procyclical implications of basel ii: can the cyclicality of capital requirements be contained?

Pb

HN

a

ARRAA

JEGGG

KBPC

1

bmfmtibahidrtutb

c

1d

Journal of Financial Stability 7 (2011) 138–154

Contents lists available at ScienceDirect

Journal of Financial Stability

journal homepage: www.elsevier.com/locate/jfstabil

rocyclical implications of Basel II: Can the cyclicality of capital requirementse contained?

enrik Andersen ∗

orges Bank, Financial Markets Department, Bankplassen 2, Norges Bank, Oslo, Norway

r t i c l e i n f o

rticle history:eceived 24 June 2009eceived in revised form 5 May 2010ccepted 7 May 2010vailable online 19 May 2010

EL classification:3221

a b s t r a c t

While the current capital adequacy framework, Basel II, aims to make banks’ capital requirements moresensitive to the underlying risk of the assets, it may also introduce an additional source of procyclicality inthe banking sector. In this paper we assess the potential cyclicality of Basel II for the entire bank portfolio.This is in contrast to previous studies which have taken into account only parts of banks’ assets, and alsoneglected the potential cyclicality of bank capital. We apply a detailed data set covering a relatively longperiod to analyse the cyclicality of both bank capital and Basel II capital requirements. Moreover, weemploy a more comprehensive system of models than applied in the existing literature. Consistent withprevious evidence, we find a substantial increase in the calculated Basel II capital requirements at the

2833

eywords:asel IIrocyclicality

same time as bank capital deteriorates in a recession scenario. However, we also find that the cyclicalityof Basel II capital requirements may be effectively contained if risk weightings are based on a sufficientlylong observation period which includes economic downturns.

© 2010 Elsevier B.V. All rights reserved.

Bbceaf

aNrtc

phadb

apital positions

. Introduction

Bank regulations may amplify the procyclicality inherent inank lending behaviour; see e.g. Kashyap and Stein (2004). Thisay particularly be the case with the current capital adequacy

ramework. Basel II aims to make minimum capital requirementsore sensitive to the underlying risk of the banks’ operations than

he former framework, Basel I. However, the risk sensitive cap-tal requirements may also reinforce the procyclicality of bankehaviour; see e.g. Borio et al. (2001) and Catarineu-Rabell etl. (2003). As bank assets and loans in particular, are assignedigher risk weightings during economic downturns, required cap-

tal will increase. At the same time capital positions will tend toeteriorate as loan losses accelerate. This may induce banks toeduce lending and increase lending margins, thereby amplifyinghe procyclicality of bank lending. Conversely, during an economicpturn lower risk weightings and excess capital holdings may con-ribute to expanding credit volumes and risk fuelling a credit-led
oom.
When the capital adequacy ratio falls below desirable levels,uts in lending is not the only option; see Benford and Nier (2007).

∗ Tel.: +47 22 31 64 66.E-mail address: [email protected].

l

ficseo

572-3089/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.jfs.2010.05.001

anks may cut dividends in order to increase the regulatory capital,ut this may be far from sufficient. Another option is to raise newapital, though this might be very costly for existing sharehold-rs when the banking sector is under pressure. Banks could alsottempt to sell part of their assets, but this cannot be the solutionor the entire sector.

Empirical research, however, strongly suggests that banks usu-lly reduce lending; see e.g. Francis and Osborne (2009). Moreover,ier and Zicchino (2005) find that the strength of the loan growth

esponse is non-linear and depends on the capital buffer availableo absorb losses. If capital buffers are below a comfort level, banksut their lending by more than if the capital buffers are ample.

It may be argued that the Basel II framework need not haverocyclical effects on the economic activity if every borroweras access to non-bank financing during downturns; see Saurinand Trucharte (2006). However, the ongoing financial crisis hasemonstrated that borrowers’ access to non-bank financing maye seriously hampered at the same time as banks contain their

ending.The contribution of this paper to the literature is threefold:

rst, going beyond the existing literature, we calculate the Basel II
apital requirements for all corporate exposures, household expo-ures, bank exposures and sovereign exposures. Considering thentire bank portfolio is important as the debt-servicing capacityf enterprises, banks, households and sovereigns may be affected
dx.doi.org/10.1016/j.jfs.2010.05.001

http://www.sciencedirect.com/science/journal/15723089

http://www.elsevier.com/locate/jfstabil

mailto:[email protected]

dx.doi.org/10.1016/j.jfs.2010.05.001

ncial

dctthTd

msimtaOscdt

odi2ti

mmbsn

ccNcefitHrabw

eitocvacrm

pfr

((T

M

tt

tatn2rtcast

ImBa

2

prianac

buSRpBS

wGs

H. Andersen / Journal of Fina

ifferently by the business cycle. Second, we analyse both theyclicality of capital positions and the cyclicality of Basel II capi-al requirements at the same time. By this we take into accounthat possible procyclical implications of Basel II also depend onow the banks’ capital positions are affected by the business cycle.hird, we apply a more comprehensive set of models and a moreetailed data set than commonly applied in the literature.1

We simulate the capital positions and minimum capital require-ents based on a suite of models. A bank model employs a stress

cenario from a macro model and estimated Basel II risk weight-ngs from an enterprise sector model, a household model, a logit

odel for banks and a loss given default (LGD) model. Contraryo most other studies, we apply the advanced IRB approach2 andllow for cyclicality in both probability of default (PD) and LGD.ur data set includes all Norwegian banks, all Norwegian joint-

tock companies and all Norwegian households. As our data setovers a relatively long observation period and includes economicownturn conditions, we are able to calculate different versions ofhrough-the-cycle (TTC) capital requirements.

Studies simulating the internal rating based (IRB) approachf Basel II find significant cyclicality in the capital requirementsetermined by internally estimated risk parameters.3 Estimated

ncreases in capital requirements in a downturn range from 30 to02%, partly depending on the macroeconomic shocks considered,he models employed and which parts of the bank portfolio werencluded.

The capital adequacy of a bank is determined both by theinimum capital requirement and the capital position. Thus, theacroeconomic effects of Basel II will also depend on how the

anks’ capital positions are affected by the business cycle. Mosttudies find that capital positions tend to deteriorate during eco-omic downturns.4

Our analysis sets out by demonstrating how risk sensitiveapital requirements can introduce an additional source of pro-yclicality in the banking sector. In a recession scenario for theorwegian economy, we document a substantial increase in thealculated Basel II capital requirements even when the banksmploy TTC rating systems where the risk parameters are based onve- and ten-year-moving averages. At the same time, bank capi-al deteriorates as banks record high losses on loans and securities.owever, the cyclicality can be effectively contained if the capital

equirements are calculated based on twenty-year-moving aver-
ge PDs and LGDs covering economic downturn conditions. This isecause a longer observation period gives new observations lowereight, and because the data from the previous economic down-
1 The literature typically tracks the rating for a hypothetical portfolio of corporatexposures using either rating transition matrices, or market indicators of probabil-ty of default (PD), as a proxy for a bank’s rating system. Changes in PDs are mappedo changes in capital requirements to estimate whether capital requirements varyver a business cycles. Simulation studies based on rating agency transition matri-es are often classified as more through-the-cycle (TTC). Studies applying differentersions of Merton’s option pricing model, such as Moody’s KMV, are often classifieds describing a point-in-time (PIT) rating system. Moody’s KMV default risk fore-ast is established as a commonly used indicator to calculate an enterprise’s defaultisk on the basis of its stock price, balance sheet information and an option pricingodel.2 Under the advanced IRB approach, banks provide their own estimates of the risk

arameters. The literature normally applies the foundation IRB approach. Under theoundation IRB approach banks provide their own estimates of the PD, while theemaining risk parameters are fixed by supervisors.

3 See e.g. Altman et al. (2005), Benford and Nier (2007), Catarineu-Rabell et al.2003), Gordy and Howells (2006), Goodhart et al. (2004), Jakubik and Schmieder2008), Kashyap and Stein (2004), Marcelo and Scheicher (2005), Saurina andrucharte (2006) and Taylor and Goodhart (2006).4 See e.g. Ayuso et al. (2004), Jokivuolle et al. (2008), Lindquist (2004), Bikker andetzemakers (2004) and Stoltz and Wedow (in press).

itLfp

a

R

wa

S

am

Stability 7 (2011) 138–154 139

urn ensures that the initial risk parameters are already high beforehe next downturn occurs.

During the recent financial crisis, market pressure forced bankso increase their capital adequacy ratios at the same time as lossesnd writedowns reduced their capital positions. This induced bankso tighten their credit standards which, in turn, amplified the eco-omic downturn. As a response, the Basel Committee at the end of009 issued proposals that will reduce the procyclicality of capitalequirements, and promote the build-up of capital buffers in goodimes.5 On the basis of a comprehensive impact assessment, a fullyalibrated set of standards will be developed by the end of 2010nd then gradually phased in. In this context, our analysis can shedome light on how the new standards should be calibrated in ordero mitigate the procyclicality of capital requirements.

This paper is organised as follows. Section 2 describes the BaselI capital requirement formula. Section 3 presents the data and

ethodology. Section 4 describes the procedures for projecting theasel II capital requirements and bank capital. Section 5 presentsnd discusses the results of our analysis while Section 6 concludes.

. Basel II and capital requirements

The Basel II framework is based on three mutually reinforcingillars: minimum capital requirements (Pillar I), the supervisoryeview process (Pillar II) and market discipline (Pillar III).6 Focus-ng on Pillar I, the Basel II framework contains three differentpproaches for calculating capital requirements for credit risk,amely the standardised approach,7 the foundation IRB approachnd the advanced IRB approach. Basel II also includes capitalharges for market risk and operational risk.

Under the IRB credit risk approaches, banks must categoriseanking-book exposures into broad classes of assets with similarnderlying risk characteristics. The classes of assets are: Corporate,overeign, Bank, Retail, Equity and Eligible purchased receivables.etail includes loans to households and loans to the smallest enter-rises. Loans to the remaining enterprises are covered by Corporate.ank covers loans and other exposures to financial institutions.overeign includes loans and other exposures to public entities.

Within the Basel II framework the formula for calculating risk-eighted assets (RWA) is a modified version of the so-calledaussian asymptotic single risk factor model of credit risk. Theolvency margin of the formula is set at 99.9%, i.e. the probabil-ty that required capital does not cover any deficits should be lesshan 0.1% over a one-year horizon. The formula is a function of PD,GD, exposure at default (EAD) and maturity (M). In addition, theormula includes a parameter for maturity adjustment (b) and aarameter for asset correlation (R).

The formula for calculating RWA for corporate, sovereign, banknd retail exposures is

WA = 12.5 × EAD ×∣∣∣LGD × N

(G(PD) +

√R × G(0.999)√1 − R

)− (PD × LGD)

∣∣∣× (1 + (M − 2.5)b)

(1 − 1.5b),

here N is the cumulative standard normal distribution functionnd G its inverse. The maturity adjustment (b) is given by b =

5 See Basel Committee on Banking Supervision (2009).6 For a detailed description of the framework, see Basel Committee on Banking

upervision (2006b).7 The standardised approach is basically Basel I with more refined risk-buckets

nd enhanced by allowing banks to apply external ratings and acknowledging riskitigation effects.

1 ncial

(srwPmr

mievubospTl

r

R ,

wbaar

R

Trcii

3

3

fastikikdt

a1r29e

o

ht1N

b1ntcid

3

troibr

jlpaa

Nisana(tbdmm

lraf(taedo

hems

40 H. Andersen / Journal of Fina

0.11852 − 0.05478 × ln(PD))2, except in the case of retail expo-ures where b is set equal to zero. The decision to adjust for maturityeflects the intuitive notion that, on the one hand, risk increasesith loan duration and, on the other hand, the likelihood that the

D will deteriorate increases when the initial PD is low and theaturity of the exposure is large. These factors suggest that capital

equirements should increase with the maturity of the exposures.Under the advanced IRB approach, banks provide their own esti-

ates of the PD, the LGD, the EAD and the M parameters. Thesenternal estimates must be grounded in historical experience andmpirical evidence. Basel II does not specify whether recent obser-ations should be given more weight or not. The length of thenderlying historical observation period for the PD measure muste at least five years. For corporate, retail and bank exposures, thene-year PD cannot be below 0.03%. The LGD and EAD estimateshould reflect economic downturn conditions and the observationeriod must be at least seven years (five years for retail exposures).he LGD estimates used for the IRB capital calculation cannot beess than the long-run default-weighted average.8

In the case of corporate, sovereign and bank exposures, the cor-elation factor (R) is given by

= 0.12

(1 − e−50PD

1 − e−50

)+ 0.24

(1 − 1 − e−50PD

1 − e−50

)− c(

1 − S − 545

)here c is equal to zero for all exposures, except in the case of SME

orrowers where the parameter is set equal to 0.04. S is expresseds total sales in million euros. For residential mortgage exposuresnd qualifying revolving retail exposures R is equal to 0.15 and 0.04espectively. For all other retail exposures R is given by:

= 0.03

(1 − e−35PD

1 − e−35

)+ 0.16

(1 − 1 − e−35PD

1 − e−35

)he formula takes only the correlation between the idiosyncraticisk of an exposure and the systematic risk into account, ignoringorrelations between the idiosyncratic risks of different exposuresn a portfolio. Thus, the formula is based on the assumption that alldiosyncratic risks are diversified away.

. Data and models

.1. Data

To calculate capital requirements according to the Basel IIramework we use Norges Bank’s proprietary database withccounting data for enterprises, households and banks. This dataet probably includes more comprehensive and relevant informa-ion for calculating Basel II capital requirements than the marketndicators used in several of the previous simulation studies. Mar-et liquidity effects, herd behaviour and several other mechanismsn the capital market may produce substantial variation in mar-et indicators that is not related to the borrowers’ probability ofefault. Moreover, market indicators are only available for a frac-ion of the banks’ borrowers.

Our data on enterprises include annual financial statements forll Norwegian joint-stock individual companies over the period988–2007. The number of enterprises submitting their financialecords to the database is up from 50,000 in 1988 to 190,000 in
007. The total bank debt of these companies accounted for around0% of the Norwegian banking sector’s loans to enterprises by thend of 2007. The data source is the Statistics on enterprises.
8 For each year considered in the observation period, the number of defaultsccurring is used to weight the final calculated average.

Vio

b

Stability 7 (2011) 138–154

For all Norwegian households we have tax return data withousehold income, debt and interest payments. This covers morehan two million households per year over the sample period986–2007. Data on standard living costs are taken from theational Institute for Consumer Research.

For all Norwegian banks we have detailed financial statements,alance sheets and capital adequacy reports9 over the period991–2008. Information on individual borrowers of each bank isot available, but the volumes of loans can be allocated to sec-ors and industries. Using this type of loan classification we canombine data from the banks’ balance sheets with detailed data onndividual enterprises and households along the sector/industryimension. The data source is the Bank Statistics.

.2. Models

We project the banks’ capital requirements and capital posi-ions using a suite of models. The literature typically employsating agency transition matrices or different versions of Merton’sption pricing model and market information to calculate the cap-tal requirements for a hypothetical credit portfolio, i.e. an artificialank. We simulate the capital positions and the minimum capitalequirements of real banks.

Chart 1 illustrates the relationship between our models. Pro-ections of key variables from a macro model, i.e. credit growth,ending rates, wages, gross domestic product, inflation, houserices and housing investments, are used as input in a bank model,n enterprise model, a household model, a logit model for banksnd a LGD model.

The macro model is an equilibrium correction model for theorwegian economy. A core macro model has been expanded by

ntroducing relationships that are central to the analysis of financialtability. The GDP equation includes feedback effects from creditnd house prices to real activity. The macro model thus inter-alises the co-movement and procyclicality of credit, asset pricesnd agent optimism discussed in the literature, see Borio et al.2001). The core model is estimated on a data set that includeshe 1988–92 banking crisis in Norway. The additional financial sta-ility equations are in general estimated on shorter samples partlyue to lack of data. Appendix A presents a stylised version of theacro model. For a more detailed review of the relationships in theacro model, see Andersen and Berge (2008).The enterprise model predicts PDs and bank loans at the firm

evel as a generalised logistic function of accounting data indicatorsepresenting earnings, liquidity, financial strength, industry, agend size of the company. The mean PD for each industry is projectedrom a base year by using autoregressive distributed lag modelsADLs) with inputs from the macro scenarios. The projected PD ishen used to predict the default risk at the firm level. The ADLsre estimated on the basis of annual data from 1989 to 2004. Thenterprise model and the projection procedures are described inetail by Nordal and Syed (2010), who also document the resultsf a back testing exercise.

The household model measures the financial margin of eachousehold, defined as the household income minus taxes, inter-st payments, repayment of debt and standard living costs. Thisargin serves as a reasonable measure of the households’ debt-

ervicing capacity. For a discussion of households’ margins, see
atne (2006, 2007). Forward projections of household income and
nterest payments are based on the macro scenarios. Repaymentf debt is assumed to be linear over twenty years. Standard living

9 The database does not include capital adequacy reports for branches of foreignanks.

H. Andersen / Journal of Financial Stability 7 (2011) 138–154 141

ship b

ca

tabcpaaLA

ve2ii2sf

eBpppmoc

4

l

al

B

r3cdbio

4

teefa

dtwcDptfNssTao4

Chart 1. The relation

osts are taken from the National Institute for Consumer Researchnd depend on key characteristics of the household.

The bank model is a non-behaviour model. It includes data fromhe banks’ annual financial statements, balance sheets and capitaldequacy reports. The bank model is used to generate paths foranks’ results and capital adequacy. The banks’ main income andost items are linked to variables in the macro model and the enter-rise model. The bank model takes loan volumes, lending rates,nd labour costs from the macro model. Projections of loan lossesre taken from the enterprise model, the household model and theGD model. For a more detailed description of the bank model, seeppendix B and Andersen and Berge (2008).

The logit model for banks provides estimates of the PD for indi-idual banks, based on indicators of the banks’ capital adequacy,arnings, liquidity, credit risk and concentration risk (see Andersen,008).10 These indicators are projected based on the macro scenar-

os before the PDs are computed by the logit model. The logit models estimated on the basis of quarterly bank data between 2000 and005. The predictive powers of the logit model were verified out-ide the estimation sample period by applying it to 11 failed banksrom the period 1990–93.

The LGD model is a simple dynamic model where the mainxplanatory factor is changes in commercial property prices; seeernhardsen and Larsen (2007).11 They found that the LGD can berojected with reasonable accuracy by this model. This is not sur-rising as banks’ lending to enterprises is often secured againstroperty. According to the estimated model, a 10% drop in com-ercial property prices leads, cet. par., to an increase in the LGD

f around 11 percentage points. Over time LGD tends towards aonstant level of 35%.

. Projections

In this section we outline our procedures for projecting the sixargest Norwegian banks’12 capital positions, Basel II exposures and

10 Probability of failure = 9.55 − 0.93Capital adequacy − 50.5Return onssets − 0.059Share of mortgages − 0.094Liquidity risk measure + 1908.3Expectedoss measure + 8.4Concentration risk measure.11 LGD(t) = 0.085 + 0.76LGD(t − 1) − 1.09�ln(commercial property prices).12 DnB NOR, SpareBank 1 SR-Bank, Sparebanken Vest, SpareBank 1 SMN and Spare-ank 1 Nord-Norge.

tn(top

mR

etween the models.

isk parameters based on the data and models described in Section. Generally, our procedures for approximating exposures and cal-ulating risk parameters are in line with the Basel II requirementsescribed in Section 2. However, as it is not possible to reproduceanks’ risk profile completely, we need to incorporate some prox-

es for Basel II capital requirements on exposures not identified inur data set.

.1. Projection of Basel II exposures

We calculate Basel II capital charges on the banks’ total loanso every sector and industry, including household exposures, bankxposures and sovereign exposures. The bulk of the empirical lit-rature only assess the cyclicality of Basel II capital requirementsor a hypothetical credit portfolio or parts of an individual bank’sssets, typically the corporate portfolio.

Exposures to the corporate sector are approximated based onata from the Bank Statistics and the Statistics on enterprises. Loanso the corporate sector accounted for 28% of the assets in the Nor-egian banking sector by the end of 2007. In the Bank Statistics the

orporate loan portfolio is composed of nine different industries.13

ata on total sales for each enterprise with bank debt makes itossible to classify debts within each of the nine different indus-ries as corporate exposures, retail exposures and SME exposuresor banks. While all enterprises with total sales less than 2 millionOK are assumed to be retail exposures, all enterprises with total

ales exceeding 400 million NOK are assumed to be corporate expo-ures. The remaining enterprises are assumed to be SME exposures.his is in accordance with the Basel II classification. Based on thisssumption, 12.8% of the loans to the enterprise sector by the endf 2007 were corporate exposures, 38.6% were SME exposures and8.7% were other retail exposures. We assume that this distribu-ion of exposures to corporates, retails and SMEs within each of theine industries is identical for each of the largest Norwegian banksand remains the same over the simulation period). This assump-
ion should be innocuous when applied to the six large banks inur sample. We proceed to calculate capital requirements for cor-orate, retail and SME exposures within each industry for each
13 Primary industries, Property management, Commercial services, Mining andanufacturing, Oil and gas, Shipping abroad, Other transport, Construction and

etail trade, hotel and restaurant.

1 ncial

ih

etdtm

fewegra

SNalas

pcew

dgtspegt

4

rattb(bi(it

OCfasP

hwit

bgbwaret

statHppdf

oiie

(fltf

msiet

rmgdpcBcoscdHeters if the distribution of PDs, LGDs and EADs is not stable throughthe economic cycle.

15 CEBS Group 1 banks are banks located in a country which are part of the Com-mittee of European Banking Supervisors, have Tier 1 capital in excess of 3 billioneuros, are diversified and internationally active. For a detailed description of the QIS5 study, see Basel Committee on Banking Supervision (2006a).


ndividual bank. We thus take into account that different banksave different exposures to individual industries, see Chart 2.

In this way we are able to differentiate the quality of borrow-rs between the banks. However, the data does not enable us toake into account the variation in bankruptcy probabilities withinifferent risk groups. Some banks may end up with less creditwor-hy borrowers than others due to lower risk aversion or poorer risk

anagement.Exposures to the households are approximated based on data

rom the Bank Statistics. Residential mortgage loans (including homequity lines of credit) accounted for 35% of the assets in the Nor-egian banking sector at the end of 2007. The remaining retail

xposures, i.e. total loans to the retail market less residential mort-ages and other retail exposures are assumed to be qualifyingevolving retail exposures. 4% of the Norwegian banking sector’sssets are classified as qualifying revolving retail.

Bank exposures and sovereign exposures are reported in the Banktatistics. While bank exposures accounted for 6% of assets in theorwegian banking sector at the end of 2007, sovereign exposuresccounted for less than 1%. Equity includes all assets posted asong-term shareholdings in the Bank Statistics. Equity exposuresccounted for less than 1% of the assets in the Norwegian bankingector.

Off-balance sheet exposures and Eligible purchased retail and cor-orate receivables are not identified in the Bank Statistics. Thus,apital requirements for these exposures are not calculated. Thesexposures accounted for a negligible share of the six largest Nor-egian banks’ total capital requirements by the end of 2008.

Altogether, we arrive at 31 different risk groups, including 27ifferent risk groups for enterprises. In the simulations below,rowth in the banks’ total loans to households and enterprises areaken from the macro model (see Chart 1). The distribution of expo-ures to each of the nine industries in the corporate portfolio will berojected by the enterprise model. Growth in residential mortgagexposures and qualifying revolving retail exposures is set equal torowth in total loans to households. Other exposures are assumedo grow at the same rate as total loan growth.

.2. Projection of Basel II risk parameters

We apply the advanced IRB approach to calculate capitalequirements for all Basel II exposures.14 Under the advanced IRBpproach, banks provide their own estimates of the risk parame-ers. Contrary to most other studies, we allow for cyclicality in bothhe PDs and LGDs. Several studies argue that the LGDs are likely toe affected by the economic cycle; see for example Altman et al.2005) and Dierick et al. (2005). Moreover, EADs may increase asorrowers make more use of their loan commitment limits dur-

ng an economic downturn; see for example Allen and Saunders2003). As the possibility of additional drawings on credit lines isncorporated in the predicted loan growth we implicitly assumehat the EAD is affected by the cycle.

PDs for loans granted to the enterprises, i.e. Corporate, SME andther retail exposures are taken from the enterprise model (seehart 1). The PDs for bank exposures are taken from the logit model

or Norwegian banks. The PD for bank exposures is set equal to the
verage probability of failure as predicted for the six banks in ourample. The PDs for Sovereign exposures are set equal to the averageD reported in the 5th Quantitative Impact Study (QIS 5) by Basel II
14 The six banks in our sample have only adopted the IRB approach for parts of theiroldings. For example, while the share of IRB exposures of total DnB NOR exposuresas 38% at end-2007, the IRB share of Nordea Bank Norge was 54%. The banks are

n the process of extending the IRB approach to their entire portfolios as set out inhe Basel II framework.

msth

wtebP

Stability 7 (2011) 138–154

anks included in CEBS Group 1.15 The PDs for the residential mort-age loans and qualifying revolving retail exposures are assumed toe proportional to the share of household debt held by householdsith a negative financial margin. The level of these PDs is calibrated

ccording to the QIS 5 study and the average Basel II parameterseported by the Norwegian banks. Thus, our PDs are initially setqual to the PDs reported by the Basel II banks and then assumedo change with households’ debt-servicing capacity.

The LGDs on corporate exposures, SMEs, bank exposures,overeign exposures and small firms defined as retail exposures areaken from the LGD model. The LGDs of residential mortgage loansnd qualifying revolving retail exposures are initially set equal tohe average LGD of the QIS 5 study, i.e. 16.1 and 55.0% respectively.owever, these LGDs are assumed to change over the simulationeriod in line with the LGD from the estimated model for enter-rises. The LGDs are not allowed to be less than their long-runefault-weighted average. This is in accordance with the Basel IIramework.

The EADs of all on-balance sheet exposures are measured grossf provisions. The effects of on-balance sheet netting are not takennto account. The possibility of additional drawings on credit liness indirectly incorporated in the predicted loan growth. The EADstimates are therefore of a PIT character.

According to the advanced IRB approach the effective maturityM) should be measured based on cash flow data. As data on cashows are not reported in the Bank Statistics, we apply the founda-ion IRB approach for measuring M. Thus M is set equal to 2.5 yearsor all exposures included in our study.

Due to data limitations we apply the simple risk weightethod16 for calculating the capital requirements for equity expo-

ures. As the share of equity holdings not publicly traded is notdentified in our data set, we apply a 350% risk weight for all equityxposures. This is in line with the simple risk weight method underhe market-based approach for equity exposures.

We calculate average risk parameters for each of the 31 differentisk groups. These aggregate bankruptcy probabilities reveal infor-ation only on the average quality of borrowers from each risk

roup, and therefore only reflect the idiosyncratic risk for each bankue to its specialisation in particular groups of borrowers. Anotherotential disadvantage of using average risk parameters is that theapital requirement formula is not linear, but concave. However,asel II urges banks to employ average risk parameters when cal-ulating capital requirements for retail exposures.17 According tour estimates 52% of the assets owned by the Norwegian bankingector were retail exposures at the end of 2007. Consequently, ouralculation of capital charges based on average risk parameters forifferent risk groups probably serve as a reasonable approximation.owever, errors may still arise from applying average risk param-

16 Under the market-based approach, banks are permitted to use two differentethods: a simple risk weight method or an internal models method. Under the

imple risk weight method a 300% risk weight is to be applied to equity holdingshat are publicly traded and a 400% risk weight is to be applied to all other equityoldings.17 According to Basel II, a retail exposure “must be one of a large pool of exposures,hich are managed by the bank on pooled basis”. Furthermore, Basel II underlines

hat “For each identified pool of retail exposures, banks are expected provide anstimate of the PD and LGD associated with the pool”. In Norwegian regulations,anks estimating PDs for individual borrowers with statistical models can set theD associated with the pool equal to the unweighted average of these PDs.


nks an

4

rsso

arniou

flct

bbmcrBr

scdcr

wtnaimo

ctc

5c

lbf

2bvooatti8

pabggt

Chart 2. The shares of loans to different industries for Norway’s five largest ba

.3. Projection of additional Basel II capital requirements

We need to incorporate some proxies for additional capitalequirements on exposures which are not identified in our dataet. The calculated cyclicality in our analysis is not expected to beignificantly affected by the fact that we do not cover the cyclicalityf every single risk exposure.

The capital requirement for operational risk is calculated byssuming that the (last reported) capital charge for operationalisk increases at the same rate as gross income. Operational risk isot expected to be a source of increased procyclicality in the min-

mum capital requirements, as historical experiences indicate thatperational risk charges actually tends to increase during economicpturns.

The calculation of capital charges for market risk cannot be per-ormed on data available in the Bank Statistics. Instead, we use theast reported capital charge for market risk at each bank. This capitalharge is adjusted for annual growth of the financial assets exposedo market fluctuations identified in the banks’ balance sheets.

The additional Pillar II capital requirement18 is approximatedy multiplying the capital charge for all the identified exposuresy annual real GDP growth (mainland Norway) from the macroodel. This ensures that the Pillar II capital charge reflects business

ycle effects. However, negative GDP growth will not result in anyeductions in the Pillar I capital requirements. This is in line withasel II which does not allow for any reductions in Pillar I capitalequirements.

The capital charge for the remaining exposures, i.e. off-balance-heet exposures, purchased receivables and exposures to foreignounterparts, and risk-reducing effects of guarantees and credit
erivatives, are endogenously determined as a residual; the capitalharge is equal to the difference between the latest Basel II capitalequirements identified in the Bank Statistics and the sum of all the
18 Under the supervisory review process (Pillar II), supervisors should addresshether the banks need to hold additional capital against credit risk concentra-

ions, liquidity risk, interest rate risk in the banking book and other risks that areot, or not fully, covered in Pillar 1. Supervisors generally expect banks to operatebove the minimum regulatory capital ratios and can require banks to hold capitaln excess of the minimum. External factors such as business cycle effects and the

acroeconomic environment should be considered. For a more detailed descriptionf the Pillar II capital charge, see Basel Committee on Banking Supervision (2006b).

ld

alpar

r

d Nordea Bank Norge. Percentage of total loans to the enterprise sector, 2008.

apital charges specified above. In the projections, we assume thathis residual grows at the same rate as the sum of all the capitalharges specified above.

. Simulations on the Basel II capital requirements andapital positions

Based on the procedures outlined in Sections 3 and 4 we calcu-ate Basel II capital requirements and capital positions for the sixanks under a four-year stress scenario. The projection period isrom end-2007 until end-2011.19

A severe shock in 2008 produces annual GDP growth of −1.6% in009, −0.5% in 2010 and 0.4% in 2011. Such a deep recession has noteen recorded in Norway since the late 1980s, i.e. during the pre-ious banking crisis. A weakening of household confidence in theirwn financial situation and the downturn of the Norwegian econ-my lead to a sharp fall in house prices. In 2011, real house pricesre about 50% lower than at the end of 2007. Consumer price infla-ion increases as a result of higher imported inflation. In responseo higher consumer price inflation, the interest rate rises rapidlyn 2008 to curb inflation. In 2009 the banks’ lending rate peaks at.1%.

Moreover, we assume that banks’ risk-willingness declines inace with heightened global liquidity and credit risk. Banks aressumed to tighten lending sharply. Lower house prices and higherank lending rates result in lower corporate and household creditrowth and weaker economic developments. At the most, annualrowth in loans to enterprises and households falls 5 and 2% respec-ively. Weaker macroeconomic developments and higher bankending rates reduce borrowers’ debt-servicing capacity. This pro-uces loan losses of 2.2% of total loans in 2011.

As in e.g. Peura and Jokivuolle (2004), we assume that banksre not able to raise new capital. The margin between deposit andending rates is constant through the projection period, but the
remium on banks’ market funding costs increases. Dividends aressumed to be 50% of after-tax profit, and are not paid if banksecord a negative profit.
19 Notice that this is a counterfactual scenario different from the one actuallyealised in 2008 and 2009.

144 H. Andersen / Journal of Financial Stability 7 (2011) 138–154

Chart 3. Non-performing loans in different industries in percent of total loans to the enterprise sector. Actual numbers 1998–2007 and PD projections for 2008–2011.

Chart 4. The share of household debt held by households with a negative financial margin. Percentage of total loans to the household sector. Actual figures 1986–2007 andprojections for 2008–2011.

Chart 5. Probability of liquidity or solvency problems for Norway’s five largest banks and Nordea Bank Norge. Actual figures 1991–2007 and projections for 2008–2011.


tage o

Ppfja1a

ls2(d

fiaromt

pBdahp2

tsr

5

saoaeto

iads

iictfibvP

ylTaeir

wrap

nrbwrthsaedata that do not capture a complete credit cycle.

As reported in Table 1, the Basel II risk-weightings increase sub-stantially for Bank exposures, Corporates, SMEs and Other retail

Chart 6. The loan-loss ratio. Percen

Initially, we employ our models to calculate PIT estimates ofDs and LGDs based on the stress scenario. Thereafter, these riskarameters are combined with the EADs and Ms and the Basel IIormula to compute capital requirements. Historical data and pro-ections of the PDs are displayed in Charts 3–5. Chart 3 shows thennual share of non-performing loans in different industries since988. The stress scenario PDs projected by the enterprise modelre also included in the chart.

In several industries the PDs increase substantially during theast years of the simulation period and reach the highest levelince the Norwegian banking crisis. The increase in the PDs from008 to 2011 varies between 2% (property management) and 774%primary industries). Hence, the effect of the stress scenario onifferent industries differs to a large extent.

The share of household debt held by households with a negativenancial margin is reported in Chart 4. As outlined in Section 4.2, wessume that the PDs for residential mortgage loans and qualifyingevolving retail exposures develop in line with this ratio. The sharef household debt held by households with a negative financialargin has been falling rapidly since 1989. However, according to

he household model the share increases in 2007 and 2008.Chart 5 displays the average probability of liquidity or solvency

roblems for the five largest domestically held banks and Nordeaank Norge. The probability of liquidity or solvency problems iserived from the logit model for Norwegian banks. The mean prob-bility of liquidity or solvency problems was 65% in 1991, at theeight of the Norwegian banking crisis. In the stress scenario thisrobability increases rapidly during the simulation period and is4% in 2011.

Chart 6 displays the historical loan-loss ratio. This ratio is equalo the historical LGD. Towards the end of the previous banking cri-is, i.e. in 1991, the loan-loss ratio was 54%. In 2002 the loan-lossatio peaked again at 25%.

.1. PIT Basel II capital requirements

We calculate Basel II capital requirements based on a PIT ratingystem and different versions of a TTC rating system. PIT modelsre based on more recent data and aim to predict developmentsver the next year, whereas TTC models aim to predict a long term
verage performance for the obligor. According to Catarineu-Rabellt al. (2003) many banks consider their rating systems as more PIThan TTC in nature. The main reason for this is the difficulty ofbtaining sufficiently long data series for use in TTC systems. This t
f total defaulted loans, 1991–2007.

s especially the case for recently established banks or banks oper-ting in new markets. In addition, banks may prefer to use recentefault data which is regarded as more relevant to the currentituation.

As a transitional arrangement the Financial Supervisory Author-ty of Norway allowed banks to employ the advanced IRB approachn 2008 and the foundation IRB approach in 2008 and 2009 even inases where the historical observation period was no longer thanwo years.20 This may have been done to promote a level playingeld. Finally, constructing a TTC system is technically challengingecause structural variables that are used to classify borrowers alsoary with the cycle. Altogether, these facts indicate that calculatingIT capital requirements is relevant.

We initially calculate PIT capital requirements based on singleear PDs and LGDs. The average PDs and LGDs included in our calcu-ation of PIT Basel II capital requirements are reported in Table C.1.he 0.03% floor is not binding in any single case. The LGDs are notllowed to be less than the average LGDs, which are assumed to bequal to the average LGDs reported in the QIS 5. This assumptions in line with the Basel II requirement that LGD estimates shouldeflect economic downturn conditions.

We compare our projections of Basel II capital requirementsith Basel I capital requirements which are based on constant

isk weights. The calculated Basel I capital requirements are onlyffected by growth in banks’ total assets as we assume no sizeableortfolio shifts during the projection period.

We find that total Basel II capital requirements grow sig-ificantly faster in the stress scenario than the Basel I capitalequirements, see Table 1. The Basel II risk-weighted assets increasey 152% during the simulation period. In 2011 the total Basel II risk-eighted assets for the six banks are 137% higher than the Basel I

isk-weighted assets. The largest relative difference between thesewo figures for a single bank is 155%. These results support theypothesis that the Basel II capital requirements might rise sub-tantially in a recession, as credit risk materialises and borrowersre downgraded. The results also highlight the fact that risk canasily be underestimated if measured over short periods and on

20 See § 49-2 of the capital requirement regulations for Norwegian financial insti-utions (2006): FOR 2006-12-14 nr 1506.


Table 1Risk-weighted assets for the five largest Norwegian banks and Nordea Bank Norge. Millions of NOK, 2008–2011.

September 2008 2008 2009 2010 2011

Reported Basel II risk-weighted assets 1295 0 0 0 0Operational Risk 67 56 57 59 63Market Risk 46 50 49 48 48Pillar II 30 36 0 0 0Retail mortgages 124 206 466 453 300Qualifying revolving retail 24 24 35 30 19Bank 157 186 545 693 682Sovereign 4 4 7 8 7Corporates 61 73 139 191 178SME 162 193 367 499 469Other retail 117 140 263 339 307Equities 67 68 66 65 65Residual 436 526 1050 1256 1126

embLpd

fi2rd

5

idl(aot

csytia0

si1im

fBwbwhw

mt

rc(tbmEtmTpiawt1

lmwcbtaafib

5

mdbndividends.21

Chart 7 displays the average capital adequacy ratio for thesix banks calculated using the Basel I approach, the Basel II PIT

21 Under Basel II, the difference between total provisions and total expected lossesmust be deducted from regulatory capital when expected losses exceed provisions.

Total Basel II risk-weighted assets 1295Basel 1 risk-weighted assets 1479

xposures. The main driver behind the accelerating capital require-ents for bank exposures is the increase in the PD mainly driven

y deteriorating earnings and falling capital adequacy ratios. TheGDs of all sectors increase rapidly during the simulation period asroperty prices fall by around 50%. This sharp fall is an importantriver behind the increase in the total capital requirements.

The share of household debt held by households with a negativenancial margin peaks in 2008 at 14%, and then falls to 13.3% in009 and below 10% in 2010 and 2011. This reduces the capitalequirements for retail mortgages and qualifying revolving retailuring the last years of the projection period.

.2. TTC Basel II capital requirements

Under the Basel II framework banks are urged to use a TTC rat-ng system. In order to calculate TTC PDs we combine historicalata with projections of the PDs. The data employed to calcu-

ate TTC PDs are shown in Charts 3–5. As in Benford and Nier2007), we set the TTC PDs equal to the five-year-moving aver-ge PDs. We also calculate TTC Basel capital requirements basedn ten-year-moving average PDs. Accordingly, we apply five- anden-year-moving average LGDs.

The risk parameters included in the TTC calculation of Basel IIapital requirements are reported in Tables C.2 and C.3. The PDstill increase rapidly during the simulation period. When the ten-ear-moving average is applied, the increase in the PDs duringhe simulation period differs from 1% (mining) to 127% (primaryndustries). The increase in the PDs based on the five-year-movingverage differs from 16% (mining) to 386% (shipping abroad). The.03% floor is not binding in any single case.

The five- and ten-year-moving average LGDs also increaseubstantially. When the ten-year-moving average is applied, thencrease in the LGDs during the simulation period differs from7% (qualifying revolving retail) to 102% (retail mortgages). The

ncreases in the LGDs are somewhat higher when the five-year-oving average is employed.Tables C.4 and C.5 report the TTC calculated risk-weighted assets

or the different exposures. Interestingly, the increase in the TTCasel II risk-weighted assets is still sizeable. The Basel II risk-eighted assets, based on the ten-year-moving average, increase

y 64% during the simulation period. In 2011 the Basel II risk-eighted assets based on the ten-year-moving average are 55%igher than the Basel I risk-weighted assets. The difference is 123%
hen the five-year-moving average PDs and LGDs are applied.
These results reveal that banks adopting TTC rating systemsay still experience substantial cyclicality in their Basel II capi-

al requirements. The calculated increases in the Basel II capital

5Espa

1562 3045 3642 32641461 1412 1383 1377

equirements are higher than the increases documented by empiri-al studies like Kashyap and Stein (2004) and Saurina and Trucharte2006). According to Kashyap and Stein (2004), TTC models leado increases in capital requirements for corporate exposures ofetween 30 and 45% over the period 1998–2002. This was a periodarked by pronounced economic slowdowns in both the US and

urope. PIT models result in increases of between 70 and 90% overhe same period. Saurina and Trucharte (2006) found that the maxi-

um TTC capital requirements were 56% higher than the minimumTC capital requirements for retail mortgages in Spain during theeriod 1990–2004. Our analysis shows increases in the Basel II cap-

tal requirements in the range of 64% (TTC) to 152% (PIT). Benfordnd Nier (2007) found that the maximum TTC capital requirementsere between 170 and 202% higher than the minimum TTC capi-

al requirements for retail mortgages in the UK during the period983–2006.

Comparing results across different studies is, however, chal-enging as they vary along a number of dimensions. First, the

acroeconomic shocks considered differ to a large extent. Whilee consider a recession scenario, other studies consider the impli-

ations of milder economic slowdowns. Second, the universe oforrowers considered varies across different studies. This affectshe results when the sensitivity to macroeconomic shocks differscross borrowers. Finally, the models employed and the Basel IIpproaches applied differ to a large degree. For instance, allowingor cyclicality in LGDs in our analysis, contrary to most other stud-es, raises the estimated increase in Basel II capital requirementsy around two thirds.

.3. Simulations on capital positions and Basel II capital adequacy

Ultimately, it is the cyclicality in the capital adequacy thatatters. Thus, the macroeconomic effects of Basel II will also

epend on how the banks’ capital buffers are affected by theusiness cycle. In our bank model the regulatory capital is endoge-ously determined by the profit net of losses and after taxes and

0% of the deduction must be made from Tier 1 capital and 50% from Tier 2 capital.xpected losses associated with equity exposures under the PD/LGD approach andecuritisation exposures are not included in the sum of total expected losses. Totalrovisions include specific provisions, partial write-offs and general provisions. Wessume that total provisions equal total expected losses.

H. Andersen / Journal of Financial

CN

awtTbfn1rts

Bsso9bi

cTay

otceopattcd

csfioioac

woTmt

aigotp

amAoTcysts

tcbbsm

6

cbcaBNmappm

ooItats

bervcm

hart 7. Mean capital adequacy in the five largest domestically held banks andordea Bank Norge. Percents of risk-weighted assets, 2008–2011.

pproach and the two Basel II TTC approaches. With constant riskeights (Basel I) the capital adequacy of the six banks falls close

o the required level of 8% in 2011 unless new capital is raised.he reported Basel I capital adequacy ratio is mainly determinedy changes in the capital position as we assume no sizeable port-olio shifts during the projection period. In the recession scenario,egative results in 2010 and 2011 reduce the capital position by6% from 2008 to 2011. Consequently, the Basel I capital adequacyatio falls from 10.0% in 2008 to 8.8% in 2011. This demonstrateshat changes in the capital position may affect capital adequacyubstantially.

We find a negative co-movement between capital positions andasel II capital weightings, i.e. capital positions deteriorate at theame time as capital requirements increase during the recessioncenario. With the PIT Basel II approach the capital requirementsf the six banks are violated already in 2009, and the ratio falls from.3% in 2008 to 3.7% in 2011. Thus, the added cyclical pressure onank capital positions caused by Basel II PIT capital requirements

s quite extreme.With the TTC Basel II approaches the added cyclicality in the

apital adequacy ratio is still sizeable. Even if the six banks employTC Basel II capital requirements based on ten-year-moving aver-ge PDs, they violate the capital requirements during the two lastears of the projection period.

One explanation behind the small difference in the cyclicalityf TTC and PIT capital requirements may be the fact that the capi-al requirements formula is concave. Thus, the effect on the capitalharge of an 800% increase in the PDs is less than eight times theffect of a 100% increase in the PDs. In addition, the PDs basedn moving averages increase substantially during the simulationeriod, see Tables C.2 and C.3. Loan losses are often unexpectednd tend to increase rapidly. Surprisingly high loan losses affecthe risk parameters even when the historical observation period isen years. Thus, Basel II may introduce significant cyclicality in theapital requirements even when the banks have access to defaultata over a full decade.

However, the cyclicality in the Basel II capital requirementsan be effectively contained if the historical observation period isufficiently long and includes economic downturn conditions. Wenally apply twenty-year-moving average PDs and LGDs, basedn the data in Charts 3–5, in order to cover the previous bank-
ng crisis. As reported in Table C.6, the risk parameters basedn the twenty-year-moving average are fairly stable. Several PDsctually fall during the simulation period as the previous bankingrisis is given lower weight. Thus, in 2011 the TTC Basel II risk-
s

ci

Stability 7 (2011) 138–154 147

eighted assets, based on the twenty-year-moving average, arenly 5% higher than the Basel I risk-weighted assets, see Table C.7.he TTC Basel II risk-weighted assets, based on the twenty-year-oving average, do not exhibit any significant cyclicality during

he simulation period.The removal of cyclicality can be explained as follows. First, with

data set covering a sufficiently long observation period the volatil-ty in capital requirements is contained as new observations areiven low weight. Second, when the risk parameters are estimatedn a data set covering economic downturns, the increase in capi-al requirements during economic downturns is curbed as the riskarameters are already high.

Our analysis demonstrates that the shorter the time seriespplied in banks’ risk models are, the more the capital require-ent swings with the business cycle. The impact can be substantial.ccording to Basel II, internal estimates of PD must be basedn a historical observation period covering at least five years.he LGD and EAD estimates should reflect economic downturnonditions and the observation period must be at least sevenears. Some countries permit the use of considerably shorter timeeries than the optimal through-the-cycle coverage. In addition,he available time series tend to not include data from reces-ions.

In sum, we find that the observation period should be longerhan ten years, and cover a recession in order to contain the cycli-ality in the Basel II capital requirements. Thus, the differenceetween the time series employed and the ideal time series shoulde taken into account when the Basel Committee calibrate the newet of standards to reduce the procyclicality of capital require-ents.

. Conclusion

A growing body of literature has assessed the potential cycli-ality of Basel II. However, only parts of banks’ assets haveeen considered in each study. In addition, the cyclicality of theapital positions is usually not taken into account. This papernalyses the cyclicality of capital positions and the cyclicality ofasel II capital requirements for the entire asset portfolio of sixorwegian banks. In addition, we apply a more comprehensiveodelling approach and a more detailed data set than commonly

pplied in the literature. A bank model developed for stress-testingurposes is combined with bankruptcy probabilities from an enter-rise model, from a logit model for banks and from a householdodel.Our analysis provides some new evidence on the cyclical nature

f Basel II. Based on a recession scenario for the Norwegian econ-my, we document a substantial increase in the calculated BaselI capital weightings even when the banks employ a through-he-cycle rating system where the risk parameters are based onten-year observation period. In our stress scenario, bank capi-

al also deteriorates as the banks record high losses on loans andecurities.

The added cyclical pressure on bank capital positions causedy Basel II may be of a larger magnitude than the pre-existingffect under Basel I. However, the cyclicality of their Basel II capitalequirements can be effectively contained if the historical obser-ation period is sufficiently long and includes economic downturnonditions. Basel II risk-weighted assets, based on the twenty-year-oving average, do not exhibit any significant cyclicality during our

imulation period.The Basel Committee will by the end of 2010 develop a revised

apital adequacy framework in order to reduce the procyclical-ty in the banking sector. Our analysis of the cyclicality of capital

1 ncial

piisnb

cmhiihqrb

A

ScaB

A

iimbit

lfiaItc

�

�

N

�

E

�

�

�

�

�

�

�


ositions and Basel II capital requirements can provide usefulndications on how the capital adequacy framework should be cal-brated in order to mitigate procyclicality. The shorter the timeeries applied in banks’ risk models are, the stronger the measureseeded to reduce the procyclicality of capital requirements wille.

An important area for further work is to develop a bank modelovering the most important aspects of bank behaviour. The bankodel employed in our analysis does not enable us to evaluate

ow banks may respond to the macro scenarios. Banks may fornstance react to a tighter capital constraint by cutting off credit tots riskier borrowers, see e.g. Jokipii and Milne (2010). On the otherand, banks may not be able to attract borrowers of high credituality when the economy is in a downturn. For a more completeepresentation we would also need to model contagion betweenanks.

cknowledgements

The author is indebted to an anonymous referee, Farooq Akram,igbjørn Atle Berg, Bent Vale and Eivind Bernhardsen for usefulomments. The views expressed in this paper are those of theuthor and should not be interpreted as reflecting those of Norgesank (the Central Bank of Norway).

ppendix A. The macro model

The core macro model is an extension of the model reportedn Bårdsen and Nymoen (2008) and Bårdsen et al. (2005).22 Its a macro-econometric model estimated on quarterly data. The

odel explicitly takes into account several channels of interplayetween output, inflation and financial stability. The equations are

n equilibrium-correction form, with backward-looking expecta-ions formation.

We present a stylised version of the model in Eqs. (1)–(13). Smalletters denote natural logarithms of the variable, � denotes therst difference operator, �j denotes the j-period difference oper-tor, and foreign variables are denoted with starred superscripts.n general, intercept terms and seasonal effects have been omit-ed from the equations for ease of exposition. The identities thatomplete the model are not reported.

Aggregate demand

yt = −0.6�yt−1 + 0.7�gt + 0.4�gt−1 + 0.1�(ph − p)t−1

+ 0.1�(cre − p)t−1 + 0.2�(crh − p)t−3

− 0.3[(yt−2 − 0.8gt−2 − 0.1(v + p∗ − p)t−1 − 0.1(crh − p)t−4

+ 0.01(RL − �)t−1] Estimation period 1991Q1–2006Q4

(1)

Exchange rate

vt = ϕ(−0.04�Rt + 0.05�R∗t − 0.1�pot − 0.07ut−1)

− 0.1[(v + p∗ − p)t−1 + 0.03((R − �)t−1 − (R∗ − �∗)t−1)

+ 0.1(po + usd − p)t−1 − �v]

Estimation period 1994Q2–2007Q2 (2)

22 The presentation of the core part of the macro model is based on Bårdsen andymoen (2008).

�

Stability 7 (2011) 138–154

Import prices

pit = 0.4�vt+1.3�pi∗t −0.4[(pi − pi∗ − v)t−1 − 0.6(p − p∗ − v)t−1]

stimation period 1990Q1–2007Q2 (3)

Unemployment

ut = 0.4ut−1 − 1.6

(�4

12

2∑j=1

yt−j − mean

(�4

12

2∑j=1

yt−j

))

− 0.03[ut−2 − 11.1�(w − p)t ] Estimation period 1979Q3–2007Q4 (4)

Wages

wt = �zt − 0.5�(wt−1 − zt−1) − 0.1[wt−2 − pt−1 − zt−2

+ 0.001ut−1 − �w] Estimation period 1978Q4–2007Q4

(5)

Consumer prices

pt = 0.3�pt−2 + 0.1�yt−1 + 0.1�(wt−1 − zt−1) + 0.1�pet

− 0.06[pt−3 − 0.65(wt−3 − zt−2) − 0.35pit−1 − �p]


Money market interest rate

Rt = 1.5(�ct − 2.5) − 0.6(Rt−1 − R∗

t−1 − 1) + 0.4�R∗t − 0.5

×

⎛⎝1

4

4∑j=1

ut−j − 2

⎞⎠ Estimation period 1991Q1–2007Q2

(7)

Banks’ lending rate

RLt = 0.8�Rt + 0.2�Rt−1 − 0.35[RLt−1 − (Rt−1 + RLM)] (8)

Household debt

(crh − pt) = −0.01(�RLt−2 + �RLt−3) + 0.3�(inc − p)t−1

+ 0.1(�(ph − p)t − �(ph − p)t−3) − 0.04[(crh − p)t−1

− 0.7(ph − p)t−4 + 0.04RLt−4 − 1.2(inc − p)t−2]


House prices

pht = 0.2�inct − 0.03�RLt − 0.02�RLt−1 + 0.03Het

−0.1[pht−1+0.05RLt−1 + 0.5ut−1.3(inc − hs)t−1−0.3crht−1]


Housing investments

j = −0.04�

(RL − 1

1∑�c

)− 0.01

(RL − 1

1∑�c

)
t 4 3
j=−1

t−j

t

3j=−1

t−j

t−4

− 0.1[(jt−1−hst−10)−(ph − p)t−4 − (inc − p)t−1 − (pj − p)t−4]


ncial

�

�

w1cf

Agpahscrsoto

beicfaeertr

nphci

set

int

(i

a(rpc(etwtup

mmritduort

rite

E(gw(

JphNifdtd

oh

i

irprice (pj − p), households’ real wage income (inc − p) as a proxy for

land costs, and the real lending rate(

RL − (1/3)∑1

j=−1�ct−j

)t−4

.

23 The constant mark-up term is suppressed. In the full econometric model,productivity (z) is an endogenous variable that depends on real wages (w − p),


Household default rate

(dh − crh) = −0.2�3(dh − crh)t−1 + 0.02�2(RL − �)t

+ 0.02�2(RL − �)t−2 − 0.5�4(ph − p)t

− 0.2[(dh − crh)t−4 − 0.4ut−3 − 0.08(RL − �)t−4

+ 1.2(inc − p)t−1 + 1.2(ph − p)t−4]


Firm default

(de − p)t = −0.3�2(de − p)t−1 + 0.02�2(RL − �)t + 0.9�ut

+ 0.7�ut−1 + 1.5�(cre − p)t−3 − 0.4�(po + usd − p)t

− 0.5[(de − p)t−3 − (cre − p)t−4 − 0.05(RL − �)t−3

− 1.7ut−2 + 0.7(v + p∗ − p)t−3 + 0.5(po + usd − p)t]


here � = 100(�4Pt/Pt−4) is the inflation rate; �u =00(�4Pu

t /Put−4) is the core inflation rate, i.e. inflation adjusted for

hanges in energy prices and taxes; �∗ = 100(�4P∗t /P∗

t−4) is theoreign inflation rate.

Growth in real aggregate demand (�y) is modelled in Eq. (1).ggregate demand is affected by the real interest rate (RL − �), realovernment expenditure (g) and the real exchange rate (v + p∗ −). Thus, a change in the nominal exchange rate would directlyffect aggregate demand. Aggregate demand is also affected byouse prices and credit. Changes in real house prices (ph − p) havehort-run effects on aggregate demand through a wealth effect ononsumption and through housing investments not captured by theeal interest rate. Real corporate credit (cre − p) affects GDP in thehort run, while real household credit (crh − p) has long-run effectsn GDP. The short-run effect is interpreted as reflecting frictions inhe credit market, while the long-run effect points towards a formf rationing of the household sector.

The exchange rate (in logs denoted v) expresses the num-er of domestic currency units per unit of foreign currency. Thequation of growth of the nominal effective exchange rate (�v)n Eq. (2) reacts to deviations from PPP (v + p∗ − p) and henceontributes to stabilising the real exchange rate. ϕ is a dummyor inflation targeting, and takes the value 0 up until 2001Q1nd the value 1 from 2001Q2. In the long run, the nominalxchange rate reflects the difference between domestic and for-ign prices and the difference between domestic and foreigneal interest rates (R − �) − (R∗ − �∗). Accordingly, domestic infla-ion is fully reflected in the nominal exchange rate in the longun.

Import prices measured in domestic currency (pi) are a homoge-ous function of the nominal exchange rate (v) and foreignroducer prices measured in foreign currency (pi*). On the otherand, import prices increase if the real exchange rate (in terms ofonsumer prices) appreciates. This is due to pricing-to-markets inmport price setting.

The unemployment rate (u) follows output growth (�y) in thehort run as an Okun’s law relationship, see Eq. (4). In addition, itxhibits slow reversion towards its equilibrium rate; an intercepterm has been omitted.

There is a pass-through of consumer price inflation (�p) to nom-nal wage growth (�w) in the short run; see Eq. (5). In each period,ominal wages adjust towards their long-run relationship wherehere is a full pass-through of consumer prices and productivity

u

flMe

Stability 7 (2011) 138–154 149

z). However, the mark-up of wages on prices and productivity isnversely related to the unemployment rate (u).23

In the short run, consumer price inflation varies with changes inggregate demand (�y) and to some extent nominal wage growth�w); see Eq. (6). In addition, it adjusts to deviation from the long-un relationship for consumer prices. In the long run, consumerrices (p) reflect a weighted average of domestic and importedosts, represented by unit labour costs (w − z) and import pricesv + p∗). It follows that the initial effect of a change in nominalxchange rate on aggregate demand would become modified overime due to the exchange rate pass-through to inflation, whichould have an effect opposite that of the nominal exchange rate on

he real exchange rate. The model also includes an equation for thenderlying, i.e. core, inflation rate (pc), which is linked to consumerrice inflation.

The three-month money market interest rate (R) follows an esti-ated Taylor-type rule in Eq. (7). Since March 2001, Norwegianonetary policy is aimed at targeting the annual core inflation

ate (�c) at 2.5%. Despite the fact that Norwegian monetary pol-cy has changed over time24 the estimated equation is stable overhe estimation period 1991–2006. The interest rate responds toeviation from target in domestic core inflation and to deviation innemployment from 2%. This unemployment gap represents theutput gap. If the interest rate deviates from the foreign interestate including a premium of 1 percentage point, this also affectshe interest rate.

Banks’ lending rate (RL) is defined to follow the money marketate. A lending margin (RLM), i.e. the margin between the lend-ng rate and the money market rate, is an exogenous variable inhe model. The coefficients of this equation are calibrated and notstimated.

The relationship explaining movements in household debt inquation (9) builds on the work presented in Jacobsen and Naug2004). Growth in household debt (�crh) reacts positively torowth in income (�inc) and housing prices (�ph), and decreasesith higher interest rate on loans (RL); see Jacobsen and Naug

2004) for further details.The model of house prices (ph) in Equation (10) is based on

acobsen and Naug (2005). The growth rate of nominal houserices (�ph) is explained by growth in nominal income (inc) andousehold expectations about their own financial situation and theorwegian economy (He), i.e. a survey-based consumer confidence

ndicator, as well as interest rate changes (�RL) and deviationsrom steady state. In steady state, house prices (ph) are mainlyetermined by income (inc) and housing capital (hc) in addition tohe interest rate (RL), the unemployment rate (u), and householdebt (crh).

The equation for gross fixed housing investment (j) is basedn Jacobsen et al. (2007), see Eq. (11). Growth in gross fixedousing investment (�j) depends on the change in the real lend-

ng rate �4

(RL − (1/3)

∑1j=−1�c

)t. In steady state, gross fixed

nvestment depends on the level of housing capital (hs) due toeplacement investment, real house prices (ph − p), real investment

nemployment (u) and a deterministic trend.24 At the very beginning of the sample, NOK was pegged to the ECU, but wentoating in December 1992. Although inflation targeting was formally introduced inarch 2001, it is a common view that this regime was gradually introduced from

arly 1999.


he ban

(dbp(irtdrpa

ia

A

tc

rdAplitqatlva

o

B

i

P

wi(

bf

N

wvaafe(raitbwhthl

Chart 8. T

The equations of default25 by households and firms in (12) and13) respectively are based on Berge and Boye (2007). Households’efault rate (dh − crh), i.e., default as a share of total householdank debt, depends on households’ real income (inc − p), unem-loyment (u), the real interest rate (RL − �) and real house pricesph − p). With respect to firms’ default, there is not homogene-ty between default and debt in the short run, only in the longun. Firms’ default, measured in real terms (de − p), depends onhe level of debt (cre − p), the real interest rate (RL − �), domesticemand proxied by the unemployment rate (u), the real exchangeate (v + p∗ − p) as a measure of competitiveness and the real oilrice (po + usd − p). The latter variable captures that the level ofctivity and investments in the oil sector affect other industries.

In addition, SMM includes estimated equations for bankruptciesn firms adapted from Jacobsen and Kloster (2005), productivity (z),nd bond rates (RB).

ppendix B. The bank model

The bank model is a non-behavioural model and consists ofhree main components: profit and loss account, balance sheet andapital adequacy calculation, see Chart 8.

The profit and loss account and the balance sheet have a recip-ocal effect on each other. Banks’ profit after tax and dividendsirectly affects banks’ equity capital included in the balance sheet.t the same time, net interest income, which is included in therofit and loss account, is determined by the size of the assets and

iabilities in the balance sheet. The balance sheet affects the cap-tal adequacy calculation. Equity capital is an important item inhe calculation of regulatory capital included in the capital ade-uacy calculation. The balance sheet also affects the risk-weightedssets for capital adequacy calculation. The bank model’s projec-ions make use of the output from the macro model on lending,
ending rates and labour costs. The model may be used for indi-idual banks or groups of banks, depending on the purpose of thenalysis. A detailed presentation of the model follows.
25 Our data on problem loans include both default and loans with a high probabilityf default as reported by the banks.

p

o

l

k model.

.1. Banks’ profit and loss account

Banks’ profit after tax and dividends is calculated in the follow-ng manner:

RO = NII + OOI − OOC − LL − T − DIV (2.1)

here profit after tax and dividends (PRO) depends on net interestncome (NII), other operating income (OOI), other operating costsOOC), loan losses (LL), tax (T) and dividends (DIV).

Net interest income (NII) which is included in the calculation ofanks’ profit after tax and dividends (see Eq. (2.1)), is calculated asollows:

II = ((L − 1 + L)/2) · LR+((OA − 1 + OA)/2) · ROA−((D − 1 + D)/2)

· RD − ((OL − 1 + L)/2) · ROL (2.2)

here the subscript −1 indicates that the variable is from the pre-ious year. Net interest income (NII) depends on net lending (L),verage lending rate (LR), other interest-bearing assets (OA), aver-ge interest rate on other interest-bearing assets (ROA), depositsrom customers and other financial institutions (D), average inter-st rate on deposits from customers and other financial institutionsRD), other interest-bearing liabilities (OL)26 and average interestate on other interest-bearing liabilities (ROL). An increase in theverage interest rate on loans and other interest-bearing assetsncreases in isolation net interest income, whereas an increase inhe average interest rate on deposits and other interest-bearing lia-ilities has the opposite effect. In addition, growth in total assetsill increase net interest income if the marginal interest rate isigher on interest-bearing assets than on interest-bearing liabili-ies. Net interest income as a share of banks’ total operating incomeas fallen over the past ten years, although the share was nonethe-

ess 67% at end-2007. It is therefore especially important that the
rojections for this income item are accurate.
Other operating income (OOI) which is included in the calculationf banks’ profit after tax and dividends (see Eq. (2.1)), is calculated

26 Other interest-bearing liabilities include short-term paper, bonds, subordinatedoans and other debt.

ncial

a

O

wd((

or

�

w−dBemfsdlt

crc

cDloe

afit

A

B

bbi

•••

s

•••••

sci

B

taspbtrbr

B

fabp

dlrea

(mwi

rirt


s follows:

OI = FEE + SD + NGS + NGFE + NGD + OGI (2.3)

here other operating income is the sum of fee income (FEE), shareividends (SD), gains/losses on securities (NGS), foreign exchangeNGFE) and derivatives (NGD), as well as other gains and incomeOGI).

Fee income (FEE), which is included in the calculation of otherperating income (see Eq. (2.3)), is calculated using an error cor-ection model estimated on quarterly data:

ln(FEE) = −4.82 − 0.37 ln(FEE − 1) + 0.60 ln(GDP − 1)

+ 1.51(R5Y−R3M)−1−0.003�FORB+0.94� ln(GDP)

+ 0.23� ln(FEE − 4) + 0.02S1 + 0.01S2 − 0.001S3

(2.4)

here ln designates the logarithm of the variable, the subscript1 indicates that the variable is from the previous quarter and �esignates the first difference, i.e. � ln(FEE) = ln(FEE) − ln(FEE − 1).anks’ fee income is estimated using GDP (GDP), the yield differ-ntial between 5-year government bonds (R5Y) and three-monthoney market rates (R3M), and the market share of branches of

oreign banks (FORB). The equation also contains the effects of sea-onal variations (S(i)). See Andersen et al. (2008) for a more detailedescription of the equation. At the end of 2007, fee income was the

argest item in other operating income, accounting for 55% of theotal.

Other operating costs (OOC) include labour costs, commissionosts, electronic data processing costs and other costs. Labour costsepresented the largest item, accounting for 55% of other operatingosts in 2007.

Loan losses (LL) have been low in recent years, but may increaseonsiderably if the financial position of borrowers deteriorates.uring the Norwegian banking crisis in the period 1988–1993, loan

osses were very high, and in 1991 loan losses were higher thanther operating costs. Projected loan losses are derived from thenterprise model, the household model and the LGD model.

Tax (T) is set at 28% of profit before tax and dividends. The aver-ge share of tax that banks have charged to income during the pastve years is slightly less than 28%, but the share varies from yearo year.

Banks’ dividends (DIV) are assumed to be 50% of after-tax profit.dividend is not paid if banks record a negative profit.

.2. Banks’ balance sheets

A bank’s balance sheet consists of an asset column listing theank’s assets and a liability column indicating how the assets haveeen funded. The assets in the balance sheet comprise the following

tems:

loans to households and enterprisessecurities and deposits in other financial institutionsother assets

A

Stability 7 (2011) 138–154 151

Loans, accounting for 67% of banks’ total assets in 2007, repre-ented the dominant asset on the balance sheet.

Liabilities comprise the following items:

customer depositsmarket fundingother debtsubordinated loansequity capital

Deposits, accounting for 62% of banks’ total liabilities, repre-ented the dominant liability on the balance sheet. Market fundingomprises bonds, short-term paper and loans from other financialnstitutions.

.3. Banks’ capital adequacy

The calculation of banks’ capital adequacy is based on projec-ions of regulatory capital and the risk-weighted assets for capitaldequacy. Banks’ regulatory capital is the sum of Tier 1 capital andubordinated debt. It is impossible, however, to identify all com-onents of Tier 1 capital. Therefore, a residual, i.e. the differenceetween the last reported figure for Tier 1 capital and the sum ofhe Tier 1 capital items in the balance sheet, is also projected. Theisk-weighted assets for capital adequacy are calculated separatelyased on projections for the assets portfolio and the correspondingisk parameters.

.4. Risk factors that can be analysed using the bank model

The bank model can be used to analyse the effect of several dif-erent risk factors on banks’ profit and financial strength. Credit riskffects banks’ loan losses (LL). Credit risk is therefore included in theank model through the effect of estimated loan losses on banks’rofit (PRO) and capital adequacy.

Liquidity risk affects banks’ funding costs. In the bank model, theeposit rate (DR) and the interest rate on other interest-bearing

iabilities (ROL) can be adjusted as a result of changes in liquidityisk. Changes in the interest rate that banks pay for funding affectsstimated net interest income (NII) as well as banks’ profit (PRO)nd capital adequacy.

Market risk affects dividends (DIV) and gains/losses on securitiesNGS), foreign exchange (NGFE) and derivatives (NGD). Changes in

arket risk are thus important to other operating income (OOI),hich is included in the calculation of banks’ profit (PRO) and cap-

tal adequacy.Banks can incur considerable losses as a result of operational

isk. In the bank model, costs resulting from operational risk arencluded in other operating costs (OOC). Losses due to operationalisk will thus lead to higher costs, lower profits (PRO) and a reduc-ion in capital adequacy in the bank model.

ppendix C.

See Tables C.1–C.7.


Table C.1PIT risk parameters, 2008–2011.

2008 2009 2010 2011

Exposure PD LGD PD LGD PD LGD PD LGD

Bank 0.0195 0.406 0.0952 0.737 0.1412 0.827 0.2358 0.720Sovereign 0.0013 0.406 0.0013 0.737 0.0013 0.827 0.0013 0.720Retail Mortgages 0.0120 0.262 0.0114 0.627 0.0085 0.7’44 0.0085 0.657Qualifying Revolving Retail 0.0369 0.558 0.0350 0.852 0.0262 0.915 0.0172 0.787

EnterprisesPrimary industries 0.0292 0.406 0.0569 0.737 0.1452 0.827 0.2549 0.720Oil and gas 0.0134 0.406 0.0217 0.737 0.0516 0.827 0.0910 0.720Mining 0.0108 0.406 0.0100 0.737 0.0167 0.827 0.0194 0.720Construction 0.0148 0.406 0.0227 0.737 0.0516 0.827 0.0871 0.720Retail trade, hotel and restaurant 0.0233 0.406 0.0298 0.737 0.0639 0.827 0.0993 0.720Shipping abroad 0.0123 0.406 0.0198 0.737 0.0471 0.827 0.0831 0.720Other transport 0.0256 0.406 0.0381 0.737 0.0829 0.827 0.1367 0.720Property management 0.0129 0.406 0.0180 0.737 0.0237 0.827 0.0132 0.720Commercial services 0.0135 0.406 0.0132 0.737 0.0344 0.827 0.0510 0.720

Equities 350% Risk weight 350% Risk weight 350% Risk weight 350% Risk weight

Table C.2TTC risk parameters. Five-year-moving average PDs and LGDs, 2008–2011.

2008 2009 2010 2011


Bank 0.0086 0.361 0.0274 0.438 0.0553 0.534 0.1011 0.608Sovereign 0.0013 0.361 0.0013 0.438 0.0013 0.534 0.0013 0.608Retail Mortgages 0.0120 0.181 0.0143 0.274 0.0154 0.391 0.0154 0.490Qualifying Revolving Retail 0.0369 0.552 0.0439 0.612 0.0472 0.685 0.0473 0.732



Table C.3TTC risk parameters. Ten-year-moving average PDs and LGDs, 2008–2011.

2008 2009 2010 2011






Table C.4Risk weighted assets for the five largest Norwegian banks and Nordea Bank Norge based on five-year-moving average PDs and LGDs. Millions of NOK, 2008–2011.

September 2008 2008 2009 2010 2011


Total Basel II risk-weighted assets 1295 1360 1756 2393 3067Basel 1 risk-weighted assets 1479 1461 1412 1383 1377

Table C.5Risk weighted assets for the five largest Norwegian banks and Nordea Bank Norge based on ten-year-moving average PDs and LGDs. Millions of NOK, 2008–2011.

September 2008 2008 2009 2010 2011


Total Basel II risk-weighted assets 1295 1335 1507 1809 2129Basel 1 risk-weighted assets 1479 1461 1412 1383 1377

Table C.6TTC risk parameters. Twenty-year-moving average PDs and LGDs, 2008–2011.

2008 2009 2010 2011






Table C.7Risk weighted assets for the five largest Norwegian banks and Nordea Bank Norge based on twenty-year-moving average PDs and LGDs. Millions of NOK, 2008–2011.

September 2008 2008 2009 2010 2011


R

A

A

A

A

A

A

B

B

B

B

B

B

B

B

B

B

C

D

F

G

G

J

J

J

J

J

J

J

K

M

L

N

N

P

S

S

T

V

Total Basel II risk-weighted assets 1295Basel 1 risk-weighted assets 1479

eferences

ltman, E., Brady, B., Resti, A., Sironi, A., 2005. The link between default and recoveryrates: implications for credit risk models and procyclicality. Journal of Business78 (6), 2203–2227.

llen, L., Saunders, A., 2003. A survey of cyclical effects in the credit risk measure-ment models. BIS Working Paper, No. 126.

ndersen, H., 2008. Failure prediction of Norwegian banks: a logit approach. Work-ing Paper 2/2008, Norges Bank.

ndersen, H., Berge, T.O., 2008. Stress testing of banks’ profit and capital adequacy.Economic Bulletin 2/2008, Norges Bank, pp. 47–57.

ndersen, H., Berg, S.A., Jansen, E.S., 2008. The dynamics of operating income in theNorwegian banking sector. Working Paper 13/2008, Norges Bank.

yuso, J., Pérez, D., Saurina, J., 2004. Are capital buffers pro-cyclical? Evi-dence from Spanish Panel Data. Journal of Financial Intermediation 13, 249–264.

asel Committee on Banking Supervision, 2006a. Results of the fifth quantitativeimpact study (QIS 5). BIS.

asel Committee on Banking Supervision, 2006b. International Convergence of Cap-ital Measurement and Capital Standards: A Revised Framework. BIS.

asel Committee on Banking Supervision, 2009. Strengthening the resilience of thebanking sector. BIS.

enford, B., Nier, E., 2007. Monitoring cyclicality of Basel II capital requirements.Financial Stability Paper No. 3, December.

erge, T., Boye, K.G., 2007. Explaining the development in default by households andfirms. Economic Bulletin 1, Norges Bank.

ernhardsen, E., Larsen, K., 2007. Modelling credit risk in the enterprise sector -further development of the SEBRA model. Economic Bulletin 3, Norges Bank,pp. 102–108.

ikker, J., Metzemakers, P., 2004. Is bank capital procyclical? A Cross-Country Anal-ysis. De Nederlandsche Bank Working Paper 009/2004.

orio, C., Furfine, C., Lowe, P., 2001. Procyclicality of the financial system and finan-cial stability: issues and policy options. BIS Paper, No. 1.

årdsen, G., Eitrheim, Ø., Jansen, E.S., Nymoen, R., 2005. The Econometrics of Macroe-conomic Modelling. Oxford University Press, Oxford.

årdsen, G., Nymoen, R., 2008. Macroeconometric Modeling for Policy. Handbookof Econometrics, vol II, Applied Econometrics. Palgrave-MacMillan.

atarineu-Rabell, E., Jackson, P., Tsomocos, D.P., 2003. Procyclicality and the newBasel accord – banks’ choice of loan rating system. Bank of England WorkingPaper, no. 181.

ierick, F., Pires, F., Scheicher, M., Spitzer, K.G., 2005. The new Basel Capital Frame-
work and its implementation in the European Union. ECB Occasional Paperseries, No. 42, December.
rancis, W., Osborne, M., 2009. Bank regulation, capital and credit supply: Measuringthe impact of Prudential Standards. UK Financial Services Authority OccasionalPaper series, No. 36, September.

V

1326 1312 1397 14421461 1412 1383 1377

oodhart, F., Hofmann, B., Segoviano, M., 2004. Bank regulation and macroeconomicfluctuations. Oxford Review of Economic Policy 20 (4), 591–615.

ordy, M.G., Howells, B., 2006. Pro-cyclicality in Basel II: can we treat the diseasewithout killing the patient. Journal of Financial Intermediation 15 (3), 395–417.

acobsen, D.H., Kloster, T.B., 2005. What influences the number of bankruptcies?Economic Bulletin 4/2005, Norges Bank, pp. 191-201.

acobsen, D.H., Naug, B.E., 2004. What influences the growth of household debt?Economic Bulletin 3/2004, Norges Bank, pp. 103-111.

acobsen, D.H., Naug, B.E., 2005. What drives house prices? Economic Bulletin1/2005, Norges Bank, pp. 29-41.

acobsen, D.H., Solberg-Johansen, K., Haugland, K., 2007. Housing investments andhouse prices. Economic Bulletin 1/2007. Norges Bank, pp. 33-46.

akubik, P., Schmieder, C., 2008. Stress testing credit risk: comparison of the CzechRepublic and Germany. FSI Award 2008 Winning Paper, Bank for InternationalSettlements.

okipii, T., Milne, A., 2010. Bank capital buffer and risk adjustment decisions. Journalof Financial Stability.

okivuolle, E, Virolainen, K., Vähämaa, O., 2008. Macro-model-based stress testingof Basel II capital requirements. Discussion Papers 17/2008, Bank of Finland.

ashyap, A.K., Stein, J.C., 2004. Cyclical implications of the Basel II standards. Eco-nomic Perspectives, Federal Reserve Bank of Chicago, Quarter 1, pp. 18–31.

arcelo, A., Scheicher, M., 2005. Capital requirements for corporate default risk. ECBMimeo.

indquist, K., 2004. Banks’ buffer capital: how important is risk? Journal of Interna-tional Money and Finance 23 (3), 493–513.

ordal. K.B., Syed, H., 2010. A model for predicting aggregated corporate credit risk.Penger og Kreditt 1/10, Norges Bank.

ier, E., Zicchino, L., 2005. Bank weakness and bank loan supply. Financial StabilityReview, December, pp 85–93.

eura, S., Jokivuolle, E., 2004. Simulation based stress tests of banks’ regulatorycapital adequacy. Journal of Banking and Finance 28, 1801–1824.

aurina, J., Trucharte, C., 2006. An assessment of Basel II procyclicality in mortgageportfolios. Presented at the Federal Deposit Insurance Corporation’s 6th AnnualBank Research Conference.

toltz, S., Wedow, M., in press. Banks’ regulatory capital buffer and the businesscycle: evidence for Germany. Journal of Financial Stability.

aylor, A., Goodhart, C., 2006. Procyclicality and volatility in the financial system:the implementation of Basel II and IAS 39. In: Gerlach, S., Gruenwald, P. (Eds.),Procyclicality of Financial Systems in Asia. IMF and Hong Kong Institute forMonetary Research, Palgrave MacMillan, New York.

atne, B.H., 2007. Financial margins in Norwegian households – an analysis of micro
data for the period 1987–2003. IFC Bulletin No. 26, pp. 40–51, Bank for Interna-tional Settlements.
atne, B.H., 2006. How large are the financial margin of Norwegian Household? Ananalysis of micro data for the period 1987–2004. Economic Bulletin 4, NorgesBank, pp. 173–180.

procyclical implications of basel ii: can the cyclicality of capital requirements be contained?

Documents