THE LIMITS TO PARTITION
by Michael C. Horowitz and Alex Weisiger
International Security
Spring 2009
Michael C. Horowitz and Alex Weisiger are Assistant Professors of
Political Science at the University of Pennsylvania.
Carter Johnson is a Ph.D. candidate at the University of Maryland,
College Park.
To the Editors (Michael C. Horowitz and Alex Weisiger write): In his
article "Partitioning to Peace: Sovereignty, Demography, and Ethnic
Civil Wars," Carter Johnson argues that partition is frequently the
best available policy response to ethnic civil wars. By creating a
new measure of the degree of demographic unmixing achieved through
partition, Johnson demonstrates that mere changes in sovereignty are
insufficient to produce peaceful outcomes. He contends, however, that
"complete" separation of the two sides will help to bring peace. He
concludes that partition is a useful tool for the international
peacemaker, with the caveat that it "should be considered, however,
only where populations are already largely separated at the time of
intervention, or where interveners are prepared to separate groups
using mass population transfers." 1 He interprets these findings as
"strong evidence for advocates of partition (p. 168)."
We believe that Johnson’s caveat is more important than it appears at
first blush, and that its practical implication is to render partition
an effectively useless policy tool. The partition prescription is
grounded in the logic of the ethnic security dilemma, but there is
empirical evidence that genuine conflicts of interest exist alongside
security dilemma fears. In this context, participants will resist the
population transfers necessary for effective partition. Given that the
international community is not going to carry out population transfers
in the face of violent resistance, effective partition is unlikely to
occur except through ethnic cleansing during war. Johnson’s data are
consistent with these observations: effective partition rarely happens,
and when it does, it is through ethnic cleansing rather than organized
population transfers. In short, successful partitions ratify military
outcomes. Furthermore, when we consider that even effective partition
has not guaranteed peace, these observations strongly suggest that
the international community should retain its current position toward
partition of tacit acquiescence in a small number of unavoidable cases,
but should not promote partition as a strategy to end ongoing wars.
The partition prescription is grounded in the logic of the ethnic
security dilemma, in which conflict breaks out in an intermixed ethnic
setting because each side fears that the other is going to attack
it and therefore sees a strategic benefit to striking first. Neutral
partition, if possible, is a logical solution to this problem because
it eliminates incentives for preemptive action and facilitates each
side’s efforts to defend itself. Given these effects, partition
logically should be welcomed by participants. 2
That observation, however, points to a significant problem with the
ethnic security dilemma and the prescription of partition. From
Kashmir to Kirkuk, ethnic conflicts have involved not only fear
of attack but fundamental disagreements about who should control
important pieces of territory. In other words, security motives
interact with predation to produce distinct but overlapping incentives
for fighting. 3 This observation has important implications for the
viability of partition as a strategy. Partition depends on outside
interveners being willing to facilitate the transfer of populations,
which they are reluctant to do given the violations of international
law involved (p. 150). If the conflict in question involves not only
security fears but competition for control of disputed territory,
the international force facilitating partition does not merely act
in the context of violence; it becomes a target. As examples such as
Somalia in 1993 and Rwanda in 1994 show, there is little historical
evidence that outside interveners are willing to suffer these sorts of
costs to facilitate an outcome from which they do not directly benefit.
Given competition over territory, voluntary population transfers
are also a chimera, which Johnson acknowledges when he states that
"all of the complete cases" of partition "involved large-scale forced
population transfers during the countries’ wars, with the possible
exception of Bangladesh." 4 If the international community will
not carry out involuntary transfers and the participants will not
accept voluntary ones, then ethnic cleansing–held up by partition
advocates as worse than their recommendation–is the only way that
population transfers will occur. 5 From this perspective, advocacy
for partition becomes advocacy for signing off on ethnic cleansing
after the fact. That is very different than arguing that partition
should be a preferable policy choice.
These observations are significant when we turn to Johnson’s
empirics. Johnson takes issue with Nicholas Sambanis’s finding
that partition has not been associated with more peaceful postwar
outcomes. He argues that this finding did not fairly test the partition
argument, because many of the cases that Sambanis coded as partitions
did not meet the standards laid out by partition advocates. 6 Johnson
categorizes Sambanis’s cases of partition into two groups: those in
which partition involved the ethnic separation that theory indicates
was necessary, and those in which ethnic unmixing did not accompany
the change in sovereignty. Consistent with theory, Johnson finds that
the greater the degree of unmixing, the better the postwar outcome,
although the strength of this finding is limited by "the statistical
problem that, as yet, there have been too few partitions" (p. 160).
The observation that there have been few "complete" partitions is more
than just a statistical problem, however. Our theoretical discussion
indicates that partition will almost never be a viable policy, at best
merely ratifying preexisting military outcomes. Closer examination
of the evidence bears that argument out. Of the seventeen cases
highlighted by Johnson in the post-World War II era, only six count
as complete partitions: the separation of Bangladesh from Pakistan
in 1971, the partition of Cyprus in 1974, the establishment of
Eritrea in 1991, the Russian-backed creation of Abkhazia and South
Ossetia in Georgia in 1993 and 1994, and the Armenian acquisition
of Nagorno-Karabakh in 1994 (ibid.). Setting aside the sui generis
case of Bangladesh, in which geographical separation between East
and West Pakistan limited opportunities for further violence, all of
Johnson’s six cases experienced significant involuntary population
transfers. The partition of Cyprus involved the disorganized mass
transfer of ethnic Greeks and Turks from the two parts of the island,
far closer to ethnic cleansing than partition. Similarly, the Armenians
cemented their conquest of Nagorno-Karabakh by ethnically cleansing
the Azeris, while the Azeris responded by forcing out Armenians in
their midst. 7 In the conflicts in both South Ossetia and Abkhazia,
large numbers of Georgians (initially the majority in Abkhazia)
were ethnically cleansed with the aid of the Russians. 8 Finally, the
Ethiopia-Eritrea conflict also featured significant ethnic cleansing,
and yet it ended not because of unmixing but because the Soviets
withdrew military support for Ethiopia, forcing it into settlement
negotiations to ratify the military outcome. 9
Moreover, the evidence that even "complete" partition limits subsequent
violence is exceptionally weak. The partition of Ethiopia was followed
only seven years later by a severe interstate war, which in two years
killed 100,000 people, roughly two-thirds of the total who died in
the decade and a half of the previous war. 10 Johnson acknowledges
the problematic nature of this case (pp. 161-162), but it raises
the further concern that even should partition prove possible it
may not provide the benefits that advocates expect. Furthermore,
the fighting in South Ossetia and Abkhazia in August 2008–after the
publication of Johnson’s article–underlines the limits of partition
in preventing violence. Given a total of six "complete" partitions
in his data set, the observation of significant fighting in three
provides little reason to think that even successful separation of
populations will prevent a return to violence.
Of course, part of the appeal of the partition hypothesis is the claim
that no other option exists. 11 Yet Sambanis’s data, on which Johnson
relies, belie this point. To evaluate the efficacy of partition,
Johnson focuses on whether war has recurred within two or five
years from the time of partition and whether lower-level violence has
occurred within those same periods. Of eighty ethnic wars, fifty-seven
did not return to war within two years of termination, and fifty-two
remained without war after a five-year period. Lower-level violence
was more common, but still thirty-two cases avoided even that through
the first two years, with thirty-one sustaining peace through five
years. Partition is not the only way to achieve peace.
We do not wish to deny that there may well be reason to acquiesce to
unpalatable outcomes after the fact. Civil wars are a rough business,
and there will inevitably be times when it is preferable to tacitly
acquiesce to a new reality than to force a return to the previous
situation or to pretend that nothing has changed. But there is a
significant difference between tacitly acquiescing when no other
option is available and actively promoting those outcomes from the
start. The knowledge that the international community will legitimate,
or at least accept, ethnic cleansing after the fact provides reason
for leaders to consider adopting that policy when they otherwise would
not have. 12 In other words, the moral hazard problems with a policy
of partition are severe.
Johnson is to be lauded for advancing and empirically substantiating
an important revision to the partition argument. We differ from
him, however, in the implications of that revision. The rarity of
"successful" partitions is, to us, an indicator that partition is
not a good prescription. Real partitions are rare because the ethnic
security dilemma is not the only source of conflict at work in ethnic
civil wars, and predation provides strong incentives to resist the
sorts of population transfers that Johnson argues are necessary for
partition to work, while even successful partition does not guarantee
peace. Far better, then, to recognize what Johnson’s evidence really
shows–that partition is truly an option of last resort.
–Michael C. Horowitz Philadelphia, Pennsylvania
–Alex Weisiger Philadelphia, Pennsylvania
Carter Johnson Replies:
I want to begin by thanking Michael Horowitz and Alex Weisiger for
their constructive comments. Scholars’ understanding of partition
has long benefited from precisely this kind of exchange, enabling
more informed decisionmaking on a topic that influences millions
of civilian lives. My response proceeds with a brief review of the
aims and conclusions set forth in my 2008 article, 1 highlighting a
fundamental area of agreement between the authors and me. I then move
to areas of disagreement, drawing on the data used in my article to
demonstrate the policy relevance of partition even if the international
community is not prepared to implement forced population transfers.
In my article I argue that "complete" partitions–those that create
a separate ethnic homeland and completely separate warring ethnic
groups–produced peaceful outcomes for at least the first five years
following the end of hostilities. This finding challenges the only
previous cross-national study that systematically examined partition
and confirmed empirically what partition advocates have long argued. 2
Giver that 60 percent of ethnic civil wars between 1945 and 2004
experienced deadly conflict renewal within the first five years,
the benefits of "complete" partition are worthy of policymakers’
consideration. 3
Horowitz and Weisiger argue that "complete" partition is essentially a
"useless policy tool" because civilians will resist forced transfers,
and "the international community is not going to carry out population
transfers in the face of violent resistance." This is an important
point, and I agree that forced population transfers are unlikely to
receive support from Western powers. I also agree that this limits
the policy usefulness of partition, a fact I acknowledged in my
article (p. 167). Still, not all forms of complete partition require
the international community to forcibly move civilians to ensure
long-term peace.
REDRAWING OF BORDERS The ethnic security dilemma argues that
ethnic groups separate from one another in wartime. Therefore,
by the time the international community chooses to act, which can
take months and sometimes years, it may not need to engage in forced
population transfers because separation may already have been largely
accomplished. Instead, the international community can redraw borders
to minimize the number of minority groups in the emerging postpartition
states.
Kosovo’s partition from Serbia provides a useful example. The 1999
conflict largely separated the ethnic Kosovar and Serb populations,
but instead of drawing the borders to reflect this reality, NATO
and its partners maintained the prewar institutional borders of
the Kosovar autonomous region, leaving a significant Serb minority
within the Western-recognized borders of today’s Kosovo. That region
experienced a large-scale renewal of deadly conflict in May 2004 and
remains dangerously unstable today; although officially under the
control of Pristina, the region is in fact under the partial control
of Belgrade. Instead of forcing the Serb populations from Kosovo,
the international community could redraw the northern border to
allow most of the Serb minority to join rump Serbia. There would
still be small Serbian enclaves in Kosovo but, as I argue below,
this is unlikely to cause widespread violence.
Evidence from ethnic civil war terminations between 1945 and 2004
suggests that even partitions that leave small numbers of an ethnic
minority behind are also successful at maintaining the peace. As I
noted in my article, six cases of "complete" partition have occurred
since 1945, but this finding was based on a highly restrictive
definition of "complete"–those reaching at least 95 percent on
the Postpartition Ethnic Homogeneity Index (PEHI) (p. 160). If all
of the PEHI scores are ranked from highest to lowest, however, one
finds that none of the ten highest-scoring countries experienced a
recurrence of conflict in the first five years after the cessation of
hostilities (Cyprus 1974, Georgia-Abkhazia 1993, Pakistan-Bangladesh
1971, Georgia-South Ossetia 1992, Ethiopia-Eritrea 1991, Azerbaijan
1994, Bosnia 1995, Yugoslavia-Croatia 1995, Israel 1948, and
India 1948). This includes countries with a PEHI score in the 50s
(p. 160). In other words, partitions that "largely" (as opposed to
"completely") separate ethnic groups are also likely to bring about
an enduring peace, making the task for the international community
much easier and therefore more practical.
THE THEORETICAL FOUNDATION OF PARTITION Horowitz and Weisiger also
argue that the ethnic security dilemma alone is the grounding logic
of partition. I disagree. Although the ethnic security dilemma is
a core component part of the partition argument, there are other
theoretical bases as well. First, as I mention in the article,
Alexander Downes draws on the realist school, emphasizing the
importance of establishing separate states to maintain long-term
peace (p. 149). 4 The creation of separate states addresses the
fear of each side over the other’s intentions even where demographic
separation is achieved. Second, establishing separate states solves
the problem of credible commitments, because neither side needs to
decommission its weapons. The problem of credible commitments has
been identified as a central obstacle in ending civil wars. 5 Third,
Thomas Chapman and Phil Roeder recently presented an institutionalist
argument to support partition, arguing that independence, not autonomy,
is most beneficial for establishing long-term peace. 6 This widened
theoretical explanation for partition might explain why complete
and near-complete partitions are successful: a combination of both
sovereignty and demography is central for the establishment and
maintenance of peace after ethnic civil war.
LONG-TERM VERSUS SHORT-TERM PEACE Finally, Horowitz and Weisiger note
that some of the conflicts I examined in my article recurred after
the five-year limit. This is a legitimate point, although it occurred
in two, not three cases: the "conflict" in Abkhazia in August 2008
caused only one soldier’s death, and therefore does not qualify as
renewed, low-level violence or war. In my article I addressed the
1998 Ethiopia-Eritrea conflict, which did cause a great number of
deaths, but also ended remarkably quickly (after two years, compared
with three decades for the civil war): indeed, interstate wars are
typically extremely short relative to civil wars, and this is one
key advantage to the "internationalization" of such conflicts. The
Georgian-Russian conflict over South Ossetia in August 2008 further
illustrates this point: active combat operations ended after only
six days. The de facto states of South Ossetia and Abkhazia have
obsequiously pandered to Russian interests since they achieved de facto
independence in 1992 and 1993, respectively, explicitly recognizing
Russia as their only security guarantee against the existential
threat posed by Georgia. By building up deterrence, these states aim
to avoid war, but this objective could be pursued only as a result
of partition. The deterrence capability provided by the region’s only
superpower cannot prevent a rash, hot-headed president from launching
an ill-fated invasion, but at least the war with South Ossetia ended
quickly, and Tbilisi is unlikely to ever attempt such actions in the
future. Further, the death count was below 1,000 (combined civilian
and combat), meaning that even its label as a "war" is open to dispute.
CONCLUSION My article aimed to build on the academic debate over
partition and to draw out policy implications from those results. I
provided a novel methodological approach that clarified the debate and
challenged conventional wisdom. I also identified several theoretical
strands that support partition as a solution to ethnic civil war.
With regard to policy, Horowitz, Weisiger, and I agree that the
international community will not sponsor forced population transfers
at this time, and this limits the ability to promote partition as a
solution to ethnic civil wars. We differ, however, in how we see the
remaining implications for partition as a policy tool. Because ethnic
civil wars separate warring ethnic groups, and partitions that largely
separate ethnic groups create peaceful solutions, policymakers need to
seriously consider drawing fresh borders to separate warring groups as
much as possible. Ethnic civil wars have an extremely high recidivism
rate during the first five years, and such violence typically harms
civilians the most. Given the promising track record of partition
during the same five-year period, it is imperative that policymakers
seriously consider this option. They could begin with several ongoing
ethnic conflicts, including those involving Israel, Kosovo, Lebanon,
Sri Lanka, and Sudan.
–Carter Johnson College Park, Maryland
FOOTNOTES 1 Carter Johnson, "Partitioning to Peace: Sovereignty,
Demography, and Ethnic Civil Wars," International Security, Vol. 32,
No. 4 (Spring 2008), pp. 140-170, at p. 165. Further references to
this article appear parenthetically in the text.
2 Chaim D. Kaufmann, "When All Else Fails: Ethnic Population Transfers
and Partitions in the Twentieth Century," International Security,
Vol. 23, No. 2 (Fall 1998), pp. 120-156.
3 Jack L. Snyder and Robert Jervis, "Civil War and the Security
Dilemma," in Barbara F. Walter and Snyder, eds., Civil Wars,
Insecurity, and Intervention (New York: Columbia University Press,
1999), pp. 15-37.
4 He then goes on to describe how forced population transfers occurred
in Bangladesh as well. Johnson, "Partitioning to Peace," p. 163.
5 In this context, it is unsurprising that population transfers
for partition have generally been extremely violent. Radha Kumar,
"The Troubled History of Partition," Foreign Affairs, Vol. 76, No. 1
(January/February 1997), pp. 22-34.
6 Johnson, "Partitioning to Peace," p. 143; and Nicholas Sambanis,
"Partition as a Solution to Ethnic War: An Empirical Critique of the
Theoretical Literature," World Politics, Vol. 52, No. 4 (July 2000),
pp. 437-483.
7 Human Rights Watch, Cynthia G. Brown and Farhad Karim, eds., Playing
the "Communal Card": Communal Violence and Human Rights (New York:
Human Rights Watch, April 1995),
8 Human Rights Watch, ed., Bloodshed in the Caucasus: Violations of
Humanitarian Law and Human Rights in the Georgia-South Ossetia Conflict
(New York: Human Rights Watch, March 1992); and Human Rights Watch,
Georgia/Abkhazia: Violations of the Law of War and Russia’s Role in
the Conflict (New York: Human Rights Watch, March 1995).
9 Edmond J. Keller, "The United States, Ethiopia, and Eritrean
Independence," in Amare Tekle, ed., Eritrea and Ethiopia: From Conflict
to Cooperation (Lawrenceville, N.J.: Red Sea Press, 1994), pp. 169-170.
10 The casualty levels undermine Johnston’s claim that postpartition
wars will be less serious. That a two-year war had almost two-thirds
the casualties of a decade-and-a-half civil war further undermines
the claim that postpartition wars will be "better" than postconflict
wars in nonpartition situations. On the casualty totals, see Meredith
R. Sarkees, "The Correlates of War Data on War: An Update to 1997,"
Conflict Management and Peace Science, Vol. 18, No. 1 (Spring 2000),
pp. 123-144.
11 Kaufmann, "When All Else Fails."
12 James D. Fearon, "Separatist Wars, Partition, and World Order,"
Security Studies, Vol. 13, No. 4 (Summer 2004), pp. 394-415. Fearon
also notes that a general policy of partition would undermine the
taboo against encouraging the breakup of opposing states, leading to
more conflictual international relations.
1 Carter Johnson, "Partitioning to Peace: Sovereignty, Demography,
and Ethnic Civil Wars," International Security, Vol. 32, No. 4 (Spring
2008), pp. 140-170. Additional references to this article appear in
parentheses in the text.
2 This line of thought was initially expressed by John Mearsheimer and
Stephen Van Evera. Mearsheimer and Van Evera, "When Peace Means War:
The Partition That Dare Not Speak Its Name," New Republic, December
18, 1995, pp. 16-21. But it was developed first by Chaim Kaufmann,
"Possible and Impossible Solutions to Ethnic Civil Wars," International
Security, Vol. 20, No. 4 (Spring 1996), pp. 136-175.
3 I defined "deadly conflict renewal" as renewal of combat leading
to at least twenty-five deaths.
4 Alexander B. Downes, "The Holy Land Divided: Partition as a Solution
to Ethnic Wars," Security Studies, Vol. 10, No. 4 (Summer 2001),
pp. 58-116.
5 Barbara F. Walter, Committing to Peace: The Successful Settlement
of Civil Wars (Princeton, N.J.: Princeton University Press, 2002).
6 Thomas Chapman and Phil G. Roeder, "Partition as a Solution to Wars
of Nationalism: The Importance of Institutions," American Political
Science Review, Vol. 101, No. 4 (November 2007), pp. 677-691.