There was a fascinating discussion on my Twitter timeline with Rob Ford, Will Jennings, Iron Economist and many other distinguished people, triggered by concerns about the Liberal Democrat revoke A50 policy. In short: the concerns expressed by some are that the Liberal Democrats might get the total majority they would need to enact this Revoke with a mere 30-35% of the vote, and that would be way short of the 50% endorsement sought by those wanting a referendum. And the fact that they could get this majority with just 35% was bolstered by the modelling I did ages ago in this popular blog post.
Which blog post I still stand by in outline, but developments since have shown up even more glitches in the system, including this feature: the LibDem seat total climbs very slowly at first, but then at some point it rockets as all sorts of seats fall. What this highlights is how having a very evenly spread vote across all constituencies is a massive disadvantage below a certain threshold, and only flips over to being an advantage when you hit the 30s in terms of vote share.
There is a corollary: learning how to concentrate your vote share is essential if you want to go above a small number of seats, as a small party: contrast the fortunes of the SNP and UKIP in the 2015 election.
Since that post, I have written a number of others exploring methods a model user might concentrate the LibDem vote share and get a different result; generally speaking, the outcome was about 20-30 more seats for the LibDems when their vote share is in the high teens/low 20s. Again, just what you would expect.
What I thought I would also share before heading off for my nighttime cocoa: that same variable becomes a disadvantage for the LibDems if they are looking for a majority. In other words, they begin to pile up pointlessly large majorities rather than gain more seats – just as hit the Tories in 1997, say, or Labour in 2017, when their votes did not go as far as they might in seats.
Here is a graphical representation: first the behaviour of party seats when there is no use of “historical LibDemmyness” in the machine (Solid line) and second the same relationship with a high degree of LibDemmyness and a little tactical voting
The dotted line suggests a much higher threshold for the LibDems is needed to get a majority – but still in the 30s. Maybe 5 percentage points higher. And none of this loopy, “450 seats plus” style outcome. Another reason to doubt whether a purely smooth swing is what we might expect.