I wasn’t expecting that.

The Times has reported its latest opinion poll. It has reported the changes in party share of the vote.

And then Peter Riddell has said,

These shifts are within the margin of error

Why’s that impressive? Because nearly every opinion poll only shows changes within the margin of error (you’ve usually got to look over a wider pattern to see statistically significant changes), but that doesn’t stop newspapers writing up their stories as if the changes in support are significant and therefore ones we can be sure actually happened.

It’s as if the newspapers think, “Look, we know the poll doesn’t really say for sure that party X is up, but frankly it’d be pretty tedious to say there’s no news from the poll we’ve just paid good money for, so let’s just gloss over that and all pretend we are sure party X is up”.

Some of the better journalist (take a bow, Julian Glover at The Guardian) often give a wider context to a poll which helps highlight what is really going on.

But I’m impressed that The Times has actually gone as far as making this point explicitly.

Ironically, for all the media’s love of US politics at other times, this is one respect in which we’ve not copied US habits and should have: the US media are generally much better at making clear that, for example, a small lead is actually a statistical dead heat.

But perhaps the UK is starting to catch up. And I suspect The Times will do rather well in my next monthly round-up of how newspapers are doing at reporting their own polls.

If you’re not familiar with statistical significant think of tossing a coin ten times. Suppose first time you get four heads and the next time you get six heads. Does that mean someone has switched the coin in between each set of ten? Of course not. It’s just chance. But if you got o heads the first time and 10 heads the second time, that’s looking more like there was a switch. It’s the same with opinion polls. Because they poll a sample of the population, a change between two polls may be down to chance. It’s only if the change is big enough we can be pretty sure it means something has actually changed.