Do Coup Attempts Beget Coups? A causal analysis with V-Dem

Author: Troy Cheng, Hongzhe Wang

---

Research question

RQ. Do coup attempts in year t increase the probability that the regime ends via a coup in year t+1?

• Treatment (A_t): e_pt_coup_attempts > 0 in year t.
• Outcome (Y_{t+1}):* from v2regendtype in t+1; coded 1 for {military coup, non-military coup, self-coup} and 0 for all other non-missing codes, including “still exists”.

Design

We estimate within-country contrasts by absorbing country and year effects and adding a time-varying confounder (civil war, e_civil_war). Models: (i) FE logit (country & year FE, clustered SEs); (ii) mixed-effects logit (country random intercept + year FE); (iii) (ii) + civil war. We report odds ratios and g-computed average risk differences.

Results

All specifications yield OR ≈ 0.7 for attempt and an average risk difference ≈ −0.02 to −0.024. Unadjusted splits look positive, but the association vanishes (and slightly flips) once we compare within countries over time.

One-sentence takeaway: After aligning time and adjusting for country/year structure (and civil war), coup attempts in year t do not raise the chance that the regime ends via a coup in t+1; the g-computed risk difference is about −2–3 pp.