We’re getting to the point where the most interesting thing about the Daily Kos Elections model is the extreme consistency from day to day. For more than two weeks now (since October 10), Hillary Clinton’s odds of winning the presidential race have been either 95 percent or 96 percent. If you’re wondering what momentous event happened on October 10, well, nothing happened, really. It’s relevant, though, because it was the Monday after Friday, October 7. In other words, that was the first day we were starting to see polls released that were in the field after the leak of the Access Hollywood tape.
Part of the problem is, there just isn’t much room for those odds to keep going up. If you’re sitting around waiting for the day when the odds hit 100 percent, unfortunately, you’ll have to keep waiting. You won’t have that level of certainty until the day after the election. If you took advanced geometry, you might remember the concept of the asymptote. That’s a curve that gets closer and closer to a line as they fade into infinity, but never quite touches it!
That applies to polling, at least as a metaphor. If you had a sample size of, say, one million people, that still wouldn’t give you absolute certainty that your poll was right, because, even then, there would still be millions more people you didn’t survey. It would give you a better margin of error than a poll of 1,000 people, but not so much of a better margin of error that it would justify the hassle and expense of contacting millions of people. (In fact, a pollster might tell you that the amount of time involved in contacting one million people would require field dates so long as to render the million-man poll worse than the thousand-person poll.)
A poll of 1,000 people gives you a margin of error of 3.1 percent; a poll of 10,000 people gives you a margin of error of 1 percent; and even a poll of 100,000 people gives you a margin of error of 0.3 percent, so you can see there’s a point of diminishing returns. And so you can also see that with the model: adding more and more polls gets you incrementally closer to being certain, but never gets you all the way there.