Assessing public debt sustainability in EU member states:a guide

Transcription

1 MPRA Munch Personal RePEc Archve Assessng publc debt sustanablty n EU member states:a gude Guseppe Carone and Kata Bert European Commsson - DG ECFIN 9. September 2014 Onlne at MPRA Paper No , posted 6. March :56 UTC

2 ISSN (onlne) ISSN (prnt) EUROPEAN ECONOMY Occasonal Papers 200 September 2014 Assessng Publc Debt Sustanablty n EU Member States: A Gude Economc and Fnancal Affars

3 Occasonal Papers are wrtten by the staff of the Drectorate-General for Economc and Fnancal Affars, or by experts workng n assocaton wth them. The Papers are ntended to ncrease awareness of the techncal work beng done by staff and cover a wde spectrum of subjects. Vews expressed n unoffcal documents do not necessarly reflect the offcal vews of the European Commsson. Comments and enqures should be addressed to: European Commsson Drectorate-General for Economc and Fnancal Affars Unt Communcaton and nter-nsttutonal relatons B-1049 Brussels Belgum E-mal: ecfn-nfo@ec.europa.eu LEGAL NOTICE Nether the European Commsson nor any person actng on ts behalf may be held responsble for the use whch may be made of the nformaton contaned n ths publcaton, or for any errors whch, despte careful preparaton and checkng, may appear. Ths paper exsts n Englsh only and can be downloaded from More nformaton on the European Unon s avalable on KC-AH EN-N (onlne) ISBN (onlne) do: /88866 (onlne) KC-AH EN-C (prnt) ISBN (prnt) do: /88909 (prnt) European Unon, 2014 Reproducton s authorsed provded the source s acknowledged.

4 European Commsson Drectorate-General for Economc and Fnancal Affars Assessng Publc Debt Sustanablty n EU Member States: A Gude EUROPEAN ECONOMY Occasonal Papers 200

5 AKNOWLEDGEMENTS Ths paper was prepared n the Drectorate-General for Economc and Fnancal Affars under the drecton of Marco But, Drector-General, Servaas Deroose, Deputy Drector-General, and Luco Pench, Drector for Fscal Polcy. The paper was prepared by Guseppe Carone and Kata Bert, wth the statstcal assstance of Etenne Sal and Eugenu Colesnc. The Debt Sustanablty Montor model used n the publc debt sustanablty analyss framework presented n the paper s the product of jont work n ECFIN.C2 (Guseppe Carone, Per Eckefeldt, Kata Bert, Vel Lane, Etenne Sal). SYMBOL smulatons used n the analyss of governments' contngent lablty rsks from the bankng sector are provded by the European Commsson's Jont Research Centre (Insttute for the Protecton and Securty of the Ctzen, Scentfc Support to Fnancal Analyss). Comments on the paper would be gratefully receved and should be sent to: Guseppe Carone European Commsson Drectorate-General for Economc and Fnancal Affars Drectorate for Fscal Polcy Unt C2: Sustanablty of Publc Fnances Offce CHAR 12/048 B-1049 Brussels e-mal: guseppe.carone@ec.europa.eu or Kata Bert European Commsson Drectorate-General for Economc and Fnancal Affars Drectorate for Fscal Polcy Unt C2: Sustanablty of Publc Fnances Offce CHAR 12/129 B-1049 Brussels e-mal: kata.bert@ec.europa.eu 2

6 TABLE OF CONTENTS 1. Introducton 6 2. Crtera used to dentfy "vulnerable" countres for "enhanced DSA" 7 3. The European Commsson's DSA framework: toolkt used Determnstc publc debt projectons Senstvty analyss around determnstc publc debt projectons Stochastc publc debt projectons The analyss of rsks related to the structure of publc debt fnancng Fnancal market nformaton The analyss of rsks related to government's contngent labltes Forecast accuracy analyss 24 Annex 1: Sample country fche for DSA 27 Annex 2: Interest rates on publc debt n the European Commsson - DG ECFIN's debt projecton model 31 Annex 3: The sgnals' approach for threshold determnaton: varables of publc debt structure, bankng sector vulnerabltes and yeld spreads 34 Annex 4: SYMBOL 36 Annex 5: Stochastc publc debt projectons based on the hstorcal varance-covarance matrx approach 39 3

7 LIST OF TABLES 1. Gross publc debt projectons (% of GDP) and underlyng macro-fscal assumptons, sample country Baselne no-fscal polcy change scenaro Gross publc debt (% of GDP) from stochastc debt projectons, sample country Dstrbuton percentles Heat map of rsks related to the structure of publc debt fnancng, sample country Fnancal market ndcators, sample country Soveregn ratngs, sample country Government's contngent labltes, sample country Heat map on government's contngent lablty rsks from the bankng sector, sample country 23 A3.1. Possble cases based on type of sgnal sent by the varable at t-1 and states of the world at t 34 A3.2. Thresholds, sgnallng power, type I and type II errors obtaned from sgnals' approach 35 LIST OF GRAPHS 1. European Commsson's (DG ECFIN) DSA framework 8 2. Gross publc debt projectons (% of GDP), sample country Baselne no-fscal polcy change and hstorcal scenaros Structural prmary balance (average and forecasted reference values) for sample country aganst probablty dstrbuton (all EU countres, ) of 3-year average structural prmary balance Change n structural prmary balance (average and forecasted reference values) for sample country aganst probablty dstrbuton (all EU countres, ) of 3-year cumulatve change n structural prmary balance Determnants of changes n gross publc debt (% of GDP), sample country Baselne no-fscal polcy change scenaro Evoluton of the maturty structure of gross publc debt (% of GDP), sample country Baselne no-fscal polcy change scenaro Gross publc debt projectons (% of GDP), sample country Baselne no-fscal polcy change, SGP nsttutonal and SCP scenaros 14 4

8 8. Gross publc debt projectons (% of GDP), sample country Senstvty tests on nterest rates, real GDP growth, nflaton, prmary balance and nomnal exchange rate around baselne no-fscal polcy change scenaro Gross publc debt (% of GDP) from stochastc debt projectons, sample country Fan chart Crtera to assess the country's vulnerablty to bankng contngent lablty rsks for the government Forecast errors on prmary balance (% of GDP) for sample country aganst EU dstrbuton of forecast errors on prmary balance (% of GDP) Forecast errors on real GDP growth for sample country aganst EU dstrbuton of forecast errors on real GDP growth Forecast errors on nflaton rate for sample country aganst EU dstrbuton of forecast errors on nflaton rate 26 A2.1. A graphcal representaton of nterest rates appled n the European Commsson DG ECFIN's debt projecton model, under the smplfyng assumpton of no change n debt 33 LIST OF BOXES 1. Debt projecton scenaros 11 5

9 1. INTRODUCTION The am of ths paper s to llustrate the methodologcal approach used by the Commsson servces (DG ECFIN/C2) to carry out, n a systematc and harmonsed way, publc debt sustanablty analyss (henceforth DSA) for EU Member States. Analysng recent and prospectve publc debt developments and rsks to debt sustanablty s crucal for EA countres and the EU as a whole to be able to formulate approprate polcy responses. To ths am, the Commsson servces (DG ECFIN) prepare on a regular bass (twce a year, followng autumn and sprng Commsson forecasts) an nternal "Debt Sustanablty Montor" report (DSM) presentng, for each Member State, a detaled publc debt sustanablty analyss, accompaned by the analyss of fscal sustanablty ndcators. 1 The DSM provdes key nformaton for regular budgetary survellance. The assessment of Member States' debt developments s ndeed a key component of fscal survellance under the Stablty and Growth Pact (SGP), the European semester and the Europe 2020 strategy. The Commsson servces' (DG ECFIN) approach to DSA results from the contnuous effort to develop a DSA framework that s n lne wth most recent methodologcal developments and practce n other nternatonal organsatons (IMF, ECB, OECD). 2 Man features of the Commsson's DSA framework are the followng: 1) Crtera are used to dentfy "vulnerable" countres from the pont of vew of publc debt sustanablty. For the latter, the DSA s "enhanced" wth a detaled wrte-up, n whch the macro-fscal assumptons used n the projectons are llustrated and debt projecton results, and rsks to debt sustanablty more broadly, are dscussed. 2) The framework s desgned n a way to allow for a comprehensve assessment of rsks to publc debt sustanablty. Senstvty analyss around baselne publc debt projectons, for nstance, s extensve, coverng downsde and upsde rsks to the man macro-fscal determnants of debt dynamcs (possbly emergng from fscal fatgue, tghtenng/relaxng of governments' fnancng condtons on the markets, shocks to GDP growth, nflaton and the exchange rate, bank-related contngent lablty shocks). 3) Varables capturng rsks potentally arsng from the structure of publc debt (publc debt by maturty, holder, currency of denomnaton) are ntegrated n the DSA through heat maps, thus usefully complementng the analyss of rsks related to the projected publc debt dynamcs. 4) The analyss of governments' contngent labltes features promnently n the DSA framework. An overvew of overall contngent labltes for the publc sector s provded based on most recent (Eurostat) data on state guarantees. Contngent lablty rsks arsng from the bankng sector are captured ndrectly through heat maps of varables that measure bankng sector vulnerabltes, as well as through model estmates of the theoretcal probablty of sgnfcant bank losses httng publc fnances n a smulated bank crss. 3 Publc debt projectons are addtonally run under a specfc bankng contngent lablty The fscal sustanablty analyss s based on the S0, S1 and S2 ndcators, respectvely capturng short-, medum- and long-term fscal sustanablty challenges. For more detals, see European Commsson (2012) "Fscal Sustanablty Report 2012" European Economy 8/2012. Recent mprovements to the Commsson servces' (DG ECFIN) DSA framework have been partly nspred by mportant methodologcal changes recently ntroduced by the IMF n ts own DSA framework. For a presentaton of the latter, see IMF (2013) "Staff Gudance Note for publc debt sustanablty analyss n market-access countres", 9 May Smulaton results are obtaned from SYMBOL (SYstemc Model of Bankng Orgnated Losses), a model that has been developed jontly by the European Commsson DG JRC, DG MARKT and academc experts. The model allows estmatng aggregate bankng losses that derve from bank defaults, accountng for banks' captal and the exstence of bankng safety net tools. For further methodologcal detals, see De Lsa R., S. Zedda, F. Vallascas, F. Campolongo and M. Marches (2011) "Modelng depost nsurance scheme losses n a Basel II framework" Journal of Fnancal Servces Research 40(3). For an applcaton of the model to the analyss of governments' contngent labltes from the bankng sector, see European Commsson (2012) "Fscal Sustanablty Report 2012" European Economy 8/2012, Secton A short explanaton on the SYMBOL model s also provded n Annex 4. 6

10 shock scenaro, f bankng contngent lablty rsks are hghlghted by the aforementoned tools. 4 5) Commsson forecast accuracy analyss on the man macro-fscal determnants of publc debt dynamcs (real GDP growth, prmary balance and nflaton) s ncluded n the DSA. 5 Ths analyss ams at provdng some ndcaton on whether forecasts, ncorporated n baselne publc debt projectons, tend to be systematcally based n one drecton or the other n a sgn of persstent optmsm or pessmsm. The paper s structured as follows. Secton 2 descrbes the crtera used to dentfy "vulnerable" countres for whch a detaled DSA wrte-up s requred by the European Commsson's (DG ECFIN) framework. Secton 3 provdes an accurate descrpton of the framework, and all the analytcal and reportng tools t encompasses. 2. CRITERIA USED TO IDENTIFY "VULNERABLE" COUNTRIES FOR "ENHANCED DSA" In the European Commsson's (DG ECFIN) DSA framework, a set of objectve crtera, based on selected varables/ndcators, s systematcally appled to all EU countres to establsh the degree of vulnerablty of the country under examnaton from the pont of vew of rsks to publc debt sustanablty. When, through ths frst screenng, a country s found to be "vulnerable", ts DSA (labelled at ths pont as "enhanced DSA") s ntegrated wth a detaled wrte-up, where macro-fscal assumptons used n the projectons are dscussed, as are the rsks to publc debt sustanablty emergng from the analyss. Addtonal ad-hoc senstvty tests around baselne publc debt projectons may be run for vulnerable countres as part of ths enhanced DSA, on top of the wde range of senstvty tests already ncluded by default n the standard DSA. EU countres are subject to an enhanced DSA, requrng a DSA wrte-up and n case ncludng addtonal customzed senstvty tests as explaned above, whenever one or more of the followng condtons hold true (see also Graph 1): 1) the country has a value of the composte ndcator of short-term fscal stress rsk, S0, above the crtcal threshold, and/or a value of the S0 fscal sub-ndex above threshold; 6 2) the country's current and/or forecasted gross publc debt 7 s at, or hgher than, 90% of GDP; 8 3) the country's current and/or forecasted change n gross publc debt over GDP s at, or hgher than, 5 p.p.; 4) the country's gross fnancng needs are at, or hgher than, 15% of GDP; For more detals see Benczur P., K. Bert, G. Cannas, J. Carbon, S. Langedjk, A. Pagano and M. Petracco (2014) "A bankng contngent lablty stress-test scenaro for publc debt projectons usng the SYMBOL model" European Economy Economc Paper, forthcomng. For detals see Gonzalez Cabanllas L. and A. Terz (2012) "The accuracy of the European Commsson's forecasts reexamned" European Economy Economc Paper No The S0 ndcator of short-term fscal stress rsk s a composte ndcator constructed usng 14 fscal varables and 14 macro-fnancal varables that are found to be good predctors of fscal stress. Thresholds of fscal rsk for the S0 ndcator, ts sub-ndexes gatherng groups of homogeneous varables (fscal and macro-fnancal varables respectvely) and the ndvdual varables ncorporated n the composte ndcator are calculated usng the non-parametrc sgnals' approach. Values of the S0 ndcator above the threshold sgnal rsks of fscal stress n the year ahead. For more detals on S0, see Bert K., M. Salto and M. Lequen (2012) "An early-detecton ndex of fscal stress for EU countres" European Economy Economc Paper No On the sgnals' approach, see Kamnsky G.L., S. Lzondo and C.M. Renhart (1998) "Leadng ndcators of currency crses" IMF Staff Papers Vol. 45, No. 1 and Kamnsky G.L. and C.M. Renhart (1999) "The twn crses: the causes of bankng and balance-of-payments problems" Amercan Economc Revew vol. 89(3), pp A short explanaton on the sgnals' approach s also provded n Annex 3. Here the reference s to general government consoldated gross debt ("Maastrcht debt"). Despte the threshold for enhanced DSA beng set at 90% of GDP, consderaton s clearly also gven n the DSA to whether publc debt s below or above the Treaty reference value of 60% of GDP. 7

11 5) the country s under a macroeconomc adjustment programme, under post-programme survellance or enhanced survellance as from the Two-Pack regulaton. 9 The thresholds ndcated above for the change n gross publc debt and gross fnancng needs have been obtaned by lowerng, for prudental reasons, the crtcal thresholds of fscal rsk derved wth the sgnals' approach. For gross publc debt, both the level and the change are consdered as useful crtera to establsh the need for an enhanced DSA. In the context of the latest economc and fnancal crss, ths would have allowed snglng out some crtcal cases where publc debt evoluton dsplayed worryng trends, though startng from relatvely low levels. Whle ndvdual varables ncluded n the set of crtera above focus exclusvely on publc fnances, the ncluson of the S0 ndcator ensures that also fscal rsks stemmng from the compettveness and fnancal sdes of the economy (and that are such to put the country at overall short-term rsk of fscal stress, as ndcated by a value of the S0 ndcator above the threshold) lead to the requrement of an enhanced DSA wth detaled wrte-up of rsks. Graph 1: European Commsson's (DG ECFIN) DSA framework Are S0 ndcator and/or S0 fscal sub-ndex above threshold? Is the current and/or forecasted gross publc debt at/hgher than 90% of GDP? Is the current and/or forecasted change n gross publc debt over GDP at/hgher than 5 p.p.? Are gross fnancng needs at/hgher than 15% of GDP? Is the country under a macroeconomc adjustment programme, under post-programme survellance or enhanced survellance? None of the above holds Any of the above holds DSA relyng on followng tools: 1. Determnstc publc debt projectons 2. Senstvty analyss around baselne publc debt projectons (on nterest rates, GDP growth, nflaton, prmary balance, exchange rate) 3. Stochastc publc debt projectons 4. Analyss of rsks related to the structure of publc debt fnancng 5. Analyss of rsks related to government's contngent labltes 6. Fnancal market nformaton 7. Forecast accuracy analyss Enhanced DSA ntegratng the standard DSA wth: 1. Customzed senstvty tests around baselne publc debt projectons 2. DSA wrte-up Regulaton (EU) No. 472/2013 of the European Parlament and the Councl of 21 May 2013 on the strengthenng of economc and budgetary survellance of Member States n the euro area experencng or threatened wth serous dffcultes wth respect to ther fnancal stablty. The logc behnd the calculaton of thresholds based on the sgnals' approach rests on the observaton that economes behave n a systematcally dfferent way n perods precedng fscal stress. Accordng to ths, tme seres of the varables for whch thresholds are to be determned, and the seres of fscal-stress epsodes recorded n the past are used together to determne an optmal fscal rsk threshold for the varable n queston, based on ts past behavour ahead of fscal stress epsodes. Such optmal threshold s determned by maxmsng the "sgnallng power" of the model,.e. ts ablty to correctly predct past fscal stress. By frst dstngushng between the two types of errors that can be made n such a predcton (.e predctng fscal stress, for a varable value beyond the threshold, ahead of no fscal stress epsode and predctng no fscal stress, for a varable value on the safe sde of the threshold, ahead of a fscal stress epsode), the optmal threshold s then determned n a way to mnmse the share of mssed (n the sense of not sgnalled) stress epsodes plus the share of non-fscal-stress epsodes wrongly sgnalled as upcomng fscal stress. A short explanaton on the sgnals' approach s also provded n Annex 3. Crtcal thresholds of fscal rsk, as obtaned through the sgnals' approach, are: 6.5 p.p. for the change n gross publc debt over GDP and 16.83% of GDP for gross fnancng needs. See Bert et al. (2012). 8

12 3. THE EUROPEAN COMMISSION'S DSA FRAMEWORK: TOOLKIT USED Ths secton descrbes n detal the way the DSA s conducted by the European Commsson servces (DG ECFIN). Apart for provdng an overvew of what are the tools used, the dfferent scenaros, senstvty and stress tests run, the objectve s to provde a clear pcture of how all these dfferent elements ft together n the DSA (see Annex 1 for the format of a sample DSA country fche dsplayng results for all tools) DETERMINISTIC PUBLIC DEBT PROJECTIONS The Commsson's DSA reles on both determnstc and stochastc publc debt projectons. Tradtonal determnstc projectons comprse a whole set of scenaros, respectvely based on Commsson's and Member States (Stablty and Convergence Programmes) forecasts, no-fscal polcy change and fscal consoldaton assumptons beyond forecasts. As wll become clearer from the explanatons that follow, these debt projecton scenaros are desgned so as to complement each other n terms of nformaton they convey on possble future debt trajectores. They are therefore conceved to be used n an ntegrated way to make assessments on publc debt sustanablty. Debt projectons run by the European Commsson are presented over a 10-year horzon ( at the tme of wrtng ths paper). Ths s deemed to be a good compromse between the need to keep publc debt projectons referred to a tme nterval that s not too long (as uncertanty naturally rses, the further projectons move nto the future), nor too short (thus allowng for a meanngful analyss of the mpact of projected age-related mplct labltes). The determnstc debt projecton scenaros used n the Commsson's framework are as follows (see Box 1 for a summary vew): 1) A baselne no-fscal polcy change scenaro, relyng on Commsson forecasts, the Economc Polcy Commttee (EPC) agreed long-run convergence assumptons of underlyng macroeconomc varables (real nterest rate, real GDP growth, nflaton rate) 12 and the assumpton of constant fscal polcy (.e. constant structural prmary balance, SPB, at last forecast value) beyond the forecast horzon. The cyclcal component of the balance s calculated usng standard country-specfc sem-elastcty parameters, 13 and the stock-flow adjustment s set to zero beyond forecasts. Ths scenaro ncorporates mplct labltes related to ageng (projected pensons, healthcare and long-term care expendture). 14 2) A no-fscal polcy change scenaro wthout ageng costs, whch dffers from the baselne nofscal polcy change scenaro above only for the excluson of age-related mplct labltes. 3) Hstorcal scenaros (whch ncorporate age-related costs) consstng of:. A hstorcal SPB scenaro, relyng on Commsson forecasts and the assumpton of gradual (3-year) convergence of the SPB to last 10-year hstorcal average beyond the forecast horzon, whle all other macroeconomc assumptons reman as n baselne scenaro (1) For GDP growth projectons agreed wth the Economc Polcy Commttee-Output Gap Workng Group are used. For the nflaton rate (GDP deflator) and the real long-term nterest rate, the long-run convergence assumptons agreed wth the Economc Polcy Commttee are used. The nflaton rate (GDP deflator) s therefore assumed to converge lnearly to 2% n the year of output gap closure (T+5) and reman constant at that value thereafter. The real long-term nterest rate s assumed to converge lnearly to 3% by the end of the projecton horzon (10 years' tme). Annex 2 provdes a more detaled analyss of how nterest rates enter the debt projecton model. Estmated sem-elastcty parameters are those endorsed by the Economc Polcy Commttee Output Gap Workng Group. These are based on European Commsson-Economc Polcy Commttee long-run projectons of age-related costs. See European Commsson (2012) "The 2012 Ageng Report. Economc and budgetary projectons for the 27 EU Member States ( )" European Economy 2/

13 . A combned hstorcal scenaro, relyng on Commsson forecasts and the assumpton of gradual (3-year) convergence of the man underlyng macroeconomc varables (SPB, nterest rate, real GDP growth) to last 10-year hstorcal averages beyond the forecast horzon. 4) A Stablty and Growth Pact (SGP) nsttutonal scenaro, where for countres under excessve defct procedure (EDP) a structural adjustment path n complance wth the fscal effort recommended by the Councl s mantaned untl the excessve defct s corrected, and thereafter an annual structural consoldaton effort of 0.5 p.p. of GDP (or 0.6 p.p. f publc debt exceeds 60% of GDP) s mantaned untl the medum-term objectve (MTO) s reached. For the other countres, the consoldaton effort to reach the MTO s centred on an annual mprovement n the SPB by 0.5/0.6 p.p. of GDP as of Ths scenaro accounts for a feedback effect of fscal consoldaton on GDP growth (a 1 p.p. consoldaton effort reducng baselne GDP growth by 0.5 p.p. n the same year). 15 Age-related costs are ncorporated n ths SGP nsttutonal scenaro. 5) A Stablty and Convergence Programme (SCP) scenaro, relyng on SCPs' macro-fscal assumptons over the programme horzon and constant fscal polcy assumpton (constant SPB at last programme year) beyond the programme horzon. The scenaros lsted above usefully complement each other n the context of country-specfc DSAs. The comparson between debt projecton results obtaned under the baselne no-fscal polcy change scenaro (1) and those obtaned under the no-fscal polcy change scenaro wthout ageng costs (2) makes t possble, for nstance, to assess the mpact of projected government's mplct labltes related to ageng on publc debt dynamcs. Hstorcal scenaros (3) provde a stress test on the long-run convergence assumptons of macroeconomc varables (structural prmary balance, nterest rate and real GDP growth) made under the baselne no-fscal polcy change scenaro. The comparson between the baselne no-fscal polcy change and the hstorcal SPB scenaros, for nstance, shows the dfference n debt dynamcs, f the structural prmary balance gradually reverted to hstorcal average after the forecasts rather than remanng constant at last forecast year (based on the defnton of nofscal polcy change). The SGP nsttutonal scenaro (4) shows the evoluton of the debt-to-gdp rato under the assumpton of fscal polcy changes over the projecton horzon, n a way to fully reflect complance wth fscal rules (EDP recommendatons; MTO convergence). The comparson wth the baselne no-fscal polcy change scenaro allows capturng the effect of fscal consoldaton (durng and beyond the forecast horzon) n lne wth fscal rules relatve to a baselne scenaro that prudentally assumes fscal polcy constant at last forecast year. Fnally, the comparson between the SCP scenaro (5) and the baselne no-fscal polcy change scenaro (1) s llustratve of the dfferences arsng by usng Member States' versus Commsson's forecasts (n both cases under a scenaro based on the no-fscal polcy change assumpton). Debt projecton results for the baselne no-fscal polcy change scenaro are presented graphcally together wth those obtaned for the no-fscal polcy change scenaro wthout ageng costs, the hstorcal SPB scenaro, and the combned hstorcal scenaro (see sample country n Graph 2). As antcpated above, the hstorcal SPB scenaro mportantly allows snglng out the role played by the no-fscal polcy change assumpton n the baselne scenaro. In the latter, the SPB s set constant at last forecast year beyond the forecast horzon, as the standard and smplest way to deal wth the fact that fscal polcy developments are unknown thereafter. On the other hand, for countres for whch the SPB s forecasted to take an unusually low/hgh value (by hstorcal standards) n the last forecast year, the assumpton that the SPB remans constant at such value also n followng years tll the end of the projecton horzon mght turn out too restrctve. Debt projecton results under no-fscal polcy change and hstorcal SPB scenaros are therefore looked at jontly n the DSA, to be able to gauge the mpact on projected debt dynamcs, were the SPB to revert to hstorcal mean beyond forecasts. Clearly, the jont analyss of results obtaned for the baselne no-fscal polcy change scenaro and the 15 Over the forecast years ( at the tme of wrtng ths paper), the feedback effect of fscal consoldaton on GDP growth apples to the dfference between the forecasted fscal effort (change n the structural balance) and the assumed fscal effort (EDP structural adjustment path or benchmark fscal effort of 0.5/0.6 p.p. of GDP). Ths s done to avod any "double-countng" as feedback effects of fscal consoldaton on growth are already featured n the forecasts over the two forecast years. 10

14 hstorcal SPB scenaro s the more mportant for countres for whch the last forecast year SPB les n the tals of the dstrbuton of the (3-year) average SPB over all EU countres n the last 15 years (hghlghtng an exceptonally low/hgh last forecast year SPB for the country under examnaton). For ths reason, the aforementoned dstrbuton s provded as complementary nformaton to debt projecton results, together wth the dstrbuton of the 3-year SPB change, from whch t can be seen whether the cumulated structural fscal effort for the country under examnaton appears to be "atypcal" or not (see sample country n Graphs 3-4). Box 1: DEBT PROJECTION SCENARIOS The debt projecton scenaros ncluded n the European Commsson's (DG ECFIN) Debt Sustanablty Montor report are the followng: 1. Baselne no-fscal polcy change scenaro (European Commsson forecasts; assumpton of unchanged fscal polcy after forecasts; Economc Polcy Commttee-agreed long-run convergence assumptons of underlyng macroeconomc varables) 2. No-fscal polcy change scenaro wthout age-related costs (same as scenaro (1) but wthout ageng costs) 3. Hstorcal scenaros (European Commsson forecasts; assumpton of gradual convergence of structural prmary balance, nterest rate, real GDP growth one at the tme and then all together to hstorcal average(s) after forecasts) 4. Stablty and Growth Pact (SGP) nsttutonal scenaro (full complance wth excessve defct procedure, EDP, recommendatons and convergence to the medum-term objectve, MTO) 5. Stablty and Convergence Programme (SCP) scenaro (SCP assumptons for man macro-fscal varables; assumpton of unchanged fscal polcy after programme horzon) Senstvty test scenaros run around the baselne no-fscal polcy change scenaro are the followng: 1. "Standard" senstvty tests on short- and long-term nterest rates (-1p.p./+1p.p. on short- and longterm nterest rates on new and rolled over debt over whole 10-year projecton perod) 2. "Enhanced" senstvty tests on short- and long-term nterest rates (-1p.p./+2p.p. on short- and longterm nterest rates on new and rolled over debt for frst 3 projecton years, followed by -1p.p./+1p.p. over remanng of projecton perod) 3. "Standard" senstvty tests on real GDP growth (-0.5/+0.5 p.p. on real GDP growth over whole 10- year projecton perod) 4. "Enhanced" senstvty tests on real GDP growth (-1 standard devaton/+1 standard devaton on real GDP growth for frst 2 projecton years, followed by -0.5/+0.5 p.p. over remanng of projecton perod) 5. Senstvty tests on nflaton (-0.5/+0.5 p.p. on nflaton rate over whole projecton perod) 6. Senstvty test on prmary balance (negatve shock to prmary balance equal to 50% of forecasted cumulatve change over the 2 forecast year; prmary balance kept constant at lower last forecast year level over remanng of projecton perod) 7. Senstvty test on nomnal exchange rate (shock equal to maxmum hstorcal change n the exchange rate, over last 10 years, appled for frst 2 projecton years) Projecton results for the baselne no-fscal polcy change scenaro are also presented n more detal n a standard table (see Table 1). To facltate the readng of results, the determnants of changes n the debt rato under the baselne s also represented graphcally (as llustrated n Graph 5), as t s the evoluton of the debt maturty structure over the projecton horzon (Graph 6). 11

15 Graph 2: scenaros Gross publc debt projectons (% of GDP), sample country Baselne no-fscal polcy change and hstorcal Source: Commsson servces Graph 3: Structural prmary balance (average and forecasted reference values) for sample country aganst probablty dstrbuton (all EU countres, ) of 3-year average structural prmary balance Source: Commsson servces Graph 4: Change n structural prmary balance (average and forecasted reference values) for sample country aganst probablty dstrbuton (all EU countres, ) of 3-year cumulatve change n structural prmary balance Source: Commsson servces 12

16 Table 1: Gross publc debt projectons (% of GDP) and underlyng macro-fscal assumptons, sample country Baselne no-fscal polcy change scenaro Gross debt rato Changes n the rato of whch Outstandng debt Rolled-over short-term debt Rolled-over long-term debt New short-term debt New long-term debt (1) Prmary balance (+ = defct) Prmary balance n structural terms Cyclcal component Cost of ageng Property ncomes (2) Snowball effect Interest expendture Growth effect Inflaton effect (3) Stock flow adjustment and one-off measures Key macroeconomc assumptons Actual GDP grow th (real) Potental GDP grow th (real) Implct nterest rate (nomnal) Inflaton (GDP deflator) Source: Commsson servces Graph 5: Determnants of changes n gross publc debt (% of GDP), sample country Baselne no-fscal polcy change scenaro Source: Commsson servces Graph 6: Evoluton of the maturty structure of gross publc debt (% of GDP), sample country Baselne nofscal polcy change scenaro Source: Commsson servces Notes: Short-term and long-term publc debt are defned as general gov't debt wth maturty below and above the year respectvely. 13

17 For our sample country, Graph 2, for nstance, shows that gross publc debt over GDP n the no-fscal polcy change scenaro leads to a lower projected debt trajectory compared to the scenaro n whch reverson to the hstorcal average SPB s assumed (due to a hgher last forecast year SPB compared to the last 10-year hstorcal average). Ths s to say that, for the country under examnaton, f fscal fatgue were to set n and reduce projected fscal consoldaton, by gradually realgnng the projected fscal stance to what observed on average for the country over the last 10 years, the projected debt rato would ncrease as shown n Graph 2. If also the nterest rate and real GDP growth converged to hstorcal averages, debt dynamcs under the combned hstorcal scenaro would further worsen. Implct labltes related to ageng do have a sgnfcant negatve mpact on the projected evoluton of ths sample country's debt rato (a debt rato that s around 5 p.p. hgher n 2024 n the baselne scenaro wth ageng costs relatve to the scenaro wthout see Graph 2 and Table 1). In terms of assessng the "degree of realsm" of the baselne no-fscal polcy change assumpton, from the plot of the dstrbuton of the 3-year average SPB over all EU countres n the last 15 years n Graph 2, t can be seen that the 2015 forecasted SPB for the sample country s broadly n lne wth the average SPB for the country (percentle ranks of 34% and 39% respectvely, as from Graph 2) and s not atypcal relatve to 3-year average SPBs recorded n the EU (n Graph 2, the whte crcle and the red rhombus ndcate respectvely the postons of the average SPB and 2015 forecasted SPB for the country, whch are close to each other, and do not le n the tal of the dstrbuton). Ths ponts to a possbly hgh "degree of realsm" of the no-fscal polcy change assumpton beyond forecasts for the country under examnaton. Projecton results under the baselne no-fscal polcy change scenaro are also plotted aganst the SGP nsttutonal scenaro n a separate chart (Graph 7). Ths makes t possble to assess how debt dynamcs would change by lftng the no-fscal polcy change assumpton after forecast horzon and assumng fscal efforts put n place by the Member State accordng to EDP recommendatons and convergence to the MTO (takng account of feedback effects from addtonal fscal consoldaton on growth). The sgnfcance of the fscal effort requred to put the debt rato on the more decsve downward path of the nsttutonal SGP scenaro dsplayed n Graph 7 can be grasped by lookng at where the mpled fscal adjustment les n the overall dstrbuton of cumulatve SPB changes over all EU countres (the trangle n Graph 4). The percentle rank tells us that n less than one thrd of the cases, over all EU countres n the last 15 years, cumulatve (3-year) fscal adjustments have been greater than that mpled by the SGP scenaro for the sample country. The fscal adjustment (cumulatve change n the SPB) forecasted for the country s even more ambtous than what mpled by the SGP scenaro (Graph 4), though the level of the SPB forecasted for 2015 (last forecast year) remans sgnfcantly below the average SPB requred by the SGP scenaro over the projecton perod (percentle ranks of 34% and 25% respectvely n Graph 3). In the plot dsplayed as Graph 7, debt dynamcs under the SCP scenaro s also shown n order to allow comparng the mpact of Member States versus Commsson forecasts (n both cases relyng on the no-fscal polcy change assumpton). Graph 7: Gross publc debt projectons (% of GDP), sample country Baselne no-fscal polcy change, SGP nsttutonal and SCP scenaros Source: Commsson servces 14

18 3.2. SENSITIVITY ANALYSIS AROUND DETERMINISTIC PUBLIC DEBT PROJECTIONS Senstvty tests are run around the baselne no-fscal polcy change scenaro to assess the possble mpact of downward and upward rsks on publc debt dynamcs. Rsks can be related to fscal fatgue, the tghtenng/relaxng of government's fnancng condtons on the markets, shocks to real GDP growth and nflaton, shocks to the nomnal exchange rate. Standard senstvty tests descrbed n ths secton am at coverng the broad nature of shocks that can affect the future evoluton of publc debt. Senstvty tests on macro-fscal assumptons used n the standard Commsson servces' DSA are desgned as follows (see also Box 1): 1) Standard senstvty tests on short- and long-term nterest rates, consstng of (permanent) negatve and postve shocks (-1 p.p./ +1 p.p.) to the short- and long-term nterest rates on newly ssued and rolled over debt appled startng from the year followng the one of last hstorcal data avalable (currently 2014) tll the end of the projecton horzon (currently 2024). 16 2) Standard senstvty tests on real GDP growth, consstng of (permanent) negatve and postve shocks (-0.5 p.p./ +0.5 p.p.) on real GDP growth appled from the year followng the one of last hstorcal data avalable tll the end of the projecton horzon. 17 3) Senstvty tests on nflaton, consstng n standard negatve and postve (permanent) shocks to the nflaton rate (-0.5 p.p./ +0.5 p.p.) appled from the year followng the one of last hstorcal data avalable tll the end of the projecton horzon. 4) Senstvty test on the prmary balance, consstng of a standard (permanent) negatve shock to the prmary balance equal to 50% of the forecasted cumulatve change over the two forecast years 18 (the structural prmary balance s then kept constant for the remanng of the projecton horzon at the lower level obtaned for the last forecast year after applyng the shock of the ndcated sze). 5) Senstvty test on the nomnal exchange rate (for non-ea countres), consstng of a shock (for two years from the year followng the one of last hstorcal data avalable) dentcal to the maxmum hstorcal change occurred n the exchange rate over the last 10 years. Ths senstvty test should receve relatvely more attenton n the DSA of countres for whch the share of publc debt n foregn currency s beyond the upper threshold of rsk (calculated usng the sgnals approach), based on last avalable data, as reported n the heat map on publc debt structure (see Secton 3.4). For countres that are dentfed as vulnerable, accordng to the crtera presented n Secton 2, and are therefore subject to the enhanced DSA, standard senstvty tests are ntegrated by more customsed scenaros desgned as follows: 1) An "enhanced" senstvty test on short- and long-term nterest rates on newly ssued and rolled over debt amed at capturng nstances of a (temporarly) more extreme worsenng of governments' fnancng condtons on the markets. Ths s done by applyng a greater postve shock (+2 p.p.) on short- and long-term nterest rates on newly ssued and rolled over debt, for three years startng from the year followng the one of last hstorcal data avalable In the European Commsson's (DG ECFIN) debt projecton model, these shocks feed nto changes n the overall mplct nterest rate (IIR), wth the sze of the change n the IIR dependng on the structure of publc debt n terms of short- and long-term debt, maturng and non-maturng debt. In ths sense, pronounced dfferences n average publc debt maturty across EU countres s one of the factors behnd the dfferental mpact of an nterest rate shock on publc debt dynamcs. As the ncrease n nterest rates only affects debt that s newly ssued or rolled over, countres wth shorter average debt maturtes are clearly more exposed to nterest rate shocks than those wth longer maturtes. The shock s symmetrcally appled to actual and potental GDP growth, so that the output gap remans unchanged. The cyclcal component of the balance (calculated usng standard sem-elastcty parameters endorsed by the Economc Polcy Commttee Output Gap Workng Group) s therefore not affected by these shocks to growth. The usual feedback effect on growth apples n ths case (-1 p.p. fscal consoldaton leadng to +0.5 p.p. n GDP growth n the same year). 15

19 (currently 2014). After the frst three projecton years, the usual +1 p.p. permanent shock tll the end of the projecton horzon would be appled also n ths case. 2) "Enhanced" senstvty tests on real GDP growth, amed at capturng the country-specfc hstorcal varablty of real GDP growth that can dffer (also substantally) from the 0.5 used n the standard senstvty tests. These enhanced senstvty tests are desgned based on a reducton/ncrease n real GDP growth by one standard devaton 19 for two years from the year followng the one of last hstorcal data avalable. After the frst two projecton years, the usual -0.5 p.p./+0.5 p.p. permanent shocks on GDP growth would be appled tll the end of the projecton horzon. 3) Fully customzed senstvty tests on ndvdual macro-fscal assumptons, when needed, capturng country-specfc rsks that requre a more talored approach. 4) A customzed combned macro-fscal shock scenaro, n whch shocks to nterest rates, real GDP growth, nflaton, prmary balance and exchange rate are combned, based on a countrytalored approach. Results from senstvty analyss around the baselne no-fscal polcy change scenaro are reported n charts, as dsplayed n Graph 8 for a sample country. A summary table reportng the underlyng macroeconomc assumptons (real and potental GDP growth, nflaton, mplct nterest rate and structural prmary balance) for each of the senstvty scenaros s always presented below the chart STOCHASTIC PUBLIC DEBT PROJECTIONS The European Commsson's (DG ECFIN) DSA ncludes stochastc projectons as the way to feature the mpact of uncertanty n macroeconomc condtons on publc debt dynamcs n a more comprehensve way. 20 Ths methodology allows gaugng the possble mpact of downsde and upsde rsks to growth on publc debt dynamcs (also accountng for the mpact on the cyclcal component of the budget balance, through the functonng of the automatc stablzers), as well as the effects of postve/negatve developments on fnancal markets, translatng nto lower/hgher borrowng costs for governments. Stochastc debt projectons produce a cone (a dstrbuton) of debt paths, correspondng to a wde set of possble underlyng macroeconomc condtons. The latter are obtaned by applyng random shocks to short- and long-term nterest rates on government bonds, growth rate and exchange rate assumed n the central scenaro. The sze and correlaton of the shocks are based on varables hstorcal behavour. 21 The methodology allows accountng for a very large number of smulated macroeconomc condtons, beyond what s concevable n the context of senstvty analyss for determnstc projectons (2000 smulatons le, for nstance, behnd the results regularly presented n the Debt Sustanablty Montor, DSM, report) The standard devaton s calculated over the last three years of hstorcal data. For methodologcal detals on stochastc publc debt projectons, see Bert K. (2013) Stochastc publc debt projectons usng the hstorcal varance-covarance matrx approach for EU countres European Economy Economc Paper No Stochastc debt projectons were presented n the European Commsson s Fscal Sustanablty Report 2012, and results are regularly updated n ECFIN/C2 nternal Debt Sustanablty Montor. Stochastc debt projectons for the EA have also been used n the assessment of the 2014 Draft Budgetary Plans (DBPs) of the EA (see Annex 2 to the Commsson Communcaton COM(2013) 900 fnal of 15/11/2013) to the am of assessng rsks to publc fnance sustanablty n the event of adverse economc, fnancal or budgetary developments (as requred by Art. 7 of Regulaton (EU) No. 473/2013). Shocks are addtonally assumed to follow a jont normal dstrbuton. 16

20 Graph 8: Gross publc debt projectons (% of GDP), sample country Senstvty tests on nterest rates, real GDP growth, nflaton, prmary balance and nomnal exchange rate around baselne no-fscal polcy change scenaro Source: Commsson servces The baselne no-fscal polcy change scenaro from determnstc debt projectons presented before s taken as the central scenaro for stochastc projectons, whch are run over a 5-year horzon (the 17

21 standard projecton horzon to obtan meanngful results from the methodology, based on the relevant lterature). The mplct nterest rate and the growth rate n the central scenaro therefore correspond to Commsson forecasts over the forecast horzon and to macroeconomc assumptons agreed wth the Economc Polcy Commttee beyond the forecast horzon. The structural prmary balance corresponds to forecasts, and s set constant at last forecast value thereafter, based on the standard assumpton made n determnstc projectons under the no-fscal polcy change scenaro (the government budget cyclcal component, on the contrary, changes under the effects of stochastc shocks to the growth rate, 22 thus changng the prmary balance). Stochastc debt projectons therefore provde a sgnfcantly renforced senstvty analyss around the baselne scenaro. The debt rato dstrbuton obtaned through stochastc projectons allows attachng probabltes to debt paths. It s possble, for nstance, to attach a probablty to the debt rato of a certan country beng hgher than a specfed value n a gven projecton year, or to the debt rato beng on a stable or declnng path over the projecton horzon. DG ECFIN s DSA ncludes the fan chart from stochastc projectons, representng the cone of the debt-to-gdp rato dstrbuton over the 5-year horzon. In the fan chart, the projected debt path under the central scenaro (around whch shocks apply) and the medan of the debt rato dstrbuton are reported respectvely as a dashed and a sold black lne at the centre of the cone. The cone covers 80% of all possble debt paths obtaned by smulatng the 2000 shocks to growth, nterest rates and exchange rates (the lower and upper lnes delmtng the cone represent respectvely the 10 th and the 90 th dstrbuton percentles), thus excludng from the shaded area smulated debt paths (20% of the whole) that result from more extreme shocks, or tal events. The dfferently shaded areas wthn the cone represent dfferent portons of the dstrbuton of possble debt paths. The dark blue area (delmted by the 40 th and the 60 th percentles) ncludes the 20% of all possble debt paths that are closer to the central scenaro. Graph 9 reports the fan chart for the a sample country, and Table 2 reports, for each of the fve years, the values of the debt-to-gdp rato at the dstrbuton percentles dsplayed n the chart. By lookng at the chart, t s possble to conclude, for nstance, that the 2018 debt rato for ths country can be expected to le roughly between 65% (the 10 th percentle) and 78% (the 90 th percentle) wth an 80% probablty. In partcular, the 2018 debt rato s projected to be hgher than 75% (the 80 th percentle) wth a probablty of around 20%. In terms of debt dynamcs, the chart shows that, n the presence of temporary shocks to nterest rates and growth, the debt rato for the country s projected to contnue rsng tll 2016 wth a 50% probablty. Graph 9: Gross publc debt (% of GDP) from stochastc debt projectons, sample country Fan chart Source: Commsson servces 22 Country-specfc sem-elastcty parameters (endorsed by the Economc Polcy Commttee Output Gap Workng Group) are used to translate shocks to the growth rate nto changes n the budget balance-to-gdp rato. 18

22 Table 2: Gross publc debt (% of GDP) from stochastc debt projectons, sample country Dstrbuton percentles Dstrbuton percentles p p p p p p p p p Source: Commsson servces 3.4. THE ANALYSIS OF RISKS RELATED TO THE STRUCTURE OF PUBLIC DEBT FINANCING The analyss of rsks related to the structure of publc debt fnancng (by maturty, credtor base, currency of denomnaton) s ntegral part of the Commsson's (DG ECFIN) DSA. Three varables are consdered to the purpose: the change n short-term publc debt (at orgnal maturty) over total publc debt, the share of publc debt held by non-resdents, and the share of publc debt denomnated n a foregn currency (for all three varables data for the last avalable year are used n rsk assessment). Clearly, changes n the share of short-term publc debt provde an ndcaton of ncreased/decreased vulnerablty of the country under examnaton n terms of government s relance on short-term market fnancng. The share of publc debt by non-resdents captures the degree of vulnerablty related to captal holdngs by non-resdents beng more volatle, whle the share of debt n a foregn currency provdes an ndcaton of rsks related to exchange rate fluctuatons. For the three varables crtcal thresholds of fscal rsk have been calculated usng the sgnals approach. 23 The applcaton of the methodology shows that, based on hstorcal events, the three varables appear to be very good leadng ndcators of fscal stress, whch further hghlghts the mportance of ncludng an analyss of these varables n the DSA. 24 Values taken by the varables are examned n relaton to the calculated crtcal thresholds to establsh whether fscal rsks related to the structure of publc debt fnancng seem to emerge under one dmenson or the other. Results of ths analyss are presented n the DSA n the form of a heat map, n whch values of the three varables (change n the share of short-term publc debt, share of publc debt by non-resdents, and share of publc debt n foregn currency) are reported: ) n red, f they are at or above the crtcal threshold of fscal rsk from the sgnals' approach; ) n yellow, f they are below the threshold, as obtaned from the sgnals' approach, but at or above a benchmark of around 80% of the same threshold, hghlghtng an ntermedate level of fscal rsk; ) n green otherwse. An example of ths heat map, relyng on upper and lower thresholds of rsk calculated as ndcated, s provded for a sample country n Table The defnton of fscal stress used n the applcaton of the methodology s borrowed from Baldacc E., I. Petrova, N. Belhocne, G. Dobrescu and S. Mazraan (2011) "Assessng fscal stress" IMF Workng Paper No. 11/100. Results obtaned by applyng the sgnals approach on the three varables dsplay an excellent n-sample performance of these varables n antcpatng fscal stress (sgnallng powers of 0.35, 0.29 and 0.24 are obtaned respectvely for the share of publc debt by non-resdents, the change n the share of short-term publc debt and the share of publc debt n foregn currency). Among fscal varables, the three publc debt structure varables appear to be among the strongest leadng ndcators of fscal stress. They also appear to be among the best-performng fscal varables also n terms of relatvely low type-ii errors (.e. error made when predctng no fscal stress ahead of a fscal stress event). Type-II errors of 0.35, 0.54 and 0.58 are obtaned for the share of publc debt by non-resdents, the share of publc debt n foregn currency and the change n the share of short-term publc debt respectvely. More detals on the results are provded n Annex 3. 19

23 Table 3: Heat map of rsks related to the structure of publc debt fnancng, sample country Publc debt structure (2013): Change n share of short-term Share of publc debt by nonresdents Share of publc debt n foregn publc debt (p.p.): (%): currency (%): -1.5 (2012) Source: Commsson servces Notes: (1) Crtcal upper and lower thresholds:. Change (yearly) n the share of short-term publc debt: upper threshold 2.76 p.p.; lower threshold 2.2 p.p.. Share of publc debt by non-resdents: upper threshold 49.02%.; lower threshold 40%. Share of publc debt n foregn currency: upper threshold 29.82%.; lower threshold 24% (2) Data on the change n the share of short-term publc debt over total debt come from ESTAT; data on the share of publc debt by non-resdents come from ECB and OECD; data on the share of publc debt n foregn currency come from ESTAT, ECB and OECD FINANCIAL MARKET INFORMATION A bref overvew of fnancal market nformaton accompanes the presentaton of results n DG ECFIN's DSA. The overvew conssts of two tables (see Tables 4 and 5 for a sample country) reportng respectvely government bond yeld spreads (2-year and 10-year benchmarks) and CDS spreads, and soveregn ratngs by Moody s, S&Ps and Ftch. For yeld spreads (2-year and 10-year benchmarks separately) 25 crtcal thresholds of fscal rsk have been calculated usng the sgnals approach (see Annex 3 for more detals). Also n ths case, we use an upper threshold correspondng to the threshold obtaned drectly from the applcaton of the sgnals' approach and a lower threshold set at about 80% of the orgnal sgnals' approach threshold. The correspondng cells n the table are hghlghted n red/yellow/green dependng on where values le relatve to these upper and lower thresholds (see Table 4). Table 4: Fnancal market ndcators, sample country CDS (bp) Fnancal m arket nform aton as of May 2014 Soveregn yeld spreads(bp)* 2-year 7 10-year 28 5-year 29 Source: Commsson servces Notes: (1) Crtcal upper and lower thresholds:. Gov't bond yeld spreads, 2-year benchmark: upper threshold b.p.; lower threshold 220 b.p.. Gov't bond yeld spreads, 10-year benchmark: upper threshold 231 b.p.; lower threshold 185 b.p. (2) Data come from Bloomberg. Table 5: Soveregn ratngs, sample country Soveregn Ratngs as of May 2014 Moody's S&P Ftch Source: Commsson servces Local currency Foregn currency long term short term long term short term Aaa Aaa P-1 AA+ A-1+ AA+ A-1+ AAA AAA 25 For the calculaton of the thresholds usng the sgnals' approach, government bond yeld spreads have been defned relatve to German and US bonds of smlar maturty for EU and extra-eu countres respectvely. 20

24 3.6. THE ANALYSIS OF RISKS RELATED TO GOVERNMENT'S CONTINGENT LIABILITIES The latest economc and fnancal crss has clearly shown the mportance of takng nto due account governments' contngent labltes, and n partcular those arsng from vulnerabltes n the bankng sector, as these can lead to rapd and substantal ncreases n gross publc debt over GDP once they materalse (the Irsh case beng an extreme example of the rsks nvolved). The ntegraton of the analyss of government's contngent lablty rsks n the DSA ndeed allows a more comprehensve assessment of rsks to publc debt sustanablty. For ths reason, a new module on contngent labltes has been ntroduced n DG ECFIN s DSA. Ths should make t possble to broadly assess related rsks n terms of the probablty of materalzaton of the events trggerng the labltes for the government and the sze of the potental labltes nvolved. Data avalablty on governments' contngent labltes s unfortunately stll lmted. The new contngent lablty module n DG ECFIN's DSA therefore reles on both drect and ndrect nformaton from avalable statstcal sources, ncludng the followng: 1) Latest data on state guarantees as percentage of GDP for the country under examnaton based on data publshed by Eurostat, 26 provdng a measure of the sze of overall contngent labltes for the government (ncludng guarantees on EFSF borrowng). 27 2) Latest data on government s contngent labltes, n percentage of GDP, drectly related to publc support to fnancal nsttutons (actvtes related to fnancal sector support that may contrbute to government labltes n the future, but are consdered as contngent on future events at the moment of the reportng), based on data that s regularly collected by Eurostat together wth the Excessve Defct Procedure notfcatons. The dsaggregaton of the data nto ndvdual tems (labltes and assets of fnancal nsttutons guaranteed by the government; securtes ssued by the government under lqudty schemes; labltes of specal purpose enttes, ncludng those to whch certan mpared assets of fnancal nsttutons were transferred) s also reported. 3) A heat map reportng values of varables that ndrectly capture short-term rsks to publc fnances from vulnerabltes n the fnancal sector (prvate sector credt flow n percentage of GDP; 30 bank loan-to-depost rato; the level and change n the share of banks nonperformng loans; the change n the nomnal house prce ndex 31 ), as well as the (countryspecfc) estmated theoretcal probablty of government's contngent labltes due to Eurostat data on state guarantees refer to explct guarantees granted at all levels of government to any non-government unts (publc and prvate corporatons, non-proft nsttutons, households and non-resdent enttes). State guarantees provded to fnancal nsttutons n the context of the economc and fnancal crss are also ncluded, as are guarantees on EFSF borrowng. The data are avalable at: Unfortunately tme seres on overall governments' contngent labltes are too short to make t possble to calculate a crtcal threshold usng the sgnals' approach. These data are taken from Eurostat supplementary tables for the fnancal crss (data collecton started wth the October 2009 EDP notfcaton). Data provded by Member States n these tables are an ndcaton of the potental maxmum mpact that could (theoretcally) arse for government fnances from such contngent labltes (see Eurostat (2013) Eurostat supplementary table for the fnancal crss. Background note, October 2013). General government guarantees on bank deposts are not ncluded n these data on contngent labltes related to fnancal sector support. It should be noted that Eurostat has already decded to ntroduce a new questonnare to the EDP related questonnares (the so called Supplement on contngent labltes and potental oblgatons to the EDP related questonnare ), ncludng tables on government guarantees, total outstandng labltes related to publc-prvate partnershps recorded off balance sheet of the government and non-performng loans of the general government (see Eurostat (2013) Decson of Eurostat on government defct and debt. Supplement on contngent labltes and potental oblgatons to the EDP related questonnare, 22 July 2013). These data wll be transmtted annually and the frst transmsson wll take place n December 2014 (the data wll be released by Eurostat n January 2015). Ths addtonal nformaton wll be ncluded n DG ECFIN s DSA once avalable. Ths varable s common to the scoreboard of the macroeconomc mbalance procedure, but t s used here n a narrower way, to capture rsks of fscal stress from vulnerabltes n the fnancal sector. The varable change n house prces has been found n the lterature to be a good leadng ndcator of bankng crses (see IMF, 2013). Results related to the change n the nomnal house prce ndex are nonetheless to be nterpreted wth cauton. Only relatvely hgh values of the varable are ndcated n the heat map as flashng red n terms of sgnallng rsks of buldng up of bubbles n the context of an early-warnng system of possble fscal stress. But n an already set n crss context, a negatve value of the varable could also pose rsks (due to the loss n value of propertes repossessed by banks) and ths consderaton need to nform the nterpretaton of the data n the rsk assessment. 21

25 bankng losses exceedng 3% of GDP, 32 obtaned from SYMBOL smulatons (under assumptons reflectng the current bankng safety net desgn). 33 Ths estmated probablty s presented n the heat map for two possble scenaros based on the assumptons of bank recaptalzaton needs up to 4.5% and 8% of rsk-weghted assets respectvely, amed at capturng the two extremes of the nterval where the probablty of government's bankng contngent labltes s lkely to fall. For the frst fve varables ncluded n the heat map (see Table 7) crtcal thresholds of fscal rsk have been calculated usng the sgnals approach, 34 and for them the heat map reles on upper thresholds of rsk correspondng to the orgnal sgnals' approach thresholds and lower threshold of rsk set at about 80% of the orgnal thresholds. For the last varable n the heat map (the probablty of government's contngent labltes from bankng losses based on SYMBOL), relatvely hgh and ntermedate values are hghlghted n the heat map based on two crtcal values that reflect major clusters n the cross-country dstrbuton of the estmated theoretcal probablty values. 35 Latest statstcs on overall contngent labltes and contngent labltes related to support to fnancal nsttutons are reported n the DSA as shown for a sample country n Table 6. In Table 7 the heat map for government s contngent lablty rsks from the bankng sector s reported n the format used for DG ECFIN's DSA. Table 6: Government's contngent labltes, sample country Government's contngent labltes Sam ple country EU State guarantees (% GDP) 1, (2012) 14.1 (2012) Labltes and assets outsde gen. gov't under guarantee : Contngent labltes of gen. gov't related to Securtes ssued under lqudty schemes support to fnancal nsttutons (% GDP) 0 : Specal purpose entty 0 : Total Source: Commsson servces Notes: (1) State guarantees (frst lne) nclude guarantees on EFSF borrowng. (2) The tem "labltes and assets outsde gen. gov't under guarantee" does not nclude guarantees on bank deposts. (3) Data are taken from ESTAT. (4) EU averages are calculated usng sub-groups of countres for whch data are avalable The expresson "theoretcal probablty" ndcates a probablty based on the Basel rsk assessment models. The Basel II crtera are such that an nsttuton s expected to suffer losses exceedng ts captal on average once n a thousand years (a confdence level of 99.9%). The regulaton acknowledges that the hgh confdence level was also chosen to protect aganst estmaton errors that mght nevtably occur from banks nternal Probablty of Default, Loss Gven Default and Exposure At Default estmaton, as well as other model uncertantes (See Basel Commttee on Bankng Supervson, 2005). In other words, the confdence level cannot be drectly nterpreted as a frequency. Laeven and Valenca (2013) dentfy 17 systemc bankng crss epsodes worldwde n the perod , and 147 epsodes snce Based on ths, t s safe to say that the Basel models tend to under-predct the actual frequency of bank defaults, whch then carres over to these model estmates. Whle theoretcal probabltes cannot be taken lterally as frequences, ther relatve magntudes can provde nformaton on whether a country s at hgher rsk than another. See Annex 4 for more detals. See Annex 3 for more detals. See Annex 4 for more detals. 22

26 Table 7: Heat map on government's contngent lablty rsks from the bankng sector, sample country Government's contngent lablty rsks from bankng sector (2013): Prvate sector credt flow (% GDP): Bank loans-to-deposts rato (%): Share of non-performng loans (%): Change n share of nonperformng loans (p.p): Theoetcal probab. of gov't cont. labltes Change n nomnal house prce due to bankng losses >3% of GDP: ndex: bank recap. at 4.5% bank recap. at 8% 0.2 (2012) (2012) 2.7 (2012) 0 (2012) % 0.14% Source: Commsson servces Notes: (1) Crtcal upper and lower thresholds:. Prvate sector credt flow (% of GDP): upper threshold 10.9%; lower threshold 8.7%. Bank loans-to-deposts rato: upper threshold %.; lower threshold 110%. Share of non-performng loans: upper threshold 2.3%.; lower threshold 1.8% v. Change n share of non-performng loans: upper threshold 0.3 p.p.; lower threshold 0.2 p.p. v. Change n nomnal house prce ndex (YoY growth): upper threshold 12.59; lower threshold 10 v. Theoretcal probablty of gov't contngent labltes lnked to bankng losses exceedng 3% of GDP (SYMBOL): upper threshold 0.2%; lower threshold 0.05% (2) Statstcal sources used: ESTAT for prvate sector credt flow; ESTAT and WB's GFDD for bank loans-to-deposts rato; ECB, IMF's FSI and WB's GFDD for share of non-performng loans; ESTAT, ECB, BIS and OECD for change n nomnal house prce ndex. (3) SYMBOL estmated probabltes of government's contngent labltes lnked to possble bank losses are provded by the European Commsson's Jont Research Centre. For countres that are dentfed as vulnerable from the pont of vew of contngent lablty rsks, the new DSA framework further requres addtonal tools to be deployed. In partcular, a country should have ts DSA ntegrated wth contngent lablty stress-test scenaros around baselne publc debt projectons when sgnfcant bank-related rsks are dentfed. The latter are deemed to arse when one or both of the followng crtera hold true (see also Graph 10): 1) at least one of a set of three varables amed at ndrectly capturng bankng contngent lablty rsks and ncluded n the heat map (prvate sector credt flow n percentage of GDP, bank loan-to-depost rato and change n the share of non-performng loans 36 ) s above the respectve crtcal threshold of fscal rsk calculated usng the sgnals' approach. 37 2) the theoretcal probablty of government's contngent labltes lnked to bank losses exceedng 3% of GDP n the country under examnaton 38 s estmated to be hgh (.e. greater than the upper threshold) under at least one of the two bank recaptalsaton assumptons. 39 Whenever any of the condtons mentoned above holds true, the country s DSA s complemented wth an addtonal stress-test scenaro for bank-related contngent lablty rsks. Based on the two crtera, for our sample country, for nstance, ths contngent lablty shock scenaro s not requred, as the three varables n queston do not sgnal hgh rsks (bank loan-to-depost rato s the only varable sgnalng medum rsks, among those concerned) and the estmated theoretcal probablty of government s contngent labltes related to bank losses exceedng 3% of GDP reaches only ntermedate values under both bank recaptalzaton assumptons (see Table 7). For countres for whch ether of the two aforementoned crtera, on the contrary, hghlght contngent lablty rsks from the bankng sector, SYMBOL estmates on the sze of the possble mpact of a severe bankng crss on the country's publc fnances (under the current regulatory scenaro) are used to desgn the bankng contngent lablty shock scenaro. 40 A bankng contngent lablty shock of the sze ndcated by SYMBOL smulaton results for the country s assumed n t+1, and the mpact on the projected path of the debt-to-gdp rato s presented as a bankng contngent lablty shock scenaro. Ths s dsplayed n the DSA n an addtonal plot, together wth the path of the debt-to-gdp rato under the baselne no-fscal polcy change scenaro The change n the share of non-performng loans, rather than the share tself, s nserted here among the crtera to be used to select countres for whch a contngent lablty shock scenaro s to be run. Ths s because the change n the share of non-performng loans s found to be a better leadng ndcator of fscal stress than the share tself (a sgnallng power of 0.28 for the former, aganst one of 0.16 for the latter see Annex 3 for more detals). See Annex 3 for more detals on results from threshold determnaton based on the sgnals' approach for these varables. See Annex 4 for more detals. Ths type of analyss was presented for the frst tme n the European Commsson's Fscal Sustanablty Report See Benczur P., K. Bert, G. Cannas, J. Carbon, S. Langedjk, A. Pagano and M. Petracco (2014) "A bankng contngent lablty stress-test scenaro for publc debt projectons usng the SYMBOL model" European Economy Economc Paper, forthcomng. 23

27 Graph 10: Crtera to assess the country's vulnerablty to bankng contngent lablty rsks for the government Is at least one of the 3 varables capturng bankng contngent lablty rsks (prvate sector credt flow over GDP, bank loan-to-depost rato and change n share of non-performng loans) above threshold of fscal rsk? Is the theoretcal probablty of gov't contngent labltes lnked to bank losses greater than 3% of GDP estmated to be relatvely hgh? None of the above holds Any of the above holds Bank-related contngent lablty stress test NOT requred Addtonal bank-related contngent lablty stress test (usng SYMBOL estmates) requred 3.7. FORECAST ACCURACY ANALYSIS European Commsson s (DG ECFIN) forecasts le behnd determnstc and stochastc publc debt projectons. It s therefore mportant to accompany DSA results wth a bref assessment of Commsson s forecast accuracy, based on the forecast track record for the country under examnaton, wth regard to the man macro-fscal varables underlyng publc debt dynamcs (real GDP growth, nflaton, prmary balance). Ths analyss s meant to show whether forecasts on the aforementoned varables for the country under examnaton are systematcally based n one drecton or the other, n a sgn of persstent optmsm or pessmsm. European Commsson's forecast accuracy analyss s regularly conducted n DG ECFIN. 41 Latest data elaboratons resultng from the analyss are presented also n DG ECFIN s DSA. The specfcaton of forecast error used to ths purpose s one of the two optons used n broader forecast accuracy analyss (the specfcaton based on the so called "year-ahead forecast"), 42 accordng to whch the forecast error for varable X n year t s defned as the dfference between the forecasted value of varable X n year t, accordng to the Autumn vntage of year t-1, and the hstorcal value taken by varable X n year t, accordng to the Autumn vntage of year t+1. Results of forecast accuracy analyss are presented n the DSA n the form of plots, where Commsson forecast errors for the country under examnaton are reported aganst the dstrbuton of forecast errors over the whole sample of EU countres, for real GDP growth, nflaton and the prmary balance respectvely. Plots for a sample country are reported n Graphs 11 to 13, where the dots represent forecast errors for the sample country n a gven year, whle the contnuous lne for the medan and the band for the nterquartle range refer to the dstrbuton of forecast errors over the sample of EU countres. The plots allow to easly vsualzng where forecast errors for the sample country le relatve to the dstrbuton of forecast errors over all countres. No systematc bases appear for any of the three varables from the plots. Reported n the graphs are also the value of the medan forecast error for the country under examnaton over the tme span dsplayed ( ) and ts percentle rank n the dstrbuton of medan forecast errors over all EU countres. For ths sample country, on average, forecast errors over the consdered tme span appear not to be "anomalous" compared to the overall dstrbuton (percentle ranks between 42% and 56% for the three varables) For latest results on European Commsson's (DG ECFIN) forecast accuracy analyss, see Gonzalez Cabanllas L. and A. Terz (2012) The accuracy of the European Commsson s forecasts re-examned European Economy Economc Paper 476. The second opton used n DG ECFIN reles on the "current-year forecast", rather than the "year-ahead forecast". See Gonzalez Cabanllas and Terz (2012) for more detals. 24

28 Graph 11: Forecast errors on prmary balance (% of GDP) for sample country aganst EU dstrbuton of forecast errors on prmary balance (% of GDP) Source: Commsson servces Notes: (1) Forecast error for the varable at year t s defned as forecast of the varable from Autumn vntage of year t-1 mnus hstorcal realzaton from Autumn vntage of year t+1. Graph 12: Forecast errors on real GDP growth for sample country aganst EU dstrbuton of forecast errors on real GDP growth Source: Commsson servces Notes: (1) Forecast error for the varable at year t s defned as forecast of the varable from Autumn vntage of year t- 1 mnus hstorcal realzaton from Autumn vntage of year t+1. 25

29 Graph 13: Forecast errors on nflaton rate for sample country aganst EU dstrbuton of forecast errors on nflaton rate Source: Commsson servces Notes: (1) Forecast error for the varable at year t s defned as forecast of the varable from Autumn vntage of year t-1 mnus hstorcal realzaton from Autumn vntage of year t+1. 26

30 ANNEX 1 SAMPLE COUNTRY FICHE FOR DSA Sample Country Publc debt projectons (% GDP) under baselne and alternatve scenaros and senstvty tests Table A1.1 Debt projectons baselne scenaro Gross debt rato Changes n the rato of whch (1) Prmary balance Structural prmary balance (kept constant at 2015 lvl) Cyclcal component Cost of ageng Others (taxes and property ncomes) (2) Snowball effect Interest expendture Growth effect Inflaton effect (3) Stock flow adjustment and one-off measures Notes: for prmary balance and structural prmary balance, a postve sgn ndcates a defct n the table above. Graph A1.1 27

31 Graph A1.2: Determnants of changes n publc debt (% GDP) Baselne Graph A1.3: - Baselne Maturty structure of publc debt (% GDP) Rsks related to the structure of publc debt fnancng Table A1.2 Publc debt structure (2013): Change n share of short-term Share of publc debt by nonresdents Share of publc debt n foregn publc debt (p.p.): (%): currency (%): -1.5 (2012) Rsks related to government's contngent labltes Table A1.3 Government's contngent labltes Sam ple country EU State guarantees (% GDP) 1, (2012) 14.1 (2012) Labltes and assets outsde gen. gov't under guarantee : Contngent labltes of gen. gov't related to support to fnancal nsttutons (% GDP) Securtes ssued under lqudty schemes Specal purpose entty Total 0 : 0 : Table A1.4 Government's contngent lablty rsks from bankng sector (2013): Prvate sector credt flow (% GDP): Bank loans-to-deposts rato (%): Share of non-performng loans (%): Change n share of nonperformng loans (p.p): Theoetcal probab. of gov't cont. labltes Change n nomnal house prce due to bankng losses >3% of GDP: ndex: bank recap. at 4.5% bank recap. at 8% 0.2 (2012) (2012) 2.7 (2012) 0 (2012) % 0.14% 28

32 Fnancal market nformaton Table A1.5 Table A1.6 Soveregn Ratngs as Local currency Foregn currency Fnancal market nformaton as of May 2014 of May 2014 long term short term long term short term Aaa Aaa P-1 2-year 7 Moody's S&P Ftch AA+ A-1+ AA+ A-1+ AAA AAA Soveregn yeld spreads(bp)* CDS (bp) 10-year 28 5-year 29 Underlyng macro-fscal assumptons Table A1.7 Macro-fscal assumptons 1. Baselne no-polcy change scenaro Prmary balance Structural prmary balance Real GDP growth Potental GDP growth Inflaton rate Implct nterest rate (nomnal) SGP nsttutonal scenaro Prmary balance Structural prmary balance Real GDP growth Potental GDP growth Inflaton rate Implct nterest rate (nomnal) SCP scenaro Prmary balance Structural prmary balance Real GDP growth Potental GDP growth Inflaton rate Implct nterest rate (nomnal) Hstorcal SPB scenaro Prmary balance Structural prmary balance Real GDP growth Combned hstorcal scenaro Prmary balance Structural prmary balance Real GDP growth Implct nterest rate (nomnal) Hgher IR scenaro (standard DSA) Implct nterest rate (nomnal) Lower IR scenaro Implct nterest rate (nomnal) Hgher IR scenaro (enhanced DSA) Implct nterest rate (nomnal) Hgher growth scenaro (standard DSA) Real GDP growth Potental GDP growth Lower growth scenaro (standard DSA) Real GDP growth Potental GDP growth Hgher growth scenaro (enhanced DSA) Real GDP growth Potental GDP growth Lower growth scenaro (enhanced DSA) Real GDP growth Potental GDP growth Hgher nflaton scenaro Inflaton rate Lower nflaton scenaro Inflaton rate Lower SPB scenaro Prmary balance Structural prmary balance Real GDP growth Potental GDP growth Exchange rate deprecaton scenaro Exchange rate deprecaton 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% Implct nterest rate (nomnal)

33 Realsm of baselne assumptons Graph A1.4: Forecast errors on prmary balance (% GDP) Graph A1.5: Forecast errors on real GDP growth Graph A1.6: Forecast errors on nflaton (deflator) Notes: the dstrbuton of forecast errors aganst whch country-specfc forecast errors are dsplayed refers to the sample of EU countres. Graph A1.7: Prob. dstrb. (EU, ) of 3-year avg struct. prmary balance (SPB) Graph A1.8: Prob. dstrb. (EU, ) of 3-year cum. change n struct. prmary balance (SPB) 30