -- too drunk to make her own concession speech, quickly followed by one legacy media anchor after another obviously near tears or in tears over Clinton's loss, followed by recriminations about how the polls had misled them. Then we had the furor about auditing or recounting the votes in several states that Clinton lost, then the ignorant rants about the popular vote, then the "Russians hacked the election" cries, and finally the attempt to suborn enough electors to at least throw the election into the House. How did we get into this mess?

The polls were not wrong.

Now, those of you who were around in 2012 will recall that I've had my own disappointments with polls. There's a reasonable explanation of how polls work in those articles, but to summarize: A poll is done by choosing a relatively few people randomly and asking them for whom they plan to vote. If we really have managed to choose randomly, and the pool of all voters is large compared to the number of our sample, then we can assume that the number of votes for one candidate or the other will be roughly in the same proportion of our sample.

To make this a little clearer, assume all the votes were counted using marbles, with red marbles for Trump and blue marbles for Clinton. After the voting is done, we'll count all the marbles to get a precise number, but for now we're more concerned with speed than exact accuracy, so we're going to take a sample of some small number of marbles and assume that's representative of the whole.

Since we know that our sample can't be perfectly representative of the whole, we compute a margin of error. Without going into the technicalities, the margin of error is basically a bet, just like most everything else in probability. In this case, if I say the poll is Clinton 45, Trump 41, with a 3 percent margin of error, I'm really saying there's 1 chance in 20 that Clinton will get above 48 percent or below 42 percent; contrariwise, there's 1 chance in 20 that Trump will get more than 44 percent or less than 38 percent.

And that's all it says.

Of course, we have to make guesses about the election in the future based on what people say, and people cussedly refuse to behave like marbles -- they change their minds, they have a sick kid, they can't get time away from work, any number of things. So pollsters also create something called a turnout model, which predicts how many people with certain characteristics will actually vote, and then fudge the percentages based on that. (Again, there's lots more detail in these 2012 articles.)