Tariffs are visible, quantifiable and negotiable in increments. Services barriers rarely are. A foreign architect is not stopped at the border by a duty. She is stopped by a licensing regime, a residency requirement for professional registration, or a qualification standard that recognises only domestic degrees. These barriers sit inside regulation that is formally non-discriminatory and often justified by real concerns about quality, safety and consumer protection.

This is the central institutional problem of services trade governance. The GATS was built around market access and national treatment commitments. Those disciplines catch measures that exclude or discriminate. But much of what impedes cross-border services supply is neither quota-like nor discriminatory. It is simply regulation, and regulating is what governments are supposed to do. Disciplining regulation without hollowing out regulatory autonomy requires a different kind of legal instrument, one that the original architecture only gestured toward.

Three decades of negotiation have produced three distinct answers. The first is a general mandate in Article VI:4 that the WTO never fully carried out. The second is a necessity test that would ask whether regulation is more burdensome than required, an idea that has repeatedly stalled. The third is the telecommunications Reference Paper, a sector-specific regime that succeeded because it did something the others did not. Comparing them shows what trade law can and cannot ask of domestic regulators. It also shows where the line between trade barrier and public policy is actually drawn.

Article VI:4 of the GATS instructs the Council for Trade in Services to develop any necessary disciplines. Their purpose is to ensure that qualification requirements and procedures, technical standards and licensing requirements do not become unnecessary barriers to trade. The provision sets out three aims. Such measures should be based on objective and transparent criteria. They should be no more burdensome than necessary to ensure the quality of the service. Licensing procedures should not in themselves restrict supply.

In effect, the provision was a built-in agenda. In 1994 the Uruguay Round negotiators identified a problem without solving it. Article VI:5 supplies an interim discipline, applying the VI:4 criteria to scheduled sectors. But it bites only where a measure nullifies or impairs commitments in a way that could not reasonably have been expected when those commitments were made. That timing qualifier has left the interim standard close to inert.

The first output was narrow. In 1998 the Working Party on Professional Services adopted Disciplines on Domestic Regulation in the Accountancy Sector. These included a requirement that measures be no more trade-restrictive than necessary to fulfil a legitimate objective. They were never formally integrated into the GATS. Instead they were held pending the conclusion of a broader round that never concluded. The Working Party on Domestic Regulation, set up in 1999 to extend the effort to all sectors, circulated successive chair's texts through the Doha years without reaching consensus.

The reasons are structural rather than technical. Services regulation spans legal professions, health, education, finance and utilities, each with its own regulatory philosophy. Horizontal disciplines must therefore be either abstract enough to mean little or specific enough to provoke resistance from particular sectors. Consensus decision-making made this worse. So did the single undertaking, which linked services to agriculture and industrial market access and turned domestic regulation into a bargaining chip rather than a project in its own right.

The breakthrough eventually came outside the multilateral consensus model. The Joint Statement Initiative on Services Domestic Regulation concluded in December 2021 among 67 members. It produced a Reference Paper that participants inscribe as additional commitments in their GATS schedules, with the benefits extended to all members on a most-favoured-nation basis. Its content is telling: objective and transparent criteria, timely processing of applications, reasonable fees, opportunities to comment, and non-discrimination between men and women. It is a procedural settlement. The substantive core of VI:4, whether requirements are more burdensome than necessary, was left out.

Takeaway

An unfinished mandate is not necessarily a failed one. Article VI:4 shows that procedural disciplines win consensus because they govern how governments decide, not what they decide.

The necessity test is the most ambitious idea in services regulatory governance and the most contested. Its logic comes from GATT Article XX and GATS Article XIV. There, a measure that would otherwise breach the agreement can be justified if it is necessary to achieve a listed objective. Under VI:4, however, necessity would work differently. It would not be a defence for a measure already found inconsistent. It would be a freestanding obligation applying to regulation that discriminates against no one.

That shift matters a great deal. Non-discrimination asks whether foreign suppliers are treated less favourably than domestic ones. A necessity test asks whether the regulation itself is justified given its trade effects. That moves trade law from comparing treatment to evaluating regulatory choices. It is a step toward the proportionality review of EU internal market law, in the line of Cassis de Dijon, and away from the negative integration model of the GATT tradition.

The case law shows what such review involves. In Korea – Beef, the Appellate Body described necessity as weighing and balancing three factors: the importance of the objective, the contribution of the measure, and its trade restrictiveness. It then compared the measure with reasonably available alternatives. In US – Gambling, it applied this framework under GATS Article XIV. It found the US measures provisionally necessary to protect public morals and placed on the complainant the burden of identifying a reasonably available alternative. The analysis is fact-intensive and inevitably asks adjudicators to second-guess regulators.

Positions in the Working Party did not split along conventional North-South lines. Several members with professional services export interests, India prominent among them, favoured strong necessity disciplines as a way to open markets for their professionals. The United States resisted firmly. Its concern was that a necessity test would expose sub-federal licensing, such as state bar admissions and medical boards, to international review. Many developing countries with limited regulatory capacity had a different worry: defending the rationality of their regulations before panels would fall hardest on the regulators least equipped to do it.

Whether necessity testing is feasible depends on what the test is for. As a screen for protectionism disguised as regulation, it largely duplicates national treatment analysis of de facto discrimination. As a check on regulation that is non-discriminatory but inefficient, it enters territory that trade agreements have historically left alone. The 2021 Reference Paper is silent on necessity, and so are many recent preferential agreements. That pattern suggests governments have reached a judgment. Transparency and procedural fairness are tolerable constraints. Substantive review of regulatory proportionality is not.

Takeaway

The real question in necessity testing is not whether regulation can be measured against its trade costs. It is who should have the authority to do the measuring.

The telecommunications Reference Paper remains the most successful instance of regulatory disciplines in services. It was agreed in 1996 during the extended negotiations on basic telecommunications, which concluded with the Fourth Protocol to the GATS. Members inscribed it, in full or in modified form, as additional commitments under Article XVIII. That mechanism allowed obligations beyond market access and national treatment without amending the agreement itself, an early demonstration of how flexible the scheduling architecture could be.

Its six sections reflect a coherent regulatory model:

• Competitive safeguards stop major suppliers from engaging in anti-competitive practices.
• Interconnection with major suppliers must be offered on non-discriminatory terms at cost-oriented rates.
• Universal service obligations are permitted but must be administered transparently and in a competitively neutral way.
• Licensing criteria must be publicly available.
• Regulators must be independent of suppliers.
• Scarce resources such as spectrum and numbering must be allocated objectively and without discrimination.

Mexico – Telecoms (2004) showed that these commitments had teeth. Mexico's international settlement regime required uniform rates, negotiated by the carrier with the largest share of outgoing traffic, effectively Telmex, together with a proportionate return system. The panel found that this regime failed to ensure cost-oriented interconnection and amounted to an anti-competitive practice under Section 1. The report also clarified the concept of a major supplier. It confirmed that the interconnection disciplines reached cross-border supply, not only commercial presence.

The Reference Paper succeeded for reasons specific to the sector. Telecommunications had an identifiable structural problem: incumbents controlled bottleneck facilities. It also had a broadly shared solution, drawn from domestic liberalisation in the United States, the United Kingdom and the European Union. Many signatories were already privatising incumbents and creating regulators, so the Reference Paper locked in reforms rather than imposing them. Crucially, it set out positive regulatory principles instead of subjecting regulation to open-ended scrutiny.

Attempts to replicate the model in energy, environmental, maritime and postal services have struggled. Those sectors lack comparable consensus on what the underlying regulatory problem is. The lesson is that sector-specific disciplines work when they codify a regulatory paradigm that has already converged, not when they try to create one. Preferential agreements such as the CPTPP and USMCA have built on the template in their telecommunications chapters. The Reference Paper's lasting legacy may therefore be as a drafting model for plurilateral and regional rule-making.

Takeaway

Trade rules discipline regulation most effectively when they codify a regulatory consensus that already exists. They rarely succeed in manufacturing one.

Taken together, these three experiences map the limits of what trade agreements can ask of services regulators. Procedural disciplines command broad acceptance. Substantive necessity review does not. Sector-specific pro-competitive principles succeed where regulatory convergence already exists.

For institutional design, the implication is that services regulation resists a single horizontal solution. A more productive path combines two layers. The first is transparency-based disciplines that apply across all sectors. The second is targeted regulatory principles, negotiated where a shared paradigm has matured, as may now be happening in digital trade, data governance and financial technology.

The unfinished Article VI:4 mandate is perhaps best read not as a failure but as a recognition built into the GATS. The line between trade barrier and legitimate regulation cannot be drawn once and for all. It has to be negotiated sector by sector as regulatory consensus evolves, and the institutions that govern services trade must be designed to accommodate that continuing negotiation.