FRTB Internal Models Approach (IMA) - Part 1: Overview and IMA Desk Approval Process

There are two approaches that BASEL (FRTB) recommends for calculating capital charge associated with Risk Weighted Assets. One is SA i.e. Standardised Approach which we discussed in 4 part series (part 1, 2, 3, 4) earlier. It is well understood, that SA is more conservative approach and results into higher capital charge requirements. The other approach, which is subject to desk level calculation and approval is IMA i.e Internal Models Approach, which will make it a potentially less expensive option. The FRTB Internal Model Approach (IMA) introduces some changes from the old way prescribed in Basel 2.5 called Internal Model Method (IMM).

It is widely acknowledged though that most banks will end up having a hybrid approach i.e. some (IMA unapproved) desks will fall under SA while others that are IMA approved will fall under IMA approach. The aggregated capital charge for market risk i.e. ACC would be

ACC = Aggregated Charge for IMA Approved Desks + Aggregated Charge for IMA Unapproved Desks which is calculated via SA + Default Risk Charge

This is new as the old Basel 2.5 Internal Models Method (IMM) calculated market risk capital in terms of Value-at-Risk (VaR) and Stressed VaR, with backtesting of the model required in order to gain regulatory approval. The key changes that FRTB IMA approach has brought into this IMM (Internal Model Method) include:

• IMA approval at the trading desk level
• Backtesting and P&L Attribution Tests (PLAT)
• Expected shortfall (ES) measure instead of 99% VaR measure/SvaR
• Proof that the risk factors used in the model are derived from sufficiently liquid instruments (known as “modellability”)
• The old blanket 10-day liquidity horizon is replaced by longer and varied risk factor based liquidity horizons designed to reflect the time period required to sell or hedge a given position during stressed markets.
• Replacement of IRC i.e Incremental Risk Charge by DRC i.e. Default Risk Charge
• Punitive stressed capital add-on inclusion for non-modellable risk factors (“NMRF”).

The cost of having money available to cover potential losses for a trading desk is lower when using IMA instead of SA. But the bank can only use IMA if the regulator approves the trading desk for IMA. Next section will focus on the process and conditions for getting approval for IMA (while I will cover other changes that IMA brings in subsequent articles).

To be able to use the IMA method, a trading desk must pass both the Backtesting and P&L Attribution tests (PLAT).

BACKTESTING

In backtesting, they compare each trading desk's predicted risk of losing money in one day (i.e 1 day VaR) at the 97.5% and 99% confidence levels, with the actual profits and losses (Actual PnL as well as Hypothetical PnL) from the last year i.e past 250 trading days. Based on the number of exceptions in last 250 day trading period, the following backtesting zones (Green, Amber, Red) may be allotted to the bank by the regulator and it would result into shown impact.

At the trading desk level (there are no zones), they simply look at how much money a desk could lose in one day (1day VaR) with 97.5% and 99% confidence levels, and compare it with their actual profits and losses from the last year. If a desk loses more money than expected more than 12 times at the 99% level, or 30 times at the 97.5% level, they have to use Standardised Approach to calculate capital charge.

Why bother? Backtesting is a way to test how well a strategy or model would have worked in the past. By looking at how the strategy would have performed using historical data, traders and analysts can determine if it's a good strategy to use in the future. If it works well in backtesting, they can feel more confident using it in real-world situations.

PLAT

There are three ways of calculating profits and losses that banks use to get approval for their trading desks:

• Actual P&L (APL) - This measures the real profits and losses of a trading desk each day, not counting fees and commissions.
• Hypothetical P&L (HPL) - This calculates what the profits and losses would have been if the positions held by the desk at the end of the previous day were revalued using current market data.
• Risk Theoretical P&L (RTPL)- This is the profits and losses that a bank's risk measurement model predicts would happen.

In 2016 (old) version of PLAT, FRTB had prescribed calculating two ratios

• Ratio of Mean Unexplained daily P&L?(difference of Risk-Theoretical P&L and Hypothetical P&L) and the standard deviation of Hypothetical daily P&L (to be between -10% and 10%).
• Ratio of variances of Unexplained daily P&L and?Hypothetical daily P&L (to be less than 20%).

If the mean ratio falls outside the range of -10% to +10%, or if the variance ratio is more than 20%, the trading desk has done something wrong. If the desk breaks the rules four or more times in the past year, they fail this test and must thereby use SA (Standardised Approach), to calculate how much money they need to have available to cover potential losses. They can come back to the original way i.e IMA, if they follow the rules and have no exceptions for 12 months.

In 2019 (new revised version) however, they revised these metrics as it was too challenging to meet old requirements. These ratios/conditions around PLA were revised to what we now know them as 'Spearman Correlation Metric' and 'KS Metric'.

Note that there is still a conceptual linkage between old and new PLA test metrics (as identified in a white paper published by KPMG snippet of which is below)

Why bother? P&L attribution process provides an assessment of how well a desk’s risk management model captures risk factors that drive its P&L. The capital calculations are done through Risk Management models, and a significant divergence between Front Office models and the Risk models could mean banks holding insufficient capital relative to the market risk they have taken with their actual positions.?

Process Flow Diagram FRTB Approaches

In next article of the series, I will attempt to illustrate process beyond step III, both in logic as well as maths and computation.

#frtb?#investmentbanking?#capitalmarkets?#capitalcharge?#sba?#sensitivity?#riskmanagement?#risk?#basel?#bcbs?#rrao?#drc #ima #pla #plat #backtesting #approval #trading