The accuracy or reliability of opinion polls is often questioned, especially when the polls are about voting intentions ahead of elections.
I’ve previously written about why 1,000 or so samples are large enough to measure views across a whole nation. But what about specifically the accuracy of voting intention figures reported by the polls? Can we trust such opinion polls?
Let’s take a look at the evidence as to the accuracy of political opinion polls.
A recent study looked at 30,000 national polls from 35 countries across just over 70 years. The verdict? Political opinion polls are pretty accurate, and if anything have been getting slightly more accurate in recent years. That’s good reason to trust the polls.
Or in more detail:
Although claims about the demise of pre-election polling have become popular in recent times, we find little basis in fact to support them. Relying on vote intention polls from more than 300 elections in 45 countries over a period of more than 70 years, there is no evidence that poll errors have increased over time. And the performance of polls in very recent elections is no exception.
But we’re particularly likely to notice polling errors because:
[Polling error] tends to be greatest for the largest parties (or candidates), which are the ones competing for power. Moreover, these errors are most consequential when elections are close, as they can be decisive for government control, as was the case in the 2015 and 2017 UK general elections and the 2016 US presidential election.
Errors also tend to be higher in the sorts of elections that, if you’re a reader from England, you are most likely to notice because:
In the subset of countries for which we have consistent poll data over a long time period, there is evidence that errors tend to be lower in PR systems, consistent with vote choices being based on partisan loyalties, which tend to be more structured.
This isn’t the only study of this type that’s been done, but they all come to similar conclusions about the accuracy of the polls. You can read more about those other studies, along with the real story of what happened to the polls in the 2016 and 2020 US Presidential elections and the 2016 Europe referendum in my book, Polling Unpacked: the history, uses and abuses of political opinion polls.
Or for more on the study quoted in this post, here it is in full.