Foreign Investment and State Conflicts in Africa

1 Introduction

Recent global and regional analyses on conflict suggest that while war rates are decreasing, other forms and agents of political violence, including militias and state-supported gangs, are becoming more common in Africa (Pinker, 2011; Raleigh, 2016). There is limited research about what motivates these other forms of political violence, compared to the extensive evidence available on the relationship between civil wars and resources (Wintrobe, 1987; Bannon & Collier, 2003; Damania & Bulte, 2003). In addition, very little is known about how a state’s external income affects the rates and forms of state violence.

Under circumstances where a state relies on external income to ensure its survival, the regime may become more beholden to outside investors, donors, and markets than to citizens, which some argue has resulted in a “dependency culture” (Williamson, 2010). This extends to the provision of security and to whom it is provided, and whether the securitization of assets and resources is prioritized over citizen safety. We define securitization as state led violence strategies to limit opposition, volatility, and harm to areas, resources and priorities. We contend that increased investment is associated with increased securitization violence and repression. Repression encompasses actions and violent events in which the state initiates the attack against unarmed civilians.^[1] Securitization includes violent attacks with groups who oppose the regime and its actions. We argue that states with a low regard for civil liberties and budget deficits may rely on state securitization violence more so than with surplus, healthy economies.

State conflict action is closely tied to the political economy of the state (Henderson, 1991). It follows that obligations to external sources of income are especially binding in cases of investment, where and when the competition to attract investment is high. Regimes that rely on investment for leadership survival may face a tradeoff between the security of investors and investments, and the security of citizens. As one of the most important state provisions, security can support external relationships, projects, and extraction to the detriment of African citizens, development, and political participation.

In this article, we provide new insight on states’ strategic conflict actions and what incentivizes increases in repressive “security”. We bridge the academic literature on regime securitization with an often-overlooked source of external income – Foreign Direct Investment (FDI) stock. We use two indicators to measure securitization violence and we assume that a state increasing its conflict engagement is doing so in order to secure its internal space. We then address whether regimes are more likely to engage in securitization violence when their level of external-based resource dependence is higher. We suggest that states attracting FDI differ in their violence profiles from those with alternative sources of state income, specifically in ways that create the perception of control and security that benefit external agents rather than citizens. Existing investments and environments of potential investment are securitized to maintain external interest and sustain regime elites. We do not assume that investment directly creates state violence, nor do we suggest here that regimes may use violence to attract investors.

Africa is an appropriate choice for this study, as many countries are exploring ways to increase investment and simultaneously have high and temporally variable rates of state conflict action, securitization, and repression. African state leaders and regimes use a variety of strategies to ensure survival: leaders may minimize grievances and the number of challengers of the regime by fostering inclusion and increasing the provision of public goods. A “good performance” strategy has support: governments that raise domestic income from taxes, for example, are generally associated with public support and peace (Moore, 2007), although few developing states have this level of state capacity. Access to resources, such as natural resource rents (Wintrobe, 1987; Damania & Bulte, 2003)^[2] or unconditional official finance from external agents, can fund and facilitate such a strategy (Kishi & Raleigh, 2015) or strategies that aim to benefit a narrow, but crucial, component on the government. In this context, whether investment can be used as a source of income for regime power strategies remains underexplored.

To test this argument, we develop a panel of 45 African states, aligning data on state conflict actions with data at the state level extracted from several sources. We document a positive and significant relationship between FDI stock and state securitization violence. Our empirical strategy exploits two objective measures of state conflict action, i.e. the share of conflict events in which the state is involved, and the number of conflict events in which the state is involved, both measured outside of civil wars and large campaigns against a national armed opponent. This allows us to distinguish defense actions from securitization. The most conservative estimate shows that an increase by 1 percent point in FDI is associated with a 0.1 percentage point increase in state securitization violence. To put such magnitude into perspective, our model would predict that a standard deviation increase in FDI stock would be associated with a rise by 7 percentage points in conflict events involving the state, ceteris paribus. We present evidence that this relationship is robust by employing several econometric techniques: the first model is a fractional probit with Mundlak correction, accounting for a set of time-varying economical, institutional, and demographic determinants of state securitization. Given that the past level of conflict in bordering countries could potentially influence the likelihood of state conflict action, we include a set of time and spatially lagged conflict-related controls. To partially address omitted variable bias, our empirical strategy accounts for unobserved characteristics with the Mundlak correction. In a second model, we use an OLS model with fixed effects, accounting for the full set of unobserved characteristics influencing the number of conflict events in which the state is involved, and we show that the result remains consistent to this alternative specification. Finally, to strengthen the causal link between FDI and state securitization violence, we also adopt an instrumental variable approach based on the cultural and historical linkages that the colonial period created between the individual state and other states belonging to the same protectorate/colonial empire in Africa. Our instrumental variable is computed as the average level of FDI stock for the states that shared the same colonial power. This instrument confirms our results in terms of magnitude and sign and is robust to alternative geographical and conflict-diffusion interpretations, as well as to other placebo tests.

We expect, however, that state securitization violence has a different economic and political cost depending on the type of regime involved. This cost might be high for states where civil liberties are historically guaranteed and effectively promoted. In this context, state securitization violence might decrease consensus toward the state, increase grievances among elites and civil society, and lead to an escalation of conflict. In the worst-case scenario, violent escalation might result in the removal of the regime. In such scenarios, regardless of the need to secure FDI, it is very costly for a state to engage in securitization. In contrast, for states with historically restricted civil liberties of their citizens – and therefore where conflict actions and state violence are, in effect, policy instruments – an increase in FDI stock would raise the likelihood of state securitization violence as its cost is contained by previous practices. This study sheds light on this mechanism by investigating the presence of strong heterogeneity in the estimates. We show that the relationship between state securitization violence and FDI is driven by states with lower levels of civil liberties, where state actions are often, though not always, linked to repression of any form of opposition. To further test the strength of the above mechanism, we ask whether states with long periods of cash shortages would increase their rate of violence. Following a cost/benefit approach with fixed values, we expect that states with a high cash deficit will increase state securitization violence action marginally to secure FDI, as FDI would have the highest impact on available fungible resources. We find evidence supporting this hypothesis, as the relationship between state securitization violence and FDI stock is significant and positive for the subsample composed of states with a long-term low fiscal budget (i.e. cash deficit). This suggests that states might consider FDI stock as an alternative source of income that regimes aim to secure in periods of budget shortage. Finally, using the number of state conflict events against militias and civilians, we highlight that the increase in conflict actions is likely explained by an increase in the number of events involving militias. As no causal correlation can be found between FDI stock and state conflict events against civilians, a common proxy for state repression, we conclude that FDI stock is generating a securitization mechanism. As our data do not allow for precisely identifying the direction and the “initiator” of a conflict event, we argue that two mechanisms could be driving this result. Firstly, states receiving FDI are more likely to engage in violence against conflict agents or civilians to signal their ability to secure an environment. Secondly, FDI stock increases the prize of having more political power; therefore reception of FDI incentivizes political militias to engage in more conflict against the state.

This article proceeds as follows: in Section 2 we explore and offer descriptive evidence of how various external income sources’ rates and patterns across Africa could be linked to effects on violence. Section 3 introduces the conceptual framework surrounding the relationship between investment and state securitization violence, and produces three testable hypotheses on this framework. Section 4 presents data and describes the main dependent and explanatory variables. Section 5 highlights the research methodology, and Section 6 presents the baseline results and the instrumented results. Finally, Section 7 discusses the implications of our results and concludes the article.

2 State conflict action and income sources in Africa

During 2012 in Marikana, a small town of the North West Province of South Africa, a group of miners were striking for a pay increase in the British-owned Lonmin mine. The initially peaceful protest was similar to many occurring in Western countries, yet resulted in a massacre. South African police forces, working in tandem with Lonmin security and the National Union of Mineworkers, fired live ammunition at strikers (Alexander, 2013). Autopsies of the forty dead miners report most were shot in the back, suggesting they were fleeing rather than attacking, and the incident has since been cited as the most lethal use of force by South African security forces against civilians since the end of apartheid (Bond & Mottiar, 2013).

Another recent example comes from the Oromo protests in Ethiopia. A string of demonstrations began in late 2015 in response to the Ethiopian government’s “master plan” that seeks to expand the borders of the country’s capital into the surrounding rural areas. Land is nationally owned in Ethiopia, and the current regime’s master plan would result in the displacement of many of the region’s two million people, largely of the Oromo ethnic group (Amnesty International, 2016). Critics of the plan argue that the “land grab” is an effort to lease large parcels of land to foreign investors from China, India, and the Middle East (Martin & Warner, 2015). While the government has agreed to alter the “master plan,” the Oromo people have continued to demonstrate (Amnesty International, 2016). Recent reports suggest that Ethiopian security forces continue to violently suppress the largely peaceful protests in the Oromia and Amhara regions (Human Rights Watch, 2016).^[3] Further, existing militias in the region experienced a marked increase in attacks by the state during this period.

Both these episodes offer anecdotal evidence that the relationship between regimes and investors are given primacy over the provision of representation and security of citizens. Under circumstances in which a regime may rely on investment as an external resource to ensure its survival, the relationship between regime and investor appears to alter how states operate across their own territory, specifically in reference to how conflict by the state is incentivized as a strategy to deal with citizens and insure continued investment.

States receiving investment must ensure a politically stable environment for their investors, or risk losing future investment. States can ensure “political stability” in a number of ways, including the absence of political violence. However, stability ought to account for whether a state can cope with violent events, not simply the frequency of their occurrence (Dowding & Kimber, 1983). This is best illuminated by the fact that “identical frequencies of a set events in different countries [do not necessarily] denote the same levels of instability” (Dowding & Kimber, 1983). Governmental longevity or the absence of structural change may also denote stability and continuity, endurance, or persistence (see Hurwitz, 1973; Dowding & Kimber, 1983), though longevity is logically independent of stability. In these cases, stability means a “predictable political environment” (Shepherd, 2010). Arguably, a state can maintain this sort of stability through ensuring its power and authority, where conflict against opponents and repression of civilians can create an illusion of continued security and control. If acts minimize the likelihood that a government would be destabilized or overthrown by unconstitutional or violent means (see Chen, Dollar & Tang, 2015), then these practices would be “stabilizing”. If securitization allows regimes to limit and coordinate elite capture (see De Waal, 2015), and continue to rule effectively, it can thereby increase their stability and lower their unexpected political risk. In turn, this may result in the security of the FDI within state borders, illustrating how, through the use and means of securitization, the state can maintain political stability for investment. There is little evidence or pathways to suggest that investment directly creates state securitization violence, or that regimes use violence to attract investors. But different sources of state income can create political environments that are more or less repressive and violent; we find that states attracting FDI differ in their violence profiles specifically in ways that create the perception of biased control and security.

African states generate the majority of their income from external and extractive means. As it is possible to observe from summaries in Table 1, states exhibit a heterogeneous combination of natural resource rents, tax revenue, foreign aid, and investment. The volatility of these dimensions is different for each country, and their wide range suggests that each country’s economy is associated with distinct patterns of stability, volatility, and conflict risk.^[4]

Those dependent on natural resource rents (i.e. rentier states) “tend to have weaker state apparatus than one would expect given their level of income because the rulers have less need of a socially intrusive and elaborate bureaucratic system to raise revenues,” and so are less accountable to their citizens (Fearon & Laitin, 2003, p. 81). Lack of accountability can increase popular grievances and can motivate rebellion (Humphreys & Weinstein, 2008). Research findings consistently link conflict risk to a state’s economic health and income composition (Collier & Hoeffler, 2004; Murshed & Tadjoeddin, 2009), including dependences on aid (Grossman, 1992) and extractive resources (Collier & Hoeffler, 2005). The strongest results suggest that exclusive income flows and distribution generate both grievances (Gurr, 1993; Stewart, 2008) and opportunities (Justino, 2009) to challenge state’s power through conflict.

A natural resource extraction-based economy has long been associated with the “resource curse” – where states with abundant natural resources see less economic growth and worse development outcomes relative to other states (Sachs & Warner, 1999). Further, this abundance can result in increased conflict, as low economic growth can create an environment in which the opportunity cost for insurgency is lowered, making civil war more likely (Collier & Hoeffler, 2005).

In contrast, states dependent on domestic taxation have uniformly positive capability and legitimacy outcomes (Bräutigam, 2008) due to building and sustaining the power of states, as well as in shaping their ties to society. Taxation can fuel state building through both fostering representative democracy as well as strengthening state capacity, both of which help to “bolster the legitimacy of the state and enhance accountability between the state and its citizens” (Bräutigam, 2008, p. 1). Dependence on tax impacts how the state seeks to obtain, use, and retain power (Moore, 2007). In contrast, if a state does not depend on tax income because of access to alternative sources (e.g. natural resource rents or foreign aid), “state elites [will be] less responsive and accountable to citizens; and, depending on the sources of non-tax revenue, may have less incentive to build up the political and organizational capacities of the state” (Moore, 2007, p. 3; see also, Fearon & Laitin, 2003). This lack of accountability can impact popular grievances (Humphreys & Weinstein, 2008). In turn, effective tax systems are central to development as they contribute to escaping aid or single resource dependency as well as “reinforce government legitimacy through promoting accountability of the government to tax-paying citizens, effective state administration and good public financial management” (Fjeldstad, Jensen & Orre, 2012, p. 4).

A significant share of developing state finance is generated through foreign aid, which is strongly associated with conflicted and post-conflict states, and in turn often operates in “response” to violence (Collier & Hoeffler, 2002). Recent research on the traditional (i.e. conditional) aid and conflict nexus suggest that high levels of aid in recipient states increase the occurrence of civil war through its allocative and distributive effects (Grossman, 1992). In fragile states, aid can often be used as a “state rent,” and “fungible” finance can be reallocated based on the needs of the state (for example, see Findley et al., 2011). The hoarding of this aid benefits select areas and communities, and can be used to support regimes’ patronage networks (for example, see Mwenda & Tangri, 2005; Findley et al., 2011). This process can fuel grievances, which can manifest into conflict.

FDI has many benefits for developing countries. First, it brings skills and technology used in training a local labor force. Also, it can boost exports and improve the competitiveness and efficiency of local producers and it can facilitate the expansion of fundamentals to growth, such as public infrastructure (Rugman & Doh, 2008). In contrast to aid and resource exports, investment is often associated with stable environments. The promise of FDI is not directly tied to increasing state capacity, legitimacy, or the delivery and availability of public goods, but is associated with the “rising tide lifts all boats” theory of economic development. Yet, where investment has not been carefully managed, these benefits – including local labor training and jobs, public infrastructure, and boosting growth – have not materialized (Schoeman, 2015).

3 Conceptual framework: state intervention and investment

Previous research demonstrates how the use of conflict and repression by regimes can effectively curtail opposition from civil society, civilians, and external parties (for some examples, see Hamilton, 2009). States may use securitization techniques to enforce their hold on power and to reaffirm their authority, and in consequence, regimes remain in power (Escribà-Folch, 2013). African regimes often securitize to eradicate competition and subordinate civilian reform and revolt in order to ensure their survival (Clapham, 1996), and intimidating, targeting, and/or killing potential opposition is effective in quelling threats from organized groups (Danneman & Hencken Ritter, 2014). Access to resources, such as investment from external agents, can aid in funding and facilitating securitization by leaders. From this perspective, we consider securitization violence a viable strategy of regimes when they will benefit from it and when the opportunities arise, rather than a central tenet of transitioning or low-liberty regimes:

This leads to:

Hypothesis 1: Increased investment is associated with increased state securitization violence.

A further refinement of Hypothesis 1 suggests that the relationship between investment rates and state conflict action is conditional on the type of government, its practices, and institutions. In states where repression is a rarely used tactic, such as in many democracies (see Davenport & Inman, 2012) and where there is a high regard for civil and political liberties of citizens, increased access to investment might suddenly increase the state’s use of violence, as the political cost of conflict action might outpace its political benefit. However, in states where conflict actions are used readily, such as under dictatorships and autocratic regimes (Escribà-Folch, 2013; Raleigh & Kishi, 2016), with low regard for civil and political liberties, an increase in levels of repressive securitization could be inconsequential compared to the impact of increased investment. This suggests:

Hypothesis 2A: Increased investment increases state securitization violence in states with a low regard for civil liberties.

Hypothesis 2B: Increased investment is not associated with changes in securitization violence in states with a high regard for civil liberties.

When developing states receive investment, accountability mechanisms are essentially non-existent.^[5],[6] This benefits both the state and the investor: access to resources with minimal accountability means that the state might use incomes to secure its position of authority, including increasing the agents and levels of state conflict action to forward goals of ensuring its position and power (Kishi & Raleigh, 2015).^[7]

A further condition relates to the a priori economic health of the state and the cost-benefit mechanism. From a regime’s perspective, investment offers an opportunity to design economic growth, curtail the reportedly excessive accountability requirements from conditional aid and official finance, relieve dependency, and potentially provide a rich source of rents through which to distribute in exchange for loyalty and service (Tomashevskiy, 2015). Access to these resources with minimal accountability can be used to supply patronage support or for “purchasing” loyalty (Tomashevskiy, 2015). Access to an economic surplus can fund patronage and “buy off” competing elites, guarding against inclusionary types of violence; this can contribute to minimizing the threat of armed challengers and uprisings [for example, see Achy’s (2013) example of Algeria; Roessler, 2011]. Leaders in this position need not solely rely on the use of securitization as a strategy to ensure political stability within their borders. This might explain why states with healthier economies often exhibit a lower likelihood of engaging in securitizing violence, but those with unhealthy, and stagnating economies, are under pressure both from within and outside their regime (Davenport & Inman, 2012). In a context of budget shortage, the occurrence of violent events would increase the cost of eventually losing FDI, as its impact on a state’s financial resources will be higher. This, in turn, will lower the cost of intervention (or increase its benefit), especially if a state cannot access other resources for patronage. This suggests:

Hypothesis 3: Higher investment levels increase state securitization violence in states with budget deficits, compared to states with surplus or healthy economies.

4 Data

A panel dataset of 45 African states, where the unit of observation is country-year from 1998 to 2014, is created to test our hypotheses. A national level analysis can determine the differences in state responses to foreign investment stock available depending on the political and economic conditions of the state itself. It provides also a comparison based on similar and dissimilar experiences with investment over a time period, whereby the immediate and longer-term effects are measureable.^[8] See Table 2 for a description of the variables and data sources.

Information on conflict events is from ACLED – the Armed Conflict Location and Event Data project (Raleigh et al., 2010). ACLED collects information on a range of political violence, distinguished by event characteristics and type of group(s) participating, with geolocation information and date.^[9] These data are available from 1997 into real-time. Multiple dependent variables are employed across this study. First, the proportions of armed events that involve state forces are extracted by country-year. Information from 1997 to 2014 is integrated into this analysis to allow for lagged controls. This dimension provides a measure of the share of events that involves state forces, including events of violent securitization and of repression. This index is bounded between 0 and 1, where 1 indicates higher securitization of the conflict environment by the state. Used as the dependent variable in an appropriate empirical strategy, this measure can be used to underscore when the state is more prone to intervene in conflicts occurring within its boundaries. As a second dependent variable, the natural log of the total number of conflict events in which the state is involved is employed.^[10] However, not all types of state events can be considered active conflict actions against political opponents. For example, during civil war periods where a rebel group is seeking to overthrow a state, state actions can be seen as an attempt to defend itself from an imminent threat, rather than as a form of active state action to thwart the rise of future challengers against the state. To avoid our estimates erroneously capturing this violence, we exclude from analysis all country-year observations that experience episodes of civil war. To determine such years, we review the useable definitions in the recent literature. While the literature lacks an agreed upon definition of civil war [as emphasized by Sambanis (2004)], it agrees at least on three main pillars for its identification, and that is: (1) the presence of rebel groups seeking to replace the coercive power of the state, and (2) who represent a challenge to state sovereignty within its boundaries; and lastly, (3) the effective resistance by the state from the rebel group (which suggests that the government is actively involved in the conflict). Following these principles, within the literature, authors often establish thresholds of fatalities to define a given event as civil war – e.g. 25 battle-related fatalities, as seen in work citing the UCDP/PRIO Armed Conflict Dataset (Gleditsch et al. 2002) or 1000 battle-related fatalities, as seen in work citing the Correlates of War (COW) dataset (Sarkees & Wayman, 2010). We adopt a new approach defining an event as a “civil war event” when the two following conditions are met: (1) rebel groups fight against the state in at least two regions within the country – this provides a proxy for whether the rebel group embodies an effective geographical challenge; and (2) during the conflict, an effective exchange of territory occurs, independent from which actor may be acquiring or losing it. Country-year observations exhibiting a civil war are excluded from analysis. As a result, the final dataset consists of an unbalanced panel.

In the final part of the analysis, we use three other dependent variables, created by drawing on data from ACLED. These variables consist in the natural log of the number of state events against: (1) all militias (communal and political), (2) political militias, and (3) civilians.^[11] All militias include local armed agents that organize for the purpose of community threats, land protection and violent incursions, and political militias; separately, we isolate political militias for investigation. These are armed, organized gangs that emerge as the conflict arm of political elites to contest each other and the state directly (see Raleigh, 2016). We use the number of events both of the combined types of groups are involved in against the state to test whether FDI stock is linked to the securitization of internal territory. According to ACLED’s methodology, civilians never engage actively in conflict; rather, they are always victims of external attacks. We therefore use this variable to investigate whether the increase in the share of state conflict action represents an active form of state repression.

As states may increase rates of conflict actions during periods of high violence, a lagged count of conflict events occurring the previous year in each given state is included, as is a spatial lag variable measuring the number of conflict events occurring the previous year in all bordering African states. A variable capturing the lagged number of actors pursuing violence against the state is also added to the specification.

Investment is measured as FDI stock, which measures the total accumulated foreign-owned assets a state has at a given time, relative to the state’s GDP.^[12] Data on FDI stock come from the United Nations Conference on Trade and Development (UNCTAD, 2015) the UN agency responsible for dealing with international trade topics. UNCTAD delivers updated data for macroeconomic analyses, which have been employed extensively in several empirical studies and exercises testing the economic effect of FDI (Alfaro et al., 2004; Choi, 2006). FDI stock rather than inflow is used here to investigate the structural role of FDI in an economy, given that investment inflows may be high in a foreign-capital dependent state simply due to a temporarily sustained economic growth rate (e.g. China in the 1980s–1990s) or may be negative in a foreign-capital dependent state due to a short and sudden economic downturn, as in New Zealand during 2001 (see Sorens & Ruger, 2012). Using FDI inflow here would incorrectly track these temporary shifts in investments, which are not assumed to be correlated with changes in the strategic behavior of the state or other actors given that the study focuses on political violence as a strategic response/incentive in the accumulation of FDI – in other words, that FDI has a structural role on the economic policy of the state. While some authors raise doubts on the quality of the FDI data due to differences in their collection methodologies or because FDI is not easily measurable by developing country statistical offices,^[13] these issues are minimized by the standardization of the methodology for data collection (UNCTAD, 2006). Further, the less volatile nature of FDI stock relative to FDI inflow makes this estimate more precise and increases the reliability of UNCTAD data. Information from 1997–2014 is integrated into this analysis.^[14]

Several other controls are added to the empirical specification. To test whether states respond differently to alternative forms of finance, we add the quantity of traditional official finance received by the state relative to GDP, which comes from AidData’s “Research Release 2.1 (1997–2012)” (Tierney et al., 2011).^[15] We might expect that the state will decrease its level of intervention when the threat of external financial sanction, through a cut in foreign aid, is credible (Carter, 2016). Information on natural resources relative to GDP is added to the specification as several studies show how states with natural resources can experience higher levels of conflict/repression (Collier & Hoeffler, 2004; Ross, 2006). Information on GDP,^[16] tax revenue relative to GDP,^[17] total population, and the Consumer Price Index (CPI)^[18] come from the World Bank Development Indicators (World Bank, 2015). While wealthier states, or states with higher levels of tax revenues, should experience less state intervention and repression, commodity prices should induce an opportunity-cost effect and increase securitization violence (see Dube & Vargas, 2013). Corruption can have a dual effect. Paying off bribes can represent an alternative to securitization, as highly corrupted states can contain opposition grievances by buying their consensus. However, highly corrupted states could also find it more profitable to repress a discontent opposition rather than including it within the decision process (Pinto & Zhu, 2009). We therefore add to the specification a “control of corruption” index, measuring the extent to which public power is exercised for private gain; this index comes from the Worldwide Governance Indicators (WGI) (Kaufmann, Kraay & Mastruzzi, 2013).^[19] Both the Civil Liberties Index, which provides a further indication of the likelihood of securitization by the state, as well as an election year dummy accounting for whether it is a presidential election year within the state from the Varieties of Democracy (V-Dem) dataset (Coppedge et al., 2015), are also included.^[20] A lagged measure of political conflict both within each state, as well as within each state’s African border states, are also added.

5 Baseline empirical strategy

The hypotheses concern whether increased investment is associated with a higher level or rate of state participation in political violence, and whether this effect is heterogeneous depending on the economic and political conditions of the recipient state. To assess the impact of FDI on the number of conflict events, we employ an OLS strategy with country fixed effects. This methodology controls for time-invariant unobserved characteristics and provides consistent estimates when analyzing a panel of observations. The selection of the appropriate empirical model is more challenging for the proportion of armed events in which the state is involved. Since this second dependent variable is bound between 0 and 1, an empirical strategy should ideally account for both its bounded form, for the unobserved heterogeneity typical of cross-section time series data, where units are observed for a n > 1 amount of time; as well as for the endogeneity between the dependent variable and the regressor variables. In this case, fitting a linear model with fixed effects on a cross-sectional dimension is discouraged as it would ignore the bounded structure of the dependent variable; this could lead to predicted values of state conflict action larger than 1 (for example, see Papke, 2005). As a further disadvantage, adopting a linear specification when estimating the conditional mean of a proportion might also neglect nonlinearities in the propensity to engage in conflict. In the case of exogenous regressor variables, studies suggest how a fractional logit model might address the fractional responses in a cross-sectional framework (for example, see Hausman & Leonard, 1997; Wagner, 2003), while better accounting for the non-linearity of the response (given the logit function). However, a fractional logit would not be able to capture the unobserved heterogeneity linked to the panel structure of the dataset. For this reason, we adopt as the baseline model the fractional probit model with Mundlak correction and time-period dummies developed by Papke and Wooldridge (2008). The Mundlak correction consists of accounting for the unobserved heterogeneity by adding the mean values of the explanatory variables for each state that vary across years to the right hand side of the equation (Mundlak, 1978). The selected model is a general linear model with a probit link function, and is estimated through quasi-Maximum Likelihood Estimation (QMLE). The fractional probit function accounts for the nonlinear nature of the dependent variable. Across all specifications, errors are clustered at the country-level to account for heteroscedasticity. This methodology better captures unobserved heterogeneity without falling into the incidental parameter problem that could arise when fitting country dummies within a probit specification. All explanatory variables are lagged at (t−1) to minimize reverse causality investigating their lagged effect on securitization. Time-period dummies control for common factors affecting all countries in a given year (for example, the business cycle effect).^[21]

Defining E(State Conflict Action_it/x_it) as the expected value of the proportion of armed events that are state events in country i at year t, the empirical model takes the following form:

where ϕ(.) is the standard normal cumulative distribution function (CDF); α₀ is the intercept, i.e. the value of state intervention when all other variables assume a value of zero; X_it−1 is a matrix of controls lagged at t−1, including FDI stock; β is the linked vector of coefficients; τ_t are the time-period dummies; γ is a vector of coefficients for Z_i, which is a matrix including the same set of controls in X but averaged at the country-level, as required by the Mundlak correction. Bayesian Information Criterion (BIC) has allowed the selection of the specification reporting the best fit (i.e. the lowest BIC value).

6 Discussion and results

6.4 Sensitivity analysis and placebo tests

The first stage of Model 1–3 in Table 5 and Table 6 shows a positive association between the instrument capturing colonial ties and state level FDI stock. However, the effect may be impacted by the geographical location of the state. The geography of colonization in Africa suggests that entire clusters of neighboring states were under the same domination, suggesting that our instrument could be highly correlated with the mean level of FDI in neighboring states. Incorrectly imputing the effect of colonial ties to something that is merely geographical would underestimate any potential spillover effect of securitization violence in neighboring states. The positive coefficient of the first stage could be determined by a positive path of growth in certain regions, encompassing several neighboring states, more than by an investors’ or firms’ decision to invest in culturally similar states. To test that the captured statistical relationship passes through colonial ties and not regional ties, we implement a placebo test fitting the mean value of FDI stock in neighboring states as the instrument. A first indication that our instrument is truly capturing common colonial ties (and not regional ties) comes from a low value of the Pearson correlation coefficient (0.24) between this new geographical FDI measure and our original instrument. Table 9 reports the result of this test and shows that no statistical linkage exists between this placebo instrument and the endogenous FDI stock; this suggests that the relationship is passing through the colonial ties of the recipient state, and not the regional ties.

Table 9:

Placebo test using FDI of neighboring states as instrumental variable for FDI stock.

	Whole sample	Low civil liberties	Low cash balance
	(1)	(2)	(3)
	First stage
FDI of neighboring states	0.133	0.340	0.070
	(0.108)	(0.160)	(0.070)
Other controls	Y	Y	Y
State FE	Y	Y	Y
Kleibergen-Paap F-statistic	1.52	4.74	1.03
P>F	0.22	0.04	0.32
	Second stage
FDI stock	0.006	0.007	0.027
	(0.010)	(0.007)	(0.035)
Other controls	Y	Y	Y
State FE	Y	Y	Y

1. Estimates are obtained fitting a 2SLS-OLS with mean FDI of other colonial partners as instrumental variable. The dependent variables are FDI stock in the first stage, and state’s proportion of armed events in the second one. The controls include: FDI stock (instrumented endogenous), Aid (lag), Tax Revenue (lag), GDP (lag), Population (lag), Natural Resources (lag), Control of corruption (lag), Armed conflict (lag), Conflict in boarder’s countries (lag), Civil liberties index (lag), Consumer price index (lag), Presidential election dummy (lag), number of actors against the state (lag), state fixed effects. Level of significance are ^ap<0.01; ^bp<0.05; ^cp<0.10 and errors are robust and clustered at state level.

As a further test, we study whether there is a common path of state conflict action between all the countries that share the same colonial domination. If this endogeneity is the case, state conflict action would be fuelling FDI stock in former colonies, and a common level of state intervention would drive our result. For this reason, we compute the average proportion of armed events that are state events for all states sharing the same former colonial power. We then use this variable as explanatory variable in an OLS with state fixed effects where, as dependent variable, we keep state conflict action. This model will indicate whether it is possible to observe any direct link between state securitization violence in a given state and its former colonial partners. As displayed in Table 10, the empirical model does not find any significant correlation between state conflict action in a given state and its former colonial partners.

Table 10:

Placebo test on the linkages between state conflict actions in ex-colonial partners (OLS).

	Whole sample	Low civil liberties	Low cash balance
	(1)	(2)	(3)
State Prop of armed events in former colonial partners (lag)	−0.15	−0.209	-0.03
	(0.116)	(0.176)	(0.104)
Other controls	Y	Y	Y
State FE	Y	Y	Y
	671	343	276

1. Estimates are obtained fitting a OLS with state fixed effects. All the specifications have the proportion of armed events that are state events as dependent variables. The controls include: mean state proportion of armed events in other former colonial partners (lag), Aid (lag), Tax Revenue (lag), GDP (lag), Population (lag), Natural Resources (lag), Control of corruption (lag), Armed conflict (lag), Conflict in boarder’s countries (lag), Civil liberties index (lag), Consumer price index (lag), Presidential election dummy (lag), number of actors against the state (lag), state fixed effects. Level of significance are ^ap<0.01; ^bp<0.05; ^cp<0.10 and errors are robust and clustered at state level.

Evidence presented above provides an overview of how states may be incentivized to increase rates of conflict action upon reception of FDI, depending on political and economic conditions. We study the sensitivity of our results, investigating the presence of thresholds for civil or political liberties and cash surpluses or deficits using a rolling-window estimation ordered across two dimensions of interests: the Civil Liberties Index and cash balance. This method suggests coefficient non-linearity without imputing the median of the three dimensions above as the point where the dataset is split. As a result, it yields to a more specific indication of the exact point of the two dimensions at which the coefficient linked to FDI stock loses statistical significance. To use a rolling window, the subsamples need to consist of a discrete number of observations, which allows for the quasi-maximum likelihood estimation (QMLE) to converge. The result, however, is robust to different window specifications. Formally, the rolling-window is applied to subsamples obtained through windows of 200 observations and steps of 10 observations – about 13 countries when considering an average of 15 observations per country. Given the consistency between the baseline specification and the instrumental variable approach, we report results from the fractional probit with Mundlak correction.

Figure 1 displays findings obtained when using a rolling estimation strategy on a restricted sample-set based on the civil liberties dimension. When the subsample displays a low level of respect for civil liberties (i.e. a low CLI value), the coefficient capturing the impact of FDI stock on state securitization violence is both positive and significant. The result holds up to a value of 0.66 CLI, where we observe a dramatic drop both in its value and its statistical significance. The graph depicts a strong effect of FDI on rates of state securitization violence when respect for civil liberties by a state is low, and no correlation between rates of state conflict actions and respect for civil liberties once the CLI level surpasses a value of 0.66, corresponding to the 63^rd percentile in the data. This suggests that investment in places with a low regard for civil rights could result in higher rates of securitization violence by the state.

Figure 1:

The effect of FDI stock on state’s conflict action at different levels of respect for civil liberties.

Figure 2 also supports earlier findings regarding the state budget. States with a low fiscal balance (i.e. a cash deficit) are more likely to use securitization violence as a means of law enforcement. The rolling estimation strategy on a restricted sample-set based on the cash balance dimension suggests that when a state’s fiscal balance is below −1.53% of its own GDP, higher levels of FDI stock induce a strong and positive shift in the rate of state conflict actions. When conducting the same analysis on a subsample of countries with more sustainable fiscal balances, no evidence of an impact is found.

Figure 2:

The effect of FDI stock on state’s conflict action at different levels of cash balance.

To conclude, Appendix 1 reports results from several robustness tests that generally confirm the findings of the main empirical strategy.

7 Conclusion

States with low regard for civil and political liberties may use increased access to external resources to engage in conflict against their political opposition, and competitors. This is especially true when growing FDI stock is found in states with a low regard for civil liberties and low access to fungible resources – i.e. those experiencing a cash deficit. Both types of common contexts see an increased rate of securitization violence in response to increased investment. The proposed link between FDI and securitization violence is that regimes may use access to external financial resources to further their power and longevity through intervention in conflict and repression, where violence against both challengers and citizens remains an effective way for regimes to secure control. This result is similar to the use of securitization violence by states with access to other forms of unconditional and unaccountable resources (Kishi & Raleigh, 2015).

The relationship between investment and securitization violence is not direct. Investors are neither seeking out states that are pariahs, nor is there any evidence to suggest that regimes are looking for investors who may turn a blind eye to the distribution of rents from investment. Instead, this research suggests that regimes calculate the costs, risks, and benefits of intervening in line with their available funding streams, and how resilient and robust such funding stream is to violence. Aid flows are vulnerable to violence; states that tax as the main source of income are unlikely to consider the costs and risks of repression and violence as worth upsetting their citizenry; extractive economies are likely to have higher rates of political violence. We find that investment, specifically in states with poor records of political and economic stewardship, is resilient to political violence by regimes towards populations. Once a highly conflict-active state receives FDI, it increases its level of state securitization violence within its borders. This link between FDI stock and state intervention is positive, significant, and robust to specifications accounting for endogeneity and unobserved characteristics.

Research considers the political structures and institutions of states to be reliable indicators of the likelihood that regimes will use violence against their citizens. Yet, the income structures of states and their financial flows incentivize how they engage with citizens and opposition. Investment is vital for African states to develop, and growing external attention and investment in non-extractive industries is closely associated with positive economic and social changes. Nevertheless, results from this research suggest that states, and their representative regimes, require transparency and monitoring in their use of investment funds and actions to secure and continue such income flows. There exist obstacles to peaceful and prosperous progress for states and citizens; these obstacles can be mitigated with oversight and public transparency. Without conditions, oversight, and monitoring of state tactics and investor behavior, African states risk “investing in violence” rather than providing an environment where citizens and future generations may benefit from the possibilities and promise of investment.