Summary
This note presents a simplified expected-loss framework for considering the calibration of pro-rata second skin-in-the-game (SSITG) requirements for central counterparties (CCPs). It compares two stylised default waterfall designs: a single-tranche design, in which SSITG is positioned after the prefunded default fund, and a pro-rata design, in which SSITG is consumed alongside the prefunded default fund.
Under the assumptions used, the analysis illustrates that the position of SSITG within the default waterfall can materially affect the expected loss borne by a CCP. For the same nominal SSITG amount, a pro-rata design generates a higher expected loss because SSITG is exposed earlier in the loss distribution.
A sensitivity analysis using alternative illustrative probabilities for Event D produces equivalent pro-rata SSITG amounts of approximately 17%–49% of the single-tranche amount. Under the central sensitivity case, in which Event D is assigned a probability of 99.5%, the corresponding amount is approximately 32%, or one third of the single-tranche amount. The analysis is based on a number of simplifying assumptions and the results should therefore be interpreted as illustrative estimates that may support policy judgement, rather than as a mechanistic calibration rule.
1: Introduction
1.1 This note sets out a proportionate approach for analysing how the placement of a pro-rata SSITG tranche within a CCP’s default waterfall may affect the expected loss borne by the CCP. The analysis is intended to support policy judgement and should not be regarded as a substitute for scenario-based modelling using actual clearing member positions, stress losses and default waterfall data.
1.2 The note compares two stylised default waterfall designs:
- a single-tranche design, in which SSITG is positioned after the prefunded default fund and before unfunded resources; and
- a pro-rata design, in which SSITG is consumed alongside the prefunded default fund.
1.3 The analysis considers whether the earlier exposure of SSITG under the pro-rata design in (2) above could allow a lower nominal SSITG amount to produce an expected-loss outcome comparable to that under (1) above. A simplified expected-loss framework is used to estimate the corresponding SSITG calibration.
2: Default waterfall structures
2.1 Table A summarises the two stylised default waterfall structures considered in this analysis.
2.2 Under the single-tranche design, losses must first exhaust the prefunded default fund before reaching SSITG. Under the pro-rata design, SSITG is exposed earlier because it is consumed pari passu with the prefunded default fund. This difference in positioning of SSITG within the default waterfall is the focus of the analysis.
Table A: Positioning of SSITG within stylised default waterfall structures
Single-tranche design | Pro-rata design | ||
Defaulter’s resources: variation margin, initial margin and default fund contribution | Defaulter’s resources: variation margin, initial margin and default fund contribution | ||
CCP’s first skin-in-the-game | CCP’s first skin-in-the-game (Event D) | ||
Prefunded default fund, ie non-defaulters’ default fund contributions | (Event A) | Prefunded default fund, ie non-defaulters’ default fund contributions, consumed pro rata with CCP’s SSITG; and CCP’s SSITG, consumed pro rata with non-defaulters’ default fund contributions (Event E) | |
CCP’s SSITG | (Event B) | ||
Unfunded resources from non-defaulters | (Event C) | Unfunded resources from non-defaulters (Event F) | |
Footnotes
- Note: The probabilities assigned to Events A and D include all loss outcomes up to the point at which SSITG is exposed. For Event A, this includes losses absorbed by the defaulter’s resources, first skin-in-the-game and the prefunded default fund. For Event D, it includes losses absorbed by the defaulter’s resources and the CCP’s first skin-in-the-game.
3: Expected-loss framework
3.1 The expected loss borne by a CCP through SSITG depends on both:
- the probability that losses reach the SSITG layer; and
- the amount of SSITG consumed when that layer is reached.
3.2 The analysis estimates expected loss using a standard probability-weighted framework:
Expected loss = Σ [probability of event × loss if event occurs]
3.3 The calculation estimates average losses to the CCP by assigning probabilities to different loss events and estimating the average SSITG loss associated with each event.footnote [1]
3.4 Where losses fall within the SSITG layer, the analysis assumes that losses are evenly distributed across that layer. Under this simplifying assumption, the average SSITG loss within the layer equals half of the SSITG amount.footnote [2]
4: Single-tranche design
4. Under the single-tranche design, three stylised events, which are marked in Table 1, are considered:
- Event A: losses are absorbed by the defaulter’s resources, the CCP’s first skin-in-the-game or the prefunded default fund, and SSITG is not used.
- Event B: losses exceed the prefunded default fund but remain within the SSITG layer.
- Event C: losses exceed both the prefunded default fund and the SSITG layer.
- The probabilities assigned to these events capture the range of loss outcomes represented by each event. In particular, Event A includes less severe outcomes in which losses are absorbed by resources positioned earlier in the waterfall, as well as losses absorbed by the prefunded default fund.
4.2 The analysis assigns a probability of 99.9% to Event A. For illustrative purposes, the remaining probability is allocated between Events B and C, with Event B assigned a probability of 0.01% and Event C a probability of 0.09%. These assumptions reflect the fact that, under a single-tranche design, SSITG is only exposed once losses have exhausted the prefunded default fund.
4.3 If the SSITG amount is denoted by x, the expected loss is:
Expected loss = 99.9% × 0 + 0.01% × (x/2) + 0.09% × x
4.4 For an illustrative SSITG amount of £100 million, the estimated expected loss is £95,000.
5: Pro-rata design
5.1 Under the pro-rata design, SSITG is consumed alongside the prefunded default fund. As a result, SSITG is exposed at an earlier stage of the loss distribution.
5.2 Three stylised events, which are noted in Table 1, are considered:
- Event D: losses are absorbed by the defaulter’s resources or the CCP’s first skin-in-the-game, and SSITG is not used.
- Event E: losses are absorbed jointly by the prefunded default fund and SSITG.
- Event F: losses exceed both the prefunded default fund and SSITG, resulting in full consumption of the SSITG amount.
- As with Event A, the probability assigned to Event D includes less severe outcomes in which losses are absorbed by resources positioned earlier in the waterfall.
5.3 For the central illustrative case, the analysis assigns a probability of 99.5% to Event D. For illustrative purposes, Event F is assigned the same 0.09% probability as Event C under the single-tranche design. The residual 0.41% probability is assigned to Event E. The different probabilities assigned to Events A and D reflect the different loss ranges represented by the two events under the illustrative assumptions. Event A includes losses absorbed by the prefunded default fund, whereas Event D does not. Under the pro-rata design, losses reaching the prefunded default fund also expose SSITG because the two resources are consumed together.
5.4 As in the single-tranche example, the analysis assumes that losses are evenly distributed within the layer in which SSITG is exposed. Under this simplifying assumption, the expected SSITG loss in Event E is equal to one half of the SSITG amount. In practice, this assumption approximates the average proportion of SSITG consumed when losses fall within the joint prefunded default fund and SSITG layer.
5.5 If the SSITG amount is denoted by x, the expected loss is:
Expected loss = 99.5% × 0 + 0.41% × (x/2) + 0.09% × x
5.6 For an illustrative SSITG amount of £100 million, the estimated expected loss is £295,000.
5.7 To assess the sensitivity of this result, the analysis also considers alternative illustrative probabilities for Event D of 99% and 99.7%, alongside the 99.5% assumption used above. These values are informed by different initial margin confidence/coverage levels, but do not imply a direct mapping between an initial margin confidence or coverage level and the probability of Event D. Event D encompasses losses absorbed by all resources preceding the joint prefunded default fund and SSITG layer.
5.8 To isolate the sensitivity of the result to the assumed probability of Event D, the probability of Event F is held constant at 0.09%, with the resulting change allocated to Event E. This is a simplifying assumption and does not imply that the probability of Event F would remain unchanged if actual initial margin coverage changed. All other assumptions are also held constant, including the simplifying assumption that average SSITG consumption where losses fall within Event E, is 50%.
5.9 Under these assumptions, Event E has a probability of 0.91%, 0.41% and 0.21% where Event D is assigned a probability of 99%, 99.5% and 99.7%, respectively. For an illustrative SSITG amount of £100 million, the corresponding estimated expected losses are £545,000, £295,000 and £195,000, respectively.
6. Results and illustrative calibration
6.1 For an illustrative £100 million SSITG amount, the single-tranche design produces an estimated expected loss of £95,000. Under the pro-rata design, the sensitivity analysis produces estimated expected losses of £545,000, £295,000 and £195,000 where Event D is assigned a probability of 99%, 99.5% and 99.7%, respectively.
- Matching the £95,000 expected loss under the single-tranche design implies equivalent pro-rata SSITG amounts of approximately 17%, 32% and 49% of the single-tranche amount, respectively. The sensitivity analysis therefore produces an illustrative range of approximately 17%-49%, with the central 99.5% sensitivity case producing an estimate of approximately 32%, or one third.
- For example, if a single-tranche SSITG amount were calibrated at 25% of regulatory capital, applying these illustrative ratios would correspond to a pro-rata SSITG amount of approximately 4%-12% of regulatory capital, with the central 99.5% sensitivity case corresponding to approximately 8%.
6.2 These results do not imply that one design is inherently preferable to the other. Rather, they illustrate that the expected loss borne by a CCP for a given SSITG amount depends on where SSITG is positioned within the default waterfall and, consequently, the range of loss outcomes across which it is exposed.
7. Limitations and use of the analysis
7.1 The analysis is deliberately simplified and is intended to provide an estimation rather than a precise calibration methodology. A more comprehensive assessment would require CCP-specific data, including stress loss distributions, clearing member positions, default fund sizing and the detailed mechanics of the relevant default waterfall.
7.2 The analysis assumes that losses are evenly distributed within the layer in which SSITG is exposed, although tail-loss distributions are unlikely to be uniform in practice. Applying the same simplifying assumption to both waterfall designs may neutralise some of the approximation error when comparing relative outcomes, but the extent to which it does so will depend on the underlying loss distribution. Alternative distributional assumptions could therefore affect the precise expected-loss ratio and illustrative calibration. Under the assumptions used in this analysis, SSITG is exposed across a wider range of loss outcomes under the pro-rata design.
7.3 This analysis assumes that the underlying loss distribution remains unchanged when different SSITG amounts are considered. In other words, the probabilities assigned to the specified loss events are assumed to be the same regardless of the SSITG amount. As a result, the analysis estimates the expected loss associated with alternative SSITG calibrations using a fixed set of loss scenarios. The results should therefore be viewed as illustrative estimates rather than a full recalibration based on CCP-specific loss data.
7.4 The sensitivity analysis varies the assumed probability of Event D while holding the probability of Event F and the other modelling assumptions constant. The alternative Event D probabilities are illustrative and do not represent a direct mapping from initial margin confidence or coverage levels, as Event D encompasses losses absorbed by all resources preceding the joint prefunded default fund and SSITG layer. Holding Event F constant at 0.09% is intended to isolate the sensitivity of the illustrative calibration to the assumed probability of Event D; it does not imply that the tail of the loss distribution would remain unchanged if actual initial margin coverage changed. The analysis also retains the simplifying assumption that, where losses fall within Event E, average SSITG consumption is 50%.
8. Conclusion
8.1 This note illustrates that the placement of SSITG within the default waterfall can materially affect the expected loss borne by a CCP. Where SSITG is consumed pro rata with the prefunded default fund, it is exposed earlier in the loss distribution and may therefore generate a higher expected loss for the same nominal amount.
8.2 Under the simplified assumptions used, the sensitivity analysis produces equivalent pro-rata SSITG amounts of approximately 17%-49% of the single-tranche amount. The central sensitivity case, in which Event D is assigned a probability of 99.5%, produces an estimate of approximately 32%, or one third of the single-tranche amount. The range illustrates the sensitivity of the estimated calibration to assumptions about the probability of losses reaching the joint prefunded default fund and SSITG layer. These estimates should be interpreted as measures of expected-loss comparability rather than full economic equivalence. Alternative waterfall designs may differ in other respects, including incentives, loss-allocation dynamics and resilience outcomes, and the precise calibration would depend on the underlying loss distribution and CCP-specific circumstances. Nevertheless, the analysis provides an illustrative range for considering how SSITG levels might be adjusted to reflect differences in default waterfall design and the timing of loss allocation.
Whilst a more precise estimation can be modelled using a specific range of stress scenarios and clearing member positions, the aim of this analysis is to provide an approximation.
In practice, losses are unlikely to be distributed uniformly across the SSITG layer, and the probability of loss would generally be expected to decrease as losses become more severe. The simplifying assumption may therefore overestimate expected loss. Applying the same assumption to both waterfall designs may reduce some of the resulting approximation error when comparing relative outcomes, although the extent to which it does so will depend on the underlying loss distribution. Alternative assumptions could therefore affect the precise expected-loss ratio and illustrative calibration. Under the assumptions used in this analysis, the pro-rata design exposes SSITG across a wider range of loss outcomes.