economic bulletin 35 bank of greece

YELLOW

BLACK

MAGENTA

CYAN

ECONOMICBULLETIN

No 35

JUNE2011

ISSN: 1105 - 9729

BA

NK

OF

GR

EEC

E

EUR

OSY

STE

ME

CO

NO

MIC

BU

LL

ET

IN N

o 3

5JU

NE

20

11

0-ESOFYLO: 1 05-06-12 08:17 1

ECONOMICBULLETIN

No 35

JUNE2011

0-ESOFYLO: 1 05-06-12 08:17 2

ECONOMICBULLETIN

No 35

JUNE2011

0-ESOFYLO: 1 05-06-12 08:17 3

BBAANNKK OOFF GGRREEEECCEE21, E. Venizelos Avenue102 50 Athens

www.bankofgreece.gr

Economic Research Department - SecretariatTel. ++30 210 320 2393Fax ++30 210 323 3025

Printed in Athens, Greeceat the Bank of Greece Printing Works

ISSN 1105 - 9729

0-ESOFYLO: 1 05-06-12 08:17 4

CONT EN T S NUMERICAL FISCAL RULES IN PRACTICEVasilios Manesiotis 7

TAX EVASION IN GREECE: AN OVERVIEWMelina Vasardani 15

THE USEFULNESS OF STRESS TESTINGEXERCISES FOR ASSESSING THE SOUNDNESSOF THE BANKING SYSTEMFaidon Kalfaoglou 25

ON-THE-JOB TRAINING IN GREECE: A BRIEF OVERVIEWDaphne Nicolitsas 47

WORKING PAPERS(July 2010 March 2011) 77

ARTICLES PUBLISHED IN PREVIOUS ISSUESOF THE ECONOMIC BULLETIN 87

0-ESOFYLO: 1 05-06-12 08:17 5

35Economic BulletinJune 20116

0-ESOFYLO: 1 05-06-12 08:17 6

NUMER I C A L F I S C A L RU L E S I N P R A C T I C E *

Vasilios Manesiotis Economic Research Department

35Economic Bulletin

June 2011

The international financial crisis caused aserious deterioration in the fiscal position ofalmost all advanced economies and a debt cri-sis in peripheral countries of the euro area.In an effort to support their financial systems,developed countries took up hefty liabilities,which led to a large increase in annual deficitsand the fast accumulation of debt to unprece-dented levels in periods of peace. The fiscalposition of several developed countries bothinside and outside the European Unionremains precarious, and achieving a soundand sustainable fiscal position is a crucialchallenge facing the countries in question.Against this background, there has been arenewed focus on strengthening the institu-tional framework of budgetary policy andadopting numerical fiscal rules for the gov-ernance of public finances.1

Obviously, the Greek sovereign debt crisis didnot result from the credit support measures;rather, it was caused by the re-evaluation ofcredit risk by markets following the globalfinancial crisis, combined with Greeceschronic fiscal imbalances, which in the pastthree decades led to the accumulation of veryhigh government debt. Imbalances resulted,among other things, from the complete lackof even a rudimentary budgetary framework,which would ensure fiscal sustainability inGreece.2

According to a survey by the InternationalMonetary Fund (IMF), in 1990 fiscal ruleswere used in only seven countries. In 2009, 80countries had fiscal rules in place, eithernational or supranational, including EU Mem-ber States (i.e. 77 out of 80 countries hadnational fiscal rules). According to the survey,there is a clear upward trend in the use ofnumerical fiscal rules and the establishment ofindependent bodies responsible for assessingthe preparation and execution of budgets, aswell as the overall fiscal policy.3

Since the 1990s, the Bank of Greece has repeat-edly recommended4 the introduction ofnational fiscal rules,5 the establishment of anindependent authority to review the govern-ment budget and the fiscal policy pursued, aswell as the adoption of budgeting methods, pro-cedures and practices that help contain publicspending and ensure fiscal sustainability. Todaymore than ever, it is imperative to adopt suchmeasures aimed at strengthening the institu-tional fiscal framework.6 The introduction ofnational fiscal rules and other measures inGreece should boost fiscal discipline and deficitreduction and could help the country regain its

7

** The views expressed in this article do not necessarily reflect thoseof the Bank of Greece. The author would like to thank the LegalDepartment of the Bank of Greece for the information providedabout constitutional provisions in European Union countries, aswell as Heather Gibson for her constructive comments. The authorassumes full responsibility for any errors or omissions.

11 For example, the OECDs recent country surveys (OECD, Eco-nomic Surveys: Germany 2010, Finland 2010, Norway 2010, CzechRepublic 2010, Austria 2010, etc.), as well as Economic Outlookreports, (OECD, 2010, pp. 256-260), present and analyse numer-ical fiscal rules and other institutions (e.g. independent fiscalcouncils) applied in many countries. The International Mone-tary Fund has also shown renewed interest on the subject. See e.g.Cottarelli et al. (2009); and Garcia et al. (2011). Lastly, the pro-posals of the European Commission and the Van Rompuy TaskForce require [...] a commitment on the part of the euro areacountries to swiftly enhance their national budgetary frame-works. See the article entitled The reform of economic gover-nance in the euro area essential elements, ECB, Monthly Bul-letin, March 2011.

22 According to a relevant European Commission survey, Greece isone of the three European Union countries that had not establisheda national numerical fiscal rule up until 2010. Law 3871/2010 pro-vides for the introduction of numerical limits for spending, but therelevant provision has not been put into effect yet. Moreover, inGreece there is no independent body to assess the budget, nor arethere any practices and procedures to govern budgeting, in orderto contain spending. See European Commission (2009), p. 88. Seealso Bank of Greece (2007), pp. 213-216.

33 Cottarelli et al. (2009), pp. 7-14.44 In the past few years, the Bank of Greece has repeatedly under-

lined the need to adopt such rules in Greece. See (i) AnnualReport 1999, p. 31; (ii) Annual Report 2001, pp. 52-53; (iii) AnnualReport 2003, pp. 70-71, for a detailed discussion on numerical fis-cal rules and certain preconditions in order to contain expenditureand fiscal deficits; (iv) Annual Report 2004, p. 65; (v) AnnualReport 2006, pp. 213-216; (vi) Monetary Policy Interim Report2008, p. 19; (vii) Monetary Policy 2008-2009; (vii) Monetary Pol-icy 2009-2010; and (viii) Monetary Policy 2010-2011.

55 I.e. fiscal rules that will complement and reinforce the respectiveprovisions of the Maastricht Treaty on deficit and debt. Accord-ing to a recent European Commission document, effectiveapplication of the EMU budgetary framework can not be expectedto derive only from provisions laid down at EU level. See Euro-pean Commission, ECFIN C4/ 28 October 2010, p. 2.

66 See e.g. Rapanos and Kaplanoglou (2010), pp. 17.

1-Meleti Manesioti: 1 05-06-12 08:17 7


credibility, which is now standing at very lowlevels. In any case, the new Euro Plus Pactprovides for more stringent fiscal rules.

1 RULES ON GOVERNMENT EXPENDITURE

For various reasons, any fiscal adjustmenteffort in Greece should begin with and focuson containing primary expenditure. First, bothinternational and Greek experience7 has shownthat fiscal consolidation can only be sustain-able when it focuses on rationalising primaryexpenditure, while a fiscal adjustment thatrelies on higher taxes or lower interest pay-ments can easily be reversed. This is consistentwith a recent OECD survey, which suggeststhat spending restraint, notably with respect togovernment consumption and transfers, ismore likely to generate lasting fiscal consoli-dation and better economic performance.8

Second, with particular regard to Greece, thepast four decades saw a failure to contain pri-mary expenditure. Even during the major fis-cal adjustment of 1994-1999, primary expen-diture continued to rise.9 As far as the currentadjustment effort is concerned, 2010 was theonly year in which ordinary budget primaryexpenditure fell by 10.9%, compared with oneyear earlier, and for 2011 an increase of 1.9%is projected in the budget.

Third, given the fierce tax competition withinthe European Union, a country like Greece,which has low competitiveness, has limitedscope for continuous increases in the tax bur-den. Besides, the tax measures taken over thepast twelve months seem to have left littleroom for further tax increases. Obviously,every possible effort must be made to curb taxevasion and collect tax arrears. This is neces-sary not only for reducing the deficits, but alsofor restoring tax fairness.

Fourth, tax revenue is much more sensitive tocyclical developments than expenditure.Therefore, maintaining a stable fiscal position(e.g. close to balance) would require frequent

changes to tax rates and/or tax bases, whichhowever is not practical. Instead, changes toexpenditure are easier in practice. As noted bythe OECD, expenditure rules are less affectedby the economic cycle.10

Fifth, a stable tax environment has been shownto minimise distortions and other adverseeffects of taxes on the economy, particularly oneconomic growth.11

Lastly, political authorities across the world areunder constant pressure by various interestgroups to increase certain expenditure cate-gories. Rules limiting a priori the possibility ofincreases in primary spending make it difficultto accommodate such requests, possibly evendiscouraging their submission.

Therefore, it is proposed to establish by law abinding ceiling on primary expenditure. In orderto be effective, this ceiling should apply toalmost all general government primary expen-diture and must have been set well in advanceof the annual budget process. It should imposea constraint on total primary expenditure, ratherthan respective limits on individual expenditurecategories (such limits can be determined at amuch later stage). This constraint should be ofa macroeconomic nature, applying to totalexpenditure and not determining individualspending categories or the budget structure. Anexpenditure ceiling put in place long before thestart of the annual budget process and of the rel-evant negotiations between the Ministry ofFinance and the other ministries would definea priori the context in which the budget will beformulated and should stabilise expectations.

Similar fiscal rules (ceilings) on governmentexpenditure are already in place in severalcountries. In Sweden, an expenditure ceiling

8

77 The large fiscal adjustment (deficit was reduced by more than 10percentage points of GDP) in 1994-1999 was based on higher rev-enue and lower interest payments, while primary expenditure con-tinued to rise. This is why the progress achieved until 2000 wasreversed afterwards. See Manessiotis and Reischauer (2001), p. 142.

88 Guichard et al. (2007), p. 7.99 See footnote 7 above.1100 OECD (2010), p. 258.1111 Kopits (2001).


35Economic Bulletin

June 2011

was introduced in 1997 and has ever since beena key aspect of the countrys institutional fis-cal framework. It is estimated to have been avery effective tool for stabilising the publicfinances.12 Additionally, the medium-termcharacter of the expenditure ceiling has helpedestablish a longer-term approach to publicfinances and brought about a better under-standing of their dynamics.

Swedens primary expenditure ceiling is setthree years ahead for fiscal year t+3 (on arolling basis). Thus, the expenditure ceiling for2013 was set in 2010 (the ceiling for 2012 hadbeen set in 2009, etc.). Finland and the Nether-lands define four-year expenditure ceilings cor-responding to the term of the government.Thus, each new government defines (at thebeginning of its term) the expenditure ceilingto apply during its four years in office.

The case of Sweden is more relevant toGreece, as it implies lower forecasting uncer-tainty (with a horizon of three years instead offour years) and, more importantly, because theceiling is not necessarily set by the governmentthat will be expected to comply with it.

Which categories of expenditure should beincluded in the ceiling?

An effective expenditure ceiling has to be com-prehensive, i.e. it must cover all primary expen-diture of general government, without excep-tions: the more expenditure categories itincludes, the harder it is to circumvent.13 Onlyinterest expenditure may be exempted, as it isnot subject to the discretion of the government.Therefore, almost all countries exclude inter-est payments from the ceiling.14 By contrast,other expenditure (which is often consideredinelastic in Greece) should not be excluded.The inelasticity of expenditure (other thaninterest payments) depends on the time hori-zon. In the medium term, almost all expendi-ture can be contained.

The only controversial category of primaryexpenditure refers to cyclically sensitive items,

such as unemployment benefits (which are,however, less volatile in Greece than in othercountries). Both the inclusion and exemptionof these items have pros and cons. Since thereis a broad consensus in the literature that thereshould be no constraints on the operation ofautomatic stabilisers, expenditure (e.g. forunemployment benefits) can be expected tofluctuate significantly in response to the eco-nomic cycle. The question is thus essentiallythe choice between a countercyclical fiscal pol-icy (involving higher expenditure) and long-term fiscal sustainability.

Lastly, the ceiling should somehow take intoconsideration the impact of inflation on gov-ernment expenditure. An appropriate adjust-ment mechanism should take into account thenature, level and volatility of inflation. In theNetherlands and Finland, the four-year ceilingson primary expenditure are determined (at thebeginning of the period) in real terms, and con-verted into nominal terms each year, depend-ing on inflation forecasts. By contrast, in Swe-den, the three-year ceilings are set in nominalterms.15

Establishing a ceiling in practice16

A realistic expenditure ceiling must take intoconsideration the prevailing fiscal and macro-economic conditions and prospects, as well asthe fiscal targets sought to be achieved. Namely,a clear relationship should be establishedbetween primary expenditure under the ceiling,revenue, and the targeted deficit or debt.

An expenditure ceiling t+n for the year t+n,consistent with a forecast for revenue Et+n forthe same year and a targeted fiscal balance (ora change in debt) Dt+n for the year t+n isexpressed:

9

1122 According to an IMF study, the adoption of numerical fiscal rules bySweden and Finland was the main element behind the fiscal adjust-ment efforts following the fiscal and banking crisis (in these two coun-tries) at the beginning of the 1990s. See Cottarelli et al. (2009), p. 17.

1133 OECD (2010), p. 258. Even the so called tax expenditure shouldbe included under the expenditure ceiling.

1144 Ljungman (2008), p. 7.1155 Ljungman, ibid., pp. 13-15.1166 Ljungman, ibid., pp. 17-19.



t+n=Et+n+t+nDt+n .o.t+n

where t+n is additional revenue from futuretax measures until the year t+n, which are notincluded in the revenue forecast (Et+n) and.o.t+n is expenditure outside the ceiling (..)(Ljungman 2008, pp. 17-19). For instance, ifinterest expenditure (outside the ceiling) is17 billion (year t+n), estimated revenue is60 billion, and expected additional revenuefrom tax reforms is 12 billion, then in orderto achieve a 2 billion surplus, the expenditureceiling should be set at 53 billion as follows:t+n=60+12217

t+n=53 billion

If the target is to achieve a budgetary positionclose to balance, then:

t+n=60+12017=55 billion

Obviously, the more accurate the revenue fore-casts, the more effective the rule.

Alternatively, the expenditure ceiling could beset in terms of a constant annual rate of change,say p. Assuming that GDP also grows at a con-stant rate g, then the p/g ratio will also be con-stant. However, expenditure as a percentage ofGDP would decline continuously if pg (if p=g, thenG/GDP is constant). The variable p may beadjustable and adaptable (according to the t+nrule), depending on GDP growth forecasts (g),so that expenditure as a percentage of GDPmay decline (alternatively, it might be possibleto use the growth rate of potential GDP). Thisrule has two desirable characteristics: it is verysimple and it facilitates forecasting the futurelevel of government expenditure.

2 RULES ON GOVERNMENT DEBT

Rules on government debt may take differentforms, and sometimes have an indirect, albeitsignificant, effect on the debt-to-GDP ratio.

According to the European Commission, in2009 thirteen EU Member States had estab-lished rules on government debt, which variedconsiderably17 across countries. In any case, theprobability to stabilise the debt-to-GDP ratiorises when fiscal consolidation focuses on cut-ting government expenditure.18

Due to its debt problems, Greece shouldimmediately adopt a rule to reduce19 the debt-to-GDP ratio to sustainable levels (e.g. 50-60%of GDP), as well as a debt brake to containthe debt ratio at such reduced level. Further-more, according to the Euro Plus Pact, allcountries are required to establish a nationaldebt rule, in addition to the reference value of60% of GDP provided for in the MaastrichtTreaty. Some countries have already adopteddebt brakes.

In the case of PPoollaanndd, a debt rule is incorpo-rated in the national Constitution, stating thatthe debt-to-GDP ratio cannot exceed 60% ofGDP. Moreover, when the debt ratio standsbetween 50% and 55%, authorities arerequired by law to take precautionary meas-ures in order to reduce it. Harder efforts toreduce the debt ratio are required if debtexceeds 55% of GDP and are further intensi-fied if debt exceeds 60% of GDP.

HHuunnggaarryy is preparing to adopt a debt rule, sim-ilar to that of Poland. Taking measures isimperative if debt exceeds 50% of GDP andefforts escalate as the debt-to-GDP ratioincreases.

As regards SSwwiittzzeerrllaanndd,20 although the relevantconstitutional reform aims at containing thedebt-to-GDP ratio, the debt rule is indirect andessentially comprises an expenditure ceilingand provides for cyclically adjusted balancedbudgets. In particular, the rule specifies a one-year-ahead ex ante ceiling on expenditureequal to forecast revenues (net of one-off

10

1177 European Commission (2006), pp. 184-188.1188 Guichard et al. (2007), pp. 15-16.1199 Recently Greece committed to reduce its debt by 1/20 each year.2200 Cottarelli et al. (2009), p. 39.


35Economic Bulletin

June 2011

items), adjusted by a factor21 reflecting thecyclical position of the economy. In essence,this helps establish an expenditure ceilingwhich is in line with a balanced budget in cycli-cally adjusted terms. If ex post results (aftercalculating real GDP) show that actual expen-diture deviates from budgeted expenditure(due to poor GDP projections), the expendi-ture overrun is recorded in a notionalaccount. If the negative balance on thisaccount approaches 6% of expenditure,authorities are required by law to take meas-ures to reduce the balance within three years.This rule helped stabilise the debt ratio untilthe beginning of the current crisis.

The GGeerrmmaann debt rule, which is also embed-ded in the Constitution, is explicit at firstglance, as it sets an annual borrowing limit of0.35% of GDP. In essence, though, it is indi-rect and quite similar to that of Switzerland.Specifically, the structural deficit of the Ger-man federal budget cannot exceed 0.35% ofGDP, while the individual federal states needto balance their cyclically adjusted budgets.22

If the structural deficit ex post exceeds 0.35%of GDP, the additional amount is stored in anotional account. If the balance of thisaccount exceeds 1.5% of GDP, then, in accor-dance with the Constitution, the governmentis required to take fiscal consolidation meas-ures. It should be noted that the relevantordinary law, which specifies further details,is more stringent than the constitutional pro-visions, as it envisages the adoption of meas-ures if the balance exceeds 1.0% of GDP23

(rather than 1.5% provided for in the Con-stitution).

In the UUnniitteedd SSttaattees, the Congress determines,in absolute terms, a ceiling on borrowing by thefederal government, currently set at $14,300billion.24 Recently, as the federal debt cameclose to the legal limit, the White House isseeking to raise that limit. However, numerousSenators and Representatives have submittedproposals that any further increase in the debtceiling should be accompanied by significantcuts in government expenditure.

The main difference between debt rules ofPoland and Hungary, on the one hand, andSwitzerland and Germany, on the other, is thatthe latter take explicitly into account the cycli-cal position of the economy, thus not hinder-ing the operation of automatic stabilisers. Thisis so because the limit applies to the structuralcomponent of the deficit, rather than its cycli-cal component.

3 CONCLUSIONS

11.. In the past twenty years, the number of coun-tries applying (national) numerical fiscal rulesin order to ensure fiscal sustainabilityincreased sharply to 80, according to the IMF,while the number of rules per country alsoincreased from 1.5 to 2.5 rules on average.

22.. Despite its serious structural fiscal problems,Greece has not yet adopted any nationalnumerical fiscal rules (other than the supra-national rules provided for in the Stability andGrowth Pact, which, however, failed to avertthe sovereign debt crisis). It is true that Law3871, which was passed in the summer of 2010,envisages the adoption of a national numeri-cal rule for expenditure. However, this provi-sion was not implemented in the 2011 budget,and it is unclear when it will be. Moreover,given that the ceiling on expenditure growth isto be determined by the Minister of Finance,this may not be the best solution for Greece,although judgment should be reserved untilthis rule is implemented in practice.

33.. In the context of creating an appropriatefiscal institutional framework to ensure thesustainability of public finances in Greece,this analysis underlines the need to adoptnumerical rules for primary expenditure andgovernment debt. Even if these rules comeinto effect at a later time, when the sovereign

11

2211 The cyclical factor is determined as the ratio of trend real GDP toexpected real GDP.

2222 Cottarelli et al. (2009), p. 40.2233 Ibid., p. 40.2244 Financial Times, 4 February 2011, p. 5.



debt crisis is over, it is advisable to legislatethem now. The debt rule in Germany will notcome into full effect before 2016; neverthe-less it was enshrined in the countrys consti-tution in 2009.

44.. These rules could deal with the main prob-lems facing public finances in Greece in thepast four decades, namely the continuousincrease in primary expenditure (and succes-sive overruns in budgeted expenditure) and thefailure to reduce the debt-to-GDP ratio. Even

when conditions were particularly favourablefor reducing the debt-to-GDP ratio (low inter-est rates, strong GDP growth, primary sur-pluses or low primary deficits, and privatisationproceeds), Greece only managed to stabilisethe debt ratio, albeit at very high levels.

55.. Finally, it should be noted that fiscal ruleslimit unexpected increases in deficits (due toexpenditure overruns) and ensure a stable taxenvironment, which contributes to strong eco-nomic growth in the long run.

12


35Economic Bulletin

June 2011

Bank of Greece (2007), Annual Report 2006, 213-216.Bank of Greece (2010), Monetary Policy Interim Report 2010.Cottarelli, C., . Kumar et al. (2009), Fiscal Rules: Anchoring Expectations for Sustainable

Public Finances, IMF Policy Paper.Debrun, X. et al. (2008), Tied to the Mast? National Fiscal Rules in the European Union, Eco-

nomic Policy, 297-362.European Commission (2006), European Economy: Public Finances in EMU 2006, 137-167.European Commission (2009), European Economy: Public Finances in EMU 2009, 87-99.Garcia, C., J. Restrepo and E. Janner (2011), Fiscal Rules in a Volatile World: A Welfare-

Based Approach, IMF Working Paper WP/11/56.Guichard, S. et al. (2007), What Promotes Fiscal Consolidation: OECD Country Experiences,

OECD Economics Department, Working Paper No. 553.Kopits, G. (2001), Fiscal Rule: Useful Policy Framework or Unnecessary Ornament, IMF Work-

ing Paper WP/01/45.Ljungman, G. (2008), Expenditure Ceilings A Survey, IMF Working Paper WP/08/282.Manessiotis, V. and R. Reischauer (2001), Greek Fiscal and Budget Policy and EMU, in Bryant,

R., N. Garganas and G. Tavlas (eds), Greeces Economic Performance and Prospects, Bank ofGreece and The Brookings Institution, 103-152.

OECD (2010), Economic Outlook, 2010/2, 256-260.Rapanos, V. and G. Kaplanoglou, Independent fiscal councils and their possible role in Greece,

Bank of Greece, Economic Bulletin, 33, 7-17.Tanner, E. (2004), Fiscal Rules and countercyclical policy: Frank Ramsey meets Gramm-Rud-

man-Hollings, Journal of Policy Modelling, Vol. 26, 719-731.

13

R E F E R ENC E S


T A X E V A S I O N I N G R E E C E : A N O V E R V I E W *

Melina Vasardani Economic Research Department

35Economic Bulletin

June 2011

Tax evasion is one of the key challenges facingthe Greek authorities in their efforts to achievefiscal consolidation. While its negative impacton government revenue is well-understood bythe public at large, its other economic impli-cations and social repercussions are less clear.This paper summarises the reasons that maketax evasion a major social problem anddescribes issues related to its definition andmeasurement. It then goes on to analyse therecently released 2009 dataset of tax returns1

as further evidence of the size and incidence oftax evasion in Greece. The paper concludeswith some policy recommendations that arelikely to improve the efficiency of tax imposi-tion and collection.

The implications of tax evasion

Tax evasion is a complex phenomenon and isstructural in nature, as it reduces the efficiencyof the economy and increases social inequality,2

by distorting the allocation and redistributionof resources both in the short and in the longrun. In greater detail, taxation and, by exten-sion, tax evasion affect disposable personalincome, thus also affecting, over a longer hori-zon, the structure of the economy, includinglabour supply and demand decisions acrossoccupations and sectors, the intetemporal allo-cation of income between consumption andsaving, prices, import volumes and growth,income outflows, etc. At the same time, taxevasion hampers the redistributive role of pro-gressive taxation, thereby increasing the taxburden on honest taxpayers and compromisingthe equal treatment of similar incomes. More-over, to the extent that it deprives the publicsector of funds that could have been used tofinance government expenditure, it drives upgovernment borrowing requirements, limits thescope for public investment, reduces the qual-ity and quantity of publicly provided goods, andhampers the functioning of a welfare state. Inthe light of the above, it becomes clear that the

cost of individual tax evasion is both directlyand indirectly diffused across society, to thedetriment of those who regularly fulfil their taxobligations.

The definition and measurement of tax evasion

The definition and measurement of tax evasionare subject to theoretical and practical con-straints, originating in the complexity of theeconomy, the tax framework and the tax col-lection mechanism. In market analyses, as wellas in articles appearing in the press, the conceptof tax evasion is very often confused with thatof the informal economy, which can be mis-leading.3 It is important to stress that these twoconcepts do not necessarily coincide (see Chart1), although tax evasion does relate in part tothe informal economy (defined as the economicactivity that is not recorded in the nationalaccounts, although it should be). Tax evasionmay occur in the context of any type of eco-nomic activity, whether recorded orunrecorded, current or non-current (such as inthe case of the real estate transfer tax and thecapital transfer tax).4 Generally, tax evasion canbe defined as the illegal practice of wilfully con-cealing income and other taxable items and ofavoiding paying tax (VAT, other withholdingtaxes, duties, levies, contributions, etc). Thecomplexity of the phenomenon is clearlydemonstrated by its many different manifesta-tions involving direct or indirect taxes, legalor natural persons, any taxation scheme (taxbrackets, separate taxation, imputed taxation),type of tax and occupational group (wage/salaryearners, professionals, traders, manufacturers,

15

** The views expressed in the paper are those of the author and donot necessarily reflect those of the Bank of Greece. Acknowledg-ments to Vasilios Manesiotis, Maria Flevotomou, Heather Gibsonand Helen Koltsida for their valuable comments.

11 Available at: http://www.gsis.gr/statistiko_deltio/statistiko del-tio_2009/statdeltio2009.html.

22 See Matsaganis and Flevotomou (2010).33 Also, tax evasion should not be confused with tax avoidance,

which is not against the law.44 See Manesiotis (1990) for a detailed discussion of the relationship

between tax evasion and the informal economy.

2-Meleti Vasardani: 1 05-06-12 08:18 15


etc.). Therefore, measuring tax evasion is adaunting task and is often subject to a prioriassumptions.

Given the link between tax evasion and theinformal economy, a considerable body ofpapers have focused in the past decades on thestudy of the phenomenon of informal econ-omy. In Greece, informal economy has beenestimated to average, over time and acrossstudies, 30% of GDP.5 Compared with otherOECD countries, it is estimated that the coun-try has one of the highest shares of informaleconomy in recent years, possibly translatinginto greater tax evasion (see Chart 2).6 Part ofthe problem is undeclared work, which isthought to be quite substantial (around 25% ofall controlled employees),7 reflecting, amongother things, a bilateral arrangement betweenemployer and employee not to declare employ-ment so that both can avoid costly social secu-rity contributions.8

The size of tax evasion in Greece had been for-merly estimated at around 4% of GDP.9 Thisfigure had been produced using differentmethodologies for direct taxes (a Cobb-Dou-

glas type revenue function) and indirect taxesdue to large differences between the two cate-gories, and had also taken account of tax eva-sion related to the informal sector. A morerecent study of personal income tax estimatedincome underreporting for the entire taxpayerpopulation at 10%, resulting in revenue lossesof 26% (because of progressive taxation).10

According to the same study, tax evasiontended to be higher among: (i) the self-employed (professionals,11 farmers), with aweighted average income underreporting ofaround 33%, implying a U-shaped distribution;

16

55 See Pavlopoulos (1987), Vavouras et al. (1990), Negreponti-Deli-vani (1991), Kanellopoulos et al. (1995), Tatsos (2001), Ernste andSchneider (1998), Schneider and Klinglmair (2004) and EuropeanCommission (2009).

66 See Schneider et al. (2010). Also see the updated tables for theperiod 1989-2009 available at: http://www.econ.jku.at/mem-bers/Schneider/files/publications/ShadowEconomy21OECD_2009.pdf.

77 See the Labour Inspectorate press release SEPE-EYPEA activ-ity statistics 2010: undeclared work (January 2011) in Greek.Also see European Commission (2009) for indirect measurementmethods for undeclared work.

88 Other factors related to undeclared work include existence of aminimum wage, employment of underage individuals and employ-ment of foreigners, etc.

99 See Kalivianakis et al. (1993).1100 See footnote 2.1111 In this paper, the term professionals refers to persons whose main

source of income is independent professional activities, trading,manufacturing activities and construction.


35Economic Bulletin

June 2011

(ii) the top- and low-income groups; and (iii) inrural areas relative to urban areas. The method-ology used compared income reported on indi-vidual tax returns with income reported in theHousehold Budget Survey. Subsequently, a tax-benefit model was used to estimate the size anddistribution of undeclared income.12

Tax statistics 2009

The analysis of tax information mainly per-sonal income tax data can also provide evi-dence of the size of tax evasion. The bulletinTax Statistics 2009, published in December2010 by the General Secretariat of InformationSystems of the Ministry of Finance, includes,in addition to personal income tax data, infor-mation and data on income (profit) tax on legalpersons,13 VAT data, as well as data on finesand capital taxes (inheritance, donation andparental gift taxes, and capital transfer or cap-ital accumulation taxes).

According to these data, the total incomeearned in 2008 was 115.1 billion. Of this

amount, 98.4 billion (or 85%) representedthe taxable income of natural persons and theremaining 16.7 billion (or 15%) the taxableincome of legal persons (see Chart 3). Thedeclared income14 of natural persons was 98.1billion. The distribution of declared personalincome by source of income shows that theoverwhelming proportion of this amount(76%) was accounted for by income fromwages/salaries and pensions. The other

17

1122 Tax-benefit microsimulation models evaluate the fiscal and redis-tributive impact of public policy by calculating benefit entitlementsand tax liabilities for a representative sample of a countrys pop-ulation. The model used in the study in question was the EURO-MOD model of cross-national microsimulation (seehttp://www.iser.essex.ac.uk/research/euromod).

1133 Legal persons include mainly socits anonymes, limited liabil-ity companies, general partnerships, limited partnerships and jointventures. By contrast, sole proprietors are natural persons.

1144 Declared income (in relation to natural persons): a taxpayersincome as reported to the tax authorities. This typically coincideswith gross taxable income (which represents the amount ofpotentially taxable income). If, however, it is lower by at least20% than income imputed on the basis of living expenditure andasset acquisition expenditure, such imputed income is used insteadin the computation of taxable income. (Net) taxable income isthe amount of income minus tax exemptions and deductions. Forinstance, donations, life insurance premiums, interest payments,family expenditure with supporting receipts and mandatory socialsecurity contributions are all fully or partially deductible from grosstaxable income. The result is used to calculate the tax due, net ofthe amounts specified in the provisions in force in connection withany tax credit (medical expenses, principal residence rent payments,tuition fees, alimony payments, etc.).



sources of income made a far lesser contribu-tion to total declared income: trade-manufac-turing (9%); construction (9%); professionalactivities (4%); agriculture, securities andincome from abroad (2%) see Chart 4. Theobservation that income from wages andsalaries accounted for the bulk of declaredincome in 2009 was found to also hold true atthe level of individual regions and prefectures.

The total tax burden on legal and natural per-sons was 13.1 billion in 2009. Of this amount,9.1 billion (or 69%) was related to personalincome tax, while the remaining 4 billion (or31%) was accounted for by corporate incometax (see Chart 5). Looking at the total tax bur-den by occupational status of the tax liableperson or by occupational group, as is theterm used in the mentioned bulletin 52.59%of the total (personal and corporate) tax bur-den was incurred by wage/salary earners andpensioners, with only 16.62% being borne bythe rest of taxpayers. This latter figure can beroughly broken down as follows: professionals(6.81%); traders-manufacturers-craftsmen(6.32%); rentiers (2.33%); and farmers(1.16%) see Chart 6.15 Legal persons bore the

remaining 30.79% of the total tax burden, a 4percentage point decrease from 2008. In short,half the total tax burden in Greece was borneby wage/salary earners and pensioners, whilemuch smaller shares were observed for theother occupational groups and for legal per-sons. Removing legal persons from the calcu-lation of total tax so that it only reflects theburden on natural persons increases the rela-tive tax burden borne by wage/salary earnersand pensioners in 2009 to 76%, roughly thelevel it stood in 2008.

The breakdown by taxable income bracket ofindividual taxpayers provides further evidenceof the size and incidence of tax evasion herethe classification per occupational status isdone per person, not per tax return, in whichcase the main source of income of the tax liableperson would play the most important role.

18

1155 The weights used in the calculation of these percentages were theshare of each occupational group in total tax, as derived from theincome tax returns of the tax liable persons, and of their childrenand spouses. Given that classification of tax returns by occupationalgroup is based on the occupational status of the tax liable person,rather than on the individual occupational status of the taxpayer,these percentages may only be considered as proxies. Tax brack-ets and other criteria are in practice applied at the individual leveldepending on the occupational status of each taxpayer.


35Economic Bulletin

June 2011

According to this breakdown, from a total of8 million taxpayers, 5 million (or 63%) weretaxed as wage/salary earners and pensioners,and the remaining 3 million (or 37%) weretaxed as non-wage/salary earners and non-pensioners (see Chart 7). The average annualincome declared by wage/salary earners andpensioners was 14,913, more than double theincome declared by non-wage/salary earnersand non-pensioners (6,354 million). From atotal of 5 million of wage/salary earners andpensioners, 2.8 million (or 53%) reported apersonal income below the tax-free thresholdof 12,000, while 0.2% reported zero income(see Chart 8). From a total of 3 million of non-wage/salary earners and non-pensioners, 2.5million (or 83%) reported a personal incomeof less than 10,000, and 47% zero income.The tax-free threshold for non-wage/salaryearners in 2009 was 10,500. Hence, 64% of alltaxpayers declared an income below the appli-cable tax-free threshold, and 17% zero income.Also, some 85% of taxpayers reported anincome below the per capita GDP of Greece(2008: 21,000). It should be noted that thesefigures had increased compared with the pre-vious year. A personal annual income of more

than 900,000 was reported by a mere 33 peo-ple in the category wage/salary earners, plusanother 27 in the remaining categories, whilea personal annual income of more than100,000 was reported by 18,942 persons(0.23%) out of a total number of 8,342,160 taxfilers.

Even when taking into consideration the fac-tors that could potentially justify low incomes,it is still difficult to explain on the basis of theGreek economic reality how 8 in 10 non-wage/salary earners can have a real incomeof less than 10,000 per year and how 3 in 5taxpayers pay no tax at all.16

Turning to legal persons, from a total of221,363 firms filing a tax return in 2009, 97,037(or 44%) reported zero profits, while the vastmajority (209,311 firms, or 95%) reported tax-

19

1166 A recent study by the Centre of Planning and Economic Research,in which tax-income elasticities are estimated, found that gener-ally high-income taxpayers tend to hide incomes, irrespective of theoccupational group they belong to. It is the self-employed, however,who have the greatest flexibility in their reporting decisions as theycan easily alter their work schedules or remuneration arrange-ments, shifting even to the underground economy. See Kaditi andNitsi (2011).



able profits of less than 120,000. Total prof-its declared in 2009 declined by 14% year-on-year to 16.7 billion, translating into a reduc-tion in the share of firms in the distribution oftax burdens. Taxable profits per firm averaged75,485, against 90,895 a year earlier.Despite a downward trend between 2006 and2009, the ratio of corporate profits-to-GDPremained significantly higher in Greece com-pared with other euro area countries, includ-ing Austria, Belgium, Germany, France andthe Netherlands (see Chart 9).17

Moreover, the tax statistics for 2009 demonstratea number of serious weaknesses inherent in bothdirect and indirect tax collection. Previous yearsVAT was assessed at 7 billion (first-time assess-ments plus assessments of unaudited tax cases),of which only 558 million (or 8%) was collected.Likewise, of the 1 billion in assessed fines in2009, only 26 million (or 3%) was collected. Inseveral cases, failure to collect the assessed taxesor fines is due to a backlog of tax cases pendingin courts. Cases such as these are typically com-plex and time-consuming. Revenue from taxes oninheritance, donations and parental giftsamounted to 606 million, of which only around45% was collected, whereas 609 million in rev-enue from capital transfer and capital accumu-lation taxes was collected almost in full.

Policy recommendations

These statistical data, along with availableempirical estimates of tax evasion and theinformal economy, indicate that tax evasion isindeed a real, chronic and extensive problemin Greece, as repeatedly suggested by variousnational and international organisations. Thecomplexity of the countrys tax system and the

20

1177 See International Monetary Fund (2010).


35Economic Bulletin

June 2011

structure of the economy (featuring a relativelylarge number of self-employed individuals)18

tend to increase the scope for tax evasion andthe forms of commission of the offence.19

Adding to the problem are a number of organ-isational and other weaknesses in tax collectionand audits. For instance, according to esti-mates by the Ministry of Finance, Greece hasthe lowest tax revenue-to-GDP ratio in theEU, approximately 5 points below average,although the countrys tax rates are close to orhigher than (in the case of indirect taxes) theEU average rates.20 Besides income tax eva-sion, evasion of other (non-income) taxes likeroad duties, which is significant in Greece,must also be taken into account.

In view of the above, it is reasonable to concludethat, to effectively counter the complex phe-nomenon of tax evasion, a broad policy packageneeds to be put in place. This should be organ-ised around three main objectives: (a) simplify-ing tax legislation and rationalising the regula-tory framework; (b) overhauling and mobilisingtax administration; and (c) strengthening taxmorale and incentives for compliance.21

(a) Simplifying tax legislation and rationalisingthe regulatory framework

Simplifying tax legislation to ensure a flexibleand rational regulatory framework and reducethe need for explanatory circulars is a key pre-condition for the effective design and imple-mentation of tax policy. For one, taxes havinglow yields and high costs (assessment costs, col-lection costs, etc.) should be reviewed, as wasthe case of some 1,600 previously existing taxescollected on behalf of third parties, whoseadministrative costs were disproportionatelyhigh (around 40%) compared with expectedyields.22 At the same time, it is equally impor-tant that the 980 tax exemptions currently inforce, as well as the taxable income determi-nation methodology, are reviewed. Forinstance, income averaging and taxation underthe tax bracket system could be considered forprofessionals in art, entertainment, sports,stock brokerage, etc. who are currently taxed

under special regimes as they present largevariations in annual income.

The existence of a plethora of tax systems andthe constant revision of related legislation cande facto cause confusion, increasing the scopefor exceptions, special provisions, irregularitiesand omissions in the implementation of taxpolicy. Moreover, lack of communicationbetween tax and judicial authorities, extremelytime-consuming tax dispute resolution proce-dures (typically lasting up to 10 years) and mildpenalties often easily convertible intofines have all been identified to act as dis-incentives for tax compliance, and thus to exac-erbate the problem of moral hazard in taxpayerbehaviour. Therefore, strengthening the penalcomponent of the dispute resolution process byspeeding up the administration of tax justiceand imposing heavier and promptly enforce-able criminal penalties has to be a priority ofany strategy aiming to combat tax evasion.

(b) Overhauling and mobilising tax adminis-tration

Moreover, a key pillar of the fight against taxevasion is the overhaul and mobilisation of taxadministration, including at auditing and col-lection level. For audits to be effective, the firstrequirement is that it must be possible to cross-check and monitor information over timethrough a centralised integrated informationsystem. One such electronic platform is the Inte-grated Information System of Controlling Serv-ices (ELENXIS) of the Ministry of Finance. sa first step, ELENXIS allows cross-checks and

21

1188 The European Commission (2010) estimates the share of self-employment in total employment in Greece to be almost doublethe EU-27 average. In 2010, this share reached 30%.

1199 Examples of various methods of tax evasion can be found in Kali-vianakis et al. (1993). Tax evasion through the use of secret offshorebank accounts or the establishment of complex corporate structuresin tax havens has reportedly been significant in recent years.

2200 See the speech given on 25 February 2011 by the Minister ofFinance George Papakonstantinou to the Greek Parliaments Com-mittee of Economic Affairs during the discussion of the draft lawon Combating of tax evasion, staffing of the auditing agencies, andother provisions within the competence of the Ministry of Finance.

2211 For a detailed discussion of policy proposals to tackle tax evasion,see Bank of Greece, Monetary Policy-Interim Report 2010, specialfeature Tax evasion and tax administration.

2222 See Balfousias (1998).



risk-based audits of tax information. To ensure,however, a systematic electronic monitoring ofthe entire range of a taxpayers assets and lia-bilities vis--vis the state, and consumer spend-ing, it would be useful to connect various bod-ies and services, including Ktimatologio SA(Hellenic Cadastre), the Ministry of Transportand insurance companies, directly to the system,and strengthen cooperation with foreignauthorities. This would also facilitate the intro-duction and widespread use of a single per-sonal identification number for each citizento be modelled on the social security numberin place in many developed countries whichin turn would allow comprehensive treatment ofthe tax affairs of the individual.

Second, it is crucial to eliminate tax corrup-tion23 and reduce personal contact betweentax auditors/collectors and taxpayers. Meas-ures likely to contribute to this effect include:(a) dismissal from service, permanent dis-qualification for re-employment in thebroader public sector and immediate prose-cution (instead of transfer to another service,as has been the case so far) of auditors/col-lectors who accept bribes, (b) complete sep-aration of audit and collection services andmonitoring of their activities by an inde-pendent entity, (c) cutting the number of localpublic revenue offices and at the same timeincreasing the powers of central services, and(d) full computerisation of core operationslike taxpayer registration and monitoring offiling of tax returns.

Other measures that might be considered arereal-time payment of VAT (i.e. as the trans-action takes place) and/or payment of profes-sionals (doctors, lawyers, electricians, etc.)fees through a private accounting company act-ing as a provider of billing services and exclu-sively by a bank, irrespective of the amount.

(c) Strengthening tax morale and incentives forcompliance

In view of the ills associated with the phenom-enon of tax evasion in Greece, creating a tax

consciousness has been a long-standing chal-lenge, although not an impossible one.Besides the announcement of the anti-evasionpolicy and the governments commitment tostrictly implement the law, transparency in theuse of collected revenue and a gradualimprovement of publicly provided goods willalso be required to prove that such revenue isactually used in the best interests of citizens, inline with the idea of reciprocity which lies at theheart of the social contract between taxpayersand the state. Moreover, according to the pre-vailing strand of literature and other empiricalstudies, the most important determinant of taxbehaviour is the probability of being detectedand punished.24 In other words, the higher theprobability, in taxpayers opinion, that tax eva-sion will be discovered and that penalties willbe imposed promptly, the greater their com-pliance. Of course, this presupposes that finesand other penalties are imposed by auditors ina fair and transparent manner. Tax officers cancontribute to improving taxpayers under-standing of the tax framework and the variousoptions available, so long as such tax coun-selling is provided in a legal and regulated man-ner. This, in turn, should help strengthen vol-untary tax compliance, freeing up administra-tive resources that could be used to target spe-cific groups representing a high risk of tax eva-sion or having a history of non-compliance.Finally, given the current public feeling ofextensive tax evasion and unfair taxpayer treat-ment, the prosecution and punishment of bigtax evaders as a means of setting an exampleand delivering social justice should also have asignificant multiplying effect.

Conclusions

This paper provides a brief overview of theproblem of tax evasion in Greece, highlightingits serious fiscal, social and economic impli-

22

2233 See Annual Report of General Inspector of Public Administration,various years in Greek.

2244 See Allingham and Sandmo (1972) and Kleven et al. (2010).According to the typical approach proposed by Allingham andSandmo (op. cit.), the three key determinants of tax evasion are:the (marginal) tax rate, the probability of being detected and pun-ished and the size of the penalty imposed.


35Economic Bulletin

June 2011

cations. In view of the multifaceted nature ofthe problem and its deeper endemic causes,anti-evasion efforts must be collective, sys-tematic and multi-sided, along the followingthree axes: simplifying legislation, improvingtax collection and audits, and strengtheningtax morale. Considerable efforts have beenundertaken by the Greek authorities in recentyears to establish an efficient tax framework,25

capable of yielding the anticipated revenueand of ensuring a transparent and equitabledistribution of the tax burden. However, lawenforcement and effective tax collection haveactually proved to be quite challenging in theface of serious institutional and organisationalweaknesses and political and socioeconomic

distortions. Cracking down on tax evasion willtherefore require greater resolve, as well asradical changes at many levels. Moreover, it isessential that a systematic and coherent effortis made at every stage of the taxation process,from the design of tax policy and the audit oftax subjects to the imposition of criminalpenalties and the creation of a sound tax con-sciousness.

23

2255 See Law 3943/31.3.2011 (Government Gazette A 66) on Com-bating tax evasion, staffing of the auditing agencies, and other pro-visions within the competence of the Ministry of Finance, Law3888/4.10.2010 (Government Gazette A 175) on Voluntary clo-sure of tax disputes, settlement of overdue debts, provisions oneffective punishment of tax evasion, and other provisions, and Law3842/23.4.2010 (Government Gazette A 58) on Restoring taxequity, addressing tax evasion, and other provisions.



Allingham, M. G. and A. Sandmo (1972), Income tax evasion: a theoretical analysis, Journalof Public Economics, 1, 323-338.

Balfousias, A. (1998), [Man-agement cost of the Greek taxation system], Centre of Planning and Economic Research, Athens[in Greek].

Enste, D. and F. Schneider (1998), Increasing shadow economies all over the world Fictionor reality?, IZA Discussion Paper No. 26, Institute for the Study of Labor (IZA), Bonn.

European Commission (2009), Study on indirect measurement methods for undeclared work inthe EU, December.

European Commission (2010), European Employment Observatory Review: Self-employmentin Europe, September.

International Monetary Fund (2010), Greece: Second Review Under the Stand-By ArrangementStaff Report; Press Release on the Executive Board Discussion; and Statement by the Execu-tive Director for Greece, IMF Country Report No. 10/372, December, Washington D.C.

Kaditi, E. and E.I. Nitsi (2011), 2010 [Eval-uation of the 2010 tax reform], Centre of Planning and Economic Research, Athens [in Greek].

Kalivianakis, K., M. Xanthakis, A. Leventis, V. Manesiotis, S. Theodoropoulos, K. Trahanas andK. Flesiopoulou (1993), , [Taxation, shadow economy and tax evasion in Greece], Foundation for Mediterranean Stud-ies, Papazisis Publishers, Athens [in Greek].

Kanellopoulos, K., I. Kousoulakos and V. Rapanos (1995), : [Shadow economy and tax evasion: estimates and eco-nomic implications], Centre of Planning and Economic Research, Athens [in Greek].

Kleven, H.J. et al. (2010), Unwilling or unable to cheat? Evidence from a randomized tax auditexperiment in Denmark, NBER Working Paper No. 15769, February.

Manesiotis V. (1990), : [A tentative exploration of the relationship between tax evasion and the shadoweconomy], in Vavouras, I.S. (ed.), [Shadow Economy], Kritiki editions, Athens,158-177 [in Greek].

Matsaganis, M. and M. Flevotomou (2010), Distributional implications of tax evasion in Greece,GreeSE Paper No. 31, The Hellenic Observatory, London School of Economics and PoliticalScience, London.

Negreponti-Delivani, M. (1991), [The economicsof the shadow economy in Greece], Papazisis Publishers, Athens [in Greek].

Pavlopoulos, P. (1987), : [A firstestimation of the shadow economy in Greece], Foundation for Economic and IndustrialResearch, Athens [in Greek].

Schneider F., A. Buehn and C.E. Montenegro (2010), Shadow economies all over the world:New estimates for 162 countries from 1999 to 2007, World Bank, Policy Research WorkingPaper Series, No. 5356, June.

Schneider, F. and R. Klinglmair (2004), Shadow Economies around the world: What do weknow?, IZA Discussion Paper No. 1043, Institute for the Study of Labor (IZA), Bonn.

Tatsos, N. (2001), [Shadow economy and taxevasion in Greece], Foundation for Economic and Industrial Research, Papazisis Publishers,Athens [in Greek].

Vavouras, I.S, K. Karavitis and A. Tsouhlou (1990), [An indirectestimation method for the informal economy: the case of Greece], in Vavouras, I.S. (ed.), [Shadow Economy], Kritiki editions, Athens, 367-379 [in Greek].

24

R E F E R E N C E S


35Economic Bulletin

June 2011 25

1 INTRODUCTION

The Committee of European Banking Super-visors (CEBS), which was renamed to Euro-pean Banking Authority (EBA) as from 1 Jan-uary 2011, was assigned by the ECOFIN Coun-cil1 to organise and conduct in the first half of2010 a second EU-wide stress testing exercise,2

in close cooperation with the European Cen-tral Bank (ECB), the European Commissionand the EU national supervisory authorities.The stress test was conducted on a sample of91 banks in 20 Member States that represent65% of the EU banking sector in terms of totalassets. The sample also included six Greekbanks, which account for 90% of the Greekbanking systems assets. The exercise receivedwide publicity and was thoroughly reported inthe press, while its objective was to assist mar-ket participants and competent authorities indrawing conclusions on the soundness of banksand their capacity to absorb shocks on markets,as well as to assess the overall resilience of thebanking system to potential adverse changes inthe economic activity.

Nevertheless, it should be noted that the Greekbanking system was stress-tested for the firsttime in 2005, when the International MonetaryFund (IMF) developed a Financial SectorAssessment Program (FSAP), which had twomain components: (i) an assessment of com-pliance with core principles for banking super-vision, and (ii) a stress testing exercise.Although the scenarios applied in that exer-cise3 could be characterised as mild comparedwith the subsequent financial crisis, its con-clusions were rather robust, as they showedthat the Greek banking system was adequatelycapitalised to absorb any internal or externalshocks. This was also corroborated both by theEU-wide stress testing exercise conducted bythe EBA and by the performance of Greekbanks so far.

A similar exercise was also conducted by theUS Federal Reserve,4 with a view to assessingthe soundness of the 19 largest US bank hold-ing companies and to distributing accordinglythe amounts of government financial assis-tance, under the Supervisory Capital Assess-ment Program (SCAP). The SCAP was astress test exercise conducted by the supervi-sory agencies in May 2009, aiming at assess-ing the capital adequacy of large banks and atdetermining whether additional capital isrequired. This capital, if needed, should beraised either in the market or through theissuance of mandatory convertible preferredshares under the U.S. Treasurys CapitalAssistance Program.

The objective of the present study is two-fold:first, to discuss the notion and the usefulnessof stress testing exercises, and second, toanalyse and compare the stress tests con-ducted by the Fed with the EU-wide exercisecarried out by the EBA. This will contributeto the understanding of the stress tests natureand facilitate all interested parties in follow-ing and correctly interpreting the repeatedround of stress tests that the EBA is con-ducting.5

THE U S E FU LNE S S O F S T R E S S T E S T I NG EX ER C I S E S FOR A S S E S S I NG TH E S OUNDNE S S O F TH E B ANK I NG S Y S T EM*

Faidon Kalfaoglou Economic Research Department

** The views expressed in this study are personal and do not neces-sarily reflect those of the Bank of Greece.

11 See Council of the European Union, Economic and FinancialAffairs, Press release, Brussels, 2 December 2009, http://www.con-silium.europa.eu/uedocs/cms_data/docs/pressdata/en/ecofin/111706.pdf.

22 The first EU-wide stress testing exercise was conducted in 2009and its results were forwarded to the ECOFIN Council in Octo-ber 2009 but were not disclosed. Conversely, the ECOFIN Coun-cil decided to disclose the results of the 2010 stress test. Seehttp://www.consilium.europa.eu/uedocs/cms_data/docs/press-data/en/ec/115346.pdf.

33 See Kalfaoglou (2006).44 See Board of Governors of the Federal Reserve System (2009a) and

(2009b).55 See The European Banking Authority up and running and prepar-

ing new EU-wide stress test, 13 January 2011,http://www.eba.europa.eu/News-Communications/Latest-news/The-European-Banking-Authority-up-and-running-and-.aspxand , 2011 EU-Wide Stress Test: Methodological Note,18.3.2011.

3-Meleti Kalfaoglou: 1 05-06-12 08:18 25

2 BANKING RISKS

Banks provide two kinds of services: brokerageand asset transformation. Brokerage servicescan be defined as those mechanisms set up bya bank to facilitate the execution of transac-tions, whereby the bank acts solely as a brokerfor a commission. As the bank does not assumeany financial position, the risks it faces are notfinancial but non-financial, e.g. operationalrisk. The second function refers to asset trans-formation, which is the ability to convert(transform) the financial characteristics ofassets and liabilities, such as maturity trans-formation, size transformation, currencytransformation, liquidity transformation, etc.This means that banks are able to offer highlyliquid products (e.g. sight deposits payable ondemand) and to invest their reserves in less liquid investments. Similarly, they can acceptdeposits in a currency and invest in another.However, asset transformation results to mis-matches between assets and liabilities, whichgenerate a position that, in turn, entails thefinancial risks facing a bank. Thus, risk takingis intertwined with the functioning of a bank,and hence the definition of banking as the artof managing risks.

Banking risks may be categorised in severalways. In academic literature various classifi-cations have been developed, depending on theassumptions and the objective of the study.The significance of any risk depends on thebusiness model of the bank and the externalenvironment in which it operates. Forinstance, market risk is more important forinvestment banks but in general has becomemore significant ever since banks started build-ing considerable trading book positions. Creditrisk has always been important, due to its link-age with the main function of a bank, i.e. thegranting of loans, and its importance is ampli-fied especially in times of financial distress.Operational risk was almost unknown untilrecently when it was found out that it laybehind the collapse of historic banks (e.g. Bar-ings) and thus, some of its components (suchas fraud risk) have gained importance amid the

financial crisis. As regards liquidity risk, thepre-crisis view that ample liquidity does notentail any risks proved to be erroneous, as itled to a relaxation of credit standards, whilethe various market innovations that haveemerged brought about parameters not ade-quately understandable (e.g. securitisation).

First of all, the discussion will focus on creditand market risks which were the object of boththe EU-wide stress test and the respectivestress testing exercise in the United States. Theremainder of this study will explore liquidityrisk, which will be analysed in line with thestress testing exercise conducted in the firsthalf of 2011.

Credit risk is defined as the risk of loss due todefault by a banks borrowers. Although abanks total assets are also exposed to creditrisk, the analysis focuses mainly on its lendingportfolio. Moreover, a bank may face creditrisk variations, such as country risk, i.e. the riskof loss due to a banks exposures towards acountry either directly (direct investment) orindirectly (cross-border exposure). In the lightof the above, country risk can be broken downinto the following components: political risk,sovereign risk, exchange rate risk, transfer risk,as well as second-round risks. Sovereign riskreflects the potential risk of loss in the eventthat the central government of the host coun-try defaults on its obligations or restructures itsdebt.

Credit risk can be described as the combinedresult of three individual risks, i.e. default risk,exposure risk and recovery risk. Default riskarises when borrowers fail to meet their con-tractual obligations, which typically leads toloan restructuring. Exposure risk is related tothe total amount exposed to credit risk. Asregards loan portfolios, the amount is repre-sented by the nominal value of credit. Themeasurement becomes complex in the case ofoff-balance sheet items, with the most commonpractice consisting in calculating credit equiv-alent value. Recovery risk is related to the pro-portion of the debt that a bank would receive



in final satisfaction of the claims on a defaultedcredit. Recovery rate depends on the value ofcollateral and/or guarantees, as well as on theseniority of the banks debt.

As a result, on the basis of the aforementionedcomponents, credit risk can be described byassessing of the following three parameters:

Probability of Default (PD), referring to theprobability of a borrower or a group of bor-rowers defaulting on their obligations vis--visthe bank. It is expressed in percentage terms.

Loss Given Default (LGD), referring to theloss incurred by a bank due to a debtors default,expressed as a percentage of total exposure.

Exposure at Default (EAD), referring to thetotal value that a bank is exposed to at the timeof default, expressed as a sum denominated inthe portfolio currency.

Having assessed all of the above parameters,banks establish provisions to cover anyexpected losses ELp due to debtors default asfollows:

ELp=iEL=

i(EADiPDiLGDi)

Expected loss is the average loss a bank expectsto sustain in a normal financial environment.As it is expected, it is not considered as abank risk but it is part of its cost, since the bankcan take appropriate measures to remedyexpected losses. These measures include pric-ing and provisioning. In the first case, interestrate margins increase in order to compensatefor the expected loss, while in the second casethe banks profitability is affected.

The estimation of expected loss is based on thebanks historical experience. Nevertheless, it islikely that the next period will be moreunfavourable than the parameter estimationperiod and losses will be higher. In this case,we refer to unexpected loss, which is embed-ded in banking risk and needs to be covered bythe so-called economic capital. Unexpectedloss is derived by statistical estimation, i.e. asa multiple of the expected loss, depending onrisk factor movements and on the risk toler-ance degree. Furthermore, if economic con-ditions are very adverse, the ensuing loss isdefined as stressed loss. All of the above canbe illustrated with a simplified chart (Chart 1).

Nevertheless, the above representation of lossdistribution is idealistic. In fact, no widely

35Economic Bulletin

June 2011 27


accepted model for the determination of lossdistribution and the level of economic capitalhas been developed so far, nor are there anytechniques to incorporate changes in historicalrelations during periods of crisis, with the mostimportant being data correlation. However, themain argument against the use of the economiccapital framework is the unwillingness of deci-sion-makers to take corrective action in linewith this framework. Therefore, for the timebeing, the economic capital framework is notsuitable for incorporating stressed loss andconsequently, the stress test results focus onexpected rather than unexpected losses.

In the light of the above, an adverse or excep-tional change in any of those three risk factors(PDi, LGDi, EADi) would generate a stressedloss SLp and this differential SLpELpshould be covered by capital.6 This is the pur-pose of stress testing, namely to determinewhether the capital base continues to cover thestressed loss and whether it still exceeds reg-ulatory minimum capital requirements.

For the estimation of credit risk parameters, arating system should be developed. This creditrisk modelling can be achieved with credit rat-ing systems for corporate portfolios or creditscoring systems for retail portfolios. Creditscoring systems are distinguished in two largecategories: application scoring models, whichtake into account data collected during theapplication process for the granting of creditproducts, as well as behavioural scoring mod-els, which are based on existing customers andtake account of information about their cred-itworthiness or payment behaviour.

Furthermore, by exploring the rationale behindthe design of these models, we distinguishthem into the following two categories: a)models in which causality is derived by theapplication of statistical techniques, and b)those in which causality stems from economictheory. The most commonly used models in thefirst category are discriminant analysis andlogistic regression, while the second categorycomprises options-based structural models at

the microeconomic level and intensity-basedmodels that consider default as a random,exogenous event and attempt to model thetime or the intensity of an eventual default,regardless of its causes.

Market risk is defined as the potential loss onthe valuation of assets included in a bankstrading book due to changes in market prices.Changes in prices include movements in inter-est rates, exchange rates, stock (or index)prices and commodity prices. The trading bookcomprises positions in financial instrumentsand commodities held for trading or for hedg-ing the risks inherent in other assets of thetrading book. The financial instruments thatare usually included in a trading book arebonds and other debt securities, positions inforeign currency, shares, positions in com-modities, mutual funds, as well as derivativeshaving any of the above instruments as theirunderlying assets.

The key components of market risk7 are generalposition risk and specific position risk. Generalposition risk refers to changes in a financialinstruments valuation due to changes in itsmarket price, whereas specific position riskrefers to changes in a financial instruments val-uation due to changes in the risk borne by itsissuer. Another important factor of market risk,which at this point is not approached from asupervisory perspective, is market liquidity risk,i.e. the risk that a financial instrument cannotbe liquidated without suffering significantlosses in its market price, and is approximatedby the market bid-ask spread.

As a result, changes in the value (Vi ) of eachasset i of the trading book can be expressed asa function of adverse changes in one of the


66 Part of the differential can be covered by operational profits (ifany), provisions, a decline in assets, deleveraging, etc. over the timehorizon of the stress test.

77 Market risk in terms of supervision is defined in a broader sensethan the usual definition provided by banks Asset ManagementDepartments. It comprises five components: position risk, foreignexchange risk, counterparty risk, settlement risk and large expo-sure risk from trading book. This study discusses only position risk,as this component was stress-tested by the EU-wide exercise.


main risk factors. Thus, if shock (Sj ) for j ={interest rates, exchange rates, share prices,commodity prices}, then:

(Vi )=(Sj )

where the valuation function f varies accordingto j. In the event that the underlying assets areinterest rates, e.g. bonds, the equation will beas follows:

(Vi )=PV PV(Sj )

i.e. the difference of the current bond valuebefore [PV] and after the shock [PV(Sj)].

Market risk modelling is easier than credit riskmodeling. Typically a valuation model or VaRmodel is used. Valuation models take intoaccount the modified duration in respect ofinterest rate instruments, the beta coefficient inrespect of equity products, and the delta andgamma coefficients in respect of option deriva-tives, while foreign exchange instruments are lin-ear. As regards the VaR model, it illustrates themaximum potential loss that a bank expects tosustain within a given period of time at a givenlevel of probability. The time period and theprobability depend on the planning period andthe degree of risk tolerance. For supervisory pur-poses the period is determined at 10 days andthe level of confidence at 99%, whereas histor-ical data covering at least 250 days must be used.

After determining the implications for eachrisk separately, the issue of risk aggregationarises. There are two alternative methods: (i)by simply adding up the individual componentsand (ii) by taking account of a potential inter-dependence between the two risks.

The interdependence issue lies in that differ-ent types of risk must be merged into a singlenumber. From a more technical perspective,this means that marginal loss distributions ofdifferent risks must be merged into a single dis-tribution. However, these distributions maydiffer substantially between them or vary interms of time span.

The interaction between credit risk and mar-ket risk has been corroborated by several stud-ies. Jarrow and Turnbull (2000)8 report on thatmatter: Economic theory tells us that marketand credit risk are intrinsically related to eachother and, more importantly, they are not sep-arable. If the market value of the firms assetsunexpectedly changes generating marketrisk this affects the probability of default generating credit risk. Conversely, if the prob-ability of default unexpectedly changes gen-erating credit risk this affects the marketvalue of the firm generating market risk.The Basel Committee, acknowledging theimportance of the interaction of credit andmarket risks, issued a working paper,9 con-cluding that a considerable underestimation oroverestimation of total risk may occur, depend-ing on the degree of both diversification andcompounding effects. The criticality of theissue led the EBA to release a consultationpaper,10 mentioning the following:

Guideline 2: In order to adequately manageconcentration risk, institutions should have anintegrated approach for looking at all aspectsof concentration risk within and across risk cat-egories (intra- and inter-risk concentration).

Liquidity risk can be defined as the failure ofa credit institution to raise, at a reasonablecost, the funds required to implement its busi-ness planning, meet its obligations and con-tinue its operation. Its obligations may ariseeither from the liability side, as its inability toservice its liabilities that fall due and/or as anunforeseen withdrawal of deposits, or from theasset side, as an unexpected utilisation by cus-tomers of committed credit lines and/or itsinability to unwind a position without signifi-cantly lowering market prices. It is clear thatliquidity risk components are a combination ofendogenous factors (e.g. investment policy)and customer behavioural factors (e.g. depos-itor behaviour towards market strains). We

35Economic Bulletin

June 2011 29

88 See Jarrow and Turnbull (2000). 99 See BCBS (2009). 1100 See CEBS (2010c).


can distinguish four risk factors: (1) the con-centration of funding sources; (2) the level ofcore deposits; (3) the level of unused creditlines; and (4) the marketability of assets (mar-ket liquidity).

The above analysis is a traditional approachto liquidity risk. However, innovations in thefinancial instruments market have shed light toadditional risk factors that are in certain casesindiscernible. For instance, the widely usedpractice of generating loans, creating securitiesbacked by these loans (securitisation) and thendistributing them to investors on the market(originate and distribute model) highlighteda new perspective in banks liquidity.

All liquidity risk factors can be quantified but,unlike other risks, there is currently no way toestimate the impact on a banks regulatory cap-ital. This means that it is necessary to deter-mine the objective of both analysing and stress-testing liquidity risk. As a rule, the analysisfocuses on the ensuing liquidity gap, as well ason the impact on liquidity ratios when a riskfactor is subject to an exceptional change.

Liquidity gap depends upon the asset/liabilitymaturity mismatch. Inflows and outflows areestimated at different time periods,11 as a resultof exceptional changes in risk factors, while theliquidity gap, either positive or negative, is cal-culated as the differential between inflows andoutflows. Liquidity gap is then compared withthe liquidity buffer, which reflects the addi-tional available liquidity in times of distress.12

It consists of liquid assets and other readilymarketable assets on money and capital mar-kets, without price discounts. The purpose ofa stress-test exercise is to determine the cov-erage of liquidity gap.

Liquidity ratios may be simple analytical ratios(e.g. loan-to-deposit) or more composite reg-ulatory ratios such as those defined in Bank ofGreece Governors Act 2614/7.4.2009 (liquid-ity coverage ratio and mismatch ratio) or thetwo new liquidity ratios (liquidity coverageratio and net stable funding ratio), which have

been recently proposed but not yetendorsed by the Basel Committee.13 Anunexpected shift in any of the risk factorschanges the ratios nominator and/or denom-inator and the analysis focuses on whether theratio continues to move within the predeter-mined limits.

It is clear that liquidity risk analysis may havesignificant elements of subjectivity, which ledthe EBA to suggest the following in its con-sultation document on stress testing:14

Liquidity risk 9. The results of the stress testsshould be used as an input for adjusting andimproving liquidity risk management.

3 A QUANTIFIED APPROACH TO RISK MANAGEMENT

The brief analysis of risks and their interde-pendence clearly reveals the revolutionarydevelopments of the last 15 years in the field offinance and banking with regard to the quan-tification of financial decisions. Several bankshave developed financial and econometricmodels, such as those described above, in orderto assess and understand their risks. Thesemodels have gained reliability because theyprovide a uniform framework for the identifi-cation, measurement, monitoring and man-agement of risks. However, at the same timethey created an illusion of safety, since in manycases it was disregarded that these models areactually a simplification of reality which,despite their increasing complexity and com-puting power, are still quite elementary to cap-ture all aspects of risks. The failure of modelsto forecast the current crisis urged AlanGreenspan to write15 that it is impossible tohave the perfect model of risk, mainly becauseof the rational behaviour hypothesis. All of the


1111 As regards liquidity risk, the critical time span is 1-30 days and isreferred to as the survival period.

1122 See CEBS (2009). 1133 See BCBS (2010). 1144 See CEBS (2010c). 1155 Alan Greenspan, We will never have a perfect model of risk,

Financial Times, 16.3.2008.


conventional models assume that participantsin the market respond rationally without incor-porating notions of animal spirits, as Kaynesput it, that prevail during periods of exceptionalchanges. Akerlof and Shiller (2009)16 haverecently attempted to decipher this notion anddetermine its components. Conventional eco-nomic theory assumes that participants in themarket act in a rational way and personal per-ceptions, passions and feelings do not influencethe process of forging economic events, whichdepend on risk factors and government deci-sions. Irrational choices, non-economic motivesand sentiments (animal spirits) have not beenput under the microscope of risk analysis. Nev-ertheless, in many cases, and particularly underextraordinary conditions when risk factorsreach extreme values, all of the above featureslead to additional uncertainty. Sometimes thisuncertainty strengthens the creation of eco-nomic events but sometimes it may lead toparalysis. Conventional risk managementdoes not take into account all of the above.

Within this framework, stress testing exerciseshave evolved. They are a risk management toolstemming from financial decision modelling,attempting to answer the question of whatwould happen, should certain exceptional butplausible risks materialise. This question arisesfrom the innate human need to anticipatefuture, and particularly unfavourable, events.We all conduct stress-test exercises in our per-sonal lives. For instance, havent we all con-templated our reaction to a potential job loss?Similarly, banks wish to plan their response topotential shocks in their economic environ-ment or, more technically, to the differentpositions of the economic cycle. In order toanswer all these questions, risk modellingtechniques were applied to quantify theresults. This is not wrong in principle, but firstof all an answer should be given to whethermodels continue to be effective in times ofstress. As already mentioned above, economicbehaviour changes in times of stress. Typically,models are based on a pattern of growth thatincludes both periods of recovery and periodsof recession, but the dynamics of each period

varies in line with the drivers of each situation.As a result, even if the perfect model did exist,it would most probably not be useful in peri-ods of stress when risk factors undergo excep-tional changes.

4 STRESS TESTING

Stress testing seeks to find out what could hap-pen to individual banks and/or the financialsystem as a whole, when and if specific excep-tional but plausible risk scenarios should occur.Its unquestionable usefulness prompted theEBA to develop a framework for conducting astress testing exercise efficiently.17 Thisframework is illustrated in Chart 2 and is bro-ken down into five areas of analysis: (1) cor-porate governance; (2) methodology; (3) sce-narios; (4) implications; and (5) correctiveaction.

(1) Corporate governance issues play a key rolein the success of a stress testing exercise. Thesenior management of each bank is responsiblefor conducting the stress test and should be ableto understand the main assumptions, approvescenarios, make sure that the stress test coversthe entire bank, adopt a reporting system andtake corrective measures in view of the impli-cations. The stress test should form part of anongoing risk management programme and relyupon adequate infrastructures, while itshypotheses, reporting lines and effectivenessmust be reviewed on a regular basis.

(2) As regards methodological issues, in con-ducting a stress test, a bank has to chooseamong two alternative techniques: the sensi-tivity approach and the scenario approach. Thesensitivity approach assumes a change in a spe-cific risk factor, e.g. a 30% decline in the shareprice index, without specifying the cause of thechange and irrespective of other risk factors,and the impact of this change is quantified.The scenario approach considers several

35Economic Bulletin

June 2011 31

1166 See Akerlof and Shiller (2009). 1177 See CEBS (2010b).


underlying risk factors, with an emphasis ontheir concurrent emergence, that are associ-ated with a specific portfolio, e.g. changes inthe loan portfolio due to simultaneous strainson interest rates and real estate prices, or aspecific event which has a horizontal effect onthe bank, e.g. changes in interest rates whichmay affect the corporate portfolio, the retailportfolio and the asset/liability relation.

The sensitivity approach has advantage overthe scenario approach due to its easy imple-mentation but it fails to fully reflect reality, asa shock is rarely limited to just one risk factor.Conversely, the scenario approach may bemore realistic but is more difficult to apply, asan intercorrelation model of risk factors isrequired. Under both approaches, shocks maybe based on historical experience or may behypothetical, reflecting theoretical future risks.

(3) With respect to the severity of scenarios, nounbiased answer can be provided.18 The stresstest assumes adverse but plausible scenarios.But what does adverse but plausible scenar-ios really mean? The answer is biased depend-

ing on the analysts sense of harshness. Dailyexperience might provide us with an answer,though not satisfactory enough, as some quan-tification is warranted. In many cases, theadverse and plausible scenario is determinedby a regulatory guideline. Namely, as regardsthe automotive industry, the safety of a car isdetermined by its performance in specific crashtests. The same applies to the launch of a newpharmaceutical product. As regards thefinancial sector, the stress testing exercise is arelatively new tool. It was used by the Inter-national Monetary Fund (IMF) under itsFinancial Sector Assessment Program (FSAP),but it was not incorporated in the decision-making process of supervisory authorities andbanks. The recent financial crisis acted as a cat-alyst for an attitude change, as it became clearthat the stress test is a powerful risk manage-ment tool. The latest example of a successfulstress testing exercise implementation was Ire-land,19 where the determination of capital


1188 See Isogai (2009).1199 See Central ank of Ireland, The Financial Measures Program

Report, 31.3.2011.


requirements for distressed banks was basedon a stress testing exercise, which was welcomeby the markets. Naturally, supervisory author-ities did not remain passive but they started,instead, to investigate the severity of scenarios.The EBA reports that:20

Guideline 10. Stress testing should be based onexceptional but plausible events. The stress test-ing program should cover a range of scenarioswith different severities including scenarioswhich reflect a severe economic downturn.

It is widely accepted that the degree of sever-ity depends on the use of the outcomes of astress test. In order to estimate capital ade-quacy, a mild recession scenario is needed,whereas when the objective of the stress test islong-term capital planning, a more adverse sce-nario is warranted. In general, stress testsshould include alternative scenarios to ensurethat the information they provide is significant.The quantification of the probability of a sce-nario materialising can usually be defined asonce every n years. The parameter n is dif-ficult to determine and once every 25 yearsis often used, which coincides with an execu-tive officers professional life. Nevertheless,adverse but plausible scenarios once every 100years or more could be used.

Besides, the degree of severity of a scenariodepends on the specific portfolio on which itapplies. A scenario may be adverse for onebank and mild for another. For instance, asregards a bank whose core business is credit toshipping, a plain recession scenario may not besatisfactory, since the business cycle of theshipping industry is known to give deeper andlagged recessions.

(4) Each bank must assess the impact of a stresstest on its assets, liabilities, profit and lossaccounts, and ultimately its regulatory capital.This, of course, implies an estimation of theexpected loss using a wide range of scenarios.

(5) Each bank must develop a corrective actionplan in respect of the stress test results to

remain functional and solvent. Correctivemeasures may include, among other things,risk mitigation techniques, a comprehensivelimit system, funding sources modifications,capital requirements planning, contingencyplanning and business continuity plans.

Three basic steps are required for the com-pletion of a stress test:

Step 1: determining macroeconomic scenarios.

Step 2: converting scenarios into risk param-eters.

Step 3: assessing the impact on a bank-by-bankbasis.

Typically, steps (1) and (2) are known as top-down approach, while step (3) constitutes thebottom-up approach; both approaches areillustrated in Chart 3. Let us investigate ea

economic bulletin 35 bank of greece

Documents

fiscal ruleswere

overall fiscal policy

use ofnumerical fiscal

fiscal sustainability

peripheral countries

developed countries

f i s c

c t i c e