Measuring Terrorism

Abstract

This study investigates the extent of reporting and nature of biases in open-source (OS) terrorism databases. We compare OS accounts with official accounts on terrorism events in Turkey (1996-2012). Results indicate (a) substantial systematic discrepancy between OS and official accounts, which we attribute primarily to underreporting in OS accounts; (b) the discrepancy is not random—incident characteristics (victim/target, offender, and incident types, temporal and spatial factors) and rational factors (especially newsworthiness) matter; and (c) severity is the strongest predictor of the probability of OS coverage.