The Causes of State Collapse: Results from a QCA Analysis * Daniel Lambach, Markus Bayer, Eva Johais

Transcription

1 The Causes of State Collapse: Results from a QCA Analysis * Daniel Lambach, Markus Bayer, Eva Johais Abstract: This paper presents preliminary results of a comprehensive investigation into the causes of state collapse. We test major causal hypotheses derived from the literature with a medium-n design involving 15 cases of post-colonial state collapse. We conduct a synchronic and a diachronic comparison with two different control groups of cases that also experienced political upheaval without collapsing. Our results cast doubt on prominent theories of state collapse and suggest that alternative factors are more important. We develop a preliminary theoretical framework to explain incidences of state collapse. Contact details: Universität Duisburg-Essen Institut für Politikwissenschaft Lotharstr Duisburg Phone: for correspondence: daniel.lambach@uni-due.de * This research was made possible by the generous funding of the German Research Foundation (Deutsche Forschungsgemeinschaft). Christian Tischmeyer, Laura Blomenkemper and Simon von Dahlen have provided helpful research assistance. We thank Benoît Rihoux, Claude Rubinson and Achim Görres for their helpful comments. We are also indebted to Thomas Richter for providing access to data on state revenus, and Lasse Cronqvist and Sascha Werthes for technical and logistical support. 1

2 1. Introduction What are the causes of state collapse? Despite two decades of research, we still know surprisingly little about the roots of this phenomenon. In this paper, we present the results of a QCA analysis of 15 cases of state collapse and two different sets of control cases (synchronic and diachronic). This project began as a brush-clearing exercise to test the multitude of theories and hypotheses about the causes of state collapse that were being raised in the literature. Hence, the first step was to test those deductively derived theories of state collapse. However, our initial results cast considerable doubt on the plausibility of these hypotheses. Therefore, we moved to a more inductive logic and sought to identify the causes of collapse for the cases in our dataset. Our findings highlight the impact of several risk factors, particulary the militarization of political actors, an extremely low per-capita income, a lack of societal cohesion based on precolonial identities and factionalist politics as well as a decline in state resources and recent political transition. In contrast, other conditions which prior theories considered to be risky e.g. informal modes of governance, personalist regimes, government obstruction of the bureaucracy and the military, or regime type offer less explanatory value. Based on these results, we propose a prelimary typology of the causes of state collapse that combines structural contexts and dynamic processes. In the next section, we review the literature about the causes of state collapse and related concepts like failed and fragile states. After that, we develop causal hypotheses that we deduced from the theoretical and conceptual literature. In the fourth section, we present our research design, focusing on case selection and our reasons for opting for QCA. The fifth section presents the data. In the sixth section, we first present the process of eliminating conditions and then show results from our comparisons. We discuss these results in the seventh section. The concluding section weighs the implications of our results for further research. 2. Prior Causal Research Even though fragile, failed and collapsed states have received a lot of attention in recent years (Bueger and Bethke 2014), there is very little systematic research on the causes of these phenomena (exceptions are Bates 2008, Carment et al. 2008, Englehart 2007, Iqbal/Starr 2007). 1 In contrast, there is a vast theoretical and conceptual literature that has put forward a broad range of causal hypotheses. In other words, there are lots of theories with little systematic empirical grounding. Our paper represents an attempt to clear some of this brush through a systematic empirical evaluation. To save space, we will restrict this review of the literature to those contributions that have at least some empirical basis. There are many single-case studies which discuss whether a particular state can be classified as fragile, failed or collapsed. 2 The heterogeneity of these cases suggests that state collapse is a complex and multicausal process. Several studies also suggest that state collapse is an equifinal process (Bennett 2004: 38-39), i.e. that states can collapse in several different ways. For instance, Englehart (2007) proposes a model of self-destructive despotism that shows how governments in 1 The focus of the literature lies on, e.g., conceptual issues (Call 2008, Grävingholt/Ziaja/Kreibaum 2012), the provision of governance in fragile states (Risse 2012, Clements et al. 2007, Meagher 2012), normative critiques of the entire discourse (Bilgin/Morton 2002, Hameiri 2007), the links between fragile states and terrorism (Menkhaus/Shapiro 2010, Simons/Tucker 2007), the implications for development policy (Hout 2010, World Bank 2011) or the efficacy of military interventions and/or statebuilding (Crocker 2003, Fukuyama 2004, Krasner/Pascual 2005). 2 These studies are too numerous to mention. Prominent edited collections of such case studies are Rotberg (2003), Schneckener (2006a) and Zartman (1995). 2

3 Somalia and Afghanistan deliberately destroyed state institutions to remove checks and balances on their exercise of power and prevent rivals from harnessing alternative sources of power. Building upon these results, Lambach (2009) has shown that this model cannot explain the collapse of the Lebanese and the Tajik state. Instead, he argues that there is a second causal model of collapse that is characterized by a polarization of the political spectrum and a militarization of political actors. This is further supported by Grävingholt, Ziaja and Kreibaum (2012, see also Call 2011) who identify separate clusters of fragile states. There have been a few attempts to analyze the causes of state collapse with large-n approaches. 3 Earlier studies were mainly concerned with developing early warning systems (e.g. Baker/Ausink 1996, Baker/Weller 1998, Norton/Miskel 1997). State failure was frequently conflated with other phenomena like civil wars or ethnic conflict so that the indicators used were relatively general signs of political instability. The largest project of this kind was the State Failure Task Force (SFTF, since 2003 named the Political Instability Task Force, PITF). This group, established in 1994, tried to identify correlates of state failure in order to develop a model that could predict state failure two years in advance with sufficient precision (Esty et al. 1995: iii). However, the SFTF was only able to identify 20 cases of state failure between 1954 and 1994 which were insufficient for the methods that it wanted to use. Therefore, the SFTF broadened its concept of state failure by including four different kinds of crisis: revolutionary wars, ethnic wars, adverse regime change as well genocides/politicides (Esty et al. 1995: 1). From phase I to phase V, the Task Force has managed to improve the accuracy of their global model from 60% to 80% (Esty et al. 1995, Goldstone et al. 2005). By now, this model has become quite useful for early warning. It is, however, less amenable for causal research due to its overly broad conceptualization of failure. 4 The lack of suitable datasets that measure the degree of stateness has long been a crucial impediment to quantitative causal research. Various projects have tried to address this lacuna, like the Failed States Index (FSI) (Baker 2006, Baker/Weller 1998, Fund for Peace 2005), the Index of State Fragility (ISF) (Carment et al. 2006, Carment et al. 2008), the State Fragility Index (SFI) (Marshall/Cole 2008, Marshall/Goldstone 2007) and the Index of State Weakness (ISW (Rice/Patrick 2008). The FSI and the ISF are explicitly designed as early warning mechanisms and also employ a very broad conceptualization of fragile statehood (for a critical appraisal see Margolis 2012). Only the ISF has been used for causal research (Carment et al. 2008). These indices are usually composed of aggregate data and other indices that cover a wide thematic spectrum from infant mortality and the deforestation rate to GDP/capita. The enormous breadth of component indicators severely limits the content validity of the overall construct (Bethke 2012). It is not clear what these indices measure it is not state fragility, but rather a general, but unspecific mélange of social, political and/or economic crisis. Causal research becomes de facto impossible since most potential causal conditions are already part of the definition of the concept that is to be explained. Furthermore, temporal coverage of these indices is limited, with only the SFI offering data for years prior to Due to these weaknesses of the major indices, quantitative research designs tend to use other variables to operationalize state fragility and state collapse. For instance, Iqbal and Starr (2007) use the standardized authority code -77 from the Polity IV dataset which denotes a period of Interregnum, a complete breakdown of political authority (Marshall/Gurr/Jaggers 2010). This code identifies 25 cases with a total of 93 country-years as failed states. Iqbal and Starr employ a 3 There has also been one attempt to use QCA for causal research. Clément (2004) conducted a study of three cases of collapse (Yugoslavia, Lebanon and Somalia). However, her study is of limited utility because the small N forced her to use very few and very general conditions. In addition, she has a functionalist understanding of statehood that is at odds with our institutionalist approach. 4 Howard (2008) uses SFTF data to develop a causal model of state failure. However, due to the overly broad definition of state failure, this is of limited utility. For critiques of the SFTF's work see King/Zeng (2001), Lambach/Gamberger (2008) and Milliken/Krause (2002). 3

4 Cox regression on a sample of 162 states from 1946 to Their model identifies a hybrid regime, an ongoing civil war, international conflict, political unrest, a low GDP and the post-cold War eta as significant risk factors associated with state failure. But proxies suffer from the same validity problems as indices. This highlights a more general weakness of quantitative approaches, namely that of a lack of sound data. Many researchers call for developing new measures of statehood (Bates 2008b: 10, Englehart/Simon 2009: 110) but there has been no convincing approach that addresses this lacuna. The problem of data quality also extends to the measurement of possible causes. Data for these measures is not missing at random instead, its availability depends on the statistical capacity of a state. Hence, we can expect a higher rate of missing data in fragile states (Bates 2008a). While these gaps can be bridged through various methods of imputation, this is only a second-best solution. Beyond the methodological issues, quantitative studies display several other weaknesses. First, quantitative approaches assume a) a unifinal causality, i.e. that variables have the same effect for all cases, b) additivity, i.e. that the causal effect of independent variables can be isolated from each other (which can be addressed via interaction terms; however, only a small number of interaction terms can realistically be included in a model), and c) a symmetric causal relationship, i.e. when A leads to B, not-a will lead to not-b (Wagemann/Schneider 2010: 378). However, our understanding of state collapse leads us to believe that these assumptions, especially the first one, are inappropriate for our research question. Second, the studies surveyed above did not test the range of hypotheses available from the theoretical literature. Out of the risk factors we extracted from the literature (see 3.), Carment, Prest and Samy (2008) and Iqbal and Starr (2007) only tested those on the impact of regime type, ethnoreligious diversity, economic openness and change in the international system. Other hypotheses, e.g. on the influence of bad governance, declining state revenues, or regional spillover effects, were not included. Instead of generating their own data, the authors relied on data that had been previously collected for other purposes. As Englehart and Simon put it: There are few variables expressly constructed for the study of state failure, forcing researchers to either engage in costly data collection or rely on pre-existing off-the-shelf-data (Englehart/Simon 2009: 110). This means that many potentially interesting explanations are not being investigated for want of suitable data. Third, this reliance on off-the-shelf data leads to an emphasis on structural variables for which data is much more readily available than for more dynamic variables. At the same time, most authors agree that the political dynamics are especially important in influencing the outome during periods of political crisis (e.g. Bates 2008b, Goldstone et al. 2005: 15-20, Rotberg 2004: 25-26). Due to a lack of data, this proposition cannot be tested quantitatively, suggesting that such studies suffer from omitted variable bias (King et al. 1994: ). 3. Risk Factors of State Collapse A review of the literature turns up a large number of claims, hypotheses and preliminary results about the causes of state collapse. Some conditions were mentioned by several authors and thus seemed more likely candidates, but no particular theory had emerged as a consensus. A few contributions (e.g. Englehart 2007, Lambach 2009) formulate more elaborate, conjunctural hypotheses. Other authors add a number of other factors, in some cases making competing causal claims, e.g. whether low or high openness to international trade made collapse more likely. 5 However, the results of our initial QCA analysis did not support these deductively derived hypotheses. Truth tables displayed a large proportion of contradictory results, calling the discriminatory power of the conditions into question and forcing us to re-evaluate our approach 5 We provide a more detailed review of the literature in Bethke and Lambach (2012). 4

5 (Ragin 1987: , Rubinson 2013). Therefore, we switched to an inductive strategy and sought to explain state collapse for the cases in our sample which consists of the majority of cases of state collapse in the post-1960 era and is therefore highly representative. In other words, we moved from an interest in effects of causes to one in causes of effects (Mahoney/Goertz 2006) or from a theory-testing to a theory-building approach. Retaining the ontological assumption that the causal structure of state collapse is complex and equifinal, we formulated two basic configurational hypotheses about the causal structure of state collapse: H 1 : No risk factor, or combination of risk factors, is a necessary condition of state collapse. H 2 : Certain combinations of risk factors are sufficient conditions of state collapse. H 2 represents a combination of INUS conditions. INUS stands for Insufficient but Nonredundant part of an Unnecessary but Sufficient condition (Mackie 1965). Mahoney describes INUS conditions as parts of larger combinations of factors that are jointly sufficient for outcomes. Thus, while an INUS condition is itself neither necessary nor sufficient for an outcome, it is part of a larger combination of factors that is sufficient for an outcome (Mahoney 2010: 131, Fn. 22) We treated potential explanatory conditions as (combinations of) risk factors that could be plugged into these configurational hypotheses. As a simple heuristic, these risk factors are grouped into political, economic, socio-cultural and international/regional ones. How we translate these risk factors into conditions is given in parentheses and the conditions are described in section 5. Political factors The first risk factors are derived from Englehart (2007) who argues that neopatrimonial, clientelistic and corrupt practices undermine the state institutions from within. A weak regime might also try to weaken alternative centers of power like security forces and the bureaucracy if it cannot control them obstruction of state institution is a survival strategy for ruling elites in this sense. P1: Systematic obstruction of the bureaucracy by the ruling elite (BUR_OBSTR) P2: Systematic obstruction of the monopoly of violence by the ruling elite (SEC_OBSTR * UNOFF_MILITIA) In his concept of the shadow state, William Reno (2000) argues that parallel and informal shadow institutions which appropriate competences and resources from the formal institutions are responsible for state fragility. P3: Strong informal political structures (INFORMAL) P4: Personalist regime (PERSONAL_RULE) There are many discussions about the effect of regime type on state collapse. Whereas Allen (1995) and Gros (1996) see autocratic regimes as prone to state collapse, Carment et al. (2008) consider hybrid regimes as more at risk. Goldstone et al. (2005) claim that hybrid regimes with a high degree of factionalism are most likely to fail. P5a: Autocratic regime (REGIME) P5b: Hybrid regime (HYBRID_REGIME) P5c: Hybrid regime with factionalism (HYBRID_REGIME * FACTIONAL) While most theories focus on strategies of regime elites or institutional configurations of the regimes themselves Lambach (2009) identifies special configurations of escalating power struggles as a cause of state collapse. These struggles are characterized by a polarization of the political system, a militarization of political actors, and a fairly equal distribution of power resources between major sides. P6: Fragmentation of the political system (POLAR * MILIT * POWER_PROP) 5

6 Economic factors Theories of neopatrimonialism (van de Walle 2004) and other rent-seeking approaches (Weede 2003) point to self-serving elite behavior as a central problem of politics in developing countries. While patronage can stabilize a political system, a centrifugal dynamic sets in once the resources needed to maintain the loyalty of rent-seeking clients are no longer available. Therefore, a decline in state revenues should make state collapse more likely in personalist regimes (Bates 2008a). W1: Decline of state revenues (GOV_REV * PERSONAL_RULE * INFORMAL) Many contributions argue that the so-called resource curse the dependency of primary commodity exports leads to a structural instability of states (Bates 2008a). Basedau and Lay (2009) cast doubt on the general validity of this hypothesis and argue that a combination of primary commodities and a high per capita income can have a stabilizing effect. An alternative risk factor would thus be the combination of a high proportion of commodity exports with a low per capita income. W2a: High proportion of primary commodity exports relative to GDP (PRIM_COMM_EXP) W2b: High proportion of primary commodity exports relative to GDP and a low per capita income (PRIM_COMM_EXP * INCOME) Socio-cultural factors Bates (2008a: 6) argues that repressive rule by an ethnic minority is a risk factor. In this scenario, the ruling minority will be in a vulnerable position due to the larger size of its opponents. Resorting to repression will stir up grievances and motivate the oppressed majority to mobilize against the state. S1: Highly repressive rule of an ethnic minority (ETH_MIN_RULE * REPRESSION) Some studies also expect that a high population density and the so called youth bulge have a negative effect on political stability (Goldstone et al. 2000, Schneckener 2006b). S2: High population density (POP_DENSITY) S3: High proportion of young people among the general population (the so-called youth bulge ) (YOUTH) Political culture is also assumed to have an effect on political stability. Buzan (1991) argues that society needs an idea or an image of the state as a legitimate political authority. Therefore, societies with a precolonial history of statehood should be more stable than those without such an historical experience (Clapham 2004, Gennaioli/Rainer 2007). S4: The lack of a native pre-colonial or pre-imperial polity (POLITY) International/regional factors The relatively high number of cases of state collapse in the early 1990s has led some (e.g. Menzel 2001) to argue that the drop of superpower patronage following the end of the Cold War had led to the collapse of states. I1: Superpower patronage drops off (AID) On a more systemic level Desch (1996) has argued that external threats can increase national cohesion through the rally 'round the flag effect and similar mechanisms. When the external threat fades, the regime loses a key justification for its calls for unity and coherence. Latent political divisions might turn into manifest political conflict. I2: Fading of an external thread (EX_THREAT) The effect of the economic integration into the world economy is a contentious issue. Some (e.g. Woodward 1999) argue that this makes countries vulnerable to external shocks and point out the 6

7 decreasing regulatory power of the state in a globalizing economy. Others expect economic integration to stabilize a state (Goldstone et al. 2005). I3a: High proportion of exports and imports relative to GDP (TRADE_OPEN) I3b: Low proportion of imports and exports relative to GDP (TRADE_OPEN) On a related note, Structural Adjustment Programs (SAPs) have received special attention. Ruf (2003) argues that SAPs have limited the agency of governments and that worsening public services and lower social spending create popular discontent which mobilizes the population against the state. I4: Structural Adjustment Programs (SAPs) In contrast to Ruf, Bussmann and Schneider (2007) argue that it is not the introduction of a Structural Adjustment Program per se which destabilizes the state. Instead they claim that liberalization programmes leave some veto players worse off who react to reform by destabilizing the state. I5: Economic liberalisation programs (LIBERAL) Finally, Lambach (2008) points to regional spillover processes which can destabilize neighbouring countries and entire regions. I6: Civil war in a neighbouring country (N_CIV_WAR) I7: State collapse in a neighbouring country (N_COLLAPSE) 4. Research Design Albeit conceiving of statehood as a continuum, we focus our research on the extreme end of dysfunctionality which we call state collapse (Lambach/Johais/Bayer 2013). We adopt Call s (2008) and Ulfelder s (2012) arguments to use this narrow term instead of the broader, less precise notions of state failure or fragility. The term state collapse characterizes a condition, not a process even though a protracted process might precede the condition of collapse. We derive the concept of state collapse as the binary opposite to a modified Weberian ideal-type of the state. This approach defines the state as an institution that is characterized by monopolies of rule-making, violence and taxations within a bounded territory and the population living therein. Employing an essentialist two-valued logic of concept formation (Goertz 2006: 35), the necessary conditions of state collapse are 1. no significant capacity to make rules; 2. no significant control over the means of violence and 3. no significant capacity to extract taxes. The three conditions are regarded as jointly sufficient to describe the phenomenon of state collapse when they occur concomitantly over a timespan of at least six months. This means that we establish a threshold between collapsed and non-collapsed states. To address the weaknesses of prior attempts at causal explanation, we employ a medium-n research design. The advantage of such an approach compared to small-n designs is the possibility of including multiple explanatory variables and lowers risks of omitted variable bias. It also allows for more representative results. Compared to large-n designs, our approach is better at dealing with the problem of equifinality (Bennett 2004: 38-40). It is also more feasible to collect original data for a medium number of cases than for a larger one. We further enhance the advantages of this medium-n approach in regard of the equifinality problem by conducting two different comparisons with different sets of control cases. In both instances, control cases were selected using a pairwise Most Similar, Different Outcome (MSDO) approach (de Meur/Berg-Schlosser 1994, Rihoux 2006). The first control group for a synchronic or horizonzal comparison consists of other countries that are structurally similar to the collapse cases and that also exhibited signs of political instability without fully collapsing. The second 7

8 control group for a diachronic, vertical comparison consists of the same countries as the collapse cases but at different points in time. While the first comparison is supposed to show the structural differences between collapsed and non-collapsed states, the second was meant to identify the more dynamic factors of state collapse by asking why collapse occurred at a particular time. 4.1 Cases In a previous paper (Lambach, Johais and Bayer 2013), we identified 17 cases of state collapse between 1960 and of these cases are included in the present analysis. 6 Cases are described in country-year format, where the year denotes the onset of state collapse (Outcome = 1). The control cases (Outcome = 0) for the synchronic comparison (see Table 1) were selected to maximize their initial similarity to the collapse cases according to the following criteria: geographic region, time period, level of political instability, size of population and territory, and level of economic development. 7 Case selection for the diachronic comparison focused on periods of significant crisis (e.g. rebellion, coup, economic crisis) that did not result in collapse (Outcome = 0). These instances of crisis took place at least 5 years prior to collapse or at least five years after the state was no longer collapsed. 8 <<Table 1>> 4.2 Method: Why (MV)QCA? We use Qualitative Comparative Analysis (QCA) as our method of analysis. QCA was developed more than 25 years ago (Ragin 1987) and has slowly but steadily gained recognition as an accepted methodology in the social sciences. Recent years have seen a rapid expansion of QCA use in research designs while the methodology has been continually expanded and refined (Marx/Rihoux/Ragin 2013: 11, Rihoux 2013: 235). 9 QCA is a tool for the systematic comparison of cases. It contains elements of qualitative and quantitative approaches, but it is grounded in the qualitative tradition of recognizing the importance and uniqueness of each individual case. It assumes that causality is complex, meaning that social phenomena have more than one cause, that these causes interact with each other in complex ways and that different instances of the same phenomenon may have different causes. QCA employs Boolean algebra to derive combinations of sufficient and necessary conditions for a particular outcome (in this case, state collapse). Causality is understood to be deterministic, i.e., a given set of conditions will always lead to the same outcome. This contrasts with the probabilistic thinking of statistical procedures who try to measure the average effect of an independent variable on a dependent variable. 6 We excluded Afghanistan 2001 and Iraq 2003 from our analysis. We consider these two cases to be outliers due to the strong impact of foreign military intervention on the stability of the state. 7 We always tried to identify the biggest period of instability in the control cases as close as possible to the beginning of the collapse in our positive case. The maximum deviation are seven years: Congo Kinshasa 1960 Nigeria 1967, Laos 1960 Cambodia 1967, Tadjikistan 1992 Uzbekistan We found no control case for Bosnia-Herzegovina because it cannot be considered a fully sovereign state since 1995, due to the ongoing de facto trusteeship by the international community. Furthermore, we only use only one diachronic control case per country, so there is only one control case (Zaire 1977) for both collapse cases of Congo- Kinshasa 1960 and Zaire The burgeoning literature on QCA is too large to be reviewed here. Articles by Berg-Schlosser et al. (2009) and Wagemann and Schneider (2010) are good introductions, while the volumes by Schneider and Wagemann (2012) and Rihoux and Ragin (2009) provide more detailed instructions. 8

9 In classic or crisp-set QCA (csqca), both conditions and outcomes are coded in binary. For instance, for a condition Population size a threshold could defined as 20 million people. Each case would then be assigned a value of 0 (state has fewer than 20 million inhabitants) or 1 (state has at least 20 million inhabitants). These values are then entered into a truth table, from which solution terms are derived using Boolean algebra. The limitations imposed by the dichotomous coding of conditions and outcomes incurs a loss of information about individual cases. Depending on the distribution of raw values among cases, it can also group together very disparate cases that happen to find themselves on the same side of the threshold. To offer more nuance, two alternative methods have been developed, fuzzy-set QCA (fsqca, Ragin 2000) and Multi-Value QCA (MVQCA, Cronqvist/Berg-Schlosser 2009). Both methods broadly use the same approach as csqca, but allow conditions with more than two values. While fsqca also allows the outcome variable to have a broader range of values, MVQCA still uses a dichotomous outcome variable. In QCA, two issues need special attention: contradictions and logical remainders. Contradictions occur when cases with different outcomes have the same set of conditions. There are various strategies for dealing with contradictions, but the general advice is that these should be prevented as much as possible (Rihoux/de Meur 2009: 48) since some or all of the contradictory cases will have to be dropped from the analysis. Logical remainders are combinations of conditions that are not associated with a particular outcome because no empirical case had this particular set of values (the problem of limited diversity, see Schneider/Wagemann 2012: ). It is permissible to make simplifying assumptions about how these combinations would have turned out if there had been an empirical case with these characteristics to arrive at a more parsimonious outcome, but care must be taken not to make unrealistic assumptions just to arrive at a better result. We chose to use QCA because there are strong indications that the causality of state collapse is complex (see 2.): Previous research had indicated that there are multiple causal paths that lead to collapse (equifinality) and that there is no single necessary or sufficient condition that can explain the occurrence of collapse, but that conditions would work in combination with one another (conjunctural causation) (Wagemann/Schneider 2010: 378). The limited number of cases would have also made statistical methods very unreliable. We initially opted for MVQCA for two reasons 10 : On the one hand, we ruled out fsqca because calibration would have been very difficult. Most of our conditions are based on a qualitative assessment of secondary sources and thresholds for two or three different values are the most we can justify more options would make coding too subjective and decrease reliability. On the other, we liked the option to use more than two values for a variable since this helps to reduce the number of contradictions in a truth table (Herrmann/Cronqvist 2009). Of course, multiple values should be used sparingly so as not to increase the number of possible configurations unnecessarily, but we had several conditions (such as REGIME, EX_THREAT or POLITY, see 5.) that were nominally scaled and had more than three expressions. Transposing a condition like regime (where we used the coding 0 = democracy, 1 = hybrid regime, 2 = autocracy) into a crisp logic would have required at least two dummy conditions (e.g., democracy yes/no and autocracy yes/no, where hybrid regimes would be indicated by a 'no' for both dummy conditions). About half of the risk factors presented in section 5 are multi-value, and most of those have just three values. Our initial theory-testing calculations were done as MVQCA analyses. However, as we moved to a theory-building, inductive approach, we found that additional values did not offer much additional explanatory power for most conditions, so we reduced the number of values per condition (e.g. for 10 MVQCA is the least used of the three major QCA variants (Rihoux et al. 2013: 176). We are aware that there is some discussion about the validity of this approach (see Vink/van Vliet 2009 and 2013, Thiem 2013 for a recent discussion on the topic). We have tried to follow best practices for using MVQCA, especially by not unduly inflating the number of values for a condition. 9

10 REPRESSION or POLITY). In the end, we had de facto conducted a crisp-set QCA. 5. Data In this section we present the conditions we derived from the literature (see 3.). Each risk factor is represented by one or more conditions. Whenever valid data was already available in other datasets, we used this as a base for our coding. Table 2 contains a list of all conditions and their possible indicators in our initial dataset. For 16 of our 27 conditions, we had to collect our own data, usually from secondary literature about cases (these are marked as qualitative research in the column Source ). This entailed the production of detailed case studies for all 43 cases. 11 To maximize the reliability of coding decisions, a set of detailed coding rules were developed which are briefly summarized in the column Description and coding rules. 12 Due to the large amount of data and the need to divide the coding work among different coders (see Lombard/Schnyder- Duch/Campagnella 2002), we decide to maximize the reliability of the coding process by evaluating intercoder reliability. We started by conducting two test cases which both coders coded independently from each other. 13 The second test case procuded a satisfactory intercoder reliability rating of 81.25%. 14 For all further case studies, which were only coded by a single researcher, we implemented a process whereby all coding decisions were subjected to review by the other researcher and the project leader, and, where necessary, recoded. <<Table 2>> 6. Results 6.1 Procedure After satisfying ourselves that the dataset did not contain any necessary conditions, 15 we quickly found that our deductively derived causal hypotheses did not produce convincing results, resulting in large numbers of contradictions instead. Moving to an inductive research strategy, we were then 11 The first case study was conducted in November 2011, the last was completed 19 months later, in June An estimated 3000 pages of case study material was collated in this process. 12 The full coding guidelines will also be made available on the project's webpage at together with the entire dataset, upon completion of the project. 13 In the initial case study of Somalia 1991, coders disagreed on the coding of five variables which gave an intercoder reliability rating of 68.75% (11/16). The second case study, Bosnia and Herzegovina 1992, showed a higher level of agreement. Both times, the findings from the separate case studies were combined in a joint report using a process of arbitration between both researchers and the project leader. Lessons learned from these initial studies also led to a revision of the coding guidelines which were retroactively applied to the case studies of Somalia and Bosnia- Herzegovina. 14 We used a simple percent agreement method. The problem of a statistical random agreement not controlled by the method is reduced by the fact that most of our variables had multiple values. For most of the variables the random agreement therefore lies at 33.33%. As Lombard, Schnyder-Duch and Campagnella point out, referring to Neuendorf (2002), reliability coefficients of.90 or greater would be acceptable to all,.80 or greater would be acceptable to most situations (Lombard/Schnyder-Duch/Campagnella 2002: 593). 15 This step must precede the actual QCA minimization that only tests for sufficiency and thus potentially eliminates necessary conditions. POWER_PROP (2) was almost necessary, being present in all cases of collapse but one (Angola 1992). Its high level of coverage (93.3%) was matched by a similar level of consistency (87.5%) in the synchronic comparison but a lower consistency (66.7%) in the diachronic comparison. However, we suspected that this was due to an undue proximity of this indicator to the concept of collapse itself. A consideration of the condition s coding rules and the indicators for the three dimensions of state collapse suggested a limited discriminatory power of this condition. Therefore, we decided to exclude it from further analysis. 10

11 faced with the task to reduce the number of conditions. Due to the catch-all nature of our literature review, we had no less than 27 different conditions in our initial dataset. Marx (2006) suggests a maximum ratio of 0.33 conditions per case (0.25, if there are 6-7 conditions, 0.2 for 8 conditions). For our comparison of 30 and 28 cases, respectively, this left a ceiling of seven conditions. Marx and Dusa (2011) provide even lower numbers if one wants to minimize the risks of random results which meant that we wanted to use no more than five conditions. The reduction of complexity involved recoding conditions from multi-value to dichotomous but was mainly focused on the complete elimination of conditions from further analysis. Our first strategy was to select a set of conditions based on the assumption that they belonged to the same theoretical explanatory model (e.g. INFORMAL, BUR_OBSTR, SEC_OBSTR, UNOFF_MILITIA and PER- SONAL_RULE representing the shadow state hypothesis) but this proved to be inconclusive. We then examined the prime implicants that consisted of no more than three conditions and that covered at least four cases. We then recoded, or omitted from further analysis those conditions that met the following criteria: First, those who only rarely appeared as part of prime implicants. Second, conditions where values occurred whose causal impact was theoretically implausible or difficult to interpret (e.g. SAP(0), NIC(0)). Third, a seemingly random distribution of values from the same condition, i.e. when different values of a particular condition could be found among the prime implicants. After a first cut of conditions, we constructed subsets of the remaining conditions through the twostep logic recommended by Schneider and Wagemann (2012: ) which differentiates between remote and proximate conditions. Whereas remote factors refer to the context and structures that enable an outcome, proximate conditions are subject to human action and propel the causal mechanism triggering the outcome. In the two-step approach, the analysis of the outcome-enabling conditions is conducted first. After that, the second step investigates the link between each of the context factors that turned out to matter for the outcome with all of the proximate factors to arrive at findings concerning the interplay between remote and proximate conditions. We also identified conditions that could be treated as substitutes based on either the empirical results, namely co-varying values of two conditions, or conceptual proximity of the conditions themselves. 16 These strategies allowed us to test sets of five to six conditions for their explanatory value by looking whether they produced any contradictory outcomes. Against the backdrop of the observed empirical relevance of these conditions, we assumed these conditions to have a theoretical relevance for explaining the occurrence of state collapse Results of the synchronic comparison 18 The synchronic comparison points to the presence of unofficial pro-government militias (Carey/Mitchell/Lowe 2013) as a sign of imminent state collapse. The consistency level of UNOFF_MILITIA(1) as a sufficient condition for state collapse amounts to 90%, the coverage to 16 We used the following pairs of conditions as substitutes of each other: 1. MILIT and UNOFF_MILITIA, 2. AID and GOV_REV. and 3. POLAR and FACTIONAL. We assume the first pair to be substitutable based on the empirical observation that we lack any case where political groups are armed systematically but unofficial pro-government militias are absent (MILIT is a necessary condition for UNOFF_MILITIA). With regard to the second pair, the reduction of government revenues and of international aid both constitute a decline of resources available to maintain the formal and informal institutions of the state. Thus, we assume these conditions to trigger similar dynamics affecting the resilience of states. POLAR and FACTIONAL were treated as substitutes as both of these highlight a prevalence of antagonistic, group-based politics. 17 All calculatuions were conducted using TOSMANA Version (Cronqvist 2011). Coverage and consistency values were calculated manually. 18 Truth tables, Venn Diagrams, robustness checks and coverage scores can be found in the Appendix. 11

12 60%. Only a single control case (Sudan 1992) displays this condition, preventing complete sufficiency. In the synchronic comparison, the five conditions that survived our elimination strategy were FACTIONAL, GOV_REV, INCOME, MILIT and LOCAL_POLITY. The combination of the five conditions generates a truth table free of contradictions. The possible combinations of these conditions that occur in the positive cases do not overlap with any of the synchronic control cases. In other words, the conjunction of these conditions is sufficient to explain the outcome. When minimizing for Outcome = 1, we get the following complex solution formula: MILIT(1) * INCOME(0) * GOV_REV(1) * LOCAL_POLITY(1) + FACTIONAL(1) * MILIT(1) * GOV_REV(0) * LOCAL_POLITY(1) + FACTIONAL(0) * MILIT(1) * GOV_REV(1) * LOCAL_POLITY(1) + MILIT(1) * INCOME(1) * GOV_REV(1) * LOCAL_POLITY(0) + MILIT(1) * INCOME(0) * GOV_REV(0) * LOCAL_POLITY(0) + MILIT(0) * INCOME(0) * GOV_REV(0) * LOCAL_POLITY(1) Each individual term of the solution only covered a few cases so this results is still too specific and idiosyncratic. As we lost the theoretical models about potential causal paths on the way and the five conditions we ended up with were the result of an inductive procedure, we were not able to make simplifying assumptions about the relation between hypothetical cases and the outcome and to produce an intermediate solution. Therefore, we decided to calculate the parsimonious solution including logical remainders: MILIT(1) * INCOME(0) + MILIT(1) * GOV_REV(1) + INCOME(0)* LOCAL_POLITY(1) + FACTIONAL(1) * MILIT(1) * LOCAL_POLITY(1) 19 This formula can be reformulated as follows: MILIT(1) * [INCOME(0) + GOV_REV(1) + FACTIONAL(1) * LOCAL_POLITY(1)] + INCOME(0) * LOCAL_POLITY(1) The solution formula shows four alternative paths that explain the outcome of state collapse in the synchronic comparison. The militarization of politics is an INUS condition in three of these paths. It appears in combination with a) extreme poverty (below 5 per cent of the global average per capita income), b) a recent decline of government revenues and c) the existence of local pre-colonial polities combined with factionalized political competition. The fourth path to the outcome includes extreme poverty and the existence of polities that existed in parts of the territory of post-colonial states or those that emerged from the breakup of multinational empires. 6.3 Results of the diachronic comparison There were no necessary conditions in the diachronic comparison. However, TRANSITION was shown to be a sufficient condition for state collapse among the cases in our dataset but only with a limited coverage of 0.4. In other words, just seven of the collapsed cases had recently undergone a political transition so this condition only partially explains the timing of state collapse in at-risk countries. As in the synchronic comparison, UNOFF_MILITIA(1) is almost a sufficient condition 19 This required six simplifying assumtions (see Appendix). There were no contradictory assumptions when compared with the parismonious solution for Outcome = 0. 12

13 with a consistency of 90% and a coverage of 60%. Again, a single control case (this time, Georgia 2003) also had unofficial militias. We used the same strategies as in the synchronic comparison to eliminate conditions. In the end, we arrived at a truth table that was free of contradictions and made up of only five conditions: MILIT, TRANSITION, REPRESSION, AID and FACTIONAL. Substituting individual conditions (e.g. POLAR for FACTIONAL or UNOFF_MILITIA for MILIT) introduced contradictions and worsened the solution coverage. Bringing in more structural conditions (INCOME, HYBRID_REGIME and TRADE_OPEN) which had initially looked promising likewise did nothing to improve the result. In most cases these only produced more elaborate versions of the five-condition solution. 20 When minimizing for Outcome = 1, QCA produced the following complex solution formula: FACTIONAL(1) * MILIT(1) * TRANSITION(1) * REPRESSION(1) + FACTIONAL(0) * MILIT(1) * REPRESSION(1) * AID(1) + FACTIONAL(1) * MILIT(1) * TRANSITION(0) * AID(0) + FACTIONAL(1) * MILIT(1) * TRANSITION(1) * REPRESSION(0) + MILIT(0) * TRANSITION(0) * REPRESSION(1) * AID(1) FACTIONAL(1) * MILIT(1) * TRANSITION(1) * AID(1) As in the synchronic comparison we computed the parsimonious solution by including logical remainders: TRANSITION(1) + FACTIONAL(1) * MILIT(1) + REPRESSION(1) * AID(1) 21 Therefore, the synchronic comparison produced a solution formula which shows three proximate paths to state collapse which was intended to complement the more structural solutions of the synchronic comparison. Political transition in the immediate pre-collapse period turned out to be a sufficient condition for state collapse and therefore is one proximate path to, or trigger of state collapse. The combination of a factionalized and militarized political competition represents another path the relevance of these two factors had already been highlighted by the results of the synchronic comparison. A third path consists of the occurrence of an intermediate to high level of repression in combination with a decline of external financial and/or military aid. 7. Discussion The data yields interesting results with regard to the initial assumptions about the causal relevance of single risk factors or combinations thereof. First of all, the data supports our conjunctural hypotheses. There are no single necessary conditions for state collapse but there are combinations of factors that are jointly sufficient for the outcome. None of these combinations covered all cases of collapse, though, neither in the synchronic nor in the diachronic comparison. It is theoretically interesting to note that a number of conditions do not show the explanatory power that we in line with the literature had expected. For instance, regime type does not have a mea- 20 Two combinations of six conditions each (TRADE_OPEN, UNOFF_MILITIA, AID, FACTIONAL, REPRESSION, TRANSITION and INCOME, MILIT, FACTIONAL, GOV_REV, REPRESSION, TRANSITION, respectively) were at least somewhat different. However, we rejected them because results tended to be more idiosyncratic due to the higher number of conditions and because the combinations of values made little theoretical sense and were difficult to interpret. 21 This required 13 simplifying assumtions (see Appendix). There were no contradictory assumptions when compared with the parismonious solution for Outcome = 0. 13

14 surable effect on the outcome neither authoritarian nor hybrid regimes seem to be particularly at risk of collapse. Furthermore, none of the conditions associated with the erosion of state institutions due to either the destruction of formal institutions (BUR_OBSTR, SEC_OBSTR) or the construction of informal networks (INFORMAL, PERSONAL_RULE) displays strong explanatory power. Also, the majority of socio-economic and international factors as well as regional or international ones have limited value for the explanation of the outcome. Our synchronic and diachronic comparisons produce partly different results, although the difference is smaller than we initially expected. We expected the synchronic comparison to highlight structural differences between collapsed and non-collapsed states while the diachronic comparison was supposed to identify dynamic differences between situations of collapse and non-collapse within the same country. Two factors the militarization of poltical actors and factionalist politics were prominent in both solutions. Both explanations also featured some form of declining access to resources, either through a decline of state revenues (synchronic) or through the reduction of international aid (diachronic). Appropriately, the remaining two conditions from the synchronic comparison's solution are an extremely low level of per capita income and the existence of local, precolonial polities, both highly structural factors. In contrast, the diachronic comparison's solution formula included two more dynamic factors, namely a recent political transition and a high level of repression. One of the most thought-provoking observations concerns the history of states. Clapham (2004) and Buzan (1991) posit that a society s common experience or idea of independent statehood are crucial for the resilience of states. Hence we expected the absence of a polity before the period of the integration into a colonial or other kind of empire (POLITY(0)) to be a risk factor for collapse. What we found was that the existence of precolonial/preimperial polities that included only parts of the territory and the population of the postcolonial/postimperial state (POLITY(1), later recoded as LO- CAL_POLITY(1)) seemed a much riskier condition. This finding highlights a facet of the argument by Clapham and Buzan, underscoring the importance of a collectively shared history and idea of statehood. In other words, the resilience of states is promoted by a society s common identity, whereas the question whether societies have any historical experience with state institutions is much less relevant for a state's trajectory. The militarization of relevant political actors features prominently in both solution formulas. This result is even more significant when the condition is compared with its closest substitute: When UNOFF_MILITIA replaces MILIT, solution coverage decreases in both comparisons. This is all the more striking in view of the fact that the existence of unofficial pro-government militias is close to a sufficient condition in both comparisons. Instead, it is the systematic armament of all relevant interest groups that has greater explanatory power as a risk factor for the collapse of a state. The importance of this deviation should not be overstated, however, as MILIT is almost a necessary condition for UNOFF_MILITIA there is only a single case in the entire dataset (Georgia 2003) that has UNOFF_MILITIA(1) without MILIT(1). Making sense of the results from the diachronic and synchronic comparison requires some interpretation. Unfortunately, cases do not fall into clearly delineated types of collapse. For the most part, there was no pattern where those cases that were covered by a given solution term in the synchronic comparison shared a different solution term in the diachronic comparison (see Appendix). For instance, both Bosnia-Herzegovina 1992 and Afghanistan 1979 were covered by the term MILIT(1) * GOV_REV(1) in the synchronic comparison. But whereas Bosnia-Herzegovina 1992 was assigned to the term FACTIONAL(1) * MILIT(1) in the diachronic comparison, Afghanistan 1979's solution term there was TRANSITION(1). To develop a theoretical framework that explained this empirical variation, we abstracted from the QCA results because they provide insights about causal relations between certain conditions and the outcome, but not into the causal mechanisms linking them. We propose a tentative typology that differentiates between three types of structural, outcome-enabling starting situations and two proxi- 14

15 mate dynamics of escalation (see Table 3). The contextual part can be understood as the fundamental risk profile of a particular state whereas the dynamic represents the series of events that precipitates the eventual collapse. <<Table 3>> In the first context, sub-national identities are more salient to citizens than their national identity. Yet, sub-national identity groups are not particularly hostile to each other but rather towards the state itself. At the center of this model is a general alienation from the state. The state offers few tangible benefits and does not contribute to the population s well-being. State performance might be hampered by to a reduction of international aid or a decline in government revenues, diminishing the legitimacy of state domination further. Alienation is futher intensified by repressive measures against the population. In the second context, sub-national identities are likewise more relevant than a common identity of the demos. In contrast to the first model, the interaction of particularistic groups is characterized by opposition and antagonism, rather than a fundamental rejection of the state itself. Societal cleavages have left a legacy of conflict and inter-group violence which may impair the performance of state institutions. The third context describes a situation where neither state domination nor the cohabitation of different identity groups in one political community is fundamentally at odds. Instead, political groups struggle about the configuration of state authority. In other words, alternative conceptions of the organization of society compete. This include such issues as the adequate representation of societal groups, the distribution of resources, or fundamental values, e.g. centralism vs. federalism, socialism vs. capitalism, or religion vs. secularism. These three models are alternative starting situations that increase the risk of state collapse. There might be some overlaps in the details but we were able to clearly match our cases with one dominant contextual model (see Table 4). It is important to note that these models are only outcome-enabling which means that they have to be linked to the outcome through a causal mechanism. We propose two short-term dynamics that drive the escalation process and ultimately precipitate state collapse. The first dynamic centers on the mobilization of armed groups. Groups of people systematically and collectively arm themselves with the intent to resist or change the status quo. To be regarded as a dynamic of escalation the level of mobilization must increase. This means either the emergence of armed groups in the first place or the enhancement of already existing armed groups, whether quantitatively or qualitatively. Mobilization requires motivation, logistics and command, control and communication (Vinci 2006: 51-57). While the first dynamic focuses on opposition to the incumbent regime, the second dynamic highlights the importance of the elite's internal politics. Elite defection describes a disruption of the center of power. The elite is comprised of all individuals that are part of the authority structure governing the country. The authority structure depends on the type of regime and is not limited to formal offices and institutions. Therefore, the group of people that make up the elite will differ from case to case. For instance, the more power is monopolized, the smaller the elite will be. Defection is an intensifying process in this dynamic which has quantitative and/or qualitative dimensions: The quantity of elite defection relates to the number of individuals that abscond from the established authority structure. The quality of elite defection refers to the level of the authority structure where defection occurs or the nature of defection (active/confrontational vs. passive exit). <<Table 4>> 15

16 The two dynamics of escalation work independently or in conjunction to trigger state collapse. We were able to identify a dominant dynamic for each case but in many cases elements of both can be found. For instance, among the elite defection cases, only Guinea-Bissau 1998 had no evidence of popular mobilization. There is no obvious pattern that particular dynamics and contexts are more or less likely to occur together. At the present stage, the combination of contexts and dynamics is a tentative theoretical framework that formalizes our ideas about the mechanisms that leads from the structural risk context to state collapse. To gain a better understanding of these mechanisms requires more in-depth study of individual cases. To this end, we will conduct process-tracing case studies with the aim of building theory from the observation of empirical cases. The case studies will focus on how the mechanisms of the mobilization of armed groups and elite defection, respectively, escalate the initial situation into state collapse. 8. Conclusion This paper represents a fairly comprehensive test of claims and hypotheses from the academic literature about the causes of state collapse. Most of the explanations thus far focus on single, often structural conditions. Very few contributions develop conjunctural hypotheses about the interplay of factors. Our QCA analysis was very useful in accounting for the complex causality of state collapse as the results indicate a multitude of paths to collapse. Our results are sobering for the field. None of the theoretical models that had been developed, even the most elaborate ones by authors like Englehart, Reno or Lambach, found much support in our analysis. The best-known cluster of explanations come from a political economy tradition and highlight the impact of informal modes of governance on the stability of state institutions. Using concepts like neopatrimonialism, clientelism, the shadow state, the rentier state and others, these explanations present political actors as rent-seekers that strip assets from the formal state. The lack of explanatory power of these models suggests that informal/hybrid/neopatrimonial political structures are not as inflexible and decrepit as they are often portrayed. We might surmise that the small number of collapse cases that were mainly due to elite defection shows that intra-elite relations based on informal and clientelistic ties are fairly stable even in difficult times. Another reason might also be that neopatrimonialism and clientelism are characteristics of fragile states where these mechanisms might help sustain low-level equilibria and thus prevent these states from collapsing. Given the limited utility of existing theories, we had to search for inductive explanations. The solutions for the synchronic and diachronic comparisons were individually instructive but required further interpretation and theory-building when combined. Therefore, we developed a typology based on the results of the QCA comparisons and our knowledge of cases. This typology differentiates between contexts and dynamics that jointly produce state collapse. At present, it is only a starting point for further exploration of the causal mechanisms that link the various conditions to the outcome. However, despite its preliminary nature, it raises new questions for the study of state collapse. The biggest and most obvious question is what these results mean for the existing body of theory. It is clear that our results would have to be replicated in other research designs before making any strong pronouncements. Nevertheless, they raise doubts about the explanatory power of many accepted causal theories in the field, especially since our findings rest on a broader empirical base than previous contributions. Several avenues for further research present themselves. The first would to be to think more systematically about the relationship between formality and informality in constituting political order. Pre- 16

17 vious accounts viewed informal arrangements as substitutes for, or parasites of formal state institutions but we should be open for alternative forms of interaction, such as contestation or complementarity (Helmke/Levitsky 2004: 728). A recent example of such an approach is the notion of political settlements which Putzel and DiJohn define as the distribution of power between contending social groups and social classes (Putzel/DiJohn 2012: 1) and which focuses on contention and bargaining among elites, between elites and non-elites, and between social groups. Our preliminary typology of contexts and dynamics of state collapse touches upon similar points. Until now, there has been little theory-building about the impact of actors and their behaviors, with most contributions focusing on structural explanations instead. Another issue would be to think about violence and statehood. The significance of militarization as risk factor and the conception of a dynamic ensuing from this finding highlights the need to further investigate the connection between violence and state collapse. On a conceptual level, we should revisit the question inhowfar, and at which level of intensity, violence should be part of the definition or operationalization of state collapse. This should include literature on civil wars and state formation which stress that violence can also create and sustain political order as much as destroy it (Taylor/Botea 2008). Such a broader way of theorizing would go some way towards undermining the fallacy that functioning formal state institutions inevitably guarantee stability, peace and development. Bibliography Allen, Chris Understanding African Politics. Review of African Political Economy 22(65): Anderson, Benedict Imagined Communities: Reflections on the Origin and Spread of Nationalism. London: Verso. Baker, Pauline H The Conflict Assessment System Tool (CAST): An Analytical Model for Early Warning and Risk Assessment of Weak and Failing States. Washington D.C.: Fund for Peace. Baker, Pauline H. and John A. Ausink State Collapse and Ethnic Violence: Toward a Predictive Model. Parameters (Spring 1996): Baker, Pauline H. and Angeli E. Weller An Analytical Model of Internal Conflict and State Collapse: Manual for Practitioners. Washington D.C.: Fund for Peace. Basedau, Matthias and Jann Lay Resource Curse of Rentier Peace? The Ambiguous Effects of Oil Wealth and Oil Dependence on Violent Conflict. Journal of Peace Research 46(6): Bates, Robert H. 2008a. When Things Fell Apart: State Failure in Late-Century Africa. Cambridge: Cambridge University Press. Bates, Robert H. 2008b. State Failure. Annual Review of Political Science 11: S

18 Bennett, Andrew Case Study Methods: Design, Use, and Comparative Advantages. In Models, Numbers, and Cases: Methods for Studying International Relations, edited by D. Sprinz and Y. Wolinsky-Namias, Ann Arbor: University of Michigan Press. Berg-Schlosser, Dirk, Gisèle De Meur, Benoît Rihoux and Charles C. Ragin Qualitative Comparative Analysis (QCA) as an Approach. In Configurational Comparative Methods: Qualitative Comparative Analysis (QCA) and Related Techniques, edited by Benoît Rihoux and Charles C. Ragin, Los Angeles: Sage. Bethke, Felix S Zuverlässig invalide - Indizes zur Messung fragiler Staatlichkeit. Zeitschrift für Vergleichende Politikwissenschaft 6(1): Bethke, Felix S. and Daniel Lambach Ursachen von Staatskollaps and fragiler Staatlichkeit: Eine Übersicht über den Forschungsstand. Duisburg: INEF-Report 106. Bilgin, Pinar and Adam D. Morton Historicising Representations of Failed States : Beyond the Cold-War Annexation of the Social Sciences? Third World Quarterly 23(1): Bueger, Christian and Felix Bethke Actor-Networking the 'Failed State': An Enquiry into the Life of Concepts. Journal of International Relations and Development 17(1): Bussmann, Margit and Gerald Schneider When Globalization Discontent Turns Violent: Foreign Economic Liberalization and Internal War. International Studies Quarterly 51(1): Buzan, Barry People, States and Fear, 2nd edition. New York: Harvester Wheatsheaf. Call, Charles T Beyond the "Failed State": Toward Conceptual Alternatives. European Journal of International Relations 17(2): Call, Charles T The Fallacy of the Failed State. Third World Quarterly 29(8): Carey, Sabine C., Neil J. Mitchell and Will Lowe States, the Security Sector, and the Monopoly of Violence: A New Database on Pro-Government Militias. Journal of Peace Research 50(2): Carment, David, Yiagadeesen Samy and Stewart Prest State Fragility and Implications for Aid Allocation: An Empirical Analysis. Conflict Management and Peace Science 25(4): Carment, David, Souleima El-Achkar, Stewart Prest and Yiagadeesen Samy The 2006 Country Indicators for Foreign Policy: Opportunities and Challenges for Canada. Canadian Foreign Policy 13(1): Clapham, Christopher The Global-Local Politics of State Decay. In When States Fail: Causes 18

19 and Consequences, edited by R. I. Rotberg. Princeton, Oxford: Princeton University Press, Clément, Caty (2004): State Collapse: A Common Causal Pattern? A comparative analysis of Lebanon, Somalia, and the Former-Yugoslavia. PhD. Dissertation at the Department of Political and Social Sciences, Catholic University of Louvain. Clements, Kevin P., Volker Boege, Anne Brown, Wendy Foley and Anna Nolan State Building Reconsidered: The Role of Hybridity in the Formation of Political Order. Political Science 59(1): Crocker, Chester A Engaging Failed States. Foreign Affairs 82(5): Cronqvist, Lasse Tosmana: Tool for Small-N Analysis [Computer Programme], Version Trier: University of Trier. Cronqvist, L. and D. Berg-Schlosser Multi-Value QCA (mvqca). In Configurational Comparative Methods: Qualitative Comparative Analysis (QCA) and Related Techniques, edited by B. Rihoux and C. C. Ragin, Los Angeles: Sage. de Meur, Giséle, and Dirk Berg-Schlosser Comparing Political Systems: Establishing Similarities and Dissimilarities. European Journal of Political Research 26(2): Desch, Michael C War and Strong States, Peace and Weak States? International Organization 50(2): Englehart, Neil A Governments Against States: The Logic of Self-Destructive Despotism. International Political Science Review 28(2): Englehart, Neil A. and Marc V. Simon Failing States and Failing Regimes: The Prediction and Simulation of State Failure. In Dealing with Failed States: Crossing Analytic Boundaries, edited by H. Starr, London: Routledge. Esty, Daniel C., Jack A. Goldstone, Ted R. Gurr, Pamela T. Surko and Alan N. Unger State Failure Task Force Report. McLean: Science Applications International Corporation. Fukuyama, Francis The Imperative of State-Building. Journal of Democracy 15(2): Fund for Peace The Failed States Index Foreign Policy (149): Gennaioli, Nicola, and Ilia Rainer The Modern Impact of Precolonial Centralization in Africa. Journal of Economic Growth 12(3): Goertz, Gary (2006): Social Science Concepts: A User's Guide. Princeton: Princeton University 19

20 Press. Goldstone, Jack A., Robert H. Bates, Ted R. Gurr, Michael Lustik, Monty G. Marshall, Jay Ulfelder and Mark Woodward A Global Forecasting Model of Political Instability. Annual Meeting of the American Political Science Association (Washington D.C.). Goldstone, Jack A., Ted R. Gurr, Barbara Harff, Marc A. Levy, Monty G. Marshall, Robert H. Bates, David L. Epstein, Colin H. Kahl, Pamela T. Surko, John C. Ulfelder, jr. and Alan N. Unger State Failure Task Force Report: Phase III Findings. McLean: Science Applications International Corporation. Grävingholt, Jörn &. Ziaja Sebastian &. Kreibaum Merle State Fragility: Towards a Multi- Dimensional Empirical Typology. DIE Discussion Paper. Bonn: Deutsches Institut für Entwicklungspolitik. (yndk_contentbykey)/anes-8sejdv/$file/dp% pdf [12. September 2013]. Gros, Jean-Germain Towards a Taxonomy of Failed States in the New World Order: Decaying Somalia, Liberia, Rwanda and Haiti. Third World Quarterly 17(3): Hameiri, Shahar Failed States or a Failed Paradigm? State Capacity and Limits of Institutionalism. Journal of International Relations and Development 10: Helmke, Gretchen and Steven Levitsky Informal Institutions and Comparative Politics: A Research Agenda. Perspectives on Politics 2(4): Hout, Wil Between Development and Security: The European Union, Governance and Fragile States. Third World Quarterly 31(1): Howard, Tiffiany O Revisiting State Failure: Developing a Causal Model of State Failure Based Upon Theoretical Insight. Civil Wars 10(2): Herrmann, Andrea Monika and Lasse Cronqvist When Dichotomisation becomes a Problem for the Analysis of Middle-Sized Datasets. International Journal of Social Research Methodology 12(1): Iqbal, Zaryab and Harvey Starr State Failure: Conceptualization and Determinants. Annual Meeting of the Interational Studies Association (Chicago). King, Gary, Robert O. Keohane and Sidney Verba Designing Social Inquiry: Scientific Inference in Qualitative Research. Princeton: Princeton University Press. King, Gary and Langche Zeng Improving Forecasts of State Failure. World Politics 53(4): Krasner, Stephen D. and Carlos Pascual Addressing State Failure. Foreign Affairs 84(4): 20

21 Lambach, Daniel Warum kollabieren Staaten? In Identität, Institutionen and Ökonomie: Ursachen innenpolitischer Gewalt, edited by M. Bussmann, A. Hasenclever and G. Schneider, Sonderheft 43 der Politischen Vierteljahresschrift. Wiesbaden: VS. Lambach, Daniel Staatszerfall and regionale Sicherheit. Baden-Baden: Nomos (Reihe Internationale Beziehungen Bd. 10). Lambach, Daniel and Dragan Gamberger A Temporal Analysis of Political Instability Through Subgroup Discovery. Conflict Management and Peace Science 25(1): Lambach, Daniel, Eva Johais and Markus Bayer An Anchor for a Dazzling Debate: Conceptualizing State Collapse. Paper presented at the 8 th Pan-European Conference on International Relations, Warsaw, September Lombard, Matthew, Jennifer Schnyder-Duch and Cheryl Campagnella Content Analysis in Mass Communication. Assessment and Reporting of Intercoder Reliability. Human Communication Research 28(4): Lucas, Viola, and Thomas Richter Taxation, Distribution and Authoritarian Regime Longevity: Evidence from the Global State Revenues and Expenditures Data Set. Paper presented at the Annual Meeting of the American Political Science Association, New Orleans, 30 August 2 September Mackie, John L Causes and Conditions. American Philosophical Quarterly 2(4): Mahoney, James After KKV: The New Methodology of Qualitative Research. World Politics 62(1): Mahoney, James and Gary Goertz (2006): A Tale of Two Cultures: Contrasting Quantitative and Qualitative Research. Political Analysis 14(3): Margolis, J. Eli Estimating State Instability. Studies in Intelligence 56(1): Marshall, Monty G. and B.R. Cole Global Report on Conflict, Governance, and State Fragility Foreign Policy Bulletin 18(1): Marshall, Monty G. and Jack Goldstone Global Report on Conflict, Governance, and State Fragility 2007: Gauging System Performance and Fragility in the Globalization Era. Foreign Policy Bulletin 17(1): Marshall, Monty; Ted Robert Gurr and Keith Jaggers Polity IV Project Regime 21

22 Characteristics and Transitions, Dataset User's Manual, 30 April Vienna: Center for Systemic Peace. Marx, Axel Toward More Robust Model Specification in Qca: Results from a Methodological Experiment." COMPASSS Working Paper. Brussels: COMPASSS. Marx, Axel, and Adrian Dusa Crisp-Set Qualitative Comparative Analysis (Csqca), Contradictions and Consistency Benchmarks for Model Specification. Methodological Innovations Online 6(2): Marx, Axel, Benoît Rihoux, and Charles Ragin The Origins, Development, and Application of Qualitative Comparative Analysis: The First 25 Years. European Political Science Review FirstView: Meagher, Kate The Strength of Weak States? Non-State Security Forces and Hybrid Governance in Africa. Development and Change 43(5): Menkhaus, Ken, and Jacob N. Shapiro Non-State Actors and Failed States: Lessons from Al- Qa'ida's Experiences in the Horn of Africa. In Ungoverned Spaces: Alternatives to State Authority in an Era of Softened Sovereignty, edited by Anne L. Clunan and Harold A. Trinkunas, Stanford: Stanford University Press. Menzel, Ulrich Der Zerfall der postkolonialen Staaten. Aus Politik and Zeitgeschichte (B 18-19/2001): 3-5. Milliken, Jennifer and Keith Krause State Failure, State Collapse and State Reconstruction: Concepts, Lessons and Strategies. Development and Change 33(5): Neuendorf, K. A. (2002): Content Analysis guidebook. Thousand Oaks: Sage. Norton, Richard J. and James F. Miskel Spotting Trouble: Identifying Faltering and Failing States. Naval War College Review 50(2): Ragin, Charles C Fuzzy-Set Social Science. Chicago: Chicago University Press. Ragin, Charles C The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies. Berkeley: University of California Press. Reno, William Shadow States and the Political Economy of Civil Wars. In Greed and Grievance: Economic Agendas in Civil Wars, edited by M. Berdal and D. M. Malone, Boulder, London: Lynne Rienner. Rice, Susan E. and Stewart Patrick Index of State Weakness in the Developing World. Brookings Institution. 22

23 Rihoux, Benoît Qualitative Comparative Analysis (Qca), Anno 2013: Reframing the Comparative Method's Seminal Statements. Swiss Political Science Review 19(2): Rihoux, Benoît Qualitative Comparative Analysis (QCA) and Related Systematic Comparative Methods: Recent Advances and Remaining Challenges for Social Science Research. International Sociology 21(5): Rihoux, Benoît, and Gisèle de Meur Crisp-Set Qualitative Comparative Analysis (Csqca). In Configurational Comparative Methods: Qualitative Comparative Analysis (Qca) and Related Techniques, edited by Benoît Rihoux and Charles C. Ragin, Los Angeles: Sage. Rihoux, Benoît and Charles C. Ragin (eds.) Configurational comparative methods. Qualitative comparative analysis (QCA) and related techniques. Los Angeles, Calif: SAGE (Applied social research methods series, 51). Risse, Thomas Governance Configurations in Areas of Limited Statehood: Actors, Modes, Institutions, and Resources. SFB-Governance Working Paper. Berlin: DFG- Sonderforschungsbereich 700. Rotberg, Robert I. (ed.) When States Fail: Causes and Consequences. Princeton: Princeton University Press. Rotberg, Robert I. (ed.) State Failure and State Weakness in a Time of Terror. Washington: Brookings Institution Press. Ruf, Werner (ed.) Politische Ökonomie der Gewalt: Staatszerfall and die Privatisierung von Gewalt and Krieg. Opladen: Leske + Budrich. Schneckener, Ulrich (ed.). 2006a. Fragile Staatlichkeit: States at Risk zwischen Stabilität and Scheitern. Baden-Baden: Nomos. Schneckener, Ulrich. 2006b. Charakteristika and Dynamiken fragiler Staatlichkeit Zur Auswertung der Fallstudien. In Fragile Staatlichkeit: States at Risk zwischen Stabilität and Scheitern, edited by U. Schneckener, Baden-Baden: Nomos. Schneider, Carsten Q. and Claudius Wagemann Set-theoretic methods for the social sciences. A guide to qualitative comparative analysis. Qualitative Comparative Analysis (QCA) and Fuzzy Sets. Cambridge: Cambridge Univ. Press. Schneider, Carsten Q. and Claudius Wagemann Reducing complexity in Qualitative Comparative Analysis (QCA): Remote and proximate factors and the consolidation of democracy. European Journal of Political Research 45(5): Simons, Anna and David Tucker The Misleading Problem of Failed States: A Socio- 23

24 Geography of Terrorism in the Post-9/11 Era. Third World Quarterly 28(2): Taylor, Brian D. and Roxana Botea Tilly Tally: War-Making and State-Making in the Contemporary Third World. International Studies Review 10(1): Themnér, Lotta and Peter Wallensteen Armed conflict, Journal of Peace Research 49(4): Thiem, Alrik Clearly Crisp, and Not Fuzzy: A Reassessment of the (Putative) Pitfalls of Multi-value QCA. Field Methods 25(2): Ulfelder, Jay. (2012): State Failure Has Failed. How About Giving State Collapse a Whirl? Available online: [01. October 2012]. UNCTAD (various years): UNCTAD Commodity Yearbook. New York: United Nations Conference on Trade and Development. Vinci, Anthony (2006): The Problems of Mobilization and the Analysis of Armed Groups. Parameters 36(1): Vink, Maarten P., and Olaf van Vliet Not Quite Crisp, Not Yet Fuzzy? Assessing the Potentials and Pitfalls of Multi-value QCA. Field Methods 21 (3): Vink, Maarten P., and Olaf van Vliet Potentials and Pitfalls of Multi-value QCA: Response to Thiem. Field Methods 25(2): Wagemann, Claudius and Carsten Q. Schneider Qualitative Comparative Analysis (QCA) and Fuzzy-Sets: Agenda for a Research Approach and a Data Analysis Technique. Comparative Sociology 9(3): van de Walle, Nicolas The Economic Correlates of State Failure: Taxes, Foreign Aid, and Policies. In When States Fail: Causes and Consequences, edited by : R. I. Rotberg, Princeton: Princeton University Press. Woodward, Susan L Failed States: Warlordism and Tribal Warfare. Naval War College Review 52 (2), S World Bank (2011): World Development Report Washington: World Bank. Zartman, I. William (ed.) Collapsed States: The Disintegration and Restoration of Legitimate Authority. Boulder: Lynne Rienner. 24

25 Appendix Synchronic comparison Table 5: Truth Table for synchronic comparison 22 case_id factional milit income gov_rev local_po lity outcome Croatia Guinea 1996, Sri Lanka Ethiopia 1974, Burundi Iran 1979, Uzbekistan 1999, Cambodia Nigeria 1967, Niger Moldova Lesotho Mali Burkina Faso Sudan Sierra Leone Uganda Somalia 1991, Chad Tajikistan 1992, Bosnia-Herzegovina 1992, Georgia Afghanistan Liberia Angola Congo-Kinshasa 1960, Lebanon Zaire Laos Guinea-Bissau Remainders are omitted from the table. 25

26 Fig. 1: Venn Diagram for synchronic comparison Results for Outcome = 0 and simplifying assumptions To check the robustness of our results, we conducted an analysis for Outcome = 0, including logical remainders for reduction. This produced the following solution formula: MILIT(0) * INCOME(1) + MILIT(0) * LOCAL_POLITY(0) + FACTIONAL(0) * INCOME(1) * GOV_REV(0) + INCOME(1) * GOV_REV(0) * LOCAL_POLITY(0) While this does not precisely mirror the solution for Outcome = 1, all conditions have the expected values: a per-capita income above 5 per cent of the global average, constant or increasing state revenues and an absence of militarization, factionalism and localized pre-colonial polities. This can be further reduced to: MILIT(0) * [INCOME(1) + LOCAL_POLITY(0)] + INCOME(1) * GOV_REV(0) * [FACTIONAL(0) + LOCAL_POLITY(0)] In other words, there are two major sets of conditions that predict the non-occurrence of state 26

27 collapse in the synchronic comparison. The first one consists of an absence of militarization and either a lack of extreme poverty or a lack of local, precolonial polities. The second one describes states that are not excessively poor and that do not suffer from declining state revenues with either non-factional politics or absence of pre-colonial polities. There were no contradictory simplifying assumptions for the parsimonious solutions for Outcome = 1 and Outcome = 0. The simplifying assumptions for Outcome = 1 were: FACTIONAL(0) * MILIT(0) * INCOME(0) * GOV_REV(1) * LOCAL_POLITY(1) + FACTIONAL(0) * MILIT(1) * INCOME(0) * GOV_REV(0) * LOCAL_POLITY(1) + FACTIONAL(0) * MILIT(1) * INCOME(0) * GOV_REV(1) * LOCAL_POLITY(0) + FACTIONAL(1) * MILIT(0) * INCOME(0) * GOV_REV(1) * LOCAL_POLITY(1) + FACTIONAL(1) * MILIT(1) * INCOME(0) * GOV_REV(1) * LOCAL_POLITY(0) + FACTIONAL(1) * MILIT(1) * INCOME(1) * GOV_REV(1) * LOCAL_POLITY(1) For Outcome = 0 they were: FACTIONAL(0) * MILIT(0) * INCOME(0) * GOV_REV(1) * LOCAL_POLITY(0) + FACTIONAL(1) * MILIT(0) * INCOME(0) * GOV_REV(0) * LOCAL_POLITY(0) + FACTIONAL(1) * MILIT(0) * INCOME(0) * GOV_REV(1) * LOCAL_POLITY(0) + FACTIONAL(1) * MILIT(0) * INCOME(1) * GOV_REV(0) * LOCAL_POLITY(1) + FACTIONAL(1) * MILIT(0) * INCOME(1) * GOV_REV(1) * LOCAL_POLITY(0) Table 6: Coverage scores of solution terms for outcome = 1 Term Cases covered RC UC MILIT(1) * INCOME (0) MILIT(1) * GOV_REV(1) INCOME(0) * LOCAL_POLITY(1) Sierra Leone 1998+Uganda 1985+Somalia 1991,Chad 1979+Liberia 1990+Laos 1960 Sierra Leone 1998+Somalia 1991,Chad 1979+Tajikistan 1992,Bosnia-Herzegovina 1992,Georgia 1991+Afghanistan 1979+Angola 1992 Sierra Leone 1998+Uganda 1985+Somalia 1991,Chad 1979+Zaire 1996+Guinea-Bissau % 13.3% 53% 33.3% 40% 13.3% FACTIONAL(1) * MILIT (1) * LOCAL_POLITY(1) Sierra Leone 1998+Uganda 1985+Congo-Kinshasa 1960,Lebanon 1975 RC = Raw coverage, UC = Unique coverage Cases that are unique to a particular term are highlighted in italics. 26.7% 13.3% Diachronic comparison Table 7: Truth Table for diachronic comparison 23 case_id factional milit transition repression aid outcome Afghanistan 1973, Somalia Angola 1975, Zaire 1977, Laos Remainders are omitted from the table. 27

28 Chad 1965, Guinea-Bissau 1980, Uganda 1971 Georgia 2003, Lebanon 2005, Tajikistan Liberia Sierra Leone Afghanistan 1979, Chad Angola Bosnia-Herzegovina 1992, Georgia 1991, Lebanon Congo-Kinshasa Zaire Guinea-Bissau Laos Liberia 1990, Somalia Sierra Leone 1998, Tajikistan Uganda Fig. 2: Venn Diagram for diachronic comparison 28