economic bulletin - core.ac.uk · bank of greece 21, e. venizelos avenue 102 50 athens economic...

YELLOW

BLACK

MAGENTA

CYAN

ECONOMICBULLETIN

No 33

MAY2010

ISSN: 1105 - 9729

EC

ON

OM

IC B

UL

LE

TIN

No

33

MA

Y 2

01

0B

AN

K O

F G

REEC

E

EUR

OSY

STE

M

ECONOMICBULLETIN

No 33

MAY2010

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

CONT ENT S INDEPENDENT FISCAL COUNCILS AND THEIR POSSIBLE ROLE IN GREECEVassilis Th. RapanosGeorgia Kaplanoglou 7

DETERMINANTS OF YOUTH UNEMPLOYMENTIN GREECE WITH AN EMPHASIS ON TERTIARYEDUCATION GRADUATESTheodoros M. MitrakosPanos TsakloglouIoannis Cholezas 21

INFLATION AND NOMINAL UNCERTAINTY:THE CASE OF GREECEHeather GibsonHiona Balfoussia 63

DETERMINANTS OF THE GREEK STOCK-BONDCORRELATIONPetros M. Migiakis 79

WORKING PAPERS(February 2009 – February 2010) 91

ARTICLES PUBLISHED IN PREVIOUS ISSUESOF THE ECONOMIC BULLETIN 107

33Economic BulletinMay 20106

33Economic Bulletin

May 2010 7

1 INTRODUCTION

The emergence and persistence of fiscal deficitsover long periods of time, leading to increasedpublic debt, seems to be a common feature inmost OECD countries, at least in the last 30years. In these countries, public debt as a per-centage of GDP rose from an average ofaround 40% in 1975 to almost 70% in 2006 (seeOECD, 2007). Fiscal deficits, once typicallysmall and controlled in most OECD countries,have risen dramatically in the last two years,due to the global financial crisis. Had theemerging deficits and increased debts been theresults of some counter-cyclical policy pursuedin order to deal with adverse developments inthe domestic and/or the world economy, theproblem might not have been that great. Butthe systematic emergence of deficits and theensuing higher debt in many developed coun-tries, and indeed over long periods of time, sug-gest the existence of other endogenous factorsthat play a significant role in this development.

According to both theoretical and empiricalstudies, lack of fiscal discipline and weak fis-cal management are such systemic factors. Forinstance, it has been established that in peri-ods of fast growth, governments, rather thantaking advantage of the economic upturn toimprove their country’s fiscal position by cre-ating surpluses and reducing its debt, tend toadopt an expansionary fiscal policy, by increas-ing public expenditure, or cutting taxes, orboth. Thus, when the economic climatereverses, the counter-cyclical policy pursuedentails the creation of high deficits and a con-siderable increase of public debt. Typical

examples can be found in many countries ofthe euro area, particularly from 2004 onwards.A period of economic euphoria was not accom-panied by adequate fiscal adjustment, so theoutbreak of the global economic crisis has ledmany of these countries into a fiscal crisis (seeVan Riet, 2010). In the worst case, the abilityto conduct a counter-cyclical policy becomesconsiderably limited, as the recent experienceof Greece clearly shows. A key question thatarises is why governments exhibit such a ten-dency to create fiscal deficits (deficit bias), andwhat mechanisms could deter this behaviour.

The main mechanism for ensuring fiscal disci-pline, implemented at times in most OECDcountries, is the imposition of various forms offiscal rules, such as numerical limits to the levelof public expenditure, or to the level of deficitas a percentage of GDP. For instance, the Sta-bility and Growth Pact, which constitutes theregulatory framework for fiscal policy in thecountries of the EU and, more specifically, ofthe euro area, sets as a medium-term objectivethe achievement of a balanced budget, and, atthe same time, sets a limit of 3% of GDP onthe annual general government deficit.2

Lately, however, attention has also been paidto the importance of another institution,namely that of independent fiscal authoritiesor councils.

I N D E P ENDENT F I S C A L COUNC I L S ANDTHE I R PO S S I B L E RO L E I N GR E E C E 1

Vassilis Th. RapanosUniversity of AthensDepartment of Economics

Georgia KaplanoglouBank of GreeceEconomic Research DepartmentandUniversity of AthensDepartment of Economics

11 The views expressed in this study do not necessarily reflectthose of the Bank of Greece. The authors would like to thankHeather Gibson and Vassilis Manessiotis for their constructivecomments.

22 The strictness and inflexibility of the Pact’s rules led to a revisionof certain provisions in 2005, so that now each member country hasa different medium-term objective; however, the Pact’s overallrationale regarding the imposition of quantitative rules on key fis-cal aggregates remains unchanged.

The theoretical investigation of such institu-tions was originally sparked by the successfulassignment of the right to conduct monetarypolicy to independent central banks (seeEichengreen et al., 1999, Calmfors, 2003,Wyplosz, 2005). In practice, independent fis-cal councils have gradually been established ina growing number of countries where,although not involved in the conduct of fiscalpolicy (according to the standards applicableto central banks), they play an important rolein enhancing the reliability and transparency ofthe respective countries’ fiscal policy.

The specific activities and responsibilities ofeach of these councils differ considerablyacross countries (see European Commission,2009); yet, as a general rule, the operation ofalmost all of them involves monitoring andassessing the fiscal policy followed, publiclymaking recommendations regarding the coun-try’s economic policy, assessing macroeco-nomic projections and examining the reliabil-ity of the forecasts for expenditure and revenuecontained in the budget, and therefore thedeficit. Another element systematicallyassessed by such councils is the long-term sus-tainability of fiscal aggregates, and especiallyof public debt.

This study attempts to analyse the reasons whythe establishment of such an independentauthority may prove extremely useful, partic-ularly in countries with considerable fiscalproblems and a history of inadequate institu-tions promoting fiscal discipline. Specifically,the next section examines in brief why gov-ernments on their own may not adopt a pru-dent fiscal policy, and why certain mechanismsalready in place for the promotion of fiscal dis-cipline may not be adequate. The third sectionattempts to concisely present the various formsof independent fiscal authorities suggested inthe theoretical literature, and also examinesthe main features of fiscal councils alreadyactive in various OECD countries. Section 4refers in particular to the role that such a coun-cil could play in Greece, as well as to the pre-requisites for its successful operation. Finally,

the last section attempts to draw some basicconclusions.

2 WHY DO PARLIAMENTARY SYSTEMS LEAD TOEXCESSIVE DEFICITS?

In public economics, a major issue that hasattracted a lot of attention, particularly inrecent years, has been to explain why there isa deficit bias in parliamentary democracies.The cumulative result of successive deficits isthe gradual accumulation of a high public debt,with major negative implications for a coun-try’s growth and social policies. Numerous rea-sons have been suggested in the theoretical lit-erature to account for this phenomenon, andtheir detailed presentation would perhapsnecessitate dozens of pages.3 Hence, any ref-erence made to these reasons in the presentstudy is more indicative rather than exhaus-tive.4

The first and most convincing explanationoffered is the “common pool” problem, anotion originally introduced by Hardin (1968)with respect to the overexploitation of freelyaccessible natural resources, and laterextended in the field of fiscal policy by numer-ous other authors, including Shepsle andWeingast (1981), von Hagen and Harden(1995), and Krogstrup and Wyplosz (2006). Putsimply, the core idea of this approach is thatwhen many individuals potentially haveaccess to a scarce resource, while at the sametime bearing only a small part of the respectivecosts incurred, there is an observed tendencytowards overconsumption of this resource. Asimilar phenomenon can often be observedwith respect to a country’s public budget aswell. Under the pressure exerted by usuallysmall but powerful interest groups, publicexpenditure is increased to the benefit of these


33 The issue has attracted a lot of attention within the strand of Eco-nomics called Public Choice. For a brief yet thorough presentationof the main postulates of Public Choice, see Mueller (2003).

44 For a more detailed analysis of the reasons creating this deficit bias,see, among others, Krogstrup and Wyplosz (2010), Wren-Lewis(2010), Reinhart and Rogoff (2010) and Rogoff and Bertelsmann(2010).

groups, which however contribute only a smallshare to the tax revenue required for financingsuch expenditure. The government and theparliament should normally understand thatsuch a policy generates problems, but in prac-tice, and to the extent they are influenced bythese groups, they give in and raise expenditureand/or reduce taxes, leading to the creation ofexcessive fiscal deficits.

Another distortion that may lead to a lack offiscal discipline stems from the fact that thetime horizon within which a government oper-ates is much shorter than that of the electorate.Elected representatives are mainly interestedin their re-election and thus in the conse-quences the policies they adopt may have onlyduring their term in power. Based on such anapproach, their consideration for the impact oftheir policies in the more distant future is less-ened, because what matters for them most istheir re-election at the next elections. If theyfeel that they can secure political gains by cut-ting taxes or increasing public expenditure inthe present, and, at the same time, that theywill not have to bear the costs eventuallyincurred in the future due to the ensuingexpansion of the deficit, the result will alwaysbe the creation of deficits, probably higherthan what the citizens would have approved(see Tabellini and Alesina, 1990, Rogoff,1990). Using a much more complex model thatinvolves two political parties with dissimilarpreferences as regards public goods and thetotal size of the public sector, Persson andSvensson (1989) arrive at the conclusion thatboth parties adopt a myopic behaviour, result-ing in higher deficits, exactly because theyassume that they are not going to stay in powerforever.

According to other theoretical approaches,elected representatives seek to maximise thestate budget, while voters lack the necessaryknowledge and information that would enablethem to effectively assess public policies. Lackof transparency in the pursuit of fiscal policyhinders the exercise of democratic control, andthis leaves room for excessive increases in

deficits and debt. Lack of transparency has sev-eral dimensions: insufficient informationregarding the various state budget aggregates,an absence of clear rules and accountabilitywith respect to the preparation of the statebudget and its implementation and ambiguityof the macroeconomic projections, on whichfiscal policy decision-making relies. These dif-ferent dimensions of a lack of transparency areinterrelated. For instance, according to vonHagen (2009), the biased fiscal and macro-economic forecasts in EU countries seem to bedirectly associated with the process of prepar-ing and implementing the state budget. Single-party governments that enjoy a strong major-ity and full power in preparing and imple-menting the budget tend to be optimisticallybiased when making such forecasts.

Particularly in the case of monetary unions,assigning the right of monetary policy conductto a supranational body reduces governments’incentive to exercise fiscal discipline. When acountry is responsible for its own monetarypolicy, the unpleasant likelihood of increasedpublic debt ultimately leading to higher inter-est rates and inflation in the future can promptgovernments to exercise a certain level of self-discipline. But within a monetary union, thisdiscipline is indirect, while at the same timeexpectations are fostered that, in case seriousfiscal problems appear for a country, supportfrom other countries of the union would bevery likely (see Beetsma and Bovenberg, 1999).

In the case of Greece, its high fiscal deficits canbe explained in several ways. The most impor-tant among them is the governments’ inabilityto control public expenditure and collect taxrevenue. But the fundamental cause has to besought in the weak institutional frameworkthat governs the operation of the public sector.The mechanisms available for budget prepa-ration and implementation and for the opera-tion of tax administration are outdated andinadequate (see Rapanos, 2007). Moreover,budgets rely on typically optimistic forecastsregarding key macroeconomic aggregates(GDP, inflation, etc.), and thus forecasts

33Economic Bulletin

May 2010 9

regarding tax revenue often end up being over-optimistic. If one adds to the above the absenceof any mechanisms for systematically moni-toring and assessing budget implementation, itbecomes clear that deficit creation is hard toavoid.5 Another element aggravating the situ-ation is that the release of fiscal data is usuallyfragmented, as well as greatly delayed, so thatultimately the citizens, the media and evenmembers of parliament themselves are not ina position to assess how budget implementa-tion is progressing, or to identify any possibledeviations from the targets set. It is obviousthat this situation runs contrary not only to thebasic principles of modern fiscal management,but also to the democratic principles of trans-parency and accountability.

A factor that could theoretically reduce thisdeficit bias is the reaction of financial markets.A downgrading of the credit rating of a coun-try that exhibits high fiscal deficits over a num-ber of years and a growing public debt leads toa deterioration in its borrowing terms.Financing the deficits and refinancing the pub-lic debt is then only achievable through thepayment of higher interest rates, a fact entail-ing a cost for the government: increased expen-diture for public debt servicing and reducedfunds available for other public expenditure. Inthe worst case, borrowing costs rise to a pointwhere they necessitate a fast and abrupt fiscaladjustment, with a negative impact on growthand employment.

But, in practice, financial markets do notappear particularly consistent or effective inplaying their disciplining role. Markets punishfiscal indiscipline in an inconsistent manner,and often only when it has already grown toparticularly large proportions. Empiricalstudies that attempt to identify the determi-nants of real interest rates in developed coun-tries conclude that the effect of fiscal variablesis in most cases rather small and uncertain.Therefore, the political cost that markets indi-rectly create for a country as a response to itsfiscal indiscipline is not high enough to deterthe latter in a consistent and systematic man-

ner (see Gale and Orszag, 2003, Balassone etal., 2004).6

To deal with deficit creation and increaseddebt, a growing number of countries have grad-ually adopted fiscal rules aimed at limiting oreliminating their deficits. This implies theimposition of a numerical limit on some keyfiscal variable, such as the fiscal deficit, pub-lic debt, public expenditure, etc. (see Kopitsand Symansky, 1998). Within EMU forinstance, the Stability and Growth Pact sets 3%of GDP as a maximum limit for the generalgovernment deficit, and the achievement of abalanced budget as a medium-term objective.Fiscal rules can indeed be very effective, par-ticularly when they provide for specific sanc-tions in case of non-compliance (see de Haanet al., 2004, Wierts, 2008). The MaastrichtTreaty and the Stability and Growth Pact didplay a catalytic role in reducing the deficits ofcountries that wanted to ensure their partici-pation in the euro area.

However, fiscal rules also exhibit some majorweaknesses. The first is that they encouragethe adoption of pro-cyclical (rather thancounter-cyclical) policies. Particularly in thecase of rules that set numerical limits to thedeficit (or the debt), such limits usually riskbeing breached in an economic downturn.Therefore, the rule generates incentives foradopting a restrictive fiscal policy when theeconomic juncture would demand the exactopposite. Correspondingly, against a backdropof robust growth, the fiscal rule is relativelyeasily respected, and thus there is no motivefor fiscal adjustment.

Fiscal rules have also been criticised for theadditional reason that they do not contributeto an improved quality of fiscal policy. Since


55 Macroeconomic projections are prepared by the Ministry ofFinance and in recent years have systematically proved to be exces-sively optimistic.

66 Recent developments in Greece, with the highly increased bor-rowing costs, seem to cast doubt on the above views. However, oneshould bear in mind that this development came immediately aftera major world economic crisis and while the fiscal deficits of majoreconomies have soared to unprecedented levels.

their sole objective is to reduce e.g. the deficit,the means by which fiscal adjustment might beachieved may include the reduction of pro-ductive public expenditure (e.g. cuts in thepublic investment budget). Finally, if fiscalrules do not reflect a broader political will forfiscal discipline, they may encourage the use ofcreative accounting in the production of fiscaldata, reduce the transparency with which thestate budget is implemented, and ultimatelyweaken democratic control of the budget (seevon Hagen and Wolff, 2006).

This problem is intensified by the fact thatcompliance with a fiscal rule usually entails theadoption of policies that will put the targetedfiscal variable on a future trajectory (e.g.achievement of a balanced budget within aspecified time horizon). But the forecastsregarding the achievement of this target areusually produced under conditions of uncer-tainty. Hence, to achieve the target with theleast effort required, governments adopt opti-mistic projections as regards the evolution ofkey macroeconomic aggregates and budgetrevenue. The recent experience of countriessubject to fiscal rules has shown that govern-ments indeed tend to be biased with respect tosuch projections, as they consistently adoptover-optimistic scenarios about the economy’sdevelopment (see Jonung and Larch, 2004 and2006).

Overall therefore, fiscal rules, althoughextensively applied, entail considerable prob-lems. According to Wyplosz (2005), rules “tendto be rigid and artificial (arbitrary limits set tothe debt or the deficit; ‘golden’ rules that relyon aggregates that can be misrepresented), sothat in the end their application is impossibleto justify in the eyes of public opinion”. Withinthe euro area, Wyplosz’s view has some merit.There is no doubt that the fiscal rules imposedby the Stability and Growth Pact contributedto the containment of deficits among euro areacountries. However, strict adherence to theserules proved impossible when EU economiesentered periods of recession or prolongedweak growth (well before the emergence of the

recent financial crisis). Yet the dilemma exists,however, in that any degree of flexibility is onlyachievable through increased complexity,something that reduces the rule’s functional-ity (see Anderson and Minarik, 2006).

3 THE ROLE OF INDEPENDENT FISCALAUTHORITIES

In the wake of the problems analysed above, inrecent years all the more emphasis has beenplaced at the global level on the complementaryrole that independent fiscal authorities (orcouncils) can play in a country’s fiscal consol-idation. This role could theoretically involve theanalysis and assessment of fiscal policy and/ora partial assignment of its conduct. The theo-retical literature has examined two types of fis-cal authorities, distinguished based on whetherthe conduct of fiscal policy is partially assignedby the elected representatives to the inde-pendent authority or not. Discussions over theestablishment of independent fiscal authoritiesempowered to make decisions that would bebinding for the government are of great theo-retical interest, as they focus on whether suchan authority could operate according to thestandards of independent monetary authorities,such as the central banks, that exist in practi-cally all developed countries.7

The present study focuses on the possiblerole of fiscal councils generally entrusted withthe analysis and assessment of fiscal policy,while the fiscal policy decision-making andimplementation competences remain with theparliament and/or the government. After all,this is the type of fiscal councils that operatein many countries today. Relying as much onthe relevant theoretical literature as on theexperience of countries that have made bestuse of the institution at issue, we will nextexamine the main activities of these councils,as well as the requirements of their success-ful operation.

33Economic Bulletin

May 2010 11

77 For a more detailed presentation of relevant theorisations, seeDebrun et al. (2009) and Wyplosz (2008).

Although in practice the type and scope of thecompetences entrusted to these fiscal councilsdiffer considerably across countries, one can dis-cern three basic areas for which such councilsare responsible (see von Hagen, 2010, Ander-son, 2009, European Commission, 2006):

1. They assess the reliability of the macroeco-nomic projections on which forecasts aboutbudget aggregates are based. Fiscal policyplanning relies on a series of assumptionsregarding the future course of the economy.Therefore, successful budget implementa-tion presupposes the adoption of realisticand prudent macroeconomic forecasts. Infact, in some countries fiscal councils aremandated to produce independent macro-economic forecasts, which are binding forthe government in the budget preparationprocess. Such are, e.g., the cases of theWIFO in Austria, the Federal PlanningBureau (FPB) in Belgium, and the CentralPlanning Bureau (CPB) in the Netherlands.

2. They analyse and assess the forecasts onpublic revenue and expenditure, and high-light possible risks of deviation from the tar-gets set in the budget. In several countries―see e.g. Calmfors (2010) for Sweden andChote et al. (2010) for the UK― fiscal coun-cils check whether the policy and the meas-ures announced by the government are com-patible with the fiscal targets the govern-ment itself has set (e.g. the deficit thatappears in the Stability and Growth Pro-gramme).

3. They supervise the budget implementationprocess throughout the year and provide rel-evant information and statistical data at reg-ular intervals.

Through its above activities and functions, afiscal council enhances the degree of trans-parency and accountability in the conduct offiscal policy and contributes to its improvedreliability. It thus ensures better informationfor parliament members and citizens alike,assesses the pragmatism of suggested policies,

and systematically monitors the progress oftheir implementation. Any deviations from thetargets set are promptly identified, whichentails a higher political cost for the govern-ment in case of any inconsistency in its poli-cies. In addition to these basic functions, fis-cal councils ―depending on their availableresources and personnel― engage in otheractivities as well, such as assessing the pro-posed policies over a short- and medium-termhorizon, examining the long-term sustain-ability of public finances, analysing the con-sequences of the state-imposed institutionaland regulatory frameworks on the operationof markets, analysing the impact of proposedtax reforms, etc.

The scope and type of their activities varyacross countries, and are defined depending oneach country’s institutional framework, tradi-tions, historical evolution, particularities, andchallenges to be met. It would be interesting tocite a few examples. The oldest fiscal councilis the Central Planning Bureau (CPB) in theNetherlands, which was established in 1947and numbers approximately 150 staff members.The CPB’s activities cover almost all the fieldsmentioned above, such as producing short- andmedium-term forecasts for key macroeconomicaggregates, analysing the institutions in certainsectors of the economy, conducting cost/ben-efit analyses regarding proposed publicinvestment projects, assessing the fiscal poli-cies pursued by the government, etc. (see Bosand Teulings, 2010). Its reports and studies arewidely accepted and constitute points of ref-erence in public debates. One particularity ofthe CPB is that it also carries out estimates ofthe fiscal cost entailed by the pre-electionpromises of political parties. These estimatesmeet with such broad acceptance that citizensliterally demand that politicians accompanyany pre-election promise with the respectiveCPB estimate as regards its fiscal cost.

The fiscal council with the highest number ofstaff members (235 people) is the CongressionalBudget Office (CBO) in the US, an independ-ent authority that reports to the US Congress.


The Congress is called upon to decide on theadoption or not of proposals submitted by theUS President. The main role of the CBO is toprovide analysis and information on all, pro-posed or alternative, policies describing theadvantages and disadvantages of each, while italso carries out studies on various issues of fis-cal interest. Its primary objective, therefore, isto improve the quantity and quality of the infor-mation on which Congress members will rely tomake decisions, without however making sug-gestions as to what they would ultimately needto vote for.

In the case of Belgium, the country’s conver-sion into a federal state in 1989 was accompa-nied by a transfer of the powers related toexpenditure programmes to local governments,while the responsibility for revenue collectionremained with the central administration. Thisfact gave rise to reasonable concerns about thefiscal problems that may emerge as a result ofany lack of coordination among the various lev-els of government. Since then, the Belgian fis-cal council (the High Council of Finance) hasbeen playing an important coordinating role,setting medium-term targets for the budgets ofboth central government and the regions.These targets form the basis of negotiationsbetween central and local government, in orderto arrive at an agreement concerning theannual target for the fiscal deficit/surplus ofeach level of government.

In Chile, the role of the independent advisorycommittee (ACRCP) is to provide the Ministryof Finance with, among other things, forecastson the potential level of world copper prices,on which a significant part of the country’spublic revenue depends. Similar councils havelately been established in other countries aswell, e.g. Sweden and Hungary, while oneshould not overlook the existence of such acouncil operating under the Canadian FederalParliament, modelled along the standards ofthe Congressional Budget Office in the US.

The above brief and quite selective reviewpoints to the conclusion that the independent

fiscal councils currently active in various coun-tries are, according to Wyplosz’s (2008) ter-minology, “soft” fiscal policy councils, in thesense that their role is mainly advisory. Buthow can one ensure the effectiveness of thistype of institution in influencing the fiscal pol-icy pursued, if it lacks the power to make gov-ernment-binding decisions? As internationalexperience shows, even a soft council can playan important role, provided that political par-ties and society in general accept its studiesand view them as a product of scientific analy-sis and not according to whether they agreewith their own ideological or political stand-points.

4 IS GREECE IN NEED OF A FISCAL POLICYCOUNCIL?

This question is vital for Greece, consideringthe size of fiscal imbalances the country cur-rently faces. Indeed, the government recentlyannounced its intention to establish an inde-pendent fiscal council that will report to Par-liament. As mentioned above, the experienceof other countries has shown that such a coun-cil may prove to be useful, given that its exis-tence enhances transparency in the conduct offiscal policy and therefore increases the polit-ical cost of ‘bad’ or inconsistent policies. In acountry such as Greece, where a critical ele-ment with respect to fiscal policy is the relia-bility of data included in the budget, the exis-tence of a body such as ―let’s name it― theParliamentary Budget Office, may contributesubstantially to the improved quality andprompt release of fiscal aggregates. Moreover,by carrying out analyses of the reliability of thegovernment’s macroeconomic forecasts, eval-uating the forecasts related to budget revenueand expenditure and monitoring budget imple-mentation, such a council will provideextremely useful information to the politicalparties and Members of Parliament, enablingthem to better carry out their work.

It is known that the reliability of Greece’s fis-cal statistics is quite low, a fact attributable as

33Economic Bulletin

May 2010 13

much to the poor quality of the data as to theirtardy release. The existence of such a councilthat will report to Parliament can exert con-siderable pressure on the executive branch,both to promptly provide data on the processof implementing the budget and to improvetheir quality. But to be able to play this role,the Parliamentary Budget Office, first of all,has to gradually and methodically build its rep-utation – a task not easy to accomplish.

The credibility of a fiscal council stems from itsindependence from government, political par-ties and other interest groups. Therefore, itmust be above all political affiliations. AsAnderson (2009) states, such a council has tobe “non-partisan”, which does not mean thatit should be “bi-partisan”. The former meansnot affiliated to any party (apolitical),whereas the latter means affiliated to and/orinfluenced by the two major (or all) parties. Itis imperative for its members to be selectedbased on merit and professional skills, and notwith a view to striking a political balance by let-ting all parties have a representative in thecouncil. In the case of Greece, it would per-haps be advisable to keep the number of coun-cil members relatively low (e.g. three) so as todissuade negotiations between the politicalparties for appointing persons somehow affil-iated to each of them. If the composition of thecouncil is the result of political arrangements,its effectiveness and credibility will be under-mined ab initio (see Rapanos, 2010). Theabsence of any political identity is a key char-acteristic of all fiscal councils that successfullyplay their role (see Steuerle and Rennane,2010, Bos and Teulings, 2010).

Another parameter that contributes to theactual independence of a fiscal council relatesto the duration of its members’ mandate. It isadvisable for this duration to exceed the gov-ernment’s term of office. In several countriesit has been set at 5 years, and council memberscan only be replaced in case of a serious breachof their duties. Their dates of appointmentcould also be different, so as to ensure thatmemberships do not expire simultaneously.

Another question that arises with respect to thedesign of such a council relates to its reportingline and its financing. The practice followed inother countries is not universal. In some coun-tries (e.g. the US and Canada) the councilreports to parliament (Congress); in others itoperates under the Ministry of Finance (e.g. theNetherlands, Chile) or reports to the govern-ment (Sweden) or is a totally independentauthority. In the case of countries with power-ful single-party governments that enjoyabsolute control over budget preparation andimplementation, the subordination of an inde-pendent fiscal council to parliament is a ratherappropriate choice (see Anderson, 2009,Schmidt-Hebbel, 2010).

In the case of Greece, according to the Con-stitution (Article 79), the Parliament votes toapprove the budget of central government rev-enue and expenditure for the following year,and, in fact, broken down by ministry. There-after, however, the Parliament has no mecha-nism available for monitoring the implemen-tation of the budget it has approved. There-fore, although it has an important institutionalrole, in practice the Parliament is unable toeffectively play it. A fiscal council reporting toParliament would therefore enhance the Par-liament’s role in the budget implementationprocess, and, by way of releasing reports andconducting open public debates, it would putpressure on the government so as to deter devi-ations from the budget targets. Hence, theintention of the Greek government to establisha fiscal council under Parliament appears to bethe right choice, as this would strike a betterbalance between the legislative and the exec-utive branches.

As regards the financing of the council, inother countries this is usually covered by gov-ernment or parliamentary funds, while a smallfraction is offered by the private sector. How-ever, financing an authority with funds fromthe government, the policies of which thisauthority is called to transparently and objec-tively assess, could potentially be a problem.An indicative recent example is that of the Par-


liamentary Budget Office (PBO) in Canada,established as recently as in 2008. The releaseof a council report that assessed the cost of theCanadian engagement in the Afghanistan warresulted in a considerable reduction of thefunds granted to the council for the followingyear. But even so, such actions obviously entaila huge political cost for the government,exactly because they are widely publicised, andare thus rarely attempted. Nevertheless, in thecase of Greece, financing the council withfunds from the Parliament’s own budget(approved only by the Parliament itself and notby the Ministry of Finance), is perhaps themost appropriate choice.

A connected issue relates to the recruitment ofspecialised and properly trained personnel forthe council. The number of people supportingthe work of fiscal councils varies significantlyacross countries, from 4 people in Sweden toroughly 250 people in the US (CBO), and thescope of activities these councils can undertakeis of course proportional to that number. Ini-tially, such a council could hire a small num-ber of experts with a background in fiscal andmacroeconomic analysis, and use, even underpart-time employment contracts, members ofthe academia. In Greece, recruiting the appro-priate scientific personnel is usually a time-consuming and costly process. For this reason,until it finally hires its own personnel, thecouncil could make best use of the humanresources available in certain other generalgovernment institutions ―such as the Centreof Planning and Economic Research(KEPE)― to provide technical support onsome fiscal issues at least (e.g. conduct macro-economic analyses and assess the consistencyof official forecasts, etc.). With respect to itsinspectional role, the council should not onlyemploy a small number of specialised experts,but should also enjoy legally secured access todata from the State General Accounting Officeand other services of various Ministries andauthorities of the broader public sector.

In light of the above, it becomes evident thatthe creation of a Fiscal Policy Council report-

ing to Parliament could contribute consider-ably to Greece’s efforts to consolidate its pub-lic finances. To deal with the current fiscal cri-sis, which has recently turned into a gener-alised economic crisis, Greece needs not onlya strict fiscal policy, but also major changes inits institutional framework for conducting fis-cal policy. One such institutional change is thecreation of such a council.

In conclusion, the experience of countries withactive independent fiscal councils shows theimmense importance of allowing such a coun-cil to build up the reputation that its assess-ments are reliable, and to convince the marketsand the public that it is unbiased and trulyindependent of any government, political partyor other interest group. In the case of Greece,an independent fiscal council could become apoint of reference, not only for Members ofParliament but also for economic analysts oreven international credit rating agencies, and―together with the Bank of Greece― couldconstitute a valid source of information on thecountry’s fiscal developments. One of themajor problems of Greece today is the lack ofcredibility of its fiscal policy and the poor qual-ity of the statistical data related to its publicfinances. Restoring such credibility is a longprocess, in which an independent fiscal coun-cil could play a catalytic role.

However, regardless of how good it is consid-ered to be or how successfully it has functionedin other countries, no institution can serve asa magic bullet (see Debrun et al., 2009, vonHagen, 2010). The ability of an independentfiscal council to effectively fulfil its taskdepends on the prevailing political will and thesupport it receives from all political parties, aswell as on the eagerness of the government andits authorities to cooperate, by providing sta-tistical data and other information to the coun-cil even if the supply of such data does not con-stitute a formal legal obligation. An inde-pendent fiscal council can contribute to theimprovement of a country’s fiscal performanceonly in the case the government itself sharesthe opinion that there is indeed a fiscal disci-

33Economic Bulletin

May 2010 15

pline problem that has to be dealt with. It is notcoincidental that many countries have estab-lished independent fiscal councils preciselyduring or right after a serious fiscal crisis. Asregards Greece, the current fiscal crisis is notthe first in its recent history. Based on theabove rationale, the opportunities for strength-ening the fiscal policy framework through insti-tutions such as that of an independent fiscalpolicy council could have already been seizedin the past.

Thus, a question arises whether the existenceof such a council would have helped the coun-try avoid, or limit to some extent, the currentfiscal crisis. Answering this question is notstraightforward. It is well-known that an insti-tution can only operate successfully if its cre-ators respect its terms of operation and all oth-ers that enjoy its services consider it to beimpartial, reliable and effective. Political, eco-nomic and other institutions do not operate ina vacuum, but within a certain political, eco-nomic and social arena, which both affects andis affected by their operation. In our opiniontherefore, even if a council existed and wasreliable, and the political system had investedin its impartiality and reliability, the council byitself would not have averted the crisis. But themembers of our political system and the pub-lic would have been provided with much moreaccurate and promptly released informationregarding the dramatic deterioration of thecountry’s fiscal aggregates, and this would haveled them to realise and respond to the problemmuch earlier.

5 CONCLUSIONS

The sustainability of a country’s fiscal positionis a prerequisite for sound growth. Accordingto a widely accepted definition, a country’s fis-cal position is sustainable when the existinginstitutions (tax system, social security andpublic healthcare system, etc.) can be sustainedover a long-term horizon without generating anuncontrolled increase in public debt. In theinternational literature, the problems created

by a high level of debt mainly refer to thefuture. As Musgrave (1988) wrote, “debt cre-ation allows the generations that live now totake resources from the generations that willlive in the future”. But as experience hasrecently proved once again, a high public debtcan also create considerable problems at thepresent time (see Reinhart and Rogoff, 2010).

Indeed in the case of Greece, what latelybecame painfully evident is that a high publicdebt is not only a problem for future genera-tions, but may even cause a chain of problemsin the economy and put at risk the country’sgrowth trajectory in the present as well. In thatsense, the systematic lack of fiscal disciplinethat leads to such situations constitutes a fail-ure of the political system (political failure),hence, one way to support its redress could bethrough reforms of the institutional frameworkthat governs the preparation and implemen-tation of the general government budget.

In the course of the past few months, Greecehas entered a profound fiscal crisis. In order todeal with this crisis, and following the dictatesof the international organisations monitoringthe Greek economy, the government hasadopted an ambitious programme of strict fis-cal adjustment. The Greek government’s com-mitment to reduce the general governmentdeficit from 13.6% of GDP to below 3% withina period of 3 to 4 years represents the greatestadjustment ever attempted by a euro areacountry. The target is ambitious, but feasible.If Greece wants to avoid reliving the history ofthe 1990s, when the period of fiscal adjustmentwas followed by a relaxation of fiscal policy andan ensuing expansion of deficits with devas-tating results, it needs to resolvedly moveahead with the reform of the institutionalframework governing its fiscal policy as well.In this direction, the establishment of an inde-pendent fiscal council could offer a key ele-ment of credibility. But one should not remainunder the false impression that such an estab-lishment is going to solve the fiscal problem.This council could play an important but onlycomplementary role in the overall effort to rad-


ically reshape the broader institutional frame-work for fiscal policy.

Decisions regarding the conduct of fiscal pol-icy in a country inescapably involve value judg-ments related to a series of questions, such asthe importance of individual liberty and free-dom of choice, solidarity with other membersof society, compassion for the disadvantaged inneed of assistance, people’s right to be com-pensated in the future in exchange for effortsmade in the present, responsibility towards andcare for future generations (see Kornai, 2010),etc. In this sense, such decisions are political

in nature, and it is not the role of an inde-pendent fiscal council to act as a substitute tothe executive. The potential value added ofsuch a council resides in its ability to analysethe short- and long-term implications ofadopted or alternative policy choices, to verifytheir consistency, and ultimately to enhancetransparency in the conduct of fiscal policy, dis-seminating information that is accessible to all,rather than to a closed circle of technocrats. Itcan thus strengthen the exercise of democraticcontrol over the executive and the latter’s obli-gation for accountability, which is a funda-mental principle of democracy.

33Economic Bulletin

May 2010 17

Anderson, B. and J.J. Minarik (2006), “Design choices for fiscal policy rules”, OECD Journalon Budgeting, Vol. 5(4), 159-208.

Anderson, B. (2009), “The changing role of parliament in the budget process”, OECD Journalon Budgeting, 2009/1, 1-11.

Balassone, F., D. Franco and R. Giordano (2004), “Market-induced fiscal discipline: is there afall-back solution for rule failure?”, Proceedings of Banca d’Italia Public Finance Workshop,Rome: Banca d’Italia.

Bank of Greece (2004), Annual Report 2003, Athens.Beetsma, R.M.W.J. and A.L. Bovenberg (1999), “Does monetary unification lead to excessive

debt accumulation?”, Journal of Public Economics, 74, 299-325.Bos, F. and C. Teulings (2010), “Lessons from the Netherlands”, paper presented at the Con-

ference on Independent Fiscal Institutions, Fiscal Council, Republic of Hungary, Budapest,March 2010.

Calmfors, L. (2003), “Fiscal policy to stabilise the domestic economy in the EMU: What can welearn from monetary policy?”, CESifo Economic Studies, 49, 319-53.

Camlfors, L. (2010), “Lessons from Sweden”, paper presented at the Conference on Independ-ent Fiscal Institutions, Fiscal Council, Republic of Hungary, Budapest, March 2010.

Chote, R., C. Emmerson and J. Shaw (eds) (2010), The IFS Green Budget: February 2010, Insti-tute for Fiscal Studies, UK.

de Haan, J., H. Berger and D. Jansen (2004), “Why has the Stability and Growth Pact failed?”International Finance, 7, 235-60.

Debrun, X., D. Hauner and M.S. Kumar (2009), “Independent fiscal agencies”, Journal of Eco-nomic Surveys, Vol. 23, No. 1, 44-81.

Eichengreen, B., R. Hausman and J. von Hagen (1999), “Reforming budgetary institutions inLatin America: the case for a national fiscal council”, Open Economies Review, 10, 415-442.

European Commission (2006), Directorate General, Economic & Financial Affairs, Brussels,04/04/2006, “Survey on Fiscal Rules and Institutions: Summary of First Results”, CFIN/LM-JAC/REP52126/06-EN.

European Commission (2009), Public Finances in EMU 2009, Brussels.Gale, W.G. and P.R. Orszag (2003), “Economic effects of sustained fiscal deficits”, National Tax

Journal, 56, 463-85.Hardin, G. (1968), “The tragedy of the commons”, Science, 162, 1243-48.Jonung, L. and M. Larch (2004), “Improving fiscal policy in the EU: the case for independent

fiscal forecasts”, Economic Papers no. 210, European Commission.Jonung, L. and M. Larch (2006), “Fiscal Policy in the EU: Are Official Output Forecasts Biased?”

Economic Policy, July, 491-534.Kopits, G. and S. Symansky (1998), “Fiscal Policy Rules”, International Monetary Fund, Occa-

sional Paper, No. 162, IMF, Washington D.C.Kornai, J. (2010), “Introductory remarks”, paper presented at the Conference on Independent

Fiscal Institutions, Fiscal Council, Republic of Hungary, Budapest, March 2010.Krogstrup, S. and C. Wyplosz (2006), “A common pool theory of deficit bias correction”, CEPR

Discussion Paper 5866.Krogstrup, S. and C. Wyplosz (2010), “A common pool theory of supranational deficit ceilings”,

European Economic Review, 54, 269-78.Mueller, D.C. (2003), Public Choice III, Cambridge University Press, U.K.Musgrave, R. (1988), “Public debt and intergenerational equity”, in Arrow, K. and M. Boskin

(eds) The economics of public debt, MacMillan, London.OECD (2007), Economic Outlook, 2007/2, 82, Paris.


R E F E R ENC E S

Persson, T. and L. Svensson (1989), “Why a stubborn conservative would run a deficit: policywith time-inconsistent preferences,” Quarterly Journal of Economics, 104(2), 325-46.

Rapanos, V. (2007), “Preparation and implementation of the state budget: European experienceand Greek reality”, Foundation for Economic and Industrial Research, Athens [in Greek].

Rapanos, V. (2010), “Greek Public Finances and the possible role of a Parliamentary BudgetOffice”, paper presented at the Conference on Independent Fiscal Institutions, Fiscal Coun-cil, Republic of Hungary, Budapest, March 2010.

Reinhart, C.M. and K.S. Rogoff (2010), “Growth in a time of debt”, CEPR Discussion PaperDP7661.

Rogoff, K. (1990), “Equilibrium Political Business Cycles”, American Economic Review, 80,21-36.

Rogoff, K. and J. Bertelesmann (2010), “The rationale for fiscal policy councils: theory and evi-dence”, paper presented at the Conference on Independent Fiscal Institutions, Fiscal Coun-cil, Republic of Hungary, Budapest, March 2010.

Schmidt-Hebbel, K. (2010), “Lessons from Chile”, paper presented at the Conference on Inde-pendent Fiscal Institutions, Fiscal Council, Republic of Hungary, Budapest, March 2010.

Shepsle, K. and B. Weingast (1981), “Political Preferences for the Pork Barrel: A Generaliza-tion”, American Journal of Political Science, 25(1), 96-111.

Steuerle, E.C. and S. Rennane (2010), “Lessons from the United States”, paper presented at theConference on Independent Fiscal Institutions, Fiscal Council, Republic of Hungary, Budapest,March 2010.

Tabellini, G. and A. Alesina (1990), “Voting on the Budget Deficit”, American Economic Review,80, 37-49.

Van Riet, A. (2010), “Euro area fiscal policies and the crisis”, ECB Occasional Paper, No. 109,April.

von Hagen, J. (2009), “Sticking to fiscal plans: the role of fiscal forecasts”, Public Choice (forth-coming).

von Hagen, J. (2010), “The scope and limits of fiscal councils”, paper presented at the Confer-ence on Independent Fiscal Institutions, Fiscal Council, Republic of Hungary, Budapest, March2010.

von Hagen, J. and I. Harden (1995), “Budget processes and commitment to fiscal discipline”,European Economic Review, 39, 771-79.

von Hagen, J. and G. Wolff (2006), “What do deficits tell us about debts? Empirical evidenceon creative accounting with fiscal rules”, Journal of Banking and Finance, 30, 3259-79.

Wierts, P. (2008), “Fiscal rules and fiscal outcomes in EMU: theory and evidence”, Thesis sub-mitted for the Degree of Doctor of Philosophy in Economics, The Centre for Euro-Asian Stud-ies, Department of Economics, University of Reading.

Wren-Lewis, S. (2010), “Comparing the delegation of monetary and fiscal policy,” paper pre-sented at the Conference on Independent Fiscal Institutions, Fiscal Council, Republic of Hun-gary, Budapest, March 2010.

Wyplosz, C. (2005), “Fiscal policy: institutions versus rules”, National Institute Economic Review,191, 70-84.

Wyplosz, C. (2008), “Fiscal policy councils: unlovable or just unloved?”, Swedish Economic Pol-icy Review, 15, 173-92.

33Economic Bulletin

May 2010 19

33Economic Bulletin

May 2010 21

SUMMARY

The problem of unemployment, especiallyyouth unemployment, is a central issue of pub-lic discourse in Greece. Youth unemploymentrates in Greece are among the highest acrossthe EU and, in contrast with other countries,a significant proportion of young unemployedpersons are tertiary education graduates. Thisstudy aims to provide a detailed analysis of thedeterminants of unemployment in Greece, withan emphasis on the variables related to theeducational qualifications of the labour force,using microdata from the Labour Force Sur-veys conducted by the National Statistical Serv-ice of Greece (NSSG) between 2004 and 2007.

The results of this study imply that the discussionabout youth unemployment is slightly misplaced.The problem is not quite one of youth unem-ployment as such, but a problem of transitionfrom education to the labour market, irrespec-tive of age. It also involves graduates of all edu-cation levels, not only tertiary education. Thedifference between tertiary education graduatesand graduates of lower education levels is thatthe unemployment rates of the first tend todecline at an acceptable level a few years aftergraduation, while for the latter the pace of thisdecline is substantially slower and unemploy-ment rates usually converge to higher levels. Fur-thermore, across the entire spectrum of tertiaryeducation (Technological Educational Institutes– TEI, Universities – AEI, postgraduate studies),the higher the level of education the lower thelong-term unemployment rate, although signif-icant variations can be observed in each educa-tion level. In fact, certain groups of tertiary edu-

cation graduates (Law school or IT graduates),at least men, face no real unemployment prob-lems after their graduation. Some other groups,however, run a high risk of unemployment butonly for a few years after graduation (graduatesfrom Schools of Physical Sciences, Mathematics& Statistics) and yet others face serious unem-ployment problems for several years after grad-uation (Physical Education & Sports, Social Sci-ences and several TEI graduates). Finally,ceteris paribus, women in general, and femaletertiary education graduates in particular, facea significantly higher probability of unemploy-ment compared to men with similar educationalqualifications. In some groups of female tertiaryeducation graduates, the estimated unemploy-ment rates are exceptionally high, even severalyears after graduation (university graduates fromHorticulture & Forestry, TEI graduates from theschools of Agriculture & Food Technology andEconomics & Management).

The results of this study reflect the conditionsin the labour market between 2004 and 2007;thus it would not be advisable to extrapolatethem to the future. This is due to the fact thatin the past fifteen years tertiary education inGreece, like in most OECD countries, hasexpanded very rapidly and the share of tertiaryeducation graduates in recent cohorts is sub-stantially higher than in previous ones. More-

DE T ERM INANT S O F YOUTH UNEMPLOYMENTI N GRE E C E W I TH AN EMPHA S I S ON T ER T I A R YEDUCA T I ON GRADUA T E S *

Theodoros M. MitrakosBank of GreeceEconomic Research Department

Panos TsakloglouDepartment of International and European Economic StudiesAthens University of Economics and Business

Ioannis CholezasCentre for Planning and Economic Research

** The views expressed in this study do not necessarily reflect thoseof the Bank of Greece. The authors would like to thank H. Gib-son, I. Kalogirou, E. Kikilias, D. Nicolitsas and G. Psacharopou-los for their useful comments. The present paper was supported bythe cultural project “Protovoulia” (Education and DevelopmentInitiative). An earlier version formed part of a broader action ofProtovoulia (“Education and Development: Connecting Educationand Employment”) and its results were presented in a special one-day conference held at Athens Goethe Institute in July 2009.

over, the recent financial crisis has broughtabout significant changes to employment andunemployment, as well as to the real economyboth at the national and the international level.Whether the increase in graduate labour supplyin Greece will lead to a more lasting rise in grad-uate employment rates, even after some diffi-culties in the first years of labour market par-ticipation, depends on several factors, such aswhether enterprises will increase their demandfor high-skilled human capital or whether theywill be willing to recruit graduate labour to fillpositions where higher education qualificationsare not required. Nevertheless, the results of thepaper suggest that the high demand of youngindividuals for tertiary education is rational, astertiary education seems to be quite effective inshielding against unemployment, even if only inthe long run. It should also be noted that schoolsassociated with lower unemployment rates areamong the most popular ones.

1 INTRODUCTION

In the last two decades, the problem of unem-ployment, especially youth unemployment, is acentral issue for public discourse in Greece. Therate of unemployment, which stood at 7.4% in1986, peaked to 11.9% in 1999, fell to 8.8% in2006 and further to 7.6% in 2008. Female unem-ployment rates are twice as high as male unem-ployment rates in every year for which data areavailable, while a constantly rising proportion ofthe unemployed are long-term unemployed, i.e.unemployed for more than 12 months. This per-centage started off from 42.5% in 1986 and sta-bilised to over 50% after 1993. In 2006, itreached 56.1%, but dropped to 49.6% in 2008.Finally, both for males and females, there is anegative relationship between age and unem-ployment rate, with unemployment percentagesfor the age groups 15-19 and 20-24 being dis-proportionally higher than those for oldergroups. In most of the years in the period underreview the unemployment rate of the (numeri-cally small) group of women aged 15-19 whoparticipate in the labour market was above 50%,whereas that of women aged 20-24 was con-

stantly higher than 30%. High unemploymentrates are also recorded among females aged 25-34 (14%-22%). Rates for the correspondingmale groups are also high, but not as high as forthe female ones, which range from 14% to 28%for the 15-19 age group, and from 15% to 22%for the 20-24 age group.

It should be clarified that the problem of youthunemployment is shared by all European coun-tries. However, as shown by recent Eurostatdata for the first quarter of 2009, based on theharmonised definition of unemployment, theunemployment rate of persons aged up to 24in Greece was 24.2%, against 18.3% of the EU-27 average. This stems from the very highunemployment rate of young women (31.8%against 19.1% of the EU average); by contrast,the unemployment rate of young men was notmuch higher than the EU average (18.3%against 17.4%).1

Therefore, it is of particular interest ―in termsof both scientific analysis and policymaking―to explore the determinants of unemploymentfor the entire labour force and more specifi-cally of youth unemployment. This can beachieved in two ways: either at the macroeco-nomic level, with a view to identifying the fac-tors that influence the aggregate unemploy-ment rate (and possibly its composition) or atthe microeconomic level, with the purpose ofassessing the determinants of the probability ofunemployment, given the total unemploymentrate. This study adopts the second approachand aims at investigating the determinants ofunemployment, focusing on tertiary educationgraduates and using the micro data fromLabour Force Surveys (LFS) conductedbetween 2004 and 2007.

In the following section, we present a briefoverview of the findings of available empiricalstudies on youth unemployment in Greece.The third section presents the LFSs used in theanalysis, the fourth section sets out the most


11 In the first quarter of 2009, compared with Greece, the only coun-tries which recorded higher unemployment rates for young peopleaged up to 24, were Spain (33.6%) and Italy (24.9%).

important empirical results of this study andthe last one summarises its findings.

2 LITERATURE REVIEW2

Certain features of the Greek labour marketoften account ―more or less― for the highunemployment rates, observed mainly in youngpeople and women. For instance, the Greeklabour market features dualism, i.e. there is anofficial labour market (subject to regula-tions/controls and offering job security) and anunofficial labour market (without job security,often with black labour, bad employment con-ditions and low compensation).3

In the past 25 years, the unemployment of ter-tiary education graduates has risen (4.3% in1981, 6.6% in 1990, 7.6% in 1997), but thisdoes not seem to affect all graduates in thesame way. Thus, according to Karamessini(2003), unemployment among graduates fromphysical education & sports, philosophy, the-ology, horticulture & forestry is higher thanaverage. Graduates from economics, social andpolitical sciences, pedagogics, physical sciencesand mathematics follow next.

In particular as regards high-skilled individu-als, there is some imbalance between supplyand demand (Liagouras et al., 2003, Kat-sanevas and Livanos, 2005), either because ofthe education system failure to adjust to theneeds of the labour market, or because of theGreeks’ insistence to acquire tertiary education(Papailias, 2006). Conversely, ISTAME (2006)considers that the main factor behind the highunemployment rates is the inadequate devel-opment of the economy’s private sector andthe subsequent low demand for highly skilledhuman capital (e.g. tertiary education gradu-ates). According to this study, most new jobsrequire low specialisation and offer low com-pensation in return. The strong demand fortertiary education is often attributable to thecomparatively low unemployment rates of ter-tiary education graduates (Kassimis, 2002,Karamessini, 2003). Keeping its focus on grad-

uates, IOBE (2007) attributes high unem-ployment on the one hand to the inability ofthe education system to adjust to rapid tech-nological progress4 and on the other hand tothe slow adjustment process of a significantportion of the economy’s productive (private)sector. Even graduates themselves consider thelack of adequate information and ―at the per-sonal level― the lack of appropriate recom-mendations as the main reasons behind unem-ployment (IOBE, 2007).

There are relatively few available empiricalstudies on the issue of youth unemployment inGreece; these can be grouped in three cate-gories according to the data they use. The firstconsists of the studies based on the single spe-cialised survey on the transition from educa-tion to the labour market that was carried outby the National Statistical Service of Greece(NSSG), under the auspices of Eurostat, in thesecond half of 2000. The second containsresearch based on field studies on the gradu-ates of specific university schools or faculties.The main advantage of such studies is that theyare targeted and thus, the interviewees supplyall necessary information for the investigationof the issue at hand. On the other side, though,such studies typically involve small subgroupsof youth population (or young graduates), thusproviding an incomplete picture. Finally, thethird category is based on Labour Force Sur-veys compiled by the NSSG, which offer theadvantage of country-wide coverage and a mul-titude of useful information for the analysis ofthe unemployment issues. Most of the avail-able empirical studies are primarily descriptiveand their conclusions are not always consistent.

The results of studies in the first category showthat men’s transition from education to their

33Economic Bulletin

May 2010 23

22 For a detailed review of the literature on the employability of ter-tiary education graduates in Greece, see Cholezas and Tsakloglou(2008). This section provides a brief overview of the main findingsof empirical studies regarding the issue of youth employment andespecially youth unemployment of tertiary education graduates inGreece.

33 See ISTAME (2006).44 According to a study conducted by the Hellenic Federation of

Enterprises – SEV (2004), half of the enterprises reported lack oftertiary education graduates with specific qualifications.

first important job ―defined as a job of at least20 hours per week, held for at least sixmonths― lasts longer compared with women(36 months on average against 29), althoughthis can be attributed to military service(Karamessini and Kornelakis, 2005). More-over, according to the same authors, graduatesspecialised in the provision of services needless time than other graduates to find animportant job. When the sample is limited tothose who graduated (any education level)between 1996 and 1999, then about one thirdof AEI graduates and one fourth of TEI grad-uates had not managed to find an importantjob until the second quarter of 2000. AmongAEI graduates, this percentage is compara-tively higher for the graduates of physics, math-ematics, IT, social sciences, mass communica-tion, humanities and pedagogics (over 40%).All in all, tertiary education graduates need onaverage more than 16 months to find an impor-tant job but, still, they are in a better positionthan those without a degree. Based on this cri-terion, graduates of physics, mathematics, IT,physical education & sports and polytechnicschools hold the worst position between all ter-tiary education graduates (over 18 months). Inrelation to gender, male tertiary educationgraduates seem to find an important job withmore delay than female ones (around 20months compared to 14).

Using the same data, Nicolitsas (2007)attempts to investigate the determinants of theduration of the transition period, applying haz-ard models. The results of her study show thatthe transition period is longer for married men,foreigners, residents in rural areas and oldergraduates, while the period is shorter for per-sons whose father is engaged in some businessactivity. The duration of the transition periodis negatively correlated with local unemploy-ment rate, graduation year and education level.Finally, female graduates of technical schools,TEI, IEK, polytechnic and medical schools finda job faster.

In the second type of research which is based onfield studies, Bitros (2002) analyses the condi-

tions for Economics graduates. According to hisconclusions, slightly more than half of the grad-uates find a job within less than a year after grad-uation. The main justification for this low per-centage, according to the graduates themselves,is the mismatch between education and labourmarket demands, as well as the lack of additionalskills. The transition of engineering schools’graduates, especially graduates from theNational Technical University of Athens(NTUA), to the labour market is the object ofa number of studies; the results of these studiesshow, on the one hand, low unemployment rates(2.2% according to NTUA, 2002) and, on theother hand, short periods of transition to the firstjob (only 5 months after graduation, accordingto NTUA, 2002, and 5.7 months according to theTechnical Chamber of Greece – TEE, 2006).

Another field study forms the basis forKaramessini’s (2006b) analysis, which focuseson graduates from the University of Thessaly.Roughly 5 years after graduation, the unem-ployment rate of that University’s graduates is7%. Ceteris paribus, women, kindergartenteachers, horticulturalists and land planning &regional development engineers face thebiggest problems in entering the labour mar-ket. However, according to the author, thetemporary nature of many graduates’ employ-ment remains a big problem, as 5-8 years aftergraduation 45% of the graduates still have nopermanent job.

The field study conducted by IOBE (2007)addresses both firms (about 200) and employ-ees (about 500), which makes it particularlyinteresting. Results show that over 90% of thegraduates found a job within 20 months aftergraduation – most of them in fields related totheir studies. Conditions seem to be more dif-ficult for graduates of social sciences, biology,horticulture and genetics, as well as for thosewho come from families with poor educationbackground, are younger graduates or post-graduates. Turning to their field of work, ittakes graduates from schools of fine arts andmathematics & statistics longer to find a jobrelevant to their degrees.


The most detailed field study is the one used byKaramessini (2007), which covers a large num-ber of Greek AEI. According to its findings, 5-7 years after graduation 84% of AEI graduatesfound a job, with minor differences betweenmen and women. Lower employment rates areobserved among graduates from schools of biol-ogy, theology, history, archaeology and politicalscience. Regarding unemployment, rates arehigher for women (7% against 5% for men and6.4% on average) and for graduates of history,archaeology and theology. Although 5-7 yearsafter graduation from AEI almost all graduates(97%) have some working experience, 18%have no important working experience and just71% have non-temporary employment (69% ofwomen and 74% of men). As regards the tim-ing of the first important job, large variationsare observed. Thus, 15% found a job beforegraduation, 19% within one month from grad-uation, 33% needed one year and the remain-ing 33% more than one year. These results painta different picture than IOBE (2007), where thepercentage of graduates who found a job withina year was much higher (91% against 67%); thisis possibly due to both the stricter definitionsused in Karamessini (2007) and the fact thatIOBE (2007) excluded unemployed graduatesfrom its sample. However, a really alarmingfinding of Karamessini (2007) is that 41% of thegraduates reported that they remained unem-ployed for more than 12 months after gradua-tion. The picture only gets gloomier for women.

Finally, the studies of Karamessini (2006a) andKaramessini and Prokou (2006) fall under thethird category, i.e. those using LFS data. Inboth studies, emphasis is placed on AEI andTEI graduates. According to their results, TEIgraduates have higher rates of participation inthe labour market in the first years after grad-uation ―probably because proportionallymore AEI graduates undertake further (post-graduate) studies― which are equalised there-after (about 6 years after graduation). Withrespect to gender, men have higher rates ofparticipation if they are TEI graduates, but dif-ferences are negligible among AEI graduates.Conversely, unemployment rates are clearly

higher for TEI graduates and women, while thegap widens as we move further away from grad-uation. These higher unemployment rates areattributed by the authors to the higher partic-ipation rate of TEI graduates in the labourmarket and to discrimination against women.The results show that new graduates face highrates of unemployment after graduation(34.4% for AEI and 36.5% for TEI graduates),which decline with time, especially for AEIgraduates. Also, they have difficulties in solid-ifying themselves to employment, as 32%(29%) of AEI (TEI) graduates still have nopermanent job 6 years after graduation.

3 LABOUR FORCE SURVEY: SHORT DESCRIP-TION AND FIRST DESCRIPTIVE RESULTS

For the purposes of this study, we used the micro-data of quarterly LFSs that the NSSG conductedduring 2004(Q1)-2007(Q3). We chose this periodbecause the methodology of collecting data forthe LFS was radically revised in 2004 and, more-over, for this period micro-data are available bythe NSSG in the form of rotating panels, as eachmember of the sample participates in the LFS forsix consecutive quarters (“waves”).

The LFS has been conducted on a quarterlybasis by the NSSG since 1998 (earlier it wasconducted only in the second quarter of everyyear) with the aim of collecting detailed dataon the employment and unemployment statusof household members aged 15 or over.

More specifically, the LFS aims at:

(a) recording the employment status of house-hold members aged 15 or over, together withgender, age and education background, atnational and regional level;

(b) studying the breakdown of employment bysector of economic activity, profession, work-ing hours etc.;

(c) monitoring the duration of unemployment,in relation to gender, age, education back-

33Economic Bulletin

May 2010 25

ground, region and some features of the latestjob, such as the sector of economic activity andthe profession;

(d) studying the employment condition ofhousehold members a year before, whetheremployed persons have a second job, etc.

In the LFS, the unemployed are definedaccording to the original definition of theInternational Labor Organization (ILO); weused the same definition in our study too. Inparticular, a person is considered unemployedif:

(i) s/he hasn’t worked for even one hour in thereference week;

(ii) s/he’s looking for a job;

(iii) s/he reports the actions s/he did to this endin the last four weeks before the study;

(iv) s/he is ready to start working the followingweek, if s/he finds a job.

The LFS quarterly sample includes approxi-mately 30,000 households (a sampling fractionof 0.85% of the total population of the coun-try), with one sixth of the sample rotating(being replaced) every quarter.

The main features of the LFS used in this studyare depicted in Tables 1 and 2. Table 1 showsparticipation rates of persons aged 15-64, bygender, level of education and breakdown byage group, for the period 2004-2007. The totalparticipation rate follows a slightly upward pathin the period under review. Male participationis significantly higher than female participation(79.0% against 54.5%, on average) and, fur-thermore, there are significant differences interns of education level and age. Female par-ticipation is higher at tertiary educational lev-els, whereas male participation shows a smalldecline between primary and secondary edu-cation and a sharp increase in tertiary educa-tion graduates. Differences between male andfemale labour force participation rates of ter-

tiary education graduates are relatively small(below 10%), especially in the age group 15-34.By contrast, for the two other education levels,large differences are noticeable in male andfemale participation rates: over 35% for pri-mary or lower education graduates and over25% for secondary education graduates.

Table 2 shows unemployment rates by gender,education level and age group between 2004and 2007. During this period, total unemploy-ment declined (from 11.3% to 9.1%). For bothmen and women, the decline was substantialbut it was more pronounced in the case ofwomen, even though the female unemploy-ment rate was constantly more than doublethat of men. Both for men and women, unem-ployment rates decline with age (with a smallexception for the male group aged 55-64). Inthe age group 15-24, unemployment rates areparticularly high. In women, unemploymentrates are markedly higher among secondaryeducation graduates, whereas in men, differ-ences in unemployment rates do not vary sub-stantially with the education level.

As it was already mentioned above, one of themost attractive features of the LFS for the pur-poses of this study, besides its large sample, isthe form of rotating panel which (theoretically)allows for the isolation of the impact of non-observable individual characteristics, with theuse of appropriate econometric techniques,and the calculation of a “net” impact that par-ticular components of the educational systemhave on a person’s probability of unemploy-ment. This isolation was not possible, however,as the intertemporal variation of the depend-ent variable and (to an even greater extent) ofmany independent variables per individual wasextremely limited. In other words, most of theparticipants were either always employed oralways unemployed over the quarters they par-ticipated in the LFS, while all other featuresremained usually unchanged throughout theirparticipation in the survey. Therefore, for thepurpose of analysis, we used the first observa-tion of each individual in the LFS of the periodunder examination.


33Economic Bulletin

May 2010 27

PPrrii

mmaarr

yy oorr

llooww

eerr ee

dduuccaa

ttiioonn

15-2

467

.637

.855

.965

.626

.449

.367

.429

.051

.857

.735

.948

.0

25-3

492

.950

.374

.392

.552

.175

.092

.354

.375

.793

.045

.573

.8

35-4

493

.357

.574

.693

.159

.275

.792

.557

.174

.691

.156

.173

.5

45-5

490

.247

.165

.689

.149

.166

.689

.746

.365

.589

.648

.465

.5

55-6

456

.824

.938

.457

.725

.038

.658

.124

.238

.259

.824

.939

.6

Tot

al77

.439

.456

.877

.339

.857

.077

.738

.356

.377

.439

.056

.5

SSeeccoo

nnddaarr

yy eedd

uuccaatt

iioonn

15-2

433

.924

.229

.034

.021

.027

.932

.220

.926

.733

.121

.127

.3

25-3

495

.467

.382

.895

.464

.881

.794

.667

.082

.494

.963

.580

.9

35-4

497

.165

.781

.697

.465

.682

.197

.965

.781

.998

.266

.883

.3

45-5

491

.050

.470

.692

.651

.073

.191

.953

.172

.391

.257

.274

.0

55-6

453

.717

.835

.953

.422

.038

.658

.722

.240

.556

.526

.142

.2

Tot

al73

.346

.660

.473

.546

.060

.673

.346

.260

.274

.647

.861

.9

TTeerr

ttiiaarr

yy eedd

uuccaatt

iioonn

15-2

486

.287

.186

.879

.188

.385

.075

.486

.482

.484

.687

.886

.4

25-3

494

.287

.890

.694

.389

.791

.895

.089

.592

.195

.087

.490

.7

35-4

498

.186

.992

.599

.087

.593

.097

.187

.792

.498

.788

.193

.3

45-5

494

.576

.987

.094

.480

.187

.894

.281

.088

.392

.779

.286

.7

55-6

465

.037

.855

.467

.742

.358

.765

.940

.756

.367

.138

.956

.9

Tot

al91

.182

.286

.691

.284

.087

.590

.683

.487

.090

.882

.286

.5

AAllll 15-2

441

.434

.237

.840

.131

.235

.838

.330

.834

.640

.031

.135

.7

25-3

494

.774

.584

.894

.774

.884

.994

.576

.385

.894

.773

.384

.2

35-4

496

.770

.783

.797

.272

.084

.596

.771

.884

.397

.272

.184

.9

45-5

491

.854

.672

.992

.057

.074

.591

.956

.974

.191

.258

.974

.6

55-6

457

.724

.640

.358

.626

.141

.760

.025

.641

.960

.527

.143

.5

Tot

al78

.953

.666

.178

.954

.566

.778

.954

.366

.679

.655

.067

.3

2004

2005

2006

2007*

Men

Wom

enAll

Men

Wom

enAll

Men

Wom

enAll

Men

Wom

enAll

Table

1Rat

e of

par

tici

pat

ion

of p

erso

ns a

ged 1

5-64

in

the

labou

r fo

rce:

LFS

(20

04 Q

1 –

2007

Q3)

* Q

1, Q

2 an

d Q

3.


PPrrii

mmaarr

yy oorr

llooww

eerr ee

dduuccaa

ttiioonn

15-2

418

.538

.923

.914

.525

.416

.925

.946

.130

.513

.527

.018

.0

25-3

49.

227

.414

.55.

616

.89.

07.

121

.111

.59.

417

.511

.4

35-4

47.

219

.412

.24.

917

.09.

86.

415

.69.

97.

216

.310

.7

45-5

44.

713

.18.

15.

211

.87.

94.

411

.87.

34.

811

.97.

8

55-6

45.

74.

95.

43.

95.

04.

34.

53.

74.

25.

64.

55.

2

Tot

al7.

014

.89.

95.

412

.07.

96.

412

.38.

66.

611

.98.

6

SSeeccoo

nnddaarr

yy eedd

uuccaatt

iioonn

15-2

420

.638

.528

.021

.637

.027

.020

.337

.026

.612

.631

.619

.7

25-3

48.

621

.813

.48.

019

.011

.98.

220

.312

.57.

116

.710

.4

35-4

44.

316

.59.

15.

614

.79.

14.

012

.97.

54.

412

.47.

5

45-5

44.

69.

56.

43.

212

.26.

23.

49.

45.

62.

78.

85.

1

55-6

44.

17.

95.

02.

65.

43.

33.

38.

04.

62.

95.

63.

7

Tot

al8.

120

.412

.77.

818

.211

.57.

317

.511

.05.

814

.89.

1

TTeerr

ttiiaarr

yy eedd

uuccaatt

iioonn

15-2

422

.436

.331

.422

.937

.732

.724

.631

.028

.920

.838

.130

.9

25-3

410

.317

.414

.212

.716

.614

.89.

716

.313

.113

.715

.314

.6

35-4

43.

09.

25.

94.

59.

16.

83.

37.

65.

32.

97.

14.

9

45-5

42.

96.

34.

11.

54.

62.

81.

32.

91.

90.

54.

22.

0

55-6

43.

02.

22.

82.

41.

92.

31.

54.

52.

31.

60.

61.

4

Tot

al6.

314

.910

.47.

213

.910

.55.

812

.49.

06.

612

.49.

3

AAllll 15-2

420

.637

.728

.420

.836

.727

.521

.735

.027

.614

.433

.822

.7

25-3

49.

219

.913

.89.

317

.612

.98.

718

.212

.79.

715

.912

.3

35-4

44.

414

.18.

55.

112

.78.

44.

111

.07.

14.

410

.77.

0

45-5

44.

19.

86.

33.

39.

65.

73.

18.

05.

02.

68.

34.

9

55-6

44.

75.

04.

83.

24.

63.

63.

44.

93.

93.

84.

13.

9

Tot

al7.

317

.111

.37.

115

.210

.46.

614

.59.

86.

213

.39.

1

2004

2005

2006

2007*

Men

Wom

enAll

Men

Wom

enAll

Men

Wom

enAll

Men

Wom

enAll

Table

2

Rat

e of

une

mplo

ymen

t of

per

sons

age

d 1

5-64

: LF

S (2

004

Q1

– 20

07 Q

3)

* Q

1, Q

2 an

d Q

3.

The next problem we encountered relates tothe definition of education groups. The LFSdivides the population into a large number ofeducation groups – sometimes quite arbitrar-ily. As this study focuses on tertiary educationgraduates, we chose to group AEI and TEIgraduates with as much detail as possible.However, in many cases this was not feasible,as the number of observations was too small.In the end, the criterion for keeping or merg-ing tertiary education groups, apart from thehomogeneity of the disciplines, was the exis-tence of a minimum number of observations(around 100 men or women, whichever lower)covering a large number of years since gradu-ation. For lower education levels, fewer groupswere formed. Moreover, it was decided toexclude from the sample a small number ofgroups that presented specific problems (e.g.graduates of special needs schools; open uni-versity and inter-disciplinary selection pro-grammes, graduates of military and lawenforcement schools and the School of Peda-gogical and Technological Education–SELETE/ASPETE, graduates from pedagogicacademies with a two-year duration of studies).

Even after these exclusions, the sample of theanalysis remains very large, as it covers a totalof 108,847 employed or unemployed individu-als (58% men and 42% women). Labour forceparticipation rates and unemployment rates ofthe sample are shown in Table 3. The varia-tions in both percentages on account of theeducation level are obvious. The rate of par-ticipation in the labour force is positivelyrelated to the education level and variationswithin a level do exist but are not large – withone exception: the participation of generallyceum graduates in the labour force is about27 percentage points smaller than that ofhigher technical education graduates (techni-cal lyceum, post-gymnasium technical school).

Table 3 also shows that the unemployment rateis lower in the small group of individuals withpostgraduate studies (7.5%) and higher in thegroup of post-lyceum and non-tertiary educationgraduates (14.3%). Significant variations are also

observed between the two larger groups of ter-tiary education graduates (AEI: 7.9% and TEI:12.3%). Even within the narrow definition ofeducation levels, there are variations. Forinstance, the unemployment rate of technicallyceum graduates is 18.4%, whereas that of gen-eral lyceum graduates, albeit higher than thenational average, is significantly lower (10.9%).Accordingly, within AEI, the unemployment rateof Law school graduates is just 2.8%, while thecorresponding rate for graduates of Social Sci-ences is 12.9%. Within TEI, the unemploymentrate of polytechnic schools graduates (StructuralEngineering, Mechanical Engineering and IT) isless than 8% and the rate of the graduates fromthe school of Agriculture & Food Technology ismore than 17%. Therefore, it is of particularinterest to analyse the probability of unemploy-ment for specific graduate groups. Moreover,these variations may be the result of factorswhich are not associated with the education sys-tem. For example, according to many studies, theprobability of unemployment is negatively cor-related with the time that has lapsed since grad-uation. Given the very rapid growth of tertiaryeducation in Greece in recent decades, it is quiteprobable that tertiary education graduates will bemuch younger on average than the graduates oflower education levels. It is, thus, particularlyinteresting to isolate the effect of the two factors(education and years after graduation).

Especially in relation to this last point, we haveto underline that public discourse typicallymakes mention of “youth unemployment”. Butis it really a youth unemployment problem ora problem of transition from education to thelabour market? Undoubtedly, 5 or even 10years after graduation, a compulsory education(gymnasium) graduate is still very young; this,however, is not necessarily true for a tertiaryeducation graduate or, even less, for a post-graduate degree holder. An initial approxi-mation to this problem is made in Charts 1 and2 where the members of the sample are dividedinto two large groups according to the numberof years that have lapsed since their gradua-tion: those with 5 or less years and those with6 or more years since graduation.

33Economic Bulletin

May 2010 29

Chart 1a illustrates the rates of unemploymentby education level for recent (up to 5 years) andolder (6 or more years) graduates. Panels 1band 1c repeat the same, splitting the sample formen and women. For reasons of comparability,in all three panels we applied the same scale tothe x-axis. In the first panel, it is clear that forall education levels, the rate of unemploymentis substantially higher for recent than for oldergraduates (27.0% against 8.6% for the total). Ingeneral, the unemployment rates of tertiaryeducation graduates are lower than those ofprimary or secondary education graduates, bothfor recent and for older graduates. However,

what is impressive with tertiary education grad-uates is the discrepancy in the unemploymentrates of recent and older graduates (ratios of3.7:1 for TEI graduates, 5.1:1 for AEI gradu-ates and 4.0:1 for postgraduates). Especially inthe case of older AEI graduates and postgrad-uates, at first sight unemployment rates do notappear to be particularly alarming (4.7% and3.8% respectively).

Charts 1b and 1c paint a similar picture forboth men and women: unemployment rates aresignificantly higher for recent graduates com-pared with older ones. However, some differ-


Pre-lyceum education 9.8 56.8 40,955Primary 9.1 59.6 27,891Lower secondary 11.1 52.3 13,064Lyceum 11.7 65.1 35,773General lyceum 10.9 60.5 27,514Technical lyceum 18.4 86.3 4,529Post-gymnasium technical school 9.4 88.2 3,730Post-lyceum non-tertiary education 14.3 85.5 9,844IEK 14.6 86.0 8,573Other post-lyceum education 12.4 82.4 1,271TEI 12.3 90.2 5,490Structural Engineering 7.5 90.3 326Mechanical & Computer Engineering 7.9 92.3 1,319Agricultural & Food Technology 17.2 89.5 463Economics & Management 15.9 90.1 1,720Medical Sciences 11.3 90.0 1,418Other TEI 12.8 82.7 244AEI 7.9 88.0 15,093Structural Engineering 5.0 91.9 1,252Mechanical Engineering 6.5 92.3 888IT 5.1 88.5 203Physical Sciences 7.3 87.2 860Mathematics & Statistics 5.7 90.0 680Medical School etc. 5.5 89.8 1,804Horticulture & Forestry 8.9 88.7 513Law School 2.8 87.4 1,120Economics & Management 7.8 85.5 2,824Social Sciences 12.9 81.8 405Humanities 11.8 86.2 2,131Languages 9.5 86.2 600Physical Education & Sports 11.8 92.1 665Pedagogics 11.0 91.2 918Other AEI 11.4 79.9 230Postgraduate studies 7.5 94.3 1,292Postgraduate degree 8.6 93.2 875Doctorate 4.9 97.1 417TTOOTTAALL 1100..77 6677..55 110088,,444477

Education level Rate of unemploymentRate of participation in

labour force N (labour force)

Table 3 Rate of unemployment and participation in the labour force of persons aged 15-64the first time they are included in the LFS sample (2004 Q1 – 2007 Q3)

33Economic Bulletin

May 2010 31

ences between the two genders are also impres-sive. In all individual groups, without excep-tion, female unemployment rates are higherthan male ones. Especially for women whohave recently graduated from lyceum, theunemployment rate is excessively high (41.2%).Male unemployment rates for older tertiaryeducation (AEI, TEI, postgraduate studies)graduates are in the area of 2.6% to 3.7%,whereas in the corresponding female groupthese rates are considerably higher (6.0% to11.3%). Moreover, although male unemploy-ment rates of recent tertiary education gradu-ates are not very different from those of lowereducation graduates, there are significant dif-ferences in women.

Charts 2a, 2b and 2c focus on tertiary educa-tion graduates and provide a more detailed pic-ture for TEI and AEI graduates, as well as forthe small but rapidly growing group of post-graduates (Kikilias, 2008). The conclusion thatis clearly drawn from these charts is that, eventhough some generalisation is possible, thereare major intra-group differences. Chart 2ashows that, among TEI graduates with highunemployment rates, there are similaritiesbetween the groups of recent and older grad-uates (with the exception of graduates from theschool of Economics & Management). How-ever, unemployment rates of recent graduatesand, to a greater extent, of older graduatesshow considerable differences between groupsof schools. For example, the unemploymentrate for older graduates from PolytechnicSchools (Structural Engineering, MechanicalEngineering and IT) is below 4%, whereas forolder graduates from the school of Economics& Management and Agricultural & FoodTechnology it is above 10%.

There are even larger differences within thegroup of AEI graduates, which is evident inChart 2b, possibly as a result of the larger num-ber of sub-groups in the sample. In all thesegroups, without exception, unemployment ratesof recent graduates are much higher than thoseof older ones. However, in some groups, unem-ployment rates are alarmingly high for older

graduates (e.g. graduates from the schools ofPhysical Education & Sports: 9.8%), whilst inother groups the corresponding rates are lowerthan those typically classified as frictionalunemployment (e.g. graduates from IT: 1.3%or Law: 1.7%). Furthermore, in the classifica-tion of schools by unemployment rates of olderand recent graduates, the groups are not reallycorrelated. For example, while the unemploy-ment rates of older graduates from the schoolsof Physical Sciences and Mathematics & Sta-tistics are very low, those of the recent gradu-ates from the same schools are among the high-est for recent graduates. Finally, Chart 2c showsthat acquiring a postgraduate or a doctoratedegree is associated with very low unemploy-ment rates 6 or more years after graduation;however, such degrees should not be consid-ered as effective shields against unemploymentin the first five years after graduation (unem-ployment rates of recent graduates are 15.8%and 12.9% respectively).

4 ECONOMETRIC RESULTS

Naturally, an individual’s probability ofunemployment is influenced by various factorsand not solely by educational qualifications.Some of these factors are directly observablewhile others are not. Furthermore, randomfactors may influence this probability. The LFSincludes a number of variables which mayinfluence the probability of unemployment.Therefore, in order to better understand thephenomenon, it is necessary to apply a multi-variate econometric probability analysis,which is the primary objective of this part ofthe study.

The descriptive results of the previous sectionprovide some indications. First, the labourforce participation rate seems to be associatedwith both the individual’s age ―and, mostlikely, with the number of years that havelapsed since graduation― and the educationlevel, as well as with gender (and, possibly,other factors such as family status). Hence,without controlling for the probability of


labour force participation, it is very likely thatthe estimated results for the probability ofunemployment are biased (from a statisticalpoint of view). In other words, it is highly likelythat a two-stage estimation method isrequired: (i) a first estimation of the proba-bility of labour force participation, and (ii) asubsequent estimation of the probability ofunemployment following the inclusion of therelevant correction term in the explanatoryvariables (Inverse Mills Ratio).

The descriptive results show that, in all likeli-hood, the variable related to the probability ofunemployment is the time interval from grad-uation and not the individual’s age per se. Thesame results imply that the time interval fromgraduation has a non-linear effect on the prob-ability of unemployment, something thatshould be taken into account in the econo-metric estimation.

Moreover, the descriptive results indicate thatboth the unemployment rate and its changeafter graduation vary significantly betweenmen and women. Therefore, it is essential toestimate different equations for men andwomen rather than introducing a dummy vari-able for women in the unemployment proba-bility equation for the entire sample.

Lastly, since the evolution of the unemploy-ment rate in relation to the years that havelapsed since graduation shows substantial dif-ferences between groups of graduates of thesame education level, graduates of differentschools should not only be distinguished byapplying dummy variables, but flexible, non-linear formulas should also be used, whichallow for different unemployment probabilitiesas a function of the time after graduation fordifferent types of education groups.

All of the above were taken into account in thespecification and estimation of the economet-ric model. The methodology used is a variationof Heckman’s (1979) two-stage method, withwhich we attempted to correct the sample’sselection bias. At a first stage, the probability

of labour force participation was estimatedusing the sample of all individuals aged 15-64who participated in the labour market. Thevariation lies in the fact that at the second stageof the estimation of the unemployment prob-ability, a binominal variable (instead of a con-tinuous one, as in Heckman’s classicalmethod) was used as a dependent variable. Thishas brought about changes in the estimatedmaximum likelihood function (Pindyck andPubinfeld, 1998, Greene, 2003). The detailedresults of these estimations and the definitionsof the variables used are presented in theAnnex.

The presentation of the results begins with theestimation of the labour force participationprobability, in which the sample comprises allindividuals aged 15-64 (79,288 men and 83,736women). The male labour force participationrate in our sample is 79.9% and the female rateis 53.8%. In both equations (men and women)the control group consists of members of child-less couples, who are Greek nationals, generallyceum graduates, residents of Athens andtook part in the LFS in 2007(Q3), whereas, onaccount of heteroskedasticity, the estimatedstandard errors of coefficients were correctedusing White’s (1980) method.

Both for men and ―markedly― women, theprobability of labour force participation isclosely linked with their family status. In com-parison with the control group, the probabil-ity of female labour force participation is sig-nificantly lower when women are married withchildren (the opposite is usually observed forfemale heads of single-parent households),while, in connection to men, the result variesdepending on the number and age of the chil-dren. In terms of education level, for both menand women, ceteris paribus, the labour forceparticipation probability for graduates of alleducation groups (except gymnasium gradu-ates) is higher compared with general lyceumgraduates. For both genders, the labour forceparticipation probability of postgraduates isparticularly high (a finding which is in line withthe predictions of human capital theory). Fur-

33Economic Bulletin

May 2010 33

thermore, for both men and women, the prob-ability of labour force participation appears tohave a close but non-linear association with thenumber of years since graduation.

Regional differences in terms of labour forceparticipation are not particularly large in men,whereas, in women, differences are moreimportant and probabilities of participationare higher in most regions (compared withAttica). As regards the degree of urbanisation,ceteris paribus, higher participation rates areobserved in semi-urban and, particularly, ruralareas. Finally, there are no significant differ-ences with respect to the year of the LFS orseasonality (the survey’s quarter). In bothequations, the coefficient showing whetherthere is bias (“arthrho”) is statistically signif-icant and implies the existence of systematicbias, which is corrected using the Inverse MillsRatio in the equations for the probability ofunemployment (in other words, the results ofthe estimation for the probability of unem-ployment presented below would not be reli-able without this correction).

After the presentation of the results on theprobability of labour force participation, wemove on to the presentation of the results fromthe estimation of the probability of unem-ployment. This time, the samples consist of63,378 men and 45,060 women. Similarly toprevious equations, on account of het-eroskedasticity, the estimated standard errorsof coefficients were corrected using White’s(1980) method. Most of the results are pre-sented in charts. Once more, the control groupcomprises members of childless couples, whoare Greek nationals, general lyceum graduates,residents of Athens and took part in the LFSin 2007(Q4). The dependent variable is theprobability of unemployment and independentvariables include the individual’s educationbackground, the number of years after gradu-ation (and its square), dummy variables relatedto household composition, the individual’snationality, the region and the degree ofurbanisation of the place of residence, the localunemployment rate, as well as the dummy vari-

ables for the year and the quarter of the LFSin which they participated. Education groupsare much more detailed, compared with theprevious equations (like the groups of Table 3),while demographic groups (household com-position) are much less detailed. In order tocapture the different unemployment proba-bilities for each education group in relation tothe time interval from the year of graduation,we introduced multiplicative terms betweenthe dummy variable of each education groupand (a) the time from the individual’s gradu-ation and (b) its square.

The dummy variables of the quarters aim tocapture the potential seasonality of unem-ployment, while the dummy variables of theyears seek to capture possible effects of theeconomic cycle on employment.5 The twogroups of variables are not statistically signif-icant, both in the female and the male function.The variable that relates to the local unem-ployment rate refers to the unemployment rateof the region of the individual’s place of resi-dence in the period of their participation in theLFS. As expected, this variable has a positiveeffect on the unemployment probability and isstatistically significant. As regards the demo-graphic features of the individual’s household,the results indicate that both for men andwomen, single parenthood as well as the pres-ence of household members other than spouseand children increase the probability of unem-ployment compared with the control groups(members of childless couples).

Before moving on to the presentation of resultsin detailed charts, it would be interesting toexamine the corresponding results for sixbroad groups according to the education level(pre-lyceum, lyceum, post-secondary non-ter-tiary, TEI, AEI, postgraduate studies). Theseresults stem from estimations that use the vari-ables mentioned above, but this time forbroader groups of graduates, and are pre-sented in Charts 3a and 3b for men and


55 However, it should be noted that the LFS used in this paper covera period of 4 years, during which the Greek economy was growingat a quite high rate.

women, respectively. These charts show theestimated rate of unemployment (verticalaxis)6 as a function of the first 20 years sincegraduation from the highest education levelcompleted (horizontal axis), after controllingfor the effect of all other variables (family sta-tus, region, urbanisation, nationality, local rateof unemployment, etc.).7 Similar charts areused in the remainder of this study.

Many interesting conclusions can be drawnfrom Charts 3a and 3b. Both men and womenwith low education level (pre-lyceum) beginwith lower rates of unemployment comparedto the other members of the sample right aftertheir graduation, but their estimated rate ofunemployment changes very slowly throughouttheir working life. Differences betweenlyceum and post-secondary non-tertiary grad-uates are negligible for women, while for menestimated rates of unemployment of post-sec-ondary non-tertiary graduates are considerablyhigher than those of lyceum graduates in thefirst 5-6 years after graduation; differences getsmall thereafter. As regards tertiary educationgraduates, estimated rates of both male andfemale unemployment for a given number ofyears since graduation are lower for holders ofpostgraduate degrees, followed by AEI grad-uates and, finally, TEI graduates. However,while estimated rates of female unemploymentof AEI graduates and holders of postgraduatedegrees converge about 20 years after gradu-ation, they never converge with the estimatedrate of unemployment of TEI graduates. Onthe contrary, the difference in the estimatedrates of male unemployment of AEI and TEIgraduates is very small about 6 years aftergraduation and converges with the estimatedrate of unemployment of postgraduate degreeholders 12-13 years after graduation, at levelsmuch lower than in other education levels.

The next seven charts, 4.1a to 4.7b, present theestimated rates of unemployment (one for menand one for women) as a function of the yearssince graduation, for homogeneous groupswithin education levels. Because the scale ofthe vertical axis (estimated rates of unem-ployment) does not always have the samelength, all charts referring to education groupsalso include, for reasons of comparability, theestimated rate of unemployment of malelyceum graduates.

33Economic Bulletin

May 2010 35

66 It should be noted that the scale which measures the rate of unem-ployment is much “lengthier” in the case of women, since their esti-mated rates of unemployment were much higher than those of men.

77 In all likelihood, 20 years after graduation is the maximum periodthat young people consider when deciding on the level and specialtyof their studies.

Charts 4.1a and 4.1b refer to groups with loweducational qualifications: primary educationand lower secondary education. It should benoted that most members of these groups areof relatively old age. The “primary education”category comprises both persons who have notfinished primary school and primary schoolgraduates who have additionally attended afew years of gymnasium. Male members oflower secondary education groups start offwith a low rate of unemployment and their esti-

mated rate of unemployment changes littlewith time, while the rate of unemployment ofprimary education graduates remains at rela-tively high levels even many years after grad-uation. For female graduates of these two cat-egories, the results are rather surprising, as theestimated rate of unemployment seems toincrease in the first few years since graduation(especially for “primary education” graduates),before starting to decline.

Charts 4.2a and 4.2b show the estimated ratesof unemployment of higher secondary and post-secondary non-tertiary education graduates.Graduates of higher secondary education aregrouped into three categories: general lyceum,technical lyceum and post-gymnasium techni-cal schools. The first category also comprisespersons who have not completed tertiary edu-cation studies; the second category consists ofgraduates from Technical Vocational Lyceums(TEL), Unified Multidisciplinary Lyceums(EPL) and Technical Vocational Institutes(TEE); and the third category comprises grad-uates from Technical Vocational Schools(TES), post-gymnasium foreman schools andpost-gymnasium mercantile marine schools.Post-secondary non-tertiary education gradu-ates are grouped into two categories: graduatesfrom (public or private) Institutes of Voca-tional Training (IEK) and graduates from otherpost-secondary education. The latter categorycomprises graduates of colleges, dance schools,tourism, (non-university) foreign languages,mercantile marine officers, etc.

Differences between men and women inCharts 4.2a and 4.2b are noteworthy. For men,differences between general and technicallyceum graduates are negligible. Graduates ofpost-gymnasium technical schools are at aslightly better position in the first few yearsafter graduation, but differences from theother levels of higher secondary education areeliminated about 6 years after graduation.Conversely, graduation from IEK or otherpost-secondary non-tertiary education estab-lishment is associated with higher rates ofunemployment than graduation from a higher


secondary education establishment. In the caseof IEK, the estimated rates of unemploymentconverge with those of higher secondary edu-cation about 10 years after graduation, whilein the case of graduates of other post-sec-ondary establishments, the estimated rate ofunemployment constantly diverges from that ofhigher secondary education graduates. As inalmost all charts, the estimated rates of femaleunemployment in Chart 4.2b are higher thanthose of male unemployment for each educa-tion group of Chart 4.1a. In women, however,the lowest estimated rates of unemploymentare recorded among general lyceum graduates.In the first 15 years after graduation, the ratesof unemployment of female technical lyceumgraduates are higher than those of post-gym-nasium technical schools graduates and, asmentioned above, both are much higher thanthose of general lyceum graduates. Estimatedrates of unemployment of IEK graduates and,to a lesser extent, other post-secondary estab-lishments are very high and, in the case ofother post- secondary establishments, divergefrom the rates of unemployment of generallyceum graduates. The above results do notcorroborate the view often expressed in pub-lic discourse on the need to boost technicaleducation in order to combat youth unem-ployment in Greece.

Charts 4.3a and 4.3b show the estimated ratesof unemployment of TEI graduates (or, pre-viously, KATEE, i.e. Centers of Higher Tech-nical and Vocational Training). These gradu-ates have been grouped into six categories. Thefirst two categories have a technical orientation(Structural Engineering and Mechanical &Computer Engineering), the third one resultedfrom the merging of the graduates of Agricul-tural and Food Technology schools, the nexttwo relate to graduates of Economics & Man-agement and Medical (or Paramedical) Sci-ences, while the last one (other TEI) comprisesLibrarians, Social Workers and Applied Artsgraduates. Due to the heterogeneity of the lat-ter and the small number of observations,results concerning the category “other TEI”must be interpreted with caution.

For both male and female TEI graduates, thelowest estimated rates of unemployment areobserved among technical school graduates. Arelatively better position between the two groupsis held by Structural Engineering graduates com-pared with Mechanical & Computer Engineer-ing graduates (for women, this is true only for thefirst 7 years after graduation). Among men, Med-ical Sciences graduates begin with extremely high

33Economic Bulletin

May 2010 37

rates of unemployment in the first few years aftergraduation, while rates of female unemploymentamong graduates of Agricultural & Food Tech-nology and, to a lesser extent, Economics & Man-agement begin with and remain at very high lev-els, even many years after graduation.

Because of the classification of AEI graduates(except for postgraduates) into a large number

of categories, the relevant results have beengrouped and are presented in three sets ofcharts. Charts 4.4a and 4.4b show the estimatedrates of unemployment of science graduates.More specifically, estimates are presented fortwo groups of technical schools graduates, i.e.(a) Structural Engineering and (b) MechanicalEngineering, as well as for (c) graduates of IT,(d) graduates of Physical Sciences and (e)


graduates of Mathematics & Statistics. Thegroup of IT graduates is numerically small andincludes relatively few individuals several yearsafter their graduation; consequently, the cor-responding results should be interpreted withcaution. The main reason we decided to groupIT graduates in a separate category is the fre-quency with which this particular group is men-tioned in public discourse as an example ofexcessive demand in the labour market. UnderStructural Engineering we have included grad-uates from schools such as Civil Engineering,Architecture, Topography, etc. MechanicalEngineering also includes graduates fromNaval Architecture, Electrical Engineering,Chemical Engineering, Mineralogy, etc.Under Physical Sciences graduates fromPhysics, Chemistry, Biology (apart from Med-ical Biology) and Geology are included.

Among both men and women, the best per-formers are IT graduates (which confirms thereferences in public discourse), followed by―at least for the first years after graduation―Polytechnic School graduates (from StructuralEngineering and Mechanical Engineering). Onthe other hand, male ―and, to a much largerdegree, female― graduates of Physical Sci-ences or Mathematics and Statistics start withextremely high estimated rates of unemploy-ment. In the case of men, these rates drop dra-matically after 5-10 years, while in the case ofwomen, the rates remain very high even 10years after graduation.

Charts 4.5a and 4.5b illustrate the estimatedrates of unemployment for five groups of AEIgraduates: (a) Medicine, (b) Horticulture &Forestry, (c) Law, (d) Economics & Manage-ment and (e) Social Sciences. Apart from Med-ical School graduates, the Medicine group alsoincludes dentistry, pharmaceutical and veteri-nary graduates, while Social Sciences includeSociology, Psychology, Anthropology, etc.

Both male and female graduates of LawSchools have low rates of unemployment (inthe case of men the rates are extremely low),even during the first years after graduation.

For Medical School graduates, the rates tendto get close to zero only many years after grad-uation. To illustrate this better, the estimatedrates of unemployment of men in this group 8years after graduation are higher than therespective rates of men who have graduatedfrom a general lyceum. Female graduates ofHorticulture & Forestry or Social Sciencesalso have high estimated rates of unemploy-ment for many years after graduation.

33Economic Bulletin

May 2010 39

Finally, the estimated rates of unemploymentof Economics & Management graduates lie inthe middle of the range.

The third category of AEI graduates is that of“Instructors”. Charts 4.6a and 4.6b show theestimated rates of unemployment for fivegroups of schools, namely Humanities, Lan-guages, Physical Education & Sports, Peda-

gogics and other AEI. The latter includes FineArts, Medical Biology, Nursing, Dietetics,Journalism, Librarianship, Home Economics,etc. As in the case of TEI, the high degree ofheterogeneity among the members of thegroup may render the interpretation of the cor-responding results almost meaningless. Thesame applies to men who have graduated fromLanguage Schools, since very few of them areincluded in the LFS sample. The Humanitar-ian Studies group includes graduates of GreekLiterature, Philosophy, History, Archaeology,Theology, Music, Theatre, etc.

Among men, those with the best prospectsappear to be the graduates of Pedagogics andamong women the graduates of Languages. Bycontrast, male graduates of Physical Educationseem to face high unemployment for manyyears after graduation. Among women, highrates of unemployment are observed for grad-uates of Humanitarian Studies and Pedagogics,especially during the first years after gradua-tion. The shape of the curve for female gradu-ates from Physical Education & Sports is rathersurprising, as it implies a rise in the estimatedunemployment rate in the first 7 years aftergraduation and a gradual decline thereafter.

Finally, Charts 4.7a and 4.7b show the resultsfor individuals with postgraduate studies, sep-arately for Master’s degree holders (MA, MSc,MBA) and doctorate (PhD) holders. In thecase of men, the estimated rates of unem-ployment of postgraduate degree holders startfrom unexpectedly high levels (7%-9% for thefirst year after graduation), but then drop rap-idly. In the case of doctorate holders, theirrates of unemployment come close to zero inless than 10 years after graduation, while in thecase of Master’s degree holders they dropbelow 5% about 5 years after graduation. Asfor women, differences between the two groupsare very difficult to ascertain. Although com-pared to most groups of female tertiary edu-cation graduates their estimated rates of unem-ployment are lower, they seem to be quite highduring the first years after graduation (for thefirst year, their estimated rate of unemploy-


ment is more than 17%), while even 20 yearsafter graduation, their estimated rate of unem-ployment remains higher than that of malelyceum graduates, in the corresponding phaseof their career.

Table 4 tackles the question “what is the esti-mated rate of unemployment 1, 3, 5 and 10years after graduation, after isolating theimpact of all other factors?” Many interestingconclusions can be drawn from the resultsreported there and we have already mentionedsome of them. Let us start from the groupingsaccording to education level. Both men andwomen with basic or lower education have thelowest estimated rates of unemployment oneyear after the completion of their studies. Thisis not surprising, as most drop-outs leaveschool because they have already found a job.Coming to 3, 5 and 10 years after graduation,the lowest rates of unemployment are regis-tered for the group with very high educationalqualifications (postgraduate degree holders).As years go by, the relative position of tertiaryeducation graduates improves and AEI grad-uates are (in general) in a better position thanTEI graduates.

At this point we should make two moreremarks. First, even if the impact of all otherfactors is eliminated, the relative position ofwomen in general ―and of female higher edu-cation graduates in particular― is much worsethan that of men. For instance, even 10 yearsafter graduation, when the estimated rate ofmale unemployment drops to 2.6% for bothAEI and TEI graduates, the rates for womenare 7.0% for AEI graduates and 9.9% for TEIgraduates. Similar differences are also evidentamong graduates of other education levels.Second, there exist very large differencesacross groups within specific education levels.Referring to tertiary education alone, one cansee the differences in the estimated rates ofunemployment in all three time periods(immediately after graduation, after 3 yearsand after 5 years) for both male and femalegraduates of Physical Sciences and Mathe-matics & Statistics on the one hand and Law

and IT on the other. However, even if a highereducation degree shields against unemploy-ment in the long run, Table 4 shows that thisdoes not apply to all graduates. For instance,even 10 years after graduation, the estimatedrates of unemployment for women who havegraduated from Horticulture & Forestry orPhysical Sciences (AEI) as well as from Agri-cultural & Food Technology or Economics &Management (TEI) are considerably higherthan 10%.

33Economic Bulletin

May 2010 41


PPrree--llyycceeuumm eedduuccaattiioonn 77..11 66..44 66..00 55..22 1177..00 1177..44 1177..44 1177..00

Primary 9.6 8.7 8.1 6.8 9.9 11.6 12.6 14.3

Lower secondary 4.9 4.8 4.6 4.3 8.6 9.2 9.5 9.8

LLyycceeuumm 88..11 66..44 55..55 33..88 2277..00 2222..55 1199..88 1144..55

General lyceum 6.6 5.5 4.9 3.6 12.2 11.5 11.0 9.7

Technical lyceum 7.3 5.7 4.9 3.5 19.2 17.5 16.3 13.6

Post-gymnasium technical school 5.6 4.9 4.4 3.5 14.6 13.9 13.4 12.0

PPoosstt--llyycceeuumm nnoonn--tteerrttiiaarryy eedduuccaattiioonn 1100..33 77..66 66..33 44..00 2244..99 2200..66 1188..33 1133..77

IEK 9.7 7.2 5.9 3.8 20.5 17.9 16.4 13.2

Other post-lyceum education 6.1 5.6 5.3 4.8 14.0 13.0 12.4 11.3

TTEEII 1122..44 77..66 55..55 22..66 2277..22 1199..77 1166..00 99..99

Structural Engineering 6.9 4.4 3.3 1.6 15.3 11.4 9.5 6.5

Mechanical & Computer Engineering 10.3 6.4 4.7 2.3 27.3 16.4 11.7 5.6

Agricultural & Food Technology 12.4 8.4 6.6 3.7 27.1 22.2 19.6 15.1

Economics & Management 11.1 7.2 5.5 2.8 20.9 18.2 16.7 13.9

Medical Sciences 20.7 10.9 6.8 1.9 24.6 17.4 13.7 7.5

Other TEI 11.7 6.7 4.7 2.1 29.7 19.1 14.3 7.5

AAEEII 1100..99 77..00 55..22 22..66 2244..88 1177..11 1133..22 77..00

Structural Engineering 6.4 4.2 3.2 1.7 19.6 12.0 8.6 4.1

Mechanical Engineering 5.8 4.6 3.9 2.8 11.5 7.0 5.1 2.7

IT 0.7 2.1 2.5 0.6 10.1 9.1 5.5 0.1

Physical Sciences 16.3 10.5 7.7 3.3 40.1 27.4 20.6 9.2

Mathematics & Statistics 24.5 11.3 6.5 1.6 44.1 25.9 17.2 5.9

Medical School etc. 8.6 7.5 6.3 3.0 18.7 12.2 9.1 4.4

Horticulture & Forestry 6.5 5.1 4.3 2.7 24.5 23.4 21.9 16.4

Law School 1.0 0.7 0.5 0.3 10.4 7.6 6.1 3.4

Economics & Management 10.9 6.6 4.8 2.3 18.3 13.6 11.2 7.0

Social Sciences 10.9 7.5 6.0 3.7 32.0 20.7 15.4 7.6

Humanities 12.9 9.0 7.0 3.6 27.3 20.3 16.5 9.3

Languages 18.7 12.3 8.0 1.3 13.2 10.4 9.0 6.4

Physical Education & Sports 12.9 10.8 9.5 6.6 9.0 11.8 12.9 11.9

Pedagogics 6.7 5.2 4.3 2.4 21.7 16.1 13.1 7.5

Other AEI 8.0 7.2 6.8 5.7 15.2 11.4 9.6 6.8

PPoossttggrraadduuaattee ssttuuddiieess 99..22 55..99 44..33 22..22 1188..44 1111..99 99..00 44..88

Postgraduate degree 8.8 6.1 4.8 2.8 17.4 11.7 9.0 5.0

Doctorate 7.2 4.4 3.0 0.9 18.2 11.8 8.9 4.7

Education level

Men Women

Years since graduation Years since graduation

<1 3 5 10 <1 3 5 10

Table 4 Estimated rate of unemployment as a function of years since graduation

Table 5 attempts to answer the question “afterhow many years would the estimated rate ofunemployment drop to 2%, 4% and 6%?”.Undoubtedly, 2% can be considered to fallwithin frictional unemployment, many econo-mists would consider 4% to fall within the nat-ural rate of unemployment and, given the cur-rent conditions prevailing in the Greek econ-omy, 6% can be considered a relatively satis-factory rate of unemployment. In Table 5 it isassumed that 40 years after graduation is themaximum time period that an individualremains in the labour market. The sample of theLFS used in our analysis includes a few personswho were less than 65 years old and reportedthat they had reached their highest educationachievement 40 years earlier. In any case, thevast majority exits the labour market in less than40 years after the completion of studies.

Women reach the 6% threshold after 18, 12and 9 years from the completion of their TEI,AEI and postgraduate studies respectively.The estimated rates of 4% are reached only byfemale graduates of AEI and postgraduatestudies, while 2% is reached only by AEI grad-uates (23 years after graduation) and not byfemale postgraduates. Men reach the esti-mated rates of unemployment much soonerthan women and, in general, the higher theeducation level, the fewer the years requiredfor the attainment of these levels. Again, thedifferences across education groups within thesame level are very large. At the tertiary edu-cation level, men are close to 2% already fromthe first year after graduation from IT and LawSchools, but never after obtaining a degree inSocial Sciences. Additionally, while the rate ofunemployment is less than 6% already fromthe first year since graduation from IT, Law,Structural Engineering (both AEI and TEI)and Mechanical Engineering, or again alreadyfrom the second year since graduation fromHorticulture & Forestry and Languages, it usu-ally takes graduates of Physical Education &Sports 12 whole years to reach the same levels.As regards women, the rate of unemploymentdrops to 2% in 10 years after graduation fromIT Schools. According to our estimates, female

graduates of Economics & Management,Social Sciences and Languages (AEI), gradu-ates of any TEI (excluding Medical Sciences)and holders of Master’s degrees and doctoratesdo not fall below these unemployment rateswithin the period of 40 years after graduation.

The last part of our empirical results is pre-sented through charts and deals with the

33Economic Bulletin

May 2010 43


BBeeffoorree llyycceeuumm 3388 1188 66 4400++ 4400++ 3388

Primary school 39 24 14 40+ 39 35

Gymnasium 32 13 1 40+ 36 29

LLyycceeuumm 2211 1100 44 4400++ 4400 2277

General lyceum 21 9 2 40+ 32 24

Technical lyceum 23 8 3 40+ 35 28

Post-gymnasium technical school 25 8 1 40+ 33 27

PPoosstt--llyycceeuumm nnoonn--tteerrttiiaarryy eedduuccaattiioonn 2211 1111 66 4400++ 4400++ 4400++

IEK (post-lyceum vocational training institute) 19 10 5 40+ 40+ 33

Other post-lyceum education 40+ 21 1 40+ 40+ 40+

TTEEII ((TTeecchhnnoollooggiiccaall EEdduuccaattiioonnaall IInnssttiittuutteess)) 1122 88 55 4400++ 4400++ 1188

Structural Engineering 9 4 1 40+ 40+ 12

Mechanical & Computer Engineering 11 7 4 40+ 14 10

Agricultural & Food Technology 18 10 6 40+ 40+ 40+

Economics & Management 14 8 5 40+ 40+ 40+

Medical Sciences 10 8 6 23 16 12

Other TEI 11 6 4 40+ 40+ 13

AAEEII ((UUnniivveerrssiittiieess)) 1133 77 55 2233 1155 1122

Structural Engineering 9 4 1 18 11 8

Mechanical Engineering 18 5 1 15 7 4

IT 1 1 1 8 6 5

Physical Sciences 13 9 7 19 15 13

Mathematics & Statistics 10 7 6 16 12 10

Medical School etc. 12 9 6 17 11 8

Horticulture & Forestry 14 6 2 23 21 19

Law School 1 1 1 15 9 6

Economics & Management 11 7 4 40+ 17 12

Social Sciences 40+ 10 5 40+ 17 13

Humanities 15 10 7 22 17 14

Languages 10 8 7 40+ 20 12

Physical Education & Sports 23 16 12 21 18 16

Pedagogics 12 6 2 21 16 12

Other AEI 40+ 21 9 40+ 40+ 13

PPoossttggrraadduuaattee ssttuuddiieess 1111 66 33 4400++ 1122 99

Postgraduate degree 14 7 4 40+ 13 9

Doctorate 7 4 2 40+ 12 8

Education groups

Men Women

Years until the estimated rate of unemployment drops to

Years until the estimated rate ofunemployment drops to

2% 4% 6% 2% 4% 6%

Table 5 Years until the estimated rate of unemployment drops to 2%, 4% and 6%

impact of three factors on the probability ofunemployment. These factors are: nationality,region and urbanisation level of the place ofresidence. The results are set forth in Charts5.1a through to 5.3b, separately for men andwomen, and involve all AEI graduates. As theterms of the equation are introduced asdummy variables, the respective curves do notintersect.

Charts 5.1a and 5.1b depict the estimatedunemployment rates for male and female grad-uates of AEI, in relation to the degree ofurbanisation of their place of residence, hold-ing constant the impact of other factors. Bothmen and women residing in Athens or anotherurban center (apart from Thessaloniki) faceconsiderably higher rates of unemployment,especially during the first years after gradua-tion. On the other hand, residence in ruralareas is associated with lower rates of unem-ployment. It is quite possible ―but cannot beascertained by our results― that the negative(positive) impact of living in a rural area (largeurban centre) on the rate of unemploymentcan be attributed to the excess demand (sup-ply) of labour in those areas. This finding mayalso be related to the high level of underem-ployment in rural areas.

Charts 5.2a and 5.2b repeat the same proce-dure for the 13 regions of the country. Theresults are quite similar for men and women.Residents in Thessaly, Western Greece andCentral Macedonia (men) or the Peloponnese(women) face higher rates of unemployment.By contrast, residing in Crete, the islands of theNorth Aegean, the remaining Central Greeceand Euboea (men) or Attica (women) is asso-ciated with lower rates of unemployment. Itshould be noted that differences in the esti-mated regional rates of unemployment areconsiderably high, especially in the first yearsafter graduation.

Finally, Charts 5.3a and 5.3b show the esti-mated rates of unemployment for AEI gradu-ates according to the years after graduation,separately for men and women, on the basis of

their nationality. The LFS sample is dividedinto three categories: Greek nationals, otherEU countries’ nationals and non-EU nationals.The estimated rate of unemployment of Greekmen is considerably higher than the rates of theother two groups. By contrast, Greek women

33Economic Bulletin

May 2010 45

and non-EU women face the same unemploy-ment rates, while the estimated unemploymentrate of women from EU countries other thanGreece is slightly higher.

5 CONCLUSIONS

According to Georganta, Kandilorou andLivada (2008), 58% of second-year students of

the Athens University of Economics and Busi-ness and the University of Macedoniareported that they decided to pursue tertiaryeducation in order to find a better-paid jobmore easily (the “better-paid” part will be thesubject of another study). Can our findings jus-tify the “more easily” part? In other words, cantertiary education shield against unemploy-ment? The results of our analysis show thatthese students took a very rational decision, asbetter educational qualifications seem to bequite effective in shielding against unemploy-ment, even if it is only in the long run. It shouldalso be noted that Schools whose graduatesexperience the lowest unemployment rates arethe most popular ones among tertiary educa-tion applicants.

This study has yielded other findings as well.Some of these findings are consistent with theresults of previous studies. However, as itbecomes clear from the review of the secondsection, empirical studies on youth unemploy-ment in Greece show divergent results. Thefirst finding of the present paper is that the dis-course on youth unemployment is rather mis-placed. The problem is not quite a problem ofyouth unemployment, but a problem of tran-sition from education to the labour market,irrespective of age. It also involves graduatesof all education levels, not only tertiary edu-cation. The difference between tertiary edu-cation graduates and graduates of lower edu-cation levels is that the unemployment rates ofthe first drop to acceptable levels a few yearsafter graduation, while for the latter the paceof this decline is substantially slower and unem-ployment rates converge to higher levels. Thisis the second significant finding of our paper.In general, the higher the education level (TEI,AEI, postgraduate studies) the lower the rateof unemployment in the long run.

The third finding relates to the very significantdifferences within education levels. Somegroups of graduates face no real problems aftergraduating (from IT and Law Schools – at leastthe male graduates), others run a serious riskof unemployment for a relatively small number


of years after graduation (Physical Sciences,Mathematics & Statistics), while others face aserious problem of unemployment for severalyears after graduation (Physical Education and& Sports, Social Sciences and several TEISchools). The last finding of the study is that,ceteris paribus, women in general, and femaletertiary education graduates in particular, facea significantly higher risk of unemploymentcompared to male graduates with similar edu-cational qualifications. For certain categoriesof female tertiary education graduates, theestimated unemployment rates are exception-ally high, even several years after graduation(AEI: Horticulture & Forestry, Physical Edu-cation & Sports; TEI: Agricultural & FoodTechnology, Economics & Management).

Would it be possible to generalise the aboveresults and extrapolate them to the future?Possibly yes, but that would be too riskybecause of ever-changing circumstances. Ter-tiary education in Greece ―like in most OECDcountries― has expanded rapidly in the past 15years. A recent OECD report concludes thatdemand for tertiary education will increase inthe years to come. The recent economic crisisand the recession could possibly push moreyoung people to remain in education as, inmost countries, the pay gap between tertiaryeducation graduates and people with lowereducation keeps widening (OECD, 2009).

In Greece, the number of places available inAEI and TEI for students accounts for morethan 80% of the corresponding (annual) demo-graphic cohort. Of course, many places remainvacant and some students never graduate. Buteven if these two factors are taken into con-sideration, the share of tertiary education grad-uates in recent cohorts is substantially higherthan in previous ones. Would this imply thatthe increased supply of graduates ―which isexpected to continue in the next years― willlead to a more permanent increase in the grad-uate unemployment rate, or that graduateunemployment is not just a temporary phe-nomenon, as described in this study? Theanswer to this depends on a number of factors,

as for example whether enterprises in Greecewill increase their demand in high-skilledhuman capital or whether they will be willingto recruit graduate labour to fill positionswhere higher education qualifications are notrequired (Kikilias, 2008). If these prospects donot materialise, and if there is no significantbrain drain, then we may witness higher ratesof graduate unemployment and the phenome-non of graduate unemployment may turn outto be more permanent (Karamessini, 2003).

Finally, there are a number of policy implica-tions stemming from our results. It is obviousthat all curricula, especially in tertiary educa-tion, have to be upgraded and linked to theknowledge society and the labour market. Theyshould provide young people with the neces-sary skills to creatively address the great chal-lenges of modern world. Their content mustfocus on the development of cutting-edge andhigh-demand skills that respond to the needsof the Greek labour market, thus contributingin effectively shielding from unemployment.Mechanisms linking education to the labourmarket could also contribute to the achieve-ment of this goal since, as confirmed by ouranalysis, the main problem is not youth unem-ployment but the transition from education tothe labour market, irrespective of age. Theeffective operation and upgrading of suchmechanisms, e.g. the National System for Link-ing Vocational Education and Training toEmployment, the Career Offices of AEI andTEI and the services of the Greek ManpowerEmployment Organisation (OAED), couldhelp curb the problem of youth and graduateunemployment. Private human resourcesmanagement firms may also play an importantrole. However, contrary to what happens inother countries, the small number of suchenterprises in Greece focuses on recruitinghigh-skilled executives primarily on behalf ofmultinational corporations. This is probablydue to the structure of the Greek economy,with its large number of small and medium-sized enterprises, which do not turn to spe-cialised (especially private) entities to covertheir needs in human resources.

33Economic Bulletin

May 2010 47

33Economic Bulletin

May 2010 49

ANNEX

Single parent with children – male 0.2322

Single parent with children and other members – male 0.6288 ***

Couple with children – male 0.0029

Couple with children and other members – male 0.3106 ***

Single male 0.1746 ***

Couple without children – male Reference group

Other type of household, single parent (without children) – male 0.5408 ***

Other type of household, couple (without children) – male 0.4359 ***

Primary or lower education 0.2393 **

Lower secondary education -0.1521 *

General Lyceum Reference group

Technical Lyceum 0.0475

Post-gymnasium technical school -0.0827

IEK 0.2024 **

Other post-lyceum education -0.0453

Structural Engineering (TEI) 0.0228

Mechanical & Computer Engineering (TEI) 0.2402

Agricultural & Food Technology (TEI) 0.3476

Economics & Management (TEI) 0.2827 *

Other TEI 0.3133

Medical Sciences (TEI) 0.6851 **

Structural Engineering -0.0210

Mechanical Engineering -0.0659

Other AEI 0.1017

IT -0.9305

Horticulture & Forestry -0.0078

Medical School etc. 0.1395

Physical Sciences 0.5213 *

Mathematics & Statistics 0.8121 **

Law School -0.8395 *

Economics & Management 0.2698 **

Social Sciences 0.2705

Humanities 0.3698

Languages 0.6157

Physical Education & Sports 0.3725

Pedagogics 0.0076

Postgraduate degree 0.1547

Doctorate 0.0456

Independent variables Coefficient

Table A1 MALE – Estimated probability of unemployment

Dependent variable: unemployed=1

*** Statistically important coefficient at importance level 1%** Statistically important coefficient at importance level 5%* Statistically important coefficient at importance level 10%


Years since graduation x Primary or lower education 0.0100

Years since graduation x Lower secondary education 0.0275 ***

Years since graduation x General Lyceum Reference group

Years since graduation x Technical Lyceum -0.0115

Years since graduation x Post-gymnasium technical school 0.0065

Years since graduation x IEK -0.0243 *

Years since graduation x Other post-lyceum education 0.0174

Years since graduation x Structural Engineering (TEI) -0.0454

Years since graduation x Mechanical & Computer Engineering (TEI) -0.0589 **

Years since graduation x Agricultural & Food Technology (TEI) -0.0468

Years since graduation x Economics & Management (TEI) -0.0524 *

Years since graduation x Other TEI -0.0783

Years since graduation x Medical Sciences (TEI) -0.1116 *

Years since graduation x Structural Engineering -0.0418

Years since graduation x Mechanical Engineering -0.0104

Years since graduation x Other AEI 0.0122

Years since graduation x IT 0.2280

Years since graduation x Horticulture & Forestry -0.0111

Years since graduation x Medical School etc. 0.0153

Years since graduation x Physical Sciences -0.0603

Years since graduation x Mathematics & Statistics -0.1527 **

Years since graduation x Law School -0.0152

Years since graduation x Economics & Management -0.0655 ***

Years since graduation x Social Sciences -0.0431

Years since graduation x Humanities -0.0385

Years since graduation x Languages -0.0410

Years since graduation x Physical Education & Sports -0.0033

Years since graduation x Pedagogics -0.0093

Years since graduation x Postgraduate degree -0.0396

Years since graduation x Doctorate -0.0469


Table A1 MALE – Estimated probability of unemployment (continued)



33Economic Bulletin

May 2010 51

Years since graduation sqrd./100 x Primary or lower education -0.0209

Years since graduation sqrd./100 x Lower secondary education -0.0498 **

Years since graduation sqrd./100 x General Lyceum Reference group

Years since graduation sqrd./100 x Technical Lyceum 0.0504

Years since graduation sqrd./100 x Post-gymnasium technical school 0.0008

Years since graduation sqrd./100 x IEK 0.0561

Years since graduation sqrd./100 x Other post-lyceum education -0.0026

Years since graduation sqrd./100 x Structural Engineering (TEI) 0.0763

Years since graduation sqrd./100 x Mechanical & Computer Engineering (TEI) 0.1438 **

Years since graduation sqrd./100 x Agricultural & Food Technology (TEI) 0.1290

Years since graduation sqrd./100 x Economics & Management (TEI) 0.1311

Years since graduation sqrd./100 x Other TEI 0.2277

Years since graduation sqrd./100 x Medical Sciences (TEI) 0.1535

Years since graduation sqrd./100 x Structural Engineering 0.1053

Years since graduation sqrd./100 x Mechanical Engineering 0.0577

Years since graduation sqrd./100 x Other AEI -0.0102

Years since graduation sqrd./100 x IT -2.0429

Years since graduation sqrd./100 x Horticulture & Forestry -0.0133

Years since graduation sqrd./100 x Medical School etc. -0.3710

Years since graduation sqrd./100 x Physical Sciences 0.0319

Years since graduation sqrd./100 x Mathematics & Statistics 0.3621 **

Years since graduation sqrd./100 x Law School 0.0713

Years since graduation sqrd./100 x Economics & Management 0.1761 ***

Years since graduation sqrd./100 x Social Sciences 0.1690

Years since graduation sqrd./100 x Humanities 0.0074

Years since graduation sqrd./100 x Languages -0.6442

Years since graduation sqrd./100 x Physical Education & Sports -0.0494

Years since graduation sqrd./100 x Pedagogics -0.0969

Years since graduation sqrd./100 x Postgraduate degree 0.1188

Years since graduation sqrd./100 x Doctorate -0.1342

Years since graduation -0.0311 ***

Years since graduation sqrd./100 0.0213


Table A1 MΑLE – Estimated probability of unemployment (continued)




Greek national Reference group

Other EU national -0.2253

Third country national -0.2265 ***

East Macedonia–Thrace 0.1591 ***

Central Macedonia 0.1442 ***

West Macedonia 0.1558 **

Epirus 0.1578 ***

Thessaly 0.1090 **

Ionian Islands 0.2150 ***

West Greece 0.0682

Central Greece and Euboea 0.0949 **

Attica Reference group

Peloponnese 0.1173 ***

North Aegean 0.0375

South Aegean 0.2027 ***

Crete 0.1552 ***

Capital region – Attica Reference group

City complex – Thessaloniki -0.1974 ***

Other urban areas -0.1154 ***

Semi-urban areas -0.2366 ***

Rural areas -0.3146 ***

First year of survey (2004) -0.0644 **

Second year of survey (2005) -0.0190

Third year of survey (2006) -0.0154

Fourth year of survey (2007) Reference group

1st quarter 0.0260

2nd quarter -0.0206

3rd quarter Reference group

4th quarter 0.0343

Regional rate of unemployment 5.2382 ***

Constant term -1.8455 ***


Table A1 MALE – Estimated probability of unemployment (continued)



33Economic Bulletin

May 2010 53

Single parent with children (0 aged up to 5, 1+ aged 6-17) – male -0.8720 ***

Single parent with children (1+ aged up to 5, 0+ aged 6-17) – male -0.4188

Single parent with children (0 aged up to 5, 1 aged 6-17) and other members – male -0.1822 **

Single parent with children (0 aged up to 5, 2 aged 6-17) and other members andSingle parent with children (0+ aged up to 5, other aged 6-17) and other members – male

0.2099

Couple with children (0 aged up to 5, 1 aged 6-17) – male -0.2698 ***

Couple with children (0 aged up to 5, 2 aged 6-17) – male -0.3715 ***

Couple with children (0 aged up to 5, 3+ aged 6-17) – male -0.5677 ***

Couple with children (1 aged up to 5, 0 aged 6-17) – male 0.9621 ***


Couple with children (1 aged up to 5, 2 aged 6-17) – male -0.1258

Couple with children (1 aged up to 5, 3+ aged 6-17) – male -0.4203 ***


Couple with children (2 aged up to 5, 1 aged 6-17) and Couple with children (2 aged up to 5, 2 aged 6-17)and Couple with children (2 aged up to 5, 3+ aged 6-17) – male

0.4779 ***

Couple with children (3+ aged up to 5, 0+ aged 6-17) – male 0.3124

Couple with children (0 aged up to 5, 1 aged 6-17) and other members – male -0.1442 ***

Couple with children (0 aged up to 5, 2 aged 6-17) and other members – male -0.2168 ***

Couple with children (0 aged up to 5, 3+ aged 6-17) and other members – male -0.4081 ***

Couple with children (1 aged up to 5, 0 aged 6-17) and other members – male 0.3859 ***

Couple with children (1 aged up to 5, 1 aged 6-17) and other members – male -0.0453

Couple with children (1 aged up to 5, 2 aged 6-17) and other members – male -0.3340

Couple with children (1 aged up to 5, 3+ aged 6-17) and other members andCouple with children (2+ aged up to 5, 0+ aged 6-17) and other members – male

0.1775

Single male -0.2780 ***

Couple without children – male Reference group

Other type of household, single parent (without children) – male 0.0752 **

Other type of household, couple (without children) – male 0.0989 ***

Primary or lower education 0.9370 ***

Lower secondary education -0.1058 ***


Technical Lyceum 0.5414 ***

TEI 0.9062 ***

AEI 0.5051 ***

Postgraduate degree 1.2741 ***

Years since graduation 0.2280 ***

Years since graduation sqrd./100 -0.4791 ***


Table B1 MALE – Estimated probability of participation in labour force

Dependent variable: labour force participant=1

Note: Wald test (rho = 0): x2(1) = 21.80 Prob > x2 = 0.0000.*** Statistically important coefficient at importance level 1%** Statistically important coefficient at importance level 5%* Statistically important coefficient at importance level 10%


Eastern Macedonia–Thrace -0.0539

Central Macedonia -0.0167

Western Macedonia -0.2810 ***

Epirus -0.1581 ***

Thessaly 0.0655 *

Ionian Islands -0.1276 **

Western Greece -0.1348 ***

Central Greece and Euboea -0.0373


Peloponnese 0.0196

North Aegean -0.1632 ***

South Aegean 0.0359

Crete -0.0572 *

Attica capital region Reference group

Thessaloniki city complex -0.0090

Other urban areas 0.0581 **

Semi-urban areas 0.2506 ***

Rural areas 0.3917 ***

First year of survey (2004) -0.0118


Third year of survey (2006) -0.0063


1st quarter -0.0257

2nd quarter -0.0220


4th quarter 0.0133


/athrho 0.5143 ***

rho 0.4733


Table B1 MALE – Estimated probability of participation in labour force (continued)


Note: Wald test (rho = 0): x2(1) = 21.80 Prob > x2 = 0.0000.*** Statistically important coefficient at importance level 1%** Statistically important coefficient at importance level 5%* Statistically important coefficient at importance level 10%

33Economic Bulletin

May 2010 55

Single parent with children – female 0.2290 ***

Single parent with children and other members – female 0.3122 ***

Couple with children – female 0.0425

Couple with children and other members – female 0.1301 ***

Single female -0.0417

Couple without children – female Reference group

Other type of household, single parent (without children) – female 0.2758 ***

Other type of household, couple (without children) – female 0.2364 ***

Primary or lower education -0.1336

Lower secondary education -0.1995 **



Post-gymnasium technical school 0.1110

IEK 0.3416 ***

Other post-lyceum education 0.0879

Structural Engineering (TEI) 0.1420

Mechanical & Computer Engineering (TEI) 0.5612 ***

Agricultural & Food Technology (TEI) 0.5574 ***

Economics & Management (TEI) 0.3562 ***

Other TEI 0.6333 **

Medical Sciences (TEI) 0.4790 ***

Structural Engineering 0.3091

Mechanical Engineering -0.0333

Other AEI 0.1390

IT -0.1075

Horticulture & Forestry 0.4771

Medical School etc. 0.2780

Physical Sciences 0.9142 ***

Mathematics & Statistics 1.0191 ***

Law School -0.0904

Economics & Management 0.2610 **

Social Sciences 0.6999 ***

Humanities 0.5627 ***

Languages 0.0490

Physical Education & Sports -0.1760

Pedagogics 0.3841 **

Postgraduate degree 0.2291

Doctorate 0.2596


Table A2 FEMALE – Estimated probability of unemployment




Years since graduation x Primary or lower education 0.0446 ***

Years since graduation x Lower secondary education 0.0246 ***

Years since graduation x General Lyceum Reference group

Years since graduation x Technical Lyceum -0.0104

Years since graduation x Post-gymnasium technical school 0.0023

Years since graduation x IEK -0.0219 **

Years since graduation x Other post-lyceum education -0.0075

Years since graduation x Structural Engineering (TEI) -0.0549

Years since graduation x Mechanical & Computer Engineering (TEI) -0.1250 ***

Years since graduation x Agricultural & Food Technology (TEI) -0.0459

Years since graduation x Economics & Management (TEI) -0.0247

Years since graduation x Other TEI -0.1123 **

Years since graduation x Medical Sciences (TEI) -0.0763 ***

Years since graduation x Structural Engineering -0.1036 ***

Years since graduation x Mechanical Engineering -0.0910 *

Years since graduation x Other AEI -0.0534

Years since graduation x IT 0.0570

Years since graduation x Horticulture & Forestry 0.0055

Years since graduation x Medical School etc. -0.0860 ***

Years since graduation x Physical Sciences -0.1088 **

Years since graduation x Mathematics & Statistics -0.1669 ***

Years since graduation x Law School -0.0504

Years since graduation x Economics & Management -0.0560 ***

Years since graduation x Social Sciences -0.1145 ***

Years since graduation x Humanities -0.0661 ***

Years since graduation x Languages -0.0385

Years since graduation x Physical Education & Sports 0.0785

Years since graduation x Pedagogics -0.0593

Years since graduation x Postgraduate degree -0.0801 **

Years since graduation x Doctorate -0.0890 *


Table A2 FEMALE – Estimated probability of unemployment (continued)



33Economic Bulletin

May 2010 57

Years since graduation sqrd./100 x Primary or lower education -0.0943 ***

Years since graduation sqrd./100 x Lower secondary education -0.0453 **

Years since graduation sqrd./100 x General Lyceum Reference group

Years since graduation sqrd./100 x Technical Lyceum 0.0116

Years since graduation sqrd./100 x Post-gymnasium technical school -0.0136

Years since graduation sqrd./100 x IEK 0.0564 **

Years since graduation sqrd./100 x Other post-lyceum education 0.0721

Years since graduation sqrd./100 x Structural Engineering (TEI) 0.1944

Years since graduation sqrd./100 x Mechanical & Computer Engineering (TEI) 0.3986 ***

Years since graduation sqrd./100 x Agricultural & Food Technology (TEI) 0.1653

Years since graduation sqrd./100 x Economics & Management (TEI) 0.1020 *

Years since graduation sqrd./100 x Other TEI 0.3469 **

Years since graduation sqrd./100 x Medical Sciences (TEI) 0.1429 *

Years since graduation sqrd./100 x Structural Engineering 0.2845 **

Years since graduation sqrd./100 x Mechanical Engineering 0.3187 **

Years since graduation sqrd./100 x Other AEI 0.2010

Years since graduation sqrd./100 x IT -2.1860

Years since graduation sqrd./100 x Horticulture & Forestry -0.2124

Years since graduation sqrd./100 x Medical School etc. 0.1787 **

Years since graduation sqrd./100 x Physical Sciences 0.1397

Years since graduation sqrd./100 x Mathematics & Statistics 0.3821 **

Years since graduation sqrd./100 x Law School 0.0733

Years since graduation sqrd./100 x Economics & Management 0.1242 **

Years since graduation sqrd./100 x Social Sciences 0.3136 ***

Years since graduation sqrd./100 x Humanities 0.0725

Years since graduation sqrd./100 x Languages 0.1115

Years since graduation sqrd./100 x Physical Education & Sports -0.4890 *

Years since graduation sqrd./100 x Pedagogics 0.0650

Years since graduation sqrd./100 x Postgraduate degree 0.2242 *

Years since graduation sqrd./100 x Doctorate 0.2503 *

Years since graduation -0.0106 **

Years since graduation sqrd./100 -0.0257 **






Greek national Reference group

Other EU national 0.0886

Third country national -0.0209



West Macedonia 0.1497 **

Epirus 0.2442 ***

Thessaly 0.2709 ***

Ionian Islands 0.1629 ***

West Greece 0.1769 ***

Central Greece and Euboea 0.2305 ***



North Aegean 0.2930 ***

South Aegean 0.1607 ***

Crete 0.2483 ***

Capital region – Attica Reference group

City complex – Thessaloniki -0.2156 ***


Semi-urban areas -0.1664 ***

Rural areas -0.2119 ***



Third year of survey (2006) 0.0129


1st quarter 0.0068

2nd quarter -0.0161


4th quarter -0.0021

Regional rate of unemployment 5.6842 ***






33Economic Bulletin

May 2010 59

Single parent with children (0 aged up to 5, 1 aged 6-17) – female 0.2829 ***

Single parent with children (0 aged up to 5, 2 aged 6-17) – female 0.1740 **

Single parent with children (0 aged up to 5, 3+ aged 6-17) – female -0.0355

Single parent with children (1 aged up to 5, 0 aged 6-17) – female 0.6509 ***

Single parent with children (1 aged up to 5, 1 aged 6-17) – female -0.1298

Single parent with children (1 aged up to 5, 2+ aged 6-17) – female -0.0771

Single parent with children (2+ aged up to 5, 0+ aged 6-17) – female -0.3168

Single parent with children (0 aged up to 5, 1 aged 6-17) and other members – female 0.2237 ***

Single parent with children (0 aged up to 5, 2+ aged 6-17) and other members – female 0.1241

Single parent with children (1+ aged up to 5, 0+ aged 6-17) and other members – female 0.1153

Couple with children (0 aged up to 5, 1 aged 6-17) – female -0.2394 ***


Couple with children (0 aged up to 5, 3+ aged 6-17) – female -0.3850 ***









Couple with children (3+ aged up to 5, 0+ aged 6-17) – female -0.7267 ***

Couple with children (0 aged up to 5, 1 aged 6-17) and other members – female -0.2126 ***


Couple with children (0 aged up to 5, 3+ aged 6-17) and other members – female -0.3372 ***

Couple with children (1 aged up to 5, 0 aged 6-17) and other members – female -0.0799

Couple with children (1 aged up to 5, 1 aged 6-17) and other members – female -0.0237


Couple with children (1 aged up to 5, 3+ aged 6-17) and other members and -0.4717 ***

Couple with children (2+ aged up to 5, 0+ aged 6-17) and other members – female -0.3842 ***

Single female -0.0037

Couple without children – female Reference group

Other type of household, single parent (without children) – female 0.2677 ***

Other type of household, couple (without children) – female 0.0292 *


Table Β2 FEMALE – Estimated probability of participation in labour force


Note: Wald test (rho = 0): x2(1) =53.40 Prob > x2 = 0.0000*** Statistically important coefficient at importance level 1%** Statistically important coefficient at importance level 5%* Statistically important coefficient at importance level 10%


Primary or lower education 0.0901 ***

Lower secondary education -0.3163 ***



TEI 1.0967 ***

AEI 0.9779 ***

Postgraduate degree 1.5844 ***

Years since graduation 0.0906 ***

Years since graduation sqrd./100 -0.2136 ***



West Macedonia 0.0554 *

Epirus 0.0815 ***

Thessaly 0.1502 ***

Ionian Islands 0.0931 **

West Greece -0.0107

Central Greece and Euboea 0.0750 ***



North Aegean -0.1056 ***

South Aegean -0.0672 **

Crete 0.2336 ***

Attica capital region Reference group

Thessaloniki city complex -0.1756 ***


Semi-urban areas 0.0261

Rural areas 0.2388 ***


Second year of survey (2005) 0.0048

Third year of survey (2006) 0.0031


1st quarter -0.0009

2nd quarter 0.0175


4th quarter -0.0319 *


/athrho 1.0758 ***

rho 0.7916


Table Β2 FEMALE – Estimated probability of participation in labour force (continued)


Note: Wald test (rho = 0): x2(1) =53.40 Prob > x2 = 0.0000*** Statistically important coefficient at importance level 1%** Statistically important coefficient at importance level 5%* Statistically important coefficient at importance level 10%

33Economic Bulletin

May 2010 61

Bitros, G. (2002), “Το εκπαιδευτικό µας σύστηµα και η αποτελεσµατική συσχέτιση θέσεων εργα-σίας και δεξιοτήτων των εργαζοµένων ―Οι Οικονοµικές Σχολές” [Our education system andthe effective matching of jobs and workers’ skills― Economic Schools], in Verevis, Α. (ed.),Το Έργο “Έρευνα” 1997-2000. Συνοπτική Παρουσίαση [The “Research” Project 1997-2000.Overview], Education Research Centre, Athens [in Greek].

Cholezas, Ι. and P. Tsakloglou (2008), “Η απασχολησιµότητα των αποφοίτων της τριτοβάθµιαςεκπαίδευσης: εννοιολογικές προσεγγίσεις και επισκόπηση της διεθνούς και ελληνικής εµπει-ρικής βιβλιογραφίας” [The employability of tertiary education graduates: conceptual approachesand review of the international and Greek empirical literature], mimeo, Department of Inter-national and European Economic Studies, Athens University of Economics and Business(AUEB), Athens [in Greek].

Georganta, Ζ., E. Kandilorou and A. Livada (2008), “Η έλλειψη δεξιοτήτων µάθησης αποτελείτροχοπέδη στην πανεπιστηµιακή γνώση” [The lack of learning skills is an obstacle to univer-sity knowledge], Technical Report 236, Department of Statistics, AUEB, Athens [in Greek].

Greene, W.H. (2003), Econometric Analysis, 5th ed., Prentice Hall, Upper Saddle River, NJ.Heckman, J. (1979), “Sample selection bias as a specification error”, Econometrica, 47, 153-62.IOBE (2007), Η απασχολησιµότητα των αποφοίτων τριτοβάθµιας εκπαίδευσης στην Ελλάδα

[Employability of tertiary education graduates in Greece], IOBE Final Report, Athens [inGreek].

ISTAME (2006), Απασχόληση: κίνδυνοι, προκλήσεις και προοπτικές [Employment: risks, chal-lenges and prospects], ISTAME Committee on Employment, Athens [in Greek].

Karamessini, Μ. (2003), “Ζήτηση και προσφορά υπηρεσιών ανώτατης εκπαίδευσης και αγοράεργασίας στην Ελλάδα” [Demand and supply of higher education services and labour marketin Greece], in Tsaoussis, D.G. (ed.), Από τη ∆ιεθνοποίηση των Πανεπιστηµίων προς την Παγκο-σµιοποίηση της Εκπαίδευσης – Εξελίξεις µιας ∆εκαπενταετίας – Η Περίπτωση της Ελλάδας[From the internationalisation of universities to the globalisation of education – Developmentsof the last fifteen years – The case of Greece], Morphology and Social Policy Centre, PanteionUniversity, Athens [in Greek].

Karamessini, Μ. (2006a), “Από την εκπαίδευση στην αµειβόµενη εργασία: εµπειρική διε-ρεύνηση της εργασιακής ένταξης των νέων στην Ελλάδα” [From education to waged work:An empirical investigation of the entry of young people in the labour market in Greece],Koinoniki Synohi kai Anaptyxi, 1(1), 67-84 [in Greek].

Karamessini, Μ. (2006b), Η Απορρόφηση των Αποφοίτων του Πανεπιστηµίου Θεσσαλίας στηνΑγορά Εργασίας [Absorption of graduates of the University of Thessaly in the labour market],University of Thessaly, Volos [in Greek].

Karamessini, Μ. (2008), Η Απορρόφηση των πτυχιούχων πανεπιστηµίου στην αγορά εργασίας[Absorption of university graduates in the labour market], Dionikos editions, Athens [in Greek].

Karamessini, Μ. and Α. Kornelakis (2005), “Από την εκπαίδευση στην απασχόληση: Η πρώτησηµαντική εργασία των αποφοίτων ΑΕΙ και η εργασιακή εµπειρία µέσω STAGE” [“From edu-cation to employment: The first important job of university graduates and the work experiencethrough the STAGE programme”, Employment 2005, Institute of Urban Environment andHuman Resources, Panteion University, Athens [in Greek].

Karamessini, Μ. and Ε. Prokou (2006), “Τριτοβάθµια εκπαίδευση και απασχόληση: επαγγελ-µατική ένταξη των αποφοίτων και εκπαιδευτική πολιτική” [Tertiary education and employment:Insertion of graduates into the labour market and education policy], paper presented at theInternational Conference: Social Policy in a changing environment: Challenges and Perspec-tives, Panteion University, Athens [in Greek].

Kassimis, C. (2002), “∆ιαδικασίες ενσωµάτωσης των νέων στην αγορά εργασίας: η περίπτωσητων αποφοίτων Κοινωνικών Επιστηµών της Ανώτατης Εκπαίδευσης” [“Processes for the inte-

R E F E R ENC E S


gration of young people in the labour market: the case of higher-education graduates of SocialSciences”, in Verevis, A. (ed.), Το Έργο “Έρευνα” 1997-2000. Συνοπτική Παρουσίαση [The“Research” Project 1997-2000. Overview], Education Research Centre, Athens [in Greek].

Katsanevas, Th. and E. Livanos (2005), “Η πορεία και τα αίτια της ανεργίας στην Ελλάδα”[Intertemporal development and reasons of unemployment in Greece], Epitheorisi ErgasiakonSheseon, 40, 4-12 [in Greek].

Kikilias, E. (2008), “Οι τάσεις, τα χαρακτηριστικά της απασχόλησης και η µετάβαση στην εργα-σία των αποφοίτων τριτοβάθµιας εκπαίδευσης στην Ελλάδα” [Trends, aspects of employmentand transition of tertiary education graduates to the labour market in Greece], mimeo, NationalCentre for Social Research, Athens [in Greek].

Liagouras, G., A. Protogerou and Y. Calogirou (2003), “Exploring Mismatches between HigherEducation and the Labour Market in Greece”, European Journal of Education, 38(4), 413-26.

Nicolitsas, D. (2007), “Youth participation in the Greek labour market: developments and obsta-cles”, Bank of Greece, Economic Bulletin, 29, 37-88.

NTUA (2002), Η απορρόφηση των µηχανικών του ΕΜΠ στην αγορά εργασίας [Absorption ofNTUA engineers in the labour market], Laboratory of Industrial and Energy Economics andNational Labour Institute, National Technical University of Athens, Athens [in Greek].

OECD (2006), Labour Force Statistics, OECD, Paris.OECD (2009), Education at a Glance 2009, OECD Indicators, OECD, Paris.Papailias, Th. (2006), Προσφορά και ζήτηση ειδικοτήτων στο ελληνικό εκπαιδευτικό σύστηµα

[Supply and demand for specialties in the Greek education system], Dionikos editions, Athens[in Greek].

Pindyck, R. and D. Pubinfeld (1998), Econometric Models and Economic Forecasts, 4th ed.,McGraw-Hill, New York.

SEV (2004), “Έρευνα για τις ανάγκες των επιχειρήσεων σε ειδικότητες αιχµής την τριετία 2005-2007” [Survey on the needs of enterprises in cutting-edge skills in the three-year period 2005-2007], SEV, Athens [in Greek]; also available at: http://www.fgi.org.gr/Uploads/largefiles/Anages_Eidikotiton_Aixmis_Teliko.pdf.

ΤΕΕ (2006), Έρευνα για την Απασχόληση των ∆ιπλωµατούχων Μηχανικών [Survey on grad-uate engineers’ employment], Technical Chamber of Greece, Athens [in Greek].

White, H. (1980), “A Heteroskedasticity – Consistent Covariance Matrix Estimator and a DirectTest for Heteroskedasticity”, Econometrica, 48(4), 817-38.

33Economic Bulletin

May 2010 63

The goal of monetary policy in most advancedeconomies is that of price stability, usuallydefined as some low rate of inflation. This isseen as crucial to providing a predictable envi-ronment, where expectations of future inflationare anchored. The rationale for such a policygoal is founded on the idea that inflation bringsboth direct and indirect costs.

Understanding the nature of these costs hasproved to be a fundamental task for theoreti-cal economists. A consensus that has emergedviews moderate anticipated inflation as beingassociated with small welfare costs. Rather, itis unexpected inflation that is the source of thecosts of inflation. Uncertainty about futureinflation is generally thought to distort therelative-price mechanism, leading to a misal-location of resources and thus harming growth.

Within the context of the literature on the costsof inflation,1 the aspect that has often been thesubject of academic debate is the relationshipbetween inflation and inflation uncertainty.Some have argued that higher inflation directlycauses inflation uncertainty, thus generatingreal indirect costs from inflation for any econ-omy (Friedman, 1977, Ball, 1992); others holdthe view that the direction of causality isreversed, with higher inflation uncertaintycausing higher average inflation (Cukiermanand Meltzer, 1986; Cukierman, 1992).

The purpose of this paper is to examine therelationship between inflation and inflationuncertainty for Greece covering the periodfrom 1981 to 2008. This relationship has notbeen studied for Greece before. Yet Greece,as a country which has experienced both peri-ods of high and variable inflation and, morerecently, low inflation, could well prove aninteresting testing ground for the competing

theories. Univariate GARCH models are usedto generate alternative measures of inflationuncertainty, using both seasonally unadjustedand seasonally adjusted data. Subsequently,Granger causality tests are employed to inves-tigate causality between inflation and inflationuncertainty and, where present, to detect itsdirection and sign.

The paper proceeds as follows: in Section 1 weconsider the extensive theoretical and empir-ical academic literature on both the costs ofinflation and the possible relationship betweeninflation and inflation uncertainty; Section 2describes our data and Section 3 outlines anddiscusses our empirical approach; Section 4presents our findings, and Section 5 providessome concluding remarks.

1 THE COSTS OF INFLATION: A SELECTEDREVIEW OF THE LITERATURE

Fully anticipated inflation can generate costs foran economy from two main sources. First thereare search costs (often referred to as “shoe-leather” costs). As inflation rises, holding moneybecomes more costly because it loses its valueover time (that is, a given amount can purchasea smaller basket of goods as time progresses).Individuals will thus spend their money closer tothe time when they receive their income andshops will hold smaller stocks, which impliesmore queuing or searching. At the same time, toeconomise on holding cash, more will be held inbanks, which implies the need for more trips tothe bank (hence the “shoe-leather” costs). Allthese consequences of effectively taxing theholdings of money balances generate costly andunproductive activities in any economy.

I N F L A T I ON AND NOM INA L UNC ER T A I N T Y :TH E C A S E O F GR E E C E

Heather D. GibsonEconomic Research Department

Hiona BalfoussiaEconomic Research Department

11 For a good early survey of the literature, see Driffill et al. (1990).

The second main cost of anticipated inflationarises from the fact that tax systems are oftennot indexed to inflation. Consider a tax systemwith various tax bands, where the tax rates areunchanged. In an economy with positive infla-tion, effective tax rates (the actual rate paid)will usually rise with inflation. This leads to dis-tortions with respect to both private savingsand labour supply decisions.

It is often argued that the costs associated withanticipated inflation are likely to be small. Thisis particularly true in a world where tax systemsare indexed and where cash transactions arelimited because of new electronic means ofpayment. However, it is unlikely that inflationcan ever be fully anticipated. This leads econ-omists to focus on the costs from unanticipatedinflation. Unanticipated inflation can generatewelfare costs because of its impact on the pricemechanism’s ability to allocate resources effi-ciently (that is, to where they will do the mostgood).

An early attempt at explaining this was pro-vided by Lucas (1973). In his model, unantic-ipated inflation causes economic agents to con-fuse absolute and relative price changes.Assume producers, for example, know theprice of the goods they produce or trade, at anypoint in time; the absolute price level in theeconomy as a whole is, however, known only inthe next period. Assume there is an unexpectedrise in the money supply. The demand for allgoods will rise and, for given output levels,prices will rise. Each individual producer expe-riences a rise in the price of the good they pro-duce. Thinking this to be a rise only in the priceof their good and not a general rise in prices,they increase output. Of course in the nextperiod they realise that the increase in priceswas a general one and not specific to the goodthey produce; their response of increasing out-put was inappropriate, since there was noincrease in relative demand.

In short, inflation causes agents to confuse realshocks, which warrant a change in quantities(output, labour supply, investment, etc.), with

nominal shocks (or monetary shocks), whichdo not. If economic agents realise that bothkinds of shocks are possible, but neither is pre-dictable, then they will tend to overreact tomonetary shocks and underreact to realshocks. It is this kind of story that leads econ-omists to argue that inflation can be costlywhen it is unpredictable – it distorts the pricemechanism in a way that changes in prices(either of goods, services or assets) no longerreflect underlying real changes in the economywhich may warrant a change in the allocationof resources to certain activities and henceinvestment in one sector rather than another.

However, if it is inflation unpredictability oruncertainty that generates costs, the interest-ing question arises as to whether inflationuncertainty is connected in some way to theactual level of inflation. Several, often con-tradictory theoretical arguments can befound in the academic literature regarding thepossible direction of a causal relationshipbetween inflation and inflation uncertainty.Okun (1971) was perhaps the first to suggestthat countries with high rates of inflation werealso likely to experience high inflation vari-ability. The underlying theoretical insight forthis argument was provided by Friedman(1977) in his Nobel lecture. He reasoned that,while high inflation generates political pressureto reduce it, policy makers are usually reluctantto do so, for fear of the possible recessionaryeffects of such an effort. This lack of strongcommitment on behalf of the monetary policyauthority creates nominal uncertainty andhigher inflation volatility, which in turn resultsin instability in consumption, production andinvestment decisions, and hence growth andwelfare losses. All these consequences justifythe public’s underlying aversion to inflation.

Friedman’s well-known insight was first for-malised by Ball (1992) in the context of a gameof asymmetric information between the mon-etary authority and the public, in which higherinflation leads to increasing uncertainty overthe monetary policy stance. Economic agentsknow there are two types of policy makers, only


one of which is willing to bear the economiccosts of reducing inflation. When inflation islow, both types of policy makers strive to keepit so, giving rise to low inflation uncertainty asperceived by economic agents. At higher lev-els of inflation, however, only one type is will-ing to disinflate, the other being reluctant to doso for fear of causing a recession. Given thatthe policy makers stochastically alternate inoffice, an increase in inflation creates uncer-tainty about future monetary policy, futureinflation and thus the general prospects of thereal economy. This line of argument is oftenreferred to in the literature as the Friedman-Ball hypothesis. Its conclusion is that inflationvery clearly causes inflation uncertainty andthe higher the rate of inflation, the greater thatuncertainty will be. Thus inflation is costly notonly because of its direct costs, but alsobecause of its indirect costs arising from itsimpact on inflation uncertainty.

Another strand of the academic literature pur-sues the idea that causality may run in theopposite direction, i.e. that in fact it is greaterinflation uncertainty which causes higher aver-age inflation. This possibility has been for-malised as a feature of models based on theBarro-Gordon framework. Seminal among thisbody of work is that of Cukierman and Meltzer(1986) and Cukierman (1992), in whose modelthe low policy credibility, the ambiguity ofobjectives and the poor quality of monetarycontrol that characterise policy makersincrease the average rate of inflation. Themonetary policy authority has a dual mandate,containing inflation and promoting economicgrowth, which is reflected in its stochasticobjective function, as is the trade-off betweenthe two objectives. However there is no com-mitment mechanism. The monetary policyauthority has an incentive to create monetarysurprises ―i.e. to generate inflation uncer-tainty― in an effort to stimulate economicgrowth; this in turn leads to increases in thelevel of inflation. The money supply process isalso assumed to have a random component,due to the monetary authority’s inability to dis-cern the most appropriate monetary policy

instrument and to precisely control it. There-fore, not only are economic agents uncertainabout the level of future inflation but, more-over, they have no way of inferring whether anincrease in the observed level of inflation isdue to a random money supply disturbance orto a shift in the government’s emphasis onunemployment. In this context, higher inflationuncertainty generates higher inflation and isevidence of an “opportunistic” or “myopic”central bank.2

It is not only the direction of causality that isquestioned – that is, whether it is inflation thatcauses inflation uncertainty or vice-versa. It isalso the sign. In both Friedman-Ball andCukierman-Meltzer, higher inflation is asso-ciated with higher inflation uncertainty. Fol-lowing a different thread, Holland (1995) pro-poses that, if the Friedman-Ball hypothesis isvalid, there may also be a secondary feedbackeffect from inflation uncertainty back to infla-tion, as a result of the policy makers’ stabili-sation efforts. As inflation uncertainty risesdue to increasing inflation, there is anincreased incentive for policy makers torespond by contracting money supply growthand contain inflation in order to reduce infla-tion uncertainty and the associated negativereal output and welfare effects. Thus, a nega-tive causal effect of inflation uncertainty oninflation is evidence of what is known as the“stabilisation motive” of the monetary policyauthority, which views inflation uncertainty asa welfare cost. It generates the result thathigher inflation uncertainty lowers inflationbecause of the monetary authorities’ response.

The final case argues that it is inflation thatcauses inflation uncertainty but that, in con-

33Economic Bulletin

May 2010 65

22 In a comparable theoretical setup, Devereux (1989) justifies thecausal link from inflation uncertainty to the level of inflation, bymodelling discretionary monetary policy as incorporating not onlyinflation but also real disturbances and the degree of endogenouswage indexation. He argues that a rise in the variance of real dis-turbances in the economy lowers the optimal degree of wage index-ation in the labour market and provides a stronger incentive for pol-icy makers to create surprise inflation, thereby increasing the aver-age rate of inflation in a “discretionary equilibrium”. Lower index-ation provides a stronger incentive for policy makers to create sur-prise inflation, and thereby inflation uncertainty, which can beargued to precede the higher average rate of inflation.

trast to the Friedman-Ball theory, higher infla-tion reduces inflation uncertainty. Pourgeramiand Maskus (1987) and Ungar and Zilberfarb(1993) argue that, as inflation increases, soagents invest more resources in forecastinginflation thus, in theory at least, reducing infla-tion uncertainty.

The different hypotheses on the link betweeninflation and inflation uncertainty have givenrise to a large empirical literature. Using USdata, Ball and Cecchetti (1990) investigate theeffects of inflation on inflation uncertainty atdifferent horizons and find that they are muchstronger at long horizons. This finding impliesincreased inflation risks for individual agentswho commit to nominal contracts, as well aslarge monetary policy swings and output fluc-tuations. Evidence in support of the Friedman-Ball hypothesis for US data is also presented inHolland (1995) and Caporale and McKiernan(1997). Grier and Perry (2000), using post-warUS data, fail to find any evidence that it is infla-tion uncertainty which causes inflation.Indeed, their estimation results indicate that,in line with Friedman’s argument, higher infla-tion uncertainty does significantly lower outputgrowth. Moreover, Hwang (2001) considers USdata for the sample periods 1926-1992 and1947-1992 and finds that, while it is inflationthat causes inflation uncertainty, the effect isnegative – in line with Pourgerami and Maskus(1987) and Ungar and Zilberfarb (1993).

Grier and Perry (1998) undertake a similarinvestigation for the G7 countries using datafor the period 1948-1993. They report thathigher inflation significantly raises inflationuncertainty in all countries. Mixed evidence isobtained for the reverse causal relationship, asincreased inflation uncertainty appears tolower inflation in the US, the UK and Ger-many, in line with Holland’s concept of a “sta-bilisation motive” for the monetary policyauthority, and to raise it in Japan and France,as predicted by the Cukierman-Meltzermodel of “opportunistic” central bank behav-iour. The authors note that these differentialresponses to inflation uncertainty are corre-

lated with Cukierman’s (1992) ratings of cen-tral bank independence, with Japan andFrance ranking less independent than the rest.3

The results of Hwang (2001) and Apergis(2004) for the G7 economies are in the samedirection – higher inflation causes inflationuncertainty but the reverse is not necessarilytrue. More recently, in the tests in Henry,Olekalns and Suardi (2007) on the G7, Fried-man-Ball effects are reported for the US, theUK and Canada.

Fountas, Ioannidis and Karanasos (2004) con-sider 6 major EU countries and report Fried-man-Ball effects in all of them. They also findthat in Germany and the Netherlandsincreased inflation uncertainty lowers inflation,while in Italy, Spain and, to a lesser extent,France increased inflation uncertainty raisesinflation. Caporale and Kontonikas (2006)empirically investigate the presence of struc-tural breaks in the relationship between infla-tion and inflation uncertainty in 12 EMU coun-tries and find that, with the introduction of theeuro, any relationship that existed appears tohave broken down in many countries.

For the UK in particular, Fountas (2001), usinga century of data, provides strong evidence infavour of the hypothesis that inflationary peri-ods are associated with high inflation uncer-tainty. Likewise, Kontonikas (2004), whose sam-ple period spans from 1972 to 2002, estimatesa positive relationship between past inflationand uncertainty about future inflation in theUK. Moreover, by controlling for the indirecteffects of lower average inflation ―which havehistorically coincided with inflation targetingperiods― he concludes that the adoption of anexplicit inflation target eliminates inflation per-sistence and reduces long-run uncertainty.

Finally, turning to recent multi-countryresearch, the results of Thornton (2007) for 12emerging market economies provide strong


.33 Berument and Dincer (2005) support these findings in a G7 sam-ple from 1957 to 2001, with the exception that in their work Japanis the only country for which increased inflation uncertainty raisesinflation.

support for the Friedman-Ball hypothesis.Daal, Naka and Sanchez (2005) consider alarge panel comprising 22 developed andemerging economies and find that, for mostcountries, inflation causes inflation uncer-tainty, while evidence for causality in the oppo-site direction is mixed. Their results also indi-cate that positive inflationary shocks have a dif-ferential impact on inflation uncertainty acrosscountries – it is mainly Latin American coun-tries, with their history of many episodes ofhigh inflation, which have the strongest impact.

In summary, the majority of the available empir-ical evidence appears to indicate that it is infla-tion that causes inflation uncertainty for mostcountries and that the effect is positive. Thisprovides support for the Friedman-Ball hypoth-esis. In contrast, the evidence for an effect run-ning from inflation uncertainty to inflationappears to be mixed, if not inconclusive.

2 DATA

Our data is the monthly consumer price index(CPI) for Greece, spanning the period 1981-

2008, as provided by the National StatisticalService of Greece. This series, which is not sea-sonally adjusted, is used to calculate CPI infla-tion in month-on-month annualised percentages.

As an alternative measure, we also calculateand use a seasonally adjusted time series forCPI inflation, by applying the Census X12technique provided by the US Bureau of theCensus (http://www.bls.gov/cpi/cpisapage.htm)and used by the US Bureau of Labor Statisticsfor the estimation of seasonally adjusted USCPI-U. The resulting constructed data seriesallows us to explore causality once seasonalvariation has been accounted for. Both timeseries are plotted in Chart 1.

3 EMPIRICAL METHODOLOGY

In order to test the theories outlined in Sec-tion 2 one requires an adequate measure ofinflation uncertainty; therein lies the firstempirical issue to be addressed in this litera-ture. Several alternative empirical approachesto proxying uncertainty can be found in the lit-erature. Early cross-country work in this area

33Economic Bulletin

May 2010 67

used simple differences in the sample standarddeviation of inflation across countries as ameasure of differences in inflation uncer-tainty.4 More recent work relies either onrolling statistical measures of variability or onthe dispersion of survey-based individual fore-casts as proxies for inflation uncertainty.5

However, neither can be said to adequatelycapture uncertainty as understood in the pres-ent context. Statistical measures of variability,such as the moving average of a variable’sstandard deviation, incorporate both pre-dictable and unpredictable variability; how-ever, only the variance of the stochastic, orunpredictable, component of a variablereflects true economic uncertainty. In the caseof inflation in particular, predictable inflationvariability can be addressed by economicagents, e.g. through indexation; unpre-dictable inflation variability, which cannot, ismore in line with the notion of inflation uncer-tainty. Survey-based measures suffer from thesame shortcoming. Moreover, while they dosummarise the range of disagreement amongindividual forecasters at a given point in time,they hold no information on the individualforecasters’ level of uncertainty regardingtheir point respective forecasts, and hence maypotentially be misleading.6

Current research largely relies on model-derived measures of inflation uncertainty, pri-marily on GARCH-type estimations.7 In con-trast to the above measures, GARCH specifi-cations provide a tool to formally model andestimate the time-varying variance of thedependent variable’s unpredictable innova-tions, i.e. a measure that corresponds well tothe notion of uncertainty used in the theoret-ical literature underlying our work and reflectsits time variation. Hence, this is our method-ology of choice.

A second empirical issue to be addressedrelates to deciding upon a way of testing forcausality. Virtually all of the aforementionedacademic literature uses Granger causalitytests to detect the presence of a causal rela-tionship between inflation and inflation uncer-

tainty. These tests, just as any other statisticalapproach, actually explore temporal prece-dence rather than causation, a point to beborne in mind when results are interpreted.An exception to this norm are Berument andDincer (2005) who, in a multi-country setup,test for causality between inflation and infla-tion uncertainty by using the Full InformationMaximum Likelihood Method with extendedlags, in addition to Granger causality tests. Inprinciple, in multi-country VAR-type estima-tions, impulse responses could also be used todraw relevant conclusions. However, neitherapproach is applicable in our single-countrysetting, nor indeed do they indicate anythingmore than temporal precedence themselves.Hence we perform Granger causality tests onour variables and try to reinforce our resultsby using many different horizons and alterna-tive proxies for uncertainty.

Three sequentially nested GARCH models ofinflation are used to obtain the correspondingestimated time-varying conditional varianceswhich are subsequently used as alternativemeasures of inflation uncertainty. These arepresented in equations (1) to (5).

A standard general GARCH (1,1) model ofinflation is presented in equations (1) and (2).


44 See, for example, Okun (1971) and Logue and Willett (1976).55 For example, Cukierman and Wachtel (1979) analyse this rela-

tionship using the moving variances of the observed rates of change,as well as the variance of inflation expectations as published in theLivingston and Michigan Survey Research Center surveys. Otherresearch using statistical measures of inflation variability includethat by Pagan, Hall and Trivedi (1983) and Silver (1988). Davis andKanago (1998) use ex-ante inflation forecasts produced and pub-lished by Business International while Davis and Kanago (2000) useex-ante forecasts made by the OECD and published in EconomicOutlook. It is noted that the findings reported by research employ-ing survey data to capture inflation uncertainty are less unanimous.

66 We note that, in any case, such measures are not available forGreece over a sufficiently long sample period.

77 There have been a number of single country studies utilising theARCH class of models to proxy inflation uncertainty, seminalamong which are those by Engle (1983), Bollerslev (1986), Evans(1991), and Brunner and Hess (1993). Others include Caporaleand McKiernan (1997), Balcombe (1999), Caporale and Capo-rale (2002) and Hwang (2001) for the USA, Fountas (2001) andKontonikas (2004) for the UK, and Bohara and Sauer (1994) forGermany. Multi-country studies include Grier and Perry (1998),Apergis (2004), Berument and Dincer (2005) and Henry,Olekalns and Suardi (2007) for the G7 countries, Fountas, Ioan-nidis and Karanasos (2004) for six European Union countries,Daal, Naka and Sanchez (2005) for twenty-two developed andless developed economies, and Thornton (2007) for twelveemerging markets.

it = α + ∑12

κ=1bkit-k + et (1)

σ2t = α + βe2

t-1 + γσ 2t-1 (2)

The conditional mean of inflation is specifiedin equation (1) as an autoregressive process oforder 12, so as to reflect the monthly nature ofour data. it denotes inflation at time t and et isthe error term which is assumed to follow aconditional normal distribution. The GARCH(1,1) representation of the conditional varianceis presented in equation (2) where σ2

t denotesthe conditional variance of inflation at time t.The model assumes the conditional variance ofinflation follows an ARMA (1,1) process, i.e.that the conditional variance at time t dependson past innovations and variances. The impli-cation is that, in predicting the variance of thisperiod’s inflation, economic agents form aweighted average of the long-term averagevariance α, the variance observed in the pre-vious period σ 2

t-1, and the size of last period’sforecast error as captured by e2

t-1. If lastperiod’s forecast error and/or variance wereunexpectedly large, economic agents willincrease their estimates of the variance for thenext period.

Equation (3) presents the alternative specifi-cation of the conditional variance implied bya Threshold-GARCH model (see Zakoian,1994, and Glosten, Jagannathan and Runkle,1993).

σ2t = α + βe2

t-1 + βnrt-1e2t-1 + γσ 2

t-1 (3)

where rt = 1 if et < 0, and rt = 0 otherwise. Inthis specification, past innovations are allowedto have a different impact on the conditionalvariance depending on their sign, i.e. positivenews may have a smaller impact on the vari-ance of inflation than negative news. In equa-tion (3) positive news has an impact of β, whilenegative news has an impact of β + βn.

Equations (4) and (5) present the conditionalvariance specification of a ComponentThreshold-GARCH (1,1) model. This formu-lation extends the previous ones by addition-

ally allowing the conditional variance to meanrevert to a time-varying level.

σ2t – mt = β(e2

t-1 – mt-1) + βnrt-1(e2t-1 – mt-1)

+ γ(σ2t-1 – mt-1) (4)

mt = α + ρ(mt-1 – ω) φ(e2t-1 – σ 2

t-1) (5)

where mt is the now time-varying long-runvolatility. The first equation describes the tran-sitory component σ2

t – mt which converges tozero, while the second equation describes thelong-run component mt which converges to ∞,at rates defined by the coefficients of theirrespective equations. The variables in the tran-sitory equation will have an impact on theshort-run movements in volatility, while thevariables in the permanent equation will affectthe long-run levels of volatility.

Having estimated these three alternative uni-variate GARCH specifications for the errorterm, we examine their relative goodness of fitto each data series and proceed to use the esti-mated conditional volatilities as proxies forinflation uncertainty. We perform pairwise bi-directional Granger causality tests betweeneach of the three estimated conditional vari-ance series and CPI inflation, seasonally-adjusted and unadjusted for different laglengths (i.e. different causality horizons) up to12 months and examine the sign of the sum ofthe auxiliary (Granger causality) equations’relevant coefficients, in order to establish thedirection of the effect where Granger causal-ity is found. Subsequently, we separate oursample at two different dates of possible struc-tural breaks and repeat the process in order toconsider whether there has been a shift in therelationships detected using the full sample.

4 EMPIRICAL RESULTS

4.1 ANALYSIS OF THE FULL SAMPLE PERIOD

A diagnostic Lagrange Multiplier (LM) test onour data indicates the presence of strongARCH effects, that is, that volatility is time-

33Economic Bulletin

May 2010 69

varying and periods of high or low volatilitytend to cluster together. Indeed, all threeGARCH specifications appear to provide agood fit to the data, as indicated by theadjusted R-squared measures reported inTable 1. The T-GARCH model appears to pro-vide a superior fit to the unadjusted data on thebasis of the Schwartz and Akaike informationcriteria; i.e. threshold effects appear to be sig-nificant in the variance of the inflation process,

while allowing for a time-varying mean vari-ance by using the C-GARCH specification notso much so. Conversely, using seasonallyadjusted data, it is the C-GARCH whichappears to provide a marginally superior fit –i.e. threshold effects in combination with atime-varying mean variance appear to providethe closest fit to the data. One may speculatethat the seasonal component of inflation ren-ders the transitory and permanent components


SSeeaassoonnaallllyy uunnaaddjjuusstteedd CCPPII iinnffllaattiioonn

GARCH 6.64 6.83 0.78

T-GARCH 6.58 6.78 0.78

C-GARCH 6.60 6.82 0.78

SSeeaassoonnaallllyy aaddjjuusstteedd CCPPII iinnffllaattiioonn

GARCH 5.87 6.06 0.55

T-GARCH 5.87 6.07 0.55

C-GARCH 5.80 6.02 0.56

Akaike Schwarz Adj. R2

Table 1 Information criteria

Note: Smaller values of the Akaike and Schwarz criteria are preferred.

of conditional inflation variance harder to dis-entangle.

The conditional variances derived from thevarious GARCH specifications are plottedagainst time in Chart 2. The volatility indicatedby the conditional variances is in line withdevelopments in the Greek economy over thesample period. In the early part of the sample,volatility was high, with peaks after the 1983and 1985 drachma devaluations, as would beexpected. Volatility also rose in the early 1990s,linked to oil price rises and large fiscal deficitswhich caused inflation to rise. Inflation wasthen propagated by the wage indexation mech-anism. From around 1994 onwards, volatilityfell gradually and has, largely, remained lowthereafter. This coincides with important pol-icy choices made by Greek governments. TheMaastricht treaty entered into force on 1November 1993. It set out the path to mone-

tary union in Europe and the Greek govern-ment publicly stated its aim to become a mem-ber. 1994 was to be a crucial year towards thatgoal. Controls on capital movements were fullyliberalised and monetary financing of the gov-ernment’s budget deficit was prohibited. Thefirst convergence programme of the Greekgovernment was adopted by ECOFIN andfocused on fiscal consolidation to meet theMaastricht Treaty criteria. From 1995, thehard drachma policy was adopted to fosternominal convergence.8 Thus, it is not surpris-ing that the behaviour of inflation volatilitychanges rather strikingly from the mid-1990sonwards. We return to this below.

Table 2 presents all Granger causality testsperformed on the full sample data. Testresults for seasonally unadjusted data are pre-

33Economic Bulletin

May 2010 71

88 See Garganas and Tavlas (2001) for a discussion of this period.

SSeeaassoonnaallllyy uunnaaddjjuusstteedd CCPPII iinnffllaattiioonn -- ffuullll ssaammppllee

Null Hypothesis: Probability (p-values)

Inflation does not Granger cause GARCH cond. variance 0.003 + 0.003 + 0.002 + 0.000 + 0.002 + 0.000 +

GARCH cond. variance does not Granger cause inflation 0.000 + 0.000 + 0.000 + 00..008877 ++ 00..008888 ++ 0.000 −Inflation does not Granger cause T-GARCH cond. variance 0.000 + 0.000 + 0.000 + 0.000 + 0.000 + 0.000 +

T-GARCH cond. variance does not Granger cause inflation 0.000 + 0.000 + 0.000 + 0.000 + 0.000 + 0.000 −Inflation does not Granger cause C-GARCH cond. variance 0.003 + 0.004 + 0.001 + 0.000 + 0.001 + 0.000 +

C-GARCH cond. variance does not Granger cause inflation 0.000 + 0.000 + 0.000 + 0.039 + 00..005500 ++ 00..005511 −

SSeeaassoonnaallllyy aaddjjuusstteedd CCPPII iinnffllaattiioonn -- ffuullll ssaammppllee



GARCH cond. variance does not Granger cause inflation 0.155 − 0.175 − 0.379 − 0.705 − 0.626 − 00..006600 −Inflation does not Granger cause T-GARCH cond. variance 0.000 + 0.000 + 0.000 + 0.000 + 0.000 + 0.000 +

T-GARCH cond. variance does not Granger cause inflation 0.129 − 0.162 − 0.365 − 0.704 − 0.615 − 00..006622 −Inflation does not Granger cause C-GARCH cond. variance 0.000 + 0.000 + 0.000 + 0.000 + 0.000 + 0.000 +

C-GARCH cond. variance does not Granger cause inflation 00..331122 + 00..331144 + 00..550099 + 00..880033 − 00..884455 − 0.035 −

Causality horizon (in months): 2 3 4 6 8 12

Table 2 Granger causality tests - full sample

Note: Figures in normal fonts denote that the null can be rejected at the 1% significance level.Figures in italics denote that the null can be rejected at the 5% significance level.Figures in bold italics denote that the null can be rejected at the 10% significance level.Figures in bold underlined italics denote that the null cannot be rejected at the 10% significance level.Plus and minus symbols indicate the sign of the sum of the relevant estimated coefficients in the corresponding auxiliary equation.

sented in the top panel while those for theseasonally adjusted series are in the bottomone. For each set of data (seasonally unad-justed and seasonally adjusted), there are 3sets of results depending on the particularGARCH specification used. Finally, for eachspecification two hypotheses are tested. First,we test the hypothesis that inflation does notGranger-cause inflation uncertainty. Rejec-tion of this hypothesis at conventional levelsof significance provides support for the viewthat causality runs from inflation to inflationuncertainty. Second, we test the hypothesisthat inflation uncertainty does not cause infla-tion and rejection in this case provides sup-port for the view that causality runs frominflation uncertainty to inflation. Using allthree conditional variance estimates on sea-sonally unadjusted data, one can reject the nullhypothesis of no Granger causality for almostall causality horizons up to 1 year, in bothdirections. Thus, inflation Granger-causesinflation uncertainty and vice versa, at con-ventional significance levels. Since the sign ofthe causal relationship is positive in both cases,these full-sample results on the seasonallyunadjusted data provide support for both theFriedman-Ball hypothesis and for the presenceof a secondary feedback effect resulting fromHolland’s (1995) “stabilisation motive”.

In the context of this application, however,using non-seasonally adjusted data may have amarked impact on our results. Indeed, inflationis known to exhibit strong seasonality. Thecomponent of inflation variability whichreflects seasonal patterns in consumptionbehaviour is of no interest to us here, but maynonetheless strongly affect our measure ofinflation uncertainty, as it is likely to filterthrough to our estimates of the conditionalvariance of inflation. This concern is the the-oretical motivation for repeating this exerciseon seasonally adjusted data of our own con-struction.

The findings presented in the second panel ofTable 2 are strikingly different to those above.While, using any of the three alternative

GARCH-specifications, one can yet againreject the null hypothesis that inflation doesnot Granger-cause its own estimated condi-tional variance for all causality horizons up to1 year, the reverse is no longer true, i.e. we can-not reject the null hypothesis that the esti-mated conditional variance of inflation doesnot Granger-cause inflation for almost allcausality horizons. In other words, based onthe seasonally adjusted data Granger causalityseems to run from inflation to inflation uncer-tainty and not vice versa. Correcting for sea-sonality appears to have completely eliminatedthe causal effect from inflation uncertainty toinflation which had previously been detected.The sign of the cumulative effect in the causalrelationship remains positive and the Fried-man-Ball effect appears even stronger.

4.2 ANALYSIS OF SAMPLE SUB-PERIODS DEFINEDON THE BASIS OF POSSIBLE STRUCTURALBREAKS

Given the substantial changes the Greek econ-omy has undergone in the nearly three decadescovered by our sample, it seems reasonable toconsider the possibility of structural breaks inour data. We consider two different structuralbreaks: the end of 1993, when the MaastrichtTreaty entered into force, and January 2001,when Greece joined the euro area. As we notedabove when describing Chart 2, a structuralbreak around 1994 is not surprising. Similarly,the joining of the euro area in 2001 radicallyaltered the framework of monetary policy for-mulation and, more generally, economic policymaking. Chow tests indeed support the exis-tence of structural breaks at both these times.

The Granger causality tests on seasonallyadjusted data for the sub-periods suggested bythese structural breaks are presented in Tables3 and 4, respectively.9 The top panels now pres-ent Granger causality tests for the first sub-sample, while those for the second subsampleare presented in the bottom panel.


99 The corresponding tests on seasonally unadjusted data are avail-able upon request.

Both sample splits appear to significantly affectour results. In both cases our earlier findingsfor seasonally adjusted data are confirmed forthe first subsample, i.e. inflation appears toGranger-cause inflation uncertainty but notvice versa. Moreover, the sign of the cumula-tive effect in the detected causal relationshipsis almost exclusively positive, in line with theFriedman-Ball hypothesis. However, in thebottom panels representing either the period1994-2008 (Table 3) or 2001-2008 (Table 4),the presence of Granger causality now appearsto be rejected in both directions and at mosthorizons.

When cast against developments in the Greekeconomy over the sample period, these find-ings appear to be intuitive. We seem to findsupport for the Friedman-Ball hypothesis dur-ing the first sample sub-period (whether that

be 1981-1993 or 1981-2000) which correspondsto a period of high and volatile inflation ratescoupled with substantial real and nominaluncertainty. This appears to be evidence of theCukierman-Meltzer hypothesis regarding cen-tral bank behaviour, that is, of a monetaryauthority which is tempted to use inflation inorder to stimulate growth.

The second subsamples correspond toGreece’s efforts to curb inflation, reduceinterest rates and meet the Maastricht crite-ria with a view to becoming accepted as amember of the EMU (1994-2008) and/or tothe EMU era itself (2001-2008). Our findings,which appear to be particularly clear for esti-mates on the periods before and after theend-1993 break, imply, in line with economicintuition, that the preparation for and cre-ation of the EMU and, later the presence of

33Economic Bulletin

May 2010 73

SSeeaassoonnaallllyy aaddjjuusstteedd CCPPII iinnffllaattiioonn -- SSaammppllee:: 11998811MM0011 22000000MM1122



GARCH cond. variance does not Granger cause inflation 0.262 − 0.326 − 0.529 + 0.791 − 0.689 − 0.129 −Inflation does not Granger cause T-GARCH cond. variance 0.000 + 0.000 + 0.000 + 0.000 + 0.000 + 0.000 +

T-GARCH cond. variance does not Granger cause inflation 0.234 − 0.313 − 0.514 + 0.789 − 0.674 − 0.130 −Inflation does not Granger cause C-GARCH cond. variance 0.000 + 0.000 + 0.000 + 0.000 + 0.000 + 0.000 +

C-GARCH cond. variance does not Granger cause inflation 0.452 + 0.495 + 0.689 + 0.891 − 0.901 − 0.114 −



Inflation does not Granger cause GARCH cond. variance 0.100 + 0.010 + 0.011 + 0.046 + 00..005522 + 0.131 +

GARCH cond. variance does not Granger cause inflation 0.258 − 0.323 − 0.496 − 0.683 − 0.828 − 0.519 +

Inflation does not Granger cause T-GARCH cond. variance 0.157 + 0.016 + 0.020 + 00..007755 + 00..008833 + 0.182 +

T-GARCH cond. variance does not Granger cause inflation 0.265 − 0.324 − 0.498 − 0.683 − 0.828 − 0.507 +

Inflation does not Granger cause C-GARCH cond. variance 0.001 − 0.000 + 0.000 + 0.002 + 0.003 + 0.005 +

C-GARCH cond. variance does not Granger cause inflation 00..119966 − 00..224455 − 00..339944 − 00..554488 − 00..668844 − 00..335544 +


Table 3 Granger causality tests - structural break in 2001


the ECB as a monetary policy authority, havecreated a strong anchor for inflation expec-tations which appears to have diluted if notdisrupted the causal relationship betweeninflation and inflation uncertainty that hadexisted in the case of Greece. It must be notedthat there is no evidence of a “stabilisationmotive” as such on behalf of the monetaryauthorities, i.e. we detect no Grangercausality from inflation uncertainty to infla-tion. This of course is not surprising, sincemonetary policy is set not on the basis ofGreek inflation but inflation in the euro areaas a whole. Neither is there evidence of infla-tion Granger-causing inflation uncertainty,something which might result from the rela-tively low level of inflation during this period.

These findings are in line with those of Grierand Perry (1998) and Fountas, Ioannidis and

Karanasos (2004), who report a relationbetween the presence and direction ofGranger causality with the degree of centralbank independence, as well as with those ofKontonikas (2004), who finds that, when theeffects of lower average inflation are con-trolled for, the adoption of an explicit inflationtarget eliminates inflation persistence andreduces long-run inflation uncertainty.

5 CONCLUSION

This paper has examined the relationshipbetween inflation and inflation uncertainty forGreece for the period 1981-2008. The ration-ale for conducting such an exercise stems fromthe theoretical observation that a significantcost of inflation arises because of its impact oninflation uncertainty.





GARCH cond. variance does not Granger cause inflation 0.765 − 0.760 + 0.785 + 0.958 + 0.952 − 0.615 −Inflation does not Granger cause T-GARCH cond. variance 0.000 + 0.000 + 0.000 + 0.000 + 0.000 + 0.001 +

T-GARCH cond. variance does not Granger cause inflation 0.756 − 0.759 + 0.775 + 0.958 + 0.952 − 0.619 −Inflation does not Granger cause C-GARCH cond. variance 0.012 + 0.024 + 0.017 + 0.031 + 0.047 + 0.175 +

C-GARCH cond. variance does not Granger cause inflation 0.879 − 0.893 + 0.952 + 0.992 − 0.980 − 0.514 −




GARCH cond. variance does not Granger cause inflation 0.350 − 0.131 − 0.125 − 0.236 − 0.482 − 0.124 −Inflation does not Granger cause T-GARCH cond. variance 0.289 + 0.262 + 0.392 + 0.538 + 00..009911 + 0.354 +

T-GARCH cond. variance does not Granger cause inflation 0.299 − 0.109 − 0.112 − 0.232 − 0.463 − 0.127 −Inflation does not Granger cause C-GARCH cond. variance 00..009999 + 0.028 + 0.037 + 00..008844 + 0.047 + 00..224422 +

C-GARCH cond. variance does not Granger cause inflation 00..887766 − 00..110088 − 00..009900 + 00..119999 − 00..449900 − 0.044 −


Table 4 Granger causality tests - structural break in 1994


The results suggest that the direction of causal-ity in the Greek case does indeed run from infla-tion to inflation uncertainty, and not vice versa.Moreover, the sign of this causal effect is pos-itive, i.e. higher inflation causes higher uncer-tainty. This is supportive of the Friedman-Ballhypothesis. Interestingly, these results breakdown in the post-1993 and post-2000 periods. In

neither period (either 1994-2008 or 2001-2008)is there much evidence of causality in any direc-tion. This is arguably a consequence of the infla-tion-targeting monetary policy adopted duringthese sub-periods, which resulted in far lowerlevels of inflation as compared to the earlierperiod and contributed to reducing uncertaintyand anchoring expectations.

33Economic Bulletin

May 2010 75

Apergis, N. (2004), “Inflation, output growth, volatility and causality: evidence from panel dataand the G7 countries”, Economics Letters, 83, 185-91.

Balcombe, K. (1999), “More evidence concerning the relationship between high and variable infla-tion: a more general approach”, Applied Economics Letters, 6, 59-61.

Ball, L. (1992), “Why does high inflation raise inflation uncertainty?”, Journal of Monetary Eco-nomics, 29, 371-88.

Ball, L. and S.G. Cecchetti (1990), “Inflation and uncertainty at short and long horizons”,Brookings Papers on Economic Activity, 1, 215-54.

Barro, R. and D. Gordon (1983), “A positive theory of monetary policy in a natural rate model”,Journal of Political Economy, 91, 589-610.

Batchelor, R. and P. Dua (1996), “Empirical measures of inflation uncertainty: a cautionary note”,Applied Economics, 28, 333-41.

Berument, H. and N.N. Dincer (2005), “Inflation and inflation uncertainty in the G7 countries”,Physica A, 348, 371-9.

Bhar, R. and S. Hamori (2004), “The link between inflation and inflation uncertainty: evidencefrom G7 countries”, Empirical Economics, 29, 825-53.

Bohara, A.K. and C. Sauer (1994), “The role of inflation uncertainty in Germany: Friedman’shypothesis revisited”, Empirical Economics, 19, 611-27.

Bollerslev, T. (1986), “Generalized autoregressive conditional heteroskedasticity”, Journal ofEconometrics, 31, 307-27.

Bomberger, W.A. (1996), “Disagreement as a measure of uncertainty”, Journal of Money, Creditand Banking, 28, 381-92.

Boyd, D. and R. Smith (2006), “Monetary regimes and inflation in 12 Caribbean economies”,Journal of Economic Studies, 33, 96-107.

Boyd, D. and R. Smith (2007), “Institutions and inflation persistence in the Caribbean”, AppliedEconomics Letters, 14, 283-6.

Brunner, A.D. and G.D. Hess (1993), “Are higher levels of inflation less predictable? A state-dependent conditional heteroskedasticity approach”, Journal of Business and Economic Sta-tistics, 11, 187-97.

Caporale, B. and T. Caporale (2002), “Asymmetric effects of inflation shocks on inflation uncer-tainty”, Atlantic Economic Journal, 30, 385-8.

Caporale, T. and B. McKiernan (1997), “High and variable inflation: further evidence on theFriedman hypothesis”, Economics Letters, 54, 65-8.

Caporale, G.M. and A. Kontonikas (2006), “The Euro and Inflation Uncertainty in the Euro-pean Monetary Union”, CESifo Working Paper No. 1842.

Cosimano, T.F. and D.W. Jensen (1988), “Estimates of the variance of US inflation based uponthe ARCH model: comment”, Journal of Money, Credit and Banking, 20, 409-21.

Cukierman A. (1992), Central Bank Strategy, Credibility and Independence, MIT Press, Cam-bridge.

Cukierman, A. and A. Meltzer (1986), “A theory of ambiguity, credibility, and inflation underdiscretion and asymmetric information”, Econometrica, 54, 1099-128.

Cukierman, A. and P. Wachtel (1979), “Differential inflationary expectations and the variabil-ity of the rate of inflation: Theory and evidence”, American Economic Review, 72, 508-12.

Daal, E., A. Naka and B. Sanchez (2005), “Re-examining inflation and inflation uncertainty indeveloped and emerging countries”, Economics Letters, 89, 180-6.

Davis, G. and Β. Kanago (1998), “High and Uncertain Inflation: Results from a New Data Set”,Journal of Money, Credit and Banking, 30(2), 218-30.

Davis, G. and B. Kanago (2000), “The level and uncertainty of inflation: results from OECD fore-casts”, Economic Inquiry, 38, 58-72.


B I B L I OGRAPHY

Devereux, M. (1989), “A positive theory of inflation and inflation variance”, Economic Inquiry,27, 105-16.

Driffill, J., G.E. Mizon and A. Ulph (1990), “Costs of inflation”, in Friedman, B.M. and F.H.Hahn (eds), Handbook of Monetary Economics, Volume II, Elsevier Science Publishers B.V.

Engle, R.F. (1982), “Autoregressive Conditional Heteroskedasticity with Estimates of the Vari-ance of United Kingdom Inflation”, Econometrica, July, 987-1007.

Engle, R.F. (1983), “Estimates of the variance of US inflation based on the ARCH model”, Jour-nal of Money, Credit and Banking, 15, 286-301.

Evans, M. (1991), “Discovering the link between inflation rates and inflation uncertainty”, Jour-nal of Money, Credit and Banking, 23, 169-84.

Evans, M. and P. Wachtel (1993), “Inflation regimes and the sources of inflation uncertainty”,Journal of Money, Credit and Banking, 25, 475-511.

Fountas, S. (2001), “The relationship between inflation and inflation uncertainty in the U.K.:1885-1998”, Economics Letters, 74, 77-83.

Fountas, S., A. Ioannidis and M. Karanasos (2004), “Inflation, inflation uncertainty and a com-mon European monetary policy”, The Manchester School, 72, 221-42.

Friedman, M. (1977), “Nobel lecture: inflation and unemployment”, Journal of Political Econ-omy, 85, 451-72.

Garganas, N.C. and G.S. Tavlas (2001), “Monetary Regimes and Inflation Performance” inBryant, R.C., N.C. Garganas and G.S. Tavlas (eds), Greece’s Economic Performance andProspects, Bank of Greece and the Brookings Institution.

Glosten, L., R. Jagannathan and D. Runkle (1993), “On the relation between the expectedvalue and the volatility of the nominal excess return on stocks”, Journal of Finance, 48, 1779-801.

Grier, K.B. and M.J. Perry (1990), “Inflation, inflation uncertainty and relative price dispersion:evidence from bivariate GARCH-M models”, Journal of Monetary Economics, 38, 391-405.

Grier, K.B. and M.J. Perry (1998), “On inflation and inflation uncertainty in the G7 countries”,Journal of International Money and Finance, 17, 671-89.

Grier, K.B. and M.J. Perry (2000), “The effects of real and nominal uncertainty on inflation andoutput growth: some garch-m evidence”, Journal of Applied Economics, 15, 45-58.

Henry, O.T., N. Olekalns and S. Suardi (2007), “Testing for rate dependence and asymmetry ininflation uncertainty: evidence from the G7 countries”, Economics Letters, 94, 383-8.

Holland, S. (1993a), “Uncertain effects of money and the link between the inflation rate and infla-tion uncertainty”, Economic Inquiry, 31, 39-51.

Holland, S. (1993b), “Comment on `inflation regimes and the sources of inflation uncertainty’”,Journal of Money, Credit and Banking, 25, 514-20.

Holland, S. (1995), “Inflation and uncertainty: tests of temporal ordering”, Journal of Money,Credit and Banking, 27, 827-37.

Hwang, Y. (2001), “Relationship between inflation and inflation uncertainty”, Economics Let-ters, 73, 179-86.

Kim, C. (1993), “Unobserved-component time series models with Markov-switching het-eroskedasticity: changes in regime and the link between inflation rates and inflation uncertainty”,Journal of Business and Economic Statistics, 11, 341-9.

Kontonikas, A. (2004), “Inflation and inflation uncertainty in the UK evidence from GARCHmodelling”, Economic Modelling, 21, 524-43.

Logue, D. and T. Willett (1976), “A note on the relationship between the rate and variability ofinflation”, Economica, 43, 151-8.

Lucas, R. (1973), “Some International Evidence on Output-Inflation Tradeoffs”, American Eco-nomic Review, 63, 326-34.

33Economic Bulletin

May 2010 77

Okun, A. (1971), “The Mirage of Steady Inflation”, Brookings Papers on Economic Activity, 0(2),485-98.

Pagan, A.R. (1984), “Econometric issues in the analysis of regressions with generated regres-sors”, International Economic Review, 25, 221-47.

Pagan, A.R., A.D. Hall and P.K. Trivedi (1983), “Assessing the variability of inflation”, Reviewof Economic Studies, 50, 585-96.

Pourgerami, A. and K. Maskus (1987), “The effects of inflation on the predictability of pricechanges in Latin America: some estimates and policy implications”, World Development, 15,287-90.

Silver, M. (1988), “Average Prices, Unanticipated Prices and Price Variability in the UK for Indi-vidual Products,” Applied Economics, 20, 569-94.

Thornton, J. (2007), “The relationship between inflation and inflation uncertainty in emergingmarket economies”, Southern Economic Journal, 73, 858-70.

Ungar, M. and B. Zilberfarb (1993), “Inflation and its unpredictability: theory and empirical evi-dence”, Journal of Money, Credit and Banking, 25, 709-20.

Zakoian, J.Μ. (1994), “Threshold heteroscedastic models”, Journal of Economic Dynamics andControl, 18, 931-55.


33Economic Bulletin

May 2010 79

1 INTRODUCTION

A central query in financial economics is theextraction of information on investors’ per-ception of the uncertainty of future marketconditions; given its direct implications forcapital allocation and risk aversion, this gen-eral topic has become critical recently. In thiscontext, the empirical finance literature hasfocused on measures stemming from financialmarkets, in order to exploit potential infor-mation contained in them in order to be usedas ‘forward-looking’ variables for market con-ditions.

A growing strand of the empirical finance lit-erature focuses on the correlation betweenbond and stock returns (hereafter ‘stock-bondcorrelation’, for brevity)1 as a tool for detect-ing asset price imbalances, thus motivating itsuse for financial stability purposes. The ration-ale behind the use of the correlations amongstock and bond returns as a proxy for marketconditions and the underlying risk can befound in the perception of the existence ofcommon pricing factors and different risk char-acteristics.

Against this backdrop, if the notion of factorsaffecting the pricing of both stocks and bonds,establishes, theoretically, the relation betweenthe information contained in the stock-bondcorrelations, their time-varying characteristicsallow for empirical examination. In thisrespect, the stock-bond correlation has beenfound to take negative values, under stressedmarket conditions. Among the reasons for thisphenomenon,‘flight-to-quality’ effects are themost prominent answer and are related to themarket’s perception of bonds as safer thanequities (see d’Addona and Kind, 2006, andEhrmann et al., 2005, among others). Flight-to-quality effects refer to increases in holdings ofsafer assets and sales of assets which are per-ceived as more uncertain; within this context,

negative correlations between returns in stockand bond markets are explained in the empir-ical literature as reflecting interrelated oppo-site transactions in the two markets (see e.g.Hakkio and Keeton, 2009, and Baur andLucey, 2009).

Furthermore, the stock-bond correlation hasbeen found to be negatively related to highvolatility regimes in equity markets (Connollyet al., 2007). Thus, the ‘decoupling’ phenom-enon between stocks and bonds appears to berelated to conditions of high uncertainty. Areasonable explanation is given by Baur andLucey (2009); investors’ tendency to “chase theyield(s)” or find shelters in turbulent periodsis revealed in the time variation and reversalsin the stock-bond correlation.

On the other hand, numerous papers havedealt with interactions among markets andfinancial segments (e.g. Baele et al., 2004,Davies, 2006, and Yang et al., 2004). Further-more, in the event that the examination focuseson the transmission of the negative effectsfrom one market to another, during turbulentperiods, this phenomenon is referred to as‘contagion’. From this perspective, there isbroad consensus in the empirical finance lit-erature that contagion refers to unforeseeablenegative returns in a market, which are causedby shocks initiated in another market andtransmitted through portfolio rebalancing andrisk aversion effects (see Beirne et al., 2009).2

As a result, previous studies relate stock-bondcorrelation to market uncertainty, while con-tagion effects arise in turbulent periods as well.

However, little emphasis has been given toexplaining the dynamics of the stock-bond cor-relation for Greece. Although investigations of

DE T ERM INANT S O F TH E GR E EK S TOCK -BONDCORRE L A T I ON

Petros M. MigiakisEconomic Research Department

11 For definitions and terminology we refer to Connolly et al. (2005),Yang et al. (2009) and Baur and Lucey (2009), among others.

22 Contagion presumes a high degree of integration in order for diver-sification effects to be inactive.

“flight-to-quality” effects have occasionallyincluded Greece in their samples (see Beber etal., 2006), there has not been a specific inves-tigation of the determinants of the Greekstock-bond correlation. In this context, weassess the extent to which the information con-tained in the stock-bond correlation isexplained by several risk proxies related to (i)stock market uncertainty, (ii) credit risk and(iii) liquidity conditions in the banking sector.Additionally, we examine whether the Greekstock-bond correlation is subject to spillovereffects from the dynamics of the relevant cor-relations of the United States, Germany andItaly.3

The present empirical analysis unfolds in thefollowing way. Section 2 continues the discus-sion of the literature, focusing on properties ofthe stock-bond correlation and financial inter-actions. Section 3 describes the empiricalinvestigation framework and Section 4 presentsthe empirical findings and discusses policyimplications. Finally, Section 5 concludes.

2 LITERATURE REVIEW

2.1 PROPERTIES OF THE STOCK-BOND CORRELATION

Early on, several works dealt with the commondeterministic process underlying the pricing ofstocks and bonds. Fama and French (1993)investigated the sources of the common pric-ing patterns between stocks and bonds; theyhave found that several risk factors generatevariations in the returns on both equities andbonds. Their assessment motivates furtherexploration of the relation between risk andthe correlation of stock and bond returns.

Campbell and Ammer (1993) show that severalcommon factors, including expected excessstock returns, the real interest rate and infla-tion, exercise causal effects on both equitiesand bonds. Interestingly, they find that shocksto expected stock returns produce variations instock and bond returns of the same direction,

while increasing expected inflation results inopposite effects on bond and stock returns.Their results highlight the fact that stocks area good inflation hedge, whereas bond yields4

are known to carry the inflation risk premium.As a result, expected inflation stands as apotential source of ‘decoupling’ effects, that is,periods in which stock returns decreases(increases) and bond returns increases(decreases) may be related to higher (lower)expected inflation.

Laloux et al. (2000) investigate the statisticalproperties of correlation coefficients stem-ming from financial time series. The authorsinvestigate the S&P500; their findings indicatethat choosing an asset allocation that min-imises the risk of the portfolio is similar tochoosing a portfolio with large weights on thesmallest eigenvalues of the correlation matrixeigenvector. In plain terms, this finding leadsto a need to examine the extent to whichdeterministic factors in the form of financialvariables can explain the correlations; themore variability of the correlation can beexplained by economically meaningful vari-ables, the less risky is the underlying portfo-lio. This finding motivates the examination ofthe stationarity properties of the stock-bondcorrelation; according to the aforementionedargument, in the event that there exist signif-icant diversification effects between stocks andbonds, as expected, their correlation should bestationary.

Overall, the stock-bond correlation measurehas been found to have small but positive val-ues whereas it is positively related to short-term rates and inflation (Yang et al., 2009).Additionally, Cappiello et al. (2003) arguedfor the time-varying characteristics of thestock-bond correlation measure. This time


33 The United States and Germany were chosen for obvious reasonsrelated to market size and their respective international and Euro-pean benchmark status. Italy is also included for two main reasons;it is the ‘high-yielders’ benchmark, according to Lane (2006), whileit could be conceived as a regional leader, in the globalisation con-cept of Krugman (1991).

44 It is clear that variable yield bonds are excluded from this refer-ence.

variation has been attributed to stock marketuncertainty by Connoly et al. (2005); ‘flight-to-quality’ effects are found to be reflected inthe stock-bond correlation, according to Baurand Lucey (2009). Specifically, it is arguedthat, in periods of turbulence in financialmarkets, investors have strong motives tochange their portfolio distribution towardslowering the weight of risky assets in the port-folio; as a result, bonds’ and equities’ returnswould diverge. This would produce a‘decoupling effect’ that would be reflected innegative stock-bond correlations. Adoptingthis point of view on the information con-tained in stock-bond correlations, Hakkioand Keeton (2009) have included the meas-ure under examination as a component to thefinancial stress index they have compiled forthe Fed.

Connoly et al. (2005) also relate decouplingeffects between stocks and bonds, as reflectedin the stock-bond correlation, with increases inthe volatility of stock markets. Their analysisargues that the probability for a negative stock-bond correlation increases in a high-volatilityregime.

Summarising the aforementioned findingsleads to several useful remarks. First, thestock-bond correlation is found to be relatedto the degree of uncertainty present in finan-cial markets at any time; the usually positivecorrelation declines, or even turns negativeunder financial markets’ turbulence. Theseeffects are in line with the explanation given bythe literature investigating the co-variates ofstock and bond returns; shocks produced byexpected stock returns move these assets in thesame direction, whereas real interest rates,expected inflation and market volatility con-tribute to ‘decoupling’ effects.

2.2 ESTIMATING MARKETS INTERACTIONS

The notion of interactions between two mar-kets refers to whether movements in one mar-ket are, at least partially, caused by movementsin another. Ultimately, under the Law of One

Price, assets with the same characteristicsshould be priced equally even across differentmarkets. This aspect, while proving to be par-ticularly useful for investigating financial inte-gration effects, also gives a proxy for estimat-ing whether the dynamics in the stock-bondcorrelation of one country are influenced bysimilar dynamics in other countries.

The methodological frameworks employedfor the examination of interactions betweenmarkets or segments are set to capture eithercausal patterns or spillover effects. In thefirst case, VARs and VECMs are used forthis purpose and Granger causality tests areapplied in order to examine the significanceand direction of the effects exercised in thepricing processes.5 Second, several GARCH(Generalised Autoregressive ConditionalHeteroskedasticity) models have been usedin order to detect variance spillover effects.In this context, correlations, estimatedthrough DCC-GARCH, (Dynamic Condi-tional Correlation-GARCH) have beenused in order to assess spillover and conta-gion effects in financial markets (e.g. Chianget al., 2007).

3 THE EMPIRICAL FRAMEWORK

3.1 DESCRIPTION OF DATA

Our analysis is based on data from three Euro-pean markets with different characteristics(Germany, Italy and Greece) and the UnitedStates, for the period after EMU, i.e. January1999 to December 2009. Both stock and bondreturns are included at weekly frequency; wedeemed that data of daily frequency wouldcontain noise, valuable for arbitrageurs but notfor our purposes, while data with a frequencylower than weekly could result in the loss ofinformation contained in infrequent events. Alldata were collected from Thomson Financial(Datastream).

33Economic Bulletin

May 2010 81

55 Interested readers can refer to Yang et al. (2004), Sarno andValente (2006), Vo (2009), among others.

Stock market returns are estimated as weeklydifferences of logged, euro-denominatedMSCI indices for the aforementioned marketsand period.6 Similarly, weekly bond returns areestimated by taking differences in yields-to-maturity of the relevant 10-year governmentbonds. Specifically,

for stocks

and

, for bonds.

Variables used for the explanation of risksreflected in the stock-bond correlation are (i)the spread between interbank rates of weeklymaturity, denominated in euro (Euribor) andUS dollars ($Libor), (ii) the spread betweenthe AAA corporate bond yield7 and the 10-yearbenchmark Bunds, and (iii) a volatility proxyfor the Greek stock market, estimated as theintraday range of the general index of theAthens Exchange (Athex).

The difference between the weekly Euriborand $Libor rates ((E-L)t), is intended to cap-ture changes in the relative banking sector liq-uidity positions between Europe and theUnited States.8 As a result, liquidity constraintsin the European banking sector would bereflected in variations in the relative spread. Inthis context, as EMU money markets arereported to be fully integrated (see Centenoand Mello, 1999) we believe that this variableis a reliable proxy for the Greek banking sec-tor as well.

The credit spread (CSt) is intended toapproximate credit risk conditions. However,due to a lack of activity in the Greek sec-ondary market for corporate bonds, relevantreliable quotations could not be found. As aresult, we rely on overall European corporatebond rates. Even though they are not set tocapture idiosyncratic credit risk conditionsfor the Greek market, they can be used asproxies for credit conditions in the Europeanbond market.

Finally, the intraday price range for the Greekstock market is intended to capture Greekstock market volatility. The intraday range isestimated as the difference between the high-est intraday value of the general stock marketindex (SH) and its lowest (intraday) value (SL).After all that has been reported in relevantempirical finance literature, this aggregate isconsidered suitable for capturing market par-ticipants’ uncertainty about the “fair-price”.9

Briefly, it should be noted that the argumentfor using the range as an indicator of investors’uncertainty is that in case the highest and thelowest bids are very divergent, this may reflectlarge differences in market’s participants’ per-spectives about fair prices for stocks. Even inthe event that it is an effect of thin tradingactivity, it still captures a significant compo-nent of market risk, i.e. liquidity risk.

Finally, we use stock-bond correlations fromother markets (namely, the US, Germany andItaly) as proxies for interactions across coun-tries. In the event that these variables exercisesignificant effects on the Greek stock-bondcorrelation, there will exist cross-country con-tagion effects, especially when the significanteffects are evident in periods of ‘decoupling’effects.

3.2 METHODOLOGY

In order to estimate the correlation coeffi-cients between stock and bond returns, we usethe dynamic conditional correlation frameworkestablished by Engle (2002). Consider the cor-relation between stock and bond returns intime t, conditional on information available upto time t-1, illustrated in equation (1) below:


!

!

66 For the estimation of stock returns under this formulation, see Cap-piello et al. (2003).

77 Extracted from the iBoxx AAA index compiled of highly liquid cor-porate bonds, with a term to maturity 10+ years, denominated ineuro.

88 Alexopoulou et al. (2009) provide a more formal discussion on theinformation expected to be reflected by the spread between short-term rates.

99 For empirical applications using the “intraday range” as a meas-ure for stock market volatility, we refer to Christensen and Podol-skij (2006) and Joshy (2008). It should be reminded that so far thereis no implied volatility index for the Greek stock exchange, as theone traded in the Chicago Board of Exchange (ticker: VIX) or theGerman equities’ implied volatility index (VDAX).

(1)

where, i stands for the countries included in thepresent analysis, Et-1(r t

s , r tb) denotes the con-

ventional covariance of stock and bond returns,and Et-1[(r t

s)2], Et-1[(r tb)2], are the respective

variances, conditioned on information avail-able until time t-1.

Afterwards, Bollerslev (1990) proposes theestimation of the conditional constant corre-lation (CCC) model specified in matrix form,as Ht=DtRDt, in which R stands for the sam-ple correlation matrix and D is the kxk diago-nal matrix of time-varying standard deviationsfrom GARCH(1,1) models with on thejth diagonal; that is the square root of the esti-mated variance.

Then, following Engle, one can relax the con-stancy assumption of the correlation matrix,thus permitting its assessment as time-varying.That is,

(2)

where,

and then,

with q–12= E (r ts , r t

b).

Having estimated pti , we then examine its sta-

tistical properties. Emphasis is given to unitroot tests, in order to specify the proper econo-metric treatment of the correlation as adependent variable. We test the null hypoth-

esis of a unit root in the data both against thealternative of stationarity, by using conven-tional Dickey-Fuller tests, and against thenear-unit-root alternative, set to capture trendstationary series, by applying the test of Elliottet al. (1996).

Then we turn to the examination of our mainquestion: what drives the stock-bond correla-tion. Conditional upon the appropriate sta-tionarity properties of the variables that enterin the regression, we estimate the relationbelow:

(3)

Equation 3 is to be estimated with the methodof least squares, with heteroskedasticity con-sistent errors by applying the Newey-Westmodification. We run several regressions ini-tially for the whole sample and then for foursub-samples, chosen exogenously according tospecific market-related conditions. The first isrelated to the EMU accession period forGreece (1999-2001); additionally, this periodis associated with the bursting of the bubble onthe domestic stock market. The second periodis the post-EMU, pre-crisis period (2001 –August 2007).10 The third sub-period is relatedto the global credit crisis (2007-2009), while thefourth sub-period begins with the date of theLehman Brothers bankruptcy (15 September2008).

Lastly, we repeat this analysis by regressingonly negative values of the Greek stock-bondcorrelation on the aforementioned explanatoryvariables. In this way, we intend to take intoaccount switches in the deterministic structureof the system, as the dependent variable movesbetween positive and negative ground. Again,the periods under examination are the samewith those described before.

33Economic Bulletin

May 2010 83

!

!

!

!

!

!

!

!

!

1100 It should be reminded that on 9 August 2007 the Eurosystem’sextraordinary liquidity provision to the banking system amountedto €95 billion after the “complete evaporation of liquidity”, as BNPParibas stated to its investors.

4 EMPIRICAL RESULTS

Table 1 presents descriptive statistics for thevariable under examination and its determi-nants. The table contains descriptive statistics,for the whole sample in the upper panel, andconventional (DF) and efficient (DF-GLS) unitroot tests in the lower panel. First, the statisticsconfirm the overall consensus in relevant lit-erature that the stock-bond correlation is pos-itive and low in value. It should be stressed thatthe figures reported are very similar to thoseobtained by Conolly et al. (2007), as far as theUnited States and Germany are concerned.Additionally, we report figures for Greece andItaly; the mean stock-bond correlations for thetwo countries and the period under examina-tion were 0.105 and 0.19 respectively.

Next, we focus on the unit root test results.Overall, our data are indicated to be station-ary, whereas in the case of the interbank ratesspread the variable is found to contain a unitroot by conventional unit root tests but the

tests of Elliott et al. (1996) confirm trend-sta-tionarity of the series.

In addition, Chart 1 illustrates the dynamicbehaviour of the Greek stock-bond correlation.The illustration of the correlation reveals thatit has ventured into negative territory at sev-eral times in the past; an effect that has beenrelated to flight-to-quality effects in the pre-vious literature. The appearance of negativecorrelations is mainly concentrated in theperiod 1999-2001 and late 2007 onwards. Wedeem that in the first case the reversal capturesthe bursting of the bubble on the Athens stockmarket, while in the second period the nega-tive stock-bond correlation is directly relatedto the credit crisis that erupted in 2007.

Next, we investigate whether the decouplingeffects that are reflected in negative stock-bondcorrelations in Greece are interrelated withsimilar decoupling effects in the United States,Germany and Italy at time t-1. For the wholesample, the results suggest that about 32.5%,


ρGR 0.105 0.137 -0.388 -0.356

ρUS 0.253 0.205 -0.115 -0.415

ρDE 0.289 0.220 0.213 -0.365

ρΙΤ 0.190 0.185 0.232 -0.223

CS 0.627 1.011 12.069 -2.799

E-L -0.304 1.36 -1.485 -0.097

SH –SL 0.007 0.005 9.657 2.597

Unit root tests

DF (lags=1) DF (lags=5) DF-GLS (lags=1) DF-GLS (lags=5)

ρGR -4.167 -3.812 -3.330 -2.894

ρUS -2.613* -2.532* -2.607 -2.526

ρDE -3.481 -3.357 -3.381 -3.273

ρΙΤ -5.295 -4.902 -5.299 -4.907

CS -4.502 -4.982 -3.993 -4.273

E-L -1.342** -1.257** -2.544* -2.654

SH –SL -10.165 -5.443 -2.919 2.340

Critical values 1%= -3.444 5%= -2.867 10%= -2.570 1%=-2.58 5%=-1.95 10%=-1.62

Mean Standard error Kurtosis Skewness

Table 1 Properties of the data

30% and 44.2% of the negative values of theGreek correlation occur when negative corre-lations have also registered for the UnitedStates, Germany and Italy, respectively. How-ever, when we restrict the period to the 2007-2009 crisis, these figures are considerably dif-ferent; in the cases of the US and Germany,negative stock-bond correlations are found withnegative Greek stock-bond correlations only29.4% and to 5.8% of the time, respectively; bycontrast, 70.6% of negative values of the stock-bond correlation in Greece coincide with neg-ative Italian stock-bond correlations. Finally,when the period is further restricted to thepost-Lehman-collapse period, the respectivefigures are 35.7%, 0% and 67.8%, respectively.

From these figures one gains an insight into theinterlinkages between the decoupling effectsfor Greek stock and bond markets with therespective markets from the United States,Germany and Italy. The results suggest that, asfar as the United States and Germany are con-cerned, there exist only weak (if any) evidence

for linkages between these effects. On theother hand, these findings initiate the need tofurther explore the links between the Italianand the Greek stock-bond correlations. Ofcourse, these figures are not adequate in orderto argue for the existence of spillover effects,but they provide strong motivation for the fol-lowing investigation of contagion effects in theGreek stock-bond correlation.

As a result, we turn to the findings of thedeterminants of the stock-bond correlationdynamics for Greece, according to the for-mulation of equation 3. Table 2 presents therelevant results. For the whole sample (1999-2009), the Greek stock-bond correlation isfound to be explained by previous observa-tions of the interbank rates spread (E-L)t-1 andthe previous week’s stock-bond correlationsfor Italy (ρt-1

IT ) and Germany (ρt-1DE ). The first

variable is found to be negatively related tothe correlation, while the coefficients of thelast two variables are positive. These effectsindicate that an increase in the spread betweenthe interbank offer rates of the euro area andthe United States explains, to a great extent,a subsequent decrease in the stock-bond cor-relation in Greece, while there exist spillovereffects from the stock-bond correlations ofItaly and Germany.

However, these findings differ considerablywhen we examine several sub-periods of oursample. Overall, the diagnostics, contained inthe lower panel of the table, indicate that theformulation under examination explains effi-ciently and to a great extent (except for the1999-2001 period) the dynamics of the Greekstock-bond correlation. Nevertheless, a morethorough inspection of the findings reveals sev-eral interesting points. Specifically, it is indi-cated that in the period after the launch of theEMU and before the official accession ofGreece, only the dynamics of German stock-bond correlation are found to significantlyaffect movements in the Greek stock-bond cor-relation. This effect has outlived the accessionof Greece in the euro area, while it is associ-ated with the emergence of significant effects

33Economic Bulletin

May 2010 85

stemming from the spread between interbankrates; a finding that we deem is quite reason-able since the liquidity conditions in the euroarea banking sector, assumed to be reflected inthe respective variable, have a direct impact onthe Greek banking sector. On the other hand,since the onset of the credit crisis in 2007, onlythe Italian stock-bond correlation is found tosignificantly explain movements in the Greekcorrelation. Additionally, the effects stemmingfrom the Italian variable are positive, indicat-ing the existence of significant spillover effects,during the credit crisis period.

This last finding can be further clarified by theexamination of the decoupling effects,reflected by negative values of the stock-bondcorrelation. The respective findings arereported in Table 3.

For the whole sample the intraday stock mar-ket’s range, assumed to capture underlyinguncertainty in the stock market, is found to sig-nificantly explain movements in the stock-bondcorrelation when it is negative. Specifically, theresults suggest that an increase11 in the stockmarket’s intraday range causes the negative

stock-bond correlation to become even morenegative in the subsequent week. Additionally,we find that the spread between the two inter-bank rates leads to positive effects on thestock-bond correlation. However, the signifi-cance of this variable might reflect mostlyeffects from the 2007-2009 period, when theUS banking system, initially, experiencedstrong deterioration in its liquidity conditions.Finally, the significant relation of the Greekand Italian stock-bond correlations is con-firmed in periods of decoupling effects in theGreek stock and bond markets as well.

However, the examination of the relation dur-ing the selected sub-periods reveals that theexplanatory variables under review do notprovide a meaningful explanation concerningthe negative values of the correlation, overthe period 1999-2007. The variables explainthe dynamics of the Greek stock-bond corre-lation only for the period which followed theonset of the credit crisis. In this context,Greek stock market uncertainty (as measured


1111 An increase in the range has been reported to reflect an increasein investors’ uncertainty. Interested readers should refer to empir-ical literature mentioned in Section 3 (footnote 8).

Note: Each cell presents the value (standard error) of the coefficient of the respective explanatory variable, as in equation 3, with the depend-ent variable being the Greek stock-bond correlation estimated by DCC-GARCH, as in equation 2. The estimation has been performedwith consistent heteroskedasticity errors, estimated with the Newey-West formula. Asterisks (**, *) denote (5%, 10%) significance.

(SH-SL)t-1-1.859

(1.194)-2.191

(1.794)-0.717

(1.387)-2.235

(1.551)-0.958

(2.060)

CSt-1-0.022

(0.012)0.087

(0.129)0.007

(0.022)-0.032

(0.021)0.019

(0.037)

(E-L)t-1-0.012**

(0.005)0.017

(0.034)-0.016**

(0.006)0.014

(0.013)0.041*

(0.023)

ρUSt–1

0.076 (0.049)

-0.159(0.138)

0.028(0.071)

-0.027(0.055)

-0.041(0.187)

ρITt–1

0.290** (0.068)

-0.121(0.156)

0.086(0.091)

0.515**(0.057)

0.304**(0.089)

ρDEt–1

0.205** (0.076)

0.321**(0.156)

0.379**(0.109)

0.132(0.099)

-0.189(0.248)

~R2 0.472 0.033 0.405 0.653 0.358

D-W 0.475 0.378 0.506 0.907 0.801

1/1/1999-9/12/2009

1/1/1999-4/12/2000

1/1/2001-9/8/2007

9/8/2007-9/12/2009

15/9/2008-9/12/2009

Table 2 Determinants of the Greek stock-bond correlation

by the intraday range) is found to have sig-nificant explanatory power for the decouplingeffects reflected in the negative values of thestock-bond correlation in the period 2007-2009, while there exist significant contagioneffects stemming from the Italian correlationover the Greek correlation for the whole cri-sis period. Furthermore, the interbankspread and the US stock-bond correlationhave a decisive impact on the dynamics of theGreek correlation, but this does not apply forthe period after the collapse of LehmanBrothers. Finally, an important finding is thatan increase in the US stock-bond correlationleads to a decrease in the Greek stock-bondcorrelation, which is highly likely to reflectdiversification effects.

In sum, it is assumed that monitoring thedynamics of stock and bond yields over timeprovides valuable insight into the underlyingconditions prevailing in the Greek financial sec-tor. Conversely, it would be useful to define theobjectives of such an exercise beforehand, inorder to ensure the efficient use of the measure.

Notably, we find that the Greek stock-bondcorrelation is affected by liquidity conditions

in the banking system and by interactions stem-ming from stock-bond correlations in othermarkets, particularly the Italian market. In tur-bulent periods only the latter effects are sig-nificant. Overall, decoupling effects reflectedin negative correlations are found to be sig-nificantly related to increased market uncer-tainty, liquidity conditions in the banking sys-tem and contagion effects mainly stemmingfrom Italy. These results are also confirmed asregards the turbulent period after 2007. Bycontrast, the AAA credit spread has not beenfound to significantly affect the Greek stock-bond correlation; this suggests that the decou-pling effects are not to be interpreted asreflecting deteriorating credit risk conditions.This finding indicates that the decouplingeffects reflected in the negative stock-bondcorrelations cannot be explained as flight-to-quality against credit risk.

5 CONCLUDING REMARKS

This paper has assessed the dynamic correla-tion between Greek stocks and bonds. Ourmain objective was to examine the determi-nants of the dynamics of the Greek stock-bond

33Economic Bulletin

May 2010 87

Note: Each cell presents the value (standard error) of the coefficient of the respective explanatory variable, as in equation 3, with the depend-ent variable being the negative values of the Greek stock-bond correlation. The estimation has been performed with consistent het-eroskedasticity errors, estimated with the Newey-West formula. Asterisks (**, *) denote (5%, 10%) significance.

(SH-SL)t-1-2.020**

(0.578)-1.125

(1.167)-0.537

(0.289)-2.150**

(0.867)-1.931*(1.067)

CSt-1-0.007

(0.004)0.042

(0.076)-0.003

(0.025)-0.016

(0.015)0.013

(0.023)

(E-L)t-10.006** (0.002)

0.030(0.022)

0.000(0.001)

0.015**(0.007)

0.007(0.012)

ρUSt–1

-0.034 (0.024)

-0.101(0.063)

-0.002(0.005)

-0.136**(0.045)

-0.058(0.087)

ρITt–1

0.128** (0.029)

0.026(0.075)

0.006(0.009)

0.261**(0.035)

0.246**(0.054)

ρDEt–1

-0.028 (-0.028)

0.161*(0.088)

0.005(0.008)

0.001(0.074)

-0.149(0.128)

~R2 0.241 0.137 0.035 0.436 0.191

D-W 0.583 0.519 1.038 0.920 0.742

1/1/1999-9/12/2009

1/1/1999-4/12/2001

1/1/2001-9/8/2007

9/8/2007-9/12/2009

15/9/2008-9/12/2009

Table 3 Determinants of negative values of the Greek stock-bond correlation

correlation. In this context, we included vari-ables reflecting stock market uncertainty,credit risk, liquidity conditions in the bankingsystem and interactions from similar correla-tion dynamics in the US, Germany and Italy.Our findings indicate that the correlation isinfluenced by the degree of uncertainty, whilenegative correlations have been explained asindicating flight-to-quality effects.

The results presented confirm the findingsreported in empirical literature concerningother markets, namely that the stock-bond cor-relation is negatively related to uncertainty andsubject to contagion effects. In particular, theexamination of the determinants of negativevalues of the stock-bond correlation suggestsflight-to-quality effects, indicating that the cor-relation is a good proxy for market uncertainty.However, our findings suggest that flight-to-quality reflected in the correlation between

Greek stocks and bonds may not be a result ofheightened credit risk.

In this context, our analysis has assessed the driv-ing factors of the stock-bond correlation. Wedeem that our findings may serve as a basis forusing stock-bond correlation as a tool for mar-ket analysis and policy decisions, for Greece.However, the dynamic nature of this measureand the changing nature of market conditionsdisallow optimism that the answer providedherein is conclusive; the main technical findingis that the dependent variable is subject to shiftsrelated to different states of volatility. Addi-tionally, apart from the specified determinantsrelated to a variety of risk perceptions, severalother explanatory variables may exist, probablyexplaining the residual variations. In this respect,we deem that macroeconomic reflections in theGreek stock-bond correlation would be an inter-esting topic for future research.


d’Addona, S. and A.H. Kind (2006), “International stock-bond correlations in a simple affineasset pricing model”, Journal of Banking and Finance, 30(10), 2747-65.

Alexopoulou, I., I. Bunda and A. Ferrando (2009), “Determinants of government bond spreadsin new EU countries”, ECB Working Paper no. 1093.

Baele L., A. Ferrando, P. Hördahl, E. Krylova and C. Monnet (2004), “Measuring EuropeanFinancial Integration”, Oxford Review of Economic Policy, 20(4), 509-30.

Baur, D. and B. Lucey (2009), “Flights and contagion – an empirical analysis of stock-bond cor-relations”, Journal of Financial Stability, 5(4), 339-52.

Beber, A., M. Brandt and K. Kavajecz (2006), “Flight-to-quality or flight-to-liquidity? Evidencefrom the euro-area bond market”, NBER Working Paper no. 12376.

Beirne, J., G. Caporale, M. Schulze-Ghattas and N. Spagnolo (2009), “Volatility spillovers andcontagion from mature to emerging stock markets”, ECB Working Paper no. 1113.

Bollerslev, T. (1990), “Modelling the coherence in short run nominal exchange rates: a multi-variate generalized ARCH model”, Review of Economics and Statistics, 72(3), 498-505.

Campbell, J. and J. Ammer (1993), “What moves the stock and bond markets? A variance decom-position for long-term asset returns”, The Journal of Finance, 48(1), 3-37.

Cappiello, L., R. Engle and K. Sheppard (2003), “Asymmetric dynamics in the correlations ofglobal equity and bond returns”, ECB Working Paper no. 204.

Centeno, M. and A.S. Mello (1999), “How integrated are the money market and the bank loansmarket within the European Union?”, Journal of International Money and Finance, 18, 75-106.

Chiang, T.C., B.N. Jeon and H. Li (2007), “Dynamic correlation analysis of financial contagion:Evidence from Asian markets”, Journal of International Money and Finance, 26(7), 1206-28.

Christensen, K. and M. Podolskij (2006), “Realized range-based estimation of integrated vari-ance”, Journal of Econometrics, 141(2), 323-49.

Connolly, R., S. Stivers and L. Sun (2005), “Stock market uncertainty and the stock-bond returnrelation”, Journal of Financial and Quantitative Analysis, 40(1), 161-94.

Connolly, R., S. Stivers and L. Sun (2007), “Commonality in the time-variation of stock-stockand stock-bond return comovements”, Journal of Financial Markets, 10, 192-218.

Davies, A. (2006), “Testing for international equity market integration using regime switchingcointegration techniques”, Review of Financial Economics, 15, 305-21.

Ehrmann, M., M. Fratzscher and R. Rigobon (2005), “Stocks, Bonds, Money Markets andExchange Rates: Measuring International Financial Transmission”, NBER Working Paper no.11166.

Elliott, G., T.J. Rothenberg and J.H. Stock (1996), “Efficient tests for an autoregressive unit root”,Econometrica, 64, 813-36.

Engle, R. (2002) “Dynamic conditional correlation – a simple class of multivariate GARCH mod-els”, Journal of Business and Economic Statistics, 17, 239-50.

Fama, E. and K. French (1993), “Common risk factors in the returns on stocks and bonds”, Jour-nal of Financial Economics, 33, 3-56.

Hakkio, C. and W. Keeton (2009), “Financial stress: what is it, how can it be measured and whydoes it matter?”, Economic Review, 2nd quarter, 5-50.

Joshy, J. (2008), “Estimation and forecasting of stock volatility with range-based estimators”, Jour-nal of Futures Markets, 28(6), 561-81.

Krugman, P. (1991), “Increasing returns and economic geography”, Journal of Political Econ-omy, 99(3), 483-99.

Laloux, L., P. Cizeau, M. Potters and J.P. Bouchaud (2000), “Random matrix theory and finan-cial correlations”, International Journal of Theoretical and Applied Finance, 3(3), 391-97.

Lane, P. (2006), “Global Bond Portfolios and EMU”, International Journal of Central Banking,2(2), 1-23.

33Economic Bulletin

May 2010 89

R E F E R ENC E S

Sarno, L. and G. Valente (2005), “Modelling and forecasting stock returns: exploiting the futuresmarket, regime shifts and international spillovers”, Journal of Applied Econometrics, 20(3),345-76.

Vo, X.V. (2009), “International financial integration in Asian bond markets”, Research in Inter-national Business and Finance, 23(1), 90-106.

Yang, J., J. Kolari and P.W. Sutanto (2004), “On the stability of long-run relationships betweenemerging and US stock markets”, Journal of Multinational Financial Management, 14, 233-48.

Yang, J., Y. Zhou and Z. Wang (2009), “The stock-bond correlation and macroeconomic con-ditions: one and a half centuries of evidence”, Journal of Banking and Finance, 33, 670-80.


9944.. Monetary time series of South-EasternEurope from the 1870 to 1914Members of the SEEMHN data collectiontask force

9955.. Modelling distortionary taxationPanagiotis Chronis

9966.. Fertility determinants and economicuncertainty: an assessment using Euro-pean panel data George Hondroyiannis

9977.. Macroeconomic implications of alterna-tive tax regimes: the case of GreeceDimitris Papageorgiou

9988.. The Greek current account deficit: is itsustainable after all?George A. Zombanakis, ConstantinosStylianou and Andreas S. Andreou

9999.. Optimum currency areas, structuralchanges and the endogeneity of the OCAcriteria: evidence from six new EU Mem-ber StatesDimitrios Sideris

110000.. Revisiting the merger and acquisition per-formance of European banks Ioannis Asimakopoulos and Panayiotis P. Athanasoglou

110011.. Bank heterogeneity and monetary policytransmissionSophocles N. Brissimis and Manthos D. Delis

110022.. An Optimum-Currency Area OdysseyHarris Dellas and George S. Tavlas

110033.. Benchmark bonds interactions underregime shiftsPetros M. Migiakis and Dimitris A. Geor-goutsos

110044.. Fiscal adjustments and asset price move-ments Athanasios Tagkalakis

110055.. Money supply and Greek historical mon-etary statistics: definition, construction,sources and dataSophia Lazaretou

110066.. The effect of asset price volatility on fis-cal policy outcomes Athanasios Tagkalakis

110077.. An alternative methodological approachto assess the predictive performance ofthe moving average trading rule in finan-cial markets. Application for the LondonStock Exchange Alexandros E. Milionis and EvangeliaPapanagiotou

110088.. The margins of labour cost adjustment:survey evidence from European firms Jan Babecký, Philip Du Caju, TheodoraKosma, Martina Lawless, JuliánMessina and Tairi Rõõm

110099.. Assessing market dominance: a commentand an extensionVassilis Droucopoulos and PanagiotisChronis

111100.. Downward nominal and real wage rigidity:survey evidence from European firmsJan Babecký, Philip Du Caju, TheodoraKosma, Martina Lawless, JuliánMessina and Tairi Rõõm

This section contains the abstracts of Working Papers authored by Bank of Greece staff and/orexternal authors and published by the Bank of Greece. The unabridged version of these publi-cations is available in print or electronic format on the Bank’s website (www.bankofgreece.gr).

CONTENTS

33Economic Bulletin

May 2010 91

WORK I NG P A P ER S(FEBRUARY 2009 – FEBRUARY 2010)

Monetary Time Series of Southeastern Europe from the 1870s to 1914

Working Paper Νο. 94Members of the SEEMHN Data Collection Task Force

The SEEMHN Data Collection Task Force isa cross-national collaboration of economistsand statisticians from six Southeastern Euro-pean (SEE) countries that put together for thefirst time a comprehensive collection of mon-etary data for the pre-WWI period. The coun-tries covered include: Albania, Austria-Hun-gary, Bulgaria, Greece, Romania and Serbia.For each country, historical aggregates are pre-ceded by an introduction which provides aninstitutional and historical narrative on theSEE region and its monetary institutions andstandards. The dataset refers to four key vari-

ables: banknotes in circulation, total reserves,exchange rates and central bank discount rates.The frequency is monthly and the time spancovers the period from 1870 or earlier to 1914.Each country uses standardised definitions foreach series to facilitate cross-country com-parison. A foreword by Michael Bordo on theparamount importance of historical data seriesis supplemented by a detailed introduction byMatthias Morys placing an emphasis on across-country historical comparison of thecountries’ monetary institutional structuresand the policy-making process.

Modelling distortionary taxation

Working Paper Νο. 95Panagiotis Chronis

The main aim of this work is to provide amethodology for analysing the distortionarycharacter of the tax system. It offers a formalproof of a general functional form which canbe used for this purpose.

In public economics, distortionary taxation isrelated, on the one hand, to the distortionsincome tax brings to the labour market – byaffecting the behaviour of tax payers viaincome and substitution effects (Stiglitz, 1988;Atkinson and Stiglitz, 1980; Cowel, 1981) orby affecting labour mobility (Goodspeed,2002) – and on the other hand, to “tax com-pliance” (Andreony et al., 1998) and “tax eva-sion” (Allingham and Sandmo, 1972; Clotfel-ter, 1983; Feinstain, 1991), as the higher thetax burden, the higher the probability of taxevasion. In the presence of tax evasion, unre-ported income escapes taxation and, as aresult, the higher the required taxes, or equiv-alently the tax rates, the greater the distor-

tionary effect of income taxation. This is alsorelated to tax incidence, since for those bur-dened by taxation it is easier to evade and payfewer taxes. Hence, the tax rate has a substi-tution effect encouraging tax evasion and anincome effect discouraging it (Yitzaki, 1974;Clotfelter, 1983).

Therefore, the choice of tax rate has a decisiverole to play in the framework of distortionarytaxation, and its relation to the notions of taxcompliance, tax evasion and tax collection. It isalso clear that the tax rate is related to the behav-iour of tax payers; it determines their behaviourregarding methods of tax burden avoidance. Asa result, within a distortionary taxation system,tax revenues are lower compared to a non-dis-tortionary system (e.g. lump-sum taxation).

Turning now to the majority of macroeconomicmodels that use taxation in their analysis, wecan distinguish two main ways taxation is intro-

33Economic Bulletin

May 2010 93


Fertility determinants and economic uncertainty: an assessment using European panel data

Working Paper Νο. 96George Hondroyiannis

duced. The first treats the tax instrument asexogenous, by assuming lump-sum taxation.Consequently, it ignores the relation betweentaxes and income, as well as its growth rate,since, under this assumption, no distortionsexist. The second way taxation is treated in theliterature is the proportional taxation scheme,where tax revenue is written as a fraction ofincome. This is often called a distortionary tax-ation system (in the sense that changes in thelevel of income, caused by the fiscal authority’stax policy, affect tax revenues).

Furthermore, the proportional taxation schemeis consistent with Ramsey’s approach, usuallyused in macroeconomics in order to study opti-mal policy problems in a dynamic setting, wherethe deviation from lump-sum taxation allowsthe benevolent government to avoid the firstbest solution, and hence a distortionary frame-work ensues. However, this approach is subjectto criticism. Golosov et al. (2006) argue that,under the Ramsey setting, the main goal for thegovernment is to mimic lump-sum taxes, whileit is not clear why the tax instrument takes aparticular form. As a result, the Ramseyapproach does not provide a theoretical foun-dation for distortionary taxation. Distortionsare simply assumed and their overall level isdetermined exogenously by the need of thebenevolent government to finance its spending.

A closer look at the above taxation schemesmakes clear that the notion of a distortionarytaxation system, in the sense described previ-

ously (including tax compliance, tax evasionand tax collection), is depicted neither in thelump-sum nor the proportional taxationscheme. In fact, including it would complicatethe analysis, since it will introduce a non-lin-ear dynamic relation that relates the behaviourof tax payers to the path of tax revenue, the taxrate and its relation to income. This dynamicrelation results in a loss of tax revenue for thefiscal authority, with a deterministic role inequilibrium (if equilibrium occurs at all) ofeconomic models that investigate the relationbetween distortionary taxation and the path ofdebt, and/or its relation to inflation and thepath of income (Woodford, 2001, 2003;Leeper, 1991, 2005; Chari and Kehoe, 2007).

To the best of our knowledge, the approachjust described is not treated in the relevant lit-erature. Hence, if one wants to include intothe analysis the consequences of a distor-tionary taxation system on the economy, it isvital to deal with the above mentioned prob-lem of non-linearity.

The purpose of this paper is to provide amethodology for dealing with this problem byusing formal analysis. For this we build andsuggest a general functional form whichrelates the behaviour of tax payers to the taxrate and the path of tax revenue, and we thenstudy its properties as well as its economicrationale. We finally use an example, in orderto show the way this functional form can beapplied to economic models.

This paper examines the determinants of fer-tility, using panel data for 27 European coun-tries. We employ panel cointegration to esti-mate fertility as a function of demographic andeconomic variables. The analysis suggests thatfertility variations in Europe are explained bychanges in infant mortality, female employ-

ment, nuptiality, old age dependency ratio, realincome per capita, real wage and output volatil-ity. Two measures, production volatility and theunemployment rate, are used as proxies tomeasure uncertainty and insecurities in moderneconomies.

33Economic Bulletin

May 2010 95

Macroeconomic implications of alternative tax regimes: the case of Greece

Working Paper Νο. 97Dimitris Papageorgiou

The empirical results for the panel data indi-cate that a downward shock to infant mortality,due to e.g. medical advances, leads to lower fer-tility. An upward shock in real GDP per capita,due to e.g. improved terms of trade, leads to anincrease in fertility. This implies a positiveincome effect on the demand for children.However, an increase in the real wage, which isa proxy for the opportunity cost of children, willdecrease fertility. This implies a negative sub-stitution effect, which is reinforced by theincrease in female employment. Hence theopportunity cost of time devoted to childcarehas increased as a result of increasing real wageand female labour force participation, and con-sequently fertility has declined. At the sametime, an increase in nuptiality leads to higherfertility, since children constitute one of themost important gains of marriage.

Finally, the empirical results support the propo-sition that economic uncertainty might be animportant determinant of the fertility decision,explaining the decline in fertility in Europe.The empirical findings suggest that ineconomies with high per capita income, anincrease in output volatility associated with loweconomic performance leads to lower fertility.Contrary to that, in economies with low incomeper capita, an increase in output volatility asso-ciated with a downturn in economic activity, oran increase in real economic activity associated

with reduction in volatility, will result in adecline of fertility. Likewise, high unemploy-ment increases labour market insecurities.Since children are long-term commitments,labour market uncertainties might dominateadults’ decisions for partnership and parent-hood, since responsible couples will decide tohave children when they are able to secure theirpresent and future income.

The empirical results have important implica-tions for the sample of European countries,with declining fertility and aging population.Fertility is found to be related to demographicand economic factors, such as infant mortality,nuptiality rate, female employment, real percapita output, real wage and uncertainty, sothat the movements in fertility rates observedover the last three decades can be accountedfor by changes in these economic variables.Therefore, the declining fertility rates in theEuropean countries can be explained by thepositive effect of a lower infant mortality rate,the negative effect of a higher price of havingchildren, due to increasing female employment,outperforming the positive effect of a higherincome on the demand for children. Finally, thedeclining fertility rates may be attributed to thegrowing uncertainties related to labour marketinsecurities, which are a major feature of mod-ern European societies, characterised by inter-nationalisation and globalisation.

This paper examines how changes in the taxmix (defined as distribution of revenue by typeof tax) influence economic activity and welfarein the Greek economy. Conducting tax policyanalysis using a Dynamic General Equilibriummodel which incorporates a detailed fiscal (tax-spending) policy structure, this is the first studythat analyses the implications of changes in thetax policy mix for the Greek economy within aDynamic General Equilibrium framework.

The results suggest that tax reforms reducingthe capital income tax rate and increasing thelabour or the consumption tax rates lead tohigher levels of output, consumption andinvestment. A reduction in the labour incometax rate met by an increase in the consumptiontax rate increases output, consumption, invest-ment and hours worked. The opposite resultsare observed when the capital tax rate increasesin order to meet the loss in labour tax revenue.


Cuts in the consumption tax rate compensatedby increases in labour or capital income taxrates have a negative impact on output andinvestment both in the short and the long run.

The results also suggest that if the goal of taxpolicy is to promote growth by replacing one

distortionary tax rate with another, then itshould reduce the tax rate on capital incomeand increase the tax rate on consumption. Onthe other hand, if the goal of tax policy is topromote lifetime welfare, then it shoulddecrease the labour income tax rate andincrease the consumption tax rate.

The present study has two aims. The first is toanalyse the degree of convergence of six newmember states (NMS) of the European Union

with the euro area, in an attempt to evaluatetheir readiness to adopt the euro. The six coun-tries are Bulgaria, the Czech Republic, Hungary,

The aim of the present paper is to shed someadditional light on the issue of the sustain-ability of the Greek current account deficit.The variety of techniques employed and con-clusions reached in the literature regarding thespecific issue points to the use of ArtificialNeural Networks as the appropriate analysistool, given a number of advantages over tra-ditional methods of analysis. The paperdemonstrates that, despite its size, the currentaccount deficit of Greece can be consideredsustainable. The conclusion takes into consid-eration that the current account has been bur-dened during the past few years by a numberof exogenous disturbances (in particular, oilprice increases as well as import payments forships to cover the high demand for global seatransport services), the effect of which isexpected to be either temporary or linked toinvestment activity. In addition, the reductionin the effect brought about by some of theseexogenous variables, to a large extent as a con-sequence of the recent international crisis, isexpected to help by bringing the current

account as a share to GDP down to possiblyone-digit levels.

The endogenous structural weaknesses of theeconomy, however, appearing in the form ofimport price inelasticity and lack of importsubstitution, together with high income elas-ticity of imports on one hand and the well-known competitiveness problems on theother, are expected to continue imposing anumber of binding constraints on the functionof the Greek external sector, and conse-quently sustaining the pressures exercised onthe current account deficit even in theabsence of any financing requirements. Thismeans that unless such imbalances are takencare of, the threat of an unsustainable currentaccount deficit will persist, despite theabsence of any financing requirements, espe-cially in an environment of recovery from therecent global crisis, with all the adverse reper-cussions that such a deficit figure might entailfor vital economic aggregates like growth andemployment.

The Greek current account deficit: is it sustainable after all?

Working Paper Νο. 98George A. Zombanakis, Constantinos Stylianou and Andreas S. Andreou

Optimum currency areas, structural changes and the endogeneity of the OCA criteria:evidence from six new EU Member States

Working Paper Νο. 99Dimitrios Sideris

Poland and Romania, which have not adoptedthe euro yet but plan to do so at some futuredate, and the Slovak Republic which adopted theeuro in January 2009. The work examineswhether these countries form an OCA with theeuro area by using the Generalised PurchasingPower Parity (GPPP) theory. The GPPP theoryproposes testing whether the real exchange ratesof a group of economies with respect to a basecurrency converge in the long run.

The second aim is to investigate whether theintroduction of the euro and the decision of thesix countries to seek to join the euro area havefostered integration of these countries with theeuro area. We argue that the decision of the sixeconomies to join the monetary union and thepolicy steps made towards convergence with it,have already promoted integration. This idea isin line with the endogenous OCA theory, whichsupports the view that countries joining a cur-rency union may satisfy the criteria of an OCAex post, even if they do not ex ante.

In the empirical work, cointegration analysis isemployed to test the GPPP hypothesis, after aninitial assessment of the stationarity of eachreal exchange rate series. The cointegrationanalysis examines the joint behaviour of therates, in three different periods: the full period1993.1-2007.12, and the periods before andafter the introduction of the euro. The results

provide evidence in favour of an OCA with theeuro area only for the period following theintroduction of the euro. The results indicatethat the group of the six economies has enjoyedreduced real exchange rate instability in thepost-euro period. This could be due toincreased trade integration of the six economieswith the EU, caused by the introduction of theeuro and the switch of the exchange rate poli-cies of most NMSs towards euro-basedexchange rate regimes.

For indicative purposes, a similar analysis withrespect to the US economy is also performed. Itsuggests that alignment of the countries with theUS is supported by the data for the period beforethe introduction of the euro, but not for theperiod following it. These results probably reflectthe weakening role of the dollar in the Europeanmarkets and the policy change in the six countries.

Overall, the findings imply that the conver-gence process with the euro area has been pro-moted in recent years probably as a result of theconvergence of the policies of the sixeconomies, the structural changes that tookplace in the economic systems of the countries,their increased trade integration with the Euro-pean Union and the significant role of the euroin the European markets. Thus, at present, thesix economies are quite well aligned with theeuro area.

The study examines the value creation ofMerger and Acquisition (M&A) deals in Euro-pean banking from 1990 to 2004. This is per-formed, first, by examining the stock pricereaction of banks to the announcement ofM&A deals and, second, by analysing thedeterminants of this reaction. The findingsprovide evidence of value creation in Euro-pean banks, as the shareholders of the targetshave benefited from positive and (statistically)

significant abnormal returns, while those of theacquirers have earned small negative but non-significant abnormal returns. In the case of theshareholders of the acquirers, domesticM&As, and especially those between bankswith shares listed on the stock market, seem tobe more beneficial compared to cross-borderones, or those when the target is unlisted.Shareholders of the targets earn in all casespositive abnormal returns. Finally, although

33Economic Bulletin

May 2010 97

Revisiting the merger and acquisition performance of European banks

Working Paper Νο. 100Ioannis Asimakopoulos and Panayiotis P. Athanasoglou

Understanding the transmission mechanism iscrucial for monetary policy. In this respect, thespecial role of banking institutions in thismechanism has been studied extensively bothat a theoretical and empirical level. The exist-ing evidence shows that banks alter their lend-ing behaviour in specific ways following achange in monetary policy. But does this reac-tion involve only changes in lending behaviour,as studied in the bulk of the literature? And doall banks in the market respond uniformly tomonetary policy changes? This paper tries toanswer these questions by analysing empiricallythe heterogeneous response of US and euroarea banks over the period 1994-2007 in termsof their lending and risk-taking decisions fol-lowing a monetary policy change. The ultimateimpact of a monetary policy change on bankperformance is also considered. In the empir-ical strategy, the distributional effects of mon-etary policy stem from a number of bank char-

acteristics, namely liquidity, capitalisation andmarket power.

The empirical analysis elucidates the sourcesof differences in the response of banks tochanges in policy interest rates by disaggre-gating, down to the individual bank level. Thisis achieved by the use of a Local GMM tech-nique that also enables a quantification of thedegree of heterogeneity in the transmissionmechanism.

As banks have a special role in the financing ofeconomic activity, their heterogeneous behav-iour is of particular importance to researchersand policy makers alike. The results of thispaper show that the extensive heterogeneity inbanks’ response identifies overlooked conse-quences of bank behaviour and highlightspotential monetary sources of the currentfinancial distress.

The theory of optimum currency areas wasconceived and developed in three highly influ-ential papers, written by Mundell (1961),McKinnon (1963) and Kenen (1969). Thoseauthors identified characteristics that poten-tial members of a monetary union should ide-ally possess in order to make it feasible to sur-render a nationally-tailored monetary policyand the adjustment of an exchange rate of anational currency. We trace the development

of optimum-currency-area theory, which, aftera flurry of research into the subject in the1960s, was relegated to intellectual purgatoryfor about 20 years. We then discuss factors thatled to a renewed interest into the subject,beginning in the early 1990s. Milton Friedmanplays a pivotal role in our narrative; Fried-man’s work on monetary integration in theearly 1950s presaged subsequent optimum-currency-area contributions; Mundell’s classic


the link between abnormal returns and fun-damental characteristics of the banks is ratherweak, it appears that the acquisition of smaller,less efficient banks generating more diversified

income is more value-creating, while acquisi-tion of less efficient, liquid banks characterisedby higher credit risk is not a value-creatingoption.

Bank heterogeneity and monetary policy transmission

Working Paper Νο. 101Sophocles N. Brissimis and Manthos D. Delis

An Optimum-Currency-Area Odyssey

Working Paper Νο. 102Harris Dellas and George S. Tavlas

The present paper formulates a multivariateframework that captures the interactionsbetween fixed income markets, by taking intoaccount the time-varying characteristics offinancial integration. The main contribution ofthis paper to the relevant literature is that itrelaxes the linearity restriction on the inter-actions between the markets examined, byallowing the system to shift between regimesaccording to underlying causal relations. Inparticular, we suggest that causal relationsamong financial markets are subject to regimeshifts, resulting in changes in the strength anddirection of interactions between them.

Our empirical analysis focuses on the interac-tions of major bond markets of the euro area,namely Germany, France, Italy and theNetherlands, and the US Treasury market, forthe period 1992-2007. First, we specify thelong-run relations of the system. Specifically,we examine the cointegration properties of theunderlying system and find that there exists asingle common trend governing the stochasticprocess of the system, thus satisfying the pre-condition imposed by previous financial inte-gration literature (Haug et al., 2000). Next, weestimate the long-run equilibrium relations, byimposing restrictions in line with the interestrates parity hypothesis. The tests confirm thatthe system is characterised by significant par-ity relations.

In the second part of the empirical analysis, weexamine whether the causal relations of the sys-

tem are subject to regime-switching effects, byapplying the Markov Switching Vector ErrorCorrection Model (VECM). Employing theMS-VECM framework proves informative as itreveals the impact of monetary unification inEurope on underlying interactions betweenmarkets. Specifically, a regime shift is found tohave occurred close to the time of the launchof the common currency, resulting in strength-ened interactions among the underlying vari-ables and significantly enhanced causal adjust-ment effects towards the long-run equilibria.

In the last part of our paper, we assess leader-follower relations by means of a variancedecomposition analysis for each of the tworegimes (Lütkepohl and Reimers, 1998). Theresults indicate significant changes in the pro-portion of the variables’ variances explained bythe Bund or the Treasury after 1999.Although these two variables are found toexplain the largest proportion of the variancein the system prior to the monetary unification,their dominant characteristics are significantlylimited after the regime shift and replaced bya three-fold benchmark status. In particular, inthe period after monetary unification, move-ments in European variables are found to becaused, to a large degree, by the Bund but onlyfor the initial period (1-2 months) after theshock. However, in the longer run, the vari-ables with the most important effects are foundto be the French OAT (for the period of 5-12months after the shock) and the Italian bond(in the interim period). These results are in

33Economic Bulletin

May 2010 99

formulation of an optimal currency area wasaimed, in part, at refuting Friedman’s“strong” case for floating exchange rates; andFriedman’s work on the role of monetary pol-icy had the effect of helping to revive interestin optimum-currency-area analysis. The paperconcludes with a discussion of recent analyti-

cal work, using New Keynesian models, whichhas the promise of fulfilling the unfinishedagenda set-out by the original contributors tothe optimum-currency-area literature, that is,providing a consistent framework in which acountry’s characteristics can be used to deter-mine its optimal exchange-rate regime.

Benchmark bonds interactions under regime shifts

Working Paper Νο. 103Petros M. Migiakis and Dimitris A. Georgoutsos

line with recent empirical findings on bench-mark status in European bond markets (Dunneet al., 2002). However, they provide a morecomplex answer to the question of ‘benchmark

status’ in the European government bond mar-ket, allowing for the characterisation of a bondas benchmark to be related to the duration ofits causality effects.

In the context of the current macroeconomicand financial market turmoil, and in the pres-ence of asset price swings, expansionary fiscalpolicy has been at the forefront. As the eco-nomic recovery gathers pace, governmentsaround the globe will start withdrawing someof the fiscal stimulus and initiating fiscal con-solidation in order to ensure sustainable fiscalpositions. These consolidation efforts will beaffected by asset price movements as the econ-omy picks up.

As has been discussed by earlier literature(Schuknecht and Eschenbach, 2004), assetprices can affect the budget via a series ofchannels. Directly, they affect the budget viacertain revenue categories, e.g. taxes related tocapital gains/losses. Indirectly, higher assetprices raise consumer confidence and con-sumption, via the wealth effect, and increasethe collection of indirect taxes. Finally, in caseof asset price busts and ailing financial insti-tutions, the state might be asked to intervenebearing some of the costs.

The objective of this paper is to improve ourunderstanding of the links between asset pricemovements and fiscal adjustments. Building onprevious literature on the determinant of suc-cessful fiscal adjustments (Alesina and Perotti,1995 and 1997; Tavares, 2004; Tagkalakis,2009), it goes one step further by investigating

whether asset price movements have any effecton the probability of initiating and successfullyconcluding fiscal adjustment. Following earlierstudies (Borio and Lowe, 2002), we use a realaggregate asset price index (from the Bank ofInternational Settlements) and its three sub-components, real residential, real commercialand real equity prices.

Our findings suggest that a pick up in assetprices increases the probability of initiating afiscal adjustment, but does not necessarily leadto a sustainable correction of fiscal imbalances.This finding is reaffirmed both for residentialand commercial property prices. This is in linewith the recent experiences of several countrieswhich faced a rapid deterioration of tax rev-enues following the collapse of residentialactivity and prices. However, increases inequity prices contribute positively both to theinitiation and the successful conclusion of a fis-cal adjustment.

Therefore, fiscal policy makers should focustheir attention on the sustainable improvementof fiscal balances by adopting measures of astructural nature. In doing so, they should alsotake on board, on top of the effect of the eco-nomic cycle, the effects that asset price move-ments have on fiscal balances, in order to havea better grasp of the actual fiscal stance and therealised fiscal outcomes.


Fiscal adjustments and asset price movements

Working Paper Νο. 104Athanasios Tagkalakis

This paper attempts to provide, for the firsttime, a survey of the construction of estimatesof the quantity of money in Greece since theestablishment of the National Bank of Greecein 1842 until the eve of WWII. Specifically, we

describe in detail the methods of constructionand the sources of data used in building theseaggregates. We discuss the data collection pro-cedure and publication practices. The endproduct is presented in a data appendix.

33Economic Bulletin

May 2010 101

Money supply and Greek historical monetary statistics: definition, construction, sources and data

Working Paper Νο. 105Sophia Lazaretou

The ongoing economic and financial marketturmoil has been accompanied by a signifi-cant fall in asset prices, following severalyears of asset price boom. These asset pricemovements affect the fiscal policy stancethrough a series of channels (see e.g.Schuknecht and Eschenbach, 2004). Directly,they affect specific revenue categories, e.g.capital gains and losses related to direct taxeson households and businesses. Indirectly, theyaffect revenue via a feedback loop fromhigher asset prices to real economic activity(higher asset prices raise consumer confi-dence and consumption, via the wealth effect)which increases the collection of indirecttaxes. Finally, in case asset price busts lead todefaults of financial institutions, the state willbe asked to intervene to preserve the stabil-ity of the financial system.

This paper, building on the work of Fatas andMihov (2003), goes beyond the previous liter-ature in examining whether asset price volatil-ity amplifies the volatility of fiscal policy out-comes. Following previous studies such asJaeger and Schuknecht (2004), we use a realaggregate asset price index (taken from theBank of International Settlements) and itsthree subcomponents: real residential, realcommercial and real equity prices. According

to our findings, there is significant evidencethat asset price volatility affects the volatilityof discretionary fiscal policy in a positive andsignificant manner. Higher residential propertyprice volatility leads to more volatile govern-ment spending and, thus, to a more volatile dis-cretionary fiscal policy stance. Equity pricevolatility affects the volatility of the fiscal pol-icy stance primarily via the government rev-enue channel.

These findings point to the need for adjustingfiscal balances, and in particular governmentrevenues, for both the economic and the assetprice cycle, in order for the policy maker tohave a clear grasp of developments in the fis-cal policy stance.

Given that asset price movements amplify thevolatility of the discretionary fiscal policystance, which in turn amplifies business cyclefluctuations and harms economic growth(Fatas and Mihov, 2003), a word of caution isneeded as regards the continuation of fiscalpolicy interventions undertaken in response tothe ongoing economic and financial marketturmoil. Such interventions also entail the riskof increasing policy volatility, generating a neg-ative effect on economic growth, thus puttinga toll on economic recovery.

The effect of asset price volatility on fiscal policy outcomes

Working Paper Νο. 106Athanasios Tagkalakis

The hypothesis of efficient markets states thatasset prices reflect all available information,hence, future prices are not predictable frompast and current information. By contrast, theso-called technical analysis (i.e. the study ofmarket action, primarily through the use ofcharts, for the purpose of forecasting futureprice trends) has been a thriving activity formore than a century. Among the trading rulesused by researchers to test market efficiency,the one employed most frequently is the so-called moving average (MA) rule. In one of itssimplest versions, two non-centred, movingaverages with different lengths are createdfrom the time series of stock prices. Buy (sell)signals are then generated when the relativelyshorter moving average crosses the relativelylonger moving average from below (above).

A usual way of measuring the predictive powerof the MA trading rule is to test for the statis-tical significance of the difference between themean return of individual trading periods char-acterised as “buy” (trading periods that accord-ing to the MA trading rule the investor’s cap-ital should remain invested in the market) andthe mean return of the whole investmentperiod, and/or the mean return of individualtrading periods characterised as “sell” (tradingperiods that the investment capital should beliquidated or sold short) and the mean returnof “buy” trading periods, and/or the meanreturn of “sell” periods and either the meanreturn of the whole investment period or zero.Although on several occasions simple t-testshave been used to test for the statistical sig-nificance of such differences, strictly speaking,the application of the t-test which assumes nor-mal, stationary and time-independent distri-butions is not legitimate since one or more ofthese assumptions is very often violated in timeseries of asset returns. To overcome this prob-lem bootstrapping techniques have been sug-

gested and this approach has been recognisedas the established one and used by mostauthors.

This paper suggests an alternative testing pro-cedure for assessing the predictive perform-ance of the MA trading rule and argues that ithas some considerable advantages over theexisting one. In principle the proposedapproach allows for the concurrent estimationof measures of the predictive power of the MAtrading rule, such as the ones mentionedabove, and at the same time takes into con-sideration the interdependence in assetreturns.

The intended testing procedure is a modifica-tion of the so-called impact assessment mod-els, originally developed by Box and Tiao. Theterm impact assessment means a test of thenull hypothesis that an event caused a changein a stochastic process measured by a timeseries. Events (also called “interventions” intime series literature) may be represented bybinary variables. However, the standard para-metric or non-parametric statistical tests usedto test differences in levels (e.g. the t-test,ANOVA, etc.) cannot be used in serially cor-related data, as was explained previously.Moreover, a change in level may not take placeinstantaneously, but gradually.

The Box and Tiao method of impact assess-ment takes into account serial correlation, aswell as gradual level shifts, and allows forsimultaneous maximum likelihood estimationof the parameters related to the level changeand those related to the serial correlation.Hence, it can be considered as a generalisationof the t-test. Thus far, the method has beenused to study the effect of events which occuron continuous lags. In this work a modificationof the Box and Tiao methodology is suggested


An alternative methodological approach to assess the predictive performance of the moving average trading rule in financial markets. Application for the London Stock Exchange

Working Paper Νο. 107Alexandros E. Milionis and Evangelia Papanagiotou

and applied in cases where the events occurdiscontinuously. This can be made possibleunder the assumption that the effect of suchevents is transient. More specifically, as waspreviously explained, the role of the MA trad-ing rule is to classify each individual tradingperiod (in this case each individual day) as“buy” or “sell”. The “buy” days are consideredto be the discontinuous interventions. In thatway the modified Box and Tiao approach isused to examine whether or not there is pre-dictive power in the MA trading rule as appliedto capital markets.

In particular, the trading rule is applied to dailyclosing prices of the FT30 Index of the LondonStock Exchange for the time period 1935-1994.The total time period that the dataset consid-ered in this work covers, as well as the partic-ular capital market, approximately coincidewith that of previous studies by other scholarswho examined the predictive performance ofthe MA trading rule for the same stockexchange index, but using different techniques.This allows for a direct comparison of theresults derived following the statisticalapproach suggested above with those of theprevious studies. The total time period isdivided into three 20-year subperiods: 1935-1954, 1955-1974 and 1975-1994; and each sub-period is also examined. Moreover, the par-ticular combinations of long and short movingaverages for the application of the MA tradingrules employed are deliberately taken to coin-cide with those of previous studies.

For this purpose two different classes of mod-els are tested. The intervention component forboth models is the simplest possible (techni-cally, a model with a zero order interventioncomponent is used for both classes) and rep-resents the relative average increase (if any) onreturns during the “buy” periods, as comparedto the “sell” periods. With the first class ofmodels (simple benchmarks) any lineardependencies among index returns are disre-garded. Therefore, results for the statistical sig-nificance of the intervention parameter may bedirectly compared with the results from an

ordinary t-test for the mean difference of “buy-sell” for index returns. The effects of lineardependencies in index returns are taken intoaccount with the second class of models that isused, and the statistical significance of theintervention component may be compared withthe results from bootstrapping techniques. Theresults derived from the application of the pro-posed methodology as compared with theresults of existing methods (such as boot-strapping and the so-called cointegrationcumulative profit test) using the same datasetare found to be, by and large, equivalent. Morespecifically, the results for statistical signifi-cance for the intervention component agreecompletely with the results derived by boot-strapping techniques for all time periods andall trading rules.

Direct comparison regarding the significanceof the intervention component is not possiblewhen the cointegration cumulative profit isconsidered, since the approach followed by thelatter is based on a different reasoning. Qual-itatively, all methods provide evidence of apronounced weakening of the predictive powerof the MA trading rule for the last sub-period(1975-1994) for the London Stock Exchange,which implies that the market graduallybecame more efficient.

An advantage of the proposed methodology isthat it may be enhanced further, as it is sus-ceptible to several improvements at both atechnical and practical level. At first, a possi-ble problem with non-normal distributions inasset returns may be alleviated by consideringthe more general class of GED distributions forthe testing procedure. Further, second-orderdependencies in asset returns such as autore-gressive conditional heteroscedasticity may alsobe taken into account by considering a generalARMA-GARCH model for the noise compo-nent of the impact assessment model. More-over, the proposed methodology is suitable toquantitatively test empirical techniques of tech-nical analysis by allowing a higher-order inter-vention component in the stochastic model. Inthat way, the time-varying effects of trading sig-

33Economic Bulletin

May 2010 103


nals may be modelled and tested. It is notedthat the proposed methodology is less compu-tationally intensive than bootstrapping, and itsapplication is relatively easy for practitioners

in the field. Hence, overall, the proposedmethodology seems promising and potentiallyhas some considerable advantages compared tothe established and the existing ones.

Apart from a decrease in base wages, firmscould use other ways of reducing labour costswhen faced by negative external shocks, e.g. cutbonuses and benefits, encourage earlier retire-ment and hire workers at lower wages thanthose who have recently quit. The adjustmentof non-wage labour costs has gained attentionin the policy debate for two main reasons.First, non-wage labour costs represent a sub-stantial part of total compensation. Since firmsare primarily concerned with total compensa-tion per employee, an assessment of the flex-ibility of non-wage labour costs is as importantas the evaluation of the degree of wage flexi-bility. Second, in an environment of stickyprices and nominal wages, non-wage labourcosts become an important tool for adjustmentto external shocks, allowing a dampening of theeffects of negative demand shocks on the firm’semployment.

Using a unique survey of firms from 12 coun-tries of the European Union conductedbetween the second half of 2007 and the firstquarter of 2008 within the framework of theWage Dynamics Network coordinated by theEuropean Central Bank, the paper examinesthe importance and determinants of alternativestrategies firms might use to adjust their labourcosts. In the survey, firms’ managers wereasked directly about their use of these policiesin the recent past. Specifically, the incidenceof the following six labour-cost saving strate-gies can be identified: reduce or eliminatebonus payments; reduce non-pay benefits;change shift assignments or shift premia; slowor freeze the rate of promotions; recruit newemployees at a lower wage level than those

who left voluntarily; and encourage earlyretirement to replace high wage employees byentrants with lower wages.

The paper’s contribution to the literature isthreefold. First, it documents comparableinformation on labour cost adjustment prac-tices beyond base wages for a large set of EUcountries and sectors. This allows the analy-sis of the relative importance of each indi-vidual strategy across countries characterisedby different sets of laws and institutions gov-erning their labour markets. The survey showsthat to reduce labour costs firms fairly com-monly use strategies other than base wagereduction: 63% of the firms’ managers saidthey had used at least one other margin ofadjustment in the recent past, and 58% hadused at least one of the six margins explicitlyidentified in the survey. The results show thatthere is substantial heterogeneity in the use ofthese strategies across countries and firms,depending on firm characteristics and labourmarket institutions.

Second, the paper examines the characteristicsof firms and the environments in which theyoperate that determine the relative importanceof each type of labour cost adjustment mech-anism. The use of each margin is related to sev-eral firm characteristics, such as the relativesize or skills distribution, as well as severalindicators of the economic environment inwhich they operate. In particular, larger firmsshow a greater flexibility with respect to usingany of these strategies in order to adjust labourcosts. Different indicators of competitionintensity suggest that firms in more competi-

The margins of labour cost adjustment: survey evidence from European firms

Working Paper Νο. 108Jan Babecký, Philip Du Caju, Theodora Kosma, Martina Lawless, Julián Messina and Tairi Rõõm

33Economic Bulletin

May 2010 105

tive environments are more likely to use someof these strategies more heavily. Furthermore,the results indicate that the presence of unionsin wage-setting is associated with a greater useof most of the strategies. A plausible explana-tion is that unions limit the flexibility of wages,pushing firms towards alternative labour-cost-cutting strategies.

Finally, the paper shows how the use of theseadjustment practices can be related to firms’experience regarding nominal wage rigidity, aswell as to the extent of wage indexation applied

by firms. The impact of unionisation on the useof alternative margins remains even after dif-ferent indicators of wage rigidity (either nom-inal wage rigidity or alternative definitions ofwage indexation) are controlled for. Interest-ingly, it is shown that firms subject to nominalwage rigidities are much more likely to useeach of the six cost-cutting strategies. This indi-cates that there is some degree of substi-tutability between wage flexibility and the flex-ibility of other labour cost components, andthat this substitutability is not limited by thepresence of unions in wage setting.

Melnik et al. (2008) propose a measure in orderto assess whether the firm with the highest mar-ket share in an industry can be characterised asdominant. A positive answer to this question isa prerequisite for further investigating the extentto which market dominance comes together withthe abuse of such a dominant position, accord-ing to Article 82 of the EC Treaty.

Additionally, Guidelines, Regulations, relatedstatements by the European Commission and

European Court Decisions point to the inves-tigation of the degree of monopoly power thatcreates or strengthens a dominant position,particularly in cases of planned mergers orbuyouts.

In this paper we investigate certain mathe-matical properties of the proposed measure,identifying some of its weak points and relat-ing it to other research papers Melnik et al.had not considered in their study.

Assessing market dominance: a comment and an extension

Working Paper Νο. 109Vassilis Droucopoulos and Panagiotis Chronis

Based on a unique firm-level survey carried outbetween late 2007 and early 2008 within theframework of the Wage Dynamics Network,the flexibility of wages across 14 countries ofthe European Union (EU) is analysed. Theobjective of the paper is to examine the extentand determinants of downward nominal andreal wage rigidity.

Downward nominal wage rigidity (DNWR) isdefined on the basis of nominal wage freezes.

Firms freezing nominal base wages at any pointduring the five-year period prior to the surveyare considered to be subject to nominal wagerigidity. Downward real wage rigidity (DRWR)is defined on the basis of wage indexation.Firms that have an automatic link betweennominal base wages and past or expected infla-tion are regarded as being subject to DRWR.The survey-based measures of downward nom-inal and real wage rigidity used in the paper areclosely related to the alternative measures

Downward nominal and real wage rigidity: survey evidence from European firms

Working Paper Νο. 110Jan Babecký, Philip Du Caju, Theodora Kosma, Martina Lawless, Julián Messina and Tairi Rõõm


derived by earlier studies on the basis of thewage change distributions.

The evidence indicates that the incidence ofboth types of wage rigidity is quite substantialin Europe – approximately 10% of firms expe-rienced wage freezes and 17% of firms appliedwage indexation mechanisms. Thus, indexation(DRWR) is much more prevalent in the EUcountries than wage freezes (DNWR). This isconsistent with other evidence on wage rigid-ity in most continental European countries, asopposed to the US and the UK. Overall, non-euro area countries of the EU are more likelyto experience wage freezes compared to euroarea countries, whereas indexation mecha-nisms are more widely used in the euro areacountries included in our sample.

Next, the paper analyses how DNWR andDRWR are related to a number of firm-leveland institutional characteristics of labour mar-kets in the countries covered by the sample.The multinomial logit estimation method isemployed, which makes it possible to assessthese relationships simultaneously for bothtypes of rigidities. The estimations indicatethat country-specific factors appear to be sig-nificant determinants of downward wage rigidi-ties, and that institutional differences betweencountries are an important factor behind thisfinding. For example, high collective bargain-ing coverage is positively related to real wagerigidity, while the estimated relationship withnominal wage rigidity is insignificant. A pos-sible interpretation of this finding is thatunions have the capacity to provide their mem-bers with information about inflation expec-tations and explain the importance of main-taining the real income level of workers. Thus,union coverage reduces the prevalence ofmoney illusion.

Analysis of the union contracts negotiated atdifferent levels (firm-level versus higher-levelbargaining contracts) implies that firm-levelcontracts are a more likely source of real wagerigidity in centralised wage-setting environ-ments. However, there is a substantial degreeof heterogeneity across countries regarding theimpact of different types of union contracts.The institutional aspects related to the diffi-culty of the employer to lay off workers are alsorelated to wage rigidity. The paper shows thatnominal wage rigidity is positively associatedwith the extent of permanent contracts. Inaddition, permanent contracts have a strongereffect on wage rigidity in countries with stricterlabour regulations.

Workforce composition also appears to play asignificant role in determining wage rigidities.Both types of rigidity are positively correlatedwith the share of high-skilled white-collarworkers; DNWR is positively related toemployees’ tenure in the firms under study.Both of these significant relationships are con-sistent with the implications of related theo-retical models. In addition, firms employinglabour-intensive technologies are more likelyto have rigid wages.

Finally, there seems to be a positive relation-ship between product market competition andDNWR, although the results are dependent onthe way competition is measured. A possiblecause of this empirical result is that in highlycompetitive industries rents should be low, andtherefore so should wages. This leavessmaller margins to reduce wages, because firmspaying low wages that are closer to a collec-tively agreed or legislative minimum level haveless flexibility than firms having a so-calledwage cushion between the minimum and theactual wage.

33Economic Bulletin

May 2010 107

Boutos Yannis, “Economic stabilisation and growth prospects”, No. 3, February 1994.Papademos Lucas, “Growth with stability: the role of monetary policy”, No. 5, March 1995.Voridis Hercules, “The special nature of banks and the transmission mechanism of monetarypolicy: A review of recent literature”, No. 5, March 1995.Filippides Anastasios, Kyriakopoulos Panayotis and Moschos Dimitrios, “Bank of Greece mon-etary policy instruments”, No. 6, November 1995.Haralabides Michael, Hardouvelis Gikas and Papageorgiou George, “Changeover to the singlecurrency: Prospects and challenges for credit institutions in Greece”, No. 6, November 1995.Karabalis Nikos, “Harmonisation of Consumer Price Indices in EU countries”, No. 7, March 1996.Public Sector Accounts Department, Studies, Planning and Operations Development Office,“Government Securities Markets”, No. 7, March 1996.Saccomanni Fabrizio, “Opportunities and challenges in the process of European monetary inte-gration”, No. 7, March 1996.Papademos Lucas, “Challenges facing monetary policy on the road to EMU”, No. 8, November1996.Information Systems and Organisation Department, “Developments in EU Payment Systems:TARGET and the Hellenic system Hermes”, No. 8, November 1996.Brissimis Sophocles and Gibson Heather, “Monetary policy, capital flows and Greek disinflation”,No. 9, March 1997.Sabethai Isaac, “From contractual earnings to labour costs: incomes policy, collective bargain-ing and inflation (1991-1996)”, No. 9, March 1997.Hall S.G. and Zonzilos Nicholas, “The output gap and inflation in Greece”, No. 9, March 1997.Pantelidis Evangelos, “The methodology of the new balance of payments compilation system ofthe Bank of Greece”, No. 9, March 1997.Papademos Lucas, “The globalisation of financial markets and the conduct of economic and mon-etary policy”, No. 10, December 1997.Brissimis Sophocles and Kastrisianakis Efstratios, “Is there a credit channel in the Greek econ-omy?”, No. 10, December 1997.Gibson Heather, “The relationship between the financial system and the real economy”, No. 10,December 1997.Hondroyiannis George and Papapetrou Evangelia, “The causal relationship between consumerand wholesale prices in Greece”, No. 10, December 1997.Hardy Daniel and Simigiannis George, “Competition and efficiency of the Greek banking sys-tem”, No. 11, June 1998.Brissimis Sophocles and Gibson Heather, “What can the yield curve tell us about inflation?”,No. 11, June 1998.Pantazidis Stelios, “Inflation, investment and economic growth in Greece”, No. 11, June 1998.Mitrakos Theodoros, “The contribution of income sources to overall inequality”, No. 11, June 1998.Garganas C. Nicholas, “Greece and EMU: prospects and challenges”, No. 12, December 1998.Brissimis Sophocles, Sideris Dimitris and Voumvaki Fragiska, “Purchasing power parity as a long-run relationship: an empirical investigation of the Greek case”, No. 12, December 1998.Manassaki Anna, “European Union transfers to Greece: historical background and prospects”,No. 12, December 1998.Lazaretou Sophia, “The drachma on the metallic monetary standards: lessons from the past”,No. 13, July 1999.Hondroyiannis George, “The causality between government spending and government revenuein Greece”, No. 13, July 1999.Mitrakos Theodoros and Tsakloglou Panos, “The distributional impact of excise duties”, No. 13,July 1999.

AR T I C L E S PUB L I S H ED I N PR EV I OU S I S S U E S O FTH E E CONOM I C BU L L E T I N

Dellas Harris, “ECB monetary policy strategy and macroeconomic volatility in Greece”, No. 13,July 1999.Papazoglou Christos, “The real exchange rate and economic activity: is the hard-drachma pol-icy necessarily contractionary?”, No. 14, December 1999.Gibson Heather and Lazaretou Sophia, “Leading inflation indicators for Greece: an evaluationof their predictive power”, No. 14, December 1999.Karapappas Andreas and Milionis Alexandros, “Estimation and analysis of external debt in theprivate sector”, No. 14, December 1999.Papademos Lucas, “From the drachma to the euro”, No. 15, July 2000.Zonzilos Nicholas, “The Greek Phillips curve and alternative measures of the NAIRU”, No. 15,July 2000.Pantazidis Stelios, “The sustainability of the current account deficit”, No. 15, July 2000.Hondroyiannis George, “Investigating causality between prices and wages inGreece”, No. 15, July 2000.Sabethai Isaac, “The Greek labour market: features, problems and policies (with emphasis onlabour market flexibility and on combatting unemployment in the context of non-inflationarygrowth)”, No. 16, December 2000.Brissimi Dimitra and Brissimis Sophocles, “The problem of unemployment in the EuropeanUnion”, No. 16, December 2000.Bardakas Joanna, “Financial deregulation, liquidity constraints and private consumption inGreece”, No. 16, December 2000.Lazaretou Sophia and Brissimis Sophocles, “Fiscal rules and stabilisation policy in the euro area”,No. 17, July 2001.Lolos Sarantis-Evangelos, “The role of European structural resources in the development of theGreek economy”, No. 17, July 2001.Papazoglou Christos, “Regional economic integration and inflows of foreign direct investment:the European experience”, No. 17, July 2001.Garganas C. Nicholas, “The future of Europe and EMU”, No. 18, December 2001.Brissimis Sophocles and Papadopoulou Dafni-Marina, “The physical introduction of euro ban-knotes and coins in Greece”, No. 18, December 2001.Zonzilos Nicholas, “The monetary transmission mechanism – the ECB experiment: results forGreece”, No. 18, December 2001.Brissimis Sophocles, Kamberoglou Nicos and Simigiannis George, “Bank-specific characteris-tics and their relevance to monetary policy transmission”, No. 18, December 2001.Tsaveas Nicholas, “Exchange rate regimes and exchange rate policy in South Eastern Europe”,No. 18, December 2001.Demenagas Nicholas and Gibson Heather, “Competition in Greek banking: an empirical studyfor the period 1993-1999”, No. 19, July 2002.Hondroyiannis George, “Economic and demographic determinants of private savings in Greece”,No. 19, July 2002.Gatzonas Efthymios and Nonika Kalliopi, “Eligible assets and management of collateral in thecontext of central bank monetary policy operations”, No. 19, July 2002.Voridis Hercules, Angelopoulou Eleni and Skotida Ifigeneia, “Monetary policy in Greece 1990-2000 through the publications of the Bank of Greece”, No. 20, January 2003.Kaplanoglou Georgia and Newbery David, “The distributional impact of indirect taxation inGreece”, No. 21, July 2003.Papapetrou Evangelia, “Wage differentials between the public and the private sector in Greece”,No. 21, July 2003.Athanasoglou Panayiotis and Brissimis Sophocles, “The effect of mergers and acquisitions onbank efficiency in Greece”, No. 22, January 2004.


Lazaretou Sophia, “Monetary system and macroeconomic policy in Greece: 1833-2003”, No. 22,January 2004.Karabalis Nikos and Kondelis Euripides, “Alternative measures of inflation”, No. 22, January 2004.Papaspyrou Theodoros, “EMU strategies for new Member States: the role of Exchange RateMechanism II”, No. 23, July 2004.Mitrakos Theodoros, “Education and economic inequalities”, No. 23, July 2004.Papapetrou Evangelia, “Gender wage differentials in Greece”, No. 23, July 2004.Gibson Heather, “Greek banking profitability: recent developments”, No. 24, January 2005.Athanasoglou Panayiotis, Asimakopoulos Ioannis and Georgiou Evangelia, “The effect of mergerand acquisition announcement on Greek bank stock returns”, No. 24, January 2005.Mitrakos Theodoros and Zografakis Stavros, “The redistributional impact of inflation in Greece”,No. 24, January 2005.Theodossiou Ioannis and Pouliakas Konstantinos, “Socio-economic differences in the job sat-isfaction of high-paid and low-paid workers in Greece”, No. 24, January 2005.Garganas C. Nicholas, “Adjusting to the single monetary policy of the ECB”, No. 25, August 2005.Mitrakos Theodoros, Simigiannis Georgios and Tzamourani Panagiota, “Indebtedness of Greekhouseholds: evidence from a survey”, No. 25, August 2005.Nicolitsas Daphne, “Per capita income, productivity and labour market participation: recent devel-opments in Greece”, No. 25, August 2005.Nicolitsas Daphne, “Female labour force participation in Greece: developments and determin-ing factors”, No. 26, January 2006.Mitrakos Theodoros and Zonzilos Nicholas, “The impact of exogenous shocks on the dynamicsand persistence of inflation: a macroeconomic model-based approach for Greece”, No. 26, Jan-uary 2006.Athanasoglou Panayiotis, Asimakopoulos Ioannis and Siriopoulos Konstantinos, “External financ-ing, growth and capital structure of the firms listed on the Athens Exchange”, No. 26, January2006.Mitrakos Theodoros and Nicolitsas Daphne, “Long-term unemployment in Greece: developments,incidence and composition”, No. 27, July 2006.Kalfaoglou Faidon, “Stress testing of the Greek banking system”, No. 27, July 2006.Pantelidis Th. Evangelos and Kouvatseas A. Georgios, “Frontier survey on travel expenditure:methodology, presentation and output assessment (2003-2005)”, No. 27, July 2006.Brissimis N. Sophocles and Vlassopoulos Thomas, “Determinants of bank interest rates and com-parisons between Greece and the euro area”, No. 28, February 2007.Simigiannis T. George and Tzamourani G. Panagiota, “Borrowing and socio-economic charac-teristics of households: results of sample surveys carried out by the Bank of Greece”, No. 28,February 2007.Papapetrou Evangelia, “Education, labour market and wage differentials in Greece”, No. 28, Feb-ruary 2007.Christodoulakis A. George, “The evolution of credit risk: phenomena, methods and management”,No. 28, February 2007.Karabalis P. Nikos and Kondelis K. Euripides, “Inflation measurement in Greece”, No. 28, Feb-ruary 2007.Kapopoulos Panayotis and Lazaretou Sophia, “Foreign bank presence: the experience of South-East European countries during the transition process”, No. 29, October 2007.Nicolitsas Daphne, “Youth participation in the Greek labour market: developments and obsta-cles”, No. 29, October 2007.Albani K. Maria, Zonzilos G. Nicholas and Bragoudakis G. Zacharias, “An operational frame-work for the short-term forecasting of inflation”, No. 29, October 2007.

33Economic Bulletin

May 2010 109

Asimakopoulos G. Ioannis, Brissimis N. Sophocles and Delis D. Manthos, “The efficiency of theGreek banking system and its determinants”, No. 30, May 2008.Balfoussia Hiona, “Stock market integration: the Athens Exchange in the European financialmarket”, No. 30, May 2008.Mitrakos Theodoros, “Child poverty: recent developments and determinants”, No. 30, May 2008.Asimakopoulos Ioannis, Lalountas Dionysis and Siriopoulos Constantinos, “The determinantsof firm survival in the Athens Exchange”, No. 31, November 2008.Petroulas Pavlos, “Foreign direct investment in Greece: Productivity and Technology diffusion”,No. 31, November 2008.Anastasatos Tasos and Manou Constantina, “Speculative attacks on the drachma and thechangeover to the euro”, No. 31, November 2008.Mitrakos Theodoros and Simigiannis George, “The determinants of Greek household indebt-edness and financial stress”, No. 32, May 2009.Papazoglou Christos, “Is Greece’s export performance really low?”, No. 32, May 2009.Zonzilos Nicholas, Bragoudakis Zacharias and Pavlou Georgia, “An analysis of the revisions offirst (flash) quarterly national account data releases for Greece”, No. 32, May 2009.


33Economic Bulletin

May 2010 111

ISSN: 1105 - 9729

economic bulletin - core.ac.uk · bank of greece 21, e. venizelos avenue 102 50 athens economic...

Documents