In the Cambridge local elections last week, just 31.3% of the registered voters went to the polls, with nearly 70% staying away. This is a pretty typical level of turnout in our local elections – here is the graph for the last few years:
One factor that clearly affects local election turnout is whether there are any other votes being held on the same day. In 2005 and 2010 there was a substantial increase in turnout because the local elections coincided with the General Election, and a smaller boost in 2011 when the Alternative Vote referendum was held. The European elections in 2004 and 2009 gave a much smaller, but still detectable, increase in turnout. But in years when no other election is held on the same day, turnout is pretty consistently in the range 30-35%.
Another factor affecting turnout is the level of campaigning by the parties. If voters receive election leaflets and are canvassed by party activists, you would expect that they would be more likely to go and vote than if they hear nothing from their local politicians. Since parties tend to concentrate their efforts on the marginal seats, with less activity in the “safe” seats, we should be able to see this factor at work in the results. This next graph shows the turnout in each Cambridge seat this year, plotted against the majority in that seat:
Seats towards the right of the graph, such as Queen Edith’s and West Chesterton, had a higher turnout, whereas in those towards the left, such as Market and Kings Hedges, fewer voters went to the polls. Similarly, seats towards the top of the graph were won with a large majority, whereas those near the bottom had a closer result. So, is there really a correlation here? There are clearly two separate groups, the six “safe” seats at the top, and the eight more marginal ones at the bottom, but you have to squint pretty hard at this graph to see a strong correlation between turnout and majority.
When statisticians don’t get the results they are expecting, they look for what are called “counfounding variables” – other factors at work that interfere with the straightforward relationship that they were hoping to portray. But there actually is a pretty good candidate for a confounding variable here – the fact that turnout among students tends to be significantly lower than the rest of the population. Much of the student vote in Cambridge is concentrated in three wards, Castle, Market and Newnham. Here is the graph again with those three wards shown separately:
Now a more convincing picture emerges, with three distinct groups:
- the higher-majority, lower-turnout wards at the top left
- the lower-majority, higher-turnout wards at the bottom right
- the student wards
Market has the highest concentration of students in the city, and it duly produced the lowest turnout this year, despite being very closely-fought. Student turnout tends to be low for several reasons:
- young people in general have a lower turnout than older people
- many students feel less involved with local government than more permanent residents
- student residences tend to be inaccessible to local party campaigners – colleges do not allow them access to knock on students’ doors, and some colleges refuse to accept election leaflets unless they are individually addressed
- many students were away for the Easter vacation during much of the election campaign
In the non-student wards, however, there is a pretty clear correlation between turnout and majority, with all the “battleground” wards having higher turnout than the “safe” wards. But does this also hold in previous years? Well, up to a point. Here’s the graph for the 2012 local elections:
Again a bit of squinting is needed to see the correlation, but it is still there. Kings Hedges and Cherry Hinton are outliers, but the remaining non-student wards still form a high-majority, lower-turnout group and a low-majority, higher-turnout group. I’m sure political activists will be relieved to know that all that door-knocking and leaflet delivery is having a noticeable effect.