404 2019 Universal registration document and annual financial report - BNP PARIBAS
5 risks and CaPital adequaCy Pillar 3
5
Market risk
➤ TABLE 79: VALUE AT RISK (1-DAY, 99%) [Audited]
In millions of euros
Year to 31 December 2019 Year to 31 December 2018
Minimum(**) Average Maximum(**) Last
measure Average Last
measure
Interest rate risk 12 19 29 24 17 20
Credit risk 8 11 17 12 11 10
Foreign exchange risk 3 7 16 6 7 9
Equity price risk 7 10 18 9 15 17
Commodity price risk 2 4 7 3 4 6
Netting effect(*) (27) (30) (30) (32)
TOTAL VALUE AT RISK 17 24 33 24 25 30
(*) Note that the minimum and maximum figures shown above for the various risk types are computed on a standalone basis (i.e. independently from each other as well as the total VaR). While the minimum or maximum for each risk type may not necessarily be observed on the same date, minimum/ maximum netting effects are not considered relevant.
(**) For minima and maxima, total VaR cannot be read as the sum of VaR by risk type.
VaR (1 day, 99%) fell in 2019 given the substantial reduction in the Group s sensitivity to market volatility, lower stock market volatility and the termination of the proprietary trading activities of Opera Trading Capital. It was low throughout 2019 and in May it hit its lowest level in ten years.
Backtesting the VaR
RISK continuously tests the accuracy of its internal model through a variety of techniques, including in particular a regular comparison over a long-term horizon between actual daily losses on capital market transactions and one-day VaR.
This backtesting consists of making a comparison between the daily global trading book VaR and the actual result except the, fees and commissions. In accordance with the regulation, BNP Paribas supplements this actual
backtesting method with a comparison between the daily VaR and the hypothetical result generated by the trading book, which is also known as hypothetical backtesting . The hypothetical result includes all components of the actual result, calculated on the previous day s positions, only incorporating changes in market parameters. A backtesting event is declared when a real or hypothetical loss exceeds the daily VaR amount. The confidence interval selected for calculating daily VaR is 99%, which in theory means the observation of two to three events per year.
The number of events is calculated at least quarterly and is equal to the highest of the number of excesses for the hypothetical and actual variations in the portfolio value.
