How to read a marginals poll
I've my usual weekly update to do later tonight or tomorrow morning, but in the meantime a quick note on the new ComRes poll of marginal seats, which has tables here. My heart always sort of falls when I see a marginal seats poll, as I know how people will misinterpret it. Twitter inevitably sees lots of comments about how Labour has a great big lead in the marginals that count, or Labour might be neck-and-neck nationally but they have a big lead in the important marginals - seemingly oblivious to the fact that the topline results and what they mean depends entirely on which marginals you are polling. What could be a bad headline result for Labour in a batch of seats they only just missed out on in 2010 could be an excellent one in a batch of target seats they need for an overall majority. The crucial thing when looking at marginal polls is the swing from the last election, and how that compares to the swing implied by the national polls.
So, using today's ComRes poll as an example, here's how you should interpret a poll of marginal constituencies. I expect this will covering old ground for many people, so I hope regular readers will forgive me stating the bleeding obvious in some cases!
The first thing is to look at what the shares of the vote were in those seats at the previous election - without knowing that, any swing or change is meaningless. In the case of ComRes, their marginal seats polls cover the 40 most marginal seats with Labour and Conservative in first and second place. That consists of 25 Conservative seats and 15 Labour ones, with shares of the vote at the last election of CON 37%, LAB 37%, LDEM 18%, UKIP 3%.
The next step is to compare the results of the previous election to what the poll shows. Topline figures in today's ComRes poll, with changes from the election, are CON 31%(-6), LAB 40%(+3), LDEM 8%(-10), UKIP 15%(+12). From being equal at the last election, Labour now have a nine point lead - the equivalent of a 4.5 percent swing from Conservative to Labour (the swing is the relative change, divided by two).
Now compare it to the national polls. Looking over the current UKPollingReport average it shows a two point Labour lead. Labour were seven points behind the Conservatives in 2010, so conveniently enough this is also a nine point relative change in the lead, and the equivalent of a 4.5 percent swing from Conservative to Labour. In other words, this poll doesn't show Labour doing better or worse in the marginals, it shows the Con-Lab marginals behaving exactly like the national GB polls suggest they will.
Of course, this is just a single poll of 1000 people anyway, so even if it did show a swing that was a point or two higher or lower in the marginals it wouldn't be worth reading much into. It could just be sample variation. Looking back over other Ashcroft and ComRes polls in Con-Lab marginal seats we've had some showing the Conservatives doing slightly better in the marginals, some showing them exactly the same, some showing Labour doing slightly better. All in all it suggests the swing in Con-v-Lab marginals is very much in line with the swing in the national polls. And that's to be expected. If history is any guide we might expect the Conservatives to do a little better as most of the battleground seats have new Conservative incumbents, but only by a very small amount - the reality is that Con-Lab marginals do tend to behave in pretty much the same way as the nation as a whole does.
