Chris Cillizza says that [...] the surge in the quantity of public polling available creates a confusing fog of numbers in which “partisans, who already live in a choose-your-own-political-reality world, can select the numbers that comply with their view of the race and pooh-pooh the data that suggest anything different.”
That’s true. But if you actually want to know what’s happening in the race the increased poll volume makes it clearer not less clear. The sense that the polls are “all over the map” is the mistake. You need to think of each datapoint as having an associated probability distribution and then look at where they overlap. [...] The fact that we now have tons of polling that averages out to [a] conclusion means the scope for “sampling error” to throw us off is, at this point, tiny. One poll showing a lead of the current magnitude would leave us with a ton of uncertainty, but a bunch of polls makes the picture pretty clear.
Matthew Yglesias