Bank of Montreal 2013 Annual Report Download - page 86

‰ capital and stress testing models for measuring capital, allocating
capital and managing regulatory capital and Economic Capital;
‰ fiduciary models for asset allocation, asset optimization and portfolio
management;
‰ major business strategy models to forecast the possible outcomes of
new strategies in support of our business decision-making process; and
‰ models driven by regulatory and other stakeholder requirements.
Model Risk is governed by the enterprise-wide Model Risk Manage-
ment Framework, which sets out end-to-end risk governance across the
model activity cycle and ensures consistency between model risk and
enterprise-wide risk appetite. The framework outlines explicit principles
for managing model risk, describes processes and clearly defines roles
and responsibilities. The Model Risk Corporate Standard, outlines the
requirements for the oversight, identification, development,
independent validation, implementation, use, monitoring and reporting
of models and model risk throughout the enterprise. Prior to use, all
models must receive approval and an assessment of their model risk by
the Model Risk and Vetting (MRV) group. All models are assigned a risk
rating as part of the vetting process, which determines the frequency of
ongoing review. In addition to regularly scheduled model validation and
vetting, model risk monitoring and oversight activities are in place to
confirm that models perform and are managed and used as expected,
thereby increasing the likelihood of early detection of emerging issues.
The Model Risk Management Forum, a cross-functional group repre-
senting all key stakeholders (model users, model owners and the MRV
group), meets regularly to provide input into the development,
implementation and maintenance of the Model Risk Management
Framework and the requirements governing all models that are used
across the enterprise.
BMO’S Risk Rating System Framework
The Risk Rating System framework utilized by BMO encompasses
various methods, processes, controls, data collection and technology to
support the assessment of credit risk of exposures. This framework also
includes the assignment of the following credit risk parameters: Proba-
bility of Default (PD), Loss Given Default (LGD) and Exposure at Default
(EAD), which are used for Regulatory Capital and Economic Capital
estimation. The principles underlying the Risk Rating System are gov-
erned by internal policies and standards.
The design and quantification of models and methodologies to
establish credit risk measures is a centralized function. An independent
validation group reviews, validates and approves these models and
methodologies prior to their implementation.
Ongoing monitoring of model performance, targeted model
reviews, annual validations and related reporting processes ensure that
the models and methodologies continue to perform as intended, and
that any material changes in operating environment, business strategy
that leads to portfolio shifts, or economic environment trigger appro-
priate and timely action. These processes are key to ensuring that BMO’s
risk rating systems continue to assign risk parameters that accurately
reflect credit risks in our various portfolios.
We employ risk rating systems for our retail portfolios (consumer
and small business) and wholesale portfolios (corporate, commercial,
bank and sovereign).
Retail Risk Rating System
Retail Risk Rating System uses an approach that rates the borrower’s
risk on a narrow range of likely expected conditions, primarily more
recent in nature (e.g. delinquency, loan to value ratio, utilization rate,
etc.). Product lines within each of the three retail risk categories
— mortgage, qualifying revolving, and other retail exposures — are
separately modelled so the risk drivers capture the distinct nature of
each product. The final segmentation scheme categorizes each exposure
within a product line into homogeneous pools of retail risk that reflect
common borrower risk drivers. Accordingly, each risk segment is then
assigned a unique combination of PD, LGD and EAD parameters,
capturing the segment-specific credit risk.
The retail risk rating system is designed to estimate values of credit
risk parameters as precisely and accurately as possible. However, the
risk parameter estimates are subject to uncertainty. In order to embed a
level of conservatism to portfolio performance projections, adjustments
are added to each parameter estimate at the segment level during the
calibration process. Additionally, the retail parameters are calibrated on
an annual basis to incorporate additional data points in the parameter
estimation process. This ensures that the most recent experience is
incorporated into the parameter assessment process.
Parameter Modelling Details (all are expressed as percentages,
between 0% and 100%)
PD: assigned to each borrower and reflects default risk over a one-year
time horizon. The PD parameter is calibrated based on BMO’s internal
default data from the period 2003 to 2012 and is meant to reflect long-
run average default rates.
LGD: assigned to each credit exposure extended to a borrower and
measures the potential economic loss at default during downturn con-
ditions. The LGD parameter is calibrated based on internal loss data from
2003 to 2011, including a specific “downturn” buffer that incorporates
the potential impact of PD and LGD correlation, and floored to the
maximum realized loss given default rate.
EAD: assigned to each exposure extended to a borrower and measures
the amount of a credit exposure that is likely to be drawn in the event
the borrower defaults. This EAD amount is derived from the EAD ratio (or
utilization given default) parameter and is based on BMO’s internal
realized loss data from 2003 to 2012. The EAD ratio parameter is cali-
brated with a long-run view, based on the average of historical realized
utilization given default rates, with a margin of conservatism added for
sources of uncertainty and to ensure the predicted EAD amount is
greater than the maximum historical realized EAD amount.
Wholesale Risk Rating System
Wholesale Risk Rating system covers the assessment of credit risk of
borrowers in non-retail asset classes (corporate, bank, and sovereign).
Relative to Retail portfolio, Wholesale portfolio is characterized by a
smaller number of larger exposures that cover a range of industries. Risk
characteristics of these borrowers are captured by developing industry-
specific risk rating models, and LGD and EAD modelling focuses on
capturing the key risk drivers of individual facility types extended to
these borrowers. Further details on each are provided below.
Parameter Modelling Details (all are expressed as percentages,
between 0% and 100%)
PD: assigned to each borrower based on its risk rating and the asset
class and reflects default risk over a one-year time horizon.
Risk ratings are assigned using the appropriate internal model. A
suite of general and sector-specific risk rating models have been devel-
oped within each asset class to capture the key quantitative and qual-
itative risk factors associated with borrowers in different industries and
portfolios. Borrower risk rating grades (BRRs) are assessed and assigned
at loan inception and reviewed at least annually. More frequent reviews
are performed for higher risk-rated borrowers, accounts that trigger a
review through a rating change or that experience covenant breaches,
and accounts requiring or requesting changes to facilities.
BMO employs a Master Scale with 14 BRRs, and for each grade
within each asset class, grade PDs are assigned to reflect the long-run
average of one-year default rates. PD estimates are based on internal
default experience over a period of more than five years that covers at
least one full economic cycle, supplemented by external benchmarking,
as applicable.
MD&A
BMO Financial Group 196th Annual Report 2013 97