Pitfalls in backtesting Historical Simulation VaR models

► We evaluate the evaluation of Value-at-Risk HS and FHS forecasts. ► We theoretically show that the popular unconditional backtest is inconsistent. ► Our results imply that multiplication factors for capital requirements are downward biased. ► A new backtesting procedure is proposed, with good power properties. ► An empirical application illustrates and confirms our theoretical results.