«Revision history Date Version Description Author 2015-12-14 1.0 Initial draft Bengt Jansson 2015-12-21 1.1 Data added Bengt Jansson 2015-12-30 1.2 ...»
Validation of Haircut Model
A validation of the Haircut Model used by Nasdaq OMX
Author Last update Version Page
2015-12-30 1.3 2 (18)
Bengt Jansson, zeb/ Risk & Compliance Partner AB
Date Version Description Author
2015-12-14 1.0 Initial draft Bengt Jansson
2015-12-21 1.1 Data added Bengt Jansson 2015-12-30 1.2 Final version Bengt Jansson 2015-12-30 1.3 Final after input Bengt Jansson Validation of Haircut model, 2014 Author Last update Version Page 2015-12-30 1.3 3 (18) Bengt Jansson, zeb/ Risk & Compliance Partner AB Table of contents
1.2 Legal environment
2. Input to the validation
2.1 Documentation at NASDAQ Clearing
2.2 Numerical analysis of Haircut Methodology
2.4 Special issues
3. Theoretical framework of the model
3.1 Background on Haircut and risk
3.2 Basic Haircut model calculations
3.3 Credit adjustments
4. Monitoring process
5. Numerical investigation
6.1 Changes from previous validation
6.2 Input to the validation
6.3 Theoretical framework of the model
6.4 Monitoring process
7.1 Appendix 1 Definitions
7.2 Appendix 2 On-demand commercial bank guarantees
7.3 Appendix 3 Collateral limits
Validation of Haircut Model
1. Background 1.1 General NASDAQ OMX Clearing AB (“NASDAQ Clearing”) provides clearing and Central Counterparty (“CCP”) services. In order to prudently manage these services NASDAQ Clearing uses a large number of different models. This report is the validation of the Haircut Model used for establishing accurate value of pledged collateral.
The purpose of a validation of models is to ensure the theoretical and empirical soundness of the models used by the CCP. The validation report should ensure transparency on the models
used by the CCP for the benefit of:
Board of Directors, NASDAQ OMX Clearing AB.
Competent Authorities Internal Audit and Audit Committee Other stake holders 1.2 Legal environment NASDAQ Clearing was at the 19th of March 2014 authorised as Central Counterpart (CCP) to offer services and activities in the Union in accordance with Regulation (EU) No 648/2012 of the European Parliament and of the Council of 4 July 2012 on OTC derivatives, central counterparties and trade repositories1.
The legal framework that governs NASDAQ Clearing is therefore the EMIR framework, Regulation (EU) No 648/20122 and supporting delegated Regulations 148/2013, 149/2013, 150/2013, 151/2013, 152/2013, 153/2013, 285/2014, 667/2014, 876/2013, 1003/2013 and the implementing Regulations 484/2014, 1247/2012, 1249/2012.
Also included is the Q&A document as per 1st October 2015.
The Regulation of particular interest for this validation is Delegated Regulation No 153/2013 “supplementing Regulation (EU) No 648/2012 of the European Parliament and of the Council with regard to regulatory technical standards on requirements for central counterparties”.
2. Input to the validation2.1 Documentation at NASDAQ Clearing
2.1.1 Previous validation of Haircut Methodology In NASDAQ Clearing’s application for being a authorised as a CCP and to offer services and activities in the Union in accordance with Regulation (EU) No 648/2012 a validation of the Haircut model was amended. All the validations of the SPAN model is important building blocks to be able to follow the development of the model. In most cases the previous 1 Usually referred to as “EMIR” 2 With changes included up to 30th November 2015 (last effective change 9th September 2015)
validation is most important but in some cases the other ones can be referred to. The full list
of validations is as follows:
“Validation of Haircut Meth. ver 1.3, 2013”, (Validation of Haircut, NOMX, 2013) “Validation of Haircut Meth. ver 1.3, 2014”, (Validation of Haircut, NOMX, 2014) 2.1.2 Clearing risk mandate This is the Risk Tolerance framework of Nasdaq Clearing. It constitutes the risk appetite basis
that controls how collate should be handled on a very top level. The full document name is:
“Clearing Risk Mandate, NASDAQ OMX Clearing AB, 150923”, (Clearing Risk Mandate, NOMX, 2015).
2.1.3 Instructions for haircut calculations This is an instruction that describes how collateral values should be calculated from mathematical point of view. The full document name is: “Instructions for haircut calculations, NASDAQ OMX Clearing AB, 151126”, (Instruction for Haircut calculation, NOMX, 2015).
2.1.4 Policies A lot of policies will be included as input to this validation. The following list will name the
most prominent policies in this aspect:
Policy for the Validation of Models: The starting point for the construction, reporting set up and the content of the validation is the policy for validating model that is
approved by the Board of Directors at NASDAQ Clearing. The full document name is:
“Model Validation Policy NOMX (150923)”, (Model Validation Policy, 2015).
Policy for setting margin parameters: NASDAQ Clearing has developed policies that
regulate how risk parameters should be estimated. The full document name is:
“Margin Parameter Policy NOMX (150826)”, (Margin Parameter Policy, 2015).
Policy for back testing of models: NASDAQ Clearing has developed policies that regulate how back testing should be conducted from a theoretical, and very general, point of view. More specific guidelines can be found for specific models. The full document name is: “Back testing Policy NOMX (150823)”, (Back testing Policy NOMX, 2015).
Policy for stress testing of models: NASDAQ Clearing has developed policies that regulate how stress testing should be from a theoretical, and very general, point of view. More specific guidelines can be found for specific models. The full document name is: “Stress Test Policy NOMX (150810)”, (Stress Test Policy NOMX, 2015) Policy for sensitivity testing of models: Nasdaq OMX has developed policies that regulate how sensitivity testing should be from a theoretical, and very general, point of view. More specific guidelines can be found for specific models. The full document name is: “Sensitivity testing and analysis Policy (130909)”,(Sensitivity testing and analysis Policy, 2015) Policy for Credit Risk: Nasdaq OMX have a policy regarding how credit risks should be handled within the company. One of the credit risks comes from the way collateral is treated and therefore this policy is interesting in part for this validation. The full document name is: “Credit Risk Policy, 20150820”, (Credit Risk Policy NOMX, 2015)
2.2 Numerical analysis of Haircut Methodology NASDAQ Clearing has ongoing numerical procedures as place to deliver numerical output from each margin model that its use. The numerical data can be roughly divided into three
Back testing data Stress testing data Sensitivity analysis data 2.3 Discussions In any validation a large part of the information received must be thoroughly discussed with the personal at the CCP. The following persons are however prime sources of information to
this model validation:
Henrik Rosén, Risk Management department of NASDAQ Clearing Karl Klasén, Risk Management department of NASDAQ Clearing 2.4 Special issues Since a models used by a CCP must, and should, be validated on a yearly basis each
validation will be updated with new issues as:
New, or changed, functionality in the methodology Changed financial environment, e.g. different volatility in the market New legislation that changes the rules thereby contradicts assumptions made in the model A section in the validation will specifically target differences between validations to facilitate reading.
3. Theoretical framework of the model3.1 Background on Haircut and risk
3.1.1 Correlation 184.108.40.206 General There are two ways of viewing collateral and the risk it is supposed to limit or mitigate. The first way is to include collateral and the derivatives3 in the calculations at the same time thus taking a simultaneous view on the risk. This includes as an example covered call strategies.
The second way is to totally separate derivatives and collateral. In this case no correlation effects are given between derivatives and collateral based on the same underlying, or highly correlated underlyings.
220.127.116.11 Integrated view on collateral and derivatives In many ways this is a more advanced and efficient view on the counterpart risk within a CCP. It is of course so that from all practical reasons the CCP has one exposure towards its counterpart. This exposure includes both derivatives and collateral at the same time. It would 3 It is of course so that collateral can cover other risk than emancipating from derivatives but in this document the phrase derivatives will be used WLOG.
as an example be very illogical to treat the unwinding of risk and positions for collateral and derivatives as two separate process in the CCP thus ignoring any correlating and hedging effects between them.
With this view in mind the parameters applied for derivatives should be the same as for collateral. This would also indicate that both collateral and derivatives should be treated as one portfolio from the viewpoint of the CCP. From the point of margin calculations this would be done on derivatives, together with collaterals, and requires that the margin calculated must be “positive” i.e. the portfolio must show a positive market value after the calculations. This indicates that the collateral covers the derivatives positions.
Three are some practical issues that must be dealt with before this can happen and that is ensuring that the margin model can accommodate the collateral within its algorithm. It must also be perfectly clear from legal point of view that derivatives and collateral can be viewed as one portfolio.
18.104.22.168 Collateral and derivatives totally separated This is the “traditional” way of handling derivatives and the corresponding collateral. From historical reasons the collateral and derivatives was kept in separate systems, there was (and to some extent still are) different legal status on the two groups that prevents a unified approach.
In this set up the collateral parameters are kept separate from margin parameters. The impact is of course that when margins are challenged in back testing and stress testing the value of the collateral after haircut is kept constant. This implies a need for haircut parameters to be set more conservative than margin parameters.
This is the method that NASDAQ Clearing uses for derivatives and collateral. This also the reason to why the haircut parameters are much higher than the corresponding4 margin parameter.
3.1.2 Time frame VaR models and margin models do try to estimate the development of derivatives (and other financial assets) for a relatively short time called ‘liquidation period5’ for CCPs. This means that VaR and margin parameters also tend to describe the development for a short period of time.
A traditional collateral model that is not part of the portfolio from risk calculation point of view must be more conservative since the collaterals, in case of a default, would be used as a buffer and not part of the calculations. For a collateral model the “liquidation period” should therefore be longer or at least the same as used margin model.
This also implies that collateral models should use a longer time span between recalculations to be more stable and not unnecessarily disturb the default process.
It is also the case that collateral haircuts must be easy to handle from an operational point of view both internally at NASDAQ Clearing but also externally at members. This also implies that changes to the haircut parameters should be done more seldom than for margin parameters.
4 Given that the same underlying instrument to the derivative are used as collateral.
5 See “Appendix I Definitions” for definition of liquidation period
Moreover, as mentioned above, depending on the setup of stress testing in the CCP, collateral haircuts may have to be estimated to withstand extreme but plausible market conditions. This typically leads to longer liquidation period and higher confidence levels than used in the margin calculations.
3.1.3 Model features So for a traditional collateral model with recalculations made with longer intervals the basic
features can be summarized as this:
No correlation between collateral and derivatives No correlation between the different collateral instruments The actual haircut calculation model should be easy to implement and communicate to stake holders The size of the haircuts should be significant to, within reasonable limits, guarantee the eligible value of the collateral portfolio It will be investigated whether NASDAQ Clearings collateral model do have these features that are required from the current usage of the model.
Do notice that the above list of features for a collateral model is dependent on the usage of the model. If collateral would be included in the risk calculations (margin calculations, stress test calculations etc.) then the requirements would change completely. Using the same model for estimating margin parameters and collateral haircuts would require extensive analysis and consideration before implementing.
3.2 Basic Haircut model calculations 3.2.1 General A haircut model is in reality several models6 and each underlying collateral type will be investigated separately.