Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Machine learning and economic forecasting: the role of international trade networks

Thiago C. Silva, Paulo V. B. Wilhelm, Diego R. Amancio

TL;DR

The paper investigates whether international trade networks improve GDP growth forecasting under de-globalization trends. It adopts a horse race of supervised regressors and SHAP-based interpretability to assess how section-level commodity trade topology enhances forecasts for about 200 countries from 2010 to 2022. It finds that non-linear models (RF, XGBoost, LightGBM) outperform linear baselines and that roughly half of the most influential predictors come from network descriptors, with Mineral-trade-density among the strongest signals and autoregressive growth dynamics consistently important. These findings highlight the value of network topology in macro forecasting and offer policy insights on navigating trade uncertainty and the rising influence of emerging economies.

Abstract

This study examines the effects of de-globalization trends on international trade networks and their role in improving forecasts for economic growth. Using section-level trade data from nearly 200 countries from 2010 to 2022, we identify significant shifts in the network topology driven by rising trade policy uncertainty. Our analysis highlights key global players through centrality rankings, with the United States, China, and Germany maintaining consistent dominance. Using a horse race of supervised regressors, we find that network topology descriptors evaluated from section-specific trade networks substantially enhance the quality of a country's GDP growth forecast. We also find that non-linear models, such as Random Forest, XGBoost, and LightGBM, outperform traditional linear models used in the economics literature. Using SHAP values to interpret these non-linear model's predictions, we find that about half of most important features originate from the network descriptors, underscoring their vital role in refining forecasts. Moreover, this study emphasizes the significance of recent economic performance, population growth, and the primary sector's influence in shaping economic growth predictions, offering novel insights into the intricacies of economic growth forecasting.

Machine learning and economic forecasting: the role of international trade networks

TL;DR

The paper investigates whether international trade networks improve GDP growth forecasting under de-globalization trends. It adopts a horse race of supervised regressors and SHAP-based interpretability to assess how section-level commodity trade topology enhances forecasts for about 200 countries from 2010 to 2022. It finds that non-linear models (RF, XGBoost, LightGBM) outperform linear baselines and that roughly half of the most influential predictors come from network descriptors, with Mineral-trade-density among the strongest signals and autoregressive growth dynamics consistently important. These findings highlight the value of network topology in macro forecasting and offer policy insights on navigating trade uncertainty and the rising influence of emerging economies.

Abstract

This study examines the effects of de-globalization trends on international trade networks and their role in improving forecasts for economic growth. Using section-level trade data from nearly 200 countries from 2010 to 2022, we identify significant shifts in the network topology driven by rising trade policy uncertainty. Our analysis highlights key global players through centrality rankings, with the United States, China, and Germany maintaining consistent dominance. Using a horse race of supervised regressors, we find that network topology descriptors evaluated from section-specific trade networks substantially enhance the quality of a country's GDP growth forecast. We also find that non-linear models, such as Random Forest, XGBoost, and LightGBM, outperform traditional linear models used in the economics literature. Using SHAP values to interpret these non-linear model's predictions, we find that about half of most important features originate from the network descriptors, underscoring their vital role in refining forecasts. Moreover, this study emphasizes the significance of recent economic performance, population growth, and the primary sector's influence in shaping economic growth predictions, offering novel insights into the intricacies of economic growth forecasting.
Paper Structure (22 sections, 1 equation, 5 figures, 3 tables)

This paper contains 22 sections, 1 equation, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Network analysis of section-level international trade networks
  3. Data description
  4. Relevant background: network measurements from the complex networks literature
  5. Topological analysis of section-level international trade networks
  6. Trade network measures as predictors: a machine learning approach to GDP growth forecasting
  7. Relevant background: supervised machine learning models
  8. Linear Regression
  9. Elastic Net (Linear Regression with regularization)
  10. Support Vector Machine
  11. k-Nearest Neighbor
  12. Random Forest
  13. XGBoost
  14. Light Gradient Boosting Machine
  15. Relevant background: the interpretable machine learning through Shapley values
  16. ...and 7 more sections

Figures (5)

  • Figure 1: This figure presents an analytical overview through five distinct plots, illustrating the evolution of key topological metrics within international trade networks from 2010 to 2022. Each plot tracks the temporal progression of a particular network measure for each of the five main section-level trade networks under scrutiny. The analysis employs a quarterly temporal resolution.
  • Figure 2: This figure illustrates the estimated error values and their associated confidence intervals for the four error metrics—Huber Loss, Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Symmetric Mean Absolute Percentage Error (SMAPE)—evaluated across the seven machine learning algorithms examined in our study.
  • Figure 3: This figure illustrates the average SHAP values for the Top 15 most influential features as determined for the XGBoost, LightGBM, and Random Forest models. The visualization provides a comparative analysis across the three models, highlighting how each model values different features in terms of their predictive power.
  • Figure 4: This figure displays a beeswarm plot of the Top 20 features from the Random Forest model, ranked by their mean SHAP values. Features are ordered vertically by importance, with their standardized SHAP values plotted horizontally, showing the contribution to prediction. Dots represent individual observations, with their distribution indicating the variability of SHAP values; dense areas suggest higher concentration, while sparse areas indicate less. The color of the dots reflects the feature values, elucidating the influence of high versus low feature values on economic growth forecasts, with positive SHAP values signaling potential growth and negative values indicating possible downturns.
  • Figure 5: This figure showcases the SHAP value dependence plots for five key features of the Random Forest model. These plots elucidate the nuanced relationships between variations in feature values and their impact on predictive outcomes. By standardizing feature values along the horizontal axis, the plots effectively demonstrate how both positive and negative deviations from the mean feature value contribute to changes in the model's economic growth predictions.