Thread by @BestForBritain

Our (massive) MRP voter analysis exercises are best-of-breed but we do get people asking how accurate they are locally at a constituency level. Gather round, while we slay that particular data dragon. Or do we..?

THREAD 1/n

bestforbritain.org/getvoting

We love data. It informs everything we do. In fact, according to this WIRED piece on tactical voting, it takes 2.5 days to crunch the data for just one of our MRP surveys, using some pretty hefty cloud computing

wired.co.uk/article/tactic…

2/n

So, when someone suggests our figures aren’t necessarily accurate at constituency level, we seek out other data to test the veracity of that claim. Got to arm ourselves against that data dragon.

3/n

Now, there isn’t that much suitable constituency-level polling to draw on. But, happily, some was published yesterday by The Observer and respected pollster Peter Kellner.

4/n

theguardian.com/politics/2019/…

Thank you @ObserverUK for the gift of local polling. But what does it show, and should we be worried?

In the 3 polled seats (more to come), the Tories lead, Lib Dems are 2nd, Labour 3rd.

5/n

So far, so uncontroversial. Our figures, though obviously not identical, show the same running order. You wouldn't expect our figures to be identical, because, for instance, we did our research before the Remain Alliance was finalised

(don't worry, we'll update again soon)

6/n

Where it gets *really interesting* is when Kellner dives into the detail, in particular looking at what happens if each seat is a 2-horse race…

Each of the 3 polled constituencies is a Remain-voting London marginal in which Lib Dem support has surged.

7/n

These are important seats in terms of halting Johnson’s barmy Brexit plans, and so much will come down to tactical voting. We really don’t want the data dragon huffing and puffing here.

8/n

CRUCIALLY, The Observer poll asks how people would vote if they thought it was a 2-horse race: either Tory v. Labour or Tory v. Liberal Democrats

9/n

The Tory-Labour race figures predict a Conservative victory. This does not make us happy but, as we said, the Lib Dems are currently polling second in each of the 3 constituencies. Which means…

10/n

…that, realistically, the race is likely to be between the Tories and Lib Dems. In that scenario, the Observer figures show Lib Dems taking all 3 seats off the Tories, decisively.

11/n

OK, back to our own MRP analysis and our tactical voting recommendations in these constituencies.

12/n

In each of these seats, we recommend that voters wanting to stop Brexit should vote for @SamGyimah in Kensington, @lucianaberger in Finchley & Golders Green, and @PaulKohlerSW19 in Wimbledon.

#GetVoting #TacticalVote

13/n

Don't forget, the VAST majority of our recommendations are FOR Labour.

14/n

getvoting.org

In summary, our tactical vote recommendation, based on national MRP figures, is backed up by the Observer’s tactical voting predictions based on local polling.

15/n

There have been a handful of other constituency polls and the results have mirrored what we’d expect with our MRP. To be frank, if they hadn’t, we’d be working very hard to find out why.

16/n

So, the data dragon is slain or at least badly injured. We’ll keep watching constituency level info closely to sense-check our own data.

For the record, no real dragons were harmed in this analysis. Data dragons suck, real dragons rock

17/17
END.