Making Sense of the Opinion Polls

A number of people have contacted me to ask what conclusions they should draw from the seemingly contradictory current opinion polls. We have recently seen all three of the major parties in the lead at different times and undoubtedly there has been a boost in the polls for the Lib Dems following the first televised leaders debate.

The first point to make is that opinion polls are a snapshot of current opinion and not a predictor of a future result. The usual wording of a voting intention question involves the assumption of “a General Election tomorrow”. As we know, there will not be a General Election tomorrow and therefore the question is somewhat unreal.

The second point is that opinion polls usually ask “Which party would you vote for…”, which is a legitimate wording but ignores the fact that many people still vote for or against a local candidate rather than a particular party. Some of those people saying they will vote Green, UKIP or BNP may not have their preferred candidate standing in their constituency and therefore if they vote they will cast their vote for someone else.

Thirdly, polls usually sample just 1-2,000 people and there is then significant manipulation of the raw data to try to make the sample representative of the population as a whole. There is much argument around the post-sample weighting process, as each pollster does it differently and gives different weight to different factors. For example some weight very heavily according to declared likelihood to vote and others less so or not at all. The small size of the sample also makes it more likely that from time to time “rogue” or “freak” results will be created. Most polls have a margin of error of 2-3% but as sample sizes shrink, the margin of error grows.

Fourthly, most polls are national and take no account of regional or local factors. Taking the published numbers and applying a uniform national swing (UNS) to work out the party seat totals is a very suspect way of projecting the election result. For example many marginal seat polls suggest that the Conservatives are doing rather better in their target seats than they are nationally. Thus UNS may suggest a hung Parliament but the actual results could deliver an overall Conservative majority. UNS is an even less reliable predictor in the current circumstances of a three horse race.

Some pollsters have a long track record and we can look at their accuracy before major elections but this is often complicated by the fact that the polls that are most examined are those published immediately before polling day and these are the ones that are often adjusted and weighted differently than their usual monthly or weekly published polls. Personally I have long thought that ICM and YouGov were the most reliable polls but changes to YouGov methodology recently have made me less sure. MORI are very erratic in their findings and others such as Angus Reid and TNS-BMRB are too new to be able to assess. BPIX have a question mark over them as unlike other major polling companies they are not members of the British Polling Council and do not publish their raw data.

If we look back at election campaigns and their polls over the last 10 or so years some consistent patterns can be seen. Almost all opinion polls overstate the Labour share of the vote, usually by only 1-2% but some far more. Many polls slightly understate the Conservatives and some understate the Lib Dems. Polls in Scotland have consistently under-estimated the Conservative position compared to actual election results.

The televised leaders debates have clearly given the Lib Dems a boost but the detailed polling data suggests that much of this support has come from the under 24 age group and those who were not intending to vote. I suspect therefore that much of the Lib Dem “surge” is in fact a polling mirage and it will not be delivered in the ballot boxes on 6th May. Even if it were to be delivered, it would seem to be fairly evenly spread across the country and therefore unlikely to produce many additional seats for the Lib Dems. The Lib Dems do not have much of an organisation on the ground in many of the seats they might hope to win and therefore are not able to get their literature through letter boxes.

I am not a betting man but if I were I would expect that the Lib Dem “surge” will continue to subside in the polls and that Lib Dem support will settle at around 24-25% before polling day. Labour will probably end up at about 28% in the final polls and achieve a bit less on polling day. Conservatives will drift up slowly to about 38-40% in the polls and end up with 39-41% in the ballot box, leading to an overall Conservative majority.

Of course I have no crystal ball (if I did I would be a rich man) and it is possible that the Lib Dems could outpoll Labour, who could experience a complete collapse in support. However, I will stick my neck out and say that the scenario I have outlined above is the most likely in my view.

I would be interested to hear the views of other readers.