There’s been much criticism of the polls and predictions in yesterday’s US election. When added to the alleged shock of Brexit and the UK 2015 election, it appears that the polling industry is on its knees.
This has led to some unfortunate conclusions. For example, that we can’t trust yesterday’s exit poll: that’s short-sighted because exit polls tend to be spot-on and contain all sorts of insight that contradicts lazy assumptions about Trump’s breakthrough being from among the disaffected poor. More ludicrously, I’ve seen today the suggestion that ‘the polls are wrong and therefore Jeremy Corbyn can win an election’.
I’m not saying there aren’t problems with polls. There are huge issues with accuracy and heated debates about methodology. But the real problems arise from the ways in which we use polls. Pollsters have become politics’ new high priests. They serve us with numbers which sound extremely accurate yet the differences in methodology provide suggested shifts and a news story about who’s up and who’s down. They give us winners. They sound scientific. And differing sides can usually find a friendly stat or two even in the same piece of research. That’s very useful in our age of post-truth spin and cheap journalism.
But although we can argue about how the numbers are calculated, we must absolutely argue about what the numbers mean. They don’t necessarily in themselves provide analysis. That in itself provides room for subjectivity. You can get far more by doing well-designed qualitative research. This has been devalued, dismissed as inauthentic focus-group politics, but it can be invaluable. If you want to know whether Corbyn could win an election (and if not, why not), it would sort you out. It’s unfashionable to say it, but New Labour understood its target voters precisely because it carried out this kind of research at a sustained and detailed level and good qual was part of what turned Labour at that time into a formidable election-winning machine. And there were plenty of signs that Brexit was imminent if you cared to look, not least in John Harris’ excellent reporting for The Guardian (you know: journalism). Most polling companies were cautious about a Remain win but the interpretation of the markets lulled many into thinking that a Remain win was in the bag.
General misunderstanding about how to analyse a poll leads to the next major problem: polls are, despite themselves, becoming players. In 2015, good polling companies were quite clear about the massive barriers facing Ed Miliband’s Labour, but news editors focused sharply on the idea of a hung parliament, and the idea took on a life of its own. A close poll just before the election causes reactions (eg. the 1992 and 2015 UK general elections), while a foregone conclusion (2001) depresses turnout. The danger of a Miliband administration, perhaps propped up by the SNP, is thought to have played its part in enabling the Conservatives to win its majority last year. To what extent did the expectation of a Remain win depress Remain votes and encourage protest Brexit votes?
The election just finished focused on personality and polls at the expense of policy. Spats between different companies took more prominence than massive issues such as climate change. 538 didn’t have the stellar accuracy of 2012 – though Nate Silver did comment a couple of times on the possibility of Hillary Clinton winning the election in terms of votes but losing the electoral college. Having said that, it’s quite possible to conclude that polls showing Clinton as a dead cert may have caused enough of her supporters to stay at home in some of the most marginal of states, or to make a protest vote. Wisconsin and Michigan would flip to Clinton if Green voters had voted Blue instead. Yes, it’s patronising to suggest that those votes were Clinton’s of right. But I suspect imperfect information results in voters making decisions that backfire, such as when in 2015 Liberal Democrat switchers boosted Labour votes but with the result that a Conservative won the seat. Under those circumstances the media’s obsession with the sport of polls, to the detriment of substantive policy coverage, leads to voters being short changed.