Modelling

Economic capital versus regulatory

capital a market benchmark
Guido Giese sets out to provide a market benchmark for the relationship between economic and
regulatory capital models currently used or developed across different financial institutions, and
to analyse the market forces driving the development of economic capital models

he impressive development of risk

management models in the past
decade was mainly driven by two
things. The regulators, aiming at increasing the security of financial markets to
avoid crises caused by insolvent banks;
and increasing market competition, with
shareholders/investors looking for the financial institution (FI) with the best riskreturn profile. The former reason led to
the implementation of the Basel I Accord
for credit risk (1988) and market risk
(1996), requiring banks to hold minimum
amounts of capital against their credit and
market risk. This will be enhanced by the
Basel II Accord in the near future, which
will incorporate operational risk into the
capital framework. The second reason initiated the development of internal riskbased (IRB) models, affecting not only an
FIs profitability but also pricing, securitisation and business strategy.

Economic capital models overview

To meet the regulators two main requirements, most FIs developed two risk
management systems one for the external risk supervision and regulatory reporting and one internal risk model as a
basis for further management decisions.
Internal models typically developed into
economic capital models, where the economic capital of an FI defines the need
for capital as a buffer against all financial
risks on a common solvency standard.
Economic capital typically functions as a
common currency for risk, overcoming
the inconsistencies in the regulatory capital adequacy framework (see below).
Since the development of economic
capital models started in the mid-1990s or
even later, market benchmarks are quite
rare, and were pursued only very recently, for example in Sarraf (2003)
analysing 61 European financial organisations; Oliver, Wyman (2001) focusing
on global banking/insurance conglomerates; CMRA (2001) investigating major
banks across the world; and ERisk (2002)
analysing economic capital-based riskadjusted return on capital (Raroc) models
in 17 major banks, to investigate bestpractice standards and the sophistication

of different models, which was found to

vary considerably.
It turned out in all surveys that more
than two thirds of major FIs in Europe and
the USA use internal economic capital
models. The key reasons named for their
implementation were (in the order of decreasing importance) regulators, rating
agencies, shareholders and equity analysts, that is, all external drivers. It is quite
surprising to see regulators as a key factor, proving that their influence and supervision already exceeds the control of
capital adequacy according to the Basel I
framework. Under the Basel II framework,
to be implemented till 2006 (Basel Committee on Banking Supervision, 2002), we
can expect this tendency to increase even
more, since under Pillar 2 the regulatory
supervision will be further enhanced, allowing regulators to require banks to estimate and hold capital requirements
beyond Pillar 1 measures. Regulators intention to increase the sophistication of
capital adequacy models is also reflected
in various studies and research papers of
the Basel Committee (Joint Forum on Financial Conglomerates, 1999).
The main features of economic capital models can by summarised as:1

Quantification of all financial risks of
an FI endangering its equity, including
credit risk, market risk, operational risk
(with two sub-categories business risk
and event risk2) and (if applicable) insurance risk. The basic idea is to quantify the
risk of an adverse change of the FIs equity E in a certain time-period T, which
is:
AE = ( assets) (liabilities )
4244
3
1424
3 14
Market / ALM /
credit risk

Insurancerisk

+ Net income (in T )

14442444
3

(1)

Business risk

Market liquidity risk of assets is included in the market risk category, whereas
liquidity risk in the sense of the banks
available liquid funds is not included in
economic capital models, since it is not
directly endangering the equity of the
bank.

Economic valuation of assets and liabilities (mark-to-market/mark-to-model

rather than book values).

Consistent measurement of different
risk-types by using a value-at-risk type
approach with the same holding period
and confidence level. The economic capital figure of an FI is the (negative) value
of E from equation (1) on a certain confidence level, where the probability distribution of E is calculated taking into
account correlations between assets, liabilities and net income streams, allowing diversification benefits when
aggregating different risk categories.

The confidence level of the economic
capital model is typically linked to the target debt rating of the FI on a one-year
time horizon, for example, a 99.9% capital level corresponds to a single-A rating.

Since the development of economic

capital models started in the mid1990s or even later, market
benchmarks are quite rare
Typical confidence levels range between
99% and 99.98% (Sarraf, 2003). The oneyear time horizon is the industry standard,
since most budgeting and allocation
processes also relate to one year.

Raroc-based performance measurement and capital allocation processes.
The Raroc of a portfolio is defined as the
excess return over the risk-free rate over
the corresponding economic capital, allowing not only for a risk-adjusted performance comparison but also enabling
the use of the results for capital allocation
decisions or pricing.
According to Capital Market Risk Advisors (2001), most FIs include market
risk (including ALM) and credit risk (76%
and 79% of all FIs respectively) in their
economic capital model, whereas only
55% of the banks include operational risk,
showing that not only the regulatory risk
framework delayed the inclusion of [Link] MAY 2003 RISK

BASEL S17

Modelling
erational risk in capital adequacy modelling. In Oliver, Wyman (2001) and Capital Market Risk Advisors (2001), the
typical risk profile of major organisations was outlined (see figure 1).
Figure 1 shows the average (stylised)
economic capital breakdown in major universal banks (where credit risk dominates), non-life insurance (where
insurance risk prevails due to the unpredictability of casualties) and life insurances
(where insurance risk is small due to the
high predictability of the mortality rate, but
market risk is high, since the stock quota
in the asset portfolio is usually very high
due to the long-term investment horizon
of life insurances). See Oliver, Wyman
(2001) and Capital Market Risk Advisors
(2001).
Most banks (more than 80%) allocate
economic capital (EC) at enterprise and
business level, but only half go down to
the desk level. Further, most banks that
use Raroc methods do so on a business
level (about 40%, where another 37% announced to do so in the close future),
whereas very view banks do so on a desk
level (about 13%), showing that Raroc
models on a business level develop rapidly to a best practice approach (Sarraf,
2003, and Capital Market Risk Advisors,
2001).

Economic capital models

implementation
Economic capital models can compensate
several weaknesses of the Basel I and II
framework, that is:

The incompleteness of financial risks
taken into account, for example, interest
rate risk in the banking book is not included in the Basel capital requirements.
Moreover, the regulatory capital model (including Basel II) views banks as silos for
assets and liabilities and neglects income
streams that can be highly correlated with
assets or liabilities, that is, business risk is
neglected.

The inconsistent risk measurement of different risk categories. A consistent method-

ology requires a full VAR approach with the

same confidence level and holding period
for all risk areas. Even under the Basel II
IRB approach for credit risk, the capital requirement calculation is not a full VAR approach, since the model is based on the
assumption of a portfolio of infinite granularity (Wilde, 2001, Wilson, 1997), neglecting portfolio diversification and risk
concentration effects. The inconsistency in
the regulatory capital framework has various consequences, for example:
Inefficient capital allocation process in
the economy due to an unfair pricing of
loans;
Regulatory arbitrage, since the same
economic risk exposure can have different regulatory capital requirements within a financial group, depending on the
entity in which it is booked.

Uneconomic aggregation of risk. Aggregation of risks has to take place under
consideration of correlation and can be
performed on the following three levels,
that is, aggregation (Oliver, Wyman, 2001)
i. within a certain risk factor, for example, market risk in the trading portfolio
(Level I)
ii. across different risk factors within a
business line for example, combining
market and credit risk within a banking
division (Level II)
iii. across different business lines, for example, aggregating banking and insurance
risk within a financial group (Level III).
Level I aggregation takes place for the
regulatory Basel model approach for market risk (BIS (1996)) and partially for credit risk under the Basel II IRB approach for
credit risk (limited by the aforementioned
granularity assumption). Aggregation of
level II and level III including correlation
and diversification effects will not be provided under Basel II.
Most economic capital models currently perform a sufficient aggregation on Level
I, for example, due to the use of VAR for
market risk (currently two out of three
banks use VAR models for market risk and
the remaining third intends to do so in the

1. Average (stylised) economic capital breakdown

Universal bank

Non-life insurance

Life insurance

future (Capital Market Risk Advisors,

2001)) or portfolio/VAR-based credit risk
tools such as KMV Portfolio Manager,
CRisk+ or CreditMetrics, which take into
account industry and geographic diversification/concentration and credit correlations.3 These approaches are already used
by more than 50% of financial institutions
within their economic capital model, and
this number is growing. For operational
risk models using VAR-type approaches
that are not yet well developed, the prevailing methodology is a heuristic scenarioanalysis, showing a clear accuracy gap
between market/credit risk and operational risk on the Level I calculation.
Industry benchmarking showed (Oliver, Wyman, 2001) that Level II and especially Level I aggregation can reveal
significant diversification benefits of (together) up to more than 50%, whereas the
benefits from Level III diversification
amounts to at most 10%, showing that
only neglecting the latter effect within a
regulatory or internal model is justifiable.
Concerning a correlated Level II and
Level III aggregation, there are broadly
two approaches (see figure 2):

A factor model (also called econometric or statistical model) that defines a
set of deep economic factors (GDP
growth, interest rates, indexes, etc) that
influence all risk types using an econometric function to model the dependency of the loss distribution function for
each risk type (market risk, credit risk,
etc) which can then be aggregated to a
group-wide loss distribution function to
calculate the total VAR (economic capital) of the bank.

Direct estimation of the correlation between risk types based on internal experience. Put simply, instead of
calculating the economic capital as a certain percentile of the banks global loss
distribution, one calculates the correlation of the N VAR figures per risk type,
requiring estimating a correlating matrix
of size N N only.
Figure 2 is an example of an economic
capital model with common risk factors for
all risk categories. The aggregation can either be based on a direct approach, calculating a total VAR as a correlation of the
single VAR figures per risk category, using
a heuristic correlation matrix ij, or on a statistical approach, aggregating the key risk
factors to a bank-wide total loss distribu1

Market
Credit
Op risk
Insurance

S18

BASEL RISK MAY 2003 [Link]

Not all FIs reached the same level of

sophistication and hence this list can be viewed
as a target most models seem to converge to
2 The risk of suffering losses due to internal or
external incidents (see Oliver, Wyman &
Company, 2001)
3 For a comparison of these methods, see
Koyluoglu & Hickman (1998)

Modelling

2. Economic capital: statistical versus direct aggregation approach

Common risk factors

Common
Risk interest
Factorsrates,
(GDP growth,
inflation,
(GDP growth,
inflation,
indexes etc) interest rates,
indices etc.)
Liability
Liability risk
Risk

Asset
Assetrisk
Risk

Credit
Credit

Market
Market

Non-life
Non-Life

V2

V1

V3

Total
Totaleconomic
Economiccapital
Capital
(direct aggregation approach)
(Direct Aggregation Approach)

VAR =

V V
i, j

tion and then taking the VAR. The direct

approach is currently the most commonly
used often with the simplification to a
simple sum of the individual VAR figures,
similar to the Basel II framework. Also, it
is usually only market and credit risk and
sometimes insurance risk that are modelled
by a statistical approach; operational risk is
usually modelled by heuristic approaches.
Banks often use a simple correlation
matrix for only three risk categories: credit, market and op risk a market average
(Oliver, Wyman, 2001) is the matrix:
1
= 0.8

0.4

0.8
1
0.4

0.4
0.4

showing that Level II and III aggregation

together recognises significant diversification benefits between credit, market and
op risk, unlike the regulatory framework.
The econometric aggregation approach is theoretically profound and can
be statistically accurate, but hard to implement, the direct approach is relatively
easy to implement but is nothing more
than a guesstimate.

ij

Use of economic capital

About 60% of financial institutions (Sarraf,
2003) use economic capital for internal
risk reporting, performance measurement
and planning/budgeting purposes, but
only about 40% include economic capital
in the pricing of products.
Almost all financial institutions included in the surveys either already use
a Raroc-based performance measurement system or intend to implement one
as soon as internal capital modelling is
complete (Capital Market Risk Advisors,
2001). The main driver for Raroc-based
performance measurement and capital
allocation are investors and shareholders, who seem to reward banks performing Raroc-based profitability and
allocation disclosure, according to ERisk
(2001). Since a bank-wide Raroc capital
allocation based on sophisticated EC
models allows us to optimise returns on
a fixed level of risk, EC can be viewed as
the best tool to find the optimal trade-off
between the conflicting interests of
shareholders on the one hand, who are
trying to avoid over-capitalisation to optimise profitability and debt holders and
policyholders on the other, who fear

Operational
Operational risk
Risk

Life
Life

Business
Business

V4

V5

Event
Event

V6

Total
Totaleconomic
Economiccapital
Capital
(statistical aggregation approach)
(Statistical Aggregation Approach)

VAR

under-capitalisation due to the implied

risk of insolvency.
An important finding in Sarraf (2003)
is the fact that more than 90% of the financial institutions under consideration
clearly stated their wish to use their internal credit capital model for regulatory
capital calculations in the future, provided they are sufficiently tested and reliable.
Even if regulators and auditors might find
good arguments against such a step, national policy makers will face an increasing pressure from the market to allow the
use of internal models for regulatory capital adequacy purposes, since once
these models are reliable it implies several advantages for an economy, that is,
a fully risk/return efficient capital allocation process and a possible competitive
advantage over those countries that keep
the old system. However, a convergence
of both regulatory capital models and economic capital models to one unified
model is not reasonable, since both approaches view risk from a different basis
regulatory capital is typically based on
the book value of assets and liabilities,
whereas economic capital models (in the
strict sense) focus on an economic value
[Link] MAY 2003 RISK

BASEL S19

Modelling
of balance sheet items.
A comparison of the absolute amounts
of economic capital and regulatory capital across different financial institutions
shown in figure 3 reveals two results (Capital Market Risk Advisors, 2001):

On average, the regulatory capital
framework seems to overestimate the financial risk of banks, because of the neglecting of diversification benefits
between different risk categories, since all
banks (13%) that reported a higher economic capital than regulatory capital also
neglect these benefits in their internal
model, so their economic capital must be
clearly overstated, especially because it
includes various risk types that the regulatory framework neglects.
The relationship between economic
capital and regulatory capital differs significantly among different financial institutions for two reasons:
i. The Basel regulatory capital framework
is inconsistent, that is, not risk-sensitive,
so the relationship between regulatory
capital and economic capital strongly depends on the banks business.
ii. There are significant differences between different banks economic capital
models.
Figure 3 displays EC versus regulatory capital (RC). Banks that reported a
higher EC than RC do not take into account diversification benefits between
different risk categories (simple sum approach see above).

Problems with economic risk capital

A main problem is the fact that apart from
market risk, where VAR models are well
established and a back-testing of the models is straightforward due to a sufficient
pool of market data, VAR-type models are
not very well developed for other types
of risk, that is, for credit risk tools such as
CreditMetrics or CRisk+ back-testing techniques were developed only recently
(Granger & Huang, 1997) and an intensive back-testing of these methods, which
is absolutely necessary to confirm the
models confidence level, is not as simple
due to the lack of data and the longer time
period considered in credit risk. In operational risk and business risk, the devel-

3. Economic capital versus regulatory capital

EC significantly less than RC

EC less than RC
EC equals RC
EC higher than RC
No EC calculated

S20

BASEL RISK MAY 2003 [Link]

opment of VAR-type models is even further behind. Hence, it is clear that the accuracy of models for the aforementioned
risk types differs considerably.
An important problem is the valuation approach that the statistical loss distribution is based on, except for assets
that are traded on a liquid market with
a well-defined market price. For nontraded assets, for example, loans, there
are two different approaches for the calculation of economic capital: the markto-market approach (for example,
CreditMetrics) calculating a market
price for loans using the yield curve for
the corresponding counterparty rating
and hence including migration and default risk into the EC calculation; and the
default model (for example CRisk+),
using the book value perspective and
modelling default risk only (Saunders,
1999). It is arguable as to whether the
pure default approach is reasonable
under the economic capital methodology, since the latter is usually based on
an economic perspective. This problem
especially arises when capital adequacy
or capital allocation and profitability issues are investigated based on economic capital, since the latter relies on an
economic picture, whereas capital is typically only known from a book value
point of view.
A major problem arising when aggregating different risk types is the use of an
adequate time horizon, which usually differs significantly among different risk categories. For example, for market risk VAR
calculations, a time-horizon of one to 10
trading days is used, for credit risk and
business risk typically one year and for

insurance risk up to 30 years (life insurance). Hence, an adequate scaling of the

time-horizon has to be developed, which
is difficult, especially for market risk, since
the assumption of a constant trading portfolio over a one-year horizon is not realistic, as traders typically react to
movements in market indexes, for example when they re-hedge portfolios. Hence,
a certain effort is necessary in quantitative modelling to take into account the behaviour of actively managed portfolios.

Conclusion
Recent market surveys demonstrate that
economic capital is evolving into the standard model for a comprehensive and consistent monitoring of the financial risks of
a financial institution, avoiding the inconsistency and incompleteness of the regulatory capital framework, including Basel II.
But the level of sophistication of internal
capital models varies significantly, and currently there seems to be no financial institution that uses a fully statistical aggregation
approach on all levels to explicitly model
all interactions between different risk types
on the basis of a common set of key risk
indicators.
A key driver for economic capital besides the influence of the regulators is
clearly investors, which seem to reward
banks using risk-adjusted profitability analysis. In this sense, a Raroc strategy based on
economic capital as a risk measure is turning out to be a key weapon for banks ability to compete in financial markets.
Guido Giese is head of market risk management, business advisory services at
KMPG Zurich

References
Bank for International
Settlements, 1996
Amendment to the capital Accord to
incorporate market risks
January
Basel Committee on
Banking Supervision, 2002
Quantitative impact study 3 technical
guidance
October

Koyluoglu H and A Hickman, 1998

Reconcilable differences
Risk October 1998, pages 5662
Oliver, Wyman & Company, 2001
Study on the risk profile and capital adequacy
of financial conglomerates
February
Sarraf H, 2003
KPMG European capital survey 2002
Credit Risk International

Capital Market Risk Advisors, 2001

Economic capital survey
Available at [Link]

Saunders A, 1999
Credit risk measurement
John Wiley & Sons

ERisk, 2002
Economic capital bank survey methodology
Available at [Link]

Wilde T, 2001
IRB approach explained
Risk May 2001, pages 8790

Granger C and L Huang, 1997

Evaluation of panel data models: some
suggestions from time-series
Discussion paper 97-10, Department of
Economics, University of California at San
Diego

Wilson T, 1997
Portfolio credit risk
Risk September 1997, pages 111117, and
October 1997, pages 5661

Joint Forum on Financial

Conglomerates, 1999
Supervision of financial conglomerates
A collection of papers, February