1.
Introduction ^
Close to 3000 international investment treaties (IIAs) protect foreign investors against the risks of expropriation, discrimination and unfair treatment.1 Today, virtually every country is signatory to an IIA. What makes IIAs matter in practice is their strong enforcement mechanism. Private investors can bring claims directly to international arbitration in order to enforce an IIA’s investment protection obligation against a host state. Over 600 of such investment treaty claims have been launched until today.2 If successful, investors can win multi-million dollars worth of damages. As a result, IIAs have come to play a central role in international economic governance.
At the same time, due to its atomized and fragmented nature, states and investors alike struggle to effectively navigate the complex IIA universe: negotiators are sorting through hundreds of agreements to find common denominators in two countries’ treaty practice; policy-makers strive to streamline a country’s investment obligations scattered in scores of treaties and litigators are comparing hundreds of agreements to find useful distinctions or analogies to advance their case. Commercial platforms and empirical scholarship only provide limited assistance to researchers and practitioners. Legal information providers (e.g. www.investorstatelawguide.com) focus on awards rather than treaties and while political scientists and legal scholars have empirically coded treaty content, their datasets are either limited in scope to few treaty features or have not been made public.3 The mappinginvestmenttreaties.com project seeks to fill the ensuing gap by providing legal analytics to assist practitioners in navigating investment law’s complexity. Using state-of-the-art text as data methods, the project reveals hitherto unknown patterns of similarities and differences in over 1600 international investment agreements and provides new web-based tools for academics and practitioners to engage interactively with the IIA universe.4
2.
Data ^
3.
Methodology ^
We use our dataset of raw treaty texts to investigate similarities and differences across agreements and articles. We proceeded in three stages. First, we developed a similarity measure to calculate distances between agreements and articles. Second, we constructed heat maps to identify differences between treaties. Third, we employed diffs to color-code word-level variations among comparable articles thereby facilitating the manual detection of textual differences.
3.1.
q-character gram representations of treaty texts and their Jaccard distance ^
To identify differences and similarities between agreements and articles, we follow an approach much akin to that employed in plagiarism detection software. First, we break down each treaty into its 5-character-long substrings and count the number of times each substring occurs in the document.5 To illustrate, the imaginary document «shall not be permitted» will contain the following 5-character substrings: «shall», «hall_», «all_n», «ll_no», «l_not», «_not_», «not_b», «ot_be», «t_be_», «_be_p», «be_pe», «e_per», «_perm», «permi», «ermit», «rmitt», «mitte», «itted» («_» signifies space). Second, we compute the Jaccard distance between two treaties based on the substrings that overlap between the pair. To continue with the above example, the document «shall not be permitted» and a second document «shall be permitted» will have similar substrings, except for «all_n», «ll_no», «l_not», «_not_», «not_b», «ot_be», «t_be_». This divergence is caused by the presence of «not» in the first document and can be quantified by counting the number of unique 5-character substrings appearing in both documents and dividing it by the total number of unique 5-character substrings in the two documents (and subtracting this figure from 1). Applying this method to our set of two documents would yield a Jaccard distance of 0.48 – a measure of dissimilarity between two documents with 1 involving two very different documents and 0 involving identical documents. Transposed to international investment treaties, this Jaccard distance allows us to determine what treaties are similar to each other and what treaties are farther apart revealing new clusters and patterns in our treaty data.
3.2.
Heat map representation of Jaccard scores ^
3.3.
Diffs and article-level comparisons of treaty texts ^
4.
Applications ^
4.1.
Systemic and country-level comparisons ^
Our similarity representations allow users to compare treaties both at the global and at the country level. The global level is suitable for identifying systemic trends. In a separate study, for instance, we use the Jaccard distance measure to trace consistency and innovation in the BIT universe demonstrating, amongst others, that developed countries tend to possess internally-coherent treaty networks suggesting that they are the system’s rule-makers whereas developing countries displaying low coherence scores are the rule-takers.6 In the same paper, we also highlighted the usefulness of tracing developments at the country level. Without requiring external input, Jaccard distances can reveal when countries shift from one treaty template to another allowing researchers to identify changes of a country’s evolving investment policy.
4.2.
Treaty- and article-level comparisons ^
On the treaty-level, Jaccard distances can trace policy diffusion processes and allow for the placing of a treaty in the wider universe of agreements. An analysis of the Trans-Pacific Partnership’s (TPP) Investment Chapter concluded in 2015, for instance, revealed that 82% of the Chapter’s main text has been copied and pasted from the United States-Colombia Free Trade Agreement (FTA) signed in 2006.7 In the same paper, we also highlighted that article-level Jaccard distances can be used to identify the articles responsible for differences between two agreements. In the case of the TPP, for instance, the National Treatment clause is virtually identical (almost 95% of similarity) to the corresponding clauses in the United States-Colombia FTA, while the Minimum Standard of Treatment clauses only share a 70% similarity suggesting that the latter but not the former is a driver of dissimilarity.
4.3.
Applications in negotiations and litigation ^
5.
Web-based Tool and Future Developments ^
In order to allow users to engage directly with our Jaccard distance representation of BITs, we have developed an interactive web-based tool that can be accessed via www.mappinginvestmenttreaties.com. The tool allows users to quickly identify differences and similarities between treaties and to interpret them substantively using diffs. Steps to expand the analysis are in the pipeline. In particular, we aim at moving from textual to semantic similarity to more accurately depict legal differences.
6.
References ^
Alschner,W. and D. Skougarevskiy, Consistency and Legal Innovation in the BIT Universe, Stanford Public Law Working Paper, No. 2595288 (2015), available at: http://papers.ssrn.com/abstract=2595288.
Alschner,W. and D. Skougarevskiy, The New Gold Standard? Empirically Situating the TPP in the Investment Treaty Universe, CTEI Working Paper, No. 2015-08 (2015).
Chaisse, J., and C. Bellak, Navigating the Expanding Universe of International Treaties on Foreign Investment: Creation and Use of a Critical Index, Journal of International Economic Law, 18 (2015), 79–115.
Manger, M.S., A Quantitative Perspective on Trends in IIA Rules, in: A. de Mestral & C. Lévesque (eds.), Improving International Investment Agreements, London, Routledge, 2011.
Spirling, A. U.S. Treaty Making with American Indians: Institutional Change and Relative Power, 1784–1911, American Journal of Political Science 56 (2012), 84–97.
UNCTAD, World Investment Report 2015: Reforming International Investment Governance, United Nations, Geneva, (2015).
- 1 UNCTAD, World Investment Report 2015: Reforming International Investment Governance, United Nations, Geneva (2015).
- 2 Ibid.
- 3 M.S. Manger, A Quantitative Perspective on Trends in IIA Rules, in: A. de Mestral & C. Lévesque (eds.), Improving International Investment Agreements, London, Routledge (2011). J. Chaisse and C. Bellak, Navigating the Expanding Universe of International Treaties on Foreign Investment: Creation and Use of a Critical Index, Journal of International Economic Law, 18 (2015), 79–115.
- 4 An earlier version of this paper was presented at the 2015 JURIX Conference.
- 5 A. Spirling, U.S. Treaty Making with American Indians: Institutional Change and Relative Power, 1784–1911, American Journal of Political Science 56 (2012), 84–97.
- 6 W. Alschner and D. Skougarevskiy, Consistency and Legal Innovation in the BIT Universe, Stanford Public Law Working Paper, No. 2595288 (2015), available at: http://papers.ssrn.com/abstract=2595288.
- 7 W. Alschner and D. Skougarevskiy, The New Gold Standard? Empirically Situating the TPP in the Investment Treaty Universe, CTEI Working Paper, No. 2015-08 (2015).