Tải bản đầy đủ - 0 (trang)
25 If an unexpected loss occurs, the identification process may not be working correctly and should be reviewed

25 If an unexpected loss occurs, the identification process may not be working correctly and should be reviewed

Tải bản đầy đủ - 0trang

Risk Identification



75



r In addition to relatively standard market and credit risks, the identification process should

r

r

r



focus on risks that might exist in other areas; these might include settlement risk, hedge risk,

convergence/divergence risk, concentration risk, local market risk, model risk, cash flow

structure risk, and so on.

It is helpful to speculate on other positions or processes that might generate large losses in

the future; if any “new pockets” of risk are identified, they should be reviewed from a control

perspective.

When risks that reside outside of a specialist’s area of expertise are identified, they should

be forwarded to those with more direct responsibility.

If a loss from an unexpected source occurs, the identification process should be reviewed,

as it may indicate a flaw.



5

Risk Quantification and Analysis

We have noted at various points that the effective risk management process relies on tools and

skills from the quantitative and qualitative sectors; the quantification stage of the risk management process brings the two together to create the strongest possible risk framework. In order

to manage exposures that have been identified, risk and trading managers need to understand

the magnitude of the risks they are facing; this can be done by using financial mathematics

to create analytics that allow for proper quantification. In some cases the quantification effort

is simple and precise. For instance, in order to determine the interest rate risk of a Treasury

bond, traders and risk officers can make use of duration and convexity formulas, which produce very exact results. In other cases the exercise is more difficult, and may be subjective.

When dealing with diversified portfolios of risk, long-dated structures or complex cross-asset

derivatives, for instance, there may be no precise way of quantifying risk exposures. While

analytics can be created to generate price and risk estimates for such exposures, they are ultimately based on financial mathematics that depend on a variety of assumptions. Errors in the

financial mathematics, or underlying assumptions, give rise to “model risk,” and the potential

for model-related losses.

Firms employing portfolio or derivative valuation tools must understand the nature of the assumptions and limitations that impact quantitative processes, and factor them into the decisionmaking framework. For instance, VAR, which has emerged as a de-facto “industry standard”

for estimating the amount of market risk inherent in a portfolio, is based on numerous assumptions that can render the measure “error prone.” In implementing a VAR methodology a firm

makes certain assumptions related to the methodology — including the distribution of returns,

magnitude/direction of correlations, and so forth — which may, or may not, be credible. If

a firm assumes that financial returns are distributed normally when they are actually characterized by skewness and “fat tails,” disaster events will happen more often than predicted

and may be much larger than expected. In addition, if a firm assumes that the correlations

between risks in the portfolio are of a particular size, any substantive change during a market

dislocation will generate results that are different than those anticipated. Such concerns are not

limited to VAR models; the same is also true for other quantitative measures. Thus, while the

quantitative discipline is of vital importance in risk management, it must be approached with

full knowledge of limitations. By recognizing the potential shortcomings, a firm can calibrate

its processes and generate useful risk measures.



5.1 RISKS DISCOVERED IN THE IDENTIFICATION STAGES

SHOULD BE DECOMPOSED INTO QUANTIFIABLE TERMS;

THIS ALLOWS EXPOSURES TO BE CONSTRAINED

AND MONITORED

Once the identification of product or business risks has occurred, the risk officer can quantify

relevant risks; this ultimately permits establishment of limits and monitoring of exposures. We



78



The Simple Rules of Risk



have indicated that during the identification stage it is important not to overlook any risks;

this holds true in the quantification stage, where it is vital to decompose risks that have been

identified into quantifiable terms. Even the most complex structures, products and deals can be

decomposed into terms that allow for quantification. For instance, if a firm is preparing to underwrite a complicated project financing, the credit officer needs to dissect the cash flow streams

and structural enhancements to determine where the credit risk resides and how large the exposure might become under particular scenarios. The risk manager considering a new complex

derivative — perhaps a compound option that is “quanto’ed” (or exchanged) into a foreign

currency — needs to identify various market risk dimensions in the initial stage (delta, gamma,

cross-gamma, volatility, and so on) and can then quantify the effect each component has on the

deal (or portfolio of deals). The quantification stage is not trivial — it is a very rigorous process

that generally requires highly numerate risk officers with considerable analytic skills. However,

even the most complex product or deal can typically be dissected into quantifiable components.



5.2 THOUGH CERTAIN RISKS CAN BE DIFFICULT

TO QUANTIFY, BASIC ATTEMPTS AT MEASUREMENT

ARE IMPORTANT IN ORDER TO OBTAIN

AN INDICATION OF RISKINESS

Though many risks are relatively easy to quantify, some can be very challenging and difficult.

When a very complex quantification problem arises, attempts at quantification must still be

made — this provides a general level of riskiness and ultimately helps in the decision-making

process. In general, techniques for quantifying credit and market risks are quite well established.

Quantifying operational risk (i.e. risk related to failure of internal control processes/platforms

(e.g. fraud, error, disaster, non-functioning systems, and so forth) is far less established and often

more challenging. Most firms lack sufficient large loss data to properly model the behavior of

such risks (e.g. lack of data does not permit the collection of a representative sample, making

construction of a loss distribution difficult); accordingly, operational risks are often difficult

to quantify. Despite the relative complexity, creating a basic process to estimate losses over

time is an acceptable way of implementing a metric; the metric can then be enhanced and

modified as data becomes more robust or new techniques are developed. For instance, a firm

may wish to collect loss data from its own operating units and combine it with publicized loss

data from the industry at large. It may then construct a distribution of operating risk based on

high probability, low severity losses (e.g. internal losses) and enhance it by incorporating low

probability, high severity losses (e.g. internal or external losses) — this gives the distribution a

more realistic appearance, incorporating “fat tails” that can skew loss estimates. Thereafter it

can determine how much capital to apply to the mean and extreme points on the curve. While by

no means perfect, such a process represents an attempt to quantify exposure that is difficult to

measure and can give a firm important insight into the magnitude of the risk it might be facing.



5.3 MODELS ARE BASED ON ASSUMPTIONS THAT MAY,

OR MAY NOT, BE REALISTIC; ASSUMPTIONS, AND

THE IMPACT THEY CAN HAVE ON VALUATION,

MUST BE WELL UNDERSTOOD

Models are the financial cornerstone of trading and risk management — they make possible

deal pricing, exposure quantification, stress testing and risk management. There is no denying



Tài liệu bạn tìm kiếm đã sẵn sàng tải về

25 If an unexpected loss occurs, the identification process may not be working correctly and should be reviewed

Tải bản đầy đủ ngay(0 tr)

×