Introduction
How does political violence shape patterns in foreign direct investment (FDI)? Violence is inherently destructive. It is antithetical to production, which both bolsters and is bolstered by the physical, social, and human capital that violence destroys. Organizations specializing in production would be better off in a world devoid of organizations that specialize in violence. The fact of the matter, however, is that the international system of states has been constructed by means of violence, and both the threat and use of violence continue to play a central role in our sociopolitical relations. Indeed, political order, as we know and understand it, is predicated on the organization of violence (Wagner Reference Wagner2007).
Although an inescapable feature of our system, violence varies in its form and function across space and time. How violence is organized and used is thought to have profound implications for societies’ economic development (Bates Reference Bates2010). This relationship is perhaps especially pronounced today, as capital becomes increasingly mobile. Multinational corporations (MNCs)—the large and rapidly growing share of firms with the means to internationalize production and distribution—enjoy the freedom to choose the environments in which they operate. They can locate capital in places where violence is organized in (relatively) less detrimental and disruptive ways.
The aim of this study is to evaluate and to understand how variations in the organization of violence influence MNCs’ investment activity. In doing so, I draw together two strands of the FDI literature, presenting a unifying theoretical framework that addresses gaps in each.
The first concerns the relationship between insurgency and FDI. There is an extensive body of research on the topic.Footnote 1 The common hypothesis is that MNCs are wary of investing in unstable and war-torn environments. The reasons are myriad and often undifferentiated in empirical models. Fighting can directly affect firms when their property or personnel are targeted (Powers and Choi Reference Powers and Choi2012; Klapper et al. Reference Klapper, Richmond and Tran2013; Dai et al. Reference Dai, Eden and Beamish2013; Camacho and Rodriguez Reference Camacho and Rodriguez2013). But even those not immediately exposed to fighting are still affected by the general destruction of infrastructure and capital, the disruptions in local supply and distribution networks, and the political uncertainty that accompanies violent contests for power (Nigh Reference Nigh1985; Braithwaite et al. Reference Braithwaite, Kucik and Maves2014; Barry Reference Barry2018; Doctor and Bagwell Reference Doctor and Bagwell2020; Joshi and Quinn Reference Joshi and Quinn2020).
The second strand concerns the relationship between repression and FDI. While there is a large literature on how MNCs respond to civil and political rights broadly construed, there has been relatively little empirical study of how they respond to governments’ abuse of physical integrity rights specifically.Footnote 2 Early scholarship posited that MNCs prefer partnering with states willing to physically suppress agitators (e.g., O’Donnell Reference O’Donnell1978; London and Ross Reference London and Ross1995; Tuman and Emmert Reference Tuman and Emmert2004). The mainstream hypothesis today is rather the opposite. The mechanisms motivating this hypothesis are imprecise and leave something to be desired. One line of reasoning, for instance, is that repression encourages political unrest, depresses human capital and productivity, and weakens foreign firms’ ability to integrate into the local market (Blanton and Blanton Reference Blanton and Blanton2007, Reference Blanton and Blanton2009). If we assume that these are unfavorable conditions for MNCs, it follows that human rights abuses discourage direct investment. Another line of reasoning is that repression deters foreign investors who wish to avoid “guilt by association” in a world increasingly committed to basic human rights and other universal values (Spar and La Mure Reference Spar and La Mure2003). While there is certainly nothing wrong with these arguments, the first only establishes a very indirect link between repression and FDI, while the second speaks to a tangential mechanism (i.e., perceptions of human rights disrepute) that can be independently measured and tested (Barry et al. Reference Barry, Clay and Flynn2013; Vadlamannati et al. Reference Vadlamannati, Janz and Isachsen Bernsten2018).
Correlational studies are decidedly not uniform in their findings for either hypothesis (e.g., Fatehi-Sadeh & Safizadeh Reference Fatehi-Sedeh and Safizadeh1989; Crenshaw Reference Crenshaw1991; Loree and Guisinger Reference Loree and Guisinger1995; Barry et al. Reference Barry, Clay and Flynn2013; Garriga Reference Garriga2016; Witte et al. Reference Witte, Burger, Ianchovichina and Pennings2017). But on balance, the evidence suggests that FDI activity decreases with both conflict and repression (Blanton and Blanton Reference Blanton and Blanton2012; Braithwaite et al. Reference Braithwaite, Kucik and Maves2014; Barry Reference Barry2018). The empirical focus, however, has been squarely on the general presence and intensity of these conditions. What’s missing is an appreciation for how the form and function of organized violence might also affect commercial interests.
I rectify this deficiency here by considering how the means of violence are distributed in the host country and by focusing on the incentives motivating different types of violent organizations. Specifically, I apply Olson’s (Reference Olson1993) concepts of “roving” and “stationary” banditry to three different types of violent actors—rebels, the state, and government agents. Combining this logic with the “political risk” theory of FDI yields some distinct hypotheses about how MNCs respond to different configurations of organized violence. I contend that MNCs’ interest in a host site decreases with the number of violent rebel groups and the frequency with which violence is perpetrated by rogue government agents; but that MNCs are more accepting of violence when it is administered willfully by the state. I test these hypotheses by analyzing patterns in FDI entry at the firm level. The data are consistent with theoretical expectations.Footnote 3
This study contributes to the literature in several ways. Narrowly, it adds another pathway by which rebellion affects FDI and also clarifies the theoretical and empirical links between repression and FDI. Broadly, it provides a framework for thinking about the relationship between violence and investment in the contemporary world—a world that is still politically structured by organizations specializing in violence, even while it is being economically “globalized” by those that specialize in production.Footnote 4
Theory
A basic framework
Organized violence serves two purposes. The first is political. An organization can strengthen its own position and influence by eliminating or otherwise incapacitating rivals. The terms of the arrangement by which political communities at any level operate are set by those with the most bargaining leverage; and bargaining leverage is ultimately a function of groups’ relative capacities to enforce or disrupt the status quo by means of violence (Wagner Reference Wagner2007). The second purpose is economic. In lieu of productive activity, violent organizations (or their membership) can threaten or use violence as a means of generating income. That is, they engage in economic “predation”—the forceful redistribution of wealth and value from those who create it to those who have the capacity to destroy it (Tilly Reference Tilly, Evans, Rueschemeyer and Skocpoll1985).
Because both activities motivate the organization of violence, groups that achieve the feat are necessarily inclined to perform them. What varies across groups, however, is the form these activities take in practice. In what is fundamentally a theory of the state, Olson provided a useful, intuitive framework for thinking about this variation (1993).
He posited two types of violent organizations. One is the “roving bandit.” This type of organization roams the countryside in search of settlements to plunder. Upon discovering one, its incentive is to engage in unrestrained theft—there is no guarantee, after all, about when or where it’ll have another opportunity to provide for itself. The second type is what he terms the “stationary bandit.” A clever predatory band will realize that, by settling down and laying exclusive claim to a settlement, it might generate a steady and reliable stream of income for itself. There is a consequent change in behavior that accompanies this transition. The “stationary bandit” is best served by engaging in a more restrained form of theft. An optimal extortion rate is one that is limited and predictable, instilling confidence in the residents that any value they produce above the extortion rate will remain theirs. Furthermore, the “stationary bandit” will protect the residents from unrestrained theft by other violent organizations. In this context, people will work harder and create more overall wealth, leaving everyone better off. For Olson, these are the conditions that ultimately give rise to political order, investment, innovation, and development.
Olson’s framework has been applied before to theories of FDI. It has featured most prominently in the literature on “political risk,” which is rooted in Vernon’s observation that foreign investors are particularly vulnerable to ex post adjustments in the extortion rate charged by host governments (1971). MNCs cannot know for sure which governments are sincere or insincere in their ex ante commitments.Footnote 5 This has implications for how FDI activity is distributed, especially among less developed countries. A popular argument is that political and legal constraints on democratically elected governments make their ex ante commitments more credible, resulting in a “democratic advantage” in FDI (Li and Resnick Reference Li and Resnick2003; Jensen Reference Jensen2003, Reference Jensen2008). In other words, the degree of certainty about the state’s de facto extortion rate varies in accordance with governing institutions (Olson Reference Olson1993); and this, in turn, shapes patterns in FDI.
While this literature makes good use of Olson’s framework, it tends to focus exclusively on the state as a source of predation. Not only does this neglect other sources (Jakobsen Reference Jakobsen2010), but it (implicitly) assumes that the market for organized violence is monopolized. We are thus left with a special case of Olson’s model—one in which the world is comprised of only “stationary bandits.” In reality, of course, there is considerable variation in how the means of organized violence are distributed. In some countries, the ability to threaten or use violence for political and economic purposes is highly centralized and belongs almost exclusively to the state; but in others, the ability to threaten or use violence is dispersed. Different configurations of organized violence lend themselves to different forms of predation and thus pose differing levels of risk to investors. I will apply Olson’s framework here to evaluate configurations involving not only the state but also rebel groups and rogue government agents.
Rebel groups and FDI
It has been commonly hypothesized that FDI activity decreases with political unrest and war (Nigh Reference Nigh1985; Li Reference Li, Fratianni and Rugman2006; Braithwaite et al. Reference Braithwaite, Kucik and Maves2014; Doctor and Bagwell Reference Doctor and Bagwell2020). This literature tends to emphasize the general destruction and disruption that accompanies fighting, particularly when it is intense and prolonged (Barry Reference Barry2018). While the intensity of fighting is certainly an important factor in the relationship between conflict and FDI, another factor that is often overlooked (neglected, even) is the incentive for economic predation that accompanies rebel mobilization. By definition, civil war involves the participation of rebel groups. Their presence in a country is thus necessarily subsumed in existing empirical studies. An application of Olson’s (Reference Olson1993) logic, however, implies that the formation of rebel organizations has an effect on investment that is independent from the intensity of fighting, and one that should therefore be isolated both theoretically and empirically.
A rebel group is a violent organization that operates outside of the existing political order. That such a group exists means that it controls a nontrivial share of the market for organized violence.Footnote 6 We can assume that it endeavors to expand its market share by violently eliminating rivals and opponents. But rebel groups also use violence for the purpose of securing wealth and resources. “Greed”-based explanations for civil war posit that revenue generation is necessary to sustain an insurgency (Collier and Hoeffler Reference Collier and Hoeffler2004). It may even be a primary motivation, if the wages of rebellion are greater than what can be earned from normal economic activity (Grossman Reference Grossman1999).
There are a variety of ways for rebels to generate revenue. Among these are plunder and extortion (Mehlum et al. Reference Mehlum, Moene and Torvik2002). Rebel organizations (or their membership) might occupy and appropriate a firm’s property, demand cash payments in exchange for “protecting” the firm’s facilities and personnel, or intercept the firm’s cargo while in transit. Whatever the method, these activities impose significant costs on the targeted firms (Klapper et al. Reference Klapper, Richmond and Tran2013; Camacho and Rodriguez Reference Camacho and Rodriguez2013), which is reason alone for MNCs to avoid countries where rebel groups operate. Indeed, foreign firms might be especially attractive targets, given that they are typically much wealthier than their local counterparts, and because they are “outsiders.”
Rebel groups are akin to Olson’s “roving bandits.” By definition, these organizations have monopolized neither violence nor theft; and the lives of their membership are constantly under threat. Securing resources is necessary for the continued survival of the organization, and members have little reason to not acquire as much wealth for themselves as possible when the opportunity presents itself. The incentive is to engage in unrestrained predation.Footnote 7 Where and when firms are targeted will also be unpredictable, making it difficult to implement effective mitigation strategies.
It is important to note that this is not an affliction of rebellion, per se, but rather one of the rebel group as a particular type of violent organization. This implies that the corresponding risk of predation increases with the number of rebel groups. When there are more rebel groups, there are more opportunities for firms to be indiscriminately and ruthlessly targeted. It also means there is more competition for resources and market share among rebel groups, which should only further weaken any incentive to practice restraint.
A simple hypothesis follows: Hosts’ FDI appeal declines with the number of rebel groups operating within their borders. Importantly, because the threat of unrestrained and unpredictable predation is a distinct mechanism, this relationship should be observed independently from the overall intensity of fighting.
The state, its agents, and FDI
While much of the literature on violence and FDI has focused on bottom-up political violence (i.e., dissent and civil war), a handful of studies have evaluated political violence from the top-down (i.e., repression). In a sharp break from traditional thinking, the mainstream hypothesis today is that FDI increases with government restraint (Blanton and Blanton Reference Blanton and Blanton2007). The strength of this link, however, remains suspect. It is not clear, either theoretically or empirically, if MNCs actually care about hosts’ repressive practices specifically, or if they are responding instead to a variety of generic sociopolitical concerns that are exacerbated by state violence (Blanton and Blanton Reference Blanton and Blanton2012; Barry et al. Reference Barry, Clay and Flynn2013; Garriga Reference Garriga2016). Combining Olson’s framework with insights from contemporary theoretical work on repression presents more precise and direct pathways connecting repression and FDI.
The modern state is complex and serves myriad functions. That said, it is still fundamentally a violent organization, and one that is unmatched in its ability to threaten or use violence as a means for realizing its interests. The degree to which violent repression is practiced varies considerably across space and time.Footnote 8 Theory has settled on two primary explanations for this variation. The first treats repression as a strategic choice. Repression has costs, but the state is willing to accept these costs when political challengers pose credible threats to the status quo (Poe Reference Poe, Carey and Poe2004; Ritter Reference Ritter2014). The second treats repression as an agency problem. Repression in this context isn’t central policy but rather reflects a government’s inability to rein in wayward offices and employees who use the tools of the state (i.e., the means of violence) to serve their own interests (Englehart Reference Englehart2009; Cingranelli et al. Reference Cingranelli, Fajardo-Heyward and Filippov2013).
The natural conclusion from this work is that not all instances of repression are the same. While an equally high level of repression might be observed among a pair of states, the contexts in which repression is being practiced could be very different. If we plug these different contexts into Olson’s framework, we see that violent repression has divergent implications for economic predation and thus for MNCs’ investment activity.
In the latter context, wherein repression reflects agency loss, the means of organized violence are dispersed. When police and military and governmental factions are able to act on their own interests with little fear of discipline or reprisal from above, they operate, in effect, as autonomous violent organizations. Like rebel groups, the incentive structure for such organizations is akin to that of the “roving bandit.” Not only will these groups use violence as a means of removing political opponents and challengers but also to enrich themselves. When there are opportunities to engage in economic predation—for instance, during periods when there is little direct monitoring or oversight by the central government—they will threaten or use violence (or, alternatively, threaten to withhold protection from theft and violence by other groups) to extort as much wealth as is feasible.Footnote 9 The targets and the timing will be relatively unpredictable, as will be the price of compliance. When the use of violence by state agents is unrestrained, so too will be their theft. In this context, the risk of predation increases with violent repression.
This does not mean, however, that the risk of unrestrained predation always increases with repression. In the other context, wherein repression is employed strategically and deliberately by the state, the means of organized violence remain centralized. Like rebels and rogue agents, the state is a violent organization that uses violent repression to incapacitate rivals and opponents. Unlike rebels and rogue agents, however, the state has much less incentive to engage in unrestrained predation. Per Olson’s framework, the state is a “stationary bandit.” It relies on a relatively limited and formalized system of extortion that is (approximately) optimized.
The optimized extortion rate varies across states and is determined by governing institutions and other factors. Indeed, as noted earlier, this has been studied extensively in the FDI literature concerning regime type and “political risk” (Li and Resnick Reference Li and Resnick2003; Jensen Reference Jensen2006). But even in countries where the de facto extortion rate is higher or less certain, the risk that the state threatens or uses violent repression for the purpose of unrestrained, unregulated, indiscriminate theft is relatively low. Despite its prominence in the “political risk” literature, outright nationalization/expropriation of multinationals’ property is rare (Kobrin Reference Kobrin1984).Footnote 10 Instead, economic predation by the state is typically restrained and systematized.
China exemplifies this distinction. A combination of autocratic governing institutions and the sheer force of its economic “gravity” results in an optimal extortion rate that is high relative to other states. This is readily observed in how China treats its foreign guests. It routinely demands technology sharing and local sourcing. The fact that these practices are routine, however, speaks to their systematic nature. MNCs know what the costs of doing business in China are beforehand. A few firms are perhaps deterred by this; but many others accept the costs and simply factor them into their long-term business projections.Footnote 11
The systematic nature of economic predation by the state doesn’t change when it deploys violence against its subjects. On the contrary, when repression is a strategic choice, its purpose is to preserve the existing political order—the very system in which the state’s economic practices are regulated and routinized. In this context, the risk of unrestrained predation does not increase with repression.
There are other reasons why state repression might coincide with a generally less-appealing business environment. These indirect or tangential mechanisms are already explored in the existing literature (Blanton and Blanton 2007, Reference Blanton and Blanton2012; Barry et al. Reference Barry, Clay and Flynn2013). But the logic proffered here fails to identify any increased risks to firms’ property and welfare resulting from willful repression by the state. It therefore offers no theoretical justification for the hypothesis that MNCs are deterred by such violence. What it does suggest, however, is that unwilful repression—the type practiced by government agents pursuing their own interests—does pose heightened risks to firms’ economic interests. The implied hypothesis is that hosts’ FDI appeal declines with the use of repression by rogue agents, but not with the use of repression by the state.
Data & method
To test these hypotheses, I analyze patterns in MNCs’ annual FDI activity between 1994 and 2018. The sample is comprised of nearly ten thousand major multinationals, paired with all host countries from the developing world that are recognized in the COW State System Membership list (2017) and for which data are available.Footnote 12 The unit of analysis is thus the company-country-year. Enterprise-level data were acquired from the LexisNexis Corporate Affiliations historical database and then aggregated at the level of “ultimate-parent” to create an MNC-level dataset recording the total number of enterprises firm i operates in country j at time t.Footnote 13 More than one hundred different “home” countries are represented in the sample.Footnote 14 Most industries are also represented, as measured at the 4-digit level of precision in the SIC schema. The median number of industries attached to each MNC during a given year is 7 (mean of 14); the median number of enterprises owned is 18 (mean of 40); and the median number of countries partnered with is eight (mean of 13).
The outcome variable employed in the analysis is binary and records entry into the host market.Footnote 15 The sample is thus limited to company–country pairs for which there is no existing FDI relationship at time t-1. Entry is observed when an FDI relationship is actualized—that is, once the number of subsidiaries operated by firm i in country j changes from zero to one or more. I also limited the estimation sample to years in which firm i is observed as entering into at least one new market in the relevant population of countries. This is to weed out superfluous observations for which entry is not even possible because the MNC in question is not actively seeking new FDI partnerships in the “developing” world. I also estimate the models of FDI exit, but these are relegated to the Appendix. The theory is attuned to the host-selection stage, when the parent MNC enjoys freedom of choice and selects from among several plausible partners based on projected risks and rewards. Theoretical expectations are less clear post-entry, when the decision-making context shifts from one of generalized cross-host comparison to site-specific management and enterprise performance (Boddewyn Reference Boddewyn1983; Tan and Sousa Reference Tan and Sousa2019).Footnote 16 I thus treat exit as an empirical curiosity—one that is secondary to the primary focus of this study, but that is still interesting and relevant insofar as the statistical results have implications for future theoretical development and research.
The argument is that hosts’ appeal to foreign investors is influenced not merely by violence but by the configuration of organized violence. I identified three types of violent actors that factor into this configuration. Rebel groups comprise one type, rogue government agents a second type, and the state itself a third type. Testing the hypotheses thus requires measuring the presence and prevalence of violence by each type in host country j during the given year. The first is straightforward. Using data from the UCDP Armed Conflict Database (Harbom et al. Reference Harbom, Melander and Wallensteen2008), I generated a variable counting the number of distinct rebel groups mobilized and engaged in armed insurgency against the host state. Directly measuring the other two is more difficult. While there are a couple of widely used datasets that record levels of government repression, these measures do not identify when the use of repression is central policy versus when it is the result of agency loss. This poses a problem because the hypothesis is that MNCs are responsive to the latter type but not the former. In lieu of direct measures, I approximate the two contexts by means of an interaction term in the statistical models.
One variable included in the interaction is an ordinal measure of repression. This is the “physical integrity rights index” from the CIRIGHTS project (Cingranelli et al. Reference Cingranelli, Richards and Chad Clay2014; Mark et al. Reference Mark, Cingranelli and Filippov2022), transposed so that it increases with government abuses.Footnote 17 The second variable included in the interaction term is an indicator of state capacity. Specifically, I use the new “absolute political capacity” metric developed by Fisunoglu et al. (Reference Fisunoglu, Kang, Kugler and Arbetman-Rabinowitz2023). It builds on the classic “relative political capacity” metric (Arbetman-Rabinowitz et al. Reference Arbeman-Rabinowitz, Kugler, Abdollahian, Kang, Nelson, Tammen, Kugler and Ronald2012) by combining data on both the government’s ability to generate inputs (i.e., absolute extraction capacity, excluding natural resource rents) and the value of its outputs (i.e., life expectancy).Footnote 18 The resulting variable (theoretically) ranges from 0 to 1, rescaled here to range from 0 to 100. Although it is a basic function of the state, collecting revenues is actually quite difficult and requires significant investment in administrative and bureaucratic infrastructure. States that demonstrate an ability to do so effectively—and to then convert those resources into real societal improvements—are typically more competent and capable.Footnote 19
I am assuming that agency loss declines with state capacity, as administratively adept governments are better able to discipline, and therefore deter, self-serving behavior by wayward employees. By interacting this variable with repression, I am assuming that physical integrity rights abuse in higher-capacity states reflects repression as strategic choice, whereas such abuse in lower-capacity states reflects at least some degree of repression by rogue agents.Footnote 20 Given the measurement scheme, the empirical hypotheses can be restated as follows:
-
H1: Pr(Entry) by firm i in country j decreases with the number of rebel groups.
-
H2: Pr(Entry) by firm i in country j decreases with repression at low levels of state capacity, but is less responsive to repression at high levels of state capacity.
The baseline probability that entry occurs is primarily a function of firm i’s overall FDI expansionism during the year—a dynamic that is particular to the MNC, and that is largely independent of the conditions characterizing country j. I thus include a variable measuring firm i’s entry rate during year t. This is calculated by taking the total number of observed entries in the “developing world” as a proportion of the total number of candidate sites (i.e., countries with which firm i does not yet have an FDI relationship).Footnote 21 This is conceptually akin to estimating MNC-year fixed effects but collapses 15,081 constants (i.e., the number of distinct MNC-year groups included in this analysis) into a single variable.Footnote 22 If new FDI ventures were randomly allocated, then this is the only variable that would associate reliably with the probability of entry. Of course, new FDI ventures are not randomly allocated; country-level factors are thus expected to inform whether country j performs above or below the baseline probability of entry by firm i, as recorded by this variable.Footnote 23
I control for a variety of other factors thought to be important determinants of FDI. These include geographic, economic, and sociopolitical characteristics of the host country that either add to or detract from its viability as an investment site.Footnote 24 The argument implies that the configuration of violent organizations operating in a country should affect MNC decision-making independently from the actual violence. It is thus important to account for other dimensions of rebellion and conflict. I use the number of battle-related deaths (recorded in hundreds) resulting from intrastate conflict during the year to measure the intensity of fighting (Davies et al. Reference Davies, Pettersson and Öberg2022). I also record the duration of conflict with a variable that counts the number of consecutive years fighting has been observed.Footnote 25 Both variables correspond with episodes of intrastate conflict recorded in the UCDP/PRIO Armed Conflict Database (Gleditsch et al. Reference Gleditsch, Wallensteen, Eriksson, Sollenberg and Strand2002).Footnote 26
Another potential confounding factor is the quality of the host’s governing institutions. Those that provide for transparency in policy-making and reliability of enforcement are particularly important for fostering a positive business environment (Jensen 2003, Reference Jensen2008; Biglaiser and Staats Reference Biglaiser and Staats2010). I use the rule of law index from the V-Dem Project (Coppedge et al. Reference Coppedge, Gerring, Knutsen, Lindberg, Skaaning, Teorell, Altman, Bernhard, Fish, Cornell, Dahlum, Gjerløw, Glynn, Hicken, Krusell, Lührmann, Marquardt, McMann, Mechkova, Medzihorsky, Olin, Paxton, Pemstein, Pernes, von Römer, Seim, Sigman, Staton, Stepanova, Sundström, Tzelgov, Wang, Wig, Wilson and Ziblatt2018), which is informed by judicial independence, respect for the constitution, access to justice, public sector accountability, impartiality in public administration, and legal transparency. It ranges (theoretically) from 0 to 1, rescaled here to range from 0 to 100.Footnote 27 A host’s financial openness and trade integration determine the ease with which prospective investors can integrate local enterprises into their existing global networks. These are measured using KOF’s “de jure financial globalization” and “de facto trade globalization” indices, respectively (Gygli et al. Reference Gygli, Haelg, Potrafke and Sturm2019), and are expected to correlated positively with entry.Footnote 28
MNCs are thought to show preference to host countries that are “closer” to home. This includes geographic proximity but also institutional similarity (Beazer and Blake Reference Beazer and Blake2018). I control for both the geographic distance and the administrative distance between home and host (Berry et al. Reference Berry, Guillén and Zhou2010). The former is measured as the number of kilometers separating each country’s geo-coordinated “centers.” The latter is a spatial metric informed by colonial history, commonality in legal systems, and religious traditions. Both increase with distance, meaning that they are expected to correlate negatively with entry.
Market size and sophistication are the most reliable correlates of FDI. I use population to measure the former and per capita GDP (i.e., development) to measure the latter. Both variables are logged. Physical size is measured by total land area in square kilometers, also logged. This variable captures the geographic scope and breadth of the host country; it also correlates strongly (∼0.8) with primary sector production. The data on population and land area are taken from the World Bank’s World Development Indicators database (2019), while data on GDP come from the UN National Accounts database United Nations (2019). Because the options for transport are necessarily more limited in island and landlocked countries alike, these conditions likely depress FDI. I include a binary indicator for each in the analyses.
A final control variable, MNC stock, records the number of foreign multinationals with FDI projects in country j. The measure is logged and was generated using the MNCs that populate the Corporate Affiliations historical database. Although this source doesn’t include the full universe of MNCs, it is large sample of many thousands and is likely a fair representation of overall FDI distribution and concentration. I anticipate it will correlate strongly with entry, as foreign investments tend to agglomerate.
All variables except entry rate are lagged 1 year.Footnote 29 This is to account for the time it takes between when the plan to enter a given market is made and when it is actually implemented. And, though it is unlikely that entry by a single firm i has an immediate causal effect on any of the host country conditions represented on the right-hand side of the equations estimated here, it is unlikelier still that it would exert influence on them in the year prior. While there are certainly valid concerns about reverse causation when studying FDI,Footnote 30 I’ve adopted a modeling strategy aimed at minimizing that particular risk.
Results
The results from a series of logit models are presented in Table 1.Footnote 31 For the sake of clarity, the estimated average marginal effects (AMEs) are reported, and as percentages (i.e., on a 0–100 scale).Footnote 32 It is important to note that, because of the structure of the dataset (i.e., company-country-year), and because the analysis isolates significant changes in the firm’s FDI position (i.e., its entry into a new market), nonzero outcomes are (relatively) infrequent. Simply put, the likelihood that firm i initiates a new FDI relationship with country j at time t is necessarily quite low.Footnote 33 Indeed, entry is recorded in 2.22% of cases. Meaningful interpretation of the results requires some extrapolation to the “bigger picture.” The AMEs should be viewed not only in terms of what they mean for the probability of investment by firm i but what they mean for the overall distribution of investment activity by the universe of firms i…n. In this light, even fractions of a percent might ultimately translate into several more (or fewer) new FDI projects in the host country. And seeing as FDI often entails a long-term commitment, this difference can have notable consequences for the host’s economy into the future.
FDI entry
Notes: Logit estimation. Average marginal effects (AMEs), reported as percentages (0–100). 1,878,689 observations. 9,709 MNCs and 131 countries, 1994–2018. Entry recorded in 2.22% of cases. Entry rate recorded in period t. All other variables lagged 1 year. Errors clustered by MNC. Two-tailed significance tests: *p < 0.05 **p < 0.01 ***p < 0.001.
Four models are presented here. A 1% increase in the entry rate corresponds to a nearly half-point increase in the probability of entry, and this is consistent across all models. It is, as anticipated, a strong effect, but it is not deterministic. There is still considerable variation in hosts’ respective prospects for selection, attributable to country-level factors.
The probability of entry by firm i increases reliably with rule of law, trade integration, financial openness, population, development, and the existing MNC stock. It decreases reliably with distance from the home country (both in terms of geography and administrative similarity) and land area. Islands and landlocked countries are also disadvantaged, all else being equal. The evidence from existing research suggests that market size and sophistication are the strongest correlates of total FDI.Footnote 34 And that is the case here, too: doubling the population size corresponds to a nearly 1% difference in the probability of entry, and doubling per capita GDP increases the predicted probability by just over one-half percentage point. MNC stock is also an especially strong correlate, as doubling the number of MNCs present corresponds to a 0.75% improvement in luring any given firm i. When one considers that, on average, entry is observed in 2.2% of cases, these estimated effects appear to be fairly powerful.
I gradually introduce the variables of theoretical interest across the four models. Model 1 includes only the fighting intensity variable. Consistent with earlier research, the probability that firm i strikes up a new FDI relationship with country j at time t decreases with the body count (e.g., Witte et al. Reference Witte, Burger, Ianchovichina and Pennings2017). That said, the magnitude of the estimated effect is relatively weak. A one-unit increase (i.e., about 100 battle deaths) reduces the viability of a potential host site by only 0.005%. Further calculations indicate that the AME attributed to 1000 additional battle deaths—a notable escalation in intensity—is 0.05%. This is roughly equivalent to the estimated improvement in host site viability corresponding to a 5% increase in population size, a 5.5% increase in MNC stock, a 7.5% increase in development level, or a three-point change on the 100-point rule of law index—all relatively minor differences in host country conditions.
The estimated effect attributed to fight intensity decreases further with the addition of fight duration in Model 2. Each additional year of fighting is estimated to reduce the probability of entry by 0.01%; 5 years of fighting reduces it by just over 0.05%, which is approximately equal to the magnitude of the effect attributed to 1000 additional battle deaths in Model 1. Though not as strong as the influences exerted by the best-performing variables, it suggests that the cumulative longevity of intrastate conflicts perhaps matters as much as, if not more than, the level of violence at any given time.
The rebel groups variable is introduced in Model 3. It is found to correlate reliably and negatively with entry—even in the presence of the fighting intensity and duration variables—as anticipated by the theory. The predicted probability is estimated to drop by nearly 0.1% with the addition of one rebel group, roughly equivalent in magnitude to the expected gain corresponding to an 8% increase in population. It drops by about 0.25% with the addition of three rebel groups, and just under 0.4% with the addition of five rebel groups—akin to the gains corresponding to a 22% and 35% increase in population size, respectively. Though not the most dominant determinant of entry found in the analysis, MNCs do seem more sensitive to the number of rebel organizations than they are to other dimensions of rebellion and conflict.
Indeed, when the rebel groups variable is included, the AMEs attributed to fighting intensity and duration both diminish substantially (and effectively drop to 0 for the former). While this wasn’t predicted by the theory, it doesn’t contradict the theory either. Indeed, what it suggests is that a nontrivial proportion of the effect that conflict exerts on FDI owes to the corresponding risk of economic predation by rebels. Only some of the effect is attributable to the general destructiveness and disruptiveness of fighting, which are the more commonly theorized mechanisms linking war to decreases in new investment.
Finally, repression and state capacity are introduced in Model 4. The results indicate that MNCs are deterred by violence perpetrated by the government. The probability of entry decreases reliably with the scale and scope of repression; and the AME attributed to each one-point change is similar in magnitude to what an approximately 3.5-point change in rule of law is expected to yield. What these results do not make clear, however, is whether these effects are generalizable across context. I theorized that MNCs care about repression when it is practiced by rogue agents acting in pursuit of their own political and economic interests, but less so when it is practiced by the central government as a means of deterring agitators from challenging the political status quo. To differentiate between these two contexts, I estimate a fifth model of entry wherein repression and state capacity are interacted. The results are displayed in Figure 1.Footnote 35 The empirical findings are consistent with the hypothesis.
Conditional marginal effect of repression on FDI entry. Note: 95% confidence intervals in shaded areas. Values on y-axes reported as percentages (0–100). State capacity set to approximately 5th, 50th, and 95th percentiles, respectively, in panel on right.
The panel on the left shows the estimated AME of repression across the range of state capacity found in the sample. While the probability of entry always decreases with an increase in government abuse, the correlation is clearly most pronounced among low-capacity states. Indeed, among the weakest states (i.e., those in the bottom 5th percentile), each one-point increase in repression is corresponds to a 0.1% decline in the country’s viability as a host site. Among those with moderate capacity, the decline drops to about 0.075%—a 25% difference in effect magnitude. It drops further still to 0.05% (a 50% difference in relative magnitude) for the highest capacity states (i.e., those in the top 5th percentile), with the upper-bound of the confidence interval approaching zero.
The panel on the right shows these same results from a different angle, plotting the predicted probability of entry across repression at three fixed levels of state capacity.Footnote 36 In the absence of repressive practices, there is no statistical difference in FDI appeal between lower- and higher-capacity states. But notable separation is found among those where repression is observed more frequently. Among administratively inept governments, the probability of entry is expected to decrease from greater than 2.25% at low levels of repression to lower than 2% at moderate levels of repression, bottoming out at nearly 1.5% at the highest levels of repression. In other words, weak states with high levels of repression are anticipated to attract 33% fewer MNCs than their less repressive counterparts. The contrast with adept states is considerable: the predicted probability of entry drops from a high of about 2.5% at low levels of repression to a low of 2.25% at high levels of repression—only a 10% difference in overall performance (and a 33% better performance than their low-capacity counterparts with equally high levels of repression).Footnote 37
These results are striking for three reasons. First, they imply that, as theorized, the perceived risks of unrestrained predation increase with repression, and especially so in environments marked by agency loss. MNCs are wary of the unpredictable, unregulated brand of theft that violent rogue agents might engage in. In contrast, they are more willing to accept repression as central policy. In this context, violence is used by the state to serve a largely political purpose, not an economic one. It is thus less threatening to foreign investors. Second, the magnitude of the effect among low-capacity states is fairly large when compared to the other variables measuring political violence employed in this study. It clearly represents a stronger deterrent than is attributed to fight intensity or fight duration and is moderately stronger than the effect attributed to rebel groups. Third, these results support the finding from other studies that accounting for state capacity is important and not only for whatever direct effects it might exert on outcomes of interest. Crucially, government strength likely moderates the effects exerted by others factors (Coan and Kugler Reference Coan and Kugler2008). In this case, that other factor is repression (and, notably, who is practicing it).
While the basic empirical model fits the data on firm entry fairly well, it struggles to fit the data on firm exit. These results are presented in the Appendix.Footnote 38 Distinctly theorizing FDI termination goes beyond the scope of this paper; and, indeed, the findings suggest that it is a question that warrants further scholarly attention and theoretical refinement (Boddewyn Reference Boddewyn1983; Tan and Sousa Reference Tan and Sousa2019). But there is one pattern that has emerged in the (limited) empirical work on the matter: MNCs’ sensitivity to general political risk factors seems to wane post-entry (Dai et al. Reference Dai, Eden and Beamish2013; Barry Reference Barry2018). This apparent, sudden shift in sensitivity reveals an intriguing possibility—that MNCs’ behavioral tendencies align more with “loss aversion” than “risk aversion,” per prospect theory (Kahneman and Tversky Reference Kahneman and Tversky1979). Of course, that’s just one plausible explanation for the disparities in the data. I’ll leave it to future research to piece together this particular puzzle.
Conclusion
The scholarship on organized violence and FDI has two problems. First, it often fails to isolate the distinct (or even direct) means by which violence threatens MNCs. Precise and targeted mechanisms are subsumed by the general conditions that characterize a violent environment. Second, it lacks a unified and coherent theoretical framework. Studies on the relationship between conflict and FDI are treated as tangentially related to, but ultimately distinct from, studies on the relationship between repression and FDI. In reality, both are merely different manifestations of organized violence; and how they influence FDI activity should therefore follow a common causal pathway.
In this paper, I set out to solve these problems. I adopted Olson’s (Reference Olson1993) model of “stationary” and “roving” banditry to serve as the overarching theoretical framework and applied it to three different types of violent organizations: the state, its agents, and rebels. The exercise yielded a few straightforward hypotheses. In short, MNCs are repulsed by configurations of organized violence that include rebel groups and rogue government agents but are indifferent to violent repression when it is willfully and deliberately employed by the state against its subjects. Testing these hypotheses against real-world data on FDI entry produced promising results. The findings are largely consistent with the theory; they also add nuance to the received understanding about the relationship between insurgency and FDI and challenge the now mainstream hypothesis that states who abuse the physical integrity rights of their subjects are punished by multinationals.Footnote 39
Interestingly, supplemental analyses revealed that the general risks associated with operating in a predatory environment have little bearing on MNC exit. Indeed, few of the variables characterizing the host country environment correlate reliably with FDI termination, which is consistent with the limited body of work on the matter (Boddewyn Reference Boddewyn1983; Benito Reference Benito1997; Barry Reference Barry2018). There is clearly a need for more research on exit moving forward (particularly in the political economy literature).
Aside from explicitly theorizing exit, there are several other ways in which the present study can (and should) be improved upon in the future. I have presented a very generalized theoretical framework and demonstrated its plausibility. But the model can almost certainly be fine-tuned by introducing contextual factors that condition the hypothesized relationships. Firms’ sensitivities likely vary with the industry of investment, for example (Blanton and Blanton Reference Blanton and Blanton2009; Barry Reference Barry2016; Witte et al. Reference Witte, Burger, Ianchovichina and Pennings2017);Footnote 40 or with their country and culture of origin (Beazer and Blake Reference Beazer and Blake2018), or with their size and experience (Jiménez et al. Reference Jiménez, Luis-Rico and Benito-Osario2014). Likewise, not all rebel (or nonstate) entities are equal in their propensity to practice unrestrained economic predation (Humphreys and Weinstein Reference Humphreys and Weinstein2006; Stewart and Liou Reference Stewart and Liou2017). Sensitivity tests revealed that MNCs do seem to discriminate between conflict environments where rebels generate income via commercial theft versus those where they don’t (see Figure B2 in the Appendix). This could be probed further by, for instance, zooming in to the subnational level and evaluating where firms’ subsidiaries are located in relation to contested versus uncontested territories (Dai et al. Reference Dai, Eden and Beamish2013) and, when uncontested, by measuring the propensities for unrestrained predation exhibited by the group holding the territory (whether rebel, state, or state-sponsored militia).
Such efforts will enhance the richness of the theoretical framework and the predictive power of the empirical model. But there are broader conclusions from the present study that are worth noting. Namely, that, in the age of international capital mobility and globalized production, there are perhaps significant long-term costs for governments that are unable to centralize and control the means of violence. Such countries will struggle to forge new alliances with the MNCs who lay claim to much of the world’s financial, technological, and intellectual capital. Curiously—and contrary to some theories of globalization (e.g., Strange 1998)—one implication is that, by rewarding and reinforcing states that are capable of monopolizing the means of organized violence and punishing those that aren’t, global capitalism will in fact buttress the international system of states, rather than threaten its demise.
But this does not mean that global capitalism will tame violence altogether. The results of this study suggest that, though MNCs are wary of violence when it is perpetrated by rebels and rogue government agents, they are notably less deterred by violence when it is wielded as a tool by a competent state against political opponents and agitators. Capital mobility may incentivize governments to rein in rival organizations and wayward employees, but it does not disincentivize the deliberate and targeted use of violence by governments themselves. At least, not without “spotlighting” by human rights activists (Spar and La Mure Reference Spar and La Mure2003).Footnote 41 Indeed, the results of this study should serve to highlight the pivotal role that civil society might need to play in shaping the global capitalist system—in transforming it from one that merely reflects the profit- and efficiency-minded interests of the world’s largest corporations to one that safeguards the basic rights and dignity of its people, too.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/bap.2023.30.
Competing interests
The author declares none.
Open access
