The Curse of Geography: How Governments Preempt Secession AttemptsRob Williams

University of North Carolina at Chapel Hill

Research Question

Puzzle:•Why is there no relationship between thegovernability of territory and rebel goal?

Answer:•Governments anticipate secession attempts inmost suitable areas and invest to prevent them

Motivation•Collected original dataset of goals for all ethnicterritorial rebel groups 1990-2013•Used geospatial data on population to measuregovernability of territory•No relationship between governability and goal

Theory•Different territories are more or less governable asa function of the resources they contain•Territories far from the reach of the state aremore governable•Population is on one of the most importantresources for governance•Marginalized ethnic groups that controlgovernable territory will be more likely to fightfor secession•Geography is relatively static and governmentsuse their knowledge of human geography topreempt conflict in areas most suited to secession•This preemption can be measured globally usingnighttime light emissions as they correlate withlocal measures of state capacity[1]

Hypothesis: The effect of territorial governabilityon nightlights should be positive and increasing indistance from the capital

Data•GeoEPR & EPR•NOAA Nightlights

•GPWv4 Population•V-Dem & WDI

Illustration

Results

●

●

●

●

●

●

●

●

●

●

●

●

●

●

Model 1 Model 2

−1 0 1 2 −1 0 1 2

(Constant)

Polyarchy

GDP

Lost Autonomy

Excluded

Area

Population Total x Capital Distance

Population Total

Capital Distance

0.60

0.65

0.70

0.75

0.80

−3 −2 −1 0 1 2

Capital Distance

Mar

gina

l Effe

ct o

f Pop

ulat

ion

Tota

l

−3

−1

0

1

2

−4 −2 0 1 3

Capital Distance

Pop

ulat

ion

−4

−2

0

2

Pre

dict

ed N

ight

light

s V

alue

ModelY ∼ N (α + γ + Xβ, σ2)β ∼ N (µβ, σβ)α ∼ N (µα, σα)γ ∼ N (µγ, σγ)

µα, µα, µβ ∼ N (0, 5)σα, σγ, σβ, σ ∼ Cauchy(0, 2.5)

Out of Sample Validation

0.0

0.5

1.0

1.5

Model 2Model 1Population DistanceMSE of models on a 1/3 held out validation set

Discussion•Governments seem to invest more in areas whichare both populous and remote•Nightlights correlated with both militaryinfrastructure and public goods•Weakly positive effect of lost autonomy suggestsstates devote attention to secession prone areas[2]

Next Steps•Medium N study of internal colonialism e.g.Aceh, Northeastern Sri Lanka, Ireland•Historical study of forced relocations e.g. Sovietdeportations, Armenian genocide

Email: jrw@live.unc.eduWeb: jrw.web.unc.edu

References[1] Ore Koren and Anoop K. Sarbahi.

State Capacity, Insurgency, and Civil War: A Disaggregated Analysis.International Studies Quarterly, Forthcoming.

[2] David S. Siroky and John Cuffe.Lost Autonomy, Nationalism and Separatism.Comparative Political Studies, 48(1):3–34, January 2015.