No. 01 // Primary strand

Internet Shutdowns as a Governance Instrument, and What Could Hold Them Accountable

Abstract

State-ordered internet shutdowns have moved, in roughly a decade, from an improvised emergency measure to a routine instrument of governance. Governments that order them almost uniformly justify them in the language of security and public order; independent measurement bodies almost as uniformly document them as tools of information control deployed around elections, protests, and armed conflict, with severe and measurable economic and humanitarian cost. This paper treats the shutdown not primarily as a human-rights violation to be condemned, which it often is, but as a governance problem to be constrained: an action whose defining feature is that the actor ordering it also owns the switch, controls the evidence, and writes the domestic law under which it is carried out. That configuration produces an accountability gap that ordinary oversight does not close. I survey the phenomenon, its evidentiary base, and the logic states use to justify it; I map seven existing and emerging accountability mechanisms across international, regional, economic, corporate, and technical domains; and I propose a layered analytical framework for assessing which mechanisms can realistically "bite" against an actor that controls the infrastructure. The paper closes with an open research agenda. The intended contribution is a governance-and-accountability lens on a subject that is still studied mostly as either a technical outage or a discrete rights abuse.

Keywords: internet shutdowns · network disruption · digital governance · accountability · critical information infrastructure · human rights online · digital authoritarianism · information controls

---

1. Introduction and research questions

On a growing number of days each year, in a growing number of countries, a government decides that the most effective thing it can do is to switch off the internet. Sometimes the switch is total: a national blackout that severs a whole population from banking, work, family, medical information, and one another. More often it is selective and deniable: a throttle that degrades connectivity to the point of uselessness while allowing officials to claim the network is "operating normally," or a block on specific platforms during the precise days a protest or an election is underway.

The academic and policy literature has grown quickly, but it has grown along two mostly separate tracks. The first is technical and empirical: measurement organizations detect, timestamp, and quantify disruptions. The second is legal and normative: human-rights bodies assess shutdowns against free-expression standards and, overwhelmingly, find them wanting. What is comparatively underdeveloped is the track between the two, the treatment of the shutdown as a governance instrument whose central analytical feature is a structural accountability gap. When a state shuts down a network, the same actor typically (a) issues the order, often informally; (b) controls the physical and logical infrastructure that carries it out, usually through licensed operators it regulates; (c) monopolizes the ground-truth data about what was disrupted and why; and (d) supplies the legal justification after the fact. Very few state actions concentrate ordering power, execution power, evidentiary control, and legal cover so completely in one place. That concentration, not the outage itself, is the governance problem.

This paper is organized around three questions:

1. Descriptive. What, precisely, is the object of study, how do we know it is happening, and what are its scale, patterns, and cost?
2. Diagnostic. Why is the shutdown so resistant to accountability, and what specifically breaks in the normal chain of constraint on state power?
3. Prescriptive. Which accountability mechanisms, legal, judicial, economic, corporate, technical, and multilateral, can realistically constrain an actor that controls the switch, and under what conditions?

A note on standpoint. I approach this as an engineer who studied the technology of network control, my bachelor's thesis examined how users evade deep-packet-inspection surveillance of their traffic, and later as a resident who lived through national-scale shutdowns from the receiving end while running a consulting practice. The paper is deliberately policy-analytical: the unit of analysis is the system and the institution, not any regime or individual. Resilience-under-constraint and accountability-for-disruption are engineering and governance questions; they are not, in this paper, a political posture.

---

2. Defining the object of study

"Internet shutdown" is used loosely in public debate to cover a family of distinct interventions with distinct governance and accountability profiles. Precision here matters, because the deniability of the softer forms is itself part of the governance instrument. Building on the working definition adopted by the #KeepItOn coalition and refined in subsequent measurement literature, I distinguish:

• Full network blackout. Connectivity is severed at the national or regional level, typically by instructing licensed operators to withdraw routes or power down infrastructure. Highly visible, maximally disruptive, hardest to deny.
• Mobile-data shutdown. Fixed-line connectivity remains while mobile data, the primary or only access method for most of the population in many countries, is cut. Targets the majority while preserving the appearance of an operating internet.
• Bandwidth throttling. Connectivity is technically present but deliberately degraded to the point where images, video, and messaging fail. The most deniable form: officials can assert the network is "up" while it is functionally unusable, and detection requires measurement infrastructure the state does not control.
• Service- or platform-blocking. Specific services, social media, messaging, VPN protocols, are blocked while general connectivity continues. Often the most durable form, outlasting the acute event and hardening into permanent censorship.
• Localized / targeted disruption. Connectivity is cut in a specific district, region, or around a specific facility, contained enough to limit political blowback while still achieving information control.

Each form maps to a different accountability challenge. Blackouts are easy to detect but justified under emergency doctrine; throttling is easy to justify but hard to prove; platform-blocking migrates from the emergency register into the ordinary-censorship register and thereby escapes the "temporary measure" framing entirely. Any accountability framework that treats "shutdown" as a single undifferentiated act will fail against at least one of these forms. [OPEN: I am developing a fuller taxonomy that also captures internet curfews, recurring scheduled shutdowns, e.g. around examination periods or nightly during unrest, which have a different normalization dynamic.]

---

3. The evidentiary base: how we know

A distinctive feature of this field is that its most reliable evidence is produced outside the state doing the shutting-down, by a small ecosystem of measurement organizations. Understanding accountability requires understanding this measurement layer, because in this domain measurement is itself a form of accountability: it is the mechanism that converts a deniable act into a documented one.

• Access Now and the #KeepItOn coalition maintain the most-cited annual dataset of documented shutdowns, drawing on technical measurement and on-the-ground verification, and publish the field's headline annual counts (accessnow.org/keepiton).
• NetBlocks operates a real-time observatory that detects and reports disruptions as they happen, frequently providing the first public confirmation of an event (netblocks.org).
• The Open Observatory of Network Interference (OONI) crowdsources censorship and blocking measurements from volunteer probes worldwide, producing open data on what specifically is being blocked (ooni.org).
• Georgia Tech's IODA (Internet Outage Detection and Analysis) monitors macroscopic connectivity at the network level, offering an independent signal that corroborates or contradicts official claims (ioda.inetintel.cc.gatech.edu).
• Cloudflare Radar and the Google Transparency Report expose aggregate traffic patterns that make national-scale disruptions visible from the outside (radar.cloudflare.com).
• Freedom House's Freedom on the Net provides the annual qualitative-comparative assessment of state information controls, including shutdowns, across dozens of countries (freedomhouse.org).

This ecosystem is the reason shutdowns can be discussed as fact rather than allegation. It also has structural limits worth naming for an accountability analysis: measurement can establish that connectivity fell and when, but attributing an order to a specific official, and distinguishing a deliberate shutdown from an incidental failure, often still requires testimony or leaked documentation the state controls. The evidentiary layer closes part of the accountability gap and leaves part of it open. [OPEN: how far can purely technical attribution be pushed before it needs corroborating human evidence? This is a live methodological question for the thesis.]

---

4. Scale, patterns, and cost

Three empirical patterns are robust across the measurement literature and matter for the governance argument.

First, shutdowns are frequent and increasing. Access Now's annual #KeepItOn reporting has documented shutdowns numbering in the low hundreds per year across dozens of countries in recent years, a level that would have been unthinkable when the practice was first systematically tracked in the mid-2010s. The trend line is one of normalization, not exception.

Second, shutdowns cluster around predictable political triggers. The measurement record shows disruptions concentrating around three moments above all: elections, protests and civil unrest, and armed conflict. This clustering is analytically important. An outage that reliably coincides with the exact windows in which a population would most want to communicate, and in which a government would least want it to, is difficult to explain as a neutral security measure and easy to read as an information-control instrument. The timing is, in effect, evidence of intent.

Third, the cost is large, measurable, and borne broadly. Independent estimates, notably the annual Cost of Internet Shutdowns accounting published by Top10VPN and earlier foundational work by Deloitte for the Global Network Initiative (2016) and by Darrell West at Brookings (2016), put the global economic cost of shutdowns in the billions of dollars annually, with individual prolonged national shutdowns running into the hundreds of millions (top10vpn.com/cost-of-internet-shutdowns). These figures capture only the measurable economic loss; they do not capture the humanitarian cost, denial of access to emergency services, medical information, banking, and the ability to document abuses, which is real but harder to price. The economic figures matter for accountability specifically because they translate a rights harm into the language of material interest, the language in which trade partners, investors, and finance ministries can be moved.

A methodological caution I hold throughout: exact annual counts and cost figures vary by source and by counting methodology (what qualifies as a distinct "shutdown," how throttling is treated, which cost model is used), and they are revised as reporting matures. I therefore treat magnitude and direction as established, hundreds of events annually, billions in annual cost, a rising trend, while treating any single precise figure as provisional and source-dependent. Over-precise citation of a contested number is itself a credibility risk in this field.

---

5. The governance logic: why states shut down networks

To constrain a practice one must first take seriously the account its practitioners give of it. States that order shutdowns rarely describe them as censorship. The recurring justifications, drawn from official statements catalogued across the measurement and legal literature, cluster into four:

1. Public safety and order. The shutdown is framed as preventing the coordination of violence, riots, or "rumors" that could inflame unrest. This is the most common and most rhetorically durable justification because it invokes a genuine state responsibility.
2. National security. During armed conflict or perceived external threat, the network is framed as an attack surface or a channel for hostile coordination, and its suspension as a defensive measure.
3. Countering misinformation. A newer justification: the shutdown is presented as stopping the spread of false or destabilizing information.
4. Examination integrity and other administrative rationales. In several states, recurring shutdowns are justified as preventing cheating during national examinations, a rationale notable for how thoroughly it normalizes the instrument by attaching it to a routine, apolitical event.

The analytical point is not that these justifications are always insincere. It is that they are structurally unfalsifiable from inside the ordering state. Because the government controls the network, the evidence, and the legal framing, its security claim cannot be independently tested through the ordinary domestic machinery, courts, legislatures, regulators, that would test other exercises of emergency power. The justification and the means of checking the justification are held by the same hand. This is why the shutdown is better understood as a governance instrument than as a discrete policy decision: it is a repeatable capability whose exercise is, by design, largely self-authorizing.

---

6. The accountability gap: why shutdowns resist constraint

Normal constraints on state power assume a separation between the actor exercising power and the actors evaluating it. The shutdown collapses several of those separations at once. I identify five structural features that together produce the accountability gap.

1. Switch control. The state, through its regulatory power over a small number of licensed operators, physically and logically controls the network. Ordering a shutdown often requires nothing more than an instruction, sometimes verbal, to a handful of firms that cannot refuse without losing their license. There is no friction, no procurement, no visible act, at the moment of ordering.
2. Evidentiary monopoly. The state holds the ground truth about what was disrupted, on whose order, and for what stated reason. Independent measurement (Section 3) partially breaks this monopoly from the outside, which is precisely why measurement organizations are treated by some governments as adversaries.
3. Legal self-authorization. Most states that shut down networks do so under broad emergency, telecommunications, or public-order statutes that vest sweeping discretion in the executive with minimal procedural constraint, no requirement of prior judicial authorization, no automatic sunset, no published reasons. The law does not constrain the instrument; it licenses it.
4. Attribution difficulty. Because orders are often informal and execution is delegated to private operators, attributing a specific shutdown to a specific decision-maker, the precondition for most accountability, is genuinely hard, and deniability (especially via throttling) is a designed feature.
5. Sovereignty framing. At the international level, the state can frame network management as an internal matter of sovereign infrastructure governance, invoking the same non-interference norms that constrain external pressure generally.

An accountability framework worth the name has to address these features specifically, not merely assert that shutdowns are wrong. The next section maps the mechanisms that exist or are emerging, and Section 8 organizes them by which feature of the gap each one actually attacks.

---

7. Mapping the accountability toolkit

7.1 International human rights law

The normative foundation is well established. The UN Human Rights Council, in Resolution A/HRC/RES/32/13 (2016), "condemn[ed] unequivocally measures to intentionally prevent or disrupt access to or dissemination of information online," and has reaffirmed the position since. The substantive standard flows from Article 19 of the ICCPR and the Human Rights Committee's General Comment No. 34: restrictions on expression must be provided by law and be necessary and proportionate to a legitimate aim. Successive UN Special Rapporteurs on freedom of opinion and expression have applied this directly to shutdowns (notably the 2017 report A/HRC/35/22), and the OHCHR's 2022 report on internet shutdowns (A/HRC/50/55) consolidated the finding that blanket shutdowns are, essentially by their nature, disproportionate. The strength of this mechanism is that it supplies an authoritative, near-consensus legal standard. Its weakness is enforcement: it constrains through legitimacy and reputational pressure, not through binding remedy, and it attacks the legal self-authorization feature only to the extent a state cares about the international normative verdict.

7.2 Regional and domestic jurisprudence

The most concrete accountability wins have come from courts, and they are worth close study because they show the gap being narrowed in practice.

• *ECOWAS Court of Justice, Amnesty International Togo and Others v. The Togolese Republic (2020).* The regional court held that Togo's 2017 shutdown was unlawful, an interference with freedom of expression, and ordered compensation and guarantees of non-repetition. A regional human-rights court imposing a binding remedy on a member state for a shutdown is a genuine precedent.
• *Supreme Court of India, Anuradha Bhasin v. Union of India (2020). Arising from the prolonged Jammu & Kashmir shutdown, the Court held that an indefinite* shutdown is impermissible, that shutdown orders must be published, reasoned, necessary, proportionate, and subject to periodic review, and that access to the internet is instrumentally protected under the constitutional guarantees of expression and trade. It stopped short of ordering restoration but imposed procedural discipline on the instrument.

These rulings attack the legal self-authorization and attribution features directly: they force orders into writing, demand published reasons, and create a reviewable record. Their limits are jurisdictional (they bind only where such courts have reach and independence) and temporal (litigation is slow; the shutdown is fast). [OPEN: a systematic comparative reading of shutdown jurisprudence across ECOWAS, Indian, East African, and European fora is one of the paper's planned empirical contributions.]

7.3 Economic accountability

Because the cost of shutdowns is measurable (Section 4), it can be operationalized as leverage. Three sub-mechanisms:

• Cost accounting as advocacy. Publishing rigorous, methodologically transparent economic-loss figures reframes the shutdown from a rights abuse (which some governments discount) into a self-inflicted economic wound (which finance ministries and investors do not), splitting the ordering government internally.
• Conditionality. Trade preferences, development finance, and aid can, in principle, be conditioned on non-disruption, attaching a concrete external cost to the instrument.
• Targeted measures. Where shutdowns accompany grave abuses, targeted sanctions on the specific officials who order them, rather than on the population, attack the attribution feature by insisting that a person, not an abstraction, made the decision.

The promise of economic accountability is that it speaks the language of material interest; the risk is bluntness and hypocrisy (conditionality applied selectively loses moral force) and the perennial problem that broad economic pressure can harm the same population the shutdown already harmed.

7.4 Corporate and intermediary accountability

Shutdowns are executed by private telecommunications operators. That makes operators a genuine, and under-used, accountability node. The Global Network Initiative (GNI) and the former Telecommunications Industry Dialogue have developed principles under which operators are expected to demand written and lawful orders, narrow the scope of compliance, and publish transparency reporting about the demands they receive (globalnetworkinitiative.org). An operator that insists on a written, legally grounded order before disrupting service does not stop a determined state, but it forces the legal self-authorization and attribution features into the open: it converts a verbal instruction into a document, and a document is the beginning of a record. The limit is obvious, operators are licensed by the state they would be resisting, so their leverage is bounded by their willingness to risk the license. Multinational operators with reputational exposure in rule-of-law markets have more room than domestic incumbents.

7.5 Measurement and documentation as accountability

As argued in Section 3, the measurement ecosystem is not merely descriptive; it is itself an accountability mechanism, arguably the most effective one currently operating, because it directly attacks the evidentiary monopoly. Real-time detection strips deniability from throttling; archived, timestamped, methodologically transparent records feed litigation (Section 7.2), cost accounting (Section 7.3), and the international verdict (Section 7.1). A great deal of the accountability that does occur is downstream of the fact that independent parties can now prove, against official denial, that the network was deliberately disrupted. Strengthening, funding, and protecting this layer is therefore not a technical footnote but a central accountability strategy.

7.6 Multilateral and coalition pressure

The #KeepItOn coalition (hundreds of civil-society organizations worldwide) aggregates advocacy, litigation support, and rapid response, converting isolated national events into a sustained international campaign with a memory. Diplomatic mechanisms, joint statements, Universal Periodic Review interventions, and specialized engagement (including within the ITU and the Freedom Online Coalition) apply reputational and peer pressure. This mechanism works on states that value international standing; it attacks the sovereignty framing by denying that shutdowns are a purely internal matter and insisting they are a legitimate subject of international scrutiny.

7.7 Domestic legal safeguards

The most durable constraint, and the hardest to achieve, is domestic legal reform that removes the self-authorizing quality of the instrument: statutes requiring prior independent (ideally judicial) authorization, published and reasoned orders, strict necessity-and-proportionality tests, automatic time limits with affirmative renewal, and mandatory post-hoc review. The Indian Supreme Court's Bhasin requirements point toward this model even where legislatures have not. Domestic safeguards are the only mechanism that attacks switch control at its root, by inserting a procedural gate between the impulse to shut down and the capacity to do so, but they depend on exactly the institutional independence that shutdown-prone states most often lack, which is the field's central paradox.

---

8. Toward a layered accountability framework

The mechanisms in Section 7 are usually discussed as a menu. I argue they are better understood as layers, each targeting a specific feature of the accountability gap identified in Section 6, with effectiveness determined by how many layers can be brought to bear and how independent each is from the ordering state. The synthesis below is the paper's principal analytical contribution and is offered as a hypothesis to be tested against cases.

Two propositions follow, and structure the rest of the research:

• P1 — The externally-independent layers do the near-term work. Measurement (7.5) and coalition pressure (7.6) are the layers least dependent on the ordering state's own institutions, and they are correspondingly the ones that function even against determined authoritarian actors. They cannot stop a shutdown, but they reliably document and cost it, which is the precondition for everything else.
• P2 — Only the state-dependent layers can prevent. Prevention, actually inserting friction before the switch is thrown, requires domestic safeguards and operator resistance (7.7, 7.4), which are exactly the layers most captured by the ordering state. This is why accountability for shutdowns is, at present, overwhelmingly ex post (documenting and penalizing) rather than ex ante (preventing). Closing that gap is the field's hardest and most important open problem.

The framework predicts, testably, that shutdown episodes will be constrained roughly in proportion to (a) how many independent layers were activated and (b) how exposed the ordering state was to the layers it does not control (international standing, trade dependence, an independent domestic court, multinational operators). Section 9 sketches the cases against which I intend to test this.

---

9. Illustrative cases

These are compressed sketches, not the full case studies the thesis will develop. Each is chosen because it stresses the framework differently.

• India (Jammu & Kashmir). The world's most-documented prolonged subnational shutdown, and the site of the Bhasin judgment. A case where a genuinely independent domestic court partially activated the legal-self-authorization and attribution layers, producing procedural constraint without full prevention. Tests the state-dependent layers in a context where they exist but are strained.
• Togo. The ECOWAS ruling case: a smaller state, exposed to a functioning regional court, where an external judicial layer imposed a binding remedy. Tests whether regional courts can substitute for absent domestic ones.
• Myanmar. Post-coup shutdowns during armed conflict, where the national-security justification is at its rhetorical strongest and international leverage at its weakest. Tests the sovereignty-framing and security-justification arguments under maximum stress.
• Ethiopia (Tigray). Prolonged conflict-linked shutdown with severe humanitarian consequences and an information vacuum that itself impeded accountability. Tests the interaction between shutdowns and the documentation of other abuses, the shutdown as a meta-abuse that disables oversight of everything else.

The cross-case question is P1/P2: does constraint track the number and independence of activated layers, as the framework predicts? [OPEN: case selection and coding scheme are still being finalized; I am weighing adding a "negative" case, a shutdown-prone state that did not shut down under pressure, to test the preventive layers.]

---

10. Research agenda and open questions

This working paper is a scaffold. The questions I am actively developing:

1. Can the layered framework be operationalized into a coding scheme that scores each shutdown episode by layers-activated and state-exposure, and predicts observed constraint? This is the empirical core of the planned thesis.
2. Where exactly does technical attribution reach its limit and require corroborating human evidence, and can measurement methods be pushed further to narrow that gap? (Section 3.)
3. *Is ex ante prevention achievable in captured institutional contexts at all, or is external ex post* accountability the realistic ceiling for the hardest cases? (Section 8, P2.)
4. What is the accountability profile of throttling specifically, the most deniable and fastest-growing form, as distinct from full blackouts? Most law and advocacy is still written for the blackout.
5. How should the humanitarian cost be priced so that it enters the same decision calculus as the economic cost, without reducing human harm to a number?
6. What is the role of diaspora and exile actors in activating the externally-independent layers when domestic civil society is silenced? (This links to my related work; see Research 1.)

---

11. Preliminary conclusions

Three conclusions are stable enough to state, provisionally, at v0.6.

First, the internet shutdown is best analyzed not as a technical outage or a discrete rights abuse but as a governance instrument defined by a structural accountability gap, one that concentrates ordering, execution, evidence, and legal cover in a single actor. Naming that configuration precisely is the precondition for constraining it.

Second, the accountability mechanisms that actually function today are overwhelmingly the *externally-independent, ex post*** ones, measurement and coalition documentation, precisely because they do not depend on the institutions the ordering state controls. This is a real achievement and an inadequate one: it documents and costs shutdowns without preventing them.

Third, the frontier problem is prevention in captured contexts: building friction before the switch is thrown, through domestic safeguards and operator resistance, in exactly the settings where those layers are weakest. This is where I believe a cyber-governance practitioner who understands both the network and the institution can contribute something that neither pure technologists nor pure lawyers reliably do, and it is the direction in which I intend to take this research.

---

References

Selected sources. This is a living reference list; entries are added as the paper develops. Where a specific figure is cited in the text it is treated as provisional and source-dependent (Section 4).

• Access Now / #KeepItOn coalition. Internet shutdowns annual reports and STOP dataset. https://www.accessnow.org/keepiton/
• NetBlocks. Internet observatory and disruption reporting. https://netblocks.org/
• Open Observatory of Network Interference (OONI). Open censorship measurement data. https://ooni.org/
• Georgia Tech IODA. Internet Outage Detection and Analysis. https://ioda.inetintel.cc.gatech.edu/
• Cloudflare Radar. Internet traffic and disruption insights. https://radar.cloudflare.com/
• Freedom House. Freedom on the Net (annual). https://freedomhouse.org/report/freedom-net
• Top10VPN. The Global Cost of Internet Shutdowns (annual accounting). https://www.top10vpn.com/cost-of-internet-shutdowns/
• UN Human Rights Council. Resolution A/HRC/RES/32/13 (2016), The promotion, protection and enjoyment of human rights on the Internet.
• UN Special Rapporteur on the promotion and protection of the right to freedom of opinion and expression. Report A/HRC/35/22 (2017).
• UN OHCHR. Internet shutdowns: trends, causes, legal implications and impacts on a range of human rights, report A/HRC/50/55 (2022).
• UN Human Rights Committee. General Comment No. 34 on Article 19 ICCPR (2011).
• ECOWAS Community Court of Justice. Amnesty International Togo & Ors v. The Togolese Republic, judgment (2020).
• Supreme Court of India. Anuradha Bhasin v. Union of India, (2020) 3 SCC 637.
• Global Network Initiative (GNI). Principles on Freedom of Expression and Privacy and implementation guidance. https://globalnetworkinitiative.org/
• Deloitte for the Global Network Initiative. The Economic Impact of Disruptions to Internet Connectivity (2016).
• West, D. M. Internet shutdowns cost countries $2.4 billion last year. Brookings Institution / Center for Technology Innovation (2016).
• Rydzak, J. Of Blackouts and Bandhs: The Strategy and Structure of Disconnected Protest in India. Stanford Global Digital Policy Incubator (2019).
• CIPESA. Research on internet shutdowns and information controls in Africa. https://cipesa.org/

---

Version history

• v0.6 (28 Jul 2026) — Added the layered accountability framework (Sec. 8) with the P1/P2 propositions and the gap-to-layer mapping table; expanded jurisprudence (7.2) and case sketches (9).
• v0.5 (Jun 2026) — Restructured around the three research questions; added the measurement-as-accountability argument (Sec. 3, 7.5).
• v0.4 (May 2026) — Added the taxonomy of shutdown forms (Sec. 2) and the accountability-gap diagnosis (Sec. 6).
• v0.3 (Apr 2026) — First full mapping of the accountability toolkit.
• v0.2 (Mar 2026) — Literature and evidentiary-base survey.
• v0.1 (Feb 2026) — Initial problem statement and research questions.

Comments, corrections, and case suggestions are welcome: contact. Consistent with the standing rule across this site, the analysis is policy-analytical, the unit of study is the system and the institution, and every contested figure is flagged rather than asserted.

← All research