Labour move up a bit

Following the Populus Poll in the Times yesterday we need to change our General Election prediction which seeks to deal with the over-statement of Labour support that has been a characteristic of the polls for nearly half a century – a factor that is being featured in a BBC radio series with Peter Snow.

The extraordinary impact of the weighting changes at Populus serve to illustrate what a dangerous area polling has now become. Our first prediction used the Populus November figures showing LAB 34: CON 33: LD 22. Yesterday Populus told us that if their new changes had been applied to that poll then the shares would have been LAB 37 (+3) : CON 31 (-2) : LD 21 (-1).

These are massive variations and could make a huge difference on the General Election spread betting markets – so punters beware.

The Politicalbetting formula to deal with what we see as the systemic over-statement of Labour is to take the average of the lowest two poll shares for the party and then deduct 1.5%. For the Tories we take the average of the top two shares and add on 1% and for the LDs we take average the top two figures and add on 0.5%. Our adjustments are in line with what actually happened at the 2001 General Election.

The latest calculation produces the following: LAB 34.5%: CON 34.5%: LD 23%..

Converting this into Commons seats we take the Martin Baxter uniform national swing figure and then switch 12 seats from Labour to the LDs for special targeting and 14 seats from Labour to the Tories to deal with tactical vote unwind. The produces LAB 318: CON 228: LD 70 seats.

If we are wrong about the pollsters over-stating Labour we promise to swallow humble pie on May 6 2005.

The spread markets are unchanged at LAB 345-353: CON 200 – 208: LD 71-75