People in the political media (myself included) have a tendency to get excited about individual polls – this result or that result are a handy snapshot to tweet to imply things are moving in whichever direction we might prefer. It’s a natural instinct in conflict with statistical good sense.
Instead, pollsters remind us each time we do it, it’s the trends you need to watch. One sample might be an outlier, or one weekend’s headlines might give a party a temporary fillip.
So it was last night when YouGov put the Labour lead at 2 per cent and ComRes put it as low as 1 per cent. Politicos tweeted away, full of glee or depression, while stattos muttered about trends – and Mike Smithson of Political Betting reminded us that:
Not all polls bad for LAB today. Earlier Populus had LAB 40 CON 33 LD 11 UKP 8
— Mike Smithson (@MSmithsonPB) January 27, 2014
It’s a good point – cherrypicking by pollster or by day is a way to confirm your own prejudices and hopes, not to accurately represent the data.
So, using the data helpfully provided by Anthony Wells, I’ve looked at the trends over the last year and a bit.
Here’s a chart of Labour’s lead in every voting intention poll result from all the main pollsters from 1st January 2013 to 26th January 2014 (the red dotted line is the trend):
It’s pretty clear – while individual polls have the potential to bounce around on a day to day basis, the overall trend is unquestionably downward.
But are some pollsters skewing the picture? After all, while some like YouGov contribute over 200 tracker polls to the dataset, others only have a couple of dozen results in there (the disparity is caused by which contracts they win – the Sun hires YouGov to do a daily poll). It’s conceivable that one company polling a lot but with flawed methodology could distort the picture.
So I picked out some of the main firms to chart their results individually:
I think you get the picture. Down, down, down – uncompromisingly down.
In fact, the only one of the seven pollsters I looked at which found anything other than a falling trend in Labour’s lead was Populus, the firm picked out by Mike Smithson above:
It’s worth noting that Wells’ dataset for Populus begins in July 2013 – so misses out a period in which most other pollsters all found the Labour lead up in double digits. Even taking the Populus figures on face value, they’re an outlier when compared to all of their competitors.
None of this is to say that the next election is in the bag, of course.
For a start, I’ve been looking at Labour’s lead – it’s depressing to note that while that measure decays, it does so mostly because Labour’s vote share is falling, while the Tory vote has increased only slightly.
Furthermore, as Paul noted this morning there are all sorts of complications that will come into play in a real election. Labour have the advantage of unfair constitutency boundaries. UKIP may well disproportionately hurt the Conservatives. Left-wing Lib Dem voters may well split disproportionately to Labour where elections are close enough to encourage tactical voting. Each party is preparing hefty campaign activity in marginal seats. And so on.
On the national stage, though, Labour’s poll lead is still only going one direction.