Out-of-sample gravity predictions and trade policy counterfactuals

2025-09-14

AI Summary

This study evaluates the applicability of gravity models for counterfactual trade policy analysis—particularly assessing regional trade agreements—with primary focus on out-of-sample predictive performance. Methodologically, it systematically compares three conventional gravity specifications (including a three-dimensional extension) against machine learning approaches—namely random forests and neural networks—in predicting bilateral trade flows, using out-of-sample forecast accuracy as the principal evaluation criterion. Results indicate that the three-dimensional gravity model delivers greater robustness in predicting aggregate trade means, whereas ensemble machine learning methods significantly outperform traditional specifications in forecasting individual bilateral flows, especially in capturing nonlinearities and heterogeneity. Crucially, this paper is the first to establish predictive performance as a key validity criterion for gravity-model-based policy simulations. By doing so, it provides a methodological benchmark and practical guidance for model selection in quantitative trade policy analysis.

Technology Category

• Machine Learning: Evaluation and Analysis
• Game Theory and Economic Paradigms: Mechanism Design
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success
• Web Mining and Content Analysis: Models for Web evolution

Abstract

Gravity equations are often used to evaluate counterfactual trade policy scenarios, such as the effect of regional trade agreements on trade flows. In this paper, we argue that the suitability of gravity equations for this purpose crucially depends on their out-of-sample predictive power. We propose a methodology that compares different versions of the gravity equation, both among themselves and with machine learning-based forecast methods such as random forests and neural networks. We find that the 3-way gravity model is difficult to beat in terms of out-of-sample average predictive performance, further justifying its place as the predominant tool for applied trade policy analysis. However, when the goal is to predict individual bilateral trade flows, the 3-way model can be outperformed by an ensemble machine learning method.

Problem

• Evaluating gravity models' out-of-sample trade prediction accuracy
• Comparing traditional gravity equations with machine learning methods
• Assessing model performance for trade policy counterfactual analysis

Innovation

• Compares gravity models with machine learning methods
• Uses random forests and neural networks for forecasting
• Proposes ensemble method for bilateral trade prediction