2016年-WTO世界贸易组织_Estimating_Trade_Policy_Effects_with_Structural_Gravity__63页_1mb

Summary of "Estimating Trade Policy Effects with Structural Gravity"

Core Content

This manuscript provides a practical guide for estimating the effects of trade policies using the structural gravity model. It emphasizes the importance of theoretical consistency and addresses several key challenges in empirical gravity estimation, offering six best practices to improve the reliability and validity of partial equilibrium estimates. The document also discusses data sources, interpretation, and applications of the structural gravity model in analyzing trade policy impacts, including regional trade agreements and unilateral measures like MFN tariffs.

Main Challenges and Solutions

Challenge 1: Multilateral Resistances (MRs)

• Description: MRs are theoretical constructs that are not directly observable, making them a challenge for estimation.
• Solutions:
  • Use exporter-time and importer-time fixed effects to account for MRs.
  • These fixed effects also absorb size variables and other country-specific characteristics.

Challenge 2: Zero Trade Flows

• Description: OLS estimation in log-linear form discards zero trade flows, leading to potential bias.
• Solutions:
  • Use the Poisson Pseudo Maximum Likelihood (PPML) estimator, which handles zero trade flows effectively.
  • PPML is recommended due to its robustness and ability to maintain theoretical consistency.

Challenge 3: Heteroskedasticity of Trade Data

• Description: Trade data often exhibit heteroskedasticity, which can lead to biased and inconsistent estimates.
• Solutions:
  • Transform trade flows into size-adjusted trade to reduce heteroskedasticity.
  • Use PPML estimator, which is more robust to heteroskedasticity than OLS.

Challenge 4: Bilateral Trade Costs

• Description: Accurate specification of bilateral trade costs is essential for reliable policy analysis.
• Solutions:
  • Include a range of observable variables such as distance, contiguous borders, language, colonial ties, and trade policy indicators (RTA and tariffs).
  • The coefficient on tariffs directly reflects the elasticity of substitution, allowing for the estimation of trade policy effects.

Challenge 5: Endogeneity of Trade Policy

• Description: Trade policy variables may be endogenous, leading to biased estimates.
• Solutions:
  • Use country-pair fixed effects or first-differencing to control for endogeneity.
  • Pair fixed effects are preferred as they account for both unobservable time-invariant trade costs and endogeneity.

Challenge 6: Non-discriminatory Trade Policy

• Description: Non-discriminatory policies like MFN tariffs and export subsidies are difficult to estimate due to their absorption by fixed effects.
• Solutions:
  • Estimate the model using both international and intra-national trade flows to identify non-discriminatory effects.
  • This approach allows non-discriminatory variables to be treated as bilateral, thus avoiding their absorption by fixed effects.

Challenge 7: Adjustment to Trade Policy Changes

• Description: Trade flows adjust gradually to policy changes, not instantaneously.
• Solutions:
  • Use panel data and allow for interval data instead of consecutive years.
  • This approach better captures the dynamic adjustment process in trade flows.

Practical Recommendations

1. Use panel data to capture time-varying effects.
2. Allow for interval data instead of consecutive years to better represent adjustment periods.
3. Include intra-national trade flows to improve the robustness of estimates.
4. Use directional (exporter and importer) time-varying fixed effects to control for unobservable factors.
5. Employ pair fixed effects to address endogeneity and time-invariant trade costs.
6. Estimate gravity in multiplicative form using the PPML estimator, which is robust to zero trade flows and heteroskedasticity.

Theoretical-Consistent Estimating Gravity Model

The authors propose a comprehensive and theoretically-consistent gravity model that integrates the structural gravity framework. This model allows for the simultaneous identification of bilateral and unilateral trade policy effects. The equation is log-linearized and expanded with an additive error term, resulting in the following form:

$$
\ln X _ {i j, t} = \ln E _ {j, t} + \ln Y _ {i, t} - \ln Y _ {t} + (1 - \sigma) \ln t _ {i j, t} - (1 - \sigma) \ln P _ {j, t} - (1 - \sigma) \ln \Pi_ {i, t} + \epsilon_ {i j, t}.
$$

The use of PPML estimator is highlighted as a key advantage due to its ability to handle zero trade flows and heteroskedasticity effectively. It is also consistent with the structural gravity model and allows for the recovery of trade elasticity directly from the tariff coefficient.

Interpretation and Aggregation

• Interpretation: Gravity estimates should be interpreted with care, especially when dealing with the effects of trade policies.
• Aggregation: Consistent aggregation methods are necessary to avoid biases in the estimation of bilateral trade costs and their effects on trade flows.

Data Sources and Limitations

• Bilateral Trade Flows Data: Available from various sources such as the UN COMTRADE database and national trade statistics.
• Bilateral Trade Costs Data: Includes tariff data and non-tariff measures (NTMs).
• Limitations: Data availability and quality can pose significant challenges, especially for non-tariff measures and sectoral trade data.

Applications

• Traditional Gravity Estimates: Used to analyze the effects of various trade determinants.
• Distance Puzzle: Addressed using the structural gravity model to capture the impact of globalization.
• Regional Trade Agreements (RTAs): Estimated as a representative form of bilateral trade policy.
• Unilateral Trade Policy: MFN tariffs are used as a representative form of non-discriminatory trade protection.

Conclusion

The manuscript concludes by emphasizing the importance of using a theoretically-consistent and robust estimation approach for trade policy analysis. It highlights the effectiveness of the structural gravity model and the PPML estimator in overcoming common estimation challenges. The authors also stress the need for careful interpretation and data handling to ensure the validity of results.