TD Bank 2014 Annual Report Download - page 80

TD BANK GROUP ANNUAL REPORT 2014 MANAGEMENT’S DISCUSSION AND ANALYSIS78
Long-run PD estimates are generated by including key economic
indicators, such as interest rates and unemployment rates, and using
their long-run average over the credit cycle to estimate PD.
LGD estimates are required to reflect a downturn scenario.
Downturn LGD estimates are generated by using macroeconomic
inputs, such as changes in housing prices and unemployment rates
expected in an appropriately severe downturn scenario.
For unsecured products, downturn LGD estimates reflect the
observed lower recoveries for exposures defaulted during the recent
2008 to 2009 recession. For products secured by residential real estate,
such as mortgages and home equity lines of credit, downturn LGD
reflects the potential impact of a severe housing downturn. EAD esti-
mates similarly reflect a downturn scenario.
Non-Retail Exposures
In the non-retail portfolio, the Bank manages exposures on an individ-
ual borrower basis, using industry and sector-specific credit risk models,
and expert judgment. The Bank has categorized non-retail credit risk
exposures according to the following Basel counterparty types: corpo-
rate, including wholesale and commercial customers, sovereign, and
bank. Under the AIRB approach, CMHC-insured mortgages are consid-
ered sovereign risk and are therefore classified as non-retail.
The Bank evaluates credit risk for non-retail exposures by using both
a borrower risk rating (BRR) and facility risk rating (FRR). The Bank uses
this system for all corporate, sovereign, and bank exposures. The Bank
determines the risk ratings using industry and sector-specific credit risk
models that are based on internal historical data for the years of 1994-
2012, covering both wholesale and commercial lending experience. All
borrowers and facilities are assigned an internal risk rating that must be
reviewed at least once each year. External data such as rating agency
default rates or loss databases are used to validate the parameters.
Internal risk ratings (BRR and FRR) are key to portfolio monitoring
and management, and are used to set exposure limits and loan pricing.
Internal risk ratings are also used in the calculation of regulatory capi-
tal, economic capital, and incurred but not identified allowance for
credit losses. Consistent with the AIRB approach to measure capital
adequacy at a one-year risk horizon, the parameters are estimated
to a twelve-month forward time horizon.
Borrower Risk Rating and PD
Each borrower is assigned a BRR that reflects the PD of the borrower
using proprietary models and expert judgment. In assessing borrower
risk, the Bank reviews the borrower’s competitive position, financial
performance, economic and industry trends, management quality, and
access to funds. Under the AIRB approach, borrowers are grouped into
BRR grades that have similar PD. Use of projections for model implied
risk ratings is not permitted and BRRs may not incorporate a projected
reversal, stabilization of negative trends, or the acceleration of existing
positive trends. Historic financial results can however be sensitized to
account for events that have occurred, or are about to occur, such as
additional debt incurred by a borrower since the date of the last set of
financial statements. In conducting an assessment of the BRR, all rele-
vant and material information must be taken into account and the
information being used must be current. Quantitative rating models
are used to rank the expected through-the-cycle PD, and these models
are segmented into categories based on industry and borrower size.
The quantitative model output can be modified in some cases by
expert judgement, as prescribed within the Bank’s credit policies.
To calibrate PDs for each BRR band, the Bank computes yearly transi-
tion matrices based on annual cohorts and then estimates the average
annual PD for each BRR. The PD is set at the average estimation level
plus an appropriate adjustment to cover statistical and model uncer-
tainty. The calibration process for PD is a through-the-cycle approach.
Facility Risk Rating and LGD
The FRR maps to LGD and takes into account facility-specific character-
istics such as collateral, seniority ranking of debt, and loan structure.
Different FRR models are used based on industry and obligor size.
Where an appropriate level of historical defaults is available per model,
this data is used in the LGD estimation process. Data considered in the
calibration of the LGD model includes variables such as collateral cover-
age, debt structure, and borrower enterprise value. Average LGD and
the statistical uncertainty of LGD are estimated for each FRR grade. In
some FRR models, lack of historical data requires the model to output a
rank-ordering which is then mapped through expert judgement to the
quantitative LGD scale.
The AIRB approach stipulates the use of downturn LGD, where the
downturn period, as determined by internal and/or external experi-
ence, suggests higher than average loss rates or lower than average
recovery, such as during an economic recession. To reflect this, aver-
age calibrated LGDs take into account both the statistical estimation
uncertainty and the higher than average LGDs experienced during
downturn periods.
Exposure at Default
The Bank calculates non-retail EAD by first measuring the drawn
amount of a facility and then adding a potential increased utilization
at default from the undrawn portion, if any. Usage Given Default (UGD)
is measured as the percentage of Committed Undrawn exposure that
would be expected to be drawn by a borrower defaulting in the next
year, in addition to the amount that already has been drawn by the
borrower. In the absence of credit mitigation effects or other details,
the EAD is set at the drawn amount plus (UGD x Undrawn), where UGD
is a percentage between 0% and 100%.
Given that UGD is largely driven by PD, UGD data is consolidated by
BRR up to one-year prior to default. An average UGD is then calculated
for each BRR along with the statistical uncertainty of the estimates.
Historical UGD experience is studied for any downturn impacts,
similar to the LGD downturn analysis. The Bank has not found down-
turn UGD to be significantly different than average UGD, therefore the
UGDs are set at the average calibrated level, per BRR grade, plus an
appropriate adjustment for statistical and model uncertainty.
Credit Risk Exposures Subject to the Standardized Approach
Currently the Standardized Approach to credit risk is used primarily for
assets in the U.S. credit portfolio. The Bank is currently in the process
of transitioning this portfolio to the AIRB Approach. Under the
Standardized Approach, the assets are multiplied by risk weights
prescribed by OSFI to determine RWA. These risk weights are assigned
according to certain factors including counterparty type, product type,
and the nature/extent of credit risk mitigation. TD uses external credit
ratings, including Moody’s and S&P to determine the appropriate risk
weight for its exposures to sovereigns (governments, central banks,
and certain public sector entities) and banks (regulated deposit-taking
institutions, securities firms, and certain public sector entities).
The Bank applies the following risk weights to on-balance sheet
exposures under the Standardized Approach:
Sovereign 0%1
Bank 20%1
Residential secured 35% or 75%2
Other retail (including small business entities) 75%
Corporate 100%
1 The risk weight may vary according to the external risk rating.
2 35% applied when loan-to-value <=80%, 75% when loan-to-value >80%.
Lower risk weights apply where approved credit risk mitigants exist.
Loans that are more than 90 days past due receive a risk weight of
either 100% (residential secured) or 150% (all other).
For off-balance sheet exposures, specified credit conversion factors
are used to convert the notional amount of the exposure into a credit
equivalent amount.