AIG 2014 Annual Report Download - page 261

ITEM 8 / NOTE 5. FAIR VALUE MEASUREMENTS

244
Sensitivity to Changes in Unobservable Inputs
We consider unobservable inputs to be those for which market data is not available and that are developed using the best
information available to us about the assumptions that market participants would use when pricing the asset or liability.
Relevant inputs vary depending on the nature of the instrument being measured at fair value. The following is a general
description of sensitivities of significant unobservable inputs along with interrelationships between and among the significant
unobservable inputs and their impact on the fair value measurements. The effect of a change in a particular assumption in the
sensitivity analysis below is considered independently of changes in any other assumptions. In practice, simultaneous changes
in assumptions may not always have a linear effect on the inputs discussed below. Interrelationships may also exist between
observable and unobservable inputs. Such relationships have not been included in the discussion below. For each of the
individual relationships described below, the inverse relationship would also generally apply.
Corporate Debt
Corporate debt securities included in Level 3 are primarily private placement issuances that are not traded in active markets or
that are subject to transfer restrictions. Fair value measurements consider illiquidity and non-transferability. When observable
price quotations are not available, fair value is determined based on discounted cash flow models using discount rates based
on credit spreads, yields or price levels of publicly-traded debt of the issuer or other comparable securities, considering
illiquidity and structure. The significant unobservable input used in the fair value measurement of corporate debt is the yield.
The yield is affected by the market movements in credit spreads and U.S. Treasury yields. In addition, the migration in credit
quality of a given security generally has a corresponding effect on the fair value measurement of the security. For example, a
downward migration of credit quality would increase spreads. Holding U.S. Treasury rates constant, an increase in corporate
credit spreads would decrease the fair value of corporate debt.
RMBS and Certain CDO/ABS
The significant unobservable inputs used in fair value measurements of RMBS and certain CDO/ABS valued by third-party
valuation service providers are constant prepayment rates (CPR), loss severity, constant default rates (CDR), and yield. A
change in the assumptions used for the probability of default will generally be accompanied by a corresponding change in the
assumption used for the loss severity and an inverse change in the assumption used for prepayment rates. In general,
increases in CPR, loss severity, CDR, and yield, in isolation, would result in a decrease in the fair value measurement.
Changes in fair value based on variations in assumptions generally cannot be extrapolated because the relationship between
the directional change of each input is not usually linear.
CMBS
The significant unobservable input used in fair value measurements for CMBS is the yield. Prepayment assumptions for each
mortgage pool are factored into the yield. CMBS generally feature a lower degree of prepayment risk than RMBS because
commercial mortgages generally contain a penalty for prepayment. In general, increases in the yield would decrease the fair
value of CMBS.
CDO/ABS – Direct Investment book
The significant unobservable inputs used for certain CDO/ABS securities valued using the BET are recovery rates, diversity
score, and the weighted average life of the portfolio. An increase in recovery rates and diversity score will increase the fair
value of the portfolio. An increase in the weighted average life will decrease the fair value.