With open parliamentary data at their disposal, parliamentary monitoring organizations, data journalists and scholars can gain significant insight into how new legislation is shaped and how parliaments work as social groups. Apart from multidimensional scaling of voting patterns, an often used technique of reducing and visualizing complex parliamentary information is social network analysis. In this post, I would like to introduce it using an example of sponsoring bills in the lower chamber of the Czech Parliament.

As the name suggests, network analysis explores networks of mutual connections (or ‘edges’) between entities (or ‘nodes’). Nodes can represent groups of people (e.g. parliamentary party groups), individuals (e.g. MPs), objects (e.g. bill proposals) or even abstract entities, such as words or ideas. Edges usually represent some form of interaction, connection, cooperation, proximity or membership.

In case of this analysis, a node represents an individual MP in the 2010-2013 Chamber of Deputies of the Czech Parliament and an edge between two MPs represents that they sponsored a bill together. The more often two MPs sponsored bills together, the stronger the connection between them. Or, to use the terminology of network analysis, the more ‘weight’ the respective edge has.

In the Czech lower chamber, each MP can sponsor a bill individually. It is common, however, that groups of MPs (often large ones) sponsor bills together. Network analysis helps to explore what determines configuration of these groups.

At KohoVolit.eu, a Czech and Slovak parliamentary monitoring organization, we update comprehensive data on the Czech legislative process are available in downloadable files daily. Therefore, it took me less than two hours to prepare data and conduct the analysis – a task that would be impractically difficult or even impossible in many parliaments.

My colleague, Michal Skop, and I use great freeware Gephi to get a beautiful picture of the network and a d3.js JavaScript library to create an interactive visualization of the network. Only MPs that sponsored at least one bill and were in the chamber for the entire term of 2010-2013 are included in the analysis. This amounts to 185 out of 217 MPs.

(Click the images below to view full size versions.)

It is apparent that MPs form three distinct clusters in the network. This corresponds to a partisan split in the chamber. A red cluster includes MPs of the Communist Party, an orange cluster includes mostly MPs of the Social Democratic party and the blue cluster includes MPs of three center-right government parties. The three government parties form recognizable sub-clusters in the blue cluster.

Clustering obviously corresponds to the fact that an MP usually sponsors bills with their colleagues from the same party. It is notable that the cluster of social democratic MPs is much tighter that the cluster of government parties. This is due to the fact that all or almost all MPs in this party usually sponsor bills together. Their mutual connection is much stronger, which in turn pushes them together in the network. This to a lesser degree applies to communist MPs. The cluster of MPs from government parties is considerably looser.

This is due to the fact that not only MPs but also the cabinet can sponsor bills in the Czech legislative system. The cabinet is in fact by far the most active and the most successful sponsor. Therefore, MPs of the government parties do not usually have to sponsor their own bills. When they do introduce a bill, it is usually sponsored by a small number of MPs, hence weaker mutual connections in the network.

Size of nodes corresponds to the measure of centrality. It captures not only how often an MP sponsors bills but also how often he or she sponsors bills together with other MPs. MPs with a high level of centrality are pushed to the center of the network. These MPs are mostly chairmen of parliamentary party groups or committees that are very active in daily workings of the chamber.

They are also usually the ones that sponsor non-controversial bills introduced by multiple parties, often cutting through the partisan split. Therefore, these MPs form stronger connections to MPs in other clusters. MPs with a low level of centrality are pushed to the edges of the network. These MPs are mostly members of the cabinet and therefore less active as bill sponsors.

This is, I think, a good illustration of how one can extract substantive knowledge of how a parliament actually works using a simple statistical methods and a freeware tool in less than one afternoon. The only necessary condition is to have the parliamentary data open.