KeyBank 2013 Annual Report Download - page 94

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.
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 observed daily profit and loss. Results of backtesting are provided to the
Market Risk Committee. Backtesting exceptions occur when trading losses exceed VaR. Actual losses did not
exceed daily trading VaR on any day during the quarters ended December 31, 2013, and December 31, 2012.
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 $1.0 million at December 31, 2013,
and $1.2 million at December 31, 2012. The decrease in aggregate VaR was primarily due to reduced exposures
in credit derivatives as well as the change from using two years of historical data to one year for the VaR
simulation, which was partially offset by an increase in fixed income VaR. Figure 31 summarizes our VaR at the
99% confidence level for significant portfolios of covered positions for the three months ended December 31,
2013, and 2012.
Figure 31. VaR for Significant Portfolios of Covered Positions
2013 2012
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 $ 1.2 $ .5 $ .8 $ .6 $ 1.0 $ .1 $ .6 $ .5
Derivatives:
Interest rate $ .5$ .2$ .3$ .2$ .3$ .1$ .1$ .2
Foreign exchange .1 — — — .1 — — —
Credit .4 .1 .3 .1 1.6 .2 .9 .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 $2.9 million at December 31, 2013.
Figure 32 summarizes our stressed VaR for significant portfolios of covered positions for the three months ended
December 31, 2013, as used for market risk capital charge calculation purposes. Stressed VaR was not calculated
for market risk regulatory capital purposes during 2012.
Figure 32. Stressed VaR for Significant Portfolios of Covered Positions
2013
Three months ended December 31,
in millions High Low Mean December 31,
Trading account assets:
Fixed income $ 3.7 $ 1.4 $ 2.4 $ 1.7
Derivatives:
Interest rate $ 1.5 $ .5 $ 1.0 $ .5
Foreign exchange .2 — .1 —
Credit 1.2 .4 .8 .4
79