riskognize

interesting things on financial risk management

BCBS - Recognising the risk-mitigating impact of insurance in operational risk modelling

Click here to download:
bcbs181.pdf (197 KB)
(download)

In my work related to Advanced Measurement Approach (AMA) operational risk modeling, the use of insurance is always one of the first questions asked. 

Today, BCBS released a document for comments titled "Recognising the risk-mitigating impact of insurance in operational risk modelling". This aims to provide clarifications and bring the global practice into alignment.

Section 4 cover the important process of supervisors' assessment of coverage and alignment of insurance and banks' operational risk profiles. The paper also calls for independent review specifically of the use of insurance and how it is modelled, in order to qualify for capital reduction.

It also consider and reject the concept of 'experience requirement', which basically only allow banks to have in place the insurance for certain period of time before starting to recognize it as mitigation for regulatory capital. I agree with the paper's rationale that experience requirement is not that useful for the type of risks that insurance is meant to cover (low frequency, high impact) as banks will not learn much more about the nature of the insurance within the period.

In the end, it comes down to - as it often does - banks being able to demonstrate to, and convince the regulator that the method used are sound, and based on reasonable assumptions and convincing data.

On the criteria of recognizing insurance mitigation, the paper clarifies many of the criteria first specified in Basel II Accord, which, while cover many areas, still raises many questions for banks which start to seriously consider incorporating insurance into their AMA models. The most challenging (but reasonable) requirements is still the mapping of insurance cover to the risk profiles.

Predicting Bank Loan Recovery Rates with Neural Networks

Abstract: This study evaluates the performance of feed-forward neural networks to model and forecast recovery rates of defaulted bank loans. In order to guarantee that the predictions are mapped into the unit interval, the neural networks are implemented with a logistic activation function in the output neuron. The statistical relevance of explanatory variables is assessed using the bootstrap technique. The results indicate that the variables which the neural network models use to derive their output coincide to a great extent with those that are significant in parametric regression models. Out-of-sample estimates of prediction errors suggest that neural networks may have better predictive ability than parametric regression models, provided the number of observations is sufficiently large.
via defaultrisk.com

This seems like a straightforward application of neural network to predict LGD. As with any usage of neural network models, It'll be more difficult to get business buy-in. If nothing else, it could help suggest some risk factors overlooked by veteran modelers, who might have tendency to pick the old work horses of factors that worked well in the past.