Open Source Margin Models
Today’s article by Risk on getting CCPs to share their margin models brings up an interesting similarity to the Linux approach to software. The Linux approach is for all the software to be publicly available, for any programmer to review and comment on, or even update and return to the public domain. The philosophy is that only full scrutiny of the code will eliminate security & reliability problems, even by letting potential hackers see the barriers they face, and to ‘crowd source’ the problem.
Given that the move to clear a large portion of the OTC market, we are therefore putting our trust in VaR like never before. What if each CCP were obliged to publish their risk management techniques, such as their Initial Margin (IM) models in public? What would be some outcomes of each CCP putting their margin model in public?
- Competition from other CCPs: Being able to replicate the IM model would enable any other CCP to look for competitive strategies to compete with the leading providers. If the IM model at competing CCPs were no longer a differentiator, then the other layers of protection and the corporate structure of the CCP would become more important. Would we see a homogenous market for clearing, where only national interests prevail?
- Regional / local spread: The current debate on whether each country should own the clearing of it’s own ‘local OTC market’, would be fuelled by the ease with which proven IM models could be copied. Currently the intellectual property underpinning not just the VaR model, but the specific parameters, scenario inputs, and approach to stress testing are not available in packaged software, which might prove possible under this ‘open source’ scenario.
- Public scrutiny: Whilst the models currently in use are subject to review by the members of each CCP, the models aren’t (to my knowledge) widely or regularly reviewed outside of the CCP and it’s membership group. Would this scrutiny benefit clearing? Much like anti-virus activity amongst vendors, increased knowledge of the technology used to measure and manage risk, might lead to wider debate and understanding of the strengths and limitations of VaR at a CCP. Perhaps debate in public with feedback to politicians is what’s needed to help them understand that centralising OTC products into a small group of CCPs isn’t necessarily a better systemic approach than currently.
- Misunderstanding: The application of VaR is complex, and beyond explanation to a typical BBC audience (although seeing Robert Peston try would be interesting). It wouldn’t be surprising if the explanation of how VaR is applied in context of a CCP were to be misrepresented, as has happened with other parts of the OTC market infrastructure. What the world doesn’t need right now is a lack of confidence in CCPs given the fragility of the world economy.
- Respect: With wider understanding might come increased respect for the role CCPs play – after all they stand between the markets and disaster. They are sometimes characterised as ‘dusty utility’ organisations, yet are operating (in one case anyway) the single largest concentration of a product with (probably) the biggest market risk simulation, on the planet.
A caveat to all this, is that the model for Initial Margin (VaR or otherwise) is only one part of the chain of protection in each CCP. You have to take into account the basics such as the settlement infrastructure, the acceptance and management of collateral assets, the market stress testing, the default waterfall, and the corporate financial and legal structure, to begin to really build a full picture of how protection is achieved.
Given that most of the emerging CCPs for OTC products haven’t yet processed a large scale default (such as Lehman), we don’t know how well the new market infrastructure will stand up to a real test. After all, users of the Lehman Prime Brokerage services thought they would receive their collateral assets back should Lehman default – they’re still waiting for the bankruptcy process to make it’s way to a conclusion.
Whilst making each CCPs IM model ‘open source’ would provide insight into one component of the protection mechanisms – copying or replicating the entire model would be a much more complex task. It seems to me that the main outcome would be a better understanding of the tools used to provide protection, and that the commercial success of CCPs would be unaffected, given that amongst the experts the techniques used to measure and manage risk are well understood across the CCP market.
Comments welcome…. Bill