We predict that President Trump is going to win the 2020 presidential election — and win big.
First, the tone of the questions. There is significant evidence from behavioral psychology that suggests that the way a question is framed predetermines the range of potential answers. In fact, Gallup has found that respondents can answer very differently to questions with the same topic even in the same survey based on the language that’s used. And the use of metaphors can even dwarf the importance of preexisting differences between Republicans and Democrats.
One of the reasons respondents do that is because of a tendency to give socially desirable answers, which was the case especially during the 2016 election. Most people don’t like confrontation, so the easiest, albeit not necessarily the best, solution is to avoid it. Right now, saying that you’re voting for Trump/Pence is often not the socially desirable answer. In fact, a recent poll by the Cato Institute suggests that nearly two-thirds of Americans say that the political climate is sufficiently harsh that they don’t want to give their genuine opinion about politics.
Second, the sample of respondents. Who responds depends on many factors, including the medium (e.g., landline versus cellphone), the location, the sample size and demographic factors. Moreover, the pool of respondents is not necessarily the same as the pool of likely voters. Even though election polls all contain a margin of error, that margin of error is unreliable if the underlying sample does not reflect the population. Researchers have also identified self-screening as the major contributing factor to the polling failures during the 2016 election cycle.
For example, distrust of pollsters also leads to lower response rates for Trump supporters. Rasmussen finds that 17 percent of likely U.S. voters who “strongly approve” of the job Trump is doing say they are less likely to let others know how they intend to vote in the upcoming election. By comparison, only 8 percent of those who “strongly disapprove” of the president’s performance say the same.
While proper sampling methodology matters more than the size, having a representative sample with enough people still helps considerably. Robert Cahaly of the Trafalgar Group notes how their work to create minimum samples sizes of 1,000 voters, added to their work to doggedly pursue the “quiet Trump voter,” led to Trafalgar being one of the most successful battleground polling firms in the country in 2016. Cahaly explains that “we don’t do a state with less than a thousand. You see these polls, 400, 500, 600 people for a state. I don’t buy that. Your margin of error is far too high.”
Third, the content of the current news cycle. What’s going on in a particular moment in time can influence voter attitudes, particularly in swing states. For example, the recent revelation of Hunter Biden’s hidden emails on his laptop, coupled with the link to his father, has come at an opportune time for the president. Moreover, if the economic recovery continues, the good news may continue putting wind in Trump’s sails.
Admittedly, no poll is perfect. That’s why RealClearPolitics takes a step forward by “averaging out” these errors across polls. But “averaging out” only works when errors are made in both directions. Here, many polls make errors primarily in one direction, so the average will still reflect some of the biases that exist.
Gallup conducted a recent survey finding that 56 percent of respondents report that they are better off than they were four years ago, which is striking given that we are in the midst of a pandemic with a recovering labor market and deepening political polarization.
We delved a little deeper into the data from Gallup, coupled with voter registration data, for eight swing states. We find that states with a higher proportion of people who report that they are thriving have a lower polling average for Biden and a higher one for Trump. What’s even more telling is that there’s a big gap between the percentage of people who say they’re thriving and the percentage who say they’re voting for Biden. For example, that gap is 5.3 percentage points in Pennsylvania.
- Moreover, if we look at voter registration in Pennsylvania, we see a similar pattern. There were 803,427 more registered Democrats than Republicans as of May, but that gap has narrowed to 700,853 as of October. In fact, Republicans have netted seven times as many registered Republican voters than Democrats since 2016. Similar trends are taking place in North Carolina and Florida.
We don’t pretend to have all the answers, but we do have some of the right questions. How are people actually feeling? Surveys are great when respondents know what they’re answering and when the sample is representative, but surveys can be misleading otherwise