The professor will see you now - too few elections
In an occasional series, Professor Philip Cowley offers a political science lesson for The House’s readers. This week: on there being too few elections
How many times has a British government recovered electorally from being consistently 15 to 20 percentage points behind in the polls? On my reading of the data the answer is zero, although it does partly depend on how you define “consistently”.
But then how many times has a government been consistently 15 to 20 percentage points behind in the polls? Excluding the current administration – and again accepting some definitional issues – I think the answer is once.
So that’s a success rate of 0 per cent, but out of just one case. How wise do we think it is to make predictions about future behaviour based on an N of one?
This is all part of a wider problem, which I discuss in my yet-to-be written magnum opus, The Small N Problem in Electoral Studies (Wishful Thinking Publications, 2028). Put simply: we just don’t have enough elections.
There have been 28 general elections in Britain since all men over 21 and most women over 30 gained the vote in 1918, and 24 since equal suffrage 10 years later. There have been just 20 since 1950, when ‘one person, one vote’ was first applied.
Even if all 20 had been fought under similar conditions, this would still be a pretty small sample on which to base predictions. And we know that they have not been fought under similar conditions. Instead there have been dramatic changes in things like the party system, campaigning techniques, and voter demographics and behaviour, all of which narrows our sensible frame of reference yet further.
Take, for example, the rise of voter volatility, as detailed in the most recent book from the British Election Study team. There is that famous quote attributed to Harold Macmillan (“Events, dear boy, events”) about the power of circumstances to derail even the best laid plans. Just because – like most of the best political quotes – it would appear to be entirely apocryphal does not undermine its value. Stuff happens. And one consequence of voters being less solidly aligned with parties is that stuff can now move voters in a way it might not previously.
How many election campaigns have seen sizeable shifts in the levels of party support? Again, it partly depends how you define sizeable, but I think the answer is just two: 2010 (when the Lib Dems surged, only to then deflate before polling day itself) and 2017 (when Labour went from 20-plus percentage points behind to denying Theresa May a majority). Should we see that as two out of the 20 since 1950 – in other words, as infrequent events that are probably unlikely? Or we could see them as two out of the last four, yet another sign of how much more volatile voters are becoming?
You might, similarly, note that the opinion polls seen in this Parliament have ranged from a Conservative lead of 26 points, to a Labour one of 39, a larger spread than in any post-war parliament. What the electorate giveth, it can take away.
And then think about the specific circumstances of contests. How many elections were fought – as the next election will be – after the incumbent party had been in office for more than a decade? Since 1945: just four. How many of those were fought by a different PM to the one who won the last election? Three (of which one resulted in a government victory). How many were fought by a different PM to the one who took over from the one who fought the last election? None, obviously.
This is, I think, why we keep reading pieces on whether the coming election will be like 1992 or 1997, when it is most probably not going to be like either. Because the number of cases that it could resemble is just so limited.
Further reading: E Fieldhouse et al, Electoral Shocks: The Volatile Voter in a Turbulent World (2020)
Get the inside track on what MPs and Peers are talking about. Sign up to The House's morning email for the latest insight and reaction from Parliamentarians, policy-makers and organisations.