What should be the default assumption?
Final Polls 1987/92/97/01/05:
Polls overstating Labour: 29
Polls getting it right: 1
Polls understating Labour: 1
Have we still got to factor in Labour over-statement?
There was a constant reminder from many of the speakers at this week’s conference on general election polling that in only one of the 31 final polls in the five general elections since 1987 has Labour been understated. The exception was ICM in 1997 when even with this it still ended up as top pollster.
For in spite of all the efforts that have been made since the polling disaster of 1992 there’s still a tendency to overstate Labour – something that we’ve covered many before on PB. The firms did better last time with NOP getting it spot on but none of them understated Blair’s party.
Until we get sets of final general election polls where the error is not going all one way then the default assumption must be that this is likely to happen again.
Things might be different on May 6th or whenever it is but it is hard going against nearly a quarter of a century of real polls being tested against real results.