The dangers of making historical polling comparisons

Like many polling nerds whenever I’ve wanted to check some past polling information I’ve gone to Mori’s “all firms” historical data tables. They are easily accessible and apparently comprehensive – except that Mori does not appear to recognise YouGov.

But there’s a huge danger of using this source which I think is affecting current perceptions of Labour’s poll leads before they got into Government – many of the numbers in the Mori table for other pollsters do not square with what the other firms actually reported at the time.

For yesterday’s discussion on the ICM survey I checked back to find out what the pollster had been reporting in October 1993 – at exactly the same stage in the 1992-1997 Parliament. On the ICM website the Guardian poll in that month showed a 3% Labour lead. Yet looking at the Mori “all firms” table the same poll was listed as showing a Labour lead of 15%.

That two online sources should record totally different figures from the same survey is absolutely staggering and shows what a minefield historical polling comparisons can be.

As far as I can make out many of the old polls of other firms in the Mori table are based on the unadjusted figures and this might explain the discrepancy.

So if you want to check out old polling data your first port of call should be the polling firm itself. For ICM polls go to ICM.

Mike Smithson