A technique used to test how well (or badly) the Value at Risk (VaR) estimates would have performed using historical data. For example, in a 10-day 99% VaR, back testing would investigate how often the loss in 10 days surpasses the 10-day 99% VaR that would have been computed for that period. If the loss occurred on no more than 1% of such periods, the methodology of calculating VaR can be considered reasonably acceptable. If the loss happened 5% of similar periods, the methodology shall be deemed malfunctioning.
November 13, 2023