One of the things I talk about a lot is what poor witnesses people are to their own behaviour – especially predicted behaviour. But why does this happen?
Here’s a great example, from an unusual source – American politics site Politico, which is covering the US presidential race in exhausting detail, and latched onto a particular study comparing voting intentions to behaviour.
What’s the big deal? We all know voting intentions don’t match behaviour. But what the study looked at wasn’t the candidate people were voting for, but whether they planned to vote at all. And it turns out that over half the people who tell pollsters they don’t expect to vote – and drop out of the survey because of it – end up voting after all.
Now – perish the thought! – I dare say some people are saying they won’t vote to get a pollster off the line. But obviously a lot of people are not judging their intentions correctly – they don’t intend to vote and yet, on election day, there they are. Why?
It’s because, the study says, they get swept up in the excitement. We are social animals, we like to copy what other people are doing, and if a lot of your neighbours and friends are going to vote, so will you even if you hadn’t really meant to.
This matters, says Politico, because an awful lot of US polling is based on “likely voters”, i.e. polls taking these claims at face value. But the implications are wider. It suggests that predicted behaviour gets far more unreliable when the prediction involves an event or something else highly social. When a Triple-A game or a new iPhone is announced, for instance, how many of the people who declare their lack of interest end up getting it anyway? Turn a launch into an event and you may find your predictions go pleasantly awry.
As for researchers, it underlines – yet again – how important it is to take the emotional context of an action into account when you’re trying to collect data around that action. It’s likely that no amount of contextual hacking in a survey can replicate the social weight and excitement of election day, but that doesn’t mean pollsters have to be quite so dry…