KeyBank 2015 Annual Report Download - page 95

are customized for specific covered positions, and numerous risk factors are incorporated in the calculation.
Additional consideration is given to the risk factors to estimate the exposures that contain optionality features,
such as options and cancelable provisions. VaR is calculated using daily observations over a one-year time
horizon, and approximates a 95% confidence level. Statistically, this means that we would expect to incur losses
greater than VaR, on average, five out of 100 trading days, or three to four times each quarter. We also calculate
VaR and stressed VaR at a 99% confidence level.
The VaR model is an effective tool in estimating ranges of possible gains and losses on our covered positions.
However, there are limitations inherent in the VaR model since it uses historical results over a given time interval
to estimate future performance. Historical results may not be indicative of future results, and changes in the
market or composition of our portfolios could have a significant impact on the accuracy of the VaR model. We
regularly review and enhance the modeling techniques, inputs and assumptions used. Our market risk policy
includes the independent validation of our VaR model by Key’s Risk Management Group on an annual basis.
The Model Risk Management Committee oversees the Model Validation Program, and results of validations are
discussed with the ERM Committee.
Actual losses for the total covered positions did not exceed aggregate daily VaR on any day during the quarters
ended December 31, 2015, and December 31, 2014. The MRM backtests our VaR model on a daily basis to
evaluate its predictive power. The test compares VaR model results at the 99% confidence level to daily held
profit and loss. Results of backtesting are provided to the Market Risk Committee. Backtesting exceptions occur
when trading losses exceed VaR.
We do not engage in correlation trading, or utilize the internal model approach for measuring default and credit
migration risk. Our net VaR approach incorporates diversification, but our VaR calculation does not include the
impact of counterparty risk and our own credit spreads on derivatives.
The aggregate VaR at the 99% confidence level for all covered positions was $1.2 million at December 31, 2015,
and $.9 million at December 31, 2014. The increase in aggregate VaR was primarily due to the increased
exposure in our credit derivative portfolio, and the composition of our fixed income portfolio. Figure 32
summarizes our VaR at the 99% confidence level for significant portfolios of covered positions for the three
months ended December 31, 2015, and December 31, 2014. During this period, none of our significant portfolios
daily trading VaR numbers exceeded their VaR limits or stress VaR limits.
Figure 32. VaR for Significant Portfolios of Covered Positions
2015 2014
Three months ended December 31, Three months ended December 31,
in millions High Low Mean December 31, High Low Mean December 31,
Trading account
assets:
Fixed income $ 1.0 $ .4 $ .6 $ .5 $ .5$ .3$ .4$ .4
Derivatives:
Interest rate $.1 —$ .1$.1$.3 —$ .1$.1
Credit .4 $ .2 .3 .4 .3 $ .1 .2 .3
Stressed VaR is calculated using our general VaR results at the 99% confidence level and applying certain
assumptions. The aggregate stressed VaR for all covered positions was $3.5 million at December 31, 2015, and
$2.6 million at December 31, 2014. Figure 33 summarizes our stressed VaR for significant portfolios of covered
positions for the three months ended December 31, 2015, and December 31, 2014, as used for market risk capital
charge calculation purposes.
81