Using System Dynamics to Model and Better Understand
State Stability
Nazli Choucri !
Daniel Goldsmith *
Stuart E. Madnick *
Dinsha Mistree!
J. Bradley Morrison*
Michael D. Siegel”
' Political Science Department
? Sloan School of Management
> Sloan School of Management and School of Engineering
Massachusetts Institute of Technology
30 Wadsworth Street, Cambridge, MA 02142, USA
‘International Business School
Brandeis University
415 South Street, Waltham, MA 02454, USA
nchoucri@mit.edu, dgoldsmi@mit.edu, jbm@kurtsalmon.com,
smadnick@mit.edu, dmistree@mit.edu, msiegel@mit.edu
Abstract
The world can be complex and dangerous - the loss of state stability of countries is of
increasing concern. Although every case is unique, there are important common
processes. We have developed a system dynamics model of state stability based on an
extensive review of the literature and debriefings of subject matter experts. We represent
the nature and dynamics of the ‘loads’ generated by insurgency activities, on the one
hand, and the core features of state resilience and its ‘capacity’ to withstand these
‘loads’, on the other. The challenge is to determine when threats to stability override the
resilience of the state and, more important, to anticipate conditions under which small
additional changes in anti-regime activity can generate major disruptions. With these
insights, we can identify appropriate and actionable mitigation factors to decrease the
likelihood of radical shifts in behavior and enhance prospects for stability.
Keywords
Model, System Dynamics, State Stability, Terrorists, Insurgency, Regime Legitimacy
1. Introduction
The loss of state stability in various parts of the world is a major threat to U.S. national
security. While every case is unique, there are common processes tending toward
instability. In its Preface, The 9/11 Commission Report states: “We learned that the
institutions charted with protecting... national security did not understand how grave this
threat can be, and did not adjust their policies, plans, and practices to deter or defeat it.”
[1: xvi]. Given current realities and uncertainties, “better preparedness” can be achieved
by identifying, controlling and managing the linkages and situational factors that fuel
hostilities and undermine stability and cohesion.
Over the course of six months', researchers from MIT and elsewhere worked with
DARPA? to develop computational social science models for understanding the nature of
state stability as well as propensities for state failure and collapse. The MIT team
developed a system dynamics model to understand and represent sources and
consequences of stability, as well as ways in which the potential for disruptions could be
reduced, managed or mitigated. This paper describes our conceptualization of how and
when threats to stability might override the resilience of the state and undermine its
overall capabilities and performance. More specifically, we isolate the conditions under
which small changes in anti-regime activity can generate major disruptions — and then
seek to identify appropriate actionable mitigation factors to reduce the potential for
undesirable shifts in behavior and enhance prospects for stability.
2. Background and C ontext
We begin by placing the key issues in context, first by noting insights from the social
sciences and then by highlighting some key system dynamics modeling features.
2.1 State Stability
The stability of a state is a process, in that states can be at different stages of ‘stability’
and subject to different pressures toward instability. There are multiple modes of
fragility as well as different paths toward a range of ‘end points.’ It is well known that
studies of state stability (and fragility) are closely connected to a wide range of issues in
the social and the computational sciences such as the analyses of civil war, political
mobilization, social disturbances, institutional development (or lack thereof), economic
performance, social cohesion, ethnic violence and a range of issue areas that bear directly
on the resilience of states and their capabilities, as well as on the pressures upon the state
and the types of threats to its integrity and stability.
One of the most recognizable indications of state instability is the onset of civil war.
While it may be impossible to predict an individual catalyst for a civil war, there are
' From April to October 2005.
? Defense Advanced Research Projects Agency
many elements that make a state predisposed towards the breakout of civil war. For
instance, the most likely states for civil war are those states that have recently undergone
another war, states whose neighbors are involved in civil war, and states that are
economically weak [2]. However, many social scientists and policy-makers identify the
best device for preventing civil war as democracy [3; 4]. Yet, Siegle, Weinstein and
Halperin show that on a global basis the evidence regarding which comes first,
democracy or development, is still contentious [5]. By the same token, Hegre, et al. show
that the greatest likelihood of civil war is not in a state which is the least-democratic, but
rather civil war is much more likely to break out in a state which is semi-democratic [6]?
Our approach to state stability takes these divergent perspectives into account, but adopts
a rather different approach.
Given the wide range of contentions regarding sources and consequences of state
stability, an initial step is to define the core proposition in order to render precision and
direction for the computational and modeling strategy. The proposition is this:
A state is stable to the extent that its resilience (capabilities) is greater than the
load (or pressures) exerted upon it.
Embedded in a high level model of state stability, this proposition helps guide
formulation of the system dynamics model.
Social scientists in general, and political scientists in particular, are in general agreement
regarding the nature of the state, its fundamental features, and its generic attributes —
irrespective of specific manifestations or characteristics shaped by time or location.
Rooted in the basic contributions of Aristotle’s Politics all states consist of, and are
governed by, a complex body of relationships and institutions framed and guided by a
constitution [7]. Many centuries later, Almond and Powell put forth a formal approach to
state capacities by defining a set of specific capabilities (namely, extractive, regulative,
responsive, distributive, and symbolic) [8]. These points support our premise that a state
is stable to the extent that the loads or pressures upon it can be managed by its prevailing
capabilities or performance capacities.
2.2 System Dynamics
System dynamics is an approach for modeling and simulating (via computer) complex
physical and social systems and experimenting with the models to design policies for
management and change [9]. The core of the modeling strategy is representation of
system structure in terms of stocks and of flows. In this connection, feedback loops are
the building blocks for articulating the dynamics of these models and their interactions
can represent and explain system behavior.
Created by Jay Forrester, system dynamics modeling (SDM) has been used as a method
of analysis, modeling and simulation for more than 50 years. SDM has been used for a
> This one-dimensional scale was originally constructed by Ted Robert Gurr in conjunction with the Polity
project. Visit http://www.cidem.umd.edu/inscr/polity/ for the complete dataset.
wide range of purposes, such as to capture the dynamic relationship of energy and the
economy [10], to model the world petroleum market over a period of thirty decades [11],
to explore dynamics of economic growth [12] to analyze the environmental implications
of international trade [13], to understand supply-chain management [14], to analyze
different policies for nation-building [15], to model software development [16], and to
examine the intricacies of the air force command and control systems [17].
SDM offers unique capabilities to contribute to social science, economics, or political
science modes of analysis. SDM recognizes the complex interactions among many
feedback loops, rejects notions of linear cause-and-effect, and requires the analyst to view
a complete system of relationships whereby the ‘cause’ might also be affected by the
‘effect’. SDM enables analysts to uncover ‘hidden’ dynamics. Moreover, SDM allows
the analyst an increased level of flexibility as SDM _ utilizes both conceptual
understanding as well as empirical data collection. As Forrester explains, “the first step
[in SDM] is to tap the wealth of information that people possess in their heads. The
mental database is a rich source of information about the parts of a system, about the
information available at different points in a system, and about the policies being
followed in decision making.” [18: 5]
The system dynamics modeling process translates these elements of causal logic into
systems of integral equations [19]. Empirical analysis is also used to explain the
relationships between individual elements in the overall system. By understanding the
dynamics of a state system, including interactions among actors, actions, structures and
processes in complex environments, one can better identify how to reinforce state
capabilities while diminishing the loads and pressures exerted upon it.
3. Modeling State Stability
3.1 Overview of Process
For modeling purposes, the first step is to define the overall domain and system of
elements tending toward state stability and the sources of instability. This step yields a
high level view that is used for framing purposes, consistent with dominant lines of
thinking in the social sciences. The value of system dynamics is that it also provides a
method for empirical model grounding as well as policy crafting.
The next step is to select and ‘drill down’ to the most important, sensitive and short-term
processes that shape the more immediate threats to stability and enhance the propensities
for instability.
The third step is to formulate the overall computational system dynamics model for
simulation and analysis drawing where possible on empirical data and observable cases.
The computational system itself consists of interconnected modules that represent
different facets of the overall processes at hand.
Drawing upon data from two real-world cases, we developed and then used the model to
make further predictions about insurgency recruiting, recognize different policy
implications, and make informed policy decisions based upon empirical measurements.
3.2. The High-Level View
We begin by creating a high-level causal loop diagram that captures the key elements of
the system in question including the major feedback loops. The diagram reflects a range
of potentially significant joint dependencies and feedback processes that explain the
dynamics of the overall stability status of the state; over time.
Economic
Performance
+
Investment in Other
Productive
Investments
State Institutional
Invesment in
ey Internal/E xternal A...
\ red eSocb-poltical
Cleayages
Regime Force
and Violence
(Regime
Anti-terrorism
Insurgents
a -
ATE Capacity ang’ Soclo pala “jen om ae
Figure 1: High-Level Diagram of State Stability. (Circled section signals the
segments for detailed system dynamics modeling)
Arrows show causal relationships between variables. A plus sign (+) indicates that a change in the first
variable (at the tail of the arrow) causes a change in the second variable (at the head of the arrow) in the
same directions. A minus sign (-) indicates that a change in the first variable causes a change in the second
variable in the opposite directions. A path that begins at any variable and traces from arrow to arrow to
returns to the original variable forms a feedback loop
Figure | presents the High-Level Diagram of overall state stability. We seek here only to
define broadly the overall domain as well as the SD model focus (encompassed within the
dashed lines). .
In many cases, there were large bodies of literature that described each of the key
relationships. Framed thus, we seek only to reflect some of the most dominant sets of
relationships reflected in the literature, by way of developing an integrated device for
representing the complexities of underlying dynamics.
3.3 Drilling Deeper
In Figure 1, the dotted line delineates several elements in the High-Level Diagram that
we select as the focus for our detailed modeling process. These focus on Regime
Legitimacy and the role of Insurgents in Anti-Regime activities.
Three factors are relevant here:
First, we chose a time horizon over which long-term ramifications might be significantly
different than short-term outcomes. We wanted to choose examples where this would be
especially pronounced.
Second, we wanted a model boundary sufficiently broad so as to allow us to explore and
explain the strengths and weaknesses of a variety of policy options.
Third, we chose to focus on system-features that would best illustrate the contributions of
system dynamics to policy analysis as well as to the social sciences.
Based on all these considerations, we selected to ‘drill down’ into the dynamics of
dissident and insurgent recruiting given the resilience of the state and its capacity to
manage anti-regime activities.
The dynamic simulation and modeling analysis of insurgency-vs.-state-resilience would
become our Proof of Concept Model.
3.4 Strategy for Proof of Concept
The starting point for modeling the potential growth of dissidents and insurgents was
recruiting. We considered and sought to represent the different causes and effects of such
recruiting. Model formulation was done in four interactive stages.
* Many of these subjects were saved in case we had the opportunity to pursue future modeling. They
ranged from analyzing instances of famine and food shortage to the mobilization and propagation of
fundamental terrorism.
(1) A thorough and comprehensive literature review in order to familiarize ourselves
with the key terms and to capture the current state of understanding of insurgent
recruitment.
(2) A set of interviews conducted with relevant experts and personnel, namely
military experts, country analysts, and scholars in order to generate further insight.
(3) Use of existing empirical work to explain relationships that would hold for a
range of states, and thus to develop generic representation of complex dynamics.
(4) Testing with country-specific information for two test cases: Country A and
Country Be Country-specific information came from personnel, online newspapers,
databases constructed by government organizations, non-governmental organizations,
and experts.
Throughout the process, we considered how different policies would affect the dynamics
of the model.
4. Proof of Concept Model
Below we present the system dynamics state stability model, beginning first with an
overview of the entire system and structure, and then addressing key components
sequentially.
4.1 System Structure - Overview
The model shows the sources and consequences of insurgent recruiting, constrained and
limited by the resilience of the regime and the extent to which the state can manage anti-
regime activities. To simplify, of the many actions that insurgents chose to perform in
order to undermine regime legitimacy, we model those associated with activites that lead
to stimulating, producing and circulating anti-regime “communications” and
“messaging”®.
Anti-regime messaging and communication are thus major mechanisms for increasing
insurgent recruitment and for mobilizing opponents to the regime. The context in which
these activities occur is partly shaped by the regime. The state’s capabilities and the
resilience of the regime operate such as to counter insurgency recruitment.
5 One of the requirements for the DARPA contract was that all teams make real predictions for two
countries. Thus Country A and Country B represent two existing countries.
° Communications and messaging are broadly defined. For example, a successful terrorist attack is a form
of “message” that can be “communicated” — even if only by word-of-mouth. Communication and
messaging are both facilitated by social networks.
Figure 2 represents the segment of Figure 1 (bounded by the dotted line) that is further
developed to constitute our proof-of-concept model. Figure 2 also includes a
segmentation of the model into sub-s
stems, such as Population Growth, Regime
Resilience, Anti-Regime Recruiting, Regime Opposition, Reducing Overt Opposition,
and Communications and Mobilization. These are examined in greater detail below.
Appeasement.
Rate
Population
Regime Opposition
‘Appeasement -
Fraction
Reducing
Overt Opposition
‘Avgerage Time
as Dissident
| _ Desired Time to H
| Remove Insurgents H
Bedoming
Disbident
Recruits Through
Social Network
Recruiting :
Normal Probabjlity
of Being Recrulted.
‘Messages on
‘Efffect of Anti-Regime
| Anti-Regime
| Resilience Recuitment
Economie social $i
pertormenies te Capa
t litical 1
itestmaey Capacly | egime
on Resilience
insurgent
Recruitment
TRemoving
‘insurgents H
Propensity to
Violent incident Commit Violence
Intensity -
Relative Strength
Incivent of Violent Incidents
Intensity intensity
Effect of Incidents on
‘Anti-Regime Messages
t Message Effect
Strength
Perceived Hl
Intensity of ‘
~Messages-|.-----------| t
Communication and Mobilization
Figure 2: Conceptual Model of Insurgent Activity and Recruitment (Simplified)
The logic of the model can account for many known patterns of insurgent recruiting in
relation to the role and influence of the regime and its resilience’. The model structure
and framework is based heavily on the social science literature and earlier studies in the
computational social sciences, as well as on “tapping into the wealth of information that
people possess in their heads” [18].
The remainder of this section presents the key components of the model and articulates
the basic logic in the course of “unfolding” the structure of the behavior of the system. A
list of some of the sources of information supporting each link in the model is presented
in Appendix A.
7 The formulated model includes about 140 equations.
4.2 Sources of Insurgents
We proceed from the assumption that in any given state with a given number of people
there are some peaceful anti-regime elements (dissidents) and there are some violent anti-
regime elements (insurgents). Thus, we divide the population into three stocks, labeled
Population, Dissidents, and Insurgents, and shown as rectangles in Figure 3. The model
shows transitions from one stock to the other with icons resembling pipes and valves*.
People in the general population may become dissidents and after some time these
dissidents may become insurgents. However, all dissidents do not necessarily become
insurgents. , Dissidents might become appeased by the state and return to the general
population. A peaceful regime change or a policy change may placate these dissidents
into regime supporters. Alternatively, dissidents who become insurgents can be removed
from the system by the state. This could occur through arrests, detentions, or state
violence. These approaches to reducing the number of dissidents and insurgents,
respectively, are shown as the “Appeasement Rate” and “Removing Insurgents” in Figure
3.
Appeasement
Rate
" [—— Removed
Population] — ssi >
mane pul Becoming |Dissidents} Insurgent insurgents} R ing Insurgents:
Dissident Recruitment Insurgents
Figure 3: Population-to-Insurgent Flow
There are other conditions affecting the Population-to-Insurgent flows. Different
elements can affect the transition of a normal member of society to a dissident to an
insurgent, as we note below.
* The valves control the rate of flow between the stocks. The factors that affect these rates are explained
later.
4.3 Removal of Insurgents
The next level of complexity is presented in Figure 4.
Desired Time to
in Remove Insurgents
Bonen oman Avgerage Time
Rate mags
as Dissident 7 indicated Force
‘Appeasement indicated Strength
Fraction Removal Rate
Removal
Effectiveness
Population issic
Pop Growth | Becoming PssdemtS T crgent ia Removing
Dissident Recruitment insurgents
Fractional
Growth Rate
Recruits Through
Social Network
Figure 4: Other Ways to Affect the Population-to-Insurgent Flow
When considering the insurgent removal, we must also consider the removal
effectiveness, the indicated force strength, and the desired time to remove insurgents in
order to assess the rate at which insurgents are removed. Policy levers, such as the
amount of resources allocated to each method, that effect these variables are exogenous
to our model,, but we can test various policy options by varying these parameters.
4.4 Anti-Regime Messaging and the Flow of Communication
In Figure 5, we expand the model to show that dissidents and insurgents take actions we
call anti-regime incidents which in turn generate anti-regime messages. Such incidents
include protests, targeted attacks, or even civil war. ‘Messages’ include both formal and
informal communications between individuals in a regime.
An anti-regime message based upon an intense incident might proclaim that the regime
violently cracked down on innocent protestors, or that the regime can no longer
effectively handle insurgent movements in a certain part of the country and is therefore
incapable of controlling the state. Messages of this sort can undermine the legitimacy of
the regime and represent an important form of loads on state capacity. The load becomes
especially strong when it is reinforced by, and diffused through, social networks, thus
facilitating recruitment further, as seen in Figure 5.
Desired Time to
Remove Insurgents
= os =
=e as Dissident Indicated Force
Appeasement Indicated Strength Removal
Fraction “he Rate LA _
ISoouliuzat A moved
‘ ‘opulation! 7
Pop Growth L Becoming Dissidents TV corgent BYEZ rgents
Dissident Recruitment Insurgents
' Propensity to
Fractional Violent incident commit Violence
Growth Rate rent nd
Recruits Through Regime intensity _—
Social Network
joc wena Opponents Relative Strength
of Violent Incidents
Normal incident incident
Intensity a
Propensity to be
Recruited
Strength of Incident
Effect oe Anu Regine Relative Incident
jessager Intensity
eaeteriand Effect of incidents on
piety Anti-Regime Messages
Frequency of
‘Anti-Regime
Messages
Message Effect
Strength
Figure 5: Message Production Generated by Dissident and Insurgent Incidents
Once again, there are many policy options available for limiting the flow of anti-regime
messages. Such options are often utilized, in countries as diverse as France, (which
imposed curfews on its citizens to quell riots in October and November of 2005) and
Thailand (where the South has been in a state of martial law since March of 2005). At
the same time, we recognize that perceptions may be different from actual behavior and
there might be differences in the effectiveness of such attempts to limit the flow of anti-
regime messages.
4.5 The Role of Perceptions and Cognition
Accordingly, in Figure 6, we take into account the perceived intensity of anti-regime
messages and the relative frequency of anti-regime messages.
10
Desired Time to
Remove Insurgents
Appeasement. ,
Avgerage Time
Rate as Dissident
Appeasement strength Removal
Fraction Removal Rate Effectiveness
ix, emoved
Removing Insurgents
Insurgents
Population 7 Dissidents
Becoming insurgent
ES
Pop Growth
Dissident Recruitment
Propensity to
Violentincident — Commit Violence
Recruits Through elie Intensity 2
Social Network ‘Gpscnants nelative strength
of Violent Incidents
Fractional
Growth Rate
Normal Incident — incident
Intensity Intensity-
Propensity to be
Recruited -
Strength of Incident
Freel bao ae Relative incident
intensity
Messages
‘Normal Frequency of Jj
‘Anti-Regime Effect of Incidents on
Messages Anti-Regime Messages
Relative Frequency of
Anti-Regime Messages
Frequency of
Perceived Anti-Regime
Intensity of Messages
Anti-Regime
Messages Scuaiiage Effect
Strength
Figure 6: Accounting for Perceptions
The effectiveness of the insurgent communications methods can either magnify the
“apparent” intensity of the Anti-Regime messages, such as through the increased use of
internet and satellite television, or diminish the “apparent” intensity if not effective. As
Anti-Regime messages become more frequent than normal, the relative frequency of anti-
regime messages increases, bolstering dissident and insurgent recruitment.
4.6 Regime Resilience
The critical constraint on insurgency expansion is the resilience of the state. The SDM
model represents state resilience through an empirically derived function of key
determinants, as indicated in the social science literature. Specifically, we combine
measures of economic performance, regime legitimacy, political capacity, and social
capacity, as shown at the lower left of Figure 7, to compute the aggregate regime
resilience, via the relationship:
ll
Regime_Resilience, = a, * 8, *7, #6, *&,
Polity Index:
Civil Liberties Index:
GDP Index:
Employment Index:
a, =|Polity, / Polity,o.o|
B, =Civil _Liberties, /Civil_ Liberties,
y, = (GDP, / Population, )/(GDP,,_, / Population,., )
6,=Emp_per_Capita,/Emp_ per _ Capita.
Literacy Index: , = Literacy, / Literacy,ogq
Appeasement. -
Rate Avgerage Time
as Dissident
ee Desired Time to
Remove insurgents
7 rd ee Removed
Pop erowtn LoPU tM Becoming Sdn igent Removing [insurgents
Dissident Recruitment Insurgents
Propensity to
Violent Incident — Commit Violence
Intensity -
Relative Strength
of Violent incidents
‘ii Protest Incident
Normal Probability Intensity, Intensity:
of Being Recruited ——"
“ Propensity to
Propensity to be Protest
FA Effect of incidents on
Emcee Regime Anti-Regime Messages
Resilience on
Recruitment Message Efect
f ‘Efffect of Anti-Regime :
Regime Messages on Perceived
Resilience Recuitment Intensity of
Economic SS Social _— Meta
Performance Capacity
Regime political
Legitmacy Capacity
Figure 7: Regime Resilience
The effects of anti-regime messages on dissident recruitment are taken into account, as
shown in Figure 7 where the “Normal Probability of Being Recruited” is modified
(increased or decreased) as a result of the “Effect of Anti-Regime Messages on
Recruitment” and the “Effect of Regime Resilience on Recruitment” to compute the
“Propensity to be Recruited.”
Considering our earlier discussion of capacities versus loads, regime resilience can be
considered as a measure of the long-term capacities of a state. The literature notes that
the resilience of a state is inversely related to the occurrence of civil war. We find
empirical support for this relationship when comparing the state resilience function to the
determinants of civil war as determined by Hegre et al. [6]. Hegre, et al. looked at all
occurrences of civil wars across the world over the last several decades. They were able
to produce a measure that determines the likelihood of a civil war breaking out.
12
As we would expect, for the case of Country A, the Relative Risk of Civil War declines
as the computed Resilience index rises, as shown in Figure 8. This suggests that insurgent
movements are being contained from further breakouts at the national level.
Resilience in Country A
jilience (Unitiess)
N
uw
Oot tot
9 6b ok © OS of ob © & © Wd
O40” gD" Pg” gD” WP OW A
PP gh GP MP US s
Year
Hegre's Relative Risk of Civil War
Relative Risk of Civil War
Figure 8: Regime Resilience vs. Relative Risk of Civil War
A detailed discussion of the factors and the calculation of the Relative Risk of Civil War
(RCW) used in Figure 8 can be found in Appendix B: “Hegre-Relative Risk of Civil War
Memo.”
4.7 Review of Loads vs. Capabilities
Regime resilience can mitigate against insurgent recruitment, thereby preventing an
increase in the load on the the system. For example, when the economy is doing well or
when the regime is perceived as having increased legitimacy, the likelihood of an
individual becoming a dissident or an insurgent becomes much smaller. The foci of the
loads and capabilities of the system are highlighted in Figure 9.
13
fn ii a
Rate ‘Avgerage Time
as Dissident
ee Desired Time to
ts
LCS v Removed
opulation
Pop Growth iA Becoming [>'ssidents} insurgent Removing Iinsurger
Dissident Recruitment Insurgents Loads
Propensity to
Violent incident commit Viovence
Recruits Through Repiee intensity {__—
Social Network
eo —_——
~~ Protest Incident of Violent incidents
Normal Probability Intensity
of Being Recruited yee intensity
Propensity to,
Protest
Propensity to be
Recruited
le
Effect of Regime
Resilience on
Recruitment
Efffect of Anti-Regime
Messages on raced,
Recuitment ee
wanna ‘Sectat > _ Anti-Regime Capabilities
Performance ‘Capacity Messages
Regime | political
Lestonacy Political
Figure 9: Identifying Loads and Capabilities
Re-examining the usage of policies specifically designed to stem message flows through
reductions in civil liberties, we see that in the short-term such policies can reduce the
messaging capabilities but in the long-term such policies can undermine the social
capacity of a state and will therefore undermine the regime resilience. This is an example
of the potentially counter-intuitive “second order” affects of certain approaches to
addressing insurgents.
As a result, while there are fewer avenues for messages to circulate, these messages are
more effective at converting individuals into dissidents and insurgents as individuals are
less happy with the regime. Therefore the state and the regime should be wary of
enforcing short-term controls with deleterious effects and should instead create and foster
policies focused on improving the capacities of the state, which in turn will balance the
loads.
5. Results of Proof of Concept Model
We have used our model to explore the impact of various changes in the exogenous
factors as well as the impact of various policy options. In this section, we will present two
sets of simulation results, for the two different countries that we studied, that show two
fundamentally different modes of behavior — long-term growth in the number of
insurgents and long-term decline in the number of insurgents In these simulations, we
explore (a) the impacts of a decline in regime resilience and (b) an example of the
impacts of different policy prescriptions, in particular, removing insurgents vs. reducing
anti-regime messages and identify tipping points in which the system shifts from one
mode of behavior to the other.
14
5.1 Decline in Regime Resilience
Recall our interest in identifying the conditions in which the loads on the system exceed
the capacities to manage the loads. At such points, regime resilience will no longer be
able to ‘fend off insurgency behavior. In Figure 10 we show several simulations, in the
base case, depicted by the blue line, the number of insurgents is declining as of 2006.
The simulations introduce various possible sudden drops in regime resilience (left graph)
and the corresponding response in the number of insurgents that are generated by those
reductions in the state capacity (right graph).
Specifically, these figures show the resulting growth in insurgency with various
reductions in long-term state capacities (known as Regime Resilience in the model)’.
Examples of situations that have caused reductions in regime resilience include election
fraud, oil shock on economy, healthcare failure, and judicial breakdown. These figures
are based upon observations, estimates, and predictions in a specific region of Country A,
where insurgency has fluctuated for the last several years.
Relece Capacities Gan puis Loads
20
10
Resilience ]
Drops
0 375
1980 1990 2000 2010 2005 2006 «= 2007S 2008 2009 2010
Time(Year) Time (Year)
Figure 10: Capacities and Loads
Under the initial “base” case” (shown as the blue lines in both the Capacities and Loads
diagrams), we see a slight reduction in insurgents and Loads on the system over time.
But as the other scenarios show, even small declines in the capacities of the state change
the direction to one of future insurgent growth and increase of the Loads on the system —
thus a ‘tipping point.’'' An initial small drop in resilience produces what appear at first
° Figure 10 was developed in color. If viewed in black and white, each of the four drops in resilience is
matched with the four increasing curves of Insurgent Loads.
'° For country A
"The numbers presented in Figures 10a and 10b on the axes are based upon a specific case study., The true
value of system dynamics does not come from empirical statements such as “if you reduce the economy
10%, you will increase insurgents by 10%,” but rather “if the economy drops even a little, one can expect
to see an increase in insurgents, and based upon previous situations, if something is not done to re-establish
the capacities, the state may tip into an unsustainable situation.”
15
to be small increases in the number of insurgents, i.e. the load on the system, but because
the drop in resilience leads to more fruitful recruiting of dissidents, the ultimate
consequence is a sustained increase in insurgent numbers. The drop in resilience has
caused the system to cross the crtitical threshold, or tipping point, such that the
reinforcing process of recruitment is now dominating the overall system behavior. This
result shows the sensitivity of the system to the choice of policy prescriptions required for
state stability.
5.2 Policy Alternatives: Removing Insurgents vs. Reducing Anti-
Regime Messages
We now examine some different policy alternatives. In Figure 11, we identify two policy
options:
(1) the state might become better at removing insurgents, or
(2) the state might improve its ability to respond to anti-regime messages,
dampening the message strength.
Increased
Removal
Effectiveness
‘Appeasement.
Rate Avgerage Time
as Dissident
Appeasement
Fraction
Desired Time to
Remove Insurgents
Removed
insurgents
Dissidents,
Pop Growth insurgent Removing
Recruitment Insurgents
Propensity to
Violentincident commit Violence
Recruits Through Régie intensity __=
Social Network
7 i: a \ Relative Strength
Sm protest Incivent _0f Violent incidents
Normal Probability Intensity, Intensity<——~
—__r
Propensity to
Propensity to be Protest
Recruited
FA Effect of Incidents on
Effect of Rime Anti-Regime Messages
Message Effect
Strength
fffect of Anti-Regime
Messages on soninrer
Recuitment wpe
economic 7 WO Soi —
Performance capeety Increased
Regime : :
Legitmacy (once Regime Voice
Figure 11: Increasing the Regime Voice vs. Increased Removal Effectiveness
Earlier, we identified that controlling the circulation of anti-regime messages (through
curfews and other civil liberty limitations) would reduce the number of anti-regime
messages in the short-term. Considering the broader system, we see that a policy
16
prescription which encourages choking messages, if done through the suppression of
liberties, would undermine regime resilience, and in the long run, such a policy
prescription would cause more harm than good. Increasing the ‘Regime Voice’ is
different. It implies that the strength of the anti-regime messages might become diluted
in an acceptable fashion.
For example, one such way to increase the regime voice is to sponsor a public campaign
against the insurgents (to undermine their legitimacy). As we observed in Jordan in
November of 2005, as a result of extreme terrorist acts by Al Qaeda, the state was able to
garner public sentiment. That can become a very successful method for controlling anti-
regime elements. There are many other ways that states can increase the regime voice to
counter the threats. In Figure 12, we compare the results of changes in each of these
inputs.
27,000
25,250
23,500
etter insurgent removal
21,750
Preventing Recruitment
20,000
2005 2006 2007 2008 2009 2010
Time (Year)
Better insurgent removal Preventing recruitment
intelligence sharing moderate rhetoric
Figure 12: The Potential Effects of Increased Removal Effectiveness (Intelligence
Sharing) versus Weakening the Message Strength (Moderate Rhetoric)
The upper line'* shows the expected number of insurgents in a “base” case'* where the
regime resilience measure is at a level such that if no action is taken, the result will be a
continued growth in insurgents.
Both of the proposed policy alternatives produce a reduction from this projected growth.
However, there are significant differences in the long-term impact. Although the use of
better insurgent removal policies'* does reduce the number of insurgents at any given
time relative to the base case, the basic trajectory is not changed and insurgents do
° The blue line in colored version.
8 For country B
‘' The red line in colored version.
17
continue to grow. On the other hand, when the state focuses on preventing recruitment
through improving rhetoric, insurgent recruitment actually slows and insurgent levels
actually decline, as shown as the lowest line'®. The policy alternative of moderating
rhetoric causes a a fundamental shift in the system behavior to one of long-term decline
in insurgent numbers,. The system has crossed the tipping point and is now headed
towards long-term stability because the policy intervention has significantly weakened
the reinforcing processes of recruitment. The recruiting processes are too weak to
overcome the capabilities of the state, so the balancing loops that maintain stability
dominate the behavior of the system.
6. Conclusion
This paper focused on particular segments of an overall modeling strategy designed to
help better understand the sources and consequences of state stability. We hope to
continue our research to create a comprehensive integrative computational model, one
that addresses all of the key features of the High Level Diagram presented earlier.
Such an effort would enable us to integrate the dynamics of insurgency and dissidence
within a more detailed and realistic representation of overall stability — all of the key
loads vs. the entire major capacities. In this paper, we illustrated the underlying interplay
between critical elements that can produce useful insights, and we have shown that
system dynamics can yield fruitful results for policy prescriptions, in this case on
combating insurgent recruitment.
System dynamics modeling enables us to understand, recognize and compare both short-
term and the long-term effects, simplifying complex dynamics, and providing a tool for
policy analysis. Scholars have been analyzing states, security, policy decisions, and
nation-building for years: the military planners and policy analysts could gain insights
from such work,.
Finally, the use of system dynamics modeling, in conjunction with evidence from the
social science is a major contribution to the development of computational social
sciences. There is a long history exploring the complexity of sovereign states in the
social sciences. Improving our understanding of the dynamics of complex systems with
specific applications to the state will enhance the predictive value of modeling for
purposes of policy and planning.
Acknowledgements
Our system dynamics modeling efforts, reported in this paper, have benefited from
interactions with, and contributions of, the subject matter experts and the other
collaborators in this DARPA project.
5 A .
' The green line in colored version.
18
References
10.
11.
12;
13.
14.
15;
The 9/11 Commission Report, 2004.
Collier, P., V. L. Elliot, H. Hegre, A. Hoeffler, M. Reynal-Querol, N. Sambanis.
(2003). Breaking the Conflict Trap: Civil War and Development Policy.
Washington DC: The World Bank.
Doyle, M. W. (1986). “Liberalism and World Politics.” American Political
Science Review, 80(December): 1151-69.
Russett, B. M. (1993). Grasping the Democratic Peace: Principles for a Post-
Cold War World. Princeton, New Jersey: Princeton University Press.
Siegle, J., M. Weinstein, M. Halperin. (2004). “Why Democracies Excel.”
Foreign Affairs, Vol. 83, No. 5 (September/October): 57-71.
Hegre, H., T. Ellingsen, S. Gates, N. Gleditsch, “Toward a Democratic Civil
Peace? Democracy, Political Change, and Civil War, 1816-1992,” American
Political Science Review, Vol. 95, Issue 1, March, 2001: 33-48. Available online:
http://www.worldbank.org/research/conflict/papers/CivilPeace2.pdf (Accessed
November 27, 2005).
Aristotle, Politics, Book 4. Translated by Benjamin Jowett. Available online:
http://classics.mit.edu/Aristotle/politics.4.four.html (Accessed November 27,
2005).
Almond, G. A. and H. B. Powell. (1966). Comparative Politics: A Developmental
Approach. Boston, MA: Little, Brown and Co.
Forrester, J. W. (1958) "Industrial Dynamics--A Major Breakthrough for Decision
Makers." Harvard Business Review, 36(4), 37-66.
Sterman, J. D. (1981). The Energy Transition and the Economy: A System
Dynamics Approach. In partial fulfillment of a Doctorate of Philosophy, Sloan
School of Management, Massachusetts Institute of Technology.
Choucri, N. (1981). International Energy Futures : Petroleum Prices, Power, and
Payments. Cambridge, MA : MIT Press.
Choucri, N. and M. Bousefield (1978). “Alternative Futures: An Exercise in
Forecasting,” in Choucri, N. and T. W. Robinson, eds., Forecasting in
International Relations. Cambridge, MA. San Francisco: W. H. Freeman and
Co.: 308-326.
Lofdahl, C. L. (2002). Environmental Impacts of Globalization and Trade: A
Systems Study. Cambridge, MA: MIT Press.
Angerhofer, B., and M. Angelides. (2000). “System Dynamic Modeling in Supply
Chain Management: Research Review.” Proceedings of the 2000 Winter
Simulation Conference, J. A. Joines, R.R. Barton, K. Kang, and P.A Fishwick,
eds. Available online: http://64.233.161.104/search?q=cache:5-p--
kFpsZMJ:www.informssim.org/wscO00papers/049.PDF+Angerhofertand+angelid
es&hl=en ( Accessed November 27, 2005).
Robbins, M. (2005). Investigating the Complexities of Nationbuilding: A Sub-
19
16.
17.
18.
19.
National Regional Perspective. In partial fulfillment of a Master of Science, Air
Force Institute of Technology.
Abdel-Hamid, T. and S. E. Madnick. (1991). Software Project Dynamics: An
Integrated Approach. Englewood Cliffs, NJ: Prentice Hall.
Lofdahl, C. L. (2005). “Envisioning C2 Systems Through Simulation: An Air
Force Example.” Presented at the 10" International Command and Control
Research Symposium: The Future of C2. Available _ online:
http://www.dodccrp.org/events/2005/10th/CD/papers/219.pdf (Accessed
November 27, 2005).
. Forrester, J. W. (1991). “System Dynamics and the Lessons of 35 Years,” in The
Systemic Basis of Policy Making in the 1990s, Kenyon B De Greene, ed.
Available online: _ http://www.echelon4.com/content%20files/forrester% 20d4224.pdf
(Accessed November 27, 2005).
Forrester, J. W. (1980). Principles of Systems, 2" Preliminary Edition.
Cambridge, MA: MIT Press.
20
APPENDIX A
Reference Sources for the Components of the Model
The purpose of this Appendix is to document the sources of the information about the
causality between variables in this model. This Appendix shows that research in both the
public and private sector supports all of the critical links in this model. The slide titles
and “yellow oval” in the subsequent pages indicate which link the associated citations
support.
Avg Time as
Dissident ——
Desired Time to
Appeasement Remove
Appeasement Fraction Insurgents
Rate
Removed
ion ———=| Dissidents Insurgents
Births aiceeials Becoming Insurgent i Removing |_Insurgents
Dissident Recruitment Insurgents ’
Propensity to
Commit Violence
Violent Incident
Intensity
Recruits Through Regime Relative Strength
Social Network Opponents honey of Violent Incidents
Normal Propensity x
to be Recruited Se Incident
See Propensity to Intensity
Propensity to be Protest
Recruited
itectornagine Effect of Incidents on
to Anti-Regime
Resilience on tee
Recruitment is
Pi set Effect of Anti-Regime
Messages on - - Technolo
; Social Recruitment Parcelved Intensity Effect
Economic Capacity of Anti-Regime
Performance Messages
Political
Regime :
Legitimacy Capacity
21
Dissidents Produce Regime
Opponents
R. Petersen, Resistance and More Dissidents Means More People Oppose
Rebellion: Lessons from the Regime
Eastern Europe, Cambridge
University Press: May, 2001.
C. Tilly, From Mobilization to
Revolution, Random House:
1978.
E. Nyberg, Insurgency: The
Unsolved Mystery, Available
online:
http://www.globalsecurity.org/
military/library/report/1991/NE
N.htm (Accessed August,
2005).
Back,
Regime Opponents Recruit
Through Social Networks
N. Berry, T. Ko, T. Moy, ]. Smrcka,}. More Regime Opponents actively recruit people
Turnley and B. Wu, “Emergent Clique joi i
Purnaton In Teraist Recruitment” to join an insurgency, increasing the number of
Sandia National Laboratories, 2005. Recruits Through Social Networks.
R. Petersen, Resistance and
Rebellion: Lessons from Eastern
Europe, Cambridge University Press:
May, 2001.
P. Heymann, “Dealing with Terrorism,”
International Security, Vol. 26, No. 3,
2001: 24-38.
F. Passy, “Socialization, Recruitment, and
the Structure/Agency Gap. A Specification
of the Impact of Networks on Participation in
Social Movements,” Available online:
http://www.nd.edu/~dmyers/lomond/passy.p
Gt (Accessed Aug. 2005).
|. Weinstein, “Resources and the
Information Problem in Rebel
Recruitment,” J ournal of Conflict
pepoluton, Vol. 49, No. 4, 2005: 598-
Back
22
Recruits Through Social Network
Leads to More People Becoming
Dissident
+ N. Berry, T. Ko, T. Moy, J. More recruits in the social network leads to a
Smrcka,}.Turnley and B. Wu, _ larger number of individuals becoming
Emergent Clique Formation In _—_ dissidents.
Terrorist Recruitment, Sandia
National Laboratories, 2005.
* R. Petersen, Resistance and
Rebellion: Lessons from
Eastern Europe, Cambridge
University Press: May, 2001.
* J. Weinstein, “Resources and
the Information Problem in
Rebel Recruitment,” ournal of
Conflict Resolution, Vol. 49,
No. 4, 2005: 598-624.
4
Back,
Appeasement Can Reduce the Number
of Dissidents
* A. Smith, ‘The Politics of Dissidents can be reduced and reintegrated into
Negotiating the Terrorist Problem _ the population, given the right conditions.
in Indonesia,” Studies in Conflict &
Terrorism, Vol. 28, 2005: 33-44.
« F.Passy, “Socialization, Recruitment,
and the Structure/Agency Gap. A
Specification of the Impact of
Networks on Participation in Social
Movements,” Available online:
http://www.nd.edu/~dmyers/lomond/pa
ssy.pdf (Accessed Aug. 2005).
* L. Kowalchuk, ‘The Discourse of
Demobilization: Shifts in Activist
Priorities and the Framing of
Political Opportunities in a
Peasant Land Struggle,’ The
Sociological Quarterly, Vol. 46,
No. 2, May, 2005: 237.
Back
23
Appeasement Fractional Movement
from Dissident to Insurgent
: ce Smith, vane Folltics, of st A greater Appeasement Fraction means a lower
egotiating the Terrorist Problem i
in Indonesia,” Studies in Conflict & Insurgent Recruliment
Terrorism, Vol. 28, 2005: 33-44.
+ F. Passy, “Socialization, Recruitment,
and the Structure/Agency Gap. A
Specification of the Impact of
Networks on Participation in Social
Movements,” Available online:
http://www.nd.edu/~dmyers/lomond/pa
Ssy.| ccesset ug. «
* L. Kowalchuk, ‘The Discourse of
Demobilization: Shifts in Activist
Priorities and the Framing of
Political Opportunities in a
Peasant Land Struggle,” The
Sociological Quarterly, Vol. 46,
No. 2, May, 2005: 237.
6
Back|
Dissidents Produce Increased
Insurgent Recruitment and Increased
Numbers of Insurgents
+ P. Heymann, “Dealing with Terrorism,” — Some Dissidents eventually become Insurgents
International Security, Vol. 26, No. 3,
2001: 24-38.
+ F. Passy, “Socialization, Recruitment,
and the Structure/Agency Gap. A
Specification of the Impact of
Networks on Participation in Social
Movements,” Available online:
http://Awww.nd.edu/~dmyers/lomond/pa
ssy.pdf (Accessed August, z
+ E. Young, The Other Rebellion:
Popular Violence, Ideology, and the
Mexican Struggle for Independence,
rama Stanford University Press:
+E. Nyberg, Insurgency: The Unsolved
Mystery, Available online:
http://www. qlobals ecurity.org/military/li
brary/report/1991/NEN.him (Accessed
August, 2005).
Back
24
Insurgents Can Be Removed From
The System |
eL. Langdon, A. Sarapu, A greater number of Insurgents means more
. people to remove, thus more Removed
and M. Wells, ‘Targeting _ insurgents per month
the Leadership of
Terrorist and Insurgent
Movements: Historical
Lessons for
Contemporary Policy
Makers,” J ournal of
Public and International
Affairs, Vol. 15, Spring,
2004.
Insurgents Can Be Removed From
The System II
M. McC lintock, The longer Time to Remove Insurgents, the
slower it is to remove insurgents, and the larger
yee or onatecraft stock of insurgents es behind the 64
Counterinsurgency and
Counter-Terrorism, 1940-
1990,” Pantheon Books:
1992.
S. Rosendorff, ‘Too Much
of a Good Thing?”
J ournal of Conflict
Resolution, Vol. 48, No.5,
Oct. 2004: 657-671.
25
More Insurgents Lead to More
L. Hovil and E. Werker, “Portrait of a More Insurgents lead to more people available
Failed Rebellion: An Account of to commit Violent Acts
Rational, Sub-optimal Violence in
Western Uganda,” Available online:
http://www. people.fas.harvard.edu/~w
erker/papers/adf.pdf (Accessed Aug.
2005)
J. Weinstein, “Resources and the
Information Problem in Rebel
Recruitment,” J ournal of Conflict
Resolution, Vol. 49, No. 4, 2005: 598-
624.
S. Metz, “Insurgency and
Counterinsurgency in Iraq,” The
Washington Quarterly, Vol. 27, No. 1,
December, 2003: 25-36.
10
Back|
More Insurgents Lead to More
Regime Opponents
A. Malaquias, “Diamonds are a__ Insurgents oppose the regime and also contribute
Guerrilla’s Best Friend: The to dissident recruitment networks.
Impact of Illicit Wealth on
insurgency piateoy hid
World Quarterly, Vol. 22, Issue
3, J une, 2001: 311-316.
M. Ross, Resources and
Rebellion in Aceh, Indonesia,
For the Yale-World Bank
project on “The Economics of
Political Violence,” J une, 2003,
Available online:
http://www.sscnet.ucla.edu/poli
sci/faculty/ross/R esourcesReb
ellion.pdf (Accessed Aug.
2005).
Back
26
Insurgents have a Propensity to
Commit Violent Acts
ae Fearon and D. Laitin, “Ethnicity,
nsurgency, and Civil War,”
American Political Science
Review, Vol. 97, No. 1, February,
2003: 75-90.
A. Croissant, “Unrest in Southern
Thailand: Contours, Causes, and
Consequences Sine 2001,”
Strategic Insights, Vol. 4, issue 2,
Feb. 2005.
M. Ross, Resources and
Rebellion in Aceh, Indonesia, For
the Yale-W orld Bank project on
‘The Economics of Political
Violence,” J une, 2003, Available
online:
http://www.sscnet.ucla.edu/polisci/
faculty/ross/R esourcesR ebellion.p
df (Accessed Aug. 2005).
A greater Propensity to Commit Violence
(#acts/ month/ Insurgent) means a greater
Violent Incident Intensity
Back|
Violent Incident Intensity and Protest
Intensity Cause Incident Intensity
C. Bob, “Marketing Rebellion:
Insurt ent Groups, International Media,
and NGO Support, International
Politics Vol. No. 3, Sep. 2001:
M. Lae ‘Protest as a Political
Resource,” The American P olitical
Science Review, Vol. 62, No. 4, Dec.
1968: 1144-1158.
S. Rosendorff, ‘Too Much of a Good
Vol. No.5, Oct. 2004: 657-671.
L. Hovil and E. Werker, “Portrait of a
Failed Rebellion: An Account of
Rational, Sub-optimal Violence in
Western Uganda,” Available online:
http://www. people.fas.harvard.edu/~w
Sos adf.pdf (Accessed Aug.
yoinge” J ournal of Conflict Resolution,
The greater the Violent Incident Intensity or the
Protest Intensity, the greater the Incident
Intensity
13
Back
27
Fluctuations in Incident Intensity Cause
Fluctuations in Effects of Incidents on
Anti-Regime Messages
* M. Lipsky, “Protest as a Political As the Incident Intensity rises, the Effect of
Resource,” The American Political —_ Incidents on Anti-Regime Messages increases
Science Review, Vol. 62, No. 4,
Dec. 1968: 1144-1158.
* C. Bob, “Marketing Rebellion:
Insurgent Groups, International
Media, and NGO Support,”
International P olitics, Vol. 38, No.
3, Sep. 2001: 311-334.
¢ J.R. Faria and D. Arce, ‘Terror
Support and Recruitment,”
Available Online:
http://www.aeaweb.org/annual_mt
g_papers/2005/0109_1300_0403.
pdf (accessed Aug. 2005).
14
Back|
Effect of Incidents on Anti-Regime
Messages Affects the Perceived
Intensity of Anti-Regime Messages
+ M. Lipsky, “Protest as a Political A stronger E ffect of Incidents on Messages
Resource,” American Political increases the Perceived Intensity of
Science Review, Vol. 62, No. 4, Anti-Regime Messages
Dec. 1968: 1144-1158.
* C. Bob, “Marketing Rebellion:
Insurgent Groups, International
Media, and NGO Support,”
International Politics, Vol. 38, No.
3, Sep. 2001: 311-334.
« A. Croissant, “Unrest in Southern
Thailand: Contours, Causes, and
Consequences Sine 2001,”
Strategic Insights, Vol. 4, Issue 2,
Feb. 2005.
« J.R. Faria and D. Arce, ‘Terror
Support and Recruitment,” Available
Online:
http://www.aeaweb.org/annual_mtg_p
apers/2005/0109_1300_0403.pdf 15
(accessed Aug. 2005).
Back
28
Perceived Intensity of Anti-Regime
Messages Produces an Effect of Anti-
Regime Messages on Recruitment
¢ A. Croissant, “Unrest in A greater Perceived Intensity of Anti-Regime
Southern Thailand: Messages means a greater Effect of
Contours, Causes, and Anti-Regime Messages on Recruitment
Consequences Sine er
2001,” Strategic Insights,
Vol. 4, Issue 2, Feb.
2005.
* C. Bob, “Marketing
Rebellion: Insurgent
Groups, International
Media, and NGO
Support,” International
Politics, Vol. 38, No. 3,
Sep. 2001: 311-334.
Effects of Anti-Regime Messages on
Recruitment Affect the Propensity to Be
Recruited
* C. Bob, “Marketing Rebellion: A greater Effect of Anti-Regime Messages on
Insurgent Groups, International Recruitment means a greater Propensity
Media, and NGO Support,” to be Recrulted
International Politics, Vol. 38,
No. 3, Sep. 2001: 311-334.
¢ A. Smith, ‘The Politics of
Negotiating the Terrorist
Problem in Indonesia,” Studies
in Conflict & Terrorism, Vol. 28,
2005: 33-44.
« A. Malaquias, “Diamonds are a
Guerrilla’s Best Friend: The
Impact of Illicit Wealth on
Insurgency Strategy,” Third
World Quarterly, Vol. 22, Issue
3, J une, 2001: 311-316.
Back
29
Economy Affects Regime
Resilience
P. Collier, L. Elliot, H. Hegre, A.
Hoeffler, M. Reynal-Querol, N.
Sambanis, Breaking the Conflict
Trap: Civil War and Development
Policy, The World Bank: 2003.
H. Hegre, T. Ellingson, S. Gates,
N. Gleditsch, ‘Toward a
Democratic Civil Peace?
Democracy Political Change, and
Civil War, 1816-1992,” American
Political Science Review, Vol. 95,
No. 1, March 2001: 33-48.
J. Fearon and D. Laitin, “Ethnicity,
Insurgency, and Civil War,”
American Political Science
Review, Vol. 97, No. 1, February,
2003: 75-90.
T. Gurr, Why Men Rebel,
Princeton University Press: 1970.
As the economy is healthier, the regime does
better. When the economy is worse, anti-regime
messages are more likely to find a captive
audience.
18
Back|
Regime Legitimacy Affects Regime
Resilience
P. Collier, L. Elliot, H. Hegre, A.
Hoeffler, M. Reynal-Querol, N.
Sambanis, Breaking the Conflict
Trap: Civil War and Development
Policy, The World Bank: 2003.
H. Hegre, T. Ellingson, S. Gates,
N. Gleditsch, ‘Toward a
Democratic Civil Peace?
Democracy Political Change, and
Civil War, 1816-1992,” American
Political Science Review, Vol. 95,
No. 1, March 2001: 33-48.
T. Gurr, “A Causal Model of Civil
Strife: A Comparative Analysis
Using New Indices,” American
Political Science Review, Vol. 62,
No. 4, Dec. 1968: 1104-1124.
D. Hibbs, Mass Political Violence:
A Cross-National Causal Analysis,
Wiley: 1973.
As the regime does good things, the resilience
becomes higher.
Back
30
Political Capacity Affects Regime
Resilience
P. Collier, L. Elliot, H. Hegre, A.
Hoeffler, M. Reynal-Querol, N.
Sambanis, Breaking the Conflict
Trap: Civil War and Development
Policy, The World Bank: 2003.
H. Hegre, T. Ellingson, S. Gates,
N. Gleditsch, “Toward a
Democratic Civil Peace?
Democracy Political Change, and
Civil War, 1816-1992,” American
Political Science Review, Vol. 95,
No. 1, March 2001: 33-48.
J. Fearon and D. Laitin, “Ethnicity,
Insurgency, and Civil War,”
American Political Science
Review, Vol. 97, No. 1, February,
2003: 75-90.
K. Deutsch, Nationalism and
aoc Communication, MIT Press:
If there are strong political institutions, a corrupt
regime may still find support in the short-run.
20
Back,
Social Capacity Affects Regime
Resilience
T. Gurr, Why Men Rebel,
Princeton University Press: 1970.
H. Hegre, T. Ellingson, S. Gates,
and N. Gleditsch, “Toward a
Democratic Civil Peace?
Democracy Political Change, and
Civil War, 1816-1992,” American
Political Science Review, Vol. 95,
No. 1, March 2001: 33-48.
R. Putnam, R. Leonardi, and R.
Nanetti, Making Democracy Work:
Civic Traditions in Modern Italy,
Princeton University Press: 1993.
D. Horowitz, Ethnic Groups in
Conflict, Yale University Press:
1968.
K. Deutsch, Nationalism and
Social Communication, MIT Press:
1953.
As there are more inequalities within a society,
more troubles are likely to occur.
Back
31
Regime Resilience Affects
Recruitment
Hp Collter, b- Ellloe, He) Hegre A. Greater Regime Resilience means a stronger
oeffler, M. Reynal-Querol, N. ili
Sambanis, Breaking the Conflict Effect of Regime Resilience on Recruitment.
Trap: Civil War and Development
Policy, The World Bank: 2003.
H. Hegre, T. Ellingson, S. Gates,
N. Gleditsch, ‘Toward a
Democratic Civil Peace?
Democracy Political Change, and
Civil War, 1816-1992,” American
Political Science Review, Vol. 95,
No. 1, March 2001: 33-48.
J. Fearon and D. Laitin, “Ethnicity,
Insurgency, and Civil War,”
American Political Science
Review, Vol. 97, No. 1, February,
2003: 75-90.
T. Gurr, Why Men Rebel,
Princeton University Press: 1970.
22
Back|
Regime Resilience Buffers the
Propensity to be Recruited
A. Malaquias, “Diamonds are a The greater the Effect of Regime Resilience on
Guerrilla’s Best Friend: The Recruitment, the larger the Propensity to be
Impact of Illicit Wealth on Recruited.
Insurgency Strategy,” Third World
Quarterly, Vol. 22, Issue 3, J une,
2001: 311-316.
A. Croissant, “Unrest in Southern
Thailand: Contours, Causes, and
Consequences Sine 2001,”
Strategic Insights, Vol. 4, Issue 2,
Feb. 2005.
H. Hegre, T. Ellingson, S. Gates,
N. Gleditsch, “Toward a
Democratic Civil Peace?
Democracy Political Change, and
Civil War, 1816-1992,” American
Political Science Review, Vol. 95, 23
No. 1, March 2001: 33-48. Back,
32
Propensity to Be Recruited
Increases Recruits through Social
Network
« N. Berry, T. Ko, T. Moy, J. As the propensity to be recruited increases, the
Smrcka, J. Turley and B. Wu, likelihood of an interaction causing a supportive
individual to become a recruit rises (and
vice versa).
“Emergent Clique Formation In
Terrorist Recruitment,” Sandia
National Laboratories, 2005.
* A. Malaquias, “Diamonds are a
Guerrilla’s Best Friend: The
Impact of Illicit Wealth on
Insurgency Strategy,” Third World
Quarterly, Vol. 22, Issue 3, J une,
2001: 311-316.
* R. Petersen, Resistance and
Rebellion: Lessons from Eastern
Europe, Cambridge University
Press: May, 2001.
Average Time as a Dissident
¢ SME Interview
¢ “12-18 months for
Thailand, and the
same with
Indonesia”
33
Desired Time to Remove
Insurgents
¢ SME Interview
* ‘The timeline in
Thailand would
probably be more
like 48 months”
« “In Indonesia 48-60
months may be a
more appropriate
time frame.”
Appeasement Fraction
* SME Interview
* ‘The appeasement fraction
for Thailand will be 75-80
ercent back to supportive.
ndonesia is more complex
with more insurgencies, with
the more mature insurgencies
there | would put the
percentage at 70-75%.”
“certainly less than 10%, in
Papua maybe less than 5%,
and in Ache | don't know...
that is a good questions...
5% is a reasonable guess.”
27
Back
34
Propensity to Protest
¢ SME Interview
¢ “In Thailand it is
between one
protest a month
and one each
quarter. One
every 1-3 months”
Propensity to Commit Violence
¢ SME Interview
* “Out of a pool of
100 armed
insurgents, in a
month they would
average 3-5
incidents of
violence.”
35
APPENDIX B
Hegre-Relative Risk of Civil War Memo
1 INTRODUCTION
The Relative Risk of Civil War (RRCW) is a measure constructed in “Toward a
Democratic Civil Peace? Democracy, Political Change, and Civil War, 1816-1992” by
Hegre, Ellingsen, Gates, and Gelditsch (2001).
The measure is a determinant of the likelihood of a civil war in a country.
This appendix presents the Hegre Method, the variables used, and our use of this method
to ‘validate’ Regime Resilience in this paper.
1.1 Hegre et al.’s Approach:
The functional form of the Relative Risk of Civil War (RRCW) equation is as follows:
RRCW = exp((exp(a)*x1) + (B*x2) + (B°*x3) + (exp(y) *x4) + (exp(S) *x5) + (€ *x6) +
(6 *x7) + (1 #x8) + (17*x9) + (6 *x10))
where
a = proximity of regime change
B = level of democracy
= level of democracy squared
y = proximity of civil war
6 = proximity of independence
€ = international war in country
¢ = neighboring civil war
1_= level of development
v= level of development squared
0 = ethnic heterogeneity.
a=)
1.2 RRCW and Regime Resilience
We have used the RRCW as a test against our own Regime Resilience measure by
searching for an inverse relationship. The logic for such a relationship is that when
Regime Resilience is high — when a country is being run well — the likelihood of civil war
is low. Vice versa, when a regime is weak or threatened, there is a higher likelihood of
civil war.
36
2, THE RRCW METHOD
2.1 Dependent Variable
The dependent variable in the Hegre research is the outbreak of civil war, as recorded in
the Correlates of War Project. This measure is constructed by examining 152 countries
from 1816-1992, with a civil war including any instance a country experiences a regime
change.'° Formally defined, a civil war is “an internal war in which ‘(a) military action
was involved, (b) the national government at the time was actively involved, (c) effective
resistance (as measured by the ratio of fatalities of the weaker to the stronger forces)
occurred on both sides, and (d) at least 1,000 battle deaths resulted;’” (Hegre, et al., 2001:
36, with inner quotation from Singer and Small, 1994: part 3).
Hegre et al. created two models: one where civil war is examined from 1816-1992, and
one where civil war is examined post-World War II (from 1946-1992). The reason behind
creating a more recent model is because of the substantial ways that both war and politics
have changed over this time period. We use the second model (post-WWII on).
2.2. Independent Variables and Control Variables
Hegre and his team include several independent variables and control variables. Hegre et
al. are interested in whether democracy and democratization stem civil war. However, to
determine the effect of democracy and democratization, they had to control for several
factors, partly including economic and social factors, as well as historical factors.
3. Operational Definitions
3.1. The Variables
Below is a brief definition of the variables and the source data used by Hegre, et al.:
e Proximity of Regime Change: This is the amount of time since the regime most
recently changed. It is a measure for how long the current regime has been in power.
It is located in the Polity IV dataset, and is listed as “durable’ or ‘duration.’ It is listed
in years, but needs to be multiplied by -1 (since a more recent regime change means
greater likelihood of civil war), divided by days, and have its exponent taken to be
consistent with the strengths of the other measures and to eradicate the negative
value. We simply multiply by -1, divide by 365 (ignoring leap years), and calculate
the exponent.
e Level of Democracy: This is a measure that represents the level of democracy in a
country, from 10 being a perfect democracy, to -10 being an autocracy. It is coded as
polity2 in the Polity IV dataset.
s Computationally, a regime change is defined as when a country undergoes a change of 2 or more on the
commonly accepted democracy-autocracy scale, which ranges from -10 (totally autocratic) to 10 (totally
democratic).
37
Level of Democracy Squared: The logic for squaring the level of democracy is that
much theory and research has suggested that strong autocracies are just as stable as
strong democracies, and that states with weak regimes are the must vulnerable to
conflict and civil war. Hegre et al. found that including this measure formed a far
more accurate model.
Proximity of Civil War: This is a measure of the number of days since the last civil
war ended. It can be found in the Correlates of War database. This value is given in
days. To make this measure consistent, we again multiply by -1, divide by 5840, and
take the exponent.
Proximity of Independence: This is a measure of the time since a country either
formed or engaged in regime change. It can be found in the Correlates of War
database. This value needs to be multiplied by -1, divided by days, and have its
exponent taken, again so as to be consistent with the strengths of the other measures.
We simply multiply by -1, divide by 365 (ignoring leap years), and calculate the
exponent.
International War in Country: This is a dummy variable for whether two countries
are fighting within the country of analysis. The measure can be found in the
Correlates of War database.
Neighboring C ivil War: This is a dummy variable for whether a neighboring country
is undergoing civil war. It is constructed by examining the Correlates of War database
to see if a civil war is taking place in a neighboring country.
Level of Development: This is a measure of the level of development, and it is
provided by the Correlates of War database. It is a measure of energy consumption.
Level of Development Squared: The level of development of a country is clearly a
decisive measure of civil war, and by squaring the value, the difference between
developed and developing countries becomes more pronounced.
Ethnic Heterogeneity: This is a measure of the ethnic composition of the country.
This measure can be found in Appendix B of Hegre, et al.’s paper, along with a list of
the civil wars that were included in the analysis. The values range from 0 to 1.
3.2. The Coefficients
Each of the values above is multiplied by a specific coefficient (x1 to x10), as determined
by regression analysis.
Below are the coefficient for the specific variables:
xl
x2
x3
x4 = 1.160
38
x5 =
x6
x7 =
x8
x9
x10= 0.800
4, CALCULATING THE RRCW
Once a country is identified and the data is located across the years of interest, the
calculation follows the above equation and directions.
Once the values are added, the exponent of the resulting summation needs to be taken in
order to complete the calculations.
The exponent is taken to pronounce the relationship, largely so that changes can be more
easily depicted, and also so that the value does not come out as negative, since a
likelihood of civil war cannot be negative.
In order to make the RRCW value useful for our purposes, we had to perform an
important transformation. The RRCW was a measure originally constructed based upon
events as the unit of analysis. For our purposes, we wanted our unit of analysis to be in
country-years, so we had to adapt the calculation method.
5. EVALUATION
Our use of civil war is only one of the possible ways of ‘validating’ our regime resilience
measure. Others include ‘Political Capacity’ from the POFED model (J. Kugler). As
such, the RRCW value was originally calculated to examine whether our measure of
Regime Resilience was empirically plausible.
We have compared the RRCW to Regime Resilience in three countries now, and we have
found the correlations to be very consistent.
References for A ppendix B
1. Hegre, Havard, Tanja Ellingsen, Scott Gates, and Nils Petter Gleditsch. 2001.
American Political Science Review 95 (1): 33-48. Available online:
http://www.worldbank.org/research/conflict/papers/CivilPeace2.pdf
2. Singer, J. David, and Melvin Small. 1994. Correlates of War Project: International
and Civil War Data, 1815-1992 [computer file] (Study #9903). Ann Arbor, MI: J.
David Singer, University of Michigan and Detroit, MI: Melvin Small, Wayne State
University [producers], 1990. Ann Arbor, MI: Inter-university Consortium for
Political and Social Research [distributor], 1993.
39
