In 2016, a whole bunch of data analysts, some of whom are still viewed as respectable commentators, attacked Nate Silver for having the temerity to point out that if polling errors were correlated across states, Donald Trump had a chance to win the election. Spoiler Alert: They were.
In 2020, there is no shortage of polling about, well, anything. But I am seeing some fretting about two particular polling results from Gallup that, upon inspection, I do not find credible at all. It is possible that in a few months, I will be exposed as someone who refused to accept perfectly valid data. Rather than assert my disbelief, however, I thought it would be useful to explain the sources of my skepticism.
The first poll shows a declining percentage of Americans saying they will take a vaccine approved by the Food and Drug Administration. According to Gallup’s Lydia Saad, “Americans’ willingness to be vaccinated against the coronavirus has dropped 11 percentage points, falling to 50% in late September. This sharp decline comes after the percentage dwindled from 66% in July to 61% in August.”
This seems pretty bad, and it is hardly the only poll to show rising public skepticism toward a vaccine. But the result is not terribly convincing for several reasons. First, the poll also shows that the decline in willingness is a function of partisanship:
As Saad notes, because the president has baselessly promised a vaccine before Election Day, “the Trump administration may have raised doubts in Democrats’ and independents’ minds about its safety.” It also explains rising GOP support for the vaccine. After Election Day, there are reasons to expect that partisan split to erode somewhat. It might erode even further if Joe Biden wins, since Democrats will trust him and Republicans will trust a vaccine produced during the Trump administration.
The other reason I doubt these findings is because of something Nobel Prize recipient Vernon Smith described as “induced value theory.” Simply put, in any kind of experimental or survey work, respondents need appropriate incentives to respond truthfully. What folks say in response to a hypothetical scenario might differ from what they do when confronted with a real version of that scenario. Or, to put it in more concrete terms, people might say they will pay more for “fair trade” coffee or “Made in the U.S.A.” products but then not do so unless confronted with real price differentials.
Simply put, if a vaccine is approved by the FDA without any shenanigans and folks like Anthony S. Fauci and Dr. Deborah Birx endorse its use, I suspect that many people who said they would not get the vaccine will get the vaccine. This will hold with even greater force as people get the vaccine and are able to return to some semblance of pre-pandemic life.
My disbelief in the Gallup result on vaccines is predicated on my skepticism about this category of survey. The second puzzling result is unique to Gallup and comes courtesy of President Trump himself. In a rare event, he tweeted out mostly accurate information:
The only error in Trump’s tweet was the timing: The Gallup result had been released in September. Otherwise, the president is correct:
If Trump pulls off an upset victory, a lot of folks will point to that Gallup result as presaging the outcome. But is it accurate?
There are reasons for doubt. For one thing, other polls asking the same thing are not nearly as positive. YouGov has used this question regularly in 2020. Their August results contradict Gallup: “Americans overall are split: about two in five (41%) now say they were better off four years ago, and 38 percent believe they are better off now.” The Financial Times-Peterson poll is somewhat more positive but shows only 35 percent of Americans saying they were better off. Neither survey echoes the enthusiasm that Gallup found.
Furthermore, Gallup’s other results also suggest that voters might be separating out their personal situation with that of the country. Earlier in September, Gallup polled Americans on whether they were satisfied or dissatisfied with the way things are going in the United States. Only 14 percent were satisfied and 85 percent were dissatisfied, which was the second-greatest response gap in the past four years. This is not a good sign for the incumbent.
Even if the Gallup result is correct, it does not necessarily equate to good news for Trump. CNN’s Chris Cillizza is not wrong when he concludes that “because Trump is Trump, he has managed to separate out voters’ positive feelings about their lives from their feelings about him. People feel good about their situations, and Trump doesn’t benefit.”
In this case, I have doubts about both Gallup’s finding and the interpretation of the finding. I could be wrong. But in the interest of transparency, at least I have explained why I might look like a fool the day after the election.