New research investigates systematic measurement errors in widely used media-based conflict event data during elections. The study shows these conventional datasets often underreport election violence incidents across the entire voting period, especially in rural areas and non-traditional hotspots.

Comparing conventional sources to domestic election monitor surveys reveals significant discrepancies between what is reported by news outlets versus on-the-ground observers.

Key Findings:

• Geocoding errors are common with media-based data

• Underreporting occurs throughout the election cycle ✘

• Rural areas and non-traditional hotspots have higher error rates

• Measurement inaccuracies can create spurious relationships or alter effect magnitudes • Existing studies may be compromised by these issues

Why It Matters:

This research highlights potential pitfalls in using media reports as the primary source of violence data during elections, suggesting that current findings might be shaped more by measurement flaws than actual patterns. The authors recommend specific areas for future investigation to address these limitations.