Practical applications of the ability to identify communities using mobile network big data

Posted on April 20, 2015  /  0 Comments

I’ve been putting a significant amount of my time in the past three months into Constitutional reforms because an unusual “policy window” or Constitutional Moment opened up as a result of the outcome of the Sri Lankan Presidential election of January 8th. The Common Candidate of the opposition included in his manifesto a series of good governance measures that had been promoted by civil society activists for a long time but with little take up. When he won, these proposals, including rebalancing the relationship between the President and Parliament, electoral reform and the Right to Information, were suddenly the highest priority items of the new government’s agenda. The catch was that everything had to be done within 100 days, because the newly elected President did not have ironclad support from the largest party in Parliament and his manifesto also included a commitment to call a General Election after 100 days, which is around now.

Considering it a citizenship duty, several of us got involved in what we considered the hardest problem, changing the electoral system. Team Leader of the Human Capital Research group Sujata Gamage led the effort doing simulations of different models. The current system had escalated the costs of campaigning and was also not conducive to lawful behavior during campaigns. What the Common Candidate had promised was to replace an electoral system based on proportional representation (with preferential voting to decide the actual MPs) with a mixed member majoritarian scheme that required some of the MPs to be selected from constituencies on a first-past-the-post (FPP) basis.

The last time Sri Lanka had such constituencies was back in 1977. Then there were 6.2 million voters in 160 electorates who elected 168 MPs. Now there are 15 million voters and it is necessary to demarcate around 140 electorates to elect around 160 FPP MPs. It is one thing to increase the number of electorates in a diverse society. But to decrease the number of electorates in the context of a larger base of voters will be a challenge given population movements, ethnic tensions, etc.

As I was pondering this puzzle, the question of whether big data could be of any assistance came to mind, especially the fascinating work we had done on the identification of communities at the national level. The same techniques can be applied to smaller geographical areas, to ascertain what the true communities (defined simplistically by me as more calls occurring within than with the outside, but actually a lot more complicated) are. Would that not be a useful contribution to deciding how the elecorates should be defined?

Here is what I wrote in a guest column in the Financial Times.

The 2011-12 Census results are available; the Survey Department can provide digital maps. The technical options can be worked up much faster than in the past. But ethnic identities and community affiliations must be taken into account. The technical options must be shaped by ideas from the public and their representatives. Public and party input must be tested in relation to previous election results, survey data and even evidence of the existence of communities based on mobile network big data.

Comments are closed.