Synopsis

Social media forecasts achieved highly accurate results in the 2014 Indonesian presidential elections. Will predictive analytics play a greater role in future electoral outcomes in Indonesia?

Commentary

IN THE midst of dispute over the quick count results for the 2014 Indonesian presidential elections, which were corroborated by the final results, a separate phenomenon was emerging: the forecast of election results through analysis of social media sentiment.

In this respect Indonesia is following the trend in other countries, where the relationship between online sentiment and electoral outcomes had been studied. Social media platforms such as Twitter and Facebook had been tracked in order to find out voter sentiment in elections such as the 2014 general elections in India, 2012 presidential elections in the United States and 2010 federal elections in Germany. It had been found that social media is a platform for political discussion and reflection of voter preferences similar to traditional election polls.

Dawn of social media forecast

In the case of Indonesia, the Jakarta gubernatorial elections in 2012 was the earliest instance where social media predictive analytics were used to gauge voter sentiment. Most opinion polls had predicted that the pair of Fauzi Bowo (Foke) and Nachrowi Ramli (Nara) would win the election, only to be proven wrong by the victory of Joko Widodo (Jokowi) and Basuki Tjahaja Purnama (Ahok).

To recap, the Foke-Nara pair won only 34% of the vote in the first round of the elections, compared to Jokowi-Ahok’s 42.6%. In the second round of elections, Jokowi-Ahok managed to secure 53.8% of the votes against Foke-Nara’s 46.2%. The event raised doubts among some quarters about the methodology employed by traditional opinion polls.

In tandem with the work of the opinion polls several independent social media consultancy firms analysed the results of Internet chatter and used algorithms to predict the winner of the elections. For instance, Provetic, a Jakarta-based data consultancy, saw online conversations as reliable predictors of local-level elections. Provetic analyses from April to July 2012 noted that during the 2012 gubernatorial election, the percentage of people talking about the different candidates on social media correlated very closely with the actual results.

Twitter buzz around the Jakarta gubernatorial election showed a strong correlation with the first round of voting: Jokowi-Ahok (40%), Foke-Nara (33%). A separate social media consultancy company, Politicawave predicted a Jokowi-Ahok victory with 40.6% of the votes in round one and 53.9% in the second round. These results corroborated very closely the actual election results.

First nation-wide social media forecast

The latest presidential election was the first time that the social media predictive analytics were being employed on a nation-wide scale. Social media platforms had reported vast amount of online interaction related to the elections held on 9 July 2014. For example, Facebook reported more than 200 million posts, comments, shares and likes related to the presidential race. By voting day, Twitter had recorded more than 850,000 conversations about Prabowo and more than one million tweets related to Jokowi.

The micro-blogging site had earlier noted that there had been close to 95 million tweets about the Indonesian elections since the beginning of the year. Given the large amount of election-related online interactions, the role of social media forecasts in the recent presidential election deserves a closer examination.

In the Indonesian presidential election, the final results announced by the General Elections Commission on 22 July 2014 (Jokowi-Kalla pair winning 53.2% of the votes and Prabowo-Hatta pair with 46.9% of the votes) corroborated very closely social media predictions. For example, Politicawave had predicted a Jokowi-Kalla victory of 53.8% on 8 July 2014. The company also believed that Prabowo-Hatta would achieve 46.2% of the votes.

Social media forecast results have also been very close to quick counts results conducted by poll companies that have passed the audit conducted by the Indonesian Association for Public Opinion (Persepi). The results released by polling firms such as Centre for Strategic and International Studies (CSIS) (Prabowo 48.1%, Jokowi 51.9%), Indikator Politik Indonesia (Prabowo 47%, Jokowi 53%), Saiful Mujani Research and Consulting (Prabowo 47.1% and Jokowi 53%), Lingkaran Survei Indonesia (Prabowo 46.4%, Jokowi 53.4%) showed a likely victory for Jokowi as well.

All these showed that social media forecast were generally in agreement with the official results and polling firms that have been cleared by Persepi. The success called for closer examination into the role of social media forecast in future elections in Indonesia.

Forecast in future Indonesian elections

With the success of predictive analytics and higher digital literacy, social media forecast will likely play a greater role in future elections in Indonesia. Several questions may be posed with regards to this development.

Firstly, will social media forecast continue to be an alternative way of electoral prediction? Is there a possibility of it becoming the mainstream way of predicting election results? Or, more likely, will it be a complementary form of analysis in future elections in Indonesia?

In any case, it is almost certain that there will be more players in the game of social media forecasts in upcoming elections. In view of this scenario, will social media forecasts undergo similar calls for rigour in terms of  their research methods and public accountability in future elections?

About the Author

Jennifer Yang Hui is an Associate Research Fellow at the Centre of Excellence for National Security (CENS), a constituent unit of the S. Rajaratnam School of International Studies (RSIS), Nanyang Technological University.