Elsevier UK Jobcode:RTF Chapter:CH06-H8304 22-3-2007 5:10p.m. Page:106 Trimsize:165×234MM
Fonts used:Ocean Sans & Sabon Margins:Top:42pt Gutter:54pt Font Size:10/12 Text Width:30p6 Depth:47 Lines
106 Risk management technology in financial services
by other mathematicians led into the non-Euclidean geometries of Carl Gauss and
From which Euclidean definitions, axioms and postulates are excluded, and
Only experimentally validated discoveries of universal physical principles are
Another significant contribution to the work of risk managers has been made by
Enrico Fermi (1901–1954), the famous physicist. Fermi asked his students at the
University of Chicago how many piano players are in the windy city. None knew the
answer. Fermi then demonstrated by successive approximation – taking into account
all factors that could impact on the number of piano players – that an order of
magnitude answer could be found.
The Fermi principle states that if the assumptions that we make do make sense,
then the errors which they possible contain become manageable.
Errors will not always be loading the results at the same side, therefore they will
more or less cancel themselves out and our estimate will have sense of balance.
The Fermi principle is very important to risk management, because the experts
controlling our bank’s exposure never really have all the data they need when a
decision is taken for damage control. The balance is provided through reasonable
assumptions. Auditors, incidentally, operate in a similar way.
Another method widely used in physics is estimation by analogy. This can be found
in several cases to underpin expert judgment. Its main aspect is the utilization of
references derived from events, for instance risks, with similar features and attributes,
in an environment not too different to the one under study.
The method by analogy typically produces a single estimate. For added value we
use bootstrapping as a simulation procedure that helps in calculating a confidence
interval for the resulting point estimate. The non-parametric approach is based on
empirical distribution of a historical data set without any assumption on the popula-
tion distribution that the information comes from.
In a parametric bootstrap, the computation of confidence intervals is based on
hypothesis which guarantees that the point estimate is mean value in a normal dis-
tribution. The downside is that when the sample is small or when the observations
come from different distributions, this approach does not mean much in terms of
6.3 Dissent, negation and reconstruction
Derived from the investigative activities of pure science, the scientific method has
been enriched over the years with management principles. Perception, cognition and
investigation of observed facts is a system effort; it is not like building a hill of sand,
where one grain falls upon another. Creative destruction lies squarely in the roots of
the scientific method – and finds its forces in dissent.