Assuming a May 7th election date, today is almost exactly 18 months to go until the 2015 UK General Election and the model predictions have been updated for this week. There hasn’t been any huge shift in the results, with the probability of a Conservative majority dropping by about 1%.
This is partly due to a small shift in the polls in Labour’s favour. On aggregate Labour were polling around 5.3% ahead of the Conservatives two weeks ago but today the difference stands at 6.2%. While it’s possible voters have responded to recent debate about energy pricing in the UK, the uncertainty associated with these polling numbers is about 1% so this is not a significant shift at this stage. The other secondary effect at play here is that the Conservatives need the polls to move in their favour in the next 18 months in order to win. As the time remaining until the election decreases the model believes this is increasingly unlikely.
Overall though the picture remains largely unchanged with the Conservatives requiring a significant shift in the polls over the next 18 months to achieve an overall majority.
The first election predictions for a May 2015 UK general election are now visible on the right of the page. The headline numbers are: the model believes there is a 55% chance of a Labour overall majority, an 18% chance of a Conservative overall majority and a 27% chance of a hung parliament. Although things look good for Labour currently, with a steady lead of about 6% in the polls, there’s still time for a lot to change and things don’t seem as hopeless for the Conservative party as some commentators are suggesting.
The biggest single uncertainty associated with election predictions at this point is how much the polls might move between now and May 2015. If things remain as they are, the model suggests the Conservatives have only around a 1% chance of achieving an overall majority. However 18 months is a long time in politics and historical data shows a 10% swing in the Conservatives’ favour is far from impossible.
It will be very interesting to watch how public opinion changes in the run up to the election. I suspect a lot will depend on the perceived state of the UK economy. Sufficient good news that the Conservatives can paint a picture of broad recovery under the current parliament is probably essential for their election chances.
In this post I’m going to describe the basic components of the election prediction model. I’ll go through each of these in more detail later but for now this will give an overview of how the prediction comes together. The model is made up of three main components:
- Aggregating polling results to get a picture of the level of support each party enjoys
- Projecting possible changes between now and May 2015
- Predicting possible election outcomes for May 2015
The first part is important because individual opinion polls only survey a small group of people and so their results have a lot of random variation. By looking at all recent polls together, it’s possible to get a clearer picture of the real situation.
Projecting poll changes is a component I haven’t seen many other UK election prediction sites attempt. Whereas most sites predict what would happen if an election occurred today, I am attempting to project what might happen in 2015. Obviously it’s impossible to predict what will happen between now and the next election, but based on historical data it is possible to ask how much the polls might plausibly move over the next 18 months.
For now the model almost exclusively considers the simple swing between the Conservative and Labour parties. This is an area in which it is weaker than some other public models and I plan to improve this in the near future. However, the rising popularity of UKIP will make modelling the effect of smaller parties especially hard for 2015, since there is little precedent for the presence of a significant fourth party in UK electoral history.
Over the next 18 months, this site will be used to release predictions of the result of the 2015 UK General Election. My background is data analysis not political science so my approach will be to rely on historical data and statistical models. The initial model posted here will be simple but will improve significantly over time.