Evaluating the predictive power of Mexico’s timely economic activity indicator: real and pseudo real-time performance

Abstract

This paper evaluates the performance of Mexico’s Timely Economic Activity Indicator (IOAE), published by the National Institute of Statistics and Geography, as a nowcasting tool for short-term economic activity. Using both real-time evidence (October 2020– December 2024) and a pseudo-real-time comparison with alternative nowcasting approaches over 2018–2024, we assess the IOAE against individual indicators, econometric models, and machine-learning methods. Nowcast accuracy is evaluated using Diebold–Mariano tests with Heteroskedasticity and Autocorrelation Consistent correction, the Superior Predictive Ability test, and the Model Confidence Set procedure. The results indicate that the IOAE consistently belongs to the set of best-performing models and tends to exhibit relatively stronger predictive accuracy, particularly at the two-month-ahead horizon. A comprehensive robustness analysis—covering alternative window schemes, numbers of factors, loss functions, and sample definitions, including evaluations conducted both with and without the COVID-19 period—supports the stability of these findings. Overall, the evidence supports the view that the IOAE provides timely and informative nowcasts of economic activity, reinforcing its usefulness as a complementary tool for short-term economic monitoring in Mexico.