Nowcasting domestic liquidity in the Philippines using machine learning algorithms

Abstract

This study utilizes a number of algorithms used in machine learning to nowcast domestic liquidity growth in the Philippines. It employs regularization (i.e., Ridge Regression, Least Absolute Shrinkage and Selection Operator (LASSO), Elastic Net (ENET)) and tree-based (i.e., Random Forest, Gradient Boosted Trees) methods in order to support the BSP’s current suite of macroeconomic models used to forecast and analyze liquidity. Hence, this study evaluates the accuracy of time series models (e.g., Autoregressive, Dynamic Factor), regularization, and tree-based methods through an expanding window. The results indicate that Ridge Regression, LASSO, ENET, Random Forest, and Gradient Boosted Trees provide better estimates than the traditional time series models, with month-ahead nowcasts yielding lower Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). Furthermore, regularization and tree-based methods facilitate the identification of macroeconomic indicators that are significant to specify parsimonious nowcasting models.