When can you make valid historical poll comparisons?

Almost everyday, it seems, someone on a thread or a journalist will seek to make polling comparisons with the 1990s and each time it happens myself and others will try to explain which this is not as easy as it appears.

For there has been such a revolution in the UK polling industry since the aftermath of the 1992 general election that you have to be very careful when you point to things that happened in the past as an explanation of what is happening at the moment.

Just look at the opening polls of the 1997 general election campaign and compare them with the result. This was the era of real Labour over-stating and, as the numbers show, something was seriously wrong.

Since then Harris and Gallup have virtually left UK voting intention polling. MORI has changed drastically and only ICM from that era is still polling in the same way. All of this is why I constantly argue that the only valid comparisons can be back to the point that the pollster introduced its current methodology.

So to make it clear when historical polling changes have happened I have produced the above simple guide. It should be noted that the process of change has continued and two of the five main firms that will be carrying out surveys in the run up to the next UK general election have made big alterations in the past two years.

Valid comparisons with ICM are possible from the mid-1990s; with YouGov and Populus since their inception but for ComRes and Ipsos-MORI you cannot go back beyond March 2007 and June 2008 respectively.

The big problem in the old days was that it appeared impossible to find samples that were politically balanced. One of the elements that has since been identified is that for one reason or another many more public sector workers seem to take part in telephone polls compared to their numbers in the population at large. This group tends to have a very different political outlook from the rest of the electorate.

The best quick guide to historical ICM polls is this table showing Guardian surveys going right back until 1984.

This page has now been included in our list of fixed links – see the panel on the right.