Model

The players in our model are a host government (G), a target firm (T), and a set of partner firms (P) that operate in the host country and are linked to the target. These partner firms may be foreign-owned or domestically owned.Footnote ²⁰ We denote the number of partner firms by parameter ρ. The game begins when Nature chooses the host government’s type, α, which is the political and/or economic pressure on the host government to break its contract with the firm.Footnote ²¹ The host government receives a benefit from breaking the contract when α > 0. A variety of circumstances might cause this: breach might enable a government to survive in power, to respond to the particular demands of a sector or industry, to cater to domestic interests, to achieve foreign policy goals, or to commit corruption (Fails Reference Fails2012; Kobrin Reference Kobrin1984; Wellhausen Reference Wellhausen2015a ). Breach might also be caused by an unexpected regime change or surge in nationalism. The benefits of breach may vary with economic conditions and the partisan or institutional structure of the government (Jensen and Johnston Reference Jensen and Johnston2011; Li Reference Li2009; Pinto Reference Pinto2013). Since expropriation can also generate costs, we allow for the possibility that breach is sometimes costly, α < 0. For example, the host government might derive ongoing political and economic benefits from FDI that would be lost with expropriation (Eaton and Gersovitz Reference Eaton and Gersovitz1984). Regardless of how we substantively interpret parameter α, it represents the firms’ uncertainty about how willing the host government is to break its contract. We remain agnostic about the government’s motives for breach, simply acknowledging that a variety of circumstances may lead a government to believe that breaking a contract will generate an overall benefit or cost.

After the host government learns its true type, each firm simultaneously chooses a level of effort, e_i ⩾ 0, to spend on protecting the target firm from expropriation. As discussed above, effort might represent a monetary payment, such as a campaign contribution or a bribe, from a firm to the government. This is a standard assumption in models of special interest groups, including the literature on endogenous trade policy (e.g., Grossman and Helpman Reference Grossman and Helpman1994, Reference Grossman and Helpman1996). We can also interpret effort more generally as costly activities like publicly supporting the host government or providing local goods and services through CSR activity, such as infrastructure and public health programs. After the target and partner firms choose their level of effort, the host government decides whether to break or honor its contract with the target. Table 1 shows the utility functions for each player and each possible outcome.

TABLE 1.

Utility Functions for the Theoretical Model

Note: e_P is the aggregate effort of partner firms.

If the government breaks its contract, it is in essence reducing the value of the target firm. In the case of outright nationalization, the government might seize a physical asset such as an oil refinery. In the case of indirect (creeping) expropriation, the host government may impose discriminatory regulation or taxation that redirects profit from the target firm to the government. We therefore assume that breaking a contract shifts value from the target firm to the government. We let parameter v > 0 represent the value of the firm’s operations in the host country.Footnote ²² If the government breaks its contract, the target firm receives a payoff of zero, and the government receives a share of the target firm’s total value plus the government’s payoff from the pressure to breach, σv + α.Footnote ²³ We refer to parameter 0 < σ < 1 as the host government’s capacity to extract value from the target’s assets. This capacity can be affected by many different factors that are government-, industry-, and firm-specific, some of which are discussed below. To allow for the possibility that partner firms may receive some economic benefits even if the government breaks its contract, we assume that a partner firm i receives payoff γ _i v where 0 ⩽ γ _i < 1. For example, when the government seizes ownership of a firm through outright nationalization, a supplier may be able to continue to sell products to the firm.

If the government honors its contract, it receives a payoff that is equal to the effort invested by the target and its partner firms, e_T + e_P . The target firm retains its total value and loses only the cost of its own effort, v − e_T . A partner firm i receives a share of the firm’s value less its own effort, (γ _i + λ _i )v − e_i where γ _i + λ _i < 1. Since γ _i v represents a partner firm’s payoff when the government has broken the contract, the quantity λ _i v represents the added benefit that i receives when the target’s contract is honored. We refer to parameter 0 < λ _i < 1 as the economic link between the target and partner firm i because larger λ _i values indicate that a partner firm has more incentive to protect the target’s contract with the government.Footnote ²⁴ This parameter can represent transactions in which the target buys and/or sells intermediate goods to partner i, as well as nonmonetary benefits, such as the ability to share expertise or technology. Even if a partner firm continues to transact with an expropriated target firm (under private or public ownership) in the future, contract breach can disrupt operations in the short-term or decrease the efficiency of the partner firm. The parameter λ _i can thus be interpreted as a partner firm’s cost when the government breaks a contract with the target firm.

Robustness

We have presented the simple model above so that we can clearly identify the mechanism that drives our main argument: an increase in the strength of economic links generates more aggregate firm-level effort to protect the target, thus reducing the probability that the host government breaks its contract with the target. Nevertheless, it is worth asking whether our results are robust to alternative assumptions and complicating factors.Footnote ²⁷

Uncertainty plays a key role in our model because breach occurs when firms underestimate the host government’s true net benefit from breaking the contract and hence choose too little effort. If firms could perfectly observe the true value of α, then the firms could almost always induce the government to honor its contract. The only situation in which they could not deter breach would be if the government is under so much pressure that it would break its contract even if all of the firms spent all of their profits on effort (i.e., e_T = v and e_i = λ _i v for all partners). Since governments often break contracts even when they are not under such extreme pressure, we must have uncertainty in our model for its observable outcomes to match real-world events. We could alternatively allow the government to send a cheap talk signal to the firms about its true value of α. However, such an extension yields the same mechanism and equilibrium outcomes, because the government would always have incentive to over-report its type (i.e., claim that its value of α is larger than the true value) in order to extract more effort. Any attempt by the government to communicate would be ineffective because rational firms would understand that the government does not have an incentive to tell the truth.

Another key component of the model is that we focus on firms that receive benefits from the target. Such partner firms want the host government to honor its contract with the target. However, a target may also harm competitor firms that want the host government to break its contract. We can easily include competitor firms in our model and allow them to invest effort in trying to get the government to break its contract. This more complex model yields the same basic results. We can also include firms that are not affected by the target, either positively or negatively. These firms would never invest effort—either for or against contract breach—because they would be unaffected by the host government’s decision.

Our model can also easily accommodate additional variation in sector, industry, and firm attributes. For example, we might imagine that the host government is more responsive to sectors, industries, and/or firms that are located in politically valuable areas or that are domestically owned, rather than foreign-owned. Similarly, some firms may find it less costly to spend effort than others. Larger firms may find it easier to hire full-time staff to manage their government relations than smaller firms. We remain agnostic about such factors, and note only that our results are robust to including sector-, industry-, and firm-level variation in the cost of effort and its benefit to the government.

Other forms of enforcement do not invalidate our results either. Suppose that even if the host government decides to break the contract, there is some probability that this action might be stymied or overturned. For example, we might imagine that the target can sometimes secure relief from a domestic or international judicial institution after the host government breaks its contract. Alternatively, the host government might be constrained by veto players or other institutional constraints that support the rule of law (Frye Reference Frye2004; Reference Frye, Aslund, Guriev and Kuchins2010). When these constraints grow stronger, firms will invest less effort in equilibrium, but Propositions 1–4 all continue to hold.

Finally, we should comment on our assumption that economic links are exogenous. If links are endogenous, then potential target firms should build economic links when possible to provide long-term protection from future contract breach. While we believe that this is one major implication of our argument, we recognize that there are inherent limits in the ability of firms to create these links. First, it is unlikely that a firm can change its economic links immediately after being targeted for a possible contract breach. In this sense, every target’s links are exogenous in the short term. Second, while links may provide protection from contract breach, they can come with other costs, such as transaction costs and economic inefficiencies. Our model provides intuition about one possible benefit of economic links, but it does not fully consider their potential costs. Third, some firms will find it difficult to construct links because of the nature of their industry or form of production. For example, firms that produce intermediate goods, such as electronic componentry, should find it easier to create economic links than firms that do not.

Empirical Implications

Our model shows that economic links play a key role in the endogenous protection of property rights. Host governments are most likely to break contracts with firms that are not integrated with others. One way to measure economic links is to consider supply chains. Supply chains create transactions that link the activities of firms in the chain. Breach at one link would likely cause other firms in the chain to reduce investments to account for expected changes in inputs. Government breach of contract may even trigger automatic exit by other firms in the chain if, for example, it limits the ability of firms to do business in the host country.

Our key comparative static, Proposition 1, considers how variation in the strength of a target’s links affects the likelihood that the government breaks its contract. One possible determinant of the size of a target’s links is its relative position within a supply chain. As a simple illustration, consider two firms, A and B, that are linked in a chain. Suppose that firm A is a supplier that sells a good to firm B, its customer. If the government breaks a contract with the supplier (firm A), the customer (firm B) is likely to experience some harm, as government interference will likely generate either higher prices for firm B or require firm B to find a new supplier. So the supplier creates a relatively small, albeit positive, benefit for its customer.Footnote ²⁸ In contrast, if the government breaks a contract with the customer (firm B), then the supplier (firm A) is likely to experience much greater harm, ceteris paribus, since demand for the supplier’s good is reduced. That is, the customer should generate a relatively large benefit for its supplier, conceptualized here as stronger economic links. By the logic of Proposition 1, this asymmetry implies that the customer (which generates larger benefits) should be less likely to experience contract breach than the supplier (which generates smaller benefits). Put differently, a host government is less likely to break contracts with “downstream” firms, which purchase more goods from other firms. Proposition 1 is the basis for our main empirical hypothesis in the next section of our article:

We also examine two additional empirical implications of our argument. First, Proposition 2 suggests that the government will be less likely to break its contract when it lacks capacity to extract value from the target firm. We measure capacity using various measures of operational expertise, thereby generating our second empirical hypothesis:

Finally, Proposition 3 states that the government is less likely to break contracts with high-value firms. As we emphasized above, this parameter refers to the total value of the firm’s operations in the host country, not worldwide. We accordingly measure a firm’s value by the size of its investment in the host’s economy, creating our final empirical hypothesis:

Cross-national Evidence

To demonstrate the plausibility of Hypothesis 1, we examine country-level data for non-OECD countries to see if economic links are associated with less government contract breach. We restrict our analysis to US-tied supply chains and government contract breach with US firms, because if home-host government politics affects investment disputes, as suggested by Ramamurti (Reference Ramamurti2001) and Wellhausen (Reference Wellhausen2015a ; Reference Wellhausen2015b ), then limiting the analysis to one home country is an appropriate starting point.Footnote ²⁹

Our ideal dependent variable would be the percentage of contracts with US firms that the host government breaks in each country-year. As the best observational proxy, we measure the number of public investment arbitrations filed by US firms against each host government while controlling for that government’s opportunities to break contracts.Footnote ³⁰ A public investment arbitration captures a contract dispute contentious and intractable enough that a firm files against an offending government in a public international tribunal. Clearly, there is a complex selection process into arbitration, making this analysis a plausibility probe rather than a definitive test of our argument. However, this analysis is useful for two reasons. First, it would be difficult to believe a theory that is inconsistent with the available observable evidence. Second, if we make the admittedly strong assumption that more arbitration is associated with more contract breach altogether, then the measure is an underestimate of the ideal dependent variable.

We collect data about public investment arbitrations brought by US firms against host governments from 2002 to 2012.Footnote ³¹ There are a total of 94 US-filed public investment arbitrations in our dataset, filed against 38 countries, ranging from 1 to 5 per country-year. Even though these arbitrations are “public,” the specific details of the cases (such as claim sizes) are usually private, because these cases are filed by firms as private actors.Footnote ³² Public investment arbitration is usually facilitated by bilateral investment treaties (BITs) and related instruments that allow investors to bring disputes at venues like ICSID, but MNCs often include investment arbitration clauses in their written contracts as well. Therefore, in our sample we include all available countries and control for the presence of a US BIT with the host country, allowing for the possibility that an investor may be able to arbitrate through a firm-specific contract, even if the host country has not signed a BIT.Footnote ³³

To account for the fact that a host country with more FDI has more opportunities to break contracts, we scale US-filed public investment arbitrations by US-origin FDI stock. Scaling allows us to meaningfully compare breach across countries with governments that differ in their opportunity to breach. This ratio is our standardized dependent variable, Arbitration. The variable ranges from 0 to 1, with a mean of 0.002 investment arbitrations filed by US firms per US$1 million in US-origin FDI.Footnote ³⁴ For the ease of the reader, we multiply this measure by 1,000 when performing the analyses in Table 2.

TABLE 2.

Economic Links Are Associated with Less Arbitration

Notes: Non-OECD countries. Robust standard errors are clustered by country.

Significance levels: * p < 0.1.

We measure our explanatory variable of interest in two ways. Our first measure of purchasing relationships is US intra-firm trade (sometimes called “related-party trade”), which measures “trade between a US entity and an entity that receives the export shipment where either party owns directly or indirectly 10 percent or more of the other party” (US Census Bureau). In other words, intrafirm trade involves exchange between entities that are economically linked through ownership stakes, capturing parts of the supply chain that operate through US MNCs. Our second measure of purchasing relationships is US intermediate goods, which measures exports of intermediate goods by in-country firms to US firms.Footnote ³⁵ This measure captures parts of global supply chains that operate outside the ownership of a US MNC.Footnote ³⁶ We lag these measures and log them to account for skewed data.Footnote ³⁷ We control for the host country’s Polity score, because the host country’s domestic institutions could affect economic links as well as FDI and the government’s willingness to breach and submit to arbitration (Jensen Reference Jensen2006; Li and Resnick Reference Li and Resnick2003).

We use two specifications. First, we analyze the data as a panel, with year fixed effects and a lagged dependent variable, which acknowledges that a government’s past experience with arbitration and levels of FDI may influence its future behavior. Second, we collapse the data and identify our model off of cross-country variation. All models are OLS regressions with robust standard errors, clustered by country. Table 2 shows our results.

In three of the four models, economic links are significantly and negatively associated with public investment arbitration relative to FDI. As a real world analogue for effect sizes, Belize, which has economic links at about the sample mean, has faced more US-filed arbitration relative to its US-origin FDI than El Salvador, which has economic links that are about one standard deviation above the mean.Footnote ³⁸

While this plausibility probe, based on novel arbitration data, is consistent with our argument, we now use novel survey evidence that more directly gets at the relationship in question.

Survey Evidence

We created and administered a survey of government affairs executives in US MNCs operating in Russia to directly test Hypotheses 1–3. These hypotheses and our empirical measures are summarized in Table 3. Russia is an appropriate setting, as expropriation of various kinds has taken place in various industries since Russia’s independence (Frye Reference Frye2004, Reference Frye2006; Gustafson Reference Gustafson2012). We focus on one host country to avoid confounders, since host country risk environments shape firms’ entry decisions (Antras Reference Antras2005; Bernard et al. Reference Bernard, Jensen, Redding and Schott2011; Dunning and Lunda Reference Dunning and Lunda2008; Henisz Reference Henisz2000; Reference Henisz2002). As was a concern in the cross-national analysis, variations in diplomatic relations could confound our findings, so we survey only US firms. A key advantage of our survey is that the respondents, predominately located in Moscow or Washington, DC, are responsible for their firms’ relationships with the Russian government. We can therefore reasonably expect our respondents to be informed about actual or potential government interference with their firms’ operations.

TABLE 3.

Empirical Implications

Note: * In our empirical analysis, the sign for these variables should be reversed because we exclude the largest category.

We emailed respondents an anonymous online survey, with two follow-up emails and one follow-up phone call to each person in the sample. Of 703 people contacted, we had 56 responses or an 8 percent response rate.Footnote ³⁹ The firms in our sample are all members of the US-Russia Business Council (USRBC), which aims to provide “significant business development, dispute resolution, government relations, and market intelligence services” to its members.Footnote ⁴⁰ The organization occasionally acts on behalf of particular members that are facing possible expropriation by the Russian government.Footnote ⁴¹ If the USRBC is effective in this mission, we would expect members to have systematically lower expectations of expropriation than nonmembers, meaning that we are blunting our ability to test our argument to the extent that nonmembers are otherwise similar to members.

How representative is our sample?Footnote ⁴² Firms span the full range of new entrants to Russia in 2013 to firms that were present in Russia in 1990; the average firm had been in Russia for 8.9 years (with a standard deviation of 6.8 years).Footnote ⁴³ Twenty-five percent of the sample reported that their firms had less than US$500,000 of investments in Russia, and 25 percent reported investments greater than US$200 million; the variety of firm sizes in our sample is useful.Footnote ⁴⁴ Table 4(a) breaks our sample down by industry. From 2010 to 2012, the biggest sectors for inward FDI to Russia (from any country) were finance and energy/mining, both of which are represented in our data (US Commercial Service 2013). To the extent that nonfinance services and manufacturing are over-represented in our sample, the “obsolescing bargain” literature suggests that we underestimate concerns about expropriation in the full MNC population. This is because such industries, with relatively mobile assets, should be less likely to face expropriation (e.g., Kobrin Reference Kobrin1987; Vernon Reference Vernon1971). We do not know the productivity of either our surveyed firms or the population. Because productivity has an effect on firms’ decisions as to how and where to invest abroad (Chen and Moore 2010; Yeaple Reference Yeaple2009), we control for a variety of characteristics of the firm—including size, sales, and industry—that likely correlate with productivity.

TABLE 4.

Survey Respondent Characteristics

We relied on a respondent’s expertise to report past experience as well as expectations about future experience with breach of contract at the firm. We prompted respondents with the statement: “Sometimes, government actions lower the value of foreign investments in ways that violate contractual obligations or general expectations of fairness.” We then ask how important this issue has been “for your firm in Russia” and how important “do you expect this issue to be in the next five years,” in four categories from “not at all important” to “very important.”Footnote ⁴⁵ Our first dependent variable, Lowers value, averages responses to the current and five-year projections. We prompted respondents a second time with the statement: “Sometimes, government actions transfer ownership of foreign investments to the state or state-owned entities.” We form our second dependent variable, Transfers ownership, out of parallel issue importance questions. Our third dependent variable, Total breach, averages the values of Lowers value and Transfers ownership.

For each issue, we went on to ask respondents how important “is this issue for your non-Russian competitors?” Respondents on average reported that their foreign-owned competitors face significantly more problems with ownership transfer (p < 0.01).Footnote ⁴⁶ This result works to our advantage. It could be that respondents are systematically underestimating their own risks, which would lead us to underestimate any effect of economic links on Transfers ownership. Or, it could be that those firms we surveyed are indeed better risk managers than their competitors, meaning that our sample would be a more difficult one in which to find effects.

To measure our main explanatory variable, we focus on a firm’s upstream economic links with Russian firms (Hypothesis 1). For our primary measure, we asked respondents, “Approximately what proportion of your supplies do you buy from Russian firms?”Footnote ⁴⁷ Results on the main explanatory and dependent variables are summarized in Table 4(b). In support of our hypothesis, the average value of the dependent variables generally decreases as supplier categories increase. The main exceptions are the average values for the eight firms reporting no Russian suppliers. However, we do not know whether these firms have only non-Russian suppliers or no suppliers at all. Our results below are robust to including or excluding these observations.Footnote ⁴⁸ We also asked how many Russian firms the respondent’s firm uses as suppliers. Fifty-one percent answered our maximum category, “10 or more,” while only one respondent reported having only one Russian supplier.Footnote ⁴⁹ We can thus have confidence that, on the whole, the economic links captured with the proportion question include benefits that are spread across multiple Russian firms. Our main explanatory variable, Proportion of Russian suppliers, is measured as a categorical variable. Our excluded category is firms with a proportion of 75 percent or more, and we compare this to firms with 0, 1–24 percent, 25–49 percent, and 50–74 percent. Thus, we expect positive coefficients: relative to the excluded category, respondents with weaker economic links should report more breach of contract.

As an ancillary test of Hypothesis 1, we examine another form of economic link: joint ventures. When a foreign-owned firm creates a joint venture with a domestic firm in its host country, it links their economic activities. This suggests that any harm that is experienced by the foreign-owned firm will spill over and hurt its domestic partners. Accordingly, we asked survey respondents whether their firm has a joint venture with a Russian entity. Our theory suggests that the coefficient for the dummy variable Joint venture with Russian entity should be negative.Footnote ⁵⁰

We examine the impact of government capacity (Hypothesis 2) using two measures. First, we create a dummy variable Interactions with SOEs that equals 1 if a firm has business interactions with state-owned entities (SOEs) at least monthly.Footnote ⁵¹ We expect that more interactions with SOEs indicate that the host government has more capacity to use foreign-owned assets, making breach a larger concern. We therefore expect the coefficient for Interactions with SOEs to be positive.Footnote ⁵² Second, we create a dummy variable called Foreign-owned market share that equals 1 if foreign-owned firms dominate the Russian market (i.e., control 50 percent or more of the industry’s market, as reported by respondents). In this case we expect a negative sign, as more foreign control in the market suggests less government capacity to successfully use expropriated assets.

To test Hypothesis 3, we use two measures that account for the value of a targeted firm. We aggregate each firm’s Size of Russian investments into four categories: less than $500,000; $500,000–10 million; $10–200 million; and over $200 million. We use firms with over $200 million in Russian investments as the excluded category. Next, we use a dummy called Sales in Russia that equals 1 if 50 percent or more of a respondent firm’s sales are in Russia. Consistent with Hypothesis 3, we expect that firms with more investment and more sales in Russia will be less concerned about a possible breach of contract.

We include several important control variables. Because the characteristics of a particular industry could be correlated with both economic links and expropriation risk, we control for whether a firm is in industries with relatively immobile assets; manufacturing; or services (the excluded category). The “obsolescing bargain” literature implies that firms in immobile industries should report systematically higher concerns about breach than those in mobile services (e.g., Kobrin Reference Kobrin1987; Vernon Reference Vernon1971). Because the assets of manufacturing firms are not easily redeployable, they may have few concerns about breach (Frieden Reference Frieden1994), although it is not ex ante clear how their breach concerns should compare to those of services firms.Footnote ⁵³ Finally, we control for a firm’s Years in Russia since 1990. We are agnostic as to whether this suggests firms should face more or fewer risks of breach.

Because our dependent variables are scales that are both bottom- and top-censored, we use Tobit regressions (with robust standard errors).Footnote ⁵⁴ Sample size varies due to respondent dropouts on particular questions, which we assume to be random. We report our results in Table 5.

TABLE 5.

Firms with Weaker Economic Links Report More Concern over Breach

Notes: Dependent variables range 1–4. † Excluded industry category: services.

Robust standard errors. Significance levels: *** p < 0.01, ** p < 0.05, * p < 0.1.

Our analysis supports Hypothesis 1. We see consistent evidence, across a variety of forms of the dependent variable, that respondents at firms with a lower proportion of Russian suppliers report significantly more breach as compared to the excluded category, which includes respondents at firms with a very high proportion of Russian suppliers. In Models (5) and (6), where breach is captured by Lowers value, respondents at firms in supplier categories from 1 to 24 and 25 to 49 percent report significantly more concerns about breach than firms with more than 75 percent Russian suppliers. In Models (7) and (8), the dependent variable is Transfers ownership. Results are again positive and significant for firms with either 1–24 percent or 1–24 and 25–49 percent Russian suppliers. Because these analyses treat the four categories “not at all important” to “very important” as a continuous four-point scale, we can interpret coefficients as such: in Models (5)–(8), respondents at firms with lower proportions of Russian suppliers are between 1.4 and 1.8 points closer to “very important” on their responses than are respondents at firms with 75 percent or more Russian suppliers. Model (9) averages the Lowers value and Transfers ownership dependent variables into one scale, Total breach, and results are consistent: respondents at firms from 1 to 24 and 25 to 49 percent Russian suppliers report significantly more breach concerns. Coefficients for firms in the 50–74 percent range are positive across all models but not significant.Footnote ⁵⁵ Our ancillary test of economic links, Joint venture with Russian entity, is uninformative.

We also find evidence that supports Hypothesis 2 regarding government capacity. Firms that regularly interact with SOEs report more concerns about breach, but the relationship is not significant.Footnote ⁵⁶ In Models (6), (8), and (9), firms that operate in industries with a large foreign-owned market share have significantly fewer breach concerns. Additionally, we find no significant support for Hypotheis 3. The coefficients for the size of firm investments in Russia are opposite to predictions, while the coefficient for sales in Russia is as predicted; none, however, are statistically significant. With regard to control variables, we find that firms in Immobile industries in fact report no more concern about breach than firms in services, while firms in manufacturing report significantly fewer concerns (Models (8) and (9)).Footnote ⁵⁷ A firm’s Years in Russia does not have a significant impact on its perceptions of breach.

Results on Hypothesis 1 are robust to using a dichotomous measure of the proportion of Russian suppliers.Footnote ⁵⁸ Results are consistent but insignificant when replacing the proportion with a measure of the actual Number of Russian suppliers.Footnote ⁵⁹ These weaker results correspond with a key element of our theory: firms face a collective action problem when they must jointly protect the property rights of other firms (Proposition 4). While each firm within a supply chain has an incentive to protect other firms in the chain, this incentive must be balanced against the temptation to free-ride on the effort of other firms in the chain.Footnote ⁶⁰

With our novel survey of government affairs representatives at US MNCs operating in Russia, we find evidence from MNC executives that a greater proportion of domestic suppliers goes together with less government breach of contract. In the next section, we use foreign firms’ experiences in another host country, Azerbaijan, to describe how domestic suppliers might provide property rights security in practice.

Case Study Evidence

We now examine relationships between the Azerbaijani government and two foreign-owned firms to demonstrate how upstream economic links can protect firms against breach of contract (Hypothesis 1).Footnote ⁶¹ These cases involve two similar firms: BP, a British oil and gas firm that leads operations in the Caspian Sea, and Barmek, a Turkish energy firm that ran the privatized power grid in Baku, Azerbaijan’s capital and largest city. These cases are not intended to serve as proof that our theoretical argument explains the behavior of all investors in all developing countries. However, we do not believe that these cases are exceptional or idiosyncratic. Azerbaijan is one of many former Soviet republics with weak political and judicial institutions that are attempting to develop their economies using foreign investment. Additionally, expropriation is a major concern for the energy industry in many developing countries.Footnote ⁶² BP and Barmek were very similar on many key dimensions that are relevant to foreign firm-government relations: they were both leaders in immobile industries core to Azerbaijan’s economy; local actors report that investors of each nationality are respected in the country; both received considerable political and press attention; and both have contributed considerably to government coffers. For example, Barmek reported that it paid US$54.6 million in taxes to Azerbaijan from 2001 to 2006.Footnote ⁶³ However, they received dramatically different treatment from the Azerbaijiani government. BP’s contract with the government remains solid as of 2015 while Barmek sued Azerbaijan for expropriation in 2006.Footnote ⁶⁴

The oil and gas industry is highly politicized in oil-rich Azerbaijan. As the leader of the Caspian Sea drilling consortium, BP regularly interacts with the government. The two sometimes disagree on issues like quotas for hiring domestic workers, the rights of foreign workers, and production levels. However, none of these disagreements have triggered public discussion of expropriation. Local interview respondents struggled to come up with an example of government interference with BP’s operations.Footnote ⁶⁵

To explain the lack of expropriation disputes, several respondents stated that BP benefits from a kind of krisha established by its suppliers. As explained above, the Russian term krisha has historically been associated with protection rackets under which firms make extra-legal payments to ensure their safety. But here, in a benign way, firms that have “strategically overlapping interests” with BP generate informal protection for BP from the Azerbaijani government.Footnote ⁶⁶ One well-placed executive explained his use of the term krisha by pointing out that BP’s “Tier 1, 2, or 3 suppliers are as strong as the firm is with the government.”Footnote ⁶⁷ A tier 1 supplier is a supplier closest in the chain to BP, while tier 2 and 3 suppliers are two or three links away, respectively. That the krisha has such a deep reach reinforces the finding from our formal model that any firm that receives a benefit from the target—no matter how far removed—can play a role in protecting the target. Further, BP is involved in joint ventures with SOCAR, Azerbaijan’s state-owned oil firm, which reinforce BP’s krisha.Footnote ⁶⁸ In short, a number of Azerbaijani firms, both large and small, receive benefits from BP’s presence and would be hurt by BP’s expropriation. Executives on the ground recognize that BP’s economic links form a krisha that protects BP, likely helping to keep BP’s regular interactions with the Azerbaijani government from generating risks of breach of contract.

The situation of BP and its krisha stands in contrast to the Azerbaijani government’s expropriation of the Turkish firm Barmek. Similar to BP, Barmek had a central role in the Azerbaijani economy: it won the 2001 privatization tender to manage the electricity grid for Baku for 25 years. Germany’s Siemens was the first foreign owner of Baku’s grid, but after the firm disagreed with the Azerbaijani government over the grid’s development, Siemens exited. Siemens’s dispute, as well as the subsequent international privatization tender process, revealed considerable information about the Azerbaijani government’s preferences over the operation of the grid. In light of this information, Barmek decided to actively manage its government relations through close personal contacts with top officials.Footnote ⁶⁹ In fact, it is widely believed that Barmek’s relationship with the Minister of Economic Development, Farhad Aliyev, helped Barmek to win the tender. When the government contained Barmek supporters, the government would have received a low benefit from expropriation (α in our model), making Barmek secure. However, Aliyev was arrested in October 2005 on charges of conspiring to overthrow the government prior to the November parliamentary elections. This change in the government’s composition increased the government’s benefit from breaking its contract with Barmek and triggered an immediate decline in Barmek’s relationship with the government. By 2006, Azerbaijan had brought 30 criminal cases against Barmek employees, accusing them of “abuse and misappropriation of money by inspectors” (Turan Information Agency 2006). These prosecutions sent a clear message: the government wanted Barmek to leave. Eventually the government decreed a forced change to Azerbaijani ownership.

We contend that this expropriation was possible in part because Barmek lacked relationships with local suppliers and hence lacked a protective krisha. Barmek’s particular ownership did not confer substantial benefits on the power plants that were its sole local suppliers, leaving supply chain links weak and the likelihood of an effective enforcement community low. Power plants had a reliable market in Baku’s grid and were still able to supply their product after Barmek’s ouster, despite the fact that the grid’s operations suffered under the government’s low operational capacity. Any costs emanating from ousting Barmek did not spread up a supply chain impacting private investment as would, say, a similar action against BP. One former government official noted the difference in the supplier base between BP and Barmek explicitly. Recalling Barmek’s lobbying efforts to save its investments, he said the firm failed because it “tried to act independently.”Footnote ⁷⁰

Despite their similarity, BP and Barmek received starkly different government treatment. BP, with a krisha formed by links to local suppliers, benefited from stable property rights, while Barmek, a firm without deep economic links, did not. Barmek sought protection directly from its government relations, but this choice did not compensate for its missing krisha. BP, despite being in a vulnerable position in an oil-rich state, benefitted from economic ties in a way that was recognized by foreign executives both in and out of the energy industry.

Clearly, BP and Barmek are not exactly the same on all characteristics except their economic links. Nonetheless, they are useful cases because they are extremely similar on many relevant dimensions—including their importance to the host economy, asset type, and political prominence—and they illustrate the causal logic of our argument (George and Bennett Reference George and Bennett2004). One meaningful difference between the two firms that could support an alternative explanation is that they have different home countries: BP is a British firm, while Barmek is Turkish. If the Azerbaijani government placed a high value on its relationship with Great Britain and little value on its relationship with Turkey, then perhaps firm nationality, rather than economic links, can explain why BP had a better outcome than Barmek (Ramamurti Reference Ramamurti2001; Wellhausen Reference Wellhausen2015b ). While we might expect Great Britain to have more influence than Turkey in many regions of the world, the opposite is true in the Caucasus region: Azerbaijan cared far more about its relationship with Turkey than its relationship with Great Britain.

Azerbaijan and Turkey are linked so tightly by their language and culture that they are often described as “brother” countries, and some prominent politicians have even proposed unification (Olson Reference Olson2002). Since its independence in 1991, Azerbaijan has cultivated a very close relationship with Turkey, in part to offset Russian and Iranian influence. Azerbaijan and Turkey have strong national security ties: Turkey provides military training to the Azerbaijani army, and Turkey supported Azerbaijan during its 2001 territorial dispute with Iran (Kamm Reference Kamm1994; Frantz Reference Frantz2001). Similarly, Azerbaijan and Turkey have strong economic ties. Shortly before the Barmek expropriation, Turkey signed a multiyear contract to buy Azerbaijani natural gas and completed construction of the Baku-Tbilisi-Ceyhan pipeline, which allows Azerbaijan to transport its oil to Turkey for sale in European markets (Ismailzade Reference Ismailzade2005; Kalicki Reference Kalicki2001). Turkey and Azerbaijan were also building a similar pipeline for the Shah Deniz gas field and planning a railway network for passengers and goods (Ismailzade Reference Ismailzade2005). More generally, Turkey is a major market for Azerbaijani exports. In the five years preceding the government’s breach of its contract with Barmek (2001–2005), Azerbaijani exports to Turkey steadily increased and had an average annual value of US$143 million (COMTRADE). The quality of the Turkish-Azerbaijani relationship has varied over time—largely in response to Turkey’s evolving foreign policy towards Armenia—but its importance has not changed.

In contrast, Great Britain has a relatively weak relationship with Azerbaijan. The political relationship between Azerbaijan and Great Britain has been strained by the European Union’s response to the Nagorno-Karabakh conflict. Western influence on Azerbaijani politics comes primarily from the US, not from Great Britain. And while Great Britain is a major provider of foreign investment via BP, it is not a significant export market for Azerbaijan. During 2001–2005, Azerbaijani exports to Great Britain steadily decreased and had an average annual value of only US$13 million, less than one-tenth of the value of Azerbaijan’s exports to Turkey (COMTRADE). Thus, if firm nationality played a key role in Azerbaijan’s decision to break contracts, we would expect a British firm to be more vulnerable than a Turkish firm. The fact that Azerbaijan broke its contract with Barmek and not with BP suggests that the bilateral ties between Azerbaijan and Turkey were not sufficient to offset the Turkish firm’s otherwise heightened political risks.