A Review of Multivariate Statistical Modelling for Analysis of NEPSE Stock Data

Abstract

The stock market serves as a key indicator of economic growth; however, its behavior is inherently complex due to nonlinear dynamics influenced by economic, political, and psychological factors, as well as the fuzzy and volatile nature of financial data. With increasing market complexity, there is a growing need for advanced forecasting models capable of delivering higher predictive accuracy to support well-informed decision-making. This article aims to systematically review the existing literature on multivariate models used for stock market forecasting, and to identify dominant variables, evaluation methods, and research gaps. This study investigates the application of multivariate statistical modeling techniques to analyze Nepal Stock Exchange (NEPSE) data. The database search initially identified 60 records, and after applying the inclusion and exclusion criteria, 30 empirical studies published during the period 2008–2025 were included in the final review. The comparative analysis indicates that Vector Error Correction Models (VECM) consistently outperform GARCH, ARIMA models in capturing long-run stock index dynamics. For short-term forecasting of stationary data, VAR models demonstrate superior performance. In highly volatile and nonlinear market conditions, machine-learning models—particularly LSTM networks—achieve the highest forecasting accuracy. The comparative analysis indicates that Vector Error Correction Models (VECM) consistently outperform VAR and ARIMA models in capturing cointegration and long-run equilibrium adjustment. For short-term forecasting of stationary data, VAR models demonstrate superior performance. In highly volatile and nonlinear market conditions, machine-learning models—particularly LSTM networks—achieve the highest forecasting accuracy. The study offers insights into the relative effectiveness of different forecasting models and highlights the need for multivariate modeling of the NEPSE index to give much richer and realistic picture.