Arming for conflict,arming for peace? How small arms imports affect intrastate conflict risk

Abstract

Although a prevalent technology of conflict, the impact of small arms imports on the risk of intrastate conflict outbreak has not been examined so far. This article argues that small arms not only enhance general military capabilities, but also contribute to state capacities necessary for conflict prevention. These two mechanisms are incorporated in a formal model of power shifts. The derived hypotheses are tested on 146 countries for the period 1993–2014. Using split-population and penalized fixed-effects logit models as innovative estimation methods for rare-events data, small arms imports are found to have no or even a risk-reducing impact.

Keywords

Small arms arms imports intrastate conflict formal conflict theory

Introduction

Most deaths in civil conflicts result from small arms (Wille and Krause, 2005, cf. Kreutz and Marsh, 2012). Exports of small arms are under public scrutiny because they are deemed to lead to more severe conflicts (Bourne, 2007; Mehrl and Thurner, 2020), destabilize whole countries and regions (e.g. Greene and Penetrante 2012) and worsen repression and human rights violations (Blanton, 1999; de Soysa et al., 2010). As Pamp et al. (2018) have shown, major conventional weapons transfers indeed raise the risk of intrastate conflicts. However, conjectures on escalating effects of governmental small arms acquisitions cannot be backed up by proper theoretical and empirical evidence in the literature—their specific impact on the risk of intrastate conflicts has not been investigated so far.

Assessments of small arms’ effects on conflict outbreaks usually focus on illicit trade and privately owned weapons (e.g. Ayuba and Okafor, 2015; Banerjee and Muggah, 2002; Chivers, 2011; Stohl and Myerscough, 2007). They mostly conclude that the availability of small arms facilitates violent uprisings of non-state actors and is thus conducive to insurgency and conflict risk. Therefore, the small arms trade is generally associated with detrimental regional and internal security developments (Wezeman, 2003). However, in countries with appropriately structured security forces like Ghana and post-conflict Liberia, governmental small arms imports might contribute to domestic security and peacekeeping (cf. Blair 2020). Furthermore, international security aid programs, e.g. for countries like Mali, Kenya, Somalia or Afghanistan, specifically address military and police state capacities by providing training and equipment to security forces, as they deem these state capacities necessary to build up peace, stability and development (cf. Brown and Grävingholt, 2016; Omelicheva et al., 2017).

Do the provision and importation of small arms to governments help to advance these goals, or do they rather have adverse effects? Should responsible supplier countries limit small arms exports as strictly as possible? Or can they play a sensible role in more comprehensive strategies integrating good governance, lawful security forces and economic development? These are essential questions for all strategies on conflict prevention, security and development. Therefore, evaluating the effects of arms imports is of paramount relevance. These questions have, however, only been investigated in the literature for the case of major conventional weapon (MCW) imports (Craft, 1999; Suzuki, 2007; Pamp et al., 2018) or without discerning between major weapons and small arms (Craft and Smaldone, 2002; Magesan and Swee, 2018).

Investigating small arms transfers separately from major conventional weapons seems warranted, from both an empirical and a theoretical standpoint. The observed trade patterns diverge substantially, as shown by a low correlation between the two types of imports.¹ At the same time, we theoretically expect differences in their demand and impact on conflict between these arms types. Small arms can play a distinct role in enhancing state capacities in peaceful times and during the emergence of conflict: not only the military but also police forces have to be equipped with small arms, which are a necessary tool for providing security and public order (cf. Karp, 2018; Bourne, 2007). While major weapons imports cause a considerable burden on the state budget, small arms have much lower unit costs. They might even lead to increased fiscal capacity when tax offices have sufficient authority to collect taxes, ensured by a sufficiently equipped police.

To capture these different mechanisms in a coherent framework, this article for the first time—to my knowledge—implements armament in a formal model of intrastate conflict, adapted from Powell (2012). In this model, power shifts in favor of the government can cause conflict when they are too large to allow compensation of the opposing rebel group. While arms imports generally lead to such shifts in power, small arms at the same time enhance conflict-preventing state capacities. Since they are also tools for securing the state's monopoly of violence, they might impede the emergence of conflict. When I integrate this mechanism into the formal model, small arms can have a neutral or even mitigating effect on conflict risk.

To evaluate the theoretical predictions, I employ innovative empirical methods on time-series cross-sectional data on 146 countries over the time span 1993–2014. Onsets of intrastate conflicts are rare events, posing particular challenges to empirical research. Usually, one would prefer to use models that can account for unobserved heterogeneity between countries. However, standard fixed effects models would lead to biased estimates, as countries without any onset, i.e. no variance in the dependent variable, have to be excluded. In fact, many countries do not exhibit any conflict over the observed period and have not done so for a long time or even since the beginning of their existence.

Cook et al. (2020) propose a penalized maximum likelihood-fixed effects estimator for such cases. This estimator uses the pool of all countries without onset as one joint category of fixed effects. Their infinity-tending parameter gets shrunk to zero by penalization. Thus, it enables one to correctly control for unobserved differences between all countries. As an alternative approach, Beger et al. (2014, 2017) develop a split-population duration estimator. It accounts for a distinction between consolidated countries that are “immune” to conflict and countries “at risk”, as they term it. The split-population model incorporates two equations: the immunity equation determines the probability at which a country-year even is at risk of a conflict; and the duration equation estimates how the covariates affect the conflict risk for the subpopulation of country-years at risk. As Pamp et al. (2018) point out, analyses of arms imports are also subject to potential reverse causality problems because the imminent risk of a conflict might prompt the state to increase its imports. Therefore, an instrumental variable (IV) approach guarantees that results are not substantially affected by endogeneity issues.

My empirical results corroborate previous findings regarding major weapons, which increase the risk of intrastate conflicts. However, small arms imports seem to have a distinct, different effect: while standard probit models do not show a statistically significant impact, instrumental variable models as well as fixed effects and split-population duration estimations exhibit a significant impact of small arms reducing conflict risk. In contrast to the destabilizing effect of major weapons, small arms are shown to not increase conflict risk and may actually have a rather stabilizing influence.

The remainder of the article is structured as follows: I begin by giving an overview of the literature on armament and intrastate conflict onset, presenting empirical results and presumed theoretical mechanisms. Next, the specific theoretical effects of small arms imports on state capacities are discussed informally. These are then formally implemented in a power shifts framework. The respective hypotheses are tested in the empirics section, which details methods, data and operationalization and delivers the results. Finally, I discuss the findings and avenues for further research.

The state of the literature: empirical evidence and potential mechanisms for arms imports affecting conflict risk

The following review of the literature first summarizes the empirical evidence regarding the effects of arms imports on the risk of intrastate conflict onset. An overview of theoretical mechanisms on the emergence of conflict then provides context for the theoretical framework laid out in the subsequent section.

The empirical evidence so far

The few studies that analyze the effects of arms imports on civil conflict onset do not provide conclusive evidence on specific arms types like small arms, as they differ in scope, methods, data used and results. While Craft (1999) and Craft and Smaldone (2002) use both inter- and intrastate conflict onsets combined as a single dependent variable, Durch (2000), Suzuki (2007), Magesan and Swee (2018) and Pamp et al. (2018) focus explicitly on intrastate conflicts. Craft (1999), Craft and Smaldone (2002), Magesan and Swee (2018) and Pamp et al. (2018) find statistically significant positive effects of arms imports on conflict risk. However, the effect is not significant in Suzuki (2007) and negative but also not significant in Durch (2000). In contrast to the repression literature (Blanton, 1999; Brender and Pfaff, 2018; de Soysa et al., 2010), small arms have not played a role in research on civil conflict outbreaks so far: only Craft and Smaldone (2002) and Magesan and Swee (2018) even include small arms in their analyses, but lumped together with major weapons, using data from the World Military Expenditures and Arms Transfers (WMEAT) database and on US Direct Commercial Sales (DCS), respectively.

Actually, only Pamp et al. (2018) and Magesan and Swee (2018) use a wide-ranging sample of both developing and developed countries and apply advanced econometric techniques to not only establish correlations but provide causal findings. Pamp et al. (2018) cover 137 countries over the time 1953–2013. Employing simultaneous equations and instrumental variable models to account for reverse causality, they present strong evidence of a positive causal effect of major weapons imports on the probability of civil conflict outbreaks. However, they do not consider the role of small arms transfers. Magesan and Swee (2018) focus on US DCS to 191 countries over the period 1970–2008. Exploiting price effects owing to US inflation, their instrumental variable models indicate that higher US sales lead to a higher risk of conflict in the buyer country. Again, their data do not distinguish between small arms and major weapons. In sum, there exists no empirical assessment of the distinct impact of small arms so far.

Theoretical mechanisms of conflict onset

Like in empirical research, governmental imports of small arms have not been the subject of systematic theoretical considerations on intrastate conflict onset so far. To scrutinize the potential effects of small arms in the next section, it is nevertheless helpful to summarize the usually proposed theoretical mechanisms associated with military endowments and arms imports in general.

Arms imports are a significant cornerstone of state capacity, which in turn is integral to analyzing conflict risk (cf. e.g. Fearon and Laitin, 2003; Fjelde and De Soysa, 2009). Hendrix (2010) discerns different types of state capacity providing valuable guidance for examining the effects of arms imports. Obviously, arms imports are expected to enhance the military capacity of the state. In a simple contest success function framework (cf. Tullock, 1980; Hirshleifer, 2000), higher military capacity can deter potential adversaries, leading to lower conflict risk (for more complex analyses of interstate conflicts see e.g. Huth, 1988; Huth and Russett, 1988; Jervis, 1989; Jackson and Morelli, 2009). Chassang and i Miquel (2010) develop a formal model for interstate conflicts to show that increases in weapons stocks might not always be deterring but can also lead to preemptive attacks, as Craft and Smaldone (2002) argue as well. Deterrence might also prove to be too costly for avoiding conflict (Coe, 2011).

At the same time, arms imports also affect a state's fiscal capacity: in particular, major weapons require substantial expenses, reducing the fiscal capacity for different spending purposes like fiscal redistribution (cf. Azam 2001), as described by the gun-vs.-butter trade-off in Powell (1993). The state's capabilities to prevent conflicts non-violently are thus reduced. In contrast, procurements of small arms might even result in a higher extractive capacity, meaning the ability to raise taxes and accrue rents, e.g. from natural resources.

Apart from direct capacity effects, arms imports affect the conflict risk owing to signaling and exacerbated commitment problems. In settings of asymmetric information, as is usually assumed for civil conflicts (Fearon, 1995; Walter, 2009), arms imports can function as a signal of resolve or of incompletely observable military state capacity. Again, this might lead to effective deterrence (cf. Huth and Russett, 1988; Jervis, 1989), but also to preemptive attacks by challengers to avoid further deterioration of the military balance in favor of the government. Powell (2006) argues that such power shifts can induce conflict if it is not possible to commit to future compensation of the challenger for the losses expected owing to the power shift. The following section adapts a formal model originally proposed by Powell (2012, 2013), incorporating these mechanisms.

Building a theory of small arms imports and conflict risk

So far, acquisitions of small arms have not been a subject of theories on the onset of intrastate conflict. In addition to the general theoretical mechanisms concerning armament and conflict outbreak laid out in the section above, I argue that small arms need to be analyzed separately because they potentially exhibit different theoretical mechanisms leading to different overall effects on conflict risk.

Under a perspective of specific state capacities, imports of major conventional weapons could be hypothesized to deter potential challengers owing to increased military state capacity. However, causal analyses actually find that major weapons imports increase conflict risk (Pamp et al., 2018). Powell (2012) provides a formal model that can help in understanding the underlying mechanisms: power shifts owing to arms importation diminish an opposition group's expectations of their future income share. Thus, the opposition has to be compensated for such power shifts. When the government cannot credibly commit to paying off the opponent in the future, this compensation cannot exceed the current income. With a power shift that is too large, the opposition cannot be compensated sufficiently, and it has to fight the government. The government's build-up of military capacity thus triggers conflict.

In general, small arms imports can be interpreted as power shifts too. However, as is argued in the next section, small arms also increase expected future wealth owing to enhanced state capacities. With this prospect of larger welfare, the impact of the power shift on the opposition group is counteracted, potentially resulting in a higher conflict threshold, i.e. reduced risk, in Powell's model.

This section sets out how small arms affect state capacities and thus intrastate conflict risk and how their effect differs from major weapons, resulting in potentially differing overall effects. Then, an appropriate adaptation of Powell’s (2012) bargaining model of conflict is presented to implement these mechanisms formally.

Small arms, state capacities and conflict

One of the most important factors to prevent conflicts is a state with sufficient capacities to solve dissent non-violently, providing security, stability and economic welfare (cf. Fearon 2003). In the following, the specifics of small arms imports compared with major weapons are described, and their differential effect on conflict-preventing, security and fiscal state capacities—and thus conflict risk—is laid out.

Small arms as distinct technology of conflict

This article argues that small arms should be considered a distinct “technology of conflict”, separate from major weapons (cf. Jia and Skaperdas, 2012; Hirshleifer, 2000; Kalyvas and Balcells, 2010). Small arms require only minimal training and are readily available to be imported at short notice (Boutwell and Klare, 1999), while major arms acquisitions often take years to be realized.² Accordingly, small arms trade flows substantially diverge from trade of MCW: Lebacher et al. (2021) find that country dyads exhibit a negative correlation between small arms and MCW transfer volume. In the data used in the following sections, the correlation between small arms and MCW imports in a country-year is only 0.24. This indicates different underlying reasons for governments to import and consequently—at least intended—different effects.

Enhanced conflict-preventing and security state capacities

Importing major weapons delivers fighting capacity in a particularly narrow sense. They are only suited for deployment in large-scale combats in intra- or interstate conflicts (or the deterrence thereof) as well as in situations preceding conflict, e.g. when repressing mass demonstrations with armored vehicles and tanks. This might in turn ignite conflict (Hultquist, 2017; Sullivan et al., 2018) but could also support demobilizing of the opposition, as shown recently by Chiang (2021), and thus reduce conflict risk. Small arms are suitable for use in the same contexts (de Soysa et al., 2010),³ but provide additional utility to governments in peaceful times. For governments, small arms are necessary instruments to provide domestic security and uphold public authority (cf. Bourne, 2007; Gobinet et al., 2011; Karp 2018).

In contrast to major weapons, small arms are not only tools of the military, but also the primary endowment of police forces (Boutwell and Klare, 1999; Karp, 2018). Importing small arms can thus lift “material constraints” in the equipment of police and security forces that otherwise lead to insecurity (Mehler, 2012: 49; Oreta, 2009; cf. Adekoya and Abdul Razak, 2017). Since these imports fulfill essential functions for governments, “richer, better-governed countries” can be expected to have higher volumes of such imports, as is in fact observed by de Soysa et al. (2009: 86). Describing the pathways to state collapse and conflict, Rotberg (2002: 129) argues that states in this situation are not able to provide necessary services to the population and “security, the most important political good, vanishes.” Insecurity provokes reactions that further undermine a state's authority and monopoly of violence because groups and organizations begin to provide armed “security” services autonomously (Atwood, 2005). Such non-state actors usually rely on poorly protected or managed government stockpiles for arming (Jackson, 2010). However, with well-equipped forces, leakage of arms to non-state actors can be prevented, inhibiting the emergence of conflict. Likewise, contagious effects of nearby conflicts can be contained (Braithwaite, 2010). Small arms imports crucially contribute to these security and police state capacities, which are central to preventing intrastate conflict (cf. Fearon and Laitin, 2003; Collier and Hoeffler, 1998). Thus, in contrast to major weapons, they might exhibit a risk-reducing impact—possibly counteracting the risk-inducing power shift effect caused by small arms imports as well.

For this conjunction of state capacity and small arms imports, Ghana provides an illustrative case. In our sample period, it is among the largest importers of small arms in Sub-Saharan Africa. As a contributor to peacekeeping missions, Ghana enhanced its military equipment, while also improving security within its borders. This strengthened the general development path resulting in a period of absence of state-based conflict (Banini et al., 2020; Birikorang, 2007).

Similar developments took place in post-conflict Liberia, where recorded small arms imports started only after the conflict was settled in 2003 and the strict arms embargo was lifted. Blair (2020) details how Liberia's police was restructured from an “instrument of repression and terror” to a stabilizing factor in the post-conflict period (cf. Joyce 2020). In the process, the Emergency Response Unit and the Police Support Unit were formed as special units which received better equipment, including small arms, and training (Friedman, 2011). They were specifically established to deal with arrests of armed criminals, violent crimes in progress, hostage situations and armed terrorist activities (Washington, 2008). In Liberia, small arms thus provided the necessary tools for the success of the restructured police and contributed to improved security and military state capacities for lowering conflict risks.

In like manner, US security aid programs specifically aim at enhancing counterterrorism efforts and general security by training and equipping the police and military forces (Berman, 2003; Omelicheva et al., 2017). Supported countries like Bahrain, Belize and Saudi Arabia imported significant amounts of small arms over the sample period of this article, without outbreaks of intrastate conflicts. However, avoidance of larger-scale conflict in these countries came at the cost of other types of violence: in Bahrain, for instance, protests in 2011 were violently dispersed by security forces, curbing an uprise that could have emerged in a conflict. Nevertheless, in the event of potentially escalating smaller-scale unrest, military and security state capacities, enhanced by small arms imports, can effectively prevent the escalation to a conflict outbreak (Chiang, 2021). In contrast, Syria, which experienced the deadliest conflict in the sample period, had only minor imports of small arms recorded.

Fiscal and economic state capacities

In a standard guns-vs.-butter framework (Powell, 1993), higher investments in the military diminish the state's fiscal capacity. It might then be less able to provide public goods and redistribution for alleviating potentially conflict-inducing grievances (cf. Azam 2001). However, small arms imports differ in their financial consequences decisively from major weapons. While the global small arms trade in 2014, the latest year with data available from the Norwegian Initiative on Small Arms Transfers (NISAT), had a volume of 5.9 billion 2012 US-dollars (including 1.1 billion 2012 US-dollars for military-grade small arms), WMEAT records a total of 188.5 billion 2012 US-dollar in all arms trade. This disparity illustrates the different dimensions between small arms and major weapons trades and financial expenses for governments. Whereas investments in major weapons thus impact national budgets substantially, small arms acquisitions pose a relatively small financial burden. Fiscal capacities that enable governments to curb economic discontent are thus not significantly affected by expenditures for small arms imports. A state's security forces, adequately equipped with small arms, may even enhance its taxing capacity by supporting tax authorities’ enforcement measures, potentially increasing government revenue (cf. Risse 2012). This can in turn foster public goods provision and reduce conflict-inducing grievances (Azam 2001; Fjelde and De Soysa, 2009; Justino and Martorano, 2018; Taydas and Peksen, 2012).

General security is also deemed an “important determinant of welfare” (Fafchamps and Minten, 2009: 831), as insecurity impedes the state's fiscal capacity, investment, competitiveness of firms, income and economic growth and thus general development (Adekoya and Abdul Razak, 2017; Atwood, 2005; Besley and Mueller, 2021; Cárdenas-Santamaría, 2007; Estrada and Ndoma, 2014; Goulas and Zervoyianni, 2013, 2015; Oriakhi and Osemwengie, 2012; Torres-Preciado et al., 2017; World Bank, 2006). With small arms enabling security forces to reduce insecurity, they might contribute to raising economic welfare in the future. Again, Ghana and Liberia provide illustrative examples with stable development and gross domestic product (GDP) growth over the observed period or, in the case of Liberia, since the end of the conflict in 2003, as a result of general stabilization. Fiscal and economic state capacities are in turn a critical determinant of conflict risk since they raise opportunity costs of conflict and provide financial capabilities to reduce grievances (Collier and Hoeffler, 1998, 2004; Jakobsen et al., 2013; Justino and Martorano, 2018; Ray and Esteban, 2017; Taydas et al., 2010).

In sum, the theoretically deduced effects of small arms on fiscal and economic state capacities are not clear-cut risk-inducing as they are for major weapons. Instead, they could have a risk-reducing impact under certain circumstances. Since the financial burden of paying for small arms importation is substantially lower than for major weapons, it has a less detrimental effect on public goods provision, allowing the reduction of conflict-provoking grievances. At the same time, the emergence of conflict can be counteracted by raised security state capacities, also supporting future economic development. In the formal model in the next section, the enhancement of state capacities by small arms imports leads to an increase in expected future utility, thereby reducing conflict risk. In contrast to the effects of major weapons, these state capacity effects partly offset or even overcompensate for the negative, conflict-inducing impact of power shifts in favor of the government.

A formal model of conflict, power shifts and state capacity

To provide a more formal assessment of the concurrence of these different theoretical mechanisms, I apply a bargaining model from Powell (2012), where power shifts induced by arms imports are at the center of the analysis. If these power shifts are too large, meaning that the government cannot compensate the opposition sufficiently, conflict occurs.

Both small arms and major weapons induce power shifts in favor of the government. These increase the risk of conflict when the opponent sees conflict as a profitable strategy to prevent the government from consolidating its new position of strength. In contrast to major weapons, small arms imports also enhance specific state capacities, contributing to improved general security and economic development. In terms of the model, these state capacities, advanced by small arms importation, raise the expected future utility after the consolidation of the government has taken place. The formal model thus provides a valuable framework to reconcile the counteracting risk-inducing and risk-reducing mechanisms of small arms imports.

In the following, the model setup and decisions in equilibrium are laid out. I then adapt the model to derive hypotheses on the effects of arms import on conflict risk.

Model setup

The model is a stochastic game where two players $i \in {G, R}$ , the government G and the rebel group R, compete for the distribution of the flow of income. The monetary utility of this income is normalized to 1 without loss of generality. However, future income and utility also depend on future state capacity c, affecting it as a multiplicative factor. Both the government and the rebel group discount future utility by the discount factor $β$ , leading to the total value of future utility $V (c) = \frac{c}{1 - β}$ over the infinite horizon.

The state is either consolidated or unconsolidated. In the unconsolidated state, the government wants to expedite consolidation, as consolidation parallels with a shift of power in favor of the government. However, the opposing rebel group can challenge the consolidation by fighting, thereby preventing the power shift and suspending consolidation for this period. Once consolidation has taken place, it will be sustained permanently.

In the unconsolidated state, the government G offers R a share of income $x_{t} \in [0, 1]$ . When R accepts, the state consolidates in a single period, and the opponent and the incumbent receive incomes $x_{t}$ and $1 - x_{t}$ , respectively. When R rejects, fighting occurs, which stalls the consolidation of the state. In fighting periods, R and G receive incomes $f_{R}$ and $f_{G}$ , respectively. However, fighting is inefficient and costly, $f_{R} + f_{G} < 1$ , which is lower than the income without conflict, which was normalized to 1.

The conflict either ends in a decisive outcome with probability d or in an indecisive stalemate with probability $1 - d$ . With a stalemate, consolidation of the state is prevented for this period, and the game simply turns to the next round without changes in the power distribution. With a decisive outcome, either the rebel group eliminates the government with probability p or the government eliminates the rebel group with probability $1 - p$ ; elimination means that the winning side will obtain the undivided total of future expected utility, while the other side gets nothing. Note that p is the conditional probability, with $d p$ being the probability of the rebels winning when a conflict occurs.

The probabilities $(d, p)$ determine the reservation values $F_{G}$ and $F_{R}$ in the unconsolidated state for the government and the rebels, respectively. These reservation values denote the discounted payoffs of fighting to the finish, i.e. fighting until a militarily decisive outcome is reached. $F_{G}$ and $F_{R}$ thus give the expected utilities when a conflict occurs. Analogously, in the consolidated state, $(\tilde{d}, \tilde{p})$ determine ${\tilde{F}}_{G}$ and ${\tilde{F}}_{R}$ . The rebel group's reservation value can be defined recursively as

F_{R} = d [f_{R} + β p V (c)] + (1 - d) [f_{R} + β F_{R}] .

(1)

The first term entails the expected payoff when a decisive outcome can be reached, where the rebels always receive their fighting income and—with probability p, if the incumbent government can be eliminated—the total discounted utility

V (c)

, starting in the next period. The second term gives the expected payoff without a decisive outcome, again comprising the fighting income and the discounted reservation value when fighting to the finish.

Consolidation in the model's terms happens with a shift of power in favor of the government, implying a lower reservation value for the rebel group, ${\tilde{F}}_{R} < F_{R}$ , owing to the reduced probability of winning a potential conflict.

Equilibrium decisions

In the consolidated state: We solve the game backwards, starting in the consolidated state. The rebels will accept the government's offer $\tilde{x}$ whenever $\tilde{x} V (c) \geq {\tilde{F}}_{R}$ . As fighting is costly, the government wants the rebel group to accept the offer, so it will always offer just enough to make R accept it: $\tilde{x} = {\tilde{F}}_{R} / V (c) = (1 - β) {\tilde{F}}_{R} / c$ . In equilibrium, there is thus never conflict in the consolidated state.

In the unconsolidated state: In the unconsolidated state, R will fight when its reservation value for fighting $F_{R}$ is greater than the future payoff $β {\tilde{F}}_{R}$ from fighting to the finish in the consolidated state, because consolidation always occurs when the rebel group decides not to fight. Again, the government wants to avoid fighting and thus can offer $\tilde{x}$ to make the rebels accept consolidation:

\tilde{x} = F_{R} - β {\tilde{F}}_{R}

(2)

However, when the shift of power is too large, i.e.

F_{R} - β {\tilde{F}}_{R} > 1

, the government is not able to fully compensate the rebel group because it cannot offer more than the utility of the total contemporary income, which we normalized to 1. As the government cannot commit to compensating its opponents in future periods, such a large power shift will result in conflict.

Effects of small arms’ and major weapons’ imports

To assess the impact of arms imports, we need to analyze how they affect the probabilities of winning and the associated reservation values. Yet in the consolidated state, any changes in $(\tilde{p}, \tilde{d})$ only influence the income share that the government has to offer the rebel group. However, there is no distribution of power possible that will lead to conflict. Therefore, arms imports have no impact on conflict risk in consolidated states.

Corollary 1: Arms imports do not affect conflict risk in consolidated states.

In the unconsolidated state, the expected payoff for the rebels in case of a power shift determines whether the rebel group can be compensated; when the reservation value after the power shift, i.e. in the consolidated state, is too small, conflict occurs. Arms imports, of both small arms and major weapons, induce an ongoing shift in the distribution of power, where we can interpret $F_{R}$ as the payoff before and without the power shift and ${\tilde{F}}_{R}$ after the power shift to $(\tilde{p}, \tilde{d})$ occurred:

{\tilde{F}}_{R} = \frac{f_{R} + β \tilde{d} \tilde{p} \tilde{V} (c)}{1 - β (1 - \tilde{d})}

(3)

As can be expected, lower probabilities of winning owing to increased arms imports lead to a lower expected payoff for the rebels,

\partial {\tilde{F}}_{R} / \partial \tilde{p} > 0

.⁴

We first turn to imports of major weapons, as their effect is unambiguous in this model. In addition to the power shift effect described above, their financial expenses also reduce the available future utility $\tilde{V} (c)$ derived from financial means and thus also the expected payoff ${\tilde{F}}_{R}$ . With a lower ${\tilde{F}}_{R}$ , the risk increases that the government cannot compensate the rebel group for the expected losses that arise owing to the power shift. Therefore, higher imports of MCW increase the risk of conflict by shifting the distribution of power and at the same time reducing the rebels’ expected utility when the state would consolidate unchallenged.

Hypothesis 1: Higher imports of MCW lead to higher conflict risk.

In comparison, small arms imports invoke different mechanisms as laid out in the preceding section: the financial burden owing to small arms acquisitions is relatively low, and they are essential tools for conflict-preventing state capacities, contributing to security and taxing capabilities, which are in turn necessary for economic welfare and development. Formally, small arms can thus be assumed to affect the expected utility $\tilde{V} (c)$ positively owing to higher state capacity c. A larger $\tilde{V} (c)$ in turn raises the reservation value ${\tilde{F}}_{R}$ , rendering conflict a relatively less valuable alternative for the rebel group. Thus, a larger $\tilde{V} (c)$ , raised by small arms imports, does not directly alleviate commitment problems but makes situations where the rebel group cannot be compensated less likely.

Combining both antagonistic mechanisms of power shifts and enhanced state capacity, our formal theory predicts small arms imports to have less of a risk-inducing effect than major weapons. If their impact via increased state capacities is large enough, the overall effect of small arms imports could even be to reduce the risk of a conflict onset.

Hypothesis 2: Higher imports of small arms increase conflict risk less than MCW imports and may even reduce it.

Testing the effect of small arms imports empirically

To evaluate the hypotheses empirically, I use time-series cross-sectional (TSCS) data on 146 countries from 1993 to 2014, covering the total time span for which reliable and comprehensive data on small arms transfers are available. The following section describes the research design and methods for analyzing the dataset as well as the operationalization and sources of the main and control variables.

Research design

Fixed effects and split-population models

TSCS data, as used in this article, usually allows unobserved unit heterogeneity to be controlled for, e.g. by applying fixed effects estimation. However, in case of the rare event of conflict onset, this advantage of TSCS cannot be exploited by standard models, as fixed effects lead to biased estimates: countries without any onset, i.e. no variation in the outcome variable, drop out of the sample, biasing the estimated coefficients (Cook et al., 2020). Aiming at using the additional information contained in the TSCS data structure, this article proposes making use of two methodological advancements relevant for the conflict onset literature: first, the penalized maximal likelihood-fixed effects (PML-FE) estimator suggested by Cook et al. (2020), and second, a split-population duration (SP-D) model (Beger et al., 2017).

The PML-FE estimator includes fixed effects $α_{i}$ for every country that experiences at least one onset over the observed period on the one hand and a common intercept $α$ for all countries that do not experience an onset on the other:

α_{i} = {\begin{matrix} α_{i} & if \sum_{t = 1}^{T} y_{i t} > 0 \\ α & if \sum_{t = 1}^{T} y_{i t} = 0 \end{matrix}

(4)

α

will tend to infinity in standard models, Cook et al. (2020) use penalization to shrink the parameter. Zorn (2005) proposed using Jeffreys’ invariant prior for penalizing estimates in the case of perfect separation, i.e. when the predictors perfectly explain the outcome, as is the case for fixed effects of countries that do not experience at least one onset (cf. Firth, 1993). Using Jeffreys’ prior for penalization introduces the assumption that each country would have an onset recorded if only the time series were long enough (Cook et al., 2020). As the penalizing term reaches its maximum with coefficients equal to zero, maximum likelihood estimation will shrink the parameters toward zero, allowing estimating fixed effects for countries without conflicts. Thus, PML-FE avoids restricting the sample to cases with at least one onset and provides more accurate estimates than standard unconditional fixed effects models (Cook et al., 2020).

In simulations, the PML-FE estimator also outperforms conditional fixed-effects models already for $T = 20$ , so the incidental-parameter bias problem (Lancaster, 2000; Neyman and Scott, 1948) should be of less concern, according to Cook et al. (2020). Nevertheless, the Online Appendix also provides the results of conditional fixed effects logit models for comparison, although these can only take countries with at least one onset during the observation period into account.

The second innovative design that I propose for dealing with the rare-events nature of conflict onsets is a split-population model. Ninety-eight countries in my sample from 1993 to 2014 do not experience any civil conflict. This corresponds to our theoretical expectations: in the formal model laid out in above, already consolidated states do not face any risk of conflict. Thus, we expect a set of countries with no risk of conflict, while other countries remain at risk. Of course, countries at risk may nevertheless not experience conflict over the sample period.

Beger et al. (2017) therefore propose a split-population duration model with the likelihood function as follows:

L = \prod_{i = 1}^{N} {π_{i} f (t_{i})}^{δ_{i}} \times {(1 - π_{i}) + π_{i} S (t_{i})}^{1 - δ_{i}}

(5)

The likelihood function consists of two parts: first, the failure rate

f (t_{i})

when the country has already experienced an onset (

δ_{i}

) and, second, the survival, i.e. duration, function

S (t_{i})

when the country's observation is right-censored (

1 - δ_{i}

). However, in the SP-D model, we distinguish subpopulations of countries being at risk of a conflict (

π_{i}

) or “immune” (

1 - π_{i}

), as Beger et al. call it, referring to the origin of the concept in medical statistics. The probability of “immunity” of a country at a specific point in time is determined empirically through the logistic link function

π_{i} = \frac{1}{1 + e^{- Z_{i} γ}},

(6)

where

Z_{i}

is the vector of country-specific covariates, which determine how likely it is that the conflict risk in a country-year is strictly positive, and

γ

is the respective coefficients vector. We thus estimate at first how likely it is that a country-year belongs to the “at risk” or “immune” subpopulation; then we assess which factors render conflict in countries at risk more likely by estimating the Weibull survival function

S (t_{i}) = e^{{(- e^{{X_{i} β}} t)}^{a}},

(7)

where

X_{i}

is the vector of covariates affecting the conflict hazard of countries at risk,

β

is its respective vector of coefficients and a is the shape parameter. Country-years estimated to have a low probability of being “at risk”, i.e. having a low

π_{i}

, receive only low weights when estimating the maximum likelihood function. Thus, the estimated effects appropriately account for the fact that the baseline risk of conflict is minimal for many countries.

Possibility of endogeneity

A potential challenge for the aim of this article—to establish causal mechanisms of small arms imports affecting conflict risk—is the possibly endogenous demand for arms: when a government perceives high odds of civil conflict in the near future, it is reasonable to expect measures to prepare for conflict, including the importation of fighting equipment. To rule out the reverse path of causality, I follow Pamp et al. (2018) to employ an instrumental variable approach. Since civil conflict onset is a dichotomous dependent variable, IV probit models are used. In Pamp et al. (2018), we propose splitting major weapons imports into two mutually exclusive subsets: imports of major weapons that are used in civil war, like tanks and helicopters, are discerned from major weapons that are not, employed in a typical civil war, e.g. ships and air-defense weapons. Then, the latter, i.e. arms not used in civil conflict, can be exploited as an instrument for the former, i.e. arms used in civil conflict.

For valid estimates, the instrumental variables have to be both relevant and exogenous. As relevance can be tested empirically, I find that a narrower set of weapons, i.e. surface-to-air missiles, has a closer association with civil-war relevant MCW for this article's shorter timeframe and is therefore used here. A positive relationship is expected because the general size of the military affects the acquisitions of all types of equipment, as they are complementary goods. As Pamp et al. (2018) argue, imports of weapons that cannot be employed in a civil war do not affect the risk of such conflicts. Importing decisions for these arms are unrelated to the potentially increased demand for arms necessary to prepare for an internal conflict. Weapons that cannot be used for providing internal security and preventing the formation of armed groups, as is certainly the case for surface-to-air missiles, have no impact on the specific state capacities determining the risk of conflict. Thus, they do not affect conflict risk through channels other than raising complementary imports of MCW and small arms, especially not after controlling for general state capacity and governance in the form of GDP and democracy scores (Fearon and Laitin, 2003). Therefore, imports of surface-to-air missiles provide an exogenous instrument. The Online Appendix presents further tests that indicate or would be able to falsify exogeneity.

Additionally, I use MCW imports of potentially hostile countries and exports of small arms as instruments. Hostile countries’ imports can affect imports of the country of interest owing to arms race dynamics (Pamp and Mehltretter, 2019) but should not affect civil conflict risk. Potential contagion effects are controlled for by including the number of close conflicts in the analyses.⁵ The relevance of small arms exports is due to their high correlation with imports, which may arise because of intra-industry trade (Thies and Peterson, 2015). At the same time, exports do not affect civil conflict risk in the exporting country—certainly not after controlling for potential indirect spillover effects. If anything, exports are expected to be substituted for by production for domestic supply in light of an internal conflict and therefore rising domestic demand. The first-stage results of the models presented in the empirical results section are available in the Online Appendix to provide evidence of the instrumental variables’ relevance. All models use robust, country-clustered standard errors. Note that PML-FE and SP-D models do not control for potential endogeneity of arms imports; results on arms imports coefficients might thus be positively biased.

Data and operationalization

Arms imports

To measure the build-up of specific military capabilities, I use data on imports of both small arms and major conventional weapons.⁶ Governmental arms imports operationalize power shifts as the theoretical mechanism in Powell (2012) directly. They are also better suited to capturing the financial flow for military purposes than measures of endowments. To capture longer-term dynamics and buildups of military equipment, the analyses use not only yearly imports, but also the average of imports of the preceding 5 years, as proposed by Pamp et al. (2018).

Small arms imports data are provided by NISAT. I exclude sporting guns and shotgun cartridges as well as primarily privately used categories like revolvers to provide a better proxy of small arms that are relevant to upholding public security or preventing an intrastate conflict.⁷ Small arms imports are measured in inflation-adjusted 2012 US-dollars and used as natural logarithm, log(SA imports),⁸ as well as in per capita terms to provide comparability with the previous repression literature (cf. Blanton, 1999; de Soysa et al., 2010). Data on imports of major conventional weapons are collected by the Stockholm International Peace Research Institute and measured in so-called “trend-indicator values”. These values are based on production costs and account for the depreciation of second-hand purchases. The variable is also used logged as log(MCW imports), which only includes major weapons types that can be used in civil conflict, following the distinction from Pamp et al. (2018) detailed above.

Note that small arms and major weapons imports are measured differently. Small arms transfers are evaluated strictly by the monetary reward for the seller, which can be affected by strategically adopted prices to support the buyer or to compensate for adversarial military gain. In contrast, major weapons imports data rather proxy the transferred military capability. Unfortunately, directly comparable measures of both types of arms transfers are not available.

Dependent variable: onset

onset is a binary variable defined as the outbreak of at least one intrastate conflict with at least 25 battle-related deaths in one calendar year, after at least 2 years of less than 25 fatalities of the respective conflict were recorded. Data on intrastate conflicts is provided by the Uppsala Conflict Data Program (Gleditsch et al., 2002; Pettersson et al., 2019). Following McGrath (2015), I code years with ongoing conflicts as missing but preserve years coded with 1 because of additional onsets of other conflicts.⁹

Control variables

All models include control variables to avoid omitted variable bias in the coefficients of interest. The control variables are lagged because a conflict onset might affect their values already in the same year. Following Cederman et al. (2013), I include the share of excluded population, which might indicate ethnic grievances that can increase conflict risk and repression (cf. Wimmer, 2013; Cederman et al., 2010, 2020; for the Ethnic Power Relations data additionally see Vogt et al., 2015). Optimally, one would include information on existing armed opposition groups, but such data unfortunately is not available for non-conflict years. However, as Jackson (2010) shows, most rebel groups start arming only with the beginning of the conflict. For onsets of recurring conflicts, the time since the last conflict spell (see below) might proxy for the existence of such groups. Missing data on existing armed groups should thus only lead to minor omitted variable biases.

As is standard in the literature (Collier and Hoeffler, 2004; Fearon and Laitin, 2003), logged versions of population and GDP per capita are included to operationalize the wealth and poverty level (Jakobsen et al., 2013) and proxy general state capacities (Fearon and Laitin, 2003). Additionally, GDP growth controls for potential tensions in dynamic economies or recessions (cf. Collier and Hoeffler, 2004). Data on population and GDP is provided by the Maddison Project Database (Bolt et al., 2018). Fearon and Laitin (2003) argue that political instability may provide opportunities for insurgents to exploit the weakness of the political system; the variable instability is coded as a change of 3 or more on the Polity IV scale over the preceding three years. Additionally, the Polity IV score and, to account for potentially more conflict-prone anocracies (cf. Fearon and Laitin, 2003), its square are included (Marshall et al., 2019).¹⁰

To capture potential contagion effects from intrastate conflicts in other countries in the region (cf. Buhaug and Gleditsch, 2008; Reid et al., 2020), I include the number of conflicts in countries in less than 3000 km distance. Interstate disputes might offer opportunities for the opposition to challenge the government. Therefore, I control for the intensity of interstate violence, using data from Major Episodes of Political Violence (Marshall, 2019).

Following the recommendations from Beck et al. (1998) and Carter and Signorino (2010) on time polynomials, the number of years since the last conflict end t, $t^{2}$ and $t^{3} / 1000$ control for temporal dependence. For the SP-D model, the immunity equation and the duration equation have to be specified by differing sets of variables to allow identification. Thus, I use only structural variables with low variance over time for the immunity equation and the complete set of variables, including the arms imports variables (and excluding excluded population, which has too low variance over time to allow convergence of the model), for the duration equation.

Empirical results

In the following section, the results of the empirical analyses of arms imports’ effects on conflict onset are presented. Standard probit as well as instrumental variable models are augmented by fixed effects and split-population models.

Probit and IV probit results

Table 1 presents probit and IV probit models with onset as the dependent variable. Odd columns are standard probit models and even ones the respective instrumental variable model. The model pairs differ in the specification of small arms imports: models (1) and (2) use the logged average of a country's small arms imports over the past 5 years; models (3) and (4) use the logged 5 year average of small arms imports per capita; and models (5) and (6) use the logged 1 year lagged imports of MCW and per-capita imports of small arms.

Table 1.

Probit and IV probit results.

	(1)	(2)	(3)	(4)	(5)	(6)
	Country SA	Country SA, instr.	SA per capita	SA per capita, instr.	SA per capita	SA per capita, instr.
log(MCW imports), 5 year average	0.053*	0.129*	0.055*	0.151*
	(0.026)	(0.061)	(0.026)	(0.071)
L.log(MCW imports)					0.025	0.147**
					(0.018)	(0.054)
log(SA imports), 5 year average	−0.017^†	−0.086*
	(0.010)	(0.043)
log(SA imports per capita), 5 year average			−0.036^†	−0.172^†
			(0.021)	(0.102)
L.log(SA imports per capita)					−0.014	−0.145*
					(0.016)	(0.073)
L.log(population)	0.160**	0.165*	0.140**	0.046	0.173**	0.015
	(0.057)	(0.078)	(0.054)	(0.099)	(0.050)	(0.094)
L.log(GDP per capita)	−0.155^†	0.011	−0.150^†	0.009	−0.147*	−0.041
	(0.084)	(0.169)	(0.084)	(0.202)	(0.074)	(0.176)
LD.log(GDP per capita)	0.962*	1.045**	0.950*	0.972**	0.945*	0.660^†
	(0.378)	(0.376)	(0.376)	(0.377)	(0.388)	(0.396)
Instability	−0.149	−0.256	−0.163	−0.262	−0.178	−0.070
	(0.429)	(0.363)	(0.433)	(0.374)	(0.457)	(0.379)
L.Polity IV	0.011	0.029^†	0.011	0.024	0.008	0.023^†
	(0.014)	(0.015)	(0.015)	(0.016)	(0.015)	(0.014)
L.Polity IV × L.Polity IV	−0.004	−0.006*	−0.004	−0.005*	−0.003	−0.004^†
	(0.002)	(0.003)	(0.002)	(0.003)	(0.002)	(0.002)
L.Excluded population	0.245	0.154	0.245	0.158	0.309	0.094
	(0.195)	(0.220)	(0.195)	(0.228)	(0.189)	(0.210)
L.No. of nearby conflicts	0.052**	0.037^†	0.053**	0.038^†	0.055**	0.036
	(0.016)	(0.021)	(0.015)	(0.022)	(0.016)	(0.022)
L.Interstate violence	0.110	0.045	0.109	0.040	0.110	0.023
	(0.202)	(0.204)	(0.203)	(0.209)	(0.199)	(0.233)
Years since last conflict	−0.128**	−0.102**	−0.128**	−0.101**	−0.135**	−0.095*
	(0.033)	(0.032)	(0.033)	(0.036)	(0.034)	(0.037)
Years since last conflict, squared	0.004**	0.003*	0.004**	0.003*	0.005**	0.003*
	(0.001)	(0.001)	(0.001)	(0.002)	(0.002)	(0.002)
Years since last conflict, cubed	−0.043*	−0.035*	−0.044*	−0.034^†	−0.046*	−0.032^†
	(0.018)	(0.017)	(0.018)	(0.018)	(0.018)	(0.018)
N	2806	2806	2806	2806	2806	2785
Pseudo R²	0.291		0.291		0.287

Notes: * (^†, **) indicates p < 0.05 (0.1, 0.01). Standard errors in parentheses. Dependent variable: intrastate conflict onset.

Odd models use probit and even models IV probit instrumenting both small arms and major weapons imports variables.

SA, Small arms; MCW, major conventional weapons.

Except for model (5), all specifications corroborate Pamp et al. (2018), finding a statistically significant effect of major weapons imports that increase conflict risk. All models exhibit negative coefficients for small arms imports. However, only in models (2) and (6) are small arms imports statistically significant. Notably, none of the specifications indicates that small arms imports would make conflict more likely—in contrast to imports of major weapons, which are consistently shown to increase the risk of intrastate conflicts. This highlights the necessity to differentiate between these specific types of weapons. Tables A1–A3 in the Online Appendix show additional results for all models with different subsets of covariates. The negative effect of small arms imports does not hinge on specific variables that the model is controlling for and is already present in a model without any controls.

The coefficient of $- 0.086$ ( $- 0.017$ ) in the IV (non-IV) model in Table 1 implies a 2.5 (0.3) percentage point decrease in conflict risk for a one standard deviation increase in logged imports. However, the risk decrease can be substantially larger in countries with higher baseline risk (see the Online Appendix for more details on marginal effects under meaningful scenarios).

Figure 1 illustrates how conflict risk is affected by increases in small arms and major weapons imports. Under a hypothetical increase of small arms imports by 25%, a substantial but not extreme rise, predicted probabilities naturally decrease, as the estimations exhibited a negative coefficient for small arms imports. The left panel of Figure 1 shows how these changes in the predicted probabilities are distributed. Most country-years would only see single-digit or near-zero reductions in conflict risk when small arms imports were 25% higher than those observed in reality. However, some countries like Turkey, Sudan, India and Pakistan were predicted to have—averaged over all years in the sample—a 9–12 percentage point lower conflict risk with such an increase in small arms imports.

Figure 1.

Changes in predicted probabilities for 25% increases in small arms and major conventional weapon (MCW) imports. Distributions of changes in predicted probabilities for country-years in sample when small arms imports (left panel) or MCW imports (right panel) were 25% higher than observed. Years without imports of small arms or MCW, respectively, were excluded.

When increasing major weapons by 25% compared with the observed values, predicted probabilities rise owing to the positive coefficient of MCW imports in our results. The right panel of Figure 1 shows the distribution of these changes in predicted conflict risk. Again, the increases are rather small in most cases, but some country-years see the conflict risk rise by more than 5 percentage points. With China, Turkey, India, Pakistan and Sudan, a similar set of countries would see the largest surge of risk with increases in MCW imports.¹¹

Returning to our theoretical framework, the evidence substantiates predictions based on Powell's model of power shifts: major weapons imports do not seem to deter potential challengers, but instead might be better interpreted as ongoing power shifts that provoke preemptive opposition attacks, as expected by Hypothesis 1. In line with Hypothesis 2, small arms are shown to affect conflict risk differently, with imports not increasing or even decreasing the probability of onset.¹²

The control variables mostly show the expected effects, except GDP growth, where high growth seems to be associated with higher conflict risk after controlling for GDP level.

Fixed effects model

Using the advantages of TSCS data, penalized maximum likelihood-fixed effects allow unobserved heterogeneity between countries to be controlled for. Table 2 gives the results of two specifications with country sum (model 1) and per capita (model 2) small arms imports. Overall, the results support the conclusions from the standard probit models. However, small arms show a more robust statistically significant effect, decreasing conflict risk as already suggested by the probit models. The coefficient of $- 0.070$ means that the conflict risk is 0.9 percentage points lower with a 1 standard deviation rise in small arms imports, giving an estimated effect size between the also non-instrumented standard probit and the IV probit models.¹³ Note that PML-FE does not instrument arms imports; possible endogeneity would bias results positively, i.e. the true effect of small arms imports might be more strongly negative than indicated by the PML-FE estimations.

Table 2.

Penalized maximal likelihood-fixed effects (PML-FE) and split-population duration (SP-D) results.

	PML-FE		SP-D
	Country SA	SA per capita	Immunity eq.	Duration eq.
	(1)	(2)	(3)
log(MCW imports), 5 year average	0.269**	0.279**		−0.105*
	(0.069)	(0.070)		(0.042)
log(SA imports), 5 year average	−0.070*			0.064**
	(0.029)			(0.021)
log(SA imports per capita), 5 year average		−0.147*
		(0.062)
L.log(population)	1.676**	1.565**	0.645**	−0.084
	(0.359)	(0.363)	(0.186)	(0.086)
L.log(GDP per capita)	−0.724*	−0.728*	−0.508^†	0.192
	(0.308)	(0.310)	(0.271)	(0.160)
LD.log(GDP per capita)	1.768^†	1.759^†		−1.859^†
	(0.966)	(0.963)		(1.011)
Instability	−0.215	−0.236		0.545
	(0.992)	(0.995)		(0.980)
L.Polity IV	−0.032	−0.035	−0.011	−0.021
	(0.032)	(0.032)	(0.045)	(0.025)
L.Polity IV × L.Polity IV	−0.011	−0.010	−0.036*	−0.007
	(0.007)	(0.007)	(0.014)	(0.006)
L.Excluded population	−0.200	−0.249	0.553
	(0.838)	(0.839)	(1.197)
L.No. of nearby conflicts	0.072	0.074		−0.085**
	(0.049)	(0.049)		(0.024)
L.Interstate violence	0.428	0.441		−1.151**
	(0.471)	(0.469)		(0.239)
Years since last conflict	−0.122*	−0.131*
	(0.062)	(0.063)
Years since last conflict, squared	0.006*	0.006*
	(0.003)	(0.003)
Years since last conflict, cubed	−0.077*	−0.082*
	(0.036)	(0.037)
Constant	−19.980**	−18.424**	−0.134	2.537
	(6.531)	(6.421)	(3.915)	(2.172)
log( $α$ )			−0.155^†
			(0.084)
Observations	2907	2907	3348
Akaike information criterion	700.273	700.550	854.945

Notes: * (^†, **) indicates p < 0.05 (0.1, 0.01). Standard errors in parentheses. Models (1) and (2) use penalized maximum likelihood-fixed effects estimation with intrastate conflict onset as dependent variable. Model 3 uses split-population duration estimation, with the immunity equation estimating whether a country-year is “at risk” and the duration equation estimating the risk for the observations at risk. Note that positive (negative) coefficients in the Weibull duration equation imply a risk-reducing (risk-increasing) effect of the variable.

SA, Small arms; MCW, major conventional weapons.

Split-population model

The split-population duration model employs two equations. The immunity equation determines the likelihood of a country-year being “immune” to conflict. For the subpopulation of country-years at risk, the duration equation determines how long peace lasts until a civil conflict onset, i.e. how large the risk is.

Model (3) of Table 2 presents results on an SP-D model using the country sum of small arms imports. The share of excluded population, level of GDP, population size and the Polity IV institutions indicator specify how likely a country in a given year is to be in the at-risk group. Arms imports only come into play when underlying tensions are present (cf. corollary 1) and are thus only included in the duration equation. The results of the duration equation strongly support the results from standard probit and PML-FE models. When sufficient risk conditions apply, imports of major weapons are associated with shorter durations until the next onset, i.e. higher onset risk. Small arms imports, in contrast, prolong the duration, i.e. decrease onset risk.¹⁴

Note that here in the SP-D model, risk of conflict does not heavily depend on peace duration, as indicated by the statistically only weakly significant $\log (α)$ . Including time polynomials in standard probit models might thus primarily control for at-risk status, as countries not at risk experience exceptionally long peace spells.

Conclusion

This article presents a theory of how small arms can affect the risk of intrastate conflict and formalizes these effects in a bargaining model. The empirical results underpin the derived hypotheses: imports of major weapons increase conflict risk, while imports of small arms have no significant effect or even reduce it. The role of arms imports in the emergence of civil conflict thus seems to be less one of deterrence, but rather primarily characterized by Powell's (2012, 2013) reasoning of power shifts. Imports of major weapons change the distribution of power in favor of the incumbent government, potentially to such an extent that it is impossible to compensate the opposition for its likely loss of power. However, small arms seem to play a different role. This article argues that small arms are also necessary tools for enhancing conflict-preventing state capacities to improve security, taxation and general economic development. Thereby, they might rather support state stability than contribute to state fragility.

In no way do the results imply that importing more small arms is unconditionally good advice to achieve development and consolidation of a state. Further research is warranted on the exact usage of small arms that can counter the conflict-inducing effects shown by major weapons imports. Presumably, it is not only the positive authority of the rule of law that prevents the emergence of conflict but also repressive, deterring authority impeding any organization of opposition. Previous results in the literature arguing that small arms imports can deteriorate respect for human rights and encourage repression (Brender and Pfaff, 2018; de Soysa et al., 2010) might indicate this.

Unfortunately, the state of data on small arms transfers is alarming: NISAT was deprived of funding in 2017, meaning that the most recent available data are from 2014 and will not be updated. Therefore, a new data collection effort is warranted to better understand the effects of the small arms trade on conflict. Findings in the literature, including this article, can only be the first indications that await confirmation in future research with more recent and reliable data.

Footnotes

Acknowledgements

I gratefully thank Paul W. Thurner and Oliver Pamp for their helpful comments and support, Nicholas Marsh for providing the small arms transfers data and three anonymous reviewers for their valuable and insightful suggestions.

Declaration of conflicting interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: I thankfully acknowledge financial support by the German Foundation for Peace Research (project “International Arms Trade and Intrastate Conflict”, grant no. FP 08/16-SP 08/12-2015) and a scholarship from the Friedrich Ebert Stiftung.

ORCID iD

Andreas Mehltretter

Supplemental material

for this article is available online.

Notes

References

Adekoya

Abdul Razak

(2017) The dynamic relationship between crime and economic growth in Nigeria. International Journal of Management and Economics 53(1): 47–64.

Angrist

Pischke

(2009) Mostly Harmless Econometrics: An Empiricist’s Companion. Princeton, NJ: Princeton University Press.

Atwood

(2005) Demanding Attention: Addressing the Dynamics of Small Arms Demand. Geneva: Small Arms Survey.

Ayuba

Okafor

(2015) The role of small arms and light weapons proliferation in African conflicts. African Journal of Political Science and International Relations 9(3): 76–85.

Azam

(2001) The redistributive state and conflicts in Africa. Journal of Peace Research 38(4): 429–444.

Bak

Chávez

Rider

(2020) Domestic political consequences of international rivalry. Journal of Conflict Resolution 64(4): 703–728.

Banerjee

Muggah

(2002) Small Arms and Human Insecurity. Colombo, Sri Lanka: Regional Centre for Strategic Studies.

Banini

Powell

Yekple

(2020) Peacekeeping as coup avoidance: Lessons from Ghana. African Security 13(3): 235–259.

Beck

Katz

Tucker

(1998) Taking time seriously: Time-series–cross-section analysis with a binary dependent variable. American Journal of Political Science 42(4): 1260–1288.

10.

Beger

Dorff

Ward

(2014) Ensemble forecasting of irregular leadership change. Research and Politics 1(3): 1–7.

11.

Beger

Hill

Metternich

, et al. (2017) Splitting it up: The spduration split-population duration regression package for time-varying covariates. The R Journal 9(2): 474–486.

12.

Berman

(2003) The provision of lethal military equipment: French, UK, and US peacekeeping policies towards Africa. Security Dialogue 34(2): 199–214.

13.

Besley

Mueller

(2021) Fiscal capacity and state fragility. In: Macroeconomic Policy in Fragile States. Oxford: Oxford University Press, pp. 235–296.

14.

Birikorang

(2007) Ghana’s regional security policy: Costs, benefits and consistency. Kofi Annan International Peacekeeping Training Centre Paper 20: 1–16.

15.

Blair

(2020) Peacekeeping, Policing, and the Rule of Law After Civil War. Cambridge: Cambridge University Press.

16.

Blanton

(1999) Instruments of security or tools of repression? Arms imports and human rights conditions in developing countries. Journal of Peace Research 36(2): 233–244.

17.

Bolt

Inklaar

de Jong

, et al. (2018) Rebasing ‘maddison’: New income comparisons and the shape of long-run economic development. GGDC Research Memorandum 174: 1–67.

18.

Bourne

(2007) Arming Conflict. The Proliferation of Small Arms. Basingstoke: Palgrave Macmillan.

19.

Boutwell

Klare

(1999) Light Weapons and Civil Conflict: Controlling the Tools of Violence. Lanham, MD: Rowman & Littlefield.

20.

Braithwaite

(2010) Resisting infection: How state capacity conditions conflict contagion. Journal of Peace Research 47(3): 311–319.

21.

Brender

Pfaff

(2018) Still tools of repression? Re-assessing the effect of arms imports on physical integrity rights. ILE Working Paper Series 22: 1–18.

22.

Brown

Grävingholt

(eds.) (2016) The Securitization of Foreign Aid. London: Palgrave Macmillan UK.

23.

Buhaug

Gleditsch

(2008) Contagion or confusion? Why conflicts cluster in space. International Studies Quarterly 52(2): 215–233.

24.

Cárdenas-Santamaría

(2007) Economic growth in Colombia: A reversal of ‘fortune’? Ensayos sobre Política Económica 53: 220–259.

25.

Carter

Signorino

(2010) Back to the future: Modeling time dependence in binary data. Political Analysis 18(3): 271–292.

26.

Cederman

Wimmer

Min

(2010) Why Do ethnic groups rebel? New data and analysis. World Politics 62(1): 87–119.

27.

Cederman

Gleditsch

Buhaug

(2013) Inequality, Grievances, and Civil War. New York: Cambridge University Press.

28.

Cederman

Hug

Schubiger

, et al. (2020) Civilian victimization and ethnic civil War. Journal of Conflict Resolution. 64(7-8): 1199–1225.

29.

Chassang

i Miquel

(2010) Conflict and deterrence under strategic risk. Quarterly Journal of Economics 125(4): 1821–1858.

30.

Chiang

(2021) Violence, non-violence and the conditional effect of repression on subsequent dissident mobilization. Conflict Management and Peace Science 38(6): 627–653.

31.

Chivers

(2011) Small arms, big problems: The fallout of the global gun trade. Foreign Affairs 90(1): 110–121.

32.

Coe

(2011) Costly peace: A new rationalist explanation for war. Working Paper, pp. 1–86.

33.

Collier

Hoeffler

(1998) On economic causes of civil war. Oxford Economic Papers 50(4): 563–573.

34.

Collier

Hoeffler

(2004) Greed and grievance in civil war. Oxford Economic Papers 56(4): 563–595.

35.

Cook

Hays

Franzese

(2020) Fixed effects in rare events data: A penalized maximum likelihood solution. Political Science Research and Methods 8(1): 92–105.

36.

Craft

(1999) Weapons for Peace, Weapons for War. The Effect of Arms Transfers on War Outbreak, Involvement, and Outcomes. New York: Routledge.

37.

Craft

Smaldone

(2002) The arms trade and the incidence of political violence in Sub-saharan Africa, 1967–97. Journal of Peace Research 39(6): 693–710.

38.

de Soysa

Jackson

Ormhaug

(2009) Does globalization profit the small arms bazaar? International Interactions 35(1): 86–105.

39.

de Soysa

Jackson

Ormhaug

(2010) Tools of the torturer? Small arms imports and repression of human rights, 1992–2001. The International Journal of Human Rights 14(3): 378–393.

40.

Durch

(2000) Constructing Regional Security. The Role of Arms Transfers, Arms Control, and Reassurance. New York: Palgrave.

41.

Estrada

MAR

Ndoma

(2014) How crime affects economic performance: The case of Guatemala. Journal of Policy Modeling 36(5): 867–882.

42.

Fafchamps

Minten

(2009) Insecurity and welfare: Evidence from county data. Journal of Development Studies 45(6): 831–863.

43.

Fearon

(1995) Rationalist explanations for war. International Organization 49(3): 379–414.

44.

Fearon

(2003) Ethnic and cultural diversity by country. Journal of Economic Growth 8(2): 195–222.

45.

Fearon

Laitin

(2003) Ethnicity, insurgency, and civil war. The American Political Science Review 97(1): 75–90.

46.

Firth

(1993) Bias reduction of maximum likelihood estimates. Biometrika 80(1): 27–38.

47.

Fjelde

De Soysa

(2009) Coercion, co-optation, or cooperation? State capacity and the risk of civil war, 1961–2004. Conflict Management and Peace Science 26(1): 5–25.

48.

Friedman

(2011) Building civilian police capacity: Post-conflict Liberia, 2003–2011. Princeton, New Jersey: Innovations for Successful Societies, Princeton University.

49.

Gates

Hegre

Jones

, et al. (2006) Institutional inconsistency and political instability: The duration of polities. American Journal of Political Science 50(4): 893–908.

50.

Gleditsch

Wallensteen

Eriksson

, et al. (2002) Armed conflict 1946–2001: A New dataset. Journal of Peace Research 39(5): 615–637.

51.

Gobinet

Davison

Alvazzi del Frate

(2011) Procurement and policy. Police use of emerging weapons technology. In: Small Arms Survey (ed.) States of Security. Cambridge: Cambridge University Press, pp. 69–99.

52.

Goulas

Zervoyianni

(2013) Economic growth and crime: Does uncertainty matter? Applied Economics Letters 20(5): 420–427.

53.

Goulas

Zervoyianni

(2015) Economic growth and crime: Is there an asymmetric relationship? Economic Modelling 49: 286–295.

54.

Greene

Penetrante

(2012) Arms, private militias and fragile state dynamics. In: Greene

Marsh

(eds.) Small Arms, Crime and Conflict. Global Governance and the Threat of Armed Violence. Abingdon: Routledge, pp. 138–159.

55.

Hanmer

Ozan Kalkan

(2013) Behind the curve: Clarifying the best approach to calculating predicted probabilities and marginal effects from limited dependent variable models. American Journal of Political Science 57(1): 263–277.

56.

Hendrix

(2010) Measuring state capacity: Theoretical and empirical implications for the study of civil conflict. Journal of Peace Research 47(3): 273–285.

57.

Hirshleifer

(2000) The macrotechnology of conflict. Journal of Conflict Resolution 44(6): 773–792.

58.

Hultquist

(2017) Is collective repression an effective counterinsurgency technique? Unpacking the cyclical relationship between repression and civil conflict. Conflict Management and Peace Science 34(5): 507–525.

59.

Huth

(1988) Extended deterrence and the outbreak of war. American Political Science Review 82(2): 423–443.

60.

Huth

Russett

(1988) Deterrence failure and crisis escalation. International Studies Quarterly 32(1): 29–45.

61.

Jackson

(2010) From under their noses: Rebel groups’ arms acquisition and the importance of leakages from state stockpiles. International Studies Perspectives 11(2): 131–147.

62.

Jackson

Morelli

(2009) Strategic militarization, deterrence and wars. Quarterly Journal of Political Science 4(4): 279–313.

63.

Jakobsen

de Soysa

Jakobsen

(2013) Why do poor countries suffer costly conflict? Unpacking per capita income and the onset of civil war. Conflict Management and Peace Science 30(2): 140–160.

64.

Jervis

(1989) Rational deterrence: Theory and evidence. World Politics 41(2): 183–207.

65.

Jia

Skaperdas

(2012) Technologies of conflict. In: Garfinkel

Skaperdas

(eds.) The Oxford Handbook of the Economics of Peace and Conflict. New York: Oxford University Press, pp. 449–472.

66.

Joyce

(2020) “A force for good”: Army-building after war in Liberia. The Journal of the Middle East and Africa 11(3): 273–293.

67.

Justino

Martorano

(2018) Welfare spending and political conflict in Latin America, 1970–2010. World Development 107: 98–110.

68.

Kalyvas

Balcells

(2010) International system and technologies of rebellion: How the end of the Cold War shaped internal conflict. American Political Science Review 104(3): 415–429.

69.

Karp

(2018) Estimating global law enforcement firearms numbers. Small Arms Survey—Briefing Papers (June).

70.

Kreutz

Marsh

(2012) Lethal instruments: Small arms and deaths in armed conflict. In: Greene

Marsh

(eds.) Small Arms, Crime and Conflict. Global Governance and the Threat of Armed Violence. Abingdon: Routledge, pp. 43–63.

71.

Lancaster

(2000) The incidental parameter problem since 1948. Journal of Econometrics 95(2): 391–413.

72.

Lebacher

Thurner

Kauermann

(2021) Censored regression for modelling small arms trade volumes and its ‘Forensic’ use for exploring unreported trades. Journal of the Royal Statistical Society: Series C (Applied Statistics) 70(4): 909–933.

73.

Magesan

Swee

(2018) Out of the ashes, into the fire: The consequences of U.S. weapons sales for political violence. European Economic Review 107: 133–156.

74.

Marshall

(2019) Major episodes of political violence (MEPV) and conflict regions, 1946–2018. http://www.systemicpeace.org/inscr/MEPVcodebook2018.pdf

75.

Marshall

Gurr

Jaggers

(2019) Polity IV project. Political regime characteristics and transitions, 1800–2018. Dataset Users’ Manual. http://www.systemicpeace.org/inscr/p4manualv2018.pdf

76.

McGrath

(2015) Estimating onsets of binary events in panel data. Political Analysis 23(4): 534–549.

77.

Mehler

(2012) Why security forces do not deliver security: Evidence from Liberia and the Central African Republic. Armed Forces and Society 38(1): 49–69.

78.

Mehrl

Thurner

(2020) Military technology and human loss in intrastate conflict: The conditional impact of arms imports. Journal of Conflict Resolution 64(4): 1172–1196.

79.

Neyman

Scott

(1948) Consistent estimates based on partially consistent observations. Econometrica 16(1): 1–32.

80.

Omelicheva

Carter

Campbell

(2017) Military aid and human rights: Assessing the impact of U.S. security assistance programs. Political Science Quarterly 132(1): 119–144.

81.

Oreta

(2009) Small arms proliferation: Impact to peace and security. Ateneo Law Journal 54: 292–297.

82.

Oriakhi

Osemwengie

(2012) The impact of national security on foreign direct investment in Nigeria: An empirical analysis. Journal of Economics and Sustainable Development 3(13): 88–96.

83.

Pamp, O & Mehltretter, A (2019) Know thy enemies, know thy friends: In search of elusive arms races. Working Paper.

84.

Pamp

Rudolph

Thurner

, et al. (2018) The build-up of coercive capacities: Arms imports and the outbreak of violent intrastate conflicts. Journal of Peace Research 55(4): 430–444.

85.

Pettersson

Högbladh

Öberg

(2019) Organized violence, 1989–2018 and peace agreements. Journal of Peace Research 56(4): 589–603.

86.

Powell

(1993) Guns, butter, and anarchy. The American Political Science Review 87(1): 115–132.

87.

Powell

(2006) War as a commitment problem. International Organization 60(1): 169–203.

88.

Powell

(2012) Persistent fighting and shifting power. American Journal of Political Science 56(3): 620–637.

89.

Powell

(2013) Monopolizing violence and consolidating power. Quarterly Journal of Economics 128(2): 807–859.

90.

Ray

Esteban

(2017) Conflict and development. Annual Review of Economics 9: 263–293.

91.

Reid

Myrick

Kadera

, et al. (2020) Conflict environments and civil war onset. Journal of Global Security Studies 6(2): 1–21.

92.

Risse

(2012) Governance in areas of limited statehood. In: The Oxford Handbook of Governance. Oxford: Oxford University Press, pp. 699–715.

93.

Rotberg

(2002) Failed states in a world of terror. Foreign Affairs 81(4): 127.

94.

Stohl

Myerscough

(2007) Sub-Saharan small arms: The damage continues. Current History 106(700): 227–232.

95.

Sullivan

Blanken

Rice

(2018) Arming the peace: Foreign security assistance and human rights conditions in post-conflict countries. Defence and Peace Economics 31(2): 177–200.

96.

Suzuki

(2007) Major arms imports and the onset of civil and ethnic wars in the postcolonial world, 1956–1998: A preliminary reassessment. Social Science Journal 44(1): 99–111.

97.

Taydas

Peksen

(2012) Can states buy peace? Social welfare spending and civil conflicts. Journal of Peace Research 49(2): 273–287.

98.

Taydas

Peksen

James

(2010) Why do civil wars occur? Understanding the importance of institutional quality. Civil Wars 12(3): 195–217.

99.

Thies

Peterson

(2015) Intra-industry Trade: Cooperation and Conflict in the Global Political Economy. Stanford, CA: Stanford University Press.

100.

Torres-Preciado

Polanco-Gaytán

Tinoco-Zermeño

(2017) Crime and regional economic growth in Mexico: A spatial perspective. Papers in Regional Science 96(3): 477–494.

101.

Tullock

(1980) Efficient rent seeking. In: Buchanan

Tollison

Tullock

(eds.) Toward a Theory of the Rent-seeking Society. College Station, TX: Texas A&M. University Press, pp. 97–112.

102.

Vogt

Bormann

Rüegger

, et al. (2015) Integrating data on ethnicity, geography, and conflict: The Ethnic Power Relations data set family. Journal of Conflict Resolution 59(7): 1327–1342.

103.

Vreeland

(2008) The effect of political regime on civil war unpacking anocracy. Journal of Conflict Resolution 52(3): 401–425.

104.

Walter

(2009) Bargaining failures and civil war. Annual Review of Political Science 12: 243–261.

105.

Washington

(2008) Elite police unit takes shape. UNMIL Focus 5(1): 4–5.

106.

Wezeman

(2003) Conflicts and Transfers of Small Arms. Stockholm: Stockholm International Peace Research Institute.

107.

Wille

Krause

(2005) Behind the numbers: Small arms and conflict deaths. In: Small Arms Survey 2005: Weapons at War. Oxford: Oxford University Press, pp. 229–265.

108.

Wimmer

(2013) Waves of War. Nationalism, State Formation, and Ethnic Exclusion in the Modern World. Cambridge, UK: Cambridge University Press.

109.

World Bank (2006) Crime, violence and economic development in Brazil: Elements for effective public policy. World Bank Report no. 36525.

110.

Zorn

(2005) A solution to separation in binary response models. Political Analysis 13(2): 157–170.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.80 MB

0.00 MB

0.32 MB