The Chișinău Declaration in the Data: Non-Refoulement and States’ Insatiable Appetite for a Restrained Court

There is a striking hidden controversy at the heart of the recent Chișinău Declaration issued by the members of the Council of Europe on May 15, 2026. All 46 member states have adopted a coordinated demand, asking that the European Court of Human Rights (the Court, the ECtHR) in particular to narrow their interpretation of Article 3 (the prohibition of torture and inhuman or degrading treatment) in cases involving irregular migrants, asylum seekers, refugees, and to recognize the sovereign right to remove foreign nationals. The controversy we are speaking about is beyond the scope of the Declaration’s text, although the text received a fair number of criticisms (see here, here, here, and here). Beyond its ambition, what makes this Declaration striking is its timing and focus. The states pressing hardest for a more restrained Court are making their request in an area where the Court has already been most deferential. For over a decade, the Court has already been doing exactly (albeit quietly) what member states are demanding in the Declaration. The ECtHR      has been issuing more state-friendly rulings when reviewing irregular immigration or asylum related cases at least under Article 3. Some of these cases directly involve the non-refoulement principle (the prohibition on returning individuals to a location where they might face torture or ill-treatment).

The evidence behind this claim comes from our analysis of a corpus of 14,196 Article 3 judgments spanning 1967 to 2025. Rather than treating Article 3 as a monolith, we disaggregate the data by claim type. This allows us to determine how the Court may have treated some Article 3 claims differently than others. We find that the Court had already become more deferential in its treatment of Article 3 claims that concern the non-refoulement principle.

As seen in the Figure 1 below, across virtually every major claim category under Article 3 (ranging from torture, detention conditions, ill-treatment during custody, intrusive detention measures, procedural failures, or provision of legal remedy, among others), the Court’s finding that a State is in breach of the provision has trended upward since 2007. Formal Mann–Kendall trend tests on the year-by-year merits violation rate find statistically significant upward slopes across Article 3 claim categories, ranging from +0.4 percentage points per year for procedural failures to +1.9 for intrusive detention measures. Non-refoulement is the lone outlier in the opposite direction, with a point estimate of a decrease in the number of findings of violations by the State, amounting to −0.8 percentage points per year. This means that violation findings increased in all but one type of Article 3 claim; precisely the one that concerned rights of immigrants and asylum-seekers.

Figure 1: Non-refoulement violation rates vs other obligations (1995–2025)

Figure 1 shows violation rates among merits claims by year, pooled across three-year rolling windows. Non-refoulement decisions are highlighted in red and the others in blue. The two lines converge briefly around 2009 before diverging sharply. Most notably, other Article 3 categories trend toward near-certain violation findings, while non-refoulement rates plateau and decline, tracing the period of sustained political pressure on the Court in this domain exactly.

Figure 2 complements this picture by plotting admissibility rates and violation rates for non-refoulement claims (red) against all other Article 3 categories (blue) across five-year periods from 2000 to 2025. On admissibility, non-refoulement claims are declared admissible at dramatically lower rates than other Article 3 categories in every single period — a gap that holds at roughly 15 to 35 percentage points throughout and is statistically significant (p < .001) in all windows. This means the Court is filtering out non-refoulement cases at the threshold stage at a far higher rate than cases involving torture, detention conditions, or other Article 3 obligations. We should also note that these figures likely overestimate the true admissibility rates for both groups. This is because the Court does not systematically publish single-judge and committee inadmissibility decisions, meaning the true filtering effect is likely even larger than what the data captures.

The violation rate panel tells an equally striking story. The non-refoulement violation rate appears relatively stable, hovering between 50% and 73% across periods, while other Article 3 categories trend sharply upward and reach 98% in 2020–2025. The gap between the two widens over time, from 18 percentage points in 2000–2004 to 28 percentage points by 2020–2025. 

Figure 2: Comparison between non-refoulement and other categories in admissibility decisions and merit judgments (five-year-period intervals)

These trends collectively show that the Court has been responsive to states and accommodated political pressure in the most contested domain of Article 3, namely the non-refoulement related cases, while expanding its reach elsewhere.

This divergence is not coincidental. In earlier work (Yildiz 2024), the pattern was documented across 2,294 judgments from 1967 to 2016. The post-2010 proportion of violation decisions in non-refoulement related cases was twelve percentage points lower relative to prior periods, while the proportion of violation decisions in other areas rose by twenty-six percentage points. Ongoing work traces the mechanism to the 2011 Izmir High-Level Conference, where Sweden, Slovakia, Turkey, and Russia explicitly called on the Court to stop acting as “an immigration appeals tribunal” (see, Yildiz & Yuksel 2025). The Court heard this complaint, also reiterated in states’ third-party interventions in cases involving the non-refoulement principle, and adjusted its treatment of such cases.

The logic of that adjustment is legible in the data. The bifurcated pattern the data documents is not evidence of a Court insulated from state preferences. Rather, it shows a reactive Court. However, this fact does not seem to be present in the way states portray the Court in public discussions or in this Declaration, for that matter. The Court, for its part, has performed its adjustments rather stealthily, by pushing forward where states are silent and holding back where they have complained loudly. The result is a two-speed working method. The Court is harder on states across the general Article 3 landscape, measurably softer in the one domain where organized political pressure has been most sustained: the non-refoulement related cases.

The Declaration, we argue, is an attempt to institutionalize a retreat the Court has largely already made. Its deeper ambition is not doctrinal but structural. Beyond the surface, member states’ demands about having more control of migration and their sovereign right to expel is a broad claim about how the Convention system should work in general. Their message is clear: Strasbourg should defer to domestic authorities especially in politically contentious cases like migration. Right now, the most concrete and detailed demands are concentrated in migration and asylum cases, but the Declaration’s broader subsidiarity arguments are not so limited. The principle of deference to domestic authorities, if absorbed by the Court as a general operating rule, has the chance to reshape its relationship with national authorities across the entire Convention (see also, Stiansen & Voeten 2020). In the future, this may not only be limited cases involving migrants, but whenever they deem a particular topic politically contentious.

Beyond subsidiarity, the Declaration contains one genuinely novel doctrinal move: the invocation of “democracy capable of defending itself” in the context of instrumentalized migration (§§37–40). That doctrine was originally developed in cases about banning extremist political parties or penalizing their members (see for example, Vogt v. Germany, no. 17851/91, 26 September 1995; Ždanoka v. Latvia [GC], no. 58278/00, 16 March 2006). Instrumentalized migration here refers to any hostile state’s attempts to facilitate or engineer migrant flows across its borders as a form of political pressure on neighboring European states. The Declaration treats this not merely as a border management problem but as a threat to democratic order itself. It emphasizes that this threat is serious enough to invoke the same constitutional logic that justified restricting rights in the party-ban cases mentioned above. By deliberately invoking this concept, the Declaration draws attention to the potential that migrant flows can be weaponized and offers states a justification to suspend (fully or partially) their ordinary Convention obligations, which would include the non-refoulement principle. Moreover, the Declaration endorses third-country processing and “return hubs” (§46), referring to the model whereby states would send asylum seekers to a third country for processing rather than handling claims domestically (e.g., UK’s Rwanda asylum plan or the Italy-Albania asylum processing deal). These are essentially not only demands for state-friendly judicial reasoning or decisions. They are requests for new categories of exceptions and soft launching of justifications for future suspension of Convention obligations.  

The Chișinău Declaration is testament to the fact that selective accommodation is no longer enough for European states. What states want is not just a softer line on non-refoulement, but a restructured institution in which subsidiarity doctrine and novel categories of exceptions together place a higher weight in legal reasoning. The data shows the Court has been willing to move in that direction before, albeit quietly and incrementally. The question, then, is will this Declaration speed up or formalize this process?

Beyond this particular case, the Chișinău Declaration is an important moment for how we theorize international courts as institutions. Existing theories often treat these courts as agents of their state principals: states create them, delegate authority to them, and then release them into the world to operate independently (see for example, Pollack 1997; Alter 2008). What episodes like this reveal is that the delegation is never truly complete. Contracting is not a singular event. States retain the capacity to return to the table, and the Declaration is precisely such a moment. The Declaration is not a casual episode of communication between governments and courts, but an active act of re-contracting. States are reshaping the Court’s working methods and priorities by making their red lines explicit, and doing so at a scale and coordination level that goes well beyond previous attempts. What this episode prompts is a question the scholarly literature will need to take seriously.

For the Court, the immediate question is how to respond. Based on what we know about the Court, its institutional survival instinct is likely to be activated next, and it will try to weather the storm through selective restraint, accommodating the pressure incrementally and stealthily, as it has done before (Yildiz 2024). But the longer-term normative and jurisprudential consequences of this move are harder to predict. Doctrinal adjustments made under political pressure have a way of outlasting the pressure that produced them. Meanwhile, the asylum seekers and migrants at the center of these cases have no place in this conversation. They are, as the African proverb says, “the grass that suffers when elephants fight”.

Methodology: analysis covers 14,196 ECHR Article 3 judgments (1967–2025). Dataset constructed via LLM classifier validated against 2,884 hand-coded cases.