Model

Management Guidance
01. Introduction
The Central Bank of United Arab Emirates is that the models described in MMG all need
(‘CBUAE’) has published two documents that to follow the principles set out in MMS. The
cover the model risk management principles. MMG covers six model types and provides
The first is the Model Management Standards guidance on how to develop and validate these
(‘MMS’), which covers the model lifecycle models. Even models not covered in MMG are
framework that applies to models. This subject to principles of the MMS (especially
was covered in our Blog 1 (October 2022). data quality, developmental rigour and the
The second one is the Model Management ability to justify modelling assumptions).
Guidance (‘MMG’), which would be covered
in this blog.

The main bridge between the MMS and MMG

2
02. Summary of common themes in MMG
The key themes that carry across from MMS and MMG), we feel that the role of the Model
to MMG are: the need and requirement for Oversight Committee is key to successfully
quality data that will be used to build models, implementing the Model Management
explainability of the modelling approach Framework. Its members are likely to need
used, independent validation and rigorous in-depth knowledge and experience of a wide
governance at each stage of the lifecycle. range of models and banking applications. A
Whereas MMG is relatively prescriptive seat on the committee is a crucial role within
(whilst MMS is generally principles based) the institution and requires significant time
when describing the specific model examples, commitment.
it does allow institutions to deviate from the
guidance if the methodologies utilised can be A key challenge in the development of
appropriately justified. the Model Management Framework and
developments within the banking industry
The main links or common themes between generated through the expanded use of models
MMS and MMG, and indeed across the model is the interconnected nature of those models
examples mentioned in MMG, are: that now require developers, validators and
those responsible for oversight to have a
• The management of data (all models broader range of modelling risks and to be
require that appropriate and good data exists much more holistic in their approach to the
so that the models represent an accepted construction of business solutions.
view of reality);
For example, a credit risk modeller now
• The desire to have all models needs to consider macro-economic effects to
independently validated by teams that have a much greater degree than in the past. The
the skills (and experience) to build the models key challenge and risk to banking institutions
if required; and, across the UAE and indeed globally will be to
attract the skills (or appointing appropriate
• The need or requirement for robust third parties) necessary (and in appropriate
governance that ensures that the models number, given the need for development,
are fit for purpose (this will include sense independent validation and governance)
checking the models, do they capture the to manage their models and the associated
business essentials, etc.). model (as required by MMS and MMG).

Given the touchpoints that the governance

requirement has with each and every stage of
the model lifecycle (highlighted in both MMS

3
03. Practical Requirements and Challenge
Before we delve into the details of MMG, it some banks may face. We will cover the
is worth highlighting what the regulator mitigation of these challenges in our
expectations are and the typical challenges upcoming webinar on this topic.

Expectation of the CBUAE

All institutions are expected to identify gaps
between their practice and the MMS and MMG
and, if necessary, establish a remediation
plan to reach compliance.

The outcome of this self-assessment and the

plan to meet the requirements of the MMS
and the MMG must be submitted to the CBUAE
no later than 6 months ( June 2023) from the
effective date of the MMS.

Banks must demonstrate continuous

improvements towards meeting these
requirements within a reasonable timeframe
depending on the complexity and the systemic complexity while maintaining their quality.
risk of each institution. This timeframe will be
approved by the CBUAE following the review Full compliance is expected from institutions
of the self-assessment. The remediation with respect to the general principles
plan and the associated timing must be described in Part I and Part II of the MMS. For
detailed, transparent, and justified. The plan the MMG, whilst alternative approaches can
must address each gap at a suitable level of be considered, the focus is on the rationale
granularity. and the thought process behind modelling
choices. Institutions should avoid material
Potential consequences of non-compliance inconsistencies, cherry-picking, reverse-
In the event that an institution (regardless engineering and positive bias, i.e. modelling
of its size) is unable to comply with the MMS approaches that deliberately favour a desired
and the MMG, it must implement a remedial outcome. Evidence of an institution defying
process. This may involve reducing the the general principles or abusing the MMS in
number and/or complexity of its models in this way will warrant a supervisory response
order to improve the quality of the remaining ranging from in-depth scrutiny to formal
models. Subsequently, the institution could enforcement action.
increase the number of models and/or their

4
Institutions that repeatedly fall short of the firms will be the cultural change, the
requirements and/or do not demonstrate change in thinking, i.e. thinking about
continuous improvements will face greater the portfolio of models as an analytics
scrutiny and could be subject to formal capability that needs to deliver service at a
enforcement action by the CBUAE. In certain quality and how model risk, if not
particular, continuously and structurally managed, can lead to significant losses
deficient models must be replaced and should b. Ongoing training of modelling teams is
no longer be used for decision making and another area that will require attention
reporting. c. The head of the model risk team will need
to have the right seniority and gravitas
Typical challenges
Based on our interaction with the market, 4. Implementation
firms are asking questions of the following
a. Set up specific Model Risk functions that
type as they begin their journey towards
need to approve independently validated
compliance with the MMS and MMG.
models prior to the committee stages of
the process. The new function will need
1. Scope and interpretation
to be resourced with skilled individuals,
a. Does the regulation apply to all models? experienced in the nature of a diverse
b. Can we phase out the implementation? array of models / business problems.
That is start with the most material Given that many UAE institutions will be
models (e.g. IFRS 9) and then progressively in the same boat finding the appropriate
incorporate other models? resources may be a challenge to some or
c. How to interpret certain aspects of the all institutions
regulation for trading book or AI models, b. Ensuring that the business-as-usual
e.g. performance monitoring processes are unaffected by the set-up
process (essentially the bank’s governance
2. Gap analysis and roadmap structures may need to be reinvented)
c. The selection of specialist third party firms
a. How much detail is required to assess the
for the construction of specific models.
current state?
The use of third-party specialists also
b. A target state needs to be articulated before
helps with upskilling internal development
the gap between the target and current
and validation teams with the latest
state can be assessed. How to describe a
techniques and industry thinking that will
target state of the modelling landscape at
aid the bank in future developments and
one’s firm over the next few years?
enhancements of the models

3. Operating model changes

a. Perhaps the biggest challenge for some

5
04.
Models Covered in
MMG
a. Rating Models

b. PD Models

c. LGD Models

d. Macroeconomic Models

e. Interest Rate Risk in the Banking Book

f. Net Present Value Models

6
a. Rating Models Governance & Strategy – The management of
the bank’s ratings models need to follow the
model lifecycle determined in the MMS with
Key Challenges models ideally based on internally collected
and stored historical data & utilise justifiable
The rating models are often the base of development, validation and monitoring
many credit risk applications (such as risk methodologies.
management, provisions, pricing, collections,
capital allocation and IFRS9) and cover retail Data Collection & Analysis – It is encouraged
and corporate risk assessment models. that the bank collects their own Ratings
Therefore, a poorly developed or managed Model development data (with the collection,
model will have effects that propagate across cleansing and manipulation processes fully
many decision areas of the bank. documented and approved). Data utilised
for modelling should ideally be at obligor
The development of such models is a well- and facility level and have sufficient volume
documented path therefore the MMG to be statistically valid. Low default volume
only suggests the following minimum techniques can be employed, where necessary,
requirements: but should be fully justified.

Segmentation – Portfolio
segmentation needs to contain
statistically homogeneous
groups of obligors, whilst
being heterogeneous to other
neighbouring segments and
generally are split by product,
customer type and difference
in historic default or credit
performance.

Default Definition – Banks

should develop and document
two default definitions, an
operational definition used
for business strategy decisions
and a second one used for the
estimation and calibration
of default probabilities
(used within regulatory
environment). Whereas these

7
definitions can be the same, the operational
definition is usually tighter than the regulatory
scenario (e.g. 60 dpd as opposed to 90 dpd).
Appropriate levels of conservatism should be
built into the definitions.

Default Estimation – Prior to modelling

a detailed understanding of the portfolios
historic default performance is required.
The historic analysis that feeds into the
understanding should cover a full economic
cycle.

Rating Scales - The Ratings Scales help banks

to map and understand the risks associated
to individual portfolios across the diverse
range of products and segments employed
by the bank. The overall scale should ensure
appropriate levels of granularity whilst
enabling robust estimates of PD within each
scale grade. External rating can be used as a
benchmark against which the internal grades
can be compared. (up or down) but must be documented with
clear reference to the credit approval &
Model Construction and Use – Whereas Retail governance structures of the bank.
models utilise standardised development
methodologies, corporate may Monitoring and Validation – Monitoring of
models
require more bespoke methods, due to the models and grades needs to be carried
portfolio complexity and (low) volumes. out on a regular basis with independent
Models can utilise both quantitative and validation carried out to assess the continued
qualitative characteristics and have the validity of the modelling assumptions, the
appropriate levels of statistical analysis characteristics within the model and the
(where possible). All methodologies need to be data used for development and monitoring.
comprehensively documented and justified Like the development, the validation needs
against any identified alternative approaches, to be approved by the appropriate oversight
independently validated and approved by the committees.
relevant committee.

Overrides – Ratings overrides are permitted

8
 b. PD Models point in the cycle the model construction
takes place, which may be difficult to estimate
(particularly if the time between peak and
Key Challenges
trough is long) making the use of scalars
difficult.
TTC models require an estimate that covers
a full economic cycle, which may be lengthy,
The TTC and PiT issues would apply to both
particularly in resource-led economies, and
PD and LGD.
many institutions tend not to hold data across
such long time periods.
The following table gives a range of definitions
that will be used in the PD section:
The PiT model needs to estimate at what

Metric Definition

PD Probability of Default

LGD Loss Given Default

PD12 The PD over a specific time horizon, usually 12 months or one year

PD Term Structure Cumulative PD over a multi-year time horizon

PD TTC PD across the whole economic cycle

Forward 1-year PD Future PD12 at some point in the future

Transition Matrix Probability of Moving from one Rating to another within a 12-month time horizon (or across
several years)

Default Rate Time Series – Similar to rating calibration needs to use a minimum of 5 years
models sufficient default history must be used data for wholesale portfolios.
to develop the PD models, with stable and
homogeneous PDs across the time horizons (a PiT PD and Term Structures – Modelling
minimum of 5 years should be used) approved decisions around PiT PD and Term
by the model oversight committee. Structures will have a material impact on
the provisions and associated management
Ratings to PD – A common technique for actions with methodology decisions being
PD estimation is the mapping of the Ratings based on desired granularity, time steps used
Grades to the Through the Cycle (TTC) PDs. and segmentation employed. From a MMG
However, the sensitivity of the PDs and grades perspective one of the following approaches
to the economic cycle need to be considered, should be used, Transition Matrices, Portfolio
indicating that Point in Time (PiT) models are Averaging / Scaling or the Vasicek framework.
developed and then calibrated to TTC. The

9
PiT PD with Transition Matrices – Transition calibrate (in particular correlation of asset
Matrices are a convenient tool with limitations value and risk factors and their interaction
but practical advantages. The modelling with macroeconomic factors). The choice of
approach incorporates credit indices to map methodology needs to be agreed by the Model
TTC to the PiT volatility of the economic Oversight Committee.
cycle and is built into the transitions within
the matrix. The matrices need to be robust Validation of PD Models - Regardless of
and forward looking. the chosen methodology the PDs should be
validated according to the MMS principles,
Portfolio Scaling Approaches – A scaling with both qualitative and quantitative
approach is simpler than the transition assessments, ensuring that a range of
matrices in that averages are modelled PD metrics at a low level of granularity.
as opposed to the dynamic nature of the Comprehensive validation reports should be
transitions and tend to favour smaller produced that address specific features of
segments. However, the drawback of the the models, compare results across several
approach is that volatility of the PD is development methodologies, deal with low
suppressed and results in underestimation. default portfolios by looking at difference
Vasicek Credit Frameworks - The Vasicek between 1 year PDs and TTC PDs, Cumulative
framework is often used to model PiT PD Default Rates etc.. Other factors such as the
term structures, however the institution Central Tendency, Back-testing, Benchmarks
should be aware of the material challenges of etc. should be part of the validation report.
the method in that it was designed to model All PDs should be economically consistent
economic capital and extreme portfolio losses
and the parameters that are challenging to

10
 c. LGD Models facility and other factors. Three versions of
the LGD (downturn, growth and long run
average) then need to be calibrated to the
Key Challenges
TTC LGD and PiT LGD.

Collections and recoveries processes are

TTC & PiT LGD – The TTC LGD measures the
continuously improving therefore a consistent
LGD independent of the economic, whereas
picture of the data is very difficult to achieve,
as PiT incorporates the economy into the
making the modelling very dependent on
models. Each model needs to have appropriate
a range of assumptions to sample the data
risk drivers that relate characteristics of the
into a representative (of the portfolio going
loan to its loss and recovery profiles. For PiT
forward) form. This may add to the modelling
models the LGD tends to be higher during
complexity and therefore increase model
recessionary times which is difficult to assess
risk.
within natural resource (e.g. oil in the UAE)
dependent economies but it is assumed this
Data Collection – Robust data collection
holds.
of loss and recovery information needs to
be detailed within the Data Management
Validation of LGD – The construction of
Framework and include obligor & facility
the LGD models should follow the lifecycle
characteristics, recovery cash flows,
stages detailed in the MMS, including the
collaterals and asset values over time.
need to independently validate the models,
both quantitatively and qualitatively, at
Historical Realised LGD – The computation
development and at 2 regular intervals
of realised LGD needs to be carried out so
thereafter. The validation scope covers
that workout period can link cashflows and
data quality, definition of default, loss and
costs to specific default events and include
recovery, methodologies employed for TTC
the LGD at the default as a percentage of the
and PiT etc.
default exposure and cashflows discount to
the default event. Institutions also need clear
processes and assumptions for dealing with
unresolved cases.

Analysis of Realised LGD – Once the realised

loss data has been extracted it should be
analysed to determine the key drivers of
loss and inform the choice of modelling
methodology. At a minimum it should be
understood when the losses occurred within
the economic cycle, the creditworthiness
of the obligor at the time of the default, the

11
d. Macro Economic Models economic expectations.

Time Series Regression Model Scope –

Key Challenges
The arenas of IFRS9 and Stress Testing are
the main uses of macroeconomic modelling
Macro modelling in the credit risk arena
where UAE financial institutions predicts PD,
is a relatively new discipline therefore
Credit Indices and LGD. Given the nature
it can be difficult to assess the range of
and complexity of the models, statistical
model alternatives effectively. Expected
techniques are often combined with
relationships between the banking metrics
judgemental approaches where key modelling
and the associated macroeconomic variables
choices are discussed and agreed by model
is often unknown potentially risking the
oversight.
wrong or incorrect variables being included
in a model alternative. Modellers need to
Data Collection – Ideally the macroeconomic
discuss and agree upon that observed trends
data cover an entire economic cycle, but as
are intuitive and in line with business and
a minimum should cover at least 5 years.

12
Fields that need to be collected include GDP, Correlation - The purpose of correlation
oil prices, house prices etc. from several analysis is to determine the strength of
independent sources. The CBUAE can supply the relationship between the dependent
the data on an interim basis. variables, e.g. PD, and the macro variables,
and to see whether the relationships show
Analysis of Dependent Variable – Default causal effects and make economic sense. A
time series should be representative of the correlation cut-off should be implemented to
current portfolio with descriptive statistics select macro variables for modelling.
and expert judgement used to determine data
suitability for modelling. Model Construction - The aim of the macro
modelling is to build relevant and robust
Variable Transformation - Variable relationship between the dependent variable
transformations will have an effect on the and several macro variables (the choice of
macro models as well as ECL (Expected variables is dependent upon the correlations)
Credit Loss), therefore any derivations using an appropriate methodology to
should be tested and documented and perform the multivariate regression analysis.
applied to both macro and dependent Performance of the models needs to be
variables. Transformations include changes assessed using a range of metrics, but also
for stationarity, lags and smoothing. business / modeller judgement.

Statistical Tests
The following tests should be used across the macro-modelling arena:

Property to be tested Description of the property to be rejected Recommended test (others

may exist)

Stationarity Absence of stationarity in each time series Augmented Dickey- Fuller (ADF)

Co-integration Absence of stationarity in a linear combination of the Engle-granger two- step method
dependent variable and each independent variable

Multicollinearity High correlation between the independent variables Variance Inflation Factor

Coefficient The coefficients are not statistically significantly different Coefficient p-value
significance from zero on a t-distribution

Autocorrelation High correlation between the error terms of the model Ljung-Box test

Heteroscedasticity Absence of relationship between independent variables and Breusch-Pagan or White test
residuals

Normality Normal distribution of the residuals Shapiro Wilk

13
• Model Selection - Model and macro • Validation of the model should be
variable selection should be based on performed an independent party, separate
clearly defined performance criteria, so from the development team, following the
that output is consistent with historic MMS principles. As the volume of macro
experience and produce accurate data is low, monitoring of macro models
predictions. The factors used in model can be done less frequently than other
selection from the pool of available types of models, annually at a minimum.
models cover statistical performance (the
model should be robust and stable on the • Scenario Forecasting – A key requirement
development, hold-out and out-of-time of the IFRS9 is for the metrics to be forward
samples)., model sensitivity, intuitive looking therefore the macroeconomic
from a business perspective, have realistic models need to include macro variables
outcomes and be implementable. Finally, that will be available going forward. Three
to test intuitiveness the model should scenarios need to be included, baseline,
be tested under downturn scenarios. upside and downside.
Model forecast uncertainty needs to be
estimated, documented and reported to
the Model Oversight Committee.

14
 Interest Rate Risk in the Banking Book
e.
(‘IRRBB’) Models

Key Challenges

The scope of the section covers both

conventional and Islamic products, with
models addressing previously issued
regulation that look at expected earning and
the value of the balance sheet. The IRRBB
model requirements relate to governance,
management, hedging and reporting.

Metrics - All interest sensitive positions

should be identified and reconciled against
the general ledger with variation in expected
interest earning captured by several metrics
including gap risk (difference between future
cash inflow and outflows), gap risk duration,
economic value of equity and net interest
income (net profit for Islamic products)

Modelling - Models to predict the interest

rate risk need to follow the MMS guidelines
and model lifecycle with modelling
assumptions are not preserve of only the
ALM or market risk function but need to be
agreed by the Model Oversight committee
with the modelling complexity determined by
the size and sophistication of the institution.
The model requirements aim to ensure the
modeller looks at computation granularity,
time buckets, option risk, commercial
margins, basis risk, currency risks, scenarios be either implicit or explicit). To model the
and IT Systems. option risk the modeller needs to consider
identifying material products, ensure
Option Risk - Option Risk is a fundamental assumptions are justified by historical data,
building block of IRRBB models as it looks at understand sensitivity, and fully document
potential changes in the future flow between the process and incorporate at a granular
assets and liabilities (where the options can level (if it is a large complex institution).
15
Interest Rate Scenario - Institutions should based on the principles of both deterministic
compute Delta EVE and Delta NII under 6 and statistical models ensuring that the
interest rate scenarios to account for interest assumptions and decisions are justified. The
rate shocks, these include parallel up and validator should consider the mechanistic
down shocks, a steeper shock (where short- construction, the financial input flows
term rates are down & long-term rates are correctly, the models are coherent and
up), flattener (short up, long down), short that the behavioural patterns are correctly
rate up-shock, short rate down-shock. The incorporated. Finally, the validation
choice of shock should be supported by should support the robust decisioning and
appropriate governance. The consideration management of interest rate risk.
of negative interest rate potential should be
incorporated into the estimates

Validation of EVE & NII - All EVE and NII

models in the IRRBB framework should be
independently validated as per the MMS and

16
 they are generally not recalibrated but the
f. Net Present Value (‘NPV’) Models
assumptions and consistent methodologies
around the inputs should be reviewed and
Key Challenges & Scope validated.

The concept of Net Present Value is used to Methodology - The modelling of NPV is split
estimate various metrics within financial into two parts, mathematical mechanistic
accounting, risk management and business considerations (well documented in
decisions around asset valuations, investment accounting rulebooks) and the choice of
value, collateral valuations and financial inputs (where institutions have a degree of
modelling to estimate cost. From an MMG autotomy).
perspective NPV comes into ECL, LGD and
CVA models. Documentation - Standalone NPV Models
need to be fully documented addressing
Governance - Standalone NPV models are the methodology, assumptions, inputs
included in the model inventory, subjected to etc. with a dedicated document for each
the model lifecycle management discussed in material valuation exercise that also details
MMS, and approved by the Model Oversight the business rationale and the prevailing
Committee. However, as deterministic model economic climate.

Validation - As the
NPV models are
considered within the
remit of the MMS a full
independent validation
of the methodology used,
assumptions made, and
inputs considered needs
to be carried out on a
regular basis. Particular
attention needs to be
paid to the application
of credit premiums due
to the degradation of
creditworthiness when
restructuring occurs.

17
Authors
Dr. Nasir M. Ahmad Dr. Horst Kausch
Managing Partner Partner and Head of Research
Basinghall Analytics Basinghall Analytics
Email: nasir.ahmad@basingha.com Email: horst.kausch@basingha.com

Matthew Freeman
Credit Risk Lead
Basinghall Analytics
Email: matthew.freeman@basingha.com

Contact
Feel free to send your queries to:

Sandip Mukherjee
Co-Founder
Aptivaa
Email : sandip.mukherjee@aptivaa.com

Aptivaa is a vertically focused finance and risk management consulting and analytics firm
with world-class competencies in Credit Risk, Market Risk, Operational Risk, Basel III, IFRS-
9, Risk Analytics, COSO, ALM, Model Risk Management, ICAAP, Stress Testing, Risk Data and
Reporting. Aptivaa has emerged as a leading risk management solutions firm having served
over 100 clients across 22 countries comprising of highly respected names in the financial
services industry.

Basinghall Analytics is a London-based risk advisory firm. We provide specialist services in

the areas of Analytics, Risk and Technology. Our leadership team has held senior modelling-
related roles at leading global institutions. We regularly work closely with European and Asian
regulators. Our core specialisms are modelling and stress testing (including climate risk). We
cover the entire model lifecycle, including development, validation, monitoring, governance
and model risk management. For more information, please visit www.basingha.com

Disclaimer:
For our analysis as represented in this document, we have used commercially available market data obtained from sources we generally believe
to be reliable. We are not giving an opinion or any other form of assurance on information from these sources. Unless otherwise noted, the values
calculated by us are derived using applicable market data parameters and generally accepted valuation methodologies.
18