Principal Components Vector Error Correction Model For Revenue Components Of Ghana

Abstract

Revenue data structure has assumed a dynamic nature, and evolving methodology for their study constitutes an interesting problem. In this regard, the study examines the various revenue components that are most influential in revenue generation and attempts to obtain a suitable multivariate time series model that characterizes the contribution of each revenue component in Ghana. Data is therefore obtained on some fourteen revenue variables from Ghana Community Management System for the study. The theory of VEC modelling, which is relevant for variables expected to be related in the long nm, is found appropriate for the study. An optimum lag order is determined at 8. The VEC(8) model produces more realistic performance measures than the initial V AR(8). By incorporating principal components extraction into the VEC model, five salient revenue dimensions are identified with no loss of information. The most dominant source is what is influenced by eIF, accounting for about 80% of the total variation in all revenue sources. The remaining 20% is explained by Volume (VOL), Total Revenue (TORE), Total Amount Exempt (TOAE) and Petroleum tax (PETAL), in that order. The VEC model is applied to project the original data onto the five components. The resulting PCA VEC model now provides a plausible econometric characterization of the data structure. The results suggest that CIF, in particular, should be protected to generate the requisite revenue.