August 22, 2006

The Geometry of Swedish Politics

I have recently put together a little report visualising the shape of Swedish politics, based on PCA analysis of parliamentary voting records and network analysis of co-sponsored bills. It is in Swedish, but here are some of my highlights:

The votes cluster very strongly - party discipline is strict, and the main difference between members of the same party is how many times they have participated, not how they voted. The shape is like the letter K: a social democrat arm, a left/green arm, an arm for the conservative party and one for the rest. The strongest component corresponds to power: whether the party belong to the ruling coalition or the opposition. The second seems to represent some form of big party-small party difference, perhaps centralisation vs. decentralisation, but it is hard to tell. The subsequent components show interparty differences that likely reflect various coalition patterns. But only the first component is truly strong, overall politics is 1-2 dimensional like British politics. It appears that the basic structure is quite stable over time, not changing shape much but possibly rotating in the ideological dimensions. There is also a tendency of convergence towards the centre.

The votes does not support any real left-right scale. One can make one that looks plausible by adding the first and second component, which produces a ranking of the MPs.

The parliamentary committees have overlapping membership showing a bit of structure, especially when dealing with law, economics or local issues.

The real fun is the bill network, which shows more of the social network of MPs. Here one can see how the big social democratic party remains aloof, with a few people linking to the other parties. The rightwing opposition forms a dense cluster, but also links extensively with the green/left parties that act as a bridge to the social democrats. Their inter-party bonds are almost as strong as the intra-party bonds. A few politicians have significant betweenness centrality.

Hopefully this study will help inspire new tools for monitoring the elected. There are quite a few interesting methods out there, such as network analysis, sites for PCA and voting visualisation, issue maps, argument maps, media bias measurement, campaign contributions and meme flow analysis.

Posted by Anders3 at August 22, 2006 01:30 PM