UK-Elect is probably the best forecasting software for the U.K. election that is available. I recently asked Tim Bickerstaff, UK-Elect’s creator, a few questions about the upcoming U.K. election.
(If you’re going to get the UK-Elect software, you can use the coupon code 270SOFT to get a 25% discount from the UK-Elect site. Good forecasting!)
Q. You’re releasing weekly forecasts at ukelect.co.uk – to what extent are the forecasts using calculations from the UK-Elect program itself, and how are these distinguished from other forecasts that are out there?
A. Yes, I’m releasing at least one forecast a week, but more normally two. It may even be three in the last week before the election.
The forecasts are entirely made using UK-Elect, and all the maps, spreadsheets etc. are created by it. I input the approximate latest poll averages for Scotland, Wales, London, and the whole of Great Britain, and then do a forecast using the latest standard “UK-Elect v9.4 method” (UK-Elect also supports a choice of other forecasting methods, such as Uniform National Swing). The only change from the default settings is that I also enable the setting that takes into account how long it is until the date of the election (May 7th). The main effect of that is usually to slightly reduce the UKIP, SNP and Labour votes and slightly increase the Liberal Democrat and Conservative votes. That setting is basically taking into account the late swings back towards previous party vote support levels that have often taken place at previous UK elections (although not all previous elections, so this swing-back theory is a little contentious).
In terms of factors taken into consideration, the UK-Elect method is probably one of the more complex ones. As an example, it uses the final candidate lists and takes into account not only which parties are standing in every consttiuency, but also whether the incumbent MP is standing again, and whether he or she won the seat for the first time at the preceding election. (This is because some academic studies have shown that a first-time incumbent attracts a relatively higher level of support than other incumbents, or new candidates.) It also – very importantly – makes separate forecasts for Scotland , Wales and London using separate percentages and combines those forecasts with an overall GB one. Because of this UK-Elect may well have been the first forecasting website/program to reveal the likely effect at Westminster of the SNP surge.
In recent months the UK-Elect forecasts have generally been close to the mean of the other forecasts, predicting similar number of MPs of the main parties to most – perhaps slightly fewer Conservative and a few more Liberal Democrat seats than some. That effect is probably because of the built-in incumbency calculations which are assuming a quite high-level of incumbency support for the Liberal Democrats. Lord Ashcroft’s constituency polls seem to confirm that the Liberal Democrat incumbency support exists, but some forecasters believe the effect is exaggerated. We will have to wait until May 7th to see who is correct.
Q. Indeed. Let’s talk a bit about the SNP surge, since it seems an important element of the composition of the next Parliament, and might even determine who will be the main party in some kind of coalition or working arrangement. Is this relatively unprecedented surge of a regional party something that you think UK-Elect can model, given the complex arrangement of 4 (or perhaps 5 or 6, depending on how they are counted) major parties voters might move from or to in Scotland? Also, can anyone who has the UK-Elect software generate their own forecasts on a regional basis, experimenting around with different numbers to see what the likely effect would be or a varying SNP surge?
A. At current levels the SNP surge isn’t especially difficult to forecast – they are actually quite close to the 50% level where they could quite conceivably win every seat in Scotland! And in truth at current anticipated support levels there are currently only three major parties in Scotland – SNP, Labour and Conservative, with the Liberal Democrats and the others far behind.
And yes, of course, with UK-Elect separate percentages can easily be used for forecasting Scotland, Wales, London etc. and an overall GB (or UK) set of percentages given as well. UK-Elect can also forecast a single region individually if required. Not only can percentages easily be adjusted to see what happens, but other factors such as the amount of incumbency support, or tactical voting, can be simulated.
Q. You mentioned Lib Dem support. Outside of the two main parties, this seems to be the other crucial piece in this election in terms of who is going to win which seats, in addition to the SNP surge. How do your forecasts try to model what’s happening there, and how can UK-Elect owners try out different options that might affect the Lib Dems?
A. The level of Liberal Democrat support in the seats that they already hold is one of the major unknowns of this election. Their overall national vote appears to be sharply down but many political experts believe that some of their incumbent MPs have built up considerable levels of personal support, enough for them to survive in current circumstances. In terms of seats this incumbency factor could be enough to save 10 or 15 of their MPs – to make the difference between an overall total of around 15 or so surviving and perhaps as many as 30. Evidence from opinion polling is mixed – a poll of Liberal Democrat seats in South West England suggested that they were on course to lose all of their 14 seats in the region. By contrast some detailed constituency polling done on behalf of Lord Ashcroft indicated that they could hold perhaps half of those seats. This difference would affect not only the Liberal Democrat seat total but also the Conservative total, as they stand to gain most of these seats.
UK-Elect users can simulate varying levels of incumbency by either trying out different forecasting methods, such as Uniform National Swing (which has no incumbency support built into it by default) and comparing them with forecasts done using the UK-Elect v9.4 method (which does support incumbency), or by directly enabling and disabling the incumbency support via a forecasting option. In addition the level of incumbency support can be specified as separate user-specified percentages for each party in each of the cases where a sitting MP is seeking re-election for the first time, or has been an MP for more than one previous parliamentary term, or is standing down. By these means many different scenarios can be simulated.
Q. Ukip and the Lib Dems are parties that seem to be on very different trajectories. In your latest forecast, you have Ukip at about 13% and 2 seats (the first they would win in a general Parliamentary election), with the Lib Dems at about 9% and 26 seats (down from their current 56). Yet, both leaders of those parties are in constituencies where they might lose – South Thanet and Sheffield Hallam. What does your latest forecast say about leaders Nigel Farage and Nick Clegg?
A. The latest forecast suggests that both Nigel Farage and Nick Clegg will be elected, although fairly narrowly. Some very recent polling evidence suggests that Farage may now be establishing a clear lead in South Thanet, although polls have disagreed previously about his chances so the trend towards him remains to be confirmed.
Q. If you were to give one word of advice to people looking at polls as we move closer to the election, what would it be?
A. Relax. Basically, remember the margin of error (usually + or- 3%) and don’t take any one poll (or even two or three polls) too seriously.