I mentioned a couple of days ago about developing a model for the 2015 election to try to get an idea of how current polling would translate into seats. I have had a couple of negative comments telling me that Electoral Calculus and UK Polling Report already do this job. Firstly I disagree that these methods will be completely accurate come the next election, secondly I see no harm in adding a new model, and thirdly, it is for my own fun as much as anything else. I will be sharing the data as I know that some are interested.
The main thing I am doing is looking at the shifts from party to party in the daily YouGov polls. Here is the data from today’s poll for example:-
The ‘2010 vote’ column shows how people voted in that election and how those votes now split – as you can see, the most drastic change is Lib Dem voters moving over to Labour – this figure is consistently around the 30% mark. Also standing out is 20% of 2010 Conservative voters moving over to UKIP – they are losing votes to the anti-EU party by a far greater margin than the other two main parties.
I have split the constituencies into different groups: Conservative/Lib Dem seats, Lib Dem/Labour seats, Conservative/Lib Dem seats, and others (usually the Scottish/Welsh nationalists, also Brighton Pavilion for the Greens). Until there is any real shift in polling, we can assume that every Labour seat from 2010 will be held (excluding any kind of bizarre local results/MP scandal etc).
I then use a formula based on the party’s 2010 vote share, and plug in the changes in vote share according to an average of the five most recent YouGov polls. This gives a greater sample without going so far back that the polling is completely out of date. The formula is adapted depending on the type of seat – the Lib Dems are given more credit in seats that they are defending, owing to their history of being hard to shift, and the Tories are given less credit in seats where UKIP did particularly well in 2010. With regards to the nationalist seats, there is not enough relevant polling, and I have allocated the seats on a rough uniform swing.
Using the most recent five YouGov polls, including today’s, we get the following figures:-
The sheet calculates new majorities for each seat – if you are interested in the prediction for a seat in particular then please let me know and I will show the data. I don’t see the point of uploading it in full as it will be updated each day.
I will be making a fortnightly or possibly weekly projection – hopefully I can look at some of the seats in more detail as time goes on.
I really do believe that the current estimates give too high a Labour majority – but aside from the opinion poll data used, there is nothing particularly scientific about this model, and it should definitely be taken with a pinch of salt.