Post on 03-May-2022
Fixed Income Risk Engine
Process Overview
│ 3
Fixed Income Risk EngineOverall process and next steps
Submission to College of Regulators for
approvalTests with Clients
Indicative Go-Live date (subject to approval by CoR)
SEPT/OCT 2021 OCT 2021 to FEB 2022 Q1 2022
• Instruments in scope of the methodology change include Italian, Spanish, Portuguese and Irish government bonds cleared in bonds and ICSD bonds clearing sections
• Current SPAN-like margining methodology will continue to apply to all the remaining bond instruments listed in the above-mentioned clearing sections (i.e. corporate bonds, as well as government bonds issued by countries other than those specifiedabove)
Tests with Clients will take place starting from October 2021 up to February 2022
NEXT STEPS
Relevant documentation and further info will be made available on CC&G website in a dedicated VAR section under Risk Management menu
General Framework
│ 5
Fixed Income Risk EngineChange in the framework
SPAN (Standardized Portfolio Analysis of Risk) VaR (Value at Risk)
Industry standard for a long time.
Its adoption is currently declining as it hardly conforms to increasingmarket complexity
Identified as the next-to-come market best practice (CC&G is workingtoward the implementation of a VaR-like model for the FI Section)
Industry Adoption
MainCharacteristics
INITIAL MARGINS
MARGIN INTERVAL
INITIAL MARGINS
SCENARIOS
PORTFOLIO PnL
Based on each underlying’s time series
Usually 10 to 20 (range based on defined Margin Interval)
Re-evaluation based on definedscenarios
SCENARIOS
PORTFOLIO PnL
Usually more than 1000 (based on lookback period)
Re-evaluation based on definedscenarios
Pros/Cons
+ Proven track record of efficiency during stressed periods
- Scenarios at instrument level
-Correlations between products are not directly managed
through scanning losses procedures
+ Scenarios at portfolio level
+Correlations between products are directly managed
through developed scenarios
+ Standardized across CCPs - Every CCP is developing its own (different) methodology
+/-Updated at discrete points in time (every time Margin Intervals are). Pro-cyclicality can be directly controlled
+/-Dinamically updated. Pro-cyclicality is not directly
controlled
│ 6
Fixed Income Risk EngineRisk Framework
Margin Component Risk Factor/sModel Engine
Mark-to-Market margin
Expected Shortfall margin
Unscaled ES
Scaled ES
Spectral Risk Measures
SS1 margin
Repo-concentration margin
Decorrelation margin
Idiosyncratic-concentration margin
CORE ENGINE
ADD-ONs ENGINE
Market Price Risk (from trade date to evaluation date)
Market Price Risk (what-if scenarios)
Under/Overestimation of Market Price Risk
(pro-cyclicality concerns vs adequate reactiveness of the model)
Tail Dilution Risk
Break-in-correlations Risk
Specific ISIN liquidity / concentration Risk
Settlement Risk
Repo Rate Risk / Repo Term Concentration Risk
Core Engine
│ 8
Fixed Income Risk EngineCore Engine: Mark-to-Market Margins
Mark-to-Market Margins are defined according to the typology of the trades to be settled (cash, repo or forward starting repo) and their
respective close-out procedure that would be put in place by the CCP in case of default of one of the two counterparties
MtM Margins computation for cash trades1
A B
p
bond
trade date settlement datet
(evaluation date)
Trade between A e B
TRADE EXAMPLE
MARGINS CALCULATION DATE
│ 9
Fixed Income Risk EngineCore Engine: Mark-to-Market Margins
CLOSE-OUT PROCEDURE IN CASE OF DEFAULT OF CTP A
CLOSE-OUT PROCEDURE IN CASE OF DEFAULT OF CTP B
CCP
Bp
bond
CCP
Cbond
p 2
t
(default A)
settlement date 2settlement date
Trade between CCP and C
ACCP
p
bond
CCP
Cp 2
bond
t
(default B)
settlement date 2settlement date
Trade between CCP and C
Operations put in place
by the CCP in order to close the original positions
│ 10
Fixed Income Risk EngineCore Engine: Mark-to-Market Margins
As the difference between the settlement date of the original trade and the settlement date of trade originated by the CCP is negligible (usually 1
business day), Mark-to-Market Margins for cash trades can be defined through the following formula:
MtMmargin = N ∗Pmarket+ AISD
100−
Ptrade+ AISD100
∗ ps
MTM MARGINS FORMULA FOR CASH TRADES
N Nominal value of the original trade
P_market Price of the traded security at evaluation date
P_trade Original trading price of the security
AI_SD Accrued interest of the security at settlement date of the original trade
ps Position sign (+ long / - short)
│ 11
Fixed Income Risk EngineCore Engine: Mark-to-Market Margins
MtM Margins computation for repo trades2
A B
p + R1
bond
A B
bond
p
spot termt
(evaluation date)
TRADE EXAMPLE
MARGINS CALCULATION DATE
In the example above spot refers to the settlement date of the spot leg of the repo trade and term refers to the settlement date of the term leg of the repo
trade. The repo interest paid by the cash borrower is represented by R1
│ 12
Fixed Income Risk EngineCore Engine: Mark-to-Market Margins
CLOSE-OUT PROCEDURE IN CASE OF DEFAULT OF CTP A
CLOSE-OUT PROCEDURE IN CASE OF DEFAULT OF CTP B
CCP
Bp + R1
bond
CCP
Cp 2
bond
t
(default A)
term
CCP
Dbond
p 2
CCP
Cbond
p 2 + R2cash
spot 2
term 2
term 1
CCP
Abond
p + R1
CCP
Cbond
p 2
t
(default B)
term
CCP
Dp 2
bond
CCP
Cp 2 + R2
bondcash
spot 2
term 2
term 1
Operations put in place
by the CCP in order to close the original
positions
│ 13
Fixed Income Risk EngineCore Engine: Mark-to-Market Margins
The net outflow for the CCP in t amounts to zero as the single flows are perfectly offset. The difference between the original settlement price (p)
and the settlement price of the close-out trade put in place by the CCP (p2) is realised at term date, as well as the difference between the repo
interest to be paid by original the cash borrower (R1) and the repo interest that refers to the close-out trade (R2), therefore MtM margins for repo
trades can be defined as follows:
MTM MARGINS FORMULA FOR REPO TRADES
N Nominal value of the original trade
P_market Price of the traded security at evaluation date
P_trade Original trading price of the security
AI_ED1 / AI_SSD Accrued interest of the security at evaluation date + 1 BD / spot settl. date
R1 / R2 Original repo interest / repo interest on the close-out trade
MtMmargin = (N ∗Pmarket+ AIED1
100−
Ptrade+ AISSD100
− (R1 − R2)) ∗ OISdiscount_factor ∗ ps
│ 14
Fixed Income Risk EngineCore Engine: Mark-to-Market Margins
MtM Margins computation for forward starting repo trades3
Forward starting repo trades differ from repo trades as the spot leg has yet to settle
TRADE EXAMPLE
MARGINS CALCULATION DATE
A B
p + R1
bond
A B
bond
p
t
(evaluation date)
termspot
│ 15
Fixed Income Risk EngineCore Engine: Mark-to-Market Margins
Operations put in place
by the CCP in order to close the original positions
CLOSE-OUT PROCEDURE IN CASE OF DEFAULT OF CTP A
CLOSE-OUT PROCEDURE IN CASE OF DEFAULT OF CTP B
CCP
B
p + R1
bond
CCP
B
bond
p
t
(default A)
termspot
CCP
C
bond
p 2 + R2
CCP
C
p 2
bond
ACCP
p + R1
bondA
CCP
bond
p
t
(default B)
termspot
CCCP
bond
p 2 + R2C
CCP
p 2
bond
│ 16
Fixed Income Risk EngineCore Engine: Mark-to-Market Margins
The net outflow for the CCP in t amounts to the difference between the original settlement price (p) and the settlement price of the close-out trade
(p2) discounted from spot to t as well as the difference between p and p2 and R1 and R2 (repo interest of the original trade and repo interest of
the close-out trade) discounted from term to t. MtM margins for forward starting repo trades can therefore be defined as follows:
MTM MARGINS FORMULA FOR FWD STARTING REPO TRADES
N Nominal value of the original trade
P_market Price of the traded security at evaluation date
P_trade Original trading price of the security
AI_ED1 / AI_SSD Accrued interest of the security at evaluation date + 1 BD / spot settl. date
MtMmargin = ((N∗Pmarket+ AIED1
100−
Ptrade+ AISSD100
)∗(OISdf2−OISdf1)−(R1−R2)∗OISdf2)∗ps
R1 / R2 Original repo interest / repo interest on the close-out trade
OISdf1 / OISdf2 Discount factor from spot / term date to evaluation date
│ 17
Fixed Income Risk EngineCore Engine: Expected Shortfall – Cashflow Mapping
MARGINED PORTFOLIO
Clearing Member
Security 1
Security 2
Security n
CASHFLOWS
Security 1
Cashflow 1.1
Cashflow 2.1
Cashflow n.1
Security n
Cashflow 1.n
Cashflow 2.n
Cashflow n.n
…
CASH FLOW MAPPING
3M
6M
1Y
2Y
3Y
Cashflow 1.x
Cashflow 2.x
Cashflow n.x
Cashflows are mapped according to their
duration and to the statistical parameters of
the vertices of the respective contiguous
verticesDuration of the cashflow
Duration of the cashflow
Duration of the cashflow
│ 18
Fixed Income Risk EngineCore Engine: Expected Shortfall – Price Scenarios
Once each cashflow has been assigned to its respective vertex of the issuer curve, for each of those vertices a time series of
price scenarios is generated to be used for the Expected Shortfall computation. EWMA volatility can be used in order to produce
scaled price scenarios time series. Price scenarios are generated on the basis of the holding period (hp) and lookback period
parameters (lp) applied to the margining model:
Scenariot =Pricet
Pricet−hp
Scenariot > 0
Unscaled scenariosScaled scenarios
│ 19
Fixed Income Risk EngineCore Engine: Expected Shortfall – Price Scenarios (scaled)
SCALED PRICE SCENARIOS (hp = holding period, lp = lookback period)
Date 0Date 1
.
.
.Date n
Price 0Price 1
.
.
.Price n
Vertex x prices
time series of prices for vertex x Computation of relative returns based on chosen hp and lp
Return 1Return 2
.
.
.Return lp
Price 1 / Price 1 – hp - 1Price 2 / Price 2 – hp - 1
.
.
.Price n / price n – hp - 1
Vertex x returnsApplication of EWMA volatility
σi = λσi−12 + 1−λ ri
2
Application of scaling factor
σlp + σi
2σi
Scaled time series of price scenarios
For each return the EWMA volatility is computed and the respective scaled factor is defined. Scaled price scenarios
are then defined as follows:
Scaled Scenariot = Returnt ∗ Scaling factort + 1
│ 20
Fixed Income Risk EngineCore Engine: Expected Shortfall – Price Scenarios (scaled)
Through the application of EWMA volatility and the subsequently defined scaling factor different weights are applied to
different scenarios. In particular, the farther the distance in time from evaluation date the smaller the weight that will be
applied to that particular scenario:
σi = λσi−12 + 1−λ ri
2 = λ(λσi−22 + 1−λ ri−1
2 ) + 1−λ ri2
Since the smoothing factor (λ) is a number between 0 and 1:
ri−12 weight in i ri
2 weight in iλ(1- λ) 1- λ<
Recent returns have higher weights than older ones
│ 21
Fixed Income Risk EngineCore Engine: Expected Shortfall – Price Scenarios (unscaled)
Unscaled price scenarios time series are defined so that the same weight is assigned to every scenario regardless of their
distance in time from the evaluation date. Unscaled price scenarios can be thus retrieved directly from the time series of the prices
of the particular vertex:
UNSCALED PRICE SCENARIOS (hp = holding period, lp = lookback period)
Date 0Date 1
.
.
.Date n
Price 0Price 1
.
.
.Price n
Vertex x prices
time series of prices for vertex x
Computation of price scenarios based on chosen hp and lp
Scenario 1Scenario 2
.
.
.Scenario lp
Price 1 / Price 1 – hpPrice 2 / Price 2 – hp
.
.
.Price n / price n – hp
Vertex x scenarios
│ 22
Fixed Income Risk EngineCore Engine: Expected Shortfall
The sum of the cashflows mapped onto the appropriate vertices of the issuer curve are re-evaluated for each one of the computed
scenarios. The subsequantial distribution of gain / losses is drawn and the tails are identified. Based on the approach chosen (single
tail / double tail) the Expected Shortfall is the mean value of the gain / losses lying on the tail(s)
TotalCashflowrevaluated = TotalCashfloworiginal ∗(scenariot −1)
For each scenario t and each vertex of the curve
Distribution of gain / lossesAverage of worst losses
Average of greatest variations
SINGLE TAIL
DOUBLE TAIL
│ 23
Fixed Income Risk EngineCore Engine: Expected Shortfall – Spectral Risk Measures
The Expected Shortfall is computed by using unevenly weighted Spectral Risk Measures (SRM): increasing weights are assigned as the
losses increase
SRM are balanced in order to:
• Ensure the risk appetite of CC&G is satisfied as the losses move further into the distribution
• Correctly tackle pro-cyclicality concerns
│ 24
Fixed Income Risk EngineCore Engine: Overview of the process
CASHFLOWS (MARGINED PORTFOLIO)
Security 1
Cashflow 1.1
Cashflow 2.1
Cashflow n.1
Security n
Cashflow 1.n
Cashflow 2.n
Cashflow n.n
…
MA
PP
IN
G
MAPPED CASHFLOWS
IR NODE 1
IR NODE 2
IR NODE n
Total mapped flows on node 1
Total mapped flows on node 2
Total mapped flows on node n
SCENARIOS
MAPPED CASHFLOWS
IR NODE 1
IR NODE 2
IR NODE n
N° lp scenarios on node 1
N° lp scenarios on node 2
N° lp scenarios on node n
MAPPED CASHFLOWS
IR NODE 1
IR NODE 2
IR NODE n
Re-evaluated cashflows on node 1
Re-evaluated cashflows on node 2
Re-evaluated cashflows on node n
ES
1
2
3
Gains / losses of the entire mapped
portfolio
ES computation (with SRM)
│ 25
Fixed Income Risk EngineCore Engine: Parameters
MODEL PARAMETERS
Tail/s approach
Cross margining
Scaling factor
Holding period
Confidence level
Lookback period
Weighting
Single tail
Intra country
99.9%
5 days
From 99.5% to 99.8% (based on CR of issuer) for scaled approach. 99.5% for unscaled floor
Anchored to 2004
Spectral Risk Measures (step 1.35)
Add-ons Engine
│ 27
Fixed Income Risk EngineAdd-on Engine: Decorrelation Margin
Starting Framework
To tackle possible break-in-correlation among nodes of the same issuer Sovereign
Curve
Potential Extension
To tackle possible break-in-correlation also among
different issuers Sovereign Curves
extendable
Relevant based on current trades volumes (almost completely on the
same country – Italy)
Not relevant based on current trades volumes
Decorrelation Margin computation
Decorrelation Margin 20% ∗(UndiversifiedES −DiversifiedES)Compliant with ESMA art. 27
(portfolio margin)
│ 28
Fixed Income Risk EngineAdd-on Engine: Repo Term Concentration Margin
Closing out a repo/forward starting repo position implies performing (at least) one repo/forward starting repo
operation of the opposite sign of the original one
Hypothetical repo interest = EUR OIS curve + spread original repo rate vs EUR OIS curveMark-to-market margin
Repo-concentration margin add-on
• What-if: EUR OIS curve scaled/unscaled
• Parameter set function of
✓ Issuer Country
✓ Amount
✓ Maturity
in order to manage concentrated repo exposures
│ 29
Fixed Income Risk EngineAdd-on Engine: Repo Term Concentration Margin - Parameters
Risk measure Expected Shortfall(takes into account tail risk)
Tail approach Double tail
Confidence level 99.5%-99.8%(aligned to core country parameter)
Lookback period All available data
Tail weighting Tail weighting to contrast the dilution of tail due to the long time series with 1.35 SRM parameter aligned to core country parameter
Holding period Function of country and repo concentration(see next slides)
│ 30
Fixed Income Risk EngineAdd-on Engine: Repo Term Concentration Margin - Parameters
Repomaturity
bands
-7d
7d – 1m
1m – 3m
3m – 1y
1y+
Repoamountbands
-500€mln
500€mln – 1€bn
1€bn – 5€bn
5€bn+
Holding periods
5
6
7
8
9
10
Country matrix
Repomaturity
band
Repoamount
band
HP set
Repo maturity(1)
Repo amount(1)
HP set (1,1)
Repo maturity(1)
Repo amount(2)
HP set (1,2)
Repo maturity(1)
Repo amount(3)
HP set (1,3)
Repo maturity(1)
Repo amount(4)
HP set (1,4)
Repo maturity(2)
Repo amount(1)
HP set (2,1)
... ... ...
│ 31
Fixed Income Risk EngineAdd-on Engine: Repo Term Concentration Margin - Parameters
Country matrix
Repo maturity band Repo amount band HP set
-7d -500€mln - *
-7d 500€mln – 1€bn - *
-7d 1€bn – 5€bn - *
-7d 5€bn+ - *
7d – 1m -500€mln 5
7d – 1m 500€mln – 1€bn 5
7d – 1m 1€bn – 5€bn 5, 6
7d – 1m 5€bn+ 5, 6, 7
1m – 3m -500€mln 5, 6
1m – 3m 500€mln – 1€bn 5, 6
1m – 3m 1€bn – 5€bn 5, 6, 7
1m – 3m 5€bn+ 5, 6, 7, 8
3m – 1y -500€mln 5, 6, 7
3m – 1y 500€mln – 1€bn 5, 6, 7,8
3m – 1y 1€bn – 5€bn 5, 6, 7, 8
3m – 1y 5€bn+ 5, 6, 7, 8, 9
1y+ -500€mln 5, 6, 7, 8
1y+ 500€mln – 1€bn 5, 6, 7, 8,9
1y+ 1€bn – 5€bn 5, 6, 7, 8, 9
1y+ 5€bn+ 5, 6, 7, 8, 9, 10
│ 32
Fixed Income Risk EngineAdd-on Engine: Idiosyncratic-Concentration Margin
Sovereign curves risk (standard, liquid, on-the-run bonds)
• Simulated price returns of each ISIN are produced on the basis of the relevant benchmark curve
• A series of deltas among actually observed price returns and the simulated once are produced
• ES on the series of deltas is calculated and applied as add-on
• The parameter set used in ES calculation is (also) function of the ISIN concentration (vs its
outstanding) in order to take into account the potential risk associated to particularly concentrated
positions
Core Model
Idiosyncratic-Concentration add-on
Specific risk of a bond type (in particular, linkers and floaters) and/or
the specific liquidity of a particular issue
│ 33
Fixed Income Risk EngineAdd-on Engine: Idiosyncratic-Concentration Margin - Parameters
Risk measure Expected Shortfall(takes into account tail risk)
Tail approach Double tail
Confidence level 99.6%-99.7%(aligned to core country parameter)
Lookback period 1y (allows to tackle most part of ISINs without recurring to approximations)
Multiplier 25%(in order to compensate for potential absence of stress events in the LP and for regulatory-compliant anti-procyclicality purposes)
Holding period Function of bond issuer/type and Clearing Memberconcentration(see next slide)
│ 34
Fixed Income Risk EngineAdd-on Engine: Idiosyncratic-Concentration Margin - Parameters
Country matrix
Bond type Concentration HP set
Bullet/Zc -5% -
Bullet/Zc 5% - 10% -
Bullet/Zc 10% - 15% 5, 6, 7
Bullet/Zc 15% - 20% 5, 6, 7, 8
Bullet/Zc 20% - 25% 5, 6, 7, 8, 9
Bullet/Zc 25%- 5, 6, 7, 8, 9, 10
Floater -5% 5
Floater 5% - 10% 5, 6
Floater 10% - 15% 5, 6, 7
Floater 15% - 20% 5, 6, 7, 8
Floater 20% - 25% 5, 6, 7, 8, 9
Floater 25%- 5, 6, 7, 8, 9, 10
Linker -5% 5
Linker 5% - 10% 5, 6
Linker 10% - 15% 5, 6, 7
Linker 15% - 20% 5, 6, 7, 8
Linker 20% - 25% 5, 6, 7, 8, 9
Linker 25%- 5, 6, 7, 8, 9, 10
│ 35
Concentration Component SubsetExceptions to Regular
Framework
Repo-Term Italian DebtNo stressed hp as a function of
concentration is used
ISIN-Concentration Italian DebtNo stressed hp as a function of
concentration is used
CCP
Overall goal of ensuringsystemic stability
Strategic Participants (Bankit and MEF) may be exemptedfor the application of Concentration add-ons on Italian debt
Fixed Income Risk EngineAdd-on Engine: Concentration Margin - Exceptions
│ 36
Fixed Income Risk EngineRisk Framework Map
│ 37
Fixed Income Risk EngineAdd-on Engine: Settlement Margin
• Margins are computed and collected at discrete points in time, while trading and settlement are almost
continuous throughout the day
• The settlement of cash and repo positions in t+1 may disrupt possible long-short position offsettings,
increasing the risk exposure of the CCP
Positions in portfolio at the moment of calculationCore Model
Settlement Margin add-on (SS1)
Max among:
• Current portfolio (intraday/end of day)
• Current portfolio assuming all t+1 trades have settled
Data
│ 39
Fixed Income Risk EngineFIRE vs MVP: Margins Comparison
6.00
8.00
10.00
12.00
14.00
16.00
18.00
24/05/2018 24/07/2018 24/09/2018 24/11/2018 24/01/2019 24/03/2019 24/05/2019 24/07/2019 24/09/2019 24/11/2019 24/01/2020 24/03/2020 24/05/2020
Tota
l M
arg
in r
eq
uir
em
ent
(bln
€)
New FI Model (Core)
New FI Model (Core + Addons)
Current Model
│ 40
Fixed Income Risk EngineFIRE vs MVP: Margins Comparison – Focus on March 2020
6.00
8.00
10.00
12.00
14.00
16.00
18.00
02/01/2020 02/03/2020 02/05/2020
Ma
rgin
re
quir
em
ent
(bln
€
)
New FI Model (Core)
Current Model
New FI Model (Core + Addons)
│ 41
Fixed Income Risk EngineFIRE vs MVP: Margins Comparison – Delta
-40.00%
-30.00%
-20.00%
-10.00%
0.00%
10.00%
20.00%
30.00%
40.00%
24/05/2018 24/07/2018 24/09/2018 24/11/2018 24/01/2019 24/03/2019 24/05/2019 24/07/2019 24/09/2019 24/11/2019 24/01/2020 24/03/2020 24/05/2020
% c
ha
ng
e v
s C
urr
ent
Mo
de
lNew FI Model (Core) vs Current Model
New FI Model (Core + Addons) vs Current Model
│ 42
Fixed Income Risk EngineFIRE vs MVP: Portfolio Backtest
│ 43
Fixed Income Risk EngineFIRE vs MVP: Portfolio Backtest – Focus on May 2018
│ 44
Fixed Income Risk EngineFIRE vs MVP: Portfolio Backtest – Focus on March 2020
│ 46
This publication is for information purposes only and is not a recommendation to engage in investment activities. This publication is provided “as is” without representation or
warranty of any kind. Whilst all reasonable care has been taken to ensure the accuracy of the content, Euronext does not guarantee its accuracy or completeness. Euronext will not
be held liable for any loss or damages of any nature ensuing from using, trusting or acting on information provided. No information set out or referred to in this publication shall
form the basis of any contract. The creation of rights and obligations in respect of financial products that are traded on the exchanges operated by Euronext’s subsidiaries shall
depend solely on the applicable rules of the market operator. All proprietary rights and interest in or connected with this publication shall vest in Euronext. No part of it may be
redistributed or reproduced in any form without the prior written permission of Euronext. Euronext refers to Euronext N.V. and its affiliates. Information regarding trademarks and
intellectual property rights of Euronext is located at www.euronext.com/terms-use
© 2021, Euronext N.V. - All rights reserved.