RISK DATA MANAGEMENT

VOL 5 | AUGUST 2015

RISK DATA MANAGEMENT VOL 5

ONLINE

FEATURED ARTICLES:

Gain a competitive advantage through good data management

NOVEMBER
AUGUST 2014
2015

Find additional risk data management articles, interviews,

and multimedia content at [Link]/RiskPerspectives

STRONG DATA MANAGEMENT AN ABSOLUTE NECESSITY

REGULATORY BIG DATA

32

Learn about trends, overall regulatory goals, and the impact of major initiatives

STRESS TESTING PROCESS FLOW: FIVE KEY AREAS

Build a streamlined process to run stress tests with ease, efficiency, and control

126

Reporting

ECB
Basel III
Data Gaps PD
Forecasting P&L

BCBS 239 Credit Risk

Finding Alpha

DFAST

CCAR

Human Data
Aggregate data

Capital Planning

Data Infrastructure

Stress Testing
Data Governance

IFRS 9
FR
Y-14Q
LGD

Credit Risk
Basel III

DFAST

Expected Default Frequency

Stochastic Scenario Generation

CCAR Granular Data PD

IFRS 4

Monte Carlo Study

The
Fed
Data Infrastructure Risk Appetite BCBS 239
PD Analytical Data

The Fed

Regulatory Big Data

Regulatory Capital

Expected Loss

Portfolio Indicators

Data Quality

FR Y-14Q

Forecasting P&L

Human Data

IAS 39

IAS 39

Risk Appetite

IFRS 4

Systemic Risk

Data Integration

FR Y-14M RWA

AnaCredit LGD Reporting

Commercial Credit Decisioning

LGD

Forecasting P&L

AnaCredit Solvency II Datamart

Basel Risk Systems LGD AnaCredit
FTP Framework Forecasting P&L
Data Management

RWA

IFRS
9
LCR FDSF

PPNR Modeling

Expected Default Frequency

Macroeconomic Scenarios

Model Risk Management

FROM THE EDITOR

Welcome to the fifth edition of Risk Perspectives, a Moodys Analytics
publication created for risk professionals, with the goal of delivering essential
insight into the global financial markets.

exposed. In Getting Human Data Right: The Hidden Advantage, Kevin Hadlock
writes about how banks often neglect the human side of their operations,
and can better manage risk arising from employee knowledge and skill.

Data is the very lifeblood of every bank. Its a shame how little attention
is paid to it, a central banks representative told me recently. As any
form of effective risk management requires powerful analytics, a robust
infrastructure and above all reliable data, this observation sheds
light on a problem that has already severely impacted
the financial services industry, and stands to impact
it further.

Banking regulations are increasingly quantitative and data-driven. In

Regulatory Spotlight, we look at key initiatives, such as BCBS 239 and the
European Central Banks analytical credit dataset (AnaCredit), and discuss the
regulatory challenges presented by PPNR and CCAR/DFAST.
The Approaches to Implementation section discusses
how to design a robust data governance process. It also
sheds light on data management in Asia-Pacific and in
structured finance.

For years, supervisors and industry experts have raised

concerns about the weak data management practices
prevalent in many banks, from poor data quality to
fragmented data infrastructure to non-existent data
governance. But nothing really changed until the
Basel Committee on Banking Supervision published its
Principles for Effective Risk Data Aggregation and Risk
Reporting in January 2013, forcing banks to enhance
their data management for good.

In the final section, Principles and Practices, my

colleagues describe the impact of better data
management on a banks operations including how data
supports critical business decisions, the impact of data
quality on credit risk modeling and stress testing, and the
ways in which a loan origination process benefits from
better data quality. In Modeling Techniques and Tools in
Scenario-based Risk Appetite Management, Pierre Gaudin
writes how regulatory stress testing requirements are
increasingly guiding financial institutions toward scenario
based-governance and risk appetite management.

Given this development in the market, we decided

to dedicate this edition of Risk Perspectives to
data management.
As mentioned above, data management comprises many aspects. With this
in mind, we have invited subject matter experts from Moodys Analytics to
share their experiences and discuss diverse aspects of data management. As
with our first four editions, this issue of Risk Perspectives offers actionable
information to assist risk professionals in their day-to-day efforts to comply
with new regulatory guidelines, master data management and infrastructure
questions, and create value for their organization through better and more
effective risk management.
In the section Rethinking Risk Management, we discuss how banks can benefit
from stronger data management, an absolute necessity for effective risk
management. We also show how banks and insurers can improve their data
quality and ultimately gain better insight into the risks to which they are

EDITORIAL

Contributing Editors

Editor-in-Chief
Dr. Christian Thun

Andrew Boddie
Mina Kang
Brian Robinson

Managing Editor
Rich McKay
Copy Editors
Allison Geller
Alexis Alvarez

DESIGN AND
LAYOUT
Creative Director
Clemens
Frischenschlager
Art Director
Oluyomi Ikotun

I am sure that our perspectives on the challenges and benefits of robust

data management will help you better understand how to address poor data
quality, fragmented data infrastructure, and weak data governance in your
own organization, and ultimately improve your risk management system to
build a more competitive business.
I encourage you to take part in this discussion and help us shape future issues
of Risk Perspectives by sharing your feedback and comments on the articles
presented in this fifth edition.
Dr. Christian Thun
Senior Director, Strategic Business Development
RiskPerspectives@[Link]

Designers
Chun-Yu Huang
Steven Prudames
Jeremy Haiting

Mary Hunt
Robert King
Michael Richitelli
Stephen Tulenko

Web Content
Managers
Meghann Solosy
Marcus McCoy

CONTRIBUTORS

ADVISORY BOARD
Michelle Adler
Gus Harris

Greg Clemens
Pierre Gaudin
Cayetano Gea-Carrasco
Kevin Hadlock
Dr. David Hamilton
Brian Heale

While we encourage everyone to share Risk PerspectivesTM and its articles, please contact RiskPerspectives@[Link]
to request permission to use the articles and related content for any type of reproduction.

Dr. Tony Hughes

Nicolas Kunghehian
Eric Leman
Dr. Juan Licari
Dr. Samuel Malone
Mark McKenna
Yuji Mizuno
Dr. Gustavo OrdonezSanz
Dr. Brian Poi
Buck Rumely
Peter Sallerson

Vivek Thadani
Dr. Christian Thun
Michael van Steen
SPECIAL THANKS
Rich Casey
Robert Cox
Shivani Kalia
David Little
Samantha Marussi
Tony Mirenda

CONTENTS

FROM THE EDITOR

Dr. Christian Thun, Senior Director, Moodys Analytics Strategic Business Development, introduces the content
of this Risk Perspectives edition, including the theme, relevant topics, and how to get the most out of it.

RETHINKING DATA MANAGEMENT

Strong Data Management An Absolute Necessity

Dr. Christian Thun

Building a Comprehensive FTP Framework for Stress Testing, Risk Appetite,

and Forecasting P&L

14

Nicolas Kunghehian

Using Analytical Data for Business Decision Making in Insurance

18

Brian Heale

Getting Human Data Right: The Hidden Advantage

24

Kevin Hadlock

REGULATORY SPOTLIGHT
Regulatory Big Data: Regulator Goals and Global Initiatives

32

Michael van Steen

IFRS 9 Will Significantly Impact Banks Provisions and Financial Statements

38

Cayetano Gea-Carrasco

What if PPNR Research Proves Fruitless?

50

Dr. Tony Hughes

Implementing the IFRS 9s Expected Loss Impairment Model: Challenges and

Opportunities
Eric Leman

53

APPROACHES TO IMPLEMENTATION
Enhanced Data Management: A Key Competitive Advantage for
Japanese Banks

60

Yuji Mizuno

Measuring Systemic Risk in the Southeast Asian Financial System

66

Dr. David Hamilton, Dr. Tony Hughes, and Dr. Samuel W. Malone

Effect of Credit Deterioration on Regulatory Capital Risk Weights for

Structured Finance Securities

76

Vivek Thadani and Peter Sallerson

Finding Alpha: Seeking Added Value from Stress Testing

84

Greg Clemens and Mark McKenna

PRINCIPLES AND PRACTICES

Modeling Techniques and Tools in Scenario-Based Risk Appetite Management

92

Pierre Gaudin

The Benefits of Modernizing the Commercial Credit Decisioning Process

100

Buck Rumely

Multicollinearity and Stress Testing

104

Dr. Tony Hughes and Dr. Brian Poi

Multi-Period Stochastic Scenario Generation

110

Dr. Juan Licari and Dr. Gustavo Ordoez-Sanz

Stress Testing Solution Process Flow: Five Key Areas

126

Greg Clemens and Mark McKenna

DATA MANAGEMENT BY THE NUMBERS

SUBJECT MATTER EXPERTS
SOLUTIONS
GLOSSARY

4
134
139
143

DATA MANAGEMENT BY THE NUMBERS

88%

$2M

100%

Of all data integration projects either fail

42% of respondents have allocated this budget

Knowledge and skills age so rapidly that the

completely or significantly overrun their

or higher for IFRS 9 compliance.

likelihood of employee error approaches 100%

budgets.

IFRS 9 Will Significantly Impact Banks

by the end of the five-year period.

Strong Data Management An Absolute

Provisions and Financial Statements. Page 38

Getting Human Data Right: The Hidden

Necessity. Page 8

Advantage. Page 24

20

59%

We know of one leading commercial bank that

The correct full model is chosen 59% of the

Main components of an Insurance Analytics

employs 20 full time workers to aggregate

time. Overall, the inflation coefficient is

Platform (IAP) architecture.

and clean data in preparation for the FR Y-14Q

statistically significant in around 67% of cases,

Using Analytical Data for Business Decision-

quarterly commercial data submission.

whereas the unemployment rate coefficient is

Making in Insurance. Page 18

The Benefits of Modernizing the Commercial

significant 91% of the time.

Credit Decisioning Process. Page 100

Multicollinearity and Stress Testing. Page 104

MOODYS ANALYTICS RISK PERSPECTIVES

150

43,700 30

From our analysis of similar initiatives and the

The final portfolio for analysis comprised

As a result, the accuracy of liquidity-monitoring

preparatory work involved, we expect that the

approximately 43,700 securities, which

models depends on the ability to evaluate

ECB will consider between 24 and 150 or more

effectively represent the structured finance

realistic credit transitions over a time horizon as

attributes per loan for inclusion in AnaCredit.

universe of non-agency transactions.

short as 30 days.

Regulatory Big Data: Regulator Goals and Global

Effect of Credit Deterioration on Regulatory

Modeling Techniques and Tools in Scenario-

Initiatives. Page 32

Capital Risk Weights for Structured Finance

Based Risk Appetite Management. Page 92

Securities. Page 76

47%

50%

1940s

Surprisingly, 14 out of 30 G-SIBs revealed that

Between June 1997 and March 1998, GDP

In the case of commercial loan volume, for

they will not be fully compliant with at least

contracted by nearly 6% in Korea, 9% in

instance, the Federal Reserve Board has

one of the Basel Committees regulatory

Thailand, and 14% in Indonesia. Equity

quarterly data stretching back to the late 1940s.

principles by the deadline in 2016.

valuations plummeted by 50% or more in

What if PPNR Research Proves Fruitless?

Enhancing Data Management is a Key

the affected countries.

Page 50

Competitive Advantage for Japanese Banks.

Measuring Systemic Risk in the Southeast

Page 60

Asian Financial System. Page 66

RISK DATA MANAGEMENT | AUGUST 2015

RETHINKING DATA
MANAGEMENT
Discusses how to establish better data management to gain a competitive
advantage, build a comprehensive FTP framework, use analytical data to
improve insurers business decisions, and manage employee knowledge
and skills.

STRONG DATA MANAGEMENT AN ABSOLUTE

NECESSITY
By Dr. Christian Thun

Dr. Christian Thun

Senior Director, Strategic
Business Development

Christian provides deep expertise on credit risk

management, Basel II, and portfolio advisory projects
and functions as a main contact for regulators and the
senior management of financial institutions.

Banks and businesses have long been plagued by poor data quality,
a result of weak technology, lack of management oversight, and
simple human error. Inferior data, too long left unchecked, has farreaching consequences not the least of which was the 2008 global
financial crisis. Banks that establish a strong data management
framework will gain a distinct advantage over their competitors and
more efficiently achieve regulatory compliance.
The data wasn't and still isnt good enough

financial crisis and the reason that the sizeable

Five years after the financial crisis, firms

investments in improving risk management in

progress toward consistent, timely, and accurate

the years preceding the crisis seem to have been

reporting of top counterparty exposures fails

in vain: There was and still is not enough good

to meet both supervisory expectations and

data about the risk to which a bank is exposed.

industry self-identified best practices. The area

of greatest concern remains firms inability to
consistently produce high-quality data.1

Effective risk management that is capable of

identifying, assessing, and prioritizing risks
is based on a sound infrastructure, powerful

This quote from the Progress Report on

analytics, and reliable data. All three ingredients

Counterparty Data by the Senior Supervisors

are interconnected and influence each other, as

Group summarizes one of the causes of the

illustrated by Figure 1.

Figure 1 Three ingredients of effective risk management

Infrastructure

Risk
Management

Analytics

Source: Moodys Analytics

MOODYS ANALYTICS RISK PERSPECTIVES

Data

RETHINKING DATA MANAGEMENT

Infrastructure comprises not only technical

issues like standard regulatory reports, the story

aspects like IT equipment, but also technical

is the same. A business or business function

organization and processes such as IT

(like risk management) that has to rely on weak

governance, roles, responsibilities, and

data is ultimately set up for failure. A bank that

internal policies.

uses good quality data has an opportunity to

Analytics refers to the wide variety of

quantitative modeling techniques that have
developed over the past 20 years to better
understand the drivers of risk and predict
potential losses resulting from credit or
market activities.
Data includes not only granular information
about risk exposure itself, but also the
taxonomies that define and categorize that
information and the data governance that
maintains the accountability and quality of
the data.

outpace its competitors.

What is good data quality?
There have been numerous attempts in the past
two decades to define data quality along a series
of dimensions, such as accuracy and consistency.3
Depending on the individual needs of an
organization, that definition can vary.
Table 1 shows the typical criteria used by
statistics providers like the Statistics and
Regulatory Data Division of the Bank of England
and Eurostat.4 Most of todays banks fall short in

As the quote from the Senior Supervisors

Group suggests, many banks use seriously
flawed data, making meaningful risk
management next to impossible. Weak data
quality is an impediment not only for risk
management, but also for the business of a
bank in general. As other risk management
experts have pointed out, If the data quality
is poor, the information will be poor and only
luck can stop the decisions from being poor.2
From important strategic decisions to mundane

at least one of these areas, giving rise to serious

concerns for risk managers.
What are the consequences of poor quality
data?
Weak data is a common deficiency in almost
all businesses. Still, some companies tolerate
a certain level of bad data rather than try to
manage or eliminate it, because the sources of
poor data quality are myriad and addressing
them one by one is a laborious, time-consuming,
and expensive exercise.

Table 1 Typical criteria used by statistics providers

Dimension

Description

Relevance

Relevance is the degree to which data meets current and potential users needs:
whether statistics and concepts (definitions, classifications, etc.) reflect these needs.

Accuracy

Accuracy in the data denotes the closeness of computations or estimates to exact or

true values: how accurately and reliably the information portrays reality.

Timeliness and
Punctuality

Timeliness of information reflects the length of time between the availability of data
and the event or phenomenon it describes.

Accessibility and
Clarity

Accessibility refers to the physical conditions in which users can obtain data: where to
go, how to order, delivery time, availability of micro or macro data formats, etc.
Clarity refers to the datas information environment: whether data is accompanied
by appropriate metadata, illustrations such as graphs and maps, and information on
their quality (including limitations in use).

Comparability

Comparability aims at measuring the impact of differences in applied statistical

concepts, definitions, and measurement tools/procedures when comparing statistics
among geographical areas, non-geographical domains, or over time.

Coherence

Coherence refers to the ability to reliably combine the data in different ways and for
different uses.

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

Sooner or later, however, bad data begins to

quality of their data and underestimate the

proliferate across systems, and discrepancies

cost of errors.12

grow rapidly, which results in a number of issues:5

Increased downtime for systems to reconcile
data
Diversion of resources from areas important
for the business

One telecommunications firm lost $8 million

a month because data entry errors incorrectly
coded accounts, preventing bills from being
sent out.13
One large bank discovered that 62% of its
home equity loans were being calculated

Slower deployment of new systems

Inability to comply with industry and quality
standards
Frustrated employees whose activities are
hampered by poor data

incorrectly, with the principal getting larger

each month.14
One regional bank could not calculate
customer or product profitability because of
missing and inaccurate cost data.15

A cumulative increase in costs

These findings provide an idea of the extent to

Quantifying the cost of bad data

There have been several attempts to quantify
the cost of bad data quality. The exact cost is
difficult to calculate, but research by academics

which weak data quality can add to a businesss

costs. Given that these examples stem from
research and industry reports that cover the
first decade of the 21st century, one must ask

Weak data quality is an impediment not only for risk management, but also
for the business of a bank in general. As other risk management experts
have pointed out, If the data quality is poor, the information will be poor
and only luck can stop the decisions from being poor.

and reports by industry experts provide a

why data quality management has not been

number of revealing examples:6

addressed more seriously by those responsible.

According to a 2010 Forbes survey, datarelated problems cost companies more

There are two main reasons for the data

quality deficit

than $5 million annually. One-fifth of the

Experts have repeatedly identified two main

companies surveyed estimated losses in

reasons for the weak data quality that plagues

excess of $20 million per year.7

many banks the lack of accountability and

Gartner research shows that 40% of the

commitment by the organizations senior

anticipated value of all business initiatives

management to address weak data, and the lack

is never achieved. Poor data quality in both

of effective technologies to monitor, manage,

the planning and execution phases of these

and correct inaccurate data when needed.

initiatives is a primary cause.8

Eighty-eight percent of all data integration
projects either fail completely or significantly

involvement, the Basel Committee on Banking

overrun their budgets.

Supervision (BCBS) outlined a number of new

Seventy-five percent of organizations have

identified costs stemming from dirty data.

10

Thirty-three percent of organizations have

MOODYS ANALYTICS RISK PERSPECTIVES

responsibilities. Boards must now determine

their risk reporting requirements and be aware
of the limitations that prevent a comprehensive

delayed or canceled new IT systems because

aggregation of risk data in the reports they

of poor data.11

receive.16 Senior management must also ensure

Organizations typically overestimate the

10

To address the lack of senior management

that its strategic IT planning process includes

RETHINKING DATA MANAGEMENT

both a way to improve risk data aggregation

By setting up stronger data governance

capability and the creation of an infrastructure

structures, banks will address the first main

that remedies any shortcomings against the

data quality weakness and define the correct

principles defined by the BCBS.

use of and accountability for that data. Data

These obligations will cover the entire value

chain of a banks data, because the BCBS
requires that senior management understand
the problems that limit the comprehensive
aggregation of risk data in terms of:
Coverage i.e., are all risks included?
Technical aspects i.e., how advanced is the
level of automation, vs. manual processes?
Legal aspects i.e., are there any the
limitations to sharing data?
To comply with these requirements, a bank
will need to establish strong data governance
covering policies, procedures, organization, and

governance should also include the scope of a

banks IT governance by considering data quality
aspects and processes.
Reaching a new data quality standard:
A step-by-step process
Because data changes constantly, continuously
monitoring it to maintain quality will only
become more and more important. Figure 2
outlines the process.17
Once data is extracted from source systems,
the next step profiling applies rules and
error checks to assess the overall quality of the
data. Profiling involves identifying, modifying
or removing incorrect or corrupt data, with the

Technology that reliably maintains data quality by automatically applying

error checks and business rules and by supporting sound audit trails and
lineage can only result in greater confidence in the data.
roles and responsibilities as part of its overall
corporate governance structure.

goal of retaining just one unique instance of each

datum. This step is also supported by the BCBS,
which requests that banks strive toward a single

Figure 2 Monitoring and maintaining data quality step-by-step process

DATA
EXTRACTION

DATA PROFILING/
QUALITY

CLEANSING &
DE-DUPING

DATA
STANDARDIZATION

QUALITY
MONITORING

ENRICHMENT

Extract data from

the various source
systems.

Make use of data

profiling techniques.

This is a dual
process cleansing
& de-duping.

Run a series of data

quality checks/rules
against the data.

Keep track of data

quality over time.

Enhance the value of

internally held data.

Loans

Use logic algorithms

and rules (both
general and specific
to the business) to
produce a picture of
overall data quality.

Identify and
modify corrupt or
inaccurate data and
remove or modify
incomplete, incorrect
or inaccurate data.

Use any of a
number of tools,
including thousands
of pre-built data
quality rules (both
general and industry
specific), enhanced
by user-defined
rules.

Use software
to autocorrect
variations based
on defined
business rules.

Enhance the value

of internally held
data by appending
related attributes
from external sources
(e.g., consumer
demographic
attributes or
geographic data).

Customer
Asset
Finance
Forecasts
Modeling
Risk

Retain only one

unique instance of
the data.

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

11

authoritative source for risk data per each

type of risk.

as the information itself are constantly

checks will evolve to avoid the repetition of

evolving, effective technology has to be

previously identified data errors. Automating

adaptable. The technology should feature

these processes will help maintain data quality

flexible processes to aggregate data in different

and even enhance it by making the data

ways and should be able to include new

more comprehensive.

information or exclude outdated information.

To address the second reason the lack of

effective technologies banks today have the
opportunity to select from a wide range of tools

It should also reflect new developments within

the organization as well as any external factors
influencing the banks risk profile, such as
changes in the regulatory framework.

and solutions that support processes to improve

Summary

and maintain data quality. Most banks already

Effective risk management relies on three key

have some components that could form the

ingredients: sound infrastructure, powerful

foundation of a data quality framework,

analytics, and reliable data. The latter especially

which they could then enhance with new

has been neglected for too long by too many.

components as required.

Although regulators have repeatedly voiced their

The requirements for effective technologies

hinge on speed, scalability, reliability, and
adaptability. Speed and scalability speak to
the ever-growing amounts of different types of

concerns, it took a financial crisis to implement

much tighter rules. As a result, banks will have
to invest heavily in their data management
architecture in the coming years.

data stored in multiple, siloed systems based on

The two main reasons for weak data quality are

entities, lines of businesses, risk types, etc. After

a lack of senior management commitment and

identifying which system contains the required

ineffective technology. To benefit from good

data, an expert must extract, standardize, and

data management, banks will need to establish

consolidate it to assess its quality.

strong data governance that sets rules and

Despite the fact that many banks employ large

numbers of employees to capture, review, and
validate data (as well as find gaps), they still
struggle with poor data quality in their core
systems, as a result of input errors, unchecked
changes, and the age of the data (particularly
with information stored in legacy systems or
compiled manually). Technology that reliably
maintains data quality by automatically
applying error checks and business rules and by
supporting sound audit trails and lineage can

MOODYS ANALYTICS RISK PERSPECTIVES

As the requirements for information as well

Based on lessons learned, the set of rules and

Implementing effective technologies

12

only result in greater confidence in the data.

18

defines clear roles and responsibilities, while

enhancing an existing data quality framework
with technologies that offer speed, scalability,
reliability, and adaptability.
Good data management will confer a
competitive advantage to those banks that
have it. The value banks can reap from setting
up a better data management framework will
be leaner, more efficient and less expensive
processes that lead to faster and more reliable
business decisions.

RETHINKING DATA MANAGEMENT

1 Senior Supervisors Group, Progress Report on Counterparty Data, p. 1, 2014.

2 Mackintoch, J./Mee, P., The Oliver Wyman Risk Journal, Data Quality: The truth isnt out there, p.75, 2011.
3 Haug, A./Albjrn, J.S., Journal of Enterprise Information Management, Vol. 24, No. 3, 2011, Barriers to master data quality, pp.
292-293.
4 Bank of England, Data Quality Framework, p. 7, 2014.
5 Marsh, R., Database Marketing & Customer Strategy Management, Vol. 12 No. 2, Drowning in dirty data? Its time to sink or
swim: A four-stage methodology for total data quality management, p. 108, 2005.
6 Marsh, R., Database Marketing & Customer Strategy Management, Vol. 12 No. 2, Drowning in dirty data? Its time to sink or
swim: A four-stage methodology for total data quality management, p. 106, 2005.
7 Forbes, Managing Information in the Enterprise: Perspectives for Business Leaders, p.2, 2010.
8 Gartner, Measuring the Business Value of Data Quality, p.1, 2011.
9 Marsh, R., Database Marketing & Customer Strategy Management, Vol. 12 No. 2, Drowning in dirty data? Its time to sink or
swim: A four-stage methodology for total data quality management, p. 106, 2005.
10 Marsh, R., Database Marketing & Customer Strategy Management, Vol. 12 No. 2, Drowning in dirty data? Its time to sink or
swim: A four-stage methodology for total data quality management, p. 106, 2005.
11 Marsh, R., Database Marketing & Customer Strategy Management, Vol. 12 No. 2, Drowning in dirty data? Its time to sink or
swim: A four-stage methodology for total data quality management, p. 106, 2005.
12 Marsh, R., Database Marketing & Customer Strategy Management, Vol. 12 No. 2, Drowning in dirty data? Its time to sink or
swim: A four-stage methodology for total data quality management, p. 106, 2005.
13 Eckerson, W. W., The Data Warehouse Institute Report Series, Data Quality and the Bottom Line, p. 9, 2002.
14 Eckerson, W. W., The Data Warehouse Institute Report Series, Data Quality and the Bottom Line, p. 9, 2002.
15 Eckerson, W. W., The Data Warehouse Institute Report Series, Data Quality and the Bottom Line, p. 9, 2002.
16 BCBS, Principles for effective risk data aggregation and risk reporting, p. 7, 2013.
17 Heale, B., Risk Perspectives, Vol. 4, Data: The Foundation of Risk Management, p. 53, 2014.
18 BCBS, Principles for effective risk data aggregation and risk reporting, p. 8, 2013.

RISK DATA MANAGEMENT | AUGUST 2015

13

BUILDING A COMPREHENSIVE FTP

FRAMEWORK FOR STRESS TESTING, RISK
APPETITE, AND FORECASTING P&L
By Nicolas Kunghehian

Nicolas Kunghehian
Director, Business Development

Nicolas provides insight on ALM, liquidity, and market

risks to help financial institutions define a sound risk
management framework.

Funds transfer pricing (FTP) is of growing concern to banks and

regulators. But what does FTP have to do with stress testing? A
comprehensive FTP framework can help organizations use the results
of stress tests to forecast their P&L across departments and lines of
business, ensuring that each units strategy aligns with that of the
greater organization.
Current FTP systems are incomplete

allocation,1 emphasizing that banks should

Virtually all banks use funds transfer pricing,

have robust strategies, policies, processes,

and yet there are no practices common to

and systems for liquidity risk management

all. Consequently, regulators are asking them

that should include adequate allocation

to improve their FTP systems. Banks are

mechanisms of liquidity costs, benefits,

developing comprehensive frameworks to meet

and risks.

these demands, but stress testing has been left

out of that framework.

only components that need to be carefully

scenarios should be part of an FTP transaction,

monitored. Due to tough economic conditions

integrating stress testing into an FTP framework

and the cost of regulatory compliance for both

is more important now than ever.

capital and liquidity ratios, the overall P&L of

structure, some FTP components can be valued

some business units has dropped dangerously

close to zero.

using market prices hence, the methodology

The main goals of a robust and consistent FTP

is the same for all banks. But costs or risks

system are to correctly allocate the P&L to

connected with some types of transactions are

each line of business and to forecast different

generally not monitored until a critical loss

costs in different scenarios. The general

occurs. This is what recently happened, for

framework will generate P&L for different

example, with higher liquidity funding costs and

departments and, depending on the way this

low interest rates.

framework is built, guide each department to

Moving to an FTP framework: goals and

challenges
The liquidity crisis severely impacted FTP
systems. Once banks took liquidity funding
costs into account, they realized that some of
their transactions were barely profitable.
In 2009, the European Banking Authority
(EBA) published guidelines on liquidity cost

MOODYS ANALYTICS RISK PERSPECTIVES

called liquidity transfer pricing are not the

Even though banks do not believe that stressed

While each bank has its own organizational

14

But the liquidity components of the FTP also

a specific P&L strategy. It is therefore critical

that this framework be aligned with a banks
overall strategy to help incentivize its teams
to make profitable business decisions and
better manage its overall P&L.
Unfortunately, moving to a comprehensive
framework will increase the costs allocated
to each line of business, as it will reveal new

RETHINKING DATA MANAGEMENT

Figure 1 FTP system process

Loan interest

Customer
Transaction

Business
Line

Funding
Unit

Interest
Rate Risk

Maturity
Mismatch

Hedged with Interest

Rate Swaps

Funded on
the Market

Regulatory
Compliance

Option Risk

Risk Capital
Allocation

Basel III Liquidity

Buffers

Credit Mitigation

Forecast of the
Liquidity Ratios in
Different Scenarios

(Prepayments, Embedded
Cap, and Floors)

Buying Caps on the

Market, Swaptions to
Hedge Prepayments

Fixed Costs

Final P&L of the

Business Line
Source: Moodys Analytics

FTP components that were either previously

hidden or not monitored. Internal costs will
be higher for all transactions. This is why some
transactions may appear to have a negative
P&L, a change banks will need to explain to all
business units. The framework will then become
more than a technical tool it will indicate the
need for a strong change management plan.
Main framework characteristics
Banks should take several factors into
account when designing a comprehensive
FTP framework.
Granularity
If a bank wants to learn which transactions
are profitable, it must calculate FTP at the
transaction level. (Aggregation is usually
not possible or advisable; funds should be
aggregated prudently, if at all.) For example,
a global FTP given to a line of business
without taking into account the way the
principal is amortized would lead to a higher
maturity mismatch.

Consistency
Another pitfall of an FTP calculation is the use
of inconsistent methodologies across a
bank. Most of the time, banks use different
methodologies for each line of business. This
can result in incentives and behaviors that are
not necessarily aligned with the firms overall
strategy. At any time, the sum of the P&L
across all lines of business must equal the P&L
of the overall firm.
Responsiveness
Finally, the framework should be updated as
frequently as possible. A system that is not
updated regularly runs the risk of lagging behind
the rate of transactions, especially as markets
tend to move very quickly.
Forecasting P&L
Once this framework has been put into place,
P&L can be calculated at the line-of-business
level. The total P&L of the bank can be divided
along lines of business, if the transfer units in

RISK DATA MANAGEMENT | AUGUST 2015

15

charge of managing the different types of costs/

than banks that do not because their FTP costs

risks are taken into account.

are higher, leading to higher client rates. In

other words, banks that calculate FTP with this

There are now different departments or

framework are accounting for an additional cost

profit centers, none of which is only a cost

center. They will each be charged for what they

that other banks might not consider.

cost, making it easier to calculate their P&L.

However, this cost is real and should be

The ability to concretely measure risk is very

measured. Take, for example, the likelihood of a

important from an analysis point of view.

customer using a line of credit that is higher than

To better drive business, it is also critical to run

simulations to forecast P&L under different

in a stressed scenario. The cost of liquidity would

be higher, ending in a loss for the treasury. If this
scenario is not part of a banks FTP framework,

scenarios, including stress test scenarios.

it should at a minimum be part of its risk

appetite framework to make the bank aware of

Integrating stress testing into an

FTP framework

the real risks presented by that scenario.

The 2008 financial crisis prompted risk

Second, it is very important to be able to

managers to focus on assessing risks under

measure and forecast the P&L of each business

stressed scenarios. Regulators, as well as the

top management within organizations, are now
asking for P&L metrics as a standard output of
any stress testing exercise.

unit for various scenarios. For example, low

interest rates tend to result in a lowering of the
commercial margin. If, on the asset side of the
business, a client asks for lower rates, the bank

If organizations want to analyze the results of

is more or less forced to comply to keep the

their stress test reports for their impact on P&L,

customer happy and stay competitive. But on

they need to build a comprehensive framework

the liability side, there is a barrier that cannot

like the one previously described. However, they

be breached: 0%. No customer would accept

are likely to run into two stumbling blocks.

negative rates for a checking account.

First, FTP is not ordinarily calculated using a

Because of this tightened margin the difference

stress testing scenario. Banks that use this

between the rate received on the asset side and

methodology will likely be less competitive

the rate paid on the liability side it is important

Figure 2 Lines of business, with ALM generally in charge of the FTP process and allocation

Investment
Banking

ALM

Retail
Banking

Funding
Unit

Corporate
Banking

Source: Moodys Analytics

16

MOODYS ANALYTICS RISK PERSPECTIVES

RETHINKING DATA MANAGEMENT

to measure FTP rates and forecast them with a

risks are unlikely to be realized. Neglecting to

high degree of accuracy.

price these risks at the P&L of the bank will lead

FTP for risk appetite

A banks senior management and operational
teams tend to view stress testing as only a
regulatory compliance or reporting obligation,
not as a benefit to their day-to-day work. But
they must acknowledge that stress testing

to incorrect incentives for operational teams and,

ultimately, affect the organizations profitability.
Risks can also be managed or monitored using
a risk appetite framework, but for consistency
the risk appetite should be reflected in the FTP
price so that all units can see the direct financial
impact of the risk on their P&L.

Ignoring certain risks is dangerous for a banks risk management operations,

even if those risks are unlikely to be realized. Neglecting to price these
risks at the P&L of the bank will lead to incorrect incentives for operational
teams and, ultimately, affect the organizations profitability.

scenarios are important for measuring extreme

In conclusion, banks will find that investing in a

events and their consequences, particularly

comprehensive framework will prove to be more

FTP components, which have historically

effective in the long run, as there will be less

been neglected (e.g., liquidity risk before the

important losses. If losses do occur, they will

subprime crisis).

already be measured and priced for all business

Ignoring certain risks is dangerous for a banks

risk management operations, even if those

units. FTP is one of the most efficient tools for

spreading risk appetite to all teams of the bank.

1 EBA, Guidelines on liquidity cost benefit allocation, October 2010.

RISK DATA MANAGEMENT | AUGUST 2015

17

USING ANALYTICAL DATA FOR BUSINESS

DECISION-MAKING IN INSURANCE
By Brian Heale

Brian Heale
Senior Director, Business
Development Officer, Global
Insurance
Brian is an insurance and Solvency II specialist who
has significant experience in technology solutions
for the global insurance industry. He has an in-depth
knowledge of the life and pensions business, coupled
with a comprehensive understanding of enterprise
technology.

Regulatory compliance is mandatory, but it doesnt have to just

be a burden. Insurers can leverage their regulatory investment to
greatly benefit their business, specifically by creating data-driven
executive dashboards. This article details the organizational and
data challenges that insurers face when harnessing the historical
and forward-thinking information needed to create interactive
dashboards. It explains how these challenges can be effectively
managed using an Insurance Analytics Platform (IAP), leading to
better decision-making at all levels of the business.
Insurers have more work to do despite progress

strategies to generate higher returns and

There is, among many insurers, a feeling of

manage costs.

regulatory burnout and disillusionment. For

the last few years, insurers around the globe
have been heavily focused on implementing
regulatory initiatives, in particular Solvency II
(and equivalent regimes), while also responding
to the implications of International Financial
Reporting Standard 4 (IFRS 4) and regulatory
stress testing.
Hundreds of millions of pounds and euros have
been spent on Solvency II projects that are
now nearing completion, but insurers need to
do more work to realize the potential business
benefits of these investments.
This continued regulatory burden, combined
with a low interest rate/low inflation
environment, is making generating value
increasingly challenging for insurance companies.
Margins are under pressure and firms have
to work much harder to remain competitive
and deliver returns to shareholders and/
or policyholders. Firms must look for new
opportunities to support growth, including

18

MOODYS ANALYTICS RISK PERSPECTIVES

Managing more effectively with better riskbased performance metrics

Our discussions with a number of insurance
CROs, CFOs, and CEOs over the last six months
or so indicate that they have the immediate
regulatory situation under some degree of
control. Therefore, their focus is turning to
running their businesses more effectively and
making informed risk-based decisions.
Business decision-making fundamentally
revolves around three high-level measures:
profitability, capital, and growth. These three
factors need to account for the entire risk profile
of the business.
There are two aspects to assessing the
interaction of these high-level measures:
understanding the historical (e.g., year-todate) performance of the firm tracked against
their strategic business plan, and modeling the
interaction of these measures over future time
horizons under differing stressed scenarios.

designing products that are aligned with this

Figure 1 illustrates the type of information to

new world and adopting alternative investment

which we believe C-suite executives need to have

RETHINKING DATA MANAGEMENT

access. Both historical and forward-looking

Creating a common vision across the entire

perspectives are critical for effective risk-based

business is not difficult at the highest level, but

decision-making.

it becomes more challenging when trying to

define the exact requirements across numerous

Equally important is having the available

information much more quickly ideally in
real time and in a format that is readily
understandable. This in essence translates to
a series of interactive executive dashboards

stakeholders. Even when the requirements are

well understood, the right solution is needed
to deliver the business benefits in a costeffective manner.

with drill-down and what-if capabilities that

Invariably, it will be ownership and

display the requisite analytical information.

implementation of this vision that is the most

difficult part. Adopting a top-down pragmatic

Four challenges of creating interactive

dashboards

approach helps firms focus on what is most

On the face of things, creating these interactive

dashboards seems relatively straightforward.
In reality, however, insurers must anticipate
multiple challenges, both in terms of data

important and avoid a boil the ocean scenario

in which considerable time and effort are spent
trying to resolve all the granular issues without
much visible business benefit.

and organization.

2. Multiple data sources and inconsistent data

1. Vision for risk-based performance metrics

While much of the historic analytical

Although insurers have a good understanding

of the type of historical information they

information (financial, risk, capital, or

investment information) required for decisionmaking may already exist within an organization,

need, the requirements tend to be siloed

within functional areas (e.g., finance or risk).

it is usually fragmented across multiple

systems and a plethora of spreadsheets. This

Figure 1 Information needed by the C-suite

CFO

CEO

CRO

CIO

COO

Chief Actuary

RISK-BASED PERFORMANCE METRICS

OTHER
EXPENSES

CLAIMS

COSTS

INVESTMENT
RETURN

NEW BUSINESS
GROWTH

REVENUE

PREMIUMS

CAPITAL BY
PRODUCT

CAPITAL BY RISK/
RISK APPETITE

CAPITAL/RISK
(REGULATORY & ECONOMIC)

CAPITAL
MEASURES

OWN FUNDS

LIABILITIES

ASSETS

BALANCE SHEET
(ACCOUNTING & SOLVENCY)

PROFITABILITY (PROFIT & LOSS)

INSURANCE
EXPENSES

BALANCE SHEET & SOLVENCY

ANALYSIS BY
GROUP/ENTITY/LOB/
PRODUCT, ETC.

NEW & EXISTING

BUSINESS

STRESS & SCENARIO

TESTING

WHAT-IF ANALYSIS

STRATEGIC PLANNING

HISTORICAL & FORWARD LOOKING

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

19

information has to be collated often using

are inconsistent, there may be questions

manual processes and even more spreadsheets

about the underlying data quality, which can

a procedure that has to be repeated each

undermine senior managements confidence

time information is required, whether for

in the provided Management Information.

regulatory or business purposes such as board

meetings. This makes producing the necessary

3. Forward-looking projections

information and metrics a difficult and time-

Historical information is the most readily

consuming job.

available to any organization. While it is

Consequently, the first challenge is extracting,

transforming, aggregating, and storing all
of the information required in a logical and
structured manner and making it easily available
to the enterprise. Many insurers have existing

important for monitoring the progress against

business metrics and targets (or for regulatory
purposes), historical information is limited in
terms of its strategic planning and decisionmaking capabilities.

Running what-if analyses can be a time-consuming process, especially

if actuarial models are involved. Having to wait days or weeks for this
information does not support the decision-making process. The lack of
accurate and timely information often means that decisions are driven by
gut feeling rather than sound analysis.

operational datamarts or warehouses, but these

Forward-looking projections by scenario

are typically based on legacy systems and are

and their corresponding what-if analyses

not necessarily suitable for storing the type of

are an important part of the C-suite toolkit.

analytical data needed at the required levels

Regulators, under the guise of processes such

of granularity.

as ORSA, are also increasingly using them.

A second problem relates to consistency across

different data sources. If data sources for a
particular use (e.g., assets, profitability, etc.)

Figure 2 ORSA results

Source: Moodys Analytics

20

MOODYS ANALYTICS RISK PERSPECTIVES

Figure 2 illustrates an insurers solvency ratio

projected over a five-year time horizon based
on a baseline scenario (most likely) and four
alternative scenarios. However, forward-

RETHINKING DATA MANAGEMENT

looking projections also present a considerable

especially if actuarial models are involved.

challenge.

Having to wait days or weeks for this

information does not support a dynamic

First, projecting an insurers balance sheet is

decision-making process. The lack of accurate

not always as straightforward as it sounds,

and timely information often means that

particularly for complex organizations or those

decisions are driven by gut feeling rather than

with complex assets and/or liabilities (e.g., path-

sound analysis.

dependent liabilities, as is common for

life insurers).

Building an Insurance Analytics Platform (IAP)

We believe that an Insurance Analytics Platform

Second, a key part of risk-based decision-

can be used to solve many of these challenges

making is the ability to measure return on

capital. This means that firms need to be able to
make projections across multiple regimes, such
as solvency regimes for capital and accounting
regimes (e.g., IFRS) for profitability.

and provide the foundation for the management

of risk-based performance metrics to support
better decision-making.
Figure 3 illustrates the conceptual architecture
of what an IAP might look like for an insurer.

4. It takes too long!

A final challenge is the length of time it takes

1. Insurance analytical repository

to generate the relevant metrics, particularly

The central core to the IAP is a dedicated

for risk-based decision-making. Running what-if

insurance analytical data repository that stores

analyses can be a time-consuming process,

the relevant asset, actuarial, finance, and risk

Figure 3 IAP conceptual architecture

Data Sources

Analytical Repository

Reporting
Engine

Outputs
Regulatory
Reports

Finance
Data

Analytical Repository

CRO
Dashboard

ETL

Policy &
Actuarial
Data

Asset Data

Results

Scenarios &
Assumptions

CFO
Dashboard

CEO
Dashboard

Projected
P&L/BS

Assumptions
Data

Reconciliation
Data

Consolidation
Engine

OLAP CUBES

Risk Data

Reporting Engine

Finance Data

What-If Analysis

Actuarial Data

Asset Data
(Holdings,
Market Data)

Modeling Engines

Actuarial
Engines

Financial
Engines

Solvency
Engines

Results

Results

Results

ORSA
Results

Risk-Based
Metrics

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

21

data (analytical data) and model run results,

enabling the generation of a range of executive

management information dependent on

reports and interactive dashboards.

the stakeholder (CEO, CFO, CRO, etc.) or

The repository acts as the common clearing

house for the analytical data across the
enterprise. The underlying data model can
be designed to support the storage of both
historical and forward-looking analytical
data for the purpose of providing drill-down
analysis of risk-based performance metrics. The
clearing house concept ensures that there
is consistency of data for the analytics, while
also ensuring complete transparency/audit
trails from the source data.
The raw data then has to be extracted,
transformed, and loaded from multiple data
sources (1 and 6 in Figure 3) before quality
and validation checks can be undertaken in
the repository. Most insurers have an existing
extract, transform, and load tool (2) for this
purpose. Importantly, the repository varies
from a typical insurance database in a number
of ways primarily in terms of the level of
granularity and ability to store complex results
such a cash flows.

comparison of the different (pre-run)

scenarios against each other or the base
scenario (e.g., compare the impact of an
extra 5% new business growth on profitability
and solvency).
Present both point-in-time and forwardlooking analytical management information.
4. Consolidation engine
Insurance companies are complex entities,
necessitating a way to easily consolidate all
the data from various sources. Thus, a key
component of the IAP is a consolidation engine.
In essence, a consolidation engine provides
a mechanism for mapping the underlying
analytical data onto the organizational structure
of the business. The engine consolidates the
granular results to present an enterprise view.
This aligns the data model to the business and
supports effective drill-down analysis.
5. Forward-looking capability
As we have already alluded to, one of the most

Generating the physical reports and dashboards

difficult challenges is projecting forward key

querying the repository, organizing the data

logically, and rendering the output in the
selected format. This is typically facilitated
by what are termed On Line Analytical
Programming cubes, which are multidimensional views of data held in the repository.
They enable high levels of granularity and
provide drill-through paths.
3. Senior management dashboards
Outputs can be generated in a variety of
formats, typically reports, spreadsheets, and
dashboards. From the perspective of decisionmakers, interactive dashboards are particularly
valuable. Such dashboards should focus on the
analytical information/metrics that are used to
manage the business and make decisions.
Provide drill-down analyses from the highlevel business metrics, offering different levels
of aggregation, drill-through, and granularity.

MOODYS ANALYTICS RISK PERSPECTIVES

functional area needs.

Provide a what-if interface to enable

2. Reporting engine
requires a reporting engine (4) capable of

22

Generate tailored views of the analytical

metrics and analytical information for strategic

purposes. Most firms have some forward-looking
capability, especially to meet the needs of ORSA
under Solvency II. The main problem is that
most insurers have not invested in the endto-end infrastructure to support the efficient
production of multi-year results across a range
of scenarios.
Given what we have seen in the banking sector
with multi-year stress testing, we expect this
will be an area that insurance companies will
increasingly look at in the coming years, which
would naturally integrate with the IAP. Even
where there is still a heavy reliance on existing
capabilities with use of spreadsheets and manual
processes, these capabilities can be integrated
into the IAP and also used to help support better
data management.
We believe that it should be possible to run a
pre-defined set of scenarios and store the

RETHINKING DATA MANAGEMENT

results in an analytical repository. This means

To meet the information challenge, firms need

that within defined parameters the CRO/CFO/

to have a clear vision of the enterprise metrics

CEOs would have real-time access to a range

required to support their business, and adopt

of results via interactive dashboards. If the

a top-down approach to ensure appropriate

information required were to be beyond the

focus on the delivery of business benefits.

scope of the pre-defined scenarios, the models

would have to be re-run.

An IAP can help firms implement their

vision. The end capability should be a flexible

More generally, we believe that insurers will

dashboard that focuses on key business

use proxy techniques to enable scenarios

metrics, can be tailored to address the

to be run more quickly without relying on

needs of different stakeholders within the

individual business units to produce results. The

organization, and provides drill-down analysis.

benefits gained through speed, accessibility,

The analytical data repository can leverage

and centralization can easily offset a reduction

the important source data via a robust

in accuracy, provided the results are good

data model designed to support the

enough. Creating a forward-looking projection

dashboards capabilities.

is a complex process and a detailed analysis is

beyond the scope of this article.

Projecting balance sheets, capital, and profits

by scenario in a timely manner to support the

Conclusion

forward-looking metrics requires significant

There is little doubt that insurers exist in

investment. However, the IAP can produce

an increasingly competitive and attritive

outputs from the manual processes that are

environment. The ability to quickly make

currently in place at most organizations. As

informed business decisions based on accurate

these processes become more streamlined,

historic and forward-looking information is

the platform must be flexible enough to cope

crucial, but that information is difficult to collate

with the changes.

as it is spread across a plethora of systems.

RISK DATA MANAGEMENT | AUGUST 2015

23

GETTING HUMAN DATA RIGHT: THE HIDDEN

ADVANTAGE
By Kevin Hadlock

Kevin Hadlock
Senior Director

Kevin is a Global eLearning Solution Specialist.

He has designed numerous distance-learning and
web-based training programs, developed and taught
many seminars, webinars, and full, blended training
programs, and has authored content used globally by
thousands of credit trainees.

With their focus on profit margins, data and risk management, and
compliance with an increasing number of regulations, financial
institutions often pay insufficient attention to the human side of
their operations. This article addresses that deficiency and explains
the sea change taking place in how risk professionals acquire human
data the quantifiable ability of employees to do their jobs well.
Risk professionals are prone to mismanaging

So when profits decline and budgets tighten,

the risks and rewards associated with employee

training gets whacked and human data suffers.

knowledge and skill levels. Improving, capturing,

As a result, performance deteriorates and risk

and driving this critical but frequently

increases. Right when the need is greatest to

forgotten data can help institutions gain a

minimize risk and maximize reward, the

competitive advantage.

needle usually gets pushed in precisely the

Knowledge and skill: The data risk

professionals forget

wrong direction!
Human beings: Great risk, great reward

After years spent working with many banks

Employees are a financial institutions double-

around the world in the credit and financial

edged sword: They represent its single greatest

sector, I have observed that few institutions

source of risk and its most profound opportunity

engage in persistent human data quantification

for reward. In spite of all an organization may

and management. They may conduct the

do to comply with regulations, establish and

occasional gap analysis, sometimes followed

monitor sound policies and procedures, and

by a burst of training activity, but the discipline

enhance its image in the marketplace, a wrong

of continuously gauging employee knowledge

or fraudulent decision by just one employee

and skills and then optimally providing training

can undo the millions of dollars and countless

and access to self-directed and socially-driven

hours invested in optimizing business prospects.

learning is almost nonexistent.

For example, a loan officer who fails to notice
This lack of an effective ongoing process leads

key credit risks may grant a loan that quickly

inevitably to negative outcomes, such as

goes into default, costing the bank several

costly one-time training ventures (where the

million dollars in lost principal and interest

training ages quickly and is too expensive to

and causing serious reputational damage. The

refresh on the fly), employees left to grasp at

profitability of dozens or even hundreds of

every informal learning source they can find

performing loans can essentially be nullified in

(frequently Wikipedia or Google), and C-level

the process.

executives who cannot see the value in spending

on training because they do not believe investing

Conversely, an astute credit analyst may

in their employees will reap rewards in kind.

determine that a borrower has upside that isnt

readily apparent and so advises decision-makers.

24

MOODYS ANALYTICS RISK PERSPECTIVES

RETHINKING DATA MANAGEMENT

The result could be the granting of a loan that

leads to a highly profitable and expanding longterm relationship that touches many areas of the

comfort have changed, often dramatically.

An example of outdated human data

bank in a positive way.

So how could this scenario play out? The poor

I have experienced both situations and have

accounting degree and, upon joining the bank, is

seen the positive and negative effects on

trained in the classic way: months of classroom

whole communities. Credit professionals in

and workbook training, followed by placement

each instance had vast amounts of data and

in a division loan center under a manager who

transparency at their fingertips. Knowing and

is a 25-year veteran. The training focuses largely

applying sound credit and risk principles, or

on EBITDA as a primary determinant of borrower

failing to do so, were the deciding factors.

creditworthiness, a view reinforced by the

Knowledge and skill: Human data that matters

So, how do credit-granting organizations
minimize the downside of employee risk,
while maximizing the upside? Is it enough to
focus solely on optimizing systems or finetuning policies? Does compliance with all the
regulations and risk management regimes put
forth by all the governments and supremely
qualified boards in the world eliminate risk
once and for all? Does best practice data

loan officer from my earlier example has an

employees manager. On the job, the new officer

is required to use a recently licensed spreadsheet
application to capture borrower financial
information and generate analysis-enabling
reports. He notices that one such piece of output
is a Uniform Credit Analysis (UCA) cash flow
report and asks his manager if it has merit. The
manager responds that she is not familiar with
the report and that it appears to have little value
in any case, so he disregards it.

management solve all ills? As much as these

Shortly thereafter, a large existing borrower

practices might help, the answer to each of

requests a renewal on a long-term line of credit,

these questions, is no as long as people

and the new officer is tasked with conducting a

are involved.

credit analysis and making a recommendation.

The probability of errors creeping into an employees work output rises at

an increasing rate the farther he or she gets from fresh, relevant human
data. Knowledge and skills age so rapidly that the likelihood of employee
error approaches 100% by the end of the five-year period.
All banks train their employees to one degree or

Using all his training and his managers guidance,

another. What too often gets left out, however,

he focuses squarely on EBITDA, failing to

is a refined, dynamic focus on knowledge and

notice that the companys performance varies

skill that are core to what I call human data.

significantly from year to year. He doesnt

Organizations tend to focus on performance,

realize that, while useful in assessing risk for

which is obviously appropriate. But it is the

generally stable companies, EBITDA often fails

rare institution that appreciates the speed of

to capture the critical nuances of enterprises in

change taking place in its employees areas

flux. So, seeing profits still in the black, he grants

of expertise.

the large renewal request (with his managers

Performance consultant and learning expert

Jane Hart notes that, the half-life of a given
stock or skill is constantly shrinking and is now
around five years.1 In practical terms, this means
that by the time an employee is seasoned and
comfortable, the definitions of seasoning and

approval), not noticing that true cash flow has

become increasingly negative for the past three
years, although he has projected it to remain
positive for the foreseeable future. Within a year,
the loan goes bad. The loan officer is now at a
loss, wondering why it happened and how he
might have predicted the failure.

RISK DATA MANAGEMENT | AUGUST 2015

25

The problem in this example is not that the bank

the higher the likelihood he or she will make

didnt invest in the new loan officers human

meaningful errors.

data, but that there was no means in place to

Figure 1 isnt defensible on a purely empirical

keep that data fresh and up to date. Both his

formal training and his managers guidance were
grounded in established analytical principals,

basis, but rather is meant to be illustrative.

I could have used any time frame for the
Distance from Professional Currency axis;

but failed to take into account what for them

I chose to use Harts five-year skills half-life

was an emerging analytical technique cash

figure, as it has researched relevance. Using a

flow analysis that would have been greatly

facilitated by the introduction of the automated
spreadsheet tool. Upon an exhaustive debrief,

straight percent scale for the Risk of Meaningful

Error axis and then assigning specific
percentages at one-year intervals is likewise

key decision makers realized that staff had

unscientific. I have drawn the progression simply

bypassed information that had been at their

fingertips. They modified the policy to require an
analysis of cash flow for all commercial loans in
the future.

to emphasize that the probability of errors

creeping into an employees work output rises at
an increasing rate the farther he or she gets from
fresh, relevant human data. Knowledge and skills

This is a true story that illustrates that human

age so rapidly that the likelihood of employee

data must be updated constantly. Furthermore,

error approaches 100% by the end of the five-

systems or processes must be put in place to

year period.

ensure that knowledge and skills are readily

updateable. In other words, professional
currency cannot be an accident. If the elevation
or adjustment of human data is left to chance,
risk will outweigh reward and the likelihood of
costly mistakes will increase.

How todays risk professionals can elevate their

own human data
Interestingly, employees themselves often
best recognize both the need for professional
currency and the implications of not having
it. They sense the urgency of having up-to-

Figure 1 illustrates that the farther away an

date knowledge and skill more than their

employee is from having current human data,

organizations do because they are the ones who

Figure 1 The relationship of professional currency to risk

100

Risk of Meaningful Error (%)

80

60

40

20

Distance from Professional Currency (Years)

Source: Moodys Analytics

26

MOODYS ANALYTICS RISK PERSPECTIVES

RETHINKING DATA MANAGEMENT

have to keep their jobs and advance their careers.

Learning and performance ecosystems

enhance performance

For them, the stakes are both high and personal.

This article is not arguing against formal training

On a related note, recent surveys and workplace

provided by the institution. Quite the contrary,

studies show that, although formal training

it remains a primary and foundational way for an

provided by the institution is useful, it provides

organization to communicate its way of doing

only a fraction of the ongoing learning that

business to its staff and will always have a role

employees need and is thus at the low end of

in workplace learning and performance. But the

the scale in terms of human data value-added.2

trends and facts identified in Table 1 actually

Additional insight provided by Hart is again

make a great deal of sense from the business

useful.3 Her research into how employees learn

side. Organizations simply do not have the

what they need to do their jobs in other words,

budgets or dexterity to keep pace with every

to maintain their professional currency is

change and nuance in an increasingly dynamic

summarized in Table 1. The findings are listed in

business world and then to nimbly and fully

order of descending usefulness.

communicate them to every staff member.

This would require expense and administrative

Although still important, formal training is at

overhead that virtually no company could

the bottom of the list for the approximately 700

efficiently take on.

survey respondents. Other means of acquiring

current human data score higher, and most of

What financial institutions can do, however,

them are either employee-directed or socially

is a much better job of creating structured

sourced. This is a multi-year trend that reflects

but open environments that combine formal

the current and ongoing reality of business.

training with self-guided and social learning,

The advent of technology-enabled access to

so that professional currency is optimized

knowledge on virtually every topic, as well as

rather than achieved coincidentally or, worse,

to other people via social networks and

accidentally.

forums, has put the means of human data

enhancement squarely in the hands of the

Perhaps the most promising approach to

employees themselves.

installing such environments is a construct

Table 1 2013 Learning in the Workplace survey results

Not
Important

Quite
Important

Very
Important

Essential

V. Imp. &
Essential

Knowledge sharing within

your team

3%

12%

30%

55%

85%

Web search for resources

(e.g., Google)

2%

17%

32%

49%

81%

Conversations and meetings

with people

2%

19%

40%

39%

79%

Personal and professional

networks and communities

3%

22%

35%

40%

75%

External blog and news feeds

8%

22%

40%

30%

70%

Content curated from

external sources

9%

29%

39%

23%

62%

Self-directed study of
external courses

14%

33%

35%

18%

53%

Internal job aids

20%

37%

26%

17%

43%

Internal company documents

13%

44%

29%

14%

43%

Company training/e-learning

25%

42%

20%

13%

33%

Source: Jane Hart Blog

RISK DATA MANAGEMENT | AUGUST 2015

27

known as a learning and performance

integrated and all-encompassing approach to

ecosystem. In their white paper on this subject,

employee training is paramount if business

Marc J. Rosenberg and Steve Foreman make the

enterprises are to compete well and survive. In

case that we must move away from individual,

other words, six weeks, or even six months, of

siloed, one-off [learning] solutions to an

new hire training alone doesnt cut it anymore

ecosystem comprised of multi-faceted learning

if it ever did.

Although they dont say it in as many words, what Rosenberg and Foreman
suggest is that optimized human data is so critical, and its insufficiency
so pervasive, that a new, more integrated and all-encompassing approach
to employee training is paramount if business enterprises are to compete
well and survive. In other words, six weeks, or even six months, of new hire
training alone doesnt cut it anymore if it ever did.
and performance options that enhance the

Once employees are turned loose in the

environments in which we work and learn.4

workplace, having a more thoughtful, dynamic

They define learning and performance

ecosystems as structures that strengthen
individual and organizational effectiveness
by connecting people, and supporting them
with a broad range of content, processes, and
technologies to drive performance. They address
six primary components of these ecosystems and
depict them in an interrelated way:
1. Talent management
2. Performance support
3. Knowledge management
4. Access to experts
5. Social networking and collaboration
6. Structured learning

fail to do so fall behind, sometimes quickly. The

results then show up in falling bottom lines
and, in the case of credit-granting organizations,
in decreasing credit quality and loan losses.
Quantifying human data as a step in
managing risk
If you work in credit and risk long enough,
you begin to see everything in numbers. You
start to believe that life must be quantified to
be understood. Thankfully, there are ways to
quantify, if imperfectly, human data.
In his report on tracking the knowledge and
skills of credit professionals, People Risk:
Improving Decision Making in Commercial
Lending,5 Ari Lehavi of Moodys Analytics

ecosystem than training. At the heart of it

explains an exam-based methodology for

all is human data the knowledge and skills

collecting metrics on human data in the area of

employees need to do their jobs effectively.

fundamental credit risk assessment. He shares

That data comes from structured learning, social

critical empirical details and broad results,

networking and collaboration, access to experts,

all of which shed light on the strengths and

and effective performance support systems.

weaknesses exhibited by lenders, analysts

It is managed, optimized, and applied most

and relationship managers at banks around

effectively and broadly over time through sound

the globe. He further breaks this information

talent and knowledge management schemes.

down geographically and by subject matter

what Rosenberg and Foreman suggest is that

MOODYS ANALYTICS RISK PERSPECTIVES

their knowledge and skills. Organizations that

There is more to a learning and performance

Although they dont say it in as many words,

28

approach in place will be critical to maintaining

(i.e., financial risk, marketplace/industry risk,

management risk, and risk mitigation).

optimized human data is so critical, and its

The most salient feature of all of these details

insufficiency so pervasive, that a new, more

about human data is that it is clearly actionable.

RETHINKING DATA MANAGEMENT

In other words, evaluating it accurately can lead

is that human data can be quantified and

to direct remediation that shores up weak areas

improved. And institutions that engage in this

and demonstrably elevates the quality of that

type of process, effectively and consistently, gain

human data. Among the reports key findings are

a current and highly useful sense of the level of

the following:

human data in the organization, both individually

The average test score across subjects

and collectively.

included in the assessment exceeded the

Investing in individuals rewards the organization

minimum subjective pass threshold by a mere

Institutions that have the foresight and will to

2%.

implement integrated learning and performance

Financial risk had the weakest relative score.

ecosystems or critical components thereof

Approximately 42% of people answered

in the near term will have the advantage

fewer than 70% of the questions correctly in

over both the medium and long terms. There

this critical subject area.

is no one size fits all answer to this, but an

Major banks around the world showed a wide

abiding appreciation for the essential nature

disparity in test performance across all areas

and inherent worth of human data, and the

of risk.

criticality of continuously optimizing it, is the

An institutions aggregate skill and knowledge

test performance correlated highly with its
relative default risk. Although there may be
other contributing factors, the lower the
banks average score, the higher its relative
default risk, as measured by Moody's Baseline
Credit Assessment rating.

foundation on which to build.

Organizations that grasp this, and then make
well-considered efforts to go beyond providing
formal training to creating a permanent learningis-performing environment subscribed to and
supported by all levels of the organization will
swing the risk/reward pendulum inexorably

This last finding is particularly enlightening

toward the reward side. This, in turn, will unlock

and reinforces the first proposition in this article

human potential and corporate profits at an

that subpar human data contributes to higher

increasing rate. Thus is the power of human data

risk in a credit-granting organization. However,

and the reason for giving it its due.

perhaps our key takeaway from Lehavis report

Jane Hart, Social Learning Handbook 2014, page 20, 2014.

Don Taylor, What will be big in workplace learning in 2015?, January 7, 2015.

Allison Rosset, Trending in Workplace Learning 2015, January 13, 2015.

3 Jane Hart Blog, April 22, 2013, [Link]

4 Marc J. Rosenberg and Steve Foreman, The eLearning Guild, Learning and Performance Ecosystems, December 2014.
5 Ari Lehavi, Moodys Analytics, People Risk: Improving Decision Making in Commercial Lending, November 18, 2014.

RISK DATA MANAGEMENT | AUGUST 2015

29

REGULATORY
SPOTLIGHT
Looks at how risk data management will impact financial institutions
preparation for regulations, including regulatory big data initiatives, PPNR
stress testing, and IFRS 9.

REGULATORY BIG DATA: REGULATOR GOALS

AND GLOBAL INITIATIVES
By Michael van Steen

Michael van Steen

Senior Director, Enterprise Risk
Solutions

Michael helps deliver advanced portfolio credit risk,

stress testing, correlation, and valuations solutions to
global banks, asset managers, hedge funds, insurance
companies, and regulatory organizations.

Big data isnt just for Silicon Valley. This article discusses the trend
of large data set capture and analysis by regulators, referred to
here as regulatory big data, by detailing the motivations and
goals of regulators and examining three significant regulatory big
data initiatives: AnaCredit in the European Union (EU), FDSF in the
UK, and FR Y-14M in the United States. It then analyzes how these
efforts complement other significant, and concurrent, regulatory
reporting and IT efforts.
A new big data paradigm emerges

supermarket chain, and a hospitals analysis of

Financial institutions worldwide are facing

patient treatment plans to determine ways to

an increased level of regulatory scrutiny

lower readmission rates.

in the aftermath of the Great Recession.

Regulatory efforts entail new and expanded
reports, templates, and calculations from
financial institutions that are often little more
than detailed presentations of information
summarized in existing regulatory disclosures
such as call reports and loan loss summaries.
Many institutions view them as just another
mile on the seemingly endless produce another
report treadmill one that significantly
increases their compliance costs.

This data generally lacks the well-defined

structure that would allow it to fit neatly
into standard relationship database tables.
Additionally, the often multi-terabyte to
petabyte size of this data is quite challenging
for a typical relational database management
system. As a result, technology companies
have developed a wide variety of new software
frameworks MapReduce, Apache Hadoop, and
NoSQL, to name a few to analyze and process
these big data repositories, often in parallel with

However, a new paradigm regulatory big data traditional database tools.

is emerging worldwide in which an institution
provides its bulk underlying granular data to
regulators for their regulatory, risk assessment,
and stress testing efforts.

32

MOODYS ANALYTICS RISK PERSPECTIVES

Regulatory big data is a catch-all term (like big

data) to describe the capture and processing of
larger sets of regulatory data via the use of both
traditional and newly developed data processing

What is regulatory big data?

tools. Although this data isnt of the same

In the technology world, big data generally

magnitude as, for example, a record of clicks on

refers to an amount of data (both structured

a popular website, its volume is much greater

and unstructured) so vast that analysis of it

than the summary regulatory reports submitted

requires new processing techniques. Examples

today. Big data tools and techniques can help

of big data include a major search engines index

regulators process the large amounts of data

of all the web pages it has crawled, the monthly

produced by financial institutions for oversight

cash register transaction receipts at a national

and compliance purposes.

REGULATORY SPOTLIGHT

Regulatory big data: Regulator goals and uses

than ever before, oftentimes down to individual

Regulators aim to use big data sets to

loans. They are also pushing for consistent

complement traditional banking supervision,

identification and organizing of counterparties

help maintain financial stability, and establish

through initiatives like the Legal Entity

monetary policy regulation procedures.

Identification (LEI) effort and more frequent

Proposals and rules from regulatory bodies such

(e.g., monthly instead of quarterly or annually)

as the Basel Committee on Banking Supervision

data submissions.

(BCBS), the Bank of England, and the Federal

Reserve (Fed), as well as speeches and articles on
the topic by regulators, have elucidated a variety
of goals for regulatory big data, among them:1

This new data would provide regulators with the

ability to:
Build a foundation for more comprehensive
and frequent institutional stress testing

Producing a rich regulatory dataset with

Measure institutional and obligor

finer granularity than current reports provide

interconnectedness and contagion risk,

Eliminating real or perceived data gaps in

by looking at obligors common to

existing reporting by requiring more frequent

multiple institutions

and less aggregated data submissions

Conduct micro- and macro-prudential risk

Reaping the benefits of advanced big

analysis, enhanced by the ability to assess

data technology without the need to use

the risk of obligors across firms and the

proprietary technology from big data firms

Aggregating an obligors total indebtedness
across multiple institutions

movement of risk factors over time

In summary, the increased quantity, and the

Establishing a tools and data submissions

likely more frequent submission, of data for

framework to provide a comprehensive

regulatory big data initiatives presents an

overview of banks and their obligors at the

opportunity for regulators to enhance their

push of a button

understanding of both the institutions they

regulate and the credit exposures of individual

To start the process of working toward these

goals, various regulatory big data initiatives have
been advocating, if not mandating, that firms

obligors across institutions.

In the following sections, we provide information

provide a much finer level of data granularity

on three current regulatory big data initiatives:

Figure 1 Regulatory big data

Financial exposures
Monthly need
amount likely
loan
individual
loans
submission challenges
extensive initiative banks
Frequent new
obligators measures stress
FDSF analytic
technology
national United summary
reports another
credit
provide
efforts sets
given
large other default
use while like set
all tools
risks States institution EU UK capital
Fed calculations
loss just Bank
dataset Bank granular
Y-14M Including
lines
several
existing
systems
initiatives subject both level
goals across workwithin
developed
example
risk
portfolio moremuch framework techniques testing
statistical database submissions requited
much information
include detailed reporting under
AnaCredit
analysis
requires

regulators

Big

instrctions

data

regulatory
Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

33

Analytic Credit Dataset (AnaCredit) in the

Eurosystem NCBs, and (b) a common granular

European Union (EU); Firm Data Submission

credit database shared between the Eurosystem

Framework (FDSF) in the UK; and Capital

members and comprising granular credit data for

Assessments and Stress Testing (FR Y-14M)

all Member States whose currency is the euro.

in the US. We discuss the features of these

initiatives, as well as the regulatory motivation
and, critically, the effect on firms subject to
these rules.

The ECB intends to leverage existing national

CCRs as a foundation for AnaCredit, but a
large amount of work still needs to be done.
For example, the ECB must roll out specific

Regulatory big data: AnaCredit (EU)

AnaCredit regulations, which are not expected

AnaCredit is an EU-proposed Central Credit

until the second quarter of 2015 at the earliest.

Register (CCR) of, initially, loans to non-financial

corporations. AnaCredits aim is to build up and
link existing national CCRs to a Europe-wide
credit data repository accessible by the
European Central Bank (ECB) and other European
nations central banks. Essentially, AnaCredit
mandates the collection of granular data from
financial institutions.
Only countries subject to the ECB (i.e., eurodenominated countries) are considering this
initiative, so non-euro zone countries such as the
UK are not taking part. AnaCredits rollout will
initially apply to an estimated 3,500 banks in the

Though they are not the final rules for AnaCredit,

the ECB/2014/6s prepare for AnaCredit
regulations shed some light on both the work
needed and the likely capabilities of this system.
The preparatory work mandated by ECB/2014/6
includes:
Identifying relevant end-user needs
Defining the granular credit data sets that will
be collected and linked
Developing a way to transmit granular credit
data securely
Developing detailed operational arrangements,

The increased quantity, and the likely more frequent submission, of data
for regulatory big data initiatives presents an opportunity for regulators to
enhance their understanding of both the institutions they regulate and the
credit exposures of individual obligors across institutions.
EU. Other lenders, including non-EU institutions
operating in the EU, do not fall under the
initial scope.
Implementation timelines have shifted since
AnaCredit was formally introduced through
the ECB decision (ECB/2014/6) on February
24, 2014;2 the latest estimate is for a phased
implementation starting in January 2018.
The objectives of ECB/2014/6 are to define:

deliverables and for monitoring progress

Addressing confidentiality, use of data, and
governance
Looking at existing European credit registers and
the overall goals of these regulatory big data
initiatives, we think it likely that, ultimately, the
AnaCredit system will include:
Data on obligors, including unique identifiers
in particular, the identifiers being developed

in a stepwise manner a long-term framework

as part of LEI enabling the linking of obligors

for the collection of granular credit data

across institutions

requirements. This long- term framework

shall include by the end of 2016: (a) national
granular credit databases operated by all

MOODYS ANALYTICS RISK PERSPECTIVES

Establishing a timetable for specific steps and

preparatory measures necessary to establish

based on harmonized ECB statistical reporting

34

given the sensitivity of the data

Amount of assets, financial derivatives, and

certain off-balance sheet items
Loan IDs, inception and maturity dates,
interest rates, and any financial guarantees

REGULATORY SPOTLIGHT

tied to loans
Analytic measurements such as loan
performance data, borrower probability
of default (PD), and exposure Loss Given
Default (LGD) estimates

scenarios, in a wide variety of stress calculations.

The FDSF requires a level of granular data and
analysis far in excess of the typical reports
submitted to UK regulators. The intense scrutiny
and significance of this exercise (which, in the

From our analysis of similar initiatives and the

event of unsatisfactory results, could prompt the

preparatory work involved, we expect the ECB

regulator to prohibit capital payouts) means that

will consider between 24 and 150 or more

each institution requires extensive audit trails,

attributes per loan for inclusion in this initiative,

detailed documentation of assumptions, full

although the exact number and composition

traceability of calculations, and the analysis of

have not yet been determined. There are several

many scenarios.

other unknowns, including:

The lower reporting threshold specifically,

As with other regulatory big data initiatives, the

data required for FDSF would be sourced from a

the euro level of individual loans that do not

variety of internal areas (e.g., individual business

need to be reported

units, each likely having multiple products and

The schedule of assets that need to be reported

associated accounting systems and assumptions)

The reporting schedule for institutions,

in myriad formats. Assembling this data is not as

particularly foreign institutions operating in

simple as appending rows to an existing table,

the EU and non-bank financial companies

however; reporting date time gaps and missing

The resulting analytic dataset although

the exact composition, rollout schedule, and
asset mix are still undecided will provide
the ECB and national central banks with a
comprehensive view of loan exposures in an
institution and of obligors across institutions.
Once assembled, AnaCredits large granular
dataset will allow regulators a view into the
institutions they regulate that is not currently
available with summary-level reports.
Regulatory big data: Firm Data
Submission Framework (UK)
The Firm Data Submission Framework (FDSF),
another example of a regulatory big data
project, is a quarterly granular reporting
requirement from the Bank of Englands
Prudential Regulation Authority (PRA). FDSF
applies to the UKs eight Systematically
Important Financial Institutions (SIFIs) and was

data are likely, as is the need to try out various

assumptions on key parameters such as expected
losses and probabilities of default. Moreover, the
calculated stress results are also likely to vary
significantly by product line and geography a
downturn in UK property prices, for example, will
likely have a significantly different impact on a
Scottish residential mortgage portfolio than on a
Greek shipping loan book.
An additional layer of complexity arises from the
potential need to map internal data structures
to those defined by the PRA. Although a banks
systems may have just a few occurrences of
key attributes as exposure, an institution has
to reconcile all of them with the PRAs precise
definition. An extensive audit trail is also
necessary so an institution can drill down and
aggregate the data as needed or requested by
the PRA.

developed to provide quantitative, forward-

As with AnaCredit, the FDSF rules define

looking assessments of the capital adequacy

regulatory big data according to the

of the UK banking system and the individual

following parameters:

institutions in it.
Like the Feds Comprehensive Capital Analysis

An extensive amount of data

Data that typically needs extensive mapping,

and Review (CCAR), on which it is loosely based,

aggregating, and cleaning prior to submission

FDSF requires that institutions collect data from

for use in stress testing calculations

all of their significant operating units and use

this data, based on PRA guidance and stress

The use of the data is quite complicated, as

RISK DATA MANAGEMENT | AUGUST 2015

35

multiple, iterative stress tests are typically

migrate out of the portfolio (e.g., are paid off

required of the banks subject to this regulation.

or have defaulted), as well as information to

In practice, however, the ability of big data tools

facilitate address-matching across loans and

to handle large datasets of partially structured

portfolio-level information.

data (i.e., not fully contained in clean relational

database tables) facilitates the FDSF initiatives.
Banks can use flat files of raw data from differing
systems, scripting languages, and statistical
packages to manage this large load of data
and complicated analytic requirements, while
maintaining a strict data quality regime.

require 137 lines of data per loan that includes a

wide range of data elements, such as origination
credit score, current credit score, probability
of default, mortgage insurer, valuation at
origination and at current time, foreclosure
status, and both actual (if any) loss given

The FR Y-14M reporting program is a regulatory

default and expected loss given default.

States. Under FR Y-14M, bank holding

companies with consolidated assets of $50
billion or more must submit detailed home
equity and credit card loan data, along with
portfolio and address data, to the Fed on the
last business day of each month.
In contrast to the existing and predominantly
summary and aggregated reporting required of
banks and bank holding companies, FR Y-14M is
a true regulatory big data program, as it requires
reporting of all portfolio and service loans
and lines in several broad portfolio categories
every month. The FR Y-14M initiative, like the
AnaCredit and FDSF regulations, aims to furnish
regulators with the tools and data necessary to
monitor very granular risk in a timely, nearcontinuous fashion.
Institutions subject to these requirements
must contend with several complex technical
challenges, given the significant amounts of
sensitive data they have to prepare and submit
every month. The Fed requires that bank holding
companies subject to these rules report all of the
following active and serviced lines and loans in
its portfolio:
Revolving, open-end loans secured by one

The Fed uses the data collected under FR

Y-14M for a wide range of regulatory purposes,
including assessing a banks capital adequacy,
supporting periodic supervisory stress tests,
and even enforcing Dodd-Frank consumer
production measures.
The sheer volume of information banks must
provide every month, the detailed and sensitive
nature of data collected, and the disparate
uses of this data everything from consumer
protection to stress testing all present new
technical challenges for banks and regulators.
Because the Feds analysis of a banks FR Y-14M
data could result in regulatory actions with
a material impact (e.g., restricting dividend
payouts), banks must take extreme care to
ensure that this data is accurate, timely,
and auditable.
Complementary efforts
Moodys Analytics research reveals that
to meet these challenges, institutions are
building complementary processes outside
their traditional credit-relational database
management systems and using new big
data analysis and formatting tools. They seek
to separate their large-scale data processing
efforts from other reporting and analysis of

to four family residential properties and

the database, all while maintaining a clear and

extended lines of credit

auditable record of data submissions.

Junior-lien closed-end loans secured by one to

four family residential properties
Domestic credit cards

MOODYS ANALYTICS RISK PERSPECTIVES

for each loan. For example, domestic first liens

Regulatory big data: FRY Y-14M Reporting (US)

big data initiative by the Fed in the United

36

An extensive amount of information is required

As an illustration of the tools and technology

available today, one institution implemented
a single, large monthly data extraction

Additionally, banks have to report detailed

script to move masses of raw data (i.e., in a

information on previously reported loans that

non-submission format and missing several

REGULATORY SPOTLIGHT

calculations) from its database to a large set of

regulatory bodies, the DGI will have a

flat files, then relied on an open source statistical

profound effect on individual firms, given

package to clean up and append the data to a

that it calls for standardized reporting

large statistical data file. The institution then

templates for large international exposures

used another set of procedures in the statistical

and an overall higher level of reporting data

program to extract and format the data for its

granularity from most financial firms.

monthly submissions and to build a detailed

log of the preparation of the data submitted.
In addition, the regulators themselves are also
launching several complementary efforts that
are transforming banks data handling and
reporting techniques.

The common theme of these initiatives is that

financial institutions have to produce much
larger volumes of data in a more consistent
and controlled way. Organizations must have
the infrastructure and the skills in place to

Because the Feds analysis of a banks FR Y-14M data could result in

regulatory actions with a material impact (e.g., restricting dividend
payouts), banks must take extreme care to ensure that this data is accurate,
timely, and auditable.
The Global Financial Markets Association
(GFMA), among others, is coordinating the
LEI initiative, wherein each single legal entity
is assigned a unique ID. LEIs will facilitate
aggregation of an obligors exposures across
institutions and easy analysis of an entitys
exposures within an organization, as well
as the use of external data on an obligor to
supplement the data an institution might
have.
The Bank for International Settlements (BIS)
has produced Principles for effective risk
data aggregation and risk reporting (BCBS
239 publication) that comments on, among
other areas, the risk infrastructure and risk
aggregation methods of larger banks. As the
publication shows, regulators recognize how

consistently produce and submit large data

sets of critical information to their regulators.
The age of regulatory big data is here
AnaCredit, FDSF, and FR Y-14M are the first of
what look to be numerous efforts by regulators
to capture more granular data much more
frequently, from the firms they regulate.
This emphasis on raw data places considerable
technical and operational burdens on
institutions. Regulatory reporting will no
longer comprise an Excel file or two emailed
every quarter but, rather, an extensive process
of assembling highly sensitive data, which
regulators can then use for critical tasks like
approving a banks dividend policy.

critical risk infrastructure and technology are

Institutions need to take a fresh look at their

at large financial institutions.

data handling, reporting, technology, and

The G20s Data Gaps Initiative (DGI) is a

security architecture to ensure that they meet

set of 20 recommendations for enhancing

these significant new challenges. The emergence

economic and financial statistics, covering

of big data tools and technologies, many of

broad topic areas such as Monitoring Risk

which are open source, can help institutions

in the Financial Sector and Financial

achieve compliance.

Datasets. Although aimed primarily at

1

Michael Ritter, Deutsche Bundesbank, Chair of the ESCB Working Group on Credit Registers, Central Credit Registers (CCRs) as
a Multi Purpose Tool to close Data Gaps, May 2014.

Anne Le Lorier, Deputy Governor Banque de France, Seventh ECB Statistics Conference, Towards the banking Union:
Opportunities and challenges for statistics, October 2014

Decision of the European Central Bank of 24 February 2014 on the organisation of preparatory measures for the collection of
granular credit data by the European System of Central Banks (ECB/2014/6)

RISK DATA MANAGEMENT | AUGUST 2015

37

IFRS 9 WILL SIGNIFICANTLY IMPACT BANKS

PROVISIONS AND FINANCIAL STATEMENTS
By Cayetano Gea-Carrasco

Cayetano Gea-Carrasco
Head of Stress Testing Services
and Advisory

Cayetano works with financial institutions on credit

portfolio management across asset classes, derivatives
pricing, CVA / Counterparty Credit Risk analytics,
stress testing, and liquidity management.

International Financial Reporting Standard 9 (IFRS 9) will soon

replace International Accounting Standard 39 (IAS 39). The
change will materially influence banks financial statements, with
impairment calculations affected most. IFRS 9 will cover financial
institutions across Europe, the Middle East, Asia, Africa, and Oceania.
IFRS 9 will align measurement of financial assets

evolving nature of risk in the banking business.

with the banks business model, contractual

In the end, a thoughtful, repeatable, consistent

cash flow of instruments, and future economic

capital planning and impairment analysis should

scenarios. In addition, the IFRS 9 provision

lead to a more sound, lower-risk banking system

framework will make banks evaluate how

with more efficient banks and better allocation

economic and credit changes will alter their

of capital.

business models, portfolios, capital, and the

provision levels under various scenarios.

9, we conducted the Moody's Analytics 2015 IFRS

classification, measurement, and impairment

9 Survey to give practitioners a snapshot of the

calculation and reporting, banks should expect

"current state" of the industry. Moody's Analytics

to be required to make some changes to the way

has also included a series of comments on best

they do business, allocate capital, and manage

practices and industry trends.

Banks will face modeling, data, reporting, and

infrastructure challenges in terms of both:
1. Reassessing the granularity (e.g., facilitylevel provisioning analysis) and/or credit
loss impairment modeling approach (e.g.,
consistency regarding the definition of default
between Basel and IFRS 9 models).
2. Enhancing coordination across their finance,
risk, and business units.
Effectively addressing these challenges will
enable bank boards and senior management
to make better-informed decisions, proactively
manage provisions and effects on capital plans,
make forward-looking strategic decisions for
risk mitigation in the event of actual stressed
conditions, and help in understanding the

MOODYS ANALYTICS RISK PERSPECTIVES

financial institutions when transitioning to IFRS

Given the IFRS 9 requirements in terms of

the quality of loans and provisions at origination.

38

To help minimize the challenges faced by

Survey Findings: Implications for Financial

Institutions
IFRS 9 will affect the business models, processes,
analytics, data, and systems across several
dimensions.
Capital, lending, underwriting, and origination
Provision levels are expected to substantially
increase under IFRS 9 versus IAS.
Further equity issuances may be needed,
with the potential for greater pro-cyclicality
on lending and provisioning owing to IFRS 9.
Capital levels and deal pricing will be affected
by the expected provisions, but must be
evaluated under different economic cycles
and scenarios.
Banks will have to estimate and book an
upfront, forward-looking expected loss over

REGULATORY SPOTLIGHT

the life of the financial facility and monitor

for ongoing credit-quality deterioration.
Rating and scoring systems may have to be
updated, especially for those banks without
Internal Ratings-Based (IRB) models.
Asset reclassification, reconciliation, and
measurement
Banks will need to reclassify assets and

Data requirements will increase to meet

IFRS 9-related calculations and ongoing
monitoring.
Retrieval of old portfolio data will also be
needed, especially for the transactions
originated before the A-IRB models have
been introduced.
IFRS 9 impairment calculation requires
higher volumes of data than IAS, which may

reconcile them with IAS. They will also need

substantially increase the performance and

to map products that can be categorized

computational requirements of a credit-loss

before the calculation (contractual cash

flow test) or create a workflow to capture

impairment calculation engine.

Financial reporting and reconciliation will

the purpose (business model test). An

be needed to align with other regulatory

additional effort could be required to

requirements.

identify those products that can be

considered out of scope (e.g., short-

Documentation and governance

term cash facilities and/or covenant-like

IFRS 9 makes the provisioning exercise a

facilities).
Institutions will have to align, compare, and
reconcile metrics consistently (e.g., Basel vs.

cross-functional activity, with coordination

needed across the risk, finance, accounting,
and business functions.

IFRS 9).
IFRS 9 is a Game Changer
Cross-coordination across risk, finance, and
business units

IFRS 9 is the International Accounting

Financial institutions will have to coordinate

financial crisis, aimed at improving the

Standards Boards (IASB) response to the

finance, credit, and risk resources for

accounting and reporting of financial assets

which current accounting systems are not

and liabilities. IFRS 9 replaces IAS 39 with a

equipped.

unified standard. In July 2014, IASB finalized

Credit impairment calculation and valuation

The IFRS 9 provision model will make banks

the impairment methodology for financial

assets and commitments. The mandatory
effective date for implementation is January

evaluate, at origination, how economic

1, 2018; however, the standard is available for

changes will affect their business models,

early adoption (e.g., via local endorsement

capital plans, and provisioning levels.

procedures).

A methodology to calculate a forwardlooking measurement will have to

be developed and/or updated (e.g.,
transformation from TTC to PiT), while
the cash flow valuation analysis must be
scenario-driven.
IFRS 9 will affect the existing
documentation and hedge accounting
frameworks.

IFRS 9 introduces changes across three areas

with profound implications for financial
institutions:
1. The classification and measurement of
financial assets
2. The introduction of a new expected-loss
impairment framework
3. The overhaul of hedge accounting models to

Data, systems, processes, reporting, and

automation
Systems will need to change significantly to
calculate and record changes requested by
IFRS 9 in a cost-effective, scalable way.

better align the accounting treatment with

risk management activities
Replacing IAS 39 with IFRS 9 will significantly
impact banks financial statements, the greatest
impact being the calculation of impairments:

RISK DATA MANAGEMENT | AUGUST 2015

39

Parallel 71%
Run

Capital
Plan

$2M

More than 82% of respondents plan

to conduct a parallel run ahead of the
implementation

Of the respondents have an

IFRS 9 roadmap in place for the
implementation

32% of respondents consider IFRS 9 a

business benefit for capital planning
activities and timely provision planning

42% of respondents have allocated this

budget or higher for IFRS 9 compliance

Data

80%

Deal
Level

Source: Moodys Analytics

Basel

More than 40% of respondents plan

to integrate IFRS 9 requirements in the
Basel infrastructure

IAS 39 A provision is made only when

there is a realized impairment. This results

specific standard.

in too little, too late provisions and does

Asia and the Middle East: More than 370

not reflect the underlying economics of the

banks (banks of significant importance)

Is a major challenge when

Of respondents will include scenario
85% of respondents plan to run
transaction.
implementing and
designing an IFRS 9
analysis in the IFRS 9 calculation
facility-level granularity for wholesale
Institutions inportfolios
the US will not be subject to IFRS
solution
IFRS 9 Aligns the measurement of

financial assets with the banks business

model, contractual cash flow characteristics
of instruments, and future economic
scenarios. Banks may have to take a
forward-looking provision for the portion
of the loan that is likely to default, as soon
as it is originated.
IFRS 9 has also several common characteristics
with the Financial Accounting Standards
Boards (FASB) Current Expected Credit Loss
(CECL) model provisioning framework to be

However, FASB will introduce a similar analytical

framework (CECL) if the current proposal is
approved under the proposed form without
major modifications.
Industry Snapshot: Current State
With all eyes on IFRS 9, Moodys Analytics
carried out our first IFRS 9 survey to help
practitioners better understand how their
peers are preparing for the implementation.
Overall, banks that participated in the survey are
accelerating their planning, budgeting processes,

Who Will Be Subject to IFRS 9?

implementation projects, given the finalization

IFRS 9 will be required for financial institutions

of the IFRS 9 standard.

and road-mapping activities for full-scale

Oceania. Specifically:

Survey Results

Companies listed on EU stock markets

regarding how they are approaching the

The survey consolidates the views of 28 banks

and EU banks must use IFRS reporting

challenges that IFRS 9 poses. Banks answered 22

standards in preparing their consolidated

questions across five main areas:

financial statements.
Europe: More than 230 banks (banks of
significant importance)
Asia, Americas (excluding the US), Oceania,
and Africa will be implementing IFRS either
through a local-endorsement process

MOODYS ANALYTICS RISK PERSPECTIVES

9 (GAAP is mandatory for those institutions).

implemented in the US.

in Europe, the Middle East, Asia, Africa, and

40

or convergence of the respective country-

1. Data
2. Analytics
3. Calculation
4. Reporting
5. Business uses

More than 82% of respondents plan

to conduct a parallel run ahead of the
implementation

Of the respondents have an

IFRS 9 roadmap in place for the
implementation

32% of respondents consider IFRS 9 a

business benefit for capital planning
activities and timely provision planning

42% of respondents have allocated this

budget or higher for IFRS 9 compliance

Basel

Data

80%

Deal
Level

More than 40% of respondents plan

to integrate IFRS 9 requirements in the
Basel infrastructure

Is a major challenge when

implementing and designing an IFRS 9
solution

Of respondents will include scenario

analysis in the IFRS 9 calculation

85% of respondents plan to run

facility-level granularity for wholesale
portfolios

REGULATORY SPOTLIGHT

Section 1 Participants
Key findings:
We gathered our survey results from a

More than 72% of the respondents are from

significant cross section of institutions of all

the risk and finance divisions at banks who

sizes, proof that IFRS 9 implementation is on

will also be the major users of IFRS 9 (from

the agenda, regardless of the size of the bank.

an impairment-calculation and financial

More than 60% of the institutions have

reporting perspective, respectively).

operations in the EMEA and APAC regions

Finance is the main stakeholder given the

where IFRS 9 will be mandatory. Institutions

financial reporting implications of IFRS 9.

in the US will not be subject to IFRS 9 (GAAP

However, the risk division closely follows

is mandatory for those banks).

finance given its role in the creditimpairment calculation.

Question 1: What are the total assets of your bank?

45%

39%

40%
35%
30%
25%

21%

21%
18%

20%
15%
10%
5%
0%
Total assets > $500bn

Total assets between $500bn and

$100bn

Total assets between $100bn and

$20bn

Total assets < $20bn

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

41

Question 2: In which region(s) does your bank operate?

36%
Globally
3%
32%

Asia (incl. Australia)

EMEA
USA/Canada

29%

Source: Moodys Analytics

Question 3: What is your role in the organization?

11%

7%
Regulatory Compliance

11%
29%

Finance
Risk
Accounting

43%

Other

Source: Moodys Analytics

Question 4: Who is the key stakeholder responsible for IFRS 9 in your organization?

7%
14%

Regulatory Compliance
Finance

14%
39%

Risk
Accounting

25%

Source: Moodys Analytics

42

MOODYS ANALYTICS RISK PERSPECTIVES

Other

REGULATORY SPOTLIGHT

Section 2 Preparing for 2018

Key findings:
More than 82% of banks surveyed have a
formal roadmap in place and plan to carry out
a parallel run ahead of the implementation
deadline.

More than 40% of the respondents are

planning to integrate IFRS 9 requirements in
the Basel infrastructure.
More than 43% of the respondents have

More than 85% of banks surveyed plan to

allocated a budget of more than $2 million

have an operational IFRS 9 solution by 2017

to meet the IFRS 9 requirements and improve

(one year before the mandatory date to be

their infrastructure and analytics.

IFRS 9 compliant).

Question 5: Do you have a formal IFRS 9 implementation roadmap at your organization?

29%
Yes
No
71%

Source: Moodys Analytics

Question 6: Are you planning a parallel run ahead of the deadline for implementation?

16%
Yes
No
82%

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

43

Question 7: If you are going to be conducting a parallel run ahead of the deadline, when will you
need an IFRS 9 solution?
46%
36%

14%
5%

2015

2016

2017

2018

Source: Moodys Analytics

Question 8: Are you planning to integrate your IFRS 9 compliance with other initiatives? Please
state which initiatives.

45%

26%
17%
9%

Basel (Risk
Systems)

Basel (Finance
Systems)

Stress Testing

Financial Planning

2%

2%

Origination
systems

Other

Source: Moodys Analytics

Question 9: What is the allocated budget for IFRS 9 implementation?

32%

< $500k
$500k-1M

43%

$1M-2M
4%
21%

Source: Moodys Analytics

44

MOODYS ANALYTICS RISK PERSPECTIVES

>$2M

REGULATORY SPOTLIGHT

Question 10: What is your timeline to be IFRS 9 compliant?

43%

29%
18%
7%

4%
2015

2016

2017

2018

Other

Source: Moodys Analytics

Question 11: Are you planning to invest in data reconciliation and aggregation platforms for
IFRS 9 provision calculation, reconciliation, and reporting?

Part of RDA

29%

No

29%

Yes

43%

Source: Moodys Analytics

Section 3 Data and Calculation

Key findings:
Gathering granular data and developing
PD and LGD IFRS 9-compliant models are

amortizing balances.
More than 63% of the respondents plan to

the major challenges to designing and

leverage their Basel IRB models for the credit-

implementing an IFRS 9 solution.

loss impairment calculation.

More than 40% of the respondents plan to

More than 50% of the respondents plan to

add the credit impairment and expected loss

run facility-level calculations for the retail

calculation engine to their Basel risk systems.

portfolio; more than 85% of the respondents

More than 82% of the respondents plan

to leverage their ALM systems to compute

are planning to run this level of granularity for

the wholesale portfolio.

RISK DATA MANAGEMENT | AUGUST 2015

45

Question 12: If you plan on investing in an IFRS 9 Provisions Engine, to which system will it be
an add-on?
39%

25%

14%
11%
7%
4%

Stress Testing

Underwriting /
Basel (Finance
Origination Systems
Systems)

Other / Not yet

defined

Accounting
Systems

Basel (Risk Systems)

Source: Moodys Analytics

Question 13: Please rank the following in order of difficulty (encountered or expected) when
designing and implementing your IFRS 9 provision and impairment solution.

3. IFRS 9
LGD Models
4. IFRS 9
EAD Models

2. IFRS 9 PD
Models

1. Granular Data

7. Deal
Bucketing

5.
Computation
Time

6. Basel
Reconciliation

Source: Moodys Analytics

Question 14: How do you plan on computing amortizing balances?

18%

Leverage Existing ALM Solution

Actuarial Formula
82%

Source: Moodys Analytics

46

MOODYS ANALYTICS RISK PERSPECTIVES

REGULATORY SPOTLIGHT

Question 15: Do you plan on building dedicated IFRS 9 provisioning models in addition to
Basel models?

39%

Yes
61%

No

Source: Moodys Analytics

Question 16: Do you plan on using an Advanced Internal Rating Model for IFRS 9 provision
calculation?

63%

37%

Yes

No

Source: Moodys Analytics

Question 17: Do you plan on incorporating scenario capabilities in the IFRS 9 provision calculation?

21%
Yes
79%

No

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

47

Question 18: What is the planned bucket allocation and provisioning calculation granularity
level for retail?

50%

29%
21%

Facility (deal) level

Sub-portfolio level

Portfolio level

Source: Moodys Analytics

Question 19: What is the planned bucket allocation and provisioning calculation granularity
level for wholesale?

86%

Facility (deal) level

7%

7%

Sub-portfolio level

Portfolio level

Source: Moodys Analytics

Section 4 Planning and Business Benefits

Key findings:
Improved timely provisioning planning and
better origination practices and capital
planning are the major IFRS 9 benefits for the
business.
More than 68% of the respondents plan to
run monthly calculations aligned with the

48

MOODYS ANALYTICS RISK PERSPECTIVES

frequency for Basel-related calculations (e.g.,

RWAs).
More than 90% of the respondents are
planning to integrate IFRS 9 scenario analysis
into capital planning, stress testing, and
origination activities.

REGULATORY SPOTLIGHT

Question 20: Will you integrate IFRS 9 scenario capabilities into the origination, capital
planning/optimization, and stress testing-related analytics or platforms?

91%

9%

Yes

No

Source: Moodys Analytics

Question 21: What is the planned IFRS 9 provisioning calculation system processing frequency?

68%

14%

11%

7%
Daily

Weekly

Monthly

Other

Source: Moodys Analytics

Question 22: What do you consider the overall business benefits of undertaking an
IFRS 9 initiative?

15%

Regulatory Compliance

10%

37%

Improved Portfolio Quality

Improved Timely Provisioning Planning
Origination Tool

27%
12%

Capital Planning

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

49

WHAT IF PPNR RESEARCH PROVES FRUITLESS?

By Dr. Tony Hughes

Dr. Tony Hughes

Managing Director of Credit
Analytics

Tony manages Moodys Analytics credit analysis

consulting projects for global lending institutions. An
expert applied econometrician, he has helped develop
approaches to stress testing and loss forecasting in
retail, C&I, and CRE portfolios.

Banks must be savvy about all the forces at work before trusting
their PPNR models. This article addresses how banks should look to
sources of high-quality, industry-level data to ensure that their PPNR
modeling is not only reliable and effective, but also better informs
their risk management decisions.
While most banks can now produce decent stress

feel slighted if their actions were dismissed as

tests for credit losses, research continues in the

irrelevant to the portfolios projections. Indeed,

important area of pre-provision net revenue

if a business experiences strong growth, how do

(PPNR). Even though PPNR is an important

they know that the upswing is a result of general

part of a banks proactive stress testing regime,

economic improvement and not a managers

researchers must consider all the factors before

improved sales procedures?

trusting the accuracy of their models or

expecting bank executives to trust them.

excluded from the model), the underlying effect

Regulators require banks to produce forecasts

of the economy on volume will be distorted

of loan and deposit volume, fees collected, and

and projections drawn from the model will be

interest rate spreads (both paid and received),

dangerously misleading.

and non-interest revenues and expenses. These

factors play an important role in determining a
banks financial position should a dire economic
scenario start to unfold.
PPNR should complement stress testing, but
the models it produces may not be as
trustworthy as they seem. For one thing, many
bank portfolios contain either scant or noisy
PPNR data. It is not atypical for a bank to be
forecasting, say, commercial loan origination
volume with only 30 or 40 time-series
observations at their disposal.
Within this context, modelers need to account
for a number of other key factors that may
influence business volume. Though the main
aim of PPNR modeling is to identify robust
macroeconomic drivers, managers would surely

MOODYS ANALYTICS RISK PERSPECTIVES

truth (and if portfolio-specific factors are

Where does PPNR fall short?

thus generating stress predictions of interest

50

If the latter explanation has even a grain of

When macro factors are mutable

Sometimes even diligent, well-designed research
finds nothing. With a huge array of macro factors
influencing the observed behavior of a portfolio,
even focused research may not lead banks to a
concrete destination.
Suppose banks diligently and intelligently
produce the best possible model given this
situation. They try to be parsimonious, using
simple but powerful techniques and employing
an intuitive behavioral framework. They then
carefully consider any statistical issues that arise
as they produce their models.
What happens, then, if the model produced
by this process the best possible model is
demonstrably unreliable or fragile?

REGULATORY SPOTLIGHT

Figure 1 FRB commercial bank assets: model structure

Total Assets

Bank credit
(71%)

Securities
(27%)

Treasury and
Agency (68%)

Cash
(20%)

Other
(8%)

Interbank Loans
(<1%)

Loans and
leases (73%)

Commercial and
industrial loans
(22%)

Other
(32%)

Real estate
loans (47%)

Consumer loans
(15%)

Other
(16%)

Trading
(<1%)

Allowance for
loan and lease
losses (<1%)

Fed funds and

reverse RPs with
banks (90%)

Derivatives with a
positive fair
value (88%)

To commercial
banks (10%)

Other
(12%)

Treasury and
Agency MBS
(70%)

Other MBS
(15%)

Revolving home
equity (13%)

Credit cards and

other revolving
plans (51%)

Fed funds and

reverse RPs with
nonbanks (25%)

Treasury and
Agency non
MBS (30%)

Other non
MBS (85%)

Closed-end
residential
(44%)

Other
(49%)

All other
(75%)

Commercial
(43%)
Source: Moodys Analytics

When quantitative research falls short, the

solution is invariably the same: collect more
data! But in the case of PPNR modeling, it is
often impossible to source more information
from within the bank. Origination volume,
the example used, is inherently a time-series
concept. Stress testers, though highly skilled,
have yet to unlock the secrets of time travel.

through many distinct business cycles.

This data is not specific to any one bank,
meaning that modeling the effect of
management actions is not possible at
this level. Despite this drawback, this
method provides the best possible avenue
through which a diligent modelers could

PPNR should complement stress testing, but the models it produces may
not be as trustworthy as they seem. For one thing, many bank portfolios
contain either scant or noisy PPNR data. Indeed, if a business experiences
strong growth, how do they know that the upswing is a result of general
economic improvement and not a managers improved sales procedures?
A time to turn to external sources

find appropriate macroeconomic drivers of

The only sensible alternative is to look for data

activity in the commercial lending space.

from external sources. In the case of commercial

Individual bank actions, under some reasonable

loan volume, for instance, the Federal Reserve

assumptions, simply do not impact industry

Board has quarterly data stretching back to the

dynamics. This means that banks can focus their

late 1940s. Using such a long series makes it

attention on identifying pertinent macro factors

easy to identify macroeconomic relationships

without having to worry about acquisitions,

RISK DATA MANAGEMENT | AUGUST 2015

51

customers switching banks, staffing shifts, or

project the banks performance under a number

changes in management strategy.

of alternative scenarios. The research is now

Modeling 30 or 40 bank-specific observations

becomes much easier when stressed industry

Is PPNR worth the effort?

variables are already in hand and the right macro

Banks may wonder whether the current

variables are understood with a high degree of

approach to PPNR modeling is very informative.

confidence. Now, stress testers can focus almost

Most banks, relying on scant internal data, have

exclusively on bank-specific drivers of observed

to cut corners or mine the data to find macro

portfolio behavior.

linkages that are likely to be spurious or, at best,

A researcher might notice, for example, that

his portfolio has been growing at a faster rate

52

MOODYS ANALYTICS RISK PERSPECTIVES

usable and relevant.

fragile. The relationships they do find are unlikely

to last through the next downturn.

than the broader industry and that the banks

Taking a realistic and holistic approach to PPNR

market share is rising as a result. He can then

modeling, then, risk modelers should look to

interrogate relevant managers on the business

the many sources available for high-quality,

side of the bank to find out why this is happening

industry-level data for PPNR components. Only

and whether the trend is likely to continue. More

by using these data will PPNR stress testing be

formally, he could seek quantitative drivers that

the basis of reliable risk management decisions

explain the banks growth anomaly and thus

and be taken seriously by bank executives.

REGULATORY SPOTLIGHT

IMPLEMENTING THE IFRS 9S EXPECTED LOSS

IMPAIRMENT MODEL: CHALLENGES AND
OPPORTUNITIES
By Eric Leman

The International Accounting Standards Board (IASB) has devoted

considerable effort to resolving issues that dramatically emerged
during the financial crisis particularly the delayed recognition
of credit losses on loans. As many believed that the incurred loss
model in IAS 39 contributed to this delay, the IASB has introduced a
forward-looking expected credit loss model. In this paper, we focus
on the impairment aspect of the IFRS 9 standard, and how banks
should now calculate credit losses to comply with the new IFRS 9
rules by 2018.
The IASB published the IFRS 9 Financial

complexity of the new systems and workflows to

Instruments in July 2014, completing its

be put in place.

response to the financial crisis by improving

the accounting and reporting of financial assets
and liabilities. It replaced the IAS 39 Financial
Instruments: Recognition and Measurement with

Under IAS 39 accounting standards, credit

losses were taken into account when the loss
occurred; hence the term incurred loss.

1. Classification and measurement: determines

recognition will follow a forward-looking

liabilities in financial statements and their

ongoing measurement.
2. Hedge accounting: launches a reformed
model for hedge accounting, with enhanced
disclosures about risk management activity.

With the new IFRS 9 standards, impairment

expected credit loss model.
According to the new model, credit exposures
will be categorized into one of three stages,
depending on the increase in credit risk since
initial recognition (Figure 1). IFRS 9 requires that
when there is a significant increase in credit risk,

3. Impairment: introduces a new expected loss

institutions must move an instrument from a

impairment model that will require more

12-month expected loss to a lifetime expected

timely recognition of expected credit losses.

loss. In making the evaluation, the institution

Impairment is the biggest change for banks

moving from IAS 39 to IFRS 9. Forecasting
expected credit losses instead of accounting for
them when they occur will require institutions
to greatly enhance their data infrastructure
and calculation engines. The timeline given by
regulators compliance by 2018 presents a
considerable challenge, especially given the

Eric specializes in banking compliance and risk

management Basel II capital adequacy (credit
risk, market risk), ALM, stress testing, and credit risk
monitoring.

Understanding the new impairment model

a unified standard that covers three areas:

how to account for financial assets and

Eric Leman
Director, Solutions Specialist

will compare the initial credit risk of a financial

instrument with its current credit risk, taking into
consideration its remaining life.
In stages one and two, the interest revenue
will be the effective interest on gross carrying
amount; in stage three it will be the effective
interest on amortized cost.

RISK DATA MANAGEMENT | AUGUST 2015

53

Figure 1 Increase in credit risk since initial recognition: three stages

Increase in Credit Risk Since Initial Recognition

Stage 1

Stage 2

Stage 3

Impairment Recognition: A Forward-looking Expected Credit Loss Model

12-month Expected
Credit Losses

Lifetime Expected
Credit Losses

Lifetime Expected
Credit Losses

Effective Interest on
Gross Carrying Amount

Effective Interest on
Amortized Cost

Interest Revenue

Effective Interest on
Gross Carrying Amount

Source: Moodys Analytics

Determining expected losses

Includes forward-looking economic forecasts

In order to calculate 12-month and lifetime

Existing internal ratings-based (IRB) Basel

expected losses, banks should apply models

models can be reused but particular attention

on credit risk (PD, LGD), balance sheet forecast

should be paid to point-in-time versus

(prepayments, facility withdraws), and interest

through-the-cycle models

rates (discount factors).

To overcome those challenges, banks should set up a dedicated group

of subject matter experts and facilitate close collaboration between the
architecture team to ensure availability of data and infrastructure and the
modeling team to ensure models are accurate and can rely on available data.
On the credit risk side, PD and LGD models are

LGD models: IFRS 9 requires an estimate of

needed to satisfy the new impairment model.

loss percentage that is consistent with the

PD models: IFRS 9 standards require an estimate

of probability of default (PD) that is consistent
with the following principles:
Considers all relevant information
Reflects current economic circumstances
(i.e., it is a best estimate rather than a
conservative estimate)
Provides the likelihood of a default occurring

54

MOODYS ANALYTICS RISK PERSPECTIVES

following principles:
Considers all relevant information and
includes a forward-looking element
Reflects current economic circumstances (i.e.,
is a best estimate rather than an economic
downturn estimate)
Considers only costs directly attributable to
the collection of recoveries

within the next 12 months or during the

Complying with IFRS 9 requirements

lifetime of the instrument

Financial Institutions will face some challenges

REGULATORY SPOTLIGHT

Figure 2 Calculation process workflow

Facilities &
Exposures

Stage
Allocation

Counterparties

Banks' Systems

Allowance &
Provision

PD Term
Structure

Risk Mitigants

LGD Term
Structure

Market Data &

Economic Forecast

EAD Profile

Accounting
System

12-Month &
Lifetime EL

IFRS 9 Impairment System

Banks Systems

Source: Moodys Analytics

to fulfilling these IFRS 9 requirements, including:

Retrieval of old portfolio data, especially
for the transactions that originated before
the advanced internal ratings-based (A-IRB)
models were introduced.
Classification of the transactions at

set up a dedicated group of subject matter

experts and facilitate close collaboration
between the architecture team (to ensure
availability of data and infrastructure) and the
modeling team (to ensure models are accurate
and can rely on available data).

origination. Products will need to be

Banks may either enhance existing solutions or

categorized a priori (contractual cash flow

use brand new products to achieve compliance.

test) or create a workflow to capture the

In either case, they should plan and execute an

classification and initial credit worthiness. An

implementation project in the next two years.

additional effort could be required to identify

those products that can be considered out of
scope (e.g., short-term cash facilities and/or
covenant-like facilities).
Management of standardized approach
portfolios (if no model is available and/or
data is not available)
Flexibility of implementations (e.g., on
models and thresholds) according to asset
classes and model availability. For instance,

Implementing a rigorous workflow

Financial institutions should ensure that their
systems can handle such granularity of data
while maintaining high quality standards. They
should use a rigorous workflow to produce these
outputs consistently (Figure 2).
Figure 2 illustrates how banks should gather
data on:

a granular approach may be needed for one

Exposures

part of a portfolio (e.g. wholesale portfolio),

Counterparties

while another portfolio (e.g., retail) may

Credit risk mitigants

require provisioning.
Historization of data for the new transactions.

From this data, banks can implement models on

To overcome those challenges, banks should

using market data and macroeconomic forecasts

PD, LGD, and exposure at default (EAD) profiles,

RISK DATA MANAGEMENT | AUGUST 2015

55

to get 12-month and lifetime expected loss

forecasts (discounted at current interest rates).
Then, based on exposure and counterparty
characteristics, allocation between stages
1, 2, and 3 sends the final EL provision to
accounting systems.
An example of such a calculation process
would include:
The interest rate of each loan is used to
calculate the discount rate.
EAD is calculated monthly for the next 360

months is the product EAD*PD*LGD divided

by the discount rate.
The EL is then summed up for the first 12
months and for the full life of the loan. These
two figures can then be used by accounting
systems.
Conclusion
IFRS 9 is the next regulatory tsunami. Like
Basel II and Basel III, it requires banks to
make huge investments in models, data, and

contractual balance of the loan, plus up to six

infrastructure for long-term implementation.

The PD is derived from a default curve

calibrated for the portfolio. The age of the
loan will give the starting point on the default
curve. This PD is then scaled to the loan, using
the Basel point-in-time PD.
The LGD is derived from the loan-to-value
(LTV) using a lookup table. The LTV uses the
value of the property covering the loan and

MOODYS ANALYTICS RISK PERSPECTIVES

eventually covered by this property.

The expected loss for each of the next 360

months, based on the amortization of the

months of arrear payments.

56

takes into account EAD from all other loans

The output of IFRS 9 will be a more resilient

financial system, capable of forecasting losses
instead of accounting them after they occur,
which will give the investor community greater
confidence and add transparency to credit losses
forecasts. Furthermore, banks will leverage
such an implementation to manage, in a more
accurate manner, their risks and forecast their
capital and profit and loss.

 2015 Moodys Analytics, Inc. and/or its licensors and affiliates. All rights reserved.

Stress testing that fits.

Moodys Analytics complete and integrated stress testing solutions
are tailored to your needs.
To learn more please contact:
[Link]/stresstestingthatfits

Essential insight serving

global financial markets

[Link]

RISK DATA MANAGEMENT | AUGUST 2015

57

APPROACHES TO
IMPLEMENTATION
Highlights best practices for effectively applying risk data management to your
organization, including improving stress testing, commercial lending efficiency,
and risk appetite management.

ENHANCED DATA MANAGEMENT:

A KEY COMPETITIVE ADVANTAGE FOR
JAPANESE BANKS
By Yuji Mizuno

Yuji Mizuno
Director, Business
Development Officer

Yuji leads the product and consulting areas of the firm

in Japan and has extensive knowledge of regulations
and risk management practices among financial
institutions. He provides clients with insight on
regulatory compliance, ALM, liquidity and
ERM frameworks.

As mass amounts of data meet ever-increasing regulation in the

world of finance, sophisticated data management has never been
more important. How a bank handles this complex problem will
make or break its position as a global player.
Japanese banks face a pivotal moment in
data management

to experience a paradigm shift from a "bigger is

With global regulatory bodies continually

flexible data infrastructure design an especially

introducing new banking regulations, the

attractive approach for Japanese banks as they

burden to be fully compliant has significantly

aim to both quickly respond to regulations and

increased. Some regulations require banks

position regulatory compliance as a profit-

to prepare new sets of data,1 while liquidity

making strategy.

always better" mindset to a more functional and

regulations, such as the Liquidity Coverage Ratio

(LCR) and Net Stable Funding Ratio (NSFR),

The more effective use of risk capital

compel banks to source data for their balance

Improving low profitability has long been

sheets. New regulatory concepts, such as the

the biggest challenge to Japans slow-growth

risk appetite frameworks and stress tests, will

economic environment. After years of rebuilding

inevitably require banks to improve their data

following the global financial crisis, Japanese

management systems even further.

banks are finally redirecting their strategies.

Banks are now asking themselves, If we are already spending an enormous

amount of time and money, why don't we make it more useful for senior
management as well? By developing effective management platforms,
they can take risks in a more aggressive but reasonable manner, ultimately
gaining the strength they need to compete with other global banks.
In response, most large Japanese banks are
currently building data management platforms
constructing large-scale data infrastructures
that address every potential business and
regulatory need. While a multi-purpose data
platform is a step in the right direction, these
ambitious projects sometimes fail at that
very task due to their sheer scope and do
everything strategy.
This complex problem has caused the industry

60

MOODYS ANALYTICS RISK PERSPECTIVES

Large banks are pursuing new means of gaining

revenue: expanding outside of Japan, such as
making acquisitions in foreign countries, or
merging with other banks in Japan. To increase
their revenue, they are also taking more risks
under a reasonable risk control regime, rather
than simply containing the risks under a
conservative limit framework.
Banks tend to regard regulatory compliance as an
annoyance, as they believe it does not generate

APPROACHES TO IMPLEMENTATION

revenue in and of itself. New stress testing

risk, and regulatory compliance at the same time

requirements have prompted many banks to

was a puzzle they needed to address.

ask, "Why do we have to use so many resources

to analyze when and how we will die?" Beyond

Large banks now know that they have to find

the best mix of solutions for these three factors.

simply maintaining a banking license, though,

Making matters difficult is the fact that they

regulatory compliance can help increase a

tend to work against each other trying to

banks profitability if the data is effectively

managed. Japan serves as an ideal case study to
this point.

achieve higher revenue may cause higher risk and

capital requirements. Moreover, new banking
regulations may not be fully consistent with one

Optimizing revenue, risk, and regulatory

compliance: a complex puzzle

another. For example, holding too many liquid

The risk appetite framework was introduced

leveraged ratio. With all of these contradictions,

to Japan in 2011. Initially, many banks did not

regulatory compliance can resemble a frustrating

know how to reconcile their need to take more

game of whack-a-mole.

assets to achieve the LCR could work against the

business risks with the frameworks ostensible

purpose of reducing them.

In attempting to win this game, Japanese

After all, the major objective of Japanese

their strength their effective organization

banks was to achieve revenue targets, not

into necessary functions and departments

just reduce risk. As banks realized that the

was actually their Achilles heel. If they worked

quest for additional revenue sources would

separately on those three factors, it would be

not be as simple as it was previously, they

almost impossible to attain an optimal solution,

gradually Japanized their implementation of

especially if they stayed in departmental silos.

banks found that what had historically been

the risk appetite framework so that it fit their

Banks discovered that risk appetite frameworks

challenging environment. This shift became

could, in fact, be a solution. They began

apparent in 2013-2014. As more complex and

implementing these frameworks as a platform

stricter banking regulations were released,

however, they realized that optimizing revenue,

for senior management to discuss how they

Figure 1 Optimization of three factors: revenue, risk, and regulation

Risk Limits

Optimization
of Three Factors
Revenue
Targets
Regulatory
Compliance

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

61

Figure 2 Risk appetite frameworks and stress tests

Risk Appetite Framework

Describe the risk that the firm must take to achieve business targets by
way of risk parameters, regulatory ratios, and KPIs.
Describe Business Plan
with Risk Appetite
Revenue Targets
Business
Plan

Stress
Scenarios

Risk Limits

Projections

Regulatory Compliance
Validate Business Plan
with Stress Tests

Stress Test Framework

Apply stress scenarios to the business plan and projects B/S, P&L,
risk parameters, regulatory ratios, and KPIs in the future.

Source: Moodys Analytics

could simultaneously optimize revenue, risk, and

regulatory compliance.
At the same time, stress testing was also
changing significantly. More banks began to
use stress testing as a tool to verify their risk

1. Aggregating all the data at a group level: As

some large banks expand their businesses
globally, it has become more challenging to
gather risk data in the same formats from all
of their global entities and business units.

appetite and establish whether or not a proposed 2. Identifying new risks throughout the group:
It is important to identify and quantify
plan could withstand stress events and still
achieve the three targets. This also helped senior
management better understand their strategys

a global organizations hidden risks and

incorporate them into its existing risk

weaknesses and adjust it if necessary.

management framework. Japanese regulators

Data management in Japan: a changing

mentality

stress testing scenarios.

Japanese banks have begun to view regulatory

compliance and data management as an
opportunity to enhance their business and
increase revenue, rather than as a mere cost of
maintaining their licenses. Along with regulatory
challenges, data management has emerged as a
critical issue for Japanese banks.

expect these emerging risks to be used for

3. Reporting those risks to senior management:
Senior management cannot effectively use
risk appetite indices or stress test results for
improving the management of their bank
unless they are reported accurately and
promptly. An MIS should incorporate a highly
automated dashboard system to quickly
share all the risk appetite indices.

Banks typically handle data management by

By overcoming these three challenges, Japanese

using a management information system (MIS).

banks can build a comprehensive data platform

An MIS is a series of IT platforms that are used

that lets them gather all the data, identify

for all stages of the process, from aggregating

emerging risks at early stages (with stress testing

data to reporting to senior management. All

results), and swiftly report Key Risk Indictors

banks must build this as a foundation for a risk

(KRIs) to senior management. Banks are now

appetite framework.

asking themselves, If we are already spending

Japanese banks typically share three common

data management challenges:

an enormous amount of time and money,

why don't we make it more useful for senior
management as well? By developing effective
management platforms, they can take risks

62

MOODYS ANALYTICS RISK PERSPECTIVES

APPROACHES TO IMPLEMENTATION

in a more aggressive but reasonable manner,

There is a general distaste in Japan for IT

ultimately gaining the strength they need to

environments that involve many different

compete with other global banks.

systems commingling like spaghetti or

The strengths and weaknesses of centralized IT

organizations
While a strong data platform is the foundation
for achieving risk management objectives,
Japanese banks tend to deal with all the data
issues at the same time by planning one huge
IT project making the work much trickier
and the risk of delays or failure much higher.
Moreover, by the time such an ambitious project
is completed, there may well be new banking

multiple overlapping data warehousing systems.

This is probably due to Japans bitter experiences
in the 1990s, when it struggled to integrate
different IT systems after mergers. Therefore,
data integration projects in Japan are commonly
comprehensive in scope and enormous in scale,
usually requiring many years to complete. But
with the extensive, complicated, and everchanging requirements of the current regulatory
regime, this centralized IT model falls short.

regulations with different data requirements to

Why is data management so challenging?

contend with.

The Basel Committee on Banking Supervision

This method has resulted in part from the

typically centralized organizational structure of
the IT group at Japanese financial institutions,
which controls the IT work for every business
line. A centralized structure enables banks
to build a consistent and comprehensive IT
infrastructure, but it lacks speed and flexibility
in implementation.

issued a report in January 2015 called Progress

in adopting the principles for effective risk data
aggregation and risk reporting, which contains an
interesting lesson about Globally Systemically
Important Banks (G-SIBs). Surprisingly, 14 out
of 30 G-SIBs revealed that they will not be
fully compliant with at least one of the Basel
Committees regulatory principles by the

Banks do not need to expand the projects scope to encompass a rebuild of

the entire database a potentially endless project but can instead simply
channel the existing data into a relay station. A data relay station is more
cost efficient, too, as it is often completed within a much shorter time
frame than a large IT project.
This structure is often contrasted with the

deadline in 2016. Some banks noted in the report

federal organizational style, which global

that this is partly due to delays in initiating or

banks outside Japan frequently use when they

implementing large-scale IT projects and the

have multiple legal entities or business units in

complexity of those projects.

a group. In a federal style, each entity or unit

has some level of independence in designing
and introducing IT platforms. To maintain order
throughout an organization, a senior IT officer
at a holding company level controls all those
activities. (With regards to data management
specifically, some large US banks have assigned
a Chief Data Officer, or CDO, although it is
still rare for Japanese banks to have a CDO.) A
federal styles strengths and weaknesses are
the opposite of those of a centralized style:
what it offers in speed and flexibility, it lacks in
consistency and comprehensiveness.

Banks meet roadblocks when they try to

accomplish multiple objectives when they have
only one core challenge. A data infrastructure
usually has more than one purpose, such
as integrating all data locations, cleansing/
reconciling data from different sources,
constructing data flows to calculation engines,
and adding calculation results to an MIS
reporting flow to senior management/regulators.
Even a simple project like constructing data
flows between several IT systems could become
delayed if managers decide to expand its scope.
Maintaining focus on the most important project

RISK DATA MANAGEMENT | AUGUST 2015

63

task is therefore paramount.

typically provided in a matrix. This helps senior

In addition to regulations requiring banks to

source new data directly, the current focus of
data management for Japanese banks lies in the
following two areas, which may involve the use
of data beyond regulatory compliance:

management understand the nature of the LCR

more effectively how it behaves under stressful
conditions and how it affects the firms liquidity.
The results should be shown in a more intuitive
way if stress testing results are to be used to
improve the banks management.

Calculation of risk appetite indicators and

reporting by an MIS
Automation of a stress testing calculation
Determining risk appetite requires banks to
collect data throughout a business unit to
calculate multiple risk indicators, such as KPIs,
and to report them to senior management/
regulators through an MIS. Several steps of data
processing are required, including aggregation,

Meeting the IT challenge: focus, flexibility,

and speed
There are two specific data challenges to
which Japanese banks should pay the most
attention. First, the data warehouse system
has to be flexible, as the requirements for data
management often change. Second, a data
management platform needs to be
implemented quickly.

calculation, and reporting. Stress testing

Figure 3 Functional database

SENIOR MANAGEMENT

REGULATORY REPORTING

MULTI-FUNCTIONAL DATABASE
MIS OR DASHBOARD
DATA PROCESSING
DATA CLEANSING/RECONCILIATION
DATA AGGREGATION

EXISTING DATA SOURCES

Source: Moodys Analytics

regimes also require banks to automatically and

promptly conduct multiple calculations.
Recent trends show Japanese banks create
multiple scenarios and conduct sensitivity
analyses on a single indicator. For example,
they simulate many patterns of LCRs based
on several different data inputs, which are

64

MOODYS ANALYTICS RISK PERSPECTIVES

These two challenges are exactly where Japanese

banks method is relatively weak. They should
instead seek a lighter IT system and focus on
one task at a time by dividing the entire project
into multiple task periods. By reasonably limiting
a projects scope, it is much more likely to
succeed.

APPROACHES TO IMPLEMENTATION

Focusing on data flow rather than data

Banks could also introduce this type of functional

storage is one way to implement an effective,

database into limited areas of business, such as

efficient data platform. Banks can maintain

stress testing. To do so, they would gather all the

existing data sources and create a data flow,

essential information, including balance sheet

which gathers the necessary data from those

items and risk parameters, from all the existing

sources and sends them to a new data relay

databases. The functional database would then

station. Banks do not need to expand the

perform data processing to create consistent

projects scope to encompass a rebuild of the

assumptions and send them to relevant

entire database a potentially endless project

calculation engines. The results would then be

but can instead simply channel the existing data

collected at the functional database again and

into a relay station. A data relay station is more

reported to senior management/regulators

cost efficient, too, as it is often completed within

through an MIS.

a much shorter time frame than a large

IT project.

Sophistication is the future of data

infrastructure

One reason a data project encounters trouble

Sophisticated data management is the

is that all these functions are built into a

foundation for both improving the management

single data platform. A data relay station

of a bank and increasing revenue in global

can be used for multiple functions, including

markets. For many Japanese banks, imitating the

data aggregation, cleansing, processing, and

best practices of foreign banks is not necessarily

reporting, which work separately from a banks

the best solution. While they struggle to find

existing data platforms. Having a functionally

the appropriate direction, they have gradually

separated data flow is less risky from an

succeeded in "Japanizing" some regulatory

operational risk perspective, too.

concepts while expanding their own concept

Under this data management structure, existing

data sources and a new data relay station are

of effective IT infrastructure beyond the

cumbersome centralized approach.

linked together. Data requirements based on

Japans banks are currently being challenged to

regulations or management needs could be

significantly improve their data management

reflected in the data relay station, not at an

systems, specifically for achieving flexibility

existing data source level, which means banks

and speed to continue growing revenues while

can only work on the data relay station in a

meeting regulatory requirements. Accomplishing

comprehensive way. Such a data relay station

this will allow a bank to become a stronger, more

can be highly functional without necessarily

competitive player on the global stage.

covering all the banking activities that require

data management.

1 Basel Committee on Banking Supervision, Principles for effective risk data aggregation and risk reporting, 2013.

RISK DATA MANAGEMENT | AUGUST 2015

65

MEASURING SYSTEMIC RISK IN THE

SOUTHEAST ASIAN FINANCIAL SYSTEM
By Dr. David Hamilton, Dr. Tony Hughes, and Dr. Samuel W. Malone

Dr. David Hamilton

Managing Director, Stress
Testing and Credit Risk
Analytics, APAC
David is Managing Director, Head of Stress Testing and
Credit Risk Analytics for the Asia Pacific Region, based
in Singapore. With two decades of experience in credit
risk modeling, credit ratings, and economic research,
he helps provide insight into the practical challenges of
credit risk measurement, management, and
stress testing.

Dr. Tony Hughes

Managing Director of
Credit Analytics

Tony manages Moodys Analytics credit analysis

consulting projects for global lending institutions. An
expert applied econometrician, he has helped develop
approaches to stress testing and loss forecasting in
retail, C&I, and CRE portfolios.

Dr. Samuel W. Malone

Director of Economic Research

Sam is Director of Economic Research in the

Specialized Modeling Group (SMG). A systemic risk
expert, forecaster, and former academic economist, his
primary focus is developing systemic risk solutions that
are actionable by individual financial institutions. He
also contributes actively to SMG consulting projects,
such as model validations and bespoke research.

Systemic vulnerabilities are an important, if often overlooked, aspect

of a financial systems stress testing regime. This article looks back at
the Asian financial crisis of 1997-1998 and applies new methods of
measuring systemic risk and pinpointing weaknesses, which can be
used by todays financial institutions and regulators.
Assessing systemic vulnerabilities: East
versus West

Assessing systemic risk has been a key part of

The ability of a countrys financial system to

In addition to regulators and central banks

withstand a severe negative shock has important

increased focus on systemic risk in the wake of

implications for its general economic and

the crisis,2 the International Monetary Fund and

social well-being. Following the global financial

the World Bank jointly initiated the Financial

crisis and the sovereign debt crisis in Europe,

Sector Assessment Program (FSAP) in 1999

stress testing has become one of the primary

to assess financial stability and perform stress

techniques for gauging the robustness of

testing of countries financial sectors. These

individual financial institutions and the financial

initiatives have been credited with helping

system as a whole. Although officially part of

Southeast Asia weather the worst of the global

stress testing mandates, such as the Federal

financial crisis (GFC) and avoid a repeat of the

Reserves Comprehensive Capital Analysis

economic devastation caused by the Asian

and Review (CCAR), measuring systemic

financial crisis.

vulnerabilities has not been emphasized.

Contagion is the word typically used to
In the West, that is. In Southeast Asia, measuring

describe how the crisis spread so virulently

and understanding the potential impact of

throughout the region. But contagion is just

systemic risk became imperative following the

another way of saying that these countries

Asian financial crisis of 1997-1998. The crisis

economies were highly interconnected, and

began in Thailand in July 1997 and quickly

thus that systemic risk in Southeast Asia was

spread to Malaysia, Indonesia, Korea, and the

high. Since the Asian financial crisis, new tools

Philippines. Singapore, a regional financial hub

and techniques have been developed to better

with an open economy, was also affected.

measure the multiple dimensions of systemic

The impact of the crisis on these countries

was staggering: in one years time, a decade of
extremely strong economic growth, the East
Asian miracle, risked being erased. Between
June 1997 and March 1998, GDP contracted by
nearly 6% in Korea, 9% in Thailand, and 14% in
Indonesia. Equity valuations plummeted by 50%
or more in most of the affected countries.1

66

MOODYS ANALYTICS RISK PERSPECTIVES

financial supervision in the region ever since.

risk. In this paper, we describe a method for

measuring the interconnectedness of financial
institutions and apply it to the ASEAN-5 group
of countries.3 Our data allow us to go back
to before the Asian financial crisis and to
compare how the different shocks to the global
financial system since then have impacted the
systemic risk of financial institutions in the
ASEAN-5 countries.

APPROACHES TO IMPLEMENTATION

Interconnectedness as a measure of
systemic risk

such as the size of non-bank deposits, the size

Systemic risk refers to a shock that results in

importance to the domestic payments system.

a broad-based failure of the financial system,

The Monetary Authority of Singapore, for

which in turn threatens to jeopardize the

example, used these measures of systemic

economy. The initial shock(s) can be exogenous

risk when it participated in the 2002 FSAP

(an oil price shock, for example) or endogenous

stress tests.4

of domestic interbank borrowing, and the

(the bankruptcy of a systemically important firm

such as PT Bank Century, for example). Whatever

Although these traditional, descriptive

the source of the shock, a high degree of

measures of systemic risk are useful and

systemic risk implies the potential for a cascade

important, size measures do not necessarily

of distress or failure among financial institutions.

uncover the risks resulting from a high degree

of interconnectedness. Size is an imperfect

The word potential is important in this

measure of systemic risk. It is vital to know

context. A high degree of connectivity among

which firms occupy important nodes in the

financial institutions is a necessary but not a

financial network for example, those that may

sufficient condition for a systemic crisis. Indeed,

be a near-monopolist in market making for a

the probability of a systemic crisis is another

particular asset class, regardless of size. Hence,

matter entirely. However, when the number

measuring too-connected-to-fail is as important

and strength of the connections between

as measuring too-big-too-fail.5

The lines connecting Thailand, Singapore, and Malaysia also tend to be

blue, meaning that the relationship between financial institutions in these
countries is positive: an increase in credit risk among financial institutions
in one of these countries has a high propensity to cause an increase in
credit risk in the others.
financial institutions in an economy is high,

Measuring financial firms interconnectedness

contagion risk across firms will also be high.

In this paper, we analyze the interconnectedness

A market shock affecting one firm can quickly

and potential for contagion among financial

spread to others through sharp drops in market

institutions using Moodys Analytics Expected

valuations and the mark-to-market impact on

Default Frequency (EDF) measures. EDF

financial institutions balance sheets. A credit

measures are probabilities of default derived

event affecting one firm can spill over through

from a contingent claims model of credit

on- and off-balance sheet exposures among

risk.6 Our systemic risk framework is built on

banks and financial counterparties. When the

a network analysis perspective. Using firm-

tinder is piled sufficiently high, a small spark can

level EDFs and the determinants of those PDs

ignite a conflagration.

market leverage and asset volatility we

measure dynamic linkages by estimating Granger

The lines connecting Thailand, Singapore, and

causal connections among all pairs of large

Malaysia also tend to be blue, meaning that the

financial institutions in the ASEAN-5 countries.7

relationship between financial institutions in

these countries is positive: an increase in credit

A time-series x is said to Granger-cause a time-

risk among financial institutions in one of these

series y if past values of x provide statistically

countries has a high propensity to cause an

significant information about future values of

increase in credit risk in the others.

y. Figure 1 illustrates the concept, plotting the

time-series of PD for two hypothetical firms,

Interconnectedness has traditionally been

x and y. The default probabilities for these two

gauged using various quantitative measures

firms look very similar, but they are out of phase;

RISK DATA MANAGEMENT | AUGUST 2015

67

Figure 1 Movements in the PD for firm x Granger-cause changes in the PD for firm y

2.5
Firm x

Firm y

2.0
1.5
1.0
0.5
0.0
1

201

401

601

801

1001

1201

1401

1601

1801

2001

Time
Source: Moodys Analytics

changes in the PD for firm x precede changes

claims-based methods to the measurement

in firm ys PD. Firms x and y are temporally

of systemic risk, although they do not take a

interconnected changes in the values of x are

network approach to estimating the dynamic

closely followed by changes in firm y. In this

linkages across financial institutions. Billio

particular example, knowing past values of firm

et al. estimate systemic risk measures based

xs PD would lead to good predictions for y. The

on Granger causality networks derived from

Granger causal link is positive and strong.

linkages identified on the basis of bivariate

Formally, Granger causality means that the

and/or coefficients in the following
bivariate vector autoregression (VAR) are
p

xt = +

i xt-1+

(i = 1)

(i = 1)

yt = +

(i = 1)

x +
i t-1

(i = 1)

study was equity returns, however, rather than

PD measures derived from a structural credit
such as EDF measures.

(1)
j yt-j+xt

If the relevant F-test of the coefficients is

significant at the 5% level, x is said to Grangercause y; whereas if the equivalent F-test of the
coefficients is significant at the 5% level, y
Granger-causes x. If both sets of coefficients are
significant, there is mutual influence between
firms x and y.
Our approach to measuring systemic risk,
which is described in more detail in Hughes and
Malone, is an extension of similar approaches
taken in the systemic risk literature.8 Gray
and Malone9 and Gray, Jobst and Malone,10
for instance, adapt and apply contingent

MOODYS ANALYTICS RISK PERSPECTIVES

system under consideration.11 Their unit of

risk model and historical data on default events

j yt-j+xt

statistically significant:

68

VAR models for pairs of entities in the financial

The paper most closely related to the

approach taken here is Merton et al., which
applies the Granger causality network
technique of Billio et al. to the expected loss
ratio (ELR)12 of firm debt for major sovereigns
and global financial institutions.13 Our analysis
can be seen as complementing theirs, in the
sense that we study network linkages among
expected default frequency measures, as well
as asset volatilities and leverage ratios, which
drive EDFs (and ELRs) in asset value-style credit
risk models.
One of the advantages of the network
approach to measuring systemic risk is that we
can estimate both the direction and strength
of the connectedness between financial
institutions.14 The direction of connectedness

APPROACHES TO IMPLEMENTATION

is determined by the statistical significance of

EDF measures have also demonstrated they

the VAR coefficients, as described previously.

exhibit superior power, compared with equity

The strength of the linkages between financial

returns, for predicting future credit events such

institutions is measured by calculating the

as defaults and restructurings.15 We can also

degree of Granger causality (DGC) described

calculate DGC measures on market leverage and

in Billio et al. DGC is simply the fraction of

asset volatility, the two primary drivers of the

statistically significant Granger-causality

EDF model, to gain insight into whether credit

relationships among all N(N-1)/2 pairs of

risk contagion is being driven by volatility or

financial institutions at any given point in

leverage spillovers.

time. Each unique pair of financial institutions

can have zero, one, or two Granger-causal
connections, thus implying a maximum of
N(N-1) possible connections that can be active
in the system. The DGC measure, therefore, lies
between zero and one. The higher the ratio, the

Empirical results
The results of our empirical analysis are based
on a dataset of financial institutions (SIC
code between 6,000 and 6,799) domiciled in
the ASEAN-5 group of countries: Indonesia,

higher the systemic risk.

Malaysia, the Philippines, Singapore, and

The DGC measure captures both upstream

institutions with at least US$1 billion in book

and downstream Granger causal linkages in

assets observed at some point over their

the system. In order to share down the DGC

available histories. The only other selection

to individual institutions, we follow Billio et

constraint placed on our data is that financial

al. in computing what is known as the Out

institutions are required to have traded equity

measure for each institution. The Out measure

and public financial statements with which to

is equal to the percentage of the rest of the

calculate Expected Default Frequency measures

other N-1 institutions in the network that are

over the time interval of our study, which begins

Granger-caused by the institution in question.

in 1995 and runs through October 2014. Our

We can think of the Out measure as capturing an

study includes 201 unique financial institutions

institutions contribution to systemic risk in the

in the ASEAN-5 countries: 36 in Indonesia, 49 in

form of dynamic downstream linkages to

Malaysia, 30 in the Philippines, 46 in Singapore,

other institutions.

and 40 in Thailand.

Thailand. We limit our dataset to financial

Size is an imperfect measure of systemic risk. It is vital to know which firms

occupy important nodes in the macro-financial network for example,
those that may be a near-monopolist in market making for a particular
asset class, regardless of size. Hence, measuring too-connected-to-fail is as
important as measuring too-big-too-fail.
There are many advantages to using Expected

Figure 2 shows the top 10 financial institutions

Default Frequency metrics and their drivers

with the highest Out measures as of October

as the basis for calculating the DGC and

2014. TMB Bank Public Co. Limited, based in

Out measures. EDF measures are forward-

Thailand, exhibits the highest Out measure of

looking PDs; in this paper, we use EDFs with a

the ASEAN-5 firms. The Out measure indicates

one-year time horizon. That allows us to kill

that the banks EDF movements Granger-

two analytical birds with one stone: we can

cause EDF movements in 30.6% of the other

measure the forward-looking likelihood of the

financial institutions in the network. Notably,

default of a firm (or of a financial system by

the statistics shown in Figure 2 suggest

aggregating EDFs across firms), and calculate

that systemic risk (measured by Out) bears

the level of systemic risk using the EDF-based

little correlation with either the probability

DGC measure.

of default or with firm size on average. The

RISK DATA MANAGEMENT | AUGUST 2015

69

Figure 2 Top 10 firms ranked by Out measure as of October 2014, with EDF level and firm size

Out

EDF

Book Assets

Financial Institution

Value

Rank

Value

Rank

Value ($ mil)

Rank

TMB Bank Public Co. Limited

0.306

0.283%

42

24,568

25

OSK Holdings Berhad

0.281

0.087%

880

115

Bank of the Philippine Islands

0.264

0.387%

72

28,868

24

UOB-Kay Hian Holdings Limited

0.256

0.304%

49

2,079

77

CIMB Thai Bank Public Co. Limited

0.248

0.333%

60

7,798

49

CitySpring Infrastructure Trust

0.240

0.107%

11

1,513

94

Bangkok Land Public Co. Limited

0.240

0.081%

1,697

87

Hong Leong Capital Berhad

0.231

0.339%

62

951

113

Bangkok Life Assurance PCL

0.231

0.343%

64

6,259

55

Bangkok Bank Public Co. Limited

0.231

10

0.237%

28

78,431

Source: Moodys Analytics

EDF-Out correlation exhibits significant and

risk of default reached a historic peak during the

informative time variation, however, as

Asian financial crisis. It is also notable that the

Figure 5 shows.

peak in the systemically weighted EDF measure

occurs one to two years prior to the peak in the

The EDF rank column shows the firms EDF rank

size weighted EDF measure. The average risk of

(sorted in increasing order) out of the 122 firms

default dropped sharply after 1998, but trended

present in the network in October 2014; bookasset rank is measured in descending order. Half
of the top 10 firms with the highest systemic

higher during the early 2000s as the dot-com

bubble burst, resulting in a recession in the
United States, and Argentina defaulted on its

risk measures as of October 2014 are based in

foreign debt.

Thailand and are about average with respect to

their EDF levels and book-asset size. Most of the

The global financial crisis, as severe as it was in

firms in the top 10 list are banks, but the rest are

the West, is a relatively minor blip in the time-

in the broker-dealer, real estate, infrastructure,

series for the ASEAN-5 nations. These results

and insurance sectors.

suggest a potentially useful and powerful way

The results shown in Figure 3 bring the impact

of the 1997-1998 Asian financial crisis into

of monitoring the future likelihood of systemic

crises.

sharp focus. The graph on the left side shows

The right side of Figure 3 shows the 12-month

the weighted average EDF level for the ASEAN-5

moving average of the DGC measure for the

countries over time. We weight the historical

network at each point in time. The right side of

EDF values using book assets (size) and by Out

Figure 3 shows the DGC measure for the network

(systemic influence) . By either measure, the

at each point in time. In this one graph, we get a

Figure 3 Aggregate EDF and DGC measures for ASEAN-5 financial institutions, 1995-3Q2014
Size-weighted EDF

Degree of Granger causality in EDF network

Systemic influence-weighted EDF

0.35

14

0.30

12
0.25

10
8

0.20

0.15

0.10

0.05

0
95

97

99

01

Source: Moodys Analytics

70

MOODYS ANALYTICS RISK PERSPECTIVES

03

05

07

09

11

13

95

97

99

01

03

05

07

09

11

13

APPROACHES
RETHINKING
TO IMPLEMENTATION
INSURANCE RISK

panoramic view of how systemic risk has evolved

Figure 3, we calculated two weighted measures

for ASEAN-5 financial institutions over the past

of leverage: using book assets (size) and the Out

20 years. The strength of interconnectedness

measure (systemic influence).

among financial institutions and the high risk of

contagion that characterized the Asian financial
crisis is captured by the peak 0.31 DGC measure.

Size-weighted leverage is nearly always higher

than systemic influence-weighted leverage,
and by a considerable margin in some time

The graph also shows that it took at least four

periods. The implication is that larger financial

years for systemic risk to subside to levels

institutions lever up more, a finding consistent

that prevailed before the Asian financial crisis.

with data for US financial institutions. A second,

Although economic growth in the countries

and perhaps more important, implication is that

most affected by the crisis bounced back

a firms size is not perfectly correlated with the

strongly after 1998, our results on systemic

spillover dimension of systemic risk contribution.

risk corroborate other macro-financial

The graph on the right side of Figure 4 shows

indicators that show that their financial

size- and systemic influence-weighted average

systems and economies took a number of

asset volatility over time. Unlike average

years to fully heal.

EDF levels and leverage values, the weighted

The DGC time-series in Figure 3 attests to the

volatility measures rise throughout but peak well

fact that the risk of credit risk spillovers arising

after the Asian financial crisis, around the time of

from the global financial crisis was virtually a

Argentinas default. Systemic influence-weighted

non-event for the ASEAN-5 group of financial

volatility is everywhere above size-weighted

institutions. Although registering a brief spike,

volatility: firms that exhibit a relatively high Out

the DGC measure continued to fluctuate around

ratio, and therefore have a high potential for

the 0.18 average that prevailed after the Asian

contagion, also exhibit higher asset volatility.

financial crisis. In contrast, the DGC measure for

large US financial institutions reached a peak of
0.61 at the height of the global financial crisis.
Intriguingly, systemic risk as measured by the

The global financial crisis exerts a stronger effect

on leverage than on volatility for ASEAN-5
financial institutions. Weighted volatility rises
and persists through the European sovereign

DGC reached its highest level since the Asian

financial crisis in July 2013. However, systemic
risk has subsided considerably since that date,
falling to its lowest level in 20 years.

debt crisis, but the magnitude of the increase

is relatively small. Weighted leverage spikes to
levels that persisted during the early 2000s but
then falls sharply to pre-global financial

Figure 4 reinforces our historical

crisis levels.

understanding of the role of leverage as one

of the key causes of the Asian financial crisis.
Here, leverage is defined as the ratio of a firms
default point to its market value of assets.18 As in

Tracking cross-sectional correlations over

time can yield additional insights into system
dynamics. Figure 5 displays two Spearman (rank)

Figure 4 Weighted average leverage and volatility, 1995-3Q2014

Size-weighted leverage

Size-weighted leverage

Systemic influence-weighted leverage

0.90

0.24

0.85

0.22

0.80

0.20

0.75

0.18

0.70

0.16

0.65

0.14

0.60

0.12

0.55

0.10

0.50

0.08

Systemic influence-weighted volatility

0.06

0.45
95

97

99

01

03

05

07

09

11

13

95

97

99

01

03

05

07

09

11

13

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

71

Figure 5 Cross-sectional correlations over time: EDF-Out and leverage-volatility

[Link]

[Link]

0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
95

97

99

01

03

05

07

09

11

13

Source: Moodys Analytics

correlations: the EDF-Out measure correlation

network. Financial institutions in Singapore

and the leverage-volatility correlation. At

and Malaysia also stand out as having a high

each time point, the correlations shown are

concentration of positive (that is, forcing)

computed using only the cross section available

Granger causal relationships.

at that time point for the system. EDF levels

and systemic influence correlations tend to be
negative during calm periods and positive during
crisis periods. Leverage and volatility correlations
are always negative.

2014. Circles represent financial institutions,

and are color coded by country of domicile. This
graph displays linkages based on the coefficients
at lag 1 in the VAR models using EDF measures

increase, as do the leverage-volatility

(equation 1 above). Red lines correspond to

correlations. The interpretation is that riskier

negative coefficients (damping effects), and blue

(that is, higher default probability) financial

lines correspond to positive coefficients

institutions increasingly drive the system

(forcing effects).

between leverage and volatility underlying

optimal leverage theories in corporate finance
(for example, the trade-off theory of the capital
structure) becomes weaker in the cross section
of firms during crises. Figure 5 also shows that
the EDF-Out correlation tends to spike at
the beginnings of crisis episodes, a pattern
that is also apparent in data for US financial
institutions around 2007-2009.19
Thailand was the epicenter of the Asian financial
crisis in 1997. The devaluation of the baht set off
a cascade of financial distress throughout the
ASEAN countries. Our study of Granger causal
connections among EDF measures reveals that
financial institutions in Thailand still represent
a concentration of systemic risk in the ASEAN-5

MOODYS ANALYTICS RISK PERSPECTIVES

of Granger causal connections as of October

During times of crisis, the EDF-Out correlations

during crises and that the negative relationship

72

Figure 6 shows the complete network map

The sets of lines connecting financial institutions

in Thailand (green), Singapore (yellow) and
Malaysia (red) are numerous, giving the graph
a very dense appearance on the right side.
The lines connecting Thailand, Singapore, and
Malaysia also tend to be blue, meaning that the
relationship between financial institutions in
these countries is positive: an increase in credit
risk among financial institutions in one of
these countries has a high propensity to cause
an increase in credit risk in the others.
Conclusion
The experience of the Asian financial crisis
sparked an intense interest in measuring
systemic risk among regulators in Southeast
Asia, with the aim of developing policy tools
to mitigate its reoccurrence. That interest has

APPROACHES TO IMPLEMENTATION

Figure 6 Network map based on Granger causal connections as of October 2014

LPI

MAYBANK

PBBANK

AEONCR

RHBCAP

CIMB

YTLREIT

P
AMMB

HLCA

HLFG

ECM
KENANGA

AFFIN

MANULFE

HLBANK

MBSB

MNRB

BIMB

HDBS

SMRA ASRI

BNII

PNLF

AFG

ALLIANZ

BNLI

ADMF

K17

S05

F17
S41

C2PU

BNGA

BVIC

C09

AU8U

BBNI

SMMA

U1 1
G07

D5IU

BDMN

PNIN

D05O39

O5RU

J91U

OSK IGB KLCC

PNBN

BSDE

E8Z

MAA

TAKAFUL

J69U

P15

F3EU

S68
U1 0

K71U

LPPS

C61U

A68U

BTPN

BBCA

J85

INPC

C38U

A7RU

NISP

BBKP

T82U

ND8U

BBRI BMRI

A17U

S35

SHNG

PNB

SPALI

MBT

BBL

P40U

F25U

KBANK
CIMBT

HEMRAJ

RCB

RLC

M44U

BAYF

ROJNA

AC

VLL

KTB

THAIRE

FLI

UBP

AGI

CHIB

ALI

MEG
PSB

BKI

TISCO

TMB

AEONTS

BLAND

SCB

SCBLIF

BPI
BDO

TCAP

BLA

SECB
SMPH

TIP

KTC

FDC

KKP

KGI

Linkage based on negative coefficients

Malaysia

Singapore

Thailand

Philippines

Indonesia

Linkage based on positive coefficients

Source: Moodys Analytics

been further reinforced by the global financial

Granger causal measures captured the contagion

crisis, which, while not having a serious effect

risk of the Asian financial crisis to the greater

on Southeast Asia, served as a salutary reminder

region extremely well. Financial institutions in

that a systemic crisis can arise in one part of the

the ASEAN-5 nations experienced historic levels

world and spread to others.

of interconnectedness and default risk during

Going forward, we expect that regulators

will require financial institutions subject to
supervisory stress tests to pay greater explicit

the Asian financial crisis. They were, however,

relatively immune to the effects of the global
financial crisis.

attention to systemic risks. In order to do so,

The tools we have explored in this research note

they must be able to quantify systemic risk in

can be of indispensable use to both financial

real time. Our empirical results showed that

institutions and regulators for estimating the

RISK DATA MANAGEMENT | AUGUST 2015

73

current and future level of systemic risk and for

identifying the sources of its changes. Regulators
can, at a glance, obtain tangible signals
indicating which institutions are most strongly
connected in the network and thus pinpoint
weaknesses in the broader financial system.

place at the wrong time.

We described a useful measure of
interconnectedness using a network approach
whose implementation is straightforward
and whose outputs are intuitive and easily
interpretable. Granger causality networks

Managers of financial institutions, meanwhile,

address one of the key aspects of systemic

can assess their counterparty risks more fully

risk: the extent of dynamic spillovers between

via consideration of joint and conditional

institutions. By using expected default

default likelihoods calculated using network-

frequency measures as the fundamental unit of

based simulations. In an environment where

observation, we are able to relate statements

financial institutions are less likely to be bailed

about the likelihood of default for particular

out, managers must take steps to guard against

institutions to measures of systemic risk

failures caused merely by being in the wrong

exposure and contribution.

1 Andrew Berg, International Monetary Fund Working Paper, WP/99/138, The Asian Crisis: Causes, Policy Responses, Outcomes,
1999.
2 Masahiro Kawai and Peter J. Morgan, Asian Development Bank Institute working paper, No. 377, Central Banking for Financial
Stability in Asia, August 2012.
3 The Association of Southeast Asian Nations is an association for regional social and economic cooperation consisting of ten
Southeast Asian countries. ASEAN was formed in 1967 by Indonesia, Malaysia, the Philippines, Singapore, and Thailand (the
ASEAN-5 countries). Five additional countries joined later.
4 Chan Lily and Lim Phang Hong, The Monetary Authority of Singapore, Staff Paper No. 34, FSAP Stress Testing: Singapores
Experience, August 2004.
5 This fact has been recognized by the central banks in the ASEAN-5 nations. Much interesting research on systemic risk from
a too-connected-to-fail perspective has been generated by the central banks of Indonesia, Malaysia and Thailand, including:
Ayomi and Hermanto (2013), Bank Negara (2013), Hwa (2013), Nacaskul (2010). Sheng (2010) also studies systemic risk using
a network approach.
Sri Ayomi and Bambang Hermanto, Bank of Indonesia Bulletin of Monetary, Economics and Banking, Systemic Risk and
Financial Linkages Measurement in the Indonesian Banking System, October 2013.
Bank Negara Malaysia Financial Stability and Payment Systems Report (2013): 46-51, Risk Developments and Assessment of
Financial Stability in 2013, External Connectivity and Risk of Contagion to the Malaysian Banking System, 2013.
Tng Boon Hwa, Bank Negara Malaysia working papers, WP1, External Risks and Macro-Financial Linkages in the ASEAN-5
Economies, 2013.
Poomjai Nacaskul, Bank of Thailand working paper, Toward a Framework for Macroprudential Regulation and Supervision of
Systemically Important Financial Institutions, December 24, 2010.
Andrew Sheng, World Bank working paper, No. 67, Financial Crisis and Global Governance: A Network Analysis, 2010.
6 Zhao Sun, David Munves, and David T. Hamilton, Moodys Analytics Model Methodology, Public Firm Expected Default
Frequency (EDF) Credit Measures: Methodology, Performance, and Model Extensions, June 2012.

74

MOODYS ANALYTICS RISK PERSPECTIVES

APPROACHES TO IMPLEMENTATION

7 Clive C. W. Granger, Econometrica Vol. 37, No. 3, 424-438, Investigating causal relations by econometric models and crossspectral methods, July 1967.
8 Tony Hughes and Samuel W. Malone, Moodys Analytics white paper, CCA Financial Networks and Systemic Risk: Concepts and
Outputs, October 2014.
9 Dale Gray and Samuel W. Malone, Chichester, England: John Wiley & Sons, Macrofinancial Risk Analysis, 2008. Dale Gray
and Samuel W. Malone, Annual Review of Financial Economics Vol. 4, No. 1: 297-312, Sovereign and Financial Sector Risk:
Measurement and Interactions, 2012.
10 Dale Gray, Andreas Jobst, and Samuel W. Malone, Journal of Investment Management Vol. 8, No. 2: 90-110, Quantifying
Systemic Risk and Re-conceptualizing the Role of Finance for Economic Growth, 2010.
11 Monica Billio, M. Getmansky, A. Lo and L. Pelizzon, Journal of Financial Economics Vol. 104, No. 3: 535-559, Econometric
Measures of Connectedness and Systemic Risk in the Finance and Insurance Sectors, June 2012.
12 ELR is defined as the expected loss, or implicit put option, component of the debt divided by its promised (or risk-free) value.
13 Robert C. Merton, Monica Billio, Mila Getmansky, Dale Gray, Andrew W. Lo, and Loriana Pelizzon, Financial Analysts Journal 69
(2): 22-33, On a New Approach for Analyzing and Managing Macrofinancial Risks, 2103.
14 Although we do not discuss it in this paper, Granger-causality networks also allow us to identify whether the relationship
between financial institutions is positive (forcing) or negative (dampening), depending on the signs of the and
coefficients in equations (1). Hughes and Malone (2015) estimate the forcing and damping effects for U.S. financial institutions.
15 Zhao Sun, Moodys Analytics ViewPoints paper, An Empirical Examination of the Power of Equity Returns vs. EDFs for Corporate
Default Prediction, January 2010.
16 To be precise, the systemic influence weights are an equally weighted average of weights based on Out, [Link], and Inverse
Closeness; the latter two measures are described in Hughes and Malone (2015).
17 Hughes and Malone, 2014
18 In the expected default frequency model, the default point is defined as the notional value of liabilities that would trigger
a credit event. For corporates, it is calculated as short-term debt plus half of long-term debt. For financial institutions, it is
calculated as 75% of reported total liabilities.
19 Tony Hughes and Samuel W. Malone, Moodys Analytics ViewPoints paper, Systemic Risk Monitor 1.0: A Network Approach,
forthcoming, 2015.

RISK DATA MANAGEMENT | AUGUST 2015

75

EFFECT OF CREDIT DETERIORATION ON

REGULATORY CAPITAL RISK WEIGHTS FOR
STRUCTURED FINANCE SECURITIES
By Vivek Thadani and Peter Sallerson

Vivek Thadani
Director, Structured Finance
Valuations & Consulting Group

Vivek is primarily responsible for developing and

maintaining analytical models for various asset
classes across the structured security space. Prior to
his current role, Vivek supported investor and asset
manager clients at Wall Street Analytics across CLO
and RMBS asset classes.

With regulatory stress testing becoming more entrenched in general

risk management, the need to understand the credit-specific drivers
of regulatory risk weights has become an important function of
risk management. This article aims to illustrate the general impact
of credit deterioration on regulatory capital risk weights in a large
dataset of multiple structured finance asset classes. For investors
and risk managers, any asset class-specific trends can help in the
investment evaluation process.
Criteria for the analysis portfolio

Peter Sallerson
Senior Director, Structured
Analytics and Valuation

Peter focuses on the CLO market at Moodys Analytics

via the further development of our Structured Finance
Portals CLO section and related research. He has
worked in many aspects of the CLO market for over
25 years.

the Federal Family Education Loan Program

deterioration on regulatory capital risk weights

(FFELP) government-guaranteed transactions

across the universe of structured finance

because the impact of credit deterioration

securities, we chose a large cohort of comparable

on these securities can be affected by policy

securities that would broadly illustrate trends

decisions. This would have introduced another

and effectively represent the universe as a whole,

dimension that was out of the scope of this

based on the following:

analysis.

1. The current outstanding notional amount

as of September 30, 2014 was at least US$1
million.
2. We excluded interest-only or combination
tranches, which would have made the
portfolio less uniform across asset classes.
Excluding these tranches also removed the
effect of cross-tranche referencing, a feature
of combination tranches.
3. To study the effect of credit deterioration
on regulatory capital, we excluded
resecuritizations that would have required
using a higher Simple Supervisory Formula
Approach (SSFA) supervisory calibration,1 so
that we could observe the effect of credit
deterioration in isolation. Further details on
the supervisory calibration parameter and its
impact on the SSFA formula can be found in
the Appendix.

76

MOODYS ANALYTICS RISK PERSPECTIVES

4. For student loan securities, we excluded

To understand the actual impact of credit

5. We included only USD-denominated

securities because credit quality can vary
significantly between the USD-denominated
securities in an asset class and similar nonUSD-denominated securities.
6. The final portfolio for analysis comprised
approximately 43,700 securities, which
effectively represent the structured finance
universe of non-agency transactions.2
As we expected, RMBS made up the largest
segment analyzed (by number of securities
and outstanding notional amount), owing to
the large size of the non-agency RMBS market.
Student loan ABS (SLABS) made up the smallest
segment.
By vintage range, the portfolio reflects broad
trends in issuance, with the large majority
of securities having been originated before
the crisis. For this analysis, pre-crisis covers

APPROACHES TO IMPLEMENTATION

Figure 1 Analysis portfolio by asset class

Number of Securities (thousands)

25

$1,000

Outstanding Notional (billions)

Number of Securities (thousands)

20

$800

15

$600

10

$400

$200

$0
SLABS

ABS

CLO

CMBS

HELOC

RMBS

Source: Moodys Analytics

securities originated in 2006 and earlier; crisis

1. 90 days or more past due

covers 2007-09; and post-crisis covers 2009

2. Subject to bankruptcy or insolvency

to the present. The large outstanding notional

proceeding

for a smaller number of securities in the post-

3. In the process of foreclosure

crisis bucket is due to the high bond factors (low

4. Held as real estate owned (REO)

seasoning) as compared to pre-crisis.

5. Have contractually deferred payments for 90

Current W parameter levels

days or more, other than principal or interest

For the SSFA for regulatory capital, the W

payments deferred on:

parameter represents the current delinquency

Federally guaranteed student loans,

and non-performing levels in a pool. As defined

in the Federal Register, the W parameter

in accordance with the terms of those

comprises loans that are:

guarantee programs

-OrFigure 2 Analysis portfolio by vintage ranges

$1,000

35
Number of Securities (thousands)

Outstanding Notional (billions)

30

Number of Securities (thousands)

$800
25
$600

20

15

$400

10
$200
5

Pre-crisis

Crisis

Post Crisis

$0

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

77

 Consumer loans, including non-federally

macroeconomic variables to exclude the

guaranteed student loans, pursuant to

dynamics of the different components of W

certain conditions

to macroeconomic stresses, which allows for

6. Are in default

a comparable evaluation of the regulatory

The SSFA formula requires normalization

impact on an entire asset class, as opposed to

deal-specific credit performance. For example, if

of a deals structure to its attachment and

detachment points, as well as normalization of
the credit risk profile to its W parameter. Hence,
for two identically structured deals i.e., two

we stressed only one macroeconomic variable,

such as home prices, we would expect a sharp
increase in the W parameter for an RMBS and

One key observation is that the stresses do not affect all asset classes
similarly; some can withstand such shocks across the rated structure better
than others. This observation is in line with what we expected that CLO
and ABS securities have, on average, better credit protection.
deals with similar attachment and detachment
points the current non-performing level
represented by the W parameter is the primary

HELOC security, but little change to a CLO

security. There could be indirect effects to the
macro-variable that drive corporate leveraged

driver of regulatory risk weight.

loan performance, but these effects would be

The risk weight is divided by 1250% to convert it

a macroeconomic shock would not affect all

to a regulatory capital charge. For this analysis,

RMBS deals uniformly, because underlying credit

we use the risk weight as the parameter to

quality differs. Stressing W directly illustrates the

analyze shocks to W. We used the risk weight

overall regulatory performance of an asset class.

minimal and likely delayed. Furthermore, such

instead of the capital charge because the risk

For the analysis, we used average levels instead

weight is the more common benchmark in

general risk management.

of medians. Although the median could be

Credit deterioration stresses to the W

parameter

average better illustrates the broad trends of

We stressed W instead of individual

deterioration does not necessarily affect all

considered a good indicator of trends, the

the stressed risk weights. Given that credit

Figure 3 Average W levels and average Aaa and Baa attachment points by asset class
Average W

Average Aaa Attach

Average Baa Attach

45%
40%

W Parameter

35%
30%
25%
20%
15%
10%
5%
0%

ABS

Source: Moodys Analytics

78

MOODYS ANALYTICS RISK PERSPECTIVES

SLABS

CLO

CMBS

HELOC

RMBS

APPROACHES TO IMPLEMENTATION

Figure 4 Average W levels and average Aaa and Baa attachment points by vintage range
Average Aaa Attach

Average W

Average Baa Attach

35%
30%

W Parameter

25%
20%
15%
10%
5%
0%

Pre-crisis

Crisis

Post Crisis

Source: Moodys Analytics

securities similarly, using a median would

for the asset classes. Specifically, the current

not demonstrate the true effect of the shock.

average W level helps identify an expectation

The effects of a stress on credit quality can

for credit deterioration shocks: changes to risk

differ, such that the median value will remain

weight in the poorer performing asset classes

unchanged but the risk weights of many

should be greater than in the better performing

securities will rise significantly. By using an

asset classes.

average, the value will move in accordance

Given that the SSFA formula assumes a fixed

with the segment overall and better

severity for the W bucket (see the Appendix),

demonstrate trends.

an alternate way to use the data is to gauge a

Figure 3 depicts the current average W levels

securitys ability to withstand credit shocks by

for the entire analysis portfolio, along with the

how much higher a security attaches (average

average attachment levels for Aaa- and Baa-

Aaa attachment and average Baa attachment)

rated securities. At a high level, it indicates

than by the average W bucket size (average W).

current performance and credit enhancement

Figure 5 W parameter components by asset class

90+ Delinquency

Bankrupt

Foreclosed

REO

Deferred Payments

Defaulted

30%
25%

W Parameter

20%
15%
10%
5%
0%
ABS

SLABS

CLO

CMBS

HELOC

RMBS

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

79

Figure 6 Average risk weight levels by asset class and original ratings

Average Risk Weight

Ba

Baa

Ba

Baa

Baa

Aa

Aaa
Aaa

Aa

Aaa

Aaa

200

Aaa

Aa

Ba

400

Aaa

Baa

600

Aa

Ba

Aa

800

Ba

Baa

Aa

1,000

Average Risk Weight %

1,200

Baa

1,400

ABS

SLABS

CLO

CMBS

RMBS

HELOC

Source: Moodys Analytics

Figure 7 Changes to average risk weight levels owing to credit deterioration shocks

Changes to Average Risk Weight

W+10%

W+20%

W+50%

90

80

Aaa
Aa
Aaa

60

Aa

50

MOODYS ANALYTICS RISK PERSPECTIVES

Ba

Ba

Baa

Baa
Ba

Source: Moodys Analytics

Ba
CLO

CMBS

RMBS

Baa
SLABS

Aaa
Aa
A

A
A
Baa

Aaa
Aa

ABS

80

Baa

Ba
10

Aaa

20

Aa

Aaa

30

Baa

40

Aa

Average Risk Weight %

70

HELOC

APPROACHES TO IMPLEMENTATION

Such a back-of-the-envelope approach allows

7.5% for the three credit deterioration scenarios.

us to quickly determine that ABS and CLO are

Using a proportional approach ensures that the

the only asset classes that have good credit

stress affects deals progressively i.e., better

protection at both the Aaa and Baa levels

performing deals are shocked by smaller stresses

(Figure 3). These levels of credit protection are a

while worse performing deals are affected by

function of how the transactions are structured

larger stresses. Figure 7 shows the results for the

and how their credit enhancement changes over

credit quality shocks.

time. Conversely, we can expect current HELOC

and RMBS performance to be fairly poor because
the average W levels are higher than the average
Aaa attachment levels.

Although the effects of the credit deterioration

stresses may appear to be minimal at the current
W levels, we note some interesting trends.
One key observation is that the stresses do

Similarly, the current credit performance of a

not affect all asset classes similarly; some

vintage segment shows good credit protection

can withstand such shocks across the rated

levels for post-crisis securities, as compared to

structure better than others. This observation

crisis and pre-crisis securities.

is in line with what we expected that CLO and

When reviewing the performance of an asset

class as defined by the W parameter, breaking
down the components of the level is also helpful.
In Figure 5, the various components highlight the
different makeup of the average W levels. This
dispersion is the primary reason for the decision

ABS securities have, on average, better credit

protection. The changes to risk weight owing
to credit quality shocks, therefore, are minimal.
Also, the performance of these securities in the
scenarios aligns well with actual performance
during the crisis.

to shock credit quality uniformly rather than by

The absolute change in stressed risk weight for

independent macroeconomic variables.

the poorer performing asset classes (such as

To gauge potential credit deterioration, we

analyzed the average risk weight by segmenting
the portfolio by asset class and the original
Moodys Investor Service rating levels, which
ranged from Aaa to B. We did not consider subratings (1 to 3).
Figure 6 shows the average current risk weight

HELOC and RMBS) are higher than the stress

applied. Although this view compares the
relative change in W to the absolute change in
risk weight a relationship that is not linear
it helps to put the risk weight changes into
context. While the change in risk weights is
higher for the poorer performing asset classes,
within an asset class such as RMBS or HELOC,

by the segments used in the credit shock.

the change in risk weight is low for lower rated

As expected, the SSFA-based risk weights are on

lower rated securities are closer to the risk

average higher for lower rating levels. For SLABS,

weight ceiling of 1250%.

securities. This is not surprising given that the

there were no securities originally rated Ba or B

that met the selection criteria. Also, for the ABS

Conclusion

buckets, very few securities were originally rated

There are a few different ways to interpret this

Ba and B and the average level skews to a low

analysis. From a regulatory perspective, the

and high value. This would not be the case for

overarching theme is that credit deterioration

one transaction the B security will always have

affects different asset classes differently. This

a higher risk weight compared to the Ba security

could be due to either the historical credit

in the same deal.

performance or the typical structure for an

asset class or both. While risk management

For this analysis, we ran three credit

professionals can use different segmentations to

deterioration scenarios using values of 10%,

analyze regulatory impact of portfolio changes,

20%, and 50% to shock the current W level.

this analysis highlights high-level trends that

For example, if a transaction had a current W

should be considered at every step of the

level of 5%, we used values of 5.5%, 6%, and

investment process.

RISK DATA MANAGEMENT | AUGUST 2015

81

Appendix SSFA Mechanics4

The Simplified Supervisory Formula Approach requires a simpler calculation and data collection
process. The trade-off for this is conservative assumptions on the losses of the underlying exposures,
which could result in potentially higher regulatory capital requirements. The SSFA calculation
requires the following input parameters:
1. Kg, which is the weighted average total base capital requirements of the underlying exposures
2. Parameter W, which is the ratio of the sum of underlying exposures that are seriously delinquent
or defaulted for regulatory purposes4
3. Parameter A, which is the attachment point of the security
4. Parameter D, which is the detachment point of the security
5. Supervisory calibration parameter p, which is set to 0.5 for securitization exposures and 1.5 for
resecuritization exposures (For this analysis, resecuritizations were excluded and the p was set to
0.5 for the entire portfolio.)
SSFA risk-based capital calculation:
Risk Weight =

KA - A
x 1250% +
D -A

KA=(1-w).KG+(0.5w)

a=-

u=D-KA

l=max(A-KA,0)

a.u
a.l
KSSFA = e - e
a(u - l)

D - KA
x 1250% x KSSFA
D -A

1
p x KA

1 Although we excluded multi-tranche resecuritizations that would have required using a higher SSFA supervisory calibration
parameter of p=1.5, we did include all single-tranche re-remics with p=0.5.
2 We analyzed the portfolio using the Regulatory Module in the Moody's Analytics Structured Finance Portal.
3 See Federal Register, Vol. 78, No. 198, October 11, 2013..
4 For more information, see Federal Register, Vol. 78, No. 198, October 11, 2013.

82

MOODYS ANALYTICS RISK PERSPECTIVES

RISK
PRACTITIONER
CONFERENCE

2015 Moodys Analytics, Inc. and/or its licensors and affiliates. All rights reserved.

RPC
15

MOODYS ANALYTICS
RISK PRACTITIONER
CONFERENCE

JOIN THE CONVERSATION AT THE

Moodys Analytics
Risk Practitioner
Conference
Our annual Risk Practitioner Conference brings
together industry experts and leading risk
practitioners from across the globe.
Register today at
[Link]/RPC2015

RISK DATA MANAGEMENT | AUGUST 2015

83

FINDING ALPHA: SEEKING ADDED VALUE

FROM STRESS TESTING
By Greg Clemens and Mark McKenna

Greg Clemens
Director,
Stress Testing Solutions

As a member of the Stress Testing Task Force at

Moodys Analytics, Greg helps clients automate their
stress testing processes providing insight about
architectures, data management, and software
solutions for risk management.

Mark McKenna
Senior Director,
Stress Testing Solutions

Mark is responsible for business and relationship

management for Moodys Analytics Stress Testing
Solutions. Prior to his current role, he was with the
Structured Analytics & Valuations Group, leading
valuations and cash flow analytics projects. He has
worked with many US financial institutions, providing
risk management tools and loss estimation models for
their retail portfolios.

How can banks measure the success of their stress testing efforts?
This article explores where banks can look for the alpha in stress
testing that is, how they can measure the performance of their
stress testing programs, identify weaknesses, and make the process
more efficient and effective.
Introduction

If stress testing were like a ski competition,

In sports, a teams success is measured by

each skier (in this case, each bank) would race

winning percentages, and an individual

down their own individual course, there would

players, by statistics such as batting average,

be no clock, and each would be declared a

yards gained, or points scored. In business,

winner if they just made it safely to the bottom

success is measured by profitability. For banks,

of the course.

that specifically means net profit from fees

and interest and return on investments.

test program succeeded beyond simply passing

alpha to gauge their performance, which is

the test. Rather than compare one banks

a measure of how much better their return is

performance to anothers, it makes more sense

versus a benchmark index. For example, if a

to compare the banks own performance from

benchmark index increases five percent, but

one year to the next.

a mutual fund generates a return of seven

percent over the same period, that fund
would have an alpha of two percent, meaning
underlying index.
But how can we measure the performance of
a banks stress testing efforts? Although its
not as simple as measuring a batting average,
there are some best practices that banks can
adopt to help them find quantifiable alpha
factors.
Stress testing is more than a pass/fail exercise
For US banks subject to the Dodd-Frank Act
Stress Test (DFAST), the results are pass/fail
for each bank; that is, for the banks that pass
the test there is no merit to being first, second,
third, or last, for that matter.

MOODYS ANALYTICS RISK PERSPECTIVES

we need to measure how well the banks stress

Portfolio managers use the concept of

that it performed that much better than the

84

To assess a banks stress testing performance,

To quantify how a banks program is

improving (taking for granted that it passed
the DFAST test), we need to find the alpha in
stress testing.
Streamlining the stress testing process
In terms of an organizational framework, a
typical stress testing process involves:
Gathering data from across the firm not
just finance, risk, and treasury but all lines
of business
Preparing an initial balance sheet using the
jump-off data
Forecasting what the balance sheet will look
like in the future in a variety of economic
scenarios using models for projections of
losses, net income, pre-provision net revenue,

APPROACHES TO IMPLEMENTATION

Processes

Figure 1 Stress testing process framework

Financial
Planning

Assess
Models

Gather
Jump-off
Financial
Data

Prepare
Scenarios

Run
Forecasts

Consolidate
Results

Prepare
Outputs and
Documentation

Publish
Reports
and Plans

Source: Moodys Analytics

cash flows, and other elements of the

an enterprise-wide exercise. Organizations

balance sheet

must access, validate, and reconcile data from

Incorporating proposed capital plans,

overlays, and expert judgment
Preparing reports and supporting
documentation to fully explain how the
forecasts were derived

across the enterprise, including all geographies,

portfolios, and instruments, irrespective of the
origin of the data.
On top of these data aggregation challenges,
firms need to widen the scope and improve the

The process seems straightforward, but in

accuracy and governance of their models and

reality it involves many stops and starts,

estimation processes for generating the forecasts

revisions, and iterations.

needed for the stress tests. These challenges are

A strain on resources: the data challenge

straining firm resources even further.

Developing and running a stress testing process

To put the required effort in perspective: We

is a daunting challenge for most banks. Multiple

believe that the largest US banks are running

regulatory reforms with complex oversight and

their stress testing programs year round, with

compliance guidelines have added to the already

upwards of four or five hundred full-time-

difficult challenges of risk management for

equivalent resources engaged in the program for

financial institutions.

much of that time.1

Government urgency to ensure that financial

Thwarting the success of these efforts is the

systems are safe and stable has prompted

data itself. Data exists in many formats, such

continued enhancements to the relatively new

as paper records, desktop files, and other

regulations even as they widen in scope. This

suboptimal data storage solutions, which adds

heightened regulatory expectation and intense

to the difficulty of efficiently meeting reporting

scrutiny come at a time when organizations

obligations. Organizations often need to build

are already under pressure to improve their

additional environments to pull data from

profitability and establish a competitive edge in

different locations for reporting purposes, as

a low-return market environment.

shown in Figure 2.

To put the required effort in perspective: We believe that the largest US

banks are running their stress testing programs year round, with upwards of
four or five hundred full-time-equivalent resources engaged in the program
for much of that time.
As a result, stress testing budgets at most

Primitive data systems, which are characterized

financial institutions have soared. These

by inconsistent standards, data duplication,

regulations have one characteristic in common:

and missing and conflicting data, lead financial

They require that financial institutions set

institutions to make major and potentially

up processes and systems to manage an

erroneous assumptions about how to reconcile

ever-growing amount of data and oversee

data. Firms must also contend with the challenge

RISK DATA MANAGEMENT | AUGUST 2015

85

Figure 2 Challenges of data standardization

Standardized
Format

_:

+ =

Source: Moodys Analytics

of ensuring consistency across reporting dates

data for each type of risk. A firm must create a

and between different reports.

structure and processes that can aggregate risk

One of the principles put forth by the Basel

Committee on Banking Supervision (BCBS)
states that organizations should strive to
determine a single authoritative source of risk

data in a way that is accurate, complete, and

transparent for its senior management, board
of directors, and regulators enabling these
stakeholders to make informed decisions.

Figure 3 A centralized datamart allows for a much simpler process

_:

+ =

Source: Moodys Analytics

86

MOODYS ANALYTICS RISK PERSPECTIVES

APPROACHES TO IMPLEMENTATION

A centralized datamart that connects different

necessarily include the strategic investments

pieces of information is key to mitigating the

critical to delivering a more sustainable and

challenges of data management and can clear a

cost-efficient business model. Compliance

path for banks to determine quantifiable stress

strategies that address only current needs could

testing performance measures.

lead to unintended downstream consequences

and additional costs. As weve shown, this

Striving for repeatability

The first step to passing regulatory requirements
is to compile the numbers. The next is to ensure
the transparency of the process, so that the
bank can explain easily and clearly where the
numbers came from and how the forecasts were
derived. The final major step is to make sure the

process isnt straightforward; it is iterative and

complicated, and requires substantial time and
labor, all of which makes it hard for banks to see
beyond the primary goal of satisfying regulatory
requirements and thus to realize secondary
benefits or find alpha in the process.

process is auditable: The bank needs to be able

Making the process repeatable, in addition to

to make sure the process can be audited to show

streamlining it, will minimize the time and cost

how everything came together, with clear and

of the exercise. It also allows for more what-if

detailed documentation showing what estimates

and sensitivity analyses, which can provide more

and models were used, what parameters and

clarity into the banks forecasts and improve its

defaults were included, how they were validated,

test results.

how overlays and expert judgment were applied,

and how all decisions were made throughout
the process.

Stress testing results: a missed opportunity

Stress testing should be seen within the wider
context of the efforts banks put into improving

The tactical approaches many organizations

their risk management capabilities. Banks face a

use to meet such increasingly complex

number of common challenges, including

regulatory and accounting requirements dont

the following:

Figure 4 Uses of a stress testing program

CCAR Banks

DFAST Banks

100%

Areas where FRB

would like to see
greater use

80%

60%

40%

20%

0%

Regulatory
Compliance

Capital
Adequacy
Measurement
and Planning

Risk Appetite
Definition

Risk
Management
and
Measurement

Limit Setting
and
Measurement

Financial
Planning /
Budgeting,
Strategic
Planning

Portfolio
Structuring

Pricing

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

87

 Significantly enhancing data and systems

stress testing into their business in these areas.

Improving risk governance and board

Many simply dont have the time or resources to

oversight
Integrating approaches to risk management
Enhancing stress testing methodology
Changing stress testing processes and
operating models
Extending reporting capability
Ideally, overcoming these challenges would
allow banks to derive value from their stress
testing programs beyond merely responding to
regulatory requirements and passing tests but
this is seldom the case. Last year, we conducted
surveys of large and mid-sized banks in the US.
As Figure 4 shows, one question we asked was
how the banks were using their stress

go beyond the regulatory requirements.

If risk managers did see ways to derive insights
from stress testing that could help them run
the bank, this would be a form of alpha. But for
the most part this isnt the case, and any value
realized beyond satisfying compliance goals
would thus be hard to quantify.
Conclusion
Making the process more efficient easier and
less time-consuming will mean that stress
testing takes up less of a banks resources. So
perhaps this is where banks can start to look
for alpha. Saving time for key resources by

testing programs.

making the stress testing process more business-

When we asked the banks what they were using

working for the regulators and have more

their stress testing results for, they all said

time to be creative and focus on the business of

regulatory compliance; most also said capital

the bank.

planning. But the degree to which banks use

stress testing results quickly drops off after that.
Not only did fewer banks mention that they
were using stress testing for other purposes, such
as financial planning and budgeting, but the ones
that did were much less emphatic about how
they used stress testing in these areas compared
to compliance.
Regulators have indicated they would like to see
banks use stress testing for other things, like risk
appetite definition, limits, and risk management
in general. Unfortunately, there seems to be a
long way to go before banks start incorporating

as-usual means a banks key people can stop

Developing automated, well-governed processes

for stress testing will lead to better results
with more transparency, auditability and
repeatability, and in less time. Banks will be even
better positioned to achieve the primary goal of
compliance, with less cost and effort.
Integrating the best practices outlined in this
article can help banks streamline their processes,
freeing up time for key resources. That is, they
may change their stress testing race course,
making it easier and faster to run through, and
find their stress testing alpha.

1 Jamie Dimon, Annual letter to shareholders, p12, April 9, 2015. He mentioned the cost of stress testing at JP Morgan.

88

MOODYS ANALYTICS RISK PERSPECTIVES

 2015 Moodys Analytics, Inc. and/or its licensors and affiliates. All rights reserved.

With regulatory and industry change accelerating, banks need

to ensure their enterprise risk management practices also
maximize opportunities, drive growth, and fuel the next big idea.
Moodys Analytics helps more than 150 global banks
manage risk, achieve regulatory compliance, and
make better informed, risk-aware business decisions.

[Link]/smallrisk2015

Essential insight serving

global financial markets

RISK DATA MANAGEMENT | AUGUST 2015

89

PRINCIPLES AND
PRACTICES
Learn effective practices for applying risk data management to your
organization, including approaches to overcoming common challenges.

MODELING TECHNIQUES AND TOOLS

IN SCENARIO-BASED RISK APPETITE
MANAGEMENT
By Pierre Gaudin

Pierre Gaudin
Senior Director, Enterprise Risk
Solutions, APAC

Pierre is in charge of strategic initiatives in Asia-Pacific

for risk appetite management, stress testing, and
origination at Moodys Analytics, assisting with subject
matter expertise, clients requirement analysis, and use
case illustrations.

To get senior stakeholders to buy in to alternative macroeconomic

scenarios, risk management and ALM teams must assemble risk
models and risk-adjusted performance measurements in their
simulation tools. Institutions must switch from a qualitative to a
quantitative approach to analysis so they can effectively define
risk appetite. This article addresses these issues, as well as building
repeatable measurements, resolving data gaps, using data flow
automation tools, and implementing processes to enforce and
monitor such measurements.
Introduction

credit default swap may be measured against

The regulatory stress testing requirements that

the probability-weighted losses incurred in a

have been published in the US, Europe, and soon

sovereign-default scenario.

in Asia-Pacific are increasingly guiding financial

institutions toward scenario-based governance
and risk appetite management. From an internal
practice perspective, management information
reports are now expected to articulate a
consistent set of profitability and risk forecasts
for different time horizons. Governance practice
is therefore shifting from a qualitative approach
to a quantified framework, which evaluates the
sustainability of the institutions compliance
and ability to deliver value to its shareholders,
through the cycle.
The overarching goal of a risk appetite
framework is to provide senior management
with a quantitative assessment of
profitability, budget, and dividends in different
macroeconomic assumptions. Such a framework
outlines a variety of scenarios (e.g., US liquidity
crisis, euro zone sovereign default, or a recession
in China) and provides potential mitigation
actions. Institutions can then make a decision,
taking into account the cost of hedging
compared to the likelihood and severity of the
scenario. For example, the cost of a sovereign

92

MOODYS ANALYTICS RISK PERSPECTIVES

Synchronizing profitability with risk forecasts in

a macroeconomic scenario presents a significant
organizational challenge. Indeed, aside from
combining simulations across risk management
and asset and liability management systems,
measurements need to account for the
consistent effects on market factors, credit
transitions, and transaction volumes across
the usual organizational silos. This challenge
sometimes attracts such focus that key project
risks are overlooked, such as data gaps.
Scenario narrative and likelihood calibration
Scenario narrative and severity have traditionally
been expressed in terms of frequency, such as
once-in-seven years market downturn, oncein-twenty-five years commodity crisis, or
once-in-a-hundred years sovereign default. This
practice, however, has reduced stress testing to a
repetitive exercise, lacking the ability to account
for the evolution of the economy from its current
state, worldwide and locally.
A more informed approach consists of centering
the construction of scenarios around a baseline

PRINCIPLES AND PRACTICES

Figure 1 Calibration of scenario likelihood around a baseline consensus

Alternative Economic
Scenarios

Event-Driven
Emerging
Markets
Hard Landing

Sovereign
Default
Shock

In Line with
Regulatory
Guidelines
Baseline:
Recession

Simulation-Based
S4:
Severe
Double Dip
1-in-25

1:100

1:25

S3:
Double
Dip
1-in-10

1:20

S2:
Mild
Double Dip
1-in-4

1:10

1:4

S1:
Stronger
Recovery
1-in-4

Forecast

1:4

Healthier
Economy

Weaker
Economy
Source: Moodys Analytics

outlook, representative of the economists

portfolio indicators, scenario definitions need

consensus, increasing the relevance of the

to be assorted with observable time-series of

stress testing exercise to the current situation.

macroeconomic and market factors that can be

Revisited monthly or quarterly using the latest

drilled down geographically to country and

macroeconomic data and economists opinions,

city levels.

alternative macroeconomic scenarios are then

built using stochastic analysis, whereby shocks
are applied throughout global macroeconomic
models and measured against the contours
of previous business cycles. As a result, each
scenario is calibrated in terms of likelihood
against the distribution (Figure 1).

This framework can also be used for regulatory

scenarios such as the Federal Reserves
Comprehensive Capital Analysis and Review
(CCAR) scenarios, the Financial Services
Authoritys Anchor scenarios in the UK, the
European Banking Authoritys stress scenarios,

Credit time-series augmentation techniques that use credit estimates

based on market prices can significantly improve credit, liquidity, and
profitability models leveraged in the risk appetite framework. These
techniques are available not only for publicly listed firms, but also for
private firms and small- and medium-sized enterprises, as well as sovereign
entities.
With this method, each scenario deemed

the International Monetary Funds scenarios,

relevant by senior stakeholders can be extracted

as well as others proposed by local authorities.

from the distribution, along with the economic

Using the same approach, banks can benchmark

narrative explaining the possible causes of the

and leverage regulatory scenarios throughout

scenario compared with the baseline outlook.

the stress testing exercise using a single

Ideally, to allow meaningful regressions to key

framework.

RISK DATA MANAGEMENT | AUGUST 2015

93

Figure 2 Interpolating internal ratings with market-price-driven credit time-series

Market value of assets or CDS spread-based ratings

Financial statement-based ratings

Sticky
Transitions

Responsiveness Market Opinion

AA

AA+

AA+

Events

Source: Moodys Analytics

Figure 3 Key portfolio indicators at different time horizons

A
Sellable
Securities
Gap

Liquidity

L
P

1M
Net Incomes
- Net Interest Income / Fees
- Operating Expenses
- Expected Loss/Provision

1Y

Profitability

Capital Adequacy
Credit /Market
Operating Losses

L
Source: Moodys Analytics

94

MOODYS ANALYTICS RISK PERSPECTIVES

Dividends
Capital

Granular
Transitions

PRINCIPLES AND PRACTICES

Resolving data gaps

need to be revisited so as to allow a proper

A key aspect emerging from the practical

scenario-driven forecast.

application of regulatory stress testing is that

risk and performance models need to establish
how credit behaviors, liquidity cash flows,
market risks, profitability, and budget forecasts
are related to macroeconomic time-series, by
leveraging the historical time-series observable
in the portfolio. While most of this data can be
well described, for most institutions the only
source of credit data is related to their internal
ratings practice, which is based on quarterly
financial statements. With at most one point per
quarter, the resulting credit time-series imply
very static and insensitive credit behaviors,
leading to significant noise both in the elasticity
models of credit transitions and in the evaluation
of correlations. This affects not only credit
forecasts, but also the subsequent liquidity
behavioral models, which are based on credit
ratings, as well as profitability adjusted for
credit losses.

Credit time-series augmentation techniques

(Figure 2) that use credit estimates based on
market prices can significantly improve credit,
liquidity, and profitability models leveraged in
the risk appetite framework. These techniques
are available not only for publicly listed firms,
but also for private firms and small- and
medium-sized enterprises, as well as
sovereign entities.
Experience shows that this approach can
deliver robust statistical regressions against
macroeconomic assumptions, as well as
correlations, providing for consistent forecasts
over different time horizons. It ensures
quality and repeatability, allowing senior
stakeholders to understand trends, acquire
reference points, and build trust in the
numbers and practice.

On the technology side, data flow automation is becoming increasingly

necessary. Institutions are streamlining their computation flows for both
internal and regulatory purposes, in areas of strategic planning, credit
portfolio management, asset and liability management, and liquidity risk
management.
Analysts often assume that the best data

Forecasting key portfolio indicators

available inherently includes such data gaps

Scenario-based risk appetite management

and time-series deficiencies. However,

leverages multiple measurements according

experience in loss forecasting shows that

to different time horizons: short-term liquidity

under-sampled historical time-series have a

compliance, medium-term net revenue,

significant impact on the consistency of model

income volatility, dividend sustainability and,

outputs. Practitioners faced similar challenges

in the long-term, capital adequacy (Figure

in modeling forecast losses in economic capital

3). These reports require a comprehensive

measurement. As a solution to data gaps, time-

description of risks that examines the

series augmentation techniques have proved

relationship between macroeconomic factors

efficient in delivering consistent reports over

and key portfolio indicators.

time. This is even more relevant considering that

regulators have used this technique to calibrate

In liquidity modeling, the behavior of a

current parametric regulatory functions.

counterparty depends highly on its own credit

Overall, project risks due to data gaps in credit

cash flow demands a precise description

time-series cannot be overstated. Traditional

of credit transitions. This is illustrated, for

modeling and simulation practices for liquidity

instance, in regulatory liquidity coverage ratio

and risk-adjusted performance measurements

(LCR) calculations, in which the eligibility of

situation. Therefore, forecasting behavioral

RISK DATA MANAGEMENT | AUGUST 2015

95

bond positions for the liquidity reserve is tied to

sensitivities are assessed by applying calibrated

the credit assessment of the issuer, and in which

shifts on market data.

inflows and outflows depend on the past-due

status of the contracts. As a result, the accuracy
of liquidity-monitoring models depends on the
ability to evaluate realistic credit transitions
over a time horizon as short as 30 days.
The use of quarterly financial statements
can lead to a significant underestimation
of volatility. Overall, credit and behavioral
models are reflected at each time horizon,
impacting liquidity, then profitability, and as
a consequence capital adequacy and dividend
sustainability (Figure 4).

In stress testing, however, whether key portfolio

indicators represent a median or a tail-risk
assessment, forecasting models typically
provide expected outcomes for each scenario
assumption. In a recession scenario, for instance,
an institution might forecast a decrease in
the LCR to 105% within a year. In this case, a
key risk managers would need to anticipate is
how narrowly distributed security prices will
be around the expected value, so as to gauge
the likelihood that the liquidity compliance
threshold will be breached, even though the
expected LCR value is compliant.

Assessing volatility
Risk measurements in the current portfolio
can generally be described as either an
expected value or a value-at-risk within a risk
distribution. Such a distribution is typically
built on through-the-cycle assumptions,
reflecting cyclical or long-run behaviors, while

Providing this additional distribution through

stochastic modeling might seem like a vast
undertaking, given the wide range of simulation
inputs (from macroeconomic and market
factors to creditworthiness and budget figures).

Figure 4 Key portfolio indicators under stress

A
Inflation Scenario
Deposit Run Off

Sellable
Securities

Liquidity

1M

P
Net Incomes
- Net Interest Income / Fees
- Operating Expenses
- Expected Loss/Provision

1Y

Capital Adequacy

Capital Adequacy
Credit / Market
Operating Losses

L
Source: Moodys Analytics

96

MOODYS ANALYTICS RISK PERSPECTIVES

Dividends
Capital

PRINCIPLES AND PRACTICES

Figure 5 Impact of credit transitions on liquidity reserves volatility

Sovereign
Default
Shock

In line with
Regulatory
Guidelines

Emerging
Markets Hard
Landing

Baseline:
Recession
S4: Severe
Double Dip
1-in-25

1:100

1:25

S3:
S2: Mid
Double Dip Double Dip
1-in-10
1-in-4

1:20

1:10

1:4

Forecast

1:4

LCR Compliance
in Stress Scenario
Level 1 & 2 Assets:
Minimum
Requirements

Distance to
Compliance

HQLA Portfolio

LCR

Source: Moodys Analytics

However, undertaking a comprehensive Monte

explain a significant part of the volatility

Carlo process across risks can lead to excessive

in the LCR. Even if high-quality liquid asset

or false precision, misaligned with the

(HQLA) positions are replaceable, it is

simulation time horizon and other key stress

worth simulating how risk can build up by

testing assumptions.

monitoring an extended set of positions

Because a key purpose of the stress testing

framework is to identify and quantify outcomes
of drastic but plausible situations, it is relevant
to focus on the key contributors to volatility
during a crisis. Spikes in market prices and

and possible replacement issuers. For this

purpose, running an analysis of credit valueat-risk on the HQLA portfolio (pre-haircut)
provides a good gauge of the LCR forecast
distribution (Figure 5).

credit downgrades explain a significant part of

Medium-term forecasts

such volatility, so a stochastic simulation of

Analyzing earnings-at-risk traditionally

macroeconomic-driven credit transitions can

encompasses gauging the adverse impact

provide a good understanding of volatility for

of interest rates and exchange rates onto

each risk and time horizon: short-term liquidity,

net interest income forecasts. By adding

mid-term profitability, and long-term

the effect of unexpected credit losses to the

capital adequacy.

earnings in each scenario, the simulation

Short-term forecasts
Credit transitions in the liquidity reserve

provides a comprehensive assessment of the

volatility in forecast incomes (Figure 6).

RISK DATA MANAGEMENT | AUGUST 2015

97

Figure 6 Impact of credit transitions on risk-adjusted earnings volatility

Sovereign
Default
Shock

Emerging
Markets Hard
Landing

In line with
Regulatory
Guidelines
Baseline:
Recession

S4: Severe
Double Dip
1-in-25

1:100

1:25

S3:
S2: Mid
Double Dip Double Dip
1-in-10
1-in-4

1:20

1:10

1:4

Forecast

1:4

Probability
Density

Credit Risk Contribution

to Income Volatility

Loss

Source: Moodys Analytics

Long-term forecasts

reference points, and build trust in the numbers

Capital adequacy is expressed through

and the practice. This process then allows

operational, market, and credit value-at-risk.

institutions to switch from a qualitative to a

Running a tail-risk simulation can help provide a

quantitative approach to risk appetite

credit loss tail distribution, thereby affording a

analysis. In the modeling exercise, data gaps

clear understanding of volatility for the forecast

have a significant impact, for which best

of capital requirements.

practices in econometric augmentation are a

Conclusion

98

MOODYS ANALYTICS RISK PERSPECTIVES

proven solution.

Overall, for the purpose of extending the

On the technology side, data flow automation

dialogue with senior stakeholders to alternative

is becoming increasingly necessary. Institutions

macroeconomic scenarios, risk management

are streamlining their computation flows

and asset and liability management teams are

for both internal and regulatory purposes, in

required to work closely together to assemble

areas of strategic planning, credit portfolio

risk models and risk-adjusted performance

management, asset and liability management,

measurements in their simulation tools. Key

and liquidity risk management. They are taking

stakeholders and supervisors need repeatable

advantage of data flow automation tools that

measurements, ensuring that the assumptions

help institutions with regulatory and internally

in the forecasts are consistent over time and

driven stress testing initiatives, handling

allowing them to understand trends, acquire

scenario libraries, driving inputs for each

PRINCIPLES AND PRACTICES

Figure 7 Impact of credit transitions on capital requirement volatility

Sovereign
Default
Shock

Emerging
Markets Hard
Landing

In line with
Regulatory
Guidelines
Baseline:
Recession

S4: Severe
Double Dip
1-in-25

1:100

1:25

S3:
S2: Mid
Double Dip Double Dip
1-in-10
1-in-4

1:20

1:10

1:4

Probability
Density

Forecast

1:4

Credit Risk Contribution

to Capital Volatility

TRC (e)

TRC(e)

K (e)

Loss

Source: Moodys Analytics

computation engine, and running parametric

an enterprise-wide and consistent limit-

regression models from macroeconomic

monitoring framework that translates risks into

scenarios into key indicator forecasts.

exposure limits in different business lines and

The final consideration, related to a

quantitative formulation of risk appetite, is the
need for processes to enforce and monitor such
measurements. Concentration monitoring and
risk appetite limit-setting are excellent starting

units, market segments, industries, geographies,

and currencies. This allows risk managers and
front offices to improve performance and control
the build-up of risk in the portfolio at the point
of origination.

points. After leveraging a data aggregation

initiative, a logical next step is to implement

RISK DATA MANAGEMENT | AUGUST 2015

99

THE BENEFITS OF MODERNIZING THE

COMMERCIAL CREDIT DECISIONING PROCESS
By Buck Rumely

Buck Rumely
Senior Director,
Client Management

Buck leads the Americas team of credit and technology

specialists at Moodys Analytics. He has helped design
credit and risk management systems for a variety of
financial, governmental, and energy firms throughout
North and South America. He has published papers on
credit risk, including Risk Magazine.

Regulators and auditors expect banks data submissions to be more

detailed than ever before. However, many banks still labor under
outdated credit decisioning systems black holes in which valuable
loan data disappears and can no longer be used for critical processes
such as stress testing. This article explains the benefits of an online
decision system to deliver higher returns on risk while making
regulatory compliance easier and cheaper.
Introduction

files where it cant be re-used for more critical

Commercial bankers understand that granting

processes, such as stress testing, covenant

a loan is an iterative and dynamic process,

monitoring, or model validation.

not a distinct event with a simple yes or

no outcome. It involves many data inputs
and outputs, as well as examination of risk
and revenue tradeoffs. A facility often evolves
substantially before finalization.

customer relationship management (CRM),

core systems, deposit and exposure systems,
financial statement spreading systems, and
scoring systems. This aggregated information

once a bank finalized the facility, made the credit

is frequently used only to facilitate credit

decision, and released the funds. Bank credit

committee decisions and is not conveniently

policies typically required an annual review, at

stored in one system. Banks thus lose invaluable

which time the bank would update the borrower

opportunities to repurpose a rich dataset for

rating and close the loan file for another year.

meaningful activities that could ultimately

Periodically and often haphazardly the

increase revenues and greatly lower compliance

banks staff checked the compliance status of

and audit costs.

ignored them). Banks rarely placed enough

clean, consistent, and quality data in a
searchable system to determine covenant
compliance without having to manually reopen
a credit file.

MOODYS ANALYTICS RISK PERSPECTIVES

document comes from many areas, including

Traditionally, a loan file was essentially closed

the loan covenants (or, being overwhelmed,

100

The data on a typical commercial loan decision

Credit decision data can help answer

regulators questions
The recent financial crisis revealed that some
banks did not electronically store data from
the credit decisioning process and lacked
systems to track covenant compliance. In

Today, the data from the loan decisioning

addition, regulatory expectations for data

process for complex commercial credit facilities

retention, storage, and reporting have grown

is still rarely aggregated in a searchable,

considerably indeed, both regulators and

reportable, and auditable system even at

auditors are increasingly requiring that banks

sophisticated banks. Instead, this data is

capture and store all key data points and

manually loaded into Excel or Word documents

collateral information associated with making

from various source systems and left in flat

a commercial loan decision. Antiquated and

PRINCIPLES AND PRACTICES

standalone systems no longer meet these

demands, much less optimize revenues.
The questions regulators ask banks might

testing and model validation

More consistent underwriting and better
return on risk

seem easy, but experienced bankers know that

We will examine each of these benefits in detail,

they can be difficult to answer, owing to the

as well as some of the challenges that banks may

limitations of their systems. Among some of

encounter when switching their systems.

the simple but challenging regulatory and audit

questions are the following:

Faster loan approvals increase loan closure

rates and productivity

How many loans are guaranteed by the same

Loan approvals often face bottlenecks, whether

guarantor, for example, by a high net-worth

from multiple approval requirements because

individual or real estate developer?

deals exceed credit authority or a key credit

How many loans comply with covenants?

How many do not?
What credit decisioning data does the bank
have for model validation?
What is the banks direct and indirect
exposure to a given customer or financial
institution?
Can the bank stress the inputs to its rating
models for enterprise stress testing?
Can the bank recreate its rationale for a
commercial loan decision?
How many loans are related to a specific
customer?
Why does the bank have multiple spreads or
credit files for a given customer? Which one
is correct?

officer is on vacation. Modern credit decisioning

systems can assign approvals to the appropriate
credit officer and reroute requests when
resources are out of the office.
Bankers can use online technology to improve
the speed and accuracy of their loan decision
making process, earning them more business.
Credit teams can add new approvers on the fly
(or the system can do this automatically) and
establish a service level, or a schedule that
outlines when the tasks required in the credit
process have to be completed. For example,
if the service level for financial statement
spreading and analysis is four hours, the deal
team can expect turnaround within that time
span. Service levels can be tracked to identify

We know of one leading commercial bank that employs 20 full time

workers to aggregate and clean data in preparation for the FR Y-14Q
quarterly commercial data submission. What if the bank had a clean,
aggregated, and reliable source of data? Both regulatory compliance costs
and data error rates would decline, and the bank would be able to more
meaningfully deploy resources to more profitable tasks.
Enter the era of modern online commercial
credit decisioning

bottlenecks and improve productivity.

Answering these questions is challenging, but

These and other process efficiencies can shorten

improving the credit decisioning process will

the cycle for credit decisioning. If a complex

make it easier and will also provide banks a

commercial credit decisioning process can be cut

number of benefits, among them:

from 22 to 12 days, productivity and profitability

Faster loan approvals, which will increase the

banks loan closure rate and throughput
Automated covenant monitoring and
reporting

will increase dramatically, which could boost

loan throughput significantly.
Automated covenants monitoring and reporting
Regulators and auditors have expanded their

Lower regulatory compliance costs

scrutiny of covenant monitoring and reporting.

Ability to re-use origination data for stress

For example, regulators are increasingly

RISK DATA MANAGEMENT | AUGUST 2015

101

Figure 1 Decision cycle length for large commercial loans: before and after modernization

Average Application to Close Cycle - Commercial Loans

25

20

Days

15

10

Before Modernization

After Modernization

Source: Moodys Analytics

issuing Matters Requiring Attention (MRAs)

policy with fewer but more effective

to commercial banks to improve systematic

loan covenants!

monitoring and compliance reporting for both

financial and non-financial covenants.

Decreased regulatory compliance costs

An efficient credit decisioning process will

data and documents in an unstructured or even

automatically capture covenants at the point

imaged process. Given that regulators now

of credit underwriting and monitor them

require more information about the rationale

throughout the life of the loan. Banks can

and all of the data involved in a commercial

select from a library of standard covenants

lending decision, however, these systems are

or customize them for a customers specific

inadequate and unable to deliver the information

risk attributes. Integration with the spreading

in the format required.

process can automatically test financial

covenants when a borrowers monthly,
quarterly, or annual financial statements are
analyzed. Portfolio- or business line-level
reports can automatically identify customers
who do not meet covenant requirements.
Banks can track the entire covenant resolution
process granting customers a grace period,
giving them the opportunity to cure the
covenant, and monitoring the cure periods so
that they can determine how to improve or
expedite the process.
Moreover, having this data and history at hand
provides banks another powerful benefit: They
can more quickly adjust their credit policies
to eliminate covenants that do not result in a
meaningful reduction of risk. Imagine a credit

102

MOODYS ANALYTICS RISK PERSPECTIVES

Historically, most banks stored credit decision

Most underwriting and decisioning systems were

designed for a single purpose credit decisioning
and not for submitting data in bulk to the
regulator. This has led to banks hiring dozens
of people to cut, paste, and audit underwriting
data for consistency and de-duplication as they
prepare information for submission to regulators.
We know of one leading commercial bank that
employs 20 full time workers to aggregate
and clean data in preparation for the FR Y-14Q
quarterly commercial data submission. What if
the bank had a clean, aggregated, and reliable
source of data? Both regulatory compliance
costs and data error rates would decline, and the
bank would be able to more meaningfully deploy
resources to more profitable tasks.

PRINCIPLES AND PRACTICES

Re-use of decisioning data for stress testing and

model validation

and boosting its commercial portfolio profits.

One essential element required for both stress

can help banks spot performance trends in

testing and model validation is good, clean,

borrowers financials and also regional or

standardized data. A modern credit decisioning

industry trends.

Readily accessible portfolio-level reporting

system can deliver this data in droves to stress

testing and quantitative teams, which can then

With reliable underwriting and loan performance

build and update more powerful and

data and an online decisioning system, a bank

relevant models.

can segment its portfolio by region or industry

and quickly analyze data and identify bright

Banks can use data validation rules in the

spots in the market. For example, during

origination process so underwriters dont

the financial crisis, many banks abandoned

mistakenly enter incorrect data. The data

commercial real estate owing to the housing

from credit decisioning can be used to build

market crash; medical office loans, however,

bottom-up or top-down stress testing models.

performed well throughout the crisis. Banks

Historical data on a borrowers financial status,

can adjust their industry and regional portfolio

loan performance, and other key data elements

composition to beat the competition in

can be quickly exported to model development

promising market segments.

environments to build more relevant models.

This lowers the costs of ad hoc data requests

Conclusion

from quantitative teams because the system

A modern, more powerful on-line credit

provides the crucial data to the modeling teams

decisioning process can help improve a banks

on the back end.

commercial portfolio performance, no matter

the economic condition, through a combination

Delivering a higher return on risk

of increased revenue, process efficiency, and

A modern credit decisioning system can help a

lower regulatory compliance costs.

bank identify trends, giving it a competitive edge

DATAMARTS OVER DATA WAREHOUSES

Many banks assume that an enterprise data warehouse will help with the creation of an online decisioning system. Despite heavy investments
in data warehouse solutions over the last 10-15 years, however, these projects have achieved only a checkered success rate. Many were too
ambitious in scope, attempting to place virtually all the bank customers data into a warehouse. Although these projects may have succeeded
in their main objective of consolidating data, they did not always succeed in delivering reportable aggregated information to commercial
business leaders and risk managers. One of our customers calls its bank data warehouse the data landfill; still others think of their data
warehouses as graveyards for data that is doomed never to be seen again.
Fortunately, new reporting tools and more focused datamarts have evolved to fill the voids left by the large data warehouses. Datamarts
and data warehouses differ in a few key attributes. Data warehouses are quite expansive, containing data from dozens to sometimes hundreds
of systems, while datamarts are highly specific means of aggregating and validating data for a specific purpose.
Datamarts also typically incorporate processes to clean and validate data. Rules can be built to automatically kick out data that doesnt
meet specified criteria. For example, an exposure aggregation process may have to be put into common currency through an automated data
aggregation and validation process before it enters the datamart.
The other major difference between a datamart and data warehouse is the reporting capability. Because datamarts are purpose-built,
standardized reports and user-friendly reporting tools can be leveraged to deliver the meaningful information users demand from the system.
Banks should utilize a data warehouse as a source of data feeding a purpose-built datamart. This will enable more informative reporting
supporting business decisions.

RISK DATA MANAGEMENT | AUGUST 2015

103

MULTICOLLINEARITY AND STRESS TESTING

By Dr. Tony Hughes and Dr. Brian Poi

Dr. Tony Hughes

Managing Director of
Credit Analytics

Tony manages Moodys Analytics credit analysis

consulting projects for global lending institutions. An
expert applied econometrician, he has helped develop
approaches to stress testing and loss forecasting in
retail, C&I, and CRE portfolios.

Dr. Brian Poi

Director, Economic Research

Brian develops a variety of credit loss, credit

origination, and deposit account models for use in both
strategic planning and CCAR/DFAST environments.
He is equally adept at developing primary models
and validating models developed elsewhere. He also
provides thought leadership and guidance on the use
of advanced statistical and econometric methods in
economic forecasting applications.

Most readers will remember being somewhat perplexed back in

their undergraduate days by a topic called multicollinearity. This
phenomenon, in which the regressors of a model are correlated with
each other, apparently causes a lot of confusion among practitioners
and users of stress testing models for. This article seeks to dispel this
confusion and show how fear of multicollinearity is misplaced and,
in some cases, harmful to a models accuracy.
Is a fear of multicollinearity justified?

it often suffices to include only one in our final

Multicollinearity is common to all non-

model of the dependent variable, even if the

experimental statistical disciplines. If we are

unknown true model actually contains both.

conducting a fully controlled experiment, we can

We are seeking to explain variations in the

design our research to ensure the independence

dependent variable using signals gleaned

of all of the control variables. In bank stress

from variations in the independent variables of

testing, the Fed and the general public, not to

the regression.

mention shareholders, would likely not approve

of banks running randomized experiments to
discern bank losses under a range of controlled
conditions. Instead, banks must do their best
to piece together the effects of a range of
performance drivers with the limited actual data
they have.

adequately capture the signal. Including both will

lead to a competition between the variables,
and they will crowd each other out. Though the
estimates will be unbiased in the more liberally
(and, indeed, correctly) specified model, the
individual coefficient estimates will have high

but we dont belong in this camp. We feel that

standard errors, and thus the probability of

multicollinearity, rather than being a problem,

obtaining a coefficient that isnt statistically

is actually what keeps risk modelers gainfully

different from zero or else has the wrong sign

employed and enjoying life. Not only would bank

would be high. If data are plentiful, on the other

stress testing, and life generally, be banal if the

hand, we can more easily distinguish the subtle

phenomenon did not exist, but interrelations

differences between the signals provided by the

between variables would not be possible. Under

two variables and include both. Multicollinearity

these circumstances, there would be no need for

is, always and everywhere, a problem that occurs

expert statisticians even bankers could conduct

due to small sample size.

live in such a cruel dystopia.)

MOODYS ANALYTICS RISK PERSPECTIVES

identical, we dont need both regressors to

Many people take a dim view of multicollinearity,

stress testing! (Personally, we wouldnt want to

104

If these signals are, for all intents and purposes,

Note that we have talked only of the

contributions of individual variables. If the aim

Multicollinearity makes estimating individual

of the exercise is forecasting for which the loss

model coefficients imprecise. Say we have two

function is specified solely in terms of forecast

highly correlated regressors. For some purposes

errors multicollinearity can be rendered a

PRINCIPLES AND PRACTICES

second-order problem. If we have two highly

Validators and examiners should carefully

correlated variables (say, r = 0.99), and we

consider the aims of the model when

compare the model estimated using both with

determining whether fear of multicollinearity is

a model estimated using just one or the other

justified for model builders.

variable, we will find that baseline projections

from the models will usually be very similar.
Although the individual contributions are
estimated imprecisely, the joint contribution
is not. If the sole aim of the model user is
forecasting (of which stress testing is a recent
but important sub-discipline), the choice
between a one- and a two-variable model is
largely immaterial. Unnecessarily including the
second regressor leads to a small efficiency loss
(i.e., one degree of freedom), but in the grand
scheme of things this is hardly worthy

Model risk and multicollinearity

Now lets consider cases where worrying about
multicollinearity can increase the prevalence of
model risk. We use risk here in the traditional
statistical sense the expected value of
statistical loss across repeated samples. The
risk function we use here, assuming squared
error loss, is a variation of that discussed in
Hughes (2012):

BL [E (yt+i - yt+i )2 ] +ADV [E(yt+i - yt+i)2]

+SA [E(yt+i - y-t+i)2 ]

of consideration.
Multicollinearity is more of a problem if the
aim of the model is to conduct some form of
structural analysis. If we are testing an assertion
about the relationship between one of our
correlated factors and the dependent variable of
interest, too much multicollinearity will tend to
drain away the power of the statistical test used
for this purpose. Tightly specifying a model and
leaving out variables that should be there will
typically distort the tests size. The upside of this
trade-off is that practitioners have more power
in conducting their tests.

where BL + ADV + SA=1 are a series of weights

that indicate the relative importance of correctly
projecting credit losses (or PDs, LGDs, volumes,
etc.) in the various Feds Comprehensive
Capital Analysis and Review (CCAR) scenarios.
Expectations are conditional on the relevant
Fed scenario actually playing out, and the
forecasts (conditional on the relevant scenario)
produced are based on the information available
at the time.
We view as reasonable the assumption that
BL= ADV = SA = 0.33 , though, admittedly, the
_BL=_ADV=_

Rather than considering multicollinearity to be a phenomenon that always

increases model risk, validators should instead try to discern the optimal
level of multicollinearity in models.
Stress testers may well be interested in
conducting this type of structural analysis. For
example, a bank may be interested in finding
out the main driver of a portfolios behavior,
unemployment, or household income. This
function should, however, be considered
separately from the broader problem of
projecting future behavior under assumed
stress. There are, to our knowledge, no
regulatory dictats against stress testers using a
horses for courses approach to model selection
and keeping a stable of models designed for
different purposes (so long as these are well
documented and well understood).

majority of banks tend to give the adverse

scenario less weight than the severely adverse
scenario under most circumstances. Fed
examiners are well known to also give the
baseline scenario considerable weight in their
deliberations. (In reality, the risk function must
also accommodate idiosyncratic scenarios
designed specifically for each bank, but we are
leaving that out of our analysis for claritys sake.)
To further set the stage, assume that the true
data generating process (DGP) is a function
only of an unknown subset of the variables
published annually by the Fed. In reality, of

RISK DATA MANAGEMENT | AUGUST 2015

105

course, this process is likely to be infinitely more

complex than implied by this simple assumption.
Suppose, unbeknown to the modeler, that
the correct specification includes only the
unemployment rate, the rate of GDP growth,
and the interest rate on ten-year treasury bills.

simulations, later in this article.

The standard fix for multicollinearity is to drop
some of the correlated regressors, but doing
so is risky because it increases the probability
of making errors like that described in (4).
If we estimate a model and find that one

The following statements about this situation

variable, intuitively viewed as important, has

are all true:

an estimated coefficient with a p-value of 0.07,

1. A model that contains only the three variables

in the DGP will minimize overall model risk.
2. Any model selection procedure established

should it necessarily be dropped? In our view,

removing the variable is riskier than keeping
it. Does the universal application of a 5%
significance level really minimize overall model

within this framework will have a non-zero

risk when the ultimate goal of the model is to

probability of selecting an incorrect model.

provide stress projections?

3. If we select a model that includes not just the

three variables, but also additional extraneous
variables, our model will still produce
unbiased forecasts in all three scenarios,
but the forecasts will not be accurate, as
discussed above.
4. If the selected model excludes one or more

Rather than considering multicollinearity

to be a phenomenon that always increases
model risk, validators should instead try to
discern the optimal level of multicollinearity
in models. Models that are specified extremely
tightly are next to useless when seeking to
understand the effects of a range of idiosyncratic

of the three variables, projections in all three

stresses on the portfolio. Likewise, models of

scenarios will be biased and inconsistent.

the kitchen sink variety are unlikely to be very

This situation could yield efficiency gains in

useful since many of the drivers will be found

parameter estimation, but these are likely

to be insignificant. The best model will be a

to be modest, given that the efficiency of

liberally specified one, but where the liberty is

a biased parameter estimate is unlikely to

not abused.

be optimal.

Our small Monte Carlo study has demonstrated in the clearest way possible
that extreme forecast bias is most likely when historical relationships shift
and key variables are removed from regressions merely because they are
insignificant.
In weighing up the relative costs of the errors
made in (3) and (4), the risk of (4) is likely
to exceed the risk of (3). From a forecasting
perspective, this must also be considered
alongside Hughes (2012) observation that input
forecast errors arent possible when computing
stress tests that are conditional on a stated
macroeconomic scenario. The implication of
these observations is that when high levels of
multicollinearity are present, the practitioner

106

MOODYS ANALYTICS RISK PERSPECTIVES

Shifts in historical correlations

A more pressing issue has to do with scenarios
involving shifts in historical correlations between
variables. What we mean here are situations
in which, for example, two variables have
historically been positively correlated but where
the Fed, in its infinite wisdom, gives us a scenario
in which the two variables move in opposition to
each other.

should still tend to err, at the margin, in favor

It is crucial that we know how to deal with these

of the more liberally specified model. We

situations, as no one knows the nature of the

will explore this question, using Monte Carlo

next stress event. Stress test models should

PRINCIPLES AND PRACTICES

be able to cope, at least reasonably well, with

act to mitigate against the effect of stress, and

unusual happenstances. Models that can only

projected real credit losses should be lower

cope with a repeat of the Great Recession and

than expected because of increases in the actual

nothing else are next to useless.

unemployment rate.

We do not need to look far to find a situation in

Such a simple data generating process can

which historical correlations shifted in this way.

throw off unrealistic results like negative

In recent years, during the 2000s and 2010s, the

default rates but we want to keep this exercise

U.S. Phillips Curve has been modestly negatively

as straightforward as possible. We first fit the

sloped. Between January 2000 and November

model containing both variables and exclude any

2014, the correlation coefficient between the

that we find to be insignificant at the 5% level

unemployment rate and the year-over-year rate

using a standard t-test; we labeled the model

of consumer price inflation has been -0.51. In

selected using this procedure Chosen.

the Feds baseline scenario published in October

We then compared the forecasting and stress

2014, the correlation between the two variables

testing performance of the chosen model with

is -0.72 across the nine-quarter forecast window,

those based on a full model. Table 1 shows

and in the severely adverse scenario, the figure

results for this simple experiment, assuming

is -0.41. In these scenarios, the Fed is saying

5,000 replications.

Table 1 Forecasting and stress testing performance: comparing the chosen and full models

Full
Chosen

E(Bias)
-0.006
-0.176

CCAR Baseline
E(RMSE) E(MAPE)
0.142
1.133
0.196
1.459

E(Bias)
0.000
0.388

CCAR Adverse
E(RMSE) E(MAPE)
0.635
1.368
1.034
1.988

CCAR Severely Adverse

E(Bias) E(RMSE) E(MAPE)
-0.004
0.467
1.310
0.060
0.602
1.573

Source: Moodys Analytics

that Phillips Curve dynamics basically mimic

those of recent history. The adverse scenario is
completely different; in this case, the correlation
is +0.97 across the nine-quarter scenario
window. To put this into context, during the
1970s considered the stagflationary nadir by
most right-thinking economists the correlation
between the two variables was a mere +0.14.

We found that the correct full model is

chosen 59% of the time. Overall, the inflation
coefficient is statistically significant in around
67% of cases, whereas the unemployment rate
coefficient is significant 91% of the time. As
might be expected, always choosing the full
model yields forecasts that suffer no appreciable
bias in any of the three scenarios. Zero bias here
means that the conditional forecasts produced

Now suppose that the true DGP for the

by the model are, on average across the nine-

probability of default (PD) for a particular

quarter forecast window, neither too high

portfolio is a function only of inflation and the

nor too low when compared to the expected

unemployment rate. We set the parameters

outcomes of the target variable.

of the model to be -2 for inflation and 2 for

unemployment, and then simulate data for
PD assuming a simple linear functional form
and normal errors. Normally, in a model of the
default likelihood of fixed repayment loans,
we would expect the unemployment rate to
be positively signed in our regression and the
inflation rate to be negatively signed. Inflation,
after all, reduces the burden of nominal principal
and interest payments as nominal income
rises at a fast clip. Inflation should therefore

The situation changes quite noticeably when we

look at the performance of the chosen model.
Predictions from this model are too low under
baseline conditions and too high in both stressed
scenarios. In the severely adverse case, the bias is
only slight, but in the adverse case the levels of
overprediction are extreme. When we consider
root mean squared prediction error (RMSE),
whereby the improved efficiency of the smaller
models may compensate for the effect of bias,

RISK DATA MANAGEMENT | AUGUST 2015

107

we find that, in all cases, using the full model

We now address those points by extending our

yields substantially smaller forecast errors than

experiment to consider a true DGP that contains

the selected model.

three factors and has five potential regressors

Because the historical correlation between the

two variables is preserved in both the baseline
and the severely adverse scenarios, we have a
pretty good shot at getting decent projections
using an incorrectly specified model that

in our variable selection choice set. The true

model, as before, contains unemployment and
inflation, to which we add GDP growth with a
parameter of -2. The choice set contains these
three variables as well as the Baa spread and the
ten-year treasury interest rate.

excludes one of the variables.

In the adverse scenario, however, the situation
changes markedly. In Feds adverse scenario,
increases in the unemployment rate, which
would normally be accompanied by declines
in inflation, are now accompanied by rising
inflation. Removing the inflation variable from
the model means that the historical effects
of inflation are conflated with correlated
unemployment effects, and the coefficient on

As before, we select a model by excluding

any variable that is found to be statistically
insignificant at the 5% level and compare
this with the strategy whereby the full model
(containing all five variables) is used every time.
Again, we are interested in the observed bias
and RMSE of the calculated projections in the
three Fed scenarios. The results are contained
in Table 2.

the unemployment variable is far higher than

In this case, the full model is potentially at a

it should be as a result. We are powerless to

disadvantage because it always contains two

capture the mitigating effect of inflation, and

extraneous variables. This has no effect on

our projections suffer alarmingly as a result.

forecast bias, however, since the estimated

One could argue that the misspecified model

here is more conservative but we think that

model encompasses the true specification.

In all scenarios, the full model suffers
effectively zero bias.

Table 2 Forecasting and stress testing performance: fuller comparison of the chosen and full models
CCAR Baseline
Full
Chosen

CCAR Adverse

CCAR Severely Adverse

E(Bias)

E(RMSE)

E(MAPE)

E(Bias)

E(RMSE)

E(MAPE)

E(Bias)

E(RMSE)

E(MAPE)

0.028
-7.078

6.308
13.070

5.143
11.927

0.035
-8.616

6.767
16.006

5.538
14.858

0.018
-9.289

7.507
16.691

6.197
15.265

Source: Moodys Analytics

misses the point. The idea of modeling should

In this case, our simple model selection

be to derive an accurate, unbiased view of reality.

procedure yields the correct model (that which

Users of models can always apply conservative

contains the three factors) only 15% of the time.

assumptions to arrive at appropriately austere

More often, one or more of the true factors is

stress test results.

missing from the selected model. In 50% of

A fuller exposition of the problem

In the preceding discussion, two features might
have immediately jumped out at the reader. The
first is that the framework is so simple that it
bears no relation to the difficult task of CCARstyle stress testing. The second point is that the
experimental set-up explicitly favors the larger
model, as it is the only correctly specified model
in the choice set.

108

MOODYS ANALYTICS RISK PERSPECTIVES

the simulations, one factor is missing; in 29%,

two of the important factors are erroneously
excluded from the model. This demonstrates
a key result of model selection that, as the
choice set expands, the probability of correct
selection declines rapidly to zero. In only 3%
of the simulations does the model include too
many factors. The full model, containing all five
variables, is selected by this simple t-statisticbased procedure a mere 0.5% of times.

PRINCIPLES AND PRACTICES

That the model selection procedure is so

they have degrees of freedom available to model

easily tricked into excluding important

them. If our advice is taken to the extreme,

factors is a likely outcome in the presence of

efficiency losses will become large enough to

multicollinearity.

outweigh any gain from a reduction in the threat

In this experiment, we find that the selection

procedure yields models that produce
projections that are consistently too high.
Bear in mind that this is a function of our

of omitted variable bias. At the margin, however,

looking at a t-statistic of 1.7, or even 1.2, should
hold few fears for model validators, so long as
inclusion of the variable is logical and intuitive.

experimental design; we could have just as

If our aim was only to conduct baseline

easily designed an experiment with bias of

forecasting, multicollinearity would be, at

the opposite sign. Looking at RMSE, we find

best, a second-order concern. Here, though,

that the model selected on the basis of t-tests

we are interested in stress scenarios, in

yields twice the forecast error of the full model

which regulators and senior managers will

always modeling strategy. Improved estimation

regularly throw curveballs involving shifts in

efficiency does little to mitigate against the

historical relationships. In this case, a fear of

proximate threat of omitted variable bias caused

multicollinearity can be positively harmful. Our

by excluding key factors on the basis of an

small Monte Carlo study has demonstrated

insignificant t-statistic.

in the clearest way possible that extreme

Conclusion
In an important sense, the results of this
analysis will be unsurprising. That the issue
of multicollinearity has little currency when
the aim of the modeler is forecasting has been
well-known for many decades. What could
be an important issue for structural analysis
using regression type models is, at the margin,
irrelevant to forecasters.
This is not to say that practitioners should go
wild and throw as many drivers into models as

forecast bias is most likely when historical

relationships shift and key variables are
removed from regressions merely because
they are insignificant. To capture nuanced
scenarios like the adverse and severely adverse
CCAR events, or bank-specific idiosyncratic
happenstances, models need to be specified
quite liberally.
Ignoring this advice will not decrease model
risk. Rather, it will raise that risk to potentially
extreme levels.

RISK DATA MANAGEMENT | AUGUST 2015

109

MULTI-PERIOD STOCHASTIC SCENARIO

GENERATION
By Dr. Juan Licari and Dr. Gustavo Ordoez-Sanz

Dr. Juan Licari

Senior Director, Head of
Economic and Consumer Credit
Analytics for EMEA
Juan and his team are responsible for generating
alternative macroeconomic forecasts for Europe and
for building econometric tools to model credit risk
phenomena.

Dr. Gustavo Ordonez-Sanz

Director, Stress Testing
Specialist Team

Based in London, Gustavo advises financial

organizations on the different aspects of stress
testing scenario analysis at the firm-wide level. This
includes: scenario generation; risk, revenue, and capital
modeling and forecasting; embedding scenario analysis
in the firms decision-making processes (risk appetite,
portfolio management, pricing, etc.); and governance,
infrastructure, and reporting.

Robust models are currently being developed worldwide to meet

the demands of dynamic stress testing. This article describes how
to build consistent projections for standard credit risk metrics and
mark-to-market parameters simultaneously within a single, unified
environment: stochastic dynamic macro models. It gives a stepby-step breakdown of the development of a dynamic framework
for stochastic scenario generation that allows risk managers
and economists to build multi-period environments, integrating
conditional credit and market risk modeling.
Introduction

in hand, we can run these forecasts through

Dynamic stress testing and multi-period credit

stress testing satellite models. This step

portfolio analysis are priority areas for risk

provides us with forward-looking, multi-period,

managers and academics. New methodologies

scenario-specific simulations for all relevant

and techniques are being developed across

risk parameters. We illustrate this process with

the globe, mainly focusing on building robust

two leading examples: default risk for a lending

models that translate macro scenarios into

portfolio (US mortgages) and mark-to-market

conditional risk parameters (so-called satellite

risk for a traded credit portfolio (rate and credit

models). But a significant challenge emerges

spread risks).

when it comes to building stochastic multiperiod environments. Dynamic simulations

can quickly get out of control when a modeler
starts increasing the sources of uncertainty and
the out-of-sample periods.

It starts with the econometrics needed to build

dynamic stochastic macro simulations (Step
1). Next, it describes our methodology for
embedding the forecasted paths with probability

framework to handle multi-period stochastic

metrics (Step 2). Satellite models are then used

simulations. The proposed methodology

to compute path-specific forecasts for credit and

hinges on macro models as the starting point

market risk parameters (Step 3).

obtain the simulations from the econometric

model, we need to embed these paths with a
probability structure. To this end, we develop
a rank-ordering mechanism that considers
several dimensions of economic performance
to produce an overall score for each scenario.
With the scenarios and their probabilities

MOODYS ANALYTICS RISK PERSPECTIVES

steps required to build the proposed framework.

In this article, we develop an innovative

in the scenario generation process. Once we

110

The structure of the article is consistent with the

The combination of conditional risk parameter

realizations and their probabilities provides
the modeler with the necessary inputs to
address dynamic stress testing questions (such
as probabilities of losses for a given stressed
scenario) and to build a stochastic framework
for multi-period credit portfolio management.

PRINCIPLES AND PRACTICES

Step 1: Simulations using a Dynamic Stochastic

General Equilibrium (DSGE) model

period optimal transition for key economic

To overcome the challenge of building multi-

of the system, connecting endogenous macro

period scenarios, we propose the use of dynamic

and financial variables with the sources of

stochastic macroeconomic models. The main

uncertainty: the shocks to economic agents and

advantage of this family of models is the

macroeconomic policies. Within these equations

ability to simulate millions of time-series of

we also obtain the arbitrage-free conditions

economic shocks that are linked to each other

for any financial assets that are priced in the

through general equilibrium conditions. In other

model. In other words, the equilibrium system

words, the simulations are consistent within

requires market-consistent pricing for all assets

periods (alternative macro series must satisfy

over time.

variables. They provide the underlying dynamics

equilibrium conditions), and the connection

across subsequent periods comes from intertemporal optimal behavior and pricing. This is
the reason why these types of models are usually
referred to as macroeconomic models with
micro-foundations.

Step 1.B: Build the system of stochastic

differential equations that represent solutions
to the model
This step is achieved by log-linearization of
equilibrium conditions around steady-state
(the long-term solution for an economic series

At the core of their set-up are optimality and

that is constant over time). The original macro

arbitrage-free pricing conditions. The following

variables get replaced by distances to the long-

sub-steps can be used to produce millions of

term values, and the non-linear system gets

simulated stochastic paths.

replaced by its first-order Taylor approximation.1

This linear system is mapped into a state-space

Step 1.A: Find the equilibrium conditions that

solve the selected DSGE model

matrix form to facilitate its estimation.

In practice, this requires the modeler to solve

dynamic stochastic optimization problems.

Step 1.C: Estimate the linear system of

stochastic differential equations

Recursive methods are leveraged in order to

Several techniques are available to estimate (or

obtain the so-called Bellman Equations.

calibrate) the stochastic system of equations.

These non-linear formulas represent the intra-

There is vast and detailed literature on how

Figure 1 Example of a standard DSGE model set-up

Workhorse example -- Three types of agents: households, firms, government

Households

Firms

Government

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

111

Figure 2 Example of arbitrage-free pricing equations and other optimality conditions

Stochastic dynamic equations: equilibrum conditions

Fixed Income
Arbitrage-free
Real Assets
Arbitrage-free
Relative Prices

Market Clearing
(Supply=Demand)

Source: Moodys Analytics

to estimate DSGE models using Bayesian

that link macro series with each other and

techniques. Fernndez-Villaverde (2010)

key stats for distributional assumptions of the

provides an overview of existing techniques and

stochastic shocks). These objects provide a very

illustrates the practical advantages of Bayesian

useful and rich platform for robust, forward-

methods in the context of dynamic stochastic

looking, dynamic, and consistent simulations of

differential equations.2

all endogenous variables.

One of the main advantages of using these

methods is that after the estimation is

Step 1.D: Use the estimated system to produce

simulations for macro and financial series

completed the modeler has access to (posterior)

This critical step involves shocking the system

distributions for all relevant parameters (betas

to produce dynamic simulations out of sample.

Figure 3 Bayesian estimation: prior assumptions (left) and posterior distributions for key parameters (right)

Source: Moodys Analytics

112

MOODYS ANALYTICS RISK PERSPECTIVES

PRINCIPLES AND PRACTICES

There are two sources of uncertainty that need

labeled +Q1, +Q2, , +Q9. (Note that this

to be considered: (a) shocks to original random

is simply our choice for this paper and not a

variables in the model (e.g., policy surprises,

preference over an approach that considers

productivity gains/losses, shocks to consumer

longer time-horizons.)

preferences, etc.) and (b) the fact that estimated

parameters are random variables (coefficient
uncertainty). Statistical properties of the

Statistical properties of simulated macro

series are illustrated in Figure 4. We include
GDP growth, unemployment rate, and home

estimated parameters are derived from their

posterior distributions. Here is where Bayesian
methods have an advantage. Obtaining the
posterior distributions allows a modeler to draw
simulated values not only for residuals and
shocks, but also for betas and other parameters.

price dynamics as leading indicators. But DSGE

models produce between 20 and 30+ economic
and financial series, depending on the specifics
of the version of the model selected. Histogram
(frequency densities) and box-plots help a
modeler understand distribution properties of

We estimate a workhorse DSGE model for the

the simulated paths. Some charts contain blocks

US economy in line with Smets and Wouters

of simulations as x-axis categories. These blocks

(2007). Throughout this exercise, we focus on

represent groups of scenarios according to their

nine consecutive quarters out-of-sample for

severity (see section 2 for a detailed explanation

the forecasting period (consistent with stress

on how the scenarios get rank-ordered). Block 1

testing CCAR practices). These periods are

groups the most optimistic forecasts while Block

Figure 4 Statistical properties for key economic factors

Figure 4.1 GDP growth, % Q/Q

GDP Growth, % Q/Q: Box-Plot, all Quarters

.1

Density
.2

.3

.4

GDP Growth, % Q/Q: Density, all Quarters

-6

-4

-2
0
GDP Growth, % Q/Q

-6

-4

-2

GDP Growth, % Q/Q

GDP Growth, % Q/Q: Forecasts per Quarter

-6

-6

-4

-4

GDP Growth, % Q/Q

-2
0
2

GDP Growth, % Q/Q

-2
0
2

GDP Growth, % Q/Q: Over Scenario Blocks

+Q1

+Q2

+Q3

+Q4

+Q5

+Q6

+Q7

+Q8

+Q9

Block 1

Block 2

Block 3

Block 4

Block 5

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

113

Figure 4.2 Unemployment rate, %

Unemployment Rate, %: Density, all Quarters

.4
0

.2

Density

.6

.8

Unemployment Rate, %: Box-Plot, all Quarters

8
10
12
Unemployment Rate, %

14

16

10

12

14

16

Unemployment Rate, %

Unemployment Rate, %: Forecasts per Quarter

12
10
8
4

Unemployment Rate, %

Unemployment Rate, %
8
10
12
14

14

16

16

Unemployment Rate, %: Over Scenario Blocks

+Q1

+Q2

+Q3

+Q4

+Q5

+Q6

+Q7

+Q8

+Q9

Block 1

Block 2

Block 3

Block 4

Block 5

Source: Moodys Analytics

5 contains stressed scenarios.

The result is a full set of dynamic, stochastic,
and forward-looking paths for macro and
financial series. The equations used to derive
these simulations rest on general equilibrium
conditions, making the projections consistent
across variables and over time. The forwardlooking attribute rests on the fact that these
scenarios will vary over the business cycle.
The DSGE model gets re-estimated with new
data and the forecasts are conditional on the
starting point of the out-of-sample period.
For notation purposes, lets refer to a given
scenario as z, wherein the vector Z contains
the whole list of macro and financial series with
values at all quarters-out-of-sample
(+Q1 to +Q9).
Step 2: Embedding the stochastic scenarios
with a probability structure
A natural next step is to derive probabilities

114

MOODYS ANALYTICS RISK PERSPECTIVES

and severities for the simulated scenarios. To

achieve this goal, we develop a multi-factor
rank-ordering mechanism that attributes
severity according to 25+ dimensions of
economic severity. The core macro variables
that are part of the calculation are: GDP growth,
unemployment rate, home price changes,
consumption dynamics, investment profiles,
interest rate movements, and inflation. For each
of these series, we rank the scenarios according
to several criteria: average and/or cumulative
values over the scenarios, maximum/minimum
targets, and volatilities (sigma vs. average).
The algorithm produces a combined score that
translates into a ranking for each scenario.
The embedded statistical structure can be
obtained by (a) calculating the severity of any
scenario from the percentage of the simulations
that score lower or higher (purely a rank
ordering exercise) or (b) grouping scenarios
based on fixed intervals for score values. All

PRINCIPLES AND PRACTICES

Figure 4.3 Home price growth, % Q/Q

Home-Price Growth, % Q/Q: Density, all Quarters

.15
0

.05

.1

Density

.2

.25

Home-Price Growth, % Q/Q: Box-Plot, all Quarters

-10

-7.5

-5

-2.5

0
hpi

2.5

7.5

10

-10

-7.5

-2.5

2.5

7.5

10

hpi

Home-Price Growth, % Q/Q: Forecasts per Quarter

5
2.5
0
-2.5
-10

-10

-7.5

-5

-5

-2.5

2.5

Home-Price Growth, % Q/Q

7.5

7.5

10

10

Home-Price Growth, % Q/Q: Over Scenario Blocks

-7.5

Home-Price Growth, % Q/Q

-5

+Q1

+Q2

+Q3

+Q4

+Q5

+Q6

+Q7

+Q8

+Q9

Block 1

Block 2

Block 3

Block 4

Block 5

Source: Moodys Analytics

scenarios within a given cluster have the same

scenarios with credit and market risk parameters.

probability, given by the relative size of the

The modeler can now leverage recent

interval (number of scenarios in the cluster

developments on stress testing methodologies.

divided by the total number of simulations).

The financial industry has produced a vast

We illustrate the rank-ordering process with

the properties of three marginal loadings and
show the distribution properties of the final,
overall score.
Step 2 provides us with a vector p(z) that has

literature on robust models that are able to

calculate risk parameters conditional on any
given macro scenario. Instead of simply running
a handful of multi-period scenarios, we can run
thousands of them through stress testing
models and obtain conditional realizations for

a single probability value per scenario. Note

credit metrics.

that it is not time-dependent, as the scoring

Simulations of credit risk parameters: US

mortgage portfolio as a leading example

algorithm has considered information across

all relevant macro series observed at all points
in time. In other words, our stochastic shocks
are represented as dynamic paths for a group
of macro and financial series. Each path has an
associated probability value p(z).
Step 3: Connecting scenarios to risk parameters
using stress testing satellite models
The last step of the process consists of linking

Following the methodology described in Licari

and Surez-Lled (2013), we leverage a vintagePD model for US first mortgages to run through
the rank-ordered macro simulations.4 The result
is a set of dynamic paths for vintage PDs from
+Q1 to +Q9 (results are illustrated in Figures 7
and 8). We need to emphasize that these PDs are
not forecasted only for existing vintages, but also

RISK DATA MANAGEMENT | AUGUST 2015

115

Figure 5 Leading marginal scores: minimum GDP cumulative growth rates, maximum unemployment rate, maximum drop on home price growth
Density Function

.15
0

.05

.1

Density

.2

.25

Box-Plot

-5

0
5
10
Max Drop in Cumulative GDP Growth, %

15

20

-5

10

15

20

Max Drop in Cumulative GDP Growth, %

Density Function

.2

Density
.4

.6

.8

Box-Plot

10
12
14
Max Unemployment Rate, %

16

18

20

10

Density Function

14

16

18

20

Box-Plot

.1

Density
.2

.3

.4

12

Max Unemployment Rate, %

-5

-2.5

2.5

7.5

10

-5

-2.5

2.5

7.5

10

Max Drop in Home-Price Growth, %

Max Drop in Home-Price Growth, %

Source: Moodys Analytics

Figure 6 Statistical properties of the standardized overall score

Overall Score (Standardized): Density Function

4
0

Density

Overall Score (Standardized): Box-Plot

.2

.3

.4

.5
Overall Score

Source: Moodys Analytics

116

MOODYS ANALYTICS RISK PERSPECTIVES

.6

.7

.8

.2

.3

.4

.5
Overall Score

.6

.7

.8

PRINCIPLES AND PRACTICES

Figure 7 Historic vs. fitted PDs over age intervals PD lifecycle (term-structure)

US First Mortgages - Vintage PD Lifecycle

Distribution Across Vintages - Different Age Groups
Age between 9 and 16 Quarters
.02
.015
.01

Default Rates (#)

.005

.006
.004
.002
0

Default Rates (#)

.008

Age between 1 and 8 Quarters

PD Rate, #

Fitted PD Rate, #

PD Rate, #

Age between 25 and 30 Quarters

.005

Default Rates (#)

.01

Default Rates (#)

.005
.01 .015

.015

.02

Age between 17 and 24 Quarters

PD Rate, #

Fitted PD Rate, #

PD Rate, #

Fitted PD Rate, #

US First Mortgages - Vintage PD Lifecycle

Historic Distributions Across Vintages

Fitted Distributions Across Vintages

.005

.005

.01

Default Rates (#)

.01
.015

.015

.02

US First Mortgages - Vintage PD Lifecycle

Default Rates (#)

Fitted PD Rate, #

1Q

4Q

8Q

12Q

16Q

20Q

24Q

30Q

1Q

4Q

8Q

12Q

16Q

20Q

24Q

30Q

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

117

Figure 8 Simulated PDs across quarters-out-of-sample (+Q1 to +Q9)

Figure 8.1 Old/seasoned vintage
US First Mortgages - Vintage PD Simulations
Simulated values for +Q1 - Old/Seasoned Vintage

300

Density

100

200

.015
.01
.005

Default Rates (#)

.02

400

500

.025

US First Mortgages - Vintage PD Simulations

Simulated values from +Q1 to +Q9 - Old/Seasoned Vintage

+Q1

+Q2

+Q3

+Q4

+Q5

+Q6

+Q7

+Q8

.006

+Q9

.01

.012

US First Mortgages - Vintage PD Simulations

US First Mortgages - Vintage PD Simulations

Simulated values for +Q9 - Old/Seasoned Vintage

100

200

Density
200

Density

400

300

600

Simulated values for +Q5 - Old/Seasoned Vintage

400

.008
Simulated PDs

.005

.01
Simulated PDs

.015

.02

.005

.01
Simulated PDs

.015

.02

Source: Moodys Analytics

Figure 8.2 Future vintage (booked in the out-of-sample period dynamic projection)
US First Mortgages - Vintage PD Simulations

US First Mortgages - Vintage PD Simulations

Simulated values from +Q1 to +Q9 - Future Vintage

1.0e+06
0

.001

5.0e+05

Density

Default Rates (#)

.002
.003

.004

1.5e+06

Simulated values for +Q1 - Future Vintage

+Q1

+Q2

+Q3

+Q4

+Q5

+Q6

+Q7

+Q8

2.00e-06

+Q9

5.00e-06

6.00e-06

Simulated values for +Q9 - Future Vintage

5000

1000

Density

Density
1.0e+04

2000

3000

Simulated values for +Q5 - Future Vintage

1.5e+04

4.00e-06
Simulated PDs

US First Mortgages - Vintage PD Simulations

US First Mortgages - Vintage PD Simulations

.0002

.0004
Simulated PDs

Source: Moodys Analytics

118

3.00e-06

MOODYS ANALYTICS RISK PERSPECTIVES

.0006

.0008

.001

.002
Simulated PDs

.003

.004

PRINCIPLES AND PRACTICES

for loans in vintages that will be originated in

volumes for these new originations. This is

Simulations of market risk parameters:

interest rates and credit spreads as leading
example

of particular importance when performing

We now study the translation of the stochastic

dynamic stress testing and multi-period

scenarios into relevant mark-to-market metrics.

portfolio credit analysis. Figure 8.2 presents the

Government bond yields and corporate credit

simulated PD values for a vintage of mortgages

spreads are presented as leading examples,

that gets originated in the first out-of-sample

following econometric techniques developed

period (+Q1).

in Licari, Loiseau-Aslanidi, and Surez-Lled

the future, together with the level of forecasted

The outputs of this exercise consist of simulated

paths of conditional PDs (for each vintage in the
mortgage portfolio and across all out-of-sample

(2013).5 The modeling methods rest on a

combination of principal component analysis
and time-series estimation techniques.

periods). These conditional PDs combined

It is worth highlighting the dynamic behavior

with the probability vector p(z) become the

of simulated term-premiums (as a proxy for the

necessary inputs for running multi-period credit

yield curve slope), as illustrated in Figure 9.3.

portfolio analysis.

The simulations produce different shapes,

Figure 9 Simulated government bond yield curves

Figure 9.1 Distributions of yields at +Q9 for different maturity points
Government Bond Yields - Leading Maturities
Distributions of Simulated Yields at +Q9
1 Year

.05

.1

Density

.15
.1

.05

Density

.2

.15

3 Months

5
10
Government Bond Yields 3m

15

10 Years

Density

.05

.05

.1

.15

.2

.2
.15
.1

Density

20

.25

5 Years

10
15
Government Bond Yields 1y

10
15
Government Bond Yields 5y

20

10
15
Government Bond Yields 10y

20

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

119

Government Bond Yield Curves

10
0

Yields, %

15

20

Box-Plots Across Maturities at +Q9

3m

6m

1y

2y

3y

5y

7y

10y

20y

30y

Source: Moodys Analytics

Figure 9.2 Yields over blocks of scenarios at +Q9 and over quarters-out-of-sample (+Q1 to +Q9)
Government Bond Yields - Leading Maturities
Simulations over Scenario Blocks (1 for Good , 5 for Stressed) at +Q9
1 Year

10

Yields, %

Yields, %

10

15

20

15

3 Months

Block 1

Block 2

Block 3

Block 4

Block 5

Block 1

Block 2

Block 5

Block 4

Block 5

10 Years

Yields, %

10

15

20
15
10
0

Yields, %

Block 1

Block 2

Block 3

Source: Moodys Analytics

120

Block 4

20

5 Years

Block 3

MOODYS ANALYTICS RISK PERSPECTIVES

Block 4

Block 5

Block 1

Block 2

Block 3

PRINCIPLES AND PRACTICES

Simulated Government Bond Yields - Leading Maturities

Box-Plots, All +Qs
1 Year
20
15
0

10

Yields, %

10
5

Yields, %

15

3 Months

+Q1

+Q2

+Q3

+Q4

+Q5

+Q6

+Q7

+Q8

+Q9

+Q1

+Q2

+Q3

+Q6

+Q7

+Q8

+Q9

+Q6

+Q7

+Q8

+Q9

10 Years

Yields, %

10

15

20
15
10

Yields, %

+Q5

20

5 Years

+Q4

+Q1

+Q2

+Q3

+Q4

+Q5

+Q6

+Q7

+Q8

+Q9

+Q1

+Q2

+Q3

+Q4

+Q5

Source: Moodys Analytics

Figure 9.3 Term-premium spreads (yield curve slope) at +Q9

Yield Curve Slope (10y vs. 3m Spread, %)

Analysis Over Quarters and Simulations
Box-Plot for +Q9

.4
0

.2

Density

.6

Distribution for +Q9

2
4
6
8
Term Premium (10y vs. 3m Spread, %)

10

2
4
6
Yield Curve Slope (10y vs. 3m Spread, %)

10
8
6
0

Yield Curve Slope

8
6
4
2

Yield Curve Slope

10

Box-Plots, all +Qs

10

+Q9 Values over Simulations

5000
10000
15000
20000
Simulation id (1 for Good , 25000 for Stressed)

25000
+Q1

+Q2

+Q3

+Q4

+Q9

+Q6

+Q7

+Q8

+Q9

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

121

Figure 10 Simulated corporate credit spreads financials and non-financials over rating classes and maturities
Figure 10.1 Scatters of simulated spreads over scenarios at +Q5

Corporate Spreads - Bb Financials

5 Quarters out of Sample - Simulated Values Over Scenarios

15
10

Bb Financials 5y

0
5000

10000

15000

20000

25000

5000

10000

15000

20000

Simulation id (1 for = GOOD , 25000 = STRESSED)

Simulation id (1 for = GOOD , 25000 = STRESSED)

10 Years

30 Years

25000

10
5
0

10

15

Bb Financials 30y

15

20

Bb Financials 10y

15
10
5
0

Bb Financials 3m

20

5 Years

20

3 Months

5000

10000

15000

20000

25000

Simulation id (1 for = GOOD , 25000 = STRESSED)

5000

10000

15000

20000

25000

Simulation id (1 for = GOOD , 25000 = STRESSED)

Corporate Spreads - Bb Non-Financials

5 Quarters out of Sample - Simulated Values Over Scenarios

5000

10000

15000

20000

25000

5000

10000

15000

20000

10 Years

30 Years

Bb Non-Financials 30y

10
5

25000

10

Simulation id (1 for = GOOD , 25000 = STRESSED)

15

Simulation id (1 for = GOOD , 25000 = STRESSED)

Bb Non-Financials 10y

10
0

5000

10000

15000

20000

Simulation id (1 for = GOOD , 25000 = STRESSED)

Source: Moodys Analytics

122

Bb Non-Financials 5y

10
5
0

Bb Non-Financials 3m

15

5 Years

15

3 Months

MOODYS ANALYTICS RISK PERSPECTIVES

25000

5000

10000

15000

20000

Simulation id (1 for = GOOD , 25000 = STRESSED)

25000

PRINCIPLES AND PRACTICES

Figure 10.2 Box plots of simulated spread curves at +Q5

Corporate Spread Curves - Financials
5 Quarters out of Sample (+Q5) - Across Rating Classes
A

.5

10

Spread, %

1.5
1

Spread, %

15

Aaa

3m

1y

3y

5y

7y

10y

20y

30y

3m

1y

3y

5y

20y

30y

7y

10y

20y

30y

25
20
15
10

Spread, %

15
10

Spread, %

10y

20

Bbb

7y

3m

1y

3y

5y

7y

10y

20y

30y

3m

1y

3y

5y

Corporate Spread Curves - Non-Financials

5 Quarters out of Sample (+Q5) - Across Rating Classes

A

Spread, %
.5
1

Spread, %
2
4
6

1.5

Aaa

3m

1y

3y

5y

7y

10y

20y

30y

3m

1y

3y

5y

10y

20y

30y

7y

10y

20y

30y

Spread, %
4
6
8

Spread, %
5
10
15

10

20

Bbb

7y

3m

1y

3y

5y

7y

10y

20y

30y

3m

1y

3y

5y

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

123

Figure 10.3 Box plots of simulated spreads, 5-year maturity, +Q1 to +Q9
Corporate Spreads over Quarters-out-of-sample
Financials - 5 Year Maturity - Across Rating Classes
A

Spread, %
5
10

Spread, %
.5
1

15

1.5

Aaa

+Q1 +Q2 +Q3 +Q4 +Q5 +Q6 +Q7 +Q8 +Q9

+Q1 +Q2 +Q3 +Q4 +Q5 +Q6 +Q7 +Q8 +Q9

Spread, %
10
15

Spread, %
10 15 20

25

20

Bbb

+Q1 +Q2 +Q3 +Q4 +Q5 +Q6 +Q7 +Q8 +Q9

+Q1 +Q2 +Q3 +Q4 +Q5 +Q6 +Q7 +Q8 +Q9

Corporate Spreads over Quarters-out-of-sample

Non-Financials - 5 Year Maturity - Across Rating Classes
A

6
2

Spread, %

.6
.4

.2

Spread, %

.8

Aaa

+Q1

+Q2

+Q3

+Q4

+Q5

+Q6

+Q7

+Q8

+Q9

+Q1

+Q2

+Q3

+Q4

+Q7

+Q8

+Q9

+Q6

+Q7

+Q8

+Q9

10

10

Spread, %

15

8
6
4

Spread, %

+Q6

20

Bbb

+Q5

+Q1

+Q2

+Q3

+Q4

Source: Moodys Analytics

124

MOODYS ANALYTICS RISK PERSPECTIVES

+Q5

+Q6

+Q7

+Q8

+Q9

+Q1

+Q2

+Q3

+Q4

+Q5

PRINCIPLES AND PRACTICES

including inverted curves (negative premiums)

Concluding remarks

and severe scenarios with high values for the

In this article, we develop a dynamic framework

yield-curve slope.

for stochastic scenario generation. The

Corporate credit spreads are observed across

financial and non-financial sectors. Within each
sector, there is further segmentation across

proposed methodology sets the necessary

inputs for dynamic stress testing and multiperiod credit portfolio analysis.

rating classes (Aaa, Aa, A, Bbb, Bb, and B) and

Of particular relevance is the ability to build

maturities (3m, 1y, 3y, 5y, 7y, 10y, 20y, and

(simultaneously) consistent projections for

30y). Statistical properties for simulated spreads

standard credit risk metrics and mark-to-

are illustrated in Figure 10.

market parameters within a single, unified

The methods described in this section provide

a modeler with forward-looking, dynamic
simulations for risk-free rates and credit spreads.
These consistent projections together with their

environment. In other words, using stochastic

dynamic macro models as the scenario
foundation allows us to integrate conditional
credit and market risk modeling.

probabilities, p(z), represent the building blocks

for mark-to-market risk assessments for traded
credit portfolios.

1 Other techniques consider higher-order approximations in order to understand the effects of non-linear relationships. See
Schmitt-Groh and Uribe (2004) for quadratic approximation methods: Schmitt-Groh S. and Uribe M., Solving Dynamic General
Equilibrium Models Using a Second-Order Approximation to the Policy Function, Journal of Economic Dynamics & Control 28,
755775, 2004.
2 Fernndez-Villaverde, J., The Econometrics of DSGE Models, SERIEs,1:349, 2010.
3 Smets, F. and Wouters R., Shocks and Frictions in US Business Cycles, European Central Bank Working Paper Series N 722, 2007.
4 Licari, J. and Surez-Lled, J., Stress Testing of Retail Credit Portfolios, Risk Perspectives Magazine, 2013.
5 Licari, J., Loiseau-Aslanidi, O. and Surez-Lled, J., Modeling and Stressing the Interest Rates Swap Curve, Moodys Analytics
Working Paper, 2013.

RISK DATA MANAGEMENT | AUGUST 2015

125

STRESS TESTING SOLUTION PROCESS FLOW:

FIVE KEY AREAS
By Greg Clemens and Mark McKenna

Greg Clemens
Director,
Stress Testing Solutions

As a member of the Stress Testing Task Force at

Moodys Analytics, Greg helps clients automate their
stress testing processes providing insight about
architectures, data management, and software
solutions for risk management.

Mark McKenna
Senior Director,
Stress Testing Solutions

Mark is responsible for business and relationship

management for Moodys Analytics Stress Testing
Solutions. Prior to his current role, he was with the
Structured Analytics & Valuations Group, leading
valuations and cash flow analytics projects. He has
worked with many US financial institutions, providing
risk management tools and loss estimation models for
their retail portfolios.

Most banks are able to stand up to quantitative stress testing and

even prove their capital adequacy. But what many organizations lack
is a streamlined process that allows them to run stress tests with
ease, efficiency, and control. This article outlines a five-step process
that will help banks maximize their stress testing investment, making
compliance easier while improving their interior management.
Current stress testing challenges

consider the capital action plans submitted

The Federal Reserve (Fed) has required banks to

by each bank. Under CCAR, a bank submits its

run the Dodd-Frank Act Stress Test (DFAST) and

proposed capital plan for the next four quarters

the Comprehensive Capital Analysis and Review

(dividend hikes, share buybacks, etc.), and the

(CCAR) for a number of years. While they are

Fed assesses whether that bank would be able

the Feds primary supervisory mechanism for

to meet required capital ratios under shaky

assessing the capital adequacy of large banks,

economic conditions. Put simply, can a bank

these exercises remain extremely complicated.

afford to give dividends to shareholders if the

CCAR and DFAST continue to cause banks to

economy starts to falter? If the answer is yes,

spend extensive resources both time and

banks then announce their capital plans to

money to run the exercises, prepare the results,

the public.

and respond to the findings of regulators.

Over the years, there have been fewer and

DFAST tests whether banks have sufficient

fewer quantitative stress test failures. This

capital to absorb losses and support operations

may be because banks are in better condition,

during adverse economic conditions (e.g.,

because they have become familiar with the

housing crash, unemployment increase, severe

test, or perhaps a bit of both. Meanwhile

GDP drop, or stock market crash) while using a

the qualitative assessment has become

The true aim of the stress testing exercises, however, is not that banks
demonstrate that they can pass tests like the severely adverse scenario per
se, as no one actually expects such a scenario to come to pass, but that
banks demonstrate that they have the ability to weather a storm, whatever
it may be.

126

MOODYS ANALYTICS RISK PERSPECTIVES

standardized set of capital action assumptions.

an exceedingly important component of the

(The assumptions keep each banks current

regulatory program. In any case, the pressure

dividend and do not include share repurchase

is still on banks to demonstrate that they

plans.) CCAR tests banks under similar adverse

can manage their risks while running their

economic scenarios, but in this case regulators

businesses.

PRINCIPLES AND PRACTICES

The DFAST and CCAR exercises are a part of the

process itself and the way they run the bank.

Feds effort to ensure that banks have robust

CCAR applies only to the largest Bank Holding

processes for determining how much capital they

Companies (BHCs), but the challenge also

need to maintain access to funding and continue

applies to the next-tier banks that only need

to serve as credit intermediaries, even under

to run the DFAST exercise. While the CCAR

stressed conditions. The most onerous test the

and DFAST are US exercises, the issue is no less

banks must pass is called the severely adverse

relevant for the banks in Europe and around

scenario, which features a severe recession with

the globe.

rising unemployment and steep declines in the

stock market, housing prices, commercial real
estate, and GDP.

A solution to managing the stress testing

process flow
There are five major areas, or components, of the

The true aim of the stress testing exercises,

stress testing process.

however, is not that banks demonstrate that

they can pass tests like the severely adverse

1. Bringing all the data together

scenario per se, as no one actually expects

2. Preparing the preliminary balance sheet

such a scenario to come to pass, but that banks

demonstrate that they have the ability to

forecast
3. Conditioning the forecast with credit losses

weather a storm, whatever it may be.

4. Completing the remaining calculations,

The challenge for banks is to institute a process

capital ratios, and RWA forecasts to prepare

for running stress tests faster and with more

the required reports and capital plan

control, to meet the changing demands of

5. Overlaying a common framework for model

regulators while also improving both the

risk management

Figure 1 Stress testing process workflow

Source Systems

Budgeting & Planning Actuals

Starting Aggregated Balance Sheet

General Ledger

Credit

Dimensions & Hierarchies

Forecasted Balance Sheet

Granular Contract Level

Credit Loss
Models

Forecasted Balance Sheet

NIE/NIR | NII

Credit Loss Forecasts

Op Loss
Forecasts

ALM
Credit Adjusted
Cash Flows

Loss Adjusted Forecasted

BS | NIE/NIR | NII | PPNR

NIR / NIE Models

ALM
Initial Cash Flows

Loss Adjusted Forecasted

BS & NII

a g e m e n t Re

NIR / NIE Models

Planned Capital
Actions

/ M an

Origination &
Deposit Models

vie w

LOB Forecasts
Forecasted Balance Sheet
Adjusted Balances

LO
B

ury/ Management
Rev
reas
T
/
iew
B
LO

ALM System

Dimensions & Hierarchies

Granular Balance Sheet - Reconciled to GL

Op Loss
Modeling

Regulatory

Dimensions & Hierarchies

Dimensions & Hierarchies

Economic
Scenario Models

Operational Data

Forecasted RWA Capital Ratios

Regulatory
Reporting

Management
Reporting

Capital
Planning

Non-CCAR Downstream Models

Model Risk Management
Model Inventory

Document Repository

Board Review

Moodys Ecosystem

Internal

Moodys/Internal

Source System/Data

Models

System/Process

Source: Moodys Analytics

RISK DATA MANAGEMENT | AUGUST 2015

127

Figure 2 Stress testing process workflow: Central data definitions

Source: Moodys Analytics

1. Central data definitions

Organizations must access, validate, and

In the first step, all the necessary data from the

reconcile data across the enterprise. On top of

various areas and business units involved in the

process are brought together. The data model
should support forecasting throughout the
process at the most granular level, supported
by multiple hierarchies and dimensions across
all areas.
The data model for the process should be
thought of as the single source of truth for
stress testing. Organizations should take three
main points into account when creating this
model:
A central risk, finance, and treasury datamart
is needed to support a large range of models
and reporting requirements
They should leverage investment in
current systems, infrastructures, and data
warehouses
Data quality and reconciliation against
production systems are important
considerations
The ongoing, complex, and ever-changing
regulations are pushing IT budgets at most
financial institutions, requiring systems that
can handle an increasing amount of data at a
granular level.

128

MOODYS ANALYTICS RISK PERSPECTIVES

the data aggregation challenges, banks need to

improve the scope, accuracy, and governance
of their ballooning data.
Organizations must access, validate, and
reconcile data from across the enterprise,
including all geographies, portfolios, and
instruments, irrespective of the origin of the
data. A few points to bear in mind:
The data model should support a repeatable,
transparent, and auditable process
Data is not complete in each source system
Data is stored at different levels of granularity
in different systems
These challenges are straining firm resources
even further. Institutions are looking for ways to
improve data quality, streamline and standardize
data flows, improve the efficiency and accuracy
of regulatory reporting, support validation
requirements, improve auditing capabilities, and
supplement management reporting. They must
satisfy both the regulators and their boards
about the accuracy, scalability, and sustainability
of the data structure and the processes used for
data management.

PRINCIPLES AND PRACTICES

Figure 3 Stress testing process workflow: Preliminary balance sheet forecast

Source: Moodys Analytics

2. Preliminary balance sheet forecast

The second step involves preparing the
initial balance sheet forecast. With increased
regulatory expectations for scenario design,
banks need a process that is user friendly, as well
as auditable, transparent, and repeatable.
This requires:
Clear understanding of the key forecast
drivers and their relation to the current state
Granular balance sheet with all jump-off data
A common, central source of data that allows
different areas to view data in the way they
are accustomed (hierarchy and dimensions)
Stress testing forces institutions to
complement traditionally expert judgmentdriven planning processes with quantitative
approaches to produce forecasted cash flows.

The approach needs to incorporate:

A material risk identification process
An effective challenge process for
management and the board
Policies that lay out expectations for all
functions involved in the capital adequacy
process
Data infrastructure and system integration is a
fundamental problem at most banks. Banks have
been going from one short-term fix to another
using SharePoint and Excel as go-betweens for
multiple systems. Instead, they need a longerterm vision for how to build an infrastructure
that enables effective stress testing, featuring:
Integration of multiple systems
Auditability of the results
Coordination across finance, treasury, and risk
groups

RISK DATA MANAGEMENT | AUGUST 2015

129

Figure 4 Stress testing process workflow: Credit loss adjusted balance sheet forecast

Source: Moodys Analytics

3. Credit loss adjusted balance sheet forecast

data management, auditability, and regulatory

The third component of the process takes

reporting. Steps in this process include:

the preliminary balance sheet forecast and

adjusts it for credit losses and other forecast
considerations, including Pre-Provision Net
Revenue (PPNR), risk-weighted assets (RWA) for
market risk, and operational risk losses.
Methodologies to project loss estimation, PPNR,
and RWA are in various degrees of development.
Most are housed in a range of formats (SAS, R,
Matlab, and Excel, etc.), making documentation,
validation, and the challenge process more
difficult.
To facilitate loss-adjusted forecasting, the
process needs to incorporate customizable
workflow and reporting functions, including

130

MOODYS ANALYTICS RISK PERSPECTIVES

Implementing a workflow that connects

with banking systems to determine the role
and functionality of each component in the
process
Determining methodologies to project
loss estimation, PPNR, and RWA, including
documentation, validation, and the challenge
process
Projecting losses through the banks asset
and liability management (ALM) system for
forecasted cash flows
Leveraging the banks current models and
systems and managing the process through the
workflow streamlines the stress testing and
capital planning processes.

PRINCIPLES AND PRACTICES

Figure 5 Stress testing process workflow: Capital planning and reporting

Source: Moodys Analytics

4. Capital planning and reporting

The fourth component takes the adjusted
balance sheet and prepares the results to be used
for the various regulatory reports, management
reports, and capital plans.
Existing reporting solutions are not well suited
for the complexities of data aggregation, edit
checks, and management reviews needed
for both CCAR and DFAST regulatory and

Supporting management reporting needs,

including board and effective challenge
documents
Banks need the ability to perform what-if
and sensitivity analyses and make comparisons
across forecasts to facilitate capital planning.
They need to ensure the efficient alignment of
financial plans, models, and forecasts across
lines of business (LOBs), departments, and the

management reporting requirements.

board.

A comprehensive solution must be capable of:

The process needs to facilitate capital planning

Handling many issues around connection

of regulatory reports and including internal

points and hand-offs

and reporting by reconciling multiple sets

dashboards tailored to each group of end users.

Managing regulatory edit checks, changes,

and report linkages (e.g., between 14A
and 9C)

RISK DATA MANAGEMENT | AUGUST 2015

131

Figure 6 Stress testing process workflow: Model risk management

Source: Moodys Analytics

5. Model risk management

status of the validation or independent review of

Finally, the entire process should include an

each model or methodology (e.g., completed, in

overlay framework for model risk management.

A common framework for managing a
model throughout its life cycle is critically
important as banks strive to meet deadlines
from external agencies. The framework should
streamline the process of creating, managing,
deploying, and monitoring the banks analytical
models, while facilitating sensitivity analysis
around key assumptions and making it easier to
identify sources of uncertainty.
The Feds 2015 CCAR Summary Instructions
and Guidance document states: BHCs are
required to provide the Federal Reserve with an

progress). 1
This suggests that the process needs to include
a model risk management framework covering
stress loss, revenue, and expense estimation
models, which all in turn should be tailored to
the task. Banks need a model risk management
framework tailored to the task of stress
testing at their own specific bank, not just a
standardized database/document repository.
The model management framework should be
repeatable and make it easy to register, validate,
deploy, monitor, and retrain analytic models. As
such, it should include the following capabilities:

Stress testing forces institutions to complement traditionally expert judgmentdriven planning processes with quantitative approaches to produce forecasted
cash flows.
inventory of all models and methodologies used
to estimate losses, revenues, expenses, balances,
and RWAs in CCAR 2015. The inventory should
start with the FR Y-14A line items and provide
the list of models or methodologies used for
each item under each scenario and note the

132

MOODYS ANALYTICS RISK PERSPECTIVES

Perform common model management tasks

such as importing, viewing, and attaching
supporting documentation
Facilitate the creation of a model and
document repository (including model
ownership, validation issue tracking,

PRINCIPLES AND PRACTICES

upstream/downstream models, and status of

process that can be governed with more control.

each model in the inventory)

The stress testing process flow outlined in this

Serve as a document repository with

article supports the intersections between risk,

all relevant model documentation and

finance, treasury, and regulatory compliance,

comments

while leveraging existing investments in current

Track and flag issues arising in non-CCAR/

DFAST models that impact the submission
Indicates model risk management ownership,
roles, and responsibilities (e.g., validate,
approve, etc.) as prescribed by regulatory
guidance
A model management framework enables
banks to meet the objectives set forth by the
FRB, OCC, and FDIC in the same system as
the process automation, thereby reducing

systems and models used for stress testing.

An effective flow:
Provides the governance of a process that is
repeatable, transparent, and auditable
Is a part of managing the bank
Allows increased frequency of stress testing
Leverages investment in current data
warehouse, systems infrastructure, and
existing models

the burden of multiple systems and allowing

If banks implement a similar process, their stress

for consistent and tractable expert judgment

testing program will become more business-as-

overlay capture.

usual, freeing up valuable resources and making

Building a better stress testing process

the entire program more accurate and efficient.

Banks need a better and faster stress testing

The Federal Reserve, Comprehensive Capital Analysis and Review 2015: Summary Instructions and Guidance, 2015.

RISK DATA MANAGEMENT | AUGUST 2015

133

SUBJECT MATTER EXPERTS

Greg Clemens

Cayetano Gea-Carrasco

Greg is a member of the Stress Testing Task Force

Cayetano is the Head of Stress Testing Services

at Moodys Analytics. He helps clients automate

and Advisory at Moodys Analytics. He has

their stress testing processes, providing insight

extensive experience working with financial

about architectures, data management, and

institutions on credit portfolio management

software solutions for risk management. Greg

across different asset classes, analysis of

has in-depth knowledge of the risk management

structured credit portfolios, derivatives pricing,

business, coupled with a strong technology

counterparty credit risk analytics, stress testing

development background, and has extensive

(CCAR, CLAR), liquidity management, and

experience developing innovative and strategic

enterprise risk management.

Director, Stress Testing Solutions

technology solutions.

Cayetano holds a BSc and MSc in

Prior to joining Moodys Analytics, Greg had

Telecommunication Engineering, a Master in

senior roles in Oracle Financial Services and

Economics and Finance, and a MSc in Financial

Citigroup, where he ran a large risk technology

Mathematics, with distinction, from Kings

development group.

College London.

[Link]@[Link]

[Link]-carrasco@[Link]

[Link]/GregClemens

[Link]/CayetanoGea-Carrasco

Pierre Gaudin

Kevin Hadlock

Pierre is in charge of strategic initiatives in

Kevin is a Global eLearning Solution Specialist

Asia-Pacific for risk appetite management, stress

and supports all training-related activities

testing, and origination at Moodys Analytics,

throughout the Asia-Pacific region. He has

assisting with subject matter expertise, client

designed numerous distance-learning and

requirement analysis, and use case illustrations.

web-based training programs, developed and

He joined Moodys Analytics via Fermat in

taught many seminars, webinars, and full,

2004, and he has lead several Basel RWA

blended training programs, and has authored

implementations in Europe and Asia.

a substantial body of content used globally by

He also developed Fermat Education services,

thousands of credit trainees.

Senior Director, Enterprise Risk Solutions

designing and delivering regulatory compliance

and ALM training to clients and system
integrators worldwide.

Senior Director

Prior to his current role, Kevin headed up

elearning and blended training development for
the organization globally, and was the architect

Pierre has 15 years of experience delivering

for the companys wholesale credit elearning

software implementation services for capital

solutions. Kevin had 13 years of experience in

markets and risk management solutions at major

the banking industry before joining Moodys

commercial banks in Paris, Bruxelles, London,

Analytics in January 1990. Kevin holds a BA

Hong Kong, and Singapore. Pierre has a Masters

degree from the University of Utah in Business

Degree in Systems and Networks Engineering

Management.

from Suplec, France.

[Link]@[Link]
[Link]/PierreGaudin

134

Head of Stress Testing Services and Advisory

MOODYS ANALYTICS RISK PERSPECTIVES

[Link]@[Link]
[Link]/KevinHadlock

Dr. David T. Hamilton

Managing Director, Head of Stress Testing and

Credit Risk Analytics, APAC
Based in Singapore, David has two decades
of experience in credit risk modeling, credit
ratings, and economic research. He helps provide
insight into the practical challenges of credit risk
measurement, management, and stress testing.
He has lectured on credit risk topics at Columbia
Business School, the Stern School of business
at New York University, and the City College of
New York (among others). He is on the editorial
board of the Journal of Credit Risk and holds
a BA in economics and classical studies from
Texas A&M University and a PhD in financial
economics from the City University of New York.
[Link]@[Link]
[Link]/DavidHamilton

Brian Heale

Senior Director, Business Development Officer

(Global Insurance)
Brian is an insurance market and Solvency
II specialist who has significant experience
in the technology solutions and issues for
the global insurance industry. He has an
in-depth knowledge of the practical aspects
of the insurance business, coupled with a
comprehensive understanding of enterprise
technology in relation to the development

Dr. Tony Hughes

Managing Director of Credit Analytics

Tony oversees the Moodys Analytics credit
analysis consulting projects for global lending
institutions. An expert applied econometrician,
he has helped develop approaches to stress
testing and loss forecasting in retail, C&I, and CRE
portfolios and recently introduced a methodology
for stress testing a banks deposit book.
Tony was formerly the lead Asia-Pacific
economist for Moodys Analytics. Prior to that,
he held academic positions at the University of
Adelaide, the University of New South Wales,
and Vanderbilt University. He received his PhD
in Econometrics from Monash University in
Melbourne, Australia.
[Link]@[Link]
[Link]/TonyHughes

Nicolas Kunghehian

Director, Business Development

Nicolas is responsible for providing thought
leadership on ALM, liquidity, and market risks
for the EMEA region to help financial institutions
define a sound risk management framework.
Nicolas worked as an ALM and risk manager in
two French banks for more than six years before
joining Fermat in 2005, which was acquired by
Moodys Analytics in late 2008.

and implementation of core administration,

Nicolas holds a degree in Mathematics and

actuarial/risk, data, and Solvency II reporting

Economics from the Ecole Polytechnique and a

systems.

degree in Finance and Statistics from the Ecole

Brian has previously worked with a number of

major insurers and technology and consulting

Nationale de la Statistique et de lAdministration

Economique.

companies across the world. He has run

[Link]@[Link]

administration, product development, and sales

[Link]/NicolasKunghehian

divisions, and also has considerable experience in

strategic planning.
[Link]@[Link]
[Link]/BrianHeale

RISK DATA MANAGEMENT | AUGUST 2015

135

Eric Leman

Dr. Samuel W. Malone

Eric has served as a Solutions Specialist of

Sam primarily develops systemic risk solutions

banking compliance and risk management

that are actionable by individual financial

since 2009, helping organizations make

institutions and also actively contributes to

more informed risk management decisions

consulting projects, such as model validations

specifically Basel III capital adequacy (credit risk,

and bespoke research, taken on by the

market risk), asset and liability management,

Specialized Modeling Group (SMG).

Director, Solutions Specialist

stress testing, and credit risk monitoring.

Director of Economic Research

Sam has taught and consulted at top institutions

He joined Fermat in 2006 (acquired by

in Europe and South America, including Oxford,

Moody's Analytics in 2008), working as an

the University of Navarra, and the Central Banks

Implementation Consultant for three years in

of Venezuela and Peru. He is a coauthor of the

Europe, Middle East, and the US. Before joining

book Macrofinancial Risk Analysis, published

Fermat, Eric worked for over four years on

in the Wiley Finance series with a foreword by

implementation services at CSC and internally at

Nobel Laureate Robert Merton, as well as the

Danone. Eric holds an engineering degree from

author of multiple academic journal articles in

Ecole des Mines de Saint Etienne, a prestigious

economics and applied math published in outlets

French Grande Ecole.

such as the Journal of Applied Econometrics,

[Link]@[Link]
[Link]/EricLeman

the International Journal of Forecasting, and

the Annual Review of Financial Economics. He
holds undergraduate degrees in mathematics
and economics from Duke University, where he

Dr. Juan M. Licari

Senior Director, Head of Economic and Consumer

Credit Analytics

studied as an A.B. Duke scholar and graduated

with summa cum laude Latin honors, and MPhil
and doctoral degrees in economics from the

Juan and his team are responsible for generating

University of Oxford, where he studied as a

alternative macroeconomic forecasts for Europe

Rhodes Scholar.

and for building econometric tools to model

credit risk phenomena. His team develops

[Link]@[Link]

and implements risk solutions that explicitly

[Link]/SamMalone

connect credit data to the underlying economic

cycle, allowing portfolio managers to plan for

Senior Director, Stress Testing Solutions

solutions are leveraged into stress testing and

Mark is responsible for business and relationship

reverse stress testing practices.

management for Moodys Analytics stress testing

Juan communicates the teams research and

methodologies to the market and often speaks
at credit events and economic conferences
worldwide. He holds a PhD and an MA in
Economics from the University of Pennsylvania
and graduated summa cum laude from the
National University of Cordoba in Argentina.
[Link]@[Link]
[Link]/JuanLicari

136

Mark McKenna

alternative macroeconomic scenarios. These

MOODYS ANALYTICS RISK PERSPECTIVES

solutions. Prior to his current role, he was with

the Structured Analytics & Valuations Group as
a solutions specialist leading valuations and cash
flow analytics projects with structured finance
participants. Additionally, he worked with
numerous US financial institutions providing risk
management tools and loss estimation models
for their retail portfolios.

Mark has been with Moodys for more than 23

portfolio and counterparty credit risk modeling

years, always in client-facing roles. His past

and management. Gustavo holds a degree in

responsibilities have included relationship

Theoretical Physics from the University Autonoma

management for the largest and most

of Madrid and a PhD in Physics from Radboud

strategically important money managers in the

University Nijmegen.

US. Early in his career, he managed a team of

relationship managers and product specialists for

[Link]-sanz@[Link]

the research, data, and analytics group. He holds

a BA degree in economics and political science
from Fairfield University.
[Link]@[Link]

Yuji Mizuno

Director, Business Development Officer

Dr. Brian Poi

Director, Economic Research

Brian develops a variety of credit loss, credit
origination, and deposit account models for use
in both strategic planning and CCAR/DFAST
environments. He is equally adept at developing
primary models and validating models developed

Yuji leads the product and consulting areas

elsewhere. He also provides thought leadership

of Moodys Analytics Japan and has extensive

and guidance on the use of advanced statistical

knowledge of regulations and risk management

and econometric methods in economic

practices among financial institutions. He

forecasting applications.

provides clients with insight on regulatory

compliance, ALM, liquidity, and ERM
frameworks. He also functions as a main contact

Before joining Moodys Analytics, Brian was an

econometric developer and director of professional

for Japanese regulators and financial institutions.

services at StataCorp LP, a leading provider of

Before joining Moodys Analytics in 2009, he

and MA in Economics from the University of

worked for ex-Sanwa Bank, Bank of America

Michigan after graduating magna cum laude from

Securities, Aozora Bank, and JP Morgan

Indiana University.

securities. He holds a Bachelor of Law from

statistical analysis software. He received his PhD

Tokyo University.

[Link]@[Link]

[Link]@[Link]

Buck Rumely

[Link]/YujiMizuno

Dr. Gustavo Ordonez-Sanz

Director, Stress Testing Specialist Team

Senior Director, Client Management

Buck leads the Americas team of credit and
technology specialists at Moodys Analytics. He
has helped design credit and risk management

Based in London, Gustavo advises financial

systems for a variety of financial, governmental,

organizations on the different aspects of stress

and energy firms throughout North and South

testing scenario analysis at the firm-wide level.

America.

This includes: scenario generation; risk, revenue,

and capital modeling and forecasting; embedding
scenario analysis in the firms decision-making
processes (risk appetite, portfolio management,
pricing, etc.); and governance, infrastructure, and

Prior to joining Moodys Analytics, Buck was a

partner in a leading technology consulting firm. He
began his career in software systems development
with a national consulting firm, and is a graduate

reporting.

of Indiana Universitys Kelly School of Business.

Gustavo also has experience in credit

Magazine, the Journal of the International Energy

He has published papers on credit risk in Risk

RISK DATA MANAGEMENT | AUGUST 2015

137

Credit Association, and the Journal of National

Petroleum and Energy Credit Association.
[Link]@[Link]

Peter Sallerson

Senior Director, Structured Analytics and Valuation

Peter focuses on the CLO market at Moodys

Dr. Christian Thun

Senior Director, Strategic Business Development

Christian is responsible for providing thought
leadership on credit risk management and
strategic business development in the EMEA
region and functions as a main contact for
regulators and senior management of financial
institutions.

Analytics via the further development of our

Structured Finance Portals CLO section and

With almost 20 years of experience, Christian

related research. During his long career, he has

has worked with numerous financial institutions

worked in the Corporate Finance department

in the EMEA region on Basel II implementation,

at Nomura Securities, was a Managing Director

risk management, stress testing, and portfolio

and one of the founders of the CLO effort at Bear

advisory projects, and in the process has

Stearns, and ran CLO origination at Mitsubishi

become an internationally-known expert on

UFJ Securities (USA).

credit risk management.

Peter graduated Phi Beta Kappa and magna cum

[Link]@[Link]

laude with a BA in Economics from Middlebury

[Link]/ChristianThun

College and his MBA is with honors from the

University of Chicago Graduate School of
Business.
[Link]@[Link]
[Link]/PeterSallerson

Vivek Thadani

Director, Structured Finance Valuations &

Consulting Group
Vivek is primarily responsible for developing and
maintaining analytical models for various asset
classes across the structured security space.
Prior to his current role, Vivek supported
investor and asset manager clients at Wall Street
Analytics across CLO & RMBS asset classes.
Vivek holds a BE in Computer Science from
Mumbai University and a MEM from
Dartmouth College.
[Link]@[Link]

Michael van Steen

Senior Director, Enterprise Risk Solutions

Michael helps deliver advanced portfolio credit
risk, stress testing, correlation, and valuation
solutions to global financial institutions and
regulatory organizations. He is the practice
lead for origination services in the Americas,
developing and managing services around
stress testing, lending workflows, pricing, and
limit setting. Previously, Michael worked in the
portfolio credit risk area of Moodys Analytics,
delivering over 40 portfolio analysis projects
that covered economic capital, stress testing,
portfolio optimization, correlation estimation,
retail pooling, and portfolio valuation.
Michael has BS and MS degrees in Engineering
from the University of California, Berkeley,
and credit-related coursework at the Stanford
Graduate School of Business and at the Kellogg
School of Management.
[Link]@[Link]
[Link]/MichaelvanSteen

138

MOODYS ANALYTICS RISK PERSPECTIVES

MOODYS ANALYTICS RISK MANAGEMENT

SOLUTIONS
LEVERAGE POWERFUL SOLUTIONS FOR ENTERPRISE-WIDE RISK MANAGEMENT
Moodys Analytics offers deep domain expertise, advisory and implementation services, in-house economists, best-in-breed modeling
capabilities, extensive data sets, and regulatory and enterprise risk management software. Our risk management solutions:
Improve strategic business planning and facilitate meeting regulatory requirements
Assist with defining both macroeconomic and business-specific scenarios
Offer a comprehensive and granular credit risk, economic, and financial data set
Help model the impact that macroeconomic cycles, regulatory directives, and/or outlier events may have on an institutions risk profile
Deliver an integrated stress testing software solution to calculate stressed performance indicators across the risk and finance functions
For more information, contact our integrated risk management experts at RiskPerspectives@[Link].
INFRASTRUCTURE

Regulatory Reporting for Solvency II

Scenario Analyzer

Integrated with our end-to-end insurance regulatory capital solution, the

Coordinates the stress testing process across the enterprise, centralizing a

easy-to-use and cost effective Regulatory Reporting module produces

wide range of Moodys Analytics, third-party, and proprietary models.

accurate management and regulatory reports in local supervisors

RiskAuthority

commonly used formats and languages.

Delivers comprehensive regulatory capital calculation and management

B&H Capital Modeling Framework

for Basel I, II, and III, including the risk-weighted asset (RWA) calculations

Comprised of three products: B&H Economic Capital Calculator, B&H

required for CCAR reporting.

Proxy Generator, and B&H Risk Scenario Generator. Together, they

RiskAnalyst
Provides comprehensive and consistent view of your firms counter-party
risk by combining financial spreading, credit analysis and robust data
storage using one flexible, secure enterprise platform.
RiskOrigins
Risk-focused, workflow-driven software platform that allows commercial
lenders to streamline and standardize the commercial credit underwriting
process, incorporating Moodys Analytics industry-leading risk

provide a rapid and highly flexible approach to calculating economic

capital for complex liabilities in the life insurance industry, and establish
the processes required by Solvency II and ORSA for assessing insurance
capital solvency.
B&H Proxy Generator
Creates proxy functions that can be used to meet a range of business
needs, such as interim valuation, capital calculations, and hedge
effectiveness.

assessment capabilities which help lenders demonstrate compliance

B&H Property & Casualty ESG

to regulators.

Provides automated Economic Scenario Generation solutions for P&C

Regulatory Reporting Module

Create, validate, and deliver monthly, quarterly and annual CCAR (FR
Y-14) and DFAST reporting requirements. Fully integrated with our

insurers with modeling capability and calibration content. It is specifically

developed to meet the needs of P&C insurers, as part of an internal
capital model or for asset portfolio risk management.

enterprise risk platform, this module creates and delivers reports in the

B&H Defined Benefit ALM

required formats.

Is a sophisticated ALM solution underpinned by the B&H Economic

RiskIntegrity
Provides a comprehensive and modular solution to manage all aspects
of Solvency II compliance, ranging from centralized data management,
internal model development, solvency capital requirement calculations,

Scenario Generator and calibration content. Provides a comprehensive

risk modeling framework that allows advisors, asset managers, and inhouse pensions teams to measure and manage the risks facing defined
benefit pension funds.

risk type aggregation, and integrated regulatory and business reporting.

RISK DATA MANAGEMENT | AUGUST 2015

139

B&H PensionsLite

time series from the best national and private sources, as well as key

Differs from B&H Defined Benefit ALM in that it has a simpler, more

multinational data sets.

accessible user experience that enables interactive meetings with trustees

or corporate sponsors, and a slightly less sophisticated modeling engine.

Is the worlds largest and cleanest database of private firm financial

RiskFrontier

statements and defaults, built in partnership with over 45 leading

Produces a comprehensive measure of risk, expressed as Credit VaR

financial institutions around the world.

or Economic Capital, which comprises the basis for deep insight into
portfolio dynamics for active risk and performance management.
SCENARIOS
Global and Regional Macroeconomic Scenarios
Delivered by a team of over 80 experienced economists, who offer

Exposure at Default (EAD)

Data is derived from a subset of the CRD Database and is compiled of
10+ years of usage data for estimating and calculating EAD. The EAD
database contains quarterly usage and Loan Equivalency Ratio data for
both defaulted and non-defaulted private firms since 2000.

standardized alternative economic scenarios, supervisory scenarios, and

PD Time Series Information

bespoke scenarios customized to your specific needs for 49 countries, as

Offers time series of observed default rates and calculated PDs, covering

well as US states and metro areas.

more than two economic cycles. This data is collected and calculated for

B&H Economic Scenario Generator

both public and private firms.

Provides Monte Carlo simulation paths for the joint behavior of financial

Credit Migration Data

market risk factors and economic variables. The automated, market-

Enables users to construct detailed credit migration (transition) matrices.

leading Economic Scenario Generation for life insurance offers modeling

This detailed private firm data allows users to be more granular with

capabilities and calibration content for many levels of user sophistication.

segmentations across industry, region, and asset size using several

B&H Market-Consistent Economic Scenario Generator

different PD rating calculation methodologies.

Addresses the challenges of market-consistent liability valuation

Credit Cycle Adjustment Data

required in Solvency II. It includes the award-winning B&H Scenario

Combines financial statement ratio information of private firms with

Generator (ESG).

credit cycle factors in the public equity markets to derive a dynamic,

B&H Scenario Service

through-the-cycle PD measure.

Is an alternative to ESG software and provides insurers with scenario sets

Structured Finance (SF) Data

on an annual, semi-annual, or quarterly basis. Insurers and reinsurers can

Offers loan, pool and bond level performance data for RMBS, CMBS,

decide between the control and flexibility of a software installation or the

ABS and CDOs. SF Data can be used for bottom-up mortgage stress

simplicity and ease-of-use offered by the scenario service.

testing model creation and calibration. SSFA data and calculations are

DATA

also available.

RiskFoundation

Default and Recovery Database

Integrates your enterprise financial and risk data to calculate regulatory

Allows users to look at how default experience varies at different points

capital, economic capital, ALM, liquidity, counterparty risk, and for a

in the economic cycle, and which factors made default experience in

global view of your exposures.

each economic cycle unique. The data includes detailed rating histories,

Moodys Content Licensing Services

Provides a suite of comprehensive data covering all current Moodys
Investors Service issuer and issue-related ratings.

30-day post default pricing, and three views into ultimate recovery.
INVESTMENT ANALYSIS/SURVEILLANCE
Moody's CreditView

Global and Regional Macroeconomic Scenarios

Research and data to assist risk practitioners with investment analysis,

Delivered by a team of over 80 experienced economists, who offer

creation of internal risk scores and meeting due diligence requirements.

standardized alternative economic scenarios, supervisory scenarios, and

bespoke scenarios customized to your specific needs for 49 countries, as
well as US states and metro areas.

MODELS
CreditCycle
Provides retail credit portfolio insights into the expected and stressed

Global Economic, Financial, and Demographic Data

performance of existing and future vintages, enabling loss forecasting

Provides a comprehensive view of global economic conditions and trends.

and stress testing.

Our database covers more than 180 countries with more than 260 million

140

Moodys Analytics Credit Research Database (CRD)

MOODYS ANALYTICS RISK PERSPECTIVES

CreditEdge Plus

SERVICES

Bridges the equity, bond, and credit derivative markets, enabling an

Enterprise Risk Solutions Services

in-depth understanding of their impact on credit risk.

Provide stress testing, model validation, and implementation services.

Stressed EDFs

Valuation and Advisory Services

Estimate PDs for public firms using a range of macroeconomic scenarios,

Provide stress testing, model validation, and implementation services

including EBA and user-defined scenarios.

for all structured finance assets.

Commercial Mortgage Metrics (CMM)

B&H Calibration Service

Is the leading analytical model for assessing default and recovery risk

Is a market-leading calibration service that provides prompt quarterly

for commercial real estate (CRE) loans. CMMs stress testing capabilities

updates for Moody's Analytics models, reflecting the latest market

leverage Moodys Analytics Economic and Consumer Credit Analytics,

conditions and economic outlook. Also provides a range of calibration

Federal Reserves CCAR, and custom scenarios.

types for specific applications.

GCorr

Outsourcing Services

Moodys Analytics Global Correlation Model (GCorr) is an

Our Copal Partners unit is one of the worlds leading providers of

industry-leading granular correlation model used to calculate each

outsourced high quality research and analytics services to institutional

exposures contribution to portfolio risk and return for improved

customers. Copals workflow processes are designed to ensure the

portfolio performance.

highest possible level of data integrity and auditability while delivering

GCorr Macro

rigorously verified research and analysis.

Macro Stress testing with GCorr Macro produces instrument-level stress

TRAINING & CERTIFICATION

expected losses across multiple asset classes to help manage credit risk.

Moody's Analytics designs, develops and facilitates training programs

LossCalc
Calculates the Loss Given Default (LGD) for loans, bonds, sovereigns,
municipals and preferred stock using a range of Asset Classes and a
Comprehensive Database of Defaulted Instruments.
Portfolio Analyzer (PA)
Is a loan level capital allocation and risk management tool providing
stressed PDs, LGDs, and prepayments for RMBS, auto ABS, mortgage

and solutions for financial services institutions and individuals interested

in banking, finance, and personal and organizational development.
Company Learning Solutions
Instructor-led Training
Our subject-matter expertise and course selection can be tailored to the
needs of multiple levels of an organization from new hires that need
basic training to experienced professionals seeking an update on current

and auto loans under the Feds CCAR scenarios and custom scenarios.

thinking or insights into new markets.

RiskCalc Plus

eLearning and Blended Solutions

Enables clients to calculate forward-looking PDs for private firms across

different regions and industries and measure how borrowers would be
affected by stressed scenarios versus a baseline scenario.

Offer a suite of training solutions for financial institutions that enable

companies to minimize people risk, with customized learning solutions
using practical learning methodologies and the Moodys Analytics
expertise in credit and finance.

RiskFrontier
Produces a comprehensive measure of risk, expressed as Credit VaR

Individual Learning Solutions

or Economic Capital, which comprises the basis for deep insight into

Public Seminars

portfolio dynamics for active risk and performance management.

Moodys Analytics offers over 200 open-enrollment courses throughout

WSA Platform
Is a risk and portfolio management tool used for stress testing structured

the year, in major financial centers around the world, on fundamental

through advanced topics in finance and risk.

finance transactions. Moodys Analytics maintains a global structured

eLearning Programs

finance deal library. WSA integrates macroeconomic, credit models,

Provide the industrys most comprehensive eLearning curricula for

pool, and loan level performance data to forecast cash flows, PDs, LGDs,

corporates, commercial and investment banks, as well as asset managers

and prepayments.

and regulators. Our programs are accessible online and on-demand from
any web-enabled computer.

RISK DATA MANAGEMENT | AUGUST 2015

141

CONNECT WITH US
Get More Insight from
Moodys Analytics
There are many ways to view our content online. Register just once and youll
gain free access to all of our valuable insight, including Risk Perspectives articles,
webinars, white papers, and more.
View previous editions of Risk Perspectives at
[Link]/RiskPerspectives
Sign up for our newsletters at
[Link]/Register
Learn about our events at
[Link]/Events
Stay in touch with our Subject Matter Experts at
[Link]/SME
Follow our social channels for the latest insight at
[Link]/Connect

About Risk Perspectives

Each edition of Risk Perspectives magazine explores an industry or regulatory topic
in depth, presenting a wide range of views, best practices, techniques, and
approaches, all with one larger goal in mind to deliver essential insight to the
global financial markets.

ABOUT US
Moodys Analytics offers award-winning solutions and best practices

We help organizations answer critical risk-related questions, combining

for measuring and managing risk through expertise and experience in

best-in-class software, analytics, data and services, and models

credit analysis, economic research, and financial risk management. By

empowering banks, insurers, asset managers, corporate entities,

providing leading-edge software, advisory services, data, and research,

and governments to make informed decisions for allocating capital

we deliver comprehensive investment, risk management, and workforce

and maximizing opportunities. Through training, education, and

solutions. As the exclusive distributor of all Moodys Investors service

certifications, we help organizations maximize the capabilities of their

content, we offer investment research, analytics, and tools to help debt

professional staff so they can make a positive, measurable impact on

capital markets and risk management professionals worldwide respond

their business. More information is available at [Link].

to an evolving marketplace with confidence.

142

MOODYS ANALYTICS RISK PERSPECTIVES

GLOSSARY OF TERMS
ALM

Asset and Liability Management

FSB

Financial Stability Board

BCBS

Basel Committee on Banking Supervision

FTP

Funds Transfer Pricing

BHC

Bank Holding Companies

GAAP

Generally Accepted Accounting Principles

BIS

Bank for International Settlements

GFC

Global Financial crisis

BoE

Bank of England

GFMA

Global Financial Markets Association

CCAR

Comprehensive Capital Analysis and Review

G-SIB

Globally Systematically Important Banks

CCP

Central Counterparty

HQLA

High-Quality Liquid Asset

CCR

Central Credit Register

IAP

Insurance Analytics Platform

CDS

Credit Default Swap

IASB

International Accounting Standards Board

CECL

Current Expected Credit Loss

IFRS

International Financial Reporting Standards

CLO

Commercial Loan Origination

IRB

Internal Ratings-Based

CRE

Commercial Real Estate

LCR

Liquidity Coverage Ratio

CRM

Customer Relationship Management

LEI

Legal Entity Identifier

CRO

Chief Risk Officer

LGD

Loss Given Default

DFAST

Dodd-Frank Act Stress Test

LTV

Loan-to-Value

DGC

Degree of Granger Causality

MRA

Matters Requiring Attention

DGI

Data Gaps Initiative

NSFR

Net Stable Funding Ratio

DSGE

Dynamic Stochastic General Equilibrium

ORSA

Own Risk Solvency Assessment

D-SIB

Domestically Systematically Important Banks

P&L

Profit and Loss

EAD

Exposure at Default

PD

Probability of Default

EBA

European Banking Authority

PPNR

Pre-Provision Net Revenue

ECB

European Central Bank

PRA

Prudential Regulation Authority (UK)

EDF

Expected Default Frequency

REO

Real Estate Owned

EL

Expected Loss

RMBS

Residential Mortgage-Backed Security

ELR

Expected Loss Ratio

RMSE

Root Mean Squared Prediction Error

ERM

Enterprise Risk Management

RWA

Risk-Weighted Asset

FASB

Financial Accounting Standards Board

SIFI

Systemically Important Financial Institutions

FDSF

Firm Data Submission Framework

SLABS

Student loan Asset Backed Security

FFELP

Federal Family Education Loan Program

SSFA

Simplified Supervisory Formula Approach

FRB

Federal Reserve Board

UCA

Uniform Credit Analysis

FSAP

Financial Sector Assessment Program

VaR

Value-at-Risk

RISK DATA MANAGEMENT | AUGUST 2015

143

 2015 Moodys Corporation, Moodys Investors Service, Inc., Moodys Analytics, Inc. and/or their licensors and affiliates (collectively, MOODYS). All rights reserved.
CREDIT RATINGS ISSUED BY MOODYS INVESTORS SERVICE, INC. (MIS) AND ITS AFFILIATES ARE MOODYS CURRENT OPINIONS OF THE RELATIVE FUTURE CREDIT RISK
OF ENTITIES, CREDIT COMMITMENTS, OR DEBT OR DEBT-LIKE SECURITIES, AND CREDIT RATINGS AND RESEARCH PUBLICATIONS PUBLISHED BY MOODYS (MOODYS
PUBLICATIONS) MAY INCLUDE MOODYS CURRENT OPINIONS OF THE RELATIVE FUTURE CREDIT RISK OF ENTITIES, CREDIT COMMITMENTS, OR DEBT OR DEBT-LIKE SECURITIES.
MOODYS DEFINES CREDIT RISK AS THE RISK THAT AN ENTITY MAY NOT MEET ITS CONTRACTUAL, FINANCIAL OBLIGATIONS AS THEY COME DUE AND ANY ESTIMATED
FINANCIAL LOSS IN THE EVENT OF DEFAULT. CREDIT RATINGS DO NOT ADDRESS ANY OTHER RISK, INCLUDING BUT NOT LIMITED TO: LIQUIDITY RISK, MARKET VALUE
RISK, OR PRICE VOLATILITY. CREDIT RATINGS AND MOODYS OPINIONS INCLUDED IN MOODYS PUBLICATIONS ARE NOT STATEMENTS OF CURRENT OR HISTORICAL FACT.
MOODYS PUBLICATIONS MAY ALSO INCLUDE QUANTITATIVE MODEL-BASED ESTIMATES OF CREDIT RISK AND RELATED OPINIONS OR COMMENTARY PUBLISHED BY MOODYS
ANALYTICS, INC. CREDIT RATINGS AND MOODYS PUBLICATIONS DO NOT CONSTITUTE OR PROVIDE INVESTMENT OR FINANCIAL ADVICE, AND CREDIT RATINGS AND MOODYS
PUBLICATIONS ARE NOT AND DO NOT PROVIDE RECOMMENDATIONS TO PURCHASE, SELL, OR HOLD PARTICULAR SECURITIES. NEITHER CREDIT RATINGS NOR MOODYS
PUBLICATIONS COMMENT ON THE SUITABILITY OF AN INVESTMENT FOR ANY PARTICULAR INVESTOR. MOODYS ISSUES ITS CREDIT RATINGS AND PUBLISHES MOODYS
PUBLICATIONS WITH THE EXPECTATION AND UNDERSTANDING THAT EACH INVESTOR WILL, WITH DUE CARE, MAKE ITS OWN STUDY AND EVALUATION OF EACH SECURITY THAT
IS UNDER CONSIDERATION FOR PURCHASE, HOLDING, OR SALE.
MOODYS CREDIT RATINGS AND MOODYS PUBLICATIONS ARE NOT INTENDED FOR USE BY RETAIL INVESTORS AND IT WOULD BE RECKLESS FOR RETAIL INVESTORS TO
CONSIDER MOODYS CREDIT RATINGS OR MOODYS PUBLICATIONS IN MAKING ANY INVESTMENT DECISION. IF IN DOUBT YOU SHOULD CONTACT YOUR FINANCIAL OR OTHER
PROFESSIONAL ADVISER.
ALL INFORMATION CONTAINED HEREIN IS PROTECTED BY LAW, INCLUDING BUT NOT LIMITED TO, COPYRIGHT LAW, AND NONE OF SUCH INFORMATION MAY BE COPIED OR
OTHERWISE REPRODUCED, REPACKAGED, FURTHER TRANSMITTED, TRANSFERRED, DISSEMINATED, REDISTRIBUTED OR RESOLD, OR STORED FOR SUBSEQUENT USE FOR ANY
SUCH PURPOSE, IN WHOLE OR IN PART, IN ANY FORM OR MANNER OR BY ANY MEANS WHATSOEVER, BY ANY PERSON WITHOUT MOODYS PRIOR WRITTEN CONSENT.
All information contained herein is obtained by MOODYS from sources believed by it to be accurate and reliable. Because of the possibility of human or mechanical error as well as
other factors, however, all information contained herein is provided AS IS without warranty of any kind. MOODYS adopts all necessary measures so that the information it uses in
assigning a credit rating is of sufficient quality and from sources MOODYS considers to be reliable including, when appropriate, independent third-party sources. However, MOODYS is
not an auditor and cannot in every instance independently verify or validate information received in the rating process or in preparing the Moodys Publications.
To the extent permitted by law, MOODYS and its directors, officers, employees, agents, representatives, licensors and suppliers disclaim liability to any person or entity for any
indirect, special, consequential, or incidental losses or damages whatsoever arising from or in connection with the information contained herein or the use of or inability to use any such
information, even if MOODYS or any of its directors, officers, employees, agents, representatives, licensors or suppliers is advised in advance of the possibility of such losses or damages,
including but not limited to: (a) any loss of present or prospective profits or (b) any loss or damage arising where the relevant financial instrument is not the subject of a particular credit
rating assigned by MOODYS.
To the extent permitted by law, MOODYS and its directors, officers, employees, agents, representatives, licensors and suppliers disclaim liability for any direct or compensatory losses
or damages caused to any person or entity, including but not limited to by any negligence (but excluding fraud, willful misconduct or any other type of liability that, for the avoidance
of doubt, by law cannot be excluded) on the part of, or any contingency within or beyond the control of, MOODYS or any of its directors, officers, employees, agents, representatives,
licensors or suppliers, arising from or in connection with the information contained herein or the use of or inability to use any such information.
NO WARRANTY, EXPRESS OR IMPLIED, AS TO THE ACCURACY, TIMELINESS, COMPLETENESS, MERCHANTABILITY OR FITNESS FOR ANY PARTICULAR PURPOSE OF ANY SUCH RATING
OR OTHER OPINION OR INFORMATION IS GIVEN OR MADE BY MOODYS IN ANY FORM OR MANNER WHATSOEVER.
MIS, a wholly-owned credit rating agency subsidiary of Moodys Corporation (MCO), hereby discloses that most issuers of debt securities (including corporate and municipal bonds,
debentures, notes and commercial paper) and preferred stock rated by MIS have, prior to assignment of any rating, agreed to pay to MIS for appraisal and rating services rendered
by it fees ranging from $1,500 to approximately $2,500,000. MCO and MIS also maintain policies and procedures to address the independence of MISs ratings and rating processes.
Information regarding certain affiliations that may exist between directors of MCO and rated entities, and between entities who hold ratings from MIS and have also publicly reported
to the SEC an ownership interest in MCO of more than 5%, is posted annually at [Link] under the heading Shareholder Relations Corporate Governance Director and
Shareholder Affiliation Policy.
For Australia only: Any publication into Australia of this document is pursuant to the Australian Financial Services License of MOODYS affiliate, Moodys Investors Service Pty Limited
ABN 61 003 399 657AFSL 336969 and/or Moodys Analytics Australia Pty Ltd ABN 94 105 136 972 AFSL 383569 (as applicable). This document is intended to be provided only to
wholesale clients within the meaning of section 761G of the Corporations Act 2001. By continuing to access this document from within Australia, you represent to MOODYS that you
are, or are accessing the document as a representative of, a wholesale client and that neither you nor the entity you represent will directly or indirectly disseminate this document or
its contents to retail clients within the meaning of section 761G of the Corporations Act 2001. MOODYS credit rating is an opinion as to the creditworthiness of a debt obligation of
the issuer, not on the equity securities of the issuer or any form of security that is available to retail clients. It would be dangerous for retail clients to make any investment decision
based on MOODYS credit rating. If in doubt you should contact your financial or other professional adviser.

144

MOODYS ANALYTICS RISK PERSPECTIVES

Reporting

ECB
Basel III
Data Gaps PD
Forecasting P&L

BCBS 239 Credit Risk

Finding Alpha

DFAST

CCAR

Human Data
Aggregate data

Capital Planning

Data Infrastructure

Stress Testing
Data Governance

IFRS 9
FR
Y-14Q
LGD

Credit Risk
Basel III

DFAST

Expected Default Frequency

Stochastic Scenario Generation

CCAR Granular Data PD

IFRS 4

Monte Carlo Study

The
Fed
Data Infrastructure Risk Appetite BCBS 239
PD Analytical Data

The Fed

Regulatory Big Data

Regulatory Capital

Expected Loss

Portfolio Indicators

Data Quality

FR Y-14Q

Forecasting P&L

Human Data

IAS 39

IAS 39

Risk Appetite

IFRS 4

Systemic Risk

Data Integration

FR Y-14M RWA

AnaCredit LGD Reporting

Commercial Credit Decisioning

LGD

Forecasting P&L

AnaCredit Solvency II Datamart

Basel Risk Systems LGD AnaCredit
FTP Framework Forecasting P&L
Data Management

RWA

IFRS
9
LCR FDSF

PPNR Modeling

Expected Default Frequency

Macroeconomic Scenarios

Model Risk Management

RISK DATA MANAGEMENT

VOL 5 | AUGUST 2015

RISK DATA MANAGEMENT VOL 5

ONLINE

FEATURED ARTICLES:

Gain a competitive advantage through good data management

NOVEMBER
AUGUST 2014
2015

Find additional risk data management articles, interviews,

and multimedia content at [Link]/RiskPerspectives

STRONG DATA MANAGEMENT AN ABSOLUTE NECESSITY

REGULATORY BIG DATA

32

Learn about trends, overall regulatory goals, and the impact of major initiatives

STRESS TESTING PROCESS FLOW: FIVE KEY AREAS

Build a streamlined process to run stress tests with ease, efficiency, and control

126