At Applied Development Economics (ADE) seminar hosted by the Lahore School of Economics, Dr Ali Hasanain, Associate Professor and Head of the Department of Economics at the Lahore University of Management Sciences, presented findings from his work with colleagues from LUMS and University of California on political violence in Pakistan.
The authors used machine learning techniques to extend and update an existing dataset on political violence (i.e. the BFRS data). The process involved real-time archiving of around 1,624,201 articles from Pakistani English newspapers between 1999 till date. Using a topic modelling algorithm (based on “Injuries and Casualties”, “Violence in the World”, and “Terrorism”), the authors narrowed down articles to 200,000. The authors conducted further analytics at topic level and entity level to create academic data.
Using this automated data, the authors provided a comprehensive geographical distribution of political violence incidents and associated causalities in the country between 2002 to 2019. The greatest prevalence of political violence has been in Sind and central and northern Punjab. On the positive side, there has been a substantial decline in the incidence of political violence incidents, injuries and killings since 2014.
These downward trends are observed across all categories of killings including assassinations, attacks on law enforcement personnel, military/paramilitary personnel as well as terrorism. Furthermore, using sophisticated data validation techniques, the authors also demonstrated that their automated data extraction techniques do a fairly accurate job at identifying political violence in Pakistan.
