Politics and Power

Abstarct

P in PMESII stands for Political. Perhaps P was placed first merely to make the acronym easier to pronounce. However, more likely P’s position of prominence was intended to signify its relative importance in international affairs.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

P in PMESII stands for Political. Perhaps P was placed first merely to make the acronym easier to pronounce. However, more likely P’s position of prominence was intended to signify its relative importance in international affairs.

Analysts, in particular, count redistribution of political power as one of the most notable effects of an international intervention. There is strong motivation for this; for example, a diplomatic intervention or international information campaign may shift power toward political groups which support the position favored by intervening groups; an economic or humanitarian intervention can strengthen political groups that control or, at least, take credit for aid; or a military intervention (a blockade, weapons provisioning, or invasion) can affect the military power of various political parties in a region, for better or worse.

Furthermore, the importance of politics does not end here. Political developments are frequently the very cause (vis a vis the result) of intervention. In response to economic or environmental shortfalls, for example, politics, through its decision-making processes, attempts to ameliorate competition for scarce resources and, in doing so, often produces conflict – ranging from the trivial (a local school board divided over the location of a new schoolhouse) to the catastrophic (a war between superpowers). Conflict, in turn, exerts an influence on the political and economic decisionmaking of intervention planners by creating uncertainty (e.g., regarding elections, economic trends, and overall stability) – often sufficient uncertainty to blur the boundary between profit and loss, or between victory and defeat scenarios.

Clearly, it is in the interest of intervention analysts to be able to understand and rigorously model political dynamics, but, aside from anecdotal instances, political scientists have received little support from the modeling community. However, this is beginning to change as the field blossoms. This chapter discusses recent work in this area, assesses the challenges faced, and provides a flavor of what is on the horizon. As with the rest of this book, it is hoped that this introduction stimulates further research as well as interest by intervention practitioners.

1 The Challenge of Political Modeling

As with many young disciplines, the world of political models is labyrinthine and balkanized; it comprises a plethora of different methodologies to analyze different aspects of similar phenomena. War, for example, is one of the primary foci of political science, as it is the most dramatic and destructive event that occurs in the political arena. As such, political scientists have long attempted to model different aspects of intra and interstate hostilities, ranging from the probability of war being initiated to the breadth, severity, and duration of a conflict, and to predicting the victor (Midlarsky 2000; Kugler and Lemke 1996; Mearsheimer 2001; Small and Singer 1982; Bueno de Mesquita and Lalman 1992; Gilpin 1981; Waltz 1979; Kahneman and Tversky 1979; Gilbert and Troitzsch 1999; Mearsheimer 2001; Bahaug and Gates 2002; Bahaug and Lujala 2004; Abdollahian and Kang 2008; Arbetman and Johnson 2008; Abdollahian et al. 2009; Levy and Thompson 2010). These inquiries result in a variety of models for different aspects of conflict and cooperation from probability through planning to reconstruction, but as yet, political scientists have no general, integrated model of when individuals, organizations, states, or collections of states will cooperate or fight.

The chief impediment to the construction of general, integrated political models is the broad theoretical question of determining which influences are most crucial in affecting political outcomes. Do individuals shape events or does history constrain individuals? This issue is known as the “levels-of-analysis problem”; i.e., whether the analyst should examine the individual decision-maker, organizational and interest group mobilization, national preferences, or the structure of the international system. Former House Speaker Tip O’Neil famously noted that “all politics are local” (O’Neill 1993); a useful (but less eloquent) corollary may be that all politics are the aggregation of preferences and power at each level of analysis.

Analytical questions regarding political modeling generally fall into three categories: micro-level dynamics (the expected actions and interactions of individuals, groups, or governments), intranational structural dynamics (the subnational, structural factors that politically propel a nation, such as economic prosperity, democratization, or other national indicators), and international structural dynamics (the cross-national comparisons of national factors). Each of these categories is best assessed using a particular methodology and the appropriate theoretical assumptions.

As Table 1 indicates, agent-based modeling is best applied to the near-term, micro-level dynamics of how individuals, groups, or nation-states interact. This bottom-up approach allows for detailed granularity in understanding how individuals interact in a given political environment. A vast literature (Schelling 1960; Axelrod 1986) on rational choice and microeconomic theories explains the drive behind individual-level behaviors and interactions (discussed later in this chapter). For example, how do specific insurgent groups gain support of likely sympathetic target populations: through coercion, influence, or the distribution of public goods? (Fearon and Laitin 2003).

Table 1 Political Methodology & Applications

Intra- and international dynamics have traditionally been captured at the structural level (Organski 1958; Waltz 1979; Goldstein 1988; Rasler and Thompson 1994; Tammen et al. 2000; Mearsheimer 2001; Lemke 2002; Doran 2003). Using equation-based dynamic modeling, most political theories (Richardson 1960; Intriligator and Brito 1984; Muncaster and Zinnes 1988; Saperstein 1994) at this level focus on indicators of national attributes and how those indicators interrelate. Here, a multitude of theoretical and empirical research exists (Zinnes and Gillespie 1976; Nicholson 1998; Brown 1995; Kadera 1995; Abdollahian 1996, 2008, 2009). Game-theoretical approaches (Schelling 1960; Powell 1987; Fearon 1994; Zagare and Kilgour 2000; Abdollahian and Alsharabati 2003) can be applied to any of the aforementioned, albeit, generally, with lower levels of fidelity unless highly tailored – and thus less generalizable – to a larger variety of political circumstances.

Of course, macro inputs, such as international events and government action, influence micro, or individual and group outcomes. Individuals, groups, and nations interact embedded in an environment that is defined, shaped, and constrained by macro dynamics of our political milieu. Changing national attributes, such as decreasing economic production or highly unstable political environments, can significantly alter micro-level interactions and even individuals’ decision calculus. For example, during domestic political disturbances, decision time horizons of individuals as well as companies become shorter in the face of increasing uncertainty, driving more selfish behavior and eroding trust (Axelrod 1986).

A political modeler should account for the nexus between intra- and interstate dynamics that influence the decision calculus of the individuals and groups at the micro level and vice versa. While local political interactions of individuals are influenced by national and international conditions, the sum of those interactions can shape national and international conditions as well. Currently, there are good political models at the micro level (Bueno de Mesquita 1985; Kugler and Feng 1997; Rasler and Thompson 1994) and at the macro level (Grossman 1991; Fearon and Laitin 2002; Collier and Hoeffler 2004), but very few bridge the gap. For that reason, this chapter surveys political modeling theories from several subdisciplines in political science, spanning macrostructural theories of conflict, deterrence, war, and political economy to micro-level theories of political motivations and decision-making. While each of these literatures defines a portion of political interactions, together they outline the phenomenology of conflict and the boundaries of our current knowledge.

2 Theoretical Building Blocks

The first step in the construction of a political model is to select an appropriate theoretical foundation to inform and validate the underlying assumptions about the political behaviors to be modeled. A classification system to assess the applicability of various political theories is outlined to aid the reader. This system includes unit of analysis, model type, assumptions, key variables, structure of the environment, and the core logic of how the variables are related, in addition to main implications, empirical support, and shortcomings of these theories. Tables 2 and 3 describe several micro-, intra-, and international political theories and their relevant discriminating attributes.

Table 2 Macro Theories and their attributes

Table 3 Micro Theories and their attributes

Political phenomena occur in a multidisciplinary environment, including not only the specific political factors but also the economic, sociological, psychological, and even technological factors that can motivate political behavior. To explain terrorism or failed states, for example, one single political theory will not suffice. In the absence of any grand, unified field theory, analysts must combine best-in-breed theories. Theories are building blocks that researchers combine in various ways to model different political phenomena. In order to do so, the inputs and outputs of the theoretical blocks must be consistent and interlocking so that they can be combined in meaningful ways.

Once the foundation of a political model has been laid with the building blocks of theory, however, an analyst must determine what will be constructed upon that foundation. What is the artificial environment in which individuals, groups, or nations will interact? For example, the methodological engines that model individual behavior can be game theoretical (Intriligator and Brito 1984; Zagare and Kilgour 2000; Powell 1987), microeconomic (Schelling 1960; Fearon 1994), or rule-based expert systems (Bennett and Stam 2000; Abdollahian and Alsharabati 2003; Gilbert and Troitzsch 1999). If using a conceptual representation of political bargaining space, then theories such as the median voter theorem (Kim and Morrow 1992; Bueno de Mesquita 1980), subjective expected utility comparisons (Edwards 1996; Camerer and Lowenstein 2003), and Arrow-Pratt risk aversion (Pratt 1964) or Prospect Theory (Battalio et al. 1990; Cacey 1994; Kahneman and Tversky 1992; Levy and Levy 2002) are among many that are commonly used by political modelers. The next section examines some of the typical approaches for creating artificial environments or models in which political phenomena at the micro-, intra-, and interstate levels may be simulated and tested.

3 Key Approaches to Conflict and Cooperation

Below are surveyed a few of the typical best-in-breed political science approaches, detailed above in Tables 2 and 3. We first explore the macro analysis of conflict and cooperation by using nation-states as the unit of analysis to understand conflict behavior among and between nations. We then turn to a few main theories that explain the nation-state itself and how its economic factors, demographics, and other national indicators can lead a nation to war or peace. Finally, several key theories that drive micro-, individual-level behavioral dynamics will be mentioned to explain how preference, behavior, and perception of individual leaders in a nation can be combined to anticipate politics, peace, or conflict.

3.1 Macro Interstate Approaches

One of the earliest models of relations between states was the so-called arms race model originally introduced by Richardson (1960). Such models investigated the dynamics of nations’ armament buildups by using coupled differential equations. Richardson’s equations posit a simple deterministic relationship between two states based on action and reaction, in which a small buildup by one side would lead to a larger counter by the opponent. Thus, an arms race could produce a wide gap in capabilities; this gap in capabilities was expected to prompt conflict. The equations define a precise movement of armaments through time, in which the pace of armament or disarmament is a function of how far one nation is away from its long-run equilibrium point. Depending on the initial conditions, the equilibrium can be peace or war. An “equilibrium point” in a dynamic system is a solution for the equations that does not change with time. The “initial condition” is the value of the variables at the onset of the simulation. Thus, the rate of armament buildup is expressed as:

$$ \rm{d}x/\rm{d}t=ay-bx+g$$

$$ \rm{d}y/\rm{d}t=cx-dy+h$$

where

• 1. dx/dt (for nation x) and dy/dt (for nation y) are the rates of armament

• 2. x and y are the amount of armaments

• 3. a and c are “threat” parameters

• 4. b and d are “fatigue” parameters

• 5. g and h are “grievance” parameters

Based on the value of the model’s parameters, the nations experience either “runaway” (i.e., unchecked) armament or disarmament (based on the model’s initial conditions) or convergence at an equilibrium point. Stability in the system is determined by whether states place relatively more emphasis on the threat of the other nation’s arms or on fatigue from armament buildup and expenditure. Although the Richardson model represents only an early attempt to simulate and predict political behavior, his work influenced later scholars.

A major extension of this perspective led to the evolution of deterrence theory at the macro interstate level. One of the cornerstones of deterrence theory was laid early in the Cold War by Brodie (1959), who urged a focus on deterrence rather than victory, as nuclear warfare is “too large-scale, too menacing to all our hopes.” The expectation that nuclear terror can credibly compel potential opponents to avoid confrontations is rooted in the high cost of nuclear war; thus, the implication of deterrence theory is that nuclear arms races ensure peace. This classic notion behind Mutually Assured Destruction (Huth and Russett 1990; Waltz and Sagan 1995) – that nuclear proliferation leads to highly stable international conditions – was refined by deterrence scholars such as Intriligator and Brito (1981). They assume that when nations anticipate that the costs of war will exceed a threshold above which said nations are not willing to initiate conflict, nations will fight only in self-defense. When a second threshold is exceeded, a nation is no longer willing to confront the opponent, and that nation will be deterred from war or yield to the aggressor’s demands. Therefore, the possibility of war exists only when costs are “acceptable.” Unstable conditions occur when contending actors have only conventional capabilities and cannot impose sufficient costs to deter opponents. In their research, this scenario is divided into four stages. The first is called the “Cone of War,” in which nuclear parity stabilizes world politics. Second is “Massive Retaliation,” in which one side initiates a nuclear buildup. Third is the “Balance of Terror,” in which both nations have nuclear capabilities, but their arsenals are not large enough to assure retaliation if the other side attacks preemptively. This stage is tenuously unstable, until a credible second-strike capability is developed. The fourth and final stage is “Mutually Assured Destruction,” in which equality of nuclear capabilities and secure second-strike capabilities on both sides minimizes the likelihood of war because the costs become unacceptably high. The basic model is as follows:

$$ \rm{dM}a/\rm{d}t=-a\rm{M}a-{b}^{\prime }b\\rm{M}b\times fb$$

$$ \rm{dM}b/\rm{d}t=-b\\rm{M}b-{a}^{\prime }a\\rm{M}a\times fa$$

$$ \rm{dC}a/\rm{d}t=\left(1-{b}^{\prime }\right)b\\rm{M}b\times vb$$

$$ \rm{dC}b/\rm{d}t=\left(1-{a}^{\prime }\right)a\\rm{M}a\times va$$

where:

Country: a, b

Initial time: t = 0

M(t): missile Stocks

C(t): casualties

α, β: the rate of firing a country’s missiles

α′, β′: proportion of counterforce (against enemy missiles) attack

(1 − α′), (1 − β′): proportion of counter value (against enemy cities) attack

f: effectiveness of missiles against enemy missiles

v: effectiveness of missiles against enemy cities

The implication of dynamic deterrence is that during the development of nuclear capabilities, one nation cannot fully deter the other and war is possible. It is in the last stage at which mutual destruction is assured that the cost of war prevents both nations from initiating war. As one nation acquires weapons unilaterally, however, the path to the last stage is very unstable. Deterrence is therefore not entirely stable; additionally, because terrorists and other violent nonstate actors do not have a “return postal addresses,” deterrence is not credible toward such threats. Deterrence is the ability to prevent attack by a credible threat of unacceptable retaliation. The calculation of the cost of war is the main rationale in deterrence; thus, the theory is difficult to apply to violent nonstate actors. One nonobvious insight from the Intriligator and Brito theory is that a strategy that assures retaliation and minimizes communication among contenders may produce conflict.

Organski (1958) proposed that nations would fight when they are dissatisfied with international norms and hold equal capabilities. This is based on the assumption that hierarchy exists in the international system. Here, hierarchy is defined as a system in which a dominant nation (the “defender” or “dominant power”) is at the top of an international power hierarchy, with “great powers,” “middle powers,” and “small powers” under the dominant power. In this power transition theory (PTT), power is measured relatively (in comparison to other states) based on demographic and industrial indicators, where Relative Power = GDP × Population × Political Capacity where Political Capacity measures the state’s domestic control, a ratio between anticipated and actual tax receipts. Additionally, the dominant power enforces the status quo of the international system, while lesser powers are either satisfied or dissatisfied with the status quo.

PTT anticipates interstate dynamics by analyzing this relative power distribution across the international system and the member’s satisfaction with the status quo. Under conditions of parity and dissatisfaction, the theory predicts the highest probability of international conflict; when nations dissatisfied with the status quo accrue enough power to challenge the dominant nation, PTT postulates that war is most likely. For example, in the middle of the Cold War, a PTT-based analysis (Organski and Kugler 1980) concluded that the conflict in Europe would not be repeated because of integration, that the USSR would fall from the rank of competitors by 2000, that China would emerge as the leading challenger to the United States, and that the political center would shift from the West to Asia by the end of this century. The dominant power is committed to defending the international treaties and norms that constitute the status quo, which reflects the dominant power’s preferences (as it is the most powerful nation within the international hierarchy).

Using a system of symmetric, coupled nonlinear differential equations, Abdollahian and Kang (2008) formalized and tested a system-dynamic model to identify to what extent and degree policymakers can maintain stability in rival dyads, such as the U.S.–China case. Their model explores some of the structural conditions of how conflict or cooperation affects the growth and transition from the PTT literature. The work suggests specific, strategic policy prescriptions for managing conflict or cooperation and highlights the nonlinear and nonmonotonic effects of foreign policy actions.

The entire power parity model system of nonlinear ODEs is the combination of the following equations:

$$ \frac{\rm{d}{P}_{\rm{D}}}{\rm{d}t}={B}_{\rm{D}}{P}_{\rm{D}}\left(1-\left({P}_{\rm{D}}+{P}_{\rm{C}}\right)\right)-{H}_{\rm{D}}{C}_{\rm{C}}$$

$$\frac{\rm{d}{C}_{\rm{D}}}{\rm{d}t}=-{S}_{\rm{D}}{P}_{\rm{D}}{C}_{\rm{C}}\frac{1}{\sqrt{2ps}}\mathrm{exp}-\frac{1}{2}{\left(\frac{{P}_{\rm{D}}/{P}_{\rm{C}}-1}{s}\right)}^{2}\frac{1}{2.50599}$$

$$ \frac{\rm{d}{P}_{\rm{C}}}{\rm{d}t}={B}_{\rm{C}}{P}_{\rm{C}}\left(1-\left({P}_{\rm{D}}+{P}_{\rm{C}}\right)\right)-{H}_{\rm{C}}{C}_{\rm{D}}$$

$$ \frac{\rm{d}{C}_{\rm{C}}}{\rm{d}t}=-{S}_{\rm{C}}{P}_{\rm{C}}{C}_{\rm{D}}\frac{1}{\sqrt{2ps}}\mathrm{exp}-\frac{1}{2}{\left(\frac{{P}_{\rm{C}}/{P}_{\rm{D}}-1}{s}\right)}^{2}\frac{1}{2.50599}$$

where:

${P}_{\rm{D}}$is the systemic power level of the dominant nation.

${P}_{\rm{C}}$is the systemic power level of the challenger.

${B}_{\rm{D}}$is the national growth rate coefficient of the dominant nation.

${B}_{\rm{C}}$is the national growth rate coefficient of the challenger.

${H}_{\rm{D}}$is the dominant nation’s cost coefficient for competition.

${H}_{\rm{C}}$is the challenger’s cost coefficient for competition.

${C}_{\rm{C}}$is the conflict level that the challenger targets toward the dominant nation.

${C}_{\rm{D}}$is the conflict level that the dominant nation targets toward the challenger.

${S}_{\rm{D}}$is the foreign policy stance of the dominant nation toward the challenger.

${S}_{\rm{C}}$is the foreign policy stance of the challenger toward the dominant nation.

$s$is the parity variance condition coefficient.

The variables in the power parity model include systemic power levels, conflict levels, foreign policy stances, and the value of the parity ratios for a rival dyad. The parameters in the power parity model are the national growth rates, the cost of competition, and the parity variance condition. By varying the parameter values and initial conditions of variables for rival dyads, an analyst can explore the performance of the dynamic model under various circumstances, not only reconstructing historical relationships between dyads but also forecasting simulations. Figure 1 demonstrates the policy results using U.S.–China data (Abdollahian and Kang 2008).

Fig. 1

U.S.–China dyad

Figure 1 depicts a scenario in which China adopts a highly hostile foreign policy stance (${s}_{\rm{c}}=-0.9$), and the U.S. policy response is allowed to vary. Notice that at aggressive U.S. foreign policy response values, the effects of competition on systemic power levels produce a significant, detrimental impact on both countries. As the United States begins to question the rise of China, small changes in the firming of the American policy stance produce sharp increases in dyadic conflict. Hence, the structural stage is set for prompting early conflict initiation and war escalation. At the other extreme, an acquiescent foreign policy stance toward China produces sustainable levels of systemic power for a while, although a Chinese overtaking is guaranteed within about 15 years. In this case, after China surpasses the United States in systemic capabilities, possible minor conflicts or hostile incidents are still expected between the two countries. At a neutral foreign policy stance, levels of U.S. conflict remain low throughout the transition period as a result of small changes in the U.S. conflict equation. Under these simulation conditions, only a neutral U.S. policy stance can secure the window of opportunity for peace and stability.

3.2 Macro Intrastate Approaches

The systematic empirical research on intrastate and nonstate conflict has a very long and distinguished record based on innumerable case studies and a vast array of alternative propositions accounting for the rise of nation-states and the emergence of institutions and dissatisfaction in the polity (Brinton 1952; Crenshaw 1995; Huntington 1968; Gurr 1974; Diamond 1992; Inglehart 1997; Welzel et al. 2003). These contributions span the fields of not only political science and economics but also sociology and cultural anthropology. Tilly (1975) in a classic assessment and Poggi (1990) summarize systematically the process of state formation. Barnett (2004) links the motivations of international terrorism to economic modernization of states. Lemke (2009) shows that when a nation emerges from a cooperative aggregation of states – such as the unification of the Italian states in the nineteenth century or expansion of the United States – that legacy leads to relatively stable and evolving governance. On the other hand, when the birth of a nation or its reconstruction is associated with serious conflict, insurgencies and developmental lags are introduced. Nation-building propositions can profit from such long-term assessments, but understanding the political motivations and mechanisms, let alone modeling causation, remains submerged in such summaries. For a recent general review, see Midlarsky (2009).

Davies (1962) was among the first to systematically relate insurgency with an inverse U curve of development. He demonstrated that internal instability was not likely among the least and most developed societies but maximized among the less developed societies, particularly those undergoing fast economic development. A large related literature subsequently developed relating opportunity to the likelihood of insurgency to predict conflicts based on the economic and political incentives or constraints they face (Grossman 1991; Collier and Hoeffler 2002, 2004; Elbadawi and Sambanis 2002; Fearon and Laitin 2003; Fearon 2004; Barnett 2004; Hegre and Sambanis 2004; Abdollahian et al. 2009). A number of explanatory variables, including ethnicity, culture, absolute deprivation, language, and race are discarded, while economic well-being, the strength of political institutions, and reliance on commodity exports – mainly oil – are consistently associated with conflict. Thus, affluent societies that have institutionalized effective governance, and do not rely on exports of commodities such as oil, are least likely to experience insurgencies. Unfortunately, this literature is, however, confounded by the lack of reliable and consistent historical data for most countries. The revival of interest in insurgency studies in the early 2000s refocused researchers on how nonstate actors can generate intrastate instability. Previous studies had demonstrated a very weak link between civil war and the initiation of international conflict (Tanter 1966; Rosenau 1964), and for that reason much of the earlier research focused on intrastate conflict at the state-society level. Demands to link intrastate with substate actors and interstate conflict forced most researchers to rely on national rather than subnational data for their exploration. Collier and Hoeffler (2004), for example, show that “greed” rather than “grievance” is associated with the initiation of intrastate conflicts. The causal relation from “grievance” and “dissatisfaction” to domestic instability is established but not directly related to the source of conflict. The concern here is that substate actors and their representative populations that have “grievances” or are “dissatisfied” with fiscal performance or political governance are not directly identified. Rather, differences across nations help to determine the likelihood of intranational instability. Buhaug and Gates (2002) among others challenge such results showing that applying aggregate measures says little about whether conflicts are located in these areas. This leaves a void where integration of political models across levels of analysis can help.

A second major contribution to the emerging understanding of intrastate instability is driven by the contribution of Fearon and Laitin (2003) that found, contrary to most case study results, little relation between conflict and ethnic, religious, linguistic, or cultural differences. Instead, they show that rough conditions identified by Guevara (1968) are cross-national correlates of insurgency and guerrilla activity. Based on these aggregate results, geography and the flow of populations are used to explain intrastate conflict (Hendrix and Glaser 2007; Salehyan and Gleditsch 2004). The standard argument is that rough terrain confers tactical advantages on insurgents by mitigating the advantages enjoyed by state armies, which can mobilize disenfranchised groups to rebel. Outnumbered and outgunned insurgents can avoid direct engagement and gain access to safe havens where they can recruit and replenish supplies. Collier and Hoeffler (2002) find that mountainous, forested terrain aids in insurgencies. Likewise, Fearon and Laitin (2003) find a significant positive relationship between rough terrain and the onset of conflict. Most likely, climate and terrain provide the preconditions for effective insurgencies but are not and cannot be the variables that cause intrastate conflict as geography does not significantly vary over time. Raleigh (2004) is right on point when he challenges the aggregate approaches, arguing that modern insurgencies are as likely to be based in urban areas as rural ones that provide excellent safe havens. Moreover, safe havens in neighboring states where porous borders exist would make rough terrain irrelevant. He argues that weak states, defined as those with low GDP per capita and weak political institutions, are limited in their ability to project political authority regardless of terrain. Fragile states rather than geography place governments at risk, and this is shown in more recent work that suggests economic, demographic, natural resources, and political factors trump geographic variables (Humphreys 2005). Controlling for these foreign safe havens and economic and political development shows that rough terrain, as defined by Fearon and Laitin (2003), is not a significant predictor of conflict onset (Bahaug 2002, 2004; Rodrik 2004; Engerman and Sokoloff 2002).

Recent work focuses on the causes of domestic instability as based on insights from the long case study record and on intrastate and stakeholder data that emphasize differences within a national unit. Cederman (2004) links ethnic groups that inhabit mountainous terrain suggesting national formation. Using agent-based modeling techniques, he finds that violent separatist movements are much more likely to occur in mountainous terrain and tropical climates that provide shelter for guerrilla activities when ethnic groups are hierarchically organized and not otherwise. This type of approach to identify the pathways of a nation moving down the road to intrastate conflict requires a detailed, subnational level analysis at the provincial, district, or individual level. Our approach starts with the respecification of successful models that account for international conflict at the intrastate level.

3.2.1 Applications: Relative Political Capacity

For a detailed example, we first examine a key political indicator of relative political capacity and then explore one structural model of domestic political economy called the Politics, Fertility, and Economic Development model [POFED] (Feng et al. 2000). Relative political capacity (RPC) is the ability of a government to extract resources from its population as evidenced by relative performance of actual versus expected tax collection efforts for a given level of economic development (Arbetman and Kugler 1997). Recent work (Arbetman and Johnson 2008) on the dynamic effects of changes in political capacity suggests that as a government loses its ability to extract resources and advance its goals, the potential for competitors willing to fill that gap rises. Unexpectedly, as the political capacity of the challenger rises, a competitor usually replaces the government. If the competitor gains footing, the political capacity rises anticipating the lowering in instability. When the new government establishes control and achieves normal levels of political performance, this cycle of instability comes to a close. The pattern suggests a relationship between a government’s level of political capacity, changes in the level of political capacity, and intrastate instability. Not only do the levels of political capacity matter but also the rates of change as shown in Fig. 2.

Fig. 2

Political capacity

Here, we see political capacity changing from positive to negative, in both level and rate, from 1996 to 2005 for a particular nation with the associated size of participants in demonstrations and those killed or wounded in such (Kugler et al. 2008). Disaggregating political capacity to the provincial level within nations shows even more clearly the areas from which a national government will be challenged. Arbetman and Johnson (2008) show that without a strong central government presence, provincial governments face a political challenge from groups that are themselves capable. Such information is essential in assessing to what degree providing economic assistance to an area – such as Darfur in Sudan – would limit casualties without destroying the central political foundations required for continual stability.

Clearly, instability results from the interaction between economic growth and political capacity, so the linkages between political capacity and domestic political economy are crucial. Here, we look at an example of a structural model of detailed domestic intrastate politics, POFED which highlights the effects of growing political resources, economic constraints, and demographic pressures on the promulgation of conflict (Feng et al. 2000; Kugler et al. 2005). The model grew out of extant literature on modernization, human capital formation, institutional capacity, and economic development as a means of tracing the dynamic interrelationships between productivity, fertility, political effectiveness, and social stability. By using a statistically validated system-dynamics approach for the behavior of individuals and policymakers in a dynamic world focusing on antecedents for state failure and insurgency, POFED accounts for the political and economic structural environments that cause a country’s living standards and political position to grow or decline.

The interplay of political, economic, and demographic indicators is modeled in POFED to anticipate the impact of interventions in fragile states and can thus be used to identify direct policy levers to mitigate state fragility. In addition to identifying policy, investment, and business actions that impact structural conditions to decrease state fragility, POFED can provide detailed tactical leverage points when applied at higher geospatial resolutions, such as provincial or district level analysis. As shown in Fig. 3, POFED has five major components that capture factors of state fragility and the effects of potential intervention: Income (y), Fertility (b), Human Capital (h), Instability (S), and Political Capacity (X). POFED is a dynamic general-equilibrium model based on the intersection of political and economic maximization function. The models show maximizing behavior of individuals seeking to maximize their lifetime utility by choosing how much to consume and save and how many children to have, while policymakers choose the tax rate and the amount of public investment and military spending to maximize their chances of remaining in power. In equilibrium, prices move endogenously so that supply balances demand for all goods and investments in the economy, while policymakers set fiscal policies. Further, the political-economic market equilibrium is a dynamic curve in the phase space tracing the evolution of the political-economic system. The model is specified as:

Fig. 3

POFED model components

$$ {b}_{t}=B{y}_{t}$$

$${y}_{t+1}=A{y}_{t}{(1-{S}_{t})}^{a}{c}_{t}{h}_{t}^{1-a}$$

$$ {h}_{t+1}=w\ {h}_{t}/{b}_{t}$$

$$ {S}_{t+1}={S}_{t}d{c}_{t}({c}_{t-1}/{c}_{t})$$

$${c}_{t+1}=C{y}_{t}{c}_{t}{p}_{t}^{f}/{b}_{t}^{{g}^{f}}$$

These equations show that birth rates b depend on income y; and that income depends on past income and political conditions, x. h shows the generational feedback on the creation of human capital, while political instability, S, has a temporal feedback and depends on external policy p _s. Similarly, political capacity, x, depends on per capita income y, external policy p, and births b.

The intuitive logic of POFED is as follows. In addition to well-established economic determinants, the fundamental political variable of political capacity alters fertility decisions, human capital accumulation, and economic development. In fact, fragile developing societies are defined by a decline in per capita income, by the potential for falling into the poverty trap, and by the low or declining capacity of governments (Guillaumont and Jeanneney 1999, Kugler and Tammen 2010). Robust societies with higher levels of political capacity extract more than anticipated from their economic endowment and allocate such resources efficiently to advance the government’s priorities; fragile societies that fall below average political capacity levels of similarly endowed societies fail to do so. Some of the key general policy prescriptions are summarized as follows:

• Sufficient political capacity is a necessary precondition for income growth for poor countries.

• Income growth is self-reinforcing: when birth rates fall, human capital rises and political instability declines.

• The poverty trap is self-reinforcing; when birth rates rise, political instability increases and income falls.

• External aid, policy interventions, and domestic policies can increase stability and promote income growth.

• Increasing political capacity and change in political capacity reduce political instability.

• Income falls and birth rates rise when political capacity is lower and political instability is higher.

• There are thresholds of political capacity and political instability driven by economic performance that can cause a state to fail.

3.3 Micro Individual Approaches

The preceding sections have examined how conflict and cooperation between, among, and within nations can be modeled based upon structural theories using the nation-state as the unit of analysis. The effects of structural variables, such as fertility, unemployment, or public opinion, on political outcomes, however, are often realized gradually over time. If the analytical question being modeled concerns the near future, or if modeling the subnational interactions of individuals or groups, then a micro-level theoretical approach is best applied.

The major difficulty in the political analysis of individuals is that human behavior is inherently difficult to predict, and we are all different from diverse cultures, religions, and political persuasions. Thus, our political preferences are not universal. Although physical security is the fundamental objective of all states, the relative priority of policy considerations ancillary to security differs widely between societies and between individuals. The deeper analytical challenge, however, is in modeling not only the competition between individual preferences, but also the origin of those preferences. For instance, how do individuals order their political preferences in the face of risk? How do individuals determine the utility of different actions amidst uncertainty? Focusing on individuals as primary actors in policymaking, several theories of microeconomics have been employed for micro-level modeling under the rubric of positive political theory. For example, the idea that risk and uncertainty may play a pervasive role in economic analysis was originally suggested by Frank Knight in his 1921 study of insurance markets. In Risk, Uncertainty, and Profit, Knight observed that the distinction between risk and uncertainty was based on whether risk can be expressed in a specific mathematical probability. If so, then risk becomes insurable, and if not, it becomes an unmeasurable probability (i.e., uncertain). This notion has also been incorporated into the study of political modeling, leading to explanations of differences between individuals’ calculations about the utility expected from taking certain actions.

Jonn von Neumann and Oscar Morgenstern (1944) investigated the motives of an individual making a decision under risk in the Theory of Games and Economic Behavior. In their theory, individuals are assumed to be facing a “choice set” of alternative probability distributions (or, in another expression, lotteries). Von Neumann and Morgenstern enable us to model the agent’s preference over alternative probability distributions by differentiating between a gamble and a lottery. In a state of nature, an individual’s utility is dependent on uncertainty as well as on the monetary payoffs. However, under uncertainty, the decision-maker is forced into a gamble in which it is impossible to ensure that every decision maximizes his utility. Conversely, a lottery enables individuals to calculate the probabilities assigned to choice. A simple lottery is denoted as follows:

$$ p\times x+\left(1-p\right)\times y$$

The above equation can be translated as “the individual receives prize x with probability p and prize y with probability (1 − p).” The prizes may be money, bundles of goods, or even further lotteries. The expected utility property says that the utility of a lottery is the expectation of the utility from its prizes. We can compute the utility of any lottery by taking the utility that would result from each outcome, multiplying that utility times the probability of occurrence of that outcome, and then summing over the outcomes. However, when the probabilities are assumed to be subjective instead of objective, then probabilities are degrees of belief in a proposition rather than a set of events that is inherent in nature. Thus, individuals should update the calculation of probabilities in light of evidence. Bayes’s theory of subjective probability nullifies Knight’s distinction by reducing all uncertainty to risk through the use of beliefs expressible as probabilities (Earman 1992). This theory argues that even if states of the world are not associated with recognizable, objective probabilities among gambles, decision-makers still behave as if utilities were assigned to outcomes, probabilities were attached to states of nature, and decisions were made by taking expected utility.

After the axiomatization of the expected utility hypothesis by von Neumann and Morgenstern, Milton Friedman and Leonard Savage (1948) advanced the concept of univariate risk propensity by analyzing economic issues within an expected utility framework. The spectrum of risk (with risk aversion at one end and risk acceptance at the other) is demonstrated once again by a gamble. For risk-averse individuals, the utility of the expected value (i.e., the “ante”) is higher than the expected utility of the gamble due to the inherent risk of loss. A level of certain wealth provides the same utility as does participating in this gamble. We call this the “certainty equivalent” of the gamble: the amount a person would take for certain rather than play the gamble. The individual will be willing to pay anything up to some value relative to the ante to avoid participating in the gamble. We call this the “risk premium,” the amount that a person would pay to avoid playing the gamble. This explains why people buy insurance. Even when these costs are paid, the risk-averse person is as well off as he would be if forced to face the world (or the gamble) uninsured. At the opposite end of the spectrum is the risk-loving individual who prefers a lottery to its expected value.

Kahneman and Tversky (1979), however, showed that individual decisions are made evaluating gains and losses separately rather than in consideration of aggregate totals. This occurs because people perceive improvements or deterioration in their welfare differently; individuals may also misperceive the probabilities underlying their decisions. The two main propositions of this theory are (1) that individuals make decisions based on changes in wealth rather than their total wealth (which is in direct contradiction to expected utility), and (2) that risk aversion does not universally prevail, as some individuals are risk-seeking regarding loss.

The perception of utility in risk assessment demonstrates how expected utility theory can be applied to models of decision-making amid uncertainty. Bueno de Mesquita (1985) asserts that once a crisis develops, calculations of net gain accurately account for the escalation and termination of disputes. He shows that once a crisis starts, the analysis of a nation’s net gains by an individual leader distinguishes between asymmetric and symmetric wars, anticipates when wars will be limited and when they will escalate, determines when confrontations will remain bilateral and when they will become multilateral, and indicates how a war will terminate.

One approach to modeling political phenomena amidst the uncertainty produced by competition between individuals with diverse preferences is the “stakeholder” family of models. Stakeholders are individuals who either have the power to influence an outcome of a decision or are deeply interested and thus active in an issue. Stakeholder models assume that individuals are “utility-maximizing rational agents” – in which utility-maximizing means that individuals will seek to enact their preferences, and rational means that those preferences are ordered (i.e., outcome A > outcome B > outcome C). One of the earliest predictive stakeholder approaches was the Prince model, originally constructed by William Coplin and Michael O’Leary (1972). A rational agent model, Prince attempts to predict political events based on the interests of the parties significant to the outcome of the event. The Prince model requires an informed observer to evaluate the orientation toward certainty of position regarding, power over, and salience of an issue to stakeholders capable of influencing event outcomes. This pencil-and-paper model lacks the fidelity of later extensions, but it was the first reasonable way to diminish uncertainty regarding the outcomes of political competition between individuals and their respective policy preferences.

3.3.1 Senturion: A Micro Dynamic Model of Politics

Several of the approaches discussed in previous sections have been formalized into dynamic or computational models (Bueno de Mesquita 1985; Bueno de Mesquita and Stokman 1994; Kugler and Feng 1997). Here, we discuss in detail one example of such an approach, Senturion (Abdollahian et al. 2006), a tool that can help policymakers and analysts predict political events and anticipate domestic or international political stability levels, as well as analyze specific investment decisions in which political matters affect outcomes. Senturion is a simulation system that analyzes the political dynamics within local, domestic, and international contexts and predicts how the policy positions of competing interests will evolve over time. The underlying methodology relies on several micro-level theoretical blocks. The set of rules used by Senturion synthesizes several classes of political science and microeconomic theories drawn from game theory, decision theory, spatial bargaining, and microeconomics. Unlike a statistical or probabilistic approach to predictive modeling, Senturion employs a set of micropolitical algorithms in sequence. Each theory provides a functional component for modeling how agents interact to model the “pulling and hauling” of political processes.

Given a particular issue, such as the attitude of stakeholders toward providing government-subsidized health care or the attitude of stakeholders toward U.S. military operations in Afghanistan, the Senturion approach facilitates subject matter expert (SME) identification of the positions of critical stakeholders on policy issues, weighs their potential influence, and assesses the strength of their commitment or advocacy of a policy position. This SME-generated data input captures a snapshot of the current political landscape. Given a particular landscape, several theoretical building blocks are useful to simulate complex human behavior and animate that landscape forward for predicting politics. We can build models of “heavy” agents, those with several initial political attributes, with the following six qualities adapted from Gilbert and Troitzsch (1999):

• Knowledge and beliefs: Agents have priors on the political environment in which they are situated. In the Senturion approach, the initial data on the political landscape generated by stakeholder attributes, such as opponents’ and supporters’ policy positions and potential to influence, is known among all other stakeholders (Coplin and O’Leary 1972).

• Inference: Agents can also make inferences from their knowledge about which potential actions to take and which ones are more credible than others as well as anticipating how other agents will react. Here, notions of risk are used to drive potential misperceptions of agent inferences, as social modeling necessitates the inclusion of political perceptions and misperceptions.

• Social models: Senturion uses the notions of Black’s political median (1958), Arrow-Pratt risk aversion (Pratt, 1964), and game-theoretical models (Bueno de Mesquita 1985; Lalman and Bueno de Mesquita 1989) to model various types of stakeholder interaction games theoretically at different points.

• Knowledge representation: Agents update their beliefs about their own political effectiveness based on how successful their efforts are with other agents.

• Goals: Each stakeholder has a preferred policy outcome that he or she is trying to achieve. Senturion assumes that agents are rational utility maximizers trying to achieve their desired political outcomes subject to being part of a winning coalition.

• Language: Senturion uses the medium of political proposals, moving or shifting from one political position to another, based upon real or perceived political pressure to represent the language of agents’ interactions.

The Senturion approach models the intuition behind each stakeholder’s political calculus in political discussions by breaking down the process into subelements that can be modeled. Each element models a particular part of the decision process, and by combining the elements sequentially, the approach can anticipate how all stakeholders will interact to arrive at a particular decision or political outcome. The approach is a dynamic and recursive estimation of how stakeholders will interact and the resulting compromises and coalitions that will form in response. Table 4 lists Senturion’s component theories and their various attributes.

Table 4 Senturion theories and elements

Figure 4 provides an overview of Senturion. The initial stakeholder environment is defined as a policy issue of political interest; for example, the range of feasible levels of budgetary allocation in dollars that stakeholders will compete to influence. This is where individuals vie and compete to influence the ideas and actions of others to support their own claims and political positions. This one-dimensional environment is populated with agents that represent stakeholders that have a potential to influence the particular political issue. These can be individuals, political parties, governments, or members of society. Each of these stakeholders has different attributes, such as a preferred political outcome that locates them in the spatial context, as well as a separate potential to influence that outcome weighted by their salience on the particular issue. This creates a snapshot of the political landscape that is quite similar to polling data in American politics or consumer preferences in market surveys.

Fig. 4

Senturion process

The second step is to apply micro rules and equations to the agents given their individual attributes, the knowledge and beliefs they form given that particular snapshot of the landscape to influence their social interactions to animate the landscape and ultimately their anticipated behavior. Thus, Senturion first starts with locating the political center of gravity, called the median from the spatial bargaining context. Intuitively, the political center outlines the place where compromise can most likely occur. If one knows what the winning compromise position will be, then we can begin to deduce several other key ideas.

Here, the median position is recognized as the safest position politically, while positions far from the median are more risky. If it is known which stakeholders are willing to take risks, they may be willing to take bigger gambles to get what they want or “hold out,” while if they are not willing to take risks, they may be more willing to “sell out.” This assumes that more extreme stakeholders are willing to take risks while stakeholders near the political center are willing to make deals in order to achieve an agreement.

Risk-taking propensities subsequently distort how stakeholders will view each other. With these distortions, Senturion estimates the pulling and hauling of the political process by a behavioral game tree. The game structure looks at the anticipated gains or losses of every pair of stakeholders on the particular issue, identifying where offers or compromises will be exchanged between two stakeholders. It then looks at the entire network of proposals among all stakeholders given the pairwise game-theoretical interactions in order to anticipate which stakeholders will revise their positions to produce the third step of iterative dynamics.

Given that stakeholders’ positions may change, how has this changed the median? If the median has changed, how have risk profiles changed, with associated impact on perceptions, proposals, and resulting position shifts? Senturion iterates the process to simulate the evolution of political dynamics over time.

One benefit of this approach is that it provides a consistent framework for objective analysis of stakeholder perceptions rather than relying solely on individual expert opinions about political outcomes. Moreover, as with any simulation tool, the specific dynamics of stakeholder proposals surrounding particular political issues can be examined in order to first gauge whether outcomes are politically feasible, second to determine possible strategic options for optimizing political outcomes using knowledge about the stakeholder dynamics, and third to anticipate unintended consequences (second and higher order effects) of actions.

3.3.2 Defining the Political Landscape and Generating Data

The process starts with representing agents in a political state space as opposed to physical environments (Gilbert and Troitzsch 1999). Adopted from economics and positive political theory, Senturion draws from spatial analysis the unidimensional issue(s) that comprise a particular political or strategic problem (Luce and Raiffa 1957; Riker and Ordeshook 1968; Ordeshook 1986). Following Feder’s (1994) and Stokman’s (2000) processes in collective decision-making, Senturion decomposes any strategic decision problem into its requisite parts in order to define one or multiple issue spaces to populate with agents. Agents are then populated on the landscape with varying attributes given subject matter inputs as described below.

• Desired issue position

• Potential power or influence over the political outcome

• Salience or importance of particular political issue relative to other concerns.

As described above, issues are unidimensional ranges of political outcomes, such as support for a particular reform policy, levels of preferred taxation, or stability. Power is defined as an actor’s capability to affect outcomes, position is each actor’s desired issue outcome, while salience measures the importance or how much of the actor’s agenda the issue occupies (Coplin and O’Leary 1972; Feder 1994; Bueno de Mesquita and Stokman 1994; Kugler and Feng 1997). Thus, stakeholders now have particular influence, importance and positional attributes that could be assigned and scaled to arrive at a relative ranking of political viability but not actual political ­outcomes. This “snapshot of the political landscape” (Fig. 5) shows stakeholders’ initial attributes and is subsequently processed and animated by computational processes.

Fig. 5

Stakeholder data

3.3.3 Overview of Senturion Algorithms

Given the generation of the stakeholder political landscape, the Senturion algorithm computes several key components used in various steps. It computes Votes and Forecast, Risk, and Power, expected utility values in a game tree, the resulting Perceptions, subsequent Proposals, Learning from interactions and finally a discount function to determine if agents will continue to interact. Figure 6 outlines the general algorithm and process.

Fig. 6

Senturion algorithm

Votes and Forecast capture the support that every stakeholder gets from every other stakeholder. These are used to compute the Median position that is the safest position politically. Stakeholder votes are simply computed by weighing each stakeholder’s potential to influence multiplied by their particular salience or importance of the issue to the stakeholder. Thus, the Forecast is the most preferred ­stakeholder position given all votes. Black (1948) originally proposed the median voter theorem, which identifies the median as the winning position on unidimensional continua among all other alternatives. Enelow and Hinich (1984), Bueno de Mesquita (1985), and Hinich and Munger (1997) suggest how to weight agents’ votes in the context of policy applications.

Risk is a key concept that introduces distortions among stakeholder knowledge of the particular political landscape and allows for the incorporation of perceptions to help drive different political dynamics. Risk is computed for each agent to determine the perceptual prism through which the agent views other individuals and introduces distortions to the way individuals will interact. Simon (1955) outlines the evaluation of alternatives in terms of gains and losses relative to a reference point such as the status quo. Thus, risk-taking attitudes can be different above or below this point. Risk-taking propensity is assumed to be individually symmetric around losses and gains. If an agent is risk-acceptant on gains, he or she is also risk-acceptant on losses, maintaining the same risk tendency on either side. Newman (1982) shows how to calculate risk for multiple stakeholders. Thus, every agent balances his or her interests of obtaining policy satisfaction versus the security of being part of a winning coalition (Morrow 1986; Lamborn 1991). It then follows to connect risk propensities back to the status quo, or in this particular case, Black’s weighted median for a particular distribution of the political landscape. Thus, stakeholders with positions farther from the median tend to be risk-acceptant while agents close to the median tend to be risk-averse.

Power measures the level of influence of each stakeholder given the likelihood of third-party support. Power is a dyadic value established by Singer et al. (1972) that introduces the notion of relative influence of stakeholders compared to all other stakeholders in a particular political process. Stoll and Ward (1989) explore this concept in detail with alternative measurements that produce effective relative measures of capabilities.

A generalized game tree of political interactions is specified and solved given the expected gains or losses for every pair of stakeholders from each stakeholder. Game theory allows us to specify the social model of political interactions among agents (Harsanyi 1968; Camerer 2003). Kadane and Larkey (1983) and Shubik (1983) show how actors choose to maximize utilities in a rational manner; a potential solution arises in decomposing a large n-person game with n-parallel two-person games. Senturion employs a generalized game that all stakeholders face in their interactions with all other stakeholders. Every agent interacts with every other agent and considers the possibility to challenge or not to challenge his or her opponent depending on the relative expected gains. When two agents decide not to challenge each other, the result is a stalemate. When two agents decide to challenge each other, the result is conflict. When one agent decides to challenge the other while his opponent does not want to challenge, then the result is a potential political offer that may or may not be accepted by the other stakeholder, as shown in Fig. 7.

Fig. 7

Game tree

Perceptions map out the net gains or losses in every dyad of agents, anticipating stakeholder interactions as peaceful, mutually conflictual, or in favor of one or the other. Based on the assumption that agents act according to their perceptions of a given political environment, each agent’s perceptions are paired together to produce the anticipated behavioral interaction between two stakeholders. Inspired by Jervis’s (1978) work on misperceptions, both Lalman (1988) and Morrow (1986) define a continuous outcome approach to map behavioral interactions. Thus, different game-theoretical outcomes are translated into a perceptual mapping that identifies the behavior relations of every stakeholder versus every other stakeholder on a particular political issue.

Proposals translate the particular stakeholder perceptions back onto the specific policy landscape as offers, pressures, and moves to which stakeholders are subject. As stakeholders may have any combination of positive and negative net EU gains given our perceptual mappings, we must sort through the network of all behavioral relations to identify the push and pull of political dynamics. This is the kind of communication that takes place during agent interactions in an agent-based computational approach. Lalman (1988), Bueno de Mesquita and Stokman (1994), and Kugler and Feng (1997) outline various conditions for stakeholder interactions. An offer is made when a stakeholder believes that there is some positive gain to be made, although this offer may or may not be credible based upon differences in risk perceptions. An offer is made when the driver perceives himself as being able to secure some gains by imposing on or bargaining with the target. In the former case, the driver makes the target move all the way to his position. In the latter case, the driver makes the target move closer but not all the way.

Stakeholders may also learn or update their beliefs about the political landscape given their interaction with all other pairs of stakeholders. Given stakeholder dynamics, information is transmitted through proposals that may create learning among all stakeholders. Stakeholders subsequently can update their beliefs about the political landscape, such as which offers were successful, that can affect future proposals depending on various rules.

Finally, a discount function determines whether the process is iterated again depending on whether stakeholders see gains from further interactions. As stakeholder dynamics may cause some stakeholders to move, their positions on the issue continuum are changed, and thus Senturion animates the evolution of the political landscape through simulated time over several iterations. This process could iterate indefinitely, but that would not accurately mimic the dynamics of political processes, as rules for termination of ABMs vary widely depending on the specific application area. Intuitively, stakeholders will stop the political process when they see no further value from continued interactions (Laibson 1997).

3.3.4 Case Study: Iraq Elections, 2005

This section details the findings from a project focused on support for the January 2005 elections in Iraq, using only open-source data from SMEs from the intelligence community and the Washington Institute for Near East Policy (Abdollahian et al. 2006). Table 5 summarizes Senturion predictions based on data collected by the end of December 2004 and compares them to actual events that unfolded over the following 2 months.

Table 5 Senturion predictions compared to actual events

Senturion predicted that a change of approach that made neutral Iraqis feel safer, by either coalition forces or the insurgency, would have allowed either the United States or insurgents to gain the support of neutral Iraqis. A major question facing analysts and decision-makers before the election was the role that other nations in the region might play, and how to assess the reactions to the election of major players in the international community. France, Russia, and Germany were expected to coalesce and increasingly support the election, but their impact would be minimal on Iraqi attitudes.

In assessing the insurgency in Iraq, Senturion provided two conclusions. Zarqawi and other foreign insurgents had very little leverage to undermine support for the election at this point. On the other hand, domestic insurgents, composed mainly of former regime elements, had most of the leverage in this situation in the months before the election. However, they did not recognize the extent of their potential influence. As with any simulation tool, this approach can also be used to test alternative political courses of action. The assumptions, policies, and tactics of U.S. stakeholders can be simulated to identify first- and second-order consequences and then adjusted to find the optimal approach to a particular situation. Moreover, because this approach calculates the perceptions of stakeholders, it can also identify circumstances when perceptions of key stakeholders are inaccurate. At times, such knowledge may form the basis for a plan of action to exploit such limits of perception.

To anticipate political reactions in advance of the 2005 elections, several courses of action to improve the situation were explored. First, a way to persuade Sunni tribal elements to moderate their opposition to the election was identified. Second, a way to obtain support from some former regime elements was sought. Finally, the implications of adjusting the force structures in Iraq were also explored by varying U.S. power levels. A reduced coalition military presence in Iraq would not have appreciably affected the attitudes of Iraqi stakeholders. However, increased coalition military strength in Iraq would have improved the attitudes of Iraqi stakeholders toward the election in the short term, by making them feel more secure.

4 Practical Advice

• Begin modeling by determining the specific political issue, event, or risk in question. For instance: Is it the relations between two countries or a large bloc as a whole? Or, is it the emergence of insurgent groups in general or the effects that a particular group may have on domestic capital formation in a specific region? Recognize that there will always be political variables or phenomena that are outside the project’s scope.

• Consider the purpose of the model in order to select the modeling approach. If the client faces an investment decision abroad and wishes to know the probability of an armed conflict in the region, an interstate or structural approach is applicable. If the concern is with the potential enactment of a specific policy that may affect the business, then an agent-based, micro-level approach may be more likely to succeed.

• Determine the desired tradeoff between the level of predictive accuracy and explanatory power. A simple, parsimonious model with a few theoretical building blocks may not be best for predictive accuracy, while an elaborate, complex model that combines multiple theoretical building blocks may increase predictive accuracy at the expense of explanatory power.

• Use Tables 1 and 2 in this chapter to select competing and complementary building blocks for the model. Competing blocks explain similar phenomena with different assumptions and are difficult to combine in the same model. Complementary blocks can be combined in parallel or in series.

• When building a model, construct a flowchart of the political or decision-making process of interest to help visualize and sequence the theoretical building blocks.

• When integrating and testing a model, beware of nonlinear subprocesses whose feedback can drive and overcome the output not only of the next building block but also of the entire system. Here, scaling of variables can be useful to dampen the potential impact across building blocks.

5 Summary

Political modeling generally falls into three categories: micro-level dynamics (the expected actions and interactions of individuals, groups, or governments), intranational structural dynamics (the subnational, structural factors that politically propel a nation, such as economic prosperity, democratization, or other national indicators), and international structural dynamics (the cross-national comparisons of national factors). Each of these categories is best assessed using a particular methodology, e.g., agent-based modeling, dynamic modeling, or game modeling. Our classification system helps assess the applicability of various political theories. It includes unit of analysis, model type, assumptions, key variables, structure of the environment, and the core logic of how the variables are related, in addition to the main implications, empirical support, and shortcomings of these theories. Macro interstate approaches include models that investigate the dynamics of political competition by using coupled differential equations, e.g., deterministic relationship between two states based on action and reaction, in which a small buildup by one side would lead to a larger counter by the opponent. An example of a macro intrastate approach is the POFED model, a dynamic general-equilibrium model based on the intersection of political and economic maximization function: individuals seek to maximize their lifetime utility by choosing how much to consume, while policymakers choose the tax rate and the amount of public investment and military spending to maximize their chances of remaining in power. Micro individual approaches focus on individual decisions of key political actors, leaders, and organizations, where models often use expected utility and must address the issue of risk. Senturion is an example that combines several micro-level theoretical blocks drawn from game theory, decision theory, spatial bargaining, and microeconomics. This simulation-based tool is used to predict political events, anticipate domestic or international political stability levels, and analyze specific investment decisions where political matters affect outcomes. In one case study, Senturion analyzed the 2005 Iraqi elections, and correctly predicted that an increase in Coalition military forces would improve Iraqi attitudes toward the election.

6 Resources

• 1. Pointers to collections of data that can be used to initialize and validate political modeling of conflict and cooperation

  The Correlates of War Project (COW) Cross-National and Cross-Time Conflict Dataset Hosting Program http://www.correlatesofwar.org/

  The Groningen Growth and Development Center (GGDC) Economic Historical Statistics (Angus Maddison) http://www.ggdc.net/

  World Development Indicators (WDI) Economic Statistics including more than 800 indicators. (The World Bank) www.worldbank.org/data

  International Financial Statistics (IFS) International Statistics on All Aspects of International and Domestic Finance. (International Monetary Fund) http://www.imfstatistics.org/imf/

  Government Finance Statistics (GFS) Annual Finance Statistical Data on General Government and Its Subsectors. (International Monetary Fund) http://www.imfstatistics.org/imf/

  UN data Gateway to Statistical Information from Databases of the UN and Member States. (United Nations Statistics Division) http://data.un.org/

  EUGene Expected Utility Generation and Data Management Program (D. Scott Bennett and Allan C. Stam, III) http://eugenesoftware.org/

  Global Terrorism Database (GTD), University of Maryland Information on Terrorist Events around the World since 1970 (National Consortium for the Study of Terrorism and Responses to Terrorism, START) http://www.start.umd.edu/start/data/

  Polity IV Project Dataset on Political Regime Characteristics and Transitions, 1800–2008 http://www.systemicpeace.org/polity/polity4.htm

  Center for the Study of Civil Wars, PRIO Data on Armed Conflict: http://www.prio.no/CSCW/Datasets/Armed-Conflict/

  Geographical and Resource Datasets http://www.prio.no/CSCW/Datasets/Geographical-and-Resource/

  Economic and Socio-Demographic Data http://www.prio.no/CSCW/Datasets/Economic-and-Socio-Demographic/

  Data on Governance http://www.prio.no/CSCW/Datasets/Governance/

  Uppsala University Uppsala Conflict Database Project (UCDP) http://www.pcr.uu.se/gpdatabase/search.php

  James Fearon and David Laitin, Stanford University Ethnicity, Insurgency, and Civil War (replication data) http://www.stanford.edu/~jfearon/data/apsr03repdata.zip

  James Fearon, Stanford University Ethnic and Cultural Diversity by Country http://www.stanford.edu/~jfearon/data/egroupsrepdata.zip

  The Minorities at Risk (MAR) Project, University of Maryland Dataset on Conflicts of Politically-Active Communal Groups http://www.cidcm.umd.edu/mar/data.asp

  The Kansas Event Data System (KEDS) Project, University of Kansas Political Event Data focusing on the Middle East, Balkans, and West Africa http://web.ku.edu/~keds/index.html

• 2. Useful books, guides, handbooks, collections of instructional materials relevant to political modeling

  Ronald Tammen et al., Power Transitions: Strategies for the twenty-first century http://www.cqpress.com/product/Power-Transitions-Strategies.html

  Manus Midlarsky, Handbook of War Studies I, II. The Interstate Dimension III: The Intrastate Dimension http://www.press.umich.edu/titleDetailDesc.do?id=348477

  Yi Feng, Democracy, Governance, and Economic Performance: Theory and Evidence http://mitpress.mit.edu/catalog/item/default.asp?tid=9932&ttype=2

  Marina Arbetman, J. Kugler, Political Capacity and Economic Behavior http://www.amazon.com/Political-Capacity-Economic-Behavior-Interdependence/dp/0813333644

  Stathis Kalyvas, Logic of Violence in Civil War http://www.cambridge.org/us/catalogue/catalogue.asp?isbn=0521854091

  P. Collier, Nicholas Sambanis, Understanding Civil Wars (v 1 + 2) http://extop-workflow.worldbank.org/extop/ecommerce/catalog/product-detail?product_id=3995594&

  Barbara Walter, Reputation and Civil War http://www.cambridge.org/us/catalogue/catalogue.asp?isbn=9780521763523

  Douglas Lemke, Regions of War and Peace http://www.cambridge.org/us/catalogue/catalogue.asp?isbn=0521809851

  Andrew Kydd, Trust and Mistrust in International Relations http://press.princeton.edu/titles/8091.html

• 3. Professional or academic organizations, NGOs, and foundations that are relevant to the political modeling

  American Political Science Association http://www.apsanet.org

  International Studies Association http://isanet.ccit.arizona.edu/

  Peace Science Society (International) http://pss.la.psu.edu/

  Inter-University Consortium for Political and Social Research (ICPSR), University of Michigan http://www.icpsr.umich.edu/icpsrweb/ICPSR/

  Military Operations Research Society (MORS) http://www.mors.org/

  Center for the Study of Civil Wars (CSCW), PRIO http://www.prio.no/CSCW

  The MacMilan Center, Program on Order, Conflict Violence (Yale University) http://www.yale.edu/macmillan/ocvprogram/

  Uppsala Conflict Data Program http://www.pcr.uu.se/research/UCDP/

• 4. Pointers to journals, newsletters, and other periodic publications particularly relevant to political modeling

  American Political Science Review http://journals.cambridge.org/action/displayJournal?jid=PSR

  American Journal of Political Science http://www.ajps.org/

  International Interactions http://www.tandf.co.uk/journals/titles/03050629.asp

  International Studies Quarterly http://www.wiley.com/bw/journal.asp?ref=0020-8833

  Journal of Conflict Resolution http://jcr.sagepub.com/

  Journal of Peace Research http://jpr.sagepub.com/

  Conflict Management and Peace Science http://cmp.sagepub.com/

• 5. Pointers to conferences and workshops that study political modeling

  International Studies Association Annual Meeting http://www.isanet.org/

  Peace Science Society (International) Annual Meeting http://pss.la.psu.edu/

  American Political Science Association Annual Meeting http://www.apsanet.org/

  Midwest Political Science Association Annual Meeting http://www.mpsanet.org/