the concept of integrated risk management encompassing
all risk categories has become fashionable, but with little
understanding of how this should be achieved in practice.
.WZ;M`IUXTM;;\PMZM;Q[;VW;ZMTQIJTM;UWLMT;;TM\;ITWVM;LI\I;;WV;
the correlation of operational risk with other risk categories.
Thus, in the absence of any better method or understanding, simple summation is generally proposed and applied.
While the “thou-shalt-integrate” instinct is fundamentally
correct, integration should be done in a proper and logical
way. A close examination of most OR systems shows that
these systems are not risk analysis systems in the traditional
sense of the word. The principal aim of OR systems is to
eliminate or reduce risk, and therefore they are similar to
UIVIOMUMV\;\WWT[;_Q\P;IV;MUXPI[Q[;WV;_WZSAEW_;IVL;KWV-
\ZWT;; ;]KP; \W WT[; \aXQKITTa; LMÅVM; XZWKM[[M[; IVL; QLMV\QNa; SMa;
QVLQKI\WZ[;;_PQKP;UMI[]ZM;\PMQZ; Y]ITQ\a;;.QZM;XZM^MV\Q WV;;NWZ;
M`IUXTM;; ZMY]QZM[; IUWVO; W\PMZ; \PQVO[; XTIKQVO; ÅZM; M`\QV-
guishers at critical locations, making sure that they function
correctly and that staff are trained to use them properly.
<PM; KTI[[QÅKI\QWV; WN ; 7:; [W]ZKM[; ¸; SMMXQVO; \ZIKS; WN ;
various remedial or preventive actions and measuring their
effectiveness — is a core task of OR systems. It is indeed
sensible to invest available managerial skills and resources in
operational loss prevention, whereas modeling operational
losses stochastically is meaningless.
Experience shows that managerial skills and techniques
PI^M;\PMQZ;TQUQ\[;;IVL;KW[\;JMVMÅ\;KWV[QLMZI\QWV[;\PMZMNWZM;
U][\;IT[W;XTIa;I;ZWTM;;.QO]ZM;;;;IJW^M;;ZQOP\;;[PW_[;\PM;UIZ-
ginal operational loss given a certain level of resources required to mitigate it and the corresponding avoidance cost.
As the resource level increases, the total loss decreases (at
least ideally), but at a marginally decreasing rate. On the
other hand, the marginal avoidance cost increases at an optimal point, where the two lines intersect. Investing fewer
resources than the optimum leads to higher losses, while
dedicating more resources is not worthwhile, since the cost
of avoiding a loss is higher than the loss itself.
1V; .QO]ZM;;;; \PM; L WUIQV; TMN\; WN ; \PM; WX\QU]U; Q[; UIVIOML;
_Q\P;\PM;_WZSAEW_;\WWT[;NW]VL;QV;\aXQKIT;7:;[a[\MU[;; 1\;Q[;KWU-
monly known as qualitative operational risk — or, as we prefer
to describe it, the operational part of operational risk. To the
right of the optimum is the residual loss after resources are optimized. This is known as quantitative operational risk — or, as
we would also call it, the risk part of operational risk.
While the operational part has nothing to do with risk management proper, the risk part is comparable to market, credit or
,);;;;;;;I@JB;GIF=<JJ@FE8C;;;; < : < D 9 < I ; ) ' ' 0
=`^li\;+ 1;Fgk`dXc;DXeX^\i`Xc;<]]fik;]fi;
; Fg\iXk`feXc;I`jb
Cost
Loss
Avoidance
Optimum
Operation part/Qualitative
Resources
Risk part/Quantitative
insurance risk. It is also the element that must be integrated into
IV; W^MZIT T; ZQ[S; NZIUM_WZS;; *I[MT; K WZZMK\Ta; LMÅVM[; WXMZI\Q WVIT;
risk in this sense as the risk of loss resulting from inadequate or
failed internal processes, people and systems, or from external
events. This, however, should not take the focus from the main
function of operational risk management: the avoidance such
risks from the start.
How, then, should the risk part of operational risk be inte-
OZI\ML;QV\W;IV;W^MZITT;-: 5;NZIUM_WZS';<PM;TWOQKIT;ÅZ[\;[\MX;Q[;
to examine the loss database, where loss events (accidents, etc.)
and the subsequent cash payments are recorded. This information can be used in two different ways:
1. For deriving key indicators. The number of losses
and their severity is an indicator of the effectiveness of resources dedicated to operational loss prevention. Too many losses
may be an indication that the resources spent on avoiding risk
are left of the optimum.
2. For ERM input. A frequency (how often a loss event
occurs) and severity (the cost of a loss event) distribution can be
empirically derived for each loss category.
While key indicators are helpful in determining the managerial optimum, frequency and severity are the key factors that
TQVS; WXMZI\QWVIT; ZQ[S; UIVIOMUMV\; \W; QV\MOZI\QWV;; .ZMY]MVKa;
and severity are also used in non-life insurance for risk modeling; the only distinction is that non-life insurance additionally
works with loss triangles;, which provide information about
how payments are expected over time. Ideally, the loss database
would also contain value and liquidity data and could be used
as a basis for statistical analysis — but in practice this is not the
case. Anecdotal evidence suggests that this is due to the low
quality of frequency and severity information, where adding
