Deconstructing the 2015 Sri Lanka presidential election through data analysis

Friday, 13 February 2015 00:00 -     - {{hitsCtrl.values.hits}}

All-island estimate     By Dr. Ranjiva Munasinghe and Ruwanthi de Silva On 8 January the people of Sri Lanka elected Maithripala Sirisena as the new President of the Democratic Socialist Republic of Sri Lanka by a majority of 51.28%. Former President Mahinda Rajapaksa came in second, tallying 47.58% of the votes. Post-election, a lot of excellent pieces have been written analysing the results, and are widely accessible from the online newspapers and other websites. In this short piece, we hope to examine the election results from different perspectives.   Estimating the Sinhalese vote for each candidate Before and after the election there has been a lot of back and forth on Mahinda Rajapaksa securing the Sinhala vote and Maithripala Sirisena winning via the minority segment. We sought to quantify these claims through a numerical data analysis to estimate the percentage of Sinhala and minority votes for each candidate. According to our findings the overall Sinhala vote appears to be split 58-41 in favour of Mahinda Rajapaksa, meanwhile Maithripala Sirisena obtained an 85-13 margin of the overall minority vote. The methodology to estimate the most likely split assumes the valid votes casted are distributed as per the 2012 census data in conjunction with statistical techniques. The all-island estimate is obtained by aggregating the district estimates. In some districts such as Jaffna both candidates “received” 0% of the Sinhala vote – this is due to the fact that Jaffna has a very small Sinhala population.   The social media President The hotly-contested presidential election of 2015 brought forth many firsts, one of which being that it was the first presidential election in Sri Lanka to be so widely discussed in social media. The social media coverage and engagement of all events that unfolded during the presidential election was tremendous and has undoubtedly influenced the end result. It was perhaps the only forum available for open and unrestricted debate, as evidenced by the posts made. The following analysis of the final two weeks of the presidential election provides a better understanding of the social media effect.       Share of voice Share of voice measures the amount of posts that were generated for each candidate on twitter and Facebook over the two-week period 24 December-7 January. The numerous crossovers, the declaration of support by the TNA and the Muslim congress generated much of the conversation towards the end of December, while posts on the election violence against the common opposition and President Sirisena’s first speech as he declared himself the common candidate were shared multiple times during the first week of January.Much of the conversation generated around the former President Mahinda Rajapaksa reached a crescendo on 29 December as a result of the attempted ‘Bollywood push’. The involvement of Salman Khan and Jacqueline Fernandez was widely tweeted and re-tweeted on Twitter. Mentions of corruption and violence were prominent in relation to the chatter on the former President. The overall word cloud (wordle) for the election further highlights the clear dominance by the common candidate in the social media sphere. The importance of the word is represented by font size and colour.     Sentiment The sentiment around the conversations and discussions for both candidates are also shown. While the sentiment for the former President was relatively evenly spread among positive and negative, the common Opposition candidate had a distinctively higher percentage of positive conversations. Based on the sentiment, the Social Net Promoter Score (SNPS) for each candidate was calculated to determine the number of promoters each candidate had.The negative SNPS score denotes the percentage of detractors. The top three key reasons for the negative SNPS for Mahinda Rajapaksa were due to:1.Widespread disgust over the Bollywood push – this was perceived to be an indication that the former President had lost touch with reality. 2.The dichotomy of the presidential campaign between the north and the south was widely perceived negatively. 3.Election violence and the excessive use of State media were vilified in social media. The reasons contributing to the positive promoter score for the common candidate are as below: 1.Commitment to good governance and change 2.TNA, Muslim congress support and other crossovers 3.Election violence towards the common candidate The analysis clearly indicates that the current President gained immense coverage on social media and was the winner in the social media space.       Gradient map view of results Post-election the typical electoral district map displayed a two-tone colour scheme of the results – unfortunately this did not capture the differential in voter percentage. We believe a more accurate depiction can be shown using a gradient (or heat) map. Figure 1: Gradient Map of MY3-MR Percentage Vote Differential 2015 Sri Lanka Presidential Election The darker the shade of orange, the larger the margin for Maithripala Sirisena, whereas the darker the shade of blue, the larger the margin for Mahinda Rajapaksa. Districts such as Gampaha, Puttalam and Badulla which are grey indicate an even differential – in this case there is an orange green tinge indicating a small margin for Maithripala Sirisena.   The 100 days The 100 day plan of the current Government is being closely watched and monitored by many. Since 9 January, this topic has generated 873 posts on Facebook and Twitter. Given that it has been only three weeks since Maithripala Sirisena took office, there is still high optimism that the 100-day plan will be implemented, with around 86% of the sentiment being positive. It is interesting to note that the sentiment was largely positive during the first 1.5 weeks, but conversation on doubts on the viability of constitutional reform in 100 days have surfaced around the mid-point of this period resulting in a negative sentiment of 14%. The reinstatement of Chief Justice Shirani Bandaranayake also contributed to the positive sentiment as many perceived this to be the first positive step in the restitution of the independence of the Judiciary. The most commonly-used words in relation to the 100 day plan is given in a word cloud.(Argyle X is a boutique data science firm providing cross-industry solutions in analytics, business intelligence, data visualisation and social media monitoring. Now in its second year of operation, ArgyleX works with large corporations to provide analytics services using existing data structures and industry information to develop predictive models for business leaders who need to make data-driven decisions on sales, marketing and operational improvements to their businesses. For more details, please visit or e-mail [email protected]) (Dr. Ranjiva Munasinghe is Managing Director of Argyle X. Ruwanthi de Silva is Social Media Analyst of Argyle X.)