The long-awaited report from the Bank for International Settlements (BIS) on the banking book risk weighted asset (RWA) comparison1 was published earlier this month. The report includes the results of the bank’s Autumn 2012 bottom-up portfolio benchmarking exercise, an effort that compiled risk estimates on over 1,000 entities from 32 international IRB banks. This exercise is a key part of the global standard setter’s broader efforts to quantify the impact of Basel II and examine differences in its application across geographies.
We think the BIS should be applauded for moving beyond top-down portfolio analysis, and seeking to control for some (but not all) of the muddying factors. However we highlight certain implied conclusions that merit further discussion.
The BIS report contrasts two top-level findings: considerable agreement across banks as to the relative default risk of obligors in the hypothetical portfolio, but variation in the estimation of absolute risk. This implies that relative conservatism or aggressiveness in risk estimation is a function of bank-level biases as to overall risk levels, rather than differing views about individual obligors. The BIS conclude that this variation could result in banks’ capital ratios varying by as much as 1.5% to 2% around a 10% benchmark ratio. It is worth pointing out that the composition of the hypothetical portfolio makes this first conclusion almost inevitable. The BIS selected a list of large obligors, about which there is a substantial amount of publicly- available information, as a practical consideration to ensure enough overlap between contributors.2 However it is in portfolios for which there is less available information that one would expect more variation between banks with regards to relative default risk.
It is self-evident that in a framework such as Basel, the very goal of which is to reduce systemic risk by encouraging diversity of credit risk views, there will at any time be an even split between banks whose estimates appear more aggressive and those whose estimates appear more cautious than the average. We feel the report reflects a growing belief that diversity has somehow gone too far. If this is the BIS’s conclusion, a discussion is needed as to where the tipping point lies. Indeed, the BIS specifically comment, “The study did not attempt to identify an appropriate or acceptable level of variation of RWA in the banking book”3. Without such a discussion we forsee a race to zero on the diversity front, and our concerns are reinforced by the references in the report to supervisor-imposed risk benchmarks, floors and caps, and even fixed values. This risks negating the intent of Basel.
The report devotes a fair amount of space to comparing average risk weights under IRB to those that would be obtained using the standardized approach4. The data suggests a nuanced picture, with the sovereign portfolio revealing higher risk weights under IRB than under standardized. Conversely, corporate and bank portfolios have lower risk weights than the standardized approach. These results are intuitive given the migration of credit risk towards sovereigns through the financial crisis and illustrate a key strength of the IRB approach: its capacity to reflect cyclical developments in risk
The BIS’s conclusions are relatively muted. The main short-term recommendations are “enhanced disclosure and additional guidance”. We believe this reflects appropriate concern about the limitations of the data gathered. As with any snapshot, the distribution of individual datapoints may or may not reflect an ongoing trend: we just don’t know, and so the BIS are right to be cautious. What is rather predictable is that there will be more of these benchmarking exercises, which are highly resource-intensive for participants, often involve the submission of theoretical outputs, and where banks themselves have precious little access to outputs.
In our view, all of this strongly reinforces the case for banks and regulators to have access to robust ongoing benchmarking, where methodologies and data practices can be openly discussed and challenged using an extensive and representative dataset. If the BIS can leverage credit risk data accumulated by almost 8,000 risk analysts at the 32 participating banks, banks should be able to do so too. Credit Benchmark was founded to fill this information gap, and will provide an independent source of credit risk benchmarking data that reflects the views of those at the coal-face of risk management.