Backtesting the evaluation of Value-at-Risk methods for exchange rates

Purpose - Small- and medium-sized enterprises can be highly affected by losses caused by exchange rate changes. The aim of this paper was to find the optimal Value-at-Risk (VaR) method for estimating future exchange rate losses within one year. Design/methodology/approach - The analysis focuses on five VaR methods, some of them traditional and some of them more up to date with integrated EVT or GARCH. The analysis of VaR methods was concentrated on a time horizon (1–12 months), overestimation predictions and six scenarios based on trends and variability of exchange rates. This study used three currency pairs EUR/CZK, EUR/USD and EUR/JPY for backtesting. Findings - In compliance with the backtesting results, the parametric VaR with random walk has been chosen, despite its shortcomings, as the most accurate for estimating future losses in a medium-term period. The Nonparametric VaR confirmed insensitivity to the current exchange rate development. The EVT-based methods showed overconservatism (overestimation predictions). Every parametric or semiparametric method revealed a severe increase of liberality with increasing time. Research limitations/implications - This research is limited to the analysis of suitable VaR models in a long- and short-run period without using artificial intelligence. Practical implications - The result of this paper is the choice of a proper VaR method for the online application for estimating the future exchange rate for enterprises. Originality/value - The orientation of medium-term period makes the research original and useful for small- and medium-sized enterprises.