SHORT-TERM FORECASTING OF BANK STOCK PRICES IN AN EMERGING FRONTIER MARKET: EVIDENCE FROM RANDOM WALK, ARIMA AND VAR MODELS IN UZBEKISTAN

Abstract

This study examines the short-term forecasting dynamics of bank stocks listed on the Republican Stock Exchange “Toshkent” using daily data from 9 September 2022 to 17 April 2026 (883 price observations; 882 logarithmic returns). Four banking issuers – HMKB, SQB, IPKY and ALKB – are analysed jointly with the Uzbekistan Composite Index (UCI) within a unified empirical framework that combines benchmark Random Walk, univariate ARIMA and multivariate VAR(4) models. The Augmented Dickey–Fuller test confirms stationarity of all return series, while ARCH–LM tests reveal pronounced volatility clustering in UCI, HMKB, SQB and ALKB. Annualised volatility ranges from 60.41% (SQB) to 153.41% (IPKY), evidencing substantial heterogeneity in idiosyncratic risk. The VAR(4) system is stable (maximum modulus of characteristic roots 0.6377), and forecast error variance decomposition reveals strong own-shock dominance ranging from 94.79% to 98.00% at the 20-day horizon, while Granger causality tests indicate that bankspecific shocks rather than the aggregate index drive system dynamics. Outof-sample comparisons based on RMSE, MAE and MAPE select VAR(4) for ALKB, HMKB and IPKY, and ARIMA for SQB. The findings document weak cross-asset information transmission, persistent volatility clustering and heterogeneous predictability – stylised facts characteristic of frontier equity markets – and provide direct policy guidance for ongoing capital-market reforms in Uzbekistan.