Opinion polls: taking the pulse or setting the tempo?

Campaign posters lining a street in Geneva, October 2019. Keystone / Martial Trezzini

Switzerland has a shorter history than the United States when it comes to using public opinion polls to predict vote outcomes. But in recent years they’ve become a fixture of the electoral landscape. How useful are they?

Though elections in Switzerland are a week away, it almost feels like the results are already in. The right-wing People’s Party will remain the biggest group, polls say, though the Greens will make gains and the overall situation shift slightly to the left.

It’s difficult to imagine the pollsters getting it wrong. If internationally they have been hit by some shocks in recent years – Trump and Brexit, this year’s Australian electionsExternal link – in Switzerland predictions are “surprisingly accurate”, says political scientist Georg Lutz from the University of Lausanne.

To illustrate: over the past two years, a total of 13 popular votes – referendums and people’s initiatives, generally yes-no issues – were held in Switzerland. All were called correctly by GfS BernExternal link, the polling group mandated by the Swiss Broadcasting Corporation for such votes. The rightward shift in the 2015 elections was also correctly forecast, even if the extent of the People’s Party dominance was underestimated.

Yet Lutz still says that methodologies leave a lot to be desired. Survey response rates are often around the 10% mark, he says – that’s low, even for a country with weak voter turnout. And some polls are run on an “opt-in” basis, whereby respondents sign up to panels in advance – this can skew representativity and again make it harder to account for low turnout at election time.

Finally, he says, it’s easy to forget that polls are just a snapshot of a certain moment in time, and can’t predict political changes or the swings in public mood that mark a campaign until its final moments. 

As the American writer E.B. White once said in 1948, “although you can take a nation’s pulse, you can’t be sure that the nation hasn’t just run up a flight of stairs”.

Helping hand

Michael Hermann, the director of the Sotomo groupExternal link, which conducted the recent election barometer published by swissinfo.ch, is more defensive.

Apart from the fact that the surveys carried out by the big two polling groups in the country are consistently accurate, they also fulfil a democratic function by helping media and citizens prepare for vote day, he says.

Lukas Golder, co-director of GfS Bern, takes this further: “if there were no published polls by regulated groups, there would be unpublished polls by political actors”, he says. These groups would then run their campaigns accordingly, without citizens being aware of the context.

The democratic argument for polls has always been debated. In the US, where polling has a long historyExternal link (it started, not coincidentally, in the media, when newspapers attached cut-out coupons that readers would fill out and post back) poll pioneer George Gallup believed his surveys took politics away from the bureaucratic machine and gave it back to the people.

Sceptics, meanwhile, have argued that “public opinion” doesn’t existExternal link, and that questions can be engineered – consciously or unconsciously – to prompt specific responses. Others have questioned the influence of polls, especially on big cable news channels in the US, who sometimes use them to decide which candidates get to take part in televised debates.

Do polls influence how we vote?

Lutz says that in theory polls can sway the electorate in a few ways: for example, voters could jump on the “bandwagon effect” by following whoever seems destined to win; or they could adhere to the “underdog effect” – rallying behind somebody lagging behind.

Hermann mentions the “options” that such polls can give. For example, when voters see in advance that a well-known but divisive figure running in their constituency has no chance, then rather than use their vote strategically to block him or her, they can use it productively to boost a candidate they actually support.

However, empirical evidence that polls directly sway voters in the build-up to elections – a difficult thing to measure – is “inconclusive”, Lutz says. Rather, it’s possible that the indirect impact of such polls is strongest.

In short, polls set the tone for how media, public, and politicians contextualise an upcoming vote: an early surge in support for one particular party will naturally result in a flurry of coverage – and thus exposure – for that party. Likewise, as Lutz says, signs that the People’s Party are set to lose a couple of percentage points will influence not only how that party organises its election strategy, but also the questions journalists ask them.

One problem with this, he says, is it can lead to a sort of “horserace” journalism that’s more interested in results than in policies – something Golder agrees with, saying that the main influence of a poll is “how media transmit the result”.

Hermann reckons it’s also important to remember that electorates (at least the Swiss electorate) tend to start out more left-leaning during a political campaign, before becoming more and more pragmatic and establishment-minded towards vote day – echoing the “snapshot” point that surveys are not set in stone.

In fact, over the course of a Swiss campaign, polls can vary widely, often swinging over a range of 20 percentage points. So much so, that pollsters – who have gotten used to the more irrational sides of voter behaviour – sometimes even call a result that runs against their official final predictions.