The Limits of Studying Networks with Event Data: Evidence from the ICEWS Dataset

Journal of Global Security Studies, Vol. 3, No. 4, October 2018, pp. 498-511

36 Pages Posted: 2 May 2019

See all articles by Kai Jäger

Kai Jäger

King’s College London - Department of Political Economy

Date Written: 2018

Abstract

Machine-coded event datasets have become popular in conflict research. I argue that systematic media biases render news-based event data unsuitable for studying anti-government networks of insurgents and political parties. Insurgent networks are too secretive to be captured by media reports, whereas alliances among regular political parties are too constant to be considered newsworthy. I analyze the data accuracy of Metternich et al.’s (2013) network study of insurgents and political parties in Thailand, which is based on the most comprehensive event dataset currently available – the ICEWS project. Based on simple evaluation criteria, I show that most of the network data entries are incorrect, leading to a depiction of the networks that is unrelated to real-world cleavages in Thailand. While my hand-coded event dataset captures relatively more network-relevant information than ICEWS, the comparison confirms that journalists specifically underreport cooperative events among insurgents and parties. In addition, the ICEWS project provides unreliable counts of conflictual events in Thailand. Using alternative conflict measurements from the Deep South Watch dataset, and a dummy variable based on established periods of unrest, I show that violent activities in Thailand’s Deep South declined during periods of conflict between pro- and anti-Thaksin groups. Conflicts were unrelated to network fragmentation, contradicting Metternich et al.’s primary finding.

Keywords: Event Data, Media Bias, Machine Coding, Network Analysis, ICEWS, Intra-State Conflict, Thailand

Suggested Citation

Jäger, Kai, The Limits of Studying Networks with Event Data: Evidence from the ICEWS Dataset (2018). Journal of Global Security Studies, Vol. 3, No. 4, October 2018, pp. 498-511, Available at SSRN: https://ssrn.com/abstract=3367909 or http://dx.doi.org/10.2139/ssrn.3367909

Kai Jäger (Contact Author)

King’s College London - Department of Political Economy ( email )

Strand Building
London
United Kingdom

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
49
Abstract Views
519
PlumX Metrics