Santosh,

I really don’t have much to say on the topic of alpha and CVaR. My main point would be that a simple beta to the market is likely to be insufficient to capture all the potential concentration risk. Some estimation based on a richer set of drivers, including possibly industry and regional indicators, is likely to be needed for a CVaR estimate to be justified in treating the idiosyncratic risk as reliably uncorrelated across different obligors. Given that we continue to argue about what alpha really encompasses, it doesn’t strike me as likely that we can model it effectively in a risk exercise such as CVaR.

Hi David,

Can you please elaborate more on feeding alpha into a full-blown credit VaR exercise

Concerning Santosh’s inquiry regarding EPE:

(For the uninitiated, “Alpha” in this instance is the ratio a bank must apply to its simulated Expected Positive Exposure (EPE) for derivatives to arrive at its loan equivalent amount. The idea is that the exposure uncertainty around the expected amount will increase the extreme loss potential compared to a static exposure of the same size.)

Own estimation of Alpha is quite a complex undertaking, since it involves simulating exposure and default simultaneously subject to supervisory review and approval of a bank’s assumptions, simulation methods and their implementation. It is likely that most banks could reduce regulatory capital somewhat by undertaking this exercise, since the mandated Alpha of 1.4 is really remarkably conservative in my experience. Despite that, I think the ancillary benefits are distinctly limited in this case. Unless this will feed into a full-blown credit VaR exercise, I don’t think it will contribute much to improving either strategic or tactical business decisions. Given the limited secondary benefits, I incline toward deciding on whether to attempt an own Alpha estimation effort based purely on the likely cost versus the perceived value of any prospective reduction in regulatory capital.

Hi David,

I full agree with your points that,

(a). CCR exercise should not be just a regulatory exercise (for that matter not just CCR, but the entire Basel II framework)

(b). Committee’s proposal is not to just reduce minimum regulatory capital, but to provide the risk sensitivity to CCR

I do also agree that only analytical approximations are not sufficient to measure the CCR, but the Quasi MC techniques. However, I feel building an MC engine is not a rocket science these days. By considering the benefits out of risk sensitiveness provided by effective EPE, Banks can always think of going for the Internal Models Method provided by the committee.

Can you please provide your thoughts on ‘own estimation of Alpha’ as part of the CCR measurement.

Santosh,

Lest my mild criticism of the Basel Committee for being late to the table on the CCR modeling issue be misinterpreted, I fully support the Committee’s move. I also think they have done a sensible job of fitting the dynamic modeling approach to CCR measurement into the larger regulatory capital scheme. My main point is that such measurement techniques should not be just a regulatory exercise. Indeed, the Basel use-test requires that the methods applied to calculating regulatory capital are rooted in the analysis used for internal risk measurement.

Beyond that, I feel the biggest benefit of the Committee’s proposal will not be to reduce minimum regulatory capital, as much as this is likely to be one effect. The core benefit is to discriminate much more accurately among various counterparty exposure amounts regardless of how simple or complex the bilateral portfolios may be. If a proper simulation approach is used for internal limit setting and exposure control, it is optimal for this to be the metric used on the trading desk to determine credit limit availability for a new trade. If it is applied globally, it has the further benefit of providing a sensible basis for a credit risk charge that affects traders’ price quotes.

For such a system to work effectively, certain analytic shortcuts are not just possible but essential. Only then will it be feasible to return metrics to traders in a timely fashion consistent with the necessary speed of decisions in many derivative markets. That said, the ever growing complexity of transactions makes exclusive reliance on analytic methods virtually impossible. Threshold changes in payments for such trades as range floaters and the complexity of modeling exposure to physically settled options beyond their exercise date make Monte Carlo simulation the ultimate benchmark against which purely analytic or hybrid results need to be measured.

Hi David,

Thanks again for your response!!!

Again, I am not fully convinced. The accord does not specify that Banks have to use only simulation based models for effective EPE estimation, but also provides the flexibility to use analytical approximation based models (please refer paragraph 27 of Annex 4 of June 2006 accord).

Next, may be in the olden days building a Mote Carlo engine would be expensive, however I feel with the advances in the technology and the algorithms like ‘Mersenne twisters’ are making the life easier.

So, the accord move towards CCR may be little late, but in my view it is a good beginning with the measures like effective EPE.

Please provide your views on the above points…

I agree with Santosh that banks should be able to receive some regulatory capital relief from the EPE approach and that recognition of collateral will contribute to this. I simply question whether this would be a sufficient reason, in and of itself, to spend the money to build or purchase a sophisticated future exposure simulation system. I think the key motivation for building such a system is, and always has been, to assess such credit exposure in a sensible and sophisticated way so as to make better day-to-day tactical credit decisions. Many banks still use MTM + add-ons as their approach to estimating exposure to individual counterparties. While this was arguably a defensible shortcut to estimating a total exposure figure for a full book, it was never adequate for individual counterparty assessments. It often gives hugely inconsistent results depending on the complexity of the trading patterns involved and often is DIRECTIONALLY incorrect relative to the exposure implications of a new trade.

While I’m at it, I should mention that the detailed run-off pattern takes on special importance in a collateralized portfolio. Often the impact of run-offs can have as much impact on potential unsecured exposure while waiting to receive new collateral as does market volatility. This in no way offsets the risk reduction implicit in using collateral. It is simply to say that it is easy to understate the residual unsecured exposure unless careful attention to the specific run-off pattern is addressed in the Monte Carlo simulations.

Hi David,

Thanks a lot for your thoughts on the CCR. I do agree with most of your thoughts; however also disagree with few of those. I do agree with your point of the committee being late on coming up with the internal models method for CCR guidelines. However, the internal models method (EPE) suggested by the committee provides benefits to the banks to move from Current Exposure Method to EPE because of the following reason:

EPE approach considers simulated Expected Exposures (EE) only up to one year, whereas the add-on in the Current Exposure Method (CEM) considers the entire life of an exposure. So, in most of the cases EPE provides lesser capital requirement than the Current Exposure Method (CEM)

Next point is that the approach suggested by the Basel II considers the simulated values of both exposure as well as collateral, which is sophisticated enough to measure the counterparty credit risk (CCR).

Please provide your thoughts on the above points…

Regarding Santosh’s question on counterparty credit risk (CCR), this is clearly a growing problem. The Basel Committee was very late to the table in addressing the issue. I even argue that the prolonged delay in allowing anything beyond mark-to-market plus add-ons as the method for calculating regulatory capital for CCR was an obstacle to banks trying to improve their systems for this important and growing source of risk. It was too easy to take the attitude that “If MTM plus add-ons is good enough for the banking regulators is should be good enough for us.”

While regulators need to push for as extensive transaction coverage as possible and for sophisticated simulation, I think it is a mistake to place too high a threshhold on the share of transactions that must be included to qualify for the expected positive exposure (EPE) approach to calculating regulatory capital for CCR. This runs the risk of discouraging banks from moving in this direction in the first place, especially if the main motivation is reducing the level of regulatory capital rather than improving risk management information and decision making. (See my July 2006 Risk magazine column Dangerous Perfection at: http://www3.sungard.com/SunGardFinancial/menus/documents/risk_managers/200607%20dangerous%20perfection.pdf )

A better approach would be to demand that transactions not included in the simulation analysis be modeled with conservative overrides and aggregated without the benefit of diversification effects.

Basel should be actively promoting a sophisticated simulation-based approach to CCR measurement rather than appearing to resist such an approach, which has often seemed to be the attitude in the past. Frankly, I think such an approach should be mandatory for major derivative trading houses and for others where CCR is a significant part of total credit risk.

Regarding Tim Trent’s comment, it is very clear that not enough is being done to assure data security on laptops. I also agree that fully encrypted data storage on all hard drives, including those on laptops, is an essential step. While a hardware solution may be the ultimate answer, there are minimally intrusive software solutions that could solve much of the problem if consistent deployment was backed up by serious employee sanctions for noncompliance. (For one example of such software see http://www.cypherix.com/ ) These solutions also would be much more cost effective (i.e. cheaper) than a wholesale replacement of hardware.

Apologies for the long silence. I was in the office for two days in early June but then was back on the road until June 22.

Regarding Santosh’s latest inquiry, it seems to me that the vital role of concentration in risk analysis has been generally accepted for over 50 years. The early work of Harry Markowitz on modern portfolio theory can be traced as far back as 1952. For all the controversy it engendered, the work of Markowitz and his co-Nobel Prize winners William Sharpe and Merton Miller drove home one crucial idea. Financial risk is inherently a portfolio concept. The risk of any individual position is meaningless in the abstract without reference to its portfolio context.

I would argue that lack of explicit treatment of concentration risk is the biggest shortcoming in the Basel II credit risk framework. By explicitly assuming broad diversification, the framework ignores this central source of risk, especially extreme tail risk. Careful treatment of diversification is a major difference between the Basel II framework and most internal credit risk portfolio models. Needless to say, this weakness of the Basel II requirements is not due to lack of understanding on the part of national supervisors or the Basel Committee. The problem is that the framework has been severely criticized for being too complex as it is. Introducing effective treatment of diversification involves a broad range of further issues that would magnify this existing complexity.

A logical expectation is that Basel III, when if emerges, will embody a significant move toward broad use of internal models. Clearly the Basel Committee felt that such models were not sufficiently advanced in 1999, when the Basel II Accord was first proposed, for them to proceed along the lines of the market risk amendment. (The situation was clearly different in the early 1990s when the Committee was considering amending the capital accord to cover market risk. The use of distributional analysis along the lines of Value-at-Risk was well advanced and widely practiced before the initial prescriptive proposal for treatment of market risk was proposed in 1993. Indeed, I argue it was this advanced state of internal practice that convinced the Committee to accept an internal models approach in April of 1995.) Now that the discussion around Basel II have dragged on for eight long years (and counting) credit portfolio models have advanced significantly in both analytical sophistication and in available market data to support objective analysis. It is very possible that if the Committee was starting discussions today, it might well consider inclusion of a far greater role for internal models. Doing so at this stage, however, would only extend application of the embarrassingly crude terms of Basel I still further.