Domestic Determinants of International Institutional Design: The Case of Bilateral Investment Treaties. November 2008

Transcription

1 Domestic Determinants of International Institutional Design: The Case of Bilateral Investment Treaties Daniel J. Blake Department of Political Science The Ohio State University November 2008 Abstract Proponents of the rational design school in International Relations posit that international institutions are purposefully designed to help states overcome obstacles such as problems of collective action, asymmetric information, adverse selection and time-inconsistent preferences that inhibit cooperation and promote collectively sub-optimal outcomes. While this functional understanding of institutions can explain important differences in institutional design and development, there remain many cases where functionally similar institutions exhibit subtle but important differences in design and content. This often unexplained variation reflects the literature s tendency to focus more on the problem-solving role of institutions and less on theorizing states preferences over precise institutional outcomes. In this paper, I attempt to partly fill this gap by arguing that domestic political institutions play a central role in determining state preferences over institutions and develop two lines of argument that address: a) the ease with which domestic institutions facilitate credible commitments; b) the institutionally derived incentives that leaders have to ensure a steady flow of public vs. private goods to their supporters. Focusing on the degree of policy flexibility that formal institutions afford governments, I argue that: a) as autocratic governments are less able to credibly commit to future policies, they will form institutions that constrain policy decisions to a greater degree; b) as democratic governments, in contrast to autocratic leaders, need to provide a greater ratio of public to private goods to stay in power, they will be more likely to build in to institutions flexibility for policies that facilitate the provision of public goods, whereas autocratic governments will seek to maintain flexibility for policies that allow them to secure private gains for small groups of core supporters. I evaluate these propositions in the case of bilateral investment treaties (BITs) using an original data set consisting of measures of the degree of policy flexibility in four different investment treatment provisions in a random sample of 324 BITs. (Preliminary Draft: Comments Welcome, Please Do Not Cite Without Permission) This paper was prepared for presentation at the 3rd Annual Meeting of the International Political Economy Society, Philadelphia, November,

2 Introduction The rational design of institutions research program in International Relations has provided scholars with a clear framework through which to understand variation in design features across institutions. This framework is unambiguously functionalist in nature; political leaders create institutions to solve problems. Rationalist explanations typically link the dynamics of a problem - its situation structure - to the institutional solutions that result (Hasenclever, Mayer & Rittberger 1997). Although some progress has been made in recent years, what has yet to be answered definitively in extant research is why institutions that aim to solve the same or similar problems, should be structured differently in different domains. For example, Abbot and Snidal (2000) argue that high levels of institutional legalization will be favored by states in environments where the potential for opportunistic defection is high. However, we see variation in the design of trade agreements where the potential for opportunistic defection is similarly high (e.g. Yarbrough & Yarbrough 1987, Pevehouse & Buhr 2005). Moreover, Kahler (2002), argues that, in the case of global monetary institutions, existing functionalist or demand-side explanations fail to explain a crucial portion of observed institutional variation over time (p.42) because they fail to acknowledge that the preferences of states can vary. Kahler s criticism is well founded in this regard as is his assertion that domestic politics and political economy offer a superior baseline for defining state preferences in functionalist models (p.42). In notable surveys of the field, several prominent scholars have stressed that an important link exists between domestic politics and political institutions and international institutions (Kahler 2000, Martin & Simmons 1998), but very little research has been conducted into the effect of domestic political institutions on the design of international institutions. In International Political Economy, much focus has instead been on the role of domestic institutions in framing foreign economic policy (e.g. Gowa 1988, Lohmann & O Halloran 1994, Mansfield, Milner & Rosendorff 2000, McGillivray 1997, Milner & Kubota 2005). One notable exception is Bernhard and Leblang s (1999) investigation of the effect of democratic electoral and legislative institutions on governments exchange rate mechanism preferences. More recent studies have linked domestic political institutions to the probability of a state joining a preferential trading arrangement and the 2

3 legalization of preferential trading agreements dispute settlement mechanisms (Mansfield, Milner & Pevehouse 2007, Pevehouse & Buhr 2005). However, these studies remain the exception, rather than the norm. This paper takes up Kahler s suggestion to look to domestic politics to determine the preferences of state leaders with respect to the design of international institutions. In particular, I look to domestic political institutions, and primarily regime type, to explain why some states choose to conclude highly constraining bilateral investment treaties (BITs) that limit policy flexibility while others choose to conclude agreements that contain exceptions and loopholes that allow governments greater freedom to influence the environment in which multinational enterprises (MNEs) operate. The government-mne relationship is characterized by time-inconsistency of preferences with governments unable to credibly commit to guarantee investors secure property rights and favorable investment conditions in the long run. Many have argued that the constraining power of democratic political institutions can help overcome this credible commitment problem. Others have suggested that institutions at the international level, in the form of investment treaties can raise the costs governments incur for reneging on past promises to foreign investors and thus enhance their abilities to credible commit to treat MNEs favorably. This suggests that BITs are in a position to substitute for domestic institutions that are not conducive to facilitating credible commitments. Arguments in the extant literature extol the virtues of democratic institutions vis-à-vis the ability of government to make credible commitments (e.g. Jensen 2003) and lead us to expect that autocratic governments seeking to credibly commit, and their treaty partners seeking credible commitments from autocrats, would favor more constraining BITs that greatly limit policy flexibility. The influence of domestic institutions on leaders institutional design preferences extends potentially beyond considerations of credible commitments. Bueno de Mesquita et al. (2003) identify a clear distinction between democratic and autocratic governments in terms of the goods they must provide their supporters to stay in power. Democratic leaders seeking the support of a larger constituency favor the provision of public goods while autocratic leaders maintain power by rewarding their smaller group of supporters with private goods. Recent scholarship has applied this insight to explain trade liberalization and trade volatility in response to leadership turnover (McGillivray 3

4 & Smith 2004, Milner & Kubota 2005) but not to the design of international institutions. I argue that the logic of political survival extends to leaders preferences over institutional design and to the relative degree of policy flexibility and constraint that institutions establish. Specifically, autocratic governments should seek to preserve greater policy flexibility in areas where policy is conducive to the provision of private goods. In contrast, democratic governments should seek to maintain greater control over policy levers that facilitate the provision of private goods. Thus, while entering into an institution such as an investment treaty entails an enhanced level of commitment and constraint on future policy, leaders should seek to temper these constraints in areas where policy tools are most integral to their ability to stay in power. In the sections that follow, I develop these two lines of argument that link domestic institutions to leaders preferences over institutional design, and then, ultimately to the institutions we observe, in the issue area of foreign direct investment (FDI) and BITs. I then test these propositions employing an original data set that contains measures of policy flexibility and constraint for multiple BIT components. I find a robust relationship between regime type and policy flexibility in BITs, however, statistical tests show that the relationship between greater institutional constraints at the domestic level are positively correlated with constraints at the international level. This suggests that BITs establish hard constraints when they are least necessary. However, I argue that these results should not necessarily lead to such a pessimistic conclusion. Furthermore, I find support for the political survival argument as democratic host states are more likely to be sign BITs where there is greater flexibility over tax and fiscal policies which are essential for public goods provision whereas BITs with greater flexibility over national treatment which eases the ability of host governments to privilege local firms are more likely to be signed by autocratic host states. Commitment, Democracy and Foreign Direct Investment The problem of time-inconsistency occurs in dynamic environments where policymakers have incentives to promise one thing and then do another after other actors have made decisions and/or taken actions based on these earlier promises (Maggi & Rodriguez-Clare 1998, Persson & Tabellini 2000). The obsolescing bargain that characterizes much FDI reflects such a situation (Vernon 1971, 4

5 Moran 1995). Under the obsolescing bargain, multinational enterprises (MNEs) are often in a position to choose to locate operations in one of multiple possible locations around the world. Governments have significant incentives to attempt to attract these enterprises as their investments are a potential source of employment, tax revenue and technology transfer. Furthermore, FDI is an important source of foreign exchange in developing countries facing balance of payments problems. Given the potential benefits of incoming FDI, potential host governments are often forced to compete to attract MNEs by offering inducements to investors before they invest in the hope that such benefits will tempt investors to locate in their state. These inducements range from tax holidays and export duty exemptions to investment grants and wage subsidies (Li & Resnick 2003). Thus, before location, bargaining power is with the firm. After entry, however, bargaining power shifts to host state because FDI is often characterized by a relatively high degree of irreversibility and sunk costs (Jensen 2003, Moran 1999, Stasavage 2002). Thus, governments, having made significant concessions in the original bargain with investors, may now be tempted to renege on their promises and establish a new bargain involving policies that direct a greater share of the foreign enterprise s revenue towards the government which it then may chose to redistribute to important constituents. Examples of such policy changes include: changes to performance requirements; direct expropriation, nationalization, and confiscation of foreign owned assets; and changes in capital taxation and regulation (Henisz 2000, Jodice 1980, Moran 1995). Investors, wary of the potential for such changing circumstances may be reluctant to invest. Thus, FDI-seeking governments have an incentive to find a way to credibly guarantee that the original terms of their investment contracts will be upheld. Many have argued that political institutions at the domestic level influence the ease with which governments can make credible commitments and that democratic regimes possess those institutional traits that enhance credibility. The first of those traits is institutional checks and balances that give multiple actors a veto over the executive s policy decisions and thus make policy change more difficult and costly (e.g. Cowhey 1993, North & Weingast 1989, Snidal & Thompson 2003). In this line of argument, the ease of policy reversal is a decreasing function of the number of veto points and democratic leaders typically have to overcome a greater number of vetoes than autocratic lead- 5

6 ers (Leeds 1999). The second trait is the need for democratic leaders to secure the support of a large winning coalition, typically through popular elections, to maintain power. This is because publics tend to disapprove of leaders who back down from previous positions and break their commitments resulting in punishment at the polls (Fearon 1994). Such audience costs are higher in democratic than in non-democratic regimes because democratic leaders are more responsive to public opinion as it is upon that opinion that their political survival rests (Gartzke & Gleditsch 2004, Mansfield, Milner & Rosendorff 2002). These arguments regarding the virtues of democratic institutions vis-a-vis credible commitments have been applied to the domain of foreign direct investment. Nathan Jensen (2003) argues that the institutional checks on policy reversal and heightened audience costs characteristic of democracies impairs the ability of the executive to reverse earlier commitments of favorable treatment towards foreign MNEs. With respect to audience costs, Jensen forcefully argues that, if governments make agreements with multinational firms and renege on the contracts after the investment has been made, democratic leaders may suffer electoral costs because commitment violations scare away further FDI (p.595). He does not deny that reneging on commitments to FDI may be politically beneficial in the short run as increased income flows from policy changes can be directed to core political supporters. However, this is equally likely to be the case in autocracies as it is in democracies. Meanwhile, Li and Resnick (2003) stress that democratic governments are better able to commit to securing the property rights of foreign investors. Drawing on earlier work by North and Weingast (1989) and Olson (1993), they argue that the effective political representation of diverse groups, meaningful legislature, electoral constraints and independence of the judiciary found in established democracies helps them attain strong property rights and uphold contracts, thereby protecting investments from predatory banditry by the executive. It is precisely because democracy itself rests on such institutions that established democracies display high levels of property rights protection. The causal logic underlying this argument falls under the umbrella of the institutionalized veto points perspective as a meaningful legislature and an independent judiciary in this argument act as institutional checks on the executive. Working from the reasonable assumption that ceteris paribus investors will locate in places where 6

7 policy shifts are less likely and property rights are more secure, Jensen, and Li and Resnick find that countries with strong property rights and democratic regimes attract more investment after controlling for other economic and political factors that influence investment location. However, Li and Resnick find find that once controlling for property rights protection, democracies do not attract more FDI. This suggests that the institutional veto points upon which effective property rights rests plays a greater role than audience costs in enabling governments to credibly commit. This is perhaps not surprising because audience costs are a function of issue salience as monitoring government behavior is a costly exercise (Gartzke & Gleditsch 2004). A decision to go to war or escalate an international crisis has high visibility and is salient for most of the selectorate in a democracy and thus there may be, as Fearon (1994) argues, genuine audience costs from backing down from a previous position. On the other hand, citizens will not find it rational to become familiar with the minutiae of investor-state contracts and monitor the behavior of their elected leaders to see if they renege on their contractual obligations. Furthermore, while backing down in a crisis can have audience costs, as Jensen acknowledges there can be domestic audience gains from backing out of commitments to foreign investors. Indeed, many large scale acts of nationalization of foreign assets, from Nasser to Morales, have had a powerful populist dimension to them. Finally, in an issue area as complex as international and contract law, it is difficult for voters to independently determine, in an environment of competing claims by their government and foreign investors, whether or not their leaders have indeed violated previous commitments 1. Thus, at best, the role of audience costs in helping governments to make credible commitments to FDI is limited. However, the effect of institutional checks and balances is still significant. Bilateral Investment Treaties as Commitment Devices International institutions are diverse in their membership, design and objectives and not all institutions are designed with the purpose of enhancing the credibility of commitments. However, the ability of international institutions to perform such a function has achieved broad acceptance 1 This may not always be the case, especially in instances of large scale nationalization and expropriation. However, such acts have been much less common since the 1980s (Kobrin 1984) and smaller violations of investment contracts, or creeping expropriation, which are harder to observe and interpret are increasingly the norm. 7

8 in the institutionalist literature in International Relations. International institutions do this by making the consequences of reneging on commitments more costly to governments. In some cases they do this by increasing the probability that non-compliance with commitments will be detected by clearly specifying standards of conduct and facilitating the monitoring of state activity (Keohane 1984, Oye 1986). This increases the likelihood of detection and the risks that states will incur reputation and audience costs from non-compliance (Dai 2005). Some institutions such as the World Trade Organization s dispute settlement procedure contain a mechanism that allows audience costs to be accompanied by material costs from sanction or retaliation for violating commitments. Furthermore, institutions help facilitate issue-linkage which can raise the costs of defection in one issue area because defection may then be punished through loss of benefits in linked issue areas (Martin 1992). The principal institution(s) governing foreign direct investment is an increasingly dense network of thousands of bilateral investment treaties (BITs). Since 1959, over 2600 BITs have been concluded by more than 175 countries (UNCTAD 2008). Attempts to consolidate these agreements into a single institution have been unsuccessful with states preferring to preserve their autonomy in negotiating with individual investment partners 2. While BITs are concluded between governments, they primarily govern the conduct of governments towards their treaty partners private investors engaged in economic activity within their territory. In terms of issues covered, BITs are remarkably similar with the vast majority outlining government obligations to foreign investors in the following areas (UNCTAD 2007): 1. Promotion of Investment 2. Absolute Standards of Treatment 3. Relative Standards of Treatment (e.g. Most-Favored Nation Obligations) 4. Transfers of Capital 5. Compensation for Losses Due to War and Civil Unrest 6. Conditions and Compensation for Acts of Expropriation 7. Pledge to Honor All Other Obligations to Foreign Investors 2 The abortive attempt led by the Organization for Economic Cooperation and Development (OECD) to develop a Multilateral Agreement on Investment, abandoned in 1997, is the most prominent example of this (Kobrin 1998). 8

9 8. State-State Dispute Settlement 9. Investor-State Dispute Settlement The first seven of these areas serve the institutional function of establishing acceptable standards of conduct. They generally aim to afford investors fair and equitable treatment, protect investor s property rights and promote an environment in which MNEs are free from discriminatory and predatory behavior by host governments. The pledge to honor other obligations is particularly important because it brings private contracts concluded between investors and governments under the protective umbrella of the treaty. The last provision listed, investor-state dispute settlement, is the primary source of increased costs for violating these standards of conduct. Most BITs dispute settlement clauses grant private investors the right to take host governments to arbitration directly (i.e. without the approval or assistance of their home governments) if they feel they have violated provisions of the BIT or, in some cases, if they have any investment related grievance. If the arbitrators find in favor of the investors, they have the power to order offending governments to pay compensation to the investors. The costs of arbitration to governments are real. Legal costs alone are tens of millions of dollars which is substantial for developing countries with small economies. Arbitration awards can also be substantial. For example, in 2003, a Stockholm based tribunal awarded the Dutch-American firm, Central European Media (CME), 350 million dollars in compensation for what it judged was violation of a BIT between the Netherlands and the Czech Republic (Peterson 2003a, Peterson 2003b). This effectively doubled the Czech governments already record budget deficit for that year and led Czech leaders to consider raising the nation s value added tax (Peterson 2003b). Thus the ability for BITs to generate material costs for commitment violating governments is genuine. Furthermore, such large awards and the actions governments need to take to pay them can also have the effect of triggering audience costs as the media is likely to publicize the outcome and constituents are forced to foot the bill. Thus, the costliness and more public nature of arbitration, as provided for in BITs, help to make violation of investment commitments salient to domestic audiences. A final manner in which BITs may raise the costs of breaking commitments to foreign investors is through reputation costs. Todd Allee and Clint Peinhardt (2008a) have argued that arbitration 9

10 of investment disputes, as provided for in BITs, sends negative signals to foreign investors about the investment environment in states that appear before arbitration which results in reduced FDI flows to that state. This dimension of commitment enhancement may be particularly relevant for autocratic governments because, as highlighted above, the effect of audience costs in autocratic regimes is attenuated. The argument that BITs increase the costs of reneging and therefore make commitments to fair and equitable treatment of foreign investors more credible has been made before as has the argument that states sign BITs with a view to attracting foreign investment by credibly guaranteeing to provide a secure investment environment (Buthe & Milner 2005, Guzman 1998, Elkins, Guzman & Simmons 2006). Simmons et al. (2006) go further to argue that the competition to attract FDI between governments has spilled over into the negotiation of BITs such that there is a competitive diffusion of BITs across host states. This suggests that governments, host and home alike, believe that BITs can substitute for imperfect domestic institutions and lock in government promises of favorable treatment towards foreign investors. Institutions, Commitment and BIT design BITs are typically signed by states that have asymmetric FDI flows. Usually one BIT partner is a developed and/or large economy that exports a significant amount of FDI while the other is a developing economy that exports little FDI but rather seeks foreign capital. Thus, based on economic size, development and FDI activity one can often identify the likely exporter of FDI (the home state) and the likely importer of FDI (the host state) in a BIT relationship. Although BIT provisions are largely reciprocal, applying equally to both treaty partners, the asymmetric nature of FDI flows between BIT partners mean that when designing and concluding a BIT the host and home state seek, in practice, to constrain the host state s policy flexibility. This is done by using the provisions in the BIT to make policy shifts, in a wide range of areas that can adversely investment, more costly. The host state seeks this constraint to attract foreign investment while the home state seeks to protect the interests of its investors. The preceding discussion of the commitment virtues of democratic institutions suggests that the 10

11 need to enhance governments capacities to credibly commit are greatest for states with autocratic institutions. Thus autocratic leaders in host states who seek to make credible commitments to foreign investors would seek highly constraining BITs given that their existing policy constraints are limited. Similarly, home governments concluding BITs with autocrats would seek treaty provisions that more tightly constrain policy in order to more effectively limit policies that can negatively affect the profitability of their firms and citizens enterprises. From this argument we can derive the following expectation: H1: BITs in which the host government faces fewer domestic institutional constraints (i.e. autocratic regimes) are more likely to contain provisions that place greater constraints on government policy. Domestic Institutions and the Political Economy of FDI The previous section outlined how the commitment capacities inherent in different domestic political institutions can affect leaders preferences with respect to BIT design. However, the need and ability to establish credible commitments is not the only manner in which domestic institutions can shape preferences over international institutional design in general, and BITs in particular, with respect to policy constraint and flexibility. In this section I present an alternative perspective that focuses on the political economy of inward FDI and leaders desires for political survival. One of the central arguments to emerge from Bueno de Mesquita and colleagues (2003) The Logic of Political Survival (and related works) is that the size of the winning coalition (W) required for an incumbent to stay in power affects the type of goods that it is most efficient for a leader to supply in order to maintain support. Leaders can choose to use their resources to provide public or private goods with the former benefitting all members of society and the latter benefitting only the winning coalition. When W is small, the incumbent leader can maintain its support most effectively by providing supporters with private goods. On the other hand, when W is large, it is more efficient to provide public goods and effective public policy. If we adopt Bueno de Mesquita et al. s plausible assumption that incumbents wish, above all, to stay in power, then they will care about preserving their flexibility and freedom to use policy levers that allow them to generate the appropriate stream 11

12 of goods to their winning coalition. Thus, with respect to international institutions such as BITs that seek to curtail policy freedom, we should expect the following: Leaders of autocratic regimes will prefer to build greater flexibility into institutions for policies that are conducive to the provision of private goods. Leaders of democratic regimes will prefer to build greater flexibility into institutions for policies that are conducive to the provision of public goods. For an institution such as a BIT to be significant, it needs to effectively constrain government behavior. However, as Rosendorff and Milner (2001) argue with respect to the World Trade Organization, office-seeking politicians will seek to balance constraint with flexibility to allow them to attend to the needs of their supporters without rupturing the entire institutional framework. What I suggest here is that the areas of flexibility that leaders seek may, at least in some cases, reflect the kind of policy levers that they find most effective in ensuring political support. I acknowledge that policies often have both private and public components (McGillivray & Smith 2004) and that leaders, regardless of the size of W, will often provide some mix of public and private goods (see Grossman & Helpman 1994). Nevertheless, the relative importance of public and private goods to leaders will vary according to the size of the winning coalition. Public Goods, Private Goods and FDI What type of good is FDI? As noted earlier, inward FDI is often greatly desired by governments, particularly those in capital scarce states in the developing world, because FDI can be a source of employment, technology, foreign exchange and tax revenue. However, much of the literature on the consequences of FDI for the host state, particularly in the short-run, indicates that benefits and costs of inward FDI are typically concentrated in certain sectors of the economy and regions of the host country (Li & Resnick 2003). For example, the net nationwide impact of FDI on employment in a host country can be very small, but if FDI chooses to concentrate in a particular region, employment in that region may rise 12

13 at the expense of other regions 3 (Graham & Krugman 1991). On the other hand, if foreign MNEs displace local firms in a particular part of the host state, and are more efficient than their local predecessors, they not only put local capital owners out of business but can also reduce the demand for labor in that sector and in that region of the host country (Hanson 2001). In addition, it has been argued that efficient foreign MNEs can drive locally owned competitors in the same sector to become more efficient (Li & Resnick 2003)e.g., however, they can also force local competitors to produce at below optimum capacity by reducing their market share and cause them to become more inefficient and to suffer profit losses (Aitken & Harrison 1999). However, while displacing local firms in the same sector, foreign firms may increase employment and productivity in upstream and downstream sectors if they use domestic firms to source intermediate goods and distribute finished products (Gorg & Greenaway 2003, Markusen & Venables 1999). Finally, some have argued that beneficial productivity spillovers from MNEs tend to concentrate in those local firms and industries that have a high quality of physical and human capital and thus are capable of absorbing such spillovers effectively (Gorg & Greenaway 2003, Li & Resnick 2003). This concentration of gains and losses from FDI in sectors and regions causes FDI to take on the character of a private good. The benefits of FDI in terms of employment and technology transfer are concentrated in small groups of firms, individuals and regions. Thus, government policy that eases investment provides a private good to these groups. Likewise, any government policy that seeks to protect or support those firms and workers that are challenged or displaced by FDI provides a private good to these groups. One might reasonably question whether or not there is no public good to be found in FDI, as there is in trade, in the form of a consumer dividend. The existence of such a dividend depends on the nature of inward FDI. If FDI seeks to service local markets, then the greater efficiency of MNEs potentially results in a greater variety of better made products at lower prices for consumers in the host country. In this case, the benefits of FDI are widespread but the costs to local firms that compete unsuccessfully with MNEs are still narrow. Not all FDI is market seeking, however, and much investment into the developing world seeks primarily to extract natural resources or 3 The development of export processing zones in developing countries may encourage such regional concentration. 13

14 use relatively cheaper inputs in manufacturing for export back to the home state and/or other countries 4. In both of these cases, the costs and benefits of FDI are narrowly concentrated. Host State Preferences Over BIT Provisions What are the policy implications of the private nature of FDI? Firstly, democratic governments gain little directly from FDI in terms of public goods creation beyond some limited increases in consumer welfare from market-seeking MNE operations. Therefore, they seek to redistribute and spread some of the gains from FDI that accrue to MNEs and to domestic winners from FDI. They do this through using FDI to generate tax revenue. Recall that governments seek to attract FDI to generate employment, technology transfers and tax revenue. For democratic leaders seeking to provide public goods, tax revenue is the most important of these three benefits of FDI. By taxing foreign enterprises the government channels a share of the profits gained by the firm and its employees, and domestic commercial partners, towards the provision of public goods. Moreover, these public goods can include social welfare programs and labor market intervention that help to compensate domestic losers from FDI. Thus, we should expect the following: H2: BITs in Which the Host State is More Democratic are More Likely to Contain Greater Flexibility for Policies that Facilitate Public Goods Provision and Redistribution (i.e. Tax and Fiscal Policy) Autocratic governments on the other hand, care less for redistribution of the gains from FDI and instead focus on promoting the interests of the small group that makes up their core supporters. Most host states that sign BITs are developing countries and in autocratic developing countries the selectorate from which the winning coalition is chosen frequently consists of the richest individuals, and hence those who own the most capital (Milner & Kubota 2005). Thus, given the propensity of FDI to displace and crowd out domestic capital, there is a high probability that members of the winning coalition will be potentially harmed by competition from highly competitive MNEs. In order to maintain the support of such capital owners, autocrats prefer to have sufficient policy autonomy to grant them special privileges and supports such as subsidies or exemptions from 4 Recall that in most cases cases the host state in BITs are developing economies. 14

15 certain regulations, that will allow their firms to survive and compete. Such protection is clearly a private good and is only effective if granted to local producers and not foreign MNEs. This leads to the following hypothesis: H3: BITs in Which the Host State is More Autocratic are More Likely to Contain Greater Flexibility for Policies that Facilitate Private Goods Provision and the Protection of Narrow Interests (i.e. Policies Privileging Domestic Firms and Producers) In terms of BIT design, such privileging of domestic firms is done by denying foreign firms national treatment (i.e. treatment that is no less favorable than that accorded to domestic enterprises). One might question the need to preserve flexibility over national treatment as governments can simply deny entry to foreign firms that can displace capital owned by important political supporters. This is indeed a possible solution but it may not always be optimal. Firstly, capital owners in upstream and downstream industries as well as those participating in joint ventures with inward FDI may benefit from FDI. Thus, autocratic leaders seeking to provide capital owners in the winning coalition with private goods may not want to deny entry to MNEs as some supporters benefit, but they may also want to protect those domestic firms that are vulnerable, especially if that vulnerability is expected to be short-lived or temporary, as in the case of infant industries. Testing the Arguments: Data and Methodology Dependent Variable: The Degree of Policy Flexibility/Constraint in BIT Provisions The growing literature on BITs has tended to treat them as homogenous in all meaningful respects 5 (e.g. Buthe & Milner 2005, Neumayer & Spess 2005, Elkins, Guzman & Simmons 2006). Moreover, even studies that acknowledge some degree of variation across BITs assert that they are essentially the same apart from provisions that relate to the states pre-consent to investor-state dispute settlement (Yackee 2007). One argument that has been forwarded to justify this approach is that as most treaties contain most-favored nation provisions guaranteeing investors the best treatment a state offers to investors from any other country, the most favorable treaty a state signs applies to all of its BIT partners 5 For a noteworthy exception to the trend of treating all BITs alike see Allee and Peinhardt (2008b). 15

16 (Yackee 2007). This argument is misleading in several respects. Firstly, most-favored nation provisions in treaties only require that the signatories grant investors from their treaty partner treatment no less favorable than they grant to investors from any third state. This does not preclude variation in the level of treatment different states choose to promise foreign investors. Secondly, most-favored nation provisions can include exceptions that are particular to certain BITs. Thus, the promise of most-favored nation may not always be comprehensive. Finally, there often needs to be some degree of situational congruence for most-favored nation provisions to apply. Situational congruence is unlikely to always be present, especially when one considers that firms from different countries will invest in different industries under different contractual terms making straightforward comparisons of the treatment of firms form different countries difficult, and possibly inappropriate. While bilateral investment treaties share the same purpose of creating a secure environment for investment and are undoubtedly similar in structure, they do vary in subtle but important ways. The theoretical discussion above has focused on the relative degree of policy flexibility and constraint in BITs. I argue that these variations in BITs are substantive and have significant implications for the policy flexibility and autonomy that governments have towards foreign investors. In short, governments can and build into treaties policy flexibility along several dimensions and this often takes the form of exceptions to general principles, such as the principle of national treatment. In some cases common provisions may be omitted entirely from the text of a treaty. Measuring Policy Flexibility in BITs I developed an original coding scheme to measure the relative level of flexibility and constraint in several BIT provisions. Development of the coding scheme was guided by the United Nations Conference on Trade and Development s (UNCTAD) (2007) Bilateral Investment Treaties : Trends in Investment Rulemaking, an impressively detailed clause by clause review of trends in BIT drafting and their implications for states and investors. The subset of provisions chosen to be coded relate to the following: 1) Admission of Investment; 2) Most Favored Nation Treatment; 3) National Treatment; 4) Tax/Fiscal Policy. The coding scheme for each of these provisions is presented in tables 1-4, with larger numbers indicating greater policy flexibility for host governments. 16

17 Code Table 1: Admission of FDI Content of Treaty Provision 0 Investors are guaranteed some form of most favored nation and/or national treatment 1 No requirements to admit investment or Admission is contingent on domestic legislation and policy Table 2: Most Favored Nation Treatment of FDI Code Content of Treaty Provision 0 Most-Favored Nation Without Exceptions 1 Most-Favored Nation With Exceptions or No MFN Commitment Table 3: National Treatment of FDI Code Content of Treaty Provision 1 National Treatment Without Exceptions 2 Most-Favored Nation With Exceptions or Contingent on Domestic Legislation 3 No National Treatment Commitments Code Table 4: Tax/Fiscal Policy Content of Treaty Provision 0 No mention of tax or fiscal policy in the context of treaty exemptions for double taxation agreements 1 Tax/fiscal policy exempt from most-favored nation and/or national treatment provisions OR exempt from all parts of the treaty 17

18 Figure 1: Differences in National Treatment and Tax Provisions in a Sample of 324 BITs These four areas were selected for several reasons. Firstly, they all have a direct influence on the freedom host governments have to implement and adjust policies towards foreign and domestic investors and manipulate the allocation of costs and benefits of foreign investment activity. Secondly, they are less open to variations in legal interpretation and stylistic variations than other BIT provisions making attempts at cross-treaty comparison more reliable. The third reason is that there is genuine variation across BITs in these areas, whereas in some others there is little meaningful variation worth investigating. For instance, there is almost universal consensus regarding the acceptable conditions for expropriation are and how investors should be compensated (UNCTAD 2007). Furthermore, the coding scheme is relatively simple and treaty provisions are broken up into easily differentiable categories. Again this facilitates reliable and easy comparisons across treaties written in multiple different languages. A random sample of 324 BITs were coded and figure 1 illustrates the variation in content in two areas: national treatment and tax/fiscal policy. Information on the universe of BITs, the dates of BIT signings and BIT texts were obtained from UNCTAD s online treaty database and the Kluwer Arbitration Online s investment treaty database 6. 6 Kluwer Arbitration s source for the BIT texts is the Penn State Institute of Arbitration Law and Practice. 18

19 The measures for each BIT provision are tested in two ways. Firstly, a combined index of all four measures are examined as a general measure of BIT policy constraint. The index is constructed by normalizing each measure and then summing across all four measures. This general measure is used primarily to test the first hypothesis that links the need to enhancement commitment credibility to BIT design. The second approach uses the national treatment and tax/fiscal policy measures individually as originally coded. These individual measures are employed to test hypotheses two and three regarding public and private goods provision. As indicated above, tax and fiscal policy facilitates public goods provision whilst exceptions to national treatment promotes the provision of private goods Independent Variables In addition to the main independent variable of regime type, several control variables that are common in the literature on BITs and FDI are also included in the statistical tests. Furthermore, variables that attempt to capture and control for the bargaining power and dynamics between the home and host state are also included. While the arguments presented above have focused on the preferences of the host state and, to a lesser degree, the home state, a BIT s design is the outcome of a bargain between the two parties. Thus, it is important to control for factors than can influence the bargaining aspect of the treaty design. The monadic independent variables are those in the model that measure characteristics of the individual members of each dyad. For the monadic variables a distinction is drawn between host and home states. The host state is the member of the dyad with the lower GDP per capita 7. All time varying independent variables are lagged one year. Monadic Independent Variables The host state s regime type is the main independent variable in the model. The home state s regime type is also included. Both are measured using the Polity IV Autocracy-Democracy Score from -10 to 10 (Marshall & Jaggers 2006). Polity s measure of regime durability is also included for the host state. It is the number of years since a significant shift in the host state s regime 7 This practice mimics the approach employed by Simmons et al. (2006) 19

20 type (defined by a 3 point or greater change in the polity scale). Regime durability may effect the design of BITs as states with long-lasting regimes may have well-known institutions on which rest established reputations for the treatment of FDI that therefore reduce the challenge of making credible commitments, and the need for highly constraining BITs. Thus a positive relationship between durability and policy flexibility is expected. Several monadic economic variables are employed in the model including the logarithmic transformation of the host and the home country s GDP per capita. The log of the host country s GDP is used as a proxy for the level of economic development. This is included as a control variable because the level of development can potentially effect the contents of BIT provisions, although the direction of the effect is not clear. For example, a more economically developed host may be less willing to push for a less constraining BIT as they believe that domestic firms are capable of competing with foreign investors and as human and physical capital is developed enough for local firms to derive substantial gains from technology and productivity spillovers. On the other hand, developed economies are desirable investment locations and so home states may be more willing to concede greater flexibility to a more developed BIT partner to ensure that an agreement is concluded that offers some protection to investors. The home country s level of development is also included as a more developed home country may have bargaining leverage because it has a greater degree of technological sophistication and know-how that its firms can potentially transfer to the host state and this may result in more constraining BIT provisions. GDP data is obtained from the Gleditsch (2002) Expanded Trade and GDP Data v5.0 the years and from the World Bank s World Development Indicators for the years A further economic variable is included in the model, FDI outflows as a share of GDP for the home state. This is another measure of bargaining leverage as home states that export larger amounts of FDI should have greater bargaining power with host states that are keen to attract that investment. Thus a negative relationship with policy flexibility is expected. FDI flow data is obtained from the World Bank s World Development Indicators. A variable which is the count of the number of BITs signed in the same region as the host state in the past five years is also included in the analysis. Simmons et al. (2006) find strong evidence of 20

21 the diffusion of BIT signing through processes of competition and learning. This measure attempts to control for the possible affect that such processes may have on the design of BITs. The assumption is that learning and, in particular, competition will be strongest with one s regional neighbors and that as the number of BITs signed in the region grows, this creates greater pressure on states to conclude more BITs. If this has any effect on BIT design, it should be to make BIT provisions more constraining as host states are more willing to concede policy flexibility in order to sign the BIT sooner rather than later. Data on regional membership is taken from Hadenius and Teorell (2005). Dyadic Independent Variables A further indicator of bargaining power is included at the dyadic level. Following measures used in the literature on material capabilities, a variable called wealth ratio is constructed by taking the home state s GDP (corrected for inflation) and dividing it by the sum of the home and the host state s GDP. The larger the ratio of home state GDP to dyadic combined GDP, the greater economic power and leverage the home state has over the host state and the more likely that BIT provisions will be more constraining. Neumayer (2006) has found support that political links between states affects the probability they will sign a BIT. It is possible that such links will spillover and influence BIT design. Therefore two dummy variables for whether or not the home and host state share an alliance and whether or not they have a shared colonial heritage are included in the analysis. Data on alliance membership and colonial heritage are taken from Leeds (2005) and Hadenius and Teorell (2005) respectively. The density of economic links between home and host state may also influence BIT design. Ideally, one would be able to control for the level of bilateral FDI flows. However, the availability of bilateral FDI data is inconsistent at best and there is very little data for dyads involving non- OECD states. Therefore, to measure economic ties, the level of dyadic trade between the home and host country is included using data obtained from Barbieri, Keshk and Pollins (2008). Furthermore, economic interactions typically decline with the distance between actors. Therefore, as more crossborder direct investment is to be expected between members in close proximity (i.e. within the 21

22 same region), there is a possibility that this expectation of dense economic ties will affect BIT design. Therefore, I include a dummy variable which equals one when both home and host state are located in the same region. At this stage, I am agnostic about the nature and direction of any such effect. Again, data on regional membership is taken from Hadenius and Teorell (2005). Method Multiple estimation approaches are adopted as the nature of the dependent variable varies. The combined general index of flexibility in investment treatment provisions is a continuous variable and therefore is evaluated using standard OLS estimation. The individual measure of national treatment is a three point ordinal scale and is analyzed using a generalized ordered logit/partial proportional odds estimation technique (Fu 1998, Williams 2006). The generalized ordered logit is preferred because in this instance, as is often the case, the model violates the parallel lines, or proportional odds, assumption that is necessary when using ordered logit estimation. However, not all variables in the model, including the primary independent variable of interest, regime type, violate this assumption. Therefore, a partial proportional odds model is fitted that employs a parallel lines constraint on those variables that do not violate the assumption but does not do so on those variables that do. For the variables that do not violate the assumption, the model estimates a single coefficient for comparisons between all outcomes. Finally, the dummy tax policy constraint measure is analyzed using a standard logit estimator. In all cases, standard errors are corrected for clustering on the home state. There are two reasons to cluster on the home state. Firstly, there are approximately one third as many home states as there are host states in the sample suggesting a greater potential for correlation of standard errors. Secondly, the most active BIT signing home states often have a model BIT on which they endeavor to base their investment treaties. By working from the same model, there is an increased likelihood of correlation of errors for BITs concluded by the same home state. 22

23 Results Table 5 presents the results for the general index of policy flexibility/constraint which pools together measures for all four types of treaty provision. The results do not offer support for the first hypothesis as while the relationship between host state regime type and the general index of policy flexibility is statistically significant at the 0.05 level, higher levels of host state democracy are positively correlated with policy flexibility. This suggests that bits in which the host state is more democratic, are more likely to have a higher general level of policy constraints. Such a result may be cause for pessimism because it is precisely those governments which are already constrained domestically that submit to tighter BIT constraints. Thus, BITs most constrain host governments when such constraints are least needed. A more positive conclusion regarding the effect of BITs can be drawn if one considers the marginal effect of BITs on states capacities to credibly commit. The commitment challenge is greater for unconstrained autocrats and thus BITs, even though they establish moderately weaker constraints on policy for autocratic leaders, still represent a marked increase in policy constraint for autocrats. Viewed in this light, it is not surprising that democratic governments sign more constraining BITs as they are already more constrained and better equipped to make credible commitments to foreign investors. Thus democratic host states, and home states seeking enhanced commitment from democratic hosts, would design more constraining BIT provisions. Otherwise the marginal effect of the BIT on the host state s commitment capacity would be minimal, and the purpose of the BIT unclear. With respect to other variables in the model, host state regime durability is statistically significant in the direction that we would expect (i.e. older host state regimes negotiate more flexibility into their treaties). Meanwhile, the home state s FDI outflows, which is one indicator of home state bargaining leverage is statistically significant at the 0.1 level and, as expected, a higher level of FDI outflows is correlated with a higher degree of policy constraint. The result for host state regime type presented in the model of the general index of flexibility is mirrored in the model of national treatment (table 6). Once again as the level of host state autocracy decreases and the level of democracy increases, the probability that a BIT will contain 23

24 Table 5: OLS Model of Combined General Measure of Policy Flexibility/Constraint Covariate Estimate p-value Polity (Host State) (0.017) Polity (Home State) (0.055) Regime Durability (Host) (0.006) Wealth Ratio (0.610) LG(GDPpc) (Host) (0.206) LG(GDPpc) (Home) (0.657) FDI Outflows/GDP (Home State) (0.045) Alliance (0.276) Colony (0.585) Dyadic Trade (0.000) Regional BITs (5 yrs.) (Host) (0.001) R N 241 Standard errors are robust to clustering on the home state 24

25 more constraining national treatment provisions increases. Figure 2 illustrates this relationship very clearly. The blue line illustrates the positive relationship between host state democracy and the predicted probability that a BIT will contain national treatment provisions with no exceptions, which is the lowest level of policy flexibility on the coding scale (see table 3). On the other hand the predicted probability of a BIT allowing greater policy flexibility in the form of national treatment provisions with exceptions or no national treatment guarantees at all is a decreasing function of the level of democracy and an increasing function of the level of autocracy (see the red and green lines in figure 2). This relationship is statistically significant at the 0.01 level and provides support for the argument that autocratic regimes will seek to preserve flexibility of policies that facilitate the provision of private goods to important supporters (H3) as exceptions to, and the absence of, national treatment provide host states with an opening to privilege domestic firms and capital vis-á-vis foreign investors. The estimates presented in table 6 also show that the relationship between FDI outflows from the home state and host state regime durability, and policy flexibility identified in the general index model, holds in the national treatment model as the does the relationship between the number of bits signed in the host state s region in the past five years. As the number of bits signed in the recent past in a host state s region increases, the probability that a BIT to which that state is a party, as a host state, will contain promises of national treatment without exceptions increases. This suggests that host states do face diffusion pressures to sign BITs and are therefore potentially more willing to accept greater constraints. The results of the logit model of tax are quite different. All variables except the host state s regime type are insignificant (it is significant at the 0.1 level) but in this model the coefficient is positive. This suggests that in matters relating to tax and fiscal policy, BITs in which the host state is more democratic are more likely to exhibit a higher degree of policy flexibility (i.e. taxation matters do not apply to one or more integral parts of the treaty). This relationship is represented graphically in figure 3 and supports the argument that links democratic governments needs to provide public goods to their preferences to build-in to BITs greater discretion and flexibility in their use of tax and fiscal policy levers (H2). 25

26 Table 6: Generalized Ordinal Logit Model of National Treatment Policy Constraints Comparison: NT Without Exceptions & All Others Covariate Estimate p-value Polity (Host State) (0.018) Polity (Home State) (0.056) Regime Durability (Host) (0.006) Wealth Ratio (0.653) LG(GDPpc) (Host) (0.238) LG(GDPpc) (Home) (0.451) FDI Outflows/GDP (Home State) (0.077) Alliance (0.322) Colony (0.541) Dyadic Trade (0.000) EU Candidate (1.095) Regional BITs (5 yrs.) (Host) (0.001) N 241 Comparison: (NT With and Without Exceptions) & All Others Covariate Estimate p-value Polity (Host State) (0.018) Polity (Home State) (0.064) Regime Durability (Host) (0.006) Wealth Ratio (0.594) LG(GDPpc) (Host) (0.238) LG(GDPpc) (Home) (0.451) FDI Outflows/GDP (Home State) (0.086) Alliance (0.322) Colony (0.541) Dyadic Trade (0.001) EU Candidate (0.959) Regional BITs (5 yrs.) (Host) (0.001) N 241

27 Figure 2: Predicted Probability of a BIT Containing Greater Flexibility/Constraint in National Treatment Obligations Figure 3: Predicted Probability of a BIT Containing Greater Flexibility for Tax/Fiscal Policy 27