Where Do the Peacekeepers Go?1
MICHAEL GILLIGAN
Department of Politics, New York University
and
STEPHEN JOHN STEDMAN
Department of Political Science, Stanford University
What determines where and when the United Nations (UN) sends peacekeepers in
civil wars? This is an important topic for at least two normative reasons. First, it is a
necessary prerequisite for judging the extent to which the organization lives up to
its aspirations for being a truly global body, capable of working to preserve
international security and relieve suffering without preference to a state’s choice of
government, location, resources, or historical connection to the great powers.
Second, given various attempts to suggest criteria or benchmarks for humanitarian
intervention, it is important to know which cases are selected for intervention in the
absence of such criteria.
The procedures and standards of the UN provide little guidance as to the actual
decisions of the Security Council regarding when and where peacekeepers will be
deployed. Peacekeepers are deployed with reference to Chapters 6 or 7 of the UN
Charter. Although these chapters differ with regard to the use of force or pacific
means to resolve disputes, they agree that the prerequisite for their enactment is a
threat to or an endangerment of ‘‘the maintenance of international peace and
security.’’ The question remains, why does the Security Council consider some civil
wars threats to international security? The charter is silent on what constitutes a
threat to international security, and the Security Council has shown enormous
flexibility in invoking the language of threat to justify the deployment of
peacekeepers. If previous deployments provide any indication, then one must
wrestle with why civil wars in Mozambique, Somalia, Guatemala, and Sierra Leone
were deemed essential for the promotion of international security, whereas civil wars
in Kashmir, Sudan, Chechnya, and Algeria were judged as peripheral to security.
Despite the importance of the issue, the amount of systematic research on the
topic is extremely small in comparison with dozens of other topics that come under
the heading of international cooperation. Partly as a consequence, unsupported
claims by journalists, policymakers, and even some academics about where the UN
sends peacekeepers have proliferated over the years. A common assertion is that
peacekeepers go where the permanent members of the Security Council (or in
some versions where the United States) have important national interests.
Alternatively, it is claimed that peacekeeping is imperialism in disguise and,
therefore, peacekeepers are sent where great powers have an economic interest in
access to raw materials and primary commodities. A different version holds that
peacekeeping since the 1990s embodies an ethos of democracy-building and that
the great powers, who have an interest in increasing the number of democracies in
the world, choose cases in which democracy is in short supply yet has the potential
to take root.
1
The authors want to thank George Downs and James Fearon for their extensive advice and comments.
r 2003 International Studies Review.
PublishedbyBlackwellPublishing,350MainStreet,Malden,MA02148,USA,and9600GarsingtonRoad,OxfordOX42DQ,UK.
International Studies Review (2003) 5(4), 37–54
The issue of where peacekeeping missions are deployed has important policy
ramifications. For example, in General Assembly debates in 2000 over the reform of
UN peace operations, many Third World member states asserted that deployment
of peacekeeping missions reflect the interests of the Security Council and,
therefore, systematically ignore Africa and favor Europe. And even though such
member states complain about what they see as the biases of the current system,
they oppose reform of the system for fear that under new mission selection criteria
Africa would be further marginalized (Stedman 2001a). What is noteworthy is that
neither claim is based on systematic evidence or analysis.
After examining where and when the UN has sent peacekeepers, and where it
has not, we conclude that the UN acts in ways that corroborate its humanitarian and
security missions, but the pattern also shows a distinct bias toward conflict in
Europe and the Western Hemisphere. Our results show that one of the best
predictors of UN intervention is the number of deaths in a conflict, a finding that
speaks well to the UN’s mission to address costly human suffering. On the other
hand, the UN has not been evenhanded in how it responds to deaths. The UN acts
more swiftly when the deaths occur in Europe than in Africa, and it acts more
swiftly in Africa than in Asia. Furthermore, the UN seems guided by considerations
of power, cost, and risk: it responds to civil wars in weaker states, measured by the
size of the country’s army, at a much higher rate than it does to civil wars in
stronger states.
We begin by describing the extant literature on determinants of UN peacekeeping
in civil wars and showing the theoretical and methodological flaws in that
literature. Very little systematic research has been done on where and when the UN
deploys peacekeepers. Moreover, the research that exists suffers from several
problems, including selecting on the dependent variable, combining every interest
of a permanent member state into a catch-all category called the ‘‘national interests
of the Permanent 5,’’ treating legal conditions for Security Council interventions as
causal explanations for interventions, and treating deployment as an either-or
variable. We then discuss our research methodology. To overcome the problems
listed above, we used a data set of all post–Cold War civil wars and survival time
techniques to predict the duration until the UN intervenes in a given civil war along
with the hazard rate of UN intervention given various independent variables. We
then present our findings and conclude with a brief discussion of their implications
for future research and policy assessment of the UN.
Prior Research
Relatively little research has been done on the determinants of where the UN
chooses to deploy peace operations. Journalistic accounts often accuse the UN of
displaying a bias toward the interests of the United States or the Permanent 5
members of the Security Council in its decisions to intervene. For example, Phylis
Bennis (1996:84) argues:
In the real world any UN decision to intervene or any UN decision to legitimize
or endorse any country’s unilateral intervention against another country will
reflect the dominant power of the intervening side and the relative importance of
the subject nation . . . anyone who believes that the real motivation for outside
governmental military intervention (UN endorsed or otherwise) is the alleviation
of civilian hardship is suffering from a serious delusion of benevolence.
Similarly, David Gibbs (1997) hints at classic imperialistic motives behind decisions
to deploy peacekeepers, arguing that even in such seemingly unselfish interventions
as Somalia, one can easily find more grubby motivationsFsuch as access to
primary commodities.
Where Do the Peacekeepers Go?38
Although most academics do not go as far as Bennis or Gibbs and sometimes
present a more nuanced account of motivations, they also tend to emphasize the
extent to which any decision to deploy peacekeepers must partially serve the
national interests of the permanent members of the Security Council (the P-5).
Chantal De Jonge Oudraat (1996:518–519), for example, argues that the choice of
where the UN goes is determined by ‘‘the extent to which the interests of one or
more of the members of the P-5 are engaged in the case in question; and the extent
to which the conflict is believed by the P-5 to constitute a threat to international
peace and security.’’ She defines whether a case is a threat to international peace on
the basis of whether the war spills over its borders to pose a larger regional threat.
In a broader but similar vein, Laura Neack (1995) examines the extent to which
state participation in peacekeeping operations, the geographical distribution of
operations, and accounts of success or failure are the result of an idealistic
commitment to the global community or a consequence of states’ national interests.
On the basis of an analysis of eighteen operations that took place between 1948 and
1990, she concludes that the findings support a realist interpretation; that is, ‘‘states
whose interests were better served by the continuation of the status quoFthat is,
states of the advanced industrialized West and non-Western states that have enjoyed
some prestige in the international status quoFhave dominated UN peacekeeping’’
(Neack 1995:181).
Peter Jacobsen (1996) argues that national interest is not the sine qua non for
intervention that many believe, but that the power of what is known as the CNN
effect is also less than many believe. He reaches these conclusions by examining the
impact of five explanatory factors in a focused comparison of five peace
enforcement missions. He finds evidence of two distinct motivational patterns,
one of which is driven by self-interest in the form of concern about consequences of
unambiguous interstate aggression and one of which is animated by humanitarian
sentiment precipitated by massive human suffering. The two motivations operate in
different ways. States are willing to take far higher levels of casualties when national
interests are at stake than they are when only humanitarian motivations are
present. The CNN effect is real, but it only operates when there is both a high
probability of success and low probability of casualties.
Andreas Andersson (2000) also disputes the argument that the national interests
of the permanent members invariably account for where the UN chooses to go.
Noting the broad geographical distribution of interventions and the many
deployments that appear to be independent of any permanent member’s direct
interest, he argues that the Security Council has been guided primarily by a desire
to promote democracy in the world. He interprets this motive as an idealistic
interest of the permanent members related to the democratic peace hypothesis. If
democracies never go to war with other democracies, then for reasons of
international security the permanent members should desire to increase the
number of democracies in the world.
Taken in its entirety, the literature on the determinants of peacekeeping has a
number of problems. At the theoretical level, the tendency to lump every interest of
any permanent member into a category called the ‘‘national interests of the
Permanent 5’’ threatens to obscure as much as it reveals. For example, the ongoing
debate between realists and liberals has less to do with whether the permanent
members are motivated by some national interest than the specific role of security
and trade interests. More broadly, it makes a great deal of difference whether the
interest we are talking about is a security or trade interest as opposed to a
humanitarian interest.
The evidence used to establish the fact that permanent member interests are
paramount often approaches tautology. Because any permanent member can veto a
Security Council resolution and their acquiescence is a legal requirement for any
operation, it is impossible to disagree with statements like: ‘‘peacekeeping missions
MICHAEL GILLIGAN AND STEPHEN JOHN STEDMAN 39
do not get off the ground without great power support’’ (Durch 1993:36).
However, no research is necessary to support such a statement. It is true by
definition. The more important issue is which cases get great power support. We
want to know what motivates great power interest, and how, if at all, this has
evolved over time. This information will allow us to think about the extent to which
permanent member interests have distorted the role of the UN in relation to that
which was envisioned in its charter and is embodied in the statements of principles
that have been issued over the years.
Similarly, if one thinks of great power support as a supply-side explanation for
intervention, then one can examine the willingness of local parties to request or
consent to the deployment of a peacekeeping mission as a demand-side explanation
for intervention (Durch 1993:18–20). It is well established that the Security Council
is much more likely to dispatch peacekeepers to a country that has consented to a
mission; what we want to know is which states are more likely than others to request
or at least to consent to such an intrusion. Two possible demand-side explanations
deserve attention. First, stronger states may be better able to resist pressure on
them to consent to intervention or to raise the costs of possible intervention enough
to deter UN intervention. Second, the war aims of the rebels in civil wars may affect
both the willingness of a state to consent to a mission and the willingness of the UN
to deploy a mission. For example, states may be less willing to concede to a UN
intervention if rebels are attempting to secede, out of fear that the mission may
freeze a de facto partition of territory.2
The supply-side explanation may be that the
permanent members may be less willing to intervene in conflicts in which the state
sovereignty principle is threatened, given that two of the permanent members,
China and Russia, are themselves threatened by secessionist movements.3
Another theoretical limitation of the literature is that the rationales behind
various hypotheses tend to be less well-developed than elsewhere in the
international relations literature, often raising as many questions as they answer.
Take, for example, Andersson’s (2000) provocative thesis that peacekeeping
decisions reflect a desire on the part of the permanent members to spread
democracy. Although it is reasonable to assume that this is a priority for democratic
states, it is far from clear why a desire to spread democracy should lead to a higher
priority being placed on intervention in nondemocratic states than on intervention
in existing democracies that are threatened with destruction. After all, success in the
former seems far less likely than in the latter, which involves trying to sustain
something that already exists. Although Andersson doubtless has reasons for
believing why an emphasis on promoting democracy would focus on the bird in the
bush rather than the one in the hand, he never makes those reasons explicit. As a
consequence the reader cannot help but wonder whether Andersson’s positive
results stem less from a preoccupation on the part of the permanent members to
convert states to democracy than from the fact that most civil wars take place in
nondemocratic states. It is also unclear why China, a nondemocracy, would, as a
permanent member of the Security Council, accede to interventions aimed
primarily at promoting democracy.
A lack of theoretical development is also evident in the literature that tries to
determine which of a number of models best explains, on the basis of self-interest
versus humanitarian-idealistic motives, where the UN intervenes. A single indicator
often represents an entire perspective, and authors have a tendency to assume that
because a given indicator is central to a particular perspective, any evidence that the
variable is important demonstrates automatically the validity of that model and the
weakness of rival models.
2
We thank David Laitin and James Fearon for suggesting that war aims may influence deployment of missions
and for providing us with the coding of this variable from their own data set on civil wars.
3
We thank Andrew Mack for this suggestion.
Where Do the Peacekeepers Go?40
The difficulty is that even though an indicator may be especially central to one
model, it may also be correlated with a variable that plays an important role in
another model or is actually included in a rival model by virtue of its playing a very
different function. As an example of the first problem, ‘‘massive human suffering’’
is a variable that is often associated exclusively with a humanitarian or idealist
model of motivation, and the connection makes a lot of sense. If eliminating or
reducing massive human suffering is not the central humanitarian goal, what is?
The difficulty is that massive human suffering, although not a primary realist goal,
tends to be correlated with the scale of conflict, which is a variable that a realist
strategist potentially cares about a great deal.4
The presence or absence of a peace treaty provides another example of how the
same variable can play different functions in different theories. From a realist
perspective, the existence of a treaty is important because it signals the conditions
under which a compromise among the conflicting interests of the permanent
members is more likely than it would be otherwise. This factor was especially
important during the Cold War, when the absence of a treaty would likely have
suggested that either the allies of the Eastern or the Western bloc believed they
currently held the advantage. For the realist, the existence of a treaty is also
important because it indicates that the intervention will be more successful and
involve lower costs than a prospective mission where there is no treaty. For the
idealist, the existence of a treaty makes a mission more attractive because it allows
the UN to function as a rule-driven and unbiased peacemaker that is defending the
norm of neutrality in an increasingly partisan world.
These interpretations raise the prospect that from an empirical standpoint
distinguishing between the idealist and realist explanations will not be as easy as it
first appears. Things can be clarified somewhat if we can find a more precise
measure of the scale of conflict than massive suffering and use this in the model, but
multicollinearity is still likely to create interpretation problems. If no separate and
reliable indicator of scale of conflict apart from massive human suffering is
available, then the same indicator will have to be used in both model specifications.
The treaty variable must be used in both.
In such cases, it is important to identify secondary and tertiary attributes of the
two models that overlap less. This task is not easy given that the variety of reliable
data collected for all, or even most, peacekeeping missions is modest. However,
sometimes it is possible to expand the set of model indicators beyond what is
normally employed in such a way that more clearly differentiates among the
different approaches. For example, even though considerable overlap exists
between the realist and humanitarian models with respect to the impact that
massive suffering will haveFif for quite different reasonsFtheir predictions with
respect to the effect of the conflict location are more distinct. For a realist, the
location of a conflict plays a critical role in determining its importance because it
plays a significant role in determining the degree of threat the conflict poses to a
state’s vital interests. This relationship stands in sharp contrast to the humanitarian
or idealist model in which the location of the conflict should be largely irrelevant.
But again even here modesty is required. A finding that the UN is more willing to
intervene in Europe than in Africa may be a function of heightened international
media attention and, therefore, greater demands for humanitarian action for
conflicts in Europe.
The literature on determinants of peacekeeping also tends to suffer from several
methodological problems. By far the chief among these is a tendency to select cases
on the basis of the dependent variable and, by so doing, to restrict the sample to
4
Thus, De Jonge Oudraat (1996:519–522) uses a scale of conflict as a proxy for Security Council judgment about
whether a civil war is a threat to international peace and security.
MICHAEL GILLIGAN AND STEPHEN JOHN STEDMAN 41
peacekeeping operations that the UN has chosen to undertake. Characteristics that
these cases share in common are then argued to be the keys to understanding UN
motivation. Cases of civil wars or interstate aggression in which the UN has chosen
not to intervene are simply ignored, so the reader has no idea whether they too
possess these same characteristicsFwhich would invalidate the inferences about
what is responsible for the decision to intervene. Given that it is obviously
important to know whether this is a problem or not, expanding the cases to include
all situations in which the UN could have potentially intervened is critical. Doing so
would also solve the related problem that the number of cases used in many of the
analyses is often far too small to support a reliable conclusion in a statistical sense.
A subtler problem has to do with how one measures interventionFthe
dependent variable. Almost invariably, when this is dealt with at all (recall most
authors restrict their sample to interventions), it is dealt with as a dummy variable.
Either the UN has initiated a peacekeeping mission in a given situation or it has not.
However, it is important to know not only whether a peacekeeping mission has
been initiated but also when it has been initiated. Everything else being equal,
prompt intervention not only has a greater potential to save lives, but the timing of
interventions can potentially tell us much more about what motivates and
constrains the UN than can the analysis of where intervention has taken place
and where it has not.
Research Approach
In this piece, we do not propose to test alternative theories of UN intervention. As
we state above, there are powerful limitations on spelling out and testing theories of
intervention that derive from the larger traditions of international relations
theories. Rather, as a first step to understanding UN intervention, we hope to be
able to provide some empirical findings on intervention that will allow us to
evaluate claims about where and when the UN intervenes.
To do so we examine the set of post–Cold War civil warsFcivil wars that began
after 1988 or those that began before 1988 but were still active after 1988. We have
strong reasons for believing that pooling Cold War and post–Cold War samples
would be inappropriate. During the Cold War, when there were sixty-three civil
wars, the UN undertook only four missions in civil wars: two in Cyprus, one in
Korea, and one in the Congo. In the post–Cold War era, there were sixty civil wars,
and the UN launched nineteen missions. Clearly, a five-fold increase in the rate at
which the UN has allocated missions to civil wars suggests that the mission selection
procedures have changed dramatically with the end of the Cold War. In our results
below, we provide some evidence that our choice to concentrate on the post–Cold
War era was the appropriate one.
We chose to examine the following independent variables associated with the
various claims about UN deployment of missions. The variable ‘‘Deaths’’ is military
and civilian deaths from the conflict in millions. We use it as a measure of
humanitarian need for an intervention in the conflict. ‘‘Treaty’’ is a dichotomous
dummy variable that equals one if the combatants had signed a peace treaty and
equals zero otherwise. We use this variable to control for the possibility that the UN
is more likely to intervene in a conflict if the combatants have signed a peace treaty.
‘‘Cold War’’ is a dichotomous dummy variable that equals one if the civil war in
question began before 1989 and equals zero otherwise. It is used to control for the
substantially lesser rate of UN involvement during the Cold War. We also include
‘‘Primary commodity exports’’ in millions of Summers and Heston real GDP units
to test if the UN intervenes more in countries that are important sources of raw
materials. To test the claim that the UN is more likely to send missions to countries
that are democracies, we use a dummy variable that equals one if the civil war
occurred in a ‘‘Democracy,’’ as specified in the Alvarez et al. (1997) data set, and
Where Do the Peacekeepers Go?42
zero otherwise.5
We also test if the UN intervenes more in countries with which the
permanent members have close ties, using a dummy variable that equals one if the
civil war occurred in a country that is a ‘‘Former colony’’ of a permanent member of
the Security Council and zero otherwise. To control for any regional differences in
the way the UN allocates missions, we include dummy variables for four regions:
‘‘Latin America and the Caribbean,’’ ‘‘Europe,’’ ‘‘Asia,’’ and the ‘‘Middle East.’’
Africa is the excluded region (and, therefore, the baseline case). Prewar
‘‘Population’’ (in millions) and ‘‘Size of government army’’ (in thousands of troops)
are included to see if the UN is more likely to avoid conflicts in large, militarily
powerful countries. Finally we wanted to examine if the rebel’s war aims affected
the likelihood that the UN would intervene. Therefore, we included ‘‘Rebels’ Aims:
Control of Government,’’ a dummy variable that is coded one if the civil war was a
contest over control of the country’s central government and zero otherwise, and
‘‘Rebels’ Aims: Autonomy,’’ a dummy variable that is coded one if the rebels were
fighting to achieve national autonomy and zero otherwise. A third variable that is
not reported below was coded one if the rebels have a mixture of the two motives
and zero otherwise, and it serves as our residual category.6
Unless otherwise noted,
we use the same data used by Michael Doyle and Nicholas Sambanis (2000), which
is available on their World Bank web site (www.worldbank.org/research/conflict/
papers/peacebuilding/index.htm). We have made some minor changes to the data
(which are discussed in the appendix to Gilligan and Stedman 2003).
We address our question of where peacekeepers are sent by employing survival
time techniques. In other words, our dependent variable is the time elapsed
(measured in years) until the UN intervenes in a given civil war. Transformation of
the Doyle and Sambanis data into survival time data was relatively straightforward
with a few exceptions, which we describe in Section 4 of the appendix in Gilligan
and Stedman (2003). As a further check for robustness, we also estimate our results
using logit on a dichotomous dependent variable equal to one if the UN sent an
observer, traditional peacekeeping, or enforcement mission to the conflict, and zero
otherwise. We also estimate ordered probit on a categorical dummy variable that is
equal to one if the UN sent an observer mission, two if the UN sent a traditional
peacekeeping mission, three if it sent an enforcement mission, and zero otherwise.
As a final robustness check, we estimate a tobit model using the cost of the UN
operations in millions of US dollars. Tobit is appropriate in this case because
operation cost is bounded by zero: the UN could not spend less than zero dollars on
the mission. Summary statistics for these variables and the independent variables
described above are reported in Table 2 of Gilligan and Stedman (2003).
We regard survival time techniques to be the most appropriate for our purposes
because the data are right censored. Most of the wars in our sample ended before
the UN intervened. We do not know what the UN would have done had those wars
persisted. Logit or probit estimates with right censoring overestimate standard
errors, which might cause us to reject inappropriately some of the hypotheses we
discussed in the previous section. Besides, such methods throw away valuable
information present in the data. Earlier intervention in a conflict suggests that the
UN was more concerned about that conflict for some reasonFinformation that is
useful to us as we seek to uncover the determinants of UN intervention (Peterson
1991; Box-Steffensmeier and Jones 1997).
In theory at least, survival time techniques are able to give us the same
information that we could retrieve from logit or probit analysis on noncensored
data, given that duration data are just an aggregation of binary data over time (Alt,
King, and Signorino 2000). In addition to predicting the duration until the UN
5
These data have been updated through 1999. See also Przeworski et al. (2000).
6
We thank James Fearon and David Laitin for sharing these measures with us. See Fearon and Laitin (2003).
MICHAEL GILLIGAN AND STEPHEN JOHN STEDMAN 43
intervenes in a given civil war, survival time analysis allows us to predict the hazard
rate of UN intervention, conditional on our independent variables. This rate is the
instantaneous rate at which the UN intervenes in a conflict given that it has not
done so up to that point. Furthermore, like logit and probit, survival time
techniques allow us to generate a predicted probability that the UN will intervene in
a conflict given that it has not done so up to that point. Both of these values give us
precisely the information that we want given that we are interested in determining
which types of conflicts the UN is more prone to help pacify. The coefficients that
we present in the tables below represent the change in the hazard rate for a oneunit
change in the respective independent variable. Because hazard rates are
sometimes difficult to interpret, we also report expected probabilities of UN
intervention in a few examples.
Findings
The findings of our survival time analysis are presented in Tables 1 and 2. In most
cases, we estimate the model on the post–Cold War sample of civil wars, but in
columns 4 and 5 of Table 2 we estimated the model on the sample of 120 post–
World War II civil wars to check the robustness of our results with respect to choice
of sample. In Table 3 we check the robustness of survival time results with respect to
other possible choices of the dependent variable and estimation techniques. Logit
results using the dichotomous dependent variable are reported in columns one and
two, ordered probit estimates using the categorical dependent variable are
reported in columns three and four, and tobit estimates using the cost of the
operation are reported in columns five and six. We do not regard the results in
Table 3 to be as trustworthy as those in Tables 1 and 2 because of the censoring
problem mentioned above, but we present them as a robustness check. The results
in Tables 1 through 3 point to ten broad findings, which we summarize below.
FINDING 1: The more severe a conflict, measured by the number of deaths,
the more likely the United Nations is to intervene.
This was the most robust result in our analysis. The coefficient on this variable
was significant at the 5 percent level or better in a one-tailed test in all specifications.
Figure 1 graphs the probability that the UN will intervene in a conflict in a
particular year given that it has not intervened in that conflict up to that point. It
graphs this function for several of the independent variables of interest.7
For the
moment we are concerned with the effects of deaths. To see the effects of deaths on
these probabilities compare the line marked ‘‘baseline’’ with the line marked ‘‘hi
deaths.’’ The ‘‘hi deaths’’ case assumes one standard deviation more deaths
(roughly 630,000). The probability of UN intervention in the tenth year in the
baseline case is only about 1 percent whereas the probability of UN intervention in
the high deaths case is by that point almost 50 percent.
We also can examine the effects of deaths on the duration until the UN
intervenes in a conflict. Take the case of Cambodia. The UN intervened in that civil
war after 13 years. Our model predicts a UN intervention after 13.1 yearsFvery
close to the actual value. If the number of casualties in that war had been one
million rather than three million, our model predicts that the UN would have
intervened after 130 years, quite a dramatic difference. For all practical purposes,
the model is telling us the UN would not intervene in this conflict with that level of
casualties.
7
To generate the predicted probabilities in Figure 1, we used the specification in column 1 of Table 2. The
baseline case assumes the average number for each of the continuous variablesFthat is a value of 0.182, or 182,000
deaths, and of 246, or 246,000, for the size of the government army. It assumes a value of zero for each of the
dummy variablesFthat is a civil war in an African country that began after the end of the Cold War.
Where Do the Peacekeepers Go?44
It is perhaps worth dwelling on one potential threat to the validity of this finding.
We only have data for the total number of deaths in each war, so our measure
cannot distinguish between pre-intervention deaths and postintervention casualties,
the latter of which obviously could not have affected the UN’s decision to
intervene. The two most prominent cases in which a large number of deaths
TABLE 1. Determinants of UN Intervention in Civil Wars, Weibull Regression Estimates
1 2 3 4 5 6
Deaths 2.0295 3.0852 2.9434 2.9420 2.2540 2.1893
(1.2216) (1.1296) (1.1416) (1.1500) (1.0600) (1.0248)
Treaty 1.1158 1.1875 1.1221 1.1716 1.2519 1.5417
(1.0756) (0.9624) (0.9725) (0.9669) (1.0262) (0.9895)
Cold War À4.1432 À3.7221 À3.4533 À3.4445 À3.1813 À3.3444
(1.1087) (1.0483) (1.0029) (1.0008) (0.8924) (0.8882)
Asia À3.4439 À7.9059 À7.6614 À7.6519 À5.8119 À5.4776
(4.2757) (3.1674) (3.2350) (3.2628) (3.0038) (2.8878)
Europe 4.1256 4.8936 4.7730 4.4492 2.9077 2.7158
(1.9828) (1.9074) (1.8873) (1.4365) (1.0290) (1.0167)
Middle East 2.3387 3.4074 3.1866 3.0535 1.6107 1.7654
(1.6294) (1.3306) (1.2865) (1.2144) (1.1433) (1.1376)
Latin America & 2.0649 3.2937 3.3198 3.3213 2.8875 2.4878
Caribbean (1.3167) (1.0447) (1.0903) (1.0748) (1.0143) (0.8818)
Democracy À1.2015 À1.3850 À1.5730 À1.5717 À1.0295 —
(1.2516) (1.2080) (1.2101) (1.2086) (1.1667)
Size of Government
Army
À0.0112 À0.0232 À0.0222 À0.0222 À0.0249 À0.0214
(0.0145) (0.0136) (0.0140) (0.0139) (0.0119) (0.0108)
Primary
Commodity
Exports
À0.0007 À0.0006 À0.0006 À0.0006 — —
(0.0004) (0.0004) (0.0004) (0.0004)
Former colony of À1.5961 — — — — —
P5 member (1.3798)
Rebels Aims: À1.3225 À0.9511 — — — —
Control of
Government
(0.9238) (0.8549)
Rebel’s Aims: À1.6395 À1.3248 À0.3590 — — —
Autonomy (1.6158) (1.6114) (1.3462)
Constant À3.8903 À4.9204 À5.5640 À5.6256 À5.0674 À5.3625
(1.8199) (1.5822) (1.4699) (1.4656) (1.3922) (1.4037)
pn
2.9483 2.9750 2.8496 2.8553 2.4587 2.4520
(0.5953) (0.6088) (0.5716) (0.5714) (0.4787) (0.4804)
n 59 59 59 59 59 59
LL À18.73 À19.37 À19.9 À19.9362 À23.1385 À23.5955
n
This parameter affects the shape of the Weibull distribution. A value significantly greater than one implies that the
hazard rate of UN interventions increases over time. (Standard errors are in parentheses; coefficients that are
significant at 5 percent are in bold).
MICHAEL GILLIGAN AND STEPHEN JOHN STEDMAN 45
occurred after the UN intervened are Bosnia and Rwanda, but other possible cases
might include Somalia, Liberia, and Angola. To address this issue we re-estimated
the results excluding these cases. Those results, which are presented in column 6 of
Table 2, are significantly stronger, suggesting that this finding is not an artifact of
cases in which high casualties occurred after the UN intervened.
TABLE 2. Determinants of UN Intervention in Civil Wars, Weibull Regression Estimates
1 2 3 4 5 6
Deaths 2.3675 2.7676 1.7284 2.9039 1.8888 6.2470
(0.9936) (1.0993) (0.7460) (1.4194) (0.9756) (3.2167)
Treaty — — 1.5191 0.2385 0.4645 —
(0.6357) (0.6928) (0.6132)
Cold War À2.8052 À2.9380 À3.4496 À2.0870 À2.1049 À5.1880
(0.7795) (0.7965) (0.8110) (0.6317) (0.5886) (1.5175)
Asia À6.0691 À6.9939 À3.2757 À7.6499 À5.1590 À15.880
(2.8528) (3.1297) (2.1334) (4.5614) (2.7107) (8.7279)
Europe 1.9144 1.7591 0.5912 2.0191 2.4252 2.7533
(0.9073) (0.9098) (0.8211) (1.0715) (0.7691) (1.3742)
Middle East — — — 0.2458 0.7338 —
(1.4409) (0.9450)
Latin America & 2.4800 2.4072 1.3002 2.7922 2.2718 4.4511
Caribbean (0.8300) (0.8719) (0.6351) (1.1668) (0.8033) (1.3029)
Democracy — — — À1.8079 À1.1694 —
(1.1196) (1.0692)
Size of À0.0216 À0.0198 — À0.0211 À0.0219 À0.0319
Government Army (0.0099) (0.0111) (0.0144) (0.0099) (0.0147)
Primary Commodity — — — À0.0008 — —
Exports (0.0005)
Former Colony — — — À0.2393 — —
of P5 member (1.0005)
Rebels Aims: — — — À0.7264 — —
Control of Government (0.8649)
Rebels Aims: — — — 1.5951 — —
Autonomy (1.2925)
Population — À0.0879 — — — —
(0.0853)
Constant À3.5994 À3.1013 À4.9693 À3.1404 À3.4420 À5.7617
(0.8704) (0.9551) (0.8928) (1.6058) (0.8825) (1.4540)
pn
2.1888 2.1585 1.9187 2.2101 1.8564 3.3079
(0.4141) (0.4093) (0.3699) (0.3980) (0.3187) (0.7671)
n 59 59 60 120 120 51
LL À25.263 À24.4807 À32.547 À31.1902 À36.9215 À11.0152
n
This parameter affects the shape of the Weibull distribution. A value significantly greater than one implies that the
hazard rate of UN interventions increases over time. (Standard errors are in parentheses; coefficients that are
significant at 5 percent are in bold).
Where Do the Peacekeepers Go?46
TABLE 3. Logit, Ordered Probit, and Tobit Estimates of Determinants of UN Intervention in Civil Wars
1 2 3 4 5 6
Dependent
Variable
Intervention Intervention Level of
Intervention
Level of
Intervention
Cost of
Operation
Cost of
Operation
Estimator Logit Logit Ordered
Probit
Ordered
Probit
Tobit Tobit
Deaths 14.7701 8.5798 1.5856 1.7124 2115.44 2121.297
(6.1055) (3.7189) (0.7237) (0.7703) (961.5721) (1019.151)
Treaty 1.4118 — 0.4290 0.3944 342.6481 245.3126
(1.3953) (0.6983) (0.8138) (520.8233) (589.735)
Cold War À4.5214 À1.8528 À0.3404 À0.3106 À69.2868 À389.539
(1.5777) (1.2405) (0.4101) (0.3892) (429.8699) (464.526)
Asia À27.8552 À17.1798 À3.5112 À3.9823 À4197.08 À4261.95
(12.9734) (7.1102) (2.1245) (1.9616) (3100.257) (2877.673)
Europe 3.5594 0.5786 1.2131 1.1333 2592.16 1511.751
(2.5702) (1.3041) (0.8618) (0.8089) (1023.153) (618.5604)
Middle East 1.6555 À0.4342 À0.5045 357.9556 64.83039
(1.8512) (0.7457) (0.8389) (932.1456) (700.0333)
Latin
America &
18.8602 8.8585 1.7923 1.9522 1165.335 1470.08
Caribbean
(7.0047) (2.6421) (0.7749) (0.6904) (736.6521) (683.0769)
Democracy —n
—n
À1.2409 À1.3050 À1303.89 À1027.35
(0.8021) (0.7662) (751.5265) (876.7648)
Size of À0.0549 À0.0455 À0.0091 À0.0099 À8.10424 À12.2676
Government
Army
(0.0374) (0.0165) (0.0039) (0.0036) (6.139003) (7.483431)
Primary À0.0015 — 0.0000 — À0.7572 —
Commodity
Exports
(0.0010) (0.0001) (0.326267)
Former Colony À2.3804 — À0.1619 — À272.548 —
of P5 Member (1.9404) (0.5958) (753.9577)
Rebel’s Aims: À1.2668 — À0.1983 — À712.179 —
Government
Control
(2.3748) (1.1624) (577.3001)
Rebel’s Aims: À3.2666 — À0.3424 — À678.274 —
Autonomy (3.2571) (1.3228) (946.6207)
Cutpoint 1nn
4.7983 1.5479 À0.6022 À0.3371 742.296 À201.284
(3.7305) (1.0762) (1.5849) (0.9020) (901.017) (723.204)
Cutpoint 2 — — 0.0705 0.3295 — —
(1.6861) (0.9899)
Cutpoint 3 — — 1.6848 1.9434 — —
(1.8894) (1.2534)
n 59 59 59 59 59 59
LL À10.2319 À13.6880 À34.7824 À34.9473 À159.05 À163.35
n
The democracy variable is excluded from this specification because it would not converge with the variable
included.
nn
In the Logit and Tobit results in columns one, two, three and four, ‘‘Cutpoint 1’’ is actually the constant.
MICHAEL GILLIGAN AND STEPHEN JOHN STEDMAN 47
FINDING 2: The United Nations is significantly less likely to intervene in civil
wars in countries with large government armies.
This was the strongest result in our analysis, in the sense that the size of
government army variable produced the largest changes in the probability of a UN
intervention. This result is illustrated in Figure 1. The line marked with squares
represents the ‘‘small army’’ case. Instead of an army of 246,000 assumed in the
baseline case, we assumed an army of only 100,000. The standard error of the
government army variable is approximately 330,000 so this is a change of less than
one-half of a standard error from the baseline case. The effects of this relatively
small change are profound. The probability that the UN would intervene in the
conflict in the ninth year given that it had not intervened up to that point is virtually
one whereas the probability that the UN would intervene in the baseline case in that
year is only about 1 percent. We interpret these results to mean that strong states
are less likely to welcome UN intervention in their internal affairs, and the UN, for
reasons of cost and risk, is much less willing to send troops to a country with large
armed forces without that country’s permission.
This result is not simply due to the fact the UN avoids missions in large countries,
and large countries happen to have large armies. In column 2 of Table 2 we
included a measure of each countries’ population in millions. The coefficient is
negative, suggesting that the UN has a slight bias against large countries, but it is
not at all significant. Meanwhile, the coefficient on size of government army and its
statistical significance changed little as a result of the inclusion of the population
variable. The UN seems to care more about the military might than the size of the
countries in which it intervenes.
FINDING 3: The probability of a UN intervention in a given war increases as
the war drags on.
This finding is shown by the estimated value of the shape parameter p. A value of
that parameter greater than one implies that UN interventions occur with an
increasing hazard rate. The parameter in all our post–Cold War sample results is
greater than two and significantly different from one at high levels. For example
FIG. 1. Probability of UN Intervention under Various Conditions.
Where Do the Peacekeepers Go?48
the estimate of p in our baseline specification in column 1 of Table 2 implies that the
UN has roughly a twenty-five times higher rate of intervening in a war after ten
years than after one year, other things being equal. Obviously, this result is also
illustrated by the fact that all the lines in Figure 1 are monotonically increasing.
FINDING 4: There is no evidence that the United Nations intervenes in
secessionist conflicts at a different rate than it intervenes in attempts to take
over control of the government.
In column 1 of Table 1, the coefficients on the two variables ‘‘Rebels’ Aims: Control
of Government’’ and ‘‘Rebels’ Aims: Autonomy’’ are very similar in magnitude, and
neither is significantly different from zero. The UN does not seem to intervene in
secessionist movements at a higher or lower rate than in civil wars aimed at control of
the central government. The two coefficients continue to be at very low statistical
significance in column 2 for which the insignificant variableF‘‘Former colony of a
permanent member’’Fis removed. In column 3, for which we dropped ‘‘Rebels’
Aims: Control of Government’’ and used only ‘‘Rebels’ Aims Autonomy,’’ the
coefficient on the latter still has very low significance levels. These results suggest that
we can be quite confident that no strong relationship exists between the war aims of
the rebels and the likelihood that the UN will intervene in a civil war.
FINDING 5: There is evidence of regional bias in the UN’s selection of
missions, but the worst bias is against Asia, not Africa.
Africa is the residual category among our regional dummy variables; so if there
were systematic bias against missions to Africa, we would expect the signs on all the
regional dummy variables to be negative. Three of them are. The UN is
consistently more likely to allocate missions to Europe and Latin America and
the Caribbean than it is to Africa. The coefficient on the Middle East dummy is
positive but not consistently significant across specifications. The coefficient on the
Asia dummy, however, is negative and highly significant. Civil wars in Asia are the
least likely of all the regions to receive UN help. For example, the civil war in
Cambodia lasted thirteen years before the UN intervened. Our model predicts 13.1
years. Our estimates on the regional dummy variables tell us that the UN would
have intervened in an identical war in Africa in less than a year (approximately 0.8
years), after about four months (roughly one-third of a year) if the war had been in
Europe, and after only about three months (roughly one-quarter of a year) if the
war had been in Latin America.
These regional effects are illustrated in Figure 1. Recall that the baseline case
corresponds to a conflict in Africa. The line with the asterisk markers graphs the
probability of a UN intervention in the given year for a European conflict with all
other variables at the same values as those for the baseline case. Notice that the
probability of a UN intervention in the fifteenth year of a conflict, given that one
has not occurred until that point, is less than 5 percent in the baseline (African) case,
whereas it is about 15 percent in the European case. Conflicts in Latin America and
the Caribbean are even more likely to receive a mission than those in Europe,
ceteris paribus, as shown by the line marked with circles. The Asia case is very
difficult to see from Figure 1 because the probability of UN intervention remains
very low (virtually indistinguishable from the horizontal axis) for the whole range of
years. These three lines illustrate the rather striking differences in the propensity of
the UN to intervene in each of the four regions.
FINDING 6: There is no strong evidence that the United Nations is more
likely to intervene in a conflict when the combatants have negotiated a peace
treaty, but this is probably due to multicollinearity.
MICHAEL GILLIGAN AND STEPHEN JOHN STEDMAN 49
This result is quite a surprise given the conventional wisdom that the UN requires
a peace treaty before it will send in a peacekeeping mission. The variable is the
proper sign in all specifications but simply not significant in any of them. A look at a
contingency table like the one in Table 4 reveals that there is a significant bivariate
relationship between the presence of a treaty and whether or not there was an
intervention. A bivariate Weibull regression using only the treaty variable (not
shown) also reveals a strong bivariate relationship in the right direction. The
problem arises in the multivariate results. The culprit appears to be multicollinearity
between the treaty and size of government army variable. The
correlation coefficient between treaty and size of government army is À0.23.
When size of government army is excluded from the equation, as we do in column
3 of Table 2, the standard error on the treaty coefficient falls by about a third,
making the coefficient statistically significant. It would be premature to claim that
the UN does not react positively to a negotiated peace treaty. These relatively weak
results of the treaty variable are due to correlation between treaty and size of
government army and a small sample size.
FINDING 7: There is no evidence that the United Nations intervenes more in
countries with high primary commodity exports.
In fact, the coefficient on primary commodity exports was the wrong sign.
Clearly we can reject the hypothesis that the UN is intervening in some
neocolonialist fashion because of permanent member interests in securing steady
flows of raw materials.
FINDING 8: There is no strong evidence that the United Nations intervenes in
democracies at a lower rate than it does in nondemocracies.
The coefficient on ‘‘Democracy’’ is negative, suggesting that the UN has a lower
hazard rate for intervening in democracies, but the coefficient is significant in only
one of the specifications in Table 1. A different measure of democracy, a five-year
average of Ted Gurr’s democracy measure, produced even weaker results, so to
save space we do not report them. These weak results, combined with the logical
flaws inherent in the argument that produced this hypothesis, make us confident
that we can reject the claim that the UN sends fewer missions to democracies.
FINDING 9: There is no evidence that the United Nations intervenes in
former colonies of permanent members of the Security Council at a higher
rate than it does in other areas.
In fact the coefficient on ‘‘Former permanent member colony’’ is negative in the
specification in column 1 of Table 1. Because of its wrong sign and insignificance,
we remove it from the analysis in the remaining specifications.
FINDING 10: The end of the Cold War had a profound effect on the UN’s
propensity to conduct peacekeeping missions.
Obviously this result is no surprise. The coefficient on the ‘‘Cold War’’ variable
tells us that conflicts that began during the Cold War had a significantly lower
TABLE 4. Bivariate Relationship between Treaty and Intervention
Treaty No Treaty
Intervention 16 3
No Intervention 11 30
Where Do the Peacekeepers Go?50
hazard rate of UN intervention. The variable was highly significant in all of the
specifications in Tables 1 and 2. Our estimates tell us that it would have been
extraordinarily unlikely for a conflict like the one in Cambodia, with three million
deaths, to last as long without UN intervention if it began in the post–Cold War era.
Had that civil war begun in the post–Cold War era, our model predicts that the UN
would have intervened in only 3.7 years rather than in the 13 years it actually took.
This result is also shown in Figure 1. The line marked with triangles assumes the
same variable values as the baseline case, except that the conflict began during the
Cold War. The line follows the horizontal axis very closely showing that the
probability that the UN would intervene in a conflict that began before the end of
the Cold War was lower in any given year than a conflict that began after the end of
the Cold War.
Discussion
The results that we report here suggest an image of a UN that attempts to balance
between dictates of power and concerns of principle. With regard to the latter, the
UN seems to respond, other things being equal, to civil wars that involve the
greatest human catastrophes. The number of deaths that a conflict produced was
the most robust predictor of UN intervention across all our specifications. In
making its decisions, however, the organization responds to basic power
considerationsFboth in terms of the importance that the Security Council places
on different regions and in its reluctance to intervene in states that have greater
lethal capacity. The result is that the UN has not been evenhanded in how it
responds to deaths. More people must die in Africa than in Europe, and even more
must die in Asia than in Africa, before the UN will intervene. Furthermore, the UN
does not intervene at the same rate in countries with large government armies. Our
estimates indicate that civil wars in countries with just average size armies have very
low probabilities of receiving a UN mission, even after several decades of conflict.
Our findings suggest two further research questions: (1) What are the causes of
regional bias in UN interventions, and (2) is there a differential quality of care
among those cases in which the Security Council intervenes?
Regional Bias
While it is understandable that the UN is loathe to become involved in civil wars in
relatively powerful countries (measured by size of army), an interesting area of
future research would be to try to understand why the regional differences exist.
We can speculate that both supply- and demand-side reasons help explain the
discrepancy. On the supply side, several answers suggest themselves. First, the
permanent members, especially the United States, France, and the United
Kingdom likely see wars in Europe, as opposed to wars in Africa, as more
important to their vital security interests. Indeed, the single variable that clearly
differentiates between explanations based on security interests of the permanent
members as opposed to humanitarian interests is the location of the conflict. Just as
plausibly, one can posit that these same members act in response to their own public
opinion, and domestic demands for intervention are likely to be stronger when the
wars are in Europe. This result may be attributable to the CNN effect or to the
greater presence of immigrants from those countries, who then lobby for
intervention. Or it could be that wars in Europe produce more asylum seekers to
the major European powers, and hence a greater sense of urgency to terminate the
war.
All these explanations sound plausible for the results comparing intervention in
Europe as opposed to Africa, but they seem problematic for understanding the Asia
result. Given the rising prominence of that region for the security interests of the
MICHAEL GILLIGAN AND STEPHEN JOHN STEDMAN 51
permanent members, as well as the increase in immigration from Asian countries to
the United States and Europe, one would predict more interventions in Asia than
Africa. It is possible that the answer lies in the demand side for intervention. During
the Cold War, consent of warring states was viewed as a prerequisite for the
deployment of a mission, and in the 1990s consent remained a key consideration in
intervention decisions. One way of addressing the problem of regional bias would
be to explore why Asian states that suffer from civil wars are less likely to provide
consent for intervention than states in Africa in similar conditions. Our results
suggest that a possible explanation for the greater proclivity of African states to
consent to intervention may be their greater inability to resist pressures for
intervention and to claim the prerogatives of sovereignty. A different answer might
focus on regional norms and institutions. Whereas the Organization of African
Unity gradually changed its stance in the 1990s toward conditional acceptance of
interference in state sovereignty to encourage conflict resolution and war
termination, the Association of Southeast Asian Nations (ASEAN) and states in
the larger Asian region have consistently resisted the suggestion that humanitarian
principles take precedence over state sovereignty and noninterference (Stedman
2001b).
Quality of Care
Understanding where the UN intervenes in civil wars is only a first step in
evaluating whether the organization lives up to its universal aspirations. A second
question is whether the UN provides an equitable quality of care to the countries in
which it intervenes. Whereas the first question is akin to evaluations of public health
systems that focus on who is admitted for care, the second question is akin to
evaluations of such systems in terms of the quality of care, once people are admitted
into the system.
The results we have discussed so far suggest the patterns by which the UN
allocates missions to civil conflicts. As a measure of the quality of the mission, we use
the cost of the UN operation in millions of dollars. Table 5 reports summary
statistics for this measure for all nineteen interventions in the postwar period and in
various regional subsamples. The average cost of a mission in Europe is
significantly higher than the cost of a mission in Africa, Latin America, and the
Caribbean, or the Middle East. Interestingly the one mission that the UN did
conduct in Asia was also quite expensive.
The specifications in columns 3 and 4 of Table 3 use cost of operation as the
dependent variable.8
The tobit coefficients confirm some of the more important
patterns we discussed above. As before, the UN seems to react strongly to the
number of casualties. The UN spent more resources, other things being equal, on
civil wars that had large numbers of casualties. The bias in favor of missions to
Europe continues to be evident. The strong bias against Asia was no longer evident
in these specifications, probably because the one mission that was sent to Asia was
also quite expensive. Size of government army was less significant in these results,
although using a log of government army (results not reported) did produce
significant results.
Although these findings are preliminary, they suggest the need for further
investigation into the question of quality of care in UN interventions. On the one
8
Again we need to interpret these results with caution because the data exhibit right censoring. That is, we have
no measure of what the UN would have spent on missions in wars that ended without an intervention had those
wars continued. Furthermore, the data in this case are also left censored, not in time, but in the different sense that
the UN could not spend less than zero dollars on a mission. As such, all missions in which the UN did not intervene
take on the same value even though some may have been less a priority for the organization than others. To combat
this latter form of censoring, we estimated the results using tobit, which controls for the left censoring problem.
Where Do the Peacekeepers Go?52
hand, using UN resources as a proxy for quality is misleading, as each of the cases
of intervention receive supplementary resources from regional organizations and
member states. In the case of Bosnia, for example, the resources provided directly
by the United States and the European Union dwarf the funds directly allocated by
the UN. On the other hand, the fact that the European cases are much more likely
to receive dramatic supplements by member states should raise greater concern
about discrepant levels of UN funding by region. In either case, such an
investigation would have to address the possibility that missions in Africa and
Latin America and the Caribbean cost less simply because the missions in those
regions required fewer resources.
Conclusion
By examining where and when the UN sends peace missions, our research
provides important evidence to gauge whether the institution is fulfilling its
humanitarian and security missions, and what biases it shows in attempting to live
up to its mandate. The results contain good news and bad news. The UN is
probably more sensitive to its humanitarian mission than it is given credit for, but its
behavior confirms the claims by many critics inside and outside the organization
that it shows distinct biases toward different regions and regimes.
Our findings present fundamental problems for an institution that has pledged
to uphold universal norms and especially for its current secretary general, Kofi
Annan, who argues that humanitarian intervention to stop horrific atrocities and
abuses, whether they be the result of civil war or governmental policy, should be a
universal imperative. The results suggest that considerations of power are at least as
important as considerations of sovereignty in constraining the UN’s universalism.
References
ALT, JAMES, GARY KING, AND CURTIS SIGNORINO. (2000) Aggregation among Binary, Count and
Duration Models: Estimating the Same Quantities from Different Levels of Data. Political Analysis
9:1–24.
ALVAREZ, MICHAEL, JOSE´ ANTONIO CHEIBUB, FERNANDO LIMONGI, AND ADAM PRZEWORSKI. (1997)
Classifying Political Regimes. Studies in Comparative International Development 31(2):3–36.
ANDERSSON, ANDREAS. (2000) Democracies and UN Peacekeeping Operations, 1990–1996. International
Peacekeeping 7:1–22.
BENNIS, PHYLIS. (1996) Calling the Shots: How Washington Dominates Today’s U.N. New York: Olive
Branch Press.
BOX-STEFFENSMEIER, JANET M., AND BRADFORD S. JONES. (1997) Time Is of the Essence: Event History
Models in Political Science. American Journal of Political Science 41:1414–1461.
DE JONGE OUDRAAT, CHANTAL. (1996) The United Nations and Internal Conflict. In International
Dimensions of Internal Conflicts, edited by Michael E. Brown. Cambridge, MA: MIT Press.
TABLE 5. Summary Statistics on Operations Cost by Region
Observed Mean SD Minimum Maximum
Full Sample 19 591.44 1101.53 4.57 4642.83
Africa 8 480.24 5550.42 16.40 1686.42
Europe 3 1583.66 2649.45 27.04 4642.83
Asia 1 1620.96 F 1620.96 1620.96
Latin America & Carribean 5 196.07 185.16 4.57 443.41
Middle East 2 21.55 12.99 12.37 30.74
MICHAEL GILLIGAN AND STEPHEN JOHN STEDMAN 53
DOYLE, MICHAEL W., AND NICHOLAS SAMBANIS. (2000) International Peacebuilding: A Theoretical and
Quantitative Analysis. American Political Science Review 94:779–802.
DURCH, WILLIAM J. (1993) Getting Involved: The Political–Military Context. In The Evolution of UN
Peacekeeping: Case Studies and Comparative Analysis, edited by William J. Durch. New York:
St. Martin’s.
FEARON, JAMES D., AND DAVID D. LAITIN. (2003) Ethnicity, Insurgency and Civil War. American Political
Science Review 97:75–90.
GIBBS, DAVID. (1997) Is Peacekeeping a New Form of Imperialism? International Peacekeeping 4:
122–128.
GILLIGAN, MICHAEL J., AND STEPHEN JOHN STEDMAN. (2003) Where Do the Peacekeepers Go? With
Re-coding Appendix and STATAs Variable Names. Available at http://www.nyu.edu/gsas/dept/
politics/faculty/gilligan/peacekeepers.pdf.
JACOBSEN, PETER VIGGO. (1996) National Interest, Humanitarianism or CNN: What Triggers UN
Peace Enforcement after the Cold War? Journal of Peace Research 33:205–215.
NEACK, LAURA. (1995) UN Peace-keeping: In the Interest of Community or Self? Journal of Peace
Research 32:181–196.
PETERSON, TROND. (1991) The Statistical Analysis of Event Histories. Sociological Methods and Research
19:270–323.
PRZEWORSKI, ADAM, MICHAEL ALVAREZ, JOSE´ ANTONIO CHEIBUB, AND FERNANDO LIMONGI. (2000)
Democracy and Development. New York: Cambridge University Press.
STEDMAN, STEPHEN JOHN. (2001a) The Future of Peace Operations. In A Global Agenda: Issues Before the
56th
General Assembly of the United Nations. Lanham, MD: Rowman and Littlefield.
STEDMAN, STEPHEN JOHN. (2001b) United Nations Peacekeeping and the Future of the U.S–Japan
Alliance. Paper prepared for United States–Japan Foundation Project on the Future of the
Alliance.
Where Do the Peacekeepers Go?54