Can you trust the following figures?
Yes, I hope. Here’s why.
The following figures are based on an online survey of Liberal Democrat members, using a similar methodology and set of security features to the survey I carried out in the last leadership election. That got the result almost spot on and performed better than other attempts to measure the contest.
After weeding out fake entries, attempted duplicate voting, ex-members and the like, 2,209 people completed the survey in full, including the leadership voting intention questions (which asked them to rank all 11 of the party’s MPs other than Tim Farron – a question design intended to ensure people gave thought to their answers rather than just picked a name or two and moved on).
Because of these sorts of protections, I’m confident that the poll isn’t in the voodoo poll category. But it is a self-selecting sample rather than a random or quota sample based on the full membership. To check the robustness of the figures, I’ve therefore used two safeguards in addition to knowing that the overall methodology worked last time.
One, I’ve checked the views of many Lib Dem members in the last few days separately from the survey to have a sanity check to compare against the survey. Those views are in line with the survey results.
Two, I’ve checked to see how different weighting of the survey would affect the results, e.g. if the figures were re-weighted based on the answers to the gender question or to how people voted in the 2015 leadership contest. The findings do not vary greatly within a reasonable range of weightings, with one possible caveat noted below. I’ve not weighted the actual figures in the end as up-to-date figures for some of the weightings are not available (yet) and knowing how accurately members recall how they voted in a previous leadership election is an area full of unknowns. (People do not report accurately how they voted in previous general elections when polled months or years afterwards so it is likely recall of a party leadership contest is similarly off.)
There will most likely be another survey once the candidates are known and the contest underway. If so, that will be an attempt to predict who is going to win. This survey is more a scene-setter before the start of the race and for that later ‘prediction’ survey I’d revisit the question of weightings.