Trade Agreements and Prosperity: An Evidence-Led Framework

Executive judgment

Trade agreements are legal infrastructure for economic coordination. At their best, they replace discretionary border power with published commitments: tariff schedules, non-discrimination rules, customs procedures, disciplines on domestic regulation, and mechanisms through which states can challenge non-compliance. This predictability changes commercial behaviour. It allows firms to price long-term investments, reorganise supply chains, enter unfamiliar markets and contract across borders with less political risk.

The economic case is consequently broader than the textbook proposition that countries gain by specialising according to comparative advantage. Modern production depends on imported inputs, cross-border services, data, finance, logistics and knowledge. A border restriction on one stage of production is therefore frequently a tax on another domestic industry. Conversely, a well-designed agreement can reduce several layers of cost at once: the tariff, the delay, the duplicated conformity assessment, the opaque licensing decision and the uncertainty that rules may change after capital has been committed.

Yet the label “free trade agreement” is not proof of economic quality. A preferential agreement departs from the World Trade Organization’s most-favoured-nation principle by giving selected partners better access than others. It may create trade by replacing inefficient domestic production, or divert trade by replacing a more efficient non-member supplier with a less efficient member. It may simplify regulation, or convert the incumbent members’ regulatory model into an expensive condition of entry. It may expand development options, or narrow them through rules whose implementation costs fall most heavily on states and firms with the least administrative capacity.

Zwarte Peper’s position: open trade should be the presumption, not because every existing agreement is benign, but because voluntary exchange, competitive entry and access to global knowledge are powerful engines of human progress. The appropriate response to concentrated adjustment costs is targeted domestic reform—not permanent border protection that raises costs for everyone and restricts the development choices of trading partners.

A high-quality agreement should therefore pass three tests. It should be economically additional, creating value beyond trade that would have occurred anyway. It should be legally legible, replacing discretion with rules that firms and citizens can understand, invoke and contest. And it should be development-compatible, enabling participation by economies and businesses with different starting capacities rather than reserving nominal market access for those already able to bear the compliance burden.

Method and scope

This analysis combines treaty law, empirical trade economics and implementation evidence. It distinguishes four questions that public debate often collapses into one:

1. Legality: Is the preferential arrangement permitted by WTO law and consistent with the parties’ other obligations?
2. Economic effect: Does it create trade, productivity and welfare gains after accounting for diversion, compliance costs and distributional effects?
3. Institutional quality: Are the commitments clear, enforceable and administered through due process?
4. Execution: Can firms actually qualify for, understand and use the negotiated preferences?

These questions must remain separate. WTO compatibility is a minimum legal condition, not a cost-benefit conclusion. A positive modelled change in aggregate income does not establish that every sector or region benefits. A commercially attractive tariff schedule is of limited value when restrictive origin rules or unrecognised product tests prevent firms from claiming it. Equally, a difficult transition for one industry does not demonstrate that society gains from maintaining a permanent tariff at the expense of downstream producers and consumers.

1. What a trade agreement legally does

The WTO’s central organising principle is non-discrimination. Under the General Agreement on Tariffs and Trade, a tariff advantage given to one member is ordinarily extended to all WTO members. Preferential agreements are authorised exceptions. Goods agreements principally rely on Article XXIV of GATT 1994; services integration is governed by Article V of the General Agreement on Trade in Services; and certain arrangements among developing economies may rely on the 1979 Enabling Clause.¹ The WTO database recorded 386 notified regional trade agreements in force in July 2026.²

This legal character matters. An agreement is not a general declaration that trade is free. It is a negotiated allocation of rights, exceptions, transition periods and enforcement options. Its commercial consequences are found in schedules and annexes as much as in headline commitments. Counsel assessing an agreement must read at least six layers together:

The distinction between a customs union and a free-trade area illustrates the operational stakes. A customs union applies a common external tariff; a free-trade area allows each party to retain its own external tariff. The latter therefore needs preferential rules of origin to prevent goods from entering through the lowest-tariff member and circulating freely. Those rules are necessary to the preference, but their complexity can consume its value. A five-percentage-point tariff preference is commercially irrelevant if certification, supplier segregation and record-keeping cost six.

Nor are modern agreements confined to border measures. The World Bank’s Deep Trade Agreements framework identifies provisions across services, investment, competition, subsidies, public procurement, intellectual property, labour and environmental policy.³ These fields contain genuine cross-border spillovers and can benefit from cooperation. They also reach deeply into domestic governance. The legitimacy of “deep” integration therefore depends not only on whether a rule promotes commerce, but on whether it addresses a demonstrable transnational problem, preserves proportionate regulatory autonomy and can be implemented by all parties.

The institutional case for agreed constraints is not merely theoretical. Recent research using global tariff choices to infer governments’ implicit welfare weights finds that the formal and informal trading system causes states to internalise a substantial—although incomplete—share of the effects their policies impose on other countries.¹⁹ Reciprocity is therefore valuable not as managed retaliation, but as a mechanism through which governments account for costs that unilateral political processes tend to ignore.

2. The economic case: from comparative advantage to institutional credibility

Specialisation, scale and consumer welfare

The foundational logic remains robust. When relative productivity differs, exchange allows resources to move toward activities in which they create greater value. Consumers obtain more variety or lower prices; exporters gain access to larger demand; and competition pressures domestic firms to improve. The gain does not require one country to be absolutely more productive than another. It arises from differences in opportunity cost.

This account is sometimes dismissed as too abstract because real economies contain unemployment, market power, logistics constraints, taxes and political risk. Those complications change the size and distribution of gains; they do not eliminate the cost of forcing production into higher-cost domestic uses. A tariff can preserve the visible output of the protected industry while obscuring losses borne by consumers, downstream manufacturers and exporters facing retaliation. The correct empirical analysis therefore follows value through the entire production network rather than counting jobs at the protected factory gate.

Imported inputs and global value chains

Nearly half of world trade is associated with global value chains, in which goods and services cross borders at multiple stages of production. The World Bank’s World Development Report 2020 concludes that these networks supported rapid growth and income convergence, while emphasising that infrastructure, predictable policy and complementary reforms determine whether participation produces durable development.⁴ The relevant unit is no longer simply the finished import. It is the domestic value that firms can add when they obtain competitively priced machinery, components, software, design, finance and logistics.

This creates an important policy inversion. Measures described as protecting national production may weaken it when they raise the price of essential inputs. The more fragmented the value chain, the more often a border cost is compounded. Trade facilitation, interoperable documentation and predictable customs valuation can therefore matter more than the remaining headline tariff.

Investment, knowledge and credible commitment

Trade agreements also affect the fixed costs of market entry. A firm deciding where to build a plant or locate a services hub must estimate whether inputs can cross borders, whether licences will be granted on equal terms, whether data and staff can move, and whether an abrupt policy reversal will strand the investment. Binding commitments can lower that risk premium. They also create channels through which regulators exchange information and firms can raise recurring implementation failures.

Knowledge transfer is not a one-directional process from high-income to lower-income economies. Production networks distribute engineering methods, managerial practice, digital capabilities and specialised services among firms in every region. India’s role in information technology and business services, East and Southeast Asia’s manufacturing ecosystems, Africa’s rapidly evolving digital and payments markets, and Latin America’s agricultural and clean-energy capabilities all illustrate that expertise is globally dispersed. A trade regime that treats some economies principally as markets for finished exports misses the larger opportunity: mutual access to ideas, talent, suppliers and experimentation.

What the empirical literature supports—and what it does not

Estimating agreement effects is difficult because countries that expect to trade more are also more likely to sign agreements. Baier and Bergstrand’s influential panel analysis addressed this endogeneity problem and found that free-trade agreements materially increase members’ trade over time.⁵ More granular results show why effects differ. Caliendo and Parro’s sectoral model of the North American Free Trade Agreement estimated welfare gains from tariff reductions of 1.31 per cent for Mexico and 0.08 per cent for the United States, alongside a small estimated loss for Canada; the same policy produced different outcomes because production structures, initial tariffs and input linkages differed.⁶

Deeper agreements can create value beyond tariffs. World Bank research estimated that, within its sample and specification, country pairs signing a deep agreement experienced a 44 per cent increase in trade. Importantly, non-discriminatory provisions concerning competition, subsidies and standards could also benefit non-members, whereas preferential tariff margins still produced trade diversion.⁷ This is not a warrant to maximise the number of treaty chapters. It is evidence that the content and implementation of commitments matter more than the agreement’s label.

3. Development and distribution: openness is not an adjustment policy

The strongest global evidence does not support the proposition that retreat from trade will make poorer economies more inclusive. The WTO’s World Trade Report 2024 associates stronger trade participation with faster convergence among low- and middle-income economies, while also finding that gains have been uneven across countries, regions and people.⁸ “Trade works” and “trade alone is sufficient” are different claims. Infrastructure, education, competition, finance, labour mobility and social protection influence who can seize new opportunities.

The distributional challenge is real. The China-shock literature found that US labour-market adjustment to a large import shock was geographically concentrated and remarkably slow. Wages and participation remained depressed for years in more exposed local markets.⁹ That evidence corrected an overly frictionless account in which displaced workers immediately move to expanding industries and regions. It does not establish that the appropriate remedy is a permanent economy-wide tariff. It establishes that governments should identify exposure before liberalisation, make benefits portable across regions and employers, and build adjustment institutions that operate at the speed of commercial change.

Protectionism frequently fails this test of fit. It taxes a broad population to deliver an opaque benefit to selected producers, often without conditioning support on productivity, innovation or transition. Research on the United States’ 2018 tariffs found essentially complete pass-through into the domestic prices of imported goods and estimated a reduction in real income of USD 1.4 billion per month by the end of that year.¹⁰ Retaliation also reduces export access. The protected sector may gain, but the instrument imposes costs unrelated to whether households or regions actually need support.

The development asymmetry lies increasingly behind the tariff

For many exporters, the principal barrier is no longer the customs duty. It is the capacity to understand, demonstrate and finance compliance with technical, health, environmental, security and documentation requirements. These measures can serve legitimate public purposes. They become economically exclusionary when they are duplicative, opaque, more trade-restrictive than necessary or designed without regard to feasible routes for foreign compliance.

UN Trade and Development reported in May 2026 that non-tariff measures exceeded tariffs as a source of trade cost in 88 per cent of country cases examined. It estimated that least-developed economies forfeit around 10 per cent of exports to G20 markets because they cannot meet applicable requirements, and that greater transparency could reduce NTM-related trade costs by roughly 19 per cent.¹¹ These figures reframe development policy. A nominal zero tariff does not create meaningful access when the exporter lacks a recognised testing facility, cannot obtain an explanation of rejection or must repeat certification in every destination market.

A development-compatible agreement should address this asymmetry directly. It should publish requirements in usable formats, provide enquiry points and advance rulings, recognise equivalent outcomes where appropriate, permit accredited testing outside the importing country, and connect implementation periods to verified capacity. Technical assistance should not be a diplomatic appendix detached from market-access obligations; it should be part of the agreement’s operating model.

Rules-based does not mean one-size-fits-all

Predictable rules protect weaker parties from raw bargaining power, but rules can also encode the preferences of those who had greater negotiating capacity. The answer is not to abandon a rules-based order. It is to improve how rules are justified and made. Provisions should respond to evidence of a cross-border problem, use the least restrictive effective instrument, allow functionally equivalent compliance and include review mechanisms when technology or economic conditions change.

This distinction is especially important for developing economies. Time-bound sequencing, capacity-linked commitments and narrowly framed safeguards can support transition. Indefinite protection, discretionary licensing and local-content mandates often entrench incumbents, raise input costs and delay competitiveness. Policy space should enable governments to solve genuine public problems—not preserve unreviewable power to allocate economic opportunity.

4. How trade agreements fail

Trade diversion disguised as liberalisation

A preferential tariff can redirect imports toward a member even when a non-member is the more efficient supplier. Consumers may pay less than under the original tariff but more than they would under non-discriminatory liberalisation. Governments should therefore model trade creation and diversion separately, test the agreement against a realistic multilateral or unilateral-liberalisation counterfactual, and examine effects on input sourcing from non-members.

Rules of origin that firms cannot use

Origin rules determine the economic nationality of a product assembled from inputs crossing several borders. WTO materials recognise that the diversity and complexity of preferential rules increase trading costs and may reduce utilisation of preferences.¹² The recurring mistake is to celebrate tariff elimination without measuring preference use. Negotiators should publish utilisation rates by tariff line, explain low-use sectors, allow reasonable cumulation, simplify documentary proof and design rules around actual supply chains rather than an imagined vertically integrated factory.

Regulatory cooperation that becomes regulatory export

Shared standards can reduce duplication and support health, safety and environmental outcomes. However, harmonisation around one party’s process may impose high fixed costs without improving the regulated outcome. This disadvantages smaller firms and jurisdictions whose institutions achieve equivalent protection differently. Mutual recognition, equivalence, international standards and outcomes-based requirements should be evaluated before insisting on identical rules.

Exceptions without process

Every serious agreement preserves room for public health, security, environmental protection and emergency safeguards. The issue is not whether exceptions exist, but how they are governed. Vague necessity claims, secret evidence and indefinite measures turn exceptions into escape routes from negotiated commitments. Even sensitive action should ordinarily be published, reasoned, targeted, time-limited and reviewable. Where disclosure would create a specific security risk, an independent process can test the basis without placing all evidence in the public domain.

Enforcement that is too remote from the commercial problem

State-to-state dispute settlement is indispensable, but formal litigation is slow and politically selective. Many losses arise from recurring administrative friction: a customs classification that changes by port, an unexplained delay, a regulator that will not recognise an accredited report. Agreements need implementation committees with deadlines, transparent agendas, business-accessible problem reporting and escalation paths. The measure of institutional quality is not the number of committees created; it is the proportion of problems resolved before firms abandon the market.

Domestic policy that socialises transition and privatises access

Governments sometimes sign an agreement while preserving domestic barriers to entry, weak competition policy and privileges for connected firms. Imports then become cheaper without ensuring that savings reach consumers; export licences and finance remain concentrated; adjustment programmes arrive after closures rather than before exposure. Trade policy cannot substitute for domestic institutional quality. It can, however, create transparent benchmarks against which such failures become harder to conceal.

5. Facilitation, services and AI: the next development frontier

As tariffs fall, the speed and legibility of administration become central. The WTO estimated that full implementation of the Trade Facilitation Agreement could reduce trade costs by an average of 14.3 per cent, with the largest gains in poorer countries.¹³ OECD indicators suggest that reforms since the TFA’s conclusion have already reduced trade costs by almost 5 per cent on average, with substantial further potential from better border processes.¹⁴

These gains come from institutional engineering: online publication, advance rulings, risk-based inspection, pre-arrival processing, appeal rights, coordinated border agencies and single windows. Automation is particularly valuable when it reduces opportunities for selective delay and generates an auditable trail. But digitising a bad process does not make it good. A system that requires the same unnecessary approvals through a portal has automated friction rather than removed it.

Services are no longer secondary to goods

Logistics, finance, telecommunications, engineering, design and software are both traded outputs and inputs into almost every physical product. WTO and World Bank analysis reports that services account for roughly half of global trade measured in value-added terms, and that developing economies’ share of world services exports rose from 23.5 per cent in 2005 to around 35 per cent in 2022.¹⁵ This creates routes to development that do not replicate the historical manufacturing sequence of today’s rich economies.

Services commitments must therefore be operational. Market access on paper may be neutralised by nationality conditions, local establishment, unrecognised qualifications, restrictions on payments or uncertain data rules. Negotiators should map the end-to-end delivery chain for priority services and test whether each required input—people, data, capital, credentials and digital authentication—can legally move.

Trade rules should enable rapid, accountable automation

AI can lower the fixed cost of translation, product classification, standards discovery, contracting and regulatory compliance. It can help smaller firms navigate markets previously accessible only to multinationals with large legal and customs teams. WTO modelling for the World Trade Report 2025 projected that global trade could be 34–37 per cent higher by 2040 across scenarios involving different degrees of technological and policy convergence; these are conditional model results, not forecasts.¹⁶

The main risk is a new form of exclusion. If AI-enabling compute, chips, cloud services, data and technical knowledge are fragmented into geopolitical blocs, lower-capacity economies may face both higher input costs and reduced access to markets. IMF modelling similarly finds that geoeconomic fragmentation tends to impose proportionally larger income losses on emerging and developing economies.¹⁷

The appropriate regulatory default is permission to innovate, paired with targeted accountability for demonstrable risks. Trade chapters should support electronic documents and signatures, interoperable identity and payment systems, cross-border data use subject to proportionate safeguards, and technology-neutral regulation. Blanket localisation, duplicative AI licensing and nationality-based restrictions should require a specific, evidenced justification. Where risks differ by use—medical diagnosis, credit, border control, marketing—the law should regulate the risk of the application rather than treating every computational model as equally hazardous.

International cooperation is valuable where it reduces contradictory compliance obligations. Joint work by the IMF, OECD, UNCTAD, World Bank and WTO identifies infrastructure, skills, capacity support and a workable regulatory environment as connected conditions for development through digital trade.¹⁸ Cooperation should not, however, freeze one region’s early regulatory choices into a global template before evidence accumulates. Sandboxes, mutual recognition of assurance methods, common incident vocabularies and machine-readable regulatory information can provide coordination while preserving experimentation.

6. Seven design principles for a prosperity-enhancing agreement

1. Start with the commercial constraint, not the negotiating template

Map where trade is actually lost: tariff peaks, border delay, licensing, standards, payments, transport, data, visas or lack of information. The agreement should concentrate political and administrative capital on constraints with measurable value.

2. Prefer broad, simple liberalisation over selective rents

Wide product coverage, short negative lists, transparent reservations and simple origin rules reduce lobbying over exclusions. Where transition is necessary, the staging schedule should be time-bound and tied to a published adjustment plan.

3. Treat implementation capacity as part of legal design

Different starting capacities are facts, not moral failures. Obligations should specify the institution, data, technology and skills required for implementation. Capacity support should have milestones and ownership, while avoiding indefinite exemptions that deny firms the benefit of reform.

4. Regulate outcomes proportionately

Health, safety, privacy and environmental goals are legitimate. Requirements should be evidence-based, non-discriminatory and no more trade-restrictive than necessary to achieve the stated outcome. Equivalence and mutual recognition should be available when another system delivers comparable protection.

5. Build due process into administration

Publication, notice, reasons, response periods, independent review and reliable contact points convert treaty language into usable rights. Digital systems should create auditability and consistent treatment rather than merely move discretion online.

6. Separate adjustment support from permanent protection

Identify exposed workers, firms and places through data; make training and income support available before the shock; remove barriers to new firm formation, housing and mobility; and evaluate outcomes. Do not force all consumers and downstream producers to finance an unmeasured industrial policy through indefinite tariffs.

7. Measure use, not signatures

Post-entry-into-force dashboards should publish preference utilisation, customs release times, services approvals, recognition decisions, SME participation, disputes and distributional outcomes. Review clauses should trigger correction when the agreement is unused or its burdens fall systematically on lower-capacity firms and economies.

7. A decision scorecard for counsel, executives and policymakers

The following scorecard converts the analysis into an initial diligence tool. It is not a substitute for product- and sector-specific legal advice.

For businesses, the output should be a landed-cost and market-entry model by product or service, not a general view that an agreement is “positive”. For governments, it should be a distributional and institutional implementation plan, not a forecast of aggregate exports alone. For counsel, it should be a traceable chain from treaty commitment to domestic law, administrative procedure, evidence requirement and remedy.

Conclusion: open markets require better institutions

The enduring case for open trade is not that markets adjust without friction or that every negotiated text advances freedom. It is that voluntary exchange across borders expands the set of people, ideas, technologies and opportunities available to society. Protection suppresses those possibilities while concealing its costs in higher prices, weaker competition and foregone development.

A rules-based trading order remains essential precisely because economic power is unequal. Smaller states and firms benefit when access does not depend solely on political leverage. But the quality of that order depends on whose costs are visible, whose knowledge informs the rules and whether compliance is practically available outside the world’s richest institutions. Global perspective is therefore not an ornamental commitment to diversity. It is necessary for accurate policy design.

The next generation of agreements should be judged by execution: whether they reduce the combined cost of tariffs, administration and uncertainty; whether they allow services, technology and knowledge to circulate; whether legitimate regulation is proportionate and contestable; and whether developing economies can use the access ostensibly granted to them. Agreements that meet those tests can support prosperity. Those that fail them should be reformed, however ambitious their titles or extensive their chapters.