Dynamics and predictability in informal currency markets: The case of the Cuban Peso

Abstract

We investigate the short-term dynamics and predictability of the Cuban informal currency market, a critical case study for understanding emerging foreign exchange markets in countries with informal financial systems. Using social media messages of sell/buy intentions as a proxy for real market activity, we define a reference price for this informal market based on the Walrasian auction to capture market price trends. We explore how market fluctuations correlate with public announcements and news events, with a particular focus on understanding why overshooting events occur and how they can be anticipated. While the inherent inefficiency of these markets implies some level of predictability, standard methods fall short in capturing trend changes during overshooting episodes. To address this, we employ advanced Artificial Neural Networks (GRU-type), fine-tuned through bootstrapping, to generate accurate short-term forecasts. Our findings highlight that inefficiencies in informal markets create exploitable patterns, and that a neural network — carefully calibrated and optimized — is essential for anticipating overshooting events. This study contributes empirical evidence to the understanding of informal market dynamics and underscores the importance of developing predictive tools tailored to emerging foreign exchange markets.