Social liberals and anyone on the progressive side of politics were not the only people to have a terrible night on Tuesday 8 November. The polling industry again found itself under fire as Donald Trump defied their predictions and defeated Hillary Clinton to become president-elect. Following the complete misreading of the 2015 British general election and the broad consensus that Remain would win the European Union referendum, the shocking US result suggests that western democracies have now become so volatile, that polls can no longer tell us anything useful – our world is now post-poll as well as post-truth.
I believe this is too extreme a reaction. For a start, the pollsters did not get it all that wrong on either Brexit or Trump. During the referendum campaign polls conducted by phone suggested a narrow win for Remain, but several online polls gave Leave the lead in the weeks running up to June 23. The media gave less coverage to those polls, perhaps because they did not fit with the expectations of a commentariat whose natural habitat is inside the Remain-supporting bubble of London.
On the other side of the Atlantic, polls had consistently pointed to a Clinton victory. In the event, Hillary won the national vote broadly in line with predictions, but the more important state-by-state polling had exaggerated her support in the crucial ‘Rust-belt’ states. Psephologist Nate Silver – lauded for the accuracy of his predictions in 2012 – had forecasted an overall Clinton win, but he had been concerned enough about the fallibility of his models that he still gave Trump a 29 per cent chance of becoming president. The gap in electoral college votes masked the true closeness of the result; as Silver pointed out afterwards if around 100,000 votes in three key states had gone the other way then Hillary would be on her way to the White House.
The British general election of 2015 was a greater disaster for pollsters, who got the result wrong by a full 7 per cent. This prompted a wholesale revision in methodology, which produced more accurate predictions for the local government elections last May (the referendum was a trickier problem due to the lack of voting history).
Nevertheless, it hard to deny that polls are prone to error. But if we start ignoring them completely, then what replaces them? How do we find an alternative way of judging the mood and voting intention of the electorate? Getting out on the doorstep and talking to voters yourself is the most obvious answer. But only the most diligent canvassers are able to speak to enough people to build up a big enough picture. And while longstanding activists will be able to glean insightful changes in voter behaviour in their local areas, they will not necessarily know whether these shifts are repeated nationwide.
Meanwhile, those longstanding activists are in the minority; most of those who criticise polls will be judging the world through the lens of what their friends and acquaintances tell them. This is an exceptionally poor way to gauge general opinion as both online and offline, people tend to befriend those who are like themselves. If the world consisted only of my Facebook feed then prime minister Ed Miliband would have just left a meeting of the 400 strong parliamentary Labour party to attend a reception with president-elect Clinton before taking a call from chancellor Merkel to discuss how Britain moves towards ever closer union within the EU. At least pollsters are aware that their samples may be unrepresentative and take action to correct this, but many users of social media seem to be unaware of how distorted a picture they are getting – the echo chamber is adept at filtering out any reference to the fact that it even is an echo chamber.
The scepticism about polling holds particular dangers for Labour. However wrong the polls have been, they have never been wrong by a margin of 14 per cent (the Tories current projected lead over Labour). Moreover, every time they have been wrong they have over-estimated the popularity of the progressive or liberal option. Errors were caused by the fact that the type of voter that pollsters find it hardest to get hold of (and therefore include in their samples) are those with the least interest in mainstream politics. And these disengaged voters are increasingly drawn towards the authoritarian or rightwing positions. In other words, we should be praying that the polls are now correct – otherwise Labour’s position is worse than it appears.
The desire to accurately predict the future is as old as humanity – as is our frustration that we constantly fail at it. The lesson of 2016 is that whenever there is a close election, we should stop relying on the polls to tell us what will happen. But this does mean they are not useful. They can give us an insight into the thoughts and intentions of voters we otherwise find it hard to reach, they can tell us how attitudes are changing, how they differ across geographical areas and demographic segments. They can lead us out of the echo chamber and force us to confront uncomfortable – but necessary – truths. Polls cannot always tell us what the outcome of an election will be, but the data they yield helps us get the outcome that we want.
———————————
Christabel Cooper writes a regular column on the Progress website. She tweets at @ChristabelCoops
———————————