Polling is dead. Do something different.

Polling has become counter-productive for political campaigns. It is slow, inaccurate and distracting. There are better ways to read, anticipate and meet the needs of voters.

In future, campaign managers will spend less on polling. Winning campaigns already use faster tools to test ideas and track public engagement. This article looks at the limitations of polls and how to cure our addiction to polling.

YouGov’s final poll for the US General Election in 2020 had Biden winning a landslide: 382 Electoral College votes including Florida and Texas.

2016 was an infamous year for polling - the year that gave us Brexit, President Trump and the rejection of the Peace Agreement in Colombia. Through a number of high-profile predictions that did not go to plan, pollsters appeared to lose touch with public opinion and voting intention. Then, the polling companies sold clients on revised modelling and weighting. By 2020, pundits and campaigns were cautiously trusting pollsters again, in terms of vote predictions and decisions about public opinion. The fixes don’t appear to have worked. As we stand now, polling has scant credibility as a predictor of voting intention and provides a patently false picture of how voters rank issues that affect them. A specific recent example being the unremarked fact that American voters placed more importance on their economic prospects than the way the Coronavirus pandemic was being mis-handled by President Trump.

CNN Exit Polls after the 2020 Election showed the Economy and Crime and Safety were both major vote motivators - with over 70% of votes cast on that basis going for Trump

As analyst and journalist John Heileman memorably put it in the post-US-Election edition of HBO’s The Circus, “Polling is f***ed. It is completely, and utterly and systemically f***ed. People in politics have a big decision to make about… how do we trust this stuff again.”

The simple answer is - why bother? There are better ways to win a campaign (more on this later) but before moving on it is worth emphasising that polling isn’t just unhelpful, it is actively bad for campaigns. There are four main reasons for this:

Reason #1: Polling is slow

Polling your response to an issue allows your opponent to define that issue. It doesn’t matter how good an idea like the Green New Deal is, if the first five times somebody hears about it they are hearing a negative message. Right wing campaigners have learned this lesson and they almost never poll responses; they A/B test them, throwing up five or six variant takes onto social media and seeing what gets the best reaction. This takes hours, not days. To give the right wing credit, they are also better at just ‘shooting from the hip’. Even an answer that someone totally disagrees with is more appealing than one that has obviously been polled.

Reason #2: Polling data is bad

Phone polling only reaches a tiny proportion of very old people, and online polling also has terrible problems of skew based on the audience that chooses to engage. Even worse than the problems of selection are those of confirmation bias; people tell you what they think you want to hear. This is why pollsters see a lot of support for progressive positions on civil rights and the environment, when in fact everyone is obsessed with their money and their job.

Reason #3: Pollsters ask the wrong questions

This is not because pollsters are stupid. In fact, they often sneak in some really good questions. But, the campaign always forces them to ask about politics, and what they specifically think about this or that politician. The truth is that most people don’t know, and don’t care about politicians. In terms of actually winning an election it is so much more useful to understand what voters do care about, and then have your candidate speak to that.

(We should acknowledge, one of the reasons polling will survive this crisis is that many analysts also ask the right questions, and extract a huge amount of value from getting some feedback on policy and how it is being communicated).

Reason #4: Focussing on the horse race causes paralysis

Of course, having paid for a poll it is natural for campaigns to focus on how their candidate is polling. For many political gurus the horse race is the essence of polling, and provides the only salient information. Which makes the fact that polls are inaccurate on this front even more corrosive. Of course what polls are meant to do, is to help you figure out ways to change the race. Pollsters may have been especially hampered in 2020 by the inability to complement polling data with qualitative focus groups. But let’s not make excuses. They also didn’t react to clear data showing Trump winning on the most-winning of all topics: the economy and jobs.

3 ways to cure your polling problem

So what’s the cure for campaign managers who are jonesing for poll data but knowing that it gets them nowhere? There are three steps to success:

Step 1. Don’t spend money on polling

If your race is important or newsworthy, media networks will do it for you. Polling is really slow, so why pay out to have the information at noon when you can turn on the TV at 6pm?

Step 2. Handicap your race

You can still make spending decisions based on where you are doing relatively well, but you have to assume you are down many, many points on what pollsters are telling you. Remember, these are TV polls, you are not paying for them.

Step 3. Focus on issues that matter

Proper research, using machine learning to analyse population level data, can give you much more accurate information than any poll if you are interested in public opinion and what voters want or need. It is crazy to ask people about this stuff because they will always tell you what they think you want to hear, or what they just saw in the news. If you want to know what is really worrying them, look at their expressed needs (social media) and their latent needs (search data). Then work out how to have your candidate speak to those problems, and those desires.

Will polling survive? Yes. Mainly because the big beasts who run political campaigns are creatures of habit and also because polling companies do a great job interpreting data to make useful recommendations - the kind of insights a campaign needs. But in terms of understanding what matters to voters - and what impact that is going to have on turnout and voting intention - polling is dead.

Signify is an ethical data science company. We work with some of the biggest brands and human rights organisations in the world to understand and respond to the things people really care about. To learn more about our methodologies, please get in touch.