The Negotiation, Diffusion, and Legacy of NAFTA Chapter 11: An Empirical Eulogy

Abstract

Following a three-year post-termination transition period to bring investor-state arbitration disputes, the investment protections afforded by Chapter 11 of the North American Free Trade Agreement (NAFTA) finally expired in June 2023. Chapter 11 was one of the most litigated, cited, commented, and copied investment treaties. An important, but largely ignored, part of its legacy is how the making of NAFTA Chapter 11 shaped its subsequent successful diffusion. Combining traditional legal assessment with computational text-as-data analysis, this article shows how the give and take during the negotiations generated buy-in on the part of Mexico and Canada and emulation by Latin American countries who helped to spread NAFTA Chapter 11 language globally. The link between the making and diffusion of NAFTA Chapter 11 highlights the power of negotiated compromise: sharing the pen with others may sometimes be the most effective way to write the rules that come to shape the world.

Résumé

Après une période de transition de trois ans pour soumettre des différends d’arbitrage investisseur-État, les protections en matière d’investissement offertes par le chapitre 11 de l’Accord de libre-échange nord-américain (ALÉNA) ont finalement expiré en juin 2023. Le chapitre 11 a été l’un des traités d’investissement les plus contestés, cités, commentés et copiés. Une partie importante, mais largement ignorée, de son héritage est la façon dont l’élaboration du chapitre 11 de l’ALÉNA a contribué à sa diffusion réussie par la suite. Combinant l’évaluation juridique traditionnelle avec l’analyse du texte en tant que données, cet article montre comment les concessions mutuelles au cours des négociations ont généré l’attachement du Mexique et du Canada et l’émulation des pays d’Amérique latine qui ont contribué à diffuser le langage du chapitre 11 de l’ALÉNA à l’échelle mondiale. Le lien entre l’élaboration et la diffusion du chapitre 11 de l’ALÉNA met en évidence le pouvoir du compromis négocié: le partage de la plume avec d’autres peut parfois être le moyen le plus efficace d’écrire les règles qui dominent le monde.

1. Introduction

It marked the end of an era. On 30 June 2023, the transition period to bring investor-state dispute settlement (ISDS) claims under Chapter 11 of the North American Free Trade Agreement (NAFTA) expired.¹ With ninety-one disputes and more than 400 million dollars of awarded damages, NAFTA has been among the most litigated international investment agreements (IIAs).² NAFTA’s successor, the Canada-United States-Mexico Agreement (CUSMA), which entered into force on 1 July 2020, lacks a general ISDS mechanism and only provides for substantively circumscribed investor-state arbitration between the United States and Mexico.³ To soften the blow, CUSMA’s Annex 14-C gave investors covered by NAFTA’s investment protections a grace period of three years to bring outstanding claims. With this period now expired, the era of frequent and widespread intra-North American ISDS is over.⁴

More broadly, the termination of the remnants of NAFTA Chapter 11 marked the end of an international investment agreement that had shaped the field like no other. NAFTA was the first investment treaty to be litigated heavily. Its awards filled early textbooks and were read by students and scholars of investment arbitration around the world. NAFTA was also the first IIA between developed countries enforced via ISDS. US investors suing Canada were its dominant users dispelling the myth that ISDS would primarily involve developing country respondents. Finally, many of the most consequential normative questions, ranging from host states’ right to regulate to the impact of interpretations by the contracting parties, first emerged in NAFTA. In many ways, NAFTA’s Chapter 11 was the petri dish in which today’s investment arbitration system formed.

Investment lawyers will continue working on NAFTA claims for years to come. Eighteen disputes were filed during the three-year transition period of CUSMA’s Annex 14-C.⁵ Some raise complex substantive questions of relevance beyond North America. For instance, the first such NAFTA “legacy claim,” Koch Industries v Canada, concerns the disbandment of Ontario’s cap-and-trade system after a change in provincial government and will likely inform debates about climate change legislation elsewhere.⁶ Other disputes fall into a jurisdictional grey zone involving measures adopted after NAFTA’s termination on 1 July 2020. This includes a claim against US President Joe Biden’s politically sensitive revocation of the Keystone XL pipeline in early 2021 — a project partly financed by the Canadian province of Alberta.⁷ Post-dating NAFTA’s termination, these claims could be rejected by future tribunals for a lack of jurisdiction.⁸ Even so, plenty of claims filed under CUSMA’s Annex 14-C relate to events prior to NAFTA’s termination and will ensure that at least some NAFTA litigation will last well into 2020s or even 2030s.

Although disputes based on NAFTA will continue for some time, the formal termination of Chapter 11 closes the curtain on an epoch and provides an opportunity to evaluate the legacy of NAFTA Chapter 11. Many of NAFTA’s “firsts” are well known, and perhaps no other IIA has attracted more scholarship. Yet, as this article will show, an important aspect of NAFTA Chapter 11’s legacy remains under-appreciated. Combining two different datasets on the negotiation of NAFTA Chapter 11 and on the uptake of its language in other investment treaties, this article shows how the making of NAFTA Chapter 11 shaped its diffusion. The negotiation of Chapter 11 required significant compromises from all three NAFTA parties. Canada and Mexico had initially been sceptical of the United States’ proposal to include an ISDS-enforceable investment chapter into NAFTA. The two states then became the most enthusiastic proliferators of Chapter 11 language in subsequent agreements with third parties. Chapter 11’s design innovations vetted through intense NAFTA bargaining came to inspire a new generation of investment treaties globally. As this article shows, the give-and-take negotiations were crucial for securing the buy-in of Mexico and Canada, which, in turn, was instrumental for NAFTA Chapter 11’s lasting impact.

NAFTA Chapter 11’s legacy underscores the power of compromise and offers important lessons for negotiators and international lawyers beyond investment law. Treaty making with an eye to national interest alone may produce short-term gains, but the NAFTA Chapter 11 experience suggests that negotiations where all states can claim wins and no state dominates can secure broad-based support and long-term gains. Unilateral treaty design innovation has spread poorly in investment treaty law. In contrast, the case study of NAFTA Chapter 11 suggests that compromise-forged innovations arising from inter-state negotiation can inspire emulation by others. This article thereby recalls the virtues of compromise for international lawyers at a time when states compete on zero-sum terms over who is to write the rules of future economic governance.

This article, first, provides a brief introduction to international investment law and highlights the known importance of NAFTA Chapter 11 in that space. Second, it introduces the dataset and methodology to trace the emergence and diffusion of Chapter 11 empirically. Third, it applies this methodology to revisit the making of NAFTA Chapter 11, highlighting the compromise-laden nature of negotiations. Fourth, it tracks the diffusion of NAFTA Chapter 11 over time and space. Fifth, it contrasts the lasting impact of NAFTA to that of unilateral treaty innovations. Finally, it concludes with reflections on the under-appreciated power that can emanate from negotiated compromise.

2. International investment agreements and NAFTA

IIAs are chiefly concerned with the protection of assets of national investors abroad.⁹ Whereas, historically, bilateral investment treaties (BITs) have governed such foreign investment relations, more recently, free trade agreements (FTAs) with investment chapters include investment protection clauses. NAFTA was the first of such agreements that made investment protection standards enforceable through investor-state arbitration and thereby sparked a surge in the growth of investment claims.

A. BITs

How to protect private assets abroad from undue foreign government interference has been a long-standing concern for capital-exporting countries.¹⁰ Foreign investors are perceived as particularly vulnerable to discrimination, expropriation, and unfair treatment by their host states without credible recourse due to ineffective or potentially biased domestic courts.¹¹ Consequently, they look to their home states for protection.¹² Even though customary law provided a minimum standard of treatment that could be enforced by the host state by exercising diplomatic protection of its nationals,¹³ this standard was vague, contested, and depended on the home state’s discretion for its enforcement.¹⁴ After the Second World War, the BIT became the instrument of choice to protect investors more effectively.¹⁵ BITs were international treaties that created obligations to protect the investments made by nationals of the other party against uncompensated expropriation, unfair treatment, or discrimination. Whereas early BITs equally depended on the home state to enforce these commitments through the inter-state dispute settlement mechanisms under the treaty, BITs from the late 1960s onwards began to include consent to ISDS.¹⁶ Aggrieved foreign investors could thus directly bring a claim of violation against the host state before international investment arbitration tribunals. The International Centre for Settlement of Investment Disputes (ICSID), created in 1965 under the auspices of the World Bank, was the most popular of such fora.¹⁷

Although formally reciprocal, most BITs were de facto highly asymmetrical. First, they were primarily concluded between capital-exporting developed countries and capital-importing developing countries.¹⁸ In exchange for promising to protect investment stock, developing countries, in turn, expected an influx of capital.¹⁹ The logic was that the additional protection afforded would make investment less risky and, thus, cheaper.²⁰ That promotional aspect of investment protection, however, has thus far not been empirically proven.²¹ Second, BITs were asymmetrical because they closely followed the treaty model of developed states.²² Most capital-exporting states approached negotiations with detailed treaty templates that were often signed on to by developing states without meaningful negotiation.²³ Empirical research has shown the existence of pronounced rule-taker and rule-maker dynamics that result in developed states having highly consistent treaties that closely follow a national treaty template, whereas developing states have signed on to a patchwork of differently worded treaties.²⁴

BITs proliferated globally particularly after the end of the Cold War. In the late 1990s, on average, four BITs were signed each week; by the 2010s, their total number had grown to nearly three thousand treaties.²⁵ Since the early 2000s, however, the conclusion of new BITs has slowed down, and, recently, the termination of old BITs has outstripped the conclusion of new ones.²⁶ In parallel, a second type of investment treaty has become more important — FTAs with investment chapters.

B. FTAs with investment chapters

For most of the post-Second World War era, international investment and trade regulations were handled by different sets of instruments. While BITs protected the property of investors abroad, the multilateral trading system — first, through the General Agreement on Tariffs and Trade (GATT) and, later, the World Trade Organization (WTO) — together with bilateral or regional custom unions and FTAs were liberalizing trade.²⁷ Both regimes began to interact more closely through the emergence of free trade agreements with investment chapters. Even though trade agreements had sporadically contained investment-related provisions, it was only with the 1988 Canada-United States Free Trade Agreement (CUSFTA) that an entire chapter was devoted to the protection of foreign investment.²⁸ In contrast to its contemporaneous BITs, however, the agreement fell short of providing investors with access to international investment arbitration to enforce the commitments made under the treaty. It was only with the conclusion of NAFTA and its Chapter 11 in 1992 that an agreement emerged, which provided protection equivalent to that found in BITs.²⁹ Since NAFTA, around four hundred treaties with investment provisions, most of them FTAs with investment chapters, have been concluded.³⁰

The creation of NAFTA Chapter 11 marked a turning point for several reasons. First, NAFTA broke with the paradigm of concluding investment protection treaties that were enforceable through international arbitration only amongst asymmetric partners. Even though ISDS in Chapter 11 was initially proposed by the United States to discipline Mexico,³¹ it had the effect of also providing US investors in Canada and Canadian investors in the United States with direct recourse to international tribunals. This was all the more remarkable since the traditional justification for investment arbitration, as discussed above, rested on biased or ineffective domestic court systems most closely associated with less developed countries. Second, NAFTA provided a blueprint for how investment and trade regulations could be regulated through the same instrument.³² This joint governance had become of growing relevance as trade and investment transactions became increasingly intertwined through the rise of global value chains.³³ FTAs with investment chapters could respond to these new business realties more holistically.³⁴ NAFTA was thus a watershed in the evolution of investment treaties deviating from the practice of North-South BITs and created new links between trade and investment governance.

C. Rise of investment arbitration claims

NAFTA’s legacy is not limited to inaugurating a new type of investment treaty. It can also be credited with giving rise to widespread investor-state arbitration.³⁵ Prior to NAFTA, investment treaties had routinely included ISDS clauses, but they had scarcely been used. The first treaty-based ISDS claim, AAPL v Sri Lanka,³⁶ was launched in 1987. In the following decade, only ten more claims had been filed. This situation changed when industrious lawyers in Canada and the United States began to realize the potential of NAFTA Chapter 11. In the three years from when the first case was submitted under NAFTA in 1997 — Ethyl Corporation v Canada — ten NAFTA cases were filed.³⁷ NAFTA quickly became the most litigated investment treaty. Until July 2023, over 1,361 treaty-based ISDS cases were filed, ninety-one of which were based on NAFTA Chapter 11, second only to the Energy Charter Treaty, with its fifty-three member states, which involved 133 claims.³⁸

The first wave of NAFTA awards proved highly formative for the field, creating a focal point for academics, students, and practitioners. Investment arbitration did not exist as a field of study or practice until the late 1990s and early 2000s. Early NAFTA cases then sparked extensive academic commentary.³⁹ Casebooks and textbooks on investment arbitration began emerging in the 2000s and relied extensively on early NAFTA awards.⁴⁰ Practitioners and tribunals began citing and reasoning based on the growing case law under NAFTA in new investment disputes.⁴¹ Even outside NAFTA, tribunals made frequent reference to NAFTA awards, turning Chapter 11 into the most cited treaty source of jurisprudence.⁴² This interest in NAFTA was helped by the wealth of normative controversies that surfaced in early Chapter 11 cases. The first NAFTA case — Ethyl Corporation v Canada — involved the banning of a harmful gasoline additive and highlighted the potential for conflict between investment protection and governmental regulation in the public interest.⁴³ Subsequent NAFTA cases involved interpretations of core investment protection concepts that are found in most investment treaties such as fair and equitable treatment or indirect expropriation.⁴⁴ In 2005, the Methanex v United States award, for example, cemented the oft-cited distinction between a state’s exercise of police powers and an indirect expropriation.⁴⁵ Early NAFTA cases also grappled with core questions that would occupy the fields for decades from transparency and participation in ISDS to the role of contracting parties following the NAFTA parties’ authoritative interpretation in 2001.⁴⁶ The surge of investment litigation under NAFTA ensured that the treaty and the awards it produced shaped investment arbitration scholarship and practice.

In summary, NAFTA Chapter 11 produced many firsts. It pioneered the new form of FTAs with ISDS-enforceable investment chapters, including between developed states. It triggered the emergence and development of investment arbitration as a field of study and practice. Many of the interpretative questions of lasting relevance first surfaced in Chapter 11 cases, producing awards that would be cited for decades. These firsts form an important part of the legacy of NAFTA Chapter 11. They are also fairly well known. The remainder of this article turns to an under-appreciated aspect of NAFTA’s legacy — namely, how its compromise-laden negotiation contributed to the global impact of its treaty design. The negotiation history of NAFTA Chapter 11 has attracted much scholarly interest.⁴⁷ Similarly, lawyers and political scientists have looked at the diffusion of NAFTA Chapter 11 and its impact on later treaty practice.⁴⁸ While these studies have generated important insights, they have tended to rely on qualitative analysis alone and fail to link the two developments. This article adopts a different approach. It uses quantitative text-as-data methods to provide a holistic, empirical assessment of the making of NAFTA Chapter 11 and its subsequent diffusion. It also shows how the negotiation and diffusion of the language in NAFTA’s investment chapter are deeply intertwined, which, in turn, casts the legacy of Chapter 11 in a new light and offers valuable lessons on the virtues of compromise beyond international investment law.

3. Dataset and methodology

The empirical analysis of this article relies on a painstakingly curated corpus of NAFTA Chapter 11 negotiation drafts as well as existing corpora of investment treaty texts. These texts were then used to trace the negotiation and diffusion of Chapter 11 using textual similarity measures. While this approach does have its limitations, coupled with traditional qualitative research, it can generate new, revealing insights.

A. Corpus of NAFTA Chapter 11 negotiation drafts

For most of NAFTA, interviews with negotiators are the only means to reconstruct the making of NAFTA because its negotiation documents have never been made public. The exception to this is Chapter 11. Country proposals relating to NAFTA Chapter 11 were published in April 2002 after the arbitral tribunal in Pope & Talbot v Canada requested their release to guide the interpretation of NAFTA Chapter 11.⁴⁹ As a result, forty-two original negotiation drafts between December 1991 and April 1993 that led to the creation of NAFTA Chapter 11 are now in the public domain.⁵⁰ The first draft of December 1991 contains the language originally submitted by the parties, roughly grouped together by themes. From the second document of January 1992 onwards, the negotiation drafts become integrated with bracketed texts attributing individual proposals to each of the three NAFTA countries (see Figure 1).

Figure 1. Extract from Chapter 11 Draft, 15 April 1992.

Note: Text without brackets indicates consensus text. Bracketed text is preceded by a superscript that indicates the country that proposed the language. Article numbers were only added in later drafts, and the structure changed during negotiations.

Once compared against each other over time, these country-specific proposals offer a unique window into how the talks unfolded. The differences between initial proposals help identify where country positions diverged or aligned. Furthermore, comparing a country’s original language to the final text allows one to quantify which country’s initial language most impacted the final Chapter 11 text. Unfortunately, in their original format, the NAFTA proposals do not lend themselves easily to such comparison. To create a usable corpus of negotiation drafts, the text had to be edited laboriously. First, not all attribution brackets were closed, and some were left unattributed to a specific state. By comparing successive drafts, these ambiguities could be resolved. Second, the structure of NAFTA changed considerably in the course of negotiations, which made it impossible to compare drafts over time without first aligning their structure. Moreover, the numbering of articles changed several times, adding further complexity. To make comparisons meaningful, each text segment in each negotiation draft was manually assigned to a final NAFTA article.⁵¹

The result of these processing steps is a text corpus of 117 negotiation drafts (thirty-nine successive drafts for each of the three NAFTA parties) from January 1991 to April 1993.⁵² This repository allows for the comparison of the different versions of all NAFTA Chapter 11 articles across drafts and party proposals. The dataset and a text comparison tool have been made available alongside this article.⁵³

B. Corpus of investment treaty texts

To investigate the subsequent imprint of the negotiated NAFTA Chapter 11 text on the investment treaty universe at large, this article draws on two datasets: (1) the Text of Trade Agreements (ToTA) corpus and (2) the Electronic Database of Investment Treaties (EDIT). They have complementary strengths. ToTA comprises the text of 448 FTAs signed between 1948 and 2016 and notified to the WTO, plus the text of the Trans-Pacific Partnership (TPP).⁵⁴ ToTA has two advantages. First, it streamlines treaty structures to make it easier to compare specific chapters — for example on investment — across treaties. Second, it contains texts in English, French, and Spanish. The latter matters in the context of NAFTA whose Spanish version impacted treaties concluded among Latin American countries in Spanish. EDIT, in the version used for this article, contains the full text of 3,239 BITs and 378 other investment treaties signed between 1950 and 2020.⁵⁵ It is the most comprehensive full text corpus of IIAs and annotates major content features in all texts. For non-English text, it contains a machine-translated version. On the one hand, this makes it easier to search for core investment standards across all IIAs. On the other hand, it can inflate the difference between texts with different original languages due to the machine translation. As a result, while ToTA was used to calculate textual similarity scores, EDIT was used to track frequency and occurrence of specific clauses.

C. Automated text comparison

Computational approaches to text analysis can bring unprecedented scale to legal empirical analysis.⁵⁶ Scholars have begun to deploy it to investigate the jurisprudence of courts and tribunals in order to understand the workings of legal institutions and to map international legal regimes.⁵⁷ Among the range of available computational methods, the automated comparison of treaty texts through similarity metrics has enjoyed a particular boom.⁵⁸ Conceptually, automated text comparison is useful for two reasons. First, verbatim copying from other treaties is frequent in trade and investment treaties.⁵⁹ Negotiators do not start drafting treaties from scratch but, rather, recycle and adjust already existing wording. Automatic text comparisons are uniquely suited to detect this copy and pasting and to trace the path-dependent evolution of legal language over time.⁶⁰ Second, textual similarity is an observable expression of other dynamics of interest to scholars such as power and influence. Existing research comparing negotiated outcomes to model agreements or prior treaties has been able to extrapolate which country “held the pen” during the negotiation.⁶¹

On the practical side, automated text comparison has the advantage of being relatively easy to implement once text is available. Different algorithms exist to compare the similarity of documents, and they vary in sophistication.⁶² At its simplest, however, similarity can be quantified by counting the share of words that two documents have in common. Mathematically, the share of common elements between two element sets is known as Jaccard similarity. Two identical documents will have a Jaccard similarity of 100 percent (and a Jaccard distance of 0 percent) when all textual components overlap between the pair.⁶³ This article quantifies textual similarity using such an approach based on Jaccard similarity but with a minor tweak. Given that word order matters in legal documents, simply counting the ratio of shared words may inflate their legal similarity. The two terms “not guilty to …” and “guilty not to …” contain the same words but have very different meanings due to their varying word order. One way to incorporate word order is by counting overlapping character strings rather than words.⁶⁴ The term “not guilty to …” represented in successive five-character strings (that is, “not_g,” “ot_gu,” “t_gui,” “_guil,” “guilt,” “uilty,” “ilty_,” “ilty_t,” “lty_to”) will thus have a different representation than “guilty not to …” (that is, “guilt,” “uilty,” “ilty_,” “lty_n,” “lty_no,” “ty_no,” “y_not,” “_not_,” “not_t,” “ot_to”) due to their varying word order. In the above example, the two terms of have a word-based Jaccard similarity of 100 percent but a five-character-string Jaccard similarity of only 20 percent (only the three shaded text components are common to both terms).

Textual similarity measures become meaningful once they are compared and placed into context. A larger similarity between the initial draft tabled by the United States and the final NAFTA Chapter 11, as compared to the initial texts of Mexico or Canada, for example, would suggest that the United States had a greater textual imprint on the negotiation outcome. Textual similarity can also track the pace of negotiations by tracing how draft texts inch ever closer to their final wording and identify phases and breakthroughs in negotiations. Aside from helping to understand negotiation dynamics, textual similarity can also trace treaty design diffusion.⁶⁵ When two treaties are very similar to each other, it is likely that the later treaty was influenced by the text of the earlier one. Prior research from the investment treaty context suggests that a Jaccard similarity of around 50 percent between two treaties is indicative of some textual imprint via copy and pasting because five-character-string Jaccard similarity is very sensitive to small differences between texts.⁶⁶ Accordingly, this article has used a similarity of around 50 percent as the initial threshold to look for treaties that display an imprint of NAFTA Chapter 11.⁶⁷

D. Limitations

To be sure, textual similarity implemented through such a procedure is an imperfect measure of the actual legal similarity of two treaties. On the one hand, the approach risks inflating differences. The algorithm compares text tokens, and, as a result, synonyms, for example, are treated as differences. The mere existence of different country names and formatting conventions also increases dissimilarity. On the other hand, the approach also risks inflating similarity. Two texts can be conceptually far apart, while using overall similar words. Whether two otherwise identically worded commitments are enforceable, for instance, may come down to one word — “shall” rather than “should.” Yet the algorithm treats all text tokens the same and fails to capture these nuances in semantics and legal significance. Finally, countries may copy selectively rather than wholesale. The language that is purposefully not taken or edited heavily — that is, the differences — can be more meaningful than any similarities.

There are technical and methodological ways to mitigate these shortcomings. More recent natural language-processing techniques can encode semantics and context, allowing researchers to distinguish between meaningful and mere cosmetic differences in text.⁶⁸ However, to uncover whether subsequent negotiators copied heavily from NAFTA, a cruder, token-based approach suffices. Moreover, textual similarity counts can be coupled with traditional qualitative analysis to validate quantitative findings and to correct blind spots. Throughout this article, differences and similarities that were automatically detected through algorithmic approaches were scrutinized manually to evaluate their legal significance. Finally, while selective copying took place and will be acknowledged in the analysis that follows, the focus of this article will be on the imprint of NAFTA Chapter 11, leaving questions of its tailoring in subsequent treaties to future research.

4. The making of NAFTA Chapter 11

The making of NAFTA has long captivated legal and political science scholars. Maxwell Cameron and Brian Tomlin’s The Making of NAFTA: How the Deal Was Done delivers a rich and thoroughly researched account of the path leading to the agreement.⁶⁹ Legal scholarship has also devoted considerable effort to understanding how NAFTA was created. Meg Kinnear, Andrea Bjorklund, and John Hannaford’s commentary of NAFTA Chapter 11 provides a careful review of the different drafts that were tabled during negotiations for each of the articles.⁷⁰ This is important because, under customary international law, the travaux préparatoires are a subsidiary means of treaty interpretation.⁷¹ Text-as-data analysis, coupled with contextual analysis, can offer an alternative lens to understanding how the language of NAFTA Chapter 11 emerged. At the outset of negotiations on NAFTA Chapter 11, it was far from a foregone conclusion that an agreement could be struck. The United States, Mexico, and Canada all approached the investment chapter from different backgrounds, resulting in vastly different objectives and textual preferences. It was the consensual give and take in negotiations that made NAFTA Chapter 11 possible. The negotiations of NAFTA Chapter 11 can be considered an outlier in a treaty universe marked primarily by asymmetric bargaining and dominant unilateral treaty design preference.

A. Negotiation context

Investment protection was an issue area hitherto largely reserved to BITs. The United States then proposed to include an investment chapter in NAFTA. ⁷² The United States’ objectives were twofold. On the one hand, it sought to offer access and protection to its investors in relation to the Mexican market; on the other hand, it aimed to impose further investment disciplines on Canada.⁷³ US-Mexican bilateral investment relations had been marked by friction for most of the twentieth century. Mexico had engaged in widespread expropriations of US oil companies in the 1930s, which spoiled the investment climate for decades.⁷⁴ Moreover, while the United States advocated the existence of an international minimum standard of treatment enshrined in customary international law, which, amongst others, entitled foreign investors to “prompt, adequate, and effective” compensation in case of expropriation, Mexico rejected the existence of such a standard and instead espoused the “Calvo Doctrine,” including in its 1917 Constitution, according to which foreigners did not enjoy greater rights than nationals.⁷⁵ Consistent with these opposing positions, the United States sought to buttress the protection of foreign investors through international law — first, through Friendship, Commerce and Navigation treaties and, from the 1980s onward, through its BIT program — whereas Mexico did not conclude any investment treaties.⁷⁶

These positions, however, began to edge closer in the mid-1980s. Faced with a severe economic and debt crisis, Mexico began embracing a widespread program of economic reform and liberalization.⁷⁷ NAFTA was part of that agenda in order to lock in domestic reform and attract American capital to Mexico.⁷⁸ With these objectives in mind, Mexico was prepared to change its stance on investment law while, at the same time, remaining within the bounds set by the Mexican Constitution. This greater openness of Mexico made the country more susceptible to US proposals, but, as Cameron and Tomlin point out, it also created a challenge for US negotiators to exploit this change of position while not giving the impression that NAFTA would lead to a massive outflow of US capital to Mexico, which would make the agreement harder to sell in the United States.⁷⁹

US-Canadian investment relations had not been easy either. Under Prime Minister Pierre Trudeau, Canada engaged in a restrictive policy towards foreign investment in the 1970s, imposing requirements on the use of inputs, screening projects, and making investment approvals subject to a “benefit-to-Canada” test.⁸⁰ The United States therefore pushed for investment disciplines as part of the CUSFTA. ⁸¹ Even though the agreement did impose strict national treatment obligations and constraints on the use of performance requirements, it continued to allow Canada to screen foreign investors under the Investment Canada Act. ⁸² As part of NAFTA, the United States therefore wanted to remedy this shortcoming in order to improve access to the Canadian market for US capital. Canada, on the other hand, was eager to defend the CUSFTA status quo in its investment relations with the United States while securing higher protection for its investors in Mexico.⁸³

Finally, Canada-Mexican investment relations, at the time of NAFTA negotiations, were minimal. Canada had little investment stock in Latin America more generally.⁸⁴ Whereas NAFTA was also an opportunity to open new investment possibilities, Canada’s primary incentive for joining NAFTA talks after they had been started on a bilateral level between Mexico and the United States was to prevent a hub-and-spoke system in North America in which Mexico had better access to the US market than Canada.⁸⁵ It was against this policy background that NAFTA negotiations on investment started in December 1991.

B. The initial positions of the three NAFTA parties

The parties tabled their first proposals in late 1991. This yet unconsolidated negotiation text of December 1991 offers the best window into the original starting positions and textual preferences of the parties. Whereas the United States approached the negotiations with its BIT model, which contained extensive protection standards enforceable through ISDS, Canada used the CUSFTA investment chapter as a blueprint.⁸⁶ Mexico, not having signed any prior investment treaties, submitted purposefully assembled language while letting its position mature as the negotiations proceeded.

i. United States

The United States tabled language principally drawn from its BIT practice. By the time NAFTA talks commenced, the United States had concluded seventeen BITs, the most recent of which was with Argentina in November 1991. The lengthy first article in the December 1991 proposal submitted by the United States was dedicated to the establishment and treatment of investment and closely tracked the language of the earlier US-Argentina BIT, notably Article II and Article IV(3).⁸⁷ While many of its core investment protection norms, ranging from fair and equitable treatment to the compensation for losses, made it into the final NAFTA text, albeit with textual tweaks, others did not. An obligation to provide effective means for enforcing claims commonly found in early US BITs, for instance, was part of the US proposal, but not of the final NAFTA Chapter 11 text.⁸⁸ Other articles of the US proposal also closely trailed its BIT practice. From its language on expropriation and freedom of capital transfer to the public order and security exception, the proposal mirrored its BIT language.

The United States argued in a later dispute that NAFTA’s investment chapter is “no more than a BIT dropped into a free trade agreement.”⁸⁹ Nowhere is this mindset more apparent than in Article 12 of its NAFTA Chapter 11 draft on post-termination coverage, where the proposal stipulates that the chapter’s obligation shall remain effective for another ten years after the termination of NAFTA. Such survival clauses are common in BITs but rare in FTAs. Ultimately, the proposal was dropped, but it is still ironic that, whereas the United States wanted parts of NAFTA to remain in force for a decade post-termination, the country advocated for the very opposite in the CUSMA negotiations. There, the United States had proposed an automatic expiry of the agreement after five years unless the parties agreed to an extension, which after negotiations became a sixteen-year sunset provision in CUSMA Article 34.7.⁹⁰

When it comes to the enforcement of investment protection, the US proposal clearly recycled prior BIT language. The ISDS clause in Article 7 of the US draft is taken almost verbatim from Article VII of the Argentina-USA BIT with a textual similarity of 93 percent and linguistic variations amounting to stylistic rather than substantive changes. Apart from the strict time rules in which decisions have to be rendered, the US proposal also copied from the Argentina-USA BIT on inter-state arbitration. This heavy reliance on prior BIT practice on enforcement also marks a strong deviation from the CUSFTA text, which did not contain an ISDS clause and provided for a general inter-state dispute settlement outside the investment chapter. Whereas the United States crafted its proposal around prior BIT language rather than the recent investment chapter with Canada, it followed CUSFTA in one limited aspect. Earlier US BITs had only contained rudimentary language on performance requirements.⁹¹ Part of the achievement of CUSFTA had been to limit Canada’s ability to link investment approvals to additional export or local purchase requirements. Article 2 of the US proposal drew on CUSFTA language in order to preserve and expand the standard set in that agreement. In conclusion, the United States came to the table with language largely taken from its prior BITs signed with developing countries, supplemented by a few obligations from its earlier deal with Canada.

ii. Canada

Canada, in contrast, wanted to preserve the bargain struck in CUSFTA to the greatest extent possible. While it was in Canada’s interest to secure strong protective obligations for its investors in Mexico, similar to those found in Canadian BITs, these capital-exporting concerns were outweighed by capital-importing considerations.⁹² Specifically, Canada sought to retain the ability to screen incoming investment under the Investment Canada Act. CUSFTA and not prior Canadian BITs thus heavily inspired the Canadian proposal. Key in Canada’s proposal was language from Article 1608(1) of CUSFTA that carved out decisions taken under the Investment Canada Act from dispute settlement. Also, Articles 404–06 of the December 1991 Canadian proposal on performance requirements, expropriation, and transfer respectively all recycled CUSFTA language. CUSFTA’s clause on monitoring of incoming investments in CUSFTA Article 1604 was similarly incorporated into the Canadian proposal.

Two additional considerations set the Canadian proposal apart from the American draft. First, Canada approached the investment chapter more holistically. Rather than considering it as a BIT dropped into a FTA, Canada sought to create linkages between issues. Its 1991 December proposal thus contained a general national treatment clause in Article 105, which applied to “goods, services and service providers, investors and suppliers.” Similarly, it included a general national security exception clause in Article 110, which differed from the investment-specific public order exception of the American proposal. Second, Canada linked investment closely to competition policy. Articles 408–11 of its December 1991 proposal dealt with monopolies, state enterprises, technology consortia, and competition matters. The United States, in contrast, only mentioned state enterprises in its original draft and noted that corresponding language may be inserted elsewhere in NAFTA. In short, Canada and the United States seemed to approach the investment with very different mindsets.

iii. Mexico

Mexico’s initial proposal is perhaps the most curious case of the three NAFTA states. In contrast to Canada and the United States, Mexico did not have a prior investment treaty practice to draw from. As Cameron and Tomlin explain, this created a negotiation dynamic by which the other NAFTA parties sought to sway Mexico towards their preferred language.⁹³ At the same time, Mexico also faced competing policy considerations. On the one hand, for historical and constitutional reasons, it could not endorse the highly investment protection-friendly American proposal.⁹⁴ On the other hand, economic considerations — specifically, the need to lock in domestic reforms and to attract foreign investments — made the Canadian CUSFTA approach less attractive. The language that Mexico submitted in December 1991 reflected this ambivalence. Across a wide range of issues, it copied directly from CUSFTA, even more so than Canada. From the scope and coverage clause (Article 2101) to national treatment (Article 2102), performance requirements (Article 2104), monitoring (Article 2105), transfers (Article 2106), and existing legislations (Article 2107), it closely tracked the structure and language of CUSFTA Chapter 16. At the same time, Mexico was also selective in its copying. Most importantly and in line with its historical and constitutional legacy, Mexico omitted CUSFTA Article 1605 on expropriation.

In again other respects, Mexico submitted sui generis language that was protective of national sovereignty. Article 2108 of its December 1991 proposal carved out investment disputes from the scope of NAFTA dispute settlement. Furthermore, Article 2109 provided highly state-centric language giving parties the broad right to deny investors protection on national security grounds. Beyond that, Mexico’s proposal left several issue areas unaddressed. The proposal did not contain language on general standards of treatment or most-favoured-nation treatment. Moreover, it also did not explicitly deal with neighbouring fields such as competition policy or taxation that were mentioned in the US and Canadian drafts. By not having had a prior investment treaty policy, Mexico’s position, more than that of the other two, was thus only beginning to emerge during the NAFTA negotiations.

C. The Negotiations of NAFTA Chapter 11

Negotiations are complex processes. The benefit of computational text analysis is to abstract from the minutiae relating to the wording of specific clauses and to trace broader negotiation patterns and dynamics. In that spirit, this section tracks negotiations related to (1) the protection of investment; (2) ISDS; and (3) exceptions. The common denominator on all three fronts is that the NAFTA parties had to depart from their starting positions to arrive at novel, compromise language that would give rise to a new breed of investment agreements.

i. Investment protection

When it comes to investment protection, disagreement between the parties turned both on the general treatment of investors and on the expropriation of assets in particular. CUSFTA had not contained any investment treatment provision beyond prescribing national treatment. At the outset of negotiations, the United States was then the only country to have submitted language on the general treatment of investment on what would later become NAFTA Article 1105. Figure 2 helps to navigate this narrative by tracing the first time each country submitted language on a NAFTA Chapter 11 provision.

Figure 2. First emergence of NAFTA proposals by parties.

Notes: This figure tracks the first time that a country submitted language on what would become a final NAFTA article, including what would become clauses in other chapters. Note that initial proposals by some states have been submitted at the same time — hence, the symbols will overlap.

The US proposed language reflected the highly protective US BIT standard. Early in the talks, during the round of negotiations in Dallas in February 1992, the parties scoped common ground. The 21 February draft shows that, whereas Mexico was accepting the US proposal on “fair and equitable treatment,” Canada agreed on the US proposal in relation to the compensation for losses in case of armed strife. In the course of the next few months, the parties’ positions converged further. In the 1 May 1992 draft, the obligation to provide “effective means” for asserting claims domestically was dropped, and the parties arrived at a consolidated text centred on the “fair and equitable treatment” standard. That text was then refined further in the ensuing months. Most notably, on 22 August 1992, the parties reversed the clause’s language turning the obligation to accord “fair and equitable treatment … in accordance with international law” into “treatment in accordance with international law, including fair and equitable treatment.”

In the Dallas round, progress was also made on the expropriation clause, which would become Article 1110. While Canada and the United States had initially submitted largely concordant language on expropriation, Mexico was against the inclusion of such a provision. In February 1992, it then showed flexibility to agree to such a clause but insisted for historical and constitutional reasons that a reference to “prompt, adequate and effective compensation” for expropriation must be dropped.⁹⁵ In the 13 February 1992 draft, the Mexican proposal instead simply referred to “compensation … [that] shall be paid within a reasonable period of time.” By 21 February 1992, the parties had agreed on a workaround that effectively enshrined “prompt, adequate and effective compensation” but in a different textual guise, referring to compensation based on “fair market value” that should be assessed through valuation criteria including “going concern value” and “asset value” and that had to be “paid without delay.” This formulation was seen as ensuring NAFTA’s compatibility with Mexican constitutional requirements and political sensibilities, while providing market-rate compensation.⁹⁶

The compromise expropriation text from February 1992 was expanded in June of the same year likely in response to negotiation developments in other chapters. A clarification was added, for example, that the expropriation clause did “not apply to the issuance of compulsory licenses” on patented goods, which were instead governed by NAFTA’s intellectual property chapter. More tweaks were subsequently made until the clause reached a close-to-final stage in September 1992. Textual similarity brings these different stories behind the negotiation of Article 1105 and 1110 to the fore in the form of a heatmap in Figure 3. The axes are the chronologically ordered Chapter 11 texts proposed by each NAFTA state, and each field of the heatmap is the comparison between the two texts with bright fields indicating low similarity and dark fields indicating high similarity. Reading the heatmaps from top left (start of negotiations) to bottom right (end of negotiation) illustrates how differently the negotiations progressed. On Article 1105 of NAFTA, the parties started with highly diverging positions (the top left mosaic), then gradually converged in April and May, before collectively fine-tuning the language by the summer of 1992. In comparison, on Article 1110, Mexico’s initial January proposal was a clearly visible outlier, and, confronted with a common US and Canadian position, agreement was quickly found around a joint text in February 1992. Collective fine-tuning, however, lasted longer and was only completed in the early fall of 1992.

Figure 3. Similarity of NAFTA proposals converging over time.

Notes: This figure tracks the textual similarity of country submitted language for each draft. Drafts are ordered chronologically, and the axes are symmetrical. High similarity is indicated by dark tiles, while high dissimilarity is indicated by bright tiles. Convergence is thus visible as a progression from red checker patterns that indicate each party’s party towards red areas. This progression differs notably between NAFTA Article 1105 and 1110.

ii. Investor-state arbitration

Investor-state arbitration was amongst the most controversial issues during the negotiation of Chapter 11. While the United States envisaged ISDS modeled on its BIT practice and proposed language to this effect at the outset of the negotiations, the other two NAFTA parties opposed it. For Mexico, it meant giving rights to foreigners that national investors did not enjoy and, hence, a break with its traditional adherence to the Calvo Doctrine.⁹⁷ For Canada, the inter-state dispute settlement under CUSFTA with carve-outs for its investment screening law worked just fine. For the first few months, the issue was relegated to the sidelines as only the United States had tabled a draft text on what would become NAFTA Article 1120 (see Figure 2). This opposition came to a head in March 1992 when Mexico proposed alternative language. Initially, as noted, Mexico had excluded dispute settlement on investment altogether. On 6 March 1992, Mexico then submitted its first language on ISDS, relegating the issue to domestic courts and stipulating: “Each Party shall provide investors of the other Parties access to an impartial judicial system with authority to enforce the rights of investors established under this Agreement.”

Canada had long refrained from submitting original language on dispute settlement. In the 4 June 1992 draft, it then aligned its position with that of the United States, accepting the enforcement of Chapter 11’s obligations through ISDS while presenting alternative language. The Canadian ISDS proposal was highly detailed and, with eighteen paragraphs and more than twenty-five hundred words, more than 2.5 times longer than the US proposal. This then marked a shift towards debates ranging from whether investor-state arbitration should be integrated into NAFTA to how it should be done and set the tone for what would be the most detailed and complex ISDS architecture of investment treaties up to that point. Much in contrast to the convergence on substance that had occurred during the 1992 spring meetings, it was only during the final round of intense negotiations in Washington in August 1992 that Mexico finally agreed to ISDS via international arbitration. By that time, the substantive part of Chapter 11 (Section A) had been largely concluded, and trust had been built amongst the parties. Mexico then worked together with the United States and Canada towards framing the new ISDS architecture.

The initial US proposal mirrored early BITs by containing only a single article on ISDS, creating a bare-bones procedures reminiscent of commercial arbitration that left the details to the arbitrators and disputants. In contrast, the newly emerging Section B of NAFTA Chapter 11 placed investment arbitration in a wider architecture of public law and public international law control mechanisms.⁹⁸ Among the novelties falling into the latter camp were Article 1128, which allowed non-disputing NAFTA parties to make submissions in ISDS disputes, and Article 1132, which enabled the three NAFTA states, as masters of their treaty, to issue joint interpretations binding on ISDS tribunals.⁹⁹ Innovation more reminiscent of a public law genre included the possibility to consolidate claims in Article 1126 and the transparency commitment by Canada and the United States (but initially not Mexico) to publish awards implicating them in what would become an annex to Article 1137. A final set of innovations clarified the different steps of the ISDS process from the conditions that must be met before a claim can be submitted to interim measures of protection and the types of remedies that can be afforded.¹⁰⁰

In short, what came out of the negotiations was a very different mechanism from what the United States had initially proposed. While, at its core, it was still investor-state arbitration, the manifold innovations jointly elaborated by the three NAFTA parties had created a more elaborate procedure that went beyond anything that existed in contemporaneous BITs and with elements more inspired by public law and public international law than by commercial arbitration.

iii. Exceptions

Whereas the United States had been able to sway Canada and Mexico when it came to ISDS, Canada was able to convince Mexico and, more reluctantly, the United States to accept exceptions reminiscent of CUSFTA. From the beginning to the end of negotiation, Canada was steadfast on demanding investment screening under the Investment Canada Act to be excluded from dispute settlement. This exclusion was finally codified in the annex to Article 1138 with Mexico registering a similar reservation. On other fronts, Canada had also argued for exceptions but was less successful in persuading its peers. As part of its January 1992 proposal, Canada wanted to make NAFTA’s investment subject to a general exception clause similar in language to GATT Article XX to protect non-economic values such as environmental or health concerns. Whereas the final NAFTA text contains an equivalent exception in Article 2101, this carve-out does not extend to Chapter 11. The NAFTA parties instead agreed between May and June 1992 to morph the exception into what would become Article 1114 on environmental measures. Whereas this article did not excuse non-compliance with an investment protection obligation, it clarified the ability of governments to ensure that investment activities were conducted in an environmentally sensitive manner as long as the government conduct is otherwise consistent with Chapter 11. NAFTA thereby set a precedent incorporating right-to-regulate language into investment treaties.

D. Evaluating the negotiations of NAFTA Chapter 11

NAFTA was a true negotiation. It was a giving-and-taking process on all sides that resulted in an agreement that looked different from what each of the states proposed going in. The necessary compromises created something new, an agreement that looked different from the types of investment agreements that Canada and the United States had concluded previously. While this may sound like a mundane, self-evident conclusion, it is not. In fact, NAFTA is very much an anomaly when looking at investment treaty negotiations generally. Investment treaty negotiations are often dominated by economically powerful rule makers that shape the terms of the treaty.¹⁰¹ For instance, the BITs the United States signed in the 2000s mirrored the language of the 2004 US Model BIT to around 95 percent.¹⁰² This suggests that the United States, and not its treaty partners, influenced the drafting of these agreements. Similarly, the language of the TPP investment chapter overlaps with the most similar US treaty to 81 percent; in fact, all the most similar agreements to the TPP are those of the United States.¹⁰³ This indicates that the TPP displayed a strong US handwriting.

NAFTA is different. In spite of the uneven bargaining power — according to World Bank data, the combined size of the Mexican and Canadian economy amounted to merely 15 percent of the US economy in 1992 — the final outcome reflected an evenly distributed give and take. Figure 4 depicts the textual similarity between the three country’s January 1992 texts with NAFTA Chapter 11. No state “won” or dominated the negotiations.¹⁰⁴ Instead, Canada and the United States are almost equidistant with a comparatively low similarity of around 40 percent. That quantifies the more qualitative impression conveyed above that all states had to significantly depart from their initial textual preferences.¹⁰⁵

Figure 4. Textual similarity of January 1992 proposals to NAFTA Chapter 11.

Notes: This figure depicts the textual similarity of the January 1992 Chapter 11 drafts submitted by Canada, Mexico, and the United States to the final NAFTA Chapter 11 texts. All states had to depart significantly from their initial proposals.

The text of NAFTA Chapter 11 also embodies a convergence of preferences and innovation formed from compromise, which sets it apart from other investment treaties before it. As noted, investment treaties typically reflect the model treaties of the dominant capital-exporting countries. Innovation, then, is primarily the product of changes in domestic politics triggering a change in model agreements rather than the compromises struck at the inter-state bargaining table. Differently put, innovation in investment law is often unilateral. Again, NAFTA was different. With no state dominating the negotiations of Chapter 11, a dynamic developed to jointly elaborate a new consensus standard especially around ISDS. This process is depicted as a heatmap in Figure 5, which like Figure 3 compares the chronologically ordered negotiation proposals of the three states, but this time for the full Chapter 11 drafts. What started as a checkerboard at the outset of the negotiations on the top left of the figure, as the proposals of each country diverged, became a joint area in the lower right corner with the texts converging around new language that differed from each of the unilateral drafts.

Figure 5. Convergence across proposals over time.

Notes: This figure tracks the textual similarity of country submitted language for each draft. Drafts are ordered chronologically, and the axes are symmetrical. High similarity is indicated by dark tiles, while high dissimilarity is indicated by bright. Convergence is thus visible as a progression from a mosaic pattern in the top left towards continuous dark areas in the bottom right.

NAFTA Chapter 11 was a product of convergence through compromise and innovation. The remainder of this article will show that these attributes of NAFTA Chapter 11 negotiations — compromise and innovation — were crucial for the successful diffusion of the NAFTA Chapter 11 text to other countries. The former generated the buy-in for NAFTA parties — specifically, Canada and Mexico — to endorse the NAFTA texts in their own negotiations with third parties. The latter created an alternative to existing BIT standards, vetted through the negotiation by the NAFTA parties, which states could turn to when seeking to innovate their practice.

5. The diffusion of the NAFTA Chapter 11 model

Scholars looking at the evolution of investment treaties have observed a spread of the “NAFTA model,”¹⁰⁶ a “NAFTA-ization,”¹⁰⁷ or a “convergence towards NAFTA.”¹⁰⁸ Political scientists speak of “policy diffusion” when prior policy choices in one country systematically condition policy choices in other countries.¹⁰⁹ As this section will underscore empirically, if defined as extensive copying of its text (around 50 percent or more), then the language of NAFTA Chapter 11 indeed diffused very successfully to other IIAs across the globe. Writing in 1993, Daniel Price was not far off the mark when he stated that “[NAFTA] Chapter 11 is the most comprehensive investment accord to date…. The chapter ought to set a standard for further multilateral and bilateral investment accords.”¹¹⁰

Why has the design of NAFTA Chapter 11 diffused so successfully? Political scientists distinguish four causal mechanisms of diffusion: coercion, learning, competition, and emulation.¹¹¹ They all may help explain the success of Chapter 11. The status of the United States as a global hegemon and NAFTA’s predominant patron added importance to the treaty, and, in bilateral negotiations, NAFTA states may have insisted on the adoption of its text (coercion). Chapter 11 was the first ISDS-enforceable investment chapter in a FTA and thus served as a template for third states aiming to conclude similar agreements (policy learning). Frequent litigation provided NAFTA Chapter 11’s provisions with a degree of interpretive predictability that language in BITs — yet to be tried and tested in dispute settlement — lacked, which turned NAFTA language into an attractive policy alternative (competition). Finally, the emergence of a new text vetted in bargaining between three diverse states may have inspired other countries to endorse its language as their own (emulation).

Although the narrative below will single out emulation as an important driver, this article is primarily concerned not with the exact causes but, rather, with the pathways of diffusion. These pathways provide clues that link diffusion back to the negotiation of Chapter 11. Diffusion depends on actors — here, states — that transfer NAFTA design. By identifying the main agents in NAFTA Chapter 11’s diffusion, this article will draw inferences on their motives and the role that the initial negotiation context played. Diffusion can occur along two pathways. Direct diffusion happens when countries advance a treaty model, which they used in the past, in new negotiations. An example of direct diffusion would be Canada using NAFTA as a template for its FTA negotiation with Chile.¹¹² Indirect diffusion, in contrast, happens when countries borrow from a template used by third states. For example, China, without having signed a FTA with the NAFTA parties, could use NAFTA as a template in negotiations with Vietnam. Both types of diffusion can generate second order effects. The countries newly exposed to NAFTA — here, Chile and Vietnam — could adopt NAFTA-style language and diffuse it to other states.

This section empirically tracks the diffusion of the NAFTA Chapter 11 text to elucidate the role its negotiation history may have played. It finds that NAFTA diffused in clearly visible phases dominated by direct diffusion paths. First, NAFTA Chapter 11 became the investment treaty blueprint for Mexico and Canada but not for the United States, and the treaties signed by the two countries set in motion a cascade of direct diffusion that led to the proliferation of the NAFTA Chapter 11 model globally. Only once the NAFTA Chapter 11 text was already entrenched can one make out a phase of indirect diffusion. This finding underscores the importance of Mexico and Canada, followed by Latin American states, as the engines for NAFTA’s diffusion.

A. NAFTA’s impact on the investment treaty design of the NAFTA parties

NAFTA Chapter 11’s impact varied strongly across the three NAFTA parties. It was most profound for Canada and Mexico, which reoriented their BIT and FTA practice around NAFTA. Only for the United States, which continued its pre-NAFTA BIT program with few modifications, did NAFTA initially remain an outlier.

i. Canada

NAFTA marked a major shift in Canada’s investment treaty program. When signing its first BIT in 1989 with Russia, Canada mirrored the practice of European countries concluding short and simple treaties that were more limited in scope than the contemporaneous US BITs.¹¹³ Following the conclusion of NAFTA in 1992, however, the length of Canadian BITs doubled as innovations from NAFTA’s Investment Chapter 11 were introduced into its new BITs.¹¹⁴ Canadian BITs from its 1994 treaty with Ukraine onwards contained (1) NAFTA-inspired language on liberalization, performance requirements, and freedom to hire senior management, which were absent in European or earlier Canadian BITs;¹¹⁵ (2) procedural innovations from NAFTA’s ISDS such as the power of tribunals to issue interim measures or the ability of state parties to make binding determinations on taxation measures;¹¹⁶ and (3) exclusions such as carve-outs for cultural industries (also found in NAFTA Article 2106) and GATT XX-type general exceptions,¹¹⁷ which Canada had unsuccessfully sought to integrate into NAFTA.

NAFTA’s imprint is even more visible in Canada’s FTA practice. In 1996, Canada signed a FTA with Chile.¹¹⁸ Its Chapter G is identical to the structure of NAFTA Chapter 11 but without Article G-15 on energy regulatory measures. Shortly after the FTA with Chile, however, Canada halted its investment treaty program after an American investor sued Canada for the first time. When Canada picked up its investment treaty again in 2006, it incorporated lessons learned from the NAFTA litigation, fine-tuning NAFTA’s language and resolving ambiguities, without departing from the structure of NAFTA Chapter 11.¹¹⁹

ii. Mexico

As noted above, NAFTA was the first investment treaty signed by Mexico. Subsequently, the country became a frequent user of both BITs and FTAs using NAFTA as inspiration. In contrast to Canada, which concluded BITs and FTAs with developing countries, Mexico signed BITs with major developed countries to attract further foreign capital. Confronted with German, UK, or Swiss model BITs, Mexico, however, fought for the inclusion of NAFTA language in elegant ways. Consider the 1995 Switzerland-Mexico BIT. While its first fourteen articles, apart from Article 5 on performance requirements, read like a conventional Swiss BIT, the annex contains major NAFTA innovations. First, it integrates a detailed ISDS architecture reminiscent of NAFTA Chapter 11 into the treaty, including provisions requiring loss to gain standing for arbitration, the consolidation of arbitral claims as well as authoritative joint interpretations by the treaty parties.¹²⁰ Second, in its annexed protocol, the treaty contains language taken from NAFTA Article 1114 on environmental measures and several exclusions specifically tied to NAFTA concessions.¹²¹

Hence, even though Mexico, in its bargaining with developed countries, could not set the terms of the agreements, it did manage to insert significant NAFTA-inspired passages. As a result, for its developed country contracting partner, the treaties with Mexico tended to be outliers within their otherwise highly consistent treaty networks.¹²² Mexico had more negotiation clout in its FTA negotiations with South American countries. Here, as detailed below, Mexico used NAFTA Chapter 11 as a template for its FTA negotiations. In short, the NAFTA experience had profoundly affected Mexico’s treaty practice.

iii. United States

The only NAFTA party on which Chapter 11 initially had only a minor impact was the United States. The country’s negotiators treated Chapter 11 as an outlier, a product of a specific negotiation context, rather than a benchmark for future treaty making.¹²³ Prior to NAFTA, the United States already had a thriving BIT program. Starting in 1982, it had concluded seventeen BITs following a relatively consistent, albeit occasionally revised, model treaty. During the two years of NAFTA negotiations, the United States signed another eight BITs with states in the former Eastern Bloc and South America. These BITs had served as templates for the original US Chapter 11 proposals, which, as discussed above, differed starkly from the final NAFTA text. Yet, whereas Mexico and Canada oriented their investment treaty programs around NAFTA language, the United States only made minor adjustments to its Model BIT post-1994, including revisions of its preamble and the addition of new prohibited performance requirements.¹²⁴ Particularly on ISDS, the 1994 US Model BIT did not seek to align BITs with the more elaborate architecture of NAFTA Chapter 11.

Chapter 11 thus remained an outlier in US investment treaty practice. Kenneth Vandevelde calls NAFTA an anomaly in the otherwise consistent US treaty practice: “Throughout the 1990s, BIT negotiations … proceeded largely as if NAFTA did not exist. NAFTA was treated as a unique agreement that was the product of unique circumstances.”¹²⁵ The NAFTA “anomaly” is well illustrated in Figure 6, which depicts the US treaty network as a similarity heatmap.

Figure 6. Similarity heat map of US BITs and FTA investment chapters.

Notes: This figure tracks the textual similarity of US BITs and investment chapters. Drafts are ordered chronologically, and the axes are symmetrical. High similarity is indicated by red tiles, while high dissimilarity is indicated by yellow tiles.

The status of NAFTA, however, changed in the late 1990s when the United States was hit by its first claims under NAFTA. While it had previously approached BITs from a capital-exporter perspective, it now came to consider these treaties from the perspective of a capital importer and, thus, a potential respondent in ISDS proceedings.¹²⁶ Through the bipartisan 2002 Trade Promotion Authority legislation, Congress enshrined this change in US policy by requiring a set of ISDS improvements inspired by NAFTA practice and by mandating that foreign investors should not enjoy greater rights in the United States than foreign investors.¹²⁷ Moreover, the United States began negotiations in the early 2000s to conclude new FTAs, which were supposed to also include an investment chapter.¹²⁸ Chapter 11 was thus a natural template as the only ISDS-enforceable investment chapter thus far concluded by the United States. These considerations then led to a revision of the US investment treaty practice, and a Model BIT around NAFTA and the lessons learned from NAFTA litigation was adopted in 2003.¹²⁹ This realignment with NAFTA is well illustrated in Figure 6, indicating that post-2003 FTAs and BITs bear more resemblance to Chapter 11 than prior US BITs. Importantly, however, by this time, the language of NAFTA Chapter 11 had already begun its spread outside North America.

B. The proliferation of the Chapter 11 model beyond North America

Even though all three NAFTA states ultimately realigned their investment practice with Chapter 11, the difference in timing is crucial. Whereas Canada and Mexico served as norm ambassadors of Chapter 11 from early on, the United States only streamlined its practice around NAFTA in 2003, a decade after NAFTA’s creation. The impetus for the diffusion of Chapter 11 thus came not from the United States but, rather, from its two junior partners. Through four successive waves depicted in Figure 7, the NAFTA model then spread to South America, East Asia, and, finally, globally.

Figure 7. Diffusion of NAFTA Chapter 11 design.

Notes: This figure shows the diffusion of NAFTA Chapter 11 design. A link is established between two countries if they sign a FTA with an investment chapter that overlaps textually with NAFTA Chapter 11 to at least 45 percent. Once a country has signed a NAFTA-inspired investment chapter, the country is coloured grey.

i. First wave 1994–98: exporting Chapter 11

After concluding NAFTA in 1994, Canada and Mexico embarked on FTA programs based on the NAFTA template. Canada signed a FTA with Chile in 1996. Its investment chapter was almost a carbon-copy of Chapter 11 with a striking similarity of 81 percent — the highest similarity between NAFTA Chapter 11 and any other text. Chile may have been positively predisposed to endorsing the NAFTA model because the country had almost joined the initial NAFTA negotiation.¹³⁰ Mexico was more active and had negotiated FTAs with Costa Rica and Colombia in 1994, Nicaragua in 1996, and Chile in 1998 around the same time. All four agreements mirror NAFTA’s treaty design, and their investment chapters follow Chapter 11 with a textual similarity between 48 percent (Colombia) and 63 percent (Chile).

ii. Second wave 1998–2002: NAFTA turns native in Central America

Mexico continued to sign new FTAs modelled on NAFTA with Honduras, Guatemala, and El Salvador in 2000, with 58 percent similarity. However, the defining feature of that second wave of agreements were the NAFTA-like treaties signed amongst Central American countries. Panama signed a FTA in 2002 with Costa Rica, El Salvador, Guatemala, Honduras, and Nicaragua, which included an investment chapter that resembled NAFTA Chapter 11 to 55 percent. The NAFTA investment chapter model had triggered a second order diffusion and became native to Central America.

iii. Third wave 2003–05: NAFTA Chapter 11 crosses the Pacific

The year 2003 marked a tipping point for the diffusion of NAFTA for two reasons. First, the United States began entering FTAs that included investment chapters closely modelled on NAFTA Chapter 11. Second, and more significantly in the long run, Central and South American countries began exporting NAFTA Chapter 11 across the Pacific. The NAFTA Chapter 11 model crossed the Pacific for the first time in February 2003 through the Chile-South Korea FTA, which contained an investment chapter mirroring NAFTA’s with a striking similarity of 71 percent (the second highest score in the dataset). A few months later, in August 2003, Panama signed a FTA with Taiwan, which included an investment chapter with 66 percent similarity to NAFTA Chapter 11 (the third highest score in the dataset). Over the next five years, the other Central American countries would follow concluding their own treaties with investment chapters modelled on NAFTA with Taiwan, starting with Guatemala in 2005 (63 percent similarity to Chapter 11), Nicaragua in 2006 (51 percent similarity), and El Salvador/Honduras in 2007 (60 percent similarity). Second generation adopters thus played a crucial role in the global diffusion of NAFTA.

This diffusion was further power charged through the United States, which began concluding FTAs on its own that mirrored NAFTA Chapter 11, albeit with the improvements that had been made through the experience with litigation. The United States-Singapore FTA, signed in May 2003, then displayed a 50 percent textual similarity with NAFTA Chapter 11. The FTA with Chile followed closely in June 2003, whose investment chapter was 53 percent similar to Chapter 11. Finally, Mexico signed a FTA with Japan in 2004 that contained an investment chapter that mirrored NAFTA Chapter 11 to 50 percent. By 2005, four Asian treaty partners had signed onto FTAs with investment chapters that bore a close resemblance to NAFTA Chapter 11.

iv. Fourth wave 2006–15: NAFTA goes global

The fourth wave of NAFTA’s diffusion then saw the treaty design spread globally and become especially entrenched around the Pacific. As time progressed, textual similarity with NAFTA Chapter 11 naturally decreased as new tweaks were added to investment protection language. These new branches in investment law’s genealogy tree, however, visibly trace their roots to NAFTA. ¹³¹ First, the treaty network connecting the Americas became denser. Canada signed treaties with Colombia, Peru, Honduras, and Panama that followed NAFTA’s text from 48 percent (Honduras) to 59 percent (Peru). The United States concluded FTAs with Colombia in 2006, Peru in 2006, and Panama in 2007, whose investment clauses resembled NAFTA at around 50 percent. South American states signed several FTAs with investment provisions modelled on NAFTA Chapter 11 among themselves, such as the 2006 Peru-Chile FTA (45 percent similarity) and the 2011 Peru-Mexico FTA (51 percent similarity).

Second, more FTAs were struck across the Pacific, and states in East Asia began signing PTAs modelled on NAFTA amongst themselves. Among the former, Canada signed a FTA with South Korea in 2014 (53 percent similarity to Chapter 11), Chile with Australia in 2008 (51 percent similarity to Chapter 11), and Peru with South Korea in 2011 (46 percent similarity). Intra-regionally, South Korea concluded a FTA with Australia in 2014 and New Zealand in 2015 (48 percent similarity to Chapter 11 respectively). This diffusion of the NAFTA template across the Pacific Rim then facilitated the negotiation of the original TPP, which overlapped with NAFTA to 59 percent when it was signed in 2015 and had an investment chapter with a similarity of 45 percent to NAFTA Chapter 11.

The NAFTA model also spread to other parts of the globe that have been omitted from Figure 7. Most notably, some of the agreements that China signed during that period bore strong resemblance to NAFTA, prompting Axel Berger to speak about a partial “NAFTA-ization” of Chinese investment treaties.¹³² The European Union (EU), following the transfer of competency over investment protection from its member states to the EU in the Treaty of Lisbon in 2009, also adopted language reminiscent of NAFTA. ¹³³ At least some member states had previously urged the Commission to avoid having Europe’s approach “contaminated” by NAFTA. ¹³⁴ During the EU’s negotiations with Canada over the Comprehensive Trade and Economic Partnership (CETA), however, Filippo Fontanelli and Giuseppe Bianco observed a “convergence towards NAFTA” in European investment policy.¹³⁵ The investment chapter of CETA between Canada and the EU that was concluded in 2016, while departing from NAFTA in relation to its ISDS infrastructure, was still significantly more similar to prior Canadian treaties than other prior European BITs.¹³⁶ In short, NAFTA Chapter 11 thus began serving as a model for treaty making across the globe.¹³⁷

6. Connecting diffusion to negotiation

NAFTA Chapter 11’s diffusion was in large part driven by direct pathways. In the 1990s, Canada and Mexico helped diffuse the treaty to South and Central America. These countries then helped carry the agreement across the Pacific to South Korea and Taiwan, which in turn diffused it within East Asia. It is important to underscore how unlikely this direct diffusion of NAFTA Chapter 11 would have seemed at the outset of the NAFTA negotiations. Recall that it was the United States that pushed the inclusion of an investment chapter. Canada had initially resisted US proposals for a strong investment chapter, favouring the more limited CUSFTA model instead, and Mexico was an investment-protection sceptic for historical and constitutional reasons. Mexico and Canada were thus unlikely to become enthusiastic proponents of Chapter 11. Yet, it was ultimately these two countries that propelled its rapid spread.

Similarly, it was hardly intuitive that Latin American countries would actively diffuse NAFTA Chapter 11. Like Mexico, many of them were investment law sceptics. Latin America was the intellectual homeland of the Calvo Doctrine, named after an Argentine jurist, and few of its states had concluded larger networks of investment treaties.¹³⁸ Moreover, why would a group of Central American developing states contemplating the basis for regional investment rules look to an agreement with compulsory ISDS that was first conceived to protect US investment in Mexico? And why would states like Panama or Chile then also use that text to conclude new IIAs with East Asian states, which, up to that point, had not concluded NAFTA-like investment agreements? This article suggests that the answer is found in the negotiation of the original NAFTA Chapter 11. This section contrasts recent unsuccessful unilateral diffusion of IIA design innovations to NAFTA Chapter 11’s successful diffusion. It highlights the importance of buy-in from potential diffusion agents and points to the normative pull of compromise-forged innovation. It thus suggests that the negotiation of Chapter 11 and its successful diffusion are linked. The compromise-laden nature of NAFTA Chapter 11 mobilized first- and second-order diffusers that spread its design widely.

A. Limited reach of unilateral innovation

Although investment practitioners and scholars like to emphasize the small yet meaningful differences that set investment treaty texts apart, a casual observer would be mostly struck by their commonalities. Alongside double taxation treaties, investment agreements have been called “common form treaties” to emphasize their relative homogeneity.¹³⁹ In the aggregate, standardized protection language (for example, “fair and equitable treatment”) and a path dependent evolution marked by refinements rather than replacements of core standards best describe the treaty regime.¹⁴⁰ At the same time, states have modernized their investment treaties over time, mostly incrementally, and, in a few instances, more dramatically. These innovations predominantly result from unilateral policy changes that are often (though not always) prompted by exposure to ISDS claims.¹⁴¹ Several developed and developing states hit by disputes have rethought their approach to IIAs, including by adopting revised model BITs as baseline for subsequent negotiations.¹⁴² Well-known examples are the 2003 updates of the Canadian and US Model BITs to reflect lessons learned from NAFTA litigations, which were subsequently used to conclude Canadian and US BITs.¹⁴³

More dramatic examples of unilateral innovations in recent practice include the evolving treaty practice of India and Brazil. India, which had been on the receiving end of several investment arbitrations, revised its model agreements starting in 2016 through several iterations.¹⁴⁴ This resulted in two subsequent BITs with Belarus in 2018 and Kyrgyzstan in 2019 that, among other innovations, curtail the scope of ISDS and make it subject to an exhaustion of domestic remedies.¹⁴⁵ Never having ratified an IIA with ISDS, Brazil launched a new investment treaty program in 2015 that placed investment cooperation and facilitation rather than investment protection at its centre.¹⁴⁶ It has since concluded thirteen BITs closely aligned with that model. In 2015, the European Commission put forth another major procedural innovation: ISDS through a permanent investment tribunal rather than ad hoc arbitration.¹⁴⁷ The proposal was prompted by widespread protests against CETA and parallel negotiations with the United States over a Transatlantic Trade and Investment Partnership, which led the EU Commission to propose the inclusion of a two-tier mechanism for a permanent first instance tribunal with an appeal stage in future EU IIAs to assuage concerns.¹⁴⁸ This innovation was subsequently introduced into CETA during the so-called “legal scrubbing” phase normally reserved for minor tweaks after formal negotiations with Canada had already ended.¹⁴⁹

What all three innovations have in common is that they are unilateral in nature, originating in domestic policy changes. The spread of such unilateral innovations in IIAs has typically been limited, however. Since these innovations respond to domestic policy concerns, partner countries with different domestic policy concerns often have little incentives to mirror them. This prevents the second-order diffusion that was so crucial for NAFTA’s spread as Central American countries used Chapter 11 as a template for their own agreements. Further reducing buy-in from partner countries is the asymmetry often present in investment treaty negotiations where dominant rule makers typically succeed in inserting their preferred language from model agreements into negotiated IIAs.¹⁵⁰ As a result, unilateral innovation does not tend to travel well beyond the treaty network of original innovators. Instead, like a hub-and-spoke system, unilateral innovation often remains confined to the investment treaty network of the innovator. Canada, in its 2021 Model BIT, did not include a permanent investment court structure modelled on CETA. ¹⁵¹ Indeed, none of the countries that have accepted the investment court in IIAs with the EU have used it in their own subsequent treaties as of this writing. Similarly, Brazil has successfully concluded more than a dozen BITs modelled on its innovative template, yet its partner countries have not concluded their own investment cooperation and facilitation agreements. Likewise, India’s model with its exhaustion of domestic remedies has not been copied by its treaty partners. Unilateral innovation remains confined to the national networks where it originated and does not tend to inspire widespread emulation by partner or third states.

B. How negotiated compromise mobilizes diffusers

Conversely, NAFTA Chapter 11’s language would not have diffused so successfully but for its active diffusers. Whereas the lack of uptake by treaty partners and third states tends to limit the spread of unilateral innovation, it was precisely this active first- and second-order diffusion by a set of unlikely diffusion agents that proved crucial for the proliferation of Chapter 11. NAFTA Chapter 11 came with many innovations, which at the time, were not less dramatic than the more recent reforms advocated by the EU, Brazil, or India. A staggering 77 percent of NAFTA Chapter 11’s article titles had never been used in an investment treaty, compared to only 44 percent in CETA’s investment chapter and 3 percent in the TPP investment chapter.¹⁵² The new level of detail and complexity of Chapter 11 was controversial at least among some states. As Lauge Poulsen documents, states that had been used to the shorter and simpler European treaties struggled when they first encountered NAFTA-inspired language.¹⁵³ Thus, what made, first, Mexico and Canada and, later, states in Latin America comfortable with the innovations introduced in Chapter 11?

The answer lies in the negotiation history of Chapter 11 itself, which mobilized subsequent diffusers. First, the consensual and compromise-oriented talks created a final text that was full of jointly crafted innovations that had the buy-in of Mexico and Canada. Their policy positions were transformed through the negotiation. From initial sceptics of the United States’ idea of an investment chapter, they became promoters of NAFTA Chapter 11. Conversely, the United States, the initial champion of Chapter 11, was its least enthusiastic supporter in the late 1990s. It saw NAFTA as a “unique agreement that was the product of unique circumstances” and refrained from introducing Chapter 11 language into its own subsequent BITs.¹⁵⁴ The perceptions of the negotiations thus seem to have been determinative of the initial diffusion behaviour of the NAFTA states. As a thought experiment, imagine a more acrimonious negotiation dominated by the United States that would have resulted in a draft closer to the initially proposed text modelled on US BITs. It seems plausible that such a negotiation would have produced greater initial enthusiasm on the part of the United States, perhaps leading to more early diffusion via US agreements. At the same time, it seems also less likely to have triggered a similarly active subsequent diffusion by Mexico and Canada. It would have been the latter countries that would have treated NAFTA as an anomaly confined to unique circumstances. In short, the compromise-laden negotiation that produced jointly crafted innovations is thus intimately linked to Mexico and Canada changing their perceptions on NAFTA Chapter 11 and becoming its active diffusers.

Second, the compromise-laden negotiation also motivated second-order diffusers that are typically reluctant to spread unilateral innovation. Innovation in Chapter 11 emerged from inter-state negotiations rather than from the domestic politics of a foreign state. A model BIT is a compromise of domestic political forces. In contrast, NAFTA Chapter 11 resulted from the bargaining between three very different states. NAFTA Chapter 11’s text had to accommodate the global hegemon’s desire to see its investment protected, it had to assuage the concerns of a middle power to shield its small market from its dominant southern neighbour, and it had to assure a developing country with no prior investment treaties that it would not be overwhelmed by investment claims. Through its detailed ISDS procedure, careful exclusions, and tailored protections, NAFTA Chapter 11 achieved this feat.

Innovations forged through inter-state compromise have a unique advantage over unilateral innovations when it comes to emulation. In the case of NAFTA Chapter 11, the fact that three very different countries painstakingly negotiated compromise language from diverging starting positions added a quality stamp to the text of NAFTA Chapter 11 that is absent in unilateral innovation. Negotiations are a form of vetting exercise — a survival test for contractual language. Only those provisions that are acceptable by themselves or as part of the overall package will make it into the final texts. Provided that talks do not suffer from stark power asymmetries, negotiating parties will reject provisions that are perceived to be too one-sided or too far-reaching and are thus unlikely to survive as initially proposed. Language is instead reworked to arrive at a compromise producing innovation in the process. The negotiation of NAFTA Chapter 11 illustrates these dynamics well.

NAFTA negotiations thus placed the quality stamp of inter-state compromise on its text. The jointly developed innovations thereby facilitated emulation and second-order diffusion. If the text was acceptable to Mexico, it should work for Panama too. In contrast to the limited spread of unilateral innovations, the innovations NAFTA Chapter 11 have been copied far and wide because they could generate buy-in and take-up from third states. NAFTA Chapter 11 negotiations therefore were likely crucial in shaping the perceptions of its content by third states prompting emulation. Otherwise, it is hard to explain why Latin American countries, long sceptical of investment treaties, endorsed its text in agreements amongst themselves. From extensive performance requirements that limited a country’s ability to generate targeted spillovers from investment activities to a circumvention of the local court system by providing for ISDS, the norms agreed to in NAFTA differed from Mexico’s original bargaining position, departed from long-held views shaped by the Calvo Doctrine and were not necessarily in the best interest of developing states, especially when concluding South-South IIAs. And, yet, Latin American countries readily emulated NAFTA Chapter 11 and were instrumental in its spread. The vetting and compromise-forged innovation that took place during the negotiations thus seemed to have been crucial in shaping the perceptions of NAFTA Chapter 11 as an attractive template to emulate.

7. Conclusion: the virtues of compromise

NAFTA Chapter 11 shaped modern investment treaty practice like no other agreement. Today, its language can be found in IIAs across the globe. Several factors contributed to the lasting impact of Chapter 11. The United States’ role as global hegemon may implicitly have made the adoption of its language more enticing to states, especially in Latin America and East Asia, close to the United States’ primary zone of influence. Yet, as this article has shown, the United States did not play an active part in the initial diffusion of NAFTA Chapter 11. NAFTA remained an outlier in US investment policy prior to its policy shift in 2003 by which time NAFTA Chapter 11 language had already successfully diffused.

Also making NAFTA Chapter 11’s language attractive to other states were its many firsts. NAFTA was the first FTA with an investment chapter enforceable by ISDS — a pioneer treaty from which other states interested in concluding comprehensive FTAs could copy. Furthermore, NAFTA had generated considerable litigation. Its language was tried and tested in ISDS awards, and NAFTA parties had introduced targeted refinement, drawing from their experience as respondents. NAFTA Chapter 11 was thus a ready-made alternative to states worried about their exposure to investment claims in vague European BITs.

While NAFTA’s soft power and the policy competition and learning that it enshrined likely played a role in its diffusion, they cannot explain the copy and pasting of NAFTA Chapter 11 among the developing countries in Latin America that became its early adopters and active diffusers. This article has shown that the foundations for the successful spread of Chapter 11’s language were laid already during negotiations. NAFTA was a consensus product. It emerged from trilateral negotiations involving three countries with diverse interests. The consensus-driven negotiation generated buy-in from Mexico and Canada to endorse Chapter 11 as model for their own treaties. It was these two countries, initially lukewarm about including an investment chapter in NAFTA, that became its main diffusers.

NAFTA Chapter 11 also produced innovations jointly created by all NAFTA parties. The endorsement particularly by Mexico gave Chapter 11 credibility among South American countries. Its text was not the product of domestic US politics but, rather, of inter-state negotiations. This novel treaty design, forged through inter-state compromise, reflected what Mexico considered to be in its best interest also in relation to Chile or Central America. This, in turn, convinced third states to emulate and promote its design as well. Whereas other recent unilateral innovations have not travelled far in the IIA universe, NAFTA Chapter 11’s innovation, forged and vetted through inter-state compromise, diffused across the globe and shaped the evolution of investment treaties.

In sum, NAFTA Chapter 11’s impact is linked to its consensus-driven negotiations. Because NAFTA Chapter 11 emerged as a compromise rather than another US BIT-style treaty, it generated buy-in on the part of Canada and Mexico to anchor their own investment policy in NAFTA language. And because innovation emerged from inter-state compromise, NAFTA language travelled further than innovation that was rooted in unilateral policy changes. Hence, an important legacy of NAFTA Chapter 11 lies in demonstrating the power of negotiated compromise. Unilateralism and raw expressions of national interests have been on the rise. Looming geo-economic competition has added further fuel to the fire.¹⁵⁵ Investment treaty negotiations have been no stranger to the ensuing confrontational dynamics. US President Barack Obama explained the need for the TPP with the premise that “if America doesn’t write those rules — then countries like China will.”¹⁵⁶ His successor, President Donald Trump, labelled NAFTA as “the worst deal ever” and forced a renegotiation.¹⁵⁷ Commerce Secretary Wilbur Ross described the US position in these talks in stark terms: “[W]e’re asking two countries [Canada and Mexico] to give up some privileges they have enjoyed for 22 years and we’re not in a position to offer anything in return.”¹⁵⁸ Economic negotiations have become framed in terms of winners and losers.

The notion of compromise has always been ambivalent. It invokes agreements to disagree, kicking the can down the road, “strategic ambiguity” undermining subsequent compliance, and lowest-common-denominator deals that lack ambition and fail to move the needle. These vices are associated with compromise and make it seem preferable to vehemently eye the national interest and push for outcomes shaped more by unilateral preference than by negotiated compromise. The legacy of NAFTA Chapter 11, however, recalls the virtues of compromise. In the context of NAFTA, compromise generated both a buy-in and a normative gravitational pull that unilateralism could not rival. While NAFTA spread, its successor, CUSMA, which US Trade Representative Robert Lighthizer and other US policy-makers hoped would become “a model agreement,”¹⁵⁹ thus far has not. Threatened with US “poison pill” negotiation proposals that have been too toxic for Canadian and Mexican negotiators to accept and the constant threat by US President Trump to pull the United States out of NAFTA if no satisfactory deal emerged, Canada and Mexico were forced to focus on damage control.¹⁶⁰ CUSMA’s changes to ISDS have not been taken up in subsequent agreements. At the point of this writing, the agreement seems unlikely to inspire emulation either by Canada or Mexico or by third states.

Conversely, NAFTA Chapter 11 has diffused widely and has produced tangible benefits for the United States down the road. As shown, agreements modelled on NAFTA Chapter 11 mushroomed on all sides of and across the Pacific in the early 2000s. This made it easier for the United States to conclude the TPP in the 2010s. Its signatories had, to some degree at least, already converged on NAFTA-inspired language.¹⁶¹ Hence, NAFTA Chapter 11’s legacy is more than the litigation it spurred, the case law it generated, and the imprint on subsequent treaties that it produced. Perhaps the most important lesson learned from NAFTA Chapter 11’s creation and diffusion is to showcase the virtue of negotiated compromise. Compromise can mobilize partners and can generate normative pull that shapes international rules for decades. The most effective way to write the rules of the twenty-first century may then lie in sharing the pen with others.