The Nonresponse To Donald Trump

Every good survey taker knows to be aware of tendencies in human nature that can corrupt their findings, enough sometimes to create a backward impression about the state of reality. Sometimes it’s because people say things that aren’t true. People who like to be agreeable create “acquiescence bias.” People who like to fit in create “social desirability bias.” But sometimes it’s because whole classes of people make themselves unavailable.

This is called “nonresponse bias.”

The idea is that bad news or good news or some other ambient factor can skew survey findings. When Barack Obama lost his first re-election debate to Mitt Romney, polls briefly inverted, to show a small Romney lead. Democrats naturally panicked, but… it was almost certainly illusory. It wasn’t the case that millions of voters who’d planned to vote for Obama became persuaded by the debate to switch their allegiances. It was that Obama supporters became demoralized, Romney supporters became energized, and the combination had a statistically meaningful effect on who was willing to accept calls from pollsters. Romney voters were pumped. Obama voters went to ground.

In this way, nonresponse bias can cause people who make decisions and form opinions based on survey data to badly misread the true state of affairs.

Nonresponse bias may have helped Donald Trump become president, twice. Polling Trump can be a challenge, because he and the GOP have trained American conservatives to distrust all sources of independent authority, including pollsters. It’s hard, in general, to get them to engage with people and institutions they don’t know and trust.

In a more direct way, the publication of the Access Hollywood tape depressed his polling much more than his organic support. He surely lost some voters, but what he lost more than anything is people who were willing to tell pollsters they intended to vote for him. For most of October 2016, it really seemed like he was cooked.

That widespread assumption wasn’t an inert factoid. It affected behavior on a national scale. Would James Comey have re-entered the fray of the election if it seemed like Trump was only two points behind instead of six or seven or eight or nine? What about swing voters who feared Trump but had hated Hillary Clinton on a personal level for years? Would 80,000 of them in the three Blue Wall states have stayed home