Talking Points

Polling was terrible again. It isn't going anywhere.

The presidential votes are still being counted in many states, and so far Joe Biden has a substantial lead in the popular vote and is narrowly ahead in the states necessary to win the Electoral College. That makes for a sharp contrast with the pre-election polling averages, which had Biden far ahead nationally and ahead in almost every swing state.

We will have to wait until all the votes are counted to see how bad the mistake was precisely, but at time of writing the FiveThirtyEight polling average was off in Ohio by about 8 percentage points, in Florida by about 5 points, and in Texas by about 4 points. Other states are still being counted, but similar errors look likely in many of them. The mistake looks considerably worse than the miss in 2016, and very possibly the worst one since the infamous "Dewey Defeats Truman" whiff in 1948.

No doubt experts are already investigating what went wrong. But it's hard to trust that they will correctly identify the source of error, given that they have been working on their models feverishly for the last four years and somehow only introduced new mistakes.

The logical response to two consecutive polling faceplants that disadvantaged Democrats would be for the party to abandon its laser focus on public opinion, and simply run on what seems like good policy. If you're going to be flying half-blind in terms of what sells, you might as well try to achieve something in the process.

But this is very hard to imagine. Polls on issues, policies, popularity, head-to-head matchups with conservatives, and so on — these things are central to modern politics. Perhaps half of political argument and political media (much of my own writing included, to be fair) turns on a poll in some way. Even if polls are untrustworthy, they will continue to be conducted regardless, and get wide attention. And panicky Democrats, desperate to find the "safe" choice of candidates or platforms, will probably keep listening to them.