Tuesday, June 14, 2011

Back testing and Benchmarking Part.II



Used properly, they can calibrate (or recalibrate) rating models and thus risk tools for the granting of products or assessing customer behavior, in particular by responding to the following problems:

* Expectations of risk of the entity are they consistent with the materialization of this risk (making)? This article seeks to back test the risk parameter ex post with the estimated parameter ( back testing, diagram below).
* What is the position of the entity namely entities instead?
* Expectations of risk of the entity they are consistent with those instead? Benchmarking of the estimated internally with the external reference system can respond to this question (benchmarking model).
* Achievement of risk is they consistent with those instead? (Benchmarking loss or damage).
Answering these questions provides a comprehensive framework for monitoring and supervision mechanism of risk.


If the qualitative interpretation and the decision process can be automated, the industrialization of data flow and processing (collection, data storage and consolidation of often disparate information from divergent SI) should allow an optimal analysis results and a realization of recurrent exercises. The aim is to produce a more reliable and suitable reporting activity and minimizing its cost both in terms of budget and time.

In the context of benchmarking, the nomenclature of internal data is a priori different from the external data device, comparing the risk parameters on a common nomenclature is essential. The mapping is then to map data sources through the adoption of strict rules and documented. In practice we retain the system that will optimize the granularity of the correspondence between the two sources to minimize the loss of information.

Element's overarching Back testing and benchmarking, performance feedback must meet three key points:

* Restitution little difficult in terms of statistics and mathematics, to make operational results and convert analyzes corrective actions;
* Flexibility in handling, in reviewing and reading the results;
* Annexes detailing the key elements of the analysis.

Back testing a draft and / or transverse Benchmarking is because it requires the participation of several entities of the bank. Indeed, to be valid, corrective action affecting a parameter of risk must be decided collectively between committees at the central management of risks, crafts, commercial and financial management.

The challenge of such a project is not only to ensure the accuracy and consistency of rating systems and procedures and the estimation of risk factors, but also to develop and propose a real risk management tool and decision support.