Skip to main content

A Word About Predictions

Predictions can make one look like a genius or a fool.

In reality, no one is a genius, just lucky. And arguably, the only thing foolish about predictions is making them in the first place. So, take my projections with a grain of salt.

But I do have a methodology. It's based on Multi-Level Regression and Poststratification (MRP).

For a discussion of MRP, please see my December 8 post Number Crunching the Polls Points to Big Tory Win.

Don't knock the idea or YouGov if you do not understand what this is all about.

Fearless Forecast Methodology

  • I believe in YouGov's MRP methodology and so does Martin Baxter at Electoral Calculus.
  • The Electoral Calculus model, at least as I understand it, uses Baxter's own MRP methodology based on an average of the latest polls.
  • I have a different twist. Instead of using an average of polls, Baxter's MRP methodology with an an average of YouGov polls and enter those averages into the Electoral Calculus User-Defined Regional Predictor.

YouGov does not disclose its MRP and neither does Baxter, for obvious reasons.

Why Use Baxter's Model?

  • No one else has user-defined calculator.
  • Baxter's model performed better that YouGov's.

Why Use YouGov Data?

  • YouGov and ComRes are the only pollsters that present region-by-region data.
  • Variances in ComRes tend to be extreme.
  • YouGov does frequent polls and one can average them.

YouGov Data

The data I entered into Electoral Calculus is from MRP Methodology, Tables and Figures on December 10.

Image placeholder title

YouGov Regional Data

Image placeholder title

I plugged those numbers into the Electoral Calculus Base Calculator + Scotland and then again into the regional calculator.

Scroll to Continue


The former predicted a majority of 42, the latter, a majority of 40. That's a nice cross-check. I went with the latter.

Mish's Fearless Forecast Seat-by-Seat Changes

Image placeholder title


  • Labour a Huge Loser: It picks up two seats from Plaid, a Wales independence party, but loses 38 seats elsewhere, 37 of them to the Tories the other to the Liberal Democrats.
  • SNP picks up 6 seats, 5 of them from Labour, only 1 from the Tories.
  • The Liberal Democrats pick up 5 seats, 4 from the Tories, 1 from Labour.

Tactical Voting

If the above expectations are in the ballpark, tactical voting will cost the Liberal Democrats dearly, perhaps as many as 10-15 additional seats.

Insead of 18-19% of the vote they will get 12-13% of the votes.

For What?

For a Tory majority and a handful of extra seats, with Jeremy Corbyn still potentially hanging on.

That's what.

By the way, 18% of 650 is 117.

Instead, I expect the Lib Dems to get about 17 seats. And the more tactical voting there is, the fewer pickups for the Lib Dems there will be.

YouGov Model vs Mish

The YouGov Final Projection is a Conservative Majority of 28.

As explained above, I estimate 40.

Final Thoughts

Any majority of 10 or greater is likely to be very stable and "Get Brexit Done".

Even 5 might suffice for 5 years because once animosity over Brexit is over, DUP will again side with the Tories in most legislative matters.

Mike "Mish" Shedlock