Markov Switching VAR-Artificial Neural Network Hybrid Model for Studying Monetary Aggregates in Nigeria

Abstract

This study examines changes in behaviour and structure of Nigeria’s monetary aggregates: Broad Money Supply (M2), Narrow Money Supply (M1), Quasi Money (QM), and Bank Reserves (BR) using a combined analytical approach integrating Markov Switching Vector Autoregressive (MS-VAR) models Markov Switching-Mean (MS-MEAN), Time-Varying Transition Probability Markov Switching (TVTP-MS), and Markov Switching with Exogenous Variables (MSX-VAR) with Artificial Neural Networks (ANN). Monthly data from 1997 to 2025 was sourced from the Central Bank of Nigeria Statistical Bulletin. Results indicate monetary aggregates display distinct regime-dependent behaviours marked by low and high volatility states with lasting periods. MS-Mean models show low-volatility environments correlate with extended stability, while high-volatility conditions reveal substantial variations. TVTP-MS models emphasise evolving transition characteristics, and MSX-VAR captures asymmetric volatility transmission among monetary aggregates. Hybrid MS-ANN models demonstrate minor nonlinear influences, enhancing comprehension of leftover interactions but yielding improvements in predictive performance compared to standard Markov Switching models. Results highlight liquidity cycles are significantly sensitive to different regimes, indicating monetary authorities should implement flexible and regime-conscious methods to manage Nigeria’s financial landscape effectively.