How pollsters ask questions is an eternally controversial issue. For voting intention, that often focuses upon how the answer options are presented. The approach that YouGov has always taken is to prompt for the traditional main parties, but only prompt for other parties if people select "other". A similar approach is taken by most other polling companies.
This may seem unfair to some people (and has often been a source of complaint from supporters of smaller parties), but is based on what actually gets elections right. In the past, prompting for smaller parties has tended to overstate their support when compared to actual elections, and the two-stage approach to prompting has produced more accurate results.
However, there comes a point when a small party becomes a big party, when they should be included in the main prompt. This can be a difficult decision, and one that YouGov takes time and care to call correctly, thoroughly testing any changes before they go ahead. This was the approach we took before the 2015 election when UKIP were breaking through. We regularly tested the effect of prompting on UKIP support, and, once it seemed it was no longer giving them an artifical boost, we started including UKIP in the main prompt alongside Labour, the Conservatives and Liberal Democrats.
We are at that same point with the Brexit Party now - testing the impact prompting has and what their support would be in a write-in question without any prompting for any of the parties. If we are confident that including them in the main prompt will produce more accurate results than grouping them with "others", we will update our question prompting.
However, at the same time we also need to make sure we do not overstate support for the Brexit Party. YouGov correctly predicted the outcome of last week's EU Parliament elections, including the level of support for the Conservatives, Labour, Liberal Democrats and Greens - but like many pollsters we overstated support for the Brexit Party, putting them at 37% compared to the 31.6% they actually achieved in Great Britain. Over the next few weeks, we will also be looking at the possible causes of that overstatement, and whether there was something to do with turnout, undecided voters or our weighting or sampling scheme that led to us having too many Brexit voters in our final poll.
As ever, our main consideration will be what is most likely to produce the most accurate results, not what would help one party or another, and we will continue to keep our methods under constant review.
