Nowcasting GDP - A Scalable Approach Using DFM, Machine Learning and Novel Data, Applied to European Economies

IMF Working Papers, March 11, 2022

Source details

Canonical URL: Nowcasting GDP - A Scalable Approach Using DFM, Machine Learning and Novel Data, Applied to European Economies

Authors: Jean-Francois Dauphin, Kamil Dybczak, Morgan Maneely, Marzie Taheri Sanjani, Nujin Suphaphiphat, Yifei Wang, Hanqi Zhang
Published: March 11, 2022
Series: IMF Working Papers
DOI: https://doi.org/10.5089/9798400204425.001

Describes recent work to strengthen nowcasting capacity at the IMF’s European department.
Motivates and compiles datasets of standard and nontraditional variables, such as Google search and air quality.
Applies standard dynamic factor models (DFMs) and several machine learning (ML) algorithms to nowcast GDP growth across a heterogeneous group of European economies during normal and crisis times.
Conclusion: Most methods significantly outperform the AR(1) benchmark model.
Conclusion: DFMs tend to perform better during normal times.
Conclusion: Many of the ML methods used performed strongly at identifying turning points.
Claim: The approach is easily applicable to other countries, subject to data availability.

Datasets include standard and nontraditional variables, explicitly mentioning Google search and air quality.
Focus on large and varied data inputs suitable for factor models and machine learning algorithms.

Most methods significantly outperform the AR(1) benchmark model.
DFMs perform relatively better in normal (non-crisis) periods.
Many ML methods perform strongly at identifying turning points in GDP growth.
The framework is scalable and intended for application across a heterogeneous group of European economies.

Approach is easily applicable to other countries, subject to data availability.
Explicit focus on both normal and crisis times, highlighting differential strengths of DFMs and ML methods.