TD Bank 2014 Annual Report Download - page 85

TD BANK GROUP ANNUAL REPORT 2014 MANAGEMENT’S DISCUSSION AND ANALYSIS 83
VaR is a valuable risk measure but it should be used in the context
of its limitations, for example:
• VaR uses historical data to estimate future events, which limits
its forecasting abilities;
• it does not provide information on losses beyond the selected
confidence level; and
• it assumes that all positions can be liquidated during the holding
period used for VaR calculation.
The Bank continuously improves its VaR methodologies and incorporates
new risk measures in line with market conventions, industry best prac-
tices, and regulatory requirements. During 2014, the Bank implemented
a modification to improve volatility risk modeling in VaR calculations.
To mitigate some of the shortcomings of VaR, the Bank uses addi-
tional metrics designed for risk management and capital purposes.
These include Stressed VaR, Incremental Risk Charge, Stress Testing
Framework, as well as limits based on the sensitivity to various market
risk factors.
Calculating Stressed VaR
In addition to VaR, the Bank also calculates Stressed VaR, which
includes Stressed GMR and Stressed IDSR. Stressed VaR is designed
to measure the adverse impact that potential changes in market rates
and prices could have on the value of a portfolio over a specified
period of stressed market conditions. Stressed VaR is determined using
similar techniques and assumptions in GMR and IDSR VaR. However,
instead of using the most recent 259 trading days (one year), the Bank
uses a selected year of stressed market conditions. In the fourth quar-
ter of fiscal 2014, Stressed VaR was calculated using the one-year
period that began on February 1, 2008. The appropriate historical
one-year period to use for Stressed VaR is determined on a quarterly
basis. Stressed VaR is a part of regulatory capital requirements.
Calculating the Incremental Risk Charge
The incremental risk charge (IRC) is applied to all instruments in the
trading book subject to migration and default risk. Migration risk
represents the risk of changes in the credit ratings of the Bank’s
exposures. TD applies a Monte Carlo simulation with a one-year
horizon and a 99.9% confidence level to determine IRC, which is
consistent with regulatory requirements. IRC is based on a “constant
level of risk” assumption, which requires banks to assign a liquidity
horizon to positions that are subject to IRC. IRC is a part of regulatory
capital requirements.
(millions of Canadian dollars) 2014 2013
As at Average High Low As at Average High Low
Interest rate risk $ 5.3 $ 5.8 $ 12.8 $ 3.3 $ 3.2 $ 9.7 $ 19.2 $ 2.9
Credit spread risk 4.9 6.3 8.8 3.9 6.0 6.0 10.9 2.4
Equity risk 5.1 3.7 9.6 1.5 2.5 3.6 8.8 1.8
Foreign exchange risk 1.6 2.7 5.5 0.7 1.7 1.4 5.8 0.3
Commodity risk 0.9 1.4 4 0.6 0.5 0.9 2.3 0.4
Idiosyncratic debt specific risk 13.6 15.8 20.5 12.1 14.2 16.5 23.6 11.3
Diversification effect1 (16.1) (17.8) n/m2 n/m2 (12.8) (18.8) n/m2 n/m2
Total Value-at-Risk $ 15.3 $ 17.9 $ 22.1 $ 14.2 $ 15.3 $ 19.3 $ 26.9 $ 13.7
Stressed Value-at-Risk (one day) 29.3 27.8 36.1 21.1 27.6 32.0 44.3 22.4
Incremental Risk Capital Charge (one year) 275.6 313.6 428.7 222.0 185.6 267.9 369.6 177.6
PORTFOLIO MARKET RISK MEASURES
TABLE 55
Average interest rate risk VaR decreased by $3.9 million compared to
the prior year due to reduced interest rate risk positions. Improvement
in the quality of data underlying the idiosyncratic debt specific model
introduced during 2013 coupled with a reduction in Canadian provin-
cial bond positions in the second quarter of 2014 decreased average
Stressed VaR compared with the prior year by $4.2 million. Larger
U.S. Agency and financial bond positions increased average IRC by
$46 million to $314 million compared to the prior year.
Validation of VaR Model
The Bank uses a back-testing process to compare the actual and theo-
retical profit and losses to VaR to ensure that they are consistent with
the statistical results of the VaR model. The theoretical profit or loss is
generated using the daily price movements on the assumption that
there is no change in the composition of the portfolio. Validation of
the IRC model must follow a different approach since the one-year
horizon and 99.9% confidence level preclude standard back-testing
techniques. Instead, key parameters of the IRC model such as transi-
tion and correlation matrices are subject to independent validation by
benchmarking against external study results or through analysis using
internal or external data.
Stress Testing
The Bank’s trading business is subject to an overall global stress test
limit. In addition, global businesses have stress test limits, and each
broad risk class has an overall stress test threshold. Stress scenarios
are designed to model extreme economic events, replicate worst-case
historical experiences, or introduce severe but plausible hypothetical
changes in key market risk factors. The stress testing program includes
scenarios developed using actual historical market data during periods
of market disruption, in addition to hypothetical scenarios developed
by Risk Management. The events the Bank has modeled include the
1987 equity market crash, the 1998 Russian debt default crisis, the
aftermath of September 11, 2001, the 2007 ABCP crisis, and the credit
crisis of Fall 2008.
Stress tests are produced and reviewed regularly with the Market
Risk and Capital Committee.
1 The aggregate VaR is less than the sum of the VaR of the different risk types due
to risk offsets resulting from portfolio diversification.
2 Not meaningful. It is not meaningful to compute a diversification effect because
the high and low may occur on different days for different risk types.