Structural and Variance Properties of Multivariate Autoregressive Models: Empirical Applications to Nigerian Economic Variables

Abstract

Multivariate Autoregressive models are widely used models in the study of economic variables. However, the understanding of the mean and variance parameters of a Multivariate Autoregressive Model enhances the accuracy and reliability of statistical inferences, forecasting precision, and model stability. It provides deeper insights into the model’s dynamic behavior, allowing for better decision-making in economic, financial, and epidemiological applications. Additionally, it aids in selecting the most appropriate model for specific data structures, ensuring robust and consistent predictions. Using the Nigerian macro-economic variables, the paper advances the theoretical and practical understanding of Multivariate Autoregressive Distributed Lag (MARDL) models by systematically analyzing their properties in comparison to Vector Autoregressive (VAR) models. This paper evaluates the mean and variance properties of both models to assess their relative performance. Findings indicate no significant difference in their mean properties; however, a variance analysis reveals that the MARDL model exhibits smaller variances compared to the VAR model. This suggests that the MARDL model offers greater stability, control, and consistency, making it a more reliable tool for prediction and forecasting. The results contribute to the ongoing refinement of multivariate time series modeling and its application in empirical research.