Another electoral forecasting website has been set up, this time by Dr Chris Hanretty, Dr Benjamin Lauderdale and Dr Nick Vivyan. Their first prediction is as follows:

Party	Lo	Votes	Hi	Swing
Conservatives	29.2%	33.4%	37.6%	-2.6%
Labour	28.0%	32.1%	36.1%	3.1%
Liberal Democrats	11.5%	14.6%	18.0%	-8.4%
SNP	1.9%	2.5%	3.0%	0.8%
Plaid Cymru	0.7%	1.0%	1.3%	0.4%
Greens	1.5%	3.0%	4.8%	2.1%
UKIP	6.8%	9.4%	12.4%	6.3%
Other	2.2%	4.1%	6.2%	-1.7%

Party	LoW	Seats	HiGH	Swing
Conservatives	243	299	360	-7
Labour	246	304	353	46
Liberal Democrats	3	15	33	-42
SNP	5	9	16	3
Plaid Cymru	3	3	5	0
Greens	0	0	1	-1
UKIP	0	0	0	0
Other	1	1	1	0

Their forecast pools different polling companies’ data to produce their forecast. Given my work on the Liberal Democrats, I’ll focus on them in particular.

Should the party achieve 14.6% of the vote, then I think their seat forecast for the Liberal Democrats is low. Now, the vote forecast might be high. The Polling Observatory blog have suggested that the Liberal Democrats might not yet have hit rock bottom in the polls, and might struggle to reach double figures.  However, if they achieve 14.6%, then I would strongly suggest that they will win more than 15 seats.

The authors have noted that this prediction is because a uniform swing in seats cannot really apply in the Liberal Democrats’ context. If they lose 1/3 of their vote, then they will have to lose some strongholds. This is potentially true, and my own research has suggested that the party may struggle in previously strong areas, namely the London and Scotland regions.

However, you can also guarantee that the party will be running a campaign to keep hold of as many seats in their strongest areas as possible. Local election results have been much stronger in areas where they have MPs, and even where candidates are stepping down and losing the incumbent effect, they have chosen candidates very early on, and are focusing on these areas.

This will mean that certain areas are getting ignored. Many of the areas where the party polled 20%+ in 2010 and came second place, they will finish near the bottom, especially in those areas where Labour are the closest competition. So whilst strongholds may go, more votes will be lost in these areas, I would suggest, than in their strongest seats.

It’ll be interesting to see how the forecast develops over the next few weeks and months. Alongside the Polling Observatory, Elections etc and Electoral Calculus, it’s great to see another one to keep an eye on.

UPDATE: 14 AUGUST 2014

The writers have kindly responded to my blog post with the following:

Thanks to @cjnu1 for his ideas https://blogs.ncl.ac.uk/craigjohnson1/2014/08/13/electoral-forecast-hanretty-and-vivyan/ … on why we might be overstating Lib Dem losses. In response, @chrishanretty went back to the historical record, and estimated a better swing model for the UK using data from 1918 forward. The resulting best fitting swing model improves our retrocast of 2010 http://electionforecast.co.uk/2010/index.html  but does not change our 2015 LD seat forecast! I just plotted the seats-votes relationships from the posterior distribution of our model.

Looks like LD average 30 seats at a bit more than 18% vote share, might get 30 seats starting at around 16% vote share.

Thanks to the writers for doing that. It shows a stark contrast with Stephen Fisher’s model, which predicts 26 seats for the Liberal Democrats on as small vote share as 11.8%.

I don’t envy electoral forecasters at the best of times, but I certainly don’t envy those looking to generate a model predicting Liberal Democrat/UKIP vote share and subsequent seat share in 2015. As I note above, I’d expect a 14% vote share for the Liberal Democrats to generate more seats than the 15 that the above model predicts, but that will depend entirely on where they can maintain support.