The six seats UKIP could win

April 4, 2015 § Leave a comment

Election Forecast predict with 90% confidence that UKIP will win between zero and one seats.

Election Forecast is an excellent site, and I love their model – which mixes local and national approaches and is the closest thing we have to the much more sophisticated Nate Silver style prediction models that Americans enjoy, and which sadly don’t seem to have crossed the pond. I also agree that the way in which the first-past-the post system makes it almost impossible for new and small parties to win seats means that it is very unlikely that UKIP will win more than a very small handful of seats.

However the idea that Clacton is the only seat in Britain that UKIP has a chance of winning stretches credibility. Election Forecast give UKIP a 57% chance in Clacton, a 2% chance in Thurrock, and a 0% chance everywhere else. That just doesn’t ring true. Nigel Farage might not win Thanet South but his odds are greater than zero. Mark Reckless may well not hang on in Rochester and Strood, but it is at least worth him giving it a shot.

The problem here is not Election Forecast’s model, which is the best we have, but simply that models of this sort – which are required to rely heavily on Uniform National Swing (or some variation thereof) as its primary indicator – are not well suited to dealing with small political parties. Particularly not small political parties like UKIP who might, or might not, be about to make a breakthrough. Uniform National Swing (UNS) is a good indicator – historically better than one feels it logically should be – but it’s just too blunt a tool for so delicate a task.

But given we don’t have any more delicate tools at our disposal what can we do?

I decided to start with what UKIP themselves say they are setting out to achieve, and then see how much of that is possible. Sky reported back in August that UKIP were strongly targeting 12 seats: North Thanet, Forest of Dean, Sittingbourne and Sheppey, Aylesbury, East Worthing and Shoreham, Great Yarmouth, Thurrock, South Thanet, Eastleigh, Portsmouth South, Boston and Skegness, and Grimsby. To that we would logically assume they would have added the three seats they won in by elections: Clacton, Rochester and Strood, and Castle Point.

That’s 15 targets, which already seems like far too many for a party which was third or worse in every seat it stood in in 2010.

Let’s use Uniform National Swing to see how big of a task winning those seats would be. Actually let’s not because there doesn’t seem to be a UNS spreadsheet for British General Elections available anywhere online – however there is one for Electoral Calculus’ system which is essentially a slightly altered (they would say improved) version of UNS. Their figures for some of the key UKIP targets do seem a bit high, suggesting that they may have already added in by-election figures, but let’s assume they know what they are doing (they have a very good track record).

Feeding in the results of the latest polls to this model we see that UKIP are on course to win two seats, Clacton and Thanet South, quite comfortably. For their other targets they fall short by between around 3% and 11%.

Now what I feel UNS fails to capture is that in those seats where UKIP are close, I think their chances are better than indicated because of the impact that heavily targeting a seat can have. We saw this with the Greens in Brighton Pavillion, a seat which no UNS based forecasting model would ever have given to the Greens.

By far the most successful piece of political targeting of modern times was the Liberal Democrats’ 1997 campaign. They nearly trebled their number of MPs (18 to 48) despite the fact that their share of the vote actually went down (19% to 18%). Various things combined to make that election unique and for various reasons it is certain that UKIP won’t be as successful in their targeting as the Lib Dems were. Most importantly the 1997 Lib Dems benefited from massive tactical voting to get the Tories out, and that will not happen for UKIP in remotely the same degree. Futhermore the impact of this tactical voting was that the Lib Dem figures from that election are fairly misleading: their share of the vote only went down because so many of their supporters were tactically voting Labour in Labour’s target seats.

However, as an experiment, let us assume (and I feel it is an assumption that is generous to UKIP), that UKIP’s targeting strategy is as successful as the Lib Dems’ was in 1997.

On this lovely website you can see the election results for every election since 1832 and so see how parties did with respect to their targets. Let’s take the Lib Dems 1997 targets and, for simplicity, look at them in groups of 10:

  • GROUP 1 – targets 1-10: Brecon & Radnor to Falmouth & Camborne. Winning these seats required the Lib Dems to outperform UNS by 1.13-3.31%. The Lib Dems won 80% of these.
  • GROUP 2 – targets 11-20: Devon West & Torridge to Torbay. Winning these seats required the Lib Dems to outperform UNS by 3.31-5.06%. The Lib Dems won 70% of these.
  • GROUP 3 – targets 21-30: Bristol West to Salisbury. Winning these seats required the Lib Dems to outperform UNS by 5.06-6.91%. The Lib Dems won 40% of these.
  • GROUP 4 – targets 31-40: Teignbridge to Carshalton & Wallington. Winning these seats required the Lib Dems to outperform UNS by 6.91-8.63%. The Lib Dems won 20% of these.
  • GROUP 5 – targets 41-50: Newbury to Bournemouth West. Winning these seats required the Lib Dems to outperform UNS by 8.63%-9.46%. The Lib Dems won 30% of these.
  • GROUP 6 – targets 51- and on: Newbury to Bournemouth West. Winning these seats required the Lib Dems to outperform UNS by 9.46%+. Their biggest win was Kingston on a 10.94% above UNS swing The Lib Dems won 20% of seats up to Kingston.

Now let’s go through the UKIP Targets:

  • Clacton and Thanet South – as previously mentioned, UKIP should win these seats based on Electoral Calculus’ version of UNS alone.
  • Rochester and Strood – UKIP currently need to outperform UNS by about 2.85%, making it an equivalent to a group 1 seat for the Lib Dems in 97. Give them an 80% chance of winning.
  • Portsmouth South – UKIP currently need to outperform UNS by about 4.95%, making it an equivalent to a group 2 seat for the Lib Dems in 97. Give them an 70% chance of winning.
  • Thurrock – UKIP currently need to outperform UNS by about 5.7%, making it an equivalent to a group 3 seat for the Lib Dems in 97. Give them an 40% chance of winning. (Although personally I would have thought the fact Thurrock is a Labour-Tory marginal means UKIP have no chance)
  • Grimsby – UKIP currently need to outperform UNS by about 5.9%, making it an equivalent to a group 3 seat for the Lib Dems in 97. Give them an 40% chance of winning.
  • Great Yarmouth – UKIP currently need to outperform UNS by about 6.1%, making it an equivalent to a group 3 seat for the Lib Dems in 97. Give them an 40% chance of winning.
  • Eastleigh – UKIP currently need to outperform UNS by about 6.5%, making it an equivalent to a group 3 seat for the Lib Dems in 97. Give them an 40% chance of winning.
  • Castle Point – UKIP currently need to outperform UNS by about 6.75%, making it an equivalent to a group 3 seat for the Lib Dems in 97. Give them an 40% chance of winning.
  • Boston and Skegness – UKIP currently need to outperform UNS by about 7.45%, making it an equivalent to a group 4 seat for the Lib Dems in 97. Give them an 20% chance of winning.
  • Sittingbourne and Sheppey – UKIP currently need to outperform UNS by about 9.55%, making it an equivalent to a group 6 seat for the Lib Dems in 97. Give them an 20% chance of winning.
  • Thanet North – UKIP currently need to outperform UNS by about 10.4%, making it an equivalent to a group 6 seat for the Lib Dems in 97. Give them an 20% chance of winning.
  • Forrest of Dean – UKIP currently need to outperform UNS by about 10.65%, making it an equivalent to a group 6 seat for the Lib Dems in 97. Give them an 20% chance of winning.
  • Sittingbourne and Sheppey – UKIP currently need to outperform UNS by about 9.55%, making it an equivalent to a group 6 seat for the Lib Dems in 97. Give them an 20% chance of winning.
  • Worthing East and Shoreham – UKIP currently need to outperform UNS by about 10.95%, making it an equivalent to a greater swing above UNS than any the Lib Dems managed in 1997. The biggest swing ever recorded in a General election was the 17.6% seen to unseat Portillo in Enfield Southgate in 1997 – even that was only 8.8% above UNS. Give UKIP a 0% chance of winning.

Therefore on average you would expect UKIP to win 5.9 of the seats above on this model. There’s a 40% chance that one of those would be from Labour (Grimsby) and a 40% chance that one of those would be Thurrock which is an expected Labour gain from Tory.

This gives a revised (rounded) Electoral Calculus prediction for the election overall of:

  1. Labour 293
  2. Conservatives 271
  3. SNP 45
  4. Lib Dems 11
  5. DUP 9
  6. UKIP 6
  7. Sinn Fein 5
  8. Plaid 4
  9. SDLP 3
  10. Alliance 1
  11. UCU-NF 1

This isn’t what I expect to happen. This is what I think the best possible result for UKIP could be based upon the current polls, but only if they ran a truly spectacular targeting campaign. What I think it is safe to say is that UKIP will win between 0 and 6 seats.

For various reasons I think it will be closer to zero than six – the biggest being that UKIP’s entire campaigning approach thus far has been “air war” (TV and newspapers) whereas effective targeting requires a “ground war” (door-to-door).

And then some sort of concluding paragraph.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

What’s this?

You are currently reading The six seats UKIP could win at Who rules where.

meta

%d bloggers like this: