302 The Basics of financial economeTrics
true. However, no volatility clustering model can claim the status of a law of
nature as all volatility clustering models fail to explain some essential fact.
It is often argued that profitable investment strategies can be based only
on secret proprietary discoveries. This is probably true but its importance
should not be exaggerated. Secrecy is typically inimical to knowledge build-
ing. Secrets are also difficult to keep. Industrial projects of a non military
nature are rarely based on a truly scientific breakthrough. They typically
exploit existing knowledge.
Financial econometrics is probably no exception. Proprietary tech-
niques are, in most cases, the application of more or less shared knowledge.
There is no record of major breakthroughs in finance made in secrecy by
investment teams of asset management firms. Some firms have advantages
in terms of data. Until the recent past, availability of computing power was
also a major advantage, reserved for only the biggest Wall Street firms; how-
ever, computing power is now a commodity. As a consequence, it is fair to
say that intuition in finance can be based on a vast amount of shared knowl-
edge plus some proprietary discovery or interpretation.
After using intuition to develop an ex ante hypothesis, the process of
model selection and calibration begins in earnest. This implies selecting a
sample free from biases and determining a quality-control methodology. In
the production phase, an independent risk control mechanism will be essen-
tial. A key point is that the discovery process should be linear. If at any point
the development process does not meet the quality standards, one should
resist the temptation of adjusting parameters and revert to developing new
intuition.
This process implies that there is plenty of intuition to work on when
dealing with the various issues in finance. The modeler must have many
ideas to develop. Ideas might range from the intuition that certain segments
of the financial market have some specific behavior to the discovery that
there are specific patterns of behavior with unexploited opportunities. In
some cases it will be the application of ideas that are well known but have
never been applied on a large scale.
A special feature of the model selection process is the level of uncer-
tainty and noise. Models capture small amounts of information in a vast
“sea of noise.” Models are always uncertain, and so is their potential lon-
gevity. The psychology of discovery plays an important role. These consider-
ations suggest the adoption of a rigorous objective research methodology. In
the next chapter we illustrate the work flow for a sound process of discovery
of profitable strategies.
A modeler working in financial econometrics is always confronted with
the risk of finding an artifact that does not, in reality, exist. And, as we
have seen, paradoxically one cannot look too hard at the data; this risks