Method Behind The Mapness
In a separate post I shared what I’d learned in pulling together some maps and scatter plots on UK petition data, but I wanted to write a separate note on the method used. It seemed like too much detail to go in there, but that there’s still value in publishing the approach to the problem, which I’ve laid out below.
All the petitions data came from the UK Petitions website
Signatures per 1,000 heads per year is based upon the total number of signatures recorded as from members of each constituency in the period 20 July 2015 to 7 February 2017. That as a proportion of the population, divided by a factor of 1.556 gives the annualised figure. So, for example, Aberavon had a population of 66,615 and a total of 25,682 signatures of all petitions, good for a rate of 248 signatures per year per 1,000 constituents.
Expected contribution is calculated as the proportion of signatures from UK constituencies that were made up by a single constituency. Looking again at Aberavon, those 25,682 signatures make up 0.0954% of the 26,932,858 overall across all 650 constituencies, and that establishes our baseline expected signature number.
Expected signatures is therefore derived by taking the expected contribution and multiplying through by the total for that petition. For simplicity’s sake, at the point where the anti-Trump petition hit exactly 1.8m signatures, we’d expect 0.0954% of those to come from Aberavon, which is 1,717 when rounded to the nearest whole person.
Population figures include all constituents because there is no restriction on ones eligibility to vote – minors can sign petitions equally as their parents. That means that our signatures per head metric has to use a total population denominator, and doesn’t risk being skewed by areas that have a smaller voting electorate anyway for reasons of political or social disengagement.
The range of dates for which petition data is available on the Government’s petition website only goes back as far as 20 July 2015, giving us 568 days. Totalling the number of signatures over that time is all well and good, but the period feels quite arbitrary and hard to picture, hence the conversion to an annual figure (568/365 = 1.556).
The propensity map doesn’t incorporate Northern Ireland, restricted due to the data source only covering Great British constituencies and not the full UK.