JP Morgan Chase 2003 Annual Report Download - page 69

J.P. Morgan Chase & Co. / 2003 Annual Report 67
There are also groups that report to the Chief Financial Officer
w ith some responsibility for market risk-related activities. For
example, within the Finance area, the valuation control functions
are responsible for ensuring the accuracy of the valuations of
positions that expose the Firm to market risk.
Positions that expose the Firm to market risk are classified into tw o
categories. Trading risk includes positions held as part of a business
w hose strategy is to trade, make markets or take positions for the
Firm’s ow n trading account; gains and losses in these positions are
reported in Trading revenue. Nontrading risk includes mortgage
banking positions held for longer-term investment and positions
used to manage the Firm’s asset/liability exposures. In most cases,
unrealized gains and losses in these positions are accounted for
at fair value, w ith the gains and losses reported in Net income or
Other comprehensive income.
Tools used to measure risks
Because no single measure can reflect all aspects of market risk,
the Firm uses several measures, both statistical and nonstatistical,
including:
• Statistical risk measures
- Value-at-Risk (“ VAR” )
- Risk identification for large exposures (“ RIFLE” )
• Nonstatistical risk measures
- Economic-value stress tests
- Net interest income stress tests
- Other measures of position size and sensitivity to market
moves
Value-at-Risk
JPM organ Chase’s statistical risk measure, VAR, gauges the
potential loss from adverse market moves in an ordinary market
environment and provides a consistent cross-business measure
of risk profiles and levels of risk diversification. VAR is used to
compare risks across businesses, to monitor limits and to allo-
cate economic capital to the business segments. VAR provides
risk transparency in a normal trading environment.
Each business day, the Firm undertakes a comprehensive VAR
calculation that includes both trading and nontrading activities.
JPM organ Chase’s VAR calculation is highly granular, comprising
more than 1.5 million positions and 240,000 pricing series (e.g.,
securities prices, interest rates, foreign exchange rates). For a
substantial portion of its exposure, the Firm has implemented
full-revaluation VAR, w hich, management believes, generates
the most accurate results.
To calculate VAR, the Firm uses historical simulation, w hich
measures risk across instruments and portfolios in a consistent,
comparable w ay. This approach assumes that historical changes
in market value are representative of future changes. The simu-
lation is based on market data for the previous 12 months.
The Firm calculates VAR using a one-day time horizon and a
99% confidence level. This means the Firm w ould expect to
incur losses greater than that predicted by VAR estimates only
once in every 100 trading days, or about 2.5 times a year.
All statistical models involve a degree of uncertainty, depending
on the assumptions they employ. The Firm prefers historical
simulation, because it involves fewer assumptions about the
distribution of portfolio losses than parameter-based method-
ologies. In addition, the Firm regularly assesses the quality of
the market data, since their accuracy is critical to computing
VAR. Nevertheless, because VAR is based on historical market
data, it may not accurately reflect future risk during environ-
ments in w hich market volatility is changing. In addition, the
VAR measure on any particular day may not be indicative of
future risk levels, since positions and market conditions may
both change over time.
While VAR is a valuable tool for evaluating relative risks and
aggregating risks across businesses, it only measures the poten-
tial volatility of daily revenues. Profitability and risk levels over
longer time periods – a fiscal quarter or a year – may be only
loosely related to the average value of VAR over those periods.
First, w hile VAR measures potential fluctuations around average
daily revenue, the average itself could reflect significant gains or
losses; for example, from client revenues that accompany risk-
taking activities. Second, large trading revenues may result from
positions taken over longer periods of time. For example, a busi-
ness may maintain an exposure to rising or falling interest rates
over a period of w eeks or months. If the market exhibits a long-
term trend over that time, the business could experience large
gains or losses, even though revenue volatility on each individual
day may have been small.
VAR Backt esting
To evaluate the soundness of its VAR model, the Firm conducts
daily backtesting of VAR against actual financial results, based
on daily market risk–related revenue. M arket risk–related revenue
is defined as the daily change in value of the mark-to-market
trading portfolios plus any trading-related net interest income,
brokerage commissions, underw riting fees or other revenue.
The Firm’s definition of market risk–related revenue is consistent
w ith the Federal Reserve Board’s implementation of the Basel
Committee's market risk capital rules. The histogram below
illustrates the Firm’s daily market risk–related revenue for trading
businesses for 2003. The chart show s that the Firm posted posi-
tive daily market risk–related revenue on 235 out of 260 days
in 2003, with 170 days exceeding $25 million. Losses w ere sus-
tained on 25 days; nine of those days were in the third quarter,
primarily driven by poor overall trading results. The largest daily
trading loss during the year w as $100 million.