Opinion polls and what to do with them

Opinion polls remain a key method of predicting what might happen in an election. We are grateful to newspapers and websites for continually paying for polling companies to ask roughly 1000 people how they attend to vote so that we have some idea of what might happen.

However, newspapers (understandably so) hope and expect a headline from their investment in polling. Hence, we see headlines such as ‘Support for Labour shrinks as faith in recovery grows, ICM poll finds’ from The Guardian, or ‘Tories at lowest ebb for 8 years‘ in The Sunday Times. Whilst these analyses do tell the truth of the poll’s findings, they tell only a part-story of how political party support stands at any one time. Polls will often provide contradictory evidence of what public opinion actually is, and as a consequence, people feel right in tweeting exaggerated findings from a single poll. However, we can only have so much confidence in one poll that it is close to representing its population. Polling companies sample their ‘voters’ randomly, in order for it to be representative. Often, this can generate a higher or lower sample of support for one party than a general election would do. Roughly, newspaper opinion polls can often be deemed as being within 3 points higher or lower than the figure quoted. Even this rough guideline is often optimistic, as Peter Kellner pointed out last summer. 

How then can we interpret the polling data that comes through? Thankfully, people much more capable than myself have taken on the task of employing statistical technique to bring together different polling over time and present it as a more sophisticated and detailed picture of public opinion. The two I would recommend most strongly are the UK Polling Report (currently showing the averages as 32% for the Conservatives, 37% for Labour, 10% for the Liberal Democrats and 14% for UKIP) and the Polling Observatory (showing 31% for the Conservatives, 38% for Labour, 8% for the Liberal Democrats and 12% for UKIP). They show slightly different results as they have different methodologies.

Newspapers often don’t write about the likes of UK Polling Report and the Polling Observatory. As I write this, the Independent on Sunday apparently has the following*…

Why wouldn’t you be interested in what this poll shows? It might well show that the Conservatives have a lead over Labour, or it might go the other way and show that Labour has a double digit lead. It might put the Liberal Democrats over 15% for the first time in a while, or it might show UKIP to reach 20%. Whatever it is, it sound like it is probably a one off result that we should not take too much notice of until we see a few more like it.

Blogs like UK Polling Report and Polling Observatory helps us understand the trends in polling more than one off results. These are often ignored as they provide less astounding results. In truth, not much has happened in the last year or so. Labour’s lead has fallen by roughly a couple of percentage points (this happened around the time of Margaret Thatcher’s death, but that was not necessarily the reason), but not to the extent that Ed Miliband and his colleagues should be worrying too much. The Conservatives have largely hovered around the low 30s, whilst the Liberal Democrats struggle to continually achieve double figure poll results. UKIP perhaps go up and down the most, but they are also most susceptible to more or less extensive media coverage.

In short, polls are incredibly helpful in allowing us to think about who might form the next government. However, they are a guide, and a limited one at that. Individual polling companies have different methods of interpreting the results they get, and when this is combined with a random sample of voters that can sometimes be a bit out, one poll can be skewed. Only by examining trends can we get a real idea of what is actually happening.

*Update 19-01-2014 – 11.40: The Indy on Sunday wasn’t revealing anything too marmalade dropping. The headline is as follows: ‘UKIP: the most popular party in Britain‘. The poll actually shows little change in voting intention, but then asks respondents: ‘Please indicate whether you have a favourable or unfavourable view of each of the following political leaders and parties.’ To that question, UKIP is in the lead. Whilst that is interesting, it seems a question geared more towards generating a juicy headline than getting to the heart of the effect of UKIP on British politics. For that, I’d recommend following Rob Ford and Matthew Goodwin on Twitter.

One thought on “Opinion polls and what to do with them”

Leave a Reply