The impact of mobile phones on opinion polling

Opinion pollsters using the telephone for political polls in the UK and the USA face some very different challenges.

In the UK, phoning a random selection of people and getting them to agree to take part in a survey has, for more than a decade, regularly produced samples that are too heavily weighted towards Labour sympathisers, necessitating all sorts of – at times very controversial – adjustments to be made to the raw figures in order to make the poll results more accurately representative of the population as a whole.

There isn’t an equivalent pro-Democrat bias in US phone polling, though instead in the US there is the challenge of adjusting for likely turnout levels. Although turnout in the UK is far short of 100%, and turnout weightings can have significant effects on poll results here, the problem has proved far harder to tackle in the US than in the UK.

As a result, there is relatively little overlap in polling sampling discussions between the two countries, unlike most other areas of political campaigning where there is a ready flow of ideas between the US and the UK.

There is though one challenge that you would expect phone pollsters to be increasingly facing in both countries, which is the rise of mobile phone only households and the subsequent problems that either you can ignore people with mobiles, possibly producing a biased sample, or you can try to call them, but run into problems (and costs) of much lower response rates as someone on a mobile is more likely to be doing something or be somewhere that makes the person not want to take part in a poll even though they answered the phone.

This is an issue that has been researched more extensively in the US than the UK and, until very recently, it has looked to be acceptable to save on time and money and ignore mobile phone numbers when doing phone polling because the answers from mobile phone only households were not significantly different than from the rest of the population.

However, this year’s US Presidential election looks to be breaking this pattern, with phone pollsters who include mobiles regularly finding a higher level of support for Barack Obama than those who ignore mobiles. The difference in support between these two approaches is large enough to matter in a close election: Obama’s lead is around four percentage points higher when mobile phone only households are included.

This is an issue that may well yet add yet another twist to the arguments over whose phone poll methodology is best in the UK.