Clegg performs well in the leaders debate and the next morning the LibDems start to climb in the polls; by Saturday according to BPIX they lead both the Tories and Labour. I just don't believe it. Look, nine-odd million out of forty-five million voters saw the TV debate - fewer than a quarter of electors. To produce a swing of this magnitude, every single viewer must have been converted to Europe and Libdemitude overnight and persuaded their neighbours and people on the train to vote for Clegg too.
I think the reality is rather different. I think there's a bias in the poll samples; it's the same people who watch TV political debates and QT who participate in opinion polls. The other four out of five of us claim we're just about to have dinner. Or don't have a land-line. So the polls are just telling us what a sample of those nine-odd millions think, not what the other 36 million think.
On the train home from Southampton yesterday I was chatting with a young man who had his own very well developed idea of how the TV debates should have been designed. They should have started with more contestants, he said, with the losers being voted off each week until the final week was a run-off between the two finalists. And they should have used telephone voting.
And perhaps he was right. Perhaps a game-show format is as good a way of picking a Prime Minister as an election.