A comparative study of the VECM, GARCH and Multivariate GARCH Techniques in Modelling External Debt

Abstract

The study modelled the determinants of external debt using the VECM, GARCH and multivariate GARCH models with the intent to identify and recommend the most effective approach. The study employed quarterly time series data obtained from the South African Reserve Bank ranging from the second quarter of 1990 to the fourth quarter of 2018. The results of the VECM revealed that the long run relationship among the variables showed that LCF, LEXP, LGDP have a positive long run relationship with LED. The LGE has a negative long run relationship with LED. The results of the GARCH(1,1) model showed that the variance of the series is increasing over time since the sum of the ARCH and GARCH term is greater than one. Furthermore, the results of the GARCH(1,1) model revealed that a positive and statistically significant relationship exists between LEXP and LED; LGDP and LED. There is also a negative and statistically significant relationship found between LCF and LED; LGE and LED. The results of the BEKK model revealed that only one diagonal parameter B(5,5) was statistically insignificant which meant that past conditional volatility does not influence volatility in external debt. The results also revealed that there was a unidirectional volatility transmission between LGDP and LED; LCF and LED. The model evaluation results suggested that the GARCH(1,1) is more efficient in modelling financial data in South Africa as compared to the other two techniques. The recommendations for future studies were recommended based of the findings of the study.