Election 2020: Gallup’s Timely Departure

Gallup isn’t polling Election 2020. It’s polling a lot of related things, but not voters’ preferences as between Trump and Biden. Why is that? It seems that Gallup abandoned the business of presidential prognostication after its poor (but not uniquely poor) sounding of the 2016 presidential election.

Gallup’s final poll had Clinton leading Trump by 4 percentage points, 46 to 42. A 4-point spread surely meant that Clinton would win. But Trump won, despite Clinton’s final 2-point “victory” in the mythical nationwide popular vote. (Why? See this.) Trump exceeded Gallup’s final estimate by 4 percentage points while Clinton bettered it by only 2 percentage points. Game, set, and match to Trump.

It was only the second time since Gallup’s first presidential poll in 1936 that Gallup recorded a clear failure. The first time was when Gallup had Dewey ahead of Truman 50-45 in the 1948 race.

There were several races in which Gallup’s final spread was 2 percentage points or less, so failure in those cases could be attributed to sampling error. Even in those cases, the candidate who was ahead, in Gallup’s estimation, lost only once: Gerald Ford in 1976.

Gallup’s departure signaled a realization that polling had become too fraught with uncertainty because (a) voters have become increasingly reluctant to respond to pollsters. And among those that do, there are more and more voters who are unwilling to divulge their true preferences (e.g., “shy” Trump supporters).

Election 2020 confirms the wisdom of Gallup’s departure from the field. Almost all of the pollsters, including the highly overrated FiveThirtyEight, are predicting victory for Biden. This comes in the face of huge pro-Trump crowds, a late swing toward Trump in Iowa, unusual endorsements of Trump (e.g. police unions, the Pittsburgh Post-Gazette), and Trump’s standing in an unbiased poll, that of Rasmussen Reports.

I’m not quite ready to say that Trump will win. But the outcome will be a lot closer than FiveThirtyEight et al. would like it to be. (Yes,”like” is the proper word.) I will issue my final prediction on the morning of Tuesday, November 3.