Election polling split poses dilemma

Morris says the relative anonymity of the internet could be encouraging respondents to be more open about their preferences than if questioned by telephone canvassers.

The mainstream pollsters back their claim to be accurately recording the national mood by noting they all show broadly the same results despite operating their surveys in three quite different ways.

They have refined their methods over recent years to avoid a repeat of the disaster of the 1992 election when all polling firms forecast a Labour victory, only for Conservative Prime Minister John Major to win a surprise fourth term for his party.

"When things did go wrong in 1992, one of the causes of that was that everybody was polling in exactly the same way, and all making the same mistakes," said Roger Mortimore, head of political research at Ipsos-MORI.

"Unless we are all now making different mistakes and still coming to the same answer, we ought now to be rather more confident that we are getting it right."