How to turn votes into tax free cash

So, YouGov ignore voting likelihood: Populus weight by it, Ipsos MORI and ICM filter by it. Clearly, it makes a difference, but there is no general agreement as to how much.

If front end data is a problem, then turning that into real results is a nightmare. In 1992, polls predicted a Labour win: in the end, John Major snuck back in with a majority of 21. To put that result in context, the Tory majority would have been reversed had the result in the 11 most marginal seats gone the other way: the final result was determined by just 1241 votes out of a total electorate of over 43 million.

Vote distribution also plays a major part in outcome, with Lib Dems historically suffering for having support spread across the nation. In 1983, 27.6% of the vote left Labour with 209 seats, whilst 25.4% of the vote gave the Liberal/SDP Alliance just 23 seats.

Whilst that may feel quirky, the ultimate prize for quirkiness brought about by the first-past-the-post voting system goes to Canada, where, in 1993, the governing Progressive Conservative Party saw its parliamentary share fall from a solid 169 seats out of 295 to just 2 - on 16% of the vote.

Our own Tories skirted disaster on a similar scale in 1997 – although in the UK system, both the main parties have a solid core of support that means it would take an electoral earthquake to take them much below the 200 seat mark. Despite their current poll woe, political better and author of a book on how to make money betting on politics, Mike Smithson, estimates that the worst Labour outcome is likely to lie in the 175 to 200 seat range.

There is still much potential mischief in the assumption, made by many newspapers and pundits, of "uniform swing". That is the assumption that if the polls show a national transfer of votes from Labour to Tories of 10%, it is possible to work out the most likely outcome by applying that 10% to every seat in the country. Apart from ludicrous results (such as this leading to a negative Labour vote in some seats), it is well established that swings vary by region.

At the last election, the London result was quite different from the rest of the country. More practically, Mike Smithson makes the point that uniform swing is quite wrong when applied to Scotland, where politics is now evolving very differently from the rest of the UK. It also doesn’t work when looking at what is likely to happen to the Lib Dems, whose share of the vote almost always increases as an election draws closer, and who are very good at holding on to individual seats against the national trend.

His preferred approach is to apply the uniform swing – and then to look at the seats predicted to change hands on a case by case basis. Incumbency counts, with new candidates often doing less well than their immediate predecessor. Local issues can play a part, which explains spectacular Labour losses in Wyre Forest and Blaenau Gwent at the last election. Scandal, too can affect the outcome, as Neil Hamilton found out to his cost in 1997, when he was ousted by BBC journalist Martin Bell.

For a reasonable analysis of the factors in play at election time, two sites are useful: UK Polling Report, and Electoral Calculus. If you want to test your own expertise as a political pundit, both sites offer election predictors, that enable you to work out likely seat outcomes based on a particular poll result. Electoral calculus includes the ability to vary your estimates by region and to look at results on a seat by seat basis.

However, for a daily updated analysis of what is happening both in UK politics and overseas, the premier source of informed expert information has to be politicalbetting.com, run full-time by the above-mentioned Mike Smithson.

His site will keep you abreast of the polls and, more importantly for anyone looking to lay a bet on electoral outcomes, it analyses how the odds are stacking up and identifies which bets look good and which don’t.