Why Do Some Civil Wars Last So Much Longer than Others?
Journal of Peace Research, 2004 · paper · Political science
Identifies five factors tied to civil war duration: wars from coups or revolutions, post-Soviet and anti-colonial wars are short; 'sons of the soil' wars and those funded by contraband are long. A game model explains this through government commitment problems.
Checked: PDF from web.archive.org, fetched 5 Oct 2026. Page numbers are those of that PDF.
5 findings
Substance Footnote 13: "median duration for 'SoS' and other wars is similarly dramatic: 5 versus 22 years, with 95% confidence intervals of [4,7] and [8, ∞], respectively". The values are in the reverse order of the groups named; SoS should be 22 [8, ∞] and other wars 5 [4,7].
- p. 10 Footnote 13
median duration for 'SoS' and other wars is similarly dramatic: 5 versus 22 years, with 95% confidence intervals of [4,7] and [8, ∞], respectively.
- p. 8 Figure 2, 'Sons of the soil' wars panel
Sons-of-soil wars
- p. 10 Footnote 13
median duration for 'SoS' and other wars is similarly dramatic: 5 versus 22 years
- p. 10 Body text, sons-of-the-soil section
are 23.2 and 33.7 years, respectively, as compared to 5.8 and 8.5 for the rest
Judge's note: The p.10 body text gives SoS Weibull median/mean 23.2/33.7 against 5.8/8.5 for the rest. In Figure 2 (p.8), the 'Sons of the soil' KM curve falls below 0.5 at about 22 years and the 'Other cases' curve at about 5 years. The open upper bound fits only the heavily censored SoS group. With "respectively", the footnote assigns the short median to SoS.
Second check: another model agreed
Report a problem with this finding
Substance In inequality (12), "1 − c_0^w + E Σ(1 − c_j)δ^j ≥ V_R^W", R's payoff for fighting is V_R^W. Under the game form (Figure 3: Fight → (c_t, 1−c_t) → to war period) and under eqs (3) and (5), that payoff is (1 − c_0^w) + δV_R^W.
- p. 26 Eq. (12)
1 − c_0^w + E Σ (1 − c_j)δ^j ≥ V_R^W
- p. 19 Figure 3, peace periods, Fight branch
(c_t, 1 − c_t) → to war period
- p. 25 Eq. (3)
ε(1 + δV_G^W)
Only the authors can say what is right here.
Judge's note: Figure 3 (p.19) gives the period payoff (c_t, 1−c_t) on the weak-period Fight branch, and play moves to the war period next period. Eq. (3) credits G with 1 + δV_G^W in that branch, and eq. (5) gives R 0 + δV_R^W. The proof of Proposition 1 uses the same timing: R compares δV_R^P with δV_R^W. Eq. (11) likewise uses 1 + δV_G^P, with reversion 'in the next period'. So, consistently, R's deviation payoff after an offer c_0^w is (1−c_0^w) + δV_R^W. The current share cancels, leaving E Σ_{j≥1}(1−c_j)δ^j ≥ δV_R^W. Adding (11) gives c_0^s + δ/(1−δ) ≥ 1 + δ(V_G^P + V_R^W). At c_0^s = 1 this becomes V_G^P + V_R^W ≤ 1/(1−δ), not the δV_G^P + V_R^W printed in (13)–(14) and in Proposition 3. Proposition 3 (p.21) also defines V_G^P and V_R^W as values starting from strong and weak periods, which differs from the appendix definitions. Neither reading yields (12) as printed. The condition in Proposition 3 therefore does not follow from the stated game, and only the author can say which timing is intended.
Second check: another model agreed
Report a problem with this finding
Substance Proposition 4(b) prints "If (2) holds and kG < βδ/(1 − δ)" with a positive threshold. The analogue of 4(a) and of its proof gives kG < −βδ/(1 − δ).
- p. 22 Proposition 4(b)
If (2) holds and kG < βδ/(1 − δ), then
- p. 20 Condition (1)
kG ≥ −βδ/(1 − (1 − ε)δ)
- p. 23 Decolonization discussion
kG was significantly negative
- p. 22 Proposition 4(b)
If (2) holds and kG < βδ/(1 − δ)
Only the authors can say what is right here.
Judge's note: The page 22 image shows no minus sign in 4(b) but shows one in 4(a) (kR < −δ(α+β)/(1−δ)). My reworking of the 'exactly the same sort of argument' (p.27): the best G can get by fighting to return to the equilibrium path is V̂_G^W = kG + α·0 + βδ/(1−δ) + γδV̂_G^W, which is ≥ 0 iff kG ≥ −βδ/(1−δ). Counterexample to the printed version: kG = kR = 0, α=β=γ=1/3, δ=0.9, ε=0.5. Then (1) holds (−0.3/0.55 ≈ −0.545 ≤ 0) and (2) holds (−3 ≤ 0), so Proposition 1 makes the Fight Equilibrium an SPE. Yet kG = 0 < βδ/(1−δ) = 3, so printed 4(b) claims a unique SPE in which G cedes. As printed, the proposition contradicts Proposition 1. The discussion on p.23 ties this case to kG being 'significantly negative'.
Second check: another model agreed
Report a problem with this finding
Substance The sufficiency step in the proof of Proposition 3 reads "this is possible since the left-hand sides of (11) and (12) are linear in c^s and c^w, which together with the reverse of (14) implies the claim". The reverse of (14) makes only the sum (13) feasible, not (11) and (12) together.
- p. 27 Proof of Proposition 3, sufficiency paragraph
this is possible since the left-hand sides of (11) and (12) are linear in c^s and c^w, which together with the reverse of (14) implies the claim
- p. 21 Proposition 3 statement
When this inequality does not hold, there exist subgame perfect equilibria in which government and region share power
- p. 26 Eqs (11)-(14)
Adding these two inequalities implies that
Only the authors can say what is right here.
Judge's note: I reworked the step. With the stationary construction on p.27 (c^s in strong periods, c^w in weak ones), E Σ_{j≥1} c_jδ^j = δ c̄/(1−δ), where c̄ = (1−ε)c^s + εc^w. Then (11) is c^s + δc̄/(1−δ) ≥ 1 + δV_G^P and (12) is 1 − c^w + δ(1−c̄)/(1−δ) ≥ V_R^W. The coefficients do not trade off one-for-one, so meeting the sum does not guarantee both. I checked the reviewer's counterexample (δ=.9, ε=.1, α=β=.05, γ=.9, k_G=0, k_R=.31) using eqs (7) and (9): V_G^P = 5.928 and V_R^W = 4.505, so δV_G^P + V_R^W = 9.841 < 10 = 1/(1−δ) and the reverse of (14) holds. Conditions (1) and (2) hold, and k_G + k_R < 1. (11) becomes 9.1c^s + 0.9c^w ≥ 6.335; its cheapest solution is c^s = .6962, c^w = 0. (12) requires 8.1c^s + 1.9c^w ≤ 5.4945, but at that point the left side is 5.639, so the construction is infeasible. More generally, with the paper's own deviation payoffs, R's on-path value is at most 3.734, so a weak node gives at most 1 + .9·3.734 = 4.361 < 4.505. So no peaceful SPE exists, which contradicts the 'when this inequality does not hold, there exist subgame perfect equilibria' clause on p.21. This defect is separate from the timing issue in C021.
Second check: another model agreed
Report a problem with this finding
Reference or label Reference entry: "Acemoglu, Daron & James A. Robinson, 2001. 'A Theory of Political Transitions', American Political Science Review 91(4): 938–963". For APSR, volume 91 does not fit the year 2001.
- p. 27 References
Acemoglu, Daron & James A. Robinson, 2001. 'A Theory of Political Transitions', American Political Science Review 91(4): 938–963.
- p. 27 References
American Political Science Review 91(4): 938-963
- p. 27
American Political Science Review 91(4): 938-963.
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: that this article appeared in American Economic Review 91(4): 938–963 (2001). The page image (p.27) shows the entry as quoted. The list's own APSR entries are internally inconsistent with it: Bennett & Stam 1996 is vol. 90 and Doyle & Sambanis 2000 is vol. 94, so a 2001 APSR article would be in vol. 95. Either the journal name or the volume is wrong.
Second check: another model agreed
Report a problem with this finding
14 minor (typos and the like; most checked by the judge only)
Reference or label The Nicaraguan revolution is labelled "Nicaragua 1979" on p.5 ("Iran 1978, Nicaragua 1979") but "Nicaragua 1978" on p.7 ("Cuba 1958, Iran 1978, and Nicaragua 1978").
- p. 5 Data on Civil War Duration, para 2
Nicaragua 1979
- p. 7 Coups and Popular Revolutions section
Cuba 1958, Iran 1978, and Nicaragua 1978
- p. 5 Data section, examples of social revolutions
Iran 1978, Nicaragua 1979
- p. 7 Coups and Popular Revolutions
Nicaragua 1978
Only the authors can say what is right here.
Judge's note: Both page images show these labels. Both passages use the case, paired with Iran 1978, as the example of a social or popular revolution, so they appear to refer to the same war. The case labels give different start years. Only the dataset coding can say which year is correct.
Second check: another model agreed
Report a problem with this finding
Wording Abstract: "Five factors are shown to be strongly related to civil war duration" (anti-colonial wars among them), yet Table II Model 1 gives Not contiguous 0.684 (1.62) without a star and the text says it "is not quite significant at the p = .10 level".
- p. 2 Abstract
Five factors are shown to be strongly related to civil war duration.
- p. 11 Multivariate analysis text
this estimate is not quite significant at the p = .10 level
- p. 12 Table II, Model 1
0.684 (1.62)
Judge's note: Checked the abstract on p.2, the text on p.11 (which excepts non-contiguity from the significant estimates; footnote 17 gives p = .11, or .15 for the anti-colonial-only variable) and Table II on p.12, where 0.684 (1.62) has no star. The bivariate gap (median 4.7 vs 7.6, p.9) is real, but "strongly" overstates the multivariate evidence for this one factor. This is a wording point only.
Second check: another model agreed
Report a problem with this finding
Wording Contraband: "durations for these 17 cases are 28.1 and 48.2, respectively, as compared to 6.0 and 8.8 for the rest". The mean/median ratio (1.72) differs from the comparison group's (1.47), unlike every other Weibull-based comparison in the paper.
- p. 11 Valuable Contraband section
durations for these 17 cases are 28.1 and 48.2, respectively, as compared to 6.0 and 8.8 for the rest
Only the authors can say what is right here.
Judge's note: In a Weibull model with a group dummy and a common shape, mean/median is the same for both groups. The other comparisons hold this: coup 3/2.1 vs 12.9/9.0, anti-colonial 7.3/4.7 vs 11.7/7.6, SoS 33.7/23.2 vs 8.5/5.8, and the Table I regions (about 1.39 to 1.44). For contraband the rounding range of 8.8/6.0 (about 1.45 to 1.49) cannot reach 1.72. The paper does not say whether the contraband figures come from separate fits, so the authors should confirm the specification or the values.
Second check: another model agreed
Report a problem with this finding
Wording Figure 3 war-period branches give "(0, δ/(1−δ)) → game ends" for 'Rebels win' and for G 'Not fight', but (k_G, k_R) on the other branches. Read literally, this omits the war-period payoffs that Eq. (4)/(6) credit in every war period (V_G^W = k_G + α0 + βδV_G^P + γδV_G^W).
- p. 19 Figure 3, war periods
(0, δ/(1−δ)) → game ends
- p. 25 Eq. (4)
V_G^W = k_G + α0 + βδV_G^P + γδV_G^W
- p. 26 Proof of Proposition 1
requires that G have V_G^W ≥ 0, which reduces to condition (1)
Only the authors can say what is right here.
Judge's note: My reworking: taking the figure's payoffs literally gives V_G^W = β(k_G+δV_G^P) + γ(k_G+δV_G^W), which differs from Eq. (4) by αk_G. The text on p.20 says the war-period payoffs are k_G and k_R and calls δ/(1−δ) and 0 the 'continuation payoff' after a rebel win. With that reading, Eqs. (4) and (6) and the proof of Proposition 1 (V_G^W ≥ 0 giving condition (1)) are consistent, so the stated results stand. The figure, however, mixes period payoffs with continuation values on its terminal labels.
Second check: another model agreed
Report a problem with this finding
Wording Proposition 3 defines V_G^P as "the government's value for the Fight Equilibrium starting from a period in which it is strong" and V_R^W as R's value "starting from a period in which the government is weak". The appendix defines V_G^P as G's payoff "going into a peace period" and V_R^W as R's payoff "going into a war period", and these are the quantities used in (11)–(14).
- p. 21 Proposition 3 statement
Let V_G^P be the government's value for the Fight Equilibrium starting from a period in which it is strong
- p. 25 Appendix, after Eq. (4)
where V_G^P is G's expected payoff going into a peace period and V_G^W is G's expected payoff going into a war period
Only the authors can say what is right here.
Judge's note: My reworking: in (11), the deviation payoff in a strong period is 1 + δV_G^P, where V_G^P is the value before Nature draws strong or weak (Eq. 3). Under the FE, the value from a strong period is therefore 1 + δV_G^P, not V_G^P. R's value from a weak peace period is 0 + δV_R^W, not V_R^W. Inequality (14) holds with the appendix definitions, so the result stands. The statement's verbal definitions name different quantities. Example values (δ=.9, ε=.1, α=β=.05, γ=.9, k_G=0): V_G^P = 0.19/0.03205 ≈ 5.93.
Second check: another model agreed
Report a problem with this finding
Wording The Conclusion says "imbalanced capabilities tend to reduce prospects for a negotiated settlement while balanced capabilities increase them", but the model section (p.24) says "increases in government strength tend to favor them".
- p. 25 Conclusion
imbal- anced capabilities tend to reduce prospects for a negotiated settlement while balanced capabilities increase them
- p. 24 Military Capabilities subsection
whereas increases in government strength tend to favor them
Only the authors can say what is right here.
Judge's note: On p.24 the text says rebel gains work against settlement and gains in government strength tend to favour it. It also says the effect of the government's chance of decisive victory is 'complicated' and that rebel advantages reduce the odds. The Conclusion (p.25) turns this into a symmetric claim that any imbalance reduces settlement prospects. A strong-government imbalance runs against what p.24 reports. The authors should say what 'imbalanced' is meant to cover.
Second check: another model agreed
Report a problem with this finding
Typo Missing word: "the wars arising out the breakups of the Soviet Union and Yugoslavia" should read "arising out of the breakups".
- p. 4 Introduction, item (3)
the wars arising out the breakups of the Soviet Union
Judge's note: The page 4 image shows "arising out the breakups" in item (3) of the introduction.
Second check: none (judge only)
Report a problem with this finding
Typo Northern Ireland is called "the 31-year conflict" (p.14) but dated "Northern Ireland (1969–99)" in the Table I note. Under the subtraction convention used elsewhere (e.g. Colombia, "37 years to 2000"), that is 30 years.
- p. 14 Per Capita Income paragraph
the 31-year conflict in the richest country in the sample
- p. 9 Table I footnote
Northern Ireland (1969-99)
Judge's note: Both images checked. 31 is an inclusive count of calendar years, a legitimate convention, but elsewhere the paper subtracts start year from end year. Minor consistency point.
Second check: none (judge only)
Report a problem with this finding
Typo Garbled phrase: "to check whether the effects factors are independent". A word is missing, presumably "the effects of these factors".
- p. 11 Multivariate Analysis section
to check whether the effects factors are independent
Judge's note: The page 11 image shows the phrase as quoted.
Second check: none (judge only)
Report a problem with this finding
Typo Duplicated word in footnote 21: "which are coded as here as 'ethnic'" should read "coded here as 'ethnic'".
- p. 15 Footnote 21
which are coded as here as 'ethnic'
Judge's note: The page 15 image shows footnote 21 as quoted.
Second check: none (judge only)
Report a problem with this finding
Typo "when we control for either coups (which associate with aiming at the center) and the sons-of-the-soil dynamics": 'either ... and' should be 'either ... or'.
- p. 15 Ethnic and Secessionist Wars
when we control for either coups (which associate with aiming at the center) and the sons-of-the-soil dynamics
- p. 15
when we control for either coups (which associate with aiming at the center) and the
Judge's note: The page 15 image shows the sentence as quoted.
Second check: none (judge only)
Report a problem with this finding
Typo Misspelling in footnote 31: "R is a group of putchists" and "new potential putchists" should read 'putschists'.
- p. 20 Footnote 31
R is a group of putchists
Judge's note: The page 20 image shows 'putchists' twice in footnote 31.
Second check: none (judge only)
Report a problem with this finding
Typo Proposition 4(a) reads "the game's unique subgame perfect equilbrium"; this should be "equilibrium".
- p. 22 Proposition 4(a)
subgame perfect equilbrium
Judge's note: The crop of p.22 shows the misspelling 'equilbrium' in print, so it is not an extraction artefact.
Second check: none (judge only)
Report a problem with this finding
Typo The Conclusion reads "'sons of the soils' cases"; elsewhere the paper uses 'sons of the soil' (e.g. p.24).
- p. 25 Conclusion
'sons of the soils' cases
Judge's note: The C019 crop of p.25 shows 'sons of the soils'. Page 24 has 'sons of the soil'. This is an inconsistent plural.
Second check: none (judge only)