Lots of interesting details to digest in today’s mammoth YouGov marginals poll (including, ahem, the importance which voters give to whether or not a candidate is local – in brief, it’s very important), but one point I’ve not yet seen anyone make is how heavily it undermines the seat predictions for the Liberal Democrats which come from uniform nation swing calculations made using Electoral Calculus.
These are very popular calculations, most notably regularly topping blog postings over at Political Betting, but compare the YouGov figures for the number of Liberal Democrat seats from the marginals poll with those from the national voting figures in the YouGov polls conducted at the same time and you find that the marginals polls gives the Liberal Democrats nearly double the number of seats compared with the national swing calculations (44 compared with 24).
The national YouGov polling at the time was putting the party on 16%, so it is hardly a surprise that both figures are down on the party’s current number of seats and given the range of factors such as YouGov’s consistent lower polling for the Liberal Democrats, the long period of time to the next election, the usual rise in the Liberal Democrat poll rating during the general election campaign proper and indeed the fluctuations in the party’s YouGov ratings since (up to 20% in the latest), it would be unwise to read anything much into the overall figures.
But the difference between the marginals and national polling, when both were conducted by the same organisation at the same time, does give a good reason to shout “that’s a load of nonsense!” at your computer (or just tut quietly under your breath and put on a cup of tea; your choice) next time you see an Electoral Calculus projection used to predict the number of Liberal Democrat MPs.
UPDATE: As touched on in the comments, my use of the phrase “uniform swing” as short hand wasn’t ideal as Electoral Calculus uses a modified proportional swing model. However, Electoral Calculus still essentially does a blanket calculation across the whole country, albeit one in two parts by using a standard calculation to split party supporters into strong or weak depending on the party’s share of the vote in a seat, and then applying swing calculations to each part. It is the poor predictive ability of these blanket calculations that is at issue, whether they are flat uniform swing or more complicated modified proportional swing calculations.