Persistence of

Armed Conflict:
1816 to 2008
James Fearon and David Laitin
Stanford University
For presentation at University of Maryland
October 25, 2013
 
 Roadmap
1.  Motivation
2.  Theory

3.  Data
4.  Results (is there persistence?)

5.  Preliminary Tests

6.  Conclusion
MOTIVATION  1:  CIVIL  WAR  AS  A  PREEMINENT  PUBLIC  
HEALTH  PROBLEM  
¢  Nearly  20    million  killed  as  a  result  of  civil  war  violence  
since  1945;  
¢  Though  a  decline  in  civil  war  violence  in  the  1990s,  civil  
wars  rage  today  in  Syria,  Iraq,  Libya,  Afghanistan,  
Pakistan,  South  Sudan,  DemocraSc  Republic  of  Congo,  
and  Nigeria,  with  devastaSng  consequences  in  human  
suffering.  
MOTIVATION  2:  CIVIL  WAR  AS  A  CONSTRAINT  ON  
GROWTH  
Region   Average   p  value   Average   Average  total  
Annual   years  of  war   decline  in  
Growth  hit  *   (if  some  war)   GDP  (%)  
(%)  
Asia   0.45   0.24   27.4   -­‐13  
E.  Europe   -­‐7.64   0.04   6.0   38  
LA/CA   -­‐3.67   0.00   12.4   37  
NA/ME   -­‐1.79   0.15   18.7   29  
SSA   -­‐2.92   0.02   18.2   42  

• “Hit”  is  the  coefficient  from  a  regression  of  WB  growth  rates  on  a  dummy  
for  civil  war,  country  and  region-­‐year  FE’s  (standard  errors  are  clustered  by  
country).  
MOTIVATION  3:  STUNNING  RESULTS  ON  LONG-­‐TERM  
HISTORICAL  IMPACTS  

¢  Examples:  
—  PoliScal  InsStuSons  (AJR,  “extracSve”  vs.  “inclusive”)  
—  Medieval  insStuSons  and  efficient  democracy  in  Italy  (Putnam)  
—  PlantaSon  economies  in  the  new  world  (Engerman  and  Sokoloff)  
—  Slavery  on  economic  growth  (Nunn)  
—  Culture  on  democracy  (Tabellini,  in  re.  Europe)  
—  Protestant  Missionaries  and  Rule  of  Law  (Woodberry)  
¢  Could  this  be  the  case  for  armed  conflict?    
—  Not  included  in  any  econometric  esSmaSons  of  the  correlates  of  
post  WWII  civil  wars  
¢  QuesSon  is  only  beginning  to  get  addressed  (Besley  and  
Reynal-­‐Querol,  who  find  conflict  persistence  (or  recurrence)  
from  1400-­‐1700  to  post-­‐1960!    
Questions:
•  How much of intrastate warfare today is
a reflection, or continuation, of persistence
of conflict in an earlier era?

•  More specifically: does conflict history

pre-1945 predict post-1945 armed conflict?

•  If yes, why? If no, why not?

Theory 1
We might see persistence for exogenous or
endogenous reasons…
Exogenous:
•  Geography and
population
•  Warlike culture
•  Strategic location

Endogenous:
•  Cycles of Revenge
•  Destruction of
institutions
•  Return on Investment
in military organization
Theory 2:
We might NOT see persistence because of
substantial changes in the international system…

Internationally:
•  Diffusion of civilized norms
(Elias; Pinker)
In Europe:
•  Nuclear weapons
•  European integration
•  Trade
•  Democracy
In Asia and Africa:
•  Decolonization creates new
fissures for political
competition
Theory 3:
We might also see both…

•  Some wars may be

state-destroying
and lead to more
war.

•  Other wars may be

state-building and
lead to peace
(Tilly).
 Data Sources: 1
Correlates Of War (COW):

•  List of all wars from 1816

to the present.

•  Strict Definition of war

with minimum of 1,000
battle deaths.

•  Categorization of wars:
interstate, intrastate,
extrastate, and nonstate.
 Data Sources: 2
Enumeration of Violent
Conflicts (Peter Brecke,
Georgia Tech):

•  List of violent political

conflicts from 1400 to the
present.

•  Definition of armed conflict:

31 combat deaths (but in
practice use of a variety of
explorer/ diaries,
encyclopedias, secondary
histories to record all
mentions of armed conflicts
w/out a strict accounting).
 Brecke and COW have similar regional distributions,
0.35 except maybe SSA and LA pre45

pre 45
post 45
NA/ME
0.30

Asia
0.25
share war yrs in Brecke

EEur SSA
0.20

Asia
SSA
West
0.15

NA/ME
LA/Ca
LA/Ca
0.10
0.05

West EEur
0.00

0.0 0.1 0.2 0.3

share war yrs in COW

 Brecke records much conflict in SSA that
doesn't appear at all in COW

199 China

Russi

Indon
99

UzbekAfgha Ethio South

GreecArgen
Spain
Kazak Moroc
Tanza Sudan Mexic
# Brecke war years pre 1945

Eritr
Benin Turke
Namib Unite India
Urugu
Germa
49

Kirgi Polan Alger Brazi

Italy
Ukrai
Mozam Niger
Egypt Hondu
Nicar El Sa
Ugand Malay Somal
Zaire Austr Libya
Guate Syria
Bulga
Georg Ghana
Seneg AlbanFranc
Peru
Niger Israe
Ruman Pakis
Bosni Vietn
Myanm
Guine ThailArmen
Kenya
Guine
Switz
Ivory Malag Japan Iran
PhiliCuba
AngolYugos
Lithu Maced Monte
Portu Colom
Mali
Malaw
Zambi Sierr Chad Costa Chile
Tadzh Boliv Venez
19

Zimba Belar
Azerb Latvi
Unite
Belgi
Laos CamboLesot Finla
Eston
Burun
Papua
Yemen
Irela Yemen Sri LDenma IraqLeban
Croat HungaParag ZAustr
NewYemen
Canad
Swazi Domin
Ecuad
Jorda Zanzi
Marsh
Palau
Burki
Fed. Singa Haiti
Slova
Weste
Nauru
East
Kirib
Gabon
Solom Czech
9

Bangl Mongo
Turkm Kosov
Korea Saudi
Panam
Rwand
Monac
Luxem
Malta
Camer
San M Norwa
Nethe Tunis Korea
Mauri
Equat
Slove
Congo
Moldo
Swede
Gambi
4

Togo
Liber
Centr Oman
Asia
Botsw
Comor
EEur
LA/Ca
Jamai Bhuta
Taiwa Czech
1

NA/ME
SSA
West
Tonga
Antig
Djibo
FijiV
Nepal
Guyan
Seych
Cypru
Cape
Qatar
Unite
Baham
Domin
Vanua
Brune
Kuwai
Barba
Andor
Bahra
Mauri
Liech
Maldi
Tuval
Beliz
Icela
Grena
Sao
Snt
Surin
Saint
TriniT Vietn
0

0 1 4 9 19 49 99 199

# COW war years pre 1945

DATA  SOURCES:  TWO  OBSERVATIONS  
¢  1.  Rather  than  reconcile  these  two  data  sources  for  a  F/L  
“gold  standard”  list  meeSng  pre-­‐specified  criteria,  we  
keep  the  integrity  of  both,  as  a  robustness  test,  to  see  if  
results  hold  from  two  independent  sources  of  data  on  
the  dependent  variable.  
¢  2.  Our  big  contribuSon  to  both  these  datasets  has  been  
to  geo-­‐locate  all  wars  in  both  datasets  as  to  what  
countries  (in  today’s  boundaries)  those  wars  were  
actually  fought.  
—  Why  not  the  Michalopoulus  strategy  of  virtual  countries?  
  
 
 
 

 
Well, is there persistence?
Three trends:

•  1. Large number of non and

extrastate wars in the 19th
century.

•  2. Relatively less interstate

war after WWII.

•  3. Predominantly intrastate
war post-1945.

•  Inference: new distributions

of war types over time
suggest low probability of
persistence.
 Data Reveal Significant Persistence
Persistence of armed conflict: COW data Persistence of armed conflict: Brecke data

AngolPhili Sudan India

Cambo Indon Ethio
Iraq regression line
India
AfghaEthio
Myanm
Vietn 45 degree line Iraq Phili
Mozam Cambo Israe
Sri Zaire
L Thail

54
Guate
Vietn Eritr
Chad Yemen Colom
Laos Myanm
Ugand
19

Laos Alger Angol

Chad Somal Indon
IranMozam
Nicar
Iran Sudan China
Somal China
Namib El Sa Moroc
Vietn Oman Moroc
Zimba Pakis Unite
Venez Alger

# Brecke war years post 1945

BanglBurun
# COW war years post 1945

Ugand Guate PeruTurke Sri L NamibAfgha

Niger Israe
Nicar Colom Leban
ZimbaPakisZaire South
Syria
Rwand Turke

19
Burun Malay Egypt
Malay
Liber Sierr Leban Peru
Yemen Egypt South
Russi Parag
Latvi GuineEl Sa
Nepal Lithu
Oman Eston Singa Sierr
Rwand
R^2 = .14
6

Tadzh
Eritr Korea Azerb
EquatEastPapua Kenya
Congo
Liber Niger R^2 = .20
AzerbBosni Yemen
ThailKenya Camer Unite
Cypru
Papua Taiwa Georg Argen Georg
Korea

6
Nepal TunisJorda Niger
Congo
Ivory Tunis Croat Polan
Ukrai Ruman Argen
Domin Cuba
Korea
Korea
Djibo CroatBolivBosni
Chile
EcuadCosta Ukrai Greec
Kuwai
Moldo
Camer Yugos
Syria
Guine
Guine
East Kosov
Malag Cuba Kuwai PanamHaiti Tanza Russi
2

Brune Kosov
Gabon Tadzh
Nethe Malag
Portu Libya
Hondu
Unite Mauri Seneg
Yugos
Zambi Brazi
GermaSpain
Mexic
Cypru
Yemen Zanzi Ecuad
Costa Chile
Parag
2
Bangl
Jorda DenmaRumanBoliv
Domin
Hunga
Hondu Maldi
Grena Lesot
SloveSaudiHunga
Malaw
Comor Czech Ghana
Moldo Slova Guine

Zambi
Nauru
Gabon
Saint
Icela
Mauri
Maldi
DominSao
Bahra
Grena
Comor T Nethe Turkm
KiribM
Tuval Austr
Haiti
GhanaNew Austr
Finla
Czech
Burki
Equat
MalawSan
Seych
Surin
Togo
Luxem
Centr
Gambi
Canad
Brune
Vanua
Botsw CzechTanza
Belgi
Kazak
Kirgi Saudi
Uzbek
Z
Bulga Seych
Fiji
Qatar
Mauri
Unite
Liech
Icela
Barba Jamai Finla
Austr
Yemen
Eston
Maced
Latvi Franc
Polan
Panam
Guyan
Barba
Monac
Tonga
Cape Saint Benin Slova
Japan Venez
Germa Sao
Vanua T Taiwa Centr Luxem
Beliz
Antig Malta Irela
Yemen
Zanzi
New Armen
ZMonte Benin
LesotUnite Alban Surin Kirgi
0

Malta
Switz
Trini
Palau
Djibo Slove
Swazi Germa Saint
Trini Marsh
Palau
Kirib Mali Urugu
Kazak
0

Weste Qatar
Unite
Fiji Maced
Mongo Seneg Spain
Franc Tuval
Cape Bhuta Togo
Botsw Mongo
Monac
Swede Nauru
WesteDenma
Fed.
Canad
Belgi
Swazi Austr
Germa Uzbek
Marsh
Niger
Swede Singa
Antig Norwa MaliUrugu
Libya Brazi
Italy Vietn
Saint
Tonga
Andor
Snt
Domin V Czech Turkm
NorwaBurki Japan
Lithu Bulga
Alban
Baham Snt V Bhuta Armen
Beliz
Fed.
Liech
Irela
Mauri
Jamai
Andor
Solom Belar Portu
Monte Greec Mexic BahamBahra
Guyan Gambi
SanSolom
M Belar Switz
Ivory Italy

0 2 6 19 54 147 0 2 6 19 54 147 402

# COW war years pre 1945 # Brecke war years pre 1945
PERSISTENCE  REMAINS  SIGNIFICANT  WITH  STANDARD  
CONTROLS  (COW  DATA;  DV=LOG(POST45  WARS  +  1)  
Models   1   2   3   4   5  
Log(pre45  war   0.37***   0.44***   0.25**   0.24**   0.20*  
yrs  +1)   (0.06)   (0.06)   (0.07)   (0.08)   (0.10)  
Log  (pop  in   0.17***   0.16***   0.15*  
1800)   (0.04)   (0.04)   (0.08)  
Log  (%  rugged)   0.10*   0.15†  
(0.05)   (0.08)  
Log  (total  langs,   0.06  
≈  1800   (0.08)  
Constant   0.47***   0.92***   -­‐1.04*   -­‐0.91†   -­‐0.71  
(0.11)   (0.17)   (0.52)   (0.53)   (1.03)  
N   200   200   199   197   144  
R2   0.14   0.35   0.40   0.41   0.37  
Region  FE’s   No   Yes   Yes   Yes   Yes  
OLS:  SE’s  in  parens;  †p  <.10;  *  p  <  .05;  **  p  <.01;  ***  p  <  .001  
CORRELATES  OF  WARFARE  IN  BOTH  ERAS  
DV   Pre45  wars   Pre45  wars   Post  44  wars   Post  44  wars  
Log  (#  langs  ≈   0.16*   0.09  
1800)   (0.07)   (0.08)  
ELF  (1960)   0.06   .93*  
(0.34)   (0.36)  
Log  (pop  in   0.37**   0.41***   0.22**   0.25***  
1800)   (0.06)   (0.05)   (0.07)   (0.05)  
Log  (%   0.14†   0.11   0.18**   0.16*  
rugged)   (0.07)   (0.07)   (0.08)   (0.08)  
Constant   -­‐4.37***   -­‐4.39***   -­‐1.57   -­‐2.09*  
(0.85)   (0.79)   (0.96)   (0.84)  
N   144   160   144   160  
R2   0.46   0.41   0.35   0.37  
Region  FE’s   YES   YES   YES   YES  
OLS:  SE’s  in  parens;  †p  <.10;  *  p  <  .05;  **  p  <.01;  ***  p  <  .001  
SUMMARY  OF  REGRESSION  RESULTS  FOR  EACH  ERA  
¢  On  average,  five  more  years  of  war  pre-­‐1945  associates  
with  about  one  more  year  post-­‐1945  
¢  Controlling  for  populaSon  in  1800  has  a  big  effect,  could  
be  due  to  “exposure”  (i.e.  more  people,  more  chances  
for  an  insurgency)  or  the  difficulty  of  governing  large  
populaSons  
¢  Rough  terrain  related  to  more  conflict  in  both  periods,  
but  more  so  auer  1945  
¢  Ethnic  diversity  in  1800  (see  LaiSn/Robinson)  related  to  
conflict  pre  1945,  but  not  auer.  ELF    measured  auer  
1945  related  to  total  conflict  auer  1945,  but  not  before.  
Suggests  ethnic  diversity  is  endogenous  to  state  building  
in  the  long  term.  
ConFnental  “reversals  of  fortune”  also  show  disFnct  regional  
effects    
INDEED  PERSISTENCE  IS  NOT  SIMILAR  ACROSS  REGIONS:  
REGIONAL  EFFECTS:  SUMMARY  
¢  Lots  of  warfare  in  EEur/FSU  pre45  due  to  interstate  
conflict  and  Russian  imperial  expansion.  That  ends  for  
many  territories  under  Soviet  and  post-­‐Soviet  rule;  
¢  Western  Europe  has  great  variaSon  in  warfare  pre45  
but  nearly  all  countries  close  to  zero  post44  
¢  Moderate  number  of  African  and  Asian  cases  that  saw  
no  COW  warfare  pre45,  but  have  major  conflicts  post44:  
e.g.:  Uganda,  Rwanda,  Burundi,  Liberia,  Ivory  Coast,  
Nepal,  PNG  
¢  Clearly  we  need  to  control  for,  and  look  within  regions  
 Explaining Persistence:
Some Preliminary Findings
FIRST  FINDING:  
¢  Extra State Wars pre-45 are associated with intra-
state wars post-44
¢  Intra-state wars pre-45 have no relationship with
intra-state wars post-44
¢  Conjecture: extra-state wars appear to be state-
destroying while intra-state wars have Tilly-like
state-building implications.
Strong extra pre/intra post correlation
accounts for much of the general findings:
 Most of the action:
More colonial wars pre 45 => more civil war post 45

India

Phili Ethio
54

Asia Eritr
SSA Myanm
Chad
Ugand Sudan
Somal Indon
# Brecke civil war years post 1945

Angol Mozam
Laos
Sri L Bangl ChinaVietn
Burun
Cambo
Zimba Afgha Zaire South
19

Thail
Sierr
Equat
East Papua
Rwand Liber
Niger
Pakis
6

Nepal Congo Niger

Djibo

Brune Malay
2

Lesot Kenya TanzaGuine

Comor Zambi Gabon Ghana reg line (R2 = .33)
Korea 45 degree line

Singa
Taiwa
Kirib
Maldi
Vietn
Nauru Guine
Zanzi
Weste
Fed.
Tuval Bhuta Gambi Korea Benin
Burki Malag
Sao T Swazi
Cape Mali
0

Seych
BotswSolom
CentrPalau Malaw
Tonga
Fiji
Mauri
Vanua Japan
Camer
Togo Mauri Namib
Marsh Mongo Ivory
Seneg

0 2 6 19 54

# Brecke colonial and imperial war years pre 1945

 total 535 1530

Relation between civil war yrs post 45 and colonial war yrs pre 45, Brecke data

DV = ln(civil war yrs post45) 1 2 3 4 5

ln(extrapre45 ) 0.50⇤⇤⇤ 0.50⇤⇤⇤ 0.34⇤⇤⇤ 0.34⇤⇤⇤ 0.28⇤⇤⇤
(0.06) (0.07) (0.08) (0.08) (0.08)
ln(land area) 0.01 0.01 0.07
(0.05) (0.05) (0.06)
ln(pop 1800) 0.19⇤⇤ 0.17⇤⇤ 0.18⇤⇤
(0.06) (0.06) (0.06)
ln(ruggedness) 0.16⇤⇤ 0.18⇤⇤
(0.06) (0.06)
ln(# langs in 1800) 0.21⇤
(0.08)
constant 0.45 ⇤⇤⇤
0.75 ⇤⇤⇤
1.43 ⇤
1.20 ⇤
1.17⇤
(0.12) (0.21) (0.59) (0.59) (0.59)
N 200 200 199 197 192
region fixed effects N Y Y Y Y
adj. R2 0.24 0.27 0.32 0.35 0.37
Standard errors in parentheses
†
significant at p < .10; ⇤ p < .05; ⇤⇤ p < .01; ⇤⇤⇤ p < .001
SECOND  FINDING  
¢  No  evidence  of  “persistent  enemies”  or  of  groups  especially  
conflict-­‐prone.  
¢  Three  procedures  to  invesSgate  this:  
—  A)  Read  secondary  literature  on  a  random  selecSon  of  countries  
asking  if  there  is  correspondence  across  eras  of  combatants  
—  B)  Select  and  research  on  countries  with  high  pre-­‐  and  high  post-­‐  
where  correspondence  would  be  most  likely  found  
—  C)  Add  precise  “combatant  name”  to  Breke  and  COW  datasets  to  
invesSgate  this  quesSon  staSsScally  (now  being  done).  
¢  Preliminary  analysis:  lots  of  new  conflicts  post-­‐independence  
between  groups  that  formerly  had  lixle  to  do  with  each  
other,  or  fought  together  against  the  colonial  power.  Almost  
no  evidence  of  “ancient  hatreds”  driving  recurrent  warfare.  
PERMANENT  ENEMIES  FROM  RANDOM  LIST:  ANGOLA  
Persistence  [war  years   Analysis  
pre-­‐1945  (HiLo);  post-­‐1944  
(HiLo]  

22  (Hi);  38  (Hi)   1.  Persistent  high  

2.  No  common  enemies  [1815-­‐44,  mostly  Portugal  vs.  the  
rebellious  kings;  1945-­‐present,  at  first  Portugal  vs.  
insurgencies  of  urban/based  miliSas  that  hide  in  rural  
tribal  areas;  and  then  wars  among  the  different  insurgent  
groups];  no  common  areas  of  resource  value.]  
3.  Military  organizaSons  (e.g.  the  Bailundu  kingdom)  in  
the  19th  c.  play  no  role  in  the  post  1960  insurgencies.  
UNITA  (see  Marcum)  was  based  in  Ovimbundu  territory,  
but  its  organizaSon  as  an  independence  seeking  party/
miliSa  was  all  organized  by  the  “new  men”  educated  at  
the  religious  mission  schools  who  recruited  from  forced  
labor,  and  (those  influenced  by  Maoism)  from  the  
peasantry.  
PERSISTENT  ENEMIES  FROM  RANDOM  LIST:  TANZANIA  
Persistence  [war  years   Analysis  
pre-­‐1945  (HiLo);  
post-­‐1944  (HiLo]  
70  (Hi);  3  (Low)   1.  Nine  wars  (23  years  of  war)  pre-­‐45  involving  the  Arabs,  or  the  Arab  
trade,  or  control  over  Zanzibar;  one  revoluSon  post-­‐44  in  1964  in  
Zanzibar  could  be  considered  a  persistent  organizaSon  for  conflict,  a  
persistent  enemy  (Arab  traders  vs.  inland  interests),  or  a  persistently  
valuable  port  worth  fighSng  over  (Mombasa  and  Zanzibar)  
2.  Maji-­‐Maji  rebellion  the  largest  in  the  pre-­‐1945  era;  one  of  the  
belligerents  was  the  Ngoni  who  fought  a  war  with  the  Germans  in  
1897;  but  Maji-­‐Maji  has  no  implicaSons  for  post-­‐independence  
insurgency.  Here  is  a  case  of  non-­‐persistence  of  heroic  organizaSon.  

3.  Perhaps  the  largest  death  count  was  with  the  Hehe  (Breke  has  
them  with  2  wars  in  the  1890s),  the  first  in  which  the  Germans  were  
defeated.  However,  they  did  not  parScipate  in  Maji  Maji  and  have  
not  been  an  insurgent  force  in  post  WWII  Tanganyika/Tanzania.  Here  
is  another  case  of  non-­‐persistence  of  heroic  organizaSon.  
PERSISTENT  ENEMIES  FROM  HIGH-­‐HIGH  LIST:  INDIA  
Group  fought  pre  and   Group  fought  pre,    but   Group  fought  post,  but  
post   not  post   not  pre  
1.  Sikhs  (1845,  1848,   1.  Marathas  (1817)   1.  Hindus  and  Muslims  
1984)   2.  Burma  vs  Assam   in  parSSon  (1946)  
(1821)   2.  Telanganas/Razakars  
3.  Bharatpuran  (1825)   (1947)  
4.  Jammu/Tibet  (1841)   3.  Hyderabad  (1948)  
5.  Gwalior  (1843)   4.  Naxalites  (1970)  
6.  Dards  (1852)   5.  Kashmir  (1947,  1965,  
7.  Santals  (1855)   1990)  
8.  Sepoy’s  (1857)  
9.  Moplahs  (1921)  
PERMANENT  ENEMIES  FROM  HIGH-­‐HIGH  LIST:  CHINA  
Group  fought  pre  and   Group  fought  pre,    but   Group  fought  post,  but  
post   not  post   not  pre  
1.  Tibet  (1912,  1918,   1.  Muslim  rebels   1.  State  vs.  Counter-­‐
1950,  1956)   2.  Kokand   revoluSonaries  (Cultural  
2.  Communists  vs.   3.  Taiping   RevoluSon)  
Republicans/KMT   4.  Miao  vs.  Han  
5.  Nian  vs.  Han  
6.  Yunnan  Muslims  
7.  Zaili/Jindan  sects  
8.  Hui  Muslims  
9.  Gansu  Muslims  
10. MulSple  regional  
warlords  
11. Etc…  
PERMANENT  ENEMIES  FROM  HIGH-­‐HIGH  LIST:  SUDAN    
Group  fought  pre  and   Group  fought  pre,    but   Group  fought  post,  but  
post   not  post   not  pre  
[None]   1.  Oxomans  vs.  Sultans   1.  Anya  Nya  –  
(1820)   southerners    vs.  
2.  Egypt  vs.  Mahdi   center(1963)  
(1824)   2.  SPLA  –  southerners  
3.  Egypt  vs.   vs.  center  (1983)  
Hadendowa  (1831)   3.  Dinka  vs.  Nuer  –  
4.  Egypt  vs.  Bahr  el   intra-­‐southerner  
Ghazal  (1869)   (1991)    
5.  Egypt  vs.  slavers   4.  Darfur  –  
(1878)   agriculturalists  vs.  
6.  UK,  Ethiopia,  Italy  vs.   herdsman  with  
Mahdi  (1881,  1889,   support  of  state  
1894,  1896)   (2003)  
OTHER  PRELIMINARY  FINDINGS  
¢  1.  Exogenous  factors,  as  we  saw  in  earlier  tables,  such  as  
rugged  terrain  and  populaSon  density  remain  significant  
across  both  eras  
¢  2.  Evidence  from  our  random  list  of  countries  do  not  
reveal  many  strategic  locaSons  over  which  wars  have  
been  fought  repeatedly  (perhaps  Zanzibar,  and  its  
locaSon  for  the  Arab  trading  system,  is  an  excepSon).  
—  This  can  be  explained  by  the  fact  that  what  consStutes  a  
strategic  locaSon  in  the  19th  century  (e.g.  a  river  bend  ideal  
for  a  fort)  is  different  in  the  20th  (e.g.  oil  reserves).  We  
therefore  have  had  trouble  finding  cross-­‐era  coding  rules  for  
strategic  locaSon.  
CONCLUSIONS  
¢  1.  Non-­‐trivial  within-­‐region  persistence  of  armed  conflict,  
before  1945  to  auer:  this  is  parScularly  evident  in  Asia,  SSA,  
NA/ME,  and  LA/Ca  
¢  2.  Not  completely  eliminated  by  controlling  for  populaSon,  
rough  terrain,  and  ethnic  diversity  in  1800  (but  we  can’t  be  
sure  other  durable  factors  won’t  eliminate  it).  
¢  3.  Ethnic  diversity  correlates  with  warfare  in  each  era,  but  is  
not  an  exogenous  source  of  conflict  (i.e.  it  changes  values  
across  eras)  
¢  4.  Results  are  primarily  due  to  colonial/imperial  war  
correlaSng  with  post-­‐independence  civil  war  
¢  5.  The  persistence  story  is  less  about  ancient  hatreds/
permanent  enemies,  and  more  about  the  arc  of  state-­‐
building