Nowcasting economic activity in Argentina using newspaper articles

Abstract

This paper explores leveraging unstructured textual data from Argentinian newspapers to nowcast the monthly Economic Activity Estimator (EMAE) of Argentina. Various economic uncertainty indexes are constructed by applying natural language processing techniques on local newspaper articles. The association between these indexes and the EMAE is analyzed through correlation analysis, Granger causality tests, and out-of-sample nowcasting exercises. The results suggest that the proposed indexes exhibit predictive value for nowcasting the EMAE across diverse data splits. While limitations exist regarding generalization across economic cycles, this study contributes in the viability of extracting valuable signals from news content to gain timely insights into economic trends, and highlights the potential for nowcasting key indicators from unstructured data as text mining capabilities and data availability continue expanding.