ESMA91-372-2060 4th ESMA CCP Stress Test Report

5 July 2022 | ESMA91-372-2060

Report

ESMA Stress Test Exercise for Central Counterparties

 
 
 
 

Table of Contents 
Executive Summary ......................................................................................................................... 6 
Introduction ..................................................................................................................................... 10 
1  Background ............................................................................................................................ 10 
2  Scope and Objectives ............................................................................................................ 11 
Methodological Overview ............................................................................................................... 13 
1  Design and Components ........................................................................................................ 13 
2  Overview of the Process ........................................................................................................ 13 
3  Market Stress Scenarios ........................................................................................................ 14 
4  Methodology – Credit Stress Test .......................................................................................... 19 
5  Methodology – Concentration Stress Test ............................................................................. 28 
6  Methodology – Operational risk analysis ............................................................................... 35 
Results ............................................................................................................................................ 45 
1  Analysis and Breakdown of Resources.................................................................................. 45 
2  Credit Stress Test Results...................................................................................................... 54 
3  Concentration Stress Test Results ......................................................................................... 73 
4  Operational risk analysis ........................................................................................................ 84 
Conclusions .................................................................................................................................. 121 
Annexes ........................................................................................................................................ 124 
1  List of CCPs included in the scope of the exercise .............................................................. 124 
2  Concentration Stress Test annex ......................................................................................... 125 
3  Operational risk analysis annex ........................................................................................... 136 
 
   
 
 
   

 
 
 
 
2 
List of Figures 
Figure 1: Overview of the Process ......................................................................................................... 13 
Figure 2: Evolution of 2-day (5-day) moves for benchmark products during the first days ................... 17 
Figure 3: Comparison between most severe shocks and Stress Test shocks ...................................... 17 
Figure 4: Overview of credit stress test runs .......................................................................................... 24 
Figure 5: Illustrative typical Default Waterfall ......................................................................................... 26 
Figure 6: Default Waterfall – All CCPs ................................................................................................... 47 
Figure 7: Required Margin and Default Fund – per CCP ....................................................................... 48 
Figure 8: Required Margin vs Default Fund – All CCPs......................................................................... 48 
Figure 9: Default Waterfall – per CCP .................................................................................................... 49 
Figure 10: Required vs Excess Margin .................................................................................................. 49 
Figure 11: Clearing Members – All CCPs .............................................................................................. 51 
Figure 12: Clearing Member Groups – All CCPs ................................................................................... 52 
Figure 13: Clearing Member Groups - Distribution of required margin shares ...................................... 53 
Figure 14: Cover-2 Groups per CCP – Date: March 2021 – Without Excess Margin ........................... 56 
Figure 15: Cover-2 Groups per CCP – Date: March 2021 – With Excess Margin ................................ 57 
Figure 16: Cover-2 Groups per CCP – Date: March 2021 – With Concentration impact ...................... 59 
Figure 17: Cover-2 Groups per CCP – Date: March 2021 – With Concentration and WWR impact .... 60 
Figure 18: Cover-2 Groups per CCP – Date: April 2021 – Without Excess Margin .............................. 62 
Figure 19: All CCPs Cover-2 – Date: March 2021 – Without Excess Margin ........................................ 64 
Figure 20: All CCPs Cover-2 – March 2021 – Without Excess Margin – With Concentration and Wrong-
way risk Impact ....................................................................................................................................... 65 
Figure 21: All CCPs Cover-2 – April 2021 – Without Excess Margin .................................................... 66 
Figure 22: System-wide market impact per asset class......................................................................... 74 
Figure 23: Breakdown of concentration risk per asset class ................................................................. 75 
Figure 24: System-wide reported concentration add-ons, per asset class ............................................ 75 
Figure 25: Concentration risk coverage by addons for individual CCPs ............................................... 76 
Figure 26: Comparison of market impact and concentration add-ons, commodity derivatives ............. 77 
Figure 27: Comparison of market impact and concentration add-ons, fixed income derivatives .......... 78 
Figure 28: Comparison of market impact and concentration add-ons, equity ....................................... 79 
Figure 29: Comparison of market impact and concentration add-ons, bonds ....................................... 80 
Figure 30: Clearing / settlement unavailable: reliability metrics ............................................................. 89 
Figure 31: Critical supporting functions unavailable: reliability metrics ................................................. 90 
Figure 32: Risk indicators by severity groups- clearing or settlement unavailable ................................ 92 
Figure 33: Risk indicators by severity groups- critical supporting functions unavailable ....................... 93 
Figure  34:  Expected  1y downtime and estimated 95
th
  percentile  downtime  - clearing or  settlement 
unavailable ............................................................................................................................................. 94 
Figure 35: Expected 1y downtime and estimated 95
th
 percentile downtime – critical supporting functions 
unavailable ............................................................................................................................................. 95 
Figure 36: Number of critical third-party service providers per CCP by entity type ............................... 97 
Figure  37:  Comparison  between  weighted  and  not-weighted  number  of  critical  third-party  service 
providers per CCP .................................................................................................................................. 99 
Figure 38: Risk reduction for CCPs’ clearing and settlement functions exposure to third-party service 
providers using operational risk management tools............................................................................... 99 
Figure 39: Risk reduction for critical supporting functions using operational risk management tools . 101 
Figure 40: Weighted exposure per CCP after operational risk management tools – critical third-party 
service providers .................................................................................................................................. 102 

 
 
 
 
3 
Figure 41: Weighted exposure per CCP after operational risk management tools – critical third-party 
service providers .................................................................................................................................. 103 
Figure 42: Behaviour of operational risk management tools ............................................................... 105 
Figure 43: Network of CCPs connected through third-party providers ................................................ 107 
Figure 44: Network of third-party providers connected to at least two CCPs ...................................... 109 
Figure 45: Box & Whisker plot, Number of CCPs connected by type of entity .................................... 110 
Figure 46: Interconnectedness analysis – Financial services .............................................................. 112 
Figure 47: Interconnectedness analysis – Software, IT & Telecommunications services ................... 114 
Figure 48: Interconnectedness analysis – Data providers ................................................................... 115 
Figure 49: Interconnectedness analysis – Other services ................................................................... 117 
Figure 50: Telecommunications provider outage ................................................................................. 117 
Figure 51: Intragroup entity outage ...................................................................................................... 118 
Figure 52: Financial Market Infrastructure outage ............................................................................... 118 
Figure 53: Settlement system outage .................................................................................................. 119 
Figure 54: Market impact vs. relative position size, Investment grade corporate and sovereign bonds
 .............................................................................................................................................................. 127 
Figure 55: Market impact vs. relative position size, equities and equity derivatives ........................... 128 
Figure 56: Market impact vs. relative position size, energy and commodity derivatives ..................... 129 
Figure 57: Market impact vs. relative position size, eur fixed income derivatives ............................... 131 
Figure 58: Market impact vs. relative position size, credit derivatives ................................................. 133 
Figure 59: Severity distribution – average disruption time and severity distribution estimate by groups of 
CCPs .................................................................................................................................................... 138 
Figure 60: Probability of event lasting more than 2h by severity groups ............................................. 139 
Figure 61: Comparison between Lognormal distribution and Student’s t-distribution ......................... 141 
 
List of Boxes 
Box 1: Narrative of the scenario as provided by ESRB ......................................................................... 15 
Box 2: The Market Stress Scenarios in the light of the Russia’s invasion of Ukraine ........................... 16 
Box 3: Description of the Credit Stress Test Chart ................................................................................ 55 
 
   

 
 
 
 
4 
List of Tables 
Table 1: Reverse Stress Test – Loss above Required Prefunded Resources (No Excess) ................. 68 
Table 2: Reverse Stress Test – Loss above Required & Non-Prefunded Resources (No Excess) ...... 69 
Table 3: Comparison between baseline and alternative models ........................................................... 82 
Table 4: Operational risk events by type, number of events per year, duration of events and events 
longer than 2 hours ................................................................................................................................ 85 
Table 5: Events resulting in clearing or settlement unavailable – distribution of events, scope, event type 
and impact type ...................................................................................................................................... 87 
Table 6: Events resulting in critical functions unavailable – distribution of events, scope, event type and 
impact type ............................................................................................................................................. 88 
Table 7: Market impact on representative large positions, Investment grade bonds .......................... 128 
Table 8: Market impact on representative large positions, single name equity derivatives and securities
 .............................................................................................................................................................. 129 
Table 9: Market impact on representative large positions, other equity derivatives ............................ 129 
Table 10: Market impact on representative large positions, energy commodity futures/forwards ...... 130 
Table 11: Market impact on representative large positions, agricultural commodity futures/forwards 130 
Table 12: Market impact on representative large positions, freight derivatives ................................... 130 
Table 13: Market impact on representative large positions, eua ......................................................... 130 
Table 14: Market impact on representative large positions, EUR fixed income derivatives ................ 132 
Table 15: Market impact on representative large positions, GBP fixed income derivatives ................ 132 
Table 16: Market impact on representative large positions, USD fixed income derivatives ................ 132 
Table 17: Market impact on representative large positions, credit derivatives .................................... 134 
Table 18: Grouping assumptions on total system-wide Market Impact ............................................... 134 
Table 19: Impact of the level of aggregation of cm groups positions .................................................. 135 
Table 20: Frequency Poisson distribution – estimated parameter λ .................................................... 137 
   

Acronyms used

bps Basis points

CCP Central Counterparty

CM Clearing Member

DF Default Fund

EMIR European Market Infrastructure Regulation – Regulation (EU) 648/2012 of the

European Parliament and Council on OTC derivatives, central counterparties and

trade repositories

ESMA European Securities and Markets Authority

ESRB European Systemic Risk Board

EU European Union

ETD Exchange Traded Derivatives

FX Foreign Exchange

GEST Group of Experts on CCP Stress Testing

LEI Legal Entity Identifier

NCA National Competent Authority

OTC Over the counter

P&L Profit and Loss

pp Percentage points

PoA Power of Assessment

RTS Regulatory Technical Standards

SITG/SIG Dedicated CCP Resources (“Skin in the game”)

WWR Wrong-Way Risk

For CCP codes, please refer to Annex 1.

1 Executive Summary

Reasons for publication

The European Securities and Markets Authority (ESMA), in accordance with the European Market

Infrastructure Regulation (EMIR), shall initiate and coordinate assessments of the resilience of

Central Counterparties (CCPs) to adverse market developments. This report presents the results of

the fourth ESMA CCP stress test exercise that includes both EU and Tier 2 Third Country CCPs.

Contents

In line with the methodology published in June 2021

, the exercise covers both credit and

concentration risks, with targeted improvements in the methodology compared to the previous

exercises. In addition, the exercise includes for the first time an assessment of operational risk.

Given the scope and type of this exercise, a number of limitations and uncertainties remain and have

been highlighted in the report. This is particularly true for the operational risk analysis of the exercise,

the methodology and assumptions of which have been applied for the first time. Results are therefore

presented on an anonymous basis.

As with previous exercises, the objective of the ESMA stress test exercise is to assess the resilience

of CCPs to adverse market developments. This exercise is not aimed at assessing the compliance of

the CCPs with regulatory requirements, nor at identifying any potential deficiency of the stress testing

methodology of individual CCPs. Despite the fact that it is not aimed to do so, it may expose individual

shortcomings in the resilience of CCPs, in which case ESMA will issue the necessary

recommendations.

Analysis of CCP financial resources

Section 4.1 provides an analysis of the financial resources held by the 15 in-scope CCPs, as of 19

March and 21 April 2021. This data gives an overview of the size of the industry and sets the scene

for the presentation of the core stress test results. Overall, the prefunded resources collected by

CCPs have increased compared to the previous exercises. The CCPs reported in total 423 (resp.

409) billion EUR of required margin, default fund contributions and other committed prefunded

resources for March 2021 (resp. April 2021). There was no significant structural change in the overall

share of excess collateral or allocation of resources between margin and default fund contributions.

The analysis shows that, while there was a general increase of provided resources by all clearing

members, at the same time the top clearing members increased their relative share, indicating a

concentration of clearing member activity compared to the previous stress test.

Credit Stress Test

The results of the credit stress test are presented in section 4.2. Two default scenarios have been

run, combined with a common market stress scenario. In addition to the profit and loss balance of

clearing member positions (P&L) stemming from this scenario, concentration costs and costs related

to wrong-way risk were also taken into account for one of the dates. The first scenario is a Cover-2

per CCP, where ESMA assumes the default of two clearing member groups separately at each CCP.

The second scenario is the All-CCPs Cover-2 scenario, where ESMA assumes the default of the

same two groups for all CCPs system-wide. The defaulting entities are selected as the groups which

maximize the shortfall of prefunded resources, or alternatively the groups which maximize the overall

consumption of prefunded resources. Both scenarios have been run on two different dates, 19 March

2021 (end of day) and 21 April 2021 (intraday snapshot).

Under the Cover-2 per CCP scenario, ESMA assesses the resilience of each CCP to the default of

its top-2 clearing members groups under common price shocks. The prefunded resources were

sufficient to cover the losses resulting from the core credit stress test scenarios with relatively low or

moderate % consumptions. The sensitivity analysis also indicated that the conclusions seem robust

to small changes in the baseline shocks. The impact due to concentration and specific wrong-way

risk stemming from cleared positions led to higher losses and consumption for almost all CCPs but

under the considered market scenario these were contained within the default waterfalls of the CCPs

and there was no shortfall of prefunded resources.

During the time of finalisation of the exercise, Russia’s invasion of Ukraine led to extreme market

movements for instruments across the commodities and energy markets. A brief analysis of the stress

scenarios in the light of this event is presented in Box 2.

The All-CCPs Cover-2 stress test scenario is designed to assess the resilience of CCPs collectively

to the market stress scenario. Under this scenario, the same two groups of clearing members are

assumed to be in default in all CCPs. The majority of CCPs would experience a default of at least

one of their clearing members. However, these consistent scenarios did not put significant stress to

any CCP with the % consumption of default fund-level prefunded resources being relatively low in all

cases. This indicates that while CCPs are highly interconnected through common clearing

participants, the exercise did not highlight any pairs of groups that are at the same time and under

the common tested scenario highly impactful at multiple CCPs.

Finally, in the reverse stress analysis discussed in section 4.2.3 ESMA intentionally goes beyond

what was considered as plausible for the purpose of this exercise by stepwise increasing the number

of defaulting entities and the severity of the market shocks. Having considered the reverse stress test

scenarios, ESMA has not identified any systemically relevant adverse impact as the result of small

increases in market shocks and number of defaulters. Taking as a starting point the base scenario

and two defaulting groups, the analysis shows that incremental changes in market shock severity are

more harmful than increases in the number of defaulting groups.

Concentration Stress Test

The results of the concentration stress tests are presented in section 4.3. Based on the sensitivity

data provided by CCPs, the market impact (liquidation cost) was computed for all identified

concentrated positions on one reference date (19 March 2021).

The European-wide concentration analysis shows that concentrated positions represent a significant

risk for CCPs. For most asset classes, concentrated position risk is clustered in one or two CCPs, in

line with the findings of the previous exercise.

Framework for the 2021 ESMA Stress Test Exercise for Central Counterparties:

https://www.esma.europa.eu/file/119720/download?token=PtWBdAbz

System-wide, the largest concentration risk can be found in fixed income derivatives (around 29bn

EUR). Bonds (including bonds from Repo clearing services) come next with a total concentration risk

modelled at 11 bn EUR. Concentration in commodity derivatives and in the equity segment (securities

and derivatives) is very significant as well, with around 7bn EUR each. There is a very large coverage

gap between the system-wide estimated market impact under ESMA methodology and margin add-

ons, for commodity derivatives and to a lesser extent for equity products.

The concentration risk is addressed explicitly by a majority of CCPs through dedicated margin add-

ons. Although all CCPs face market impact, 4 CCPs (KDPW, CCPA, KELER, CCG) did not report

any concentration add-ons. Since the data request date, KDPW and CCG have implemented or are

in the process of introducing concentration add-ons. KELER relies on a monitoring system to require

additional collateral in case of elevated concentration.

Operational Risk

The results of the operational risk analysis are presented in section 4.4. In his analysis, ESMA derived

insights with respect to the level of operational resilience of CCPs for 14 CCPs (one was excluded

due to the absence of historical operational events data) and took an in depth look at third-party risk.

Using information about internal incidents of CCP’s systems and third-party providers ESMA

developed two methodologies to measure operational risk from historical events. With the computed

results, ESMA identified varying degrees of operational reliability for the CCPs included in the

exercise and identified specific CCPs where further work should be conducted to understand the

drivers of these differences, the root causes of the events and the remediation actions taken.

Through the use of a hypothetical scenario, ESMA evaluated the exposures to critical third-party

providers and the ability of CCPs to reduce risk through operational risk management tools. Using

exposure indicators, differences across CCPs in their relative level of third-party risk were identified.

Further work should be conducted to evaluate the individual circumstances of these exposures and

the suitability of taking corrective action to improve operational resilience against operational shocks

affecting critical third-party service providers.

In the analysis of the network of critical third-party providers, ESMA aggregated the information

provided by individual CCPs in order to understand and assess risks from common exposures to

third-party risk. Overall, ESMA identified a number of critical third-party service providers, which have

the potential to affect the critical functions of multiple CCPs in a correlated manner. In addition, ESMA

identified the critical third-party service providers with highest systemic importance for the CCP sector

due to both the criticality of their services and their level of interconnectedness with CCPs.

Overall Results

EU and Tier 2 CCPs proved to be overall resilient under the considered components, scenarios and

assumptions. As with the previous exercise, the adverse scenario did not aim to cover all possible

market movements but was designed to provide an internally consistent narrative to assess the

resilience of CCPs to system-wide market shocks.

The concentration component highlighted once again the need for CCPs to accurately account for

liquidation cost within their risk framework. Finally, the operational risk analysis highlighted a series

of areas and entities where further work to assess differences in measured risks between CCPs

should be conducted, and where risk mitigation measures may need to be further enhanced.

During the time of finalisation of the exercise, Russia’s invasion of Ukraine led to extreme market

movements for instruments in the commodity and energy markets. ESMA, in coordination with the

NCAs, closely monitored the impact that the outbreak has had on EU and Tier 2 CCPs. The analysis

performed by ESMA confirmed that the CCPs active in commodities clearing were the most exposed,

in particular the ones with relevant positions in power and to a lesser extent gas products. Moreover,

the CCPs with a more diversified set of cleared products were not significantly affected primarily

because of the lower experienced volatility in prices of other commodity and financial products.

Overall, ESMA notes that CCPs remained resilient through the crisis, despite the increased market

volatility.

Next Steps

In line with the EMIR mandate, where the assessments expose shortcomings in the resilience of one

or more CCPs, ESMA will issue the necessary recommendations.

2 Introduction

2.1 Background

1. CCPs are systemically important, and their resilience is critical to the stability of the financial

system in the EU. By their nature, CCPs are counterparties to all their clearing members. Failure

of CCPs to mitigate risks could potentially lead to spill-over effects and may exacerbate systemic

risk. Moreover, as evidenced in previous ESMA stress test exercises, CCPs are highly

interconnected through common stakeholders, which may propagate failures in one CCP

throughout the system. Stress testing CCPs, both individually and at financial system level, is an

important supervisory tool to ensure the sector is safe and resilient to defaults of clearing member

groups and market shocks. The Stress Test is a useful tool to assess the resilience of CCPs also

from other angles, such as the capacity to withstand the costs arising from the liquidation of large

positions or the operational resilience with respect to an outage of critical third-party service

providers.

2. The ESMA stress test is different than the stress tests of individual CCPs. CCPs run daily stress

tests on the basis of stringent prudential requirements that focus on their own environment,

including participants and cleared products. By its nature, the individual CCP’s stress test cannot

consider how the default of one of its clearing members or third-party providers impacts other

CCPs. Therefore, the ESMA stress test is a critical tool in assessing the systemic implications of

system-wide events and thus the resilience of the system of European CCPs.

3. One of the objectives of Regulation (EU) No 648/2012 of the European Parliament and of the

Council of 4 July 2012 on OTC derivatives, central counterparties and trade repositories (EMIR)

is to promote central clearing and ensure safe and resilient CCPs. Therefore, ESMA shall at least

annually, in cooperation with the ESRB, initiate and coordinate assessments of the resilience of

CCPs to adverse market developments. Following the amendments to Regulation (EU) No

648/2012 in 2019, these assessments should include both EU and third-country Tier 2 CCPs.

Moreover, ESMA shall include both financial and operational risks. ESMA shall develop the

following, for application by the competent authorities:

• Common methodologies for assessing the effect of economic scenarios on the financial

position of a financial market participant,

• Common approaches to communication on the outcomes of these assessments of the

resilience of financial market participants,

• Common methodologies for assessing the effect of particular products or distribution

processes on the financial position of a financial market participant and on investors and

customer information.

4. Where the assessment exposes shortcomings in the resilience of one or more CCPs, ESMA

shall issue the necessary recommendations.

5. The present report sets out the results of the 4th ESMA system-wide stress test exercise in

Section 4, following a description of the employed methodology in Section 3. The objectives,

scope and overview of the different tests performed are presented in the following paragraphs of

this section.

2.2 Scope and Objectives

6. The objectives of the ESMA stress test exercise result directly from the legal mandate given to

ESMA under EMIR. The objectives are to:

• Assess the resilience of CCPs to adverse market developments,

• Identify any potential shortcomings in the CCPs’ resilience, and

• Issue recommendations as appropriate.

7. The overall design of the stress test framework was also guided by a number of overarching

principles. ESMA has assessed the resilience of all CCPs in scope, individually and as a system.

This was done on the basis of, as much as possible, common methodologies and criteria. The

ESMA CCP stress testing exercise is not aimed at assessing the compliance of the CCPs with

regulatory requirements nor at identifying any potential deficiency of the stress testing

methodology of the CCPs. It may however expose individual shortcomings, in which case ESMA

will issue the necessary recommendations.

8. The exercise covers 15 CCPs, including all authorised EU CCPs as well as Tier 2 CCPs.

9. The scope of the stress test exercise developed over the years. The first exercise conducted by

ESMA was focused on the counterparty credit risk that CCPs would face as a result of clearing

member defaults and simultaneous market price shocks. The second stress test introduced

several methodological improvements as well as incorporating an assessment of liquidity risk.

The third exercise included a concentration risk component, with the aim of adjusting the losses

arising from the credit stress test to account for the costs of liquidating concentrated positions.

In this fourth exercise, the assessment of liquidity risk was paused, whereas the scope includes

operational risk as a new component. The design of the new component is discussed in detail in

section 3.6 and the results in section 4.4. Also, the integration of concentration with credit is an

important new development in this fourth exercise that has further improved the detections of

vulnerabilities in the European system of CCPs. The details of the methodology are provided in

paragraph 3.4.3.3 and the results in paragraphs 4.2.1.1 and 4.2.2.1.

10. Counterparty credit risk and concentration risk are the core types of risks faced by CCPs. The

methodology has evolved to cover additional risk sources and includes (i) the integration of

concentration with credit on a mutual date, (ii) an intraday test for credit risk only on a second

date.

11. In addition, an analysis of operational risk was performed. This analysis covered a general

assessment of operational resilience of CCPs based on the analysis of past events, as well as

specific analyses on third-party risk through the use of a hypothetical scenario and an analysis

of the network of critical third-party providers.

12. While residual risks from the in-scope risk sources are analysed and highlighted in the

framework, CCPs are also subject to other types of risks that are either not covered or are

partially covered and could in isolation or in combination with assessed risks challenge their

resilience. In particular, legal and any type of business risks are outside the scope of the exercise,

because of their largely idiosyncratic nature. Also, potential shortcomings in policies and

practices of individual CCPs, such as for example in the operationalisation of default handling

procedures, can challenge their resilience but are beyond what was considered in the course of

this exercise. Finally, environmental risk may be covered in a future exercise.

13. Furthermore, this exercise does not cover all possible scenarios to which CCPs may be exposed

to. When modelling the scenarios and credit exposures, it is not possible to cover all possible

risk factors and then all possible combinations of risk factor shocks for all CCPs. Indeed, while

the architecture of this stress test is based on internally consistent scenarios, where N securities

or contracts are cleared and possibly in the same portfolio, the number of possible basis risk

movements is 2^N. The value of N is at least thousands in the case of an equity clearing service

and thousands for derivatives. This makes it impossible to apply consistently all the potentially

damaging scenarios consistently across all portfolios of CCPs.

3 Methodological Overview

3.1 Design and Components

14. This stress test exercise has the following components:

15. Credit Stress: Assess the sufficiency of CCPs’ resources to absorb losses under a combination

of market price shocks and member default

scenarios.

16. Concentration risk: Assess the impact of

liquidation costs derived from concentrated

positions.

17. Operational risk: Analyse operational

resilience with a focus on external

operational dependencies that are needed

by CCPs to provide their critical services.

18. Reverse Credit Stress: Increase the

number of defaulting entities and level of

market price shocks to identify at which point CCP resources are exhausted.

3.2 Overview of the Process

19. ESMA followed the same approach as during the previous exercises and key steps are further

discussed in the next paragraphs.

FIGURE 1: OVERVIEW OF THE PROCESS

20. ESMA issued on 7 June 2021, the framework for the fourth CCP Stress Test Exercise

presenting the scope, the methodology and the details of the project. A market stress scenario

for CCPs was built by the ESRB. During the data request, CCPs were provided with templates

as well as detailed instructions on how to calculate and report the required information, including

the calculation of P&L using market stress scenario, concentration metrics or operational risks

and events.

21. A Group of Experts for CCP Stress Testing (GEST) with representatives from all national

competent authorities for CCPs (NCAs) has been setup with the aim of contributing during the

different steps of the project. ESMA and the Bank of England also collaborated during the

different steps of the exercise involving UK Tier 2 CCPs. ESMA finally organised a workshop

with EACH that was consulted on the overall framework and more specifically on the data request

templates and the instructions.

https://www.esma.europa.eu/file/119720/download?token=PtWBdAbz

Definition of

the ST

Framework

Data

Request

Data

Validation

Data

Analysis

Reconciliati

Final

Publication

22. The data request was launched on 8 June 2021 and the CCPs were asked to deliver by 20

August 2021 the completed data templates to the NCAs for EU CCPs or both ESMA and the

Bank of England for UK CCPs.

23. The receipt of the files on 20 August 2021 was followed by the first data validation phase, where

NCAs and the Bank of England validated the submitted data against the instructions and

according to a common set of validation rules. ESMA also coded and offered to run a validation

algorithm to facilitate this task. The first data validation phase lasted until 6 October 2021. Each

Authority appointed one officer that was the single point of contact. Where needed, the appointed

officers were in contact with ESMA staff and fellow officers from other NCAs in order to facilitate

the consistent implementation of the framework across all CCPs. Moreover, in order to facilitate

the convergence of the validation practices across different authorities, ESMA staff compiled and

shared with the authorities a list of frequently asked questions, together with the respective

answers.

24. The first validation phase was concluded with the delivery of the data templates in early October

2021 to ESMA that acted as a second line of defence in terms of data quality assurance. ESMA

checked at least on a sample basis, that the reported data were consistent, reasonable and

conform to the requirements included in the instructions. It finally assessed the overall plausibility

of results, including a comparison between CCP results, to detect any outliers. The second

validation phase was scheduled to last a total of 7 weeks. While the first set of findings were

identified and addressed within this period, there were a significant number of issues that had to

be followed-up multiple times, while in some cases, the correction of issues or the progress of

the analysis raised new issues. Therefore, in practice the validation process continued in parallel

with the analysis of the data that started immediately after resolving the first issues.

25. When sufficient progress was made on data validation and analysis, the GEST set the sensitivity

parameters used in the concentration component in January 2022. ESMA calculated and

analysed the results of the stress test. The preliminary results of the stress test were first

discussed in March 2022 with the GEST (and the Bank of England for UK CCPs) and then at the

CCP Supervisory Committee in April 2022. As a final step, ESMA also reconciled in April 2022

the core stress results with each individual CCP in an effort to reconfirm their robustness. The

reconciliation exercise was focused on CCP specific data. Systemic data could not be reconciled

without revealing confidential information on other CCPs or clearing members. Again, sufficient

time and effort were devoted to the reconciliation process in line with the previous exercise, in

order to ensure that the participants had the time and information needed to confirm the

interpretation of the sourced data and the correctness of the results. To take into account the

constraints of Russia’s invasion of Ukraine, the launch of this reconciliation exercise was delayed

by a few weeks and when launched, two weeks were given to CCPs.

26. To a significant extent, the quality of the data and results still rely on the data submitted by the

CCPs and the primary checks performed by the NCAs as ESMA lacks direct access to the CCPs

and was not in a position to redo all the validation checks that have been performed by the NCAs.

3.3 Market Stress Scenarios

27. Similar to the previous stress test exercises, the ECB, in close collaboration with the ESRB and

ESMA, has developed the narrative and has calibrated the adverse scenario for the 4th stress

test exercise. The shocks were produced using the tool that is employed for the calibration of

financial shocks for adverse scenarios at the ECB and has been in use for the calibration of

financial shocks for the EBA, EIOPA and ESMA scenarios.

28. The scenario that was produced reflects the ESRB’s assessment of prevailing sources of

systemic risk for the EU financial system. It reflects the triggering of one or more of the sources

of systemic risk to the EU financial system identified by the ESRB. These risks could materialise

jointly and reinforce each other. The results were derived using a methodology that considers

the joint empirical distribution of historical observations of the risk factors deemed relevant to

CCPs to produce a coherent market risk scenario.

BOX 1: NARRATIVE OF THE SCENARIO AS PROVIDED BY ESRB

The translation of the sources of systemic risk identified by the ESRB into instantaneous shocks

following triggers initiated in various market segments is described below.

In this adverse scenario, ongoing concerns about the evolution of the COVID-19 pandemic and

its economic ramifications trigger adverse confidence effects worldwide and prolong the

unprecedented economic contraction. The worsening of economic prospects is reflected in a

global decline in risk-free rates (from what is already a historically low level). Countries’ fiscal

positions weaken, as do corporate sector balance sheets. Despite the low risk-free interest rates,

concerns about the sustainability of public and private debt resurface, leading to a sharp increase

in credit risk premia and a widening of credit spreads worldwide. Countries with large spreads

are particularly affected, whereas countries with few debt sustainability concerns experience

somewhat more muted increases in sovereign spreads. As a result, the dispersion of sovereign

bond yields across the EU increases. The reassessment of market participants’ expectations

amid declining corporate earnings results in abrupt and sizeable adjustments to financial asset

valuations. Widespread downsizing of firms and rating downgrades trigger large-scale fire sales

in the non-banking sector. Market volatility spikes, the correlation of asset returns increases, and

borrowing costs surge on the back of expectations that non-financial corporations will default.

Similarly, the global fallout in terms of economic activity and the sharp increase in non-financial

corporate bond yields weigh on global investment and global demand for raw materials, causing

an abrupt repricing of commodities. The risk of idiosyncratic failures by financial institutions

intensifies, reflecting the deterioration of the macro-financial environment, with potentially severe

consequences for the financial system as a whole.

The scenario has been obtained by choosing the mean response for each conditioned variable

in an adverse scenario where the triggering variables are stressed over a two or five day horizon

depending on the asset class. The sample chosen for the calibration spans the period from

January 2005 to December 2020.

29. The system-wide stress scenarios should not be bound to only replicate past historical scenarios,

but also use past observations in combination with a narrative that reflects the assessment of

prevailing sources of systemic risk for the EU financial system, including the two Tier 2 CCPs in

the UK, to produce shocks that model potential future market conditions.

30. When modelling the stress scenario, it is not possible to cover all possible movements of different

risk factors and their co-movements within and across asset classes. The scenario constitutes a

severe yet plausible scenario that could arise if a risk environment such as the one explained in

the narrative were to materialise.

31. Overall, it is a very difficult task to produce potential future scenarios for such a wide range of

financial variables covering all major asset classes, which are at the same time sufficiently

severe, internally consistent and plausible. The methodological tool used can combine a large

number of time series and has allowed for the calibration of a more granular scenario, covering

more than 800 risk factors. There is no single test that can ensure that all variables are jointly

sufficiently severe and plausible.

32. During the time of finalisation of the exercise, Russia’s invasion of Ukraine led to sharp and

extreme market movements. An analysis of the stress scenarios in the light of these market

moves is presented in Box 2.

BOX 2: THE MARKET STRESS SCENARIOS IN THE LIGHT OF THE RUSSIA’S INVASION OF UKRAINE

Russia’s invasion of Ukraine on the 24

of February led to turmoil in the global markets. However,

the severity of the movements was asymmetrical across asset classes with commodities being

the most impacted one.

An overview of the evolution of the market moves for benchmark products in the main asset

classes during the first three weeks of Russia’s invasion is presented in the following figure.

The energy derivatives were the most impacted: a sharp

increase in power and gas prices was observed on the day of

the invasion, which marked the maximum upward movement

over the considered period. Severe upward shocks for power

and gas derivatives were observed also in the following week,

together with a sharp increase in coal prices, while prices

significantly decreased the week after. Oil benchmark

products initially suffered from upward pressure as well, but

the shocks were less severe than in the case of power and gas

products. The role of Russia as the main EU supplier of crude

oil, natural gas and solid fossil fuels led to price increases in

energy products amid fears of reduction in Russian supplies.

The upward shocks on wheat prices, that mainly occurred during the first two weeks of the war,

were also expected because Russia and Ukraine are significant exporters of wheat.

It is worth mentioning that nickel contracts trading was suspended after having reached all-time

highs during the third week of the conflict. However, the CCPs involved in the current stress test

exercise were not directly affected.

More details on the evolution of the market moves for benchmark products during the first days

of the market turmoil are presented in the following figure. An arrow is shown on a date if the 2-

day move (ending on that date) was ‘high’

, with the colour and direction of the arrow indicating

the direction of the relevant move.

i.e. higher than 50% of the period maximum. Notice that 5-day move was considered for ‘CDS’ and ‘Swap (EUR)’.

FIGURE 2: EVOLUTION OF 2-DAY (5-DAY) MOVES FOR BENCHMARK PRODUCTS DURING

THE FIRST DAYS

After having analysed the experienced market movements for a number of benchmark products

across all asset classes, ESMA staff compared them with the shocks used under the baseline

common market stress scenario. For this purpose, ESMA staff used the maximum of 2-day

moves over the period for all benchmark products with the exception of the primarily OTC-traded

instruments (i.e. CDS and Swaps) for which the maximum of 5-day moves was used. This choice

was made to reflect the EMIR requirement in terms of the minimum number of days that the

CCPs need to consider when calculating the margin requirements for the different instruments

and to remain consistent with the methodology used to calibrate the scenario shocks. It should

also be noted that this analysis compares the shocks of the internally consistent stress scenario

with the maximum moves observed during an event that unfolded over multiple weeks. Not all

maximum moves happened on the same day, and these could not have hit a single CCP at the

same time. On the other hand, no second-round effects were considered, that could have

amplified the market moves in case a default would have happened. The figure below

summarises the results of the analysis by comparing the maximum and minimum experienced

market movements during the first days of the Russian invasion with the shocks of the stress

test scenario used for this 4

stress test exercise.

FIGURE 3: COMPARISON BETWEEN MOST SEVERE SHOCKS AND STRESS TEST SHOCKS

The comparison of the scenario shocks with the maximum market moves during the first days of

the war showed that the ESRB scenario is overall of greater or comparable severity for most

asset classes

, but of a lesser severity for some commodities, mainly in the EU energy space.

Moreover, different directions of shocks were observed in some cases, mostly in the commodities

asset class.

The divergences highlighted can be explained by the scenario design, that was modelled based

on the sources of systemic risk to the EU’s financial system that have been identified by the

Only a few outliers were observed.

ESRB. More specifically, it was built around ongoing concerns at the time of the design about

the evolution of the COVID-19 pandemic and its economic ramifications. ESRB stress test

scenarios typically model an economic downturn which is very different from shocks driven by

supply concerns experienced during the war. The stress test cannot be used to assess resilience

under specific historic events, but rather aims to assess the resilience of CCPs on a forward-

looking basis and under a specific potential future scenario. Finally, ESMA staff would like to

stress that the extreme market moves were really restricted to a few asset classes that account

for a fraction of the cleared assets overall but could potentially put significant stress to particular

CCPs.

In the context of risks linked to the clearing activity, a combination of clearing member defaults

and simultaneous extreme market moves are needed to put a CCP at risk. In principle, if clearing

members continue to post margin and meet their obligations, periods of extreme market volatility

in isolation will not pose a specific market risk to a CCP. Moreover, the clearing members are

required to collateralise on a daily basis their exposures and thus, the market risk is limited to

the potential price movement from the last collateralisation of a defaulter’s position until the time

needed to hedge or close-out the position. Therefore, in terms of market movements, and always

in combination with simultaneous defaults, it is generally the extreme short-term shocks

spanning over a period of a few days that may put a CCP at risk and not medium or long-term

moves.

The extreme market movements led to a sharp build-up of losses for many market participants,

combined with margin calls from CCPs issued to collateralise the increasing exposures, also on

an intraday basis. Despite the extreme pressure, the impact on CCPs in scope of this exercise

was overall contained: no clearing member defaults were experienced in CCPs (except a small

one) and no inherent weaknesses were found so far, although some CCPs are reviewing margin

models and lists of eligible collateral.

While being cautious about the uncertainties resulting from any attempt to conduct a comparison

of this kind, ESMA staff has analysed the impact of using the shocks that actually occurred during

the first days of the conflict as if they had materialised on the March date in 2021 used for the 4

stress test exercise. This analysis cannot lead to accurate and robust conclusions as different

positions and margins were available on the days of the conflict, instrument prices were different,

and a full revaluation of the positions was not performed. The tests that are run daily by CCPs

are better placed to assess the impact of historic market moves on corresponding historic

positions taking fully into account specificities of cleared products. The ESMA Stress Test aims

to assess the resiliency of CCPs to adverse market movements on a forward-looking basis. It

cannot be used to draw conclusions on the resilience of CCPs to specific historic events, as this

would require the exact replication of historic exposures and eventually any conclusions drawn

would be bound to a specific historic event of the past, with limited explanatory power for future

events. However, the analysis performed by ESMA confirmed that indeed the CCPs active in

commodities clearing would have been the most exposed, in particular the ones with relevant

positions in power and to a lesser extent gas products. Moreover, the CCPs with a more

diversified set of cleared products were not significantly affected primarily because of the lower

experienced volatility in prices of other commodity and financial products.

To conclude, the scenarios that were used to run the ESMA CCP stress test are overall of greater

or comparable severity to the overall actual market events in March/April with the commodity

asset class being the only relevant exception. The used scenario remains valid and informative

as it stresses all CCPs as opposed to a single historic event that would only stress specific asset

classes and specific CCPs. Nevertheless, this can be further analysed in the context of future

exercises to understand if and how one could tweak the design of this exercise and these

scenarios to also test for stresses to particular assets in a manner that remains internally

consistent.

However, one needs to be very careful when drawing conclusions, as the ESMA stress test is

subject to several limitations and assumptions. Moreover, the unpredictability of the evolution of

the conflict may lead to additional extreme moves and the CCPs need to be prepared to mitigate

the resulting risks, especially if exposed to energy or agricultural products.

3.4 Methodology – Credit Stress Test

3.4.1 Overview

33. The goal of the credit stress test is to assess the

sufficiency of CCPs’ resources to absorb losses under a

combination of market price shocks and member default

scenarios.

34. The CCPs were asked to report, for each one of their

members and for each date separately, the losses the

CCP would face if a member would default following the market shocks dictated by the common

Market Stress Scenario and the resources that would be available to cope with the default.

35. Since it is not feasible to define scenarios for each and every risk factor of all CCP-cleared

contracts, the scenarios were defined for a set of high-level risk factors across different asset

classes and the CCPs needed to translate the risk factor shocks into P&L for their cleared

products and the members’ portfolios. Therefore, the Group of Experts for CCP Stress Testing

(GEST) developed and provided together with the data request and the market stress scenario

a set of detailed instructions that explain how these are expected to be implemented. The

instructions were drafted to provide clarity and address material implementation challenges. The

instructions were shared with EACH

for consultation before the finalisation of the design. An

overview of the rules, with a focus on the improvements compared to the previous exercises, is

provided in paragraph 3.4.3.

36. After receiving the exposures of each CCP towards each clearing member, ESMA applied the

conditions and assumptions underlying the Member Default Scenarios to identify the groups of

clearing members that are assumed to be in default. The groups with the top exposures were

identified by aggregating the losses across clearing members and, where relevant across CCPs

(i.e. for the All-CCPs member default scenario). A detailed description of the member default

scenarios is provided in the following paragraph (3.4.2).

37. The results are reported in terms of losses compared to the resources that were available to

cope with the default and are subject to the assumptions and limitations as these are described

in paragraph 3.4.4.

European Association of CCP Clearing Houses

3.4.2 Member Default Scenarios

38. The member default scenarios define the conditions used to select the entities that are assumed

to be in default. In all cases, the defaulting members were selected for each stress date

individually and considering only the required margin (i.e. excluding excess). Central banks,

governments and CCPs are not included in the list of entities that may be assumed to be in

default for the purpose of this exercise. The following member default scenarios were employed:

Cover-2 groups per CCP: For each CCP, ESMA staff selects as defaulting entities the members

belonging to the top-2 (corporate) groups of clearing members for that particular CCP. The

defaulting clearing member groups are selected per CCP; hence they may be (and in most

cases are) different for each CCP and they are not considered to be in default in other CCPs.

When a group is assumed to be in default in one CCP, all clearing members that belong to the

identified corporate group are assumed to default for the same CCP. ESMA staff first looks for

pairs of groups that lead to the highest aggregate (EUR) loss beyond required margin collateral

of the defaulter and beyond the Default-Fund-level prefunded mutualised resources, including

the Default Fund, the “Skin-in-the-game” and other prefunded Default-Fund-level resources.

Hence, ESMA staff first looks for pairs of groups that could together lead to a depletion of the

prefunded resources. If such pairs of groups are not to be found (i.e. there is no shortfall of

prefunded resources following the default of two groups), ESMA staff selects the two groups

that would lead to the highest consumption of resources, measured by the aggregate (EUR)

loss beyond Required Margin. The consumption can also be measured on a relative basis (i.e.

% of resources consumed). This may lead to different results for CCPs that have more than one

default funds. The selection of defaulting entities on the basis of the relative (%) consumption

could focus on a smaller default fund that may be closer to creating a breach, instead of

selecting pairs of groups that would cause larger (in absolute terms) losses at a larger default

fund or even multiple default funds. Hence, while the core selection is done on the basis of the

absolute (EUR) consumption, we also explore cases where there may be pairs of defaulting

groups that would create a higher % consumption at such default funds. This impact is

discussed when presenting the results as it may highlight a higher sensitivity at a smaller default

fund.

39. All CCPs Cover-2 groups: Across all CCPs (full scope), ESMA staff identifies the two clearing

member groups with the highest aggregate exposure under a particular market stress scenario.

All clearing members that belong to an identified corporate group are assumed to default across

all CCPs. Under this scenario, there may be CCPs with no clearing members defaulting, if none

of the identified defaulters is a member at these particular CCPs. This scenario aims to give an

aggregate view of the impact of the simultaneous default of the same two groups of clearing

members at all CCPs. With regards to the exact condition used to select the clearing member

groups, the first choice would be to select the top-2 groups that would lead to the highest

aggregate shortfall of prefunded resources across all CCPs. However, the results did not indicate

such cases. ESMA staff is therefore reporting the results after selecting the groups that lead to

the highest aggregate (EUR) loss beyond Required Margin across all CCPs.

3.4.3 Calculation of Credit Stress Exposures

3.4.3.1 Stress Dates and modelling of the Default

40. The credit stress test was run for two reference dates, i.e. Friday, 19 March 2021 and

Wednesday, 21 April 2021. For the first date (March), the default was modelled as a weekend

default, similar to previous exercises. For the second date (April), the default event was modelled

as an intraday default and the CCPs were asked to report exposures and collateral as of a

specific time window on this date.

41. For the March date (weekend default), all payments/obligations due on Friday prior to the default

were assumed to be met in full. After the default (which occurs during the weekend), no payments

were exchanged between the CCP and the defaulting member. Trading access was revoked in

the weekend, so that no position changes were accepted after the last novation cycle of Friday.

The positions therefore reflected the positions as of Friday end-of-day, including all transactions

that were accepted for novation during Friday. All price movements were supposed to be

happening instantaneously at the time the defaults are announced.

42. For the April date (intraday default), the assumption was that the defaulting clearing members

had met all payments/obligations due before a cut-off time, excluding the settlement of any

securities transactions that were to be settled on this date. After this time, no payments would

have been exchanged between the CCP and the defaulting member. The exposures would have

included any positions assumed by the member as a result of trading/novation during this date

up to the cut-off time and any securities transactions that were due to be settled on or after this

date

. The collateral included any collateral required and collected up to this cut-off time on

21/4/2021

. The underlying assumption was that the defaulting members met all payments before

the start of the day and were allowed to trade normally until a specific time during the day. The

members would have then stopped honouring any obligations after this time. The CCP would

have stopped accepting new transactions from these members after this time and would have

declared them in default later the same day. Finally, the CCP would have launched its default

management procedures that would have allowed it to start the liquidation of the positions on the

morning of the next day.

43. The intraday member default scenario aimed to test the intraday risk management procedures

of the CCPs, including margining and settlement procedures, considering that clearing members

may have increased their exposures during the day (day trading). This member default scenario

explored for the first time the consequences of the CCP having to face the default of members

carrying these increased positions, while having available only the collateral that was required

and collected up to this time. The implementation of such a supervisory stress test scenario

posed significant implementation challenges. Increased effort was required by all

participants/stakeholders, including CCPs, NCAs and ESMA. As with all assumptions

implemented for the first time, some uncertainties remained on the modelling of the relevant

assumptions. Hence, results should be read with caution. Moreover, in order to manage the

required effort, the cut-off time was not exactly the same for all CCPs and services but was set

by each CCP according to the schedule of its margin calls subject to conditions. For this purpose,

ESMA staff defined a common target time (14:00 CET) and each CCP was asked to identify the

cut-off time to be used for the exercise as the cut-off time of its scheduled intraday margin calls

that was (a) closer to the common target time and (b) in all cases after 12:45 CET and before

15:15 CET. The cut-off time selected reflected the time that was used to take a snapshot of the

positions and collateral in order to execute the intraday margin call and not the time of executing

the margin call or the time reflecting the deadline given to members to provide the collateral.

Moreover, the CCPs were not asked to report the data at account level for this date (21/4/2021)

but only at clearing member level, respecting of course any applicable segregation rules. This

comes at the cost of the stress test results for this date not being able to reflect the additional

stress assumptions (e.g. impact from concentration and wrong-way risk).

The assumption is that the defaulting clearing member would not have settled any securities transactions on this date.

The CCPs were allowed to adjust the collateral for cases where additional margin would have been required, called and collected

by the CCP according to existing rules and procedures as a direct result of the assumption that no securities transactions were

settled by the defaulting member on 21/4/2021.

44. For both dates, it was assumed that no porting of clients occurred, hence clients’ portfolios were

covered along with the proprietary positions of the defaulted clearing members and any losses

resulting from clients’ positions were included in the reported results. In the context of the credit

stress test exercise, this is a conservative assumption as the margin allocated to client accounts

can anyway not be used to cover losses of other client or proprietary accounts of the clearing

member.

3.4.3.2 Calculation of stress P&L from closing the positions at stressed market prices

45. All positions were assumed to be closed, for each individual account, at the prices implied by the

provided market shocks which were modelled as instantaneous shocks. No further price moves

were assumed to occur.

46. The CCPs were instructed on how to identify or adjust when needed the shocks to be applied to

their own products using the provided risk factor shocks and how to calculate the P&L stemming

from those shocks. Specific rules were provided per product type or asset class to set how the

shocks were to be adjusted, e.g. for similar underlyings or different maturities. For a few assets

(e.g. dividend / inflation derivatives) for which no relevant risk factor has been provided, the

shocks were to be modelled by the CCP using the stress scenarios used for their default fund

sizing under the supervision of the NCA.

47. As a general rule, CCPs needed to operate a full repricing on the basis of the risk factor shocks

and using the pricing models they normally use for the daily valuations of positions. Wherever

they are available, the CCP needed to use actual market prices for the base price, i.e. the price

to which the shocks are to be applied. Model-implied prices were only to be accepted where

market prices are not available or not reliable.

48. Beyond the exposures using the common market shocks, the CCPs were asked to report the

exposures as well after applying a number of multipliers on the shocks (i.e. x0.7, x 1.2, x1.5 and

x2.0). Each value of the multiplier corresponds to a Reverse Stress Scenario and all shocks are

to be simultaneously scaled. For each value of the multiplier, the CCPs ran a full repricing of the

portfolios, as opposed to applying a multiplier to the result (P&L) of the scenario.

49. In the determination of losses, no hedging strategy was allowed to be acknowledged or modelled.

In other words, the CCP was assumed to not have performed any risk mitigating transactions in

order to limit the risk of the defaulting member’s positions, but it has liquidated all the defaulting

member’s positions at the stressed price and has not introduced additional transactions such as

an index trade to capture first order risk.

50. The reported losses reflected the full amount that the CCP would have collected / paid in case

of the default, i.e. not only the profit or loss due to the stress shocks (stress P&L), but also any

accumulated profit or loss that has not been settled until the default and would have to be settled

when closing the position (non-stress P&L). This includes for example a loss due to actual market

movements on Friday that should have been settled on Monday when the member would have

been assumed to be in default.

3.4.3.3 Incorporation of impact from concentration and wrong-way risk

51. The methodology of the credit stress test component has now evolved to incorporate the impact

from additional risk sources. In particular, the results for one of the dates (March) were also

calculated to reflect the impact from concentrated positions and from wrong-way risk resulting

from cleared positions.

52. The base methodology, parameters and assumptions used to calculate the P&L due to

concentration and wrong-way risk are described in detail in 3.5 and 3.4.3.4 respectively.

However, one of the challenges of including this type of risks in a supervisory stress test exercise

is that the impact is dependent on the selection of defaulting clearing members. For example, a

clearing member may have positions in instruments issued by another clearing member. The

additional loss from wrong-way risk will only impact the CCP if the two members are assumed to

default together. As a further example, the default of two clearing members that hold large same-

direction positions on the same instrument may exacerbate the impact from concentration risk

as the positions would have to be liquidated together. Hence, the impact can only be calculated

in relation to a specific pair of defaulting members, while at the same time the selection of

defaulters needs to consider this additional impact.

53. For ESMA to be able to seamlessly incorporate the additional impact from concentration and

wrong-way risk, the CCPs were asked to report for one of the stress dates (March) the required

data not only at clearing member level but also at account level. In particular, the stress P&L and

corresponding collateral were reported at clearing member and account level and the

concentrated positions only at an account level. The instructions and reporting templates were

redesigned to allow ESMA to have the information required to aggregate results from account

level to clearing member level, while incorporating the effects from these additional stress

assumptions (concentration and wrong-way risk). Beyond allowing the assessment of the impact

from concentration and wrong-way risk, the more granular reports enhanced the visibility in

calculations and together with the detailed instructions helped to further strengthen the data

validation process and the credibility of the exercise.

54. The CCPs were asked to report the data for the accounts that were active (i.e. had open positions

or provided collateral) on this date, specifying also the relationships between different accounts

and priorities in loss absorption reflecting their segregation rules in case of default. CCPs have

in general very diverse account structures that go beyond the minimum set of accounts required

by EMIR. They have different accounts that serve different purposes (e.g. position accounts,

margin accounts, collateral accounts). For the purpose of this exercise, an account was defined

as the level at which collateral can be fully offset against P&L from all positions recorded in the

same account. Hence, CCPs were instructed to report at a level that would allow ESMA to

correctly aggregate all fields from account to clearing member level by implementing the reported

relationships / segregation rules.

55. Hence, for one of the dates (March) ESMA was able to run the stress test with and without these

additional stress assumptions. In fact, three sets of results are presented:

a) Credit stress test results without concentration and wrong-way risk impact

b) Credit stress test results with concentration impact (but without wrong-way risk impact).

c) Credit stress test results with concentration and wrong-way risk impact.

56. For the second date (April), the data was reported by CCPs only at clearing member level

order to manage the overall effort. Hence, one cannot reflect the impact from these additional

stress assumptions for the April date or for the reverse stress test scenarios.

The stress results without the concentration and wrong-way risk could be calculated both by starting from the account-level

reports or by starting from the clearing member reports as CCPs reported both for the March date. This was used to confirm the

correctness of the aggregation algorithm.

In all cases, even where results were reported by CCPs at clearing member level, the reported data reflected all applicable

segregation rules, e.g. that client’s resources cannot be used to cover losses from proprietary positions.

57. The following figure provides an overview of the different results obtained.

FIGURE 4: OVERVIEW OF CREDIT STRESS TEST RUNS

58. As explained above, the concentration and wrong-way risk impact stemming from one defaulting

clearing member group may vary depending on the selection of the second defaulting group.

Therefore, results would ideally need to be computed for all possible combinations of pairs of

clearing member groups before selecting the top defaulting pair for each scenario. However, the

number of scenarios, CCPs and Clearing Member Groups implies that there are too many

combinations, that could not be exhaustively computed in a timely manner. In order to address

this issue while reasonably trying to make sure that all relevant pairs of clearing member groups

are analysed, ESMA staff implemented the following heuristic two-step approach. This was

applied when calculating the results with concentration and wrong-way risk.

• First select a subset of clearing member groups to be considered for Cover- 2.

a. Calculate results per single clearing member group taking into account stress

scenario losses and concentration impact and select the top-10 clearing member

groups impacting mutualised prefunded resources.

b. Select the top-10 clearing member groups in terms of potential (aggregate across

all members) wrong-way risk impact.

c. Combine both lists and create a combined list of clearing member groups that are

relevant from a concentration and/or wrong-way risk perspective.

• Then consider all possible pairs between the clearing member groups

belonging in the

relevant shortlist and compute results for each pair to identify the top pair based on the

member default scenarios (3.4.2)

59. It is acknowledged that this approach has some limitations. Not all possible pairs of clearing

member groups are tested. However, it would be extremely difficult, resource- and time-

consuming to calculate the combined concentration impact for each possible pair needed to

exhaustively test all possible cases. Nevertheless, in order to provide some level of comfort that

i.e. up to 20 clearing member groups per scenario leading to 190 pairs of groups to compute per scenario

The concentration-only run was based on the shortlist identified in terms of concentration risk while the concentration- and

wrong way risk- run was based on the combined shortlist of clearing member groups.

no pairs with a potentially significant impact are left out, ESMA staff has still tested all possible

combinations of clearing member groups but without considering the combined impact on

concentration costs, i.e. the concentration impact was calculated for each clearing member group

separately without considering that positions would have to be liquidated together in the market.

3.4.3.4 Estimation of wrong-way risk

60. One of the improvements incorporated in this exercise is the enhancement of the wrong-way risk

adjustment for cleared positions.

61. Ideally, when assuming that an entity is in default, one should also reflect this in the price of the

cleared instruments and collateral for all clearing members and CCPs. In the previous exercise

the CCPs were instructed to incorporate in the P&L calculations for each member this effect for

all cleared instruments issued by this specific clearing member or its affiliates. This means that,

with regard to cleared instruments, CCPs did not model the effect from the default of all entities,

e.g. when one assumed the default of two clearing members, one would only have for each

clearing member the effect from instruments issued by itself. Therefore, the scope of this

adjustment was limited and was also not applied consistently across all members.

62. The more granular reports of the new exercise allowed us to address this limitation. The CCPs

reported stress data and concentrated positions at account level for one of the dates (March).

Moreover, CCPs were asked to identify which cleared instruments (or underlyings of cleared

instruments) are issued or guaranteed by one of their clearing members or affiliates. Hence,

where reported positions referenced instruments issued by a defaulting entity or its affiliates,

ESMA was able to estimate and incorporate in the stress test results the impact that the default

of the entity will have on the positions of all clearing members.

63. In order to ensure internally consistent results, and in contrast to what was done in the previous

exercise, the wrong-way risk adjustment was applied independently of whether it would result to

a loss or a profit

. Overall, ESMA staff has noticed that incorporating a positive wrong-way risk

(right way risk) would have not significantly impacted the end results as these are always based

on a worst-case scenario assumption, e.g. selection of members that would have resulted to the

highest losses.

64. The methodology that was used to perform the wrong-way risk adjustment per type of instrument

is described below. The adjustment was calculated as the impact on top of the stress shocks,

i.e. on top of the stress P&L already reported by CCPs (e.g. if a shock of -15% is applied in the

market scenario to a stock, the P&L from the wrong-way risk adjustment as indicated below was

added to the stress P&L calculated from this shock).

• Security (e.g. stock) issued by the clearing member or one of its affiliates: assume a shock

of -50% to the position value reported in the relevant concentrated position file.

• Equity derivatives (e.g. Single stock derivatives) having as an underlying a security issued

by the clearing member or one of its affiliates: assume a shock of -50% on the position value

reported in the relevant concentrated position file.

• Corporate

bond issued by the clearing member or one of its affiliates: revalue at 40% of

face value,

e.g. for a short position equity derivatives position on a defaulted entity.

The impact was not calculated for sovereign bonds or other public bonds as reported by CCPs.

• Covered bond/MBS issued by the clearing member or one of its affiliates: revalue at 88.75%

of face value,

• Single-name CDS referencing the clearing member or one of its affiliates: revalue assuming

a recovery rate of 40%.

65. In the interest of avoiding complexity, products (e.g. derivatives, warrants, ETFs) on an index

where one of the constituents is issued by the clearing member or one of its affiliates were not

adjusted for wrong-way risk. Similarly, index CDS’s with constituents referencing the clearing

member or one of its affiliates were not adjusted for wrong-way risk.

66. In general, the direct incorporation of wrong-way risk for cleared exposures improves the

consistency and credibility of the exercise but the estimation of the relevant impact is still subject

to model risk

. As a mitigation measure, ESMA staff shared the impact estimation with the CCPs

and asked them to provide a more accurate estimate where needed together with a justification.

Finally, any wrong-way risks towards issuers & custodians of collateral and other resources were

also not acknowledged in the context of the credit stress test component.

3.4.3.5 Default Waterfall and Collateral

67. CCPs collect margins, default fund contributions and keep dedicated own resources that can be

used to cover losses stemming from a clearing member’s default. The scope and priority in use

of the different resources are set in regulation and the rules of CCPs. A typical default waterfall

is presented below, only for illustration purposes. The actual default waterfall of each individual

CCP, as this was reflected in the data reported, has been considered to calculate the absorption

of losses in the EU-wide CCP stress tests.

FIGURE 5: ILLUSTRATIVE TYPICAL DEFAULT WATERFALL

For example, for Equity Options the adjustment was based on the delta-equivalent position values reported by the CCPs. For

corporate (covered) Bonds the impact was calculated as the difference between the reported present value and 40% (88.75%) of

the face value derived from the present value using available prices. For CDS the impact was calculated as 60% of the notional

estimated using the reported sensitivity to credit spread changes and the maturity aggregated into maturity buckets.

Other Default-Fund-level or

CCP-level Resources

Default Fund Contributions

of non-defaulting members

Dedicated CCP Resources

("skin-in-the-game")

Default Fund Contribution

of Defaulter

Margin of Defaulter

3.4.3.6 Identification of Collateral

68. Concerning the default fund contributions, the reported amount reflected the required amounts,

i.e. no excess collateral reported for the default fund contribution. In terms of margin, the CCPs

were asked to report separately the minimum required collateral, not including any excess

amounts, and the total available collateral.

69. The minimum required collateral is meant to reflect a scenario where defaulting members would

have withdrawn under stressed conditions any collateral exceeding the minimum required. In

fact, any member experiencing financial difficulties would most probably post only the minimum

required collateral. Nevertheless, the CCPs have been asked to report also the actually held

(total available) collateral, including excess amounts. Therefore, although the base stress results

only considered the required collateral, ESMA staff also presented in some cases for

completeness the stress test results using the excess collateral. In order to make the two sets of

results (with / without excess) directly comparable, the same defaulting entities have been

considered and in particular, the defaulting entities have always been selected using the

minimum required collateral without the excess.

70. The required margin was identified as the sum of the margins required to be paid on the morning

of the day of the default, any payment issued and paid during this day (and up to the cut-off time

for the intraday default) as a result of margin calls and any of the collateral previously held as

excess but consumed by the member’s activity or intra-day valuations and offset against the

computation by the CCP of margin requirements during this day, the absence of which would

have led to a margin call according to the CCP’s existing rules and procedures. Moreover, for

the intraday default scenario the CCPs were allowed to consider margin that would have been

required, called and collected by the CCP according to existing rules and procedures before the

exercise cut-off time as a direct result of the assumption that no securities transactions were

settled by the defaulting member on this date.

3.4.3.7 Valuation of Collateral

71. The CCPs were asked to revalue the collateral alongside the cleared products using the market

stress scenarios shocks. Therefore, ESMA staff did not rely on the haircuts applied by CCPs.

72. Although in principle, valuing collateral using the same stress shocks improves scenario

consistency and gives us the ability to check haircut adequacy, it is not necessarily in all cases

the most conservative choice. For example, it can be that the collateral value increases following

the shocks, while when relying to CCPs’ haircuts the collateral value is always reduced.

Moreover, the CCP may have re-invested the actually provided collateral and the P&L from the

actually available resources may be different (higher or lower) than the P&L from the provided

collateral.

73. The following modelling assumptions were used. The CCPs were asked to report and use the

stressed values of margin & default fund collateral actually provided by clearing members (as

opposed to the stressed values of relevant resources following re-investment). Since the credit

stress test is based on the provided (as opposed to invested) collateral, any market or credit

risks stemming from the re-investment of collateral are not reflected in the exercise.

3.4.4 Residual Limitations of Credit Stress Test

74. As in all exercises of this scale and type, there are residual limitations.

75. The credit stress test exercise has evolved to include the impact from concentrated positions for

one of the stress dates. However, the estimation of this impact is subject to limitations, which are

described in the relevant methodology, including due to the restricted modelling of the default

management procedure, the model granularity and the uncertainties around the estimation of

the market impact parameters.

76. Investment risks, including credit risks arising from the default of an issuer or custodian of

collateral or other resources are not assessed in the exercise. The exercise does incorporate an

assessment of the market risk for provided collateral using the common market stress scenarios.

Any additional market or credit risks resulting from the re-investment of provided collateral are

not covered. These limitations are due to the fact that these risks are linked to the individual

actions and rules of the CCP and are thus difficult to model consistently across CCPs.

77. The wrong-way risk adjustment is applied for one of the stress dates and has been enhanced to

also reflect the risk that would materialise if one defaulting clearing member clears instruments

issued by another defaulting clearing member. However, the estimation of this impact is subject

to limitations, including due to uncertainties in the estimation of the recovery values. Moreover,

in the interest of avoiding complexity, the wrong-way risk effects on cleared index products are

not modelled.

78. Operational risks, including those that may lead to increased credit risks, such as the

operationalisation of default procedures, are also not reflected in the credit stress test results.

The ESMA stress test exercise includes for the first time an assessment of operational risks in a

separate component, but these are not reflected in the credit stress test results.

79. Any additional second round effects to prices following the default of entities will not be modelled

(i.e. the price shocks are the ones provided by the ESRB and the number of defaults are the

ones described above, but the two are taken exogenously). Also, the default of additional entities

due to losses accumulated from non-cleared portfolios will not be modelled because the scope

of the exercise is limited to CCPs exposures.

80. When modelling the scenarios and credit exposure, it is not possible to cover all possible risk

factors and then all possible combinations of risk factor shocks for all CCPs. That would require

modelling several thousands of risk factors and then all their co-movements. Since the exercise

has to be run on the basis of common methodology and criteria, it cannot be aimed to identify

topical deficiencies of individual CCPs. This includes for example the change of spread between

two markets. Moreover, the shocks are modelled using a very large but still limited number of

risk factors. CCPs’ models are in most cases more sophisticated and cater for additional sources

of risk, such as jump-to-default-risk for CDS.

3.5 Methodology – Concentration Stress Test

81. The objective of the concentration cost analysis is to assess the concentrated positions present

in the portfolios of CCPs, estimate the potential liquidation costs that could be derived from their

closing out, and assess the potential implications to CCP resources these positions pose.

82. For this exercise, the market illiquidity (or concentration) risk is defined as the added cost of

liquidating in the market a position (or hedging it) in a short amount of time (in practice the time

allocated to the management of a default by a CCP).

83. Initial market shocks apply to the mid-price of all positions regardless of their size and direction.

However, it is likely that CCPs would incur costs beyond this price, depending on the size of their

positions and the depth of the markets they clear.

84. Under the Article 53(3) of the RTS (Commission Delegated Regulation EU No 153/2013), a CCP

shall conduct a thorough analysis of the potential losses it could suffer and shall evaluate the

potential losses in clearing member positions, including the risk that liquidating such positions

could have an impact on the market and the CCP’s level of margin coverage.

85. Under the 2017 CPMI-IOSCO report (Resilience of CCPs: further guidance on the PFMI), a

CCP’s margin model should incorporate estimates of market liquidation costs, including bid-ask

spreads not otherwise modelled in the price returns or explicit fees paid to trading platforms or

liquidation agents. These market liquidation costs should also reflect the market impact of

liquidation activity, when applicable. When a portfolio liquidation requires the disposal of

concentrated positions or portfolios that are otherwise significant in terms of anticipated impacts

on market liquidity in the relevant product, a CCP should contemplate the possibility that

assumed market liquidation costs, such as bid-ask spreads or mid-market pricing, will not in fact

be actionable or otherwise predictable in the face of an actual liquidation.

86. ESMA incorporated the above requirements in the design of this exercise to develop a

methodology to include concentration risk in the CCP stress test exercise.

3.5.1 Scope and methodological principles

3.5.1.1 Market Scope

87. The exercise covers securities (equities and bonds) and derivatives (equity, fixed income,

commodities, credit, freight and emission allowance).

88. To limit the overall complexity, other markets have been excluded on the following grounds:

• Small volumes in CCPs (structured finance products, ETCs and ETN bond types, securitised

derivatives, CFDs)

• Highly liquid markets (Foreign exchange derivatives)

• Complex sub-asset classes decided on a case-by-case basis to limit the overall computational

complexity (volatility index derivatives, dividend derivatives, inflation and cross-currency

swaps).

3.5.1.2 Position type coverage

89. The framework design ensures that most concentrated spread positions, even market neutral

ones, are captured.

90. As the transaction costs add up, spread positions between two correlated but different

underlyings are not offsetting. For example, a large short position in one equity and a large long

position in another equity do not offset each other's costs. Likewise, electricity or commodity

derivatives with different delivery points will be captured.

91. Curve / calendar spreads on the same underlying are captured unless all components fall in the

same maturity bucket. Hedges with economic rationale such as delta hedging single stock

derivatives with the underlying stock are considered.

3.5.2 Concentration modelling overview

General principles

92. This exercise does not model the whole default management procedure, for example, there is

no attempt to factor in the impact of an auction which could lead to smaller or bigger

concentration costs. Rather, ESMA staff assessed the market impact of liquidating positions or

setting up hedges, compared these to available concentration add-ons, and combined the

findings with the credit component to improve the methodology of the stress test.

93. The market impact can be broken down in two parts:

• An exogenous factor which is the relative size of the bid-ask spread. Spreads would

represent a cost even for small positions.

• An endogenous factor, when positions are too large and cause the market to move against

them (one can think of a forced liquidation). Market impact depends on the position size

relative to the market depth, which is the ability of the market to absorb a substantial amount

without materially impacting the mid-price.

94. Exogenous liquidity adjustment is of negligible importance for the world's main futures and

currency markets, but more significant for other markets, such as credit or energy markets.

95. For large positions, market impact is usually much larger than bid-ask spreads.

96. In the context of a portfolio containing a single asset, e.g. an equity, the concept is quite

straightforward. There is only so much the market can absorb in one day before the market price

of the security moves in an adverse direction. For derivatives such as swaps or options, the

concept is more complex and market specific.

97. The importance of managing concentration risk was illustrated in a recent market event.

Following the default of a clearing member on the 11/09/2018 at Nasdaq Clearing, it was

assessed that its positions were too large to be closed in the market. The illiquidity of the

positions made the final losses to largely exceed the mark to market losses prior to the default.

Overview of the methodology

98. The computations performed by ESMA staff are based on three main data sets reported by the

CCPs and further described in section 3.5.2.1:

• concentrated positions of each of the CCP clearing members at account level,

• Average Daily Notional Amount (or Average Daily Volume) metrics for the relevant asset

classes,

• sensitivity tables estimating the liquidation costs for the different asset classes as a function

of the position size relative to the Average Daily Notional Amount (or Average Daily

Volume).

99. ESMA staff performed three main steps in the computation phase:

• aggregation of the size of the position to be liquidated under the no porting assumption, as

detailed in a)

• computation of the size of this position (or its hedge) relative to the Average Daily Notional

Amount (or Average Daily Volume), as detailed in b),

• estimation of the liquidation market impact of the position as a function of the ratio between

the size of the position (or its hedge) and the Average Daily Notional Amount (or Average

Daily Volume), by using the common ESMA sensitivity tables and as detailed in c).

100. Further details on the different steps and methodological assumptions are provided in section

3.5.2.2.

3.5.2.1 Input data submitted by the CCPs: positions, reference volumes and liquidation costs

101. ESMA requested CCPs to report the concentrated positions of each of their clearing members

at account level following prescribed aggregation rules for each asset class. The target sub-

classes are built from tables of the annex III of the Commission Delegated Regulation 2017/583

on MiFID II, dealing with transparency requirements. The segmentation criteria are

complemented where necessary to improve the granularity, with for instance, the introduction of

a delivery / cash settlement location for some commodity derivatives.

102. For each given aggregation level, each CCP also reported a common reference volume for all

its positions. This reference volume is usually reported as an Average Daily Volume (ADV) for

securities, and an Average Daily Notional Amount (ADNA) for derivatives. For securities, the

primary source for ADVs is the systematic internaliser data

computed and published by ESMA.

In most other cases, the reference volume was set using the CCP’s own submitted data, as they

reflect the markets the CCP can readily access and for which it has in place the operational

arrangements to readily execute transactions. For equity index derivatives, the cash turnover of

the underlying index is used when relevant.

103. Finally, each CCP provided sensitivity tables estimating the liquidation costs for the different

asset classes it clears. Typically, for any given asset class or sub-class, the tables give the cost

(in bps or % market value) for executing trades that are 0.5, 1, and 2 times the average daily

volume (or average daily notional amount when relevant). Market-wide baseline sensitivity tables

for each asset class were built by ESMA and discussed by the Group of Experts on CCP Stress

Testing (GEST).

3.5.2.2 Computation steps

a) Aggregation and reporting of positions

104. As instructed, CCPs calculated and reported the aggregated positions per instrument/asset

class for each clearing member house/ client account.

The porting assumption means that the CCP has committed to trigger the procedures for the transfer of the assets

and positions held to another clearing member following a client(s) request.

The “systematic internaliser calculations” data files can be downloaded from ESMA Website: https://www.esma.europa.eu/data-

systematic-internaliser-calculations

105. The first step performed by ESMA staff was computing the size of the position to be liquidated

at Clearing Member level and for each instrument/asset class, aggregating the data at account

level provided by the CCPs.

106. The aggregation was based on the following high-level principles:

• For securities, the positions are aggregated at the ISIN level.

• For other derivatives, non-linear positions (e.g. options) are aggregated with linear positions

(e.g. futures/forwards) using their delta. The vega is also reported for equity and commodity

derivatives.

• Single stock equity derivatives are aggregated with the underlyings through the net delta at

ISIN level. For the rest of derivatives, the aggregation follows class specific criteria and

maturity buckets.

• Fixed income and credit derivatives positions are reported through aggregated risk

sensitivities.

107. To limit the data volume and focus on concentrated positions, CCPs have only reported relevant

positions above class-specific thresholds for bonds, equity and equity derivatives.

108. To allow for a simpler implementation:

• the positions are valued without the impact of any market risk scenario.

• no porting of accounts is assumed.

b) Computation of the relative size of the positions to be liquidated

109. Once the size of the positions to be liquidated were determined, ESMA staff computed the ratio

between the size of each position (or its hedge) and the to the Average Daily Notional Amount

(or Average Daily Volume).

110. The ratio is an estimate of the size of the position, and it is used as entry in the baseline

sensitivity tables to estimate the liquidation market impact.

c) Estimation of the liquidation market impact

Sensitivity tables

111. Each CCP provided sensitivity tables estimating the liquidation costs (in bps or % market value)

as a function of the ratio between the size of the positions to be liquidated and the Average Daily

Notional Amount (or Average Daily Volume), as described in the previous computation step.

112. The sensitivity tables were provided by each CCP for the different asset classes it clears. The

number of CCPs providing estimates varies widely across asset classes. For instance, only 2

CCPs clear freight derivatives, but 10 CCPs clear equities.

113. The third computation step started by summarising all the sensitivity tables provided and

produce a single market-wide baseline sensitivity tables for each asset class. ESMA staff

typically chose the median contribution to reduce the influence of outliers. This step involves the

scrutiny for accuracy and plausibility as well as the removal of outliers. For instance, emission

allowance sensitivities were increased to match the energy commodity ones.

114. Yet, additional methodological choices had to be made to calibrate the sensitivity tables.

115. Sensitivity tables give the cost for executing trades that are 0.5, 1, and 2 times the average daily

volume (or average daily notional amount) and other values ranging from 0.5 to 2 are then

interpolated.

116. For positions far exceeding 200% of the reference volume, extrapolation assumptions have a

huge impact. To avoid unrealistic estimates for larger positions (in relation to the reference

volume), it was decided to extrapolate flat the market impact in basis points beyond the last point

provided by CCPs (i.e. 200% of the reference volume). In other words, the total market impact

will scale linearly with the notional but not quicker. Positions smaller than 25% of the reference

volume are not contributing to the concentration risk.

117. For credit and fixed income derivatives, note that this extrapolation was made for values beyond

500% of the reference volume (i.e. beyond the last point provided by CCPs for those asset

classes), for the same reason of avoiding unrealistic huge impacts.

118. Fixed income derivatives and Credit derivatives follow a different methodology

. A hedging cost

(as opposed to a liquidation cost) of the reported positions was computed using a limited number

of hedging instruments.

Concentration PnL at account level

119. The final step consisted in using the proper market-wide sensitivity tables for each asset class

to retrieve the market impact (in bps or % market value) of the positions to liquidate, as a function

of the relative size computed in b).

120. ESMA staff first determined the size of each position (or its hedge) relative to the average daily

volume (or such relevant parameter), and then its liquidation market impact using the baseline

sensitivity tables.

121. Finally, the concentration PnL was allocated at account level as a function of the position size

at account level and the market impact estimated. The concentration PnL computed at account

level allows to include the concentration costs into the waterfall, as described in 3.4.3.3.

122. In case of multiple clearing member defaults (as part of one or more groups), the total position

was used to get the total market impact, which was then apportioned to the different clearing

members and their client / house accounts.

3.5.3 Known limitations of the Concentration Risk Analysis

3.5.3.1 Limited scope

123. The exercise does not model the whole default management procedure. More specifically, there

is no attempt to factor in the impact of an auction which could lead to smaller or bigger

The Fixed Income derivatives and Credit derivatives methodologies are described in more details in the Annex (section 6.2.2).

concentration costs. This impact could be significant for credit and fixed income derivatives that

are modelled through their hedging portfolios.

124. Although most cleared asset classes are covered, markets like cross-currency basis swaps,

longer term foreign exchange derivatives or less liquid foreign exchange pairs were not in scope.

Contract for differences (CFDs) were also not included.

125. Therefore, concentration risk on these segments is not quantified and there may be an

underestimation of the concentration risk for default funds that include such asset classes.

126. To reduce complexity, some calendar / curve risks within asset classes are not being considered

when they are categorized within the same buckets. Likewise, for some asset classes, market

practices could allow for more aggregation than considered in the framework.

127. The liquidation of collateral is not covered to avoid making the exercise overly complex. For

instance, it would have been necessary to model the change in the order with which resources

are used for each CCP and depending on which CM is in default.

3.5.3.2 Model and calibration risk

128. The model chosen may be insufficiently accurate and / or fail to consider properly specific

features of some asset classes

. However, its results explain well some of the CCPs’ own

concentration risk models.

129. As the market impact estimates are provided by the CCPs, there is a risk to have a biased

estimation of the real risks in stressed markets. For the same asset class, estimates submitted

by different CCPs varied significantly. This already suggests a very different sensitivity toward

concentration risk of different CCPs, which impacts the overall exercise. In some cases (freight

or emissions), few CCPs contributed, and it was difficult to challenge the CCPs and get better

estimates. As previously mentioned, emission allowance sensitivities were increased by ESMA

staff to match the energy commodity ones.

130. Additionally, the concentration risk estimates are not adjusted for the impact of the any market

risk scenario.

131. In section 4.3.4.2, the model risk for securities is assessed using an alternative model.

3.5.3.3 Market depth

132. For securities, the primary source for ADVs is the systematic internaliser data

computed and

published by ESMA.

133. In most other cases, the market depth was computed using the CCP’s own submitted reference

volumes, as they reflect the markets the CCP can readily access and for which it has in place

the operational arrangements to readily execute transactions.

134. The submitted reference and aggregated position volumes could be affected by errors in the

data provided or assumptions used by CCPs. The two levels of validations (by NCAs and ESMA

For example, see Credit Derivatives methodology in annex.

The “systematic internaliser calculations” data files can be downloaded from ESMA Website: https://www.esma.europa.eu/data-

systematic-internaliser-calculations.

staff) aimed at limiting the risk of wrong computations by CCPs, but this risk cannot be completely

eliminated with ex-post desk-based verifications.

3.6 Methodology – Operational risk analysis

3.6.1 Overview

135. The objective of the operational risk analysis is to assess the level of operational resilience of

EU and Tier-2 CCPs. In this first exercise, specific focus will be put on third-party risk, with a

hypothetical scenario of an outage of a critical third-party service provider.

136. Operational resilience has been defined by the Basel Committee on Banking Supervision

(BCBS) in its “Principles for Operational Resilience”

as the ability of a regulated entity to deliver

critical operations through a disruption. The use of this definition has also been applied in the

IOSCO consultation report of “Operational resilience of trading venues and market

intermediaries during the COVID-19 pandemic”

137. The concept of operational resilience is closely related to the broader concept of operational

risk as defined in the glossary (annex H) of the Principles for Financial Market Infrastructures

(PFMI) as: “The risk that deficiencies in information systems or internal processes, human errors,

management failures, or disruptions from external events will result in the reduction,

deterioration, or breakdown of services provided by an FMI.”

138. The operational risk analysis is comprised of three parts:

1. Assessment of the general level of operational resilience of individual CCPs

– Using data of past events, ESMA staff developed metrics to evaluate and compare the

past operational performance of individual CCPs’ internal critical systems and critical

supporting functions and derive insights on their level of operational resilience.

2. Assessment of risk exposures of individual CCPs to critical third-party providers

– Using data of critical providers, risks related to each critical provider and available

protective tools ESMA staff developed metrics to evaluate and compare the exposure

of individual CCPs to third-party risk and their resilience to a hypothetical unavailability

scenario.

3. Assessment of concentration or systemic risks in the network of critical third-party providers

– Using data of critical providers, risks related to each critical provider and available

protective tools ESMA staff analysed the interconnections between individual third-

BCBS Principles for Operational Resilience (2021) (https://www.bis.org/bcbs/publ/d516.pdf)

Operational resilience of trading venues and market intermediaries during the COVID-19 pandemic

(https://www.iosco.org/library/pubdocs/pdf/IOSCOPD694.pdf)

party providers and multiple CCPs that could lead to correlated operational risk events

across entities.

139. For the operational risk analysis, only 14 out of the 15 CCPs that could be in scope are included,

due to the absence of historical data of operational risk events for 1 CCP.

3.6.2 Assessment of the general level of operational resilience of individual CCP

3.6.2.1 Overview

140. Measuring operational resilience involves understanding the universe of risks an entity can face

and how well it would be able to avoid disruption or minimize downtime; both elements represent

a challenge, as the universe of risks cannot be easily determined ex-ante and the ability of the

CCP to withstanding these may or may not be known depending on whether there is data

available to reach meaningful conclusions.

141. The starting point for the analysis of operational resilience was to look at past historical events

to gather evidence. Although there is no defined way to measure operational resilience, it is

known that the deterioration or disruption of the CCP’s ability to perform its services would have

an effect on the performance indicators of the service an FMI provides, which is something for

which the discipline of reliability engineering has mathematical tools that can be used to measure

reliability and availability of CCP’s services.

142. From a user perspective, the CCP provides a specific service that is expected to be available

and perform in an expected manner, during agreed business hours. A disruption is a lack of

availability or a degradation in the quality of the service.

143. Linking operational resilience and availability, it can be affirmed that the level of operational

resilience of an entity will determine the availability of the service in the face of adversity and

threats to disruption. Firms with a high level of resilience against specific risks or scenarios will

be able to maintain higher levels of availability of the service when confronted with these risks

than firms with lower levels of resilience against those risks or scenarios.

144. Understanding this relationship, ESMA staff can start the analysis by observing the historical

incidents of EU and Tier-2 CCPs and develop reliability and availability indicators, as they will

provide information of the level of operational resilience of CCPs against the events they have

experienced during the selected historical timeframe.

3.6.2.2 Reliability measurement

145. From a reliability engineering perspective, a service is considered as a repairable system. This

means that the system may experience outages, leading to downtime, but it is repairable, and

after some time (repair time) the service will be restored.

146. The mathematical modelling of repairable systems is performed with probabilistic models and

statistical methods. The reliability (the ability of a system to operate for a specific period of time)

of a repairable system is described by the following elements:

• A stochastic process (usually Poisson, Renewal…) that describes the frequency of outages

across time.

• A probability distribution of the time to repair of individual outages (usually Weibull,

Exponential…).

147. The occurrence of outages and their associated time to repair generates downtime of the service

or systems in scope.

148. Availability metrics describe relationships between downtime, uptime and total time of operation:

• Availability: The fraction of time that the service is in operating condition in relation the total

time where it is expected to be operational.

• Unavailability: The fraction of time that the service is in downtime in relation the total time

where it is expected to be operational.

3.6.2.3 Reliability and availability metrics used in the report

149. For this analysis ESMA staff used two approaches:

1. Metrics based on average times and expected unavailability:

o Mean time between failures (MTBF): The average time between service breakdowns.

o Mean time to repair (MTTR): The average time taken to recover from a failure.

o Expected 1 year unavailability: The expected downtime in a one-year period using

MTBF and MTTR.

2. Estimation of percentile metrics:

a. Model based estimation of percentile metrics of unavailability.

150. As part of the results, a description of the model’s methodology, its parameters, assumptions

and results are provided.

151. Both approaches are used together to analyze the risk characteristics of individual CCPs.

3.6.2.4 Information collected from operational incidents experienced by CCPs

152. In order to calculate the reliability and availability metrics, information of past incidents provided

by CCPs is used. Incidents were measured at an internal CCP level by using:

• Data of incidents of that have affected the clearing service or could have affected it had it

occurred at a different time (incidents and near misses).

• Incident time is measured between the start of the incident until remediation.

153. The measurement with an internal approach is different from the measurement at customer

experience level. ESMA staff adopted an internal approach measurement in order to develop

indicators that allow us to detect issues in a preventive manner.

154. The internal approach implies that incidents included in the measurement may or may not be

responsibility of the CCP, as they can include outages of other FMIs that would have a knock-

on effect on the CCP and are out of its control. It also implies that the measured incident

remediation time will be equal or higher than the incident time experienced by customers.

155. The instructions given to CCPs were to report events where any of the following conditions were

to report:

All operational risk events whether generated by an internal or external cause during the past

five years since the reference date for the operational risk analysis (the most recent date from

the two dates specified in the credit component) are in scope. Events shall be reported when

any of the below conditions are met:

1. A CCP’s critical clearing services or supporting functions are affected during business hours,

with any of the levels of impact defined in the “Types of impact” considered.

2. The CCP experienced a direct financial loss greater than 50,000 €.

3. A third-party provider is unavailable during business hours (even if it doesn’t create an outage

in the CCP, due to not being needed in that specific moment) for a period greater or equal to 30

minutes.

156. Incidents in parts of the clearing process that would have a role before a trade is accepted by

the CCP creating the legal obligation for the CCP would not be in scope of the exercise. For

example, the failure of a trading venue in sending trade information would be considered

previous to the start of the obligations of the CCP and would not be reported in this exercise.

157. It must be noted that for this exercise we only consider two types of states for systems / third-

party providers:

a. Available: The system / third-party is operating normally

b. Non-available: The system / third-party is suffering an outage that leads to a lack of

availability or a degraded state

3.6.2.5 Dimensions of operational risk considered

158. When considering operational risk events in the context of operational resilience, one needs to

use time as the quantitative variable (as opposed to monetary quantities, which are the

quantitative variable used when assessing the financial consequences of operational events)

and take into account different aspects:















159. The  describes the probability of events and the probability distribution

of the time duration of events.

160.  is qualitative in nature; buckets are used in order to be able to aggregate

information using three different  representing different levels of severity

describing the consequences for the CCP:

1.Clearing / settlement unavailable

161. Description: Immediate critical impact to the ability of the CCP to perform the clearing function

for clients in any clearing service, product, or currency. Evidenced by:

• Inability to accept / receive trades.

• Inability to make necessary computations to calculate payments / settlements (due to any

element involved being unavailable, including infrastructure or data unavailability /

inaccuracy).

• Inability to complete full process of payment or settlement to final client (due to any element

that forms part of the chain).

• Non-availability of cash or securities that are needed to fulfil payment or settlement under

normal conditions (no default assumed).

• Non-availability of risk management function leading to the inability of the CCP to perform

the clearing or fulfilment of payments / settlements.

• Non-availability of liquidity provider that is needed to perform operational functions under

normal conditions (no default assumed) leading to the CCP's inability to perform the clearing

function.

• Non-availability of operational infrastructure leading to inability to perform the clearing

function.

• Interoperability link disrupted.

2.Critical supporting function unavailable

162. Description: Impacts a critical supporting function of the CCP without direct disruption of the

clearing function for the clients meanwhile it is remediated. Evidenced by:

• Non-availability of Business Continuity, disaster recovery or cybersecurity capabilities.

• Non-availability of risk management functions limiting the ability of the CCP to monitor /

manage risk (but allowing the CCP to clear and fulfill payments / settlements).

• Non-availability of default management capabilities.

• Non-availability of capabilities, infrastructure or data for monitoring or supervision purposes.

3.Other Service Level Agreement breach

163. Description: Any other impact that would have a lower level of severity than the two above-

described categories and that would be different from no impact.

164. Lastly, 



is defined as the percentage of clearing activity that was affected by

the incident. For this analysis, we build a proxy of activity impacted by using the margin that can

be linked to the activity disrupted, in the following manner:















3.6.3 Assessment of risk exposures of individual CCPs to critical third-party providers

3.6.3.1 Objective

165. The aim of this exercise is to quantify the level of third-party risk for each individual CCP and

understand how the CCPs would be able to cope with a hypothetical scenario involving the

outage of a critical third-party provider.

3.6.3.2 Overview

166. ESMA staff used concepts and methods from the disciplines of reliability engineering and

operational risk management to describe the logic behind the approach for CCP third-party risk

indicators and the methodology used for the hypothetical scenario. The purpose in this section

is to provide an intuition of the logic behind this methodology, not to build a complete model for

computing results, so where relevant ESMA staff will use mathematical methods that describe

probabilities of a single outage to simplify things.

3.6.3.3 Third-party risk from an individual entity

167. From the perspective of a CCP, the third-party risk to which it is exposed is related to the number

of third-party entities that provide critical services needed for the CCP to perform its functions,

their reliability (‘probability of failure’ and ‘severity of failure’ probability distributions), the potential

impact on the CCP’s processes from the non-performance of any of these entities (‘type of

impact’), the exposure of the CCP to each third-party entity (‘exposure’) and the operational risk

management tools the CCP has in place to reduce risk (‘tools’).

168. The risk to the CCP from each individual third-party entity without taking into account any risk

management tools can be linked to the following factors:

 













169. For the  dimension, ESMA staff applied the same approach as in the section on

measuring operational resilience at CCP level. ESMA staff separated the analysis for the two

relevant categories of “Type of Impact” (Clearing / settlement unavailable and Critical supporting

function unavailable) due to their different qualitative nature.

170. For the 



dimension, exposures to individual entities received a weight by

adjusting for the percentage of CCP’s clearing activity serviced by 



171. With respect to the  of individual entities, quantitative estimations were not produced

due to the current limitations of available data; rather exposures were identified by the type of

entity, using three broad categories linked to their regulatory status.

• FMI group: Financial Market Infrastructures, payment systems, settlement systems, central

banks.

• Other financial: Regulated financial institutions excluding those in the first group (credit

institutions, insurance undertakings or investment firms …).

• Non-financial: Entities outside of the financial regulatory perimeter (technology providers,

data providers…).

3.6.3.4 Third-party risk from the aggregated exposure to the network of critical third-party

providers

172. The starting point is the exposure towards a single entity, for which ESMA staff has identified

the main factors of interest and described the approach in the previous section:

 













173. From a mathematical perspective, reliability is defined as the probability of non-failure across

time. For one failure, the reliability of a simple component with respect to time (







) is described

as:









  







174. with 







being the probability function of failure sometime up to time t.

175. Critical third-party service providers are entities that are all needed at different points of the

operational process of a CCP in order for it to deliver its clearing services; in this sense if any of

the critical providers has an outage, then the CCP would experience an outage with a specific

 for a percentage of its activity linked to the 



of the third-

party provider experiencing the outage.

176. A system composed by a set of components that are all needed in order for the system to work

can be described as a series system, and the reliability of a series system is described as:























































177. In other words, the reliability of the CCP can be expressed as the product of the reliability of

each of the component of the system: if one component suffers an outage, the entire system

suffers an outage.

178. From the observation of this simple reliability model, one can come up with some initial

observations:

• The probability of non-failure decreases as the number of components increases (assuming

the components have a non-zero probability of experiencing outages), so third-party risk of

an entity is an increasing function of the number of individual critical third-party providers.

The longer and more complex the system, the higher the probability of a system outage for

a given level of reliability of individual components.

• Third-party risk depends on the level of reliability of its individual components.

179. Taking into account these observations, ESMA staff set as general objective for CCP risk

indicators to quantify the number of entities to which individual CCPs have exposure and the

level of expected reliability of these entities.

3.6.3.5 The effect of operational risk management tools

180. From an operational risk management perspective, CCPs can act proactively to enhance their

resilience with respect to failures from critical third-party providers by using preventive tools or

mitigation/protective tools.

181. For the area of third-party risk, ESMA staff considered as preventive tools any Policies,

Procedures or Controls that can reduce the probability of failure of individual third-party entities

or that could help reduce the detection time contributing to a lower recovery time. The main tools

in the area of third-party risk are the selection of highly reliable critical third-party service

providers and the enforcement/monitoring of high operational standards in the selected third-

party entities.

182. These tools affect the individual reliability of the components in the CCP’s system of critical

third-party providers. When a CCP chooses high quality providers, then the reliability of the

individual components will be high leading to a reduced overall third-party risk.

183. The problem with respect to the quantification of the effect of preventive tools is that one can’t

easily measure their effect, as one doesn’t know what the situation would be should the

prevention tools be different or non-existent (one can only measure the reliability of an existing

provider). Information about the quality of the third-party providers can be extracted using the

analysis of historical information of events with the methodology of section 1 ‘Assessment of the

general level of operational resilience of individual CCPs’.

184. With respect to mitigation tools, they are categorized into two groups:

• Building resilience by setting up a redundant external third-party provider that acts as a

backup, e.g., receive critical data from two providers, have contractual arrangements with

more than one financial entity for a specific critical service, have a redundant data centre,

and duplicated telecommunications lines.

• Building resilience by developing a specific internal tool that can act as a substitute of the

critical third-party service provider, e.g., being able to build data in house in case of

emergency, having alternative communication means, a UPS system for electricity outages,

and relevant and effective legal provisions.

185. The effect of mitigation tools can be directly measured as each of the elements is observable.

Their effect is that building resilience through mitigation tools reduces the probability of failure of

each specific component in the system of third-party dependencies. Each component in our

series system model can be considered a unique service for which the CCP can have one or

more providers operationally set-up; increasing the number of alternative providers (or internal

tools) increases the reliability of an individual component (









in the model) due to the fact that

a simultaneous failure of all redundant options is needed in order to produce an outage in the

component and subsequently affect the reliability of the system.

186. Mitigation tools have been the focus of the analysis, as they are the most relevant for the

hypothetical scenario of outage affecting a critical third-party provider. Once an outage is

assumed, prevention tools cannot protect a CCP from the consequences, but mitigation tools

can protect from one (or multiple) outages.

3.6.3.6 Modelling the behaviour of the risk management tools

187. CCPs were asked to report the impact from a hypothetical outage affecting each critical third-

party provider without taking into account any mitigation risk management tools and also the

residual impact after taking into account mitigation tools.

188. For the reporting of the residual impact, CCPs were allowed to report the impact after taking into

account mitigation tools:

• There was not residual impact,

• The  would change, and/or

• The 



would be reduced.

189. Critical third-party providers that perform similar functions and act as redundant options of each

other were linked together, so that they are not considered independent entities where the failure

of any of them leads to an impact to the CCP but rather groups of providers where all of them

need to fail in order to cause an impact to the CCP (they are all considered as one exposure with

multiple providers).

3.6.3.7 Hypothetical scenario

190. In the hypothetical scenario it was assumed that any critical third-party provider could suffer an

outage independently of its level of perceived resilience, and that individual outages for each

third-party provider were independent (the fact that third party ABC has an outage has no impact

on the probability of third-party DEF suffering an outage at the same time). It was also assumed

that all reported tools work as intended. The scenario simulated one single outage (so

redundancy configurations of two providers are sufficient to absorb the shock).

191. The outcome of the scenario was the identification of the critical third-party providers to which

the CCP would be exposed in case of this scenario materializing: they represent single points of

failure tied to specific consequences ().

192. For the risk indicators, the exposure to each third-party entity was weighed taking into account

the percentage activity of the CCP they service. This weighted exposure metric was calculated

as:





  





193. With   measured using relative margin quantities

linked to the clearing activity that is serviced by the third-party entity:







 







3.6.4 Assessment of concentration or systemic risks in the network of critical third-

party providers

3.6.4.1 Overview

194. This section provides insights into the topology of the network of CCPs’ third-party providers

and performs a risk assessment of potential systemic implications from entities interconnected

to more than one CCP.

195. This analysis is divided into three sections:

• Overview of the network: overall topology of the network and analysis of most

interconnected entities.

• Analysis for specific types of services: analysis of interconnectedness segmented by types

of services.

• Evidence of events affecting multiple CCPs: analysis of past events that have affected

multiple CCPs.

3.6.4.2 Measuring interconnectedness per third party provider

196. Interconnectedness was measured by the number of CCPs connected to a third-party provider

as a share of the total number of CCPs in the stress test for which there is incident data available

(number of connected CCPs / Total number of CCPs (14)).

3.6.4.3 Assessing the risk of each interconnection

197. The risk of the interconnections was characterized using the information derived from the results

of the hypothetical scenario (described in section 3.6.3.4). Different colours were used to

represent the different levels of impact severity that the CCP would have experienced in case of

an outage of a critical third-party provider (the categories are described in section 3.6.2.5). The

colours follow a “traffic light” approach as described below:

• Risk level 0 (Grey colour): The CCP has a preventive/protective tool in place to prevent

any risk in case of an outage of the third-party.

• Risk level 1 (Green colour): In case of an outage of the third-party, the CCP would

experience an impact leading to a deterioration of its ability to achieve a specific Service

Level Agreement (Type of impact: Other Service Level Agreement breach).

• Risk level 2 (Orange colour): In case of an outage of the third-party, the CCP would

experience an impact of type: critical supporting function non available.

• Risk level 3 (Red colour): In case of an outage of the third-party, the CCP would experience

an impact of type: clearing or settlement function non available.

198. The analysis used this information both in the graph charts and in the network charts. The

interconnectedness indicators have been constructed such that:

• The percentage of interconnectedness is (number of connected CCPs / Total number of

CCPs (14)).

• The size of each colour part of the interconnectedness bar is (number of connected CCPs

to Risk level X / Total number of CCPs (14)).

3.6.4.4 Assessing the risk of each node

199. The risk characteristics of entities have not been calculated, three broad categories linked to

their regulatory status were used.

• FMI group (Purple colour node): Financial market infrastructures, payment systems,

settlement systems, central banks.

• Other financial (Blue colour node): Regulated financial institutions excluding those in the

first group (such as credit institutions, insurance undertakings or investment firms).

• Non-financial (Black colour): Entities outside of the financial regulatory perimeter (such as

providers of cloud, data or electricity).

200. In case of the hypothetical groups, all nodes are in grey colour as the groups encompass more

categories than one.

3.6.4.5 Scope of operational outages and hypothetical groups of critical third-party providers

201. Operational outages can affect the whole entity (LEI level), part of the activity of the entity, or

groups of third-party entities in the hypothetical cases of events involving interconnections

between entities (leading to propagation of operational risks) or common points of failure (such

as reliance on common systems or third-party providers).

202. The analysis focused on interconnections at LEI level, for which the results have been

calculated. Additional results involving hypothetical scenarios of groups of entities are provided.

203. The hypothetical scenarios involving groups of entities have a connecting relationship as to test

aggravated hypothetical scenarios of correlated outages within the whole group. These

relationships are based on legal relationships between entities belonging to the same group,

relationships found through their website, or possible reliance on shared infrastructure. These

relationships are built using publicly available information and expert knowledge without detailed

information on the operational aspects of the relevant entities. As such, they should be

considered as hypothetical scenarios for analytical purposes without being necessarily plausible.

4 Results

4.1 Analysis and Breakdown of Resources

4.1.1 Overview, objective of this analysis and limitations

204. The CCPs included in the scope of the exercise reported data on the required and available

financial resources used to run the credit stress test. The analysis of reported resources as

presented in this section is used to set the scene for the core credit stress test, provide an

overview of the size of the industry, the breakdown of activity by individual CCPs or participants

and identify significant changes or potential trends in the activity or risk management practices.

205. Although one can identify different practices and risk management techniques being

implemented by different CCPs, the purpose of presenting this data is not to benchmark

individual CCPs. Different CCPs clear different products with distinct characteristics. The size of

resources alone cannot indicate the effectiveness of the CCP’s risk mitigation arrangements.

The resilience of CCPs to adverse market developments is assessed using the core stress

results.

206. The data on financial resources was available for the two reference dates of the credit stress

test exercise, i.e. 19

March and 21

April 2021. It should be noted that some of the financial

resources available to CCPs, such as margin amounts, may vary significantly between different

periods depending on the activity and volatility of underlying markets. Therefore, the analysis

cannot be used to draw conclusions on the size and breakdown of resources held at other times.

207. Finally, the presented data is not always directly comparable with similar data reported in the

context of previous stress tests because the underlying data has changed, and definitions have

in many cases evolved to accommodate the scope of the present exercise. For example, the

presented amounts now correspond to pre-stress values (as opposed to post-stress values

reported in the previous exercise) and also include interoperability margin that is forwarded to

the linked CCPs. Finally, the stress test exercise includes in its scope one CCP less (LME)

compared to the previous exercise.

4.1.2 Default Waterfall

208. The amount of resources comprising the default waterfall of CCPs has overall increased

compared to the previous stress exercises. The total amount (and % share) allocated to each

tranche of the default waterfall across all CCPs in scope of this exercise and for each of the two

reference dates can be seen in Figure 6.

209. The CCPs reported in total approximately 423bn EUR of required margin, default fund

contributions and other committed prefunded resources for March and 409bn EUR for April. The

required margin alone corresponded to more than 90% of these resources and the total amount

was slightly higher in March (392bn EUR) compared to April (377bn EUR). Keeping in mind the

limitations of comparing amounts reported for different stress test exercises, it can still be noted

that the required margin has significantly increased since the previous exercise. In particular, the

required margin amount has increased by close to 26%

over the two-year period from March

2019 to March 2021.

210. Moreover, the amount of mutualised resources contributed by clearing members to the Default

Funds of all CCPs was 30.5bn EUR in March and 31.7bn EUR in April. The total amount of

default fund contributions has also increased, but less so compared to margins, if compared to

what was calculated in the previous stress test (+9% for March). Finally, the amounts of

dedicated own resources (skin-in-the-game of 0.6bn EUR) and other committed prefunded

resources (approximately 0.1bn EUR) have not materially changed since the previous exercise.

This comparison corresponds to the margins of the CCPs included in the scope of the present exercise, i.e. ESMA staff has

accounted for the fact that the present exercise includes one less CCP in its scope.

211. The data collected for the purpose of this exercise does not allow to identify the exact reasons

behind the apparent increase of total resources. However, it can be reasonably assumed that

the volatility experienced during the covid-19 crisis has played a role in this direction and which

may be further accentuated by an overall increase of the clearing activity.

212. ESMA staff has not identified any significant change in the relative allocation of prefunded

resources between the different tranches of the default waterfall, i.e. margin, skin-in-the-game

and mutualized resources.

213. The sizes of the two main tranches of the default waterfall (margin and default fund) are

presented below for each CCP. As also noticed in the previous exercises, the top CCPs are

significantly larger compared to the remaining CCPs and especially the top CCP (LCHUK)

accounted for approximately 48% of the total required margin and for 30% of the total default

fund collateral.

FIGURE 6: DEFAULT WATERFALL – ALL CCPS

214. Finally, the data shows that the allocation of resources between margin and default fund

contributions is not always proportional across different CCPs. This is also shown in the figure

below (Figure 8) illustrating (in logarithmic scale) for each CCP the required margin against the

default fund size.

FIGURE 7: REQUIRED MARGIN AND DEFAULT FUND – PER CCP

FIGURE 8: REQUIRED MARGIN VS DEFAULT FUND – ALL CCPS

215. The composition of the default waterfall of individual CCPs is illustrated in Figure 9

. Different

CCPs would rely more on margins or mutualised resources. As explained above, this alone

cannot be used to draw any conclusions on the resilience of a CCP as it can be the result of a

conservative calibration of any of the two tranches. If compared to the previous exercise, there

One CCP (ICENL), reported zero margin requirements and zero exposures.

is no significant structural change. It seems that, overall and with only a few exceptions, the

CCPs that were relying to a larger extent on mutualized resources, continue to do so.

216. Overall, it is again confirmed that smaller CCPs tend to have a larger share of their coverage

stemming from mutualized resources. This could be partly explained by the fact that since all

CCPs have to meet as a minimum and independently of their size a cover-2 requirement, the

risk-sharing (mutualised) part is expected to generally be smaller for CCPs that have a larger

number of clearing members and smoother allocation of exposures across their top participants.

Of course, other factors that are linked to the cleared products also play a significant role such

as the comparison of the severity of adverse (to be covered by margins with a minimum of

99%/99.5% confidence level) versus stress market conditions (such as most extreme historical

moves to be covered by mutualised resources).

FIGURE 9: DEFAULT WATERFALL – PER CCP

217. In the context of the credit stress test, the assessment of the resilience of the CCPs will be

based on the required margin collateral, as it can be assumed that a member in distress would

not have posted any excess collateral, while excess margin of non-defaulting members can

anyway not be used to cover losses. Nevertheless, for completeness the CCPs were asked to

also report the (post-haircut) amount of collateral that was actually provided in order to give an

indication of the actually available margin collateral on a particular date even where this is in

excess of the minimum margin requirement. Figure 10 illustrates the required margin amount in

comparison to the actually provided (post-haircut) collateral value, i.e. including excess margin

amount.

FIGURE 10: REQUIRED VS EXCESS MARGIN

218. Overall, the excess margin collateral corresponds to a relatively small percentage (14%-15%)

of the total provided margin collateral. If compared to the previous exercise, the share of excess

collateral has not changed significantly (was 16%). Also, on a per CCP basis, overall, the CCPs

that were collecting a large amount of excess collateral continue to do so with only a few

exceptions. Although not always the case, it can again be noted that it is mostly the smaller CCPs

and especially the ones that clear cash equities that have an exceptionally large share of excess

collateral. This can be attributed to the fact that exposures can change significantly from one day

to the other. Members prefer to over-collateralize their exposures, in order to avoid having to

provide additional collateral on an intraday basis.

4.1.3 Clearing Members

219. ESMA staff has also analysed the distribution of required margins and default fund contributions

across the clearing participants providing them in order to investigate if the increased financial

resources were contributed by only a few top members, or if there was a general increase of

provided resources by all participants. The analysis presented below indicates that while all

clearing members have in general provided more financial resources, the top participants have

done more so and have increased their relative share.

220. In this exercise ESMA staff identified approximately 750 clearing members (single entities) being

a member in at least one CCP, which is smaller than the number of entities identified in the

previous exercise

. However, there is no strong evidence of this being a general trend as for 6

out of the 15 CCPs the number of clearing members actually increased.

221. Of course, many of the reported entities are at the same time members at multiple CCPs. For

example, 6 entities are at the same time clearing members at 10 CCPs or more. The number of

clearing members and cumulative share (percentage), in terms of their aggregate contribution to

the Default Fund of all CCPs and in terms of the aggregate margin required again from all CCPs

(Figure 11) are presented below. For comparison, one can see in light grey the relevant share of

resources under the previous (3

) stress test exercise and in yellow the corresponding share

under the 2

stress test exercise

222. First, it is observed that more members have very high margin requirements. In particular, the

number of clearing members having each a total margin requirement (across all CCPs) greater

than 10bn EUR has increased further, i.e. 7 members in this exercise compared to 5 in the 3

exercise and none in the 2

exercise. Similarly, more members have now a required margin

amount greater than 5bn EUR

and these account in total for 57% of the margin provided to all

Approximately 800 single entities were identified in the previous exercise. Please note that this includes members from one

additional CCP (LME). After the removing the entities that were only members at this CCP, the total number of entities was close

to 790.

It is noted that the shares reported here stemming from different exercises are not 100% comparable, including because the

and 3

exercises reference one additional CCP.

24 in the present exercise compared to 16 in the 3

exercise and 14 in the 2

CCPs compared to 47% in the previous exercise and 39% in the 2

exercise. This observed

increase could be driven by the general increase of required resources, i.e. members in general

required to provide more resources due to the overall increase of CCP margin, as the comparison

is done across constant buckets. However, this trend is further analysed below and it seems that

while the general increase of resources has pushed the top participants to higher levels, at the

same time, these participants have also increased their relative share. The clearing member

(single entity) with the highest required margin amount across all CCPs had a total margin

requirement of approximately 18bn EUR.

223. At the same time, no significant change is observed to the distribution of default fund

contributions. Focusing on the highest contributors, e.g. members with an aggregate contribution

across CCPs of more than 500m EUR, it can be seen that the number of members and their

share have not changed significantly compared to the previous exercise

FIGURE 11: CLEARING MEMBERS – ALL CCPS

17 members with a contribution of 500m EUR or more in both exercises (2021 and 2019) accounting for 42% of the default

funds in 2021 and 46% in the previous exercise.

224. This analysis was also run at the group level, after adding the resources provided by all affiliates

within a single corporate group. As expected, the level of concentration increases further (Figure

12), while the amounts calculated for the top groups confirm the conclusions drawn above.

225. The top Clearing Member Groups have contributed more resources, both in absolute and in

relative terms. In particular, 5 clearing member groups have each provided more than 20bn EUR

in required margin across all CCPs, compared to 3 groups in 2019 (3rd exercise) and none in

2017 (2nd exercise). These groups account now for approximately 39% of total margin

requirement, compared to 24% in the precious exercise. At the same time, there is no significant

change in the number of groups than have a required margin that is higher than 10bn EUR

Hence, it seems that the members belonging to the top groups increased their share of margin

compared to the remaining members, thus meaning that clearing activity becomes more

concentrated between main players. This is confirmed in Figure 13 where it is shown that the

top-5 groups provided 39% of the total required margin in 2021 vs 36% in 2019 (3rd exercise).

The group with the maximum margin requirement across CCPs had a total margin requirement

of approximately 36 billion EUR.

226. Furthermore, no significant change was observed in the distribution of default fund contributions,

similarly to what was observed at single entity level. The group with the largest aggregate

contribution to default funds had a total default fund contribution across CCPs of 2.5 billion EUR.

FIGURE 12: CLEARING MEMBER GROUPS – ALL CCPS

10 groups (61% of margin) in 2021 compared to 11 groups (62% of margin) in 2019.

FIGURE 13:

CLEARING MEMBER GROUPS - DISTRIBUTION OF REQUIRED MARGIN SHARES

4.2 Credit Stress Test Results

227. This section presents the full range of credit stress test results assessing the sufficiency of

CCPs’ resources to absorb losses under a combination of market shocks and member default

scenarios. From a credit risk perspective, a combination of clearing member defaults and

simultaneous severe shifts of risk factor prices, including those due to high concentration or

specific wrong-way risk, is needed to put a CCP at risk. If clearing members continue to post

margin and meet their obligations, periods of extreme market volatility in isolation will not pose a

specific market risk to a CCP. Similarly, defaults of clearing members without simultaneous

adverse price shocks should not put a CCP at risk

. Under normal market conditions, the CCPs

will have the resources to withstand multiple defaults. Hence, from a credit risk perspective and

with the exception of investment risks, only simultaneous defaults and extreme, adverse shifts

of market prices could pose potential risks to a CCP.

228. First, “cover-2 per CCP” results are presented (4.2.1) where two clearing member groups are

assumed to be in default separately at each CCP and then “All CCPs cover-2” results are

discussed (4.2.2), where the default of two groups across all CCPs is assumed, i.e. the same

two groups for all CCPs. All these results are separately presented for the two dates that are

covered in this exercise

. The methodology used in the credit stress test, including the design

and assumptions of the market and member default scenarios, is detailed in Section 3.4. Where

needed to provide additional insight, results are discussed using alternative defaulters’ selection

conditions and also the estimated impact from increased shocks is explored in the form of a

sensitivity analysis.

229. One of the innovations of this stress test exercise is that for one of the dates ESMA staff has

included in the calculations the impact due to concentration and specific wrong-way risk

stemming from cleared positions. In order to be able to include those additional potential costs,

results have been recalculated starting from the data reported at individual account-level and

P&L calculations have been propagated through the account structure and default waterfall of

each CCP. So, for the March date and the baseline market shocks results are also reported after

adding the concentration and wrong-way risk impact.

4.2.1 Cover-2 per CCP Credit Stress Test Results

230. The “Cover-2 groups per CCP” member default scenario is designed to independently assess

the resilience of each CCP to the Market Stress Scenario, focusing on the worst outcome for

each CCP.

231. In accordance with the methodology, ESMA staff selects for each CCP individually two (2)

corporate groups and assume that all the clearing members belonging to those 2 groups would

default in the same CCP. The selected clearing member groups and defaulting entities will be

different for each CCP and are not considered to be in default in other CCPs. The results for

each CCP come from an independent selection of defaulting groups that don’t propagate to other

CCPs, therefore the interpretation should be limited to the assessment of the resilience of

individual CCPs under a common market stress scenario.

Clearing members post margins and default fund contributions scaled to a very high confidence level. This should make sure

that CCPs have sufficient resources to manage a default of a clearing member in normal market conditions, and close out the

resulting open positions in a stable market before suffering a loss.

19 March and 21 April 2021.

BOX 3: DESCRIPTION OF THE CREDIT STRESS TEST CHART

The credit stress test results are always presented in the form of a panel, showing for each CCP the

following (from bottom to top):

Amounts of default waterfall consumption (in mil. EUR)

Loss covered with DF, SITG and other DF-level Resources: Amount of stress loss (in million EUR)

covered with the Default Fund (including defaulting and non-defaulting members’ contributions),

dedicated CCP resources (“skin-in-the-game”) and other prefunded and committed Default-Fund-

level Resources that the CCP may have. Where the CCP has more than one Default Fund, this

amount is the sum of amounts calculated per Default Fund. It is illustrated in green in the chart.

Loss covered with other CCP-level Resources: Amount of stress loss (in million EUR) covered with

other prefunded and committed CCP-level resources, where applicable. The CCP-level resources are

resources that can be used across default funds where the CCP has more than one default funds. It

is illustrated in yellow in the chart.

Loss covered with PoA: Amount of stress loss (in million EUR) that would need to be covered with

non-prefunded resources (powers of assessment). Where the CCP has more than one Default Fund,

this amount is the sum of amounts calculated per Default Fund. Only the non-defaulting members are

assumed to provide additional non-prefunded resources. It is illustrated in red in the chart.

Loss after PoA: Amount of stress loss (in million EUR) left uncovered after using prefunded and non-

prefunded resources. This amount is again the sum of all uncovered amounts where the CCP has

more than one Default Funds. It is illustrated in black in the chart.

% Consumption of Resources

% Consumption of the Default Fund (including the defaulters’ contributions), the skin-in-the-game and

other prefunded and committed Default-Fund-level Resources that the CCP may have. For CCPs that

have more than one default funds, the maximum % consumption is presented.

% Consumption of Powers of Assessments (called only from non-defaulting members). For CCPs

that have more than one default funds, the maximum % consumption is presented

Two flags

A flag indicating (in red) whether non-prefunded resources would have to be used.

A flag (top of the panel) indicating (in black) whether there would be uncovered losses after using

also non-prefunded resources.

4.2.1.1 Cover-2 per CCP Results for March 2021

Cover-2 Groups per CCP (no cross defaulting)

Date: March 2021 – Without Excess Margin

FIGURE 14: COVER-2 GROUPS PER CCP – DATE: MARCH 2021 – WITHOUT EXCESS MARGIN

232. The core credit stress test results for March do not indicate a shortfall of prefunded resources

for any of the CCPs in scope of the exercise. The maximum stress loss above margin is

approximately 1.5bn EUR (ECAG) and the maximum % consumption of financial Default Fund-

level resources available to cover losses beyond margin was 45% (CCPA). In terms of losses in

monetary (EUR) amounts, the largest losses are naturally calculated at the bigger CCPs with the

three largest amounts found at ECAG with 1,549 million EUR, LCHUK with 1,309 million EUR

and ICEEU with 895 million EUR. Yet all three CCPs had sufficient prefunded resources to cover

such losses, with relatively low % consumptions of available resources.

233. Since there was no shortfall of prefunded resources, the defaulters’ selection algorithm focused

on the pair of groups that would maximise the stress losses above margin. When ESMA staff

instead selected the pair of defaulters to maximise the % consumption of financial resources,

with the objective to identify cases where the scenario could put significant pressure to smaller

default funds without necessarily maximising the total amount of losses, ESMA staff did not find

any such cases and the results changed materially only for one CCP (Nasdaq) with the %

consumption increasing but only to moderate levels, i.e. from 3% to 30%.

234. Hence, the implemented market stress scenario, before accounting for any additional losses

due to concentration and wrong-way risk, has not put for the March date any of the in-scope

CCPs to significant stress and all CCPs had sufficient prefunded resources to cover such losses.

235. Overall, aggregating the independent cover-2 results of the different CCPs there is a volume of

approximately 4.1 billion EUR of losses after required margin and 3.3 billion EUR of losses after

required margin and defaulters’ default fund contributions. These amounts give an indication of

how impactful the scenario is, but it should be noted that this is not a scenario that could be

realised at the same time across all CCPs. It aggregates the worst results that were produced

per CCP and assumes the default of different groups at different CCP.

236. For completeness, results using excess margin are also presented. The excess margin consists

of collateral that was actually available at the CCPs on these particular dates. It was provided by

the clearing members in excess of the required margin amounts. The rationale of not including

excess collateral in the base scenarios is that it would not be prudent to assume that a member

in default would have actually provided on the previous day any collateral in excess of the

minimum requirement. The cover-2 per CCP results for the March date using excess margin are

reported in Figure 15. It is noted that the selection of top defaulting entities is always performed

using only the required margin collateral. The same defaulting entities are considered when

reporting the results with total (i.e. including excess) collateral.

Cover-2 Groups per CCP (no cross defaulting)

Date: March 2021 – With Excess Margin

FIGURE 15: COVER-2 GROUPS PER CCP – DATE: MARCH 2021 – WITH EXCESS MARGIN

237. After considering the excess margin, the losses would be smaller for many but not all CCPs.

The maximum loss above total margin is significantly reduced to 937 million EUR while the

maximum % consumption of Default-Fund-level mutualised resources is the same (45%)

indicating that the assumed defaulters had not provided excess margin for this CCP. Overall,

aggregating the independent cover-2 results of the different CCPs the volume of resources

above margin that would be consumed decreased to approximately 2.8 billion EUR, compared

to the 4.1 billion EUR calculated using only the required margin.

238. As with any exercise of this type, the magnitude of the market stress shocks that would be

needed to adequately reflect extreme but also plausible conditions in a forward-looking basis is

subject to uncertainties. Therefore, ESMA staff has also explored the impact of small or moderate

changes to the assumed shocks. The CCPs were asked to report the results not only for the

common market scenario shocks, but also after applying a number of multipliers on the shocks

For each value of the multiplier, the CCPs ran a full repricing of the portfolios, as opposed to

applying a multiplier to the result (P&L) of the scenario. All shocks are simultaneously scaled for

all risk factors. For the purpose of this analysis, the results were calculated after considering all

shocks increased by 20% and by 50%. Acknowledging the severity of the shocks and the fact

that it goes beyond what was considered as extreme but plausible in the context of this exercise,

it should be noted that the rationale of this analysis is not to put the focus on specific CCPs but

rather investigate if relatively small increases of the shocks could lead to systemically relevant

changes on the results of individual CCPs. A similar analysis is performed under the reverse

stress test component that is expanded in two dimensions, being the severity of the shocks and

the number of defaulting groups. The key difference is that the reverse stress analysis focuses

on the internally consistent “All CCPs member default” scenarios, i.e. select the same groups as

defaulting across all CCPs. Here ESMA staff selects the worst two groups per CCP and thus

tries to identify any systemically relevant impact at individual CCPs. For March, when moves that

are 20% or even 50% more severe than the baseline stress shocks were assumed, there would

still be no shortfall at any CCP. The 20% increase of the shocks led to a maximum loss above

required margin of approximately 2.9bn EUR, which is compared to the 1.5bn EUR of the

baseline scenario. The % consumption also remains moderate with a maximum % consumption

of default funds and other mutualised resources equal to 62%. This indicates that the conclusions

seem robust to small changes in the baseline shocks. When even more severe shocks were

assumed, i.e. increase of baseline shocks by 50%, the CCPs were subject to significant

pressure. There would still be no shortfall of prefunded resources, however, one CCP would

have had exhausted the resources dedicated to one smaller default fund and would have had to

use a very small amount (<1m EUR) of other CCP-level prefunded resources. For three other

CCPs the consumption of default fund, skin-in-the-game and other prefunded resources would

be greater than 80%. The maximum loss over required margin at a single CCP would be close

to 5.6bn EUR (double compared to the +20% scenario). As explained, considering the fact that

it goes well beyond what was considered as an internally consistent, extreme but plausible

scenario in the context of this exercise, these results raise no additional concerns.

239. The following figure (Figure 16) illustrates the baseline cover-2 per CCP results after also adding

the impact from the liquidation of concentrated positions, as this is calculated according to the

concentration component. The methodology used to incorporate the concentration cost and

select the top default parties is detailed in Section 3.4.3.3.

The multipliers used are x0.7, x 1.2, x1.5 and x2.0, implying an increase of the baseline shocks by 20% to 100%.

Cover-2 Groups per CCP (no cross defaulting)

Date: March 2021 – with Concentration impact

FIGURE 16: COVER-2 GROUPS PER CCP – DATE: MARCH 2021 – WITH CONCENTRATION IMPACT

240. As expected, the addition of the concentration impact leads to higher losses and consumption

for almost all CCPs. However, there is still no shortfall of prefunded resources with the %

consumption of default-fund level mutualised resources reaching 79% (Nasdaq). Of course, the

maximum amount of losses above required margin is increased from 1.5bn EUR (under the

baseline scenario without the concentration impact) to 2.3bn EUR. The aggregate (across all

CCPs) amount of losses above margin add to approximately 6bn EUR, compared to 4.1bn EUR

under the baseline – without concentration impact - scenario. Hence, the addition of the

concentration impact increases significantly the losses, but under the considered market

scenario, these are contained within the default waterfalls of the CCPs. This impact is further

discussed below after also adding the enhanced wrong-way risk cost.

241. The baseline “Cover-2 per CCP” results after also adding the wrong-way risk adjustment for

cleared positions are presented in the following figure (Figure 17). The methodology used to

calculate the wrong-way risk and relevant assumptions are detailed in paragraph 3.4.3.4.

Cover-2 Groups per CCP (no cross defaulting)

Date: March 2021 – with Concentration and Wrong-way risk impact

FIGURE 17: COVER-2 GROUPS PER CCP – DATE: MARCH 2021 – WITH CONCENTRATION AND WWR

IMPACT

242. When also considering the wrong-way risk on top of the concentration impact for the March

date, the losses above required margin increased for 3 CCPs without leading to a shortfall of

prefunded resources at any CCP. The impact from wrong-way risk is significant under the

considered scenario only for one CCP (Nasdaq), as it clears covered bonds. The maximum loss

above margin (2.3bn EUR) did not change as the CCP (LCHUK) showing this loss did not

experience any wrong-way risk losses under the considered member default scenario. The

maximum % consumption is now 65%, i.e. lower, and also for a different CCP compared to the

concentration-only run, which seems counter-intuitive. This is because the defaulters’ selection

algorithm now focuses on a different pair of members that maximises the total amount of losses

above margin for a different (larger) default funds leading to lower % consumption.

243. When instead selecting groups maximising the % consumption, there would be 100%

consumption of the default fund-level resources at a smaller default fund of one CCP (Nasdaq)

with very small residual losses (<2m EUR) still covered fully by the additional available prefunded

CCP-level resources. This impact was driven by additional concentration costs. Furthermore,

under this selection, there would be increased % consumption for two other CCPs, i.e. Keler

(90%) for a smaller default fund due to concentration and ATHX (6%) due to concentration and

wrong-way risk, but in both cases the loss above margin was again very small (<1m EUR).

244. Hence, the addition of the concentration and wrong-way risk impact in the considered scenarios

did not raise any systemically relevant concerns. However, it should be noted that this impact

was added to the P&L calculated from the baseline market stress scenarios. Therefore, there

may be cases where this additional cost, though significant, would be added to accounts that

would experience profits from the given market scenario, muting the impact from these additional

risks. Hence, these results cannot be used to draw conclusions on what the impact would be

under all possible extreme but plausible market scenarios. The CCPs should have dedicated risk

management measures to prudently mitigate these risks. Finally, the potential impact from

increased concentration, independently from the market scenario, is assessed in the

concentration component (4.3).

4.2.1.2 Cover-2 per CCP Results for April 2021

Cover-2 Groups per CCP (no cross defaulting)

Date: April 2021 – Without Excess Margin

FIGURE 18: COVER-2 GROUPS PER CCP – DATE: APRIL 2021 – WITHOUT EXCESS MARGIN

245. For the April date, the cover-2 per CCP scenario does not generate a shortfall of prefunded

Resources at any CCP. The maximum loss above required margin is 1.5bn EUR, very close to

what was calculated for the March date, but for a different CCP (ICEEU). The maximum %

consumption of default fund-level mutualised resources is 56% (ICEEU).

246. In general, if compared to March the impact is mixed across CCPs. For 8 CCPs the losses over

required margin are lower with the largest positive difference being +0.6bn EUR (ICEEU) and

the largest negative difference equal to -0.7bn EUR (ECAG). However, in aggregate across all

CCPs, the scenario generated similar losses over required margin, i.e. 4.2bn EUR compared to

4.1bn EUR calculated for the March date. Also, the impact on non-defaulting members is similar,

since the losses after exhausting the defaulters’ resources are 3.4bn EUR compared to 3.3bn

EUR for March.

247. Some CCPs present higher losses, but overall, it seems that the intraday default assumption as

implemented for the April reference date did not put significant additional stress on the resilience

of the system of CCPs. Having said that, it is important to note that the results are based on the

data reported by CCPs implementing the intraday default assumptions. Sourcing positions and

resources from intraday data has added significant complexity both on CCPs and authorities

validating the data, which means that, also as with any newly tested stress assumption, some

uncertainties may remain with regard to their consistent implementation.

248. Finally similar to March, ESMA staff has calculated the results with increased shocks for April in

an effort to explore whether small changes could dramatically impact the conclusions; this does

not seem to be the case. When shocks were increased by 20%, the maximum loss above

required margin increased from 1.5bn EUR (baseline scenario) to 3.2bn EUR. Increased

pressure would be noted especially for one CCP with a % consumption of prefunded resources

that was close to 81%. However, there was no shortfall of prefunded resources at any CCP, and

the conclusions seem again robust to small changes in the underlying shocks. On the other hand,

where shocks were increased significantly by 50%, multiple CCPs would be subject to significant

pressure with 4 CCPs showing consumption of prefunded resources higher than 80%. Moreover,

there would be a shortfall of prefunded resources at one CCP for approximately 0.5bn EUR that

could still be covered with additional non-prefunded resources that the CCP has the right to call

from non-defaulting members. As explained before, the purpose of this analysis is to explore the

impact of small changes and considering the fact a 50% increase would go well beyond what

was considered as an internally consistent, extreme but plausible scenario in the context of this

exercise, these results raise no additional systemic concerns.

4.2.2 All CCPs Cover-2 Credit Stress Test Results

249. The “All CCPs Cover-2” credit stress test is designed to assess the resilience of CCPs

collectively to the Market Stress Scenario, focusing on the worst outcome for the whole system

of CCPs.

250. In this scenario, all members belonging to the same two clearing member groups are assumed

to be in default at all CCPs in scope of the exercise. The selection of the top-2 defaulting groups

is based on the aggregate impact to prefunded resources, considering all CCPs. Given that the

selection of defaulters is the same for all CCPs, the results illustrate what would be the systemic

effect of the most impactful default of two clearing member groups and how it would affect each

CCP.

251. The “All CCPs Cover-2” results are reported separately for the two dates, based on the same

format that was used for the “Cover-2 per CCP” results. For the March date results after

considering the impact from concentration and wrong-way risk are also presented.

4.2.2.1 All CCPs Cover-2 Results for March 2021

252. The “All CCP cover-2” results for March after selecting the two groups that would maximize the

overall impact on prefunded resources are reported in the figure below (Figure 19).

All CCPs Cover-2 (top-2 groups CCP-wide, defaulting at all CCPs)

Date: March 2021 - Without Excess Margin

FIGURE 19: ALL CCPS COVER-2 – DATE: MARCH 2021 – WITHOUT EXCESS MARGIN

253. As expected, there is no common pair of defaulting groups that would lead to a shortfall of

prefunded resources at any of the CCPs. The individual results of each CCP are by design

equally or less severe than the results calculated under the “cover-2 per CCP” assumption. The

reason is that here ESMA staff selects the same two clearing member groups as defaulting

across all CCPs.

254. The two defaulting groups were selected to maximise the aggregate, across all CCPs, loss

above required margin in order to assess the impact on the system of CCPs. In this case, the

selected pair of defaulting groups is none of the pairs that would maximise the losses at any

individual CCP. The algorithm focuses on a pair that maximises the aggregate impact across all

CCPs. We note that by prioritizing maximization of losses above required margin in absolute

terms, the Cover 2 selection naturally leans towards combinations that are most impactful for the

largest CCPs.

255. The total loss above required margin is close to 2.2bn EUR with 1bn EUR being the maximum

loss at a single CCP. This scenario does not put significant stress to any CCP with the %

consumption of default fund-level prefunded resources being less than 20% in all cases.

However, 10 of the CCPs would have experienced at least one member default with 5 CCPs

having to use prefunded resources beyond collateral (margin or default fund contribution) of the

defaulter(s)

. While this confirms on the one hand that CCPs are highly interconnected through

common clearing participants, the exercise did not highlight any pairs of groups that are at the

same time and under the common tested scenario highly impactful at multiple CCPs.

256. Similar to the “cover-2 per CCP” scenario, ESMA staff also presents here (Figure 20) the “All

CCPs cover-2” results after considering the impact from concentration and wrong-way risk.

All CCPs Cover-2 (top-2 groups CCP-wide, defaulting at all CCPs)

Date: March 2021 - Without Excess Margin – With Concentration

and wrong-way risk Impact

FIGURE 20: ALL CCPS COVER-2 – MARCH 2021 – WITHOUT EXCESS MARGIN – WITH CONCENTRATION

AND WRONG-WAY RISK IMPACT

257. After adding the concentration and wrong-way risk impacts, the algorithm used to select the top

defaulting groups focuses on the same defaulting pair that maximises again losses across CCPs.

258. The total loss above required margin increases by approximately 0.9bn EUR to 3.1bn EUR. This

increase is only due to the addition of concentration cost, as there is no wrong-way risk impact

stemming from the default of the selected pair. There is no shortfall of prefunded resources. In

fact, the impact on the default waterfall of CCPs is still limited with the % consumption of default

fund, skin-in-the-game and other prefunded default fund-level resources being lower than 25%.

And 6 CCPs having losses beyond required margin of the defaulter(s) as shown in the figure.

259. Hence, even after adding the concentration cost, the exercise did not highlight any pairs of

groups that would at the same time and under the common tested scenario cause a significant

impact at multiple CCPs. The limitations explained before still hold. The results may be sensitive

to the underlying market scenario, should be used with caution when drawing general

conclusions and CCPs should have dedicated risk management measures to prudently mitigate

these risks. The potential impact from increased concentration, independently from the

underlying market scenario, is assessed in the concentration component (4.3).

4.2.2.2 All CCPs Cover-2 Results for April 2021

260. The All CCPs cover-2 results for the April date after selecting for the groups that would maximize

the overall impact, are presented in the figure below (Figure 21).

All CCPs Cover-2 (top-2 groups CCP-wide, defaulting at all CCPs)

Date: April 2021 - Without Excess Margin

FIGURE 21: ALL CCPS COVER-2 – APRIL 2021 – WITHOUT EXCESS MARGIN

261. The pair maximising the losses across CCPs is again none of the pairs that would maximise the

losses at any individual CCP. The calculated total loss above margin is 2.3bn EUR, very close

to what was observed for the March date before adding the concentration cost. The maximum

loss above margin is slightly higher, i.e. 1.2bn EUR, and for a different CCP (ICEEU), also leading

to a higher % consumption of 44%.

262. Under the selected scenario, 10 CCPs would experience a default of at least one of their clearing

members. However, the majority of losses would stem from two of the bigger CCPs and although

there would be losses above required margin of the defaulters for 7 CCPs, only 3 CCPs would

experience losses above the collateral (required margin default fund contribution) provided by

the defaulting parties. Hence, the exercise did not identify any pairs of groups whose default

would at the same time have a significant impact at multiple CCPs under the common tested

scenario.

263. For the April date, the concentration and wrong-way risk costs were not considered as the data

was not provided at account level. The sensitivity of the “All CCPs cover-2” results to changes in

the market shocks and number of defaulting entities is analysed in the following section as part

of the reverse stress test.

4.2.3 Reverse Credit Stress Test Results

264. The reverse stress test in the 4

stress test exercise has similar characteristics with the analysis

performed in previous exercises. For the reverse stress tests, ESMA staff performs a two-

dimensional analysis of the absorption capacity of the system of CCPs by stepwise increasing

the number of defaulting groups and the severity of the market shocks in order to identify at

which point resources are exhausted.

265. While exploring the different combinations, ESMA staff goes intentionally beyond what is

considered as plausible for the purpose of this exercise. The idea is to capture the sensitivity of

the results to the considered stress scenarios and understand how the results are affected by

changing the underlying assumptions. After all, although the baseline stress scenario is carefully

modelled to simulate extreme market conditions, it is still subject to uncertainties and limitations,

as is the case with all modelling procedures. A steep increase of the uncovered losses following

a relatively small change in the shocks could indicate a high sensitivity and raise concerns on

the robustness, considering the limitations and uncertainties.

266. With respect to the number of defaulting groups, ESMA staff considers the default of the top-n

clearing member groups, where n ranges from one (1) to five (5) groups. All entities belonging

to these groups are considered to be in default across all CCPs. The selection of defaulting

groups for each combination of severity level and number of defaulting groups is done by an

algorithm that selects the groups that maximize the losses over prefunded resources. The

selection is done independently for every combination of severity level and number of defaulting

groups. The selection of groups is performed without considering excess margin and is looking

for the greatest loss over prefunded resources. The same selected groups are then used for the

analysis of losses over non-prefunded resources.

267. The different severity levels are the result of adjusting the base scenario shocks using a number

of multipliers

. At each severity level, the shocks of all risk factors are adjusted simultaneously.

The five severity levels are the following:

x0.7: A decrease in the stress test shocks of 30%.

Base: The base scenario shocks as used for the credit stress test.

x1.2, x1.5, x2: An increase in the stress test shocks of 20%, 50% and 100% respectively.

For each value of the multiplier, the CCPs ran a full repricing of the portfolios, as opposed to applying a multiplier to the result

(P&L) of the scenario

268. The following two tables of results are presented both using only required margin:

The “Loss above Required Prefunded resources” table (Table 1) presents the aggregate (across all

CCPs) amount of losses (in billion EUR) beyond prefunded resources, as applicable. These include

required margin collateral, “skin-in-the-game”, default funds and other Default-Fund-level resources

The “Loss above Required & non-Prefunded resources” table (Table 2) presents the aggregate (across

all CCPs) amount of losses (in billion EUR) beyond prefunded and non-prefunded resources (Powers

of Assessment).

TABLE 1: REVERSE STRESS TEST – LOSS ABOVE REQUIRED PREFUNDED RESOURCES (NO EXCESS)

269. The amounts shown in the table are the losses beyond prefunded resources in billion EUR

assuming the default of the same groups across all CCPs. So, this is an extension of the “All-

CCPs cover-2” member default scenario.

270. When the severity of the shocks is only increased by 20% and without increasing the number of

defaulting clearing member groups, i.e. stay at the regulatory requirement of cover-2 defaulting

groups, there is no shortfall of prefunded resources for any of the two dates. A further increase

of the severity of the shocks, i.e. by 50%, would lead to a shortfall of 0.5bn EUR at one CCP for

April which was already discussed in the context of the sensitivity of the “Cover-2 per CCP”

results in 4.2.1.2. When the shocks are increased by 100%, the maximum shortfall of prefunded

resources that would be inflicted by the default of a pair of clearing member groups would be

5.4bn EUR for March and 5.2bn for April, and in both cases caused by one CCP.

271. In case of the default of one clearing member group (cover-1 defaulting group), a very small

shortfall can be noted when the underlying shocks are increased by 100% (doubled). For the

April date, this is a shortfall of approximately 50m EUR at one CCP. For March, the shortfall

observed in the table is simply because of not including other CCP-level resources in this reverse

The other CCP-level resources have not been considered in the reverse stress test in order to simplify the calculation, but the

impact from this assumption is assessed as immaterial in the context of the reverse stress test.

The cell is highlighted in red where there is a non-zero loss. There are cells where the loss is small (<0.1), but still greater than

0 and is thus highlighted.

x0.7 Base x1.2 x1.5 x2

1 - - - - 0.0

2 - - - - 5.4

3 - - - 1.9 10.6

4 - - - 3.8 15.3

5 - - - 5.4 20.3

1 - - - - 0.1

2 - - - 0.5 5.2

3 - - 0.1 1.9 10.7

4 - - 0.5 3.9 16.9

5 - - 0.8 6.0 21.8

March 2021

April 2021

Number of Groups Defaulting

Market Shock

Loss above Required Prefunded Resources (bil. EUR)

stress test analysis. In practice, this shortfall would be covered by the available prefunded

resources.

272. Following a small increase of the shocks, i.e. by 20%, there is no shortfall for March even when

the number of defaulting clearing member groups is increased to 5. On the other hand, for April

there is a small shortfall of 60m EUR at cover-3 (three defaulting clearing member groups).

273. Finally, without increasing the shocks (baseline common market stress scenario), there is no

shortfall of prefunded resources at any CCP for any of the dates even when 5 clearing member

groups default simultaneously.

274. In the following table (Table 2) one sees the shortfalls after accounting also for the non-

prefunded resources that the CCPs have the right to call. Of course, one should note that each

CCP uses different definitions, assumptions and conditions, when setting the maximum amounts

that can be called. These may include for example specific cool-off periods, distinction between

simultaneous and sequential defaults, limited scope of use for resources and different priorities

amongst clearing members depending on the source of the default event. Therefore, any effort

to use a harmonised modelling approach in order to analyse such a severe impact across CCPs

can only serve as a rough approximation.

TABLE 2: REVERSE STRESS TEST – LOSS ABOVE REQUIRED & NON-PREFUNDED RESOURCES (NO

EXCESS)

275. It can be seen that significantly more extreme assumptions would be needed in order to create

a shortfall of non-prefunded resources. Under the considered scenarios, a cover-5 assumption

(five defaulting clearing member groups) in combination to a 100% increase (x2) of the market

shocks was necessary in order to have a shortfall of non-prefunded resources. However, it

should be noted that this scenario would already involve a very large amount of non-prefunded

resources (up to 19bn EUR) being called from non-defaulting clearing members and used to

cover losses.

276. From the analysis of the reverse stress test results ESMA staff has not found any systemically

relevant adverse impact following small changes in the underlying stress assumptions. It is also

confirmed that incremental changes in the severity of the market shocks are generally more

harmful than increases in the number of defaulting groups. For very large increases of the

severity of the market shocks, going well beyond what was considered extreme but plausible in

the context of this exercise, the observed maximum shortfalls of prefunded resources following

x0.7 Base x1.2 x1.5 x2

1 - - - - -

2 - - - - -

3 - - - - -

4 - - - - -

5 - - - - 2.1

1 - - - - -

2 - - - - -

3 - - - - -

4 - - - - -

5 - - - - 2.6

March 2021

April 2021

Number of Groups Defaulting

Market Shock

Loss above Required & Non-Prefunded Resources (bil. EUR)

the default of two clearing member groups would not be spread across CCPs implying that there

are different pairs of defaulting groups that would maximise the shortfalls at different CCPs for

these particular dates.

277. One of the key limitations of this analysis is that second round effects are increasingly relevant

as scenarios become more extreme, beyond what can be reasonably considered as plausible.

However, as in the core credit stress test, second round effects are not accounted for. It should

be highlighted that in practice the wide-spread effects from such catastrophic events in the

financial system cannot be analysed fully only considering the CCPs and the cleared exposures.

Therefore, due to its limited scope, this analysis cannot predict the impact from such events. Its

purpose is to assess the sensitivity of the CCP stress results to relatively small changes in the

scenarios and underlying assumptions.

4.2.4 Conclusions of Credit Stress Test Results

278. In the credit stress test, ESMA staff analysed the sufficiency of CCPs’ resources to withstand

the losses resulting from hypothetical multiple clearing member defaults combined with

simultaneous extreme price changes. The core “Cover-2 per CCP” credit stress test results for

the two dates did not indicate a shortfall of prefunded resources at any of the CCPs in scope of

the exercise.

279. The implemented market stress scenario, especially before accounting for any additional losses

due to concentration and wrong-way risk, has not put any of the in-scope CCPs to significant

stress and all CCPs had sufficient prefunded resources to cover such losses. The CCPs could

have covered losses generated by the common market stress scenario with relatively low or

moderate % consumptions of available resources. For the April date ESMA staff tested using an

intraday default assumption. Some CCPs presented higher losses, but overall, it seems that this

modelled assumption did not put significant additional stress on the resilience of the system of

CCPs. Having said that, sourcing positions and resources from intraday data has added

significant complexity both on CCPs and authorities validating the data, which means that, as

with any newly tested stress assumption, some uncertainties may remain with regard to their

consistent implementation.

280. ESMA staff also performed a sensitivity analysis to explore the impact of small or moderate

changes to the assumed shocks. When assuming moves that are 20% higher than the baseline

stress shocks, there would still be no shortfall at any CCP. This indicates that the conclusions

seem robust to small changes in the baseline shocks. When the shocks were increased further

(+50% from baseline), ESMA staff noted increased pressure to CCPs including a shortfall of

prefunded resources for one CCP for one of the dates. However, considering the fact that such

assumed shocks would go well beyond what was considered as an internally consistent, extreme

but plausible scenario in the context of this exercise, these results raise no additional systemic

concerns.

281. For one of the dates, ESMA staff included in the baseline scenario calculations the impact due

to concentration and specific wrong-way risk stemming from cleared positions. This led to higher

losses and consumption for almost all CCPs but under the considered market scenario these

were contained within the default waterfalls of the CCPs and there was no shortfall of prefunded

resources. Hence, the addition of the concentration and wrong-way risk impact in the considered

scenarios did not raise any systemically relevant concerns. However, it should be noted that this

impact was added to the P&L calculated from the baseline market stress scenarios. Hence, these

results cannot be used to draw conclusions on what the impact would be under all possible

extreme but plausible market scenarios. The CCPs should have dedicated risk management

measures to prudently mitigate these risks under all scenarios.

282. Under the “All CCP cover-2” scenario ESMA staff assumed the default of the members that

belong to the same two top clearing member groups across CCPs that would maximize the

overall impact on prefunded resources. The individual results of each CCP are by design equally

or less severe than the results calculated when assuming the default of the top-2 groups selected

for each CCP. The majority of CCPs would experience a default of at least one of their clearing

members. However, these consistent scenarios did not put significant stress to any CCP with

the % consumption of default fund-level prefunded resources being relatively low in all cases.

This indicates that while CCPs are highly interconnected through common clearing participants,

the exercise did not highlight any pairs of groups that are at the same time and under the common

tested scenario highly impactful at multiple CCPs.

283. Finally, in the reverse stress test analysis, ESMA staff intentionally went beyond what was

considered as plausible for the purpose of this exercise by stepwise increasing the number of

defaulting entities and the severity of the market shocks. The results have not indicated any

systemically relevant adverse impact following small changes in the underlying stress

assumptions. It is also confirmed that incremental changes in the severity of the market shocks

are generally more harmful than increases in the number of defaulting groups. For large

increases of the severity of the market shocks, going well beyond what was considered extreme

but plausible in the context of this exercise, the observed maximum shortfalls of prefunded

resources following the default of two clearing member groups would not be spread across CCPs

implying that there are different pairs of defaulting groups that would maximise the shortfalls at

different CCPs for these particular dates.

4.3 Concentration Stress Test Results

284. The objective of this analysis is to assess the adequacy of CCPs’ resources in covering the cost

of liquidating concentrated positions. To do so, the exercise computed the market impact of

concentration risk in different asset classes (according to the methodology detailed in Section

3.5) and compared it with the concentration add-ons reported by the CCPs.

285. The sum of the market impacts of all clearing members does not represent the actual

concentration risk faced by the CCPs, as CMs would not all default simultaneously, and because

the final impact may be lowered by offsetting positions between defaulting CMs. However, the

aggregated market impact approximates what would need to be charged by CCPs to cover the

concentration risk, generally through dedicated add-ons (or other duly computed resources

included in the initial margin).

286. The market impact computation heavily relies on the system-wide baseline sensitivity tables that

ESMA staff computed for each asset class starting from the sensitivity parameters that each

CCP submitted. To increase transparency, ESMA staff reported in the Appendix 6.2.3 a selection

of the most important system-wide sensitivity tables, together with the market impact on typical

concentrated positions.

4.3.1 Overview

287. The analysis starts with an overview at system-wide level of the concentration risk, in terms of

market impact (EUR), for each asset class. In addition, details are provided about exposure of

individual CCPs to concentration risk for each asset class.

288. The analysis continues by showing the concentration add-ons (EUR) provided for each asset

class at system-wide level.

289. The market impact is then compared to corresponding concentration add-ons. First, a

comparison is performed at CCP level. Subsequently, the comparison between concentration

risk and concentration add-ons is performed at CCP level but separately for each asset class.

290. Finally, the importance of accurately estimating the concentration risk at clearing member level

in order to prevent the consumption of mutualised resources is discussed.

4.3.2 System-wide impact

4.3.2.1 System-wide market impact per asset class

291. Figure 22 shows the aggregated system-wide market impact for each asset class, across all

CCPs of the exercise.

292. ESMA calculation shows that fixed income derivatives positions contain most concentration risk,

with a total over 29bn EUR. Bonds (including bonds from Repo clearing services) come next with

a total modelled concentration risk of around 11 bn EUR.

293. Concentration in commodity derivatives and in the equity segment (securities and derivatives)

is very significant as well, with around 7bn EUR of concentration risk calculated for each asset

class.

294. The concentration risk modelled for Emission Allowances stands also out at 2.5bn EUR.

295. The low market impact for credit derivatives is most likely driven by methodological limitations

of the framework

FIGURE 22: SYSTEM-WIDE MARKET IMPACT PER ASSET CLASS

296. For asset classes only cleared by two CCPs, concentration risk is balanced between them as in

Credit Derivatives (LCH SA & ICEEU) and Freight Derivatives (ECC & ICEEU).

297. For other classes, one CCP carries most of the risk and a second one most of the remainder.

This is the case for Fixed Income Derivatives (LCHUK 81%, ECAG 16%), Commodity Derivatives

(ICEEU 87%, ECC 7%), Equity (ECAG 75%, ICEEU 11%), Emission Allowances (ICEEU 90%,

ECC 10%).

298. Only concentrated positions in bonds are spread over many CCPs (LCHSA 64%, LCHUK 17%,

CCG 10%, ECAG 9%).

299. As illustrated, for most asset classes concentrated positions are clustered in only a few CCPs.

For both CCPs clearing Credit Derivatives, the market impact is much lower than the add-ons they charge.

FIGURE 23: BREAKDOWN OF CONCENTRATION RISK PER ASSET CLASS

4.3.2.2 System-wide concentration add-ons

300. CCPs generally charge concentration add-ons to cover concentration risk. Such add-ons are

reported in Figure 24, aggregated on a system-wide basis, by asset classes. The analysis shows

that, in absolute terms, such add-ons are largest for fixed income derivatives, bonds and equity

(securities and derivatives). Commodity derivatives and emission allowances show overall lower

concentration addons.

FIGURE 24: SYSTEM-WIDE REPORTED CONCENTRATION ADD-ONS, PER ASSET CLASS

4.3.3 Comparison of concentration add-ons and market impact

4.3.3.1 CCP level coverage

301. When aggregating concentration add-ons across all asset classes, some CCPs charge more

concentration add-ons than implied by the chosen baseline model (notably LCHUK and LCHSA).

On the other hand, four CCPs charge concentration add-ons which are lower than the modelled

market impact by more than 750m EUR (ICEEU 5.8bn, ECAG 1.8bn, CCG 1.1bn, ECC 800m).

FIGURE 25: CONCENTRATION RISK COVERAGE BY ADDONS FOR INDIVIDUAL CCPS

4.3.3.2 Asset class level coverage

302. The analysis shows that the calculated market impact materially exceeds the concentration add

ons for commodity derivatives and emission allowances. In some cases, this observation also

applies to Bonds, Equity Derivatives and Fixed Income Derivatives.

303. Within asset classes, the coverage of modelled market impact risk with concentration add-ons

differs across CCPs. By normalising the market impact and the add-ons by the total required

margin, both the importance of concentration and the different treatment by the CCPs can be

visualized. This allows to draw conclusions, even for CCPs that did not report dedicated

concentration add-ons.

304. For example, add-ons exceed the market impact in commodity derivatives only for some CCPs.

The market impact would also use a very large part of the required margin the CCPs currently

charge in some cases. Keeping in mind the limitations of the exercise, this could indicate an

insufficient coverage of concentration risk.

305. For the two main CCPs clearing commodities, ICEEU and ECC, the baseline model

concentration risk is 7 to 10 times greater than their concentration addons. The gaps

representing around 17% of required margin are 778m EUR for ECC and 5.6 bn EUR for ICEEU.

The overall concentration risk for KELER is much smaller at 66k EUR.

306. Further, for emission allowances, ICEEU charges 573m EUR for a baseline model risk of 2.16

bn EUR (a gap of 1.59 bn EUR of 20% of the required margin for those products).

FIGURE 26: COMPARISON OF MARKET IMPACT AND CONCENTRATION ADD-ONS, COMMODITY DERIVATIVES

307. For fixed income derivatives, the add-ons and modelled market impacts are more in line.

However, for ICEEU and KDPW modelled market impacts exceed add-ons by 442m EUR and

KDPW of 8m EUR respectively.

FIGURE 27: COMPARISON OF MARKET IMPACT AND CONCENTRATION ADD-ONS, FIXED INCOME

DERIVATIVES

308. For equity and equity derivatives, concentration risk and add-ons seem overall to be balanced.

However, it should be noted that ECAG, the CCP with the most concentrated positions, has a

gap of 1.8 bn EUR (or 6.5% of the required margin). KDPW and KELER have gaps of 5m and

760k respectively.

FIGURE 28: COMPARISON OF MARKET IMPACT AND CONCENTRATION ADD-ONS, EQUITY

309. For bonds, CCG and KDPW did not report concentration addons for the stress testing date,

leading to gaps of 1.06 Bn EUR and 840k EUR respectively.

310. KDPW implemented concentration addons for securities, but after the stress testing date.

Hence, this change is not reflected in the results.

311. For CCG, over half of the total concentration risk is caused by the interoperable CCP and public

or public owned entities. The results also do not assess a relevant model change proposed to

introduce concentration add-ons.

FIGURE 29: COMPARISON OF MARKET IMPACT AND CONCENTRATION ADD-ONS, BONDS

312. Although all CCPs have market impact risk according to the framework, four CCPs (KDPW,

CCPA, KELER, CCG) did not report any specific concentration add-ons.

313. Notwithstanding this limitation, ESMA looked at the overall margins and compared them with

the market impact.

4.3.3.3 Accuracy of the coverage at clearing member level

314. As illustrated in the previous section, overall add-ons collected at CCP level cover the computed

concentration risk. However, a large total amount of add-ons at default fund level does not

correctly protect the mutualised resources if there is a mismatch at clearing member level

between add-ons and concentration risk. Indeed, the market impact costs stemming from a

defaulting clearing member’s concentrated positions is only covered by the individual resources

of this clearing member. Therefore, if the concentration risk is not covered properly at CM level,

mutualised resources may still be consumed.

315. As previously shown, the level of add-ons charged at asset class / CCP level can differ widely

from the market impact, but this could be explained mostly by the choice of sensitivity

parameters.

316. It is also interesting to identify outliers where the market impact uses a large proportion of the

required margin, putting the mutualised resources at risk.

317. Across CCPs, the baseline concentration risk accounts for more than 25% of required margin

for 9.5% of clearing members, and more than 50% of required margin for 2.3% of them.

318. For those clearing members, before any prior market move, a large share of the required margin

would be at risk in case of default.

319. It is therefore important for margin models to be not only conservative overall but also accurate.

320. A model that would build in some conservativeness (i.e. by using a longer margin period of risk)

would be conservative most of time for most positions. However, in case a CM builds up some

very large positions, the model may not be sufficient to cover the real concentration risk. In such

a model, the adequate coverage of the market concentration risk at one point in time may not

always demonstrate the robustness of the model to varying portfolios. This is in particular the

case for securities, as the cleared portfolios can change a lot on a daily basis.

4.3.4 Additional risks

4.3.4.1 Vega risk

321. Products such as equity derivative options are sensitive to implied volatility. During the default

management of a clearing member, this implied volatility sensitivity risk (vega) needs to be

closed off or hedged before being auctioned off. For large optional positions, hedging vega or

liquidation incurs further costs.

322. As with outright directional positions, such costs could arise from the bid ask spread or for the

endogenous market impact.

323. Vega sensitivity was reported for equity derivatives and most commodity positions. ESMA staff

computes the sensitivity of the market impact P&L for each 1 volatility point move. This sensitivity

is then compared to the baseline (directional) market impact on all clearing member positions.

324. For equity instruments, a 1 volatility point impact when hedging the vega exposure represents

10% of the baseline model market impact from the delta exposure arising from both equity

securities and derivatives. For commodities, the vega market impact is much smaller than the

concentration risk from directional positions.

1% VOL MARKET IMPACT DELTA MARKET IMPACT

Commodity Derivatives 102,213,180 € 7,391,083,142 €

Equity (Securities & Derivatives) 677,053,411 € 6,865,975,310 €

325. The concentration risk stemming from implied volatility appears only significant for equity

derivatives. Although smaller on aggregate than the market impact stemming from directional

hedging, this is not the case for all portfolios even under only the 1% volatility assumption.

326. It should be noted that the further costs incurred from the vega liquidation are not part of the

chosen baseline.

4.3.4.2 Model risk

327. It is notoriously difficult to estimate the price impact as a function of the sold volume for

hypothetical sales under stressed market conditions. Moreover, the market impact parameters

are derived from the CCPs’ own estimates, with only few contributions for some asset classes.

328. The order of magnitude of the chosen estimates has been reported for transparency.

329. To assess model risk for securities, it was decided to model the price impact of fire sales based

on the exponential specification in line with Cont and Schanning (2017)

, with a market impact

on securities calculated according to the formula:

















  













330. where 



is the impact parameter, 



the corresponding boundary parameter and 



the total

amount sold.

331. Fukker et al. (2022)

extend the exponential specification with an approach that is similar to the

CoVaR methodology of Adrian and Brunnermeier (2011)

with the exception that the q

quantile

of the security-level price impact is estimated as a function of volumes sold and the system-level

return. This is the basis for our benchmarking exercise.

332. ISIN-level parameters are available for 1403 equity securities and 3244 bonds with calibration

at quantile levels q in {0.05;0.10;0.15;0.20;0.25}. Those securities are responsible for around

75% of baseline market impact.

333. In this exercise, the tail returns at q = 0.05 is selected, given that they reflect best stressed

market conditions. It is also assumed that one can apply the same parameters when closing

short positions.

Baseline model (bn

EUR)

Alternative model (bn EUR)

Difference

Bonds

5.18

8.7

67.82%

Equity (with

derivatives)

1.29

2.16

67.73%

TABLE 3: COMPARISON BETWEEN BASELINE AND ALTERNATIVE MODELS

334. The baseline model relies on the average daily volume as the only ISIN specific parameter. The

alternative model uses a full functional form with 3 ISIN specific parameters. It does not rely on

Cont, Rama and Eric Schaanning (2017). “Fire sales, indirect contagion and systemic stress testing”.

Fukker and al (2022). Contagion from market price impact: a price-at-risk perspective.

Adrian, Tobias and Markus K Brunnermeier (2011). CoVaR. Tech. rep. National Bureau of Economic Research.

extrapolation beyond a cut-off (2 average daily volumes) and provides a cost even for the

smallest positions.

335. The impact varies across CCPs, with a large CCP having a concentration risk 250% greater

with the alternative model.

336. Although the alternative model only covers a subset of cleared securities, it evidences the

importance of model risk for concentration risk.

4.3.5 Conclusions of Concentration Stress Test Results

337. The analysis shows that concentrated positions represent a significant risk for CCPs. Moreover,

the overall risk is clustered in one or two CCPs for most asset classes.

338. ESMA calculation shows that fixed income derivatives have the most concentration risk, with a

total over 29bn EUR. Bonds (including bonds from Repo clearing services) come next with a

total modelled concentration risk of around 11 bn EUR.

339. Concentration modelled for commodity derivatives and the equity segment (securities and

derivatives) is very significant as well, with around 7bn EUR of concentration risk calculated for

each asset class. There is a very large coverage gap between the system-wide estimated market

impact and margin add-ons, for commodity derivatives and to a lesser extent for equity products.

340. For the two main CCPs clearing commodities (ICEEU and ECC),the baseline model

concentration risk is 7 to 10 times greater than the concentration addons. The gaps representing

around 17% of required margin are 778m EUR for ECC and 5.6 bn EUR for ICEEU.

341. The modelled concentration risk for Emission Allowances stands also out at 2.5bn EUR and is

not adequately covered per the ESMA methodology.

342. The concentration risk is factored in explicitly in a majority of CCPs through dedicated margin

add-ons. Although all CCPs have market impact risk, four CCPs (KDPW, CCP.A, KELER, CCG)

did not report any concentration add-ons. Since the data request date, KDPW and CCG have

implemented or are in the process of introducing concentration addons. KELER relies on a

monitoring system to require additional collateral in case of elevated concentration.

343. Margin models need to be not only conservative but also accurate. This is especially the case

for liquid markets where large positions can build up very quickly.

4.4 Operational risk analysis

4.4.1 Results of the assessment of the general level of operational resilience of

individual CCPs

4.4.1.1 Descriptive analysis of operational risk events

Overview

344. A total of 14 CCPs (out of 15 in-scope CCP, cf. 3.6.1) reported a total of 330 operational risk

events that occurred during the reporting period, implying on average around 4 events per CCP

and per year. The number of events reported by CCPs ranged between 7 (1.2 per year) to 68

(12 events per year). In terms of impact, most operational risk events relate to the operational

risk dimensions ‘clearing/settlement unavailable’ and ‘critical supporting function unavailable’

(Chart 1). Over time, the number of operational events has been rising to reach 91 events in

2020 (Chart 2). It is challenging to know whether this increase relates to better reporting by CCPs

for more recent periods, or a genuine increase in the frequency of events. More than half of the

reported events lasted more than the 2-hour target recovery time used with respect to clients,

and 83 events lasted more than 10 hours (Chart 3), although the latter did not impact the core

functions of CCPs

. Around half of the ‘clearing/settlement unavailable’ events lasted more than

two hours and for the ‘critical supporting function unavailable’ or ‘service level agreement

breaches’, the share of events lasting more than two hours amounted to more than 60% of the

cases (Chart 4). This shows that events affecting the core services of CCPs had a shorter

duration. On average, events affecting clearing or settlement had a duration of 6.5 hours

compared with around 17 hours for events affecting critical supporting functions or resulting in

SLA breaches.

Operational risk chart 1

Operational risk events by type

Most events related to clearing/settlement

and critical supporting function unavailable

Operational risk chart 2

Number of events per year

The frequency of operational events has

been rising

Around 65% (53 cases) of these very long cases resulted either in a breach of ‘other service level agreement’, and only affected

a few clients or had an impact that affected less than 10% of the CCP’s activity.

Clearing /

settlement

unavailable

Critical

supporting

function

unavailable

Other Service

Level

Agreement

breach

>2h Less than 2h

Note: Number of operational events per type of impact

Sources: CCP, ESMA.

100

2016 2017 2018 2019 2020 2021

Less than 2h >2h

Note: Number of operational events per year.

Sources: CCP, ESMA.

Operational risk chart 3

Distribution of duration of operational risk events

More than half of the events lasted more

than the 2-hour target recovery time

Operational risk chart 4

Risk events longer than 2 hours

Most long events related to the same

impacts

TABLE 4: OPERATIONAL RISK EVENTS BY TYPE, NUMBER OF EVENTS PER YEAR, DURATION OF EVENTS

AND EVENTS LONGER THAN 2 HOURS

345. In the subsequent analysis ESMA staff focuses on events related to ‘clearing/settlement

unavailable’ and ‘critical supporting functions unavailable’ because they are more critical to the

functioning of CCPs and are less likely to be affected by reporting errors. Events with a minor

impact (less than 10% of the CCP’s activity affected) are also excluded.

0.1

100

1000

1 51 101 151

SLA and low impact events

2h recovery target

Other events

Note: Duration time of operational risk events, in hours.

Sources: CCP, ESMA.

93 low impact events lasting more than 2 hours

88 other events lasting more than 2 hours

47%

68%

60%

10%

20%

30%

40%

50%

60%

70%

80%

Clearing /

settlement

unavailable

Critical

supporting

function

unavailable

Other Service

Level

Agreement

breach

Less than 2h

Note: Share of long events (>2h) per type of impact

Sources: CCP, ESMA.

Events resulting in clearing or settlement unavailable

346. The sample covering events on the unavailability of clearing and settlement services (or core

functions) includes 113 events across 14 CCPs. Most operational risks events are concentrated

in a limited number of CCPs, with the number of events per CCP ranging from 1 to 26, with 6

CCPs reporting less than 5 events. Most events lasting more than two hours are related to

payments or trade transactions and are mainly caused by third party issues. In terms of duration,

Chart 5 shows a large dispersion among CCPs. Such operational events impact foremost at CCP

level (the whole CCP activity is affected) or at clearing service level (Here, by clearing service

we refer to the clearing activity linked to a segregated default fund, for CCPs that have different

clearing services with separated default funds). This impact implies a lower severity than the

whole CCP) rather than clients or products (Chart 6), which imply a lower level of severity. For

each datapoint, additional information on the risk event type is reported, detailing whether the

event is related to a third-party issue, a technology issue or issues around transaction processing

and execution. Overall, most of the events are related to issues with third parties especially for

events lasting more than 2 hours (87% of the ‘long event’ cases against 61% for the events

lasting less than 2h). Technology accounts for 19% of the events (Chart 7), mainly concentrated

in short events (30% of the cases) rather than long events (4% of the cases). Transaction

processing and execution-related events accounted for 8% of the events reported, equally across

short and long events. The impact of these events was mainly on payment and cash

management (50% of cases) and to a lesser extent on trade acceptance (24%).

Operational risk chart 5

Distribution of events

High dispersion across CCPs

Operational risk chart 6

Scope

Clearing services and CCP most impacted

Note: The box plot shows the interquartile range (Q1-Q3) in the rectangular

area, the median is indicated by a cross and the mean by an horizontal bar.

Outlier points as shown by dots. Distribution of events duration by CCP, in

hours. Only events leading to clearing/settlement unavailable. CCP with less

than 5 events not shown.

Sources: CCP, ESMA.

less than 2h >2h

Clearing NA

CCP Clearing services Clients Currency Product level

Note: Number of events by scope for clearing/settlement unavailable.

Sources: CCP, ESMA.

Operational risk chart 7

Risk event type

Most events related to third party issues

Operational risk chart 8

Type of impact

Mainly payment and cash management

TABLE 5: EVENTS RESULTING IN CLEARING OR SETTLEMENT UNAVAILABLE – DISTRIBUTION OF EVENTS,

SCOPE, EVENT TYPE AND IMPACT TYPE

Events resulting in critical functions unavailable

347. The sample covering events resulting in the unavailability of critical functions includes 80 events

across 12 CCPs, with the number of events per CCP ranging from 1 to 20 and 6 CCPs reporting

less than 5 events. In terms of duration, Chart 9 shows a large dispersion among CCPs. Such

operational events impact foremost the CCP and clearing services rather than clients or products

(Chart 10). Most of the events — as for clearing or settlement unavailable — are related to issues

with third parties (58% of the cases). Events related to technology account for 26% of the cases,

with similar patterns irrespective of the duration of the events. The impact of these events was

mainly on the risk management function of CCPs (45% overall, 56% for long events) and on

‘other’ critical supporting functions (28%). Events related to ‘other’ include delay issues with

pricing data and connectivity issues.

less than 2h >2h

Clearing NA

Third party

Technology

Transaction processing and execution

Note: Number of events by risk event type for clearing/settlement unavailable.

Sources: CCP, ESMA.

less than 2h >2h

Clearing NA

Payment and cash management unavailable

Trade acceptance unavailable

Other

Non-cash asset management unavailable

Risk management function unavailable

Note: Number of events by impact type for clearing/settlement unavailable.

Sources: CCP, ESMA.

Operational risk chart 9

Distribution of events

High dispersion across CCPs

Operational risk chart 10

Scope

Clearing services and CCP most impacted

Operational risk chart 11

Risk event type

Most events related to third party issues

Operational risk chart 12

Type of impact

Mainly risk management

TABLE 6: EVENTS RESULTING IN CRITICAL FUNCTIONS UNAVAILABLE – DISTRIBUTION OF EVENTS, SCOPE,

EVENT TYPE AND IMPACT TYPE

Note: The box plot shows the interquartile range (Q1-Q3) in the rectangular

area, the median is indicated by a cross and the mean by an horizontal bar.

Outlier points as shown by dots. Distribution of events duration by CCP, in

hours. Only events leading to critical functions unavailable. CCP with less than

5 events not shown.

Sources: CCP, ESMA.

less than 2h >2h

Other critical

CCP Clearing services Clients Product level

Note: Number of events by scope crticial supporting function unavailable.

Sources: CCP, ESMA.

less than 2h >2h

Other critical

Third party

Technology

Transaction processing and execution

Data management

Note: Number of events by risk event type critical supporting functions

unavailable.

Sources: CCP, ESMA.

less than 2h >2h

Other critical

Risk management function unavailable

Other

Payment and cash management unavailable

Trade acceptance unavailable

Non-cash asset management unavailable

Note: Number of events by impact type for critical supporting function

unavailable.

Sources: CCP, ESMA.

4.4.1.2 Operational reliability metrics: Results

348. Using the reliability metrics described in the methodology ESMA staff computed results for both

the  “Clearing / settlement unavailable” and the  “Critical

supporting function unavailable” (see section 3.6.2.5).

349. The results are evaluated in conjunction with the percentile metrics in section 4.4.1.4.

Operational risk chart 13

Reliability metrics

Type of impact: Clearing / settlement unavailable

FIGURE 30: CLEARING / SETTLEMENT UNAVAILABLE: RELIABILITY METRICS

350. Overall, we observe high levels or reliability across CCPs; the list of CCPs is ordered using the

expected aggregated amount of downtime (or unavailability) per year, the top half of entities

exhibit very low expected downtime values, meanwhile the bottom three entities exhibit figures

that signal a higher level or risk.

351. While expected downtime is an important figure, for FMIs it is particularly important to minimize

the Mean Time to Repair, in order to achieve availability of the clearing services with a Recovery

Time Objective of two hours. In this sense we pay particular attention to CCPs with high values

of MTTR that don’t have Mean Time Between Failures values of a very high magnitude (which

would imply that events are very rare, and our conclusions may not be significant), Using these

criteria we observe high MTTR values for CCP10, although for CCP10 the high value of MTBF

indicates that the figures may be driven a very small number of events.

Critical functions

CCP MTBF_days MTTR_hours Expected_1y_downtime_hours Average activity affected

CCP14 416.7 1.0 0.6 38%

CCP06 312.5 0.8 0.7 100%

CCP02 416.7 1.3 0.8 75%

CCP08 1250.0 4.0 0.8 11%

CCP13 250.0 1.2 1.2 67%

CCP01 250.0 1.2 1.2 60%

CCP12 416.7 2.4 1.4 99%

CCP11 178.6 3.3 4.6 100%

CCP09 89.3 1.9 5.3 99%

CCP03 138.9 4.0 7.2 61%

CCP10 625.0 21.0 8.4 100%

CCP04 83.3 7.4 22.2 95%

CCP07 48.1 4.5 23.4 74%

CCP05 69.4 8.4 30.1 56%

Operational risk chart 14

Reliability metrics

Type of impact: Critical supporting function unavailable

FIGURE 31: CRITICAL SUPPORTING FUNCTIONS UNAVAILABLE: RELIABILITY METRICS

352. For critical supporting functions we observe slightly higher levels or resilience, with some CCPs

reporting not incidents.

353. Only the bottom four entities exhibit metrics that suggest that further supervisory analysis should

be performed, with CCP07 being a different case than the other three, as rather than exhibiting

incidents with long duration, the low MTBF and MTTR indicate recurrent frequent problems with

short remediation times.

4.4.1.3 Estimation of percentile metrics: methodology and results

354. The data on operational risk events can be used to assess the likelihood and impact of

disruptions using the loss distribution approach commonly used to estimate operational risks for

banks. The underlying idea is that observed events are ‘draws’ from a specific frequency

distribution (which describes the average number of events over a given time horizon) and from

a specific severity distribution (which describes the duration of the disruption time). By calibrating

the parameters of the frequency and severity distributions and running a large number of

numerical simulations, one is then able to estimate the distribution of operational events

355. The objective of these metrics is to produce measurements that reflect stressed conditions and

complement the operational metrics based on mean measurements.

356. The model methodology, its assumptions, calibration and model risk analysis are provided in

Annex 6.3.

See Shevchenko (2010) for details on the loss distribution approach. Shevchenko, P. (2010), “Calculation of aggregate loss

distributions”, Journal of Operational Risk Vol.5 (2).

Critical supporting functions

CCP MTBF_days MTTR_hours Expected_1y_downtime_hours Average activity affected

CCP11 1250.0 1.5 0.3 25%

CCP13 625.0 1.3 0.5 100%

CCP02 416.7 1.0 0.6 53%

CCP14 312.5 1.5 1.2 85%

CCP03 625.0 4.0 1.6 78%

CCP12 250.0 1.9 1.9 100%

CCP09 156.3 1.2 2.0 100%

CCP01 1250.0 10.0 2.0 68%

CCP07 89.3 3.1 8.6 41%

CCP08 250.0 15.6 15.6 51%

CCP05 250.0 17.4 17.4 87%

CCP10 62.5 10.0 39.9 75%

Results for clearing or settlement unavailable

357. For each CCP, ESMA staff obtained results which are driven by two factors: the average

frequency of operational events and the severity distribution which depends on the bucket the

CCP belongs to

. The charts below show that for CCPs in the lowest severity group, the total

disruption time in a given year would be below 2 hours on average and would remain below 10

hours in the most extreme cases (VaR and expected shortfall at 95% level). For CCPs in the

second group (low severity), total disruption time would be below 6 hours but in the most extreme

cases could range up to 21 hours. CCPs in group 3 (high severity) exhibit some variation because

some of the CCPs had more frequent disruptions than others, resulting in a higher frequency of

events and relatedly more total disruption time on average. Overall, average disruption time

would be below 10 hours for two CCPs but above that mark for one other CCP with more frequent

events. In extreme cases, total disruption time could be higher than 50 hours for at least one

CCP. Finally, two CCPs in the highest severity bucket have median disruption time above 25

hours per year, while the other CCP would have a median disruption time around 9 hours, and

all the CCPs in this bucket would have total disruption times above 50 hours in extreme cases.

Operational risk chart 15

Risk indicators for group 1 (lowest severity)

Total median disruption time below 2h

Operational risk chart 16

Risk indicators for group 2 (low severity)

Total median disruption time below 6h

In a few cases, some CCPs had a low number of events (low frequency) but these events lasted for long. Therefore, for some

CCPs in group 4 the total median disruption time in one year might be lower than CCPs in group 3 (which had a higher frequency

of events.

CCP06 CCP14 CCP13 CCP01 CCP02 Average

Average Median VaR 95% ES 95%

Note: Sum of disruption time over one-year for clearing or settlement

unavailable, in hours

Sources: CCPs, ESMA

CCP09 CCP12 CCP11 Average

Average Median VaR 95% ES 95%

Note: Sum of disruption time over one-year for clearing or settlement

unavailable, in hours

Sources: CCPs, ESMA

Operational risk chart 17

Risk indicators for group 3 (high severity)

Total median disruption time up to 20h

Operational risk chart 18

Risk indicators for group 4 (highest severity)

Total median disruption time up to 24h

FIGURE 32: RISK INDICATORS BY SEVERITY GROUPS- CLEARING OR SETTLEMENT UNAVAILABLE

Results for critical supporting functions unavailable

358. For each CCP, ESMA staff obtained results which are driven by two factors: the average

frequency of operational events and the severity distribution which depends on the bucket the

CCP belongs to

Qualitatively, the results are similar to those described in the previous section.

First, CCPs in the lowest severity group have an average total disruption time below or close to

2 hours and even in the extreme cases disruption time remains below 10 hours. Second, CCPs

in the low severity group have slightly longer disruption time but are quite close to the first group.

Third, one of the CCPs in the high severity group would have disruption time higher than 10

hours on average and more than 30 hours in extreme cases. Finally, CCPs in the highest severity

group could experience long disruption time on average (more than 12 hours) and more than 80

hours for all those CCPs in extreme circumstances.

Operational risk chart 19

Risk indicators for group 1 (lowest severity)

Total median disruption time below 2h

Operational risk chart 20

Risk indicators for group 2 (low severity)

Total median disruption time below 2h

In a few cases, some CCPs had a low number of events (low frequency) but these events lasted for long. Therefore, for some

CCPs in group 4 the total median disruption time in one year might be lower than CCPs in group 3 (which had a higher frequency

of events.

CCP08 CCP03 CCP07 Average

Average Median VaR 95% ES 95%

Note: Sum of disruption time over one-year for clearing or settlement

unavailable, in hours

Sources: CCPs, ESMA

100

120

140

160

180

200

CCP04 CCP05 CCP10 Average

Average Median VaR 95% ES 95%

Note: Sum of disruption time over one-year for clearing or settlement

unavailable, in hours

Sources: CCPs, ESMA

Operational risk chart 21

Risk indicators for group 3 (high severity)

Total median disruption time up to 8h

Operational risk chart 22

Risk indicators for group 4 (highest severity)

Disruption time higher than 80h in extreme

cases

FIGURE 33: RISK INDICATORS BY SEVERITY GROUPS- CRITICAL SUPPORTING FUNCTIONS UNAVAILABLE

4.4.1.4 Identification of entities with higher measured risk

359. In order to perform a final evaluation, ESMA staff compared the combined results from two of

the risk indicators developed: the expected 1yr downtime and the estimated 95th percentile 1yr

downtime, in order to identify CCPs that are performing worse than average in both for the

sample of CCPs.

CCP02 CCP09 CCP13 Average

Average Median VaR 95% ES 95%

Note: Sum of disruption time over one-year for critical supporting function

unavailable, in hours

Sources: CCPs, ESMA

CCP14 CCP11 CCP12 Average

Average Median VaR 95% ES 95%

Note: Sum of disruption time over one-year for critical supporting function

unavailable, in hours

Sources: CCPs, ESMA

CCP07 CCP03 Average

Average Median VaR 95% ES 95%

Note: Sum of disruption time over one-year for critical supporting function

unavailable, in hours

Sources: CCPs, ESMA

CCP01 CCP08 CCP05 Average

Average Median VaR 95% ES 95%

Note: Sum of disruption time over one-year for critical supporting function

unavailable, in hours

Sources: CCPs, ESMA

Operational risk chart 23

Scatterplot: Expected 1y downtime and estimated 95

percentile downtime

Type of impact: Clearing or settlement unavailable

FIGURE 34: EXPECTED 1Y DOWNTIME AND ESTIMATED 95

PERCENTILE DOWNTIME - CLEARING OR

SETTLEMENT UNAVAILABLE

360. Using Chart 23, for the category of critical clearing functions ESMA staff identifies three CCPs

where risk indicators signal higher risk: CCP04, CCP07 and CCP05. This implies that in terms

of availability, operational risk for clearing or settlement (‘critical clearing functions) might be

higher for those CCPs. Therefore, further scrutiny of prevention and recovery tools for those

CCPs is key.

CCP07

CCP08

CCP10

CCP09

CCP11

CCP12

CCP13

CCP14

CCP1

CCP2

CCP3

CCP4

CCP5

CCP6

Mean

0.0 20.0 40.0 60.0 80.0 100.0 120.0

Expected 1y downtime

Estimated 95% percentile downtime

Critical clearing functions

Operational risk chart 24

Scatterplot: Expected 1y downtime and estimated 95

percentile downtime

Type of impact: Critical supporting function unavailable

FIGURE 35: EXPECTED 1Y DOWNTIME AND ESTIMATED 95

PERCENTILE DOWNTIME – CRITICAL

SUPPORTING FUNCTIONS UNAVAILABLE

361. Using Chart 24, for the category of critical supporting functions ESMA staff identified three CCPs

where risk indicators signal higher risk than their peers: CCP08, CCP05 and CCP10.

362. It must be noted that due to the construction of the category “Critical supporting function”, there

is a higher degree of heterogeneity in the events included and the impact for the CCPs is of lower

importance than the events of the category with impact “Clearing or settlement non-available”,

as they do not ultimately affect the ability of customers to access clearing (e.g: incidents affecting

accuracy of internal risk management systems are relevant from a supervisory monitoring

perspective but they would typically have limited operational impact to customers). The

heterogeneity of this category suggests that further detailed analysis of individual events should

be performed in order to reach definitive conclusions about the underlying risk signaled by the

indicators.

4.4.1.5 Conclusions

363. The analysis confirms that operational risk is a substantial risk for CCPs that may impact their

resilience. During the reporting period, CCPs experienced a range of operational risk events

affecting their clearing and settlement activity or some of their critical supporting functions. The

reported data suggests that the number of events is increasing, although caution is needed in

interpreting the data. The current level of data quality and early development of methodologies

suggest caution in drawing preliminary conclusions and their use should be focused on

CCP07

CCP08

CCP10

CCP09

CCP11

CCP12

CCP13

CCP14

CCP1

CCP2

CCP3

CCP5

Mean

0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0

Expected 1y downtime

Estimated 95% percentile 1y downtime

Critical supporting functions

identifying areas of risk to direct supervisory efforts. A key finding is that most of the operational

events stem from issues in the provision of third-party services. This is subject to further analysis

in the following sections.

364. Overall, CCPs exhibit high levels of reliability, as it is expected for this type of entity. However,

there is variation across CCPs both with regards to the general level of operational risk measured

and to the presence of incidents with long remediation timeframes or recurrence.

365. Using the information about internal incidents of CCPs systems and third-party providers ESMA

staff developed two methodologies to measure operational risk from historical events for

individual CCPs. With the computed results ESMA staff identified specific CCPs for which, based

on the expected 1yr downtime and the estimated 95th percentile 1yr downtime, further specific

work is warranted to understand the drivers of these differences, the root causes of the events

and the remediation actions taken.

366. Looking forward, further work on a consistent reporting of operational risk events might be

warranted. While CCPs already disclose publicly information on availability through the CPMI-

IOSCO disclosures, a more consistent framework for collecting and reporting operational risks

events to NCAs and ESMA would improve the quality of the analysis and support the monitoring

of risks for CCPs.

4.4.2 Results of the assessment of risk exposures of individual CCPs to critical third-

party service providers

4.4.2.1 Exposures of individual CCPs to critical third-party service providers without taking

into account risk management tools

367. In line with the methodology, CCPs reported the critical third-party service providers on which

they rely in order to provide their clearing services. The total number of third-party entities and

the type of entity is shown in Chart 25.

368. From the data reported, significant variations are observed, with the minimum being around ten

providers and the maximum nearly sixty providers. All CCPs use FMI services, which are mainly

CSDs, trading Venues, settlement systems and payment systems; Also, all CCPs use non-

financial services, such as technology providers, telecommunications and utility providers or

specialized data providers; All but one CCPs use intragroup services, which directly relates to

governance and the structure or the organisation of the CCP. Most CCPs also rely on different

financial services for payments, custody, data, default management, settlement, investment and

liquidity needs.

Operational risk chart 25

Number of critical third-party service providers per CCP by entity type

FIGURE 36: NUMBER OF CRITICAL THIRD-PARTY SERVICE PROVIDERS PER CCP BY ENTITY TYPE

369. The variation in these numbers across CCPs may be driven by various factors:

• Size and complexity of the CCP: Higher operational complexity may require the reliance on

more critical third-party service providers.

• Exposure to each individual provider: When counting the number of providers, a value of

one is assigned to each of them, however some entities may have a higher number of

providers with each serving only segments of the activity, while others may rely on a more

reduced number of providers that serve the whole CCP. This will be adjusted when using

Weighted Exposure indicators later in this section in order to determine the relative

exposure towards each third-party service provider.

• Risk management strategy: Building redundancy increases the number of third-party

entities with relationship to a CCP, but its risk reducing (e.g.: A CCP may rely on one third-

party service providers for a specific function, while another CCP may decide to have two

providers for the same function, so that in case one is not available, they have another

provider operationally set-up). This will be taken into account through the hypothetical

scenario and modelling of risk management tools.

370. In the below figure, one can observe the weighted exposure per CCP and the difference with

respect to just counting the number of linked third-party service providers. The number of third-

party service providers is adjusted by weighting each service with the percentage of CCP

clearing activity that they support (in line with the methodology described in section 3.6.3.7). The

weighted metric adjusts for the exposure to each individual critical third-party provider and allows

for a better comparability across CCPs that may have different operational structures.

Furthermore, the reduction in exposure for the CCPs with a higher number of linked third-party

service providers is higher than for those with lower number of linked third-party service

providers. This is aligned with expectations, as entities with higher levels of operational

complexity will tend to have more third-party service providers that serve only specific segments

of the CCP.

Operational risk chart 26

Adjusting the number of critical third-party service providers weighting each of them with

the percentage of CCP clearing activity they service

FIGURE 37: COMPARISON BETWEEN WEIGHTED AND NOT-WEIGHTED NUMBER OF CRITICAL THIRD-PARTY

SERVICE PROVIDERS PER CCP

4.4.2.2 Reducing CCP exposures to third-party service providers through risk mitigation tools

for a hypothetical scenario of an outage affecting a critical third-party provider

371. For our analysis of exposures, we follow a similar approach with respect to “Type of Impact” as

in section 4.4.1., we analyse separately exposures with a severity in case of failure such that

critical clearing or settlement functions would be affected from those in which a critical supporting

function would be affected.

372. First, we look at the overall level of exposure of all CCPs before and after considering

operational risk management tools in order to draw general conclusions and then we provide the

exposures after considering risk management tools per CCP.

Overall risk reduction for exposures to entities that would have an impact to critical clearing or

settlement functions

Operational risk chart 27

Aggregated exposures of CCPs towards third-party service providers before and after

taking into account operational risk management tools. Exposure with impact to critical

clearing or settlement functions.

FIGURE 38: RISK REDUCTION FOR CCPS’ CLEARING AND SETTLEMENT FUNCTIONS EXPOSURE TO THIRD-

PARTY SERVICE PROVIDERS USING OPERATIONAL RISK MANAGEMENT TOOLS

100

373. This section adjusts the operational risk exposures of CCPs towards third-party service

providers by taking into account the operational risk management tools that the CCP applies to

manage and mitigate its operational risks toward the critical third-party service providers. When

calculating the third-party exposure ESMA staff starts with the total number of providers, adjust

individual entity exposures to the CCP’s activity they serve (in order to enable comparability and

adjust for relative importance) and transform the exposures to residual exposures with respect

to the application of operational risk management tools, using the assumption that the mitigation

tools work. The exposures after the application of tools are determined by using the residual

exposures after tools reported by CCPs (in line with methodology section 3.6.3.6)

374. When analysing the reduction of exposures of CCPs towards critical third-party service

providers through the use of operational risk management tools, one first observes that for the

FMI category there is very limited change between exposures before and after tools; this may

be due to the low substitutability of the services provided by these types of entities (so even if

CCPs wanted to build mitigation risk management tools it may not be possible) or the expectation

that FMIs will behave in a resilient manner.

375. The change between exposures before and after tools for intragroup entities shows a pattern

that is similar to FMIs. In this case the most probable explanation is that intragroup entities are

usually an extension of internal systems that are shared across different entities belonging to the

same corporate group, hence the logical strategy to increase resilience would be through

improvements at an internal level rather than through third-party risk mitigation tools. In any case,

different strategies with respect to corporate structure and intragroup services do not imply

different levels of risk per se.

376. For exposures to non-financial and other financial entities, one observes a substantial reduction

with respect to exposures before considering tools. This can be explained by the availability of

alternatives and substitutes in the market that the CCP can engage with, coupled with a

motivation by CCPs to build resilient operations and minimize the number of single points of

failure that depend on third-party entities out of their direct control.

Overall risk reduction for exposures to entities that would have an impact to critical supporting

functions

101

Operational risk chart 28

Aggregated exposures of CCPs towards third-party service providers before and after

taking into account operational risk management tools. Exposure with impact to critical

supporting functions.

FIGURE 39: RISK REDUCTION FOR CRITICAL SUPPORTING FUNCTIONS USING OPERATIONAL RISK

MANAGEMENT TOOLS

377. For critical third-party service providers that would impact critical supporting functions, one

observes that the bulk of the operational risk exposure of CCPs is towards non-financial entities,

with this category being composed mainly by exposures to Software, IT & Telecom services,

around 70% of exposure before the application of risk mitigation tools and around 80% after the

application of risk mitigating tools. One also observes that the level of risk reduction is high for

all categories, probably reflecting a higher substitutability of these services and ability of CCPs

to build risk mitigation tools.

Operational risk exposure per CCP after application of risk mitigating tools towards third-party

service providers that would have an impact to critical clearing or settlement functions

102

Operational risk chart 29

CCP’s weighted operational risk exposure after application of risk mitigating tools.

Exposure with impact to critical clearing or settlement functions.

FIGURE 40: WEIGHTED EXPOSURE PER CCP AFTER OPERATIONAL RISK MANAGEMENT TOOLS – CRITICAL

THIRD-PARTY SERVICE PROVIDERS

378. The exposure of CCPs toward service providers after the application of operational risk

mitigation tools exhibits a significant variance across CCPs; however, one observes the common

trend towards a dominant presence of exposures to entities in the FMI group with also a

significant presence of intragroup exposures.

379. While it can be difficult to reduce exposure to entities in the FMI and intragroup categories,

exposures to non-financial and other financial entities represent opportunities where operational

103

resilience with respect to a scenario of critical third-party provider failure can be increased if

deemed desirable.

Operational risk exposure per CCP after application of risk mitigating tools towards third-party

service providers that would have an impact to critical supporting functions

Operational risk chart 30

CCPs’ weighted operational risk exposure after application of risk mitigation tools

FIGURE 41: WEIGHTED EXPOSURE PER CCP AFTER OPERATIONAL RISK MANAGEMENT TOOLS – CRITICAL

THIRD-PARTY SERVICE PROVIDERS

104

380. The operational risk exposures of CCPs after the application of risk mitigation tools for these

critical third-party service providers exhibits a significant variance across entities, with some

CCPs able to completely eliminate their exposure to single points of failure in this category.

381. For most entities that have residual exposure after tools, the dominant exposure is with respect

to non-financial entities.

4.4.2.3 Evidence of behaviour of operational risk management tools using empirical data from

past events.

382. In the analysis above, it is assumed that CCPs’ risk management tools work perfectly. In order

to verify this assumption, ESMA staff checked the past incidents data collected and linked

operational events originating from third-party service providers to the relevant CCP risk

management tools, which comprise of the use of alternative service providers and internal tools

(tools are described in section 3.6.3.5).

383. For this analysis, we linked each event to each critical third-party provider and each with whether

there was any mitigation risk management tools and the type of tool. In order to observe any

meaningful trends, we look at the ratio incidents with respect to services (similar to the number

of critical third-party providers but takes into account that individual providers can provide more

than one service and have associated more than one mitigation tool).

384. Using the Figures below one observes that exposures protected by a tool of type “Alternative

provider” have a very low level of experienced incidents, while unsurprisingly the highest level of

incidents is in exposures for which no tool is present.

385. Results for exposures protected by an “Internal tool” show an intermediate level of risk; ESMA

staff followed up with a questionnaire to understand the reasons behind the events affecting

these services in order to understand better their nature.

386. From the compiled results (fourth chart below), one observes that in 75% of the incidents, the

Internal tool has an activation time and CCPs didn’t make use of the tool. This information points

to the fact that these types of tools would not protect against incidents with short durations and

would mostly be effective to prevent high severity events.

105

Operational risk chart 31

Number of services by type of risk

management tool protecting them

Operational risk chart 32

Number of operational events linked to

critical third-party service providers by type

of risk management tool associated with

the service provider

Operational risk chart 33

Ratio [number of incidents / number of

services] by type of risk management tool

Operational risk chart 34

Reasons behind incidents for services

protected with an internal tool

FIGURE 42: BEHAVIOUR OF OPERATIONAL RISK MANAGEMENT TOOLS

74%

17%

Switch time Simultaneous failure Other

106

4.4.2.4 Conclusions

387. The analysis of the CCPs’ operational risk exposures towards third-party service providers

provides insights about the different operational dependencies to critical third-party service

providers and how CCPs use operational risk management tools to mitigate risk.

388. We develop a methodology and risk indicators that enable us to monitor the single points of

failure with respect to a hypothetical scenario involving an outage of a critical third-party provider.

The results exhibit differences across CCPs in their relative level of third-party risk. Some of

those differences may be explained by the variation in operational complexity that cannot be risk-

reduced using mitigation tools (such as the exposure to FMIs due to the business model), but in

other cases, it may indicate there is room for increases in operational resilience. With respect to

the use of the indicators developed, it must be noted that while a higher number of third-party

exposures is indicative of higher risk (under the assumption of similar level of risk for individual

entities), the analysis performed has not estimated risk of individual entities and how that would

influence the assessment.

389. Overall, all CCPs significantly reduce their exposure to the groups of non-financial and other

financial entities, while the reduction of risk through mitigation tools for critical clearing or

settlement functions with respect to exposures to FMIs or Intragroup is very limited, which would

be consistent with the low substitutability of FMI services and the similarity of Intragroup services

to internally managed operations (for which resilience would be managed through improvements

at an internal level rather than through third-party risk mitigation tools).

390. Certain CCPs exhibit significant levels of exposure after tools to non-financial or other financial

entities. For these entities and exposures further work should be conducted to evaluate the

individual circumstances of these exposures and the suitability of taking corrective action.

391. Finally, using incidents data, ESMA staff evaluates the behaviour of the reported operational

risk management tools. From the evidence collected, one notes that exposures protected

through redundancy exhibit significantly low levels of risk, which is consistent with expectations.

The empirical results for exposures protected through tools categorized as “internal tools” are

mixed, the follow-up work performed by ESMA staff indicates that many of these tools are

probably only suitable to protect CCPs against events of long duration. Given the results, it is

recommended that supervisors emphasize the verification of testing results for these types of

tools, in order to increase the likelihood that they work as intended in case of an event of long

duration materializing.

4.4.3 Results of the assessment of concentration or systemic risks in the network of

critical third-party service providers

4.4.3.1 Overview of the network of third-party service providers

Operational risk chart 35

Network graph

Network of all third-party service providers connecting to CCPs

107

FIGURE 43: NETWORK OF CCPS CONNECTED THROUGH THIRD-PARTY PROVIDERS

392. From the data submissions to ESMA, a network displaying the ecosystem of the third-party

providers and the CCPs was constructed (Figure 43). The network is unweighted, meaning that

the links between the nodes do not have any associated weights (as opposed to a weighted

network). Further, the network is undirected in that the order of the nodes does not matter, only

the links between CCPs and third-party providers. The network’s layout simulates the forces of

attraction between the connected nodes (CCPs and the entities) to show the individual clusters

of the network. To further improve visualization, it decreases the crossings of links and evens

the node distribution in the layout.

393. The network of critical third-party providers and 14 CCPs contains 295 unique third-party

providers, with 19% of them providing services to more than one CCP and the remaining 81%

connected to a single CCP. Despite the average number of links per third party entity being 2.5

(the average degree of the whole network), the distribution of links is uneven – it is the CCPs

and only a few third-party providers bearing the high number of interconnections.

394. In the constructed network, one can visually analyse clusters - a group of nodes that are more

connected to each other than to the other nodes. Those can be identified at three levels. First,

at the micro-level, third-party providers connect mostly to the CCPs, which stems from the

definitional features of the dataset. Then, at the meso-level, 10 CCPs lie in the outskirts (CCP1,

CCP2, CCP5, CCP6, CCP7, CCP8, CCP9, CCP10, CCP11, CCP12), on average sharing 6.8

108

third party providers with its closest CCPs. Finally, 4 CCPs (CCP3, CCP4, CCP13, CCP14) form

a cluster, on average sharing 15.6 third party providers among its closest CCPs.

395. When considering only the links between entities and the CCPs, there are, on average, 27 third-

party service providers per CCP, with a maximum of 56 and a minimum of 10 third-party service

providers per CCP. If the type of entity in the whole network is considered, 27.4% belong to an

FMI group (purple), 18.6% belong to the group of other financial entities (blue nodes), and 53.8%

are non-financial entities (black). The split of the risk levels of links is 45% for risk level 0 (grey

edges), 40% for risk level 3 (red edges), 10% for risk level 2 (yellow edges), and 5% for risk level

1 (green edges). When focusing on entities connected to more than one CCP, one observes that

out of the 56 entities, 50% belong to the FMI group, 23.2% belong to the group of other financial

entities, and 26.8% are non-financial entities.

396. The focus of the following section is the analysis of the network of third-party service providers

to understand aspects of concentration risk and systemic risk in relation to critical third-party

service providers. As such, critical third-party service providers connected to single CCPs are

filtered out for the rest of the analysis and metrics, leaving third-party service providers that have

two or more connections.

Operational risk chart 36

Network graph

Network of third-party service providers connected to at least two CCPs

109

Operational risk chart 37

Most interconnected third-party entities

Top 10 entities with higher level of interconnectedness

FIGURE 44: NETWORK OF THIRD-PARTY PROVIDERS CONNECTED TO AT LEAST TWO CCPS

397. When looking at the top-10 most interconnected third-party entities one sees that most of them

belong to a single FMI group. The bar on the left indicates for each of the top-10 most

interconnected third-party entities the percentage of CCPs that have an operational dependency

on it. The colors highlight the level of risk of the operational relationship, using the information

collected through the hypothetical scenario of section 4.4.2. For example, for FMI 01 the level of

interconnectedness reaches 100%, implying that FMI01 is used by all CCPs in this sample. A

failure of FMI01 would create substantial issues for around 75% of the CCPs (risk level 3 in red)

while for the remaining 25%, the CCPs have systems in place that could be used as a backstop

(minimizing the risk to zero).

398. The high level of interconnectedness and risk in the FMI group is consistent with their role in the

financial markets. Apart from entities in the FMI group, four other types of entities show a high

level of interconnectedness (a technology provider ‘Technology provider-01’, a financial entity

‘Financial-01’, a data provider ‘Data provider-01’ and an entity whose main role is providing

intragroup services ‘ Intragroup-01’); however, when taking into account the level of risk of the

interconnections, only one of them could potentially impact simultaneously the critical functions

of more than one CCP.

399. In the graph below one can observe details about the distribution of interconnectedness by type

of entity for the subsample of entities connected to at least 2 CCPs (14% of CCPs in the sample

of CCPs). The mean interconnectedness is 22% for FMIs, 18% for non-financials and 19% for

other financial entities. There are two outliers (100%, 50%) that represent three entities (FMIs-

01, 02, 03). Apart from these outliers, the max for all groups is 29% (connected to 4 CCPs).

0% 20% 40% 60% 80% 100%

FMI-01

FMI-02

FMI-03

Technology provider-01

FMI-04

Financial-01

Data provider-01

FMI-05

FMI-06

Intragroup-01

Risk level 0 Risk level 1

Risk level 2 Risk level 3

110

Operational risk chart 38

Box & Whisker plot: Number of CCPs connected by type of entity

FIGURE 45: BOX & WHISKER PLOT, NUMBER OF CCPS CONNECTED BY TYPE OF ENTITY

4.4.3.2 Detailed analysis for specific types of services and hypothetical groups

400. In order to assess concentration issues for specific types of services, ESMA staff performed the

analysis on a more granular level considering a segmentation into four groups of aggregated

services covering a range of specific sub-services.

401. It must be noted that entities that provide multiple types of services (such as intragroup entities)

may appear in multiple categories showing the sub-set of services and interconnections that

belong to the category. The services are grouped in the following manner:

Financial services

• Clearing & risk services

• Collateral

• Custody

• Default Management Process (DMP)

• Financial messaging service providers

111

• Interoperability link

• Investment

• Liquidity provider

• Payments

• Settlement

• Trade provider/source

Software, IT &

Telecommunications

services

• IT providers

• Cyber security

• Telecommunications (abbreviated as Telco)

• Software

• Cloud services

• Support

Data providers

• Any type of entity that provides data used for pricing

or valuation purposes by the CCP

Other services

• Disaster recovery capacities

• Electricity provider

• Physical infrastructure

• Regulatory reporting

• Utility operators in commodity derivatives

• Other

Financial services

112

Operational risk chart 39

Network chart

Financial services – LEI level

Operational risk chart 40

Most interconnected third-party entities

Financial services – LEI level

Operational risk chart 41

Network chart

Financial services – Hypothetical groups`

Operational risk chart 42

Most interconnected third-party entities

Financial services – Hypothetical groups

FIGURE 46: INTERCONNECTEDNESS ANALYSIS – FINANCIAL SERVICES

402. When looking at financial services one observes that some Financial Market Infrastructures

have high degrees of interconnectedness and their failure would impact the critical functions of

multiple CCPs simultaneously, this is consistent with their role in the financial markets and their

low level of substitutability.

403. The three most interconnected FMIs have a marked systemic nature, as 100%, 50% and 43%

of CCPs are connected to these infrastructures for some of their critical functions. Some financial

0% 20% 40% 60% 80% 100%

FMI-01

FMI-02

FMI-03

FMI-04

FMI-05

Financial-01

FMI-06

Financial-02

FMI-07

Financial-03

Risk level 0 Risk level 1

Risk level 2 Risk level 3

0% 20% 40% 60% 80%100%

FMI-01

FMI Interconnection-…

FMI Group-01

FMI Group-02

Financial-01

FMI Group-03

Fin Group-01

Financial-02

Financial-03

FMI Group-02

Risk level 0 Risk level 1

Risk level 2 Risk level 3

113

entities also reach high levels of interconnectedness that result in ranges between 28% and 21%

of CCPs connected to them.

404. When aggregating in groups of related entities one observes a general increase in the

interconnectedness indicators of FMIs, this is due to many entities belonging to financial groups

with multiple FMIs. One also observes that some FMIs have operational dependencies within

the central banking system that could theoretically lead to correlated operational risk events due

to their reliance on common infrastructures. For financial entities, no change in their level of

concentration concerning the measurement at individual entity level is observed.

405. Overall, the observed level of interconnectedness for FMIs is in line with the expectations, many

FMIs are connected to multiple CCPs providing critical services and there are some components

of the financial infrastructure that have a systemic nature as they service the whole network of

CCPs either directly or indirectly. With respect to other financial entities, one observes that there

are instances where they are connected to more than one CCP, which implies that potential

correlated operational events that propagate through financial institutions is a plausible scenario.

Operational risk chart 43

Network chart

IT, Software & Telco services – LEI level

Operational risk chart 44

Most interconnected third-party entities

IT, Software & Telco services – LEI level

114

Operational risk chart 45

Network chart

IT, Software & Telco services – Hypothetical

groups

Operational risk chart 46

Most interconnected third-party entities

IT, Software & Telco services –

Hypothetical groups

FIGURE 47: INTERCONNECTEDNESS ANALYSIS – SOFTWARE, IT & TELECOMMUNICATIONS SERVICES

406. When looking at the most interconnected providers of IT services, software, or

Telecommunications services one notices one IT entity connected to 28% of the CCPs (IT

provider-01) and the rest of the entities between the 14% and 20% mark. One also observes that

for the majority of providers, CCPs either have protective tools or the type of impact would be

limited to supporting functions. Only one entity (Telco-01) would have the potential to impact

critical functions at more than one CCP in a correlated manner.

407. When aggregating in groups of related entities one observes a slight increase in concentration

across the board. However, taking into account the risk of the interconnections, for the most part

the CCPs have protective tools in place. For critical functions, the maximum risk for critical CCP

functions would be linked to one entity connected to 21% of CCPs with a Risk level 3 and three

other entities with 14% of CCPs connected each.

Data providers

0% 20% 40% 60% 80% 100%

IT Group-01

Telco Group-01

IT Provider-01

IT Group-02

Telco Group-02

Telco Group-03

Telco Group-04

FMI Group-01

IT Group-02

FMI Group-02

Risk level 0 Risk level 1

Risk level 2 Risk level 3

115

Operational risk chart 47

Network chart

Data providers – LEI level

Operational risk chart 48

Most interconnected third-party entities

Data providers – LEI level

Operational risk chart 49

Network chart

Data providers – Hypothetical groups

Operational risk chart 50

Most interconnected third-party entities

Data providers – Hypothetical groups

FIGURE 48: INTERCONNECTEDNESS ANALYSIS – DATA PROVIDERS

408. When looking at the most interconnected third parties providing data related services, one

observes that only three individual entities are connected to more than one CCP, and for the

most part risk is mitigated through CCP’s protective tools; no individual data provider with the

potential to affect the critical functions of more than one CCP in a correlated manner is observed.

0% 20% 40% 60% 80% 100%

Data provider-01

Data provider-02

Data provider-03

Risk level 0 Risk level 3

0% 20% 40% 60% 80% 100%

FMI Group-01

Data Provider-01

FMI Group-02

FMI Group-03

Data Group-01

Risk level 0 Risk level 3

116

409. When aggregating in groups of related entities no material change other than FMIs appearing

in the selection of entities is observed. FMIs are providers of data for many CCPs, however this

doesn’t raise any incremental concerns as their level of interconnectedness and criticality of

services is already higher when analysing them from the financial services perspective.

Other services

Operational risk chart 51

Network chart

Other services – LEI level

Operational risk chart 52

Most interconnected third-party entities

Other services – LEI level

Operational risk chart 53

Network chart

Other services – Hypothetical groups

Operational risk chart 54

Most interconnected third-party entities

Other services – Hypothetical groups

0% 20% 40% 60% 80% 100%

FMI-01

Intragroup-01

FMI-02

Risk level 0 Risk level 1

Risk level 2

117

FIGURE 49: INTERCONNECTEDNESS ANALYSIS – OTHER SERVICES

410. When looking at the most interconnected entities providing other types of services, one

observes in general low level of interconnectedness and risk.

411. When aggregating in groups of related entities the findings are not materially different, the only

entity which has connection to more than one CCP has a limited impact, as it is a provider related

to a specific commodity product.

4.4.3.3 Evidence of events affecting multiple CCPs

412. During the five-year period of data collected, there are four events registered that affected more

than one CCP during the same day.

Event 1: Telecommunications provider outage

413. A telecommunications provider (Telco-01) experienced an outage causing connectivity issues

to two CCPs, impacting some of their critical functions.

Operational risk chart 55

Entity interconnections – LEI level

Event 1: Telecommunications provider outage

FIGURE 50: TELECOMMUNICATIONS PROVIDER OUTAGE

414. The reported propagation of the outage is consistent with the identified connections and risks

as this non-financial entity is connected to three CCPs but one of them has built redundancy as

protective tool, mitigating its third-party risk exposure.

415. ESMA staff notes that the reported duration of the event is different between entities, with one

CCP reporting approximately 6 hours of incident time and the other CCP reporting approximately

13 hours of incident time.

Event 2: Intragroup entity outage

416. An entity belonging to a financial group with multiple FMIs and providing services to three CCPs

experienced a technology network outage impacting some critical functions of two CCPs.

0% 20% 40% 60% 80% 100%

Telco-01

Risk level 0 Risk level 1 Risk level 2 Risk level 3

118

Operational risk chart 56

Entity interconnections – LEI level

Event 2: Intragroup entity outage

FIGURE 51: INTRAGROUP ENTITY OUTAGE

417. The reported propagation of the outage is consistent with the identified connections and risks

for two CCPs, but there was no reported event for the CCP connected to this third-party provider

with reported potential impact to affect critical supporting functions. By analysing the description

of services provided, the most probable explanation is that the type of service provided to the

third entity is of a different nature, so the lack of correlation seems plausible as operational events

do not necessarily affect at whole entity level.

The reported duration of the event is similar for both entities, around 3 hours.

Event 3: Financial Market Infrastructure outage

418. A financial market infrastructure (FMI-05) experienced technical issues affecting two CCPs and

impacting some of their critical functions.

Operational risk chart 57

Entity interconnections – LEI level

Event 3: Financial Market Infrastructure outage

FIGURE 52: FINANCIAL MARKET INFRASTRUCTURE OUTAGE

419. The reported propagation of the outage is consistent with the identified connections and risks

for three out of the four CCPs connected to this provider. Using the qualitative information

provided, ESMA staff derives as possible reason that the nature of the outage was not at whole

entity level, but rather affecting a subset of entities due to common infrastructure element.

0% 20% 40% 60% 80% 100%

Intragroup-01

Risk level 0 Risk level 1 Risk level 2 Risk level 3

0% 20% 40% 60% 80% 100%

FMI-05

Risk level 0 Risk level 1 Risk level 2 Risk level 3

119

420. One notes that there are significant differences in the reported duration of the event between

CCPs (17 hours and 2.5 hours) which could be due to reporting inconsistencies or differences in

the services used and negative effects experienced.

Event 4: Settlement system outage

421. A settlement system (FMI-04) experienced technical issues affecting two CCPs and impacting

some of their critical functions.

Operational risk chart 58

Entity interconnections – LEI level

Event 4: Settlement system outage

FIGURE 53: SETTLEMENT SYSTEM OUTAGE

422. The reported propagation of the outage is consistent with the identified connections and risks,

with the event affecting the two entities with reported exposure after taking into account risk

management tools.

423. The reported duration of the event is similar for both entities, which is around 12 hours.

4.4.3.4 Conclusions from the analysis of the network of third-party providers

424. In the analysis of the network of critical third-party service providers ESMA staff aggregates the

information provided by individual CCPs in order to understand and assess risks from common

exposures to third-party risk. Through the use of the results from the hypothetical scenario of an

outage at a critical third-party provider, ESMA staff qualifies the risk of each interconnection to

better understand the impact from shocks transmitted through the network of third-party

dependencies.

425. Overall, ESMA staff observes a high level of interconnectedness and criticality in services

provided by FMIs, which is consistent with their role and low level of substitutivity. Three entities

(FMI- 01, 02, 03) have been identified as having a particularly high level of systemic importance,

as their levels of interconnectedness reach quantities of 100% and 50% of CCPs connected to

them.

426. Some financial entities could impact critical functions in more than one CCP simultaneously.

Financial entities may also have roles as clearing members. These aspects should be closely

monitored.

0% 20% 40% 60% 80% 100%

FMI-04

Risk level 0 Risk level 1 Risk level 2 Risk level 3

120

427. Some IT, Software & Telco services, including cloud services, are interconnected with multiple

CCPs, however when taking into account CCP’s risk management tools the potential impact is

substantially mitigated. There is one entity which has critical interconnections with more than one

CCP and has already caused a correlated outage in the past. This interconnection deserves

specific monitoring.

428. For data providers and other types of services, when assessing at LEI level, there does not

seem to be potential for correlated events affecting CCP’s critical functions.

429. Intragroup entities providing services to multiple CCPs should be closely monitored given their

potential to cause correlated operational risk events.

430. When analysing past events, we found no empirical evidence of events affecting groups of

entities (such as the hypothetical groups included in our analysis), only events affecting single

entities (at LEI level) have been registered in the historical timeframe evaluated.

121

5 Conclusions

431. The fourth ESMA CCP stress test aimed to assess the resilience of all 15 authorised EU and

recognised Tier 2 CCPs against adverse market developments and the default of clearing

members. In accordance with the methodology published in June 2021, this exercise covered

both credit and concentration risks, with targeted improvements compared to the previous

exercise. In addition, this exercise included a new operational risk component, which aimed to

assess the level of operational resilience of CCPs with a focus on third-party service provider

risk. As with the previous exercises, the ESRB has delivered the narrative and the adverse

scenario used for this 4th stress test exercise.

432. As with all exercises of this scale and type, it is subject to a number of limitations. While residual

risks from the in-scope components have been highlighted in the report, CCPs are also subject

to other types of risks that are either not or only partially covered in the exercise, but which could

still in isolation challenge their resilience. For example, legal and any type of business risk have

been left outside of the scope of the exercise, as well as environmental risk. ESMA remains

committed to further improve and develop the methodology and scope of the future CCP stress

tests.

433. The report analysed the financial resources held by the 15 in-scope CCPs as of 19 March and

21 April 2021. The aggregate amount of resources available to CCPs on these reference dates

was respectively 423 and 409 billion EUR, an increase compared to the previous exercise. There

was no significant structural change in the overall share of excess collateral or allocation of

resources between margin and default fund contributions. The analysis shows that, while there

was a general increase of provided resources by all clearing members, at the same time the top

participants increased their relative share, pointing to a concentration at clearing member level.

434. The credit stress test results have been computed for two default scenarios and on two

reference dates (19 March end of day and 21 April 2021 intraday). For the March date, additional

costs were considered, namely concentration costs and costs related to wrong-way risk. Under

the Cover-2 per CCP scenario, ESMA assessed the resilience of each CCP to the default of its

top-2 clearing members groups. In all cases, the prefunded resources would be sufficient to

cover the resulting losses under the core credit stress test results. The CCPs could have covered

losses generated by the common market stress scenario with relatively low or moderate

percentage consumptions of available resources. ESMA also performed a sensitivity analysis

and the conclusions seem robust to small changes in the baseline shocks. For one of the dates,

the impact due to concentration and specific wrong-way risk stemming from cleared positions

was included in the baseline scenario calculations. This led to higher losses and consumption

for almost all CCPs but under the considered market scenario these were contained within the

default waterfalls of the CCPs and there was no shortfall of prefunded resources.

435. For the All CCPs cover-2 scenario, two clearing members groups as defaulting at system-wide

level were selected, i.e. the same two clearing member groups for all CCPs. The majority of

CCPs would experience a default of at least one of their clearing members. However, these

consistent scenarios did not put significant stress to any CCP with the % consumption of default

fund-level prefunded resources being relatively low in all cases. This indicates that while CCPs

are highly interconnected through common clearing participants, the exercise did not highlight

any pairs of groups that are at the same time and under the common tested scenario highly

impactful at multiple CCPs.

436. The reverse stress tests analysis assessed the sensitivity of the credit stress results to stepwise

increases in both the number of defaulting groups and the severity of market shocks. Overall,

the analysis shows that incremental changes in market shocks severity are more harmful than

increases in the number of defaulting members. The results have not indicated any systemically

122

relevant adverse impact following small changes in the underlying stress assumptions. For very

large increases of the severity of the market shocks, the observed maximum shortfalls of

prefunded resources following the default of two clearing member groups would not be spread

across CCPs implying that there are different pairs of defaulting groups that would maximise the

shortfalls at different CCPs for these particular dates.

437. The concentration analysis showed that concentrated positions could represent a significant risk

for CCPs, with the overall risk clustered in one or two CCPs for most asset classes.

438. ESMA calculations show that fixed income derivatives have the most concentration risk, with a

total over 29bn EUR. Bonds (including bonds from Repo clearing services) come next with a

total modelled concentration risk of around 11 bn EUR.

439. Concentration in commodity derivatives and in the equity segment (securities and derivatives)

is very significant as well, with around 7bn EUR of concentration risk calculated for each asset

class. There is a very large coverage gap between the system-wide estimated market impact

under ESMA methodology and margin add-ons, for commodity derivatives and to a lesser extent

for equity products.

440. The concentration risk is factored in explicitly in a majority of CCPs through dedicated margin

add-ons. Although all CCPs have market impact risk, 4 CCPs (KDPW, CCPA, KELER, CCG) did

not report any concentration add-ons. Since the data request date, KDPW and CCG have

implemented or are in the process of introducing concentration addons. KELER relies on a

monitoring system to require additional collateral in case of elevated concentration.

441. In the operational risk analysis, ESMA derived insights with respect to the level of operational

resilience of CCPs for 14 CCPs (one was excluded due to the absence of historical operational

events data) and took an in depth look at third-party risk.

442. Using information about internal incidents of CCP’s systems and third-party providers ESMA

developed two methodologies to measure operational risk from historical events. With the

computed results, ESMA identified varying degrees of operational reliability for the CCPs

included in the exercise and identified specific CCPs where further work should be conducted

to understand the drivers of these differences, the root causes of the events and the remediation

actions taken. Further detailed conclusions are provided in section 4.4.1.5.

443. Through the use of a hypothetical scenario, ESMA evaluated the exposures to critical third-party

providers and the ability of CCPs to reduce risk through operational risk management tools.

Using exposure indicators, differences across CCPs in their relative level of third-party risk were

identified. Further work should be conducted to evaluate the individual circumstances of these

exposures and the suitability of taking corrective action to improve operational resilience against

operational shocks affecting critical third-party service providers. Further detailed conclusions

are provided in section 4.4.2.4.

444. In the analysis of the network of critical third-party providers ESMA aggregated the information

provided by individual CCPs in order to understand and assess risks from common exposures

to third-party risk. Overall, ESMA analysed a number of critical third-party service providers,

which have the potential to affect the critical functions of multiple CCPs in a correlated manner.

In addition, ESMA identified the critical third-party service providers with the highest systemic

importance for the CCP sector due to both the criticality of the services provided and their level

of interconnectedness with CCPs.. Further detailed conclusions are provided in section 4.4.3.4.

123

445. As with the three previous exercises, this year’s stress test exercise showed that EU and Tier 2

CCPs are overall resilient to common shocks and multiple defaults. However, the credit stress

test highlighted differences in resilience between CCPs under the selected market stress

scenarios, although no systemic risk has been identified. Similarly, the concentration component

highlighted the need for CCPs to accurately account for liquidation cost within their risk

framework. Finally, with respect to operational resilience, a series of areas and entities have

been identified where further supervisory attention should be put in order to assess discrepancies

in the measured levels of operational risk.

124

6 Annexes

6.1 List of CCPs included in the scope of the exercise

CCP

CCP code

Athens Exchange Clearing House

ATHX

BME Clearing

BME

Cassa di Compensazione e Garanzia S.p.A.

CCG

CCP Austria Abwicklungsstelle für

Börsengeschäfte GmbH

CCPA

Eurex Clearing AG

ECAG

European Commodity Clearing

ECC

European Central Counterparty N.V.

EUROCCP

ICE Clear Europe

ICEEU

ICE Clear Netherlands B.V.

ICENL

KDPW_CCP

KDPW

Keler CCP

KELER

LCH.Clearnet SA

LCHSA

LCH.Clearnet Ltd

LCHUK

Nasdaq OMX Clearing AB

NASDAQ

OMIClear – C.C., S.A.

OMI

125

6.2 Concentration Stress Test annex

6.2.1 Methodology worked example

446. The present section provides a worked example of market impact computation for equity. An

analogous approach is adopted for the other asset classes, with some differences for Fixed

Income and Credit (as specified in 6.2.2).

Data reported and Reference volume computation

447. For each aggregation level (ISIN for equities), CCPs reported the account position values and

the relevant Average Daily Volume (ADV).

448. For equities, the reference volume is by default taken as the systematic internaliser data

average volume, or the ADV submitted by the CCP as a fallback.

Market impact

449. The market impact (in basis points) is retrieved at {ISIN, CM} level by computing in turn the

following quantities:

• Position value to liquidate: the net positions across all accounts of this CM for this ISIN.

• Position size to liquidate: ratio of the Position value to liquidate and the reference volume.

• Only absolute positions greater than 0.25 are considered significant.

• The market impact is then interpolated linearly from the Sensitivity tables using the significant

position size to liquidate. For positions larger than 200% of the reference volume, a flat

extrapolation is applied.

450. The table below illustrates the computation.

ISIN

LEI

Position

Value

REF

VOLUME

POSITION

VALUE

TOLIQ

POSITION

SIZE TOLIQ

SIGNIFICANT

POSITION SIZE

TOLIQ

MARKET

IMPACT

ISIN 1

CM1

-5000

10000

-5000

-0.5

0.5

292

ISIN 1

CM2

15000

10000

12000

1.2

700

ISIN 1

CM2

-3000

10000

12000

1.2

700

ISIN 1

CM3

10000

-10000

-1

583

ISIN 1

CM3

-10000

10000

-10000

-1

583

451. Notice that the Market impact is retrieved by the relevant entry of the Sensitivity table.

Asset Class

Sub-Asset Class

Size cost (bps) 25%

Ref Volume

Size cost (bps) 50%

Ref Volume

Size cost (bps)

100% Ref Volume

Size cost (bps)

200% Ref Volume

Stocks

Mid cap

146

292

583

1,166

Market impact delta PnL

126

452. Finally, the market impact delta PnL is computed by scaling the market impact by the account

position, also considering the case where account positions reduce the positions to be liquidated.

ISIN

LEI

Position

Value

POSITION

VALUE

TOLIQ

MARKET

IMPACT

MARKET

IMPACT

DELTA PNL

ISIN 1

CM1

-5000

292

146

ISIN 1

CM2

15000

12000

700

1050

ISIN 1

CM2

-3000

12000

700

-210

ISIN 1

CM3

-10000

583

ISIN 1

CM3

-10000

583

6.2.2 Specific Concentration Methodologies

6.2.2.1 Fixed Income Derivatives Methodology

453. For both Fixed Income Derivatives and Credit Default Swaps, the concentration risk is assessed

through the market impact cost of setting-up a relevant hedging portfolio. The further costs

incurred from auctioning the portfolio are not considered.

454. To allow the accurate pricing and hedging of swaps, CCPs have reported the position

sensitivities to both forecasting and discounting curves on 15 maturity points spanning 1Y to

50Y.

455. In each relevant currency, it is assumed that the main risks can be adequately hedged through

hedging the exposures to both the discounting and the forecasting curves on 4 different maturity

points (2Y, 5Y, 10Y, 30Y). ESMA staff apportioned the sensitivities to the 4 hedge maturity points

on a time basis.

456. Basis swaps between OIS and IBOR were also considered as a possible hedge. The most

favourable market impact using one of the 3 possible hedging strategies {forecasting +

discounting, forecasting + basis and discounting + basis} was kept.

457. For each pillar, a concentration cost per hedge maturity is computed by using the relevant size

through interpolation.

458. A separate market impact for each of the 3 reporting sub-asset classes (Bond futures / forwards,

IR futures and FRA and Swaps) is computed.

6.2.2.2 Credit Derivatives Methodology

459. CDSs are modelled similarly to interest swap derivatives: the CDS curve is assumed to be

hedged on the 4 different maturity points (1Y, 2Y, 5Y, 10Y).

460. Given the practice of the market to use 5Y instruments as a hedge (with the exception of

distressed credits close to default but this is expected to be a minor part of the inventory of

positions), the expected cost of setting up a hedge takes the 5Y as a reference. Costs to set up

hedges for the other maturities (1Y,2Y,10Y) are defined as multiples of the 5Y reference.

461. Although the model was enriched in this exercise with a term structure for the hedging costs,

CCPs clearing CDSs use more complex models than the approach chosen by the framework.

Moreover, with the parameters provided by the 2 CCPs, the model produced much lower

concentration risk than what the CCPs charge their clearing members.

127

462. It is therefore difficult to draw conclusions from the results on that asset class.

6.2.3 Sensitivity parameters

463. Based on the data reported by CCPs, and in accordance with the methodology described in

section 3.5, system-wide sensitivity tables have been built for each sub-asset class.

464. A selection of the most important system-wide sensitivity parameters is reported below.

465. For each asset class, the Figures below show how the market impact rises when increasing the

position size (in bps).

466. Below are provided worked out examples (Table 7 to Table 17) of the market impact for

representative large positions in each asset class. This ensures transparency on the parameters

and inputs used, which are based on the CCPs’ inputs.

Bonds

467. Bonds sensitivity estimates are provided by issuer type (Corporate/Sovereign), rating

(Investment grade/ Non-investment grade) and maturity. Figure 54 and Table 7 show that, based

on data reported by CCPs, for similar positions the market impact for bonds is generally higher

for corporate than for sovereign bonds and tends to grow faster with position size for longer

maturities.

FIGURE 54: MARKET IMPACT VS. RELATIVE POSITION SIZE, INVESTMENT GRADE CORPORATE AND

SOVEREIGN BONDS

100

150

200

250

25 50 100 200

IG Corporate Bonds

Corporate Bond / IG / 1-5 y

Corporate Bond / IG / < 1y

Corporate Bond / IG / > 5y

Note: Market impact vs relative position size in bps.

Sources: ESMA

100

120

25 50 100 200

IG Sovereign Bonds

Sovereign Bond / IG / 1-5 y

Sovereign Bond / IG / < 1y

Sovereign Bond / IG / > 5y

Note: Market impact vs relative position size in bps.

Sources: ESMA

128

TABLE 7: MARKET IMPACT ON REPRESENTATIVE LARGE POSITIONS, INVESTMENT GRADE BONDS

Investment Grade Bonds

Sub-asset

Class

Maturity

POSITION

VALUE TOLIQ

(k€)

Reference

Volume (k€)

Significant

Position

Size

Market

Impact

(bps)

Market

Impact

(k€)

Sovereign Bond

1 to 5 years

-555,916

75,191

7.40

1,622

Corporate Bond

1 to 5 years

-6,572

1,681

3.91

184

121

Corporate Bond

< 1 year

5,014

3,050

1.64

119

Sovereign Bond

> 5 years

55,542

28,300

1.96

108

601

Equities

468. Equities sensitivity estimates are differentiated by capitalization size (Small/Mid/Big cap).

Overall, Figure 55 shows a similar evolution of the market impact with size across all equity

instruments (although for no clear reason, according to the data reported by CCPs market impact

for Mid Cap tends to grow faster with size).

FIGURE 55: MARKET IMPACT VS. RELATIVE POSITION SIZE, EQUITIES AND EQUITY DERIVATIVES

100

200

300

400

500

600

700

800

25 50 100 200

Equities and Equity Derivatives

Equity Derivatives / ETF futures/forwards

Equity Derivatives / Stock index futures/forwards

Stocks / Big cap

Stocks / Mid cap

Stocks / Small cap

Note: Market impact vs relative position size in bps.

Sources: ESMA

129

TABLE 8: MARKET IMPACT ON REPRESENTATIVE LARGE POSITIONS, SINGLE NAME EQUITY DERIVATIVES

AND SECURITIES

Single Name Equity Derivatives and Securities

Sub-asset

Class

POSITION VALUE

TOLIQ (k€)

Reference

Volume (k€)

Position

Size

Market

Impact

(bps)

Market

Impact

(k€)

Small cap

-447,313

207,920

2.15

602

Small cap

-42,445

24,189

1.75

528

1,599

Mid cap

11,327

6,580

1.72

570

155

Mid cap

27,744

13,070

2.12

700

1,513

Big cap

18,799

9,092

2.07

500

664

Big cap

30,679

16,710

1.84

456

431

TABLE 9: MARKET IMPACT ON REPRESENTATIVE LARGE POSITIONS, OTHER EQUITY DERIVATIVES

Other Equity Derivatives

Sub-asset

Class

POSITION

VALUE TOLIQ

(k€)

Reference

Volume (k€)

Significant

Position

Size

Market

Impact

(bps)

Market

Impact

(k€)

ETF futures/forwards

-6,906

24,250

0.28

Stock index futures/forwards

1,023,950

175,263

5.84

359

36,794

Stock index futures/forwards

-4,265,135

1,144,257

3.72

359

153,263

Stock index futures/forwards

-302,723

357,426

0.85

120

3,634

Energy and other commodity derivatives

469. Figure 56 shows that for comparable positions median sensitivities for electricity are lower than

for other commodities. Overall, for most commodities the sensitivity growths almost linearly with

the size of the position.

FIGURE 56: MARKET IMPACT VS. RELATIVE POSITION SIZE, ENERGY AND COMMODITY DERIVATIVES

100

150

200

250

300

350

400

450

500

25 50 100 200

Energy Derivatives

Electricity futures/forwards

Energy commodity futures/forwards

Note: Market impact vs relative position size in bps.

100

200

300

400

500

600

700

25 50 100 200

EUA, Agricultural and Freight

Derivatives

Agricultural commodity futures/forwards

European Union Allowances (EUA)

Freight Derivatives

Note: Market impact vs relative position size in bps.

130

TABLE 10: MARKET IMPACT ON REPRESENTATIVE LARGE POSITIONS, ENERGY COMMODITY

FUTURES/FORWARDS

Energy commodity futures/forwards

Sub-asset

Class

Segment

Underlying

Energy

Maturity

POSITION

VALUE

TOLIQ (k€)

Reference

Volume

(k€)

Position

Size

Market

Impact

(bps)

Market

Impact

(k€)

electricity

BSLD

baseload

1Y-2Y

191,731

48,656

3.94

267

5,112

natural gas

NCGG

baseload

0-1M

254

104

2.44

458

coal

COAL

1Y-2Y

349,168

85,779

4.07

458

2,120

oil

KERO

1Y-2Y

-91,340

30,555

2.99

458

4,185

oil

BRNT

1Y-2Y

2,926,389

2,671,941

1.09

300

23,730

Note: the delivery point is a segmentation criterium for electricity derivatives.

TABLE 11: MARKET IMPACT ON REPRESENTATIVE LARGE POSITIONS, AGRICULTURAL COMMODITY

FUTURES/FORWARDS

Agricultural commodity futures/forwards

Underlying

Maturity

POSITION

VALUE

TOLIQ

(k€)

Reference

Volume (k€)

Significant

Position

Size

Market

Impact

(bps)

Market

Impact

(k€)

DIRY

0-3M

-3,407

821

4.15

592

CCOA

1Y-2Y

95,546

28,601

3.34

592

131

WHSG

0-3M

128,745

140,536

0.92

242

3,178

SEAF

0-3M

394

119

3.31

592

TABLE 12: MARKET IMPACT ON REPRESENTATIVE LARGE POSITIONS, FREIGHT DERIVATIVES

Freight Derivatives

Segment

Maturity

POSITION

VALUE

TOLIQ

(k€)

Reference

Volume

(k€)

Significant

Position

Size

Market

Impact

(bps)

Market

Impact

(k€)

Dry bulk

Capesize

CPT

2Y-3Y

660

8.70

456

Dry bulk

Supramax

SPT

1Y-2Y

518

22.8

456

Dry bulk

Panamax

PTC

2Y-3Y

-8,715

108

456

397

Tanker

265,000mt

ME Gulf to China

1Y-2Y

-86,733

1,562

55.52

455

3,952

Tanker

55,000mt

ME to Japan

1Y-2Y

-4,573

150

30.42

455

208

Tanker

130,000mt

W Africa to Cont

9M-1Y

777

455

Note: freight positions are typically very large compared to the average daily notional amounts.

TABLE 13: MARKET IMPACT ON REPRESENTATIVE LARGE POSITIONS, EUA

Emission Allowances - European Union Allowances (EUA)

131

Maturity

POSITION VALUE

TOLIQ (k€)

Reference

Volume (k€)

Significant

Position

Size

Market

Impact

(bps)

Market

Impact

(k€)

0-4M

133,915

47,833

2.8

458

6,136

1Y-2Y

-1,008,797

622,562

1.62

392

39,538

2Y-3Y

408,174

196,353

2.08

458

20,865

5Y-6Y

395

10.1

458

Fixed Income Derivatives

470. For each currency, curve type (Discounting, Forecasting, OIS vs IBOR) and maturity point (2Y,

5Y, 10Y, 30Y), the highest submitted PV01 reference notional was chosen to ensure the best

possible coverage across clearing members concentrated positions.

471. A linear fit was used to generate the common sensitivity table for each currency. Figure 57

shows a similar behaviour for all maturity points of the curve, with very steep increases (almost

exponential) for large position sizes. Moreover, for comparable positions median sensitivities for

30Y are higher than for other maturities.

FIGURE 57: MARKET IMPACT VS. RELATIVE POSITION SIZE, EUR FIXED INCOME DERIVATIVES

472. The following tables show worked out examples of market impact computation for typical

representative fixed income derivatives positions.

840,972 1,681,944 3,363,889 6,727,777 16,819,443

EUR Discounting Curve (OIS)

2Y 5Y 10Y 30Y

Note: Hedging market impact vs position size in bps.

Sources: ESMA

840,972 1,681,944 3,363,889 6,727,777 16,819,443

EUR Forecasting Curve

2Y 5Y 10Y 30Y

Note: Hedging market impact vs position size in bps.

Sources: ESMA

132

TABLE 14: MARKET IMPACT ON REPRESENTATIVE LARGE POSITIONS, EUR FIXED INCOME DERIVATIVES

EUR - Fixed Income Derivatives

Curve Type

Maturity

Hedge

PV01

Hedge

Size

Market

Impact

(bps)

Market

Impact

Pnl EUR

Discounting

-6,255,485

1.86

3.92

20,441,036

Discounting

7,837,568

2.33

5.00

-404,604

Discounting

10Y

5,103,520

1.52

3.72

198,429

Discounting

30Y

15,106,506

4.49

11.82

178,497,700

Forecasting

2,999,695

0.89

1.84

27,351

Forecasting

7,769,832

2.31

4.34

5,688,355

Forecasting

10Y

8,236,009

2.45

4.88

38,903,982

Forecasting

30Y

-7,520,312

2.24

5.79

27,255,892

TABLE 15: MARKET IMPACT ON REPRESENTATIVE LARGE POSITIONS, GBP FIXED INCOME DERIVATIVES

GBP - Fixed Income Derivatives

Curve Type

Maturity

Hedge

PV01

Hedge

Size

Market

Impact

(bps)

Market

Impact

Pnl GBP

Discounting

-9,751,769

3.38

9.42

66,315,816

Discounting

835,349

0.29

1.57

1,315,221

Discounting

10Y

6,318,825

2.19

6.77

42,761,202

Discounting

30Y

-9,171,831

3.18

11.27

5,756.15

Forecasting

-5,642,366

1.96

5.71

21,448,955

Forecasting

835,349

0.29

1.57

1,315,221

Forecasting

10Y

-1,824,578

0.63

2.40

4,380,934

Forecasting

30Y

-7,586,368

2.63

9.49

6,472,623

TABLE 16: MARKET IMPACT ON REPRESENTATIVE LARGE POSITIONS, USD FIXED INCOME DERIVATIVES

USD - Fixed Income Derivatives

Curve Type

Maturity

Hedge

PV01

Hedge

Size

Market

Impact

(bps)

Market

Impact

Pnl USD

Discounting

13,089,362

3.27

13.35

174,793,884

Discounting

24,606,161

6.15

15.94

3,975

Discounting

10Y

-8,184,405

2.05

7.80

63,869,864

Discounting

30Y

11,335

2.83

12.82

659,925

Forecasting

7,310,734

1.83

4.64

81,805

Forecasting

13,635,768

3.41

9.17

15,513,998

Forecasting

10Y

10,903,113

2.73

7.74

2,663,816

Forecasting

30Y

13,714,725

3.43

11.95

136,917

133

Credit Derivatives

473. The concentration risk for credit derivatives is assessed through the market impact cost of

setting-up a relevant hedging portfolio. Such hedging is assumed to be done using the 5Y

maturity only.

474. CCPs reported identical hedging cost parameters for off-the-run and on-the-run series.

475. Overall, Figure 58 shows that the sensitivity growths in a similar way with position size for all

sub-asset classes considered. As expected, the market impact is somehow higher when the

credit quality of the underlying decreases (e.g. market impact for CDX HY is higher than for CDX

IG Main).

FIGURE 58: MARKET IMPACT VS. RELATIVE POSITION SIZE, CREDIT DERIVATIVES

25% 50% 100% 200% 500%

ITraxx Indices

ITraxx CrossOver

ITraxx Fin Snr

ITraxx Fin Sub

ITraxxMain

Note: Hedging market impact vs share of average daily reference risk

Sources: ESMA

25% 50% 100% 200% 500%

CDX Indices

CDX HY CDX IG Main

Note: Hedging market impact vs share of average daily reference risk

Sources: ESMA

25% 50% 100% 200% 500%

Single Names

HY Single Name IG Single Name

Note: Hedging market impact vs share of average daily reference risk

Sources: ESMA

134

TABLE 17: MARKET IMPACT ON REPRESENTATIVE LARGE POSITIONS, CREDIT DERIVATIVES

Credit Derivatives

Currency

Curve Type

Average Daily

Risk Per Basis

Point

Position Risk

Per Basis

Point

Relative

Hedge

Size

Market

Impact

(bps)

Market

Impact

Pnl

USD

CDX.HY off-the-run

33,080

23,361

0.71

22.81

448,122

USD

CDX.HY on-the-run

645,067

-691,480

1.07

5.95

291,145

USD

CDX.IG Main on-the-run

3,263,256

-906,974

0.28

1.34

607,924

EUR

CrossOver off-the-run

372,885

177,779

0.48

4.38

777,783

EUR

CrossOver on-the-run

654,561

-598,076

0.91

4.69

2,501,753

EUR

Fin Snr off-the-run

167,740

-274,625

1.64

3.07

10,871

EUR

Fin Snr on-the-run

571,759

-692,724

1.21

1.51

1,115,416

EUR

Fin Sub off-the-run

472,156

-287,071

0.61

6.67

1,913,469

EUR

Fin Sub on-the-run

118,134

-220,401

1.87

9.41

2,490,812

EUR

HY Single Name

22,710

17,096

0.75

14.26

243,877

EUR

IG Single Name

23,565

-59,063

2.51

2.72

160,868

EUR

ITraxxMain off-the-run

488,572

-1,558,635

3.19

5.55

8,813,460

EUR

ITraxxMain on-the-run

3,348,303

-2,259,130

0.67

1.44

2,239,382

6.2.3.1 Position overlap analysis

476. The impact of liquidating overlapping positions held by one (or more) clearing member(s) at one

(or more) CCP(s) is driven by their total net exposure, which is used to determine the market

impact in bps. Therefore, offsetting (resp. same direction) positions will reduce (resp. increase)

the cost of liquidating each position.

477. Under a real-life default scenario, it is assumed that all CCPs would liquidate the defaulting

clearing members’ positions at the same time. Similarly, a default of clearing member group

would trigger the liquidation of the positions of all its clearing members.

478. Across all asset classes, the aggregation at clearing member group level somewhat reduces

the total market impact risk.

TABLE 18: GROUPING ASSUMPTIONS ON TOTAL SYSTEM-WIDE MARKET IMPACT

by Clearing

Member

by Group

Difference

Fixed Income Derivatives

33,367

29,043

-12.96%

Bonds

11,432

11,110

-2.81%

Commodity Derivatives

7,544

7,425

-1.58%

Index Equity Derivatives

3,827

3,456

-9.69%

Single Stock Equities & Derivatives

3,462

3,409

-1.51%

Emission Allowances

2,491

2,405

-3.47%

Credit Derivatives

676

626

-7.34%

Freight Derivatives

132

128

-2.99%

135

479. The biggest offsets appear to be between clearing members of the same group in fixed income

derivatives, index equity derivatives and credit derivatives.

480. The aggregation of the positions held by clearing member groups across multiple CCPs has

little impact on the concentration risk.

TABLE 19: IMPACT OF THE LEVEL OF AGGREGATION OF CM GROUPS POSITIONS

at CCP Level

across CCPs

Difference

Fixed Income Derivatives

29,043

29,279

0.81%

Bonds

11,110

10,948

-1.46%

Commodity Derivatives

7,425

7,408

-0.23%

Single Stock Equities & Derivatives

3,409

3,540

3.82%

Index Equity Derivatives

3,456

3,366

-2.63%

Emission Allowances

2,405

2,336

-2.86%

Credit Derivatives

626

0.00%

Freight Derivatives

128

0.00%

481. As the aggregation of positions across multiple CCPs has no significant impact for all asset

classes, and to simplify the interpretation and reconciliation of the results, market impacts have

been reported assuming a liquidation at the CCP level unless specified otherwise.

136

6.3 Operational risk analysis annex

6.3.1 Modelling, calibration and model risk analysis for the model used to estimate

percentile quantities of unavailability per CCP

Modelling approach

482. In terms of modelling approaches, the frequency of operational risk events is assumed to follow

a Poisson distribution. This distribution implies that losses happen randomly through time, so

that in any short period of time  (a year on this case) there is a probability  of an operational

risk event occurring. The probability that operational risk events arise over a year is given by:

























483. Regarding the severity distribution  (which relates to the duration of disruption time), it is

assumed that the duration of events follows a lognormal distribution: 







, which has a

fatter tail than the (truncated) normal distribution. The probability density function  is given by:

































 











Estimation

484. ESMA staff applies the method outlined above separately to two datasets: the events related to

clearing and settlement unavailable and the events related to critical supporting functions

unavailable. In both cases events that affect less than 10% of the clearing activity of the CCP

are excluded.

485. For the frequency distribution, the parameter  is estimated separately for each CCP and the

two types of events, equal to the average number of events per year rounded up to the nearest

integer (since the Poisson distribution requires an integer parameter). The average is rounded

up to ensure that the analysis remains conservative.

Clearing or settlement unavailable

Critical supporting functions unavailable

Average per year



Average



CCP1

0.2

CCP2

0.6

CCP3

1.8

0.4

CCP4

CCP5

3.6

CCP6

0.8

137

CCP7

5.2

2.8

CCP8

0.2

CCP9

2.8

1.6

CCP10

0.4

CCP11

1.4

0.2

CCP12

0.6

CCP13

0.4

CCP14

0.6

0.8

Average

1.6

1.2

TABLE 20: FREQUENCY POISSON DISTRIBUTION – ESTIMATED PARAMETER Λ

486. Regarding the severity distribution, it is not possible to obtain reliable estimates of the

distribution separately for each CCP due to the shortage of data points

. In order to tackle this

limitation in the available data, CCPs are classified into four buckets (of 3 or 4 CCPs) for each

type of events based on their average disruption time (Chart 59). For each bucket as well as for

the whole sample (labelled ‘average’), the parameters of the lognormal distribution are estimated

by maximum likelihood. This approach allows us to approximate the severity distribution of

individual CCPs by using groups that have exhibited similar average disruption times.

487. Chart 60 shows the resulting distributions for the clearing unavailable events for the four buckets

as well as when using all the data (orange series). Group 1 and 2 have distributions concentrated

around the average, while group 3 and 4 have larger tails (with a larger proportion of events

above 5 hours).

488. As an illustration, Chart 61 shows an example of the estimation of the severity distribution for

the clearing and settlement unavailable events (red line) and the histogram of the actual data.

Operational risk chart 59

Average disruption time by group of CCPs

Operational risk chart 60

Severity distributions for clearing NA

As described in the previous section, some CCPs report one single event over the reporting period and most CCPs report less

than 10 events. Given the small sample size, the estimation of parameters would not be reliable.

138

Operational risk chart 61

Data for clearing NA and estimated severity distribution

FIGURE 59: SEVERITY DISTRIBUTION – AVERAGE DISRUPTION TIME AND SEVERITY DISTRIBUTION

ESTIMATE BY GROUPS OF CCPS

489. Based on the severity distributions, it is possible to calculate the probability that if an event were

to occur, it would last longer than the 2-hour recovery target. As shown in the chart below, for

CCPs in the high/highest severity groups, there is more than a 50% probability that an

operational risk event occurring could last more than the target recovery time.

Group 1 Group 2 Group 3 Group 4

Clearing or settlement NA Critical supporting functions NA

Note: Average diisruption time by type of events and by group of CCPs, in

hours.

Sources: CCPs, ESMA.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

all Group 1 Group 2

Group 3 Group 4

Note: Distribution of disruption time for clearing or settlement unavailable by

group.

Sources: CCPS, ESMA.

139

Operational risk chart 62

Probability of event lasting more than 2h

More than 50% for high severity groups

FIGURE 60: PROBABILITY OF EVENT LASTING MORE THAN 2H BY SEVERITY GROUPS

490. Finally, for each CCP and type of events 100,000 Monte Carlo simulations are run to estimate

the aggregated disruption time distribution. This distribution provides the total amount of

disruption time for each CCP and type of events over one year. From the distribution one derives

percentile risk measures of aggregated time of unavailability / disruption over a one-year period:

average and median disruption time over one year, as well as the 95% VaR (the total disruption

time in 95% of the cases) and the 95% expected shortfall (the average of total disruption time

above the 95% VaR). Although VaR and expected shortfall are usually metrics linked to monetary

quantities, this terminology is used for quantifications of time as they allow a straightforward

interpretation of the underlying mathematical formulation.

Model risk

491. The model of operational risk events presented above and that was used to derive the simulation

results rests on a number of assumptions. These are generally inherited from the Loss

Distribution Approach for operational risk, upon which the model was based.

492. One fundamental assumption is that of independence, both between the number of events

occurring in a period and their severity, and between the severity of different events. This

assumption is made for simplicity, as the introduction of a consistent correlation structure

requires careful consideration. Given the early stages of this exploration of operational risk in

CCPs and the relatively sparse dataset, calibrating the parameters required to represent a logical

correlation structure satisfying these constraints would be very challenging.

493. The other assumptions in the model are distributional assumptions, in particular the choice of

Poisson distribution for the number of events and the lognormal distribution for their severity.

The choice of Poisson distribution is very common for models of discrete events. Alternatives

such as the negative binomial are available, but they usually require the calibration of additional

parameters. With the current data limitations this is not straightforward. Considering also the

10%

20%

30%

40%

50%

60%

70%

80%

90%

Lowest

severity

Low

severity

All sample High

severity

Highest

severity

Clearing or settlement NA Critical supporting functions NA

Note: Probability of having events with disruption time longer than 2 hours in a

year, in percent.

Sources:CCPs, ESMA.

140

conservativeness in the calibration of the Poisson distribution, where the observed frequencies

were rounded up substantially, it was decided not to investigate this aspect further for the

moment.

494. On the other hand, the representation of the events’ severity via a lognormal model can be

challenged in a more concrete way. From a statistical point of view, the P-value of the Jarque-

Bera test applied to the logarithms of the severities of the ‘clearing or settlement unavailable’

events is 0.04%, while for the ‘critical supporting function unavailable’ events it is 9%. Therefore,

while the assumption of lognormality is plausible for the second dataset, it is very unlikely for the

first. However, for practical reasons one may still accept the lognormal assumption as starting

point in both cases, and then test the sensitivity of the model to alternative choices. In particular,

statistics such as the high percentiles, averages and expected shortfall previously reported may

depend on the right tail of the distribution used in the model. While the lognormal distribution has

a relatively heavy right tail, other distributional choices might produce substantially different

results.

495. To test the robustness of the simulation results to the choice of distribution for the events’

severity, alternative models based on Student  distribution were fitted to the data. To be more

precise, the logarithms of the ‘clearing and settlement unavailable’ and ‘critical supporting

function unavailable’ severity data sets were modelled with two alternative distributions each,

obtained from  distributions with 3 and 5 degrees of freedom by a linear transformation in order

to match the mean and standard deviation of the observations.

496. In other words, the events’ severity was modelled as 



where  indicates a Student 

random variable with 3 or 5 degrees of freedom, and  and  are location and scale parameters

designed to match mean and standard deviation. This is in contrast with the lognormal model













where  is a standard normal variable.

497. ESMA staff would like to stress that this does not endorse the choice of the  distribution as an

appropriate model for the dataset at hand. The purpose of the exercise is to assess the impact

of the lognormal distribution assumption on statistics derived from the model, in particular where

these depend on the tails of the distribution. In this context, the  distribution with low degrees of

freedom was chosen simply for its well-known property of very heavy tails. Therefore, the results

from the -based model should be seen as a boundary case to assess the sensitivity of the

simulation results to distributional assumptions.

498. The simulation results show that the choice of an alternative distribution with very heavy tails

has little effect on the median and 95% VaR metric. This indicates that the information in the

data, together with the other assumptions of this model (independence and modelling of the

events frequency), is sufficient to obtain reliable tail statistics at this level of confidence.

Operational risk chart 63

Clearing or settlement unavailable

Total disruption time for the Average group

Operational risk chart 64

Critical supporting function unavailable

Total disruption time for the Average group

141

FIGURE 61: COMPARISON BETWEEN LOGNORMAL DISTRIBUTION AND STUDENT’S T-DISTRIBUTION

499. The situation changes when looking further into the tails. For example, the 99% VaR is

significantly higher for the -based models compared to the lognormal. This indicates that

statistics at this level or those that may be affected by realisations in the far tails of the simulation

model, such as the average or expected shortfall measures, should be taken with caution. This

is a natural observation given the early stage of the investigation into operational risk for CCPs

and the limited amount of data available.

Use of the model and its limitations

500. After developing this model and assessing its limitations, ESMA staff considered that a 95%

percentile measurement is a useful risk metric to help understand how the entities in scope could

potentially behave in a “bad year” (1/20 year) and complement the assessment based on

average metrics that describe how is an “average year”; ESMA staff considered that usable

percentile metrics at higher levels of confidence or measures such as expected shortfall cannot

be derived reliably from the current model due to the small sample size.

100

120

140

median VaR 95% VaR 99%

Lognormal t (df=5) t (df=3)

Note: Sum of disruption time over one-year for clearing or settlement

unavailable, in hours

100

120

140

median VaR 95% VaR 99%

Lognormal t (df=5) t (df=3)

Note: Sum of disruption time over one-year for critical supporting function

unavailable, in hours