Methodology Review - Summer 2013

Anthony WellsDirector of Political and Social Research
September 02, 2013, 4:45 AM UTC

At YouGov we always keep our methodology under constant review.

Our primary aim is to make sure our results are as accurate as they can possibly be and as the British population gradually changes over time, we need to make sure our methods keep up with them. We carefully track the changing opinions and demographics of our panellists, of official data about the British population and compare how people vote in actual elections with how our panellists told us they would vote.

YouGov are also committed to being as open as possible about the methods we use, and this is an integral aspect of our membership of the British Polling Council. This week we are rolling out a minor update to our methods, the first since October 2011. YouGov's sampling methods, prompting, analysis methods and the variables we weight by all remain exactly the same. However, we have made two small tweaks to our weighting targets.

The first takes account of the latest demographic data on the age and gender of the British population. In the past we have used population statistics from the 2001 census, now demographic data from the 2011 census is fully available we have switched over to the latest figures. This is an extremely minor change in the age profile that should make no noticeable difference to our results.

The second change is a minor update to our party identification weighting. Because YouGov is a panel based company we are able to weight using party identification of panellists recorded in May 2010, weighting it to targets based in May 2010 when they could be correlated against people's general election votes and the actual election result.

Party identification is different from current voting intention (for example, at the last election many people said they identified with the Labour party, but actually didn't vote or voted Liberal Democrat) and changes only slowly over time. However, it does change and we need to reflect that for the proportion of our panel who have joined the panel more recently than May 2010. Tracking party identification at an individual level we have found an increase in the proportion of people identifying with other parties or no party, and a decrease in identification with the established political parties, particularly the Liberal Democrats. We have reflected this in our new weightings.

The changes are only minor, and have very little impact on our topline voting intention figures. On average they make negligible difference to Labour or Conservative support, reduce Liberal Democrat support by around 0.5%, and increase UKIP support by around 0.8%. Using our old weighting targets today's poll would have shown topline figures of CON 33%, LAB 40%, LDEM 8%, UKIP 12%, Others 7%, as opposed to CON 33%, LAB 40%, LD 8%, UKIP 13%, Others 7% using the new weights.

YouGov will continue to keep all aspects of our methods under constant review in the run up to the European elections and the General Election in 2015.