Why the BBC’s election statistics are better than The Guardian’s (and others)

The BBC and The Guardian are reporting markedly different figures for seat changes in this May’s local council elections. How come?

The BBC have done their own calculations while The Guardian, and some other media outlets, are using data from the Press Association (PA). Although you might expect those numbers to end up the same, they’re not doing so.

There are three reasons usually why such numbers vary between sources, so let’s take them one at a time.

First, during the first few hours of results, it’s common for different sources to be out of synch on what results they have in. But we’re pretty much past that now.

Second, do you count changes from who held the seat just before the election or from who won the seat when it was last up in a normal election (i.e. do you take last time’s result, or do you factor in defections and by-elections since)? I prefer counting since the previous time’s normal election result as that gives the truer sense of how voters have changed since last time. This isn’t what the PA does, but it’s a small factor.

Third, and this is the big factor, boundary changes. We’ve had some big boundary changes this time and the BBC and the PA have taken very different approaches. Imagine that there’s a council with 80 seats and one party, the Chocolate Lovers, won half of them, 40. Then there are boundary changes, reducing the council in size to 40 seats (boundary changes and local government reorganisation tend to reduce rather than increase the number of seats). But the Chocolate Lovers had a stunner of an election, winning every single one.

What’s the net change in seats to report?

Under the PA approach you just report the change in totals – so the Chocolate Party by going from 40 to 40 has had no net change.

And yet… those chocoholics have also gone from half the seats to all the seats. So no net change sounds an odd figure.

Which is why the alternative approach, which is harder but better and is the one done by the BBC, is to calculate what the result last time would have been on the new boundaries and use that as the benchmark for seat changes. Working out seat numbers using last time’s vote shares and this time’s boundaries requires a fair degree of estimation and very local knowledge (as vote totals are published broken down to the very local level you need). But it means you get more meaningful seat change numbers out.

In my hypothetical case, it looks like under the new boundaries last time the Chocolate Lovers would have won about 20 seats (exact number depends on the distribution of votes between the wards) and so that would have given a net seat change figure of plus 20. ‘Plus twenty’ rather than ‘no change’ is more informative as a summary figure given that they’ve gone from half to all the councillors.

With some big sets of boundary changes and alternations in councillor numbers due to the new unitary authorities, that’s why there’s a noticeable difference in the output from these two approaches this time.

Overall, as boundary changes tend to reduce councillor numbers, this means that the BBC’s figures tend to favour those parties who are gaining in seats. But they are also more meaningful. (So yes, bookmark this post and come back to quote it in future years if the Lib Dems are going backwards and I succumb to quoting the PA statistics instead!)

The PA’s methodology is explained at the foot of the story here. My understanding of the BBC’s methodology comes from discussions with them.

Sign up to get the latest news and analysis