Coup Atlas · Data Insights
Short answers. Visible evidence. A closer look at what the coup record tells us—and what it doesn’t.
1,162 records · descriptive findings · follow-up cutoff 2025-12-31
01 / Repeat Instability
Not reliably in this collection. Another actual attempt or successful coup followed 293 of 449 eligible successful coups within five years in the same country.
That is event-level recurrence, not proof that the first coup caused the next. Recent events without five years of follow-up are excluded.
02 / Military Takeovers
Yes. Of 271 eligible successful military coups, 153 were followed by another actual military coup attempt or success within five years.
Both events must be coded Military coup as their primary type. This narrower definition excludes other kinds of seizure.
03 / Overlapping Actors
No. 11 use Foreign-backed coup as their main type, but 115 have a foreign-backed actor flag. The main category hides overlap.
The counts include conspiracies and follow baseline coding. They do not independently establish which foreign government did what.
04 / Outcomes
472 of 881 actual attempts succeeded in the current collection. That is 53.6%.
The familiar 40.6% uses all 1,162 records, including conspiracies. Both calculations are valid only when the denominator is explicit.
05 / Country Histories
Through 2025, 105 of the 130 countries with recorded actual attempts had more than one.
This is conditional on appearing in our dataset. It does not mean this percentage of all countries experienced coups, or predict future events.
06 / What We Can’t Conclude Yet
Not from the current tactics sample. Only 52 of 1,162 records have detailed tactics coding.
Those selectively researched cases are not a representative global sample. Missing tactics are unknown—not evidence that a tactic was absent.
07 / Missing Is Not Zero
Only 481 records have a reported death figure. A blank cannot be treated as zero.
Even a reported zero means the contributing source recorded none. It does not establish that nobody died, or cover later repression.
New Research · Two-Case Pilot
A first look at how we could track the steps between a seizure of power and an elected presidency. These milestones come from our sourced Guinea/Gabon analysis. No global completion rate is calculated.
Mamadi Doumbouya
The new constitution permitted military leaders to stand and introduced seven-year terms.
The coup leader became the elected president—not a different civilian successor.
Brice Oligui Nguema
The transitional leader appointed officials and entered the election under the new system.
The coup leader won the presidency. The final official result was 94.85%.
Read the Analysis and Complete Source Trail →
Original handover deadlines, every revision, documented no-run pledges, candidacy rules, election observation, actual transfers of executive authority, and dates of international recognition. Unknown values must stay unknown. The two cases above illustrate the design; they cannot establish how often this happens worldwide.
Foreign protection and post-coup democracy scores are additional research tracks. They require sourced assistance records and post-event country-year data that are not yet consistently coded.
These are descriptive calculations from the current Coup Atlas collection, largely the Cline Center baseline. They do not establish causes, predict future coups, or measure the risk for countries without recorded events.
Recurrence uses events through 2025-12-31, excludes conspiracies, and only includes starting events with a complete follow-up window. A later event must be in the same country and on a later date. Overlapping windows count events, not independent countries; recent coups are excluded rather than counted as stable.
“Military only” requires both the starting and subsequent event to have the primary type Military coup. Actor flags can overlap; the primary event type cannot. Source coverage and definitions differ.
Download Calculations and Included Record IDs · Full Methodology · Dataset Documentation
Computed from the current event collection. New questions will be added only when the evidence supports an answer.