While Joe Biden comfortably won the popular vote to become President of the United States in November, many are pointing to the polls’ overestimation of his success, which did not predict the close results in swing states, losses in Congress, nor the continuing and widespread support millions of Americans still hold for Donald Trump. MSc Politics and Communication student Rosie Trainor asked Dr Nick Anstead, associate professor in the Department of Media and Communications at LSE and polling expert, his thoughts on this year’s outcome and the future of polling.
Should journalists rely less on polling in political coverage?
For a long time, academics have complained about so-called “horse-race” coverage – who is up, who is down etc. The argument is that this obscures discussion of bigger, policy-based questions. There is certainly something in this, but I think a more plausible hope is to aim for the media to be more statistically literate in their coverage of polling. Really simple things would help like explaining margin of error properly and sample size. That would be a good start.
Can getting it wrong erode public trust in media institutions?
I think the bigger issue is what it does to democratic institutions generally. In the US, we have seen polling and coverage of polls (in common with lots of other areas of public life) infected with hyper-polarisation. You only have to look at FiveThirtyEight.com’s pollster ratings to see there are a lot of cowboys out there, tailoring polls to meet the expectations of specific media and partisan segments. That is before you get onto online Voodoo polling. The problem this potentially generate is some segments of the population don’t trust the actual results after the election. This is tied in with broader mistrust of some parts of the media.
The problem this potentially generate is some segments of the population don’t trust the actual results after the election. This is tied in with broader mistrust of some parts of the media.
Is there an alternative way of understanding and communicating public opinion, especially in the context of the pandemic?
There aren’t really any other ways of understanding public opinion in the way it has have generally defined for the past 90 years or so – that is, quantitative and representative of the voting population. Evidence of this is found in the fact that, even after the election, people are still quoting polls (how many Republicans don’t believe the election result, for example).
As I have argued in my research, if we want to integrate other measures, such as social media data into the conversation about public opinion, then we need a difference definition of the term, which is more qualitative and about the circulation of ideas, rather what proportion of the population holds a particular view.
Will the 2020 US election results, with a narrower win for Biden than predicted, deal a fundamental blow to polling industry?
I would quibble slightly with the proposition of your question – it wasn’t really a narrow win for Biden. He comfortably won the Electoral College, and dominated the popular vote. He also unseated a sitting President, which is fairly rare. But that is precisely the problem. But for two factors – the polls and order the states counted in (and in particular Florida finishing its count so soon) – I don’t think we would think of this election as close.
The pollsters will continue to work on their method. The question now is what sort of problem they have – is it a Trump problem or a Republican problem? 2018, when they broadly called the midterms right, would suggest it might be a Trump problem. So if he does not appear on the ballot again, the issue might resolve itself. But I think there is a broader challenge about polling in an age of populism and unstable politics.