Private, transnational governance regimes with the power to write, interpret, and enforce commercial rules have proliferated in recent decades. As Braithwaite wrote, many countries have “become rule-takers rather than rule-makers.”Footnote 1 Because much of the scholarly work on private regulation focuses on the transnational regimes themselves, we know relatively little about the consequences of this changing international institutional landscape for domestic political development. I argue that the growth of private, transnational authority carries with it an implicit model of political and legal development, what I call the “unbundled state,” that cuts against traditional models of political development. “Unbundling” governance refers to the partial delegation of authority—that is, the power to write, interpret, or enforce rules—that has traditionally been “bundled” in centralized public institutions. Under this model, rather than supporting holistic competence building within centralized institutions, states have the option of delegating piecemeal governance tasks to actors with little accountability to domestic publics. One unintended byproduct of the growth of transnational institutions is that countries with weak state capacity may suffer from institutional stagnation and divestment as powerful domestic and foreign actors, who would otherwise have a stake in the strength of domestic institutions, instead make use of transnational substitutes for the same services. The exit of such actors from domestic institutions diminishes the incentives states face to engage in capacity-enhancing reform. I offer a theory for understanding the domestic consequences of the growth of global governance institutions and apply this theory through an empirical analysis of the deleterious effects of international commercial arbitration (ICA) on national courts.
ICA is a widely used system for privately adjudicating international commercial disputes.Footnote 2 Parties typically enter into ICA through contractual provisions stipulating that future disputes will be removed from national courts’ jurisdiction and instead sent to private arbitration. ICA allows disputants to choose the relevant procedural and substantive laws and pick the arbitrators who will hear the case, among other things. Most importantly, an award issued by an arbitration panel can be enforced through national courts almost anywhere in the world. With the expansion of commercial arbitration in the twentieth century, the practice has sparked intense debate over the role of private authority in public affairs.Footnote 3 This debate has taken on added urgency in light of the increasing deference legislatures and judiciaries around the world have granted to arbitration.Footnote 4
Regimes such as ICA have proliferated in recent decades.Footnote 5 Private governance regimes are institutions, often created by private businesses or other NGOs, with the ability to provide services or create rules that other actors follow.Footnote 6 We witness this in a variety of issue areas, including environmental regulation,Footnote 7 financial accounting standards,Footnote 8 and human rights.Footnote 9 On its face, there might be a net gain when governance tasks like contract enforcement are privatized. Perhaps contract enforcement through private arbitration, for example, simply eases the process by which firms involved in international business enforce contracts and settle disputes while no one else is made worse off. But while much of the scholarship in this area focuses on first-order outcomes within each regime's targeted policy domain, the central theoretical claim I put forward here is that the emergence of private authority can have important implications beyond these specific policy domains.
Here I first argue that the implications of unbundling for the underlying public institution—the institution whose services are being augmented or replaced by the private regime—depends on how the public and private bodies interact, specifically whether the private body acts as a complement to the public institution or a substitute for it. Substitution provides actors an “exit option” from the public institution, reducing the incentive for states to invest in maintaining or improving the quality of service. Complementarity, by contrast, implies that unbundled institutions will enhance governing quality because civil society actors, domestically or abroad, remain invested in the quality of the public institution. I then apply this framework to the case of ICA, arguing that it serves as a substitute for national courts.
Empirically, I focus on the consequences of enacting pro-arbitration domestic laws based on the Model Law on International Commercial Arbitration (hereafter, the Model Law) of the UN Commission on International Trade Law (UNCITRAL). The Model Law is a ready-made legislative text incorporating key features of the “state of the art” in ICA which allow parties to use ICA as a substitute for national courts for commercial dispute resolution. The Model Law limits judicial intervention in the arbitral process by severely circumscribing the scope of judicial oversight while at the same time requiring courts to enforce arbitration agreements and awards without substantive review unless one of a very narrow set of exceptions is met. Without such protections, national courts can intervene or obstruct arbitration proceedings. Consistent with my argument that ICA operates as a substitute for domestic courts, I find that enactment of the Model Law has a deleterious effect on the development of domestic legal institutions, particularly in countries with already weak legal institutions.
The main empirical findings presented here are as follows. First, I present difference-in-differences estimates of the effect of enacting pro-arbitration laws on subsequent legal development. I estimate that the strength of a country's legal institutions gradually decays after the enactment of such reforms. This effect is largely driven by countries whose institutions are weak prior to enactment. Disaggregating this estimate, I show that it is characterized by declines as well as unrealized improvements in the quality of domestic legal institutions in enacting countries. I then break down the index I use to measure legal capacity to find that the effect is driven by changes in the strength, predictability, and independence of the domestic judiciary and is unrelated to phenomena such as corruption or embezzlement. I find consistent results when using an instrumental-variables strategy that exploits plausibly exogenous variation in the enactment of arbitration reforms among a country's export competitors in contract-intensive trade. Finally, I present cross-national evidence consistent with an important mechanism of the theory: that arbitration facilitates firms’ exit from the public judicial system. I use new data from the Court of Arbitration of the International Chamber of Commerce (ICC) to show that pro-arbitration reforms increase the use of arbitration by domestic firms, though it is less clear that the law increases arbitration subject to domestic jurisdiction. In sum, these findings are consistent with arbitration's goal of easing the process by which firms can resolve disputes and enforce contracts outside of the judiciary, though I argue arbitration can also have negative spillover effects for a country's broader, public legal institutions.
I seek to contribute to two strands of research within the extensive scholarship on global and private governance in international relations. First, I extend the literature on ICA to examine more directly its relationship with national judiciaries.Footnote 10 In doing so, I bring new evidence to the debate between those who argue arbitration has the potential to act as a boon for domestic legal development and those who warn of negative consequences.Footnote 11 My findings complement work on the domestic effects of globalization,Footnote 12 including the private-governance strategies that firms employ to escape the reach of public institutions or to influence domestic regulatory outcomes.Footnote 13
Second, I build on scholarship on the interaction between domestic and global governance and the growing tensions between globalization and democratic institutions.Footnote 14 The privatization and export of services traditionally entrusted to public institutions is not new. In her account of how states are pushed to transform national institutions to accommodate the demands of economic globalization at the expense of domestic accountability, Saskia Sassen describes the “denationalizing of several highly specialized national institutional orders … [that are] partial and incipient but strategic.”Footnote 15 It is increasingly clear that resource-rich individuals and firms can access these strong, “denationalized” institutions while those without such resources are left to deal with subpar domestic institutions.Footnote 16 I build on this work by offering a framework for conceptualizing the second-order effects of global, private governance on domestic public institutions.
Global Governance and the Unbundled State
While early debate on the relationship between global and domestic governance institutions was often concerned with whether global institutions substituted for or complemented their domestic counterparts, more recent scholarship seeks rather to identify the conditions under which global governance institutions will substitute for or complement domestic institutions.Footnote 17 I aim to build on this growing body of work by contextualizing substitution and complementarity within the domestic institutional environment. Rather than focusing on how a global governance arrangement is (or is not) successful at accomplishing its goals, I examine the consequences global governance has for domestic public institutions.
The framework presented here highlights the importance of focusing on not just whether governance tasks are delegated to transnational institutions, but how. One possibility is that the resulting governance arrangement offers a partial, independent functional equivalent to the public institution that allows for minimal or even no state oversight (that is, a substitute). The availability of such a private “exit option” can harm the capacity of the domestic institution from which the task was delegated because it removes a constituency that would otherwise have an interest in exerting political pressure to maintain some level of quality or demand improvements.Footnote 18 An alternative possibility is that the global institution does not substitute for the tasks that were delegated from the domestic institution but in some way relies on the domestic institution to function well (that is, it is complementary to it). While beyond the empirical scope of this article, I would expect such integration with a transnational authority to sustain or even increase political pressure from interest groups to maintain governing quality and generate positive spillovers within the domestic institution. I discuss each of these possibilities in turn before applying the argument to ICA.
Unbundling As a Substitute
To start, we can think of the typical modern state as composed of largely centralized institutions that “bundle” together a wide set of governance tasks. The judiciary is a prime example of such an institution. Broadly speaking, the same court will hear cases in any number of issue areas. The same judge might sit on a national security case one week and an intellectual property case the next. Even in jurisdictions with distinct commercial courts, such as England and Wales, the judges appointed to the court are part of a broader judicial organization and often sit on other courts hearing noncommercial cases. Such an arrangement allows substantial professional movement and knowledge sharing within bundled institutions.
In addition to intra-institutional knowledge building, bundling provides simple lines of accountability linking task to institution to outcome. Bundling thereby “internalizes externalities,” which helps resolve collective action problems.Footnote 19 Because the policies of a bundled institution affect a wider range of actors, it is easier to identify negative externalities. Bundling thus eases the process of building a coalition for reform, as the policy affects a larger set of actors than it would have if it were implemented by an institution with a smaller task set. This is especially true with respect to legal infrastructure. Bundled legal institutions enhance public accountability by developing and applying broad principles to disparate cases, reducing the risk of contradictory rules forming in different issue areas.
Global governance arrangements are often much narrower by comparison. Delegation to modern global governance regimes thus tends to be partial with respect to the domestic institutions from which some governing task was delegated. Such partial outsourcing risks undermining the positive externalities of centralized, public institution building. In the context of human rights and legal development, for example, Milli Lake shows that, particularly in states with weak legal capacity, international NGOs can improve legal accountability for gender violence.Footnote 20 A potential problem arises, however, when an outside authority substitutes for or bypasses the public institution it is meant to augment. Lake discusses the possibility that NGO involvement may erode the judiciary's connections with local populations. This could concentrate the judiciary's attention on issues prioritized by the NGO at the expense of other issues that may be important to the local population. It could similarly reduce the state's incentives to invest its own resources in capacity-enhancing legal reform. While Lake finds that substituting for the state was a success with respect to the NGO's first-order goals, it could still have second-order effects that “undermine rights in other areas.”Footnote 21
We see this dynamic play out in the area of public security as well. Anna Leander argues that the international market for security forces undermines investments in public security forces, particularly in weak states.Footnote 22 Privatization leads to what she calls a “Swiss cheese” security environment, characterized by isolated pockets of stability where there exist funds to support it (such as areas where international NGOs or multinational corporations operate), further diminishing the incentives to commit public resources to enhancing the public security forces capable of bridging these gaps. Recent scholarship has also found a similarly corrosive effect of the growth of international credit markets on domestic fiscal capacity in weak states.Footnote 23
While the logic here is similar to that behind the potentially harmful effects of aid dependence on institutional outcomes,Footnote 24 my theoretical and empirical focus is more targeted. Rather than arguing that transnational governance can harm the quality of domestic institutions in general, I argue that unbundling risks undermining the broader functions carried out by the specific domestic institution that has been partially outsourced.
Unbundling As a Complement
Unbundling is not necessarily harmful to domestic institutional development, however.Footnote 25 Where private authority does not substitute for domestic institutions, domestic capacity can be maintained or even enhanced because the private authority to which some task is outsourced still depends on domestic capacity or engagement to succeed. We can see the potential for positive externalities in areas of public–private governance that promote complementarity, such as efforts to regulate the global timber trade. Bartley argues that the success of private, transnational timber regulation hinges largely on the degree to which it operates through domestic institutions and laws.Footnote 26 While some governance has been delegated to global governors, the private regulatory regime retains a stake in the capacity of the domestic institutions with and through which it operates. The logic presented here is that this form of public–private symbiosis would also have positive spillovers for the broader domestic environmental agency tasked with cooperating with and regulating the transnational timber regime. For example, public investments in the capacity to regulate timber can spill over into other areas within the agency's broader ambit, such as air or water pollution. Such positive externalities come from the fact that the transnational institution still depends on the operation of its domestic counterpart; it remains a “client” of the domestic institution.
Transnational anticorruption efforts provide another example of the capacity-enhancing potential of global governance. While debate continues within this literature on the consequences of the growth of transnational antibribery enforcement for the development of domestic capacity,Footnote 27 the framework presented here would predict an increase in domestic capacity, given the reliance of transnational actors on domestic law enforcement for investigatory assistance and illicit-payment detection.
To summarize, institutions can be thought of as bundles of tasks and authorities. Increasing economic interdependence has put pressure on state institutions to partially delegate authority to private transnational authorities. As private, transnational institutions increasingly take over tasks that were previously bundled in more general public institutions, rule-making authority becomes more diffuse, decentralized, and complex. Increased complexity risks entrenching the power of well-resourced actors, while the export of governance authority risks undermining political and legal development incentives. I expect therefore that high interdependence or complementarity between private and public bodies should have an enhancing effect on domestic institutions. Whereas I expect substitution to weaken interdependence and have a stagnating effect. In the next section, I apply this framework to the case of ICA and argue that, by substituting for courts, arbitration reduces powerful commercial actors’ reliance on the judiciary, to the detriment of the broader judiciary.
ICA As an Unbundling Institution
I argue that ICA puts public investment in domestic legal capacity at risk by reducing pressure on the state to commit political and financial resources to it. In a nutshell, ICA reduces commercial actors’ reliance on domestic courts by providing them with a private, enforceable, and extrajudicial means of resolving contract disputes. Firms also prefer arbitration because it is typically confidential and the parties control the entire process. Unlike judicial proceedings, arbitration allows the parties to determine nearly all aspects of the dispute resolution process, such as who the arbitrators are; what rules will govern the merits of a dispute (that is, what law will be used to determine the issues at stake); the rules governing the procedure of the arbitration; and in what jurisdiction the award will be enforced. For example, a Chinese firm may use an arbitration provision in a contract with an American counterparty and take them to arbitration in England; and despite its taking place in England, Chinese law may apply to the case. If the American firm loses, the Chinese firm can ask an American court to enforce the award with the same legal force as an American judicial ruling. The American court is bound to enforce the award, with only a very narrow set of exceptions.
The legal flexibility ICA provides has turned it into a crucial backbone of the global legal framework facilitating international trade and investment.Footnote 28 We can see this just in the number of cases sent to arbitration each year. There were 842 cases filed with the ICC in 2018, and hundreds more with its competitors, like the London Court of International Arbitration. While most of these disputes are likely between private parties, ICA does not handle only purely private disputes. In fact, more public–private disputes are arbitrated through ICA than treaty-based investor–state dispute settlement (ISDS) cases. And reliance on ICA over ISDS has been growing in the last few years (Figure 1). In contrast to ISDS, which is designed to manage violations of international law, ICA is equipped to resolve almost any cross-border contract dispute. ICA therefore offers a more complete substitute for a country's domestic contract enforcement institutions that applies to fully private as well as public–private disputes. While this system may create localized benefits for large firms and cross-border trade, I argue here that the way ICA is currently practiced—substituting for domestic courts—risks undermining the public legal infrastructure within countries that facilitate ICA.
FIGURE 1. Contract- versus treaty-based arbitration between private and public actors, 1995 to 2020
How ICA Hinders Legal Development
Legal capacity is in large part a political outcome.Footnote 29 The prospect of boosting international trade and economic development creates an important incentive for judicial reform, as it is widely understood that international (and domestic) commercial actors are highly sensitive to the capacity of domestic institutions.Footnote 30 And leaders are aware of this. For example, in his study of reform of the Egyptian judiciary, Tamir Moustafa quotes from an official involved in designing Egypt's newly independent (in economic matters) constitutional court in 1979 as saying that while there was domestic and international pressure, “more importantly, from the outside there was pressure from foreign investors and even the foreign embassies.”Footnote 31
Arbitration can diminish outside pressure on the state to invest in legal capacity because it is a highly effective and private alternative to contract enforcement through national courts. With their exit into arbitration, politically influential firms have less of a stake in the quality of domestic courts. Arbitration thereby reduces the incentives for political leaders to carry out politically and financially costly reforms and similarly lowers the economic costs of political intervention in the judiciary. As Tom Ginsburg argues, the availability of international arbitration therefore “may reduce courts’ incentives to improve performance by depriving key actors from a need to invest in institutional improvement.”Footnote 32
The harmful effect of substitution is likely exacerbated in countries where legal capacity is already low due to lack of funding or political support. For example, in a study of commercial arbitration in Sudan, Mark Massoud argues that the Sudanese regime promoted ICA to provide high-quality legal services demanded by foreign investors without risking spillover of liberal rule-of-law norms into the broader judiciary.Footnote 33 Arbitration thus grants key interest groups access to an effective and neutral contract enforcement institution, without undermining the regime's use of the judiciary as a tool for repression. Massoud quotes an international lawyer who states bluntly how the growth of arbitration has altered political leaders’ incentives: “Given how well-established international arbitration is and how [strong] it's become, money saying ‘look at our courts and how independent they are’ might not be money well spent.”Footnote 34 Indeed, in an international survey of in-house counsel, 92 percent of respondents said they prefer arbitration to cross-border litigation in national courts.Footnote 35
Arbitration thus targets the demands of specific economically important actors at the expense of those with less influence. It splits constituencies that would otherwise share an interest in pressuring the state to maintain or enhance public legal capacity. This reduces the political and economic costs that states face for failing to invest in reforms promoting legal education, transparency, accountability, and efficiency in domestic legal institutions.
A resource-constrained state may benefit from a private exit option for discontented actors.Footnote 36 The growth of a private alternative to national courts relieves pressure from commercial interests on leaders to implement reforms promoting judicial neutrality, predictability, and expertise. Without a private alternative, leaders face a dilemma in which they may prefer to have a strong judiciary to promote investment but fear that an independent judiciary may turn against the regime's interests in other areas.Footnote 37 Outsourcing otherwise public adjudication tasks to private substitutes can help resolve this dilemma.Footnote 38
Egypt faced this dilemma in the 1990s. After granting its Constitutional Court greater levels of independence over the prior decade, the Egyptian government cracked down on it after it tried to parlay the legitimacy it had won in the economic realm into matters like human rights.Footnote 39 Egypt enacted pro-ICA reforms based on the Model Law in 1994 and was able to maintain its reputation as an attractive site for ICA despite subsequent political interventions in the judiciary. As these examples show, ICA enables states with weak public legal capacity to provide neutral, efficient judicial-like adjudication while reducing the costs of maintaining tight control over the judiciary in other matters.
The growth of global governance disconnected from public accountability risks reducing not only international reform pressure but also domestic pressures. Because of the latitude given to contractual parties in defining what constitutes an “international” contract or dispute and the mobility of capital, domestic actors can take advantage of ICA as well. Russian oligarchs and commercial interests, for example, have taken their capital abroad to avoid domestic institutions and take advantage of British courts and international arbitration bodies like the London Court of International Arbitration or the ICC to settle disputes and enforce contracts.Footnote 40 The availability of these institutional exit options in the 1990s worsened the collective action problem plaguing Russian commercial interests when faced with an increasingly extortionate and illiberal state.Footnote 41 It became easier to simply rely on transnational contract enforcement institutions than to lobby for domestic reforms. Gulnaz Sharafutdinova and Karen Dawisha argue that the availability of high-quality transnational contract enforcement institutions not only reduces political pressure for domestic reform but also increases the incentives for domestic business elites to maintain the illiberal domestic status quo: “Business elites take advantage of weak institutions at home to make profits, while using strong institutions abroad to safeguard them.”Footnote 42 In other words, beyond simply reducing pressure on the state to improve public institutions, strengthening private institutions could even generate an antireform constituency that benefits from easier access to private substitutes for weak public institutions.
Can ICA Improve Legal Development?
Some argue that unbundling might produce positive externalities, such as competition between arbitration and courts that generates a “race to the top.”Footnote 43 For competition to produce a “race to the top,” however, there must be some mechanism by which competition creates costs that the public institution will seek to minimize or recoup. It is unclear what those costs would be. National judges do not internalize the benefits of the law they provide.Footnote 44 Therefore, there is little reason to expect courts to suffer when dispute resolution is outsourced to a third party—judges do not lose from the growth of arbitration.
According to the theoretical argument given earlier, a private authority could be designed to be complementary to a domestic institution if it remains reliant on the capacity of the domestic body to function well. Could this be the case with ICA? As I will argue, opportunities for states to regulate ICA have been declining for decades. The two main opportunities for overseeing ICA are in the design of domestic legislation and in the judicial enforcement of arbitration agreements and awards. I deal with each of these factors in turn.
In theory, there are ways a country could both promote ICA and oversee its practice, thereby promoting a potentially complementary relationship between arbitration and the judiciary. A country could grant the right to judicial review on the merits; require arbitrators to state the reasons for their decisions; require that awards be made public; and so on. Few countries do so. Instead, most countries enacting ICA reforms today base those reforms on the Model Law (often considered the “gold standard” of a modern ICA regime), which expressly limits potential mechanisms through which a court might oversee arbitration.
In the interest of promoting ICA globally, the goals of the Model Law were twofold: first, to encourage countries to adopt a globally harmonized legal regime facilitating and protecting ICA domestically; and second, to ensure such laws prevent public and judicial intervention in ICA. The Model Law severely restricts judicial intervention through various rules, including that arbitral awards cannot be appealed; that courts must enforce awards and arbitration agreements except under very limited circumstances; and that arbitrators can find their own jurisdiction (Kompetenz-Kompetenz). UNCITRAL's advocacy has been instrumental in harmonizing and increasing ICA protections around the world. Moreover, competition for trade and investment drives states toward focal standards such as the Model Law and incentivizes them to limit the scope of public oversight of the practice. Because the pressure for reform also comes from a desire to attract capital, rather than to reform the judiciary, countries have opted to adopt the Model Law with minimal revision.Footnote 45
Another mechanism for retaining firms’ reliance on courts is to carve out certain areas of law over which the judiciary has oversight or exclusive jurisdiction. But arbitrators today have wide latitude to base decisions on their own interpretations of almost any relevant rules of law. Judiciaries in major arbitration states have been gradually increasing the authority of arbitrators to interpret and apply public law. For example, through a series of interpretations of the Federal Arbitration Act, the US Supreme Court has increased arbitrators’ powers to interpret on “mandatory rules,” or rules that are supposed to apply to all contracts even if both parties would prefer to opt out of them (versus “default rules,” which parties can modify). As a consequence, actors gained an avenue for circumventing mandatory rules in areas like securities law and antitrust.Footnote 46 Similarly, European courts have given arbitrators more authority to root decisions in their own interpretations of mandatory EU law.Footnote 47 This deference has led to the transnationalization of commercial law and its decoupling from domestic law. Using both legal analysis and interviews with practitioners, Joshua Karton finds that the culture of ICA has led arbitrators to arrive at distinct, though internally consistent, interpretations of domestic law.Footnote 48
Karton's findings are particularly important in light of concerns that arbitration decreases the predictability and transparency of domestic law by preventing commercial law from developing in public view. Inconsistencies between private and public applications of public law could be resolved if arbitral awards were reviewable by a court for legal errors. But the Model Law bars such review, further limiting the opportunity for judicial oversight. A South African judge president wrote in 2005 that arbitration gave White business interests a tool for undermining attempts to integrate the judiciary: “This [commercial arbitration] is clearly an attempt to undermine the transformation of the judiciary. Arbitration does not contribute towards the development of the law.”Footnote 49 Similarly, the English lord chief justice warned that the growth of arbitration provisions in commercial contracts “has been a serious impediment to the development of the common law by the courts in the UK.”Footnote 50
A final potential avenue for complementarity written into the Model Law is an exception that allows courts to deny enforcement of awards that cut against “public policy.” The scope of this exception is very narrow, and shrinking, further limiting domestic authorities’ ability to oversee arbitration practice.Footnote 51 Recently, courts in major enforcement states have increasingly interpreted this exception to refer only to international public policy—even if it contravenes domestic law. An array of courts in important enforcement countries—the United States, Italy, India, Egypt, France, Switzerland, and others—have ruled along these lines.Footnote 52
In sum, the modern ICA regime is designed to prevent domestic oversight. It removes the state from the regulation of commercial disputes, reducing dependence on public legal institutions. Pushing the state into the background minimizes incentives that would otherwise exist to invest in costly legal reforms. I therefore expect to find stagnation or a negative association between the promotion of ICA and legal development. I test this hypothesis on a cross-national panel of countries that have implemented UNCITRAL's Model Law on ICA.
Data and Methods
Dependent variable: legal development
I adopt a definition of legal development that is tied to the capacity, efficiency, and fairness of the judiciary specifically. This definition is narrower in scope than traditional conceptions of the “rule of law” because it is primarily from the judiciary that tasks are being unbundled and delegated to ICA. In the absence of an exit option, commercial actors would have an interest in pressuring the state to improve the capacity of the judiciary to enforce contracts and resolve disputes efficiently and fairly.
I use the Rule of Law Index from the Varieties of Democracy Project (V-Dem) to measure the capacity of judicial institutions cross-nationally and over time.Footnote 53 V-Dem's index is ideal for this case because it is an aggregation of expert-coded measures primarily pertaining to theoretically relevant features of the domestic legal system, including both the independence and competence of multiple levels of each country's judiciary, along with other aspects of modern legal development, including the openness and transparency of laws, citizens’ ability to redress rights violations through courts, and the predictability of enforcement. Aside from the substantive similarity of the index to the definition of legal capacity adopted here, another benefit of the measure is that it has very wide coverage. It supports the inclusion of over 150 countries in the sample across the full length of the relevant time span (beginning in 1985, the year the UN General Assembly adopted the Model Law). V-Dem's index is preferable to other measures, such as the rule-of-law indices maintained by the World Bank or Freedom House, for conceptual reasons as well as for its more expansive temporal and geographic coverage. Those other measures refer to a much broader conception of the rule of law that incorporates outcomes that are only tenuously linked to judicial capacity, like crime, war and violence, corruption, and policing. In the robustness checks, I disaggregate the V-Dem measure and try other targeted measures of judicial capacity from the Fraser Institute.
Independent variable: protections for ICA
As a proxy for integration into the ICA regime, I collected data on the enactment of domestic legislation based on the UNICTRAL Model Law on ICA. Introduced in 1985, the Model Law is considered the “state of the art” in permissive arbitration laws. Written in the early 1980s and approved by the UN General Assembly in 1985, the Model Law was meant to correct deficiencies that international commercial and legal communities felt were hindering ICA outside of a few arbitration “hubs” like the United States and France. By 2020, over seventy-five countries had enacted national legislation based on the Model Law (Figure 2).
FIGURE 2. Rate of national legislation based on the UNCITRAL Model Law on International Commercial Arbitration, 1985 to 2020
The data were collected from the UNCITRAL's yearly “Status of Conventions” reports. These reports update UNCITRAL members when a country is recognized by UNCITRAL for having legislation based on the Model Law (and other UNCITRAL initiatives) enter into force.Footnote 54 While countries can shape domestic implementation of the Model Law as they see fit, UNCITRAL's primary goal is transnational legal harmony. UNCITRAL has an interest in maintaining the value of its legal instruments not only as guides for commercial law reform but also as heuristics for the international legal and commercial communities. For that reason, UNCITRAL will not approve a country as a “Model Law country” if it deviates too far from the text or spirit of the Model Law. As one senior legal officer at UNCITRAL writes, there is a “high degree of substantive uniformity in the implementation” of the Model Law.Footnote 55 Also, a law is not considered an enactment of the Model Law if it “contain[s] any provision incompatible with the basic philosophy of the Model Law.”Footnote 56 Consistent with these norms and the incentives countries face for harmony, an independent analysis of all Model Law countries in 2010 found extremely close similarity between jurisdictions.Footnote 57 Thus the rules governing who is eligible for UNCITRAL's imprimatur as well as independent, in-depth, legal analyses of the laws themselves demonstrate high uniformity across jurisdictions. Thus the Model Law is a “bundled treatment” in that it implements a network of rules that together facilitate the privatization of dispute resolution by shielding the process and outcomes of arbitration from judicial scrutiny, while at the same time requiring courts to enforce arbitration agreements and awards. Enactment thus implies strong restrictions on judicial intervention in arbitration.
Estimation strategy
I estimate the effect of enacting strong protections for ICA on the quality of domestic legal institutions using the PanelMatch difference-in-differences (DiD) estimator with weighted matched sets.Footnote 58 This estimator avoids potential issues with the two-way fixed effects estimator as it accommodates treatment effects that are heterogeneous across units and time and prevents mismatched comparisons between already-treated and newly treated units. The PanelMatch estimator is also better equipped to handle unbalanced panels with staggered adoption and relatively fewer pretreatment time periods than other, similar strategies, such as synthetic control methods.Footnote 59 To assess the robustness of the findings to alternative specifications, I rerun the analysis using the unbiased linear estimator proposed by Borusyak, Jaravel, and Spiess.Footnote 60
The propensity scores used to create the weights for the matched sets are estimated by regressing the treatment variable, enactment of the Model Law, on a set of covariates prior to enactment. I include three institutional covariates. First, I include a count of bilateral investment treates (BITs) in force. BITs often provide access for foreign investors to arbitration through international investor–state dispute settlement (ISDS) procedures. ISDS and ICA are dispute resolution frameworks with similar relationships to domestic courts, so having ratified BITs in the past may increase a state's odds of enacting the Model Law. Second, I include a dummy variable for whether a country has ratified the New York Convention. Third, I include the dependent variable of the second stage of the analysis, the V-Dem Rule of Law Index, in case enactment is correlated with pre-existing legal capacity.Footnote 61
I also include a set of economic covariates. Countries that are more integrated into the global economy face more pressure to provide neutral dispute resolution services and therefore may be more likely to invest in both capacity-enhancing legal reforms and transnational contract enforcement regimes like ICA. I therefore include economic variables that could influence both pressure for reform and access to legal development assistance. I include measures of total trade (imports plus exports) as a percentage of GDP, logged GDP, GDP per capita, and GDP growth to help adjust for any confounding effects of market size and economic development trajectory. These data are from the World Bank's World Development Indicators. To measure a country's dependence on FDI, I obtained data on the total inward FDI stock as a percentage of GDP from UNCTADstat. I lag all explanatory variables by one year.
The final step of the procedure is to estimate the average treatment effect on the treated in the year of enactment of the Model Law (t i) and for each of the five years thereafter (F). I apply the following DiD estimator for each time period F:
$$\widehat{{{\rm AT}{\rm T}_F}} = 1{\rm /}N \times \mathop \sum \limits_i^N \left({( {Y_{i, t_i + F}-Y_{i, t_i-1}} ) -\mathop {\sum {} }\limits_{{i}^{\prime}\in {\cal M}_i} \omega_i^{i^{^{\prime} }} ( {Y_{{i}^{\prime}, t_i + F}-Y_{{i}^{\prime}, t_i-1}} ) } \right)$$N is the number of countries in the sample that have enacted the Model Law. t i is the year in which the Model Law enters into force in country i. Y i,t and 
$Y_{{i}^{\prime}, t}$ are the rule-of-law scores for Model Law and matched non–Model Law countries. 
${\cal M}_i$ defines the matched and weighted “control groups” for each Model Law country i. Each matched set 
${\cal M}_i$ includes all countries that have not yet enacted the Model Law and will not within five years of country i's enactment. While countries may be in multiple control groups, the weights applied to them are unique to each Model Law country. 
$\omega _i^{{i}^{\prime}} $ denotes the normalized weights applied to the rule-of-law scores for matched countries (i ′) within each Model Law country's control group (
${\cal M}_i$). This equation yields an estimate of the difference between the change in the rule-of-law score from one year before the Model Law enters into force to years t i + F for Model Law countries and the weighted average of the change within each Model Law country's matched set over the same duration. I calculate this for each Model Law country, then average the results for each time period. 
$\widehat{{{\rm AT}{\rm T}_F}}$ is therefore the estimated average effect of the Model Law entering into force for each year, from the year it enters into force through each of the following five years. A more thorough description of the estimation procedure is given in Appendix D.
Results
The main results are presented in Figure 3. Figure 3a reports the estimates from the full sample. While the Model Law and control groups are indistinguishable in the year of enactment, we see an increasingly large relative decrease in the Rule of Law score for Model Law countries, consistent with the framework presented here. The difference becomes statistically significant at the 95 percent level three to four years post-enactment. It takes time for the legal and behavioral changes brought on by the Model Law to influence broader legal development in the country. Parties must opt out of national judicial institutions by negotiating arbitration clauses into their contracts. Therefore, there should be some lag as firms shift their attention away from the domestic judiciary and rules and toward transnational arbitration centers. Exit by commercial parties from national legal institutions lowers outside pressure on the state to invest in progressive reforms like revamping archaic procedures, improving judicial and legal training, raising salaries, and funding domestic law schools. This process leads to the gradual reduction in political pressure for investment in progressive rule-of-law reforms, which then allows problems in the legal system to persist and accumulate. Moreover, the economic costs of having a low-capacity or politically motivated judiciary are expected to be lower in countries that promote the use of arbitration. For any given jurisdiction, change may be bumpy because it is often through crises or cases that new information is revealed about the capacity and independence of the judiciary. And if the Model Law makes such events more likely on average, we should see a gradually increasing separation between Model Law and non–Model Law countries as a result.
FIGURE 3. Main results
Figure 3 also presents the results of a placebo test to assess the parallel-trends assumption: that Model Law countries and their matched sets would not differ in the absence of Model Law enactment. The flat line prior to enactment (years −5 through −2) in all three figures does not provide any evidence that the results are driven by pre-existing differences in the trajectories between the two groups in the years leading up to enactment.
We are most interested, however, in the effect of ICA on institutions in countries that do not already enjoy a high-capacity, consolidated legal regime. Unpacking how ICA influences domestic legal institutions in weak rule-of-law countries is important because the Model Law is often embedded within broader development efforts to promote the rule of law in countries where legal capacity is low. As Catherine Rogers and Christopher Drahozal put it, there is “an implicit promise of investment arbitration … that it will not only provide protection of foreign investors, but also foster good governance.”Footnote 62 The sample in Figure 3a includes all enacting countries, which may be biasing the results toward zero, for a couple of reasons. First, countries that enjoy robust legal systems may not be actively engaged in legal reform, so a reduction in pressure for reform would have little effect on institutional outcomes. Second, weak rule-of-law countries tend to have fewer resources, so they face higher opportunity costs when investing in reform projects. Minimizing pressure for legal reform may have a larger negative impact in those countries than in better-resourced countries. Third (and related), arbitration will diffuse concerns about illiberal interventions in the judiciary.
To examine the effect of the Model Law on countries with weaker legal infrastructure, I rerun the analysis but exclude countries that enact the Model Law but also have strong pre-existing rule-of-law institutions. I classify as “low rule-of-law” any country with a Rule of Law Index less than 0.8 at the time of enactment of the Model Law.Footnote 63 As a frame of reference, Bulgaria, a Model Law country, has hovered around 0.75 for the last decade. Another, Mexico, fluctuated between 0.5 and 0.65 over the same period. Just above the cutpoint is Greece, which had a score of 0.82 in 2017. The results for this subsample are reported in Figure 3b.Footnote 64
Comparing Figures 3a and 3b, we see that the point estimate of the average treatment effect on the treated for low rule-of-law countries is roughly double that of the full sample (though the difference is not statistically significant). We also see the same pattern of gradual institutional degradation relative to the control group. The model estimates that, on average, five years after enacting the Model Law a country is around 0.047 points below where it would otherwise have been. This comes out to a cumulative decrease over five years of roughly 15 percent of a standard deviation of the rule-of-law score in the sample. This finding is robust to alternative specifications. The difference between weighting by propensity scores or covariate balancing propensity scores is negligible (compare columns 2 and 3 of Table A5). The unbiased linear estimator yields very similar estimates (column 5 of Table A5).
Does the Model Law exert a similar effect on countries with already consolidated legal regimes? It appears not. Figure 3c plots the results for the high rule-of-law sample. Unlike in the estimates for weak rule-of-law regimes, the promotion of ICA appears to have no effect on legal development in consolidated legal regimes (the estimated effect is statistically insignificant, and very close to zero).
Are legal institutions in recent Model Law countries weakening, or are they simply not improving at the rate they otherwise would have? We can examine the first differences to see what is driving the growing divergence between Model Law and non–Model Law countries.
Figure 4 plots the estimated trajectories of the two groups separately. Corroborating the placebo tests visualized earlier, they have similar pre-enactment trajectories—both gradually improving. But after enactment the non–Model Law group continues to improve steadily (gray line), while the Model Law group shows an absolute and relative decline in the quality of domestic legal institutions (black line). This suggests that the effect is driven partly by institutional erosion within Model Law countries, but also partly by a continued improvement in the non–Model Law countries after enactment, which halts in Model Law countries. This is consistent with the theory presented here, in which the exit of international and domestic commercial actors from the domestic legal system is expected to lower pressure on governments to invest in costly reforms to improve the neutrality, competence, and efficiency of national legal institutions, while at the same time reducing the economic costs faced by regimes with weaker public legal systems.
FIGURE 4. First differences
As mentioned, the V-Dem Rule of Law Index is a composite indicator. Some of its subcomponents are of direct theoretical relevance, but others are less so.Footnote 65 To see which of the subcomponents is driving the results, I conducted a series of static DiD analyses using the robust linear estimator,Footnote 66 with the same set of covariates, but replacing the composite index with each of its subcomponents (Figure 5). For ease of interpretation, I categorize each subcomponent based on its theoretical relevance. The primary drivers are almost all theoretically relevant. While the Model Law has a null effect on judicial independence, it is associated with worse judicial outcomes: compliance with the judiciary as a whole (and, to a slightly lesser extent, the high court alone) declines after enactment, as does the availability of judicial remedies for men and women. Similarly, by removing commercial dispute resolution from public scrutiny and facilitating the importation of foreign law, the Model Law diminishes the relevance of domestic law and thereby reduces the need for the state to commit resources to improving the quality of domestic legislation. Accordingly, we also see a reduction in the transparency and predictability of domestic laws. The subcomponents with the weakest associations have to do with matters unrelated to the Model Law: corruption and embezzlement.
FIGURE 5. Estimates of the effect of the Model Law on subcomponents of the V-Dem Rule of Law Index
Robustness Checks
Alternative low rule-of-law cutoffs
To ensure that these results are not driven by how I categorize “high” and “low” rule-of-law countries, I rerun the low rule-of-law analysis using other cutpoints (Figure 6). These plots show that the estimates presented in Figure 3b are consistent across a range of other plausible cutpoints (between 0.5 and 0.9). As in the main results, we see a gradual reduction in rule-of-law scores after Model Law enactment across all analyses. Interestingly, the upward slope in each plot indicates that the effect size decreases as the mean rule-of-law score in the “low” rule-of-law group increases, suggesting further that countries with weaker legal systems prior to enactment are more susceptible to institutional stagnation.
FIGURE 6. Alternative cutoffs
Alternative measures of dependent and independent variables
Some popular “arbitration hubs” have not enacted the Model Law, often because they were the early adopters and promoters of commercial arbitration that set the standard on which the Model Law was based. They include France, Sweden, Switzerland, the United States, and the United Kingdom. Since these countries tend to have highly efficient judicial systems, their inclusion in the control group might bias the results. But excluding them from the analysis does not materially alter the estimates (Figure A3).
I also rerun the main analysis on weak rule-of-law countries using alternative measures of the rule of law created by the Fraser Institute. I find that Model Law enactment is associated with a statistically significant decline in judicial independence (Figure A6a), and negatively associated with the integrity of the legal system (Figure A6b). And as further evidence of the Model Law's impact on domestic contract enforcement, I find that, despite its negative effect on broader legal institutions, Model Law enactment increases the quality of contract enforcement within a country (Figure A6c).
Instrumental variables estimation
While I find no evidence that legal institutions in Model Law countries and non–Model Law countries are on different trajectories prior to enactment of the Model Law, my estimates could be biased if, for example, a large subset of leaders enact the Model Law in anticipation of policy that would erode the quality of domestic legal institutions independently of Model Law enactment. Here, I develop an original instrument for Model Law enactment to help allay endogeneity concerns. This instrument helps deal with such issues because it predicts Model Law enactment using variation in the rate of Model Law adoption among a country's export competitors in contract-intensive trade, which is plausibly unrelated to unobserved domestic factors that might generate a spurious correlation between Model Law enactment and legal stagnation.
The intuition motivating this instrument is that greater competition with Model Law countries in contract-intensive products will increase the incentive for countries to also enact the Model Law to improve the attractiveness of their contracting institutions in the eyes of foreign purchasers or investors. Government reports and speeches demonstrate how Model Law enactment among a country's trade competitors influences domestic considerations. In South Africa, for example, a report by the South African Law Commission advocating Model Law enactment noted that the end of apartheid led to “increased regional trade and economic links with other countries,” which made it increasingly “important that the country's arbitration law should be in line with international norms.”Footnote 67 The report then mentions regional economic competitors explicitly to justify its recommendation to enact reforms based on the Model Law.Footnote 68 Elsewhere, an official in Argentina's Ministry of Justice supported adopting the Model Law in part because a regional economic competitor, Uruguay, was taking steps to enact it. In his words, “the global market demands an increasingly uniform legal system.”Footnote 69
I instrument for Model Law enactment using the global rate of Model Law adoption weighted by how much a given home country competes with Model Law countries in contract-intensive export markets. To construct a measure of competition in contract-intensive trade, I first obtain product-level trade data spanning 1996 to 2019.Footnote 70 I then identify products at the four-digit level under the UN's Standard International Trade Classification (Rev. 3) that are “differentiated,” meaning they are not traded on an exchange or tied to a reference price.Footnote 71 I examine only exports of differentiated products because they tend to be more complex, and therefore trade in such goods is more reliant on relationship-specific contracts.
From these data, I construct an 
$N \times N \times T$ matrix containing the correlation of the value of exports at the importer-product level between each country pair ij in each year t, denoted w ijt. I replace negative correlation coefficients with 0 because I expect only positive trade similarity to induce competitive pressure.Footnote 72 I then normalize every correlation for each country i by the sum of its correlations with all other countries j, or 
$w_{ijt}^\ast{ = } {{w_{ijt}} \over {\sum _{i\ne j}^n w_{ijt}}}$. This ensures that the weights are not homogeneous across country pairs but are instead relative to each country's overall level of competition. Following this, I multiply each 
$w_{ijt}^\ast $ by 1 if the Model Law is in force in competitor country j in year t and 0 otherwise, and then sum the result. This yields the global Model Law adoption rate weighted by the level of export competition each country faces with Model Law countries in contract-intensive trade.
To illustrate the face validity of the measure, I plot the average export competition weights (before multiplying by Model Law enactment) for South Korea and Thailand (Figure 7). Comparing the upper and lower panels of Figure 7, we can see that Thailand's biggest competitors are largely restricted to South Asia, while South Korea's span the globe, with higher-weighted regions in East Asia, North America, and Europe. This difference likely reflects the relative position of each country in global value chains, as the kinds of countries from which Thailand faces the stiffest competition also tend to be lower down the value chain than Korea's. Thailand's largest export competitors are Malaysia and China, whereas South Korea's are Canada and Japan. And Korea's higher-weighted regions in Europe are driven by competition with countries like the UK, Germany, and Sweden which also specialize in manufacturing cars and high-end electronics.
FIGURE 7. Average export competition in contract-intensive trade in Thailand (above) and South Korea (below)
I estimate the effect of Model Law enactment on the capacity of domestic legal institutions using two-stage least squares (2SLS) regression in which I instrument for Model Law enactment using contract-intensive export competition (see Appendix F for details). The main results are presented in Table 1. Consistent with the DiD findings, I find a negative effect of Model Law enactment on the quality of domestic legal institutions. The effect is statistically significant in all specifications.
TABLE 1. 2SLS estimates
Notes: ExportComp is scaled to have mean 0, SD 1. “Legal” controls: NYC ratification and log of no. of BITs + 1 ratified; “Econ. int'l”: log of inbound FDI stock and trade dependence; “Econ. domestic”: log GDP per capita, GDP, and growth. All explanatory variables are lagged by one year. Model Law countries without pre-treatment data are excluded. Effective F-stat estimated using method described by Olea and Pflueger (Reference Olea and Pflueger2013). Weak-IV robust CI is the 95% confidence intervals generated from the Anderson–Rubin test (Chernozhukov and Hansen Reference Chernozhukov and Hansen2008). Full results and discussion in Appendix F. Standard errors clustered by country. *p < .10; **p < .05; ***p < .01.
One potential issue is that the F-stats for the excluded instrument hover around 10. This suggests the possibility of a weak instrument, which could introduce bias into the 2SLS estimates. I assess the robustness of my estimates to this possibility in three ways. First, re-estimating the models using the limited-information maximum likelihood estimator produces estimates that are essentially equivalent to those of 2SLS (Table A8).Footnote 73 Second, the reduced-form specification (in which I regress legal capacity on the instrument) is unbiased in the presence of a weak instrument; the reduced-form estimates are highly stable and statistically significant across all specifications (Table 2). Finally, I estimate 95-percent confidence intervals using the Anderson–Rubin test, which provides correct coverage regardless of instrument strength.Footnote 74 The weak-IV robust confidence intervals are all highly statistically significant and consistent with my theoretical expectation that enactment of the Model Law will have deleterious effects on legal development (Table 1).
TABLE 2. Reduced-form estimates
Notes: ExportCompetition is scaled to have mean 0, SD 1. OVB robustness values are derived from the method proposed by Cinelli and Hazlett (Reference Cinelli and Hazlett2020). Standard errors clustered by country. *p < .10; **p < .05; ***p < .01.
As an indirect method of validating the instrument, I reran the analysis using export competition in undifferentiated products (that is, less contract-intensive trade). This predicts neither Model Law enactment in the first stage, nor changes in the rule of law in the reduced form (Tables A9 and A10).Footnote 75
For contract-intensive export competition to be a valid instrument it must influence the quality of legal institutions in the home country only through its effect on Model Law enactment. Due to similarities between investor–state arbitration and ICA, one might worry that competition with Model Law countries might also increase the propensity of a country to ratify BITs as well, which some have argued may have harmful effects on domestic governance.Footnote 76 I find no evidence, however, that contract-intensive export competition influences BIT ratification (Table A11). Or perhaps it is simply export competition driving the results. I find no association between total export competition in differentiated products and Model Law enactment in the first stage (Table A12). Similarly, I find no effect of total export competition in the reduced form (Table A13). It is otherwise unclear how Model Law enactment among a country's competitors might influence the development of its legal institutions. While this assumption is unfortunately untestable, I can assess how my estimates would change under hypothetical violations and provide some benchmarks for thinking about how threatening a potential violation of this assumption is to my estimates. To do that, I conduct a sensitivity analysis on the reduced-form specification.Footnote 77 This will provide a sense of how robust the 2SLS estimate is to omitted variable bias (that is, violations of the exclusion restriction) given that the 2SLS coefficient is equal to the ratio of the estimated coefficients from the reduced-form and first-stage models.
The results of the sensitivity analysis are presented below the coefficient estimates in Table 2. The robustness values, RVq=1 and RVq=1,α=0.05, indicate the percentage of the variation in both export competition and rule of law that an unobserved confounder would have to account for to drive the coefficient to zero or the p-value above .05, respectively. Here is a more concrete benchmark: to drive the coefficient on export competition to zero, a potential confounder would need a partial R 2 (on both treatment and outcome) of about fifteen times that of logged trade dependence or about five times logged GDP per capita. So while one can never entirely rule out the possibility of a violation of the exclusion restriction, these statistics suggest that the results presented here are fairly robust even if the exclusion restriction is violated to some degree. In sum, the IV estimates, combined with the DiD estimates presented in the previous section, suggest a negative relationship between the promotion of arbitration and the subsequent development of a country's legal institutions.
Does the Model Law Increase the Use of Arbitration?
ICA has important political and legal implications in large part because of the structural factors discussed earlier, such as the absence of any system of appeal. But such lacks should not be confused with the absence of any system of control or supervision. There remains a limited set of tools available to national courts for overseeing arbitration (such as setting aside awards and issuing interim measures), so long as the arbitration is seated in that court's jurisdiction. Importantly, where to seat an arbitration is decided by the parties. The effect of the Model Law on the authority of national courts is therefore partly a function of the behavior of private (and public) actors in negotiating where to seat their arbitration. This means substitution is not only a legal question but also an empirical one. Earlier I argued that the structure of modern ICA reduces its dependence on public institutions; in this section I examine the behavioral implications of enactment. I find that Model Law enactment weakens dependence on courts—enactment increases the use of arbitration by nationals in an enacting jurisdiction—but I do not find consistent evidence of an increase in the rate at which that jurisdiction is selected as the seat of arbitration. This suggests that, beyond the structure of ICA, party behavior is further reducing dependence on domestic institutions and thereby decreasing pressure on states for capacity-enhancing reform.
To examine the impact of Model Law enactment on dispute resolution behavior, I gathered yearly data on the location of the seat of arbitration as well as the nationality of parties to cases managed by the ICC from 1992 to 2019.Footnote 78 The ICC is an especially useful case here for two reasons. First, the ICC typically manages very high-value disputes, so its cases tend to represent the behavior of some the largest firms and international deals. And, second, the ICC is both a highly active ICA center and a distinctly international one. The range of arbitral seats in the ICC's caseload is uniquely diverse compared to its closest competitors. Given the stature of the ICC within the field of ICA, patterns within the ICC are legally and politically important in their own right. We can interpret trends seen there as broadly indicative of shifts in ICA practice for high-value disputes.
I control for a variety of economic and institutional factors that may increase the probability that commercial disputes arise, including the size of the country's economy, its inbound FDI stock, and its dependence on trade. I control for the level of development with GDP per capita. And because disputes tend to arise more often during periods of economic downturn, I add a measure for GDP growth.Footnote 79 I also control for membership in the New York Convention and the strength of domestic legal institutions using the V-Dem Rule of Law Index. I estimate the following equation using the Poisson pseudo-maximum likelihood estimator:
$${\rm Y}_{it} = \exp ( \beta {\rm ModelLa}{\rm w}_{it} + \delta {\bf X}_{it} + \gamma _i + \omega _t) $$where Yit represents the outcome; 
${\bf X}_{it}$ is a vector of controls; and γ i and ω t are country and year fixed effects. As before, I also present results using the unbiased linear estimator,Footnote 80 though I transform the case count variables with the inverse hyperbolic sine for these analyses.
The results are presented in Table 3. Panel A presents the results for the yearly counts of ICC-managed arbitrations seated in a given country. The Model Law exhibits a positive but inconsistent effect on this number. In the full sample (columns 1–3), we see that the effect is strong in the bivariate specification, loses significance after adding economic controls, and becomes significant at the 10 percent level with the addition of institutional controls (though it fails at the 10 percent level in the Borusyak, Jaravel, and Spiess estimates). Columns 4 and 5 subset the outcome based on how the seat was determined. In column 5, the outcome is the number of cases in which the location of the seat was chosen by the parties themselves. Here again we see a weak effect. Column 4 presents results using the count of cases in which the seat was determined by the ICC rather than the parties. This suggests that the Model Law has a slightly stronger effect on the viability of the jurisdiction in the eyes of the ICC, though the pretrends are significant and in the same direction as the estimated effect in the Poisson regressions.
TABLE 3. Estimates of the effect of the Model Law on outcomes at the ICC
Notes: Values in brackets denote the probability that three yearly leading treatment indicators jointly equal 0. Full tables can be found in Appendix G. Standard errors clustered by country. *p < .10; **p < .05; ***p < .01.
We now turn to Panel B of Table 3, in which I shift the outcome from the seat of arbitration to a yearly count of the nationality of parties to arbitration at the ICC. Here we see a much stronger and stable effect of the Model Law on arbitral behavior. The estimates on the Model Law are highly significant and consistent across all specifications of the pooled sample (columns 1–3). The substantive effect is significant as well. The model with a full set of controls (column 3) estimates that enactment of the Model Law leads to an increase in a country's nationals represented at ICC proceedings of roughly 25 percent. I subset this analysis based on the party's role in the arbitration as either the complainant or the defendant. The models estimate a larger effect on the complainant side, with a 34 percent increase in the number of cases with nationals as complainants versus an 18 percent increase for defendants. This suggests that the Model Law has more influence on the behavior of domestic firms that choose to submit disputes to arbitration.
Conclusion
The findings presented here suggest that the growth of transnational substitutes for domestic institutions may have costs for the very countries they are often purported to assist. Before declaring international arbitration a success for the rule of law (because of the relative ease by which firms can enforce international contracts), we need to evaluate potential downstream effects that are likely to hit developing countries hardest.Footnote 81 The broader consequences of ICA for legal development are particularly salient in light of concerns from legal scholars regarding the growth within nondemocracies of new forms of commercial dispute resolution, including arbitration, that are meant to enhance those regimes’ legitimacy as legal service providers without extending such services to the broader public.Footnote 82
More broadly, the findings presented here lend support to the emerging body of scholarship in global economic governance that considers not just the first-order effects of global governance but also the potential second-order effects. Particularly in light of the competing findings in the literature across ISDS and now ICA, there remain important open questions regarding how transnational and domestic institutions in this and other domains interact. Future research could explore alternative mechanisms of institutional interaction, such as norm diffusion. While the dynamics of the arbitration profession tend to limit competitive pressure on courts and, because of the lucrative salaries, pull legal talent out of domestic practice rather than into it, there may be opportunities for normative diffusion in areas with more fluid transnational movement of people and ideas.Footnote 83
Progressive rule-of-law reform is most likely to succeed when commercial and other civil society groups have a joint interest in pressuring the state to invest in it. The growth of substitutive international institutions risks undermining the efforts of countries with weaker legal capacity to invest in broad-based legal reforms by giving commercial actors an exit option unavailable to others. In this vein, Judge Abdulqawi Yusuf, former president of the International Court of Justice, described the importance of “re-localizing” arbitration to promote the rule of law in counties where it is lacking.Footnote 84 This suggests a need for increased focus on methods for promoting complementarity between transnational institutions and their domestic counterparts.Footnote 85
This growth of private global governance is especially important given the complexity of political accountability in such regimes. Simple lines of accountability channeled through visible, bundled domestic institutions facilitate coalitions for reform by clarifying the causal connections between governing institutions, tasks, and outcomes. But the decentralized world of transnational authority diffuses accountability across an ever-growing array of overlapping institutions and shrouds political decision making behind the veil of expertise. The theory and empirical findings presented here suggest that the design of transnational institutions is key. My results suggest that global governance institutions that are not designed to lock in interdependence between transnational and national authorities may have the unintended consequence of atrophy in domestic institutions.
Data Availability Statement
Replication files for this article may be found at <https://doi.org/10.7910/DVN/GSVW3C>.
Supplementary Material
Supplementary material for this article is available at <https://doi.org/10.1017/S0020818323000218>.
Acknowledgments
For their valuable comments on earlier versions of this paper, I thank Elizabeth Acorn, Oumar Ba, Kathleen Claussen, Matthew Evangelista, Terence Halliday, David Lektzian, Odette Lienau, David Di Micheli, Rebecca Perlman, Paula Rettl, Aditi Sahasrabuddhe, Ken Scheve, Theo Serlin, Whitney Taylor, and Christopher Way, as well as the participants in the Cornell International Relations Workshop, the 2020 annual meeting of the Southern Political Science Association, the 2022 annual meeting of the Junior International Law Scholars Association, the 2022 meeting of the International Political Economy Society, and the Graduate Students in International Political Economy Workshop. I would also like to express my gratitude to the IO editorial team and three anonymous reviewers for their thoughtful and constructive feedback.
