Local debt expansion and vulnerability reduction: an assessment for six crisis-prone countries 1

Transcription

1 Local deb expansion and vulnerabiliy reducion: an assessmen for six crisis-prone counries 1 Paloma Acevedo, Enrique Alberola and Carmen Broo 2 1. Inroducion The raios of public and exernal deb o GDP consiue crucial indicaors in assessing he financial and fiscal vulnerabiliy of a counry. On he one hand, high raios of public deb jeopardise is susainabiliy and is solvency posiion. On he oher hand, a high proporion of exchange rae exposure in deb composiion may abruply worsen is susainabiliy in imes of financial sress, characerised by problems of access o exernal markes or by sharp exchange rae movemens. In some emerging markes, exernal and domesic deb denominaed in foreign currency (boh henceforh referred o as foreign exchange, or forex, deb) have played an imporan role in he srucure of public secor deb because hese markes could no issue deb locally or in local currency. This consrain is a phenomenon someimes referred o in he lieraure as he original sin (Eichengreen and Hausman (1999)). The decreasing rend of public deb over GDP in recen years has been accompanied simulaneously and more inensely in many counries by a decrease in he corresponding share of forex deb and has coincided wih a period of widespread appreciaion of exchange raes. 3 These counries have herefore seen his evoluion as signalling a breakhrough: heir financial prospecs are improved because heir financial vulnerabiliy is reduced. Our goal in his paper is o assess quaniaively his vulnerabiliy reducion and is reversibiliy under financial urbulence. We focus on six counries ha provide an adequae sample of emerging regions: Brazil, Colombia, Indonesia, Russian Federaion, Turkey and Uruguay. We seleced hese counries among hose undergoing crises in he las decade primarily because of daa availabiliy. In addiion, hey exemplify he generalisaion of he downward rend of public and forex deb. We chose quasi-gross public deb as he ype of deb o include in our analysis so ha we could obain a homogeneous sample of daa across he counries and deec in he daa he effec of he accumulaion of reserves which is also a cenral consideraion The opinions expressed in his documen are solely he auhors and do no represen he views of he Banco de España. We would like o hank Jose Monero, Iikka Korhonen and paricipans in he CGFS Workshop on Balance Shee Effecs and Emerging Markes Bond Spreads a he Bank of England (London), he emerging marke workshop a he Ausrian Cenral Bank, he BIS-FRB Alana Join Meeing on Recen Financing Trends in Lain America: a bumpy road owards sabiliy a Mexico Ciy and he Banco de España for he helpful commens received. The usual disclaimers apply. Conac auhors: paloma.acevedo@bde.es, alberola@bde.es, carmen.broo@bde.es. See, for insance, Iner-American Developmen Bank (IADB) (2007) for a recen general view concerning public deb in emerging counries. The choice among gross deb, ne deb and any alernaive ype of measure of deb is no rivial. As saed in IADB (2007), alhough many counries provide measures of ne deb, neing sraegies differ across counries, so ne deb does no consiue a homogeneous measure. Furhermore, gross deb does no capure he effec of inernaional reserves. See Cowan e al (2006) or IMF (2003) for alernaive deb definiions ha are differen from quasi-gross public deb. 88 BIS Papers No 36

2 Because no homogeneous daabase exiss ha perfecly suis he period of ime and disaggregaion required by his research, in all bu wo cases we colleced daa direcly from he specific deb daa releases of official insiuions. For Russia and Indonesia we used Inernaional Moneary Fund (IMF) daa. We decomposed quasi-gross deb ino foreign deb (issued in inernaional deb markes) versus local deb (issued in domesic deb markes). We hen made a furher disincion beween local deb linked o he exchange rae and local deb linked o local currency when ha disincion was available. Table 1 shows he sources and respecive links used o creae he daabase. Table 1 Daabase consrucion Counry Deb Availabiliy crisis year of daa Descripion Source Web link Brazil General governmen gross deb 1 Cenral Bank of Brazil Public secor domesic deb 2 Minisry of Finance esaisica/es_divida.asp Colombia Public secor deb 1 Bank of he Republic deuda/boleindepu18.pdf Naional governmen domesic deb 2 Minisry of Finance and Public Credi Indonesia Cenral governmen gross deb Russia Cenral governmen gross deb Turkey Public secor deb 1 Uruguay Public secor deb 1 Ar. IV (IMF) Ar. IV (IMF) Turkish Treasury Cenral Bank of Uruguay pepmaf/deudapublica/dbspg2.xls 1 Domesic and exernal deb. 2 Used o calculae he breakdown of domesic public deb. Source: Auhors elaboraion. Figure 1 displays he raio of public secor deb o GDP for hese counries in 20 and in he year of he highes ousanding deb during he pas decade, which in mos cases coincides wih episodes of financial urmoil (see Manasse and Roubini (20) or de Bolle e al (2006) for a daing of financial crises). The graph shows boh he gross deb holdings and deb ne of inernaional reserves (quasi-gross public deb). BIS Papers No 36 89

3 Figure 1 Gross and quasi-gross public secor deb, seleced counries As a percenage of GDP Toal deb Exernal deb Domesic deb Quasi-gross¹ Gross Reserves Ne of reserves Currency-linked No currency-linked BR CO ID RS TK UY BR CO ID RS TK UY BR CO ID RS TK UY BR = Brazil; CO = Colombia; ID = Indonesia; RS = Russia; TK = Turkey; UY = Uruguay 1 Domesic deb plus exernal deb ne of inernaional reserves. Source: See Table 1 for variable definiions and sources. As Figure 1 shows, Russia had he mos significan deb reducion. Quasi-gross public deb wihin he sample shrank abou 99 percenage poins (pp) of GDP beween 1999 and 20 o become negaive, due o he counry s large reserve accumulaion. 5 In Turkey and Indonesia, he reducion was 34 pp and 24 pp of GDP, respecively, from 2001 o Brazil also exemplifies hese dynamics: in 2002 is quasi-gross public secor deb was 74% of GDP, whereas in 20 i decreased o around 68% of GDP. The quasi-gross public secor deb in Uruguay and Colombia fell around 13% and 6%, respecively, from 2003 o 20. I is remarkable ha he reducion in deb has been accompanied by an overall reducion in he share of forex deb (exernal deb, domesic deb in foreign currency or deb linked o he exchange rae). Figure 2 clearly shows he reducion of he proporion of forex deb. This figure represens he evoluion of he deb composiion in erms of exernal deb, exchange rae-linked domesic deb and domesic deb in local currency for he same periods. The decline in he forex deb share is more dramaic in Brazil, Turkey and Colombia (40%, 28% and 18%, respecively). Also noable is he reducion in exchange rae-linked domesic deb in he wo Lain American counries, 7 o he poin ha exchange rae-linked domesic deb was suppressed in Brazil by Only in Indonesia did he proporion of exernal deb increase in he laer years Henceforh, for Russia only, we develop our analysis of public deb in erms of gross public deb raher han quasi-gross public deb. Oherwise, since quasi-gross deb is currenly negaive, he corresponding resuls for he res of he analysis would be misleading. For Indonesia, he year 2001 is considered o be he previous peak of public deb, mainly because of daa availabiliy, alhough according o oher papers (ie de Bolle e al (2006)), he mos recen urmoil is raceable back o See Jeanneau and Tovar (2006) for a recen discussion of he evoluion of domesic markes in Lain America and Tovar (20) for a deailed analysis of deb denominaed in local currency in he hree Lain American counries of he sample (Uruguay, Colombia and Brazil). 90 BIS Papers No 36

4 Figure 2 Quasi-gross public secor deb composiion, seleced counries 1 Exernal ne of reserves Domesic currency-linked Domesic oher Brazil Colombia Indonesia Russia² Turkey Uruguay 1 In per cen. 2 Gross public deb used for calculaions. Source: See Table 1 for variable definiions and sources. There are wo caegories of explanaions for his developmen of boh he public and he forex deb. Figure 3, which shows he evoluion of he nominal exchange rae and he sovereign spreads, illusraes he firs caegory: an inernaional conex of very favourable financial condiions have influenced public deb considerably. As we shall see, he second caegory is closely relaed o he firs, i.e. he creaion of proacive policies o manage public deb has also been significan. Regarding he favourable inernaional financial conex, some aspecs are worh qualifying. For insance, jus as exchange rae crises make deb explosive in counries wih a large share of forex deb, real exchange rae appreciaions can dramaically decrease deb raios and have an impac on he srucure of deb. This is precisely wha happened afer he crises. The exchange rae recoveries were generalised, as shown in he real exchange rae evoluions in Figure 4. The appreciaion of he Russian rouble (a 64% real appreciaion beween 1999 and 20), he Turkish lira (a 44% real appreciaion beween 2001 and 20) and he Brazilian real (26% beween 2002 and 20) were he mos significan. The Indonesian rupee is he only currency of he sample ha depreciaed from 2003 o 20 (6%) precisely he only counry where he share of forex deb increased. The posiive period for emerging financial markes is also confirmed by he dynamics of sovereign spreads ha have narrowed in a conex of increasing capial flows. In his sense, he EMBI Global Composie decreased around 900 basis poins from January 1999 o Ocober 2006, and his reducion of sovereign spreads was especially severe in emerging Europe, where in he same period i narrowed around 2000 basis poins. Anoher facor conribuing o his benign financial framework is he favourable behaviour of he GDP growh raes in all emerging regions in he conex of propiious world growh. For insance, he annual percen change of growh in 20 for such emerging regions as developing Asia, Cenral and Easern Europe and Lain America was 9.0%, 5.4% and 4.3% well above he 2.6% rae of growh for advanced economies in 20 and higher or similar o world growh (4.9%) (see IMF (2006)). As for he proacive deb managemen facor, he evoluion of public and forex deb is closely relaed o he developmen of local deb markes in local currency. Fiscal auhoriies began o aach increasing imporance o reducing in a susainable manner he vulnerabiliy of public finances, and began o creae more proacive deb policies o manage public deb in his direcion. They had learned heir lesson from pas experience, when excessive exchange rae exposure gave rise o balance shee mismaches. BIS Papers No 36 91

5 Figure 3 Sovereign spreads and exchange raes in crisis episodes, seleced counries 1 Brazil 2 Colombia 3, 4 Indonesia 4, Exchange rae (rhs) Spread (lhs) Russia 3 Turkey 3 Uruguay Noe: Daes of crisis episodes considered: Brazil (2002); Colombia and Uruguay (2003); Indonesia and Turkey (2001); and Russian Federaion (1999). 1 Naional currency per US dollar and spreads in housands of basis poins. Index. 4 Exchange rae in housands of unis per US dollar. 5 ABI index. Source: Daasream. 2 EMBI + index. 3 MBI Global The link beween benign condiions and proacive policies derives from anoher facor. The favourable financial condiions and he expeced behaviour of he exchange rae, which increased he relaive demand for local deb and he abiliy of he auhoriies o place i on he marke, also encouraged auhoriies o make discreionary changes o he deb composiion. Ineresingly, he conjuncion of hese wo facors has creaed a paradox worh menioning. In an environmen of currency appreciaion, auhoriies rying o maximise deb reducion in he shor erm have an incenive o mainain or increase he share of forex deb, as his would decrease public deb as a percenage of GDP, leading o some sor of viruous circle. Conversely, a reducion of forex deb as a resul of acive deb managemen ends o miigae deb reducion driven by exchange rae appreciaion. Coningen on financial urbulence, however, his paradox of he local deb bias can be solved. In such a case, he exchange 92 BIS Papers No 36

6 rae should sharply depreciae and, if here has previously been a reducion in he proporion of forex deb in oal deb, hen he counry is beer able o absorb he impac of he negaive scenario. The comparison beween he shor-erm coss of deb reducion and he long-erm (coningen) benefis is one of he by-producs of our analysis. 8 Figure 4 Real exchange raes, seleced counries 1 Brazil (2002) Indonesia (2001) Turkey (2001) Colombia (2003) Russia (1999) Uruguay (2003) Noe: Year of major ousanding deb in brackes; doed line hereafer = 100. Source: Economis Inelligence Uni. For our analysis, we develop a quaniaive approach o assess he effecive vulnerabiliy reducion in he deb composiion and he precise conribuion of proacive deb managemen. Firs, in he nex secion, we disenangle he conribuion of he exchange rae o he shifs in deb srucure from oher auonomous or genuine composiion effecs in he srucure of deb. Following ha, we develop a heoreical framework of deb dynamics analysis, and hen perform a counerfacual exercise based on calculaing public deb dynamics under he previous deb srucure. In his way we can assess he change in vulnerabiliy based on he difference in percenage poins of GDP beween he acual deb and he deb resuling from his counerfacual exercise. Then, we replicae he previous crisis scenario of economic and financial urbulence for he period and perform a sress es analysis of deb susainabiliy. We also use alernaive crieria o design he sress as a es of robusness. Finally, we summarise our findings. 2. Public deb composiion: disenangling price and composiion effecs This secion focuses on seing a framework for analysis of he effecs of he shifs in forex deb (he sum of exernal and domesic exchange rae-indexed deb) on oal public deb. The share of forex deb, α is defined as 8 In his paper, we focus on he susainabiliy-vulnerabiliy assessmen concerning he exchange rae-linked deb. We do no address oher opics on deb composiion such as long-erm versus shor-erm deb or nominal versus indexed deb even hough here is inense debae on hese issues. See, for insance, Alfaro and Kanczuk (2006). BIS Papers No 36 93

7 * e D α = (1) ( D + e D * ) * where e is he nominal exchange rae in he period, D is he amoun of ousanding forex deb, eiher exernal deb or exchange rae-linked domesic deb, denominaed in dollars in, and D is he ousanding domesic deb denominaed in local currency in period. 9 Wihin his framework, i is raher sraighforward o evaluae he imporance of he effec of he exchange rae and he effec due o he composiion of deb on he oal variaion in composiion. The oal variaion of he raios of forex deb o oal deb beween he final ( = 1) and he iniial ( = 0) periods of reference, ha is, (α 1 α 0 ), can be decomposed in hese wo effecs, as follows: α1 α 0 = EE + CE + ε (2) where he firs par of he righ-hand side of (2) (EE) is he exchange rae effec and CE is he composiion effec. The residual erm ε in he expression will be allocaed beween boh effecs, as we explain below. EE is he variaion in he proporion of exernal deb. I is indexed o a foreign currency domesic deb resuling from variaions in he exchange rae obained by keeping he amoun of deb unalered. Analyically: * e1d0 EE = α * 0 (3) D0 + e1d0 where he firs elemen in he righ-hand side of EE will be denoed as α 1 E. CE is he variaion of α due o he changes in he relaive volumes of he differen ypes of deb, had he exchange rae no changed: * e0d1 CE = α * 0 (4) D1 + e0d1 where, analogously o (3), he firs elemen in he righ-hand side of (4) will be denoed as α 1 C. In his las ype of effec, he effec of proacive managemen policies arises, alhough oher facors, such as he relaive demand and supply of deb insrumens, may also be prominen. The allocaion of he residual change o each facor is made according o he scheme in Figure 5. Noice ha he whole variaion in he forex deb share (ha is, α 1 α 0 ) is he area defined by coordinaes e 1 D 1 * minus e 0 D 0 * (he area shadowed wih verical lines). EE as saed in he previous noaion would be he area comprising α E 1 α 0 and CE would be α C 1 α 0 (he yellow and green shaded areas, respecively). The remaining area should be equally disribued beween EE and CE in order o accuraely represen he difference beween he vecors α 1 and α 0. 9 See Calvo e al (2002) for a pioneering analysis of fiscal susainabiliy incorporaing he currency composiion of deb. 94 BIS Papers No 36

8 Figure 5 Public deb decomposiion α 1 E α 1 E e 1 α 0 α 1 C e 0 D* 0 D* 1 Source: Auhors elaboraion. The facorial decomposiion of EE and CE is represened in Figure 6 for he six counries in erms of he percenage poins ha each facor has conribued o he reducion in he share of foreign currency deb, considering ha = 1 is 20 and = 0 is he year of he corresponding deb crisis for each counry. We use as reference for his exercise he public deb ne of reserves (quasi-gross public deb), excep for Russia, where such magniude is negaive. Figure 6 Exchange rae effec and composiion effec, seleced counries 1 Composiion effec (CE) Exchange rae effec (EE) Toal effec Brazil Colombia Indonesia Russia² Turkey Uruguay 50 1 Variaion of he raios of forex deb o oal quasi-gross public deb beween he crisis episode and he year 20, in per cen. 2 Gross public deb used for calculaions. Source: Auhors calculaions based on naional daa. Despie he srong exchange rae appreciaion, CE dominaes in all counries bu Indonesia, where i conribues o he increase in he share of forex deb. In absolue erms, CE is larges in Brazil: 34% of he 40% reducion in he forex deb share is due o CE. However, in relaive erms, i is even more significan in Turkey: 26% of he 28% reducion is CE in oher words, more han 90% of he reducion is due o CE. For he average of he five counries where he share of foreign currency deb is reduced, 85% of he reducion can be aribued o he pure composiion effec. BIS Papers No 36 95

9 3. The framework of analysis: deb dynamics Public deb susainabiliy analysis (DSA) is an increasingly widespread ool used o assess he vulnerabiliy posiion of public finances. In recen years, more aenion has been paid o his approach in policy analysis, mos noably in IMF counry assessmens. A growing number of papers also employ DSA someimes from a sochasic approach (eg see Celasun e al (2006), Hosland and Karan (20) or Garcia and Rigobon (2004)). Apar from is simpliciy, he main advanage of his mehodology for our objecives is ha i can provide an explici measure of vulnerabiliy ha can be raced over ime and is well-suied o he sress es analysis. DSA focuses on he deb dynamics equaions ha are deermined, in a simplified framework, by a raher limied number of variables. Furhermore, forecass for mos of hese variables are readily available on he marke. These forecass allow us o esablish a baseline scenario for he fuure evoluion of deb. The framework is also useful for visualizing how deb would respond o a siuaion of sress by changing he forecass using esimaes of he variables under negaive shocks. These sress ess compound alernaive scenarios; his gives an idea of he resilience of deb and herefore of he vulnerabiliy of he public finance posiion. The saring poin is he deb dynamics equaion expressed as: * (1 + r ) (1 + r )(1 + Δe ) D = PB + (1 α ) D 1 + α D 1, (5) (1 + g ) (1 + g ) where PB is he primary balance and D is he sock of public deb a he end of ime, boh expressed as a raio of GDP. The share of deb denominaed in foreign exchange is α, as we already know, while (1 α ) is he share of local currency deb; r* and r are heir corresponding real ineres raes. Foreign-denominaed exernal deb can be in foreign currency (mosly exernal deb) or indexed o he exchange rae (mosly domesic deb). Finally, Δe is he variaion in he nominal exchange rae, where a posiive Δe means an exchange rae depreciaion and g is he real rae of growh. Afer some algebra, he dynamics of public deb can be expressed as * ( r g ) ( r + Δe + r Δe ) ΔD = PB + (1 α ) D 1 + α D 1, (6) (1 + g ) (1 + g ) where, for simpliciy, we have dropped he coningen liabiliies. This equaion is he basis for he susainabiliy exercises performed in he DSA. Given he curren level and composiion of deb, for given forecass of he primary balance, he growh rae, he nominal exchange rae and he real ineres raes (domesic and foreign), i is possible o projec deb rajecories. Increases in he raio of deb o GDP derived from hese exercises provide a measure of vulnerabiliy, and a decrease in he raio suggess a reducion in vulnerabiliy. Expression (6) can be ransformed in a more convenien way by separaing he effec of he exchange rae from he res: * ((1 α ) r + α r ) g Δe + r Δe ΔD = PB + D 1 (1 α ) D 1 + α D 1. (7) (1 + g ) (1 + g ) (1 + g ) For pracical purposes, i is imporan o noe ha he real ineres raes by insrumen or currency are no usually available, so ha we have o find a way o measure he approximae real cos of local and forex deb. Daa exis on ineres paymens on public deb: IP, which can be defined as * IP = (( 1 α ) r + α (1 + Δe ) r ) D 1 = ρd 1, (8) * * 96 BIS Papers No 36

10 where, for convenience, ρ denoes he average cos of deb a ime. ρ can be calculaed in every counry hrough he daa of IP expressed as: ρ IP = D 1 For compleeness and furher convenience, also noe ha he implici local deb real rae can be solved from he definiion of r : r * ρ α (1 + Δe ) r =. (10) (1 α ) so ha if we are able o proxy for he real foreign cos of deb hrough he spread, as i urns ou we derive an approximaion of he respecive real ineres rae by counry. Subsiuing ρ in (7) yields he basic equaion for he empirical approach: D 1 ΔD = PB + [ ρ (1 α ) g + Δe α ]. (1+ g ) (9) (11) 4. Empirics: deb evoluion, deb srucure and vulnerabiliy reducions These expressions provide us wih an adequae framework o analyse wha has been going on in he counries under consideraion here. I is convenien o sar wih an illusraive example of how he differen facors impinge on he evoluion of deb and hen move o a more deailed analysis of he impac of he shifs in deb srucure on vulnerabiliy. 4.1 Conribuions o deb reducion Compuing he parial derivaives in expression (5) allows us o deermine he conribuion of each facor o he variaion of D (ΔD ) on an annual basis. To focus on he issues we are more ineresed in, we consider he decomposiion of he annual variaion of D in erms of PB (in his case, here is a one-o-one relaionship), and he annual variaion of he share of forex deb in oal public deb (α ), he exchange rae (e ) and, for he sake of simpliciy, he remaining conribuions (ineres raes and raes of growh) are aggregaed in a residual. Figure 7 illusraes he case of Brazil. The subsanial magniude of he primary balance is a powerful deb reducion driver hroughou he period. However, he more ineresing resuls are he ineracion beween he exchange rae and he share of forex deb α. From 2001 o 2002, he currency depreciaed, and here was an imporan posiive conribuion o deb of 9 pp of GDP. Thereafer, he appreciaion of he exchange rae induced a negaive conribuion o public deb in GDP. The cumulaive decrease in GDP from 2002 o 20 was 4%. In parallel, α increased in he firs, urbulen period. Owing o he conemporaneous exchange rae depreciaion, his variable added 3 pp o he deb-o-gdp raio (he green area in Figure 7). Boh facors ogeher (α and exchange rae) amouned o a 12 pp increase in deb in However, in he following years, he ineracion of currency appreciaion and reducion in forex deb had a differen resul: he conribuion of he dwindling share of forex deb is posiive because i miigaes he effec of he exchange rae appreciaion on deb reducion. Finally, he residual picks up he combined conribuion of ineres raes, growh plus oher adjusmens. BIS Papers No 36 97

11 Figure 7 Annual variaion of public deb in GDP by erm conribuions in Brazil 1 Residual α² Primary balance (PB) Exchange rae Variaion As a percenage of GDP. 2 Share of forex deb in oal deb, in per cen. 7.0 Source: Auhors calculaions based on naional daa. 4.2 A counerfacual exercise: deb reducions wihou proacive managemen policies The Brazilian example highlighs he fac ha he ineracion beween exchange rae appreciaions and reducions in he forex share can work agains deb reducion. This is he paradox of he local deb bias ha we poined ou in he inroducion. Bu we also noed laer ha he changes in deb srucure (see Figure 6) are in par mechanically driven by he evoluion of he exchange rae. In fac, we showed ha a subsanial par of he reducion in forex deb was no due o he exchange rae evoluions bu raher o pure composiion effecs, in which he proacive deb managemen policies of he auhoriies have had a cenral role. Now, wihin he deb dynamics framework, we can give a quaniaive assessmen of he (negaive) impac of proacive deb managemen on deb reducion. The quesion is sraighforward: wha would he level of deb be oday, neing ou he composiion effec ha is, looking a i wihou proacive deb managemen? Obaining he compuaions of α E 1 as saed in (3) on a yearly basis, we can deermine counerfacual deb pahs for he public deb raio. Figure 8 shows he resuls of his exercise as carried ou for all six counries. The blue line represens he acual public deb rajecory. Neing ou he pure composiion effec delivers he pah represened by he green line. The graph is compleed wih he opposie exercise, shown by he red line. In his case, we consider he pure composiion effec bu assume ha he impac brough abou by he exchange rae evoluions disappears ha is, we assume his o be he curren deb level, had he real exchange rae been kep consan. Table 2 summarises he oucomes of he counerfacual exercise for he six counries. In he case of Brazil, he acual pah displays deb falling from 74% o 68% of GDP. However, his decrease is acually much greaer, around 60% in 20, when we ne ou he pure composiion effec. The reason for his difference is ha he dwindling forex deb does no fully capialise he impac of he real exchange rae appreciaion. To sum up for Brazil, he implici loss, in erms of percenage poins of deb o GDP, derived from proacive deb managemen would mean ha he level of deb would now be a sizeable 8 pp of GDP. This can be aken as a measure of he opporuniy coss of subsiuing local deb in local currency 98 BIS Papers No 36

12 for forex deb. On he conrary, if he nominal exchange rae had remained a he 2002 levels, he quasi-gross public deb in 20 would have been around 79% of GDP. Figure 8 Acual vs counerfacual deb evoluion, seleced counries 1 As a percenage of GDP Brazil Colombia Indonesia Deb ne of reserves 80 Deb ne of exchange rae effec 51 Deb ne of composiion effec Russia 2 Turkey Uruguay Public deb ne of reserves (quasi-gross deb). 2 Gross public deb used for calculaions. 20 Source: Auhors calculaions based on naional daa. In Turkey, hese proacive policies have also been quie pronounced. There, neing he change in composiion due o he governmen s deb managemen means ha he public deb in 20 would be 10 pp of GDP less. In he res of he counries where he reducion in he share of forex deb in oal public deb has been relaively small or has no aken place (ie Colombia, Indonesia, Russian Federaion and Uruguay), he difference beween he acual pah of public deb and he public deb under consan composiion of he year of crisis is also small (his difference represens less han 1 pp of he GDP of each counry). Brazil and Turkey are clear examples of where he opporuniy coss of diminishing he share of forex deb in oal public deb are more eviden, due o boh he inense exchange rae appreciaions and he effors by he fiscal auhoriies o recompose deb in favour of local and local currency-denominaed deb. BIS Papers No 36 99

13 Table 2 Counerfacual exercise resuls for 20 Counerfacual (20) Brazil Colombia Indonesia Russia Turkey Uruguay Public deb ne of reserves/gdp 68.5% 44.4% 34.6% 14.5% 57.1% 71.9% Deb/GDP ne of exchange rae effec 74.0% 47.3% 30.4% 54.6% 65.7% 90.3% Deb/GDP ne of composiion effec 60.0% 43.8% 34.4% 14.7% 47.3% 72.3% Poins of deb/gdp due o composiion effec Poins of deb/gdp due o exchange rae effec Source: Auhors calculaions. Some imporan caveas, however, give a more nuanced view of he opporuniy coss of moving ou of forex deb. Mos imporan is ha his is a parial exercise. We are assuming ha nohing else changes, bu his is quie an assumpion. As menioned above, fiscal auhoriies could no have developed he local deb markes as swifly under more sringen financial condiions. More imporanly, he very same evoluion of he exchange rae is no far from he evoluion of deb composiion; he reducion in exernal deb, a process deepened by very acive policies in Brazil, shapes he expecaion of agens and has probably conribued o a greaer pressure on he exchange rae and fosered a higher accumulaion of reserves (and hus a bigger reducion of quasi-gross deb). From he second ype of exercise, where he exchange rae is kep unalered, some ineresing conclusions also follow. As expeced, he numbers show ha deb dynamics would have been much less favourable under he exchange raes of he year of crises for he six counries excep for Indonesia and Russia, where he nominal exchange rae has appreciaed wih respec o heir years of crises. The more damaged counry in erms of mainaining he same nominal exchange rae would have been Uruguay, which would have increased is deb 18 pp of GDP. All in all, under he perspecive aken in his secion, i migh seem ha proacive policies o reduce he share of forex deb in oal GDP have enailed coss in erms of limied deb reducion. Neverheless, his shor-erm cos mus be balanced wih he prospecive benefis derived from a less forex-dependen deb srucure in cases of financial urbulence. 4.3 Sress es: resilience in deb vulnerabiliy The sandard DSA framework based on sress esing consiss of designing a siuaion of urbulence (or sress) comparable o he mos recen crises o deermine wheher vulnerabiliy has effecively been reduced and hen conrasing i wih a baseline scenario. The firs sep, hen, is o define he baseline scenario. Wih forecass from he IMF s Aricle IV repors for he counries under sudy ( ) and LainFocus Consensus Forecas (2006) over a hree-year horizon (2006, 2007 and 2008), we obained he raw daa o projec he deb pahs. This mehodology is useful for improving he homogeneiy of he analysis and for comparing he differen oucomes wih hose provided by he IMF. Nex, we designed he sress 100 BIS Papers No 36

14 scenario o replicae he mos recen financial urmoil experienced by hese counries coinciding, as we showed above, wih he previous peak in deb. Table 3 displays he daa underlying he baseline and he sress scenarios. The changes herein are applied o all variables in deb dynamics equaion (5). 10 Table 3 Baseline scenario and sress scenario for he simulaion of deb dynamics In per cen Baseline scenario Sress scenario Brazil Exchange rae Real GDP growh GDP deflaor Ineres rae (i) Ineres rae (i*) Primary balance Implici liabiliies 4, Colombia Exchange rae Real GDP growh GDP deflaor Ineres rae (i) Ineres rae (i*) Primary balance Implici liabiliies 4, Indonesia Exchange rae Real GDP growh GDP deflaor Ineres rae (i) Ineres rae (i*) Primary balance Implici liabiliies 4, For foonoes, see he end of he able. 10 In hose isolaed cases wih no daa available for he period of crisis, we obained he negaive shock by adding o he daa in he baseline scenario one sandard deviaion of he available daa. BIS Papers No

15 Table 3 (con) Baseline scenario and sress scenario for he simulaion of deb dynamics In per cen Baseline scenario Sress scenario Russia Exchange rae Real GDP growh GDP deflaor Ineres rae (i) Ineres rae (i*) Primary balance Implici liabiliies 4, Turkey Exchange rae Real GDP growh GDP deflaor Ineres rae (i) Ineres rae (i*) Primary balance Implici liabiliies 4, Uruguay Exchange rae Real GDP growh GDP deflaor Ineres rae (i) Ineres rae (i*) Primary balance Implici liabiliies 4, Variaion of naional currency per US dollar. 2 Nominal domesic ineres rae. 3 Nominal exernal ineres rae. 4 As a percenage of GDP. 5 Recogniion of implici or coningen liabiliies. Source: Auhors calculaions based on IMF daa. Figure 9 and Table 4 presen he resuls. Looking again a Brazil as illusraive, he blue doed line represens he baseline scenario and he orange doed line sands for deb dynamics under he sress scenario. Boh deb evoluions employed he pah of α under deb composiion for he year 20. As expeced under he baseline scenario conveying he coninuaion of favourable condiions quasi-gross deb gradually decreases owards 60% of GDP, while deb increases under he sress scenario and hen sabilises above 70%. 102 BIS Papers No 36

16 Figure 9 Baseline and sress scenarios, seleced counries As a percenage of GDP Brazil Colombia Indonesia Public deb¹ Sress - curren³ Baseline scenario 5 Sress - ne² 90 Sress - counerfacual Russia 6 Turkey Uruguay Public deb ne of reserves (quasi-gross deb). 2 Sress wih deb composiion ne of composiion effec. 3 Sress wih curren deb composiion. 4 Sress under counerfacual deb composiion, defined as he year of major ousanding deb as i appears in Figure 1. 5 Curren srucure. 6 Gross public deb used for calculaions. Source: Auhors calculaions based on naional daa. Wha would he impac of he urmoil have been if he deb srucure had been kep unalered relaive o he year of crisis? The red line provides a firs, bu inadequae, approximaion. I represens he impac of he sress es wih he deb srucure ne of he pure CE (bu leaving EE operaing) and saring from he curren level of deb (20). Noice ha he evoluion is much more explosive under he curren deb srucure. Had deb managemen no been proacive, 11 he increase would have been much larger (o over 95%), as he red line shows, seing he deb on an explosive pah. The gap beween boh lines more han 20 pp of GDP over a hree-year horizon is indicaive of he imporance of a less forex-exposed deb srucure. 11 To be more precise, he 2002 deb composiion permis he exchange rae o affec he srucure bu nes ou he pure composiion effec. BIS Papers No

17 104 BIS Papers No 36 Table 4 Comparison of sress scenarios Brazil Colombia Indonesia Baseline scenario Replica Under curren srucure Under counerfacual Gap Two sandard deviaions Under curren srucure Under counerfacual Gap Aggregaed sress Under curren srucure Under counerfacual Gap For foonoes, see he end of he able.

18 BIS Papers No 36 1 Table 4 (con) Comparison of sress scenarios Russia Turkey Uruguay Baseline scenario Replica Under curren srucure Under counerfacual Gap Two sandard deviaions Under curren srucure Under counerfacual Gap Aggregaed sress Under curren srucure Under counerfacual Gap Under curren deb composiion. 2 Represens he difference beween sress under 2002 deb composiion (counerfacual) and sress under curren deb composiion. Source: Auhors calculaions based on naional daa.

19 Why is he red line misleading? We have seen in he counerfacual exercise ha neing ou pure composiion effecs would have resuled in a lower deb raio in he case of Brazil, owing o he susained real exchange rae appreciaion. Thus, we have o compare he effecive lower reducion in deb due o he proacive deb managemen policies o he prospecive gains during a financial crisis. More precisely, he green line represens he deb dynamics assuming no pure composiion effec as in he red line plus he level of deb resuling from he counerfacual exercise. This, in our view, is he righ gauge o measure vulnerabiliy reducion due o proacive deb managemen. In pracical erms, his amouns o aking he end-poin of he green line as a reference and projecing i forward under he sress scenario. The green line hus exended has a pah similar o he red line, bu i sars from a lower level. Consequenly, he difference in he deb raio is very small in he firs year, and hen widens o around 10 pp of GDP. We can ake his figure as he ne gain from Brazil s deb managemen policy. In oher words, he shor-erm coss of implemening proacive policies in order o decrease he share of forex deb in oal deb are more han compensaed for by he long-erm gains of implemening hem. For he oher five counries, he forecass under he baseline scenario are as follows. Colombia, Turkey and Uruguay decrease heir deb owards 40%, 45% and 60% of GDP, respecively (blue line). Despie he evoluion of is currency and he composiion of is deb, Indonesia also reduces is deb o 25% of GDP. Finally, he forecas for Russia is especially favourable, as he gross deb decreases sharply o 25% of GDP ha is, no only is quasigross deb negaive, bu so is gross deb. We can employ he sress scenario for he remaining counries in he same manner, alhough he resuls are less clear. Recall ha he more ineresing conclusions derive from comparison of he evoluion of deb under he deb composiion of he year of crisis (green line) and under he deb composiion of 20 (orange line). The comparison favours deb recomposiion only in Russia, where he deb level is no currenly a problem, and in Uruguay, o a lesser exen han in Brazil (see Table 4). Neverheless, in some counries, such as Colombia, he benefi of he proacive policies implemened unil 20 gives rise o a scan average decrease in deb of 1 pp of GDP accumulaed in he forecased period. In he case of Turkey he accumulaed differences afer hree years are negaive ( 1 pp GDP), alhough hey were previously posiive. In he case of Indonesia, he gap is negligible hroughou he forecas scenario. I is imporan o keep in mind here he caveas we menioned above because he direc inference from hese resuls is ha, excep for Brazil, deb recomposiion effors do no seem o pay off in erms of vulnerabiliy reducion under sress. Again, we base hese caveas on he impac of hese deb rajecories on expecaions. I is difficul o assume ha he reacion of he markes would be he same comparing he mild deerioraion implied by he orange line wih he sharp increase in deb under a less favourable deb srucure. As a consequence, he evoluion of he financial variables is reasonably expeced o be worse in he second case. This endogeneiy implies ha he compuaion of ne gains is a floor raher han a midpoin esimae. To check he robusness of he sress es, we repea he exercise, considering wo alernaive assumpions for he design of he sress scenarios. Firs, following he mehodology employed in mos of he IMF s Aricle IV, we add wo sandard deviaions from he sample series o he corresponding daa of he baseline scenario, denoed as 2SD. 12 Second, we build a scenario on he average sress scenario (average sress, for shor) for each variable of he six counries based on he hisorical crierion of he previous subsecion. 12 Two sandard deviaions from he sample of each variable from he year of crisis o 20 are added o each variable from 2006 o 2008 (boh inclusive). 106 BIS Papers No 36

20 Table 4 shows he oucomes corresponding o hese wo new crieria. The resuls in general are quie robus under he hree differen alernaives, in erms of boh he size of he shock and he direcion vis-à-vis he wo new sress scenarios designed. Indonesia and Turkey are he excepions. For Indonesia, he exercise is no robus under he assumpion of average sress, as public deb o GDP is lower under he sress scenario han under he baseline scenario. For Turkey, he gap beween he sress under he deb srucure of he year of crisis ha is, he counerfacual scenario and he sress under he deb srucure of 20 becomes negaive, implying less resilience o a negaive shock. For he oher simulaions, he resuls are almos he same and, in some cases such as he resuls of he scenario based on averages for Colombia show a lower deb under he curren composiion han under he previous composiion. 5. Conclusions In his paper, we have evaluaed he impac of he shif of public deb away from foreign currency on he vulnerabiliy of a group of seleced emerging counries, which no so long ago underwen deep financial urbulence. We firs emphasised ha he raio of public deb o GDP and, even more dramaically, he share of forex deb have been reduced in hese emerging markes in a conex of favourable financial condiions. Exchange rae appreciaions helped o reduce boh raios. However, he proacive deb managemen of fiscal auhoriies aimed a reducing he vulnerabiliy of he deb composiion has been he dominan facor in quaniaive erms in mos of hese counries. Clearly, a favourable exernal environmen and exchange rae evoluions have faciliaed his process, since expeced exchange rae appreciaion favours issuing deb in domesic currency. The developmen of local deb markes has boh benefied from and faciliaed his proacive deb managemen. The changes in deb srucure are expeced o have imporan implicaions for he reducion of financial vulnerabiliy in public finances. However, our approach o his issue has uncovered a paradox relaed o he recen bias owards local deb. By reducing forex deb hrough proacive policies, governmens have no aken full advanage of he real exchange appreciaion enjoyed by heir economies afer he crises. Oherwise, he deb raios in he analysed counries would have been lower han hey are currenly and he difference is sizeable in cerain cases. This opporuniy cos underscores he dramaic shif in deb managemen sraegies in mos of hese counries. In he pas, governmens used periods of benign exernal financial condiions o issue exernal deb, usually beyond wha would be advisable and pruden from a fiscal poin of view, and have hus se he sage for fuure financial problems and crises. They are now prepared o refrain from his empaion and even o dismiss par of he impac of he exchange rae appreciaion on he deb raios in he shor erm in order o srenghen heir underlying financial posiion. This change in sraegy hus conribues o a srucural reinforcemen of public finances and helps counries redeem hemselves from he original sin. Fiscal auhoriies and analyss mus ake ino accoun his shor-erm opporuniy cos of shifing owards local deb so ha hey can assess he ne benefis of proacive deb managemen policies. The sress ess sugges ha even afer conrolling for hese shorerm coss, here is a reducion in vulnerabiliy derived from he proacive shif owards local deb, measured in mos cases by he difference in he raio of deb o GDP in a siuaion of sress, alhough he magniude in raio for some of he counries analysed is small. In inerpreing he resuls, we have o ake ino accoun some qualificaions. Firs, here is an imporan cavea ha reinforces hese resuls. Buil ino his exercise is a cenral assumpion: he evoluion of he variables ha drive he deb-o-gdp raio is independen of he raio or BIS Papers No

21 srucure of deb. However, he behaviour of he financial variables is very much influenced by percepions of deb vulnerabiliy, a siuaion ha is rue boh in he counerfacual and in he sress ess. More precisely, wih a higher share of forex deb, he exchange rae appreciaions would presumably have been lower in he laer years and he deerioraion of he financial variables in he sress es would have been higher. Anoher qualificaion is ha he probabiliy of urbulence is expeced o increase under a deb srucure very sensiive o financial volailiy. These caveas aken ogeher imply ha he esimaed reducion in vulnerabiliy is a minimum boundary, and herefore ha he effecive vulnerabiliy reducion is higher. On he whole, we showed ha he move o local deb is posiive in erms of vulnerabiliy reducion. This finding can be considered an imporan breakhrough in emerging markes and used o improve heir resilience in he face of evenual financial shocks as well as o reduce he shocks occurrence. References Alfaro, L and F Kanczuk (2006): Sovereign deb: Indexaion and mauriy, Iner-American Developmen Bank (IADB) Working Paper no 560. De Bolle, M, B Roher and I Hakobyan (2006): The level and composiion of public secor deb in emerging marke crises, IMF Working Paper 06/186. Calvo, G, A Izquierdo and E Talvi (2002): Sudden sops, he real exchange rae and fiscal susainabiliy: Argenina s lessons, NBER Working Paper Celasun, O, X Debrun and J Osry (2006): Primary surplus behaviour and risks o fiscal susainabiliy in emerging marke counries: a fan-char approach, IMF Working Paper 06/67. Cowan, K, E Levy-Yeyai, U Panizza and F Surzenegger (2006): Sovereign deb in he Americas: new daa and sylized facs, IADB Working Paper no 577. Eichengreen, B and R Hausmann (1999): Exchange raes and financial fragiliy, in Federal Reserve Bank of Kansas Ciy (ed.), New Challenges for Moneary Policy, pp García, M and R Rigobon (2004): A risk managemen approach o emerging markes sovereign deb susainabiliy wih an applicaion o Brazilian daa, NBER Working Paper series no Hosland, D and P Karam (20): Assessing deb susainabiliy in emerging marke economies using sochasic simulaion mehods, IMF Working Paper /226. Iner-American Developmen Bank (IADB) (2006): Living wih deb. How o limi he risks of sovereign finance, in E. Borenszein, E. Levy and U. Panizza (coords), Economic and social progress in Lain America: 2007 repor. Inernaional Moneary Fund (IMF) (2003): Exernal deb saisics: Guide for compilers and users, June. ( ): Aricle IV Saff Counry Repors, various, (2006): World Economic Oulook (Financial Sysems and Economic Cycles), Sepember. LainFocus Consensus Forecas (2006), July. Jeanneau, S and C Tovar (2006): Domesic bond markes in Lain America: achievemens and challenges, BIS Quarerly Review, June. 108 BIS Papers No 36

22 Manasse, P and N Roubini (20): Rules of humb for sovereign deb crises, IMF Working Paper /42. Tovar, C (20): Inernaional governmen deb denominaed in local currency: recen developmens in Lain America, BIS Quarerly Review, December. BIS Papers No