Election 2005: What can happen?

First posted 1 May 2005

Is this outcome certain to occur, or can other factors change the result?

As in previous elections (see 2001 errors) the main sources of error in this result are:

• Poll error - the opinion polls may not be accurate
• Tactical voting - voters in marginal constituencies may behave differently
• Model error - local factors and model approximations cause prediction errors

The last of these introduces an error in the majority of about 20, and it is worth remembering that all predictions are subject to that uncertainty. Given that, we can explore the first two sources of error quantitatively to see if Labour is certain to win on 5 May.

Poll error

The pollsters' record in 2001 was not flawless. Although they correctly predicted a strong Labour lead in the polls, they overstated the size of the lead by 6.5%. Although pollsters have learned from the experience, no-one can be sure what the polling error will be this year. Divisions of opinion are also apparent, for instance:

• MORI favour restricting attention to those "absolutely certain to vote", which reduces Labour's lead by about 4%.
• YouGov use internet polling, which they believe is more accurate, and which reduces Labour's lead by about 5%.

Considering a scenario under which the pollsters are again wrong by 6%, the actual result would be:

So Labour's parliamentary majority would be cut to 52, but they would still be comfortably able to form a government.

Tactical voting

In two-way marginal seats, those who naturally support the third-place candidate are crucial. If they choose to support (or desert) one of the other candidates, the result can be affected. In previous elections there was noticeable anti-Conservative tactical voting. In 2005, organised tactical voting has been more diverse with different groups promoting anti-Blair, anti-Conservative and anti-Iraq War tactical voting. However most two-way marginal seats are contests between Conservative and Labour where the two leading candidates are likely to be pro-Blair, pro-Conservative, and/or pro-Iraq War. In such seats, the most likely outcome is for "tactical unwind" where LibDem supporters who voted Labour tactically in 2001 go back to voting Liberal Democrat, even though their candidate is unlikely to win. With some irony, this is actually non-tactical voting as it based on principle and protest rather than shrewd calculation.

The second largest category of two-way marginals is those between the Conservatives and Liberal Democrats, where classical tactical voting by Labour supporters may be expected to operate as usual.

We assume a tactical voting scenario where there is

• Tactical unwind of 2% away from Labour to the LibDems in Conservative-Labour marginals
• Tactical voting of 2% from Labour to the LibDems in Conservative-LibDem marginals

The outcome of this scenario is:

Even with tactical voting against Labour, it still would enjoy a landslide majority of 126.

Double trouble

Pushing the boundaries of the possible, we could ask what would be the effect of both large-scale pollster error and anti-Labour tactical voting. This would be a very worst-case scenario for Labour and an extreme best-case scenario for the Conservatives. The result would be:

Even in this extreme case, Labour still has a slender majority of 16 seats in parliament.

What the people say

Pollsters: Different pollsters use different methods and ask different questions. For the record, here are their most recent polls translated into seats:

Markets: We can also get information from the financial markets as to the likely result. For instance, IG Index makes a market in the Labour majority which is currently trading at 74 - 80 seats (as at 1 May 2005). We can use our seat predictor in reverse to imply party support figures from the majority. These are illustrative figures, and are not market prices for party support.

The market is clearly a little more cautious than the pollsters about Labour's lead, but still predicts a sizable majority of 77 (mid price).

Users: Users of the Electoral Calculus "make your own prediction" feature have been voting anonymously by choosing their own scenarios over the past couple of days. The average results of over 5,000 separate predicted scenarios are:

This is definitely not a scientific survey of opinion, or even a survey of what people think is likely to happen, but only an average of the scenarios that some people think are interesting. Intriguingly, the result is closer to the implied market support levels than to current opinion polls.

Electoral Calculus users have also given their average preferred tactical voting scenario:

Users are predicting a modest amount of anti-Labour and anti-Conservative tactical voting which takes 9 seats away from Labour, leaving them with a majority of just 48.

Summary

At current poll levels, there are no plausible scenarios which give either a Conservative or a Liberal Democrat majority in parliament.