US opinion polls are unreliable indicators of the Presidential election results

The argument

A key argument that Donald Trump would be re-elected held that opinion polls, which showed him trailing, could not be trusted — and that the same polls had already been proven wrong once. In 2016, national and state surveys had largely pointed to a Hillary Clinton victory, and Trump's win confounded the forecasts, particularly in the rust-belt states that decided the election. Proponents argued that the polling industry had not fixed the underlying problems, and that anyone relying on 2020's similar polling deficit for Trump was repeating the mistake of taking the numbers at face value. The argument offered a specific explanation for why the polls understated Trump's support: the 'shy Trump voter'. Proponents contended that, in a climate where backing Trump could attract social hostility, many of his supporters were reluctant to declare their preference to pollsters, leading surveys to systematically undercount him. They also pointed to the difficulty of modelling who would actually turn out, arguing that Trump's base was unusually motivated and that polls leaned on turnout assumptions that missed his hidden support. If polls could not capture voters who would not admit their intentions, then the apparent Biden lead was an illusion. From this standpoint, the visible polling picture was not evidence that Trump would lose but an artefact of flawed methods that had already failed once. The real electorate, proponents argued, contained more Trump support than the surveys revealed, and on election day that hidden support would surface as it had in 2016. Because US opinion polls are unreliable indicators of the result, this argument held, Donald Trump would be re-elected.

Premises

Counter-arguments

Critics answered that 'the polls were wrong in 2016' overstates the case — national polls that year were within normal historical error and correctly showed a close race — and that a generic charge of unreliability cannot fix the direction of any error, so it gives no reason to assume a hidden Trump majority rather than a hidden Biden one. The 'shy Trump voter' thesis had been specifically tested in list-experiment research and found little support. Treating polls as pure illusion, they argued, is unfalsifiable: any lead can be dismissed as concealed support, which is not evidence of a coming win. In the event, Trump lost the 2020 election.

Rejecting the premises

[Rejecting P1] National 2016 polling was within normal error and correctly indicated a tight race, so 'the polls were wrong' overstates a modest, well-understood miss. [Rejecting P2] 'Shy Trump voter' effects were tested directly and found little support in list-experiment studies, so systematic hidden undercounting was not established. [Rejecting P3] A generic charge of unreliability cannot fix the direction of error, so it cannot show the hidden votes favour Trump rather than his opponent.