The Geometric-VaR Backtesting Method On Cryptocurrencies

This paper investigates a recent value at risk (VaR) backtest that uses both the duration between two consecutive VaR violations and the value of the VaR: the geometric-VaR test from Pelletier & Wei (2016). Investigation is done by replicating two tables, reporting the size and power of the test, with Monte Carlo simulation. This test is then used to evaluate VaR estimates from five cryptocurrencies. Four methods are used to estimate VaRs with RiskMetrics providing the ’best’ VaR estimates for three cryptocurrencies.