The Meaning of Swing
As I’ve said before, I don’t really intend to turn this into a political blog – it’s primarily about writing and reading speculative fiction, with other geeky bits and pieces added. But today, I really couldn’t help myself. It’s the mathematician in me, I think: there is just something profoundly wrong about people who misuse statistics in order to champion their pre-determined point.
In this case, I refer to this gentleman.
Now a moment’s glance at the blog in question would suggest that the author really doesn’t see eye to eye with British Labour leader Jeremy Corbyn. Which is fair enough – people are entitled to their own opinions (personally I find it hilarious how blatant the pre-8th June posts are in predicting a landslide Labour defeat, while the post-8th June posts suddenly switch to complaining that Labour should have won a massive victory. Which is it, sir?). No, the issue is that the blogger presents us with this graph as evidence that Corbyn’s Labour is wasting votes by piling up support in safe seats (Britain, unlike New Zealand, still uses First Past the Post):
Based off this data, British Labour has just seen an extraordinary percentage gain in votes (bigger than 1997), for comparatively few seats. Oh no, claims our blogger, clearly this is due to an incompetent Labour campaign.
The basic problem here is that this is not how you measure electoral swing. There is one giant factor missing from the above graph – the simultaneous change in vote for the British Conservative Party.
To take some examples, 1997 saw Labour gain massively (+8.8%), and the Tories lose massively (-11.2%). Unsurprisingly, this translated into big seat gains for Labour. February 1974 saw Labour lose votes badly (-5.9%), but the Tories lost even more badly (-8.5%), so Labour picked up seats on a lower vote share. 1983 saw a massive Labour vote drop (-9.3%), but the damage was less than it could have been, because the Tories also lost some votes (-1.5%). And so on.
What to make of 2017? Well, yes, Labour’s vote went up massively (+9.6%), but the issue is that the Tories also went up significantly (+5.5%). This, not safe seat pile-up, is the reason for Labour’s relatively modest seat gain. Imagine a similar graph showing change in Conservative vote vs Seats – the Tories have just gained 5.5%, their biggest increase since 1979, while also losing seats. By the blogger’s reasoning, this must have represented an insane level of marginal seat neglect, which it wasn’t.
To actually measure election swing, you need to average the change in Labour vote with the change in Conservative vote:
Thus the swing to Labour in 1997 was 10% (thus massive gains), the swing to Labour in February 1974 was 1.3% (thus modest gains), and the swing to the Conservatives in 1983 was 3.9%. The swing to Labour in 2017? A tad over 2%, which is slightly better than Neil Kinnock managed in 1992 – based off which, you would expect meaningful Labour gains, but hardly a landslide.
Which, surprise, surprise, is what happened. Labour gained 30 seats net (the Tories lost 13). No need to bring in nonsense about safe seat pile-up.
There is admittedly the question of why the Conservative vote went up too – while this is straying away from the point of this post, I would invite the blogger to consider the change in UKIP vote (12.6% in 2015 to 1.8% in 2017), and then consider where, exactly, that vote might have gone in light of the 2016 referendum.