Assessing Financial Risk using Value-At-Risk (VaR) from the perspective of a third world economy, Zimbabwe’s Forex Market

Abstract

The Global Financial Depression of 2008 exposed the problems of financial risk estimations in the forex sector and impacted negatively on developing countries. In this paper, the performance of Generalised Autoregressive Conditional Heteroskedasticity (GARCH) family models are used to assess and compared in the estimation of Value at Risk (VaR). The study is based on three major currencies that are used in Zimbabwe’s multiple-currency regime against the USD.  The three exchange rates considered are, the ZAR/USD, the EUR/USD, and the GBP/USD. Three univariate types of GARCH models, with the Student’s t and the Normal error distributions, are applied to the three currency indices to ascertain the best VaR estimation formula. Evaluation tests, namely the Violation ratio, Kupiec’s test, and Christoffersen’s test are used to assess the quality of the VaR performance. The GARCH (1, 1) with t-distributed errors produced relatively more accurate computations on the VaR for EUR/USD and GBP/USD at 99% level of significance, while the backtests results were inconclusive for ZAR/USD. The GARCH(1,1) model with t-distributed errors had the lowest Akaike's Information Criterion (AIC) and Schwarz’s Bayesian Information Criterion (SBIC) values. The GARCH (1,1) with t-distributed error model is suggested in computing VaR and making other deductions on the capital required. The Global Financial Depression of 2008 exposed the problems of financial risk estimations in the forex sector and impacted negatively on developing countries. In this paper, the performance of Generalised Autoregressive Conditional Heteroskedasticity (GARCH) family models is used to assess and compared in the estimation of Value at Risk (VaR). The study is based on three major currencies that are used in Zimbabwe’s multiple-currency regime against the USD.  The three exchange rates considered are, the ZAR/USD, the EUR/USD, and the GBP/USD. Three univariate types of GARCH models, with the Student’s t and the Normal error distributions, are applied to the three currency indices to ascertain the best VaR estimation formula. Evaluation tests, namely the Violation ratio, Kupiec’s test, and Christoffersen’s test are used to assess the quality of the VaR performance. The GARCH (1, 1) with t-distributed errors produced relatively more accurate computations on the VaR for EUR/USD and GBP/USD at 99% level of significance, while the backtests results were inconclusive for ZAR/USD. The GARCH(1,1) model with t-distributed errors had the lowest Akaike's Information Criterion (AIC) and Schwarz’s Bayesian Information Criterion (SBIC) values. The GARCH (1,1) with t-distributed error model is suggested in computing VaR and making other deductions on the capital required.