Revised Model-Based Market Risk Rules Costly for Banks – Fitch

The overhaul of the internal models approach – used by most banks with large trading books to calculate market risk capital requirements – will be costly, says Fitch Ratings. The Basel Committee on Banking Supervision’s revised market risk framework, published in January and effective from 2019, fundamentally changes the approach.

The model revisions should improve risk assessment capabilities, lead to higher capital charges for hard-to-model trading positions and make it easier to compare banks’ results. But the model approval process and governance are being thoroughly revised and implementing the changes will require considerable investment in technology and risk management.

Banks will need to obtain approval for internal models desk by desk, rather than bank-wide. This will make it easier for supervisors to decline approval for a particular trading desk, if, for example, the desk is unable to satisfy model validation criteria due to back-testing failures or an inability to properly attribute profits and losses across products. But Fitch thinks costs associated with building and running the more sophisticated models will be high.

Instead of running a single bank-wide model for a range of stressed and unstressed risk factors, multiple new models will need to be built, validated and run daily. This will multiply the number of model reviews and operational runs and add to subsequent data analysis and reporting procedures. Additional risk personnel will be required for review, oversight, and reporting purposes.

The amount of regulatory capital models-based banks will need to cover potential market risks following the revisions is uncertain. The Basel Committee’s latest studies show that, for a sample of 12 internationally active banks with large trading books, all of which provided high-quality data, market risk capital charges under the revised approach were 28% higher. But for a broader sample of 44 banks using internal models, the median market risk capital requirements fell by 3% under the revised models.

Fitch thinks the result for the 12 banks could reflect greater concentrations of less liquid credit positions that require more capital, or larger trading positions lacking observable transaction prices, which are subject to a stressed capital add-on. Banks facing higher charges under the regime may re-assess whether certain activities remain profitable.

The new internal models approach replaces value at risk (VaR) with an expected shortfall (ES) measure. VaR does not capture the tail risk of loss distribution, which can arise during significant market stress. The use of ES models for regulatory capital is positive for bank creditors because they could lead to better capitalisation of tail-risk loss events and might motivate risk managers to limit trading portfolios that could lead to outsized losses.

When calculating ES measures, banks will have to use variable market liquidity horizons – to a maximum of 120 days for complex credit products, against the current fixed 10-day period. We think model inputs will be more realistic, by acknowledging that some instruments take longer to sell or hedge without affecting prices. ES will also constrain recognition of diversification and hedging benefits, extensively used in VaR models to reduce capital charges. We think this will make model outputs more prudent and force banks to better capitalise potential trading losses.

Structural flaws in the way banks calculated capital charges for market risk were exposed during severe market stresses in 2008-2009. The Basel Committee subsequently undertook a fundamental review of the trading book. The original proposals were watered down, but we think the final revised minimum capital standards for model-driven market risk are positive for creditors because improved model standards and more prudent methods employed to capture risk should mean trading risks are more accurately capitalised.

Author: Martin North

Martin North is the Principal of Digital Finance Analytics

Leave a Reply