Nowcasting GDP with Machine Learning: The Case of Indonesia

Abstract

Abstract This paper develops robust economic models for nowcasting Indonesia's GDP growth. Utilizing a comprehensive dataset comprising monthly economic and finance indicators from 2013 to 2023, we explore their association with GDP growth. The study period spans from January 2013 to December 2022 for in-sample data, with an extension to September 2023 for out-sample analysis. We employ a suite of machine learning models - Elastic Net, Random Forest, XGBoost, and SVM - selected for their complementary strengths. The models are trained and tested, yielding an RMSE range of 0.01 to 0.41 for the training period and a narrower range of 0.11 to 0.17 for the testing period, indicating robust performance across different economic conditions, including normal and pandemic periods. A key aspect of our study is the model interpretation, which includes decomposition and Shapley value analysis, providing insights into the driving factors behind GDP fluctuations. Our findings highlight the significant impact of several indicators on Indonesia's GDP. Additionally, our nowcasting exercise for Q4 2023 projects an average GDP growth of 5.13%, ranging from 5.01–5.26%. JEL classification: C53, C63, E37, E58, O40