Worcester's Weblog

MORI chairman Sir Robert Worcester analyses the latest opinion poll data.

On day two of the election, no new polls have been published at the time of writing, except for a bit of data released by British Election Study director Paul Whitley exclusively on last night's BBC Newsnight.

These raised some eyebrows, because while the BES team claim a much better design for the survey this time (a probability sample hyped in last week's Economist) compared with the flawed findings in 2001 (the furthest out of line with the final result of any poll).

Newsnight compared the 2005 'certain to vote' figure of 53 per cent in 2001 with 48 per cent in 2005, drawing the conclusion that the vote is much softer this time.

As a matter of fact, the two figures are within the margin of error comparing the findings of the two polls. If Newsnight wanted evidence of this softness, better data was provided by most of the polls published in the national papers reported in yesterday's blog and which could be read in the coverage, most of which was excellent, in yesterday's national press and especially in knowledgeable coverage of the four polls in yesterday's London Evening Standard.

The other bit of data released was Labour's share of the vote. At the start of the 2001 campaign in the BES it was 56 per cent (53 per cent was recorded at the time), while Tuesday night's figure was 44 per cent, which is well out of line with every other poll published over the past month or so. In 2001, the final BES reported finding was 55 per cent. Three other polls published over the same time as the BES in 2001 reported Labour shares of 47 per cent (both NOP and ICM) and 43 per cent (MORI). Labour's final tally was 42 per cent. No other poll was over 47 per cent the week before election day 2001.

The 'same' data, slightly different figures, is also published in the House Magazine this week, showing that 47 per cent of the BES respondents have 'decided how to vote, and among those who have decided, Labour leads with 43 per cent'.

It's good practice for reports of polls to report fieldwork dates and sample size at a minimum, and a requirement for those organisations belonging to the British Polling Council.

In 2001, there was a transparency to the BES findings that seems to be missing in this campaign, at least so far, as while a Google search finds the 2001 data with ease, there is more difficulty in obtaining any sight of much but the occasional release such as last night on Newsnight and on April 1 in the Daily Mail, both raising more questions than giving answers.

Perhaps I speak too soon, and we'll soon be seeing the open access to the BES that publicly funded survey data that we should have.