Data

The GovConReD dataset, its coding scheme, and figures released as the project progresses.

Open Data

GovConReD — Government–Constituency Relations Dataset

GovConReD is the first open-access, comprehensive, and dynamic dataset on government–constituency relations in conflict settings. It documents both attractive strategies (reforms, concessions, privileges, exemptions) and repressive strategies (curfews, forced displacement, travel bans, and more) that governments direct at the constituencies of non-state armed actors.

The dataset builds on more than 9,000 manually coded events drawn from over 1.2 million news articles (1995–2020), extended with machine learning and computational social science tools to reach its full scope.

Data coming soon
50YEARS COVERED (1980–2030)
120COUNTRIES
350ARMED GROUPS
9,000+EVENTS CODED SO FAR
What's in the Dataset
Coverage & Coding

For every recorded event, GovConReD codes a common set of features so that attractive and repressive measures can be compared on equal terms:

  • Initiator features — which government body or official announced the measure, and at what level of seniority.
  • Type — political, social, or economic.
  • Coverage — the demographic and geographic reach of the measure.
  • Institutionalization — whether the measure was written into law or policy, or remained informal.
  • Implementation pace — how quickly a measure moved from announcement to delivery.
  • Novelty, conditionality, and schedule — whether the measure was new, attached conditions, and had a defined timeline.
Descriptive Findings
Visuals from the Data

Selected figures drawn from the events coded so far. Click any figure to open it at full resolution. These will be updated as the dataset is validated period by period.

Figure 1 — Where and when concessions cluster
Figure 1 — Where and when concessions cluster

Annual coded measures for the twenty most-represented countries, 1991–2022; each country’s total coded measures are shown in parentheses.

View full resolution
Figure 2 — Geographic coverage of the coded events, 1991–2022
Figure 2 — Geographic coverage of the coded events, 1991–2022

Coded accommodative measures by country, the fifteen most-represented countries, and the fifteen most-targeted constituencies — 73 countries, 225 armed actors, and 170 country–constituency pairs.

View full resolution
Figure 3 — Types of accommodative measure
Figure 3 — Types of accommodative measure

Single-domain and multi-domain measures, how the political, sociocultural, and economic mix shifts over time, and how it varies by constituency type (N = 4,012 events).

Open source PDF
Figure 4 — Distinguishing rhetoric from delivered commitment
Figure 4 — Distinguishing rhetoric from delivered commitment

Cheap talk (rhetoric, never implemented) against credible commitment (institutionalised and implemented) over time, with country profiles for the fifteen most-represented cases.

View full resolution
Access
Getting the Data

GovConReD will be released as an open-access dataset alongside its codebook once data collection and validation are complete for each covered period. Journal replication files will be posted here as papers are published.

Researchers interested in early access or collaboration should get in touch via the Contact page.