Estimating voter turnout is critical to forecasting an election correctly, but the coronavirus poses special challenges to that undertaking. A respondent may tell a pollster that they are 'very likely' to cast a ballot, but a sudden surge in positive COVID cases in their area close to Election Day might keep them away from the polls. In traditional, largely 'in-person' elections, researchers have developed methods to 'trick' respondents into revealing if they truly are likely voters—like asking them to name their polling place or the candidates they voted for in a particular election. In 2020, those guardrails will not be as effective, with the spike in mail voting posing an additional challenge. For example, respondents who tell pollsters they plan to vote by mail may have never experienced that process before and could feel intimated by it. It's almost impossible to measure that level of anxiety—and what impact it might have on actual voting behavior.
And, there is also the issue of the counting of non-traditional ballots. No one yet knows how many mail-in ballots will be invalidated (for various reasons). There is no way for researchers to account for how the process will impact the final tally. Since many states have limited experience in dealing with millions of mail ballots, estimating the true impact of this phenomenon will be very difficult. This election may very well play out the way it has been trending for months, but the real 'October Surprise' may be the voting uncertainties posed by an unprecedented pandemic.