Party membership data has much to teach us about how well – or how badly – we campaign

By Mark Wallace
Follow Mark on Twitter.

The party membership figures available to us are only a partial picture, but analysing them against recent election results does enable us to learn some things about how we use our grassroots base.

As we reported this week, a ConservativeHome examination of the hundreds of accounting units which file records with the Electoral Commission reveals 139 units which declared a specific number of members in 2012. In order to study them against election results, we must strip out five are amalgamated areas and regions rather than specific constituency associations.

That leaves us with 134 associations.

The first thing to note is that these are not a representative sample of the party across the country. 101 are Conservative-held seats, for the obvious reason that more detailed accounts tend to be submitted by associations who have the greater infrastructure that comes with having an MP.

Only 20 are seats which the Conservatives won from other parties at the last election, suggesting that marginals are also under-represented. Indeed, the average Conservative vote share in the sample seats in 2010 was 45.5 per cent, much higher than the 36.1 per cent we received nationally.

Of the 33 which are not Tory-held, 18 are Lib Dem, 13 are Labour, 1 is Green and 1 returns an SNP MP. In most of these seats, the Conservatives are in second place.

So the sample is disproportionately drawn from those seats which have a strong Conservative presence. This is why we did not just extrapolate from these figures to a national estimate for party membership - the likelihood is that those associations not on the list will tend to be smaller than those who file full reports.

What we can do, however, is explore the correlation between Tory votes and the scale of local association membership. Bearing in mind that the comparison is imperfect, given that these are 2012 figures and the most recent General Election was in 2010, here is how it looks on the surface:

Obviously there's an inherent distortion created by using raw numbers of votes - bigger seats are likely to have more party members, so we can get a better picture by looking at the share of the vote instead:

That irons out some of the distortion caused by seat size. There's a clear correlation - clearer than you might expect given the effects of individual candidates, local issues and so on.

Still, correlation is not the same as causation. We can't know if having more members generates more votes, or if it's simply the case that a certain proportion of people are joiners, meaning membership could be a symptom driven by how Tory an area naturally is.

Looking at the number of votes received divided by the number of members in the local association reveals that there are very few seats where there are less than 30 voters per local party member:

The seats on the left are mostly those where membership is extremely low due to being a safe Labour or Lib Dem seat, but a core vote is still turning out, generating vast ratios of voters to members. Interestingly, that trend peters out quite swiftly, with seats creeping closer to that 30:1 ratio the more members they have. It could show that each member is turning out 30 people, or just that one person in every 30 is prone to joining organisations - in all likelihood it is a combination of the two factors.

While we can't prove a causal link for the impact of each member, what we can do is compare the performance of different constituency associations. The following chart shows the seats we held at the last election in blue, the seats held by other parties in grey and the new Tory gains in orange.

As you can see, there were several remarkable successes where Conservative candidates gained seats despite having fewer than 200 members in the constituency - in one, there were fewer than 100.

Either they are particularly active members, their candidates were particularly charismatic or the local campaign used the approach Benedict McAleenan proposed of signing up activist supporters without requiring they join the party. One way or another, the data allows us to pick out seats whose tactics should be closely studied in order to learn from their success.

Similarly, a number of those grey, enemy dots are seats where Conservative membership is relatively high, but the share of the vote received was surprisingly low. These should ring alarm bells - why were such comparatively plentiful resources so ineffective on the ground?

Of course, we would know more if we had wider data in terms of the number of associations and deeper data in terms of the change in membership over time. There are campaign lessons to be learned here, seats to be won and defeats to stave off. Publishing the full party membership data may be seen as a PR risk, but it should also be seen as an opportunity to improve the way we fight elections.