Gathering portfolio information from source systems

Thus far, we have taken a bottom-up approach to the mapping problem; starting
with the fundamental assets, we have been progressively expanding the range of
assets that can be incorporated into the VaR calculation. At this point, we switch to
a more top-down approach, which is closer to the point of view that is needed in
implementation. Mathematically speaking, the VaR calculation, at least in the form
presented here, is relatively trivial. However, gathering the data that is needed for
this calculation can be enormously challenging. In particular, gathering the portfolio
information required for the mapping process can present formidable problems.
Information on the positions of a financial institution is typically held in a variety of
heterogeneous systems and it is very difficult to gather this information together in
a consistent way.
We recommend the following approach. Each source system should be responsible
for providing a description of its positions in terms of DEAFs. The VaR calculator is
then responsible for converting DEAFs to exposure vectors for the fundamental
assets. The advantage of this approach is that the DEAFs provide a well-defined,
simple, and stable interface between the source systems and the VaR computational
engine. DEAFs provide an unambiguous financial specification for the information
that source systems must provide about financial instruments. They are specified in
terms of basic financial-engineering abstractions rather than system-specific concepts,
thus facilitating their implementation in a heterogeneous source-system environment.
This specification is effectively independent of the particular selections made
for the fundamental assets. Since the latter is likely to change over time, it is highly desirable to decouple the source–system interface from the internals of the risk
system. Changes in the source–system interface definitions are expensive to implement,
as they require analysis and changes in all of the source systems. It is therefore
important that this interface is as stable as possible.
In the presence of heterogeneous source systems, one reasonable approach to
gathering the necessary information is to use a relational database. In the following
two subsections, we outline a simple two-table design for this information.