RDP 2015-02: Central Counterparty Loss Allocation and Transmission of Financial Stress 5. Policy Implications

The analysis in Section 4 gives rise to policy messages in three broad areas: (i) the trade-off between liquidity and solvency risk as collateral coverage increases; (ii) CCP default fund resources and loss allocation mechanisms; and (iii) network analysis. These policy conclusions acknowledge the limitations of the model identified in Section 3.1 – i.e. that we do not capture the extent to which derivative positions, liquidity holdings and capital positions are endogenous to the clearing market structure; and that we do not directly observe the bilateral matrix of positions.

5.1 Liquidity and Solvency Stress – the Trade-off

Using actual data on banks' OTC derivatives positions, the analysis in this paper confirms the finding in Heath et al (2013) that there is a trade-off between liquidity risk and solvency risk. Particularly in Scenarios 1 and 2, where non-central clearing persists, we present evidence consistent with the u-shaped relationship identified in that paper: the incidence of solvency stress declines sharply as initial margin coverage increases, but at the same time the incidence of liquidity stress steadily rises. The results for Scenarios 1 and 2 are particularly relevant to the extent that progress towards CCP clearing of OTC derivatives has been slower than anticipated and some product classes may continue to be predominantly non-centrally cleared for some time.

Netting efficiency is a critical determinant of the shape of this trade-off. The analysis confirms that multilateral netting via CCPs can substantially reduce banks' variation margin obligations, even in extreme market conditions, making liquidity stress less likely (see Section 4.2.1). It can also lower the collateral requirements associated with banks' OTC derivative positions.

Nevertheless, even with CCP clearing, extreme price changes can give rise to high variation margin obligations that trigger liquidity stress among some participants. In interpreting this result, it is important to reiterate that the initial liquidity positions assumed in the analysis in Section 4 prevailed under a different clearing structure. Banks' liquidity holdings would be expected to increase materially in an environment with higher collateralisation, and to vary according to the clearing structure. The key message is that robust liquidity regulation is crucial, such as has been introduced in the form of the Liquidity Coverage Ratio under Basel 3.

5.2 CCP Default Fund Resources and Loss Allocation

We recognise in our framework that while CCP clearing concentrates risk in a single node in the network, that node cannot generally be a direct source of stress in the system. A CCP does not generally assume financial risks other than those arising from the positions that it clears for its participants. Accordingly, typically the only circumstance in which a CCP may experience stress is if one or more of its participants defaults. If this arises, the adequacy of the CCP's financial safeguards is critical to ensuring that any stress is contained. The results in Section 4 support the view that a CCP designed and operated in accordance with the PFMIs can be expected to promote stability in the financial network.

As discussed, the PFMIs specify that initial margin should cover at least 99 per cent of potential future price changes and that a (large, internationally systemically important) CCP should maintain additional pre-funded default fund resources to meet the Cover 2 standard. This nevertheless leaves the possibility either that more than two participants enter stress and/or the market conditions prevailing at the time of participant default are more extreme than those considered in the CCP's stress tests. As noted in Section 4.2, realised market conditions could be more extreme not only in terms of the magnitude of the price move across products relative to that assumed in calibrating default fund resources, but also in terms of the assumed co-movement between products and the assumed closeout period. In such circumstances, the CCP's pre-funded financial resources could be exhausted.

To ensure that it did not then become insolvent and cease its provision of critical infrastructure services, the CCP would, in accordance with the PFMIs, allocate any uncovered losses to its participants. In our analysis, allocating uncovered losses back to participants is the principal channel by which a CCP could transmit stress back to the wider system.

In the results presented in Section 4.2, there is little evidence of such flow-on solvency stress arising from loss allocation. There could nevertheless be circumstances in which stress transmission did occur – for instance, this is observed in our analysis when we consider a six standard deviation price change or multiple sequential participant defaults. Even in these circumstances, our analysis suggests that losses would be sufficiently widely dispersed that stress would be well contained.

Since our analysis is focused on 41 large banks, however, it does not capture the extent to which the allocation of losses via VMGH could impose stress on non-banks, such as investment funds and other end users of derivatives who may have more directional positions. Equally, however, extending the network beyond large banks could, by dispersing uncovered losses more widely, potentially leave the system even better able to absorb stress.

Precisely how stress would transmit in the event of an extreme shock is highly dependent on the particular scenario at hand: the particular loss allocation mechanism applied; the distribution and direction of positions across participants; the magnitude of price changes across product classes and their co-movement; and the financial position of participants at the time of the shock.

Given the multi-dimensionality of the problem, it is inherently difficult for CCP participants to estimate their contingent liability when uncovered losses arise. It is nevertheless important that – subject to confidentiality constraints – CCPs provide sufficient transparency about their exposures, risk models and frameworks to assist participants in modelling and managing their potential obligations in the event of loss allocation.

To this end, CPMI and IOSCO have developed a public quantitative disclosure framework for CCPs (CPMI-IOSCO 2015). This includes required disclosures around margin models and coverage of pooled financial resources. Such transparency is a very welcome development. CCPs could, however, perhaps go beyond the disclosure framework to make additional information available specifically to assist participants in their understanding of ‘tail-of-tail’ risks. Additional transparency would be particularly useful in the areas of stress testing and model validation.

5.2.1 Stress testing

Stress testing is at the core of a CCP's risk framework, and central to any analysis of the adequacy of a CCP's pre-funded resources and its capacity to absorb rather than transmit stress. There are two ways in which the comparability and interpretation of outputs from CCPs' stress tests could be further improved:

• The PFMIs appropriately allow CCPs discretion in establishing what constitutes an extreme but plausible scenario to be used in calibrating stress tests and pooled financial resources. This allows stress tests to be tailored to a CCP's particular product and participant profile. It also allows for innovation in stress-testing techniques over time. It is crucial that participants and regulators understand the range of scenarios used by CCPs for similar products to facilitate analysis and allow better comparison of resilience and loss-absorbing capacity across CCPs.
• To promote transparency and comparability further, regulators could consider the feasibility of regulatory stress tests of CCPs' exposures on similar lines to those that have been carried out in the United States and Europe for banks. This could be challenging, given that CCPs face very different market and operating environments and have different product and participant profiles. Nevertheless, where feasible, periodic regulatory stress tests could be a useful tool for benchmarking by both regulators and participants, and for the exercise of market discipline (see also, Bailey (2014)). Consistent with local oversight arrangements, however, CCPs should retain discretion to tailor the stress tests used in their risk management processes.

5.2.2 Model validation – reverse stress testing and sensitivity analysis

Since the circumstances in which stress is transmitted back into the system are those in which a CCP's pre-funded financial resources have been exhausted, participants should be made aware of scenarios in which this could arise. The PFMIs require that CCPs carry out ‘reverse’ stress tests to gauge the circumstances in which pre-funded resources could be exhausted. Transparency around the outcomes of such tests would assist participants (and regulators) in understanding the potential range of circumstances in which uncovered losses could arise and loss allocation mechanisms could be invoked.

CCPs are also required under the PFMIs to analyse the sensitivity of their margin models to key assumptions, such as closeout periods, the sample period used to estimate the price distribution, or any floors applied in the margin-setting process. As an example of sensitivity to key model assumptions, our analysis considers simultaneous extreme price movements across all five derivative product classes. It is assumed, however, in calibrating initial margin and default fund resources that the covariance of price changes is zero. Such extreme realisations of co-movement are a potential trigger for stress in scenarios involving multi-product CCPs, which reveals the importance of prudent recognition of price change covariances between products in calibrating a CCP's financial protections. This is reflected in the PFMIs, which require that portfolio margin offsets be applied only where ‘the risk of one product is significantly and reliably correlated with the risk of the other product’ (CPSS-IOSCO 2012).

5.3 Network Analysis

The contagion analysis in this paper emphasises the importance of understanding how, in the event a shock did arise, stress could be transmitted through the system. As trade repositories deliver more detailed information to support such data-intensive analysis, and as methodologies and techniques in this area are further refined (perhaps building on the techniques in this paper), regulators could consider complementing regulatory stress tests with system-wide ‘big data’ network and contagion analysis. Wendt (2015) makes a similar recommendation. Metrics, such as the eigen-pair method described in Markose (2012), provide a good first approximation of the stability of the network and which particular institutions may warrant closer attention. This could be a useful tool for regulators.