Time Series Modeling and Forecasting of NEPSE Index: An ARIMA/SARIMA Approach with Market Efficiency Evidence

Forecasting stock market behavior remains an important issue for investors, policymakers, and financial researchers, particularly in emerging markets where market volatility and informational inefficiencies may influence investment decisions. Despite growing interest in forecasting the Nepal Stock Exchange (NEPSE) index, existing studies have primarily focused on conventional ARIMAbased approaches without sufficiently incorporating rolling-window out-of-sample evaluation and benchmark comparison within the context of weak-form market efficiency. Addressing this gap, the present study examines the forecasting performance of the NEPSE daily closing index using an ARIMA/SARIMA-based time series framework and compares the selected model with a naïve random walk benchmark. The study utilized 1,157 daily observations of the NEPSE index and applied descriptive statistics, Augmented Dickey-Fuller stationarity testing, autocorrelation analysis, SARIMA model selection procedures, residual diagnostic testing, and rolling-window one-step-ahead forecasting techniques. Based on model selection criteria, seasonal structure, and diagnostic adequacy, SARIMA (3,1,0) (2,0,0) [5] was selected as the final forecasting model. The findings revealed that the SARIMA model generated slightly lower forecasting errors than the naïve benchmark during the Positive Science-Economics