Standardising risk or the risk of standardisation?

If adopted and implemented globally, these proposals would have broad ramifications across the banking world, including fundamental shifts in how banks view and price risk in their businesses. The two documents are inextricably entwined.

"The intricate web of measures might look complex (but), they actually sacrifice sophistication in the management and measurement of risk."
Brad Carr, Brad Carr, Senior Policy Adviser, Institute of International Finance

The concept of the Capital Floor would mean that where some banks are currently accredited by their regulators (such as APRA, the Australian Prudential Regulation Authority) to use sophisticated risk models, they would be subject to new ‘floors’ based on the Standardised Approach to capital, hence the links.

Such is the scope of the change implied by the mix of a new capital floor, revisions to the Standardised Approach and the introduction of Basel 3’s Leverage Ratio, that some suggest it represents a shift to “Basel 4” or at least “Basel 3+”.

The issue is while the intricate web of measures might look complex, they actually sacrifice sophistication in the management and measurement of risk, particularly for the larger banks that are of greater systemic importance.

Far from continuing the evolution of risk models, this would unfortunately make banks’ key performance measures and strategic decision-making less sensitive to their underlying risk. Rather than Basel 3+, more akin to ‘Basel 1.5: Back to the Future’.

Standardisation and Simplification

The Basel Committee has been careful to not pre-empt the level of its proposed floor (and has canvassed feedback on the floor’s mechanics) but it is significant it looks to use the Standardised Approach as its reference point.

The Standardised Approach has been the domain of banks with smaller balance sheets or narrower business models. The new floor proposal means it suddenly comes into play as a major (and potentially predominant) factor in the capital requirements for all banks, regardless of size or complexity.

This means we need to look closely at the specifics within that approach. Under the status quo, the Standardised Approach recognises assessments from credit ratings agencies for loans to those sovereign, institutional and corporate entities that are externally rated.

For unrated exposures, there is no variability for individual borrowers’ risk characteristics – all corporate and SME loans are assumed to have the exact same level of risk, as are all mortgages. As such, the current Standardised Approach resembles the Basel 1 conditions that prevailed pre-crisis. There is no real discrimination even though we know the risk characteristics can be radically different.

The Basel Committee’s proposed revisions to the Standardised Approach would make a number of changes, firstly by removing the use of external credit ratings altogether. This move is motivated by the widespread criticism of ratings agencies’ assessments of structured securitisations and some sovereigns during the crisis, notwithstanding the fact their ratings of banks and corporates were mostly vindicated.

Each class of borrower would instead have a series of blunt risk-weights applicable, via a simple matrix.

For instance, for corporates and SMEs, the applicable risk-weighting on each loan would be determined by the combination of their size (expressed in terms of their amount of revenue) and their leverage, as follows:

{CF_IMAGE}

Whereas banks’ internal models consider a multitude of variables in assessing each corporate’s creditworthiness, this approach prescribes a highly simplified approach with just those two variables. Such a simplification overlooks factors such as industry type, profit margins, operating cost-base, exposure to volatility and management capability.

It is also debateable whether the enterprise’s scale is really one of the top two determinants of risk, and if it is, whether revenue is the best measure of that scale.

Similarly for mortgages, risk-weightings would be determined by a simple matrix of each mortgage’s Loan to Valuation Ratio (LVR) and the Debt Service Coverage ratio (DSC), as follows:

{CF_IMAGE}

The striking thing in this scenario is while there is a reasonably high sensitivity to the relative property value, there is a very small sensitivity to the borrower’s means to repay the loan. Indeed, to the extent that there is any sensitivity, there is a pronounced cliff effect, with the only variation coming as we move from one side of the 35 per cent DSC mark to the other.

A risk-sensitive approach needs to include more factors and to do so with more granularity. This comes back to why the status-quo applies the Standardised Approach for smaller banks with simpler business models and why larger banks have been expected to invest in (and then use) more sophisticated modelling of their risk.

Moreover, besides the excessive simplification, the choice of variables, the lack of sensitivity and the cliff effects, these matrices will still have the challenges of international inconsistencies, most notably with variations in accounting rules meaning leverage is calculated differently in some countries.

Likewise, the underlying risk in mortgage portfolios is influenced by national treatments in local laws on recourse to the borrower beyond the property and taxation factors such as whether interest is deductible.

Over-riding Risk-Based Models
The Basel Committee’s proposals have been prompted by calls from some quarters for simpler approaches to bank capital measurements and greater consistency in banks’ internal risk models - notwithstanding the underlying variances between jurisdictions.

It’s true the variance between banks’ models does need to be reduced but some variance is warranted. Models give simulated estimates, there is no single correct view of risk – not least because banks have different loan origination practices and criteria, different loan portfolio histories and concentrations, and different policies and approaches to managing and collecting on defaulted borrowers.

The IIF Risk Weighted Asset Task Force has identified 78 recommendations of specific instances where banks’ credit risk modelling practices could be harmonised. Implementing these recommendations would narrow the variances in banks’ models significantly, to just those specific factors that relate to the legitimate differences between banks’ risk management approaches.

But while some convergence is needed, complete standardisation brings its own dangers.

Whereas a diversification in risk measurement reduces the correlation of losses among banks by encouraging them to take different risks, if a single standardised risk measurement methodology is established as the binding constraint on all banks, it steers them into the same loan portfolios, increasing systemic risk.

This makes the calibration of a floor absolutely critical. If a floor is set at a moderate level, it can be a useful complement in detecting and over-riding extreme outliers, without impeding risk-sensitive measures. But if calibrated too high, it becomes the binding constraint and over-rides accurate and sensitive views of risk for all.

Notably, the effect of a floor is felt on banks’ lower risk and short term exposures. The blunter profile of standardised risk-weights means they will commonly over-state the risk on stronger credits, whilst under-stating riskier lending, as follows:

{CF_IMAGE}

Consequently, if a floor is calibrated at a higher level and it does become the binding constraint on a bank, the incentive can be to move out along the credit curve, shifting the lending profile towards riskier assets.

A Better Way Forward
A better approach would be to enhance banks’ models and harmonise many of their modelling assumptions and practices, converging outcomes whilst maintaining risk-based models as the key driver of banks’ capital requirements.

Particular models are more accurate and reliable in some areas than others. But where there are concerns about specific modelling parameters – such as where there is limited history of past defaults for particular borrower types – regulators can impose specific constraints, rather than over-riding all modelling on a broad-brush basis.

For example, Jeremy Rudin, head of Canadian regulator OSFI, noted recently:

“Our view is that the rules on capital requirements should be based primarily on risk; however, we must be aware of the limitations of risk-measuring models…Therefore, we are prepared to consider new restrictions on the use of models when the data on severe losses are limited”

Meanwhile, there are a lot of areas where banks’ models have proven effective and a recent Moody’s review, Proposed Bank Rating Methodology, September 9, 2014, supported the predictive performance of RWA based on banks’ internal models through the crisis. In those cases, models make sense.

We must be clear about what we want: less complexity but not at the risk of over-simplifying the measurement of risk. We don’t want to compromise risk-sensitivity or generate false incentives.

Brad Carr is a senior adviser at the Institute of International Finance.