 |
Web-scraped time-changing Belief Maps for British MPs
It can be quite tricky to discern what individual UK parliamentary politicians think about a range of issues. Most parliamentary votes are whipped, which means that the parliamentary parties instruct Members of Parliament (MPs) how to vote on specific issues (and woe betide Members that disobey). You would have thought that one might discover the views of individual MPs during the build-up to elections but, again, the party machines exercise control over most policies, apart from those that only have local impact.
One way of discovering MPs' opinions is to ask them. However, in general, this is tricky and acquiring a coherent response to a range of standard questions would be virtually impossible.
However, there is a source of data that covers most politicians and reflects their genuinely held beliefs on a variety of issues. That source is the Early Day Motion database that can be accessed at the Early Day Motion website
The description on the Early Day Motions website reads
"Early Day Motions (EDMs) are formal motions submitted for debate in the House of Commons. However, very few are actually debated. EDMs allow MPs to draw attention to an event or cause. MPs register their support by signing individual motions."
To get some idea of what Early Day Motions are about,will we examine one here. Early Day Motion 111 in the 2012-13 parliamentary session was entitled
"West Ham United's Promotion" which was a motion congratulating the West Ham United football (soccer) team for their promotion to the UK Premier League."
The Primary Sponsor was David Amess a Conservative MP (for Southend West and Basildon since 1983) and the Sponsors (supporters and signers of the motion) were John McDonnell (Labour, for Hayes and Harlington), Alan Meale (Labour, Mansfield), Steve Rotheram (Labour, Liverpool Walton, until 2015), Jim Shannon (Democratic Unionist, Strangford, Northern Ireland).
There are several notable features about this particular motion. First, it is congratulatory and not specifically to do with any of the policy areas usually associated with politicians (e.g. health, defence, education, welfare, etc). Second, the politicians are from different areas of the country (the south-east, London, the north Midlands, the northwest and Northern Ireland) and from different politican parties. Although David Amess supports West Ham it is known that Rotheram and McDonnell are Liverpool supporters. In summary, five MPs signed this motion and this Motion (observation) provides information that links these five MPs.
The power of Early Day Motions is that they are not whipped and are a reflection of the true thoughts and beliefs of MPs. They are also spontaneous and each such Motion gives a snapshot of which MPs communicate with each other, from all sides of the House. A single Motion does not, on its own, provide a lot of information. However, there can be thousands of Motions per year (e.g. in 2015-16 there were 1457) and it is the combined information contained within these Motions that enables us to build up a comprehensive picture of MPs' opinion.
One approach, that we adopted in the mid-2000s (published in 2008), is to analyze this set via multidimensional scaling. For this, we needed to calculate the distance from every MP to every other MP. This can be done easily from the lists of the Motions, who signs them and a concept known as Jaccard's distance. Some MPs sign no motions, usually Government Ministers, Whips and Parliamentary Private Secretaries, but some regular MPs too. These MPs are discarded from our analyses (However, in time local analyses, where the status of MPs changes they can sometimes be included, for example, before an MP becomes a Minister).
From Jaccard's distance we can use multidimensional scaling to reconstruct a configuration of all the MPs who signed a Motion in a given session. We can present an approximate solution in two dimensions which gives a map of all MPs in space such as in the following picture.
Each point in the scaling solution corresponds to a MP. The scaling solution often works better in higher dimensions, but usually a low number is acceptable and two dimensions often works fine most of the time. The figure above actually contains a three-dimensional projection, which can be seen as you play the video. As can be seen from the figure, the various groupings and alliances amongst MPs are easily spotted via the colours (scheme explained below). Note, the coloring DOES NOT inform the location of the MPs on the map, the colours were only put on afterwards.
Unfortunately, the EDM website does not provide access to their underlying complete data tables. However, it is possible to obtain information from all of the Early Day Motions by scraping the web-site. Computer code written in R, perl, awk and sed enabled extraction of all of the vital details from the web page of each Early Day Motion. This enabled lists detailing which MP signed which Early Day Motion. Additionally, for other analyses, we also extracted information as to the nature of each EDM, whether it was primarily about health, education, defence or a range of other topics. Two PhD students were employed to independently read through each EDM and classify each one. Where there was disagreement, a referee separately read the Motion and made a final decision. Of course, some Motions cover more than one topic. For example, Motions that discuss the financing of the health service overlap both the health and the economy policy areas, and so Motions received a primary and a secondary classification. Further details of the full analysis and what was achieved can be found in the accompanying paper.
Given a particular configuration that plots all the MPs in a two-dimensional space (as in the video above) we can additionally assign each one a colour. The colours we choose correspond to the popular colours often used by the parties themselves. For example, blue for Conservatives, red for Labour, yellow for Liberal Democrat, green for the Green party, and so on. Where the MPs form like-party clusters we can form shapes (or convex hulls) that reflect the party groupings.
These shapes can be seen in the video below, which shows how they evolve over a period of years. The approximate date is shown in the video at the bottom left. Within the video can be seen a bubble with the letter W. This corresponds to the position of the MP Shaun Woodward, who was initially a Conservative MP that joined Labour in 1999. Arguably, one can see Woodward initially positioned in the blue shape and then, over time, moves to the red shape.
Further analyses might reveal associations between MPs that were not suspected. However, it is important to realize that, in these analyses, such associations could exist due to MPs coincidentally sharing similar views and not necessarily because there was extensive communication between them, or anything else untoward. From such configurations it is possible to draw approximate boundaries between parties and identify MPs on boundaries, or groups who form the core of parties. Such analyses might provide useful information to others who wish to know where MPs stand on certain issues. For example, to whips who wish to identify potential waverers or journalists who seek information on which MPs express views outsite of their party's mainstream.
At the moment, such an analysis is not publically or routinely carried out. However, if an analysis was carried out along these lines, and MPs were identified, then the MPs themselves might well change their behaviour or stop signing Motions. This is both the strength of data science and, ultimately, can be its downfall if the system you are studying changes because you have studied it!
|
|