GDP nowcasting with Machine Learning and Unstructured Data

In a context of ongoing change, “nowcasting” models based on Machine Learning (ML) algorithms deliver a noteworthy advantage for decision-making in both the public and private sectors due to their flexibility and ability to drive large amounts of data. This document introduces projection models designed for real-time forecasting of the monthly Peruvian GDP growth rate. These models integrate structured macroeconomic indicators with high-frequency unstructured sentiment variables. The analysis spans from January 2007 to May 2023, encompassing a comprehensive set of 91 leading economic indicators. Six ML algorithms were rigorously evaluated to identify the most effective predictors for each model. The findings underscore the remarkable capability of ML models to yield more precise and foresighted predictions compared to conventional time series models. Notably, Gradient Boosting Machine, LASSO, and Elastic Net emerged as standout performers, demonstrating a prediction error reduction of 20% to 25% when contrasted with AR and various specifications of DFM. These results could be influenced by the analysis period, which includes crisis events featuring high uncertainty, where ML models with unstructured data improve significance.