RBS 2012 Annual Report Download - page 407

RBS GROUP 2012
405
In determining whether an instrument is similar to that being valued, the
Group considers a range of factors, principally: the lending standards of
the brokers and underwriters that originated the mortgages, the lead
manager of the security, the issue date of the respective securities, the
underlying asset composition (including origination date, loan to value
ratios, historic loss information and geographic location of the
mortgages), the credit rating of the instrument, and any credit protection
that the instrument may benefit from, such as insurance wraps or
subordinated tranches. Where there are instances of market observable
data for several similar RMBS tranches, the Group considers the extent
of similar characteristics shared with the instrument being valued,
together with the frequency, tenor and nature of the trades that have
been observed. This method is most frequently used for US and UK
RMBS. RMBS of Dutch and Spanish originated mortgages guaranteed by
those governments are valued using the credit spreads of the respective
government debt and certain assumptions made by the Group, or based
on observable prices from Bloomberg or consensus pricing services.
The Group primarily uses an industry standard model to project the
expected future cash flows to be received from the underlying mortgages
and to forecast how these cash flows will be distributed to the various
holders of the RMBS. This model utilises data provided by the servicer of
the underlying mortgage portfolio, layering on assumptions for mortgage
prepayments, probability of default, expected losses and yield. The
Group uses data from third-party sources to calibrate its assumptions,
including pricing information from third party pricing services,
independent research, broker quotes, and other independent sources. An
assessment is made of third party data source to determine its
applicability and reliability. The Group adjusts the model price with a
liquidity premium to reflect the price that the instrument could be traded in
the market and may also make adjustments for model deficiencies.
The fair value of securities within each class of asset changes on a
broadly consistent basis in response to changes in given market factors.
However, the extent of the change, and therefore the range of reasonably
possible alternative assumptions, may be either more or less
pronounced, depending on the particular terms and circumstances of the
individual security. The Group believes that probability of default was the
least transparent input into Alt-A and prime RMBS modelled valuations
(and most sensitive to variations).
Commercial mortgage-backed securities (CMBS)
CMBS are valued using an industry standard model and the inputs,
where possible, are corroborated using observable market data.
Collateralised debt obligations (CDO)
CDOs purchased from third-parties are valued using independent, third-
party quotes or independent lead manager indicative prices. For super
senior CDOs which have been originated by the Group no specific third-
party information is available. The valuation of these super senior CDOs
therefore takes into consideration outputs from a proprietary model,
market data and appropriate valuation adjustments.
A collateral net asset value methodology using dealer buy side marks is
used to determine an upper bound for super senior CDO valuations. An
ABS index implied collateral valuation is also used to provide a market
calibrated valuation data point. Both the ABS index implied valuation and
the collateral net asset value methodology apply an assumed immediate
liquidation approach.
Collateralised loan obligations (CLO)
To determine the fair value of CLOs purchased from third parties, the
Group uses third party broker or lead manager quotes as the primary
pricing source. These quotes are benchmarked to consensus pricing
sources where they are available.
For CLOs originated and still held by the Group, the fair value is
determined using a correlation model based on a Monte Carlo simulation
framework. The main model inputs are credit spreads and recovery rates
of the underlying assets and their correlation. A credit curve is assigned
to each underlying asset based on prices from third party dealer quotes
and cash flow profiles, sourced from an industry standard model. Losses
are calculated taking into account the attachment and detachment point
of the exposure. Where the correlation inputs to this model are not
observable, CLOs are classified as level 3.
Other asset-backed and corporate debt securities
Where observable market prices for a particular debt security are not
available, the fair value will typically be determined with reference to
observable market transactions in other related products, such as similar
debt securities or credit derivatives. Assumptions are made about the
relationship between the individual debt security and the available
benchmark data. Where significant management judgment has been
applied in identifying the most relevant related product, or in determining
the relationship between the related product and the instrument itself, the
instrument is classified as level 3.
Equity shares
Private equity investments include unit holdings and limited partnership
interests primarily in corporate private equity funds, debt funds and fund
of hedge funds. Externally managed funds are valued using recent prices
where available. Where not available, the fair value of investments in
externally managed funds is generally determined using statements or
other information provided by the fund managers.
The Group considers that valuations may rely significantly on the
judgments and estimates made by the fund managers, particularly in
assessing private equity components. Given the decline in liquidity in
world markets, and the level of subjectivity, these are included in level 3.
Derivatives
Derivatives are priced using quoted prices for the same or similar
instruments where these are available. However, the majority of
derivatives are valued using pricing models. Inputs for these models are
usually observed directly in the market, or derived from observed prices.
However, it is not always possible to observe or corroborate all model
inputs. Unobservable inputs used are based on estimates taking into
account a range of available information including historic analysis,
historic traded levels, market practice, comparison to other relevant
benchmark observable data and consensus pricing data.