Exploring Long-memory Dynamics in Nigerian Commercial Banks' Lending Rates: A Comparative Analysis of ARIMA, ARFIMA, and FIGARCH Models

This study investigates the dynamics of commercial banks’ maximum lending rates in Nigeria using short-memory ARIMA and long-memory models such as ARFIMA and the FIGARCH models. The data for the study spanned from January 1997 to May 2024. The results indicate that while ARIMA models adequately capture short-run autocorrelation, they struggle to address non-stationarity and long-run dependence. In contrast, ARFIMA models reveal a large, long-run dependence, with fractional difference (d) values ranging from −0.021 to 0.431, indicating both continuous and discontinuous persistent volatility behavior in commercial banks’ maximum lending rates in Nigeria. Similarly, the ranges of values of d in the FIGARCH models are less than zero (0.580, 0.564, 0.484), except in the estimation of the FIGARCH model using the maximum lending rate of banks in Nigeria, where the coefficient of the fractionally integrated root d (d -FIGARCH) is 1.450. The d-FIGARCH coefficient is greater than zero whereas others are less than zero. This is evidence of asymmetric reaction to shocks. This means that the series tends to reverse. The ARFIMA (1, 0.021, 2) model emerged as the best model based on model selection criteria, confirming the superiority of long-memory models in capturing the slow deterioration of commercial banks’ maximum lending rates to shocks. The superiority of the ARFIMA (1, 0.021, 2) model highlights the importance of long-memory models in capturing the continuous and dynamic behavior of commercial banks’ maximum lending rates in Nigeria. This is crucial for the commercial banking sector in Nigeria. This is because accurate forecasting enables informed decisions by investors, borrowers, and financial institutions. Understanding commercial banks’ maximum lending rates dynamics also helps policymakers develop effective monetary policies. Long-memory models such as ARFIMA consider historical patterns and anomalies, thereby reducing forecast errors. By using ARFIMA (1, 0.021, 2), stakeholders can better navigate Nigeria’s complex commercial banks’ lending rates system. Therefore, long-memory models are essential for understanding the persistence and mean-reversion dynamics of commercial banks’ maximum lending rates in Nigeria, providing valuable insights for forecasting and policy decisions.