s are Welcomed!
1
On-the Ground Assessment of United Nations Peacekeeping: Micro-Level Evidence
from Surveys with Ex-combatants in Burundi.

By

Eric Mvukiyehe
Columbia University

Abstract

Previous statistical studies have established that the presence of peacekeepers increases the likelihood that
a country will transition from war to peace and that postwar peace will be more likely to last. These
empirical studies, however, have suffered from at least two different, but interrelated shortcomings: (i)
selection and endogeneity problems; and (ii) under-specification of causal mechanisms. Both problems
have complicated the endeavor of establishing causality. While recent methodological advances have
allowed scholars to address the former problem, very little effort has been made to address the later. Part
of the problem is that these empirical studies have taken a macro-level approach. The theories that guide
these studies, however, are actually micro-level theories because they explain how peacekeeping alters
individual combatants incentives such that they are willing to lay down arms and support the peace
process. But the macro-level tests that have been conducted thus far do not allow us to examine if and
under what conditions such mechanisms are relevant. The aim of this paper is to fill this empirical gap
that is, to evaluate whether the empirical effect attributed to peacekeeping interventions in civil wars can
be accounted for by the proposed causal mechanisms. I use micro-level data, combining survey interviews
with ex-combatants (including members of the current military and national police) and in-depth interviews
with key political and military elites from a recently completed pilot study in Burundi. Preliminary results
suggest that exposure to UN peacekeepers was positively associated with attitudinal outcome measures
capturing the extent of confidence building between belligerentsa finding consistent with the information
provision mechanismbut did not have any association with whether combatants violated ceasefire
agreements. The results also suggest important factional differences.

I. Introduction
Does peacekeeping work? Very few topics in international relations have sparked as
much interests among policy analysts and scholars in recent years than the role of
peacekeeping interventions in civil wars. While the theoretical literature has been scant
and fragmented, the empirical literature has made significant contribution to our
understanding of this topic. Particularly, large-N, statistical studies have consistently
established that the presence of peacekeepers increases the likelihood that a country will
transition from war to peace and that postwar peace will be more likely to last (Doyle and
Sambanis 2000; Hartzell et al 2001; Fortna 2003; 2004). These empirical studies,
however, have suffered from at least two different, but interrelated shortcomings: (i)
selection and endogeneity problems; and (ii) under-specification of causal mechanisms.
Both problems have complicated the endeavor of establishing causality. While recent
methodological advances have allowed scholars to address the former problem, very little

This is a Draft! Comments are Welcomed!
2
effort has been made to address the latter problem. That is, existing quantitative studies
have failed to specify (let alone to empirically evaluate) the potential causal mechanisms
linking the presence of peacekeepers to the durability of peace.
Part of the problem is that these empirical studies have taken a macro-level
approach. The theories that guide these studies, however, are actually micro-level
theories because they explain how peacekeeping alters individual combatants incentives
such that they are willing to lay down arms and support the peace process. But the macro-
level tests that have been conducted thus far do not allow us to examine if and under what
conditions such mechanisms are relevant. The aim of this paper is to fill this empirical
gapthat is, to evaluate whether the empirical effect attributed to peacekeeping in its
role of helping to establish and prolong peace following a civil war can be accounted for
by the proposed causal mechanisms. I use micro-level data, combining survey interviews
with ex-combatants (including members of the current military and national police) and
in-depth interviews with key political and military leaders from a recently completed
pilot study for a survey project entitled Wartime and Post-conflict Experiences in
Burundi: An Individual Level Survey.
1
Preliminary results suggest that exposure to UN
peacekeepers was positively associated with attitudinal outcome measures of trust
between belligerentsa finding consistent with the information-provision mechanism
but did not have any association with whether combatants violated ceasefire agreements.
Preliminary results also suggest important factional differences.
Micro-level data allow us to do two things. First, they minimize the unit
homogeneity assumption that bedevil most large-N, macro-level studies, and thereby help
improve causal inferences. In other words, it is relatively easy to isolate the effect of
peacekeeping because factors such as the type of war, the issues over which war is being
fought, and many other context-specific variables are held constant. The idea is that we
increase our confidence in the theories and empirical results established at the macro-
level if we find that the purported relationships also obtain at the micro-level. Second,

1
This is a collaborative research project with Cyrus Samii (Columbia University) and Gwendolyn Taylor
(New York University) and supervised by professors Macartan Humpreys (Columbia University) and
Michael Gilligan (New York University). Generous financial support for the pilot project was provided by
the Folke Bernadotte Academy a Swedish government agency dedicated to improving the quality and
effectiveness of international conflict and crisis management. Financial support for the scaled-up project is
being provided by grants from the Folke Bernadotte Academy and the United States Institute for Peace.

This is a Draft! Comments are Welcomed!
3
micro-level data allows us to directly (or indirectly) test the potential causal mechanisms
that are posited in peacekeeping theories, but are difficult to test using macro-level data.
The rest of the paper proceeds as follows. The next section discusses theoretical
arguments linking the presence of peacekeepers and the durability of peace and gives a
brief review of the empirical literature in support of these arguments. Section three
identifies and discusses potential causal mechanisms behind theoretical accounts on the
effectiveness of peacekeeping, which are often assumed rather than laid out explicitly in
existing studies. Section four discusses a micro-level research design with particular
focus on the operationalization of variables. Section five introduces the data from a
recently conducted individual-level survey in Burundi, presents statistical analyses,
discusses preliminary results, and offers some conclusions. I conclude by suggesting
avenues for future research in section six.

II. Literature Overview
This section offers a brief overview of the theoretical arguments linking peacekeeping to
peace and discusses some of the empirical evidence in support of these arguments. The
question here is: what difference does it make to have peacekeepers present rather than
absent in the peace process?
2
While there has been very little theoretical work to shed a
light on this question, recent empirical studies have offered interesting insights. Perhaps
the predominant theoretical explanations for why peacekeeping works is that it: (i)
enables combatants to overcome commitment problems in agreeing to ceasefires and
disarmament and demobilization processes (Walter 2001); and (ii) prevents defectors and
spoilers from endangering the peace process (Stedman 1997; Fortna forthcoming).
Underlying these claims is the notion that peacekeepers act as guarantors of political and

2
Different empirical studies on the effectiveness of peacekeeping in the peace process (especially large-N,
macro-level studies) focus on different dependent variables. Doyle and Sambanis (2000), for example,
focus on the impact of peacekeeping on peace-buildinga concept they operationallize along three criteria:
a) whether war ended; b) whether violence ended; and c) the level of democratization after the end of the
war. Fortna (2003; 2004), on the other hand, focuses on the durability of peace in the aftermath of the civil
warnarrowly defined with reference to whether and how long ceasefire agreements last. While different,
the dependent variables in both studies mirror two operational indicators long suggested by Diehl (1994)
and Stedman et. al (2002): (i) conflict abatement; and (ii) conflict resolution. The former measures the
extent to which a negative peace has been established (i.e. whether belligerents have stopped fighting).
The second measures the degree to which belligerents have resolved the underlying causes of the conflict
including the implementation of agreements. This paper focuses on aspects of both.

This is a Draft! Comments are Welcomed!
4
military settlements during the peace process (Ratner 1995, 26).
3
According to Ratner, a
guarantor is a powerful actor that agrees to undertake its best efforts to preserve a
political situation involving other actors. The guarantor commits to preserve the
settlement, ensure that each party fulf