Goldman Sachs 2012 Annual Report Download - page 91

Management’s Discussion and Analysis
Value-at-Risk
VaR is the potential loss in value of inventory positions due
to adverse market movements over a defined time horizon
with a specified confidence level. We typically employ a
one-day time horizon with a 95% confidence level. We use
a single VaR model which captures risks including interest
rates, equity prices, currency rates and commodity prices.
As such, VaR facilitates comparison across portfolios of
different risk characteristics. VaR also captures the
diversification of aggregated risk at the firmwide level.
We are aware of the inherent limitations to VaR and
therefore use a variety of risk measures in our market risk
management process. Inherent limitations to VaR include:
‰VaR does not estimate potential losses over longer time
horizons where moves may be extreme.
‰VaR does not take account of the relative liquidity of
different risk positions.
‰Previous moves in market risk factors may not produce
accurate predictions of all future market moves.
When calculating VaR, we use historical simulations with
full valuation of approximately 70,000 market factors.
VaR is calculated at a position level based on
simultaneously shocking the relevant market risk factors
for that position. We sample from 5 years of historical data
to generate the scenarios for our VaR calculation. The
historical data is weighted so that the relative importance of
the data reduces over time. This gives greater importance to
more recent observations and reflects current asset
volatilities, which improves the accuracy of our estimates of
potential loss. As a result, even if our inventory positions
were unchanged, our VaR would increase with increasing
market volatility and vice versa.
Given its reliance on historical data, VaR is most effective in
estimating risk exposures in markets in which there are no
sudden fundamental changes or shifts in market conditions.
Our VaR measure does not include:
‰positions that are best measured and monitored using
sensitivity measures; and
‰the impact of changes in counterparty and our own credit
spreads on derivatives, as well as changes in our own
credit spreads on unsecured borrowings for which the fair
value option was elected.
Model Review and Validation
Our VaR model is subject to review and validation by our
independent model validation group at least annually. This
review includes:
‰a critical evaluation of the model, its theoretical
soundness and adequacy for intended use;
‰verification of the testing strategy utilized by the model
developers to ensure that the model functions as
intended; and
‰verification of the suitability of the calculation techniques
incorporated in the model.
Our VaR model is regularly reviewed and enhanced in
order to incorporate changes in the composition of
inventory positions, as well as variations in market
conditions. Prior to implementing significant changes to
our assumptions and/or model, we perform model
validation and test runs. Significant changes to our VaR
model are reviewed with the firm’s chief risk officer and
chief financial officer, and approved by the Firmwide
Risk Committee.
We evaluate the accuracy of our VaR model through daily
backtesting (i.e., comparing daily trading net revenues to
the VaR measure calculated as of the prior business day) at
the firmwide level and for each of our businesses and major
regulated subsidiaries.
Stress Testing
We use stress testing to examine risks of specific portfolios
as well as the potential impact of significant risk exposures
across the firm. We use a variety of stress testing techniques
to calculate the potential loss from a wide range of market
moves on the firm’s portfolios, including sensitivity
analysis, scenario analysis and firmwide stress tests. The
results of our various stress tests are analyzed together for
risk management purposes.
Sensitivity analysis is used to quantify the impact of a
market move in a single risk factor across all positions (e.g.,
equity prices or credit spreads) using a variety of defined
market shocks, ranging from those that could be expected
over a one-day time horizon up to those that could take
many months to occur. We also use sensitivity analysis to
quantify the impact of the default of a single corporate
entity, which captures the risk of large or
concentrated exposures.
Goldman Sachs 2012 Annual Report 89