NEXUS OF UNIVARIATE AND MULTIVARIATE MODELS FOR FORECASTING INTEREST RATES IN PAKISTAN USING VAR-COV ANALYSIS

Abstract

Interest rate forecasting serves as a vital tool for economic decisions within Pakistan's changing monetary environment. Our research assesses how well different forecasting models such as Naïve, Exponential Smoothing, ARIMA, and NARDL predict interest rate fluctuations. The study uses quantitative research methods and data from 2001 to 2023 to evaluate model accuracy using root mean squared error (RMSE), mean absolute error (MAE), and Theil’s U-statistic. NARDL shows superior performance in forecasting because it effectively captures asymmetric economic relationships unlike traditional models while ARIMA fails to handle nonlinear financial dynamics. The variance-covariance (Var-Cov) analysis demonstrates that hybrid models combining ARIMA with Exponential Smoothing and NARDL produce the most dependable interest rate forecasts. The study highlights the necessity of combining structured econometric approaches with advanced hybrid methods to improve prediction accuracy and lower forecasting errors while supporting monetary policy development and financial risk management in Pakistan.

Keywords:

Interest rate forecast, ARIMA, NARDL, Exponential Smoothing, Naïve Model, Variance-Covariance Analysis.