L’EMbeDS DS^3: "Nowcasting GDP in a data-poor environment"

The DS³ organizing group, alongside the L'EMbeDS Department of Excellence of the Sant'Anna School for Advanced Studies, is launching a new online seminars series devoted to frontier research in data science, its applications and its implications across disciplines.  

The L'EMbeDS Data Science Seminar Series (L'EMbeDS DS³) hosts national and international scholars to discuss cutting-edge methodology, applications in economics, the social sciences -- and beyond, societal implications and governance issues. 

For announcements and further information please visit our web page and join our Google Group. 

The seminar will feature Samuele Centorrino (International Monetary Fund), who will present a talk entitled: “Nowcasting GDP in a data-poor environment” 

>>Join the event via the following link

ABSTRACT:  

The timely and high-frequency measurement of economic activity is a necessary condition for the formulation of evidence-based policies. However, it remains largely inaccessible in low-income, fragile, and conflict-affected states, where quarterly GDP estimates are unavailable or published with substantial lags. Non-traditional data sources provide a viable avenue for constructing real-time proxies of economic activity. However, machine learning estimators, the canonical tools for exploiting such high-dimensional data, exhibit systematic performance degradation in these contexts: the combination of short time-series and conventional cross-validation schemes induces overfitting and pronounced deterioration in out-of-sample predictive accuracy. This paper addresses this methodological gap by proposing a block-bootstrap cross-validation procedure tailored to hyperparameter tuning of machine learning models under data-scarce conditions. We apply this framework to nowcast quarterly economic activity across a heterogeneous sample of developing and low-income economies, and establish its out-of-sample validity through Monte Carlo simulations and an empirical application to several African countries. The proposed methodology extends the nowcasting toolkit to environments characterized by data paucity, institutional fragility, and weak structural fundamentals, with direct implications for economic surveillance, crisis response, and resilience-building in data-poor contexts.