The Dynamics of Violence in Vietnam: An Analysis of the Hamlet Evaluation System (HES)
| Published date | 01 May 2009 |
| DOI | 10.1177/0022343309102656 |
| Date | 01 May 2009 |
| Author | Matthew Adam Kocher,Stathis N. Kalyvas |
| Subject Matter | Articles |
335
© The Author(s), 2009. Reprints and permissions:
http://www.sagepub.co.uk/journalsPermissions.nav,
vol. 46, no. 3, 2009, pp. 335–355
Sage Publications (Los Angeles, London, New Delhi,
Singapore and Washington DC) http://jpr.sagepub.com
DOI 10.1177/0022343309102656
The Dynamics of Violence in Vietnam: An
Analysis of the Hamlet Evaluation System (HES)*
STATHIS N. KALYVAS & MATTHEW ADAM
KOCHER
Department of Political Science, Yale University
The authors analyze a unique data source to study the determinants of violence against civilians in a
civil war context. During the Vietnam War, the United States Department of Defense pioneered the
use of quantitative analysis for operational purposes. The centerpiece of that effort was the Hamlet
Evaluation System (HES), a monthly and quarterly rating of ‘the status of pacification at the hamlet
and village level throughout the Republic of Vietnam’. Consistent with existing theoretical claims, the
authors find that homicidal violence against civilians was a function of the level of territorial control
exercised by the rival sides: Vietnamese insurgents relied on selective violence primarily where they
enjoyed predominant, but not full, control; South Vietnamese government and US forces exercised
indiscriminate violence primarily in the most rebel-dominated areas. Violence was less common in the
most contested areas. The absence of spatial overlap between insurgent selective and incumbent indis-
criminate violence, as well as the relative absence of violence from contested areas, demonstrates both
the fundamental divergence between irregular and conventional war and the need for cautious use of
violent events as indicators of conflict.
Introduction
One of the most exciting recent develop-
ments in the field of civil wars studies is the
explosion of ‘micro-level’ research, denot-
ing an empirical move toward subnational
research designs that rely on the analysis of
finely disaggregated data; and a theoretical
move toward the specification of microfoun-
dations. The shift toward the micro level
reflects a desire to improve the specification
of causal mechanisms underlying statisti-
cal correlations and to address problems of
measurement and interpretation that cannot
be easily resolved at the cross-national level
(Kalyvas, 2008; Cramer, 2007; Cederman
& Girardin, 2007; Humphreys, 2005).
In particular, three areas have mobilized
researchers.
The first investigates the logic of partici-
pation and recruitment: who joins, how,
when, what type of organization – and
likewise, who leaves and how? (Arjona &
Kalyvas, 2008; Humphreys & Weinstein,
2008; Oyefusi, 2008; Blattman & Annan,
2007; Guichaoua, 2007; Kalyvas & Kocher,
2007b; Weinstein, 2007; Viterna, 2006;
Wood, 2003). The unit of analysis in these
studies is the individual, and data consist
mainly of ex-combatants surveys.
The second area relates to the dynamics
and location of conflict: where and how do
* Author names are ordered alphabetically. For com-
ments on the article, we thank Emily Meierding, partici-
pants in the Comparative Politics Workshop, University
of Chicago, and three anonymous reviewers. Replication
data, jobs, and a web-only appendix may be accessed at
http://www.prio.no/jpr/datasets. Correspondence: stathis.
kalyvas@yale.edu or matthew.kocher@yale.edu.
journal of PEACE RESEARCH volume 46 / number 3 / may 2009
336
rival sides clash, and what explains the type,
intensity, and duration of these clashes?
(Ziemke, 2008; Do & Iyer, 2007; Hegre &
Raleigh, 2007; Bohara, Mitchell & Nepal,
2006; Buhaug & Rød, 2006; Murshed &
Gates, 2005; Deininger, 2004; Restrepo,
Spagat & Vargas, 2004; Trejo Osorio,
2004). The unit of analysis here is spatial,
including regions, administrative districts,
or geographic ‘polygons’. Data are culled
from a variety of sources, including newspa-
pers and local NGOs, and the analysis often
employs GIS (Geographic Information Sys-
tems) methods.
The third area of research is based on
the conceptual disaggregation of violence
and conflict. The focus is primarily on
the patterns of homicidal violence, as dis-
tinct from the logic of conflict in general
( Bundervoet, 2009; Hagelstein, 2008;
Hultman, 2008; Kocher, Pepinsky &
Kalyvas, 2008; Balcells, 2007; de la Calle
Robles, 2007; Eck & Hultman, 2007;
Weinstein, 2007; Humphreys & Weinstein,
2006; Kalyvas, 2006; Straus, 2006; Verwimp,
2006; Barron, Kaiser & Pradhan, 2004; de
la Calle Robles & Sanchez Cuenca, 2004;
Verwimp, 2003; Azam & Hoeffler, 2002).
This research area has generated consid-
erable diversity in units of analysis and
specific foci. Some studies have analyzed
spatial variation of violence while keeping
conflict, actors, time, and type of violence
constant. Others have focused on types of
violence (direct vs. indirect; homicidal vs.
non-homicidal; selective vs. indiscriminate;
strategic vs. expressive; violence used to con-
trol vs. violence used to exterminate), types
of perpetrators (incumbents vs. insurgents;
insurgents across countries), types of victims
(co-ethnics vs. non-co-ethnics), or types
of conflict (irregular vs. conventional civil
wars). The proliferation of questions is a
natural reflection of the complexity of vio-
lence; at the same time, it also reflects the
field’s youth.
Our goal is twofold. First, we take
advantage of unique, and up to now
unequaled, data to investigate violence
against civilians during the Vietnam War:
the Hamlet Evaluation System (HES).1
These data, collected by the US military,
include measures of both selective and
indiscriminate violence and levels of terri-
torial control across all Vietnamese hamlets
for a substantial temporal cross- section of
the conflict. We are unaware of any other
dataset that includes data so finely grained
and covering so many units. Second, we use
these data to conduct an out-of- sample test
of a central claim in the literature on civil
war violence. Specifically, we investigate
whether the spatial distributions of selective
violence by the Vietcong and of indiscrimi-
nate violence by the South Vietnamese and
the US military, conform to the empirical
predictions of Kalyvas’s Logic of Violence in
Civil War (2006). This is a rare opportunity
to evaluate an existing theory on a different,
yet fully comparable and more comprehen-
sive, body of data.
Our analysis shows that selective violence
by the Vietcong was much more common
in hamlets that were predominantly, but
not fully, controlled by them than it was in
hamlets that were fully under Vietcong con-
trol, hamlets that were contested between
the rival sides, or hamlets under predomi-
nant or full government control. We also
find that government bombing and shell-
ing most heavily affected hamlets that were
under total Vietcong control. The two types
of violence we analyze happened in differ-
ent places; furthermore, they did not occur
in the most contested territory, the type of
place that most resembles the front line of
1 We obtained electronic copies of the HES data held by the
National Archives and Records Administration (NARA), along
with photocopies of supporting documentation. The data were
archived as zoned-decimal files. We converted them to ASCII
with help from Fay Booker, Data Librarian at Social Science
Research Computing in the University of Chicago. Our analy-
ses were carried out with STATA 10 and ArcGIS 9.2.
To continue reading
Request your trial