Polling and Why Nate Silver Shouldn’t Be Burned in Effigy

I understand why Democrats were wary of polls after the significant errors in 2016. I also understand why the polling errors in 2020, particularly in high profile states like Florida, have many swearing off polls or calling for Nate Silver’s head on a platter. But data-driven journalism outlets like FiveThirtyEight and the Upshot are still valuable even when they make substantial errors.

Why? Because the alternative to data-driven campaign journalism is the pure drivel and gossip-mongering that dominated traditional campaign journalism. We still see versions of this done by people like Chuck Todd.

It’s all campaign rumors from inside sources and speculation about grand strategy, with little connection to reality. Plus, there’s no accountability because fact-checking gossip is essentially impossible.

While polling aggregators aren’t perfect, they help the public in three specific ways:
1) They avoid panic or overconfidence due to single outlier polls.

2)They introduce probabilistic thinking to a large audience.

3)They help identify systemic polling errors, not just in single elections, but over time.

In 2016, pollsters failed to capture the proper vote share of non-college educated white voters. There may be more dynamics at play, like a growing gender gap cutting across race, that will explain 2020 results in hindsight

Nate Silver and the polling industry need to take a good, long look at their models and assumptions. They need to make their thought process public so we can assess whether or not they’re making the adjustments necessary to avoid similar errors in the future.

But personally, I am thankful that people like Nate Silver brought probabilistic thinking into the mainstream so we can even have these conversations in public. The fact that we’re upset that many outcomes lay outside the margin of error demonstrates how far public discourse has come since data-driven journalism became popular.

But if I have to choose between criticizing FiveThirtyEight about systemic errors in the polling industry or trying to read tea leaves and chicken entrails with Chuck Todd, I’m gonna pick Nate. Even if he can be kind of a dick sometimes.