Polls, polls – and incredibly bad reporting of opinion polls

Yesterday’s papers carried news of a YouGov poll that put Labour 7 points ahead of the Tories. Today’s Times(£) excitedly splashes on a poll that puts the Tories one point ahead of Labour. That suggests a massive 8 point turnaround in just 24 hours. I don’t buy it – any of it, whether it’s Labour having a large lead of the Tories having a small one. What’s actually happening is that there’s statistical variation around a mean point – a Labour lead of 3/4 points – which is roughly where we’ve been for some months now.

That’s not to say that there hasn’t been a narrowing of the polls (if after a Tory conference that was broadly well received and a Labour conference that was broadly flat there wasn’t some poll narrowing, that’d be odd), but we won’t know that until we’ve seen numerous polls over a number of days (and preferably from a number of pollsters).

Unfortunately though, that’s not how the poll (and polls in general) get reported in the papers.

I spent years reporting polls wrong. I would happily talk about polling “surges”, attributing X poll result being  directly attributable to Y event. It seems many in the press are still doing likewise. It’s understandable I suppose, because if your paper has commissioned a poll in order to generate *SHOCKING NEWS*, and hacks write the poll up as “statistical variation around a mean point” then they’re going to incur the wrath of their editors sooner rather than later. But it does mean that journalists sometimes do harm to our political discourse and our understanding of what’s actually happening by writing stuff that is palpable nonsense.

To be clear – none of this is to say that I’m in any way relaxed about the Tories taking their first YouGov poll lead in 30 months. For me, that such a poll is even statistically possible within the margin of error means that Labour and the Tories are polling too close together for comfort. But I won’t be altering my view of where the polls are until I see a trend. Reacting to a single data point is terribly exciting, but it’s also pretty irrational.

(Update: How journalism works. Front pages for the one poll that puts the Tories ahead. Little interest in this new Populus poll that shows Labour’s lead more than double – Lab 38 (+2), Con 33 (-1), LD 8 (+1), UKIP 13 (-1), Oth 8 (-1). There is no trend suggesting a Tory surge, not yet anyway).

Which brings us onto the Sun, who have a Survation poll on the Heywood and Middleton by-election. The poll shows Labour on 50% of the vote, with UKIP in second place on 31% of the vote. That’s a pretty high UKIP vote, but it’s a 19 point Labour lead. The Sun booms(£) that the polls shows “voters abandon Ed Miliband” at the by-election, but the poll actually shows Labour’s vote *increasing* by 10 points since 2010. The Tories were on 33% in 2010, and now fall to just 13% according to the Survation poll. As Survation’s own press release states:

“UKIP in our poll have taken an eye-popping 45% of Conservative 2010 voters and 19% of Labour 2010 voters in this seat”.

Whilst it’s concerning that Labour voters are shifting to UKIP, Tory voters in this constituency are shifting at a much more significant rate. The Sun *should* be writing that almost half of all Tory voters have abandoned the Tories and David Cameron. That would be factually accurate. But instead they’ve completely ignored what the poll actually says to have a pop at Ed Miliband that’s not justified by the numbers. That’s just bad journalism. Survation go on to say:

“Labour’s “firewall” here are those 23% of voters who picked the Lib Dems in 2010 – the lion’s share of which, so far, push Labour’s vote HIGHER now than their 2010 share – a factor that will be some comfort for the Labour party considering their chances in LAB/CON marginal seats for 2015.”

Strangely, the Sun omitted such information from their own reporting of the poll. Presumably because “Labour look likely to retain Labour-held seat” wasn’t the story they wanted to write.

Of course there is, as anyone who has been out campaigning across the UK a significant problem with former Labour voters abandoning the party and switching to UKIP. The Fabians have some important research out on this, as reported in the Independent today. But to claim that a constituency poll which sees the Labour vote *increase* is bad news for Miliband and Labour just beggars belief…