Trade Policy Analysis: How the Structural Gravity Model Reveals Hidden Economic Logic

For decades, trade policy analysis relied on a seemingly straightforward question: “What happens to trade flows when a tariff changes?” The answer, however, has never been simple. Traditional tariff-focused approaches—computing trade creation and diversion using partial equilibrium models or simple before-and-after comparisons—consistently fail to capture the complex web of trade costs that shape global commerce. Logistics bottlenecks, regulatory divergence, geopolitical frictions, and even the indirect effects of a trade deal between third parties can dramatically alter bilateral trade outcomes.

In response, a growing number of economists and policymakers have turned to the structural gravity model as the analytical backbone for modern trade policy analysis. This framework, grounded in microeconomic theory, provides a unified way to disentangle the multiple forces driving trade flows. A landmark reference in this field is the PDF guide by Yotov and colleagues (available at yotoyotov.com/files/book.pdf), which distills advanced techniques for applying the model to real-world questions—from trade wars to regional integration. This article unpacks what makes the structural gravity model indispensable for analysts who need to move beyond headline trade statistics and understand the deeper economic logic at work.

[IMAGE: Diagram comparing a simplistic tariff-line chart (showing only tariff rates and import volumes) with a multi-node trade flow network (nodes as countries, edges with varying thickness, arrows showing direction). Caption: “Tariff-focused analysis vs. the network perspective of structural gravity.”]

---

The Structural Gravity Model: Core Economic Logic

At first glance, the name “gravity model” might seem like a physics metaphor stretched too far. Yet the intuition is remarkably robust: just as Newton’s law states that gravitational force between two objects is proportional to their masses and inversely proportional to the distance between them, trade between two economies is proportional to their economic sizes (typically GDP) and inversely proportional to the trade costs separating them.

The formula is deceptively simple:

\[

X_{ij} = \frac{Y_i E_j}{Y} \left( \frac{t_{ij}}{\Pi_i P_j} \right)^{1-\sigma}

\]

where \(X_{ij}\) is exports from country \(i\) to country \(j\), \(Y_i\) and \(E_j\) are the economic sizes (output and expenditure), \(t_{ij}\) represents bilateral trade costs (tariffs, distance, non-tariff barriers), and \(\Pi_i\) and \(P_j\) are the multilateral resistance terms—the model’s crucial innovation.

Multilateral resistance captures the idea that trade between two countries depends not only on their own barriers but also on the barriers each faces with all other trading partners. If country A raises tariffs on imports from country B, B’s exporters may redirect goods to country C, lowering prices there and making C more competitive relative to B. This substitution effect, often invisible in partial equilibrium analysis, can explain why a trade agreement between two large economies might reduce trade between non-members—a phenomenon that traditional tariff-line models routinely miss.

The structural gravity model, by incorporating multilateral resistance, reveals the hidden logic of international trade networks. It shows that trade policy is never a bilateral affair; every barrier or liberalization ripples through the entire system.

[IMAGE: Equation of the structural gravity model with annotated variables (trade, GDP, distance, tariffs) plus a scatter plot showing a downward-sloping relationship between trade intensity and distance, with countries labeled. Caption: “The gravity relationship: trade decays with distance, but multilateral resistance modifies the slope.”]

---

Applying the Model to Real-World Trade Policy Scenarios

The real power of the structural gravity model lies not in abstract theory but in its ability to answer concrete policy questions. Here are three high-stakes applications:

Tariff Impacts: Direct vs. Indirect Effects

When a country raises tariffs, the direct effect is a decline in imports from the targeted partner. But the model shows that this is only part of the story. Higher tariffs on one source raise the average trade costs for the importing country, shifting its multilateral resistance terms. As a result, imports from all other partners may increase (trade diversion), and the exporting country’s competitiveness in third markets may erode. The model allows analysts to separate these direct and indirect effects, providing a complete picture of how a tariff change alters global market shares.

Trade Agreements: Ex-Ante Estimation of Creation and Diversion

Regional trade agreements such as RCEP, USMCA, or the African Continental Free Trade Area are designed to reduce barriers among members. Using the structural gravity framework, analysts can estimate ex-ante—before implementation—how much trade will be created between members and how much will be diverted away from non-members. By simulating counterfactual scenarios (e.g., removing tariffs within the bloc while keeping external barriers constant), the model generates welfare estimates that account for changes in consumer prices, producer surplus, and government revenue. This is far more informative than the rudimentary “trade creation minus trade diversion” arithmetic of old textbooks.

Supply Chain Resilience: Propagation of Trade Costs

Supply chain disruptions—whether from sanctions, port closures, or pandemics—do not stay in one place. A two-week delay at a major container port increases shipping costs for all goods transiting that route, which then feeds into production costs for factories downstream. The structural gravity model can quantify how a shock to trade costs in one node propagates through the network. For instance, the model has been used to estimate how U.S. tariffs on Chinese intermediate goods raised costs for American manufacturers that rely on Chinese components, while simultaneously opening opportunities for suppliers in Vietnam and Mexico. This capability makes the model a key tool for policymakers concerned with supply chain resilience and diversification.

[IMAGE: World map with highlighted trade blocs (e.g., RCEP, USMCA, EU) and arrows showing trade flow redirection away from a central region. Caption: “Trade agreements redirect flows. Structural gravity quantifies the winners and losers.”]

---

Advanced Techniques from the Guide: Data, Estimation, and Interpretation

The Yotov guide is not just a theoretical exposition; it is a practical manual for analysts who want to implement the structural gravity model with real data. Key methodological choices can make or break the credibility of the results.

Panel Data and Fixed Effects

The recommended approach uses panel data (multiple years for each bilateral pair) and a rich set of fixed effects. Exporter-time and importer-time fixed effects absorb all country-specific shocks (GDP changes, exchange rate fluctuations, institutional reforms) and automatically control for multilateral resistance terms. Pair fixed effects (exporter-importer) capture time-invariant bilateral frictions such as distance, common language, or colonial ties. This structure allows the researcher to isolate the impact of time-varying trade policy variables (tariffs, non-tariff measures, trade agreements) without omitted variable bias.

PPML Estimator vs. OLS

One of the guide’s central recommendations is to use the Poisson Pseudo-Maximum Likelihood (PPML) estimator instead of ordinary least squares (OLS). Trade data contain many zeros—country pairs that do not trade at all. OLS on log-linearized gravity (where zero is undefined) either discards those pairs or adds arbitrary small numbers, biasing results. PPML handles zeros naturally and also corrects for heteroskedasticity. Empirical studies consistently show that PPML produces more reliable estimates of trade cost elasticities.

Endogeneity Concerns

A common criticism of gravity models is reverse causality: does a trade agreement cause more trade, or do countries that already trade a lot tend to sign agreements? The guide addresses this with techniques such as using lagged variables, instrumental variables (e.g., predicted trade costs based on historical political relations), or dynamic panel methods. Understanding and addressing endogeneity is essential for making causal claims about trade policy.

Interpreting Coefficients

Once the model is estimated, the coefficients on tariffs or trade agreement dummies can be converted into meaningful economic quantities. For example, a coefficient of -5 on a tariff variable implies that a 1% increase in the ad valorem tariff reduces bilateral trade by approximately 5% (given the elasticity parameter \(\sigma-1\)). More advanced users can translate these trade flow changes into welfare effects (equivalent variation) using the model’s structure. The guide provides formulas and Stata code for these calculations.

[IMAGE: Flowchart of estimation steps: (1) collect panel data with trade, GDP, distance, tariffs, RTAs; (2) specify fixed effects (exporter-time, importer-time, pair); (3) run PPML regression; (4) extract coefficients and compute counterfactuals. Caption: “From data to policy insight: the gravity estimation workflow.”]

---

Case Studies: What the Model Reveals That Simple Analysis Misses

The US-China Trade War

Between 2018 and 2020, the United States imposed tariffs on over $300 billion of Chinese imports. A naive before-and-after comparison would show U.S. imports from China falling sharply—which they did. But what happened to the trade flows that China redirected toward other markets? Gravity model analysis by several research teams (including Yotov’s own work) revealed that Chinese exports to third countries like Vietnam, Mexico, and Germany increased significantly, partly offsetting the loss from the U.S. market. Moreover, U.S. importers shifted sourcing to alternative suppliers, but many of those suppliers themselves used Chinese inputs. The structural gravity model captured these supply chain reconfigurations, showing that the net welfare cost to the U.S. was larger than tariff revenue gains—a conclusion impossible to reach with tariff-line data alone.

Brexit and UK–EU Trade

The UK’s departure from the EU introduced new non-tariff barriers (customs checks, regulatory divergence) that went far beyond tariff changes. Traditional models predicted a 5–10% fall in UK–EU trade. Gravity models that explicitly account for the cost of non-tariff measures (using data on border delays, sanitary standards, and rules of origin) estimated the drop at closer to 15–20%. They also found that the UK’s trade with non-EU countries increased modestly, but not enough to compensate. The result informed the UK government’s recalibration of its trade strategy, including negotiations for new bilateral agreements with Australia and New Zealand.

Regional Integration in Africa

The African Continental Free Trade Area (AfCFTA) aims to create a single market for goods and services. Gravity models have been used to simulate its potential impact under different scenarios of tariff elimination and non-tariff barrier reduction. A striking finding: full tariff removal alone would increase intra-African trade by only 10–15%, but combining it with a 50% reduction in non-tariff barriers (such as cumbersome customs procedures) could boost trade by over 40%. The model also highlights that the biggest gains will go to countries that invest in trade facilitation, while landlocked countries may benefit less unless transit infrastructure improves. These insights guide policymakers on where to focus complementary reforms.

[IMAGE: Two-panel chart: left panel shows US imports from China and from Vietnam over 2017-2021, with a sharp drop in China line and a rise in Vietnam line; right panel shows a gravity-model simulation of trade diversion effects under three tariff scenarios. Caption: “The US-China trade war: what actually happened vs. what the gravity model predicted.”]

---

Conclusion: Why the Structural Gravity Model Is the New Standard

The structural gravity model is not just another regression technique—it is a framework that aligns economic theory with empirical rigor. For analysts grappling with trade policy analysis in a world of multilateral resistance, shifting supply chains, and increasingly complex non-tariff barriers, it offers the only reliable way to untangle cause and effect. The Yotov guide provides a complete toolkit, from data preparation to counterfactual welfare analysis, empowering policymakers and strategists to move beyond simplistic headlines.

As global trade becomes more interconnected and more volatile, the demand for rigorous, theory-driven analysis will only grow. The structural gravity model, once a niche academic tool, is now an essential part of any serious trade analyst’s arsenal.

For a deeper dive into the methodology, data requirements, and Stata implementation, consult the original PDF guide by Yotov et al. at yotoyotov.com/files/book.pdf.