The pit-falls of using the pollster’s historical all-firms table

With the first major shift in party popularity since 1992 a lot of people on the site are looking back at the data from the final years of the John Major government to find comparisons with what we are seeing at the moment.

Many are asking how the Tories are doing now compared with how Labour was performing in the polls in the first year or so after the 1992 election. Unfortunately there are dangers using the main polling source from that period – the “all polls table” from Mori.

I discovered this a few weeks ago after I got into a furious argument with two people on the site when I made the assertion that in spite of “Black Wednesday” in September 1992 the Tories were still ahead at the start of 1993. This was based on the ICM Guardian poll for January 1993 which had the parties on CON 39: LAB 37: LD 18.

Those who were questioning my data were using, apparently, the Mori historical all polls table which showed shares of CON 38: LAB 42: LD 15 from the same survey.

We had the same with yesterday’s thread when a number of contributors on the site were using the Mori table as their source to compare with this month’s ICM poll in the Guardian. According to the pollster’s website the comparative poll in 1993 had CON 32: LAB 38: LD 24. The Mori site figures for the same poll were CON 29: LAB 43: LD 23.

Well I think that I’ve worked out why there’s the discrepancy. In a footnote at the bottom of the MORI table it states “ICM/NOP/Harris/ORB unadjusted figures, not headline figures”. This refers to the ICM September 1994 poll but I assume it reflects what they had been doing the year before. At that stage other pollsters had embraced a form of past vote weighting to ensure that unbalanced samples were corrected – something that Mori still do not use.

With all historic polling comparisons it’s vital to compare like with like and the best source of consistent historical data is the ICM polling archive. There’s an excellent table here of all ICM polls for the Guardian since 1984.

From this it’s clear that the Tories are not doing quite as well as Labour were in 1993 – but they are not far behind.

The Mori all-firms’ list, which on many occasions appears to have unadjusted figures for ICM until just before the 1997 General Election, gives a wrong impression and should be avoided.

Bromley By-election betting has opened.

Mike Smithson