Thursday was a long and interesting day, as I was standing
for re-election to my local Borough Council (Gedling, on the outskirts of
When I am wearing my politician’s hat I use several techniques which are common to market researchers, but with a different slant. During the campaign period (approximately the two months before polling day), we spend much of our time canvassing. The popular misconception that that we canvass people (either face-to-face or via telephone) in order to try and change people’s minds. Although one occasionally meets a voter with a specific query which might affect their voting, that is not the main role of canvassing.
Canvassing is, basically, data collection. What we want to know is which way are you likely to vote, and to an extent whether you are likely to vote (more on this point later). This information can be used in several ways. Firstly, in the early stages of the campaign we use this information to try and project the result for each ward. Most parties have scarce resources and they want to concentrate their effort on the marginal wards, the ones where they are neither certain to win nor certain to lose. These are the areas which generate the highest return on effort (our ROI).
The next stage of using the information is to target communications depending on people’s views. For example, in my Ward the result 4 years ago (i.e. at the last local election) was that one LibDem (me) and one Labour were elected, with the Conservatives coming a poor third. Therefore, we had two key messages to communicate. To people minded to vote LibDem we wanted to encourage them to actually vote (only about 30%-40% of people vote in local elections), and as part of this we wanted to remind them how close the result was last time, and to remind them of all the good work we have done for the area. For those inclined to support the Conservatives the core message was that is they vote Conservative (a right of centre party), they are, in effect, making it easier for Labour to win (Labour have traditionally been seen as left of centre, and to the left of the LibDems).
The other key element in the process is identifying who is
likely to vote. In the
During election day the electioneering process moves into over-drive. Outside each of the polling stations (in my ward there are two polling stations) volunteers ask voters to tell them their polling numbers (about 99% are happy to do this). This numbers are conveyed back to the committee room where the list of everybody who has indicated they will vote LibDem is listed. The list of people who has voted is checked against the list of supporters, and those who have voted are crossed of the list of supporters. At this point other volunteers visit the people on the list of supporters to remind them that today is polling day, and to offer lifts to the polling station.
The final burst of statistical activity occurs at the count. Before the votes are counted they are first verified. Verified means counting the votes inside each ballot box to ensure that they match the number issued. Once the ballots are verified all the papers for a ward are mixed together and counted. However, we politicians are interests in our strength within each ballot box area, not just our strength overall. So, volunteers watch the process and count samples of papers as they are unfolded to make estimates of the support within each box.
Reviewing the process, I am struck the similarities and dissimilarities between the way market researchers and politicians use statistical processes. In the early part of the campaign the politician uses sampling to estimate support. However, the sampling is not normally based on random probability sampling. We modify the population to be those with a history of voting and those new to the register. Our concentration tends to be on switch analysis rather than simple sample predictions, i.e. we use previous records to assess whether things have moved towards us or away from us since the last election.
During the campaign we attempt to identify the entire population of likely supporters, and the entire population of key other groups (e.g. Tory supporters, Greens etc.). During the early part of the account we assess what we think is happening, but with an immediacy that most marketers never face, within minutes or at most a few hours of the statistical process we have either been re-elected or deposed.
And how did the election turn out? I am glad to report that I was re-elected, with an increased share of the vote, with an increased turnout. Did I win because of the polling day operation? I don’t think so. If a result is close then the polling day operation is vital. If the result is reasonably close then the campaign is the key factor. But in most cases it is the work that happens during the previous four years that makes the difference – at least that is my theory, and I have won all 11 local government elections I have stood in since 1983.