The Liberal Democrat eruption is not finished yet

The Guardian article "Liberal Democrat eruption is not finished yet" (April 25th, 2010) offers a corrupted version of the median voter theorem, suggesting that but one in twenty-five marginal voters decides the ruling party in Britain.

Elections are determined by remarkably few voters. These are those voters who choose to vote (in the past two elections only six in 10); who have little party loyalty (about one in five); and who live in marginal seats. As few as one voter in 25 decides who rules the land.

This might appear to be true, and it might appear that the marginal voter is one twenty-fifth of the eligible voting population. However, it's not clear that this is actually the case. For any given election, there may be 75% of the population supporting the candidate that wins. Two-thirds of that 75% have an incentive to free-ride on the other 25% plus one that can win the election for them without having to vote, something that might be considered to be costly.

In such a situation, the defection of every single individual who was going to vote for the winning party would mean nothing--there is a vast resovoir of previously supra-marginal, now-marginal agents willing to vote for the winner. Voting is endogenous to chances of winning, and the Guardian appears to have missed this.