Every preferential trade agreement faces the same uncomfortable question: does it actually improve welfare, or does it merely reroute commerce through politically convenient but economically inferior channels? This is not a theoretical curiosity. With over 350 regional trade agreements currently in force and notified to the WTO, the distinction between trade creation and trade diversion has become the central analytical battleground for evaluating whether the proliferating architecture of preferential liberalization strengthens or undermines the multilateral trading system.

The framework for this evaluation originates with Jacob Viner's 1950 analysis of customs unions—a contribution that overturned the then-prevailing assumption that any movement toward freer trade, even on a discriminatory basis, necessarily improved welfare. Viner demonstrated that preferential arrangements generate two opposing effects, and that the net welfare outcome depends on which dominates. This insight remains the intellectual foundation upon which every serious assessment of trade agreements rests.

Yet the original Vinerian framework, elegant as it is, was designed for a world of homogeneous goods and simple tariff preferences. Modern trade agreements encompass services liberalization, regulatory convergence, investment disciplines, intellectual property standards, and increasingly, digital trade provisions. The analytical challenge has evolved accordingly. Evaluating whether a contemporary mega-regional agreement creates or diverts trade requires methodological sophistication that Viner could scarcely have imagined—and institutional judgment about effects that no model fully captures.

Viner's foundational insight was deceptively simple: when two countries form a customs union or preferential trade agreement, the resulting tariff discrimination generates two distinct effects. Trade creation occurs when the agreement shifts production from a higher-cost domestic source to a lower-cost partner-country source. This is unambiguously welfare-improving—consumers gain access to cheaper goods, and resources are reallocated toward more efficient uses. It is the effect that proponents of any trade agreement instinctively emphasize.

Trade diversion, by contrast, occurs when preferential tariff treatment causes imports to shift from a lower-cost non-member supplier to a higher-cost partner-country supplier. The partner gains market share not because it is the most efficient producer globally, but because the tariff preference artificially tilts the competitive landscape. The importing country loses tariff revenue without gaining a genuinely cheaper source of supply. This is the effect that critics of discriminatory liberalization rightly flag as the systemic risk embedded in every preferential arrangement.

The welfare outcome of any given agreement depends on the relative magnitude of these two effects. Viner's analysis established that customs unions are more likely to be trade-creating when member countries' economies are large, diversified, and competitive with each other rather than complementary—and when pre-agreement tariff levels against non-members are high. In such configurations, the scope for displacing inefficient domestic production is substantial, while the risk of displacing efficient third-country imports is comparatively limited.

Subsequent refinements by Meade, Lipsey, and others introduced consumption effects and terms-of-trade considerations that Viner's original production-focused analysis had neglected. Lipsey demonstrated that even in cases of net trade diversion on the production side, consumer gains from lower prices on intra-bloc trade could generate positive welfare outcomes. Kemp and Wan later showed that a customs union could always be designed to improve welfare if the common external tariff was set to hold imports from non-members constant—though this theoretical possibility has rarely governed actual negotiating practice.

What the Vinerian framework ultimately provides is not a prediction but a diagnostic structure. It forces analysts and policymakers to ask the right question: is this agreement displacing genuinely inefficient production, or is it merely redirecting trade flows in ways that serve political rather than economic logic? The answer is almost never purely one or the other, which is precisely why rigorous empirical estimation matters.

Takeaway

A preferential trade agreement is not inherently good or bad for welfare—its value depends entirely on whether it displaces inefficient domestic production (trade creation) or redirects imports away from efficient outside suppliers (trade diversion). The analytical discipline lies in measuring which effect dominates.

Translating the Vinerian framework from theory to measurement requires econometric tools capable of isolating the causal effects of preferential agreements from the countless other factors that influence bilateral trade flows. The gravity model of trade has emerged as the workhorse methodology for this purpose. In its modern specification, the gravity model relates bilateral trade volumes to economic mass (GDP), geographic distance, and a set of trade cost variables—among which preferential trade agreement membership features as a key explanatory dummy or set of dummies.

The methodological evolution of gravity-based estimation has been substantial. Early naive approaches suffered from omitted variable bias, endogeneity, and the failure to account for multilateral resistance terms—the insight, formalized by Anderson and van Wincoop, that bilateral trade depends not only on bilateral trade costs but on trade costs relative to all other partners. Modern gravity specifications employ structural foundations, country-pair fixed effects, and time-varying exporter and importer fixed effects to address these concerns. Baier and Bergstrand's influential work demonstrated that properly accounting for endogeneity roughly doubles the estimated trade-creation effects of free trade agreements.

Distinguishing trade creation from trade diversion within the gravity framework requires careful specification. Analysts typically include separate variables for intra-bloc trade (to capture creation effects) and for trade between members and non-members (to capture potential diversion). A well-identified model shows whether increased intra-bloc trade comes at the expense of imports from third parties. The results are frequently nuanced: many agreements show statistically significant trade creation with modest or insignificant diversion, though the magnitudes vary enormously across agreements, sectors, and time horizons.

Computable general equilibrium (CGE) models offer a complementary approach, particularly for ex ante evaluation of proposed agreements. CGE frameworks model entire economies with multiple sectors, factors of production, and trading partners, allowing analysts to simulate the effects of tariff changes, rules of origin, and non-tariff barrier reductions. The GTAP (Global Trade Analysis Project) framework has become the standard platform for such exercises. CGE models excel at capturing inter-sectoral resource reallocation and terms-of-trade effects that reduced-form gravity estimates cannot isolate.

Neither approach is without significant limitations. Gravity models rely on historical variation and may not capture structural breaks or anticipation effects associated with agreement negotiations. CGE models depend heavily on assumed elasticities, market structures, and closure rules—choices that can materially alter welfare estimates. The most credible empirical assessments triangulate across methods, combining the statistical rigor of gravity estimation with the structural coherence of general equilibrium analysis, while remaining transparent about the uncertainty inherent in both.

Takeaway

No single empirical method delivers a definitive verdict on whether a trade agreement creates or diverts trade. The most reliable assessments combine gravity model estimation with general equilibrium simulation, and they are always honest about the sensitivity of results to specification choices and modeling assumptions.

The original trade creation-diversion framework was built for a world where trade agreements principally reduced tariffs on goods crossing borders. Contemporary agreements operate on fundamentally different terrain. The most consequential provisions in mega-regionals like the CPTPP, the EU-Japan EPA, or the RCEP involve services market access, regulatory coherence, investment disciplines, and increasingly, digital trade rules. Applying the Vinerian lens to these dimensions requires conceptual extension that the profession is still actively developing.

Services liberalization presents particular analytical challenges. Trade in services is governed primarily by regulatory barriers rather than tariffs—licensing requirements, recognition of professional qualifications, restrictions on foreign equity participation, data localization mandates. When a preferential agreement reduces these barriers on a discriminatory basis, the potential for services trade diversion exists but is harder to quantify. A mutual recognition agreement for professional qualifications between two partners, for example, may effectively exclude third-country professionals without any tariff equivalent to measure. The empirical literature on services trade creation and diversion remains thinner and less settled than its goods-trade counterpart.

Regulatory harmonization and mutual recognition provisions raise a distinct set of evaluation challenges. When agreement partners align product standards, testing procedures, or conformity assessment mechanisms, the trade-facilitation benefits can be substantial. But if the harmonized standard diverges from international norms—or if mutual recognition is extended only to partners—the result can function as a non-tariff barrier against non-members. Richard Baldwin's concept of the spaghetti bowl of overlapping rules of origin captures one dimension of this problem, but the regulatory dimension adds layers of complexity that standard trade creation-diversion accounting struggles to accommodate.

Perhaps the most significant analytical frontier involves dynamic effects that the static Vinerian framework deliberately excluded. Modern agreements may generate welfare gains through investment attraction, technology transfer, pro-competitive market restructuring, and institutional upgrading that dwarf the static allocation effects. The EU's deep integration agreements with Central and Eastern European countries prior to enlargement are frequently cited as cases where dynamic transformation effects overwhelmed whatever static trade diversion occurred. Yet dynamic effects are notoriously difficult to attribute causally to specific agreement provisions rather than to concurrent policy reforms or global economic trends.

The institutional implication is clear: evaluating trade agreements solely through the trade creation-diversion binary is necessary but no longer sufficient. A comprehensive assessment framework must incorporate services and regulatory effects, dynamic productivity impacts, and the systemic consequences for the multilateral architecture. The WTO's Committee on Regional Trade Agreements has long struggled with this challenge, lacking both the methodology and the political mandate to deliver authoritative assessments. Until evaluation frameworks catch up with the ambition of modern agreements, the governance gap will persist—and with it, the risk that preferential architectures serve political convenience over genuine economic welfare.

Takeaway

The trade creation-diversion framework remains indispensable, but applying it only to tariffs on goods captures a shrinking share of what modern trade agreements actually do. The hardest and most consequential evaluation challenges now lie in services, regulation, and dynamic effects—precisely where our measurement tools are weakest.

Viner gave the profession its most enduring diagnostic question: does a preferential arrangement create trade or divert it? Seven decades of theoretical refinement and empirical innovation have made the question easier to investigate but no easier to answer definitively. Every major agreement generates both effects simultaneously, and the net welfare outcome depends on magnitudes that are sensitive to method, data, and specification.

The deeper institutional challenge is that evaluation capacity has not kept pace with the proliferation and deepening of preferential agreements. As provisions extend into services, regulation, digital trade, and investment, the analytical frameworks designed for tariff preferences on goods require fundamental extension. The profession is building these tools, but the gap between what agreements now do and what we can rigorously measure remains substantial.

For trade lawyers and policymakers, the imperative is to design agreements with trade creation in mind—broad sectoral coverage, low and liberal rules of origin, external tariffs that do not rise—while investing seriously in the institutional capacity to evaluate what these agreements actually deliver. Preferential liberalization is not going away. The question is whether we govern it with analytical rigor or political faith.