Nowcasting world trade with machine learning: a three-step approach - Working Paper Series - No. 2836

Aug 2, 2026

Upload by Hazel

The paper nowcasts world trade using machine learning, contrasting tree-based methods (random forest, gradient boosting) with regression-based counterparts (macroeconomic random forest, linear gradient boosting). Regression-based approaches deliver higher and more consistent performance than tree methods and a range of traditional benchmarks including OLS, Markov-switching, quantile regression, and PCA-OLS. A flexible three-step pipeline—pre-selection, factor extraction, and machine learning regression—improves predictive accuracy. The approach also scales seamlessly beyond world trade forecasting.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages52

Reading Time2 h 11 min

AccessFree to download

Abstract

Non-technical summary

Data and motivation

Machine learning approach and results

Three-step forecasting framework

Chinn, Menzie David; Meunier, Baptiste; Stumpner, Sebastian Working Paper Nowcasting world trade with machine learning: A threestep approach ECB Working Paper, No. 2836 Provided in Cooperation with: European Central Bank (ECB) Suggested Citation: Chinn, Menzie David; Meunier, Baptiste; Stumpner, Sebastian (2023) : Nowcasting world trade with machine learning: A three-step approach, ECB Working Paper, No. 2836, ISBN 978-92-899-6121-9, European Central Bank (ECB), Frankfurt a. M., [https://doi.org/10.2866/744676](https://doi.org/10.2866/744676) This Version is available at: [https://hdl.handle.net/10419/278668](https://hdl.handle.net/10419/278668) Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor maybe saved and copied foryour personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. Working Paper Series Menzie Chinn, Baptiste Meunier, Sebastian Stumpner Nowcasting world trade with machine learning: a three-step approach No 2836 Disclaimer: This paper should not be reported as representing the views of the European Central Bank (ECB) . The views expressed are those of the authors and do not necessarily reflect those of the ECB. Abstract We nowcast world trade using machine learning, distinguishing between tree-based methods (random forest, gradient boosting) and their regression-based counterparts (macroeconomic random forest, linear gradient boosting). While much less used in the literature, the latter are found to outperform not only the tree-based techniques, but also more “traditional” linear and non-linear techniques (OLS, Markov-switching, quantile regression). They do so significantly and consistently across different horizons and real-time datasets. To further improve performances when forecasting with machine learning, we propose a flexible three-step approach composed of (step 1) pre-selection, (step 2) factor extraction and (step 3) machine learning regression. We find that both pre-selection and factor extraction significantly improve the accuracy of machine-learning-based predictions. This three-step approach also outperforms workhorse benchmarks, such as a PCA-OLS model, an elastic net, or a dynamic factor model. Finally, on top of high accuracy, the approach is flexible and can be extended seamlessly beyond world trade. Keywords: Forecasting, big data, large dataset, factor model, pre-selection JEL classification: C53, C55, E37 Non-technical summary1 Real-time economic analysis often faces the fact that indicators are published with significant lags. This problem is encountered for world trade in volumes: the earliest indicator is published by the Dutch Centraal Plan Bureau (CPB) roughly eight weeks after month end – meaning that March 2023 data is available around May 25th. Since these data are widely used among economists, this poses a challenge policy-wise as decisions should rely on timely information about the current business cycle. In the meantime, a number of early indicators are available. The purpose of this paper is to exploit such informa

FAQs

What problem does the paper address in world trade analysis?

Indicators for world trade volumes arrive with substantial publication lags, making timely policy-relevant assessment difficult. The paper aims to generate advanced estimates ahead of official releases.

How do the paper’s machine learning methods differ?

It compares tree-based methods (random forest, gradient boosting) with regression-based techniques (macroeconomic random forest, linear gradient boosting). The regression-based methods perform better on the paper’s dataset.

What are the three steps in the proposed forecasting approach?

Step 1 performs pre-selection of informative predictors; Step 2 extracts and orthogonalizes selected variables into a few factors; Step 3 uses machine learning regression on those factors.

Aug 2, 2026

Upload by Hazel