Based on the results of Granger causality tests, we conclude that terrorist attacks generally lead to government interventions within a couple of months and that government actions did not result in significant retaliatory effects from the terrorist groups. Our Granger causality tests also revealed attack patterns across regions. For example, attacks in Batticaloa increase the likelihood of subsequent attacks in Trincomalee and Vavunia.
We assembled 2,924 terrorist attacks in Sri Lanka from 1975 to 2010 in the final dataset, collecting about 120 variables for each. Out of all attacks, we attributed 1,587 cases to the LTTE; 110 cases were suicide attacks. We also created a separate dataset consisting of 156 government-related events that we judged to be critical to the long term conflict with the LTTE in Sri Lanka. Among these government-related events, 22 were peace talks; 40 were related to government elections; 19 were passing of new legal acts; and 44 were other government actions. The geo-spatial analysis shows that different terrorist groups operated within different regions, outside of their bases. The LTTE began its activities in the northern region while other Sri Lankan terrorist organizations were mostly active in the central and southern regions. The LTTE aggressively and ruthlessly struck out against rival organizations. They first eliminated organizations in the North and then gradually turned their focus to the South. Our analysis of census data shows that regions of Sri Lanka with more terrorist activities also have more urbanized populations, a higher proportion of young adults (between 20-24), larger Sri Lanka Tamil and Hindu populations, and smaller percentages of the population who live under the poverty line. After the elimination of the LTTE, places with more former terrorist attacks also have more criminal incidents. To better understand the connections between government actions and terrorist activities, we did a monthly analysis of both government actions and terrorist attacks. Based on the results of Granger causality tests, we conclude that terrorist attacks generally lead to government's interventions within couple of months and that government actions did not result in significant retaliatory effects from the terrorist groups. Our Granger causality tests also revealed attack patterns across regions. For example, attacks in Batticaloa increase the likelihood of subsequent attacks in Trincomalee and Vavunia.
This study used data collected from multiple sources, including terrorist attacks in the Global Terrorism Database, Sri Lanka census information, election-related violence data collected by the Centre for Monitoring Election Violence, and government-related events data. All data cover the time period from 1975 to 2010. We geocoded all data to the district level to enable spatial analysis of terrorist activities over time and the relationship between terrorism and regional and local characteristics. We used the Geographic Information System (GIS) to map terrorist attacks and their concentration across 25 districts. GIS also helps to show the geographic movements of terrorist activities over time. We used Granger causality tests to examine the longitudinal relationship between terrorist attacks and government actions as well as terrorist trends between districts. Granger causality tests are commonly used to track local relationships between time series data to examine the direction of influence between different series of events. We also applied multivariate methods to model how place characteristics help predict terrorist attacks, especially suicide attacks. Our quantitative analysis was informed by a set of qualitative interviews conducted with government officials, opposition leaders, former members of the LTTE, researchers and public policy experts in Sri Lanka.