Lib Dem Voice

Does the alphabet matter when it comes to Liberal Democrat internal elections?

There is consistent evidence that in public elections people with names higher up the alphabet, and hence higher up the ballot paper, do slightly better than people lower down. It is not a major effect, though in a marginal seat a small difference can mean you win rather than lose, and seems to be strongest where ballot papers are more complicated, e.g. if all three seats in a council ward are up for election.

So it is no great surprise that after the Liberal Democrat federal committee and interim peers panel election results were declared on Saturday, several people made comment about how the results looked to benefit people higher up the ballot paper. The large number of candidates means these were just the sort of contest where you might expect an effect to come into play, and with the final few places on committees or panels often decided by small margins, there might be an important prompt here for reform.

But what does the evidence say?

The following graphs show for each committee the first preferences gained by a candidate (calculated as a percentage of the total number of votes cast in that contest) versus the candidate’s position on the ballot paper. Also on each graph are two trend lines – a simple linear trend line and also a second-order polynomial trend line (as it is possible that, for example, candidates near the top or bottom of a ballot paper do better than those in the middle):

Federal Executive

Federal Policy Committee

Federal Conference Committee

Interim Peers Panel

The evidence then looks pretty weak. However, in all the cases there are some major outliers which distort the results. I of course mean that in the nicest possible way as I’m one of those distorting outliers, so here is the data again, but this time with the top two candidates in each contest removed:

Federal Executive

Federal Policy Committee

Federal Conference Committee

Interim Peers Panel

The change in the graphs thanks to the removal of just two data points from each is a healthy reminder about the vagaries of trying to read too much into data and the conclusion still is that there is no strong, consistent effect here. Just possibly there is a slight edge for those higher up the ballot paper (three of the four straight line trends are down and the fourth is flat) but the evidence is weak (look at the conflicting trends of the polynomial lines).

Which, as someone who finished second on first preferences behind a candidate many places ahead of me on the ballot paper is a bit of a shame :-)

If you would like the data to analyse it further beyond my basic first cut, or would be willing to type up data from other years to extend the data set, let me know.

Comment moderation policy