Volatility Modelling of Nigerian Bank Stocks: Evidence from Real and Simulated Data using GARCH and Machine Learning

Abstract

This study examines the volatility dynamics of selected Nigerian banking stocks us- ing both real and simulated share price data. Daily closing prices of five major Nige- rian banks Access Holdings Plc (ACCESSCORP), Guaranty Trust Holding Company Plc (GTCO), Fidelity Bank Plc (FIDELITYBK), FBN Holdings Plc (FBNH), and United Bank for Africa Plc (UBA) are analyzed over a common sample period from January 1, 2015, to December 17, 2025. Log-returns are computed and volatility is modelled using the GARCH (1,1), EGARCH (1,1), and TGARCH (1,1) models under Normal and Student-t er- ror distributions. To assess the robustness of volatility behaviour, simulated return series are generated from the empirical properties of the real data under a Gaussian assumption and analysed alongside the observed data. Furthermore, machine learning models includ- ing Random Forest regression, Long Short-Term Memory (LSTM) networks, and hybrid GARCH LSTM approaches are employed to enhance volatility forecasting performance. Results indicate that real data exhibits higher volatility persistence and thicker tails than simulated data across most banks. Among the GARCH-family models, EGARCH with Student-t innovations provides superior performance in capturing asymmetric effects and extreme market movements. Machine learning models further improve forecasting accu- racy, particularly during periods of financial stress. The findings offer important insights for investors, risk managers, and policymakers in emerging financial markets.