What if the US election polls are wrong?

There is precious little time for Trump to change the trajectory of the election. But, as Sean O’Grady explains, history shows why it’d be foolish to write him off just yet

Neither candidate will be taking the polls as gospel, given historical precedent (AP)

Joe Biden and his supporters must be looking at the opinion polls and wondering if things might be just a little too good to be true. They may be, for three distinct reasons, all possibly true and – possibly – working to the unlikely advantage of Donald Trump.  

First, they could just be wrong. Even as a snapshot of present opinion – because 3 November is, after all in the future – they might be inaccurate for “technical” reasons. As has been the case periodically in general elections, the pollsters’ methodologies move out of kilter with social trends, distorting the results. Thus, for example, polling the same standard target sample of 1,000 people through face-to-face contact (in the street and/or door to door) will miss some people; doing it by phone may miss others; using the internet might well not reach older citizens, and so on. The pollsters’ methods of “weighting” samples by demographics might be out of date, using, say, older census data. They may have failed to notice strong unexpected trends in willingness to turn out to vote, or some other long-established correlation might have uncoupled itself without them fully noticing. In 2016, for example, the state-based polls in places such as Wisconsin, Iowa and Ohio failed to catch the switch to Trump among working-class so-called left-behind voters, both in terms of moving from a Democrat allegiance and/or bothering to go and vote. There may also have been an underestimate of this at the national level, but in the US system of the electoral college and the importance of swing states it became a crucial (if still relatively modest) misreading of public mood.