Analysing the police and crime commissioner election data

UK readers will know that elections for police and crime commissioners (PCCs) took place earlier this week, the first time we have had such elections in England and Wales. They will also know that some combination of lack of knowledge, apathy and disagreement with the whole process led to our collective lowest voter turnout for a long time.

In the UK, election results are typically reported in terms of the percentage of votes cast. This makes reasonable sense in cases where the turnout is good; one can make the assumption that the votes cast are representative of the electorate as a whole. Where voter turnout is very low, as was the case this time, I think it makes far less sense. People have stayed away for a reason and I think it is more interesting to look at the mandate that newly elected commissioners have received in terms of the overall electorate.

I took a quick look at the PCC results data that is available from the Guardian data blog and came up with the following chart:

The result makes for pretty uncomfortable viewing, at least for anyone worried about how well democracy served England and Wales in this case.

The data itself comes in the form of a Google Spreadsheet containg a row for each police force and columns for the winning party and the turnout (as votes and as a percentage), followed by various columns for each of the candidates.

The data does not contain a column indicating the size of the electorate in each area, so I added a column called 'Potential vote' and reverse engineered it from the 'Turnout, votes' and 'Turnout, %' columns.

The data also does not contain a single column for each of the winners - instead, this data is spread across multiple columns representing each of the candidates. I created three new colums called 'Winner', 'Voted for winner' and 'Voted for winner as % of turnout' and manually copied the data across from the appropriate cells. (There is probably an automated way of doing this but I couldn't think of it and there's not that many rows to deal with so I did it one at a time).

From there, it is pretty straight-forward to populate columns called 'Voted for other' ('Turnout, votes' less 'Voted for winner') and 'Didn't vote' ('Potential vote' less 'Voted for winner' less 'Voted for other') and to turn these into corresponding percentages of the electorate.

Voila!

As an aside, the data does not contain any information about spoiled ballot papers, of which there were alledgedly a large number in this election. A revised version of the spreadsheet today contains columns for this data but they are currently unpopulated, so I assume that this information is coming. I think this would make a useful addition to this chart.

If you are interested, my data is available here.