Addressing the economic data crisis: Bayesian statistical learning for real-time forecasting and decision support.
Principal investigator and first supervisor, with C. Bueno Macedo Medeiros and G. Koop; PhD student: Thiago Da Cruz Alves Santos
Student Excellence Award (SEA), John Anderson Research Studentship Scheme (JARSS), University of Strathclyde, internally allocated studentship worth GBP 148,160.00, October 2026 - September 2029.
Summary. Recent episodes have exposed a broader crisis in official economic statistics: data arrive with delays, are revised, and are measured inaccurately, while decision makers in policy institutions and firms still have to act in real time. This project develops flexible and fast forecasting methods that combine modern Bayesian machine learning with a wider information set than forecasters typically exploit, including conventional time series as well as survey- and text-based indicators. The research is organised in three chapters: (1) Bayesian mixed-frequency (MIDAS) regressions that handle high-dimensional data released at different frequencies for forecasting GDP growth and inflation; (2) Bayesian machine learning MIDAS regressions that incorporate unconventional data sources to improve measurement at lower frequencies; and (3) an extension of this framework to real-time measures of macroeconomic vulnerability and tail risks. The aim is a unified framework for macroeconomic measurement, forecasting, and risk assessment that provides policymakers with timely signals of economic instability and supports decisions under uncertainty and imperfect information.
