tomatosquid's posterous

This political map is based on candidates responses to Democracy Club's questionnaires (http://www.democracyclub.org.uk/twfy/parties/)

To create the map I used Multidimensional Scaling (MDS) as implemented in the Matlab Statistics Toolbox. First I entered the raw agreement scores for each party's candidates for each of the fifteen questions. Only parties with 200 or more members are included in the summary data I used. For example, for the statement "The British government interferes too much with business" the average level of agreement amongst Green Party respondents was 37% whereas for Conservatives it was 87% - so on this issue they are 50% "apart" from one another. The Euclidean distance between parties is the square root of the sum of the squares of the fifteen individual values.

The differences between each pair of parties on the fifteen questions define these distances between their views in a fifteen dimension space - but this is hard to visualise. MDS finds a projection of the data such that the distances are approximately preserved in fewer dimensions - here the first two are used to locate the parties relative to one another. The old left-right dimension is apparent (and luckily it seems to run from left to right on the figure) I suspect the other dimension might capture differences in attitudes to individual freedom of action v. social responsibility. The 2D map is fairly intuitive, but slightly misleading. The "true" distances that go into the calculation are shown below:

Significantly it appears that the Lib Dems are roughly equidistant to Tories and Labour, though this depends on the issues covered in the questionnaires which may not represent the topics that the parties themselves care about most strongly (don't do it Nick).

Disclaimers. this quick analysis was done with the average data, rather than individual candidates' responses. I am not an expert in MDS, and it is Friday night.