A strong body of research exists following the rational design approach that links the design of institutions to the structure of the problems they are created to solve.Footnote 1 One of the most common of these problems is governments' inability to make credible commitments to other states and nonstate actors when their incentives to abide by those commitments are inconsistent over time. International institutions can help governments make their commitments more credible by incorporating features that increase the reputation, audience, and material costs of commitment violation. However, even when comparing across institutions created to solve the same credible commitment challenges, critical features vary in institutional design. This suggests that factors beyond problem structure can also shape how governments design international institutions. One major factor, particularly in the case of credible commitment institutions, is the time horizon of governments.
A government's time horizon—how far into the future it expects to govern—matters to institutional design because the features of institutions that generate credibility also generate sovereignty costs for governments by constraining the freedom with which they can develop economic and social policies, which can be integral to their constituents' welfare and to their efforts to remain in power. Governments will submit to such constraints only if they feel the benefits of enhanced credibility outweigh the drawbacks of sacrificing policy autonomy. However, once a government has signed up to an institution, political and economic conditions can change, so the institutional commitments it agreed to become much less beneficial and potentially costly. At such points in time, governments would prefer to have greater autonomy to develop policy responses to the changing conditions. When designing an institution, governments with longer time horizons will be more concerned by the potential for conditions to change unfavorably because they anticipate being in power in the future and having to govern in the presence of shifting circumstances. Therefore, governments with longer time horizons, while endeavoring to design institutions with features that will make their commitments more credible, are also more likely to seek designs that afford them greater policy autonomy to respond to a changing environment.
I develop and test this argument in the context of the design of bilateral investment treaties (BITs), the most widespread and powerful international institutions governing foreign direct investment (FDI). Because BITs largely seek to overcome the same underlying credible commitment problem that governments have toward foreign investors, there are many similarities in their design. However, in some treaties governments have built in areas of greater policy autonomy. The most notable instance of this is when governments, while committing to afford foreign investors equal or better treatment than domestic investors (national treatment), carve out areas of policy in which they can deviate from this commitment. Viewed through the lens of legalization, these policy carve outs scale back the legalization of obligation in BITs' national treatment commitments.Footnote 2 Because carve outs afford governments greater autonomy to respond to shifting circumstances, I argue that national treatment obligations in BITs should be less legalized, with a greater number of carve outs, when concluded by governments that have longer time horizons. Through analysis of an original data set of the number of national treatment carve outs in a random sample of 342 BITs, I find a strong positive relationship between the time horizons of governments that are likely to be net importers of FDI and the number of national treatment carve outs in the treaties they sign. This relationship holds across regime types and is robust to controlling for selection into treaties.
The article shows that while problem structure does explain certain aspects of institutional design, government time horizons also shape design preferences and outcomes, particularly with respect to design features that promote governments' abilities to respond to changing circumstances. This is significant because while scholars have acknowledged that time horizons may play a role in framing governments' institutional design choices, they have failed to theorize or test this role fully. Moreover, rationalist theories of institutional design have been criticized for implicitly assuming excessively long time horizons. I address this critique by treating time horizons as a variable, and explaining how and why variation in the length of a government's time horizon can systematically affect its approach to the design of an international institution. The article's robust findings provide evidence that scholars interested in explaining international institutional design would do well to consider time horizons' role in shaping governments' design preferences, particularly when focusing on features that can afford governments greater or lesser policy autonomy to respond to changing conditions in institutions where enhancing the credibility of commitments is a core objective. With respect to BITs, this article challenges the common assumption that all BITs are analytically equivalent.Footnote 3 Leveraging original data, I show that the design of one of the most central aspects of BITs—the commitment to national treatment—varies systematically in ways that are economically and politically significant. These findings not only further scholarly understanding of BIT design, but they should also provoke careful reconsideration of approaches, based on assumptions of homogeneity across treaties, that are employed in studies of the legal, political, and economic causes and effects of BITs.
BITs and National Treatment: Credibility Versus Autonomy
An actor encounters the problem of making credible commitments when it has incentives to renege in the future on commitments made in the present.Footnote 4 When other actors are aware of this, it can prevent beneficial cooperation and exchange. This problem is manifest in the relationship between multinational enterprises (MNEs) and host governments.Footnote 5 Governments wish to attract FDI because it is a potential source of employment, tax revenue, technology, and foreign exchange. To better attract MNEs, governments promise secure and nondiscriminatory investment environments and offer concessions such as tax holidays and minimal performance requirements.Footnote 6 However, such promises are not credible because FDI is often characterized by irreversibility and sunk costs—after an MNE has located its operations in a particular country, it is costly and difficult for it to relocate elsewhere.Footnote 7 Aware of this, governments face incentives to move away from the initial promises they made to foreign investors and reshape the policy environment for foreign firms in a manner that extracts greater benefits from MNEs' operations for themselves and their supporters.Footnote 8 Common policy changes include direct expropriation or nationalization of foreign-owned assets and changes to performance requirements, capital taxation and regulation, tariffs, and fees.Footnote 9 The dilemma for governments seeking FDI is that MNEs are aware of the potential for government policy to shift ex post and therefore will refrain from investing.Footnote 10 Thus, governments have an incentive to find a way to make their commitments more credible to MNEs.
Any solution to a credible commitment problem must alter decision makers' ex post incentives—by raising the benefits of adhering to commitments and/or raising the costs and difficulty of breaking them—such that they are more likely to abide by their commitments in the future. This manipulation of incentives is frequently accomplished through institutions, particularly legalized institutions.Footnote 11 In the case of FDI, governments have turned to BITs as an institutional solution to the credible commitment problem.Footnote 12 Indeed, in the absence of a global investment institution and with more than 2,850 BITs concluded since 1959, BITs have become arguably the most widespread and important international institutions governing FDI.Footnote 13 While other international instruments exist to help regulate MNE-host government relations, BITs go further in terms of their geographic scope, substantive breadth and the capacity for direct enforcement of commitments afforded to investors. For example, the agreement of the World Trade Organizations (WTO) on Trade Related Investment Measures (TRIMs) addresses only trade-related aspects of FDI and, unlike most BITs, the WTO does not afford investors direct access to dispute-settlement procedures. Furthermore, unlike the Energy Charter Treaty, BITs' coverage is not limited to a particular sector or industry. While private contracts offer a potential alternative to BITs for managing government-MNE relationships, a contract is far more limited in scope because it governs only those activities pertaining to a single investment relationship, and direct investor-government contracts are uncommon for smaller-scale investments or projects undertaken by small- and medium-sized firms, which make up the majority of MNEs operating in the world today.Footnote 14 The high institutional profile of BITs in the area of FDI is reflected in the growing number of studies that have sought to establish their effects on investment flows, and while these effects continue to be debated, several recent, methodologically sophisticated studies have found that BITs do significantly increase investment.Footnote 15
BITs contain several design features that help make governments' commitments to afford investors a favorable investment environment more credible. First, they contain a series of obligations of favorable treatment to foreign investors including, among others, commitments to treat foreign investors from their treaty partners in a fair, equitable, and nondiscriminatory manner, and to pay appropriate compensation in cases of expropriation. Crucially, these obligations are highly legalized, that is, they are legally binding to a high degree.Footnote 16 The high legalization of obligations undertaken in treaty commitments is important because it increases their credibility by raising the costs and difficulty of ex post commitment violation.Footnote 17 For example, failure to abide by treaty commitments undertaken formally under international law is costly because states suffer reputation costs that affect all of their future conduct within the international law regime.Footnote 18 In the case of BITs, these reputation costs can be localized: if a state frequently reneges on its BIT commitments, it will gain a reputation among investors for not respecting its international legal and investment treaty commitments. Consequently, the state's BITs will be much less effective in generating investment. Furthermore, formal legalization of international commitments makes them public and raises their profile domestically. This creates domestic compliance constituencies—groups that benefit from the commitment and/or adjust their expectations and behavior around it—that can sanction governments for reneging on their commitments by, for example, withdrawing financial or electoral support.Footnote 19 In the case of BITs, such audience costs would likely emerge from domestic firms and workers who rely on foreign investment for business and employment and fear that their government's failure to adhere to its commitments to foreign investors will cause foreign firms to refrain from investing further in their country.Footnote 20
The credibility-enhancing effect of high legalization of obligation in BIT commitments is reinforced by other BIT features. In particular, most BITs afford investors direct access to binding international arbitration if they feel their legal rights have been violated by a host government, thus offering investors a pathway through which they can ensure that commitments undertaken in BITs are enforced. Moreover, these arbitration panels have authority to award compensation to MNEs if governments are found guilty of breaching their treaty obligations. In addition, investment treaties typically have lengthy duration and grandfather clauses that ensure treaty obligations continue to apply to investments even if a treaty is abrogated by a host government, thus making BIT commitments more credible over the long term.
While BITs help to make governments' commitments more credible by making it more difficult and costly for governments to deviate from them, they do so at the cost of policy autonomy because treaty commitments narrow the set of policies governments are able to adopt that affect the activities of MNEs.Footnote 21 This curtailing of governments' autonomy to determine policy by international institutions is often referred to as sovereignty costs, and it can hinder governments' attempts to maintain domestic political support and stay in power.Footnote 22 The political and economic significance of constraints on policy autonomy is particularly evident in the case of the commitment to national treatment in BITs, which has been described as “perhaps the single most important standard of treatment enshrined in international investment agreements.”Footnote 23
The principle underlying national treatment is that a host government must “make no negative differentiation between foreign and national investors when enacting and applying its rules and regulations.”Footnote 24 This commitment helps to ensure MNEs have an even playing field with local firms in a host country. However, national treatment necessarily prevents governments from affording local firms better treatment than their foreign counterparts, which can hinder a range of policies and objectives that governments may wish to pursue. Most directly, national treatment commitments limit the policy measures that governments can adopt to privilege domestic firms and protect them from competitive MNEs in the domestic market. For example, the Canadian government's attempts to protect domestic chemical management firms against competition from U.S.-owned firms have fallen foul of its national treatment commitments under the investment chapter of the North American Free Trade Agreement (NAFTA).Footnote 25 National treatment commitments also insulate foreign investors from any policies that are specifically targeted at them and that diminish the returns on their investment. For example, particularly in developing countries, governments may want to implement performance requirements targeted exclusively at some or all foreign firms to promote the transfer of advanced technologies, the location of production in certain underdeveloped territories or to employ domestic inputs in their production to promote employment and economic development.Footnote 26 In the likely case that such performance requirements are not imposed on local firms, they are a violation of national treatment. Finally, legalized national treatment obligations also constrain policies geared toward social and environmental objectives. For example, governments may wish to offer subsidies to local firms to develop more environmentally friendly technologies (as the Danish government did in the electricity sector) or to discriminate in favor of firms owned by previously oppressed domestic social groups (as the South African government does to promote greater equity in the involvement of Black South Africans in the national economy).Footnote 27 Such measures, however, also potentially contravene national treatment commitments in BITs.
In sum, while governments want to attract foreign investment and its attendant economic benefits by making commitments such as national treatment more credible through BITs, doing so causes such commitments to more tightly constrain the space governments have to implement policies in pursuit of economic and social objectives that are potentially important to their domestic constituents.
Carve Outs: Building in Policy Autonomy
The legalization of international obligations can vary considerably. At the high end of the legalization spectrum, international treaties contain legally binding commitments that are expressed in unconditional terms, without reservation. At the other end are international agreements containing informal and explicitly nonbinding commitments. Between the two extremes, one of the most common ways in which governments can limit the degree of legal obligation is by making obligations contingent, including escape clauses or allowing for states to file reservations.Footnote 28
In the case of national treatment, some governments have concluded treaties with highly legalized national treatment obligations. In such cases the commitment to national treatment is stated in general terms, without any limit or exception, causing governments' policy options to be highly constrained because they are bound, in all circumstances, to treat foreign investors at least as well as they treat local investors. In contrast, some governments have sought to scale back legalization and build in greater policy autonomy through the use of carve outs.Footnote 29 A carve out exists when both treaty partners exempt certain activities or categories of policy such as acts taken in the interests of national security and public order, protection of health or the environment, or taxation from some or all treaty obligations. When applied to the national treatment commitment, carve outs afford governments greater policy autonomy by allowing them to deviate from their commitment to nondiscrimination between local and foreign firms in a number of significant policy areas. However, by creating space for legitimate deviation from national treatment, carve outs also diminish the credibility of the overall commitment to nondiscrimination. Thus, as legalization of obligation promotes credibility, scaling back legalization weakens it.
Figure 1 illustrates that the number of national treatment carve outs distributed in a random sample of BITs is not uniform. Although the majority of investment agreements with national treatment provisions contain no carve outs, just less than half of the treaties in the sample (49.5 percent) contain one or more carve outs. While carve outs may often be few in number, their significance for policy autonomy is still considerable because most cover a very broad area (for example, taxation, environment, security), which encompasses a large number of policy issues and activities. This makes the presence of even a single carve out significant for policy autonomy, and in practice governments have been able to use carve outs to pursue policies that otherwise contravene their broader BIT commitments, including the commitment to national treatment. For example, Argentina has successfully invoked a security and public order carve out in its BIT with the United States to exempt from its treaty obligations several actions it took during its 1999–2000 financial crisis.Footnote 30 In another example, Ecuador successfully deviated from its commitment to national treatment through a taxation carve out in its BIT with Canada when it halted value-added-tax refunds to the Canadian oil company EnCana.Footnote 31
FIGURE 1. Distribution of the number of national treatment carve outs in a sample of 279 treaties
The variation one observes in the number of carve outs in Figure 1 indicates that governments choose to pursue credibility and policy autonomy differently in the design of their national treatment commitments by including carve outs to differing degrees. Why do governments' preferences over the degree of legalization of their national treatment commitments diverge in this manner? The answer lies in the important role that time horizons play in shaping governments' willingness to forego policy autonomy.Footnote 32
Time Horizons and Preferences over Legalization of Obligation
Several scholars have identified time horizons as a critical factor shaping government policy preferences. Much recent scholarship has focused on explaining leaders' willingness to trade off policies that yield short-term benefits with those that yield longer-term gains.Footnote 33 In the international context, proponents of cooperation theory have stressed the importance of the shadow of the future in helping states to cooperate with one another.Footnote 34 Although there is comparatively little direct investigation into the role of time horizons in institutional design, they are a potentially important explanatory factor because governments with longer time horizons are more likely to care more about the long-term effects of institutions than governments with short time horizons.Footnote 35
The discussion here focuses on governments, not states, because while in a formal, legal sense states create and join international institutions, in practice institutions are designed and created by the governments of the day.Footnote 36 Furthermore, because governments must formulate and enact policies, and rely on policy to increase their chances of political survival, they experience the benefits of credibility through legalization as well as the costs of institutional constraints on policymaking autonomy. Governments typically pay these costs at moments when they wish to pursue policies that contradict their institutional commitments. One might argue that such costs are likely to be small considering that governments will design and join only those institutions in line with their prevailing policy preferences.Footnote 37 However, circumstances can change over time causing governments' preferences over policies to shift in a direction that is incongruent with prior commitments.
The potential for circumstances to shift is present in many areas, and particularly so in the case of FDI and BITs.Footnote 38 Governments cannot be certain how their national economies will perform in the long run, how important domestic producers will fare or what the long-run effects of foreign-owned firms on the local economy and industries will be. Governments can also encounter unforeseen crises and developments.Footnote 39 Argentina's recent experience is illustrative. To attract greater foreign investment in public utilities in the early 1990s, the Argentine government promised foreign investors a tariff rate indexed to the U.S. dollar. A severe financial crisis that began in 1999 caused the peso to devalue significantly and led the government to abrogate its tariff commitments to avoid natural gas and other utilities becoming too expensive for local homes and businesses.Footnote 40 Foreign investors have interpreted this as breaches of commitments made under Argentine BITs and have instituted a raft of arbitration cases against Argentina with compensation claims totaling more than $17 billion.Footnote 41
A government's time horizon shapes how concerned it is by the potential for circumstances to shift in a manner that generates strong incentives for it to pursue policies prohibited by its prior institutional commitments. In political terms, a government's time horizon is a function of how long into the future it expects to govern. If a government anticipates losing power in the near future with little prospect of returning to power, its time horizon will necessarily be short. In such cases, a government will be less concerned about having to confront changing circumstances. Any political and economic costs that are incurred in the future when institutional commitments limit policy responses to changing conditions will be paid by future governments. On the other hand, a government that expects to be in power well into the future has a longer time horizon and anticipates being in power when circumstances shift. Indeed, the further into the future a government is in power, the more likely it is that it will have to contend with economic and political shocks, and more gradual changes. Therefore, a government with a longer time horizon will be much more concerned about institutional constraints on its autonomy to develop policy responses to changes in its environment. It follows that governments with longer time horizons will be more likely to limit the legalization of their institutional obligations to preserve some autonomy to pursue policies that deviate from their broader institutional commitments, if such policies are expedient or necessary to respond effectively to shifting circumstances. In the case of national treatment in BITs, this means governments with longer time horizons should seek more carve outs for key policy areas in order to give them greater space to respond to changes free from their commitment to nondiscrimination.Footnote 42
However, limiting legalization through more carve outs diminishes the credibility of the overall national treatment commitment. Why are governments with longer time horizons willing to accept this diminished credibility? Governments do not value credibility for its own sake. Rather they seek to make credible commitments to receive certain political and material benefits that a credible commitment brings; in the case of national treatment and BITs, the expected outcomes include protection for investors, increased investment, growth, and employment. If a government finds in the future that circumstances have changed such that its national treatment commitment is not producing sufficient benefits for its constituents, it would prefer to have greater autonomy to pursue discriminatory policies if necessary to meet the demands and needs of its supporters without triggering the material, reputation, and audience costs that accompany violating a BIT. As a government with a longer time horizon anticipates encountering changing conditions, which may cause the benefits of adhering to its national treatment commitment to diminish markedly, it will be more likely to promote greater autonomy to deviate from that commitment through carve outs. By contrast, a government with a shorter time horizon does not anticipate confronting such changing conditions and will therefore be more likely to seek to maximize credibility by eschewing carve outs.
One might question why, if a government with a longer time horizon is truly concerned about preserving autonomy to respond to changing conditions, it does not do so by abstaining from legalized commitments such as national treatment or BITs altogether. A carve out enables a government to undertake a formal commitment to foreign investors, while also preserving autonomy to deviate—in a limited number of important policy areas—from that commitment if and when necessary to respond to shifting conditions. However, to abstain from a commitment altogether is to forego all the possible economic and political benefits that could be derived from making a legalized commitment (even if its credibility is diminished to some extent by carve outs). A government is likely to do this only if it perceives no benefit to undertaking the commitment given prevailing economic and political conditions. When this is the case, however, neither a government with a long nor short time horizon will benefit from undertaking the commitment and therefore both will refrain from doing so. Therefore, while some governments do refrain from signing BITs or from signing BITs with national treatment commitments, these decisions are not shaped by governments' time horizons (as the empirical analysis will bear out).Footnote 43
Scope of the Argument and Other BIT Features
As I noted, in addition to the legalization of obligation, other dimensions of BITs such as third-party enforcement of commitments through international arbitration and grandfather clauses also help generate credibility and constraints on policy. Should time horizons also shape preferences over their design? Empirically, there is little variation across BITs in the design of these features with respect to their impact on policy autonomy. In the sample of BITs used in this study, 87 percent permit investors to bring any dispute that may arise under a BIT to international arbitration. Moreover, data collected by Peinhardt and Allee show that fewer than 2 percent of BITs require firms to go through domestic courts before gaining access to international arbitration.Footnote 44 With respect to withdrawal from BIT commitments, in this sample of BITs, 95 percent of BITs have an initial lock-in period of ten years or longer, meaning that the treaty remains in force for a minimum of ten years from the date of entry into force. Furthermore, 95 percent of BITs in the sample provide for commitments to be grandfathered in following treaty termination for a period of ten years or longer. This means that even when a treaty is abrogated or altered, the provisions of the original treaty apply to all investments made while that treaty was in force for ten years or more into the future after the date of termination. The broad scope of issues that can be brought to international investor-state arbitration, and the difficulty of withdrawing quickly from treaty commitments in the vast majority of BITs, serve to limit government policy autonomy significantly since they help ensure that government policy commitments taken in BITs will be enforced and that those commitments will bind governments for many years into the future even if a government wishes to withdraw from them.
What explains the relative uniformity in the design of arbitration and withdrawal clauses and why have governments with longer time horizons not sought to build in autonomy in these dimensions of BITs? Borrowing from the rational design perspective, the most plausible explanation is that these features are critical to solving the credible commitment problem BITs are created to address. A key reason why governments struggle to make credible commitments is that foreign firms often have no direct and effective method of enforcing those commitments. This is overcome through dispute-settlement procedures that enable MNEs to take governments directly to independent international arbitration where arbitrators are empowered to issue binding financial awards. Arbitration proceedings also play a pivotal role in helping to generate the reputation and audience costs associated with the legalization of obligation because an arbitration panel's finding that goes against a host state is an objective and often high-profile signal to international and domestic audiences that a host government has broken its formal, legal commitments to a foreign investor.Footnote 45 Another core challenge to credibility is that states often experience government turnover, bringing to power governments with different preferences who may wish to revoke or alter commitments made by previous governments to foreign investors. BITs help to counteract this and create long-term credibility of commitments by grandfathering in treaty commitments for many years. Considering that strong dispute-settlement provisions and grandfather clauses are key to the credibility enhancing effects of BITs, it is unsurprising that governments with long time horizons seeking credibility through BITs have refrained from seeking greater policy autonomy by weakening them.
Beyond the specific context of BITs, when governments face incentives to renege on commitments, an international institution struggles to make those commitments more credible if it does not provide for strong enforcement provisions or it allows governments to easily withdraw from or alter those institutional commitments. Therefore, governments with longer time horizons seeking to both make credible commitments through an international institution while also preserving some autonomy to respond to changing circumstances need to build that autonomy into how they structure those commitments, such as in how legalized their obligations are. This insight relates to the distinction between transformative flexibility—flexibility regarding the termination or redesign of an institution—and adaptive flexibility—flexibility within an existing institutional framework.Footnote 46 If transformative flexibility is high, governments are able to easily withdraw from or alter institutions in which they have enshrined commitments made to other actors, and this necessarily weakens the credibility of those commitments considerably. Therefore, for institutions such as BITs that are principally created to overcome a credible commitment problem, one should expect that governments with longer time horizons will seek to build in policy autonomy to respond to changing circumstances primarily through adaptive flexibility rather than transformative flexibility.
Bargaining: Home States and Host States
BITs are often 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.Footnote 47 It is common practice to distinguish between these two treaty partners and refer to them as the home state and host state, respectively. The asymmetric nature of FDI flows between BIT partners means that the host state's policy freedom is circumscribed by greater legalization of obligation to a much greater degree. Moreover, as net exporters of dyadic FDI, home states possess a stronger incentive to protect the interests of their investors abroad, which may make them more amenable to greater legalization of obligation (fewer carve outs) and the expanded protection and benefits it offers to MNEs. Therefore, although national treatment provisions are reciprocal, I expect the relationship between time horizons and preferences over the legalization of obligation outlined here to apply primarily to host state governments.
Given that home states have less reason to be concerned about possible constraints on their own policymaking autonomy imposed by BITs, and that they have an incentive to seek better protections and conditions for their investors abroad, should one necessarily expect the legalization of obligation in national treatment commitments to reflect the preferences of host states when their time horizons are long? The final text of an investment treaty is the product of negotiations between the host and home state, and each state's ability to influence the final draft will be a product of its relative bargaining power.Footnote 48 Insights from bargaining theory highlight that in a bargain between two actors, the longer an actor's time horizon (the less it discounts the future), the better able it will be to move the final bargain closer to its preferred outcome.Footnote 49 Fearon makes a related point arguing that an actor with a longer time horizon cares more about the long-term consequences of bargains reached in the present and will therefore drive a harder bargain, potentially prolonging negotiations.Footnote 50 In the BIT context, a host government with a longer time horizon will have an incentive to drive a harder bargain and drag out negotiations because it knows it will have to govern well into the future under the constraints in BITs it agrees to in the present. Home states will be wary of allowing this to happen because the longer they fail to sign a BIT, the longer their investors go unprotected by any of the BIT's provisions. Thus, all else equal, a longer time horizon furnishes a host state with greater leverage to ensure that its preferences over the legalization of its national treatment commitments are reflected in the final treaty text. Consequently, time horizons are potentially a powerful explanatory variable for institutional legalization and design, capable of explaining both governments' preferences and bargaining leverage.
To summarize, the argument is that limiting the legalization of obligation of national treatment commitments affords governments' greater policy autonomy to respond to changes in political and economic conditions. Governments with longer time horizons are more likely to value this autonomy because they anticipate being in power longer into the future and having to develop policy responses to changing circumstances. Because legalization of obligation in national treatment provisions is limited, and policy autonomy promoted, through carve outs, the core hypothesis emerging from the argument is: BITs will contain more carve outs to national treatment commitments when signed by governments that have longer time horizons. The asymmetric nature of FDI flows between BIT partners means that this expectation is likely to hold primarily in the case of host state governments.Footnote 51
Data and Measurement
Dependent Variable: Legalization of National Treatment
In order to evaluate the hypothesis, the contents of a sample of 342 BITs signed between 1960 and 2006 were hand-coded.Footnote 52 These BITs consist of all treaties signed by the members of a random sample of 3,600 dyads, taken from the universe of dyads in 2006, whose treaty texts are publicly available.Footnote 53 Information on the universe of BITs, the dates of BIT signings, and BIT texts were obtained from the online treaty database of the United Nations Conference on Trade and Development (UNCTAD), Kluwer Arbitration Online's investment treaty database, the U.S. Department of State's website, and WorldTradeLaw.net's trade agreement depository.
To measure the legalization of national treatment obligations, I identified six possible policy categories that can be carved out of national treatment commitments—taxation, security/public order, public health, environment, economic sectors, and miscellaneous—and counted the number that were carved out of each national treatment commitment in the sample of treaties. A tax carve out was coded as present only if taxation or fiscal matters other than or beyond double taxation treaties or other international agreements pertaining to taxation were carved out. Sectoral carve outs apply to particular industries or sectors of the economy and miscellaneous carve outs are those that do not fall into one of the other five categories. The coding approach was holistic and involved examination of the entire treaty, and not just national treatment clauses. Accordingly, carve outs specific to national treatment were included in the count, as were carve outs that applied to the whole treaty or to parts of the treaty that included the national treatment clause since these latter carve outs also apply to the national treatment commitment.Footnote 54
Independent Variable: Time Horizons
A government's time horizon is based on its expectations of being in power in the future and such expectations are shaped significantly by the institutions within which it competes for power. Accordingly, democratic and autocratic governments' time horizons must be conceptualized and measured separately.Footnote 55 In ascertaining a democratic government's time horizon I focus on parties, rather than individual leaders, because in most democracies political organization and competition is centered on parties who contest regular elections. Moreover, policies typically reflect platforms agreed by party members so government policy preferences in democracies are, to a significant extent, the preferences of the main party in government and its time horizon is the critical component of a democratic government's time horizon.
What distinguishes a party with a long time horizon from one with a short time horizon is its degree of institutionalization.Footnote 56 Highly institutionalized parties are those that are widely regarded as legitimate and effective organizations with roots in society, and that are capable of recruiting members and placing them in positions of power.Footnote 57 They are also long-lasting and contain multiple generations of individuals, which causes them to be highly similar to overlapping generations' organizations.Footnote 58 This means that in well-institutionalized parties, the most senior cohorts are nearing the end of their political careers and do not expect to be in government in the future. Thus, they have very short time horizons and little individual incentive to consider the long-run consequences of the decisions they take for government. However, younger generations within the party expect to play a role in politics and compete successfully for power longer into the future and thus they shoulder the costs of any decisions made in the present by the senior party leadership that adversely affects their ability to compete for power and govern effectively in the long term (for example, signing on to highly constraining international agreements). Therefore, these younger cohorts have longer time horizons and strong incentives to offer various forms of support and bribes to senior cohorts (for example, offering or threatening to withdraw legislative or campaign support) to deter them from making decisions that will adversely affect younger cohorts and the party in the future.Footnote 59 In this manner, the longer time horizons of younger cohorts engender longer time horizons in the party as a whole in the case of highly institutionalized parties. Weakly institutionalized parties on the other hand are often short-lived and struggle to recruit younger cohorts, which causes them to have short time horizons.
Party institutionalization is a complex phenomenon engaging legitimacy, societal roots, and organizational strength. To measure institutionalization, I follow Simmons and use party age as a proxy for party institutionalization, because “illegitimate, poorly organized, and ideologically unattached parties are unlikely to persist over time.”Footnote 60 Thus, the older a party is, the more institutionalized one can assume it to be and the longer time horizon it should have. Accordingly, I measure a democratic government's time horizon as the log of the age of the largest party in the governing coalition. In presidential systems, I consider the president's party to be the largest party in the government. Data on party age and type of government (presidential versus parliamentary) are obtained from the World Bank's Database of Political Institutions (DPI).Footnote 61
Compared to democracies, formal opportunities for removing the leadership are typically much more limited in autocracies, where the process of leadership selection is closed and less frequent. This means that autocratic governments seldom have a realistic chance of regaining power once they have been removed. It follows that an autocratic government's time horizon will be closely aligned with its assessment of its likelihood of being removed from power and when an autocratic government perceives that its hold on power has begun to weaken, its time horizon contracts.
Extant scholarship has identified several institutional and economic variables that influence the likelihood of autocratic survival including the presence of legislative institutions and the type of authoritarian regime.Footnote 62 If one assumes that autocratic leaders are sensitive to their regimes' vulnerability to collapse, this knowledge of the factors that affect regime survival can allow one to empirically estimate the probability of survival and use this estimate as a proxy for autocratic governments' time horizons.Footnote 63 I employ this approach and use the predicted probabilities generated by Wright through a survival model of autocratic regime duration, which includes several political, economic, and institutional predictors of regime survival.Footnote 64 A higher predicted probability of survival reflects a strong and stable regime, which should engender longer time horizons in its leaders. By contrast, a lower probability of survival indicates that the leadership has a fragile hold on power, which should be associated with a shorter time horizon.
Control Variables
The “home” state BIT partner is likely a net exporter and the “host” state is likely to be a net importer of dyadic FDI. I identify a host and home state based on the level of economic development with the host state being the treaty partner that has the lower gross domestic product (GDP) per capita.Footnote 65
While time horizons can shape a state's bargaining leverage, I control for further potential sources of bargaining power. Similar to Allee and Peinhardt, I control for relative economic size of BIT partners using a variable called host share of dyadic wealth, which is the host state's GDP divided by the host and home state's combined GDP. The larger the value for this variable, the greater the economic bargaining power of the host relative to the home state. I also follow Allee and Peinhardt and include a measure of the host state's GDP per capita growth.Footnote 66 One could argue that when host governments are faced with low growth, they are more willing to concede to greater constraints on policy autonomy in BIT negotiations as they seek to attract FDI to stimulate their economies. GDP data are obtained from Gleditsch's Expanded Trade and GDP Data v5.0.Footnote 67
It is likely that the age of a democracy and the age of its largest party in government will be highly correlated.Footnote 68 To ensure that the effects of party institutionalization and not regime institutionalization are being captured with the party age variable, I control for the log of the age of the regime (in years) for democratic states using data on regime age from Cheibub, Gandhi, and Vreeland as well as Polity.Footnote 69 I also control for the log of the number of years that a host state has been independent because some have argued that newly independent states are more likely to be sensitive to the sovereignty costs of constraints on policymaking autonomy.Footnote 70 Dates of independence are obtained from the ICOW Colonial History Data Set.Footnote 71
The design of BITs has evolved over time—it is possible that variation in the legalization of BIT obligations is a product of secular trends toward greater legalization of obligation since BITs have become more prevalent, and the global investment climate has become more liberal.Footnote 72 Therefore, I also control for potential trends in the legalization of national treatment over time through an annual time counter. In addition, some studies have asserted that a qualitative difference exists between treaties that developing countries conclude with each other and those they conclude with developed states.Footnote 73 To control for this, I include a dummy variable that takes the value of 1 when the BIT is signed between a country in the developed North and in the developing South.Footnote 74
Finally, I control for the effect of governments' differing needs to enhance credibility through BITs. Several scholars have argued that greater domestic institutional constraints on the executive help make commitments to MNEs more credible so I control for a government's ex ante level of credibility, and thus its credibility needs, through Polity's seven-point measure of executive constraints.Footnote 75 The level of constraints should be positively correlated with national treatment carve outs if credibility needs shape carve-out design.
Results
The main results, presented in Table 1, consist of four negative binomial regression models of the number of national treatment carve outs. To account for possible interdependence of observations caused by the appearance of multiple BITs in the data signed by the same states (but with different partners), two-way clustered standard errors are used with clustering on the home and host states respectively.Footnote 76 Each model contains a different combination of host- and home-state regime types. The sample is split according to regime type because different scales are used to measure time horizons for democratic and autocratic governments.Footnote 77
TABLE 1. Negative binomial regression models of national treatment carve outs
Notes: Two-way clustered (on host and home state) robust standard errors in parentheses.
* p < .1,
** p < .05,
*** p < .01.
In the first two models, the host state is autocratic and the measure of time horizons is the probability of regime survival. The signs on the time horizon coefficients are positive and statistically significant at the 0.05 level. This means that for autocratic host states, as the government's time horizon lengthens (the probability of regime survival increases), it concludes BITs with more national treatment carve outs. For democratic host states, the results are presented in columns (3) and (4) and indicate a positive and statistically significant relationship (at the 0.01 level) between party age and the number of national treatment carve outs in BITs. Because party age proxies for democratic time horizons, these results indicate that older parties with longer time horizons are more likely to conclude BITs with a greater number of national treatment carve outs. The substantive effect of time horizons is also significant. A decrease in the probability of regime survival for autocratic host states from 99 percent to 90 percent causes the expected number of carve outs to drop from 0.76 to 0.24, holding all other variables at their mean and mode values (for dichotomous variables). With respect to democratic host states, the expected number of carve outs when the governing party is a mature thirty years old is one, while this number drops by almost 50 percent to 0.55 when the governing party in the host state is only a relatively young three years old, holding all other covariates fixed.Footnote 78 There is no clear statistically significant relationship between the home state government's time horizon and the number of carve outs. This likely reflects the home state's relatively weaker concern for policy autonomy and the incentive it has to protect its investors in the host state through tighter BIT constraints.
None of the other control variables in Table 1 display a consistently statistically significant relationship with the number of national treatment carve outs across all host regime types. In the case of democratic host states, there is evidence of a trend over time in favor of greater numbers of carve outs; however, there is no support for such a relationship in BITs when the host state is not a democracy. Interestingly, across all models the GDP-based measures of bargaining leverage register a statistically significant estimate on only one occasion, suggesting that nontime-horizons-based sources of bargaining power do not exert a clear effect on the design of national treatment provisions.
Across all models, the coefficients for the host's constraints on the executive are negative and fail to achieve conventional levels of statistical significance. While these results provide little support for the argument that carve outs are calibrated to states' needs for credibility, the results regarding host state age of democracy may do otherwise. The age (or institutionalization) of the host state's democracy exhibits a positive and statistically significant relationship with the number of national treatment carve outs. Several scholars have argued that democracies possess institutions that afford MNEs a credible and secure investment environment.Footnote 79 Therefore, one possible explanation for this result is that democracies' greater credibility applies less for young democracies than older, more-established democracies, causing the former to have a greater need for credibility and to consequently design BITs with fewer national treatment carve outs. However, the results here are only suggestive of such a relationship and further investigation of the ties between democratic institutionalization, credibility, and institutional design could be a fruitful direction for future research.
From the results in Table 1, one can conclude that host state time horizons are a key driver of the legalization of obligation in national treatment provisions and exert an effect that is uniquely robust across BITs signed by democratic and autocratic host states. To test this robustness further, I undertake two additional statistical analyses. First, I estimate two hurdle models of national treatment in BITs. While national treatment is typically regarded as one of the standard commitments found in BITs, a minority of treaties contain no national treatment provisions at all. Therefore, one might be concerned that focusing on only national treatment carve outs represents a truncated analysis and that ignoring whether or not a national treatment clause is present introduces bias into the results. A hurdle model helps to overcome this by jointly estimating a logistic regression model of whether or not a national treatment commitment is present and a truncated negative binomial regression model of the number of national treatment carve outs for those treaties in which a national treatment clause exists. A conventional hurdle model typically distinguishes between zero and any positive count value.Footnote 80 However, it is possible to have a national treatment commitment and no carve outs. Thus, I create a new dependent variable for the hurdle model that takes the value of 0, if there is no national treatment commitment, and the value of the number of national treatment carve outs plus one if there is a national treatment commitment. Thus, the measure takes a value of 1, if a national treatment commitment is present but there are no carve outs. When the dependent variable is set up this way, a hurdle model is appropriate because, in order to get a count of carve outs, a treaty must “jump the hurdle” of having a national treatment commitment in the first place. Results for the hurdle models using the covariates from columns (1) and (2) in Table 1 are presented in Table 2. No models are estimated for BITs with democratic host states since only a very small number of BITs in the sample signed by democratic host states do not possess a national treatment commitment, which makes statistical estimates unstable.
TABLE 2. Hurdle models of national treatment carve outs—autocratic host governments
Notes: Standard errors in parentheses. *p < .1,
** p < .05,
*** p < .01.
In both hurdle models, government time horizons for autocratic host states continue to exert a statistically significant effect in the expected direction on the number of carve outs, but do not have a significant effect on whether or not a national treatment commitment is present. This accords with the earlier discussion of why time horizons shape governments' preferences over carve outs but not their preferences over the inclusion of national treatment commitments. With respect to both the number of carve outs and the decision to include a national treatment commitment, host share of dyadic wealth exerts a negative and statistically significant effect. GDP per capita growth is also negatively and statistically significantly associated with the decision to include a national treatment provision. Given that these indicators of economic leverage do not achieve statistical significance in most of the other tests presented here, these results suggest a potentially complex relationship exists between economic bargaining power and treaty design that future investigation focusing on negotiating dynamics in particular BITs may help to flesh out.
As a final robustness check, I take up the issue of selection. Existing studies have shown that several factors systematically influence whether or not two states choose to conclude a BIT.Footnote 81 Therefore, it may be important to control for selection into BITs to accurately gauge the effect of time horizons on the degree of legalization of national treatment obligations. In this regard, I follow the approach taken by Stone in his study of the design of International Monetary Fund (IMF) agreements by first estimating a partial observability bivariate probit model of BIT signing, modeling the decisions of the host state and home state to sign a BIT separately.Footnote 82 I use the random sample of 3,600 dyads noted earlier to estimate the model; the unit of observation is the dyad-year for the years 1960–2006. Following Stone, when a BIT is signed between the two members of a dyad, the dyad is dropped from the estimation for subsequent years. I then replicate the models presented in Table 1 including marginal predicted probabilities of the home and host state signing a BIT with each other, derived from the bivariate probit model of BIT signing.Footnote 83
In total, I estimate four bivariate probit models of BIT signing that reflect the host-home regime profiles in Table 1. These models include the time horizon, regime type, regime age, years since independence, and economic growth variables used in the main models above, as well as several other covariates that other studies have found to be associated with BIT signing: whether or not the host state has a common law system, whether or not dyad members have diplomatic representations with each other, and whether or not the host state was a former colony of the home state or they share an alliance.Footnote 84 I also include two measures to capture competitive diffusion effects; the total number of BITs one's potential treaty partner has concluded, and the share of countries in one's region with which one's potential BIT partner has already concluded a treaty.Footnote 85
The reestimated models from Table 1Footnote 86 with predicted probabilities of BIT signing are presented in Table 3 and display a very similar pattern of results.Footnote 87 The signs on the estimates for the host state government time horizon variables remain unchanged and are statistically significant at the 0.05 or 0.1 levels. The effect of democratic regime age also remains robust, while the statistical significance of time for BITs with democratic host states weakens.
TABLE 3. Negative binomial regression models of national treatment carve outs (controlling for selection)
Notes: Bootstrapped standard errors in parentheses. Significance levels determined according to bias corrected and accelerated (BC a) confidence intervals.
* p < .1,
** p < .05.
Overall, the results of the statistical analyses provide firm support for the hypothesis I developed because host state government time horizons display a strong, positive, and uniquely consistent relationship with the number of national treatment policy carve outs. This relationship holds across regime types and is robust to controlling for selection into BITs. Because the legalization of national treatment obligations is limited through the use of carve outs, the results also provide support for the more general argument that governments are more likely to limit legalization of obligation and build in policy autonomy to respond to changing conditions in credible commitment institutions when they have longer time horizons.
Conclusion
While some studies have noted in passing that governments' time horizons may influence the design of international institutions, there are few that have sought to theorize and empirically investigate this relationship fully.Footnote 88 This is a significant shortcoming in the extant literature because a failure to appreciate the role of time horizons has been used to critique functionalist explanations of institutional design. One leading proponent of this perspective, Pierson, argues that an institution's long-term functioning can rarely explain its design because politicians have short time horizons and care only for the short-term effects of their decisions and not their long-term consequences. However, Pierson does acknowledge that if governments' time horizons are longer, it is more plausible that governments will care about institutions' long-term effects. It follows, therefore, that “the issue of time horizons should be treated as a variable with real implications for questions of institutional origins and change, and therefore as a subject deserving serious study.”Footnote 89
I take up Pierson's challenge and explain how international institutions can differ when designed by governments with varying time horizons. As a government's time horizon lengthens, it seeks to design institutions that afford it greater policy autonomy to respond to changing circumstances over time. In the context of BITs this is achieved by limiting the legalization of obligation through carve outs. Focusing in particular on the commitment to national treatment and through rigorous analysis of an original data set of carve outs, I find robust empirical support for this argument. Indeed, in the case of the legalization of obligation in national treatment, time horizons prove to be the most robust explanatory factor among a host of other institutional and economic variables.
The primary implications of these findings for future research are twofold. First, scholars studying the causes and effects of BITs need to acknowledge that while there are many important similarities across investment treaties, there are also substantive differences that are politically and economically significant. Differences in BIT design have been largely ignored in the extant literature in political science. However, this study shows that meaningful and systematic differences exist and therefore variation in BIT provisions should not necessarily be treated as random noise. Instead, with both the volume of FDI covered by BITs and the number of international arbitration cases brought under investment treaties growing, scholars should seek to probe further into the reasons behind differences in the design of BITs and their consequences for investment policy and investor activity. While this study takes an important step toward empirically identifying and explaining key design differences in a broad sample of BITs, much remains to be done. In particular, future research that focuses on a small number of theoretically significant BITs with the aim of developing a detailed and comprehensive understanding of the process of negotiation over BIT provisions and the motivations of the government actors involved would add considerably to the existing stock of knowledge.
The second major implication arising from this study is that scholars of international institutional design more broadly, and particularly those who adopt a rationalist approach, need to consider how governments' varying time horizons fit into their theoretical frameworks. Scholars may find it instructive to fully incorporate time horizons into their analyses. The specific theory presented in this article, although developed in the context of national treatment in BITs, is potentially generalizable and may be particularly relevant to studies that are focused on the core issues of credible commitment and sovereignty costs in international institutional design. Even if scholars choose not to explicitly integrate time horizons into their work, my findings suggest time horizons cannot be ignored. Rather, scholars should seek to acknowledge and explain any assumptions they make about governments' time horizons and whether or not such assumptions impose scope conditions on their arguments. Indeed, consideration of time horizons' role can potentially refine leading extant theories of institutional design. For example, several scholars have argued that states will seek more flexible international institutions when faced with uncertainty.Footnote 90 The framework developed in this article would lead one to expect that when flexibility relates to policy autonomy, and particularly autonomy to respond to shifting circumstances, this argument is more likely to hold when governments have longer time horizons, because governments with short time horizons should care less for preserving flexibility and policy autonomy.
Further investigation of the role of government time horizons in institutional design in other institutional contexts will play a crucial role in helping better understand when, as well as how, time horizons shape institutional design preferences. This study has revealed that time horizons shape design features that promote greater policy autonomy to respond to changing conditions by permitting governments to deviate from broader substantive commitments—a form of adaptive flexibility. By contrast, time horizons do not shape transformative flexibility or the broad design of enforcement provisions such as dispute-settlement clauses. This is because of the credible commitment problem that both governments with long and short time horizons seek to solve through BITs, and the importance of limited transformative flexibility and strong enforcement provisions to overcoming this problem. Further research will help to reveal the extent to which this pattern is unique to BITs or institutions focused on overcoming a credible commitment problem. Similarly, it will help identify if and when governments with longer time horizons may choose to build in policy autonomy through greater transformative flexibility or weaker enforcement provisions.
