Country Risk, Macroeconomic Fundamentals and Uncertainty in Latin American Economies

Transcription

1 Ali Acosa Daniel Barráez Danyira Pérez Mariana Urbina Counry Risk, Macroeconomic Fundamenals and Uncerainy in Lain American Economies Absrac This paper analyzes he relaion beween he counry risk and is macroeconomic deerminans for Argenina, Brazil, Mexico and Venezuela, during he period, using a Markov-swiching sur model esimaed by Bayesian echniques. Two independen regimes for each counry were idenified. The firs one, associaed wih periods of sabiliy and favorable inernaional condiions, in which he variables under consideraion behave as repored in he lieraure. On he oher hand, he second regime emporarily coincides wih periods of high domesic and inernaional uncerainy. Our findings sugges ha he changes in he analyzed relaion depend on he origin of he uncerainy. If he uncerainy s source is associaed wih exernal shocks, such as inernaional crises, he financial markes volailiy gains relevance, while he solvency and liquidiy variables are less relevan; if he causes of uncerainy are domesic, he laer are he key variables o explain he sovereign risk. A. Acosa <aliacos@bcv.org.ve>, Banco Cenral de Venezuela; D. Barráez <dbarraez@bcv.org.ve>, Banco Cenral de Venezuela and Universidad Cenral de Venezuela; D. Pérez <daperez@bcv.org.ve>, Banco Cenral de Venezuela; and M. Urbina <marianaurb@gmail. com>, Universiy of Minnesoa. Monearia, July-December, 2015

2 Keywords: counry risk, emerging economies, macroeconomic fundamenals, Markov-swiching regime, Gibbs sampler. jel classificaion: C11, C15, C21, F INTRODUCTION The impac of macroeconomic fundamenals on sovereign defaul risk has been sudied in he radiional lieraure (Sachs, 1985; Edwards, 1986; González-Rozada, 2006; Uribe and Yue, 2006; Hilscher and Nosbusch, 2010) hrough linear models. Ineres has grown recenly in exploring nonlinear relaions beween sovereign defaul risk, is macro deerminans and global variables, in differen ypes of economies. In advanced economies, discussion on he high levels reached by sovereign deb and he susainabiliy of fiscal policy, have demonsraed he imporance of a nonlinear relaion beween deb size and borrowing coss, as well as he nonlineariies caused by uncerainy regarding he ype of economic policy coordinaion designed o address he deerioraion in fiscal accouns. Troy e al. (2010) sudied he consequences of rising public deb in developed counries in an environmen wih a fiscal limi, concluding ha uncerainy regarding he way economic policies are combined generaes a nonlinear relaion beween deb and inflaion. Huixin (2012) presens a sudy on he ineracions beween sovereign risk premia and fiscal policy under condiions of fiscal limi in developed counries, finding a nonlinear relaion beween sovereign risk premia and he level of public deb in line wih he empirical evidence. Greenlaw (2013) analyzed he ipping poins of sovereign deb markes for 20 advanced counries during he period The auhors find evidence of nonlineariies in he relaion beween borrowing raes on sovereign deb and is proporion on GDP of he economies sudied. These auhors poin ou ha sovereign ineres raes rise much more quickly when deb levels are high. In emerging economies, linear models have encounered some difficulies in explaining he evoluion of counry risk 148 Monearia, July-December, 2015

3 over he las wo decades in erms of heir macroeconomic fundamenals and global variables due o facors such as poliical or economic uncerainy, conagion effecs, among ohers. Acosa, Barráez and Urbina (2014) proposed a Markov regime swiching model (Hamilon, 1989) o sudy he case of Venezuela. These auhors sugges ha in he process of forming expecaions regarding a counry s capaciy o pay is debs, agens will no weigh he differen macroeconomic deerminans consanly over ime. They idenified wo emporary regimes in which he linear relaion beween he fundamenals and sovereign defaul risk clearly varies regarding he emporal regime. These regimes emporarily coincide wih periods of high and low economic uncerainy. A number of he research papers menioned above poin o uncerainy as one possible cause of he nonlinear relaion beween sovereign deb and fundamenals. This empirical work sudies he break in he linear relaion beween sovereign defaul risk and is deerminans (macroeconomic fundamenals and global variables) for Argenina, Brazil, Mexico and Venezuela during he period. To his end, we focus on exploring how uncerainy influences his break, according o he even ha generaes i, wheher i is domesic (corresponds o specific evens in each economy) or exernal (linked o inernaional ype evens). For his purpose, a Markov swiching regime model is implemened, which unlike ha he one presened by Acosa, Barráez and Urbina (2014), has a Seemingly Unrelaed Regression (sur) srucure esimaed wih Bayesian simulaion echniques (Kim and Nelson, 1999). The proposed model allows emporary saes or regimes o be specified for each counry, while carrying ou he esimaaion joinly, his enables o exploi he correlaion ha migh exis beween he shocks o he sovereign defaul risk in each counry. This paper verified he presence of nonlineariies beween sovereign defaul risk and is deerminans, and idenified wo emporary regimes in each counry similar o ha repored by Acosa, Barráez and Urbina (2014): a firs regime, linked o A. Acosa, D. Barráez, D. Pérez, M. Urbina 149

4 periods of relaive sabiliy or low uncerainy, where he relaion beween counry risk and fundamenals behaves according o he repored in he lieraure, and a second regime associaed o periods of high uncerainy. The mos significan finding of his research demonsraes ha he changes occured in he relaion beween he risk and is explanaory variables depends on he causes of uncerainy in boh regimes. If he source of uncerainy is associaed wih exernal evens, such as inernaional crises, financial marke volailiy gains relevance, while solvency and liquidiy variables are less relevan, such as in he case of Mexico and Brazil. If he causes of uncerainy are of domesic origin, he opposie occurs, such as observed in Argenina and Venezuela. In he case of he laer, he resuls coincide wih he findings of Acosa, Bárraez and Urbina (2014). I is imporan o menion ha he subprime crisis is he only common shock in he high uncerainy regime for all he counries, excep Brazil, in whose case he relaion of sovereign defaul risk wih is deerminans remained in he sabiliy regime. Such behavior is probably explained by economic policy measures adoped in his counry o address he crisis. From he model esimaed in his paper we obained counry risk elasiciies wih respec o heir deerminans in each regime. These elasiciies are useful because hey allow for assessing economic policies aimed a reducing sovereign defaul risk. The paper is divided as follows. Secions 2 and 3 describe he main aspecs of he daa and economeric mehodology. The fourh secion presens he empirical model esimaed. The fifh shows he resuls. Finally, he conclusions are given. 2. DATA The daabase used for he esimaion conains quarerly informaion for he period , for Argenina, Brazil, Mexico and Venezuela. The embi+ index calculaed by jp Morgan, obained from Bloomberg, was used as a measure of sovereign defaul risk for 150 Monearia, July-December, 2015

5 each counry included in he sudy. The variables considered as counry risk deerminans are divided ino hree groups: macroeconomic fundamenals, solvency and liquidiy variables, and global indicaors. The firs group consiss of he growh rae of real gdp, inflaion and exchange rae variaions. The second group includes inernaional reserves, commodiy prices and exernal deb as percenage of gdp. The hird group of variables includes global indicaors, such as marke volailiy and inernaional ineres raes. In he case of macroeconomic fundamenals, daa for gdp growh rae, inflaion and he exchange rae are aken from imf saisics for Argenina, Brazil and Mexico. In he case of Argenina, he price index regisered by PriceSas was also used. In he case of Venezuela, hese variables were obained from Cenral Bank saisics, excep for he parallel marke exchange rae employed for calculaing he spread wih respec o he official rae as a measure of exchange rae imbalance, which is obained from differen sources. Regarding liquid and solvency indicaors, inernaional reserves were expressed as monhs of impors obained from imf saisics. Daa relaed o exernal deb was obained from he saisics of he respecive minisries of finance and saisics insiuions of each counry. This variable was expressed as a percenage of gross domesic produc. Wih respec o global indicaors, marke volailiy is capured hrough he Chicago Board Opions Exchange Volailiy Index (vix). The 3-monh Unied Saes Treasury bill rae, obained from Federal Reserve saisics, was used as he measure of inernaional ineres raes. Commodiy prices were incorporaed via he commodiy price index, obained from he imf for Brazil and Mexico. The commodiy price index published by he Banco de la República Argenina was used for Argenina, while for Venezuela he price series of he Venezuelan oil baske, obained from he Venezuelan Energy and Oil Minisry, was employed. These crieria for selecing he indexes were based on he srucure of expors, aking ino accoun he mos represenaive commodiies of each counry. A. Acosa, D. Barráez, D. Pérez, M. Urbina 151

6 Before saring he esimaion, uni roo ess were carried ou o deec saionariy in he series. Thus, he es of Levin, Lin and Chu (o verify he exisence of common uni roo processes), and hose of Pesaran and Shin, W-Sa, adf Fisher and pp Fisher, were employed o prove he exisence of individual uni roo processes. All he variables were ransformed ino logarihmic differences, excep he coefficiens (exernal deb/gdp, reserves/impors) and he ineres rae, which are assumed a saionary in levels. Selecion of hese economies was made considering he mos represenaive Lain American counries in erms of he size of he economies for which he embi+ is elaboraed (jp Morgan calculaes he embi+ for 16 counries, six of which belong o Lain America). The sudy period was chosen aking ino accoun he availabiliy of saisical daa. 3. METHODOLOGY The muliple srucural changes in Lain American economies would seem o sugges ha a linear model for explaining defaul risk for each of he counries considered would be an inappropriae simplificaion. Thus, nonlinear Markov regime swiching models seem o be more appropriae for adjusing o his ype of behavior. The insabiliy in regression models is frequenly associaed o changes experienced by he equaion s parameers from one sample period (regime) o anoher. If knowledge is available on when such regime changes occur, and he subgroups of he sample are well defined, he Chow F-es can be applied o prove he exisence of he srucural change hypohesis. However, in many cases very lile informaion is available abou he occurrence of such srucural changes, hen, in addiion o he esimaion of he model s parameers, he srucural breaks of he equaion mus also be inferred as unobservable variables. The sur mehodology was used in order o joinly esimae he Markov regime swiching model, his provides informaion 152 Monearia, July-December, 2015

7 abouhe correlaion beween he shocks o which risk is exposed in each considered counry. The Markov-swiching sur model can be wrien as follows: 1 y = x β + e, i, i, is, i, wih = 1,, T observaions for each of he i = 1,, N equaions (counries). y i, denoes he sovereign defaul risk observaion a ime of equaion i, x i, is a 1 k i vecor ha conains he explanaory variables of equaion i a ime, β is, represens he respecive vecor coefficiens of equaion i a ime, which has he following srucure: 0 1 βs = βi ( 1 si, )+ βi si,, s i, = 0 or 1 (regime 0 or 1). s i, is he unobservable variable ha governs he regime change of equaion i, during regime 0 he parameers of his equaion are given by β i0, while during regime 1 hey would be given by β i1. Up unil now nohing has been said regarding he characerisics of random errors in he model. e = ( e e e 1,, 2,,, N, ) is defined o allow error correlaion beween cross-secion unis, we should assume ha e ~ N( 0N, Σ ) for = 1,, T, where Σ is a co-variance marix N N. Then, he likelihood funcion o maximize is given by: T ln ( L)= f( y s, s,, s, 1 = 1s1= 0s2= 0 sn = 0 1, 2, N, ψ 1 N ) f( s i, ψ ) i = 1 where 1 1 f ( y s1,, s2,,, sn,, ψ 1 )= exp y x N 1 βs 2π 2 2 Σ 2 ( ) 1 ( ) Σ 1 ( y x ) s β, x1, 0, 0 β1, s s 1, x y = ( y y yn ) x = 0 2, 0 1, 2,,, = β 2, s,, βs, s = s x N, β N, s s N, Ψ 1 represens he daa available a ime 1. Finally, an assumpion mus be imposed on he sochasic behavior of he unobservable variables s i,, which will allow o deermine f( s i, ψ 1 ). If i is assumed ha hese follow a firs-order Markov random process, he specificaion of a Markov regime swiching model A. Acosa, D. Barráez, D. Pérez, M. Urbina 153

8 will have been compleed. Inference of hese variables is carried ou hrough he Hamilon filer (1989). When he model presened in Equaion 1 depends on muliple cross-secion unis, each wih explicaive variables, he number of parameers o be esimaed increases considerably and maximizaion of he likelihood funcion expressed in Equaion 2 by classical mehods becomes a very complicaed ask. Moreover, Bayesian mehods provide several imporan advanages ha avoid difficulies relaed o numerically maximizing he likelihood funcion wih resricions on he parameers imposed by economic heory. The use of prior densiies, in addiion o including informaion no conained in he sample ino he esimaion process, allows for working wih smaller sized samples han hose required by frequenis mehods, which is of paricular ineres in our case. Regarding he esimaion echnique, he Bayesian simulaion algorihms proposed by Kim and Nelson (1998) were employed for esimaing he model. The idea is o use Gibbs sampling o obain he poserior disribuion of he parameers β 0 i, β 1 i, Σ, i = 1,, N and he sae vecors s i, from which heir mean and variance can be inferred, hereby avoiding direc maximizaion of he likelihood funcion. Gibbs sampling only requires poserior simulaion of he condiional disribuions of each parameer. Assuming a mulivariae normal prior disribuion for he vecor of parameers ( ) β = 0 β 0 β 0 β 1 β 1 β 1 β 1 2 N 1 2 N N B0, V 0 he poseriorcondiional disribuion f( βψ T, Σ, s 1,, s 2,,, sn, ) will be given by β ~ N( B1, V1 ), where: Σ B V 1 1 = V + Σ, 1 0 ( ) 1 1 1( 0 0 ) 1 1 = V V B + X Σ Y, = Σ I T ( : Kronecker produc operaor), 154 Monearia, July-December, 2015

9 Y1 yi, 1 = Y 2 = yi, 2, Y i, i = 1,,N YN yit X1 0 0 = 0 X 2 0 ( ι K S ), X N ( : elemen-wise produc operaor) X i xi, 1 S1 x i S =, 2 S = 2,, Si x it S N si, 1 = si, 2 N, K = = = ki, i 1,, N, i 1 sit ι K denoes a K 1 vecor ha only conains ones. To simulae he variance-covariance marix Σ, an inverse Wishar priori disribuion Σ ~ IW ( D 0, δ 0 ) will be assumed, where D 0 and δ 0 represen a N N scale marix and degrees of freedom, respecively. The resuling poserior condiional disribuion f( Σ ΨT, β, s 1,, s 2,,, sn, ) is of he same funcional form: Σ ~ IW ( D 1, δ 1 ), where: D1 = D0 + EE ei, 1 ei E = ( E E EN ) Ei =, 2 1 2,, i = 12,,, N, δ 1 = δ 0 + T. e it, To simulae he poserior disribuion f( S i βσψ,, T ) we use he Carer and Kohn (1994) resul, which indicaes ha: A. Acosa, D. Barráez, D. Pérez, M. Urbina 155

10 T 1 ( )= ( ) ( ) f Si β, ΣΨ, T f sit, β, ΣΨ, T f si, si, + 1, β, ΣΨ, T, i = 1,,N, = 1 where each of hese disribuions are obained by implemening he Hamilon (1989) filer [for furher deails on his resul see Carer and Kohn (1994) or Kim and Nelson (1999)]. 4. EMPIRICAL MODEL The base model esimaion is given by: EMBI, = θ + θ PIB + θ π + θ R + θ Tc i S S i S i S i S i 5 + θ Vix S + θ Ti + θ D + θ PMP + ε i S i S i S i i, S εi, s N (, 0 σ 2 s ), i i i θs = θ S θ S 0( 1 )+ 1, S = 0 o 1 (regime 0 or 1), where subindexes i and denoe he counry and he ime respecively, represens real gdp growh rae; π i he inflaion rae; R i is inernaional reserves expressed in monhs of impors; Tc is he variaion of he exchange rae; 3 Vix i is he cboe volailiy index; Ti i is he hree-monh US Treasury bill ineres rae; D i is exernal deb as percenage of gdp and is he variaion of commodiy prices. +, 5. RESULTS The resuls of he parameer esimaion are shown in Table 1 of he Annex. Two regimes were idenified for each counry. The firs regime, which we will call regime L (low uncerainy), is relaed o periods of sabiliy, economic growh and favorable inernaional condiions. The second one is he regime H (high uncerainy), which emporarily coincides wih periods of domesic and inernaional urbulence. The mehodology employed allows regimes 156 Monearia, July-December, 2015

11 o be independen beween counries and hey do no necessarily coincide in emporaliy. In regime L, for Argenina, Brazil and Mexico all of he deerminans considered are saisically significan and he signs of he coefficiens were as expeced, excep for he gdp growh rae in he case of Argenina. Ou of he macroeconomic fundamenals, gdp growh rae has a negaive sign, while inflaion and he exchange rae have posiive ones. Ou of solvency and liquidiy variables, deb has a posiive sign, while inernaional reserves and commodiy prices have negaive ones. Ou of he global variables, he vix has a posiive sign. In his regime, he counries sovereign defaul risk behaves as expeced in he lieraure. Unlike he res of he counries, in Venezuela during regime L, risk is mainly deermined by a small number of variables, being he mos imporan oil prices and financial marke volailiy, confirming he resuls obained by Acosa, Barráez and Urbina. This finding reflecs he imporance of oil revenues for he Venezuelan economy and he sensiiviy of he yield curve of deb insrumens o oil price shocks (Chirinos and Pagliacci, 2015): in periods of low uncerainy, sovereign deb risk percepion is essenially linked o oil prices. In conras o regime L, regime H emporarily coincides wih periods of high uncerainy where disurbances of inernaional scope are presen, such as he Russian crisis, he Argenine deb crisis and he subprime crisis, in addiion o domesic evens ha adversely affeced risk premia. In he case of Mexico and Brazil, he periods of high uncerainy are mainly associaed wih major exernal disurbances, while in Argenina and Venezuela his regime basically coincides wih domesic ype evens. Now we are going o analyze he resuls for each of counry in regime H. In Mexico regime H is observed during he periods 1998Q1-1998Q3 and 2007Q2-2009Q2, coinciding wih he Russian and subprime crises, respecively. Negaive economic growh raes, depreciaion of he Mexican peso and increased sovereign defaul risk were recorded in boh periods. A. Acosa, D. Barráez, D. Pérez, M. Urbina 157

12 Wih respec o he laer period, i is imporan o menion ha ou of he counries of he region, Mexico was mos affeced due o he synchronizaion of is business cycle wih ha of he Unied Saes. Wih respec o he coefficiens of regime H, inflaion, gdp and he exchange rae cease o be significan. The inernaional reserves and vix coefficiens are greaer han hose esimaed for his variables in regime L. Unlike regime L, he sign of he coefficien for he Unied Saes Treasury bill raesis negaive in regime H, which reveals he imporance of Unied Saes moneary policy in invesors valuaion of Mexican deb. In he case of Brazil, regime H, covering he period 2002Q2-2004Q2, was characerized by a significan deerioraion in he erms of rade due o a decline in rade flows wih Argenina as a consequence of he deb crisis affecing ha counry. In addiion, he burs of he speculaive bubble in 2000 and he evens of Sepember 2001 generaed volailiy in inernaional markes. This unfavorable inernaional environmen caused a slowing of economic aciviy. During his period sovereign bond spreads surpassed 2,000 basis poins (bp) and he real suffered a sharp depreciaion. Wih respec o he coefficiens of regime H as in he case of Mexico, a group of variables ceased o be significan: inflaion, exernal deb, gdp and Unied Saes ineres raes. The exchange rae and he vix increased heir weigh as risk deerminans compared o hose obained by regime L. While he coefficiens of inernaional reserves and commodiy prices are similar o hose of regime L, he signs of he coefficiens are as expeced a priori. During 1998Q4-1999Q2 he presence of macroeconomic imbalances were seen afer he collapse of he Plan Real, which increased risk premia. However, he mehodology employed did no associae his period wih regime H, given ha his regime depends on he behavior of global indicaors. During he subprime crisis, no change of regime was observed eiher in sovereign defaul risk for Brazil, which remained in regime L, despie he increase in he regime swiching probabiliy 158 Monearia, July-December, 2015

13 (Figure 1). Such behavior could be explained by he effeciveness of economic policy measures (mainly moneary and fiscal) miigaing he impac of he crisis. Figure 1 PROBABILITY OF REGIME SWITCHING FOR BRAZIL Probabiliy Q Q Q Q Q Q4 2008Q3 2010Q Q Q4 Source: own elaboraion. To assess wheher moneary policy in Brazil influenced he evoluion of sovereign defaul risk during he subprime crisis, a Taylor rule was esimaed. i π = r + a( π π )+ b( y y)+ ε, where i is he moneary policy ineres rae of he Banco Cenral do Brasil (selic), r is he long-erm ineres rae, π is he inflaion arge, π π is he difference beween he acual inflaion rae and he arge, y y is he gdp gap and ε is he moneary policy shock. A. Acosa, D. Barráez, D. Pérez, M. Urbina 159

14 In order o es wheher he policy mesures influenced he fac ha soveriegn defaul risk remained in he regime of low uncerainy during he crisis period, he residuals of he Taylor rule (which express he orienaion and magniude of moneary policy) were capured for esimaing a logisic model on he regime swiching probabiliy. Figure 2 shows he probabiliy of a change in counry risk regime during he implemenaion of an expansive moneary policy measure (Figure 2a), comparing i wih a counerfacual exercise assuming he implemenaion of a neural moneary policy (null shocks in he Taylor rule), which is shown in Figure 2b. In his regard, i can be seen how he regime change probabiliyduring he crisis is higher in he abscence of moneary policy measures, i.e., moneary policy conribued o say in he low uncerainy regime during he subprime crisis. I is imporan o poin ou ha fiscal policy acions were included in he logisic model, and counerfacual exercises were carried ou similarly o hose menioned above, employing variables such as ax revenues and expendiures. However, hese variables were no saisically significan, i.e., no saisical evidence was found o suppor he premise ha fiscal policy influenced he presence of he Brazilian economy in he low uncerainy regime during he crisis. On he oher hand, regarding Mexico and Brazil, inernaional reserves is only variable of he solvency and liquidiy group whose coeffcien increased during his regimeis in he case of Mexico. The oher wo of his group remained unchanged or even los heir significance, such as in he case of deb in Brazil, while he coefficien of he global variable vix is higher in his regime for boh economies. In he cases of Argenina and Venezuela, regime H consiss of hree periods, mainly associaed wih adverse domesic economic and poliical evens. In Argenina, he firs period (2001Q4-2005Q2) was marked by he public deb crisis of December 2001 and he subsequen social and poliical evens ha led o he resignaion of he hen presiden. The economy suffered a subsanial conracion during his period, accompanied by a significan fall in 160 Monearia, July-December, 2015

15 Figure 2 PROBABILITY OF BRAZIL COUNTRY RISK REGIME SWITCHING, GIVEN THE IMPLEMENTATION OF ECONOMIC POLICY MEASURES (PANEL A) OR NOT (PANEL B) PANEL A PANEL B Fuene: own elaboraion. A. Acosa, D. Barráez, D. Pérez, M. Urbina 161

16 inernaional reserves, depreciaion of he exchange rae and a cessaion of exernal public deb paymens. The second period (2008Q4-2009Q2) coincides wih he oubreak of he subprime crisis ha affeced various counries in he region. The inernaional crisis caused a slowing of economic aciviy, a decline in he erms of rade and a depreciaion of he currency in Argenina. During his period, fears in he financial markes increased wih respec o he Argenine governmen s abiliy o mee deb and ineres paymen commimens mauring in Thus, alhough he iniial disurbance was of exernal origin, i passed hrough he domesic economy, affecing fundamenals, and solvency and liquidiy variables. The hird period (2012Q1-2013Q4) was characerized by he applicaion of economic policy measures, he mos imporan ones being hose relaed o renaionalizaion of a majoriy share in he oil company Repsol ypf, foreign currency conrols on domesic operaions (mainly in he real esae secor) and he reducion of foreign currency hoarding by residens. Wih respec o he coefficiens, hose for deb and he exchange rae cease o be significan in regime H. The mos imporan changes are expressed in he size of he coefficien of commodiy prices and he consan erm, which represen almos double and riple he esimaes for regime L, respecively. This reflecs he growing imporance ha agens give o his liquidiy indicaor in response o he drop in inernaional reserves. For Venezuela, he firs period ha ook place during 1998 coincides wih he collapse of he fixed exchange rae regime and capial conrols implemened since 1994, and he sar of a sysem of exchange rae bands in July In he inernaional conex, several evens occured during his period, such as he Asian crisis in July 1997, he Russian crisis of 1998 and he fall of inernaional oil prices o hisorically low levels. During he second period (2002Q1-2003Q2), imporan evens such as he aemped coup d éa of April 2002 and he subsequen oil srike in December of he same year, which had economic and poliical repercussions. In he economic field, he subsanial fall in inernaional reserves led o he applicaion of a new fixed exchange rae regime and capial conrols. 162 Monearia, July-December, 2015

17 The hird and final period (2005Q3-2013Q4) was characerized by high risk premia deriving from domesic evens such as socialis economic iniiaives (naionalizaion of privae companies: seelmakers, cemen producers, and food processing firms, among ohers). In he inernaional conex, 2008 saw he defaul of Ecuador and he subprime crisis, which led o a conracion in he global economy, significan marke volailiy and a decline in oil prices. All he aforemenioned considerably increased Venezuela s risk premium. In regime H, Venezuela mainains is aypical behavior; all he variables are significan, excep gdp and he exchange rae. I should be menioned ha he exernal deb and inernaional reserves in regime L are no significan, while in his regime he former of hese is he variable wih he larges coefficien. The absolue values of he coefficiens for he vix and for commodiies are smaller han in regime L. The behavior of Venezuela s sovereign defaul risk obeys o he specific characerisics of an oil economy, in periods of low uncerainy agens focus on oil prices and inernaional marke volailiy o form heir risk percepions, while in periods of high uncerainy hey consider oher variables, besides hose menioned early. For Argenina and Venezuela, where he high uncerainy regime is associaed wih domesic evens, he coefficien of he vix decreased compared o hose of he regime L in boh cases. Regarding he solvency and liquidiy group of variables, only he coefficienes of oil prices for Venezuela and deb for Argenina decreaseor lose heir significance, he ohers increase heir weigh or remain he same. In Argenina s case, alhough wo differen measures of inflaion were used, none of hem were significan, regardless of he regime. The same was also observed for Venezuela. The resuls obained allow exracing some characerisics ha are common o all he economies. In general erms, he resuls sugges ha a change of regime in he relaion beween counry risk and is deerminans depends on he origin of he uncerainy. If he uncerainy s source is associaed wih exernal shocks (such as inernaional crises), financial marke volailiy gains relevance, whereas he solvency and liquidiy variables A. Acosa, D. Barráez, D. Pérez, M. Urbina 163

18 Figure 3 MARKOV REGIME SWITCHING BY COUNTRY (1998Q1-2013Q4) ARGENTINA Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q1 BRAZIL 1998 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q1 EMBI+ Regime A Regime B S Noe: lines in dark grey corresponds wih he observed EMBI+ by each counry, lines in ligh grey and he black ones correspond o he high and low uncerainy regimes, respecively. Shaded area is associaes o high uncerainy and allows observing he regime swiching easily. Source: own elaboraion. 164 Monearia, July-December, 2015

19 Figure 3 (con.) MARKOV REGIME SWITCHING BY COUNTRY (1998Q1-2013Q4) 8.0 MEXICO Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q VENEZUELA 1998 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q1 EMBI+ Regime A Regime B S Noe: lines in dark grey corresponds wih he observed EMBI+ by each counry, lines in ligh grey and he black ones correspond o he high and low uncerainy regimes, respecively. Shaded area is associaes o high uncerainy and allows observing he regime swiching easily. Source: own elaboraion. A. Acosa, D. Barráez, D. Pérez, M. Urbina 165

20 are less relevan.while, if he riggers of uncerainy are of domesic origin, he laer are he key variables. On he oher hand, o assess he robusness of he base model, alernaive models were esimaed ha ook ino consideraion oher conrol variables such as he degree of openness, governmen effeciveness, poliical sabiliy/absence of violence and regulaory qualiy. The firs is measured as he raio beween oal impors plus oal expors and gdp, while he res of hem are indexes prepared by he World Bank. The model specificaionha includes degree of openness is he same as in he baseline model, bu excludes he raio of inernaional reserves o impors and he exernal deb as a percenage of gdp because of problems of collineariy. The resuls of he esimaion of his model are shown in Table 2 of he Annex. I can be seen ha he model is robus afer his variable is incorporaed given ha he regime changes regisered and he majoriy of he parameers do no change significanly compared o he baseline model. This measure of openness was significan for Argenina and Brazil, wih a posiive sign in regime H and a negaive one in regime L for boh counries. This indicaes ha he more open hese economies are during periods of high uncerainy, he more sovereign defaul risk is affeced due o fears of conagion. Wih respec o he oher variables considered for esimaing he alernaive models, none of hem were significan excep for governmen effeciveness in he case of Argenina in regime H, wih a negaive sign as would be expeced. The model esimaed allows for sovereign defaul risk elasiciies o be derived wih regard o heir deerminans in each regime. By simulaing percenage increases in he respecive exogenous variable, he resuling percenage variaions in he endogenous variable are couned in order o obain he desired elasiciy. These elasiciies, shown in Table 3 of he Annex, are useful for elaboraing policies aimed a miigaing he impac of crises on sovereign defaul risk. The Table shows, for insance, how an increase of 1% in he exchange rae leads o an increase of 0.49% in sovereign defaul risk for he case of Mexico in he low uncerainy regime. 166 Monearia, July-December, 2015

21 6. CONCLUSIONS The resuls of his research poin o he fac ha he relaion beween sovereign defaul risk and is deerminans for he counries considered has been disurbed by differen ypes of evens. In he inernaional conex, hese evens are relaed o he economic and financial crises ha occurred during he sudy period: Russian crisis, Argenine deb crisis and he subprime crisis. In he domesic environmen, hese evens are linked o macroeconomic imbalances, poliical insabiliy and socialconflics. The nonlineariy associaed wih his behavior was capured by esimaing a Bayesian Markov-swiching sur model. This mehodology allowed wo independen regimes o be idenified for each counry. The firs regime, named regime L (low uncerainy), is relaed o periods of sabiliy, economic growh and favorable inernaional condiions. The second, regime H (high uncerainy), emporarily coincides wih periods of inernaional and domesic urbulence. The resuls sugges ha in he period of high uncerainy, agens give more imporance o some key variables for forming heir risk expecaions. Such variables depend on he causes of he uncerainy. If he source of uncerainy is associaed wih exernal evens, such as inernaional crises, financial marke volailiy becomes imporan, such as in he case of Mexico and Brazil. If he riggers of uncerainy are of domesic origin, he key variables are he liquidiy and solvency indicaors of he counry in quesion, as observed in Argenina and Venezuela. In he case of Venezuela, he resuls coincide wih he findings of Acosa, Barráez and Urbina (2014), despie he differences wih respec o he frequency of he saisical daa used. I should be poined ou ha he subprime crisis is he only common even in regime H for all he economies, excep Brazil, in whose case he relaion beween sovereign defaul risk and is deerminans remained sable in regime L as a resul of effecive economic policy measures (mainly moneary and fiscal). A. Acosa, D. Barráez, D. Pérez, M. Urbina 167

22 Annex A Table 1 Counry Coefficien Poserior mean RESULTS OF ESTIMATES FOR THE BASE MODEL Regime H Regime L Poserior 90% confidence bands Poserior mean Poserior 90% confidence bands poserior Consan (12.52;14.57) 7.19 (4.36; 10.01) gdp ir/i 0.28 ( 0.40; 0.14) ed/gdp 2.85 (2.31; 3.38) Argenina vix 0.38 (0.27; 0.49) 0.46 (0.37; 0.54) Treasury Bill 0.21 (0.16; 0.26) 0.15 ( 0.17; 0.12) ipmp 1.35 ( 1.46; 1.23) 0.64 ( 0.86; 0.41) er 2.29 (1.66; 2.91) Consan 5.81 (2.40; 9.07) 7.95 (7.76; 8.13) π 3.23 (2.27; 4.22) gdp 1.21 ( 1.91; 0.51) Brazil ir/i 0.21 ( 0.31; 0.11) 0.22 ( 0.24; 0.20) ed/gdp 2.24 (1.53; 2.96) vix 0.66 (0.41; 0.92) 0.27 (0.23; 0.31) 168 Monearia, July-December, 2015

23 Treasury bill 0.06 ( 0.07; 0.05) ipmp 0.60 ( 0.89; 0.29) 0.59 ( 0.62; 0.56) er 0.66 (0.38; 0.93) 0.28 (0.18; 0.38) Consan 4.98 (3.78; 6.19) 4.44 (4.23; 4.66) π 3.77 (3.06; 4.49) gdp 1.15 ( 1.75; 0.56) ir/i 0.88 ( 1.41; 0.36) 0.31 ( 0.42; 0.19) Mexico ed/gdp 1.78 (0.58; 3.01) 3.18 (2.82; 3.53) vix 0.85 (0.65; 1.05) 0.36 (0.30; 0.41) Treasury bill 0.10 ( 0.14; 0.06) 0.03 (0.02; 0.04) ipmp ( 0.59; 0.24) 0.30 ( 0.34; 0.26) er 0.49 (0.31; 0.66) Consan 7.37 (7.01; 7.71) 6.24 (5.80; 6.67) gdp ir/i 0.18 ( 0.26; 0.10) ed/gdp 0.82 (0.40; 1.23) Venezuela vix 0.31 (0.25; 0.34) 0.56 (0.46; 0.66) Treasury bill 0.31 ( 0.32; 0.29) 0.03 (0.007; 0.04) ipmp 0.29 ( 0.33; 0.24) 0.44 ( 0.49; 0.37) er 0.24 (0.11; 0.36) A. Acosa, D. Barráez, D. Pérez, M. Urbina 169

24 Table 2 RESULTS OF ESTIMATIONS FOR THE MODEL INCORPORATING DEGREE OF OPENNESS Counry Coefficien Poserior mean Argenina Brazil Regime H Regime L Poserior 90% confidence bands Poserior mean Poserior 90% confidence bands Consan (12.81; 14.04) 9.54 (8.47; 10.57) gdp vix 0.36 (0.25; 0.48) 0.8 (0.71; 0.89) Treasury bill 0.22 (0.17; 0.28) 0.18 ( 0.20; 0.16) ipmp 1.47 ( 1.61; 1.34) 0.95 ( 1.18; 0.71) er 0.1 ( 0.19; 0.01) 1.04 (0.20; 1.88) Degree of openness 1.01 (0.01; 2.02) 2.83 ( 3.83; 1.84) Consan 4.23 (3.32; 5.13) 6.46 (6.03; 6.90) π 8.09 (5.55; 10.61) gdp vix 1.12 (0.91; 1.32) 0.33 (0.25; 0.40) Treasury bill 0.08 ( 0.12; 0.04) 0.03 (0.008; 0.04) ipmp 0.35 ( 0.57; 0.11) 0.51 ( 0.59; 0.41) er 0.53 (0.27; 0.80) 0.32 (0.10; 0.53) Degree of openness 3.76 (0.00; 7.70) 1.54 ( 2.77; 0.32) 170 Monearia, July-December, 2015

25 Mexico Venezuela Consan 4.87 (3.70; 6.06) 4.83 (4.52; 5.11) π 9.27 (6.10; 12.31) 8.93 (7.94; 9.86) gdp ( 1.67; 0.22) vix 0.41 (0.27; 0.55) 0.54 (0.46; 0.60) Treasury bill 0.08 ( 0.12; 0.04) 0.01 (0.00; 0.02) ipmp ( 0.53; 0.34) er 0.82 (0.28; 1.37) - - Degree of openness Consan 7.88 (7.61; 8.13) 5.73 (5.17; 6.30) gdp vix 0.25 (0.19; 0.30) 0.69 (0.55; 0.81) Treasury bill 0.33 ( 0.34; 0.31) 0.02 (0.00; 0.04) ipmp 0.38 ( 0.44; 0.31) 0.41 ( 0.53; 0.29) er (0.09; 0.43) Degree of openness A. Acosa, D. Barráez, D. Pérez, M. Urbina 171

26 Argenina Brazil Table 3 COUNTRY RISK ELASTICITIES RELATIVE TO THEIR DETERMINANTS Percenage change of embi + afer a 1% increase in he variable Counry Variable Regime H Regime L gdp Ir/I 0.28 d/gdp 0.93 vix Treasury bill ipmp er 2.32 cpi 3.23 gdp 1.21 Ir/I d/gdp 0.41 vix Treasury bill Bren er Monearia, July-December, 2015

27 Mexico Venezuela cpi 3.76 gdp 1.15 Ir/I d/gdp vix Treasury bill Bren er 0.49 gdp Ir/I 0.17 d/gdp 0.14 vix Treasury bill Bren c 0.24 A. Acosa, D. Barráez, D. Pérez, M. Urbina 173

28 References Acosa, Ali, Daniel Barráez and Marín Urbina (2014), Un modelo no lineal para el riesgo país en Venezuela, paper presened a cemla s Researchers Nework, Mexico, November Chirinos-Leañez, Ana María, and Carolina Pagliacci (2015), Macroeconomic Shocks and he Forward Yield Curve: How Imporan is Moneary Policy?, Macroeconomics and Finance in Emerging Marke Economies, Vol. 8, No. 3, pp Davig, Troy, Eric M. Leeper and Todd B. Walker (2011) Inflaion and he Fiscal Limi, European Economic Review, Vol. 55, No. 1, pp Edwards, Sebasian (1986), The Pricing of Bonds and Bank Loans in Inernaional Markes: An Empirical Analysis of Developing Counries Foreign Borrowing, nber Working Paper Series, no Eichengreen, Barry, and Ashoka Mody (2000), Wha Explains Changing Spreads on Emerging Marke Deb: Fundamenals or Marke Senimen?, in S. Edwards (ed.), Capial Flows and he Emerging Economies: Theory, Evidence and Conroversies, Universiy of Chicago Press. González Rozada, Marín, and Eduardo Levy Yeyai (2006), Global Facors and Emerging Marke Spreads, iadb Working Paper, No Greenlaw, David, James D. Hamilon, Peer Hooper and Frederic S. Mishkin (2013), Crunch Time: Fiscal Crises and he Role of Moneary Policy, nber Working Paper Series, No Hamilon, James D. (1989), A New Approach o he Economic Analysis of Nonsaionary Time Series and he Business Cycle, Economerica, Vol. 57, No. 2, March, pp Hilscher, Jens, and Yves Nosbusch (2009), Deerminans of Sovereign Risk: Macroeconomic Fundamenals and he Pricing of Sovereign Deb, Kamakura Corporaion. Huixin, Bi (2012), Sovereign Defaul Risk Premia, Fiscal limi, and Fiscal Policy, European Economic Review, Vol. 56, No. 3, pp Kim, Chang-Jin, and Charles R. Nelson (1999), Sae-space Models wih Regime Swiching: Classical and Gibbs-sampling Approaches wih Applicaions, mi Press, Cambridge, Massachuses. Pagliacci, Carolina, and Daniel Barráez (2010), A Markov-Swiching Model of Inflaion: Looking a he Fuure during Uncerain Times, Análisis Económico, Vol. xxv, No. 59, pp Sachs, Jeffrey, and John Williamson (1985), Exernal Deb and Macroeconomic Performance in Lain America and Eas Asia, Brookings Papers on Economic Aciviy, Vol. 16, No. 2, pp Uribe, Marín, and Vivian Z. Yue (2006), Counry Spreads y Emerging Counries: Who Drives Whom?, Journal of Inernaional Economics, No. 69, June, pp Monearia, July-December, 2015