百媚导航

SOIC Study Methodology

Definitions and Methodology
This page captures research and analysis by SSI professors, USAWC faculty and students, and research assistants into the causes and outcomes of internal conflicts since 1945 as part of the ongoing Study of Internal Conflict (SOIC). Research and analysis will also include topics in Unconventional and Irregular Warfare.

Outline (quick links)

• Study Overview
• Conflict Inclusion Criteria
• Methodology
• Research Utilizing SOIC
• Defining the Dependent Variable
• Defining the Independent Variables
• Interrelationships of Independent Variables
• Key Findings to Date
• Additional Study Findings
• Glossary

 

The Study of Internal Conflict (SOIC) research project at the Strategic Studies Institute began in 2014 to implement a data-driven approach in determining the actual causes of government failure in internal conflict. Since 2014, it has formed the core of the U.S. Army War College elective course S12232, 鈥淎 Problem from Hell: Civil War and Insurgency鈥� taught by Dr. Chris Mason. Each student in the course each year has conducted at least one case study from the list of SOIC conflicts (see Conflict Inclusion Criteria, below) as a part of their course written requirements. The SOIC has also been an internship option at the Army War College for several years. More than a dozen undergraduate and graduate interns have worked as research assistants conducting conflict case studies under the direction of program director Dr. Chris Mason. To date, the Study has completed more than sixty individual conflict case studies, each involving three to four weeks of guided research on a particular internal conflict. The research study and the theoretical underpinnings of the study were recently described in some detail in an article in the Summer 2021 issue of Parameters, the U.S. Army War College Quarterly.

For consideration of a conflict in the research, the Study of Internal Conflict uses the Correlates of War Project database as the starting point. Since 2014, the SOIC has systematically analyzed all the conflicts in the Correlates of War Project database and identified all human conflicts in the modern era that meet the following six criteria:

1. The conflict was an intrastate conflict, not an interstate conflict. In other words, the conflict was primarily a war inside one country, not a war between two or more countries. Conflicts frequently have external actors who support one side or another and essentially fight by proxy. External intervention does not preclude a conflict from being included, so long as the conflict was not in essence a war fought between two separate countries. Colonial wars of independence were excluded from the SOIC research project, as these were deemed essentially wars between a colonial power and one or more groups of colonized subjects, i.e., between two different countries.
2. Since the purpose of the study is to determine why governments fail, the second criterion for inclusion in the study is that the government of the state was itself a party to the conflict. Because there are many kinds of civil conflict (for example, feuds between groups in remote areas that are not fought for control of the government), those internal conflicts listed in the Correlates of War database to which the government was not a party are excluded from the study.
3. Because of the definition of 鈥済overnment victory,鈥� discussed in detail later, the conflict must have ended at least two years prior to the date of determination.
4. The internal rebellion sought either full control of the government or the creation of an independent breakaway country. Internal conflicts in which the purpose of the rebel group was not political power through self-determination were excluded. This excludes a relatively small number of conflicts, but some internal conflicts have been fought over control of specific resources or other issues.
5. The fifth criterion is that at least one thousand persons died (combatants and civilians) during a continuous twelve-month period at some point during the conflict, and as a direct result of the conflict. To keep the study size manageable and relevant to U.S. government interests, the multitudinous smaller conflicts in the Correlates of War databases have not been included up to this point in the study.
6. The conflict began after the end of World War II, defined as August 1945. This date was selected as the political and military watershed in global affairs that shaped the modern global order, established current international borders, and created the tactics and types of weapons used in modern insurgencies.

The underlying methodology for the study is a straightforward regression analysis in which researchers sought to find correlative relationships between selected independent variables and a single dependent variable. In this sense, the purpose of the study is to forecast outcomes upon the dependent variable based on a consistent set of one or more independent variables, each of which correlate to the failure of the dependent variable in at least 90 percent of all fifty-three currently active case studies (conflicts).¹ Researchers initially evaluated a large number of independent variables (see Defining the Independent Variables, below) to isolate those which correlated with the failure of the dependent variable in at least 90 percent of all cases analyzed. The five discrete independent variables now in use in the study are all of those political-military factors which were found to be present in at least 90 percent of all cases in which state governments failed to attain the stated conditions of victory (see Defining the Dependent Variable, below). Of all the independent variables considered, only these five were found to reach the 90 percent threshold (see Defining the Independent Variables, below). No other independent variables actually showed a statistically strong (consistent) correlation with outcomes. 

¹ A number of early case studies were conducted on conflicts which were subsequently determined not to meet the six criteria for inclusion in the Study.

1. Mason, Chris. 鈥淐OIN Doctrine is Wrong.鈥� U.S. Army Parameters, 2021.
2. Miller, John. 鈥淭he Katanga Secession, the Five Factors Model, and Counterinsurgency (COIN) Theory.鈥� Small Wars Journal, 21 November 2024.
3. Mason, Chris. 鈥淢easuring and Quantifying State Fragility.鈥� Found in Resilience and Resistance: Interdisciplinary Lessons in Competition, Deterrence, and Irregular Warfare, edited by Robert Burrell. Joint Special Operations University Press, 2025.
4. Burrell, Robert and John Collison. 鈥淎 Guide for Measuring Resiliency and Resistance.鈥�  Small Wars & Insurgencies, 14 December 2023.
   
   

For the parameters of the Study of Internal Conflict, the established dependent variable is termed 鈥済overnment victory.鈥� The definition of government victory has two components which must both be true. Government victory parameters are assessed as having been met if the end state of the conflict is a political condition in which: (1) the same government which was in control of the apparatus of state power at the time of the start of the internal conflict or its natural successor remained in power eighteen months after the end of the conflict; and, (2) the integrity of the state boundaries at the start of the conflict remained substantially intact. These two victory conditions correlate to the two rebel intentions as established by Inclusion Criterion 4 above (i.e., the intent either to take over all state power or to establish an independent country; such as East Timor, for example). In other words, the definition of 鈥済overnment victory鈥� infers the rebels did not succeed in their objective and the government remained in power.

The study began with a list of more than forty social, economic, political and geographical parameters (such as, for example, whether the population was primarily rural or urban, the predominant religion, overall poverty levels) to identify all political, military and economic factors which correlate with government failure in at least 90 percent of all the internal conflicts studied. A 90 percent correlation rate was established as the threshold for determining sufficiently predictive relationships using observational data. Of all the potential factors studied which might plausibly have correlated to conflict outcomes, only five have been identified which meet or exceed the 90 percent correlative threshold:

1.  Less than 85 percent of the total population located their personal identities at the level of the nation, i.e., expressed a national identity as defined by political science. (See the Glossary below for a full set of research definitions).
2. Less than 85 percent of the total population of the country believed the government in power to be a legitimate holder of state power again as defined by political science. (See the Glossary below for a full set of research definitions).
3. Less than 85 percent of population was fully isolated from meaningful contact with the rebel group. (See Glossary, below, for the definition used for 鈥渕eaningful contact鈥�).
4. The existence of persistent external sanctuary for militarily significant numbers of the rebel group. (See Glossary, below, for the definition used for 鈥榩ersistent sanctuary鈥� and 鈥渕ilitary significant numbers鈥�).
5. The lack of sustainable, pre-existing security forces under the control of the government at the outset of the conflict (See Glossary, below, for the definition used).

Although the threshold for statistical correlation was set at 90 percent, in actuality, each the five individual factors described above has been found in at least 94 percent of all internal conflicts. No additional factors crossed the 70 percent correlation threshold and were not considered sufficiently correlative for predictive purposes. The two binary Boolean independent variables (the existence of external sanctuary and the lack of a pre-existing security force under government control) were actually each independently fatal in 100 percent of all internal conflicts that met the six criteria for inclusion in the study (see Conflict Inclusion Criteria, above). In the cases of protracted conflicts spanning several years, researchers sought multiple benchmarks across time throughout the course of the conflict (from which the percentages of the population holding a national identity and believing the government to be the legitimate holder of state power could be deduced). In doing so, the study discovered that in no case since 1945 has either of these two benchmarks ever increased during the course of the conflict.

Each independent variable is considered to be self-standing, in that its discrete impact on the dependent variable does not depend on any other independent variable. However, some factors were clearly interrelated, in the sense that they frequently appeared together. In other words, in most cases, more than one of the independent variables (defeat factors) was present when a government was unable to survive the internal conflict with it pre-conflict borders intact or largely intact. In particular, a lack of national identity and a lack of perceived government legitimacy were often nested together, as might be expected. As in calculating probable medical outcomes across multiple morbidity factors, co-existing multiple independent variables existing within a conflict are cross-multiplied to determine the mathematical probability of regime survival. Therefore, the presence of multiple negative independent variables in one conflict dramatically decreases the likelihood of government survival. In fact, most conflicts were found to have more than one of the five independent variables in play at the same time. In some cases studied, all five factors were present. Naturally, the more factors which are present, the higher the statistical probability of government defeat, but the presence of a single factor is fatal at least 94 percent of the time.

The key findings of the study are centered on two distinct clusters of empirical data, both of which have significant implications for U.S. foreign policy. The first finding is that study results show conclusively that governments fail against internal rebellions for one or more of five fundamental structural reasons, and the outcomes of internal conflicts are heavily dependent on these five preexisting political-military conditions. Each of the five factors was found in government failure in at least 94 percent of all fifty-three conflicts, and only two of the five are susceptible to military action. Further, two of the five factors are simple binary variables, while the remaining three factors are mathematically quantifiable to a useful degree of accuracy, creating thresholds that correlate to government defeat with a remarkable degree of consistency and accuracy.

The empirical data prove that only two of the five factors can be altered in any meaningful way after the onset of hostilities. In essence, whether a government will or will not be successful in suppressing an internal rebellion depends predominantly on whether these five factors are present at the start of the conflict. Thus, collectively, the five independent variables (or factors) constitute a predictive model of probable internal conflict outcomes with a reliability that startled researchers. Cases of 鈥渟uccessful counterinsurgency鈥� often cited by proponents of COIN doctrine were found to be simply cases where all five political-military factors were already in favor of the existing government at the outset of the conflict. The second key finding, which was not anticipated at the start of the project, is that in seven years of research, no evidence has been found to support the basic assumption which underpins such elements of current counterinsurgency doctrine as 鈥渃lear, hold, and build,鈥� 鈥減acification鈥� and 鈥渘ation building.鈥�

The assumption upon which these doctrinal elements are based is that measures can be taken by the counterinsurgents which will increase popular support for the state government, or decrease support for the insurgent movement, and/or increase the legitimacy of the state government. The measures usually include spending on reconstruction and civic infrastructure projects such as schools, hospitals, health clinics and roads. On the contrary, SOIC has compiled a body of studies and surveys from around the world that show either no increase in government support (or legitimacy), or show a reduction in support and/or an increase in conflict and instability in the vicinity of the civil projects. This finding should spur a reconsideration of U.S. military counterinsurgency doctrine that places such transactional civil affairs efforts at the heart of such operations.

In looking for this data, the researchers also searched for evidence that popular support (one potential indicator of government legitimacy) had ever increased at any level (state, provincial, district or local) associated with the types of government action usually undertaken to reduce support for the rebels and increase government legitimacy, actions typically referred to collectively as 鈥渃ounterinsurgency.鈥� Such evidence could take the form, for example, of surveys taken before and after development projects, or a reduction in the level of enemy-initiated violence over a period of time, or multiple studies done in a locality over a period of several years. No such evidence was found. However, a large number of reports, surveys and other statistical evidence show the absence of such an increase, and several cases of a decrease in support for the government and an increase in insurgent violence or local community conflict. In other words, there is a great deal of statistical proof that typical counterinsurgency methods have no effect or even a negative effect, but zero evidence across the fifty-three conflicts studies that they have had a positive effect (i.e., increased popular support or government legitimacy) in any quantifiable degree anywhere in the world. This finding has profound implications for the methods used by governments and external actors to defeat internal rebellions.

External Sanctuary
鈥淓xternal sanctuary鈥� is defined as a land border with a neighboring country that the rebel combatants could and did cross in militarily significant numbers without excessive difficulty or danger. Militarily significant numbers are considered to be platoon-size elements crossing with their weapons and equipment, rather than individuals exfiltrating and infiltrating across a guarded border. Once across the border, for the destination to be considered 鈥渟anctuary,鈥� the rebels must be reasonably certain of safety from detention, repatriation, or pursuit. Certainty of external support in the form of funding, new equipment, training, or new recruits is not required for the cross-border area to be considered 鈥渟anctuary,鈥� but it must be across an international border. Determining whether external sanctuary existed in a given case was relatively easy in actual practice, as during the course of the study, there have been no cases in which adjudication was required whether the criteria were met. In other words, all cases have been clear-cut. If sanctuary was available at the start of the conflict, but the government succeeded in sealing off its borders to eliminate the sanctuary during the conflict, the study defines this as a 鈥渘egative鈥� finding for external sanctuary despite its earlier existence, and it appears as 鈥淵es 鈫� No鈥� in the data matrix.

Government Victory
 鈥淕overnment victory鈥� is defined as the government which was in power at the start of the conflict or its natural successors (i.e., not a change of government) remained in power eighteen months after the end of hostilities, and the territory of the country remained unchanged from the start of the conflict. If the rebels succeeded in bringing down the government or in achieving their goal of either establishing an independent breakaway country, even if the government remained in power, the government cannot be said to have won the conflict.

Legitimacy of Government
The study uses established political science principles in defining legitimacy of government. Princeton University Encyclopedia for example defines it as: 鈥淭he belief that a rule, institution, or leader has the right to govern. It is a judgment by an individual about the rightfulness of a hierarchy between rule or ruler and its subject and about the subordinate鈥檚 obligations toward the rule or ruler.鈥� The level established by the study for analysis is 85 percent of the population holding such a belief.

National Identity
The study uses established political science principles in defining 鈥渘ational identity鈥� as citizens locating their personal identities at the level of the state and not at a sub-state stratum such as their tribe, clan, ethnicity, linguistic or religious group. The level established by the study for analysis is 85 percent of the population holding such a state or nation-level identity.

Population Protection
Although population protection is a bedrock principle of counterinsurgency doctrine, it was necessary to define what the SOIC means by the term in order for researchers to assess whether or not a population was in fact being protected. The definition the study uses states that a population is protected if 鈥�85 percent or more of the civilian population is effectively sealed off from meaningful contact with the rebel group.鈥� Examples of meaningful contact would be the printing and distribution of leaflets, pamphlets or other propaganda material in any significant degree, the persistent ability of a rebel group to make radio broadcasts which reached any significant number of listeners, the ability of rebels to give speeches in public or address groups of more than a few individuals in private, the ability of rebels to credibly threaten, intimidate or bribe government officials, or to assassinate government workers or leaders. This definition acknowledges that it is not possible to create a completely airtight firewall between civilians and rebels that will prevent 100 percent of rebel propagandists or terrorists from slipping past government security cordons 100 percent of the time. Thus 鈥淧opulation Protection鈥� is defined as the reliable, continuous maintenance a high level of public security which prevents all but rare interaction between the civilian population and rebel elements.

Sustainable, pre-existing security forces
This term is defined for the study as an existing security force at the outset of the conflict which was funded by the government, wore some sort of uniform, observed some sort of hierarchical rank structure in which the soldiers largely obeyed the orders of their superiors, was capable of transporting itself and sustaining itself in the field, and which generally carried out the orders of the Head of State or the appointed government official responsible for security. In the case of military rule, the standing security force had to be subordinate to the military element holding state power.

 

---