Data Sources

Individual Source Event Data

The current xSub release features event data on 195 countries from 22 sources, including

• American Bar Association / Darfur

• Armed Conflict Location and Event Data Project (ACLED)

• Beissinger / Protest

• Beissinger / Riot

• Beissinger / Ukraine

• Empirical Studies of Conflict Project (ESOC) / Afghanistan / WITS

• Empirical Studies of Conflict Project (ESOC) / Iraq / SIGACT

• Empirical Studies of Conflict Project (ESOC) / Iraq / WITS

• Empirical Studies of Conflict Project (ESOC) / Mexico / Drug-Related Murders

• Empirical Studies of Conflict Project (ESOC) / Mexico / Homicides

• Empirical Studies of Conflict Project (ESOC) / Pakistan / BFRS

• Empirical Studies of Conflict Project (ESOC) / Pakistan / WITS

• Lankina Russian protest event dataset

• National Violence Monitoring System (NVMS) Indonesia

• Northern Ireland Research Initiative (NIRI)

• Political Instability Task Force / Worldwide Atrocities Dataset

• Social Conflict Analysis Database (SCAD)

• UCDP / Georeferenced Event Dataset

• Zhukov / North Caucasus

• Zhukov / Chechnya

• Zhukov / Libya

• Zhukov / Ukraine

Multiple Source Event Data

In addition to the data collections above, xSub features several pooled, multiple-source datasets, integrated and disambiguated with the MELTT software package. These datasets combine events from our 21 collections, helping to offset under-reporting in each source with information from another. For details about MELTT’s underlying methodology, please see

• Donnay, Karsten, Eric Dunford, Erin C. McGrath, David Backer and David E. Cunningham. “Integrating Conflict Event Data,” Journal of Conflict Resolution 63, no. 5 (May 2019): 1337--1364. https://doi.org/10.1177/0022002718777050

We currently provide 8 sets of integrated data, organized by the spatial, temporal and event taxonomy filters used to remove duplicate events:

• Spatial filter (which events are ‘co-located’):

  • 1 km: contemporaneous events located within a 1 km radius of each other count as a single event

  • 5 km: contemporaneous events located within a 5 km radius of each other count as a single event

• Temporal filter (which events are ‘contemporaneous’):

  • 1 day: geographically co-located events reported within 1 day of each other count as a single event

  • 2 days: geographically co-located events reported within 2 days of each other count as a single event

• Dyad type:

  • A: directed (includes only events where initiator and target are both known)

  • B: undirected (all events, including those with some ambiguity over which actor was initiator vs. target)

  NOTE: ACLED and UCDP-GED contain only undirected dyad information. Most other sources include directed dyad information. Accordingly, type-A dyads exclude ACLED and UCDP-GED from the integrated event counts.

In all, we offer eight different sets of integrated, multiple-source data:

• 1 km, 1 day, dyad type A (MELTT1km1dA)

• 1 km, 1 day, dyad type B (MELTT1km1dB)

• 5 km, 1 day, dyad type A (MELTT5km1dA)

• 5 km, 1 day, dyad type B (MELTT5km1dB)

• 1 km, 2 days, dyad type A (MELTT1km2dA)

• 1 km, 2 days, dyad type B (MELTT1km2dB)

• 5 km, 2 days, dyad type A (MELTT5km2dA)

• 5 km, 2 days, dyad type B (MELTT5km2dB)

Of these, 1km1d has the most stringent standard for duplicates (resulting in more unique events), and 5km2d has the loosest (resulting fewer unique events).