Regulatory Frameworks
Table of Contents
1. Summary
Several regulatory frameworks compute a capital or margin number from a portfolio, and each prescribes its own vocabulary for doing so. That vocabulary reaches the model directly: the classifications analysed in asset classes and the sensitivity names analysed in the greeks are largely inherited from these frameworks rather than invented.
The frameworks are prescriptive rather than descriptive. Their enumerations exist to index published tables of risk weights and correlations, so a framework's classification is fixed by the methodology document and cannot be adjusted for modelling convenience.
| Framework | Computes | Imposes |
|---|---|---|
| ISDA SIMM | Initial margin on non-cleared derivatives | Risk classes, product classes, risk types |
| FRTB, standardised | Market risk capital | Risk classes, and delta, vega and curvature explicitly |
| FRTB, internal models | Market risk capital | Expected shortfall, liquidity horizons, desk-level approval |
| SA-CCR | Counterparty credit risk exposure | Asset classes and hedging sets |
| BA-CVA, SA-CVA | Capital for credit valuation adjustment risk | Risk factors and buckets |
2. CRIF
The Common Risk Interchange Format is the record format these frameworks exchange sensitivities in. ORE documents it precisely: "A CRIF record is a row of the CRIF file outlined in the document: ISDA SIMM Methodology, Risk Data Standards. Version 1.36: 1 February 2017."
One CRIF row carries a trade id, a trade type, netting set details, a
product class, a risk type and an end date, together with
framework-specific fields. The FRTB fields are creditQuality,
longShortInd, coveredBondInd, trancheThickness and bb_rw.
CRIF matters to the model beyond its role as a file format, because it shows how the frameworks themselves separate classification from qualification. The risk type names a class and a sensitivity; the additional fields carry the properties. That separation is the basis of the argument in risk classes that a framework's apparent extra classes are properties.
A CRIF record declares which framework produced it. ORE holds
RecordType { SIMM, FRTB, SACCR, Generic } for that purpose, and
CapitalModel { Empty, SACCR, SACVA, FRTB } for the capital model a
row belongs to.
3. ISDA SIMM
SIMM is ISDA's Standard Initial Margin Model. The ORE user guide names it in full and lists it as an analytic added in ORE v10, citing the "ISDA SIMM Methodology, version 2.5A".
It computes initial margin on non-cleared over-the-counter derivatives. Its purpose is agreement rather than accuracy alone: a single published methodology means two counterparties to a trade compute the same margin figure, which removes a class of dispute that independent models would create.
The calculation is sensitivity-based. Each party produces sensitivities in CRIF form, and the model applies published risk weights and correlations to them.
3.1. What SIMM imposes
ProductClass carries RatesFX, Credit, Equity and Commodity
as substantive values, with Rates and FX separately for the IM
Schedule extension.
RiskClass carries InterestRate, CreditQualifying,
CreditNonQualifying, Equity, Commodity and FX.
RiskType is the finest of the three, crossing a class with a greek
and, on credit alone, a qualifying flag.
The ordering of RiskClass is load-bearing. Its declaration warns
that "the ordering here matters. It is used in indexing in to
correlation matrices for the correlation between risk classes", which
is why the enumeration behaves as an index rather than as a taxonomy.
The IM Schedule is the alternative to the sensitivity-based
calculation, computing margin from notional and maturity instead. It
is why ProductClass carries Rates and FX separately and why
RiskType carries PV.
4. FRTB
FRTB is the Basel Fundamental Review of the Trading Book, the market risk capital framework.
4.1. The two approaches
FRTB is not one calculation. It offers two, and the difference decides what a bank must be able to produce.
| Approach | Who prescribes the model | What the bank must supply |
|---|---|---|
| Standardised (SA) | The regulator, entirely | Sensitivities, in the prescribed classes |
| Internal models (IMA) | The bank, subject to approval | An approved model, and the evidence it stays approved |
Under the standardised approach the methodology is fixed: positions are mapped to prescribed risk classes, sensitivities are weighted with published risk weights and aggregated with published correlations. Under the internal models approach the bank uses its own model, and what the regulator prescribes instead is the set of tests the model and the desk must keep passing.
The sections below describe the standardised approach, which is what reaches this corpus: it is the approach that consumes sensitivities in the prescribed buckets, and so the approach whose vocabulary the model must be able to produce.
4.1.1. What the internal models approach adds
Three of its features matter here even though the approach itself does not.
The measure changes. Value at risk is replaced by expected shortfall, computed over a set of liquidity horizons rather than a single holding period, so a risk factor's assumed time to exit becomes a parameter of the capital number.
Risk factors are partitioned by whether they can be modelled at all. A factor passing the risk factor eligibility test is modellable and enters the expected shortfall calculation; one failing it is non-modellable and is capitalised through a stress scenario instead. That partition is a property of observed market data rather than of the trade, which is why it belongs with market data requirements rather than with any trade classification.
Approval is granted per trading desk, not per bank. A desk that fails its profit and loss attribution test or its backtesting falls back to the standardised approach while the rest of the bank continues under its models. The consequence for the model is that a desk is a regulatory unit and not merely an organisational one: capital is computed, attributed and lost at that grain, so a desk identifier must be carried on a position rather than inferred from a book hierarchy.
4.2. The sensitivities-based method
Its six risk classes are: general interest
rate risk (GIRR), credit spread risk (CSR), equity (EQ),
commodity (COMM) and FX, with credit spread divided three ways by
securitisation status.
FRTB names its greeks explicitly rather than by suffix, so its risk
types read GIRR_DELTA, GIRR_VEGA, GIRR_CURV and so on. Curvature
requires two scenarios rather than one number, which ORE holds as
CurvatureScenario { Empty, Up, Down }.
Two further charges sit outside the sensitivities-based method, remain
within the standardised approach, and are not sensitivities at all. The default risk charge (DRC) is split the
same three ways as credit spread. The residual risk add-on (RRAO)
applies at one per cent or a tenth of a per cent.
5. SA-CCR
SA-CCR is the standardised approach for counterparty credit risk, which computes exposure at default for derivative positions rather than a margin figure.
It classifies positions into asset classes and then into hedging sets
within them, which is a third way of grouping alongside SIMM's product
classes and FRTB's risk classes. ORE carries SA-CCR risk types in the
same enumeration as the other two, and RiskType marks FX and PV
as shared with SIMM and the IM Schedule respectively.
6. The CVA approaches
Credit valuation adjustment risk has two standardised treatments,
which ORE lists among its recent analytics additions: the basic
approach (BA-CVA) and the standardised approach (SA-CVA). ORE's
CapitalModel enumeration carries SACVA alongside SACCR and
FRTB.
7. Reporting regimes
A second group of regimes shapes identity rather than risk. They prescribe which identifiers a trade and a security must carry, and appear in this corpus for that reason rather than for any classification they impose.
| Regime | Requires |
|---|---|
| EMIR | Trade reporting to a repository, using the UTI |
| Dodd-Frank, through the CFTC | Swap reporting, originally using the USI |
| MiFID II | Transaction reporting, using the ISIN |
Trade identifiers covers UTI and USI; standard financial identifiers covers ISIN and its siblings.
8. What ORE implements
The ORE user guide records SIMM as an analytic added in v10, and lists regulatory capital for counterparty credit risk and CVA risk, covering SA-CCR, BA-CVA and SA-CVA, among its recent additions. It also lists dynamic SIMM based on path-wise sensitivities computed with algorithmic differentiation.
FRTB appears in the CRIF record types and risk types rather than in the user guide's analytics list.
9. Open questions
- Whether ORE Studio needs to hold any of these classifications directly, or only to map onto them when producing a CRIF file. The second is sufficient for reporting and avoids adopting a vocabulary that changes when a methodology version changes.
- How a methodology version is recorded. SIMM's risk weights and correlations change between versions, so a margin figure is meaningless without the version that produced it, and nothing in this corpus records one.
- Whether the reporting regimes belong on this page at all. They impose identity requirements rather than risk vocabulary, and share little with the capital frameworks beyond being regulatory.
10. Sources
ORE's enumerations and the CRIF description are read from
OREAnalytics/orea/simm/crifrecord.hpp and
OREAnalytics/orea/simm/simmconfiguration.hpp. The CRIF definition is
quoted from the doc comment there, which cites the ISDA SIMM
Methodology Risk Data Standards, version 1.36 of 1 February 2017.
The analytics list and the expansion of SIMM are read from
Docs/UserGuide/userguide_intro.tex, and the SIMM methodology version
from Docs/UserGuide/references.tex.
The two FRTB approaches, and the internal models features named above, are as described in Basel's Minimum capital requirements for market risk, the standard the framework is published as. No chapter is cited and the text was not read for this page, so the account is a summary from general knowledge rather than a verified reading; the division into a standardised and an internal models approach is not in doubt, but the detail should be checked against the standard before being relied on. The gap was pointed out by a LinkedIn post from QFI Bootcamps on FRTB SA versus IMA, which is a promotional article and is recorded here as the prompt rather than as an authority.
The regulatory background — what each framework is for, which rules drive it, and how the reporting regimes relate — is not cited. ISDA's published FAQ was unreachable and no encyclopaedia article was available, so those statements rest on general knowledge rather than on a verified source and should be checked before being relied on.
11. See also
- Regulatory Risk Buckets — the aggregation buckets these frameworks weight and correlate against.
- Asset Classes — the taxonomy these frameworks each cut differently.
- Risk Classes — what a product is sensitive to.
- The Greeks — the sensitivity names these frameworks prescribe.
- Trade identifiers — UTI and USI, required by the reporting regimes.
- Knowledge — the hub of all knowledge notes.