Impact statement
Governments have argued for more than three years over how far a treaty to end plastic pollution should reach, mostly about chemistry, trade and money. This article draws attention to a body of law that has barely entered the conversation but could determine the treaty’s worth in practice: the roughly 3,300 investment agreements that let foreign companies sue states before private arbitral tribunals when a new rule cuts the value of their assets. Petrochemical and plastics producers sit squarely inside their protection. Mapping the measures now on the table against four decades of arbitral decisions, we show that the tools most likely to cut plastic at source – production caps, polymer and additive bans, producer-responsibility levies and waste-trade limits – are also the ones most open to compensation claims. We set the two trend lines side by side: production on course for about 1,235 million tonnes a year by 2060, and an arbitration caseload growing by close to 80 known cases a year. This is not a prediction that states will lose such cases. Investment law contains real defences for evenhanded, evidence-based regulation, and states have won important cases. The point is that fear of a claim can work before any writ is filed, nudging officials to soften or drop a rule, the regulatory chill familiar from climate, tobacco and pesticide policy. A treaty with teeth must treat investment law as seriously as trade law, building in the carve-outs and safeguards that keep the right to protect public health from being priced out of reach.
Introduction
‘The polluter should, in principle, bear the cost of pollution’. So the assembled governments declared at Rio in 1992, writing the polluter-pays principle into Principle 16 of the Rio Declaration (United Nations, 1992). Three decades later, a second body of law threatens to run that maxim backwards: under the investment treaties that most states have signed, a company whose polluting business is curtailed by new regulation can demand to be paid for the interruption, and, not rarely, it is. What if one of the hardest obstacles to a treaty that ends plastic pollution is not written in any environmental text at all, but sits, largely unexamined, in some three thousand investment agreements signed to protect capital?
The scale is familiar. Humanity has produced more than eight thousand million tonnes of virgin plastic, most now waste and only a fraction ever recycled (Geyer et al., Reference Geyer, Jambeck and Law2017); annual production climbed from about 2 million tonnes in 1950 to 413.8 million tonnes in 2023 (Ritchie et al., Reference Ritchie, Samborska and Roser2023; Plastics Europe, 2024); and the OECD projects plastic waste rising from 353 million tonnes in 2019 to more than a billion by 2060, with leakage doubling on the way (Jambeck et al., Reference Jambeck, Geyer, Wilcox, Siegler, Perryman, Andrady, Narayan and Law2015; Borrelle et al., Reference Borrelle, Ringma, Law, Monnahan, Lebreton, McGivern, Murphy, Jambeck, Leonard, Hilleary, Eriksen, Possingham, de Frond, Gerber, Polidoro, Tahir, Bernard, Mallos, Barnes and Rochman2020; Lau et al., Reference Lau, Shiran, Bailey, Cook, Stuchtey, Koskella, Velis, Godfrey, Boucher, Murphy, Thompson, Jankowska, Castillo, Pilditch, Dixon, Koerselman, Kosior, Favoino, Gutberlet, Baulch, Atreya, Fischer, He, Petit, Sumaila, Neil, Bernhofen, Lawrence and Palardy2020; OECD, 2022a, 2022b). The instruments that might have contained this – conventions on the sea, on chemicals and on waste – were built for other purposes and have not held (Dauvergne, Reference Dauvergne2018; UNEP, 2021; Mendenhall, Reference Mendenhall2023), and reviews agree the response remains fragmented (Vince and Hardesty, Reference Vince and Hardesty2017; Cowan and Tiller, Reference Cowan and Tiller2021; Tiller et al., Reference Tiller, Booth and Cowan2022).
Recognising the gap, in March 2022 the United Nations Environment Assembly agreed to negotiate a legally binding instrument covering the full life cycle of plastic (UNEA, 2022), with reduction, reuse and the elimination of ‘chemicals of concern’ all in scope; scientists set out what a strong agreement would need, chiefly a firm cap on how much plastic the world makes and binding rules on what goes into it (Simon et al., Reference Simon, Raubenheimer, Urho, Unger, Azoulay, Farrelly, Sousa, van Asselt, Carlini, Sekomo, Schulte, Busch, Wienrich and Weiand2021; Fletcher et al., Reference Fletcher, Couceiro, Farrelly, Mauad, Muthumbi and Tiller2023). Yet by August 2025, at the resumed fifth session in Geneva, the talks broke up without agreement, foundering on exactly those upstream questions (UNEP, 2025), and the next full negotiation is not expected before late 2026 (IISD, 2026; UNEP, 2026). Studies of the positions show why the fault lines run deep: states cluster by whether they favour upstream, midstream or downstream measures, and those preferences track fossil-fuel dependence and domestic polymer production (Dreyer et al., Reference Dreyer, Hansen, Holmberg, Kielhöfer, Olsen and Stripple2025).
Almost all the attention to why an ambitious treaty is so hard to reach has gone to the visible politics – lobbying, coalitions and the consensus rule that lets a determined minority hold the pen. Far less has been said about a legal architecture that operates in the background and binds long after a treaty is signed. By the end of 2023, states had signed 3,291 bilateral investment treaties and investment chapters (UNCTAD, 2024a). They share a promise: to treat foreign investors fairly, not to expropriate without compensation and, importantly, to let the investor enforce that promise directly before an international arbitral tribunal rather than in the host state’s own courts (Van Harten, Reference Van Harten2008; Schill, Reference Schill2010; Bonnitcha et al., Reference Bonnitcha, Poulsen and Waibel2017; Sornarajah, Reference Sornarajah2017), an obligation many assumed in the 1990s with little understanding, merely to secure capital (Poulsen, Reference Poulsen2015). These treaties protect exactly the foreign investors, petrochemical firms, resin manufacturers, and packaging multinationals whose business a plastics treaty would curtail, giving rise to two legal orders that overlap yet rarely coexist without tension (Viñuales, Reference Viñuales2012).
The concern is not merely that a company might win, but that the credible possibility of an expensive claim can change what governments are willing to do. Kyla Tienhaara named this ‘regulatory chill’. She distinguished an internalisation chill, where officials build arbitration risk into their own advice, from a threat chill, where a company raises the prospect of a claim to move a decision, and a cross-border chill, where one state’s experience deters its neighbours (Tienhaara, Reference Tienhaara2009, Reference Tienhaara, Brown and Miles2011, Reference Tienhaara2018). In each, a measure is softened, delayed or shelved before any tribunal convenes. The idea is contested, since a chilled decision leaves few fingerprints (Côté, Reference Côté2014), but it is not merely theoretical: Canadian officials have admitted weighing arbitration risk when designing environmental rules (Van Harten and Scott, Reference Van Harten and Scott2016), and Moehlecke’s study of anti-smoking laws in 92 countries found the 2 policies actually challenged under investment law were adopted markedly more slowly than 7 that were not (Moehlecke, Reference Moehlecke2019).
This article asks whether plastics policy is exposed to a similar predicament, and what would reduce that exposure. Its contribution sits at the junction of three literatures that have run in parallel. Regulatory-chill scholarship is rich on climate, tobacco and pesticides but has not reached plastics (Côté, Reference Côté2014; Tienhaara, Reference Tienhaara2018; Moehlecke, Reference Moehlecke2019). Work on environment and investment arbitration is doctrinal and case-driven, and rarely quantifies the trajectory of the system generating the risk (Viñuales, Reference Viñuales2012; Titi, Reference Titi2014). The plastics-governance literature analyses treaty design, state preferences and regime interaction without treating investment agreements as part of the regime complex at all (Raubenheimer and Urho, Reference Raubenheimer and Urho2020; Mendenhall, Reference Mendenhall2023; Dreyer et al., Reference Dreyer, Hansen, Holmberg, Kielhöfer, Olsen and Stripple2025). We join them in four moves. First, we build a typology of the measure families a treaty is likely to require and map each against the state conduct tribunals have already assessed, producing an exposure assessment rather than a case census. Second, we set the public numbers on both sides on a common horizon by methods a reader can reproduce. Third, we extend the chill argument backwards from implementation to negotiation. Fourth, we weigh reform pathways against the timing those numbers imply. The claim is not that investment law makes ending plastic pollution impossible; it is that a treaty drafted as though investment law were absent may find its sharpest instruments blunted, and that the remedy is available if negotiators reach for it in time.
Methods
This study combines a doctrinal legal mapping with a descriptive quantitative analysis. Neither part generates new data; the contribution lies in reading two well-documented but rarely connected bodies of evidence against each other. Terminology is kept distinct throughout, because the indicators are not interchangeable. Primary plastics production denotes annual output of new polymer material, the series plotted in Figure 1; plastics use is the OECD’s broader indicator, which adds secondary plastics and is therefore larger in any given year; plastic waste is an end-of-life flow, and the plastic scrap trade a subset of it. One caveat follows from the splice described below: the Plastics Europe total is not virgin-only, since roughly a tenth of the 413.8 million tonnes reported for 2023 was recycled or bio-based (Plastics Europe, 2024).
Global plastics production, 1950–2023, in million tonnes per year, with a log-linear trend fitted to the 2000–2023 observations and extrapolated to 2060 on the assumption of unchanged policy. The shaded band is the 95% confidence interval for the fitted trend; it is not a prediction interval for realised output in any future year, which would be wider. Circular markers show the OECD baseline for global plastics use (460 million tonnes in 2019; 1,231 million tonnes in 2060). Plastics use is a broader indicator than production because it includes secondary plastics, so the two series are compared as a benchmark and are not equivalent. The observed series splices Geyer et al. (Reference Geyer, Jambeck and Law2017), as compiled by Ritchie et al. (Reference Ritchie, Samborska and Roser2023), with Plastics Europe (2024), whose global total includes a recycled and bio-based share of roughly a tenth in 2023. Sources: Authors’ calculations from Geyer et al. (Reference Geyer, Jambeck and Law2017), Plastics Europe (2024) and OECD (2022b). Underlying values are given in Supplementary Data S1.
Figure 1. Long description
A line chart. The horizontal axis is the year, 1950 to 2060, labelled every twenty years; the vertical axis is global plastics production in million tonnes per year, 0 to 1,400. A solid dark blue line shows observed production: about 2 million tonnes in 1950, 35 million by 1970, 213 million by 2000 and 413.8 million tonnes in 2023, the final observed point, labelled. The rise steepens throughout, with brief dips in the mid-1970s, 2008 and 2020. From 2023 a dashed red line continues the log-linear trend fitted to the 2000 to 2023 data, reaching about 530 million tonnes in 2030, 703 in 2040, 932 in 2050 and 1,235 in 2060. A pale pink band around it is the 95 per cent confidence interval for the trend, widening to roughly 1,114 to 1,370 million tonnes by 2060. Two gold circular markers show the OECD baseline for plastics use: 460 million tonnes in 2019, just above the observed line, and 1,231 million tonnes in 2060, labelled, sitting almost exactly on the extrapolated trend.
The legal mapping drew on three bodies of primary material. The first was the regulatory measures in play, the Chair’s draft treaty texts, the pre-session submissions analysed in the negotiation literature (Raubenheimer and Urho, Reference Raubenheimer and Urho2020; Dreyer et al., Reference Dreyer, Hansen, Holmberg, Kielhöfer, Olsen and Stripple2025) and national laws, such as single-use bans, deposit and producer-responsibility schemes and scrap-import restrictions, from which we built a typology of six measure families a treaty is likely to require. The second was arbitral jurisprudence. Rather than attempt a census of the more than 1,400 known claims (UNCTAD, 2025), we searched the UNCTAD Investment Dispute Settlement Navigator and published awards for decisions meeting three criteria: the challenged measure was a health or environmental regulation rather than a contract or tax dispute; its legal form corresponds to one of the six families, a product or substance ban, a restriction on waste movement, a permit refusal or withdrawal, a forced closure or a charge on producers; and the tribunal reasoned on the regulatory measure itself. Cases meeting all three were retained however they were decided, so the set contains leading state victories as well as investor successes. The analogues stand in for plastics measures because the fact patterns are structurally alike, not because the products are; the set is illustrative of the doctrinal terrain rather than a representative census. The third body was the treaty texts: representative agreements and their clauses on fair and equitable treatment, indirect expropriation and the ‘umbrella’ of investor protections, with the newer treaties and reform proposals that write the right to regulate back in Gaukrodger and Gordon (Reference Gaukrodger and Gordon2012) and Titi (Reference Titi2014). We then asked of each measure family: has conduct of this type already been challenged, and with what result? The output is a qualitative exposure assessment (Table 1), not a prediction of who would win any particular case. Searches of the investment-law and plastics-governance literatures returned work on tobacco, climate and mining and work on plastics treaty design, but nothing joining them; we state that as the limit of our search rather than as a settled fact about the field.
Exposure of plastics-control measures to investor–state claims, with analogous decided cases, 1987–2024
Table 1. Long description
Table 1 appears in the article “How International Investment Law Could Freeze Global Action on Plastic Pollution” by Haider, Khan and Szeberenyi. Its caption reads: “Exposure of plastics-control measures to investor-state claims, with analogous decided cases, 1987–2024.” The table sets out six families of plastics-control measures and, for each, the investment-treaty standards most likely to be invoked against it, the closest arbitral analogues, and the authors’ indicative rating of its exposure to investor-state claims.
The ratings are the authors’ qualitative assessment from the mapping described in the Methods. They record legal contestability, the closeness of the arbitral analogues and the breadth of arguable claims a measure would create. They are not predictions of the probability that a claim of that type would succeed. Cases are cited by short name, and full citations appear in the reference list.
Structure. The table has four columns and seven data rows below a header row. The columns, from left to right, are: Measure family; Standards most likely invoked; Closest arbitral analogues; and Indicative exposure (legal contestability). Each row describes one measure family, and the cells in that row are read from left to right.
Row 1. Caps or moratoria on primary polymer production. Standards most likely invoked: indirect expropriation, and fair and equitable treatment (legitimate expectations). Closest arbitral analogues: Rockhopper v. Italy and Lone Pine v. Canada. Indicative exposure: High.
Row 2. Bans on specific polymers and chemicals of concern. Standards: indirect expropriation, fair and equitable treatment, and national treatment. Analogues: Methanex v. United States, Chemtura v. Canada and Ethyl v. Canada. Indicative exposure: Moderate, and lower where measures are non-discriminatory and science-based.
Row 3. Extended producer responsibility, levies and plastics taxes. Standards: fair and equitable treatment (stability), and arbitrary or discriminatory measures. Analogues: Philip Morris v. Uruguay and Philip Morris Asia v. Australia. Indicative exposure: Moderate.
Row 4. Restrictions on the plastic waste trade. Standards: national treatment and fair and equitable treatment. Analogue: S.D. Myers v. Canada. Indicative exposure: Moderate to high where a protectionist effect is arguable.
Row 5. Refusal, non-renewal or withdrawal of permits, and forced closures. Standards: indirect expropriation, and fair and equitable treatment. Analogues: Metalclad v. Mexico, Tecmed v. Mexico, Vattenfall v. Germany (II), and RWE and Uniper v. Netherlands. Indicative exposure: High.
Row 6. Downstream measures: collection targets, design rules and reporting. Standards: rarely litigated. Analogues: no close analogue identified. Indicative exposure: Low.
Summary of the pattern. Two measure families carry a rating of High, namely caps or moratoria on primary polymer production and permit refusals, non-renewals, withdrawals and forced closures. Both are linked to indirect expropriation and fair and equitable treatment claims and to several decided cases. Bans on specific polymers and chemicals, producer responsibility schemes, levies and taxes are rated Moderate, and restrictions on the plastic waste trade are rated Moderate to high where a protectionist effect is arguable. Downstream measures are rated Low, with no close arbitral analogue identified.
Note: The exposure ratings are the authors’ qualitative assessment from the mapping described in the Methods. They record legal contestability, the closeness of the arbitral analogues and the breadth of arguable claims a measure would create, and are not predictions of the probability that a claim of that type would succeed. Cases are cited by short name; full citations appear in the reference list.
The quantitative layer is drawn entirely from public sources, so every number can be checked against its origin: the 1950–2015 production series of Geyer et al. (Reference Geyer, Jambeck and Law2017) via Our World in Data (Ritchie et al., Reference Ritchie, Samborska and Roser2023), spliced with Plastics Europe’s estimates for 2018–2023 (Plastics Europe, 2024); the OECD’s Global Plastics Outlook to 2060 (OECD, 2022a, 2022b); UNCTAD’s year-end totals of known treaty-based cases (1,401 by end-2024) and its statistics on outcomes, standards and amounts (UNCTAD, 2024b, 2025); and the OECD’s waste-trade monitoring series based on UN Comtrade records (Brooks et al., Reference Brooks, Wang and Jambeck2018; OECD, 2025).
We use deliberately simple forecasting methods. Production is extrapolated from the 2000 to 2023 observations by a log-linear fit, equivalent to compound growth of 2.9% a year; the band in Figure 1 is the 95% confidence interval for the fitted trend, and not a prediction interval for realised output in any future year, which would be wider. The cumulative caseload is extrapolated by a linear trend over 2016–2024, equivalent to 79 new known cases a year; the band in Figure 2 is a scenario range spanning 55–80 new cases a year, chosen to cover filing rates as first reported and after later revision, and is not a statistical interval of either kind. We prefer these methods because any reader can reproduce them in a spreadsheet, and because the production extrapolation can be set beside the OECD’s structural projection, which shares a 2060 horizon. That comparison is a benchmark rather than a validation: the OECD models plastics use, a broader indicator, so proximity shows only that a naive trend and a general-equilibrium model imply a similar order of magnitude.
Cumulative number of known treaty-based investor–state arbitration cases, 1987–2024, with a linear extrapolation to 2035. Points are year-end totals reported in successive UNCTAD ISDS Navigator updates (2016–2024); the 2009 value is derived from UNCTAD’s statement that 27% of the 1,401 cases known by end-2024 were initiated in 1987–2009. The dashed line is a linear trend fitted to the 2016–2024 counts, equivalent to 79 new known cases a year. The shaded band is a scenario range assuming 55–80 new known cases a year; it is not a statistical confidence or prediction interval. Counts are of known cases only, and cumulative totals incorporate retroactive additions as confidential proceedings become public. Source: Authors’ calculations from UNCTAD (2025). Underlying values are given in Supplementary Data S1.
Figure 2. Long description
A line chart. The horizontal axis is the year, about 2009 to 2035, labelled every five years; the vertical axis is the cumulative number of known investor-state dispute settlement cases, 0 to 2,500. A solid dark blue line with circular markers shows year-end totals from UNCTAD ISDS Navigator updates: 378 cases in 2009 and 767 in 2016, both labelled, then unlabelled markers at roughly 940 in 2018, 1,105 in 2020, 1,190 in 2021, 1,255 in 2022 and 1,330 in 2023, ending at 1,401 cases in 2024, labelled. The slope is near constant from 2016 onward, with no sign of flattening. From 2024 a dashed red line continues a linear trend fitted to the 2016 to 2024 counts, equivalent to 79 new known cases a year, reaching about 2,280 by 2035. A pale pink wedge beneath it widens to the right, spanning a scenario range of 55 to 80 new cases a year and annotated 2,000 to 2,280 by 2035. The fitted trend runs along the upper edge of that wedge. The band is a scenario range, not a statistical interval.
This study has three limitations. Arbitral awards create no formal precedent; they merely inform later proceedings. Regulatory chill is not directly observed but must be inferred from admissions, timing and documented internal caution (Moehlecke, Reference Moehlecke2019). And extrapolation assumes that tomorrow resembles today: recent-year counts are revised upward by about 15% as confidential arbitrations surface, though a wave of treaty terminations could bend the path down (UNCTAD, 2025). A strong plastics treaty would itself be the kind of structural break our production trend cannot anticipate, which is, in a sense, this article’s central contention.
Results and discussion
Two rising curves
The production curve is the closest thing environmental policy has to a straight line on a logarithmic chart. Output was about 2 million tonnes in 1950, 35 million by 1970, 213 million by 2000 and 413.8 million tonnes in 2023 – a 200-fold increase within a single human lifetime (Geyer et al., Reference Geyer, Jambeck and Law2017; Plastics Europe, 2024). Growth has slowed, but only to a rate that still doubles output every 25 years or so. Composition matters as much as volume: packaging, consumer products and textiles generate about two-thirds of all plastic waste (OECD, 2022a), the very segments single-use bans, deposit schemes and producer-responsibility levies target and where recycling, at 6% of plastics use in 2019 and only 12% by 2060 on the OECD’s baseline, never catches the primary curve without policy help (OECD, 2022b). Growth is also overwhelmingly a developing-world story, so the countries hosting most of the new plastic are those least equipped to pay an arbitral award (OECD, 2022b).
Figure 1 shows the series with our fitted trend. Continuing the 2000–2023 trajectory yields roughly 530 million tonnes of production in 2030, 703 million tonnes in 2040, 932 million tonnes in 2050 and 1,235 million tonnes in 2060, with a 95% confidence interval for the trend of about 1,114–1,370 in the final year. This is not a prediction, but a statement of where the observed trajectory points if nothing bends it. The OECD’s baseline for plastics use in 2060, built by a multi-sector general-equilibrium model rather than a ruler, is 1,231 million tonnes (OECD, 2022b). The two figures are close, but the indicators differ in coverage, since plastics use includes secondary plastics, and the agreement is therefore a benchmark rather than independent confirmation of our extrapolation. What it does establish is that two very different methods point to the same order of magnitude, which puts the burden on anyone who believes the curve will flatten by itself. Where the plastic is made shapes who sits on which side of the table: new capacity is concentrated in Asia, the Gulf and the United States, much of it integrated backwards into refineries as oil majors hedge against declining fuel demand (Plastics Europe, 2024). States heavily invested in polymer capacity are thus potential claimants through their companies’ foreign subsidiaries, while countries hosting foreign-owned capacity are most exposed if a treaty later obliges them to strand it. Exposure is not distributed evenly, nor in ambition’s favour.
The other curve has climbed nearly as steadily, if less visibly. The first treaty-based investor-state case was registered in 1987; by end-2009, there were roughly 380 known cases; by end-2024, there were 1,401, against 135 states plus the European Union, about three-quarters of them initiated from 2010 onwards (UNCTAD, 2025). Who sues whom is as revealing as how often: in 2024 about 55% of new cases were brought against developing countries, which hold a third of the world’s inward investment stock, while developed-country claimants brought roughly 80% – claims running along the same north–south axis as the treaty’s deepest financing dispute (UNCTAD, 2025). Two features of the load matter most for plastics. The first is its sectoral tilt: at least 249 fossil-fuel-related claims have been filed; the Energy Charter Treaty alone is the source of 172, and more than half of 2024’s cases concerned energy or extraction (Di Salvatore, Reference Di Salvatore2022; UNCTAD, 2024b, 2025). The plastics value chain begins in the same wells and crackers; firms ordered to shut polymer plants will rely on the same counsel and rehearse the same arguments. The second is the age of the instruments: about 85% of 2024 cases invoked treaties signed before 2010 (UNCTAD, 2025). The old promises are still binding, and claimants choose the treaty that suits.
Figure 2 plots the cumulative totals and extends them: a linear fit to the 2016–2024 counts implies 79 additional known cases a year, and the 55–80 scenario range puts the caseload between roughly 2,010 and 2,280 by 2035, with the fitted trend near the upper bound. Nothing favours the lower end: filing rates have been stable for a decade, the treaty stock is intact and third-party funding has made large claims easier to bring (Pelc, Reference Pelc2017; UNCTAD, 2025). Not every argument succeeds; fair and equitable treatment, the elastic standard housing ‘legitimate expectations’, was invoked in 835 known cases but found breached in only 280 (UNCTAD, 2025), yet that most arguments fail is cold comfort, for a claimant needs only one tribunal to accept one, and the average successful claim is now worth nearly a quarter of a billion dollars (Gaukrodger and Gordon, Reference Gaukrodger and Gordon2012). Modern awards compensate not only sunk costs but discounted future cash flows, so applied to a polymer plant with a design life of decades, a production cap becomes a claim for every tonne it will no longer make.
Outcomes complete the picture and carry a sectoral warning (Figure 3). Across all 1,050 concluded cases, 38% were decided for the state, 29% for the investor with damages and 17% settled (UNCTAD, 2025). In concluded fossil-fuel cases, the state’s position is markedly weaker: 22% ended in its favour, 31% were won by investors and 32% settled (UNCTAD, 2024b). Settlement is not neutral, as Ethyl shows; it often means paying the claimant, reversing the measure or both, so in the sector closest to plastics, investors obtain some satisfaction in nearly two disputes out of three. Set that beside the OECD’s finding that only its Global Ambition package brings leakage near zero, for 0.8% of global GDP (OECD, 2022b). This is the empirical heart of the argument: the scenario that solves the pollution problem is the one that strands the most private capital, while the legal system that prices stranded capital grows by about 79 known cases a year and wins or settles most of its fossil-fuel encounters. Two curves are rising, and a treaty is where they meet.
Outcomes of concluded treaty-based investor–state cases, as a percentage of concluded cases: all sectors (1987–2024, n = 1,050) compared with concluded fossil-fuel-related cases. Categories follow UNCTAD’s classification. For fossil-fuel cases, the residual category combines discontinued proceedings with decisions finding breach but awarding no damages, which UNCTAD reports separately only for the full population. Percentages are as reported and may not sum to 100 because of rounding. Sources: UNCTAD (2025) for all cases; UNCTAD (2024b) for fossil-fuel cases. Underlying values are given in Supplementary Data S1.
Figure 3. Long description
Two stacked horizontal bars, each spanning the full chart width, divided into segments totalling 100 per cent of concluded cases. A legend gives five outcome categories: decided in favour of the state, decided in favour of the investor, settled, discontinued, and breach found but no damages. The upper bar, labelled all concluded cases, n equals 1,050, reads from left to right: 38 per cent for the state, 29 per cent for the investor, 17 per cent settled, 14 per cent discontinued, and a narrow segment of about 2 per cent for breach without damages. The lower bar, labelled concluded fossil-fuel cases, reads: 22 per cent for the state, 31 per cent for the investor, 32 per cent settled and 15 per cent discontinued, with no separate breach-without-damages segment because UNCTAD reports that category only for the full population. Side by side, the state’s share falls from 38 to 22 per cent while investor wins and settlements together rise from 46 to 63 per cent.
Where the treaty meets the treaties
The mapping yields a clear and uncomfortable pattern, summarised in Table 1. The measures likely to do most for plastic pollution, those that reduce production and toxicity at the top of the chain, have the closest parallels in the arbitral record, while downstream measures that tidy up waste have almost none. This matters because the science is emphatic that upstream action is what counts (Lau et al., Reference Lau, Shiran, Bailey, Cook, Stuchtey, Koskella, Velis, Godfrey, Boucher, Murphy, Thompson, Jankowska, Castillo, Pilditch, Dixon, Koerselman, Kosior, Favoino, Gutberlet, Baulch, Atreya, Fischer, He, Petit, Sumaila, Neil, Bernhofen, Lawrence and Palardy2020; Simon et al., Reference Simon, Raubenheimer, Urho, Unger, Azoulay, Farrelly, Sousa, van Asselt, Carlini, Sekomo, Schulte, Busch, Wienrich and Weiand2021).
A firm limit on primary plastics production is, for many scientists, the single most important thing a treaty could do (Simon et al., Reference Simon, Raubenheimer, Urho, Unger, Azoulay, Farrelly, Sousa, van Asselt, Carlini, Sekomo, Schulte, Busch, Wienrich and Weiand2021), and what investment law is least comfortable with, because a cap deliberately strands capacity built in reliance on running it. Two energy disputes show the risk. In Rockhopper v. Italy, a tribunal held Italy’s refusal of a production concession, after it reintroduced a coastal drilling ban, to be an unlawful expropriation and ordered roughly €190 million, more than the company had spent, since the award compensated lost future profits (Rockhopper v. Italy, 2022); its annulment in 2025 does not undo the lesson. In Lone Pine v. Canada, a company sued over Quebec’s fracking moratorium beneath the St Lawrence, and the tribunal rejected the claim only after a decade (Lone Pine v. Canada, 2022). A cap on polymers invites the same argument, for indirect expropriation asks whether a measure has deprived an investor of the substance of its investment, which is what a hard output ceiling does to the marginal plant. Petrochemicals are the downstream cousin of an industry already generating that docket: 13 new fossil-fuel cases were filed in 2024, one seeking US$45 billion over Australia’s refusal of a coal project, and Germany faces a claim over its legislated coal phase-out (UNCTAD, 2025).
Restricting what plastic is made of, banning a hazardous additive or an unrecyclable polymer, looks like ordinary product regulation, and the arbitral record is more reassuring here, though not unreservedly. States have successfully defended chemical and product bans: Canada prevailed when Chemtura challenged its ban on the pesticide lindane, upheld as a valid exercise of police powers by a competent regulator (Chemtura v. Canada, 2010) and the United States defeated Methanex’s challenge to California’s ban on the fuel additive MTBE, on the ground that a non-discriminatory regulation for a public purpose is not an expropriation (Methanex v. United States, 2005). However, the punishment is the process: Canada had earlier settled Ethyl Corporation’s claim over a different additive, MMT, withdrawing the measure and paying compensation (Ethyl v. Canada, 1998). A ban that targets a named producer’s product, or that a tribunal reads as protectionist rather than precautionary, remains at risk, and even a successful defence takes years and millions (Pelc, Reference Pelc2017).
Extended producer responsibility schemes and plastics taxes rarely expropriate outright, so the pressure point shifts to fair and equitable treatment, read to protect an investor’s legitimate expectations of a stable regime and the most-invoked and most-found standard in the system. Producer responsibility is where national policy is actually moving: packaging schemes now operate or are legislated on every continent, and in the OECD’s modelling, taxes and producer responsibility do much of the work separating the policy scenarios from the baseline (OECD, 2022b). As these instruments spread, they touch the revenues of every packaging multinational, creating in each jurisdiction a class of protected foreign investors with a treaty to complain under; demand-side attitudes may be shifting (Lukács et al., Reference Lukács, Szeberényi and Papp-Váry2023; Raether et al., Reference Raether, Végvári and Szente2025), but the producer-side exposure is unchanged. The tobacco parallel is the sharpest lesson. When Uruguay required large graphic warnings and one presentation per brand, Philip Morris sued under a Swiss treaty, and in 2016, a tribunal upheld the measures as a legitimate exercise of the power to protect public health (Philip Morris v. Uruguay, 2016); a parallel challenge to Australia’s plain packaging was rejected on jurisdiction (Philip Morris Asia v. Australia, 2015). These are wins for the right to regulate, yet the chill still operated: New Zealand delayed its own plain-packaging law until the Australian litigation resolved, and said so (Tienhaara, Reference Tienhaara2018; Moehlecke, Reference Moehlecke2019). A small state weighing a levy on multinational packaging firms may make the same calculation, and few will announce it.
Limiting the movement of plastic scrap is live policy, given effect through the 2019 Basel amendments that brought most plastic waste under a prior-consent regime (Basel Convention, 2019). The trade has halved since 2014, to about 6 million tonnes in 2023, but it has been redirected rather than eliminated (Brooks et al., Reference Brooks, Wang and Jambeck2018; OECD, 2025). Each new restriction is a flashpoint, and the precedent is pointed: in S.D. Myers v. Canada, a tribunal found Canada’s temporary ban on PCB-waste exports discriminatory in effect, designed to protect Canadian processors, and Canada paid (S.D. Myers v. Canada, 2000). A state that restricts scrap imports to protect its environment while its own recyclers benefit is doing nothing wrong, but it is walking into that fact pattern. The same doctrine reaches permits and closures; landfill-permit refusals were held expropriations in Metalclad and Tecmed, and coal and nuclear phase-outs have drawn Vattenfall, RWE and Uniper claims (Metalclad v. Mexico, 2000; Tecmed v. Mexico, 2003; RWE and Uniper v. Netherlands, 2021; Vattenfall v. Germany II, 2021), so ordering a polymer plant to close, however sound the reasons, sits in the same neighbourhood.
The ledger has another side, and it should be set out with the same directness. Investment law is not indifferent to the power to regulate. The police-powers doctrine holds that a non-discriminatory measure adopted in good faith for a public purpose and applied with due process is an exercise of regulatory authority rather than a taking that must be paid for: Methanex stated it, Chemtura applied it from the perspective of a competent regulator acting on its mandate, and Philip Morris v. Uruguay carried the reasoning into health measures adopted on a contested evidentiary record. Non-discrimination protects the state that regulates even-handedly; what defeated Canada in S.D. Myers was differential effect on a foreign competitor, not the environmental objective it announced. Additionally, the newer treaty generation writes the point into the text, general exceptions modelled on Article XX of the General Agreement on Tariffs and Trade, express right-to-regulate clauses and annexes providing that non-discriminatory measures protecting legitimate public welfare objectives do not constitute indirect expropriation except in rare circumstances (Titi, Reference Titi2014; Sands and Peel, Reference Sands, Peel, Fabra and MacKenzie2018). A plastics measure that is evidence-based, applied to domestic and foreign producers alike and adopted with notice and a transition period is defensible, and states have defended such measures successfully. Exposure, on our reading, is therefore not the likelihood of losing. It is the breadth of arguable claims a measure creates and the cost of answering them.
We found no decided case in which the challenged measure was a plastics-control measure as such: the exposure described here is prospective, built by analogy and a sceptic could say the plastics docket stands at zero. However, that is what the tobacco docket looked like in 2009, and the fossil-fuel docket in the 1990s. Claims follow regulation, and the measures that would generate plastics claims are largely the ones the deadlock has so far prevented anyone from adopting.
Chill inside the negotiating room
Regulatory chill is usually assumed to operate after a treaty, when a government hesitates to use a power it holds; the plastics negotiations raise a further possibility: that the shadow of investment law is shaping the treaty before it is written. A state with significant polymer producers, and one that has given them treaty protection, would have a reason beyond raw economics to resist a binding cap; losses are easier for an investor to characterise as a compensable taking once a cap is adopted internationally and enforced at home. On that reading, a negotiator’s insistence on voluntary, downstream, nationally determined measures need not be only industrial protectionism; it could also be a hedge against future liability. We advance this as a hypothesis and not as a finding, and the evidentiary boundary should be marked clearly: delegations’ internal calculations are not on the public record, and we do not claim to have observed them. What can be observed is consistent with the hypothesis without establishing it. State positions cluster around fossil-fuel and polymer interests (Dreyer et al., Reference Dreyer, Hansen, Holmberg, Kielhöfer, Olsen and Stripple2025), and the sessions have stalled on the upstream measures Table 1 rates as most contestable while the least contestable attract broad support (IISD, 2026; UNEP, 2026). Other explanations account for the same pattern – revenue, employment, energy-transition politics and nothing here distinguishes between them. The hypothesis earns its place by also explaining why governments with few domestic producers hesitate: they host foreign-owned plants whose owners can sue. There is also a timing trap, and it does not depend on motive at all. The exposure that matters is already in place, in treaties signed before 2010, and will not lift during a session. Each year of delay is a year in which more petrochemical capacity is financed and covered by existing treaties, adding to the very bill of stranded assets that makes states pause. Delay is not a neutral resting state; it enlarges the problem that produced it.
Reform pathways
If the diagnosis is right, the treatments are known; the question is how much each can deliver (Table 2). An express carve-out would declare that a non-discriminatory, good-faith plastics measure cannot found a compensation claim; model language exists, from the health exceptions recent treaties carry to clauses excluding tobacco control from investor-state dispute settlement altogether (Titi, Reference Titi2014). The drafting is not the obstacle; reach is, for a plastics treaty binds only its parties and cannot sweep away the thousands of agreements already in force (Bonnitcha et al., Reference Bonnitcha, Poulsen and Waibel2017). This is why a second pathway matters: since 2017, the UNCITRAL Working Group III has pursued the first broad multilateral reform of investment-dispute settlement in a generation, a standing mechanism, appellate review and rules on adjudicator conduct, costs and funding (UNCITRAL, 2018) – whose language a plastics instrument could borrow. A third, immediately available, is for treaty parties to issue joint interpretive statements and explicit right-to-regulate and police-powers language directing tribunals to defer to genuine health and environmental measures; Methanex and Chemtura show this is feasible, and the reform is only to make it mandatory rather than discretionary (Van Harten, Reference Van Harten2008; Titi, Reference Titi2014). The fourth is to change the agreements themselves: exit is possible, as the UK and EU withdrawals from the Energy Charter Treaty in 2025 show (CIEL, 2024; Council of the European Union, 2024), but that treaty’s 20-year survival clause, still protecting existing investments after withdrawal, marks the limit. The numbers add a fifth consideration the literature underweights: speed. On our projections, a further 600–880 known cases will be filed by 2035, and every cracker financed in the interim enlarges the protected asset base, so slow reforms arrive after much exposure is locked in. That argues for front-loading what works at once: carve-out language in the instrument itself, and joint interpretive statements by willing states under their existing treaties, which cost nothing and need no one’s consent but their own. No single pathway suffices; used together – carve-out for clarity, alignment for reach, interpretation for the tribunals, reform for the root – they could keep the right to protect public health from being priced out of the negotiators’ hands.
Reform pathways for insulating plastics measures from investment claims, with their principal strengths and limits
Table 2. Long description
Table 2 appears in the article “How International Investment Law Could Freeze Global Action on Plastic Pollution” by Haider, Khan and Szeberenyi. Its caption reads: “Reform pathways for insulating plastics measures from investment claims, with their principal strengths and limits.” The table compares four reform pathways by the mechanism through which each would work, its principal strengths and its principal limits. The caption states that the pathways are not mutually exclusive and that the assessment is doctrinal and qualitative, drawn from the treaty texts and reform documents cited in the Methods.
Structure. The table has four columns and four data rows below a header row. The columns, from left to right, are: Pathway; Mechanism; Strengths; and Limits. Each row describes one pathway, and the cells in that row are read from left to right.
Row 1. Express carve-out in the plastics instrument. Mechanism: declares that good-faith, non-discriminatory plastics measures do not found compensation claims. Strengths: political clarity; guides tribunals; costless to adopt with the treaty. Limits: binds only treaty parties, and older investment agreements remain available to claimants.
Row 2. Alignment with the UNCITRAL Working Group III process. Mechanism: procedural reform, comprising a standing mechanism, appellate review, adjudicator conduct, and cost and funding rules. Strengths: multilateral momentum, and it addresses systemic defects. Limits: slow; procedural rather than substantive; outcome uncertain.
Row 3. Interpretive statements and right-to-regulate clauses. Mechanism: joint interpretations and explicit police-powers language directing deference to bona fide health and environmental measures. Strengths: available immediately under existing treaties, and consistent with the reasoning in Methanex and Chemtura. Limits: tribunals retain interpretive discretion, and uptake is uneven.
Row 4. Renegotiation, termination or withdrawal from investment agreements. Mechanism: removes or narrows the underlying cause of action. Strengths: proven feasible, since the exits of the United Kingdom and the European Union from the Energy Charter Treaty took effect in 2025. Limits: survival clauses protect existing investments for up to 20 years, and treaty shopping persists.
Summary of the pattern. Each pathway is paired with a stated limit. The carve-out binds only treaty parties, the Working Group III process is slow and procedural, interpretive statements leave tribunals with discretion, and withdrawal is qualified by survival clauses and treaty shopping.
Note: Pathways are not mutually exclusive; the assessment is doctrinal and qualitative, drawn from the treaty texts and reform documents cited in the Methods.
Conclusions
We began with a question: could one of the hardest barriers to ending plastic pollution be a body of law written to protect capital rather than to poison the sea? The mapping suggests it is real, if uneven. The measures that would do the most good – caps on production, bans on the worst polymers and additives, levies that make producers pay and limits on the waste trade – are the ones with the closest analogues in four decades of investor-state awards. That is a statement about legal contestability rather than about who would win: the police-powers doctrine, non-discrimination and the newer right-to-regulate clauses give a well-designed measure a real defence and states have won such cases. The numbers make the stakes concrete: production is on a path toward roughly 1,235 million tonnes in 2060, while the caseload grows by about 79 known cases a year and, in the sector nearest to plastics, delivers investors a win or a settlement in most concluded cases. The risk is rarely that a state loses outright; more often, the threat does its work first and a rule is softened before any claim is filed. That the evidence for such chill is circumstantial does not make it absent.
Three lines of research follow. The hypothesis that treaty exposure shapes negotiating positions is testable, by proxying state-level exposure with the stock of foreign-owned petrochemical capacity and applicable-treaty coverage and setting it against coded negotiating positions of the kind Dreyer et al. (Reference Dreyer, Hansen, Holmberg, Kielhöfer, Olsen and Stripple2025) have assembled. The chill mechanism can be studied prospectively rather than forensically, by tracking announced producer-responsibility and ban proposals in countries that do and do not face treaty exposure, extending Moehlecke’s design from tobacco to plastics. The mapping itself would gain from systematic coding of petrochemical-sector claims, which UNCTAD folds into chemicals and manufacturing.
The old maxim that treaties must be kept, pacta sunt servanda, was never meant to freeze the equally old principle that one must use one’s own so as not to harm another, sic utere tuo ut alienum non laedas. Determined states can act within investment law, as Uruguay showed; however, they should not have to gamble tens of millions to do the obvious. Ending plastic pollution will take confronting investment law as squarely as trade law, while the pen is still in hand.
Open peer review
To view the open peer review materials for this article, please visit http://doi.org/10.1017/plc.2026.10067.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/plc.2026.10067.
Acknowledgements
The authors are grateful to Dr Jessica Jones, and also thank the two reviewers and the handling editor for comments that materially improved the manuscript.
AI statement
The authors declare that no AI-LLM tools were used in the writing of this article; Grammarly Premium was used for language polishing.
Author contributions
A.H. conceptualised the study, led the legal analysis and drafted the manuscript. A.K. contributed to data collection, quantitative analysis and revision of the manuscript. A.S. contributed to the legal mapping, interpretation of the findings and critical revision of the manuscript. All authors approved the final version.
Competing interest
The authors declare none.
Data availability statement
No new data were created in this study. All quantitative series analysed are publicly available from the cited sources: global plastics production from Geyer et al. (Reference Geyer, Jambeck and Law2017) as compiled by Our World in Data and from Plastics Europe (2024); projections from OECD (2022b); plastic waste-trade statistics from OECD (2025); and investor–state arbitration statistics from the UNCTAD ISDS Navigator and IIA Issues Notes (UNCTAD, 2024b, 2025). The numerical values used to construct Figures 1–3, with their sources, are provided as Supplementary Data S1. All primary legal sources, arbitral awards, treaty texts and negotiation documents are publicly available through the cited repositories.
Ethics and approvals
This research did not involve human participants, animal subjects or personal data. It draws exclusively on publicly available documentary and statistical sources, published arbitral awards, treaty texts, negotiation documents and official statistical series and no ethical approval was therefore required.
Funding statement
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Open access
Comments
No accompanying comment.