The Pixar Way: 37 Quotes on Developing and Maintaining a Creative Company (fr...
Misleading financial modelling and risky spreadsheets (sept. 2016)
1. Misleading financial modelling and risky spreadsheets
(sept. 2016, compl. jan. 2017)
(Jean-Luc Marsat)
Among criticisms directed to globalisation, and subsequently to the global financial system, those
dealing with the ‘’no-control no-limit’’ philosophy, resulting in deregulation and immoderate
openness, appear to be the most significant ones to light up obvious weaknesses.
Miscellaneous features, including unanticipated containment in case of emergency (e.g. world
market crash) and no-track record of securities transactions, have quite likely contributed to the
proliferation of damages in the 2007-8 crisis.
Those damages lay mainly upon erroneous risk assessments contained in financial models being
widely used and diffused, providing results or predictions eventually known as wrong but till then
used, and in their turn reused as inputs for other models.
Not all models did amplify the trouble. Actually tools for financial modelling can be distributed
into two main categories :
- the most common one, and universally used, is the spreadsheet, almost exclusively worked out
thanks to MS Excel
- another one comprises all software programs, tools and extensions written for a provisional
purpose, possibly resorting to simulation programs (Matlab, Simulink), customer-tailored, more
IT-minded (as per method and efficiency rules) and more or less proprietary software.
Concerned troubles are concentrated in the first category. One must admit that spreadsheet
models are quite often of insufficient quality, and pave a « kingly route » to pandemic spreading
of fall-downs, as it occured in 2008.
These issues have already emerged and been discussed, not only by consulting and training firms
focussed on Financial modelling (F1F9, Corality, ...), but also by such a well-known academic
author as Nassim Nicholas Taleb, in the famous essay ‘The Black Swan. The Impact of the Highly
Improbable’ (2007).
All agree on that point : so soothing might be the appearance, as powerful as software programs
they seem to be, spreadsheet models can lead to great harm.
But another point should be right now emphasized : spreadsheets ARE actually software
products.
They are easy to learn and to build, and users don’t feel they are handling objects of the same
kind as ‘‘usual‘‘ software (i.e. code lines written in ad-hoc programming languages with their own
grammar rules), but the true nature of these objects is software too.
In fact problems and induced risks are similar ones in both categories. Those troubles identified
in spreadsheets occur also in ‘‘usual‘‘ IT, e.g. :
. hard-coding,
. code structure to be improved
. scratch testing (if any),
. bad or poor documentation (this problem is not specific to IT : it can be experienced also with
calculation documents attached to common industrial equipment such as steam turbines).
Of course, spreadsheets have their specific problems too (dependant cells, long chain formulas
hardly understood due to non-structured display, ...).
The core matter is that spreadsheet modelling demands the same high-level quality as other IT
solutions, but fails to meet the requirements because spreadsheets are not viewed as
risk-generating items. They seem harmless.
_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
‘Misleading financial modelling and risky spreadsheets’ (J-Luc Marsat, sept.’16-jan.’17) 0126.2336 1
2. Nevertheless, just as time and effort is required to master languages and concepts of ‘‘usual‘‘
software, to set up working rules and methods, spreadsheet modelling requires the same quite
as much and needs to be cautiously designed and operated.
Spreadsheets are not just tools for amateurs, they truly are professional means of IT engineering.
Now, some other issues should also be tackled while not often mentioned in papers or just
reminded, but having a true impact on quality :
- What about the reliability of financial functions in Excel ?
If we can’t access and have a look on the (‘closed’) code underlying them (MS property ...), we can’t
check whether this code is properly written (by MS) or not (nobody is obliged to trust the
Microsoft company and to give them any blessing in advance).
Besides, financial functions are poorly documented through F1 Help : lack of precision,
wordiness, at such a point that the mathematical expression (using formal Descartes’s notation)
is not even given ...
- Is it sensible using an interpreted language (VBA) to write macros ?
B in ‘VBA’ stands for Basic ... When we attended our first IT courses (long ago ...), most of us wrote
one’s first progam in the simplest language then available : Basic !
To-day, even for macros and routines, one could expect a more efficient and powerful way of
coding.
This is where lies an advantage of LibreOffice/OpenOffice over Excel : the possibility of using
Python rather than VBA.
- Should it not be wise to institute the position of ‘‘model manager‘‘ ?
Ensuring the coherency and consistency of the whole spreadsheet, centralizing all modifications
or improvements, updating the attached documentation would thus be assigned to a single
person and performed without disrupture in time or quality.
Finally, the best practice in spreadsheet modelling is not a mere matter of training (though
essential), it also involves job organisation inside businesses, selection of appropriate tools,
quality-oriented procedures (review, audit), to give the essentials.
The hereabove comments are exclusively given with regard to the IT angle. Of course troubles
which have occurred in the past and were aimed at by authors like Taleb are not restricted to this
domain.
Software quality is a prerequisite, but having it in mind should not wipe off other shortcomings,
common to all types of financial models and more specific to their financial core, such as :
superficial knowledge about securities under process, rough study of cascade phenomena, ... (to
be completed from any available field experience !).
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
2 ‘Misleading financial modelling and risky spreadsheets’ (J-Luc Marsat, sept.’16-jan.’17) 0126.2336