Rates of foreign direct investment (FDI) to Africa are increasing, yet little is known about how this will affect the political environment. One possibility, explored here, is that increasing levels of FDI within developing states will incentivize state conflict activity. Using an instrumental strategy, we show that in states with a low regard for civil liberties, or with unhealthy economies (i.e. states with a cash deficit), increased access to investment is associated with a higher number of conflict actions by the state. We argue that access to investment can push regimes into using violent strategies to secure their internal environment and to ensure their survival, specifically in their engaging in conflict against opposition and armed combatants. This underscores the need for extensive monitoring of state behavior following the receipt of investment, similar to the oversight of conditional aid.
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. 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) 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.
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). 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.
|Income streams||Mean||Range||Percentage of GDP (range)||States with low levels||States with high levels|
|Domestic tax||$4.90 billion/year||$0.01–93.85 billion/year||15.87% (1.16%–58.11%)||– DR-Congo
– Libya (in recent years)
|Extractive resource rents||$4.86 billion/year||$0.005–89 billion/year||15.50% (0.32%–77.29%)||– Djibouti
– Republic of Congo
|Foreign aid (conditional)||$0.644 billion/year||$0.004–11.43 billion/year||9.47% (0.01%–147.18%)||– Libya
|Foreign direct investment (stock)||$7.54 billion/year||$0.001–174.71 billion/year||35.42% (0.06%–811.20%)||– Burundi
– Republic of Congo & Mozambique (in recent years)
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).
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., 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).
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.
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. See Table 2 for a description of the variables and data sources.
|Variable||Description||Coverage of the data source||Mean (SD)||Min–Max||Data source|
|FDI Stock||FDI stock accumulated by a given state in a given year as proportion of GDP||1970–2014||35.42 (70.24)||0.06–811.20||UNCTAD, 2015|
|State proportion of armed events||Proportion of organized, armed conflict events involving state forces over total number of armed events that occurred in a given state in a given year||1997–2015||0.42 (0.34)||0–1||ACLED; Raleigh et al., 2010|
|Number of state events against militias (ln)||Natural log of the number of conflict events between state and militias||1997–2015||0.92 (1.25)||0–6.07||ACLED; Raleigh et al., 2010|
|Number of state events against political militias (ln)||Natural log of the number of conflict events between state and political militias||1997–2015||0.84 (1.19)||0–6.07||ACLED; Raleigh et al., 2010|
|Number of state events against civilian (ln)||Natural log of the number of conflict events between state and civilians||1997–2015||1.00 (1.15)||0–5.37||ACLED; Raleigh et al., 2010|
|GDP||GDP of a given state in a given year (in billion USD)||1960–2014||26.26 (58.41)||2.06–5218||World Bank, 2015|
|Population||Total population of a state in a given year (in thousands of people)||1960–2013||19,162 (25,485)||687.90–
|World Bank, 2015|
|Natural resource exports||Natural resource rents (i.e. sum of oil, natural gas, coal [hard and soft], mineral, and forest rents) of a state in a given year as a proportion of GDP||1960–2013||15.50 (15.68)||0.32–77.29||World Bank, 2015|
|Tax revenue||Tax revenue of a given state in a given year as a proportion of the GDP||1990–2013||15.87 (8.88)||1.16–58.11||World Bank, 2015|
|Consumer price index||Consumer Price Index (i.e. “the cost to the average consumer of acquiring a basket of goods and services) of a given state in a given year||1960–2014||83.72 (29.67)||11.98–298.51||World Bank, 2015|
|Control of corruption index||Captures “the extent to which publ#ic power is exercised for private gain, including both petty and grand forms of corruption, as well as ’capture’ of the state by elites and private interests”||1998, 2002–2012||2.09 (0.50)||0.97–3.95||Worldwide Governance Indicators (WGI), Kaufmann, Kraay & Mastruzzi, 2013|
|Traditional official finance||Total amount of non-Chinese official finance received by a state in a given year||1960–2012||9.47 (11.08)||0.01–147.18||AidData; Tierney et al., 2011|
|Presidential election dummy||Dummy variable capturing a presidential election year in a given state||1997–2014||0.16 (0.36)||0–1||V-Dem, Coppedge et al., 2015|
|Prior armed conflict||Number of organized, armed conflict events that occurred in a given state in the previous year||1997–2015||50.93 (111.06)||0–840||ACLED; Raleigh et al., 2010|
|Number of agents against state||Number of single agents that battled against the state in a given year||1997–2015||3.70 (4.27)||0–32||ACLED; Raleigh et al., 2010|
|Prior bordering conflict||Spatial lag variable capturing number of organized, armed conflict events that occurred in bordering states in the previous year||1997–2015||321.94 (427.6)||0–4334||ACLED; Raleigh et al., 2010|
|Civil liberty index||Index capturing the absence of physical violence committed by government agents and the absence of constraints of private liberties and political liberties by government||1900–2014 (with gaps)||0.61 (0.19)||0.10–0.94||Coppedge et al., 2015|
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. 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. 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. 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. 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, 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.
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). 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, tax revenue relative to GDP, total population, and the Consumer Price Index (CPI) 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). 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. A lagged measure of political conflict both within each state, as well as within each state’s African border states, are also added.
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).
Defining E(State Conflict Actionit/xit) 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; Xit−1 is a matrix of controls lagged at t−1, including FDI stock; β is the linked vector of coefficients; τt are the time-period dummies; γ is a vector of coefficients for Zi, 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).
Table 3 reports the findings from the fractional probit with Mundlak correction. Because of the non-linearity specification, coefficients are reported as marginal effects when all other variables are set to their mean values.
|Civil Lib<median||Civil Lib >median||Cash Balance <median||Cash Balance >median|
|FDI stock (lag)||0.001c||0.005a||0.000||0.005b||0.000|
|Tax revenue (lag)||0.000||0.002||0.003||0.004||−0.005|
|Natural resource rents (lag)||0.006a||0.003||0.012b||0.003||0.010b|
|Control of corruption (lag)||0.056||0.129||−0.081||−0.124||0.112|
|Armed conflict (lag)||0.000||0.000||0.000||0.000||−0.000c|
|Conflict in border countries (lag)||0.000||0.000||0.000||−0.000c||0.000|
|Civil liberties index (lag)||0.03||−0.078||0.602||0.551c||−0.069|
|Consumer price index (lag)||0.001||0.001||0.002||0.002||−0.001|
|Presidential election dummy (lag)||0.020||0.001||0.047||0.055||−0.007|
|Number of actors against State (lag)||0.003||−0.001||0.004||0.005||0.006|
Empirical specification is a fractional probit with Mundlak correction. All the specifications have the proportion of armed events that are state events as dependent variables. The control includes: FDI stock (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), year dummies (not reported). Mundlak’s controls include the country-average of all the above mentioned variables excluding Aid Level of significance are ap < 0.01; bp < 0.05; cp < 0.10; robust standard errors are clustered at country-level.
Model 1 reports the effect of FDI on state conflict action when using the entire sample. An increase in FDI received by a state is positively linked to a consequent raise in the proportion of armed events involving the state. Keeping all other variables at their mean values, an increase of 1% in FDI stock is correlated with an increase of about 0.1 percentage points in the proportion of armed events involving the state. This estimate implies a sizeable impact. For example, considering the FDI stock distribution, a state moving from the 25th percentile (10.24%) to the 75th percentile (38.28%, such as Mozambique from 2012 to 2013, or Mauritania from 2011 to 2012), would experience an increase in the rate of state conflict actions by 2.8 percentage points, keeping all other variables at their means.
As mentioned above, different conditions may incentivize a state to engage in conflict actions when observing an increase in the FDI hosted. Political and economic incentives have a role in determining the direction, size, and significance of state responses to increased investment. To account for this, the data are pooled in Models 2–5 according to whether the state has “liberty” and “healthy economy” dimensions above or below the median value of the dataset. The dimensions considered are: (1) the Civil Liberties Index (CLI), and (2) their Cash Balance (CB) (Coppedge et al., 2015; World Bank, 2015). To identify states that historically restrict liberties or often carry a cash deficit, we first average these dimensions at the state level and then define the two clusters. Since this classification is fixed in time, we run the specification on two equally sized clusters, instead of pooling the sample using the interaction terms.
Models 2–3 divide the sample above and below the median of the CLI index. The former reflects liberal freedom, where freedom is considered a right of individuals (Coppedge et al., 2015). The indicator ranges between 0 and 1, where 1 denotes a high regard for civil liberties, the absence of physical violence committed by government agents, and the absence of any constraints on private and political liberties by the government. We expect states below the median – such as Sudan, Libya, and Egypt – are more likely to engage in conflict as an instrument for gaining FDI. States such as Ghana, South Africa, and Benin display a higher level of respect for civil liberties (mean CLI > 0.86).
When considering the subset of countries with a low CLI, Model 2 finds evidence of a positive and extremely significant correlation between receiving FDI and subsequent rates of state securitization violence. In marginal terms, keeping all other variables at their mean values, a state receiving 1% more FDI at time t−1 reports an increase of 0.5 percentage points in its rate of state securitization violence at time t. Again, the magnitude of this result needs to be weighted by both the within-country and between-year variation of FDI receipt. More than 25% of country-year observations in the dataset experienced positive growth in FDI stock; for 5% of observations, the increase is larger than 13% of their own GDP. Keeping all other variables at their mean values, this would translate to an increase in the rate of securitization violence by at least 6.5 percentage points. This evidence supports the hypothesis that FDI stock in a state with low regard for civil liberties increases the rate of state securitization violence in that country.
Model 3 tests the effect of FDI stock on the rate of state conflict action on the set of countries with the highest regard for civil liberties. As expected, Model 3 does not report any significant impact of investment on state intervention, suggesting that not all states raise their rate of intervention following investment. In some environments, respect for civil liberties has allowed for the development of an alternative set of policy instruments to ensure stability and contemporaneously secure foreign investment. In many democratic states, respect for civil liberties allows for public grievances to be aired through nonviolent demonstrations. Cederman, Wimmer, and Min (2010), for example, discuss how the heightened grievances of mobilized groups, stemming from exclusion and powerlessness, can fuel individuals to take up arms against the state. Being able to interact with the state through nonviolent demonstrations can mean that individuals need not seek violent means to have their grievances heard; it may also mean that these grievances may be addressed by the state without violence being evoked.
Hypothesis 3 contends that states experiencing a strong fiscal deficit for a longer period of time as a result of structural determinants – such as a high level of tax evasion, or a low ability to enforce accountability – may be incentivized to use securitization violence as an effective policy instrument. In states experiencing a fiscal deficit, the choice between “buying consensus” or engage in conflict against opposition will more often result in the latter. To test this hypothesis, the sample is divided according to the median of cash surplus/deficit, expressed in terms of a ratio relative to each state’s GDP, as reported by the World Development Indicators (World Bank, 2015). Unfortunately, more than 17% of data points in the dataset have missing values. Imputing cash surplus/deficit for such a large subset of observations would result in misleading results. These missing data points fall out from the analysis, decreasing the total number of observations, as reported in Models 4 and 5. States with the lowest balance relative to GDP are Gambia (−6.87%), Egypt (−6.61%), Senegal (−5.50%), and Zimbabwe (−5.15). States with the largest cash surplus relative to their GDP are Niger (12.53%), Republic of Congo (8.07%), and Angola (6.44%).
Model 4 corroborates Hypothesis 3, confirming that economic conditions may incentivize state securitization violence rates. Findings suggest that states with restricted budgets intervene more often when receiving FDI to secure their political environment. In terms of magnitude, the correlation between FDI and securitization violence rates is similar to that obtained for the subsample of states with low respect for civil liberties. An increase of 1% in FDI stock is correlated with an increase in the rate of securitization violence by 0.5 percentage points. This value is similar to that of Model 2, suggesting that economic and political incentives have a similar weight in shaping state decisions to intervene in conflict. States with a safe public balance do not seem to employ securitization violence as a political instrument (see Model 5 in Table 3).
Estimates on the control variables report the expected sign and magnitude, especially when running our specification on the subsample with low civil and political liberties. GDP is positively linked to an increase in the rate of state involvement in political violence when respect for civil and political liberties by the state is low. This suggests that leaders of wealthier states, with minimal respect of civil liberties, will still use securitization violence as a tactic to secure their position. The level of natural resource rents is also positively associated with an increase in state securitization violence (Models 1, 3, and 5). Model 1 suggests that an increase in natural resource rents by 1% is followed by an increase in state intervention in conflict of 0.6 percentage points. Again, increasing the value of an income stream has a positive effect on state securitization violence. A similar result – even if smaller in magnitude and less consistent across the different models – is identifiable for aid, especially when the level of respect for civil liberties is low. Lastly, the BIC is reported in the second to last row of Table 3 and reports the goodness of fit of the model. As expected, when splitting the sample, BIC increases dramatically from its original value (−4075.43).
Table 4 reports the results from the OLS-FE specification, where the dependent variable is the natural log of the conflict events in which the state is involved. Also in this case, the level of conflict increases sharply when FDI increases. However, coherently with our hypotheses and with the above results, this occurs only for states with a low respect of civil liberties (Model 3) and with low cash balance (Model 4). In terms of magnitude, an increase of 1 percent in FDI is correlated with an increase in the number of events by about 0.8 percent for states with low respect for civil liberties. The magnitude of the coefficient is still higher for the countries with cash deficits, as in this case the predicted increase in conflict events is equal to 1.8 percent. Also in this case, estimates of the control variables are coherent across the specifications. For example, aid is negatively correlated with the level of conflict against the state, but only for countries with high respect for civil liberties and a moderate cash balance. Amongst others, it is worth noting that both GDP and CPI are positively correlated to the level of conflict in which the state is involved. The same results hold for the lag of the total number of conflicts within the state, which is a good proxy for predicting the level of state events the following year.
|Civil Lib<median||Civil Lib>median||Cash balance<median||Cash balance>median|
|FDI stock (lag)||0.002||0.008b||0.000||0.018b||−0.001|
|Tax revenue (lag)||0.000||−0.019c||0.013||−0.029||0.007|
|Natural resource rents (lag)||0.007||0.000||0.009||0.001||0.016|
|Control of corruption (lag)||0.043||0.058||−0.163||−0.629c||0.476|
|Armed conflict (lag)||0.002a||0.002a||0.001c||0.002b||0.000|
|Conflict in border countries (lag)||0.000||0.000||0.001c||−0.000c||0.000|
|Civil liberties index (lag)||1.517||2.358||−0.685||2.212||−0.893|
|Consumer price index (lag)||0.009a||0.006||0.009a||0.008c||0.006b|
|Presidential election dummy (lag)||0.048||−0.186||0.248b||0.190||0.029|
|Number of actors against state (lag)||0.066a||0.044||0.049||0.080b||0.057|
Empirical specification is a OLS with Fixed Effects. All the specifications have the natural log of the conflict events involving the state as dependent variable. To include the observations with value equal zero, it is added 1 to the entire series before taking the log. The control includes: FDI stock (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; robust standard errors are clustered at country-level.
The estimates document a positive correlation between FDI stock and state conflict action under the assumption that both the lagged value of FDI stock and the remaining controls are exogenous. However, an alternative explanation for the above findings is that the behavior of investors/firms is affected by the level of securitization violence at t−1, or by a third factor simultaneously influencing both dimensions (omitted variable bias). In this subsection, we address and solve any concerns of endogeneity by adopting an instrumental variable strategy. This approach requires a variable to be correlated with the endogenous independent variable (FDI stock), but that is uncorrelated to the dependent variable (state’s proportion of armed events).
Our instrumental strategy is based on the common cultural, institutional, and legal traits inherited through the colonization of states by Western powers. Most African states were part of colonial empires/protectorates between the 19th and 20th centuries; the exceptions are Ethiopia and Liberia. The experience of colonization deeply marked the institutional system, the legal framework, the language, and other social features of the colonized state. For example, former French colonies – such as Algeria, Burkina Faso, Cameroon, Gabon, Niger, Republic of Congo, Togo, and Tunisia – currently share a similar semi-presidential system. Many of these states recognize French as their second official language (e.g. Cameroon, Ivory Coast, Niger, Republic of Congo), and their populations are often fluent in French, with family ties to residents of France. A similar argument holds for former British, Italian, and Portuguese colonies, where external rule was extensively and forcibly imposed. These cultural, institutional, legal, and family ties favored the creation of special economic relationships and linkages both between the former colonies and their colonizers. For example, in studying FDI investments in Vietnam, Makino and Tsang (2011) show that historical ties explain FDI allocation decisions. Neumayer and Spess (2005) offer evidence that bilateral investment treaties increase the flow of FDI to developing countries, and their empirical strategy accounts for the colonial origin of recipients. Other evidence suggests that France and the UK have been more likely to stipulate a bilateral tax treaty, aimed at stimulating FDI, with their former colonies (Lejour, 2014). Being a former colony increases the probability of having a bilateral tax agreement with the former colonizer by 84 percent (Lejour, 2014). It is possible to argue, therefore, that colonial ties have established a direct channel of preferences in investment, along with a set of cultural and social characteristics, that together are influencing investment in states belonging to the same protectorate or former empire.
These characteristics not only create cultural linkages between former colonizers and colonies, but some theories indirectly suggest that the linkages may arise also between countries colonized by the same Western country through the firms’ preferences in investment (Torres, de Lemos & Fidas, 2012). For example, firms engage in foreign operations by first investing in more culturally and geographically similar countries (Forsgren, 2002). This may suggests that part of FDI stock shifts can be predicted by accounting for the shift in FDI stock in states with similar colonial ties. We build an instrumental variable that takes into account these colonial ties, yet is not correlated with the share of events involving state violence in a given state. We avoid computing the share of investment to a given state from the former colonizer state, as this would likely be correlated with the dependent variable. In contrast, we adopt as an instrument for country i: the average level of FDI stock in the states that shared the same colonial protectorate, as follows:
where the value of our instrument for country i, at time t, belonging to the former colonial protectorate c, is equal to the mean FDI stock of the other countries j under the same protectorate. Since we hypothesize that common cultural and institutional traits inherited by colonial ties are driving investors and firms to invest in a subset of countries, we expect that this instrument will be positively correlated with FDI stock in state i. The positive sign of the coefficient and its statistical significance are crucial for the validity of our instrumental strategy. We run some placebo tests to test the validity of our exclusion restriction, i.e. to be sure that the relation we are capturing goes from the ex-colonial partners to the state’s share of armed events, and not the reverse. Adopting this instrumental strategy excludes Ethiopian and Liberian observations from our sample. Table 4 reports results from a Two Stage Least Square (2SLS-OLS) strategy on the whole sample (Model 1), on the subsample of states reporting low civil liberties (Model 2), and with high cash-deficit (Model 3).
Since the instrumental variable controls for endogeneity, we control for the level of FDI stock and its instrument at time t, keeping lagged the remaining regressor variables. In terms of independent variables, we add the entire set of controls from the baseline specification. All the specifications include state-fixed effects, which allow us to take into account within panel correlation between observations and for unobserved characteristics. Errors are robust and clustered at the state level. Note that our proportional dependent variable brings advantages and drawbacks. Advantages include being able to account for state conflict action independently from the number of events occurring in a given country. This means that the number of events that occur will not affect the estimated relationships, and therefore we can exclude the notion that the latter are correlated to the coefficient of interest. Drawbacks include that the magnitude of the coefficients need to be accurately compared with those in Table 3, as the dependent variable remains bounded and the 2SLS-OLS is a linear estimator. Table 5 shows that the instrument is positively and significantly correlated to the endogenous FDI stock (Model 1). In terms of magnitude, an increase by 1 percent in the mean level of FDI stock in the former colonial partners is associated with an increase by 0.11 percent in FDI stock. This result is robust to the inclusion of a large set of controls. Since the empirical strategy assumes heteroscedasticity in the errors, we report results from a Kleibergen-Paap weakness statistic in the first-stage panel (upper part of Table 5); this test excludes any weakness in the instrument. The second-stage of the model generally confirms the estimates from the baseline strategy. An exogenous increase in FDI stock of 1% induces a significant increase in states’ conflict action of 0.5 percentage points (Model 1). This suggests that states respond to changes in FDI investment by engaging in a larger share of political violence to secure their internal space for this external wealth. Models 2 and 3 report again that the linkage is extremely pronounced for states with low respect of civil liberties (+0.6%) and with low cash balance (+0.8%).
|Whole sample||Low civil liberties||Low cash balance|
|First stage results|
|FDI other colonies||0.115a||0.131a||0.06b|
|Second stage results|
Estimates are obtained fitting a 2SLS-OLS with mean FDI of other colonial partners as instrumental variable in the first stage. Instrumented variable is FDI stock for all models. All the specifications have the proportion of armed events that are state events as dependent variable in the second stage. The controls include: FDI stock (endogenous instrumented), 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.
Table 6 reports the results obtained when using the natural log of the number of conflict events in which the state is involved as the dependent variable. The results remain consistent in this case. According to the first model, an increase in FDI stock leads to an increase by 3.2 percent in conflict events in which the state is involved. As Models 2 and 3 show, this effect is sharp and higher for countries with a low level of respect for civil liberties (+3.9%), or with cash deficits (+4.2%). Compared with the estimates from Table 4, these new results show that the effect is stronger than the one predicted, not accounting for time-varying heterogeneity.
|Whole sample||Low civil liberties||Low cash balance|
|First stage results|
|FDI other colonies||0.115a||0.131a||0.06b|
|Second stage results|
Estimates are obtained fitting a 2SLS-OLS with mean FDI of other colonial partners as instrumental variable in the first stage. Instrumented variable is FDI stock for all models. All the specifications have the natural log of the conflict events involving the state as dependent variable in the second stave. To include the observations with value equal zero, it is added 1 to the entire series before taking the log. The controls include: FDI stock (endogenous instrumented), 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.
Table 7 displays the second-stage of the 2SLS-OLS run on the set of countries with a high regard for civil liberties and high cash balance. As expected, these coefficients are not significant, indicating that FDI does not have a significant statistical impact on the decision to engage in conflict when this action is associated to high costs and low benefits.
|High civil liberties||High cash balance|
Estimates are obtained fitting a 2SLS-OLS with mean FDI of other African colonial partners as instrumental variable for FDI stock. All the specifications have the proportion of armed events that are state events as dependent variable in the second stage. 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.
Despite the above evidence pointing to an increase in state participation in conflict when receiving foreign aid, it still does not provide a clear indication whether these conflict events involve a process of state securitization or state repression. Note that securitization can derive from several conditions. For example, a state may be incentivized to attack opponents more often as a signal to investors that state actions are designed to secure their investments. However, reception of FDI stock could also increase the prize of engaging in conflict against the state. Militias may be more likely to engage in conflict against a regime in an effort to acquire more political and economic power. Repression, in contrast, often involves direct conflict actions against weak or unarmed agents, such as civilians. State forces may be incentivized to repress unarmed civilians as it not only signals strength and the willingness to protect its investment, but also aids in thwarting stronger opposition forces from developing.
Ideally, to test whether the increase in the share of conflict actions ought to be attributed to securitization or repression, data would need to indicate the “initiator” of the conflict action. Unfortunately, no reliable data exist with this information. However, an indication in this sense could be provided by running the IV-OLS specification with the number of state events against militias and civilians as dependent variables. A positive effect on the first of the two dependent variables (state conflict events against militias) would underscore that a process of securitization is occurring; a positive impact on the third variable (state events against civilians) would indicate that the state is engaging in active repression. As specified in the data section, these dimensions are log-linearized, and therefore the coefficients need to be interpreted as a percentage change. Table 8 reports the IV estimates obtained and suggests that FDI stock is positively and significantly associated with an increase in the number of state events against militias (Model 1), and specifically against political militias (Model 2). In terms of magnitude, an increase of 10% in FDI stock determines a positive shift of 37% in the number of conflict events against militias, which corresponds to an increase of approximately 7 events per year for a state in the top 10th percentile. As no significant effect is observed for state events against civilians (Model 3), we conclude that the increase in the share of state conflict actions is likely driven by a securitization strategy, determined either by states’ willingness to signal a “controlled environment,” or by militias’ interest in controlling more political and economic power. Other research has suggested that states have a number of strategies that they can use to repress civilians while alluding responsibility for those actions, including the use of pro-government militias to attack civilians (see Raleigh & Kishi, 2016). This may be a factor in the relative weakness of the repression hypothesis here.
|State vs. Militia||State vs. Political Militia||State vs. Civilian|
|First stage results|
|FDI other colonies||0.115a||0.115a||0.115a|
|Second stage results|
Estimates are obtained fitting a 2SLS-OLS with mean FDI of other colonial partners as instrumental variable in the first stage. Instrumented variable is FDI stock for all models. Column 1 has ln(number of state vs. militia events+1) as dependent variable. Column 2 and 3 have ln(number of state vs. political militia events+1) and ln(state vs. civilian events+1) as dependent variables. The controls include: FDI stock, 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.
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.
|Whole sample||Low civil liberties||Low cash balance|
|FDI of neighboring states||0.133||0.340||0.070|
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.
|Whole sample||Low civil liberties||Low cash balance|
|State Prop of armed events in former colonial partners (lag)||−0.15||−0.209||-0.03|
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 63rd 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 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.
To conclude, Appendix 1 reports results from several robustness tests that generally confirm the findings of the main empirical strategy.
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.
Funding and Disclosure Statement: this work has been supported by the European Research Council Young Investigator Grant – GEOPV. No potential conflicts of interest with respect to the research, authorship, and/or publication of this article are declared.
Achy, L. (2013). The Price of Stability in Algeria. Carnegie Papers Carnegie Middle East Center. Search in Google Scholar
Alexander, P. (2013). Marikana, turning point in South African history. Review of African Political Economy, 40(138), 605–619. Search in Google Scholar
Alfaro, L., Chanda, A., Kalemli-Ozcan, S. & Sayek, S. (2004). FDI and economic growth: the role of local financial markets. Journal of International Economics, 64(1), 89–112. Search in Google Scholar
Amnesty International. Amnesty’s International Report 2015/2016 2016. Search in Google Scholar
Bannon, I. & P. Collier (Eds.) (2003). Natural Resources and Violent Conflict: Options and Actions. Washington, DC: World Bank Publications. Search in Google Scholar
Bond, P. & Mottiar, S. (2013). Movements, protests and a massacre in South Africa. Journal of Contemporary African Studies, 31(2), 283–302. Search in Google Scholar
Bräutigam, D. (2008). Introduction: taxation and state-building in developing countries. Bräutigam, D, O.-H Fjeldstad & M Moore (Eds.), Taxation and State-Building in Developing Countries. UK, Cambridge: Cambridge University Press. Search in Google Scholar
Carter, B. (2016). Repression and foreign aid in autocracies: exploiting debt relief negotiations in post-cold war Africa AidData Working Paper 29. Search in Google Scholar
Cederman, L.-E., Wimmer, A. & Min, B. (2010). Why do ethnic groups rebel? new data and analysis. World Politics, 62(1), 87–119. Search in Google Scholar
Chen, W., Dollar, D. & Tang, H. (2015). Why is China Investing in Africa? Evidence from the Firm Level Brookings Working Paper Series. Search in Google Scholar
Choi, C. (2006). Does foreign direct investment affect domestic income inequality? Applied Economics Letters, 13(12), 811–814. Search in Google Scholar
Clapham, C. (1996). Africa and the International System. Cambridge, UK: Cambridge University Press. Search in Google Scholar
Collier, P. & Hoeffler, A. (2002). Aid, policy, and growth in post-conflict societies Policy Research Working Paper 2902. Search in Google Scholar
Collier, P. & Hoeffler, A. (2004). Greed and grievance in civil war. Oxford Economic Papers, 56(4), 563–595. Search in Google Scholar
Collier, P. & Hoeffler, A. (2005). Resource rents, governance, and conflict. Journal of Conflict Resolution, 49(4), 625–633. Search in Google Scholar
Coppedge, M., Gerring, J., Lindberg, S. I., Skaaning, S.-E., Teorell, J., Altman, D. & Zimmerman, B. V-Dem Dataset v5. Varieties of Democracy (V-Dem) Project 2015. Search in Google Scholar
Damania, R. & Bulte, E. (2003). Resources for sale: corruption, democracy and the natural resource curse CIES Discussion Paper 0320. Search in Google Scholar
Danneman, N. & Hencken Ritter, E. (2014). Contagious rebellion and preemptive repression. Journal of Conflict Resolution, 58(2), 254–279. Search in Google Scholar
Davenport, C. & Inman, M. (2012). The state of state repression research since the 1990s. Terrorism and Political Violence, 24(4), 619–634. Search in Google Scholar
De Waal, A. (2015). The ‘political marketplace’ & the roots of persistent conflict. Fletcher Security Review, 2(1), 1–6. Search in Google Scholar
Dowding, K. M. & Kimber, R. (1983). The meaning and use of ‘political stability’. European Journal of Political Research, 11, 229–243. Search in Google Scholar
Dube, O. & Vargas, J. F. (2013). Commodity price shocks and civil conflict: evidence from Colombia. Review of Economic Studies, 80, 1384–1421. Search in Google Scholar
Escribà-Folch, A. (2013). Repression, political threats, and survival autocracy. International Political Science Review, 34(5), 543–560. Search in Google Scholar
Fearon, J. D. & Laitin, D. D. (2003). Ethnicity, insurgency, and civil war. American Political Science Review, 97(1), 75–90. Search in Google Scholar
Findley, M., Powell, J., Strandow, D. & Tanner, J. (2011). The localized geography of foreign aid: a new dataset and application to violent armed conflict. World Development, 39(11), 1995–2011. Search in Google Scholar
Fjeldstad, O. H., Jensen, S. K. & Orre, A. (2012). Taking stock of the tax reform process in Angola, and why tax incenfives should be avoided. CMI Angola Country Brief, 2(1), 1–4. Search in Google Scholar
Forsgren, M. (2002). The concept of learning in the Uppsala internationalization process model: a critical review. International Business Review, 11(3), 257–277. Search in Google Scholar
Gleditsch, N. P., Wallensteen, P., Eriksson, M., Sollenberg, M. & Strand, H. (2002). Armed conflict 1946–2001: a new dataset. Journal of Peace Research, 39(5), 615–637. Search in Google Scholar
Grossman, H. I. (1992). Foreign aid and insurrection. Defence and Peace Economics, 3(4), 275–288. Search in Google Scholar
Gurr, T. R. (1993). Why minorities rebel: a global analysis of communal mobilization and conflict since 1945. International Political Science Review, 14(2), 161–201. Search in Google Scholar
Hamilton, A. (2009). Political oppression in Sub-Saharan Africa. Human Rights & Human Welfare, 1(1), 39–57. Search in Google Scholar
Hausman, J. A. & Leonard, G. K. (1997). Superstars in the National Basketball Association: economic value and policy. Journal of Labor Economics, 15(4), 586–624. Search in Google Scholar
Henderson, C. W. (1991). Conditions affecting the use of political repression. Journal of Conflict Resolution, 35(1), 120–142. Search in Google Scholar
Human Rights Watch. Ethiopia: No Let Up in Crackdown on Protests 2016 <https://www.hrw.org/news/2016/02/21/ethiopia-no-let-crackdown-protests>. 21 February 2016. Search in Google Scholar
Humphreys, M. & Weinstein, J. M. (2008). Who fights? the determinants of participation in civil war. American Journal of Political Science, 52(2), 436–55. Search in Google Scholar
Hunter, D. & Bridgeman, N. L. (2008). Narrowing the accountability gap: toward a new foreign investor accountability mechanism. Georgetown International Environmental Law Review, 20(2), 187–236. Search in Google Scholar
Hurwitz, L. (1973). Contemporary approaches to political stability. Comparative Politics, 5(3), 449–463. Search in Google Scholar
Justino, P. (2009). Poverty and violent conflict: a micro-level perspective on the causes and duration of warfare. Journal of Peace Research, 46(3), 315–333. Search in Google Scholar
Kaufmann, D., Kraay, A. & Mastruzzi, M. Worldwide governance indicators. Worldwide Governance Indicators Project 2013 <http://info.worldbank.org/governance/wgi/index.aspx#home>. Search in Google Scholar
Kishi, R. & Raleigh, C. (2015). Chinese Official Finance and Africa’s Pariah States ACLED Working Paper Series. Search in Google Scholar
Koenig-Archibugi, M. (2004). Transnational corporations and public accountability. Government and Opposition, 39(2), 234–259. Search in Google Scholar
Lejour, A. (2014). The foreign investment effects of tax treaties (No. 1403). Search in Google Scholar
Makino, S. & Tsang, E. W. (2011). Historical ties and foreign direct investment: An exploratory study. Journal of International Business Studies, 42(4), 545–557. Search in Google Scholar
Martin, M. & Warner, G. (2015). Crackdown Turns Deadly in Ethiopia as Government Turns against Protestsers. NPR, . 19 December 2015. <http://www.npr.org/2015/12/19/460414825/crackdown-turns-deadly-in-ethiopia-as-government-turns-against-protesters>. Search in Google Scholar
Moore, M. (2007). How Does Taxation Affect the Quality of Governance? IDS Working Paper 280. Search in Google Scholar
Mundlak, Y. (1978). On the pooling of time series and cross section data. Econometrica: Journal of the Econometric Society, 46(1), 69–85. Search in Google Scholar
Murshed, S. M. & Tadjoeddin, M. Z. (2009). Revisiting the greed and grievance explanations for violent internal conflict. Journal of International Development, 21(1), 87–111. Search in Google Scholar
Mwenda, A. M. & Tangri, R. (2005). Patronage politics, donor reforms, and regime consolidation in Uganda. African Affairs, 104(416), 449–467. Search in Google Scholar
Neumayer, E. & Spess, L. (2005). Do bilateral investment treaties increase foreign direct investment to developing countries? World Development, 33(10), 1567–1585. Search in Google Scholar
Papke, L. E. (2005). The effects of spending on test pass rates: evidence from Michigan. Journal of Public Economics, 89(5), 821–839. Search in Google Scholar
Papke, L. E. & Wooldridge, J. M. (2008). Panel data methods for fractional response variables with an application to test pass rates. Journal of Econometrics, 145(1), 121–133. Search in Google Scholar
Pinker, S. (2011). The Better Angels of Our Nature. New York, NY: Viking. Search in Google Scholar
Pinto, P. M. & Zhu, B. (2009). Fortune or Evil? The Effect of Inward Foreign Direct Investment on Corruption. International Studies Quarterly, 60(4), 693–705. Search in Google Scholar
Raleigh, C. (2016). Pragmatic and promiscuous explaining the rise of competitive political militias across Africa. Journal of Conflict Resolution, 60(2), 283–310. Search in Google Scholar
Raleigh, C. & Kishi, R. (2016). Leadership survival and hired guns: repression, counterbalancing, and pro-government militias ACLED Working Paper Series. Search in Google Scholar
Raleigh, C., Linke, A., Hegre, H. & Karlsen, J. (2010). Introducing ACLED: armed conflict location and event data. Journal of Peace Research, 47(5), 1–10. Search in Google Scholar
Roessler, P. (2011). The enemy within: personal rule, coups, and civil war in Africa. World Politics, 63(02), 300–346. Search in Google Scholar
Ross, M. (2006). A closer look at oil, diamonds, and civil war. Annual Review of Political Science, 9, 265–300. Search in Google Scholar
Rugman, A. M. & Doh, J. P. (2008). Multinationals and Development. New Haven, CT & London: Yale University Press. Search in Google Scholar
Sachs, J. D. & Warner, A. M. (1999). Natural resource abundance and economic growth. Development Discussion Paper number 517a. Cambridge: Harvard Institute for International Development. Search in Google Scholar
Sambanis, N. (2004). What is civil war? conceptual and empirical complexities of an operational definition. Journal of Conflict Resolution, 48(6), 814–858. Search in Google Scholar
Sarkees, M. R. & Wayman, F. (2010). Resort to war: 1816–2007. Washington, DC: CQ Press. Search in Google Scholar
Schoeman, M. (2015). 25th AU Summit: Political Risk and FDI in Africa. South African Institute of International Affairs: Trade Beat, . 25 June 2015. <http://www.thetradebeat.com/opinion-analysis/25th-au-summit-political-risk-and-fdi-in-africa>. Search in Google Scholar
Shepherd, B. (2010). Political stability: crucial for growth?. Onslow, S (Ed.), Resurgent Continent? Africa and the World. London UK: London School of Economics. Search in Google Scholar
Sorens, J. & Ruger, W. (2012). Does foreign investment really reduce repression? International Studies Quarterly, 56(2), 427–436. Search in Google Scholar
Stewart, F. (2008). Horizontal inequalities and conflict: an introduction and some hypotheses. Horizontal Inequalities and Conflict (pp. 3–24). United Kingdom, London: Palgrave Macmillan UK. Search in Google Scholar
Tierney, M. J., Nielson, D. L, Hawkins, D. G., Timmons Roberts, J., Findley, M. G., Powers, R. M. & Hicks, R. L. (2011). More dollars than sense: refining our knowledge of development finance using AidData. World Development, 39(11), 1891–906. Search in Google Scholar
Tomashevskiy, A. (2015). Political investments: the political effects of international capital flows.. Unpublished dissertation. Search in Google Scholar
Torres, M., de Lemos, F. F. & Fidas, P. (2012). Between neighbours’ heat and colonial ties’ comfort: does distance really matter. Business, Society and Politics: Multinationals in Emerging Markets (28, pp. 105). Search in Google Scholar
UNCTAD, G. (2006). World investment report. Search in Google Scholar
UNCTAD. World Investment Report 2015: Reforming the International Investment Governance, UNCTAD/WRI 2015. Search in Google Scholar
United Nations. Norms on the Responsibilities of Transnational Corporations and Other Business Enterprises with Regard to Human Rights. U.N. Doc. E/CN.4/Sub.2/2003/12/Rev.2 2003. Search in Google Scholar
Wagner, J. (2003). Unobserved firm heterogeneity and the size-exports nexus: evidence from German panel data. Review of World Economics, 139(1), 161–172. Search in Google Scholar
Williamson, C. R. (2010). Exploring the failure of foreign aid: The role of incentives and information. The review of Austrian economics, 23(1), 17–33. Search in Google Scholar
Wintrobe, R. (1987). The Political Economy of Dictatorship. New York: Cambridge University Press. Search in Google Scholar
World Bank. World Development Indicators. World Bank 2015 <http://data.worldbank.org/data-catalog/world-development-indicators>. Search in Google Scholar
The online version of this article offers supplementary material (DOI: https://doi.org/10.1515/peps-2017-0007).
©2017 Walter de Gruyter GmbH, Berlin/Boston