Finding nth ordered loss (percentile) in historical simulation (HS) VaR
This is a review of three methods for finding the 95th percentile Value at Risk (VaR) in historical simulation (HS). In my opinion, none are wrong, all are valid. Please let me know if you disagree? I personally prefer Kevin Dowd's approach because I think it is the most elegant fit to the expression "5% of the time we expect the loss to exceed the VaR."
(spreadsheet at our website)