The global investment treaty regime, consisting of thousands of international investment agreements (IIAs),Footnote 1 is undergoing a period of unprecedented contestation and reform. This includes widespread efforts by governments in both developed and developing countries to scrutinize and reformulate their IIA policies. These changes seem driven by the fear that international obligations excessively restrict the flexibility of host states (where investments are located) in important areas of public policy, including environmental regulation, public health, social policy, and national security. Thus, concerns about the IIA regime are a specific instance of the “sovereignty cost” questions that arise with many legalized international institutions.Footnote 2
We see similar political turbulence in other areas of the world economy as governments struggle to reconcile the pressures of economic globalization with the need to preserve domestic policy space and social stability.Footnote 3 In the context of IIAs, these problems are exacerbated by the binding investor-state dispute settlement (ISDS) provisions found in most agreements, which allow foreign investors to challenge their hosts in international tribunals and to claim compensation, often in the range of tens or even hundreds of millions of dollars, if their rights have been violated. ISDS has grown increasingly controversial, drawing the ire of populist politicians and civil society groups around the world. It is often among the most hotly contested issues in contemporary international economic negotiations. Understandably, the political and legal implications of ISDS have received significant attention from scholars.Footnote 4
Dissatisfied states face a range of options for altering their IIA obligations. For new treaties, they can seek more favorable terms and update the “model” text that many use as a template for negotiations. For the thousands of existing agreements, states must choose between two main strategies: termination or renegotiation. Unilateral termination of an IIA is possible only after its initial term expires (typically ten years) and is often subject to “survival clauses” that restrict the legal benefits while still presenting political and economic risks.Footnote 5 It was therefore rare but has gained popularity in recent years.Footnote 6 A more common option is renegotiation, which produces immediate effects and permits a more tailored recalibration of IIAs. It has the additional virtue of allowing states to reclaim flexibility without abandoning existing agreements or appearing to undermine the broader regime.Footnote 7 Renegotiations first occurred about twenty-five years ago and have accelerated in recent years, resulting in about 200 renegotiated IIAs.
We focus on renegotiation (and, to a lesser extent, termination) to investigate the politics of sovereignty and change in the IIA regime. Previous scholarship shows that a state's experience with ISDS spurs learning and systematic changes to IIA policies.Footnote 8 We significantly extend this line of research by examining how exposure to ISDS influences changes in actual treaty obligations, utilizing a new data set on IIA design we developed to investigate these alterations systematically. Others have used treaty text as data to compare the content of economic agreements.Footnote 9 However, rather than focus on degrees of overlap in text per se, we are interested in capturing particular and theoretically meaningful dimensions of these treaties. Specifically, our coding scheme allows us to determine, across a range of provisions, whether a treaty provides for more or less “state regulatory space” (SRS), a concept that captures key tradeoffs between investment promotion and sovereignty. In cases of renegotiation, we measure how much SRS is increased, decreased, or preserved by the revised treaty, thereby shedding light on the concerns that drive treaty design and change.
How do states react to their experience with ISDS in subsequent treaty design? We develop hypotheses that consider both the extent of state involvement in disputes and their outcomes—whether states win or lose—to generate predictions about when they are likely to seek greater levels of SRS. We also ask whether states are more likely to seek change by modifying the substantive provisions of IIAs or the provisions related to ISDS itself. In the process, we consider whether termination, as an alternative to renegotiation for adjustment, follows similar patterns.
In the next section, we explain the concept of SRS, link it to growing political concerns about investment arbitration, and present the hypotheses. The third section explains our data and research design and the fourth presents the results of our analysis. A concluding section suggests implications for the IIA regime, the politics of globalization, and research on treaty design and change.
Arbitration and State Regulatory Space
States hoping to attract foreign direct investment (FDI) face a credible commitment problem. Foreign investors, worried that host governments will expropriate or otherwise undermine their investment in the future, favor locations with a lower risk of interference. By establishing legalized rules and binding arbitration, IIAs help host governments tie their hands to fair and predictable treatment of FDI.Footnote 10 This hands-tying comes with a downside, however, in the form of diminished policy autonomy for host governments, which face a variety of political and economic priorities that may conflict with foreign investors’ rights.Footnote 11 States may be willing to forego potential investments in return for a freer hand at the domestic level. As in many areas of international law, the resulting tradeoff—between constraining international agreements and sovereign flexibility—is central to the decision to seek IIAs and choices over their design.
We capture this tradeoff with the concept of SRS, defined as the ability to freely legislate and implement regulations in given public policy domains.Footnote 12 We conceive of SRS as a continuum where at one extreme governments have maximum flexibility to pursue domestic policies. At the other extreme, they are highly constrained by IIA rules and the specter of ISDS deters regulation in the public interest. SRS is a function of both substantive provisions, which determine state obligations regarding FDI, and procedural provisions related to ISDS, which outline the consequences for violating substantive rules. From both theoretical and political perspectives, SRS is arguably the most important dimension along which IIAs vary.
To understand when states are more likely to seek greater SRS, we focus on the most controversial aspect of IIAs when it comes to host state flexibility: ISDS. Beginning in the mid-1990s, the number of investor arbitration cases began to rise, accompanied by high-profile disputes that entailed large awards. To date, there are over 800 known cases of ISDS based exclusively or partially on IIAs, involving 114 countries.Footnote 13 This system of arbitration has sometimes surprised states with inconsistent and expansive rulings, raising legitimate fears of a “regulatory chill” if broadly interpreted substantive obligations, combined with the threat of arbitration, deter governments from pursuing important public policy goals.Footnote 14 Merely being challenged with an ISDS claim can be enough to affect the reputation of host governments and deter investors.Footnote 15 As a result, there has been growing public concern and even a “backlash” in some quarters, partly driven by populism and economic nationalism, against investment arbitration.Footnote 16
These concerns have spurred lively debates about IIA reform among both scholars and practitioners. Some reforms focus on substantive provisions and how they can be designed to protect investors while giving host states more room to regulate.Footnote 17 India, for example, is using a model text for its bilateral investment treaties (BITs) that narrows the definition of an investment, excludes the “most favored nation” principle, and replaces the “fair and equitable treatment” standard with a narrower list of prohibited measures.Footnote 18 Other reforms relate directly to dispute settlement provisions. The European Commission proposed a new approach to ISDS that replaces ad hoc panels with a more centralized “investment court system,” elements of which are reflected in the EU's recent economic agreements with Canada and Vietnam. Many newer IIAs contain more modest elements of ISDS reform—limiting the rules subject to arbitration or making it harder to access—and others exclude ISDS entirely.Footnote 19
These trends suggest that states are reacting to IIAs in general and ISDS in particular by seeking to preserve their sovereignty. Indeed, states involved in ISDS cases appear more likely to renegotiate or terminate their IIAs.Footnote 20 Important questions remain, however, about what motivates these practices. We offer a set of hypotheses to determine systematically whether states seek change because of sovereignty concerns that arise through ISDS experiences, and to shed light on the more precise dynamics driving renegotiation. Our underlying assumption is that states react rationally to life in the IIA regime and update accordingly (if imperfectly). As Beth Simmons notes, it may have been difficult to accurately assess the consequences of these treaties when they were originally signed and before they were interpreted via arbitration.Footnote 21 Over time, however, states learn about the costs and benefits of their legal commitments and revise their design preferences accordingly, a process that should translate into new treaty provisions. Indeed, modern IIAs are generally pursued with careful attention to detail.Footnote 22
We propose, most generally, that being involved in investment disputes informs governments about potential sovereignty costs and prompts a search for greater SRS. This could be true even for claimants’ home states because any involvement in ISDS could expose states to the details and possible consequences of IIAs. However, we expect the effect to be stronger when a state is not merely involved but is the target of claims brought by investors, which more vividly highlights the actual or potential costs of litigation.Footnote 23 We explore these possibilities with our first hypothesis:
H1
States that have been involved in more investment disputes are more likely to renegotiate agreements with greater SRS. This effect is stronger for states that have been involved often as respondents.
We consider not only the extent of a state's involvement with ISDS cases but also their outcomes. Presumably, a state that has often been on the losing end of claims is more likely to infer that its existing IIAs are not doing enough to preserve policy space in the face of investor complaints, prompting a reconsideration of the central tradeoff—sacrificing sovereignty to promote FDI—underlying investment treaties.
H2
States that lose a higher number of investment disputes are more likely to renegotiate agreements in the direction of greater SRS.
Finally, we consider whether shifts in SRS follow a different pattern if we look only at changes to ISDS provisions, as opposed to changes in substantive provisions. This distinction allows us to investigate whether concerns over ISDS per se—as the most high-profile aspect of IIAs—are driving the desire to renegotiate, as claims of a backlash against investor arbitration might suggest. By modifying ISDS provisions, political leaders can more directly and visibly address widespread populist concerns about sovereignty and international courts.Footnote 24
H3
When states seek to change SRS, they are more likely to do so by modifying ISDS (rather than substantive) provisions.
To be clear, although current discussions of IIA reform are often fueled by a perceived loss of regulatory space, this does not motivate all renegotiations. Evidence suggests that arbitration can be used to resolve disputes efficiently and that lessons from ISDS experiences are sometimes positive.Footnote 25 In fact, some states have used renegotiation to improve investor protections, even at the expense of SRS. Germany, for example, renegotiated several BITs to add “high-standard” ISDS provisions, and China reformed its IIA policy to embrace more constraints on host states, which it implemented in new and renegotiated treaties.Footnote 26 In other instances, renegotiation has addressed issues not directly related to SRS. For example, some IIAs signed by Central European countries in the 1990s were amended in the 2000s simply to accommodate their admission to the EU. This countervailing evidence suggests a plausible null hypothesis, where ISDS experiences are not associated with greater SRS in renegotiated treaties.
Finally, as we noted, treaty termination provides an alternative means for reclaiming SRS. It is unclear whether termination is primarily a substitute for renegotiation or whether it is a significantly different strategy that results from distinct concerns, in which case it would require its own theoretical explanation. The strategies of renegotiation and termination are often intertwined in practice and are difficult to distinguish analytically. In some cases, a renegotiation process ends with termination of the original IIA, and in other cases, termination is actually part of a long-term renegotiation strategy.Footnote 27 Thus, although it might be tempting to consider termination as a more dramatic step that seeks to undermine IIAs rather than reform them, this is not necessarily the case. Indeed, termination of a single IIA is unlikely to change the perceptions of foreign investors, who respond more to the overall number and orientation of a state's IIAs as the relevant signal.Footnote 28
Whether it is driven by similar or distinct processes, we should account for termination in our research design. We therefore test our hypotheses with alternative models that include terminated IIAs along with renegotiated ones. This allows us to assess the robustness of our findings and to explore whether seeking SRS through termination is explained by different factors.
Research Design
To test the hypotheses, we investigate change from the original IIA to the one replacing it (or to no IIA in the case of termination). The unit of analysis is thus the individual treaty. Given our interest in the nature and direction of change, we include all modified or terminated IIAs as the relevant set of agreements. The dependent variables are operationalized as continuous measures of change in SRS so our main analyses employ ordinary least squares (OLS) estimation with robust standard errors.Footnote 29 This section elaborates on the dependent variables, sample and data, then outlines the independent and control variables. Summary statistics and bivariate correlations of these variables are reported in the online appendix.
Dependent Variables
To gauge change in SRS, we first need to measure this underlying concept. To do so, we build on the IIA Mapping Project, a text-coding scheme developed by the United Nations Conference on Trade and Development (UNCTAD).Footnote 30 This scheme examines the most important substantive and procedural provisions of agreements and codes them on the inclusion, exclusion, or degree of various elements. The Mapping Project is designed for “raw” comparative purposes, not with SRS in mind. We have therefore adjusted the coding criteria to reflect our research interests.
We classified all relevant provisions in forty-two categories, which are grouped under eight broader dimensions of IIAs that are central to SRS. An outline of these dimensions and categories is provided in the appendix, and the full coding scheme and a more detailed justification of our approach are presented in the online appendix. The SRS score for a given treaty ranges from 0 for limited SRS (less policy space), to 1 for greater SRS (more policy space). Terminated IIAs reflect maximum policy space and therefore score a value of 1 on all aspects of SRS.Footnote 31 As we note in our discussion of H3, states might be especially motivated to alter ISDS provisions over substantive ones. To test for this, we use two versions of the dependent variable, one that includes only substantive provisions and one that includes only ISDS indicators.
The dependent variables, labeled delta srs substantive and delta srs isds, capture the difference between the SRS value of the original treaty and its renegotiated replacement. Recalling that the SRS score varies from 0 to 1, the values on these variables can range from + 1 to -1, where positive values indicate an increase in SRS and negative values indicate a decrease (and 0 indicates no change).
In one example, the 1959 Germany-Pakistan BIT—the first “modern” BIT—had an SRS value of 0.35 on substantive provisions and a corresponding value of 1 on ISDS provisions (reflecting the fact that it lacked such provisions). It was replaced in 2009 with a BIT that has SRS values of 0.25 and 0.07 on substantive and ISDS provisions, respectively. Thus, the values for this observation are -0.10 for delta srs substantive and -0.93 for delta srs isds. In contrast, the 1996 Canada–Panama BIT, with SRS scores of 0.30 and 0.26 on substantive and ISDS provisions, respectively, was replaced in 2010 with a more “progressive” BIT, with corresponding scores of 0.49 and 0.40. Thus, the values for this observation are 0.19 for delta srs substantive and 0.14 for delta srs isds.
The values for terminated IIAs are almost always positive, of course, because the change is from an agreement that includes at least some constraints on the host government to no agreement at all. For instance, the 1984 Norway–Malaysia BIT, with an SRS value of 0.255 on substantive provisions and 0.056 on procedural ones, was terminated in 2001. The values of delta srs substantive and for delta srs isds for this observation are therefore 1 – 0.255 = 0.745 and 1 – 0.056 = 0.944. We report the results with and without termination.
The data set includes 247 IIAs in total. Of these, 177 are instances of renegotiation and the remaining seventy are cases of termination. With respect to the former, we account for IIAs in force that were either replaced by a new treaty (a BIT or an investment chapter in a free trade agreement (FTA)) or amended by a protocol.Footnote 32 The online appendix provides additional information on data collection and processing, coding procedures, and the sample. Figures 1 and 2 report histograms for delta srs substantive and delta srs isds, respectively, broken down by renegotiated and terminated IIAs. They demonstrate the substantial variation on these two variables.
Figure 1. Histogram of Delta SRS Substantive, by type of change
Figure 2. Histogram of Delta SRS ISDS, by type of change
Weighting and Disaggregation
Our measure of SRS includes all forty-two provisions in our coding scheme and weights them equally. In the absence of compelling theoretical guidance on how to weigh different provisions, equal weighting is arguably the least arbitrary option. Nevertheless, it is certainly possible that some provisions have a disproportionate impact on SRS. We take a number of steps to address this concern.Footnote 33
First, we use factor analysis to extract the principal components, then use them as dependent variables. The results, reported in the online appendix, remain intact. Second, focusing on delta srs isds, which subsumes a relatively small number of more homogenous provisions, we construct an alternative version of the dependent variable that assigns different weights to the provisions, using our judgment to determine their relative importance. For example, we assign greater weight to limitations on the scope of ISDS and the requirement to turn to a domestic court as a prerequisite to international arbitration, compared to provisions that increase the transparency of ISDS proceedings. The online appendix presents relative weights of the various categories and several regression models that account for variation in this variable. They show that the different weighting has no meaningful effect on the results.
Concerns related to weighting may be especially pronounced with respect to srs substantive, which includes several dimensions of provisions that address a variety of issues, such as preambular language, definitions, standards of treatment, substantive obligations, and flexibility. In general, we believe that aggregation across these provisions is appropriate. IIAs are a “package deal” more than a collection of discrete components. For example, how an investor and investment are defined affects the impact of an IIA's standards of treatment. Alternatively, preambular language might have important implications for a state's right to regulate in certain policy areas by providing an interpretative basis for arbitration. Moreover, an examination of IIA reform efforts demonstrates that they concern a wide variety of substantive areas and are not confined to a handful of supposedly “key provisions.”Footnote 34 From this perspective, a large number of seemingly modest changes to IIA provisions can amount to a consequential shift in SRS overall.
Nevertheless, to explore possible differences across provisions, we unpack the aggregate measure of delta srs substantive in two ways for additional analysis. First, because one might argue that the preamble does not generate formal legal commitments, we exclude this category from the dependent variable. Second, we identify four sets of substantive provisions that are clearly relevant to SRS and have been subject to substantial reform efforts in recent years. These are: (1) the definition of investors and investments; (2) standards of treatment; (3) expropriation and compensation; and (4) flexibility (i.e., denial of benefits, security and public policy exceptions, and other carve-outs).Footnote 35
Independent Variables
Our hypotheses suggest that involvement in ISDS is likely to prompt treaty renegotiation that results in higher levels of SRS. We consider two sets of ISDS-related variables, discussed in turn. For of all these variables, we employ data from UNCTAD's Investment Dispute Settlement Navigator.Footnote 36
Participation in disputes
We count the overall number of investment claims that both parties were involved in from the year their shared IIA was signed until the year it was renegotiated or terminated. As we discussed earlier, we have reason to believe that states draw different lessons depending on their role, so we construct two separate variables to capture the logic behind Hypothesis 1. The first, labeled dispute respond, is the total number of investment disputes both parties were involved in as a respondent. The second, dispute claimant, is the total number of investment claims filed by investors from both parties.Footnote 37 For example, the Germany–Democratic Republic of Congo (DRC) BIT was signed in 1965 and renegotiated in 2010. Over these forty-five years, Germany was a respondent in two disputes and the DRC in four. The value for dispute respond is therefore 6. During these years, German investors were claimants in twenty-five cases and Congolese investors were in none. The corresponding value for dispute claimant is thus 25.Footnote 38 The bivariate correlation between the two variables is fairly low (r2 = ~0.15), so we include both in the same statistical models.
Dispute outcome
UNCTAD classifies each completed case with a known outcome into one of five categories: decided in favor of state; decided in favor of investor; decided in favor of neither party; settled; and discontinued. Given our expectation in Hypothesis 2 that rulings in favor of the claimant will have the most pronounced impact on changes in SRS, pro-investor ruling is the total number of cases decided in favor of the investor for both IIA parties. With respect to the Germany–DRC BIT mentioned earlier, for example, the DRC lost two cases and Germany lost none, resulting in a value of 2 on this variable. Pro-investor ruling is highly correlated with dispute respond (r2 = ~0.80) so they are reported in separate models.
The count of dispute outcomes requires two qualifications. First, some ISDS outcomes are confidential and one study finds that politically consequential disputes are more likely to remain undisclosed through official channels.Footnote 39 These may be the very cases that push governments to renegotiate or terminate their IIAs. Second, even when the outcome is known, what counts as a “win” or a “loss” is open to interpretation.Footnote 40 The results on dispute outcomes should therefore be treated with caution.
Confounding Factors
Extant research on the investment agreements suggests several additional factors that influence IIA policy and design. First, the parties’ relative level of development might affect the content of IIAs.Footnote 41 We account for this possibility with north-south iia, a dummy variable that scores 1 if the treaty involves an economically developed country, and 0 otherwise.
Next, the IIA regime has evolved in distinct waves.Footnote 42 Following the more neoliberal, pro-investor approach of the 1980s and 1990s, there was a shift toward more concern with state flexibility around the mid-2000s. Thus, period is a dummy variable that scores 1 if the year in which the treaty was renegotiated or terminated is 2005 or later, and 0 otherwise.Footnote 43 Turning to the type of IIA, replacing a BIT with an FTA with an investment chapter, rather than another stand-alone BIT, requires negotiating a wider range of issues, sometimes with more states. Chapter in fta is a dummy variable that scores 1 if the renegotiated treaty is an FTA, and 0 otherwise.
Finally, South American countries like Ecuador and Bolivia were some of the earliest to reject the IIA regime, and the United States and Canada have been prominent advocates for reform. Western hemisphere is a dummy variable that scores 1 if at least one IIA party is from the Americas, and 0 otherwise. States acceding to the EU have had to revise or terminate their IIAs to conform to EU rules. Thus, new eu member is a dummy variable that scores 1 if the treaty involves at least one party that joined the EU in the 2000s, and 0 otherwise.
Results and Discussion
Table 1 presents four OLS models with dispute respond and dispute claimant as the main independent variables. Model 1 estimates the effect of these and other independent variables on delta srs substantive for the subsample of renegotiated IIAs. Model 2 includes terminated IIAs but is otherwise similar to model 1. Models 3 and 4 replicate Models 1 and 2 for the dependent variable that includes only ISDS provisions, delta srs isds. Table 2 is similar to Table 1 except that it replaces dispute respond and dispute claimant with pro-investor ruling as the main independent variable.Footnote 44
Table 1. The sources of Delta SRS, OLS with dispute participation variables
Notes: Figures in parentheses are t statistics. *p < .10; **p < .05; ***p < .01 (two-tailed test).
Table 2. The sources of Delta SRS, OLS with pro-investor ruling variable
Notes: Figures in parentheses are t statistics. *p < .10; **p < .05; ***p < .01 (two-tailed test).
Table 3 presents twenty additional models that disaggregate SRS to examine subsets of substantive provisions. Models 9 to 12 report the effect of ISDS participation and outcomes on the version of delta srs substantive that excludes preamble-related indicators, with and without terminated IIAs. Models 13 to 16, 17 to 20, 21 to 24, and 25 to 28 repeat these model specifications for delta SRS in relation to the definition of investors and investment, standards of treatment, expropriation and compensation, and flexibility, respectively. All models include the set of controls listed in Tables 1 and 2 (these are not presented for the sake of clarity).
Table 3. The sources of Delta SRS—subsets of SRS Substantive, OLS
Notes: Figures in parentheses are t statistics. All models include the following control variables: north-south iia, period, western hemisphere, chapter in fta, and new eu. *p < .10; **p < .05; ***p <. 01 (two-tailed test).
The results offer substantial support for the expectation, captured in H1, that experience with investment disputes is associated with efforts to reclaim regulatory space. They also point to several important nuances in these relationships. As expected, states with much experience as respondents in investment disputes seek greater SRS in the substantive provisions of their treaties. As models 1 and 2 in Table 1 demonstrate, dispute respond is positive and statistically significant whether terminated IIAs are included in the sample or not. The results reported in the online appendix indicate that this conclusion is robust to alternative measures of the dependent and independent variables as well as different model specifications. The substantive effect of this variable is also sizable. Based on model 2 in Table 1, an increase in one standard deviation of dispute respond (about ten claims) increases substantive SRS by about 0.11. To put these numbers in perspective, they suggest that it takes about twenty-five claims filed against both parties to move from a traditional IIA, such as the 2001 Germany–Morocco BIT where substantive SRS equals 0.13, to a much more progressive IIA, such as the 2015 Turkey–South Korea BIT where substantive SRS equals 0.43. Thus, as the number of claims they face mounts, states attempt to regain regulatory flexibility in the substantive provisions of renegotiated IIAs.
The effect of dispute respond on ISDS provisions is weaker. Although the effect of this variable on delta srs isds remains positive and statistically significant in models that include terminated IIAs, it loses statistical significance in models that exclude them. This is apparent in model 3 in Table 1 and in additional models presented in the online appendix. This suggests that states with greater involvement in ISDS are more likely to focus on SRS in substantive rather than procedural provisions in their treaty renegotiations, going against the logic of H3. The distinct results when terminated IIAs are included is interesting. Perhaps states pursue the more comprehensive option of termination when they have faced political demands that focus on curtailing ISDS.
The effect of dispute claimant is less robust than for dispute respond. dispute claimant is positive and statistically significant in models that include terminated IIAs, regardless of the type of provisions. When they are excluded, however, this variable is statistically significant only in some models pertaining to substantive SRS and is never significant in models using srs isds. Substantively, based on model 2 in Table 1, an increase in one standard deviation of dispute claimant (about sixteen claims) increases substantive SRS by about 0.085 (thus, 25 percent less than dispute respond). The positive effect of this variable, when terminated IIAs are included, might seem surprising at first: why would home states to investors benefiting from ISDS be interested in terminating their IIAs? One possible answer is that they are not, but that their partners are. That is, states whose investors filed multiple claims are more likely to see their partners denounce their agreements. For example, after Western European countries refused to renegotiate their treaties with India, which was demanding greater SRS, India moved to unilaterally terminate them. Even in the models that include terminated IIAs, however, the substantive and statistical effects of dispute claimant are smaller than for dispute respond, suggesting that the latter has a greater impact on governments’ inclination to pursue greater SRS. This might emanate from greater sensitivity to the direct costs associated with claims against them than to the indirect benefits resulting from the legal remedies available to their home investors. On the whole, consistent with H1, the empirical analysis indicates that being on the receiving end of ISDS has a more pronounced effect than being a home country to claimants in terms of subsequent treaty change.
Regarding H2 and the effect of dispute outcomes, the results are rather mixed. As Table 2 reports, pro-investor ruling (implying host state losses) is positive and statistically significant in the two models pertaining to substantive SRS and the model pertaining to ISDS SRS and including terminated IIAs. However, this variable is statistically insignificant in the model (model 7) using delta srs isds and excluding them. The substantive effect of this variable appears meaningful, albeit smaller than for dispute respond. Based on model 6 in Table 2, an increase in one standard deviation of pro-investor ruling (2.55) increases substantive SRS only by about 0.04. At the same time, it appears that the effect of dispute outcome is vulnerable to some (but certainly not all) alternative model specifications. For example, pro-investor ruling is statistically insignificant when only the cases lost by the more powerful party are counted (see online appendix). As we discussed earlier, incomplete information on the outcome of some disputes may lead to underestimating their impact.
Turning to the dependent variables that reflect subsets of substantive provisions, it appears that excluding the preamble category does not affect the results. All three explanatory variables remain positive and statistically significant at a 90 percent level of confidence or higher. Thus, the results are not driven by changes in this particular category. The results for the other four subsets of provisions offer several insights. First, dispute respond and dispute claimant remain positive and statistically significant in models that include terminated IIAs, while pro-investor ruling becomes statistically insignificant in all models except the one pertaining to flexibility. This is consistent with the generally weaker results on dispute outcome. When terminated IIAs are excluded, dispute respond remains positive for provisions related to definitions and flexibility, but not for standards and expropriation and compensation. Dispute claimant, on the other hand, is insignificant in all models. In the models in which both variables are statistically significant, the substantive effect of dispute respond is, again, somewhat stronger than the effect of dispute claimant. These results reinforce the conclusion that being a target of ISDS claims has a greater effect on SRS in renegotiated IIAs, compared to being a home state to a claimant.
The results also shed light on which provisions are most often renegotiated in the aftermath of investment disputes. Surprisingly, perhaps, states appear to tackle “auxiliary” provisions, such as how investors and investment are defined, and provisions that qualify their obligations with flexibility and carve-outs, rather than the standards and obligations themselves. These seemingly secondary provisions can have substantial implications for SRS and, indeed, they have been the subject of many IIA reforms.Footnote 45 Further examination of this variation and the factors behind it is a promising avenue of future research.
The control variables behave mostly as expected. Period is positive and statistically significant in most models, consistent with the observation that states have sought to “rebalance” the global investment regime in recent years. Western hemisphere is also positive and almost always statistically significant, indicating that the Americas have led the way in seeking greater SRS through treaty renegotiation and termination. New eu member changes signs and levels of statistical significance across models. A closer look at treaties associated with countries that acceded to the EU in the 2000s shows that renegotiation led to little, if any, change in SRS. Looking at different types of IIAs, chapter in fta is positive and statistically significant in the models that exclude terminated IIAs, suggesting that FTAs with investment chapters that replace older BITs usually embody higher levels of SRS than do renegotiated BITs. Finally, north-south iia is statistically insignificant in most models.
A look at the entire set of results related to investment disputes suggests three important conclusions. First, states are especially responsive to their experience as respondents to investment claims. As their exposure to such claims increases, they appear more likely to seek greater SRS—that is, to reclaim their sovereignty—through the renegotiation or termination of their IIAs. The impact of serving as a home to claimants, and of dispute outcomes, is also positive but less pronounced and robust. Second, the models that include terminated IIAs produce, on the whole, stronger results, suggesting that in the aftermath of ISDS involvement states often prefer to terminate entire treaties rather than adjust their content. Third, in the models that exclude terminated IIAs, we observe an effect on substantive provisions but not on ISDS provisions. This indicates that even in the aftermath of investment disputes, parties to IIA renegotiations appear relatively content with the ISDS procedures but pursue greater regulatory space in substantive rules. Disaggregation of substantive provisions suggests that respondent states are especially keen on narrowing the definition of investors and investments and increasing their policy flexibility. These findings tell a more nuanced story than dramatic accounts of a “backlash” against investor arbitration imply and show that there is room to recalibrate a variety of provisions to address sovereignty concerns.
Conclusion
Although various observers and stakeholders associate change in the rules of the IIA regime with dissatisfaction over investor-state arbitration and the regime's presumed pro-investor bias, little systematic evidence has been brought to bear on the timely question of what motivates these efforts and what treaty obligations are in fact changed.
Using new data on the degree of state regulatory space, or SRS, reflected in about 250 renegotiated and terminated IIAs, we examine the relationship between exposure to investment arbitration, on the one hand, and change in the content of existing treaties, on the other. This allows us to relate the burgeoning literature on ISDS to important debates about the design and legalization of IIAs.Footnote 46 The empirical results indicate that, indeed, experience with ISDS leads to greater SRS in renegotiated treaties and, even more clearly, to their termination. This effect varies, however, with the nature of involvement in ISDS and with respect to different types of treaty provisions. The number of cases brought against a state has the strongest impact on changes in SRS. Being the home to a claimant and losing cases also matter but these effects are weaker. In addition, the results indicate that when states renegotiate their IIAs they focus more on the substantive rules than on changing ISDS provisions. This suggests that governments are less concerned about the institutional arrangements for settling disputes than with recalibrating the protections they guarantee.Footnote 47 These intriguing findings call for further analysis of how states learn from and react to ISDS and of what factors prompt them to focus on changing certain provisions.
As tactics to shape the IIA regime, renegotiation and termination should become even more important because the initial duration of hundreds of IIAs are set to expire in the next several years.Footnote 48 Although our findings are based specifically on renegotiated and terminated treaties, understanding these practices should give us insight into the preferences and design choices that are driving the evolution of the global investment regime more generally. Focusing especially on the role of ISDS in motivating change should produce lessons for how states can successfully navigate this politically controversial aspect of IIAs, balancing investor rights and sovereignty concerns in the process. These lessons could also apply to other domains of economic globalization that increasingly force states to reconcile their international and domestic priorities and their economic and political interests. For governments seeking to preserve the benefits of globalization while distributing its impacts in a more balanced way, recalibrating international commitments can be a useful strategy alongside changes at the domestic level.Footnote 49
Our study also suggests more general lessons for those interested in international agreement design and change. The literature on agreement design has evolved considerably, aided by the availability of new, more detailed data.Footnote 50 Arguably, deliberate and well-informed choices to change existing agreements tell us even more about design preferences than initial negotiations, which often take place with less information about the consequences of alternative design choices.Footnote 51 In terms of the specific factors that lead to change, the lessons from IIAs should apply especially well to other regimes with strong dispute settlement mechanisms and that entail significant sovereignty costs for states. Our findings are more likely to generalize to other bilateral and regional settings, without the larger numbers of states and distinct strategic dynamics of multilateral agreements. For dissatisfied states, withdrawal may be the only viable option in these settings, although, like termination, it risks undermining some long-term benefits of membership.Footnote 52
Supplementary Material
Supplementary material for this research note is available at <https://doi.org/10.1017/S0020818319000195>.
Appendix
Table A1. Coding state regulatory space in IIAs: variables, dimensions, and categories
Acknowledgments
The authors’ names appear in random order and indicate equal authorship. For helpful comments, we are grateful to Patrick Bayer, Axel Berger, Xinyuan Dai, Ben Graham, Soo Yeon Kim, Krzysztof Pelc, Rachel Wellhausen, and Wen-Chin Wu, as well as two anonymous reviewers and the editor. We thank Elisabeth Tuerk and UNCTAD for kindly sharing and assisting us with their Mapping Guide. Earlier versions of this research were presented in the International Political Economy Society virtual workshop, the Political Economy of International Organizations conference, the APSA annual meeting, and the Brown Bag Lunch program at the International Centre for Settlement of Investment Disputes. We thank participants in those forums for their feedback. Or Avrahami, David Chriki, Alex Fischer, Moshe Goldman, Lea Gutowicz, Valentine Herzl, Liza Holodovsky, Nir Kosti, Maayan Morali, and Keren Sasson provided valuable research assistance.
Funding
This research was supported by a Marie Curie FP7 Integration Grant within the seventh European Union Framework Programme.
