Title: Practical NTM Estimation for CGE Modelling: Best Practice Gravity and Techniques for Computable General Equilibrium Analyses

In the realm of computable general equilibrium (CGE) modelling, the accuracy of spatial interactions hinges critically on the treatment of trade costs and flows between regions. This blog post synthesises practical approaches for estimating non-terminal marginal trade costs (NTM) within CGE frameworks, drawing on best practice gravity specifications to deliver robust, policy-relevant insights.

Why NTM estimation matters in CGE models
NTMs capture the friction and heterogeneity in bilateral trade that are not explained by distance alone. They reflect a blend of ad valorem equivalents, transport costs, border frictions, policy barriers, and other institutional factors that influence the ease with which goods and services move across borders. In CGE models, the gravity equation is the workhorse for modelling trade flows between regions or countries. Getting NTM estimation right is essential for:
– Accurately characterising trade elasticities and inter-regional spillovers.
– Deepening the realism of welfare and structural impact assessments.
– Enabling credible counterfactuals under policy reform, integration, or tariff changes.

Best practice gravity: a concise framework
A robust gravity specification in a CGE context usually involves a layered, theory-consistent approach:

1) Core gravity specification
– Trade flows are modelled as a function of economic mass (GDP or value-added proxies) and trade costs between origins and destinations.
– The standard functional form often uses an exponentiated(log-linear) structure, e.g., trade_ij = exp(α_i + β_j + γ_lnTij), where Tij captures bilateral trade costs.

2) Bilateral trade costs and NTM components
– NTMs are embedded in Tij as multiplicative cost shifters: Tij = base_cost_ij × NTM_ij.
– Base costs capture distance, shared language, colonial ties, common currency, and infrastructural proxies.
– NTM_i j reflects policy barriers (tariffs, non-tariff measures), regulatory distance, quality differentials, and non-policy frictions.

3) Homogeneity and symmetry considerations
– Ensure consistency with the underlying CGE structure: bilateral costs should align with the unit of trade (commodity, sector, or value-added). Depending on data richness, NTMs can be sector-specific or aggregate.
– Decide on symmetry: whether NTMs are identical in both directions or allow asymmetry due to preferential arrangements or border controls.

4) Inclusion of multilateral resistance terms
– Following the literature on the Anderson–van Wincoop framework, include multilateral resistance components in the gravity specification to capture the relative cost of trading with each partner given the global trade network.
– In empirical practice, this often translates into fixed effects or composite indices that absorb country-level and sector-level price pressures, reducing biased estimates of bilateral NTM effects.

5) Data sources and construction
– Use a combination of trade data (disaggregated by sector where possible) and proxy variables for trade costs: distance, shared language, colonial ties, common currency, time zones, trade agreements, infrastructure indicators, and policy indicators.
– Construct bilateral NTM indices by combining policy coverage (tariffs, import quotas, technical barriers), regulatory distance metrics, and quality-adjusted measures (e.g., product standards alignment).

6) Estimation strategies
– Use a gravity estimator that accommodates multilateral resistance terms and zero-trade flows when present.
– Employ Poisson Pseudo-Maximum Likelihood (PPML) or zero-inflated variants to handle zeros and heteroskedasticity common in bilateral trade data.
– Include exporter and importer fixed effects to absorb time-invariant bilateral characteristics and multilateral resistance.
– Consider random effects or panel specifications if the data span multiple periods and regions.

Practical steps for estimating NTM within a CGE context
– Step 1: Define the trading universe and harmonise sectoral classifications. Align the CGE model’s sectors with the granularity available in trade data to the extent possible.
– Step 2: Assemble bilateral trade data. If sectoral data are sparse, start with aggregate trade flows and progressively disaggregate as data permit.
– Step 3: Construct core gravity variables. Compute GDP (or value-added) measures for origin and destination, and assemble bilateral distance and shared characteristics (language, borders, trade agreements).
– Step 4: Build NTM measures. Combine policy indicators (tariffs, non-tariff barriers, standards alignment), regulatory distance metrics, and quality proxies into a composite NTM index for each i–j pair (and sector where feasible).
– Step 5: Estimate the gravity model with multilateral resistance. Use PPML with exporter and importer fixed effects, and include multilateral resistance terms through denser fixed effects or explicit indices.
– Step 6: Validate and diagnose. Check for robust elasticities, assess the sensitivity to sectoral disaggregation, and test alternative specifications for NTM construction.
– Step 7: Integrate into the CGE model. Map the estimated NTM effects onto the CGE’s trade cost structure, ensuring compatibility with the model’s units (e.g., ad valorem equivalents, iceberg costs).

Best practice tips and common pitfalls
– Beware zero trade flows: PPML is preferred over OLS on log-linear forms due to the prevalence of zeros in bilateral trade data.
– Guard against multicollinearity: Trade cost shifters such as distance, common language, and shared border can be highly correlated; use caution in interpretation and rely on fixed effects to absorb persistent factors.
– Ensure policy relevance: When translating NTM estimates into CGE parameters, consider the interpretation window. Ad valorem equivalents should align with the model’s pricing and transport cost structure.
– Address data limitations: If sector-level NTM data are unavailable, begin with broad sector aggregates and gradually introduce more granular measures as data enable.
– Conduct scenario testing: Use the gravity-estimated NTMs to simulate reforms (e.g., tariff reductions, alignment of standards) and compare outcomes against benchmarks to gauge model responsiveness.

Implications for policy analysis
A well-specified NTM estimation within a gravity framework enhances CGE modelling by:
– Improving the realism of trade frictions and their sectoral spillovers.
– Providing more credible welfare and production impact assessments under policy changes.
– Supporting evidence-based policy design for trade integration, regulatory harmonisation, and regional development strategies.

Concluding thoughts
Estimating NTMs through a best-practice gravity approach offers a practical, transparent pathway to strengthen CGE analyses. By carefully structuring the bilateral trade cost components, incorporating multilateral resistance, and grounding the estimation in robust data and estimation methods, analysts can deliver CGE results that are both policy-relevant and methodologically sound. This, in turn, supports more informed decision-making in trade policy, regional integration, and structural transformation efforts.

July 14, 2026 at 11:12AM
研究：估计非关税措施（NTMs）：DBT 工作论文
https://www.gov.uk/government/publications/estimating-non-tariff-measures-ntms-dbt-working-paper
DBT 关于使用最佳实践重力方法进行可计算一般均衡（CGE）建模的实用NTM估计的论文。”

阅读更多中文内容： Practical NTM Estimation for CGE Modelling: A DBT-Informed Approach with Best Practice Gravity