Resumen de Short-term real-time forecasting model for spanish GDP (Spain-STING): new specification and reassessment of its predictive power

Ana Gómez Loscos, Miguel Ángel González Simón, Matías José Pacce

• The predictive power of short-term forecasting models was impaired by the increased volatility observed in most economic indicators following the outbreak of COVID-19. This paper sets out a revision of the Spain-STING model (one of the tools used by the Banco de España for short-term forecasts of quarter-on-quarter GDP growth) with a view to improving its predictive power in the wake of the pandemic. In particular, the revision entails three main changes: (i) the correlation between the indicators included in the model and the estimated common component is now coincident for all of the indicators, rather than leading in the case of some of them; (ii) by using a stochastic process to model the variance in the estimated common component, such variance may now vary over time; (iii) the set of indicators has been revised in order to include only those that provide the most relevant information when it comes to predicting post-pandemic GDP growth. These modifications yield a substantial improvement in the predictive power of Spain-STING in the post-pandemic period, and maintain such predictive power for the pre-pandemic period.