Banks and the Domestic and International Propagation of Macroeconomic and Financial Shocks

Transcription

1 MPRA Munich Personal RePc Archive Banks and he omesic and Inernaional Propagaion of Macroeconomic and Financial Shocks Rober Kollmann CARS, Universié ibre de Bruxelles and CPR 2010 Online a hps://mpra.ub.uni-muenchen.de/70349/ MPRA Paper No , posed 29 March :37 UTC

2 Banks and he omesic and Inernaional Propagaion of Macroeconomic and Financial Shocks Rober Kollmann CARS, Universié ibre de Bruxelles and CPR May 25, 2010 This paper incorporaes a bank ino a dynamic sochasic general equilibrium model. The bank collecs deposis and makes loans o an enrepreneur, subjec o a regulaory bank capial requiremen. The presence of he bank dampens he response of real aciviy o TFP shocks, bu i magnifies he effec of credi losses. An unanicipaed credi loss reduces he bank s capial, which raises he spread beween loan and deposi raes, and riggers a sizable, bu shor-lived, fall in real aciviy. When he bank operaes inernaionally, hen a loan defaul shock in one counry riggers a sizable fall in boh domesic and foreign oupu. J codes: F36, F41, G21, F34 Key words: banks, inernaional business cycles, financial crisis Address: CARS, CP 114, Universié ibre de Bruxelles; 50 Av. Franklin Roosevel; B-1050 Brussels, Belgium. rober_kollmann@yahoo.com I am very graeful o Charles ngel and Gerno Müller for helpful discussions. For useful suggesions, I also hank Werner Roeger, Zeno nders and workshop paricipans a he Konsanz Seminar on Moneary Theory and Policy, and a he Insiue for Advanced Sudies (Vienna). I hank he Naional Bank of Belgium and he U Commission for financial suppor. 1

3 1. Inroducion Sandard macroeconomic models developed before he curren financial crisis absraced from banks and oher financial inermediaries. The curren financial crisis has revealed he limiaions of his class of models. The crisis was riggered by credi losses in he US morgage marke. These credi losses lowered he capial of US and foreign banks acive in he US marke, hus leading o an increase in he credi spreads, and a persisen fall in real aciviy world-wide. This paper presens a SG model wih banks ha accouns for hese phenomena. I consider a closed economy, before analyzing a wo-counry world. There are hree (represenaive) agens: (i) a household ha works and invess her savings in bank deposis; (ii) a banker who lends o an enrepreneur; (iii) he enrepreneur accumulaes capial and produces a final good (using capial and labor). eposis provide liquidiy services o he household. The bank faces a regulaory capial requiremen, and hus parially finance loans using own funds (equiy). Hence, he loan rae exceeds he deposi rae. The ineres spread is a decreasing funcion of he bank s excess capial (i.e. of bank capial held in excess of he mandaory level). In he srucure here, an unanicipaed credi loss lowers he bank s capial, and raises he loan/deposi rae spread. ssenially, an unanicipaed fall in he bank s wealh worsens he financial fricion, which leads o a fall in invesmen, employmen and oupu. In calibraed model versions, he deposi rae falls, in response o he credi loss, and household consumpion rises; his raises he wage, and riggers a fall in employmen and oupu, and a fall in invesmen. By conras, in a model varian in which households direcly lend o enrepreneurs (wihou using financial inermediaries), a credi loss has (virually) no effec on he loan rae, and oupu and invesmen change much less. Numerical simulaions sugges ha he magnificaion of he real effecs of credi losses, due o financial inermediaion, can be sizable. However, financial inermediaion dampens he response of oupu and invesmen o produciviy (TFP) shocks. A posiive TFP shock raises household income, and hus he household holds more deposis, i.e. he bank s excess capial falls. This riggers a widening of he loan/deposis ineres rae spread, which dampens he expansion of 2

4 lending, compared o a seing wih fricionless lending, and explains he more mued response of invesmen and oupu. The wo-counry varian of he model assumes a global bank: he bank collecs deposis from local and foreign households, and makes loans o local and foreign enrepreneurs. Credi losses in one counry rigger a world-wide widening of loan/deposi rae spreads, and a world-wide fall in lending and oupu. The effec on real aciviy is very similar across counries. To be added: discussion of relaed lieraure. Goodfriend and McCallum (2007); Van den Heuvel (2008). Recen quaniaive closed economy SG models wih banks: de Walque, Pierrard and Rouabah (2010); Gerali, Neri, Sessa and Signorei (2010); Roeger (2009). Value added here: emphasis on ransmission of credi loss shock; analyical resuls. Open economy: evereux, Yeman (2010) assume inernaional invesors subjec o leverage consrain, hard o inerpre as banks; simpler echnology (eg no capial accumulaion) ifference: my paper assumes banks, focus on credi losses, full business cycle model 2. The closed economy model The closed economy model assumes hree (represenaive) infiniely-lived agens: a household, a bank and an enrepreneur. There is a final good ha is used for consumpion (by each of he hree agens), and for capial accumulaion (by he enrepreneur). All agens are price akers. The household The household consumes he final good, provides labor o he enrepreneur and invess her savings in bank deposis. Her dae budge consrain is: where C and C W N R, (1) 1 W are consumpion and he wage rae, respecively (he final good is used as numéraire). N are hours worked. 1 are he bank deposi held by he household, a he end of period. a -1). R is he gross ineres rae on deposis, beween -1 and ( R is se 3

5 The household s expeced life-ime uiliy a dae is: s N [ u( C ) ( 0 s u 1 s) N s} s, (2) wih N 1, 0; u( x) ( x 1) /(1 ), wih 0 is an increasing and concave funcion. The household maximizes (2) subjec o he resricion ha her period-budge consrain holds a and a all subsequen daes. Ruling ou Ponzi schemes, he household decision problem has hese firs-order condiions: R u'( C )/ u'( C ) u'( )/ u'( C ) 1, (3) N u'( C ) W. (4) The bank In period, he bank receives deposis 1 and she makes a (one-period) loan 1 o he enrepreneur. The bank faces a capial requiremen: her dae capial 1 1 should no be smaller han a fracion of asses 1. A capial requiremen of his form can eiher represen a legal requiremen (Basel II), bu i migh also resul from pressure by deposiors (o ensure bank solvency). I assume ha he bank can hold less capial han he required level, bu ha his is cosly (e.g. because he bank hen has o engage in creaive accouning). e x ( 1 1) 1 (1 ) 1 1 denoe ha bank s excess capial a. The bank bears a cos ( x ) as a funcion of x, wih (0) 0 and ' 0, '' 0. Hence, ha cos is decreasing and sricly convex. When he bank sricly mees is capial requiremen, hen he cos is zero (a posiive cos only arises when x 0; when x 0, hen he bank receives a benefi). A, he bank also bears an operaing cos ( 1, 1) ha is increasing and linear in deposis and loans 1, 1. The bank s period budge consrain is: where R (, ) ( (1 ) ) d R (1 ), (5) B d is he profi (dividend) generaed by he bank a. R is he gross loan ineres rae beween -1 and. 0 1 is an exogenous sochasic loan defaul rae: a, he 4

6 enrepreneur only pays back a fracion 1 of he conraced amoun R. R is se a -1. However, he effecive rae of reurn on he loan, ne of defaul, is only realized a. The banker does no have access o oher asses, and hus she consumes her dividends. Her expeced life-ime uiliy a is: s B u d s 0 s ( ).The banker maximizes life-ime uiliy subjec o curren and fuure budge consrains. Ruling ou Ponzi schemes, ha problem has hese firs-order condiions: B B R ' 1 1 u '( d 1 )/ u'( d ) 1, and (6) R (1 ) u'( d )/ u'( d ) 1 (1 ) ', (7) B B , where, and, are he marginal coss of deposis and loans, respecively and ' '((1 ) ) 1 1. By acceping more deposis a, he banker can increase her dae consumpion, a he cos of a reducion of consumpion a +1. Specifically, when he bank raises deposis 1 by 1 uni (holding consan loans), hen her capial falls by one uni, which raises by ' 0; in addiion she incurs a marginal operaing cos,. S Hence, he banker s marginal benefi of deposis (in uiliy erms) is u'( d ){1 ', }. B The discouned expeced marginal cos of deposis o he bank is R 1 1 u'( d 1). A a maximum of he bank s decision problem, he expeced marginal benefi equals he marginal cos. If he bank raises loans by one uni a (holding consan deposis), hen his lowers her dae dividend by 1 ', (1 ). The bank s effecive (gross) real rae of reurn on loans is hus R ' 1(1 1 ) /{1, (1 ) }, which explains he uler equaion (7). The enrepreneur The enrepreneur accumulaes physical capial, and she uses labor and capial o produce he final good. The law of moion of he capial sock is: K 1 (1 ) K I, (8) 5

7 where K is he capial sock used in producion a ; 0 1 is he depreciaion rae of capial, and I is gross invesmen. Final good oupu, denoed Y, is produced using a Cobb-ouglas echnology: Y K N (9) 1 ( ) ( ), wih 0 1. Toal facor produciviy is an exogenous random variable. where The enrepreneur s period budge consrain is: R K W N d K N K, (10) 1 (1 ) 1 1 ( ) ( ) (1 ) d is he enrepreneur s dividend income a. The enrepreneur consumes her dividend income. Her lifeime uiliy a is given by life-ime uiliy subjec o (10) yields hese firs-order condiions: s 0 s u( d ). Maximizaion of 1 1 ( u '( d 1)/ u'( d 1)){ 1 K 1 N 1 1 } 1, (11) R (1 ) ( u'( d )/ u'( d )) 1, (12) s W (1 ) K N. (13) Marke clearing Marke clearing for he final good requires: Y C d d I (, ) ( (1 ) ). (14) B Ineres rae spreads and bank capial Noe ha, in conras o much recen heoreical research on financial fricions (eg Kiyoaki and Moore (2007)), he model here assumes ha all agens have he same subjecive discoun facor, and ha he enrepreneur does no face a collaeral consrain. In models of he Kiyoaki-Moore ype, here are no financial inermediaries; enrepreneurs are less paien han households; enrepreneurs face a collaeral consrain for deb (enrepreneurs deb canno exceed a fracion of heir physical capial sock), which allows o ensure exisence of a saionary equilibrium. This paper assumes a bank ha faces a flexible ype of collaeral consrain (i bears a resource cos when deposis 6

8 fall below a fracion of he bank asses), bu he oher agens do no face collaeral consrains his allows o focus on he effecs of he bank capial resricion. As deposis provide liquidiy services o households, and as financial inermediaion is cosly, he deposi rae is lower han he loan rae, in he presen model. e R 1 R (1 ) be he expeced effecive gross loan rae (i.e. loan rae, ne of defaul). Up o a cerainy-equivalen approximaion, he bank s uler equaion (7) B B implies R ' 1 1 u'( d 1 )/ u'( d ) 1, (1 ). Thus (using (6)), R ' ' 1/ R 1 {1, (1 ) }/{1, }, and hence: R R '( (1 ) ) 0. (15) 1 1,, 1 1 Holding consan he marginal coss of deposis and loans (,,, ), a rise in excess bank capial 1(1 ) 1 lowers hus he (effecive) loan/deposi ineres rae spread R R (recall ha '' 0). 1 1 Up o a linear approximaion, a dae shock o he expeced (exogenous) loan defaul rae a +1, 1, has no effec on he expeced effecive loan rae R 1 observed in equilibrium, and hence no effec on consumpion, oupu, loans or deposis; such a shock only affecs he conracual loan rae R 1 (e.g. when he expeced defaul rae rises by 1 percenage poin, he conracual rises by approximaely 1%). Only unanicipaed changes in he defaul rae affec he real economy. An unanicipaed increase in he dae defaul rae, 1 0 brings abou a wealh ransfer from he bank o he enrepreneur. As shown below, such a ransfer can have a sizable effec on oupu, when he bank faces a capial requiremen. To provide inuiion for his effec, I now analyze in greaer deail he opimizing behavior of he bank. I do his for he special case where he bank has log uiliy ( 1). I is sraighforward o show ha, in ha case, he bank s dae consumpion equals a fracion 1 of her beginning-of-period (ne) wealh: B d (1 ){ R (1 ) R } ; (16) 7

9 hence, (from he budge consrain (5)), end-of-period wealh plus coss equal a fracion of beginning-of period wealh: (, ) ( (1 ) ) { R (1 ) R } Up o a linear approximaion (around seady sae loans and deposis), he lef-hand side of his expression equals 1(1 (1 ) ') 1(1 ') 1 R 1R. As R 1 (from he enrepreneur s uler equaion (12)), we have A R R R, (17) { (1 ) } Shocks in period only affec A 1 and is affeced. Hence, A 1 and unanicipaed TFP shocks: B d o he exen ha beginning-of period wealh B d only respond o unanicipaed credi losses, bu no o d d (1 ) R ( ). B B 1 1 A A R ( ) An unanicipaed credi loss lower A 1 and B d. The reducion in he banker s end-orperiod wealh (by a fracion of he credi loss) is much larger ha he reducion in consumpion (fracion 1 of he loss). To undersand why his maers for real aciviy, recall ha he loan/ deposi ineres rae spread is a decreasing funcion of excess bank capial x 1(1 ) 1. Noe ha ( ) x (1 ) A R (1 ) 1 (1 ) A The simulaions below se 0.1 and show ha A 1 and x are highly posiively correlaed in response o credi loss shocks. As an unanicipaed credi loss a dae lowers he bank s end-of-period wealh, A 1, i riggers a fall in excess bank capial x, which raises he loan/deposi ineres rae spread (his resul is robus o assuming risk aversion differen from uniy). As poined ou above, he financial fricion hus becomes more severe when an unanicipaed credi loss occurs. An unanicipaed TFP shock raises he household s wage income and hus increases her holdings of deposis. On impac, he shock has no effec on he banker s end-of-period wealh, and hus he increase in deposis lowers he bank s excess capial, hus riggering a rise in he loan/deposi ineres rae spread, which explains why (as 8

10 shown below), he presence of he bank dampens he effec of he TFP shock on real aciviy. 4. Calibraion I consider a baseline calibraion wih log uiliy, 1. The elasiciy of oupu wih respec o capial is se a 0.3. One period represens 1 quarer in calendar ime. Accordingly, I se he depreciaion rae of physical capial a (a sandard value used in quarerly models). I se seady sae TFP as 1. I se he required bank capial raio a 10% (Basel II requiremen: 8%). The calibraion assumes ha he deposi rae and he effecive loan rae (ne of defaul) are 2% and 4% per annum. The annual loan defaul rae is se a 3%, so ha he loan rae is 7.12% per annum. (The seady sae defaul rae does no affec real aciviy.) On a d quarerly basis, he seady sae ineres raes are hus: r 0.496%, r 0.985% and d d r 1.757%, respecively (where r R 1, r R 1, r R 1). I hus se he subjecive discoun facor a (as R 1). The bank s uler equaions (6),(7) imply R 1 ' and R 1 (1 ) ' ; any combinaion of marginal coss,, ' consisen wih hese condiions generaes he same firs-order dynamics. I assume ha he marginal coss, are consan across ime (and equal o seady sae values, ). I assume ha, in seady sae, he bank s excess capial is zero, and ha he enrepreneur s deb represens 20% of he physical capial sock (or 43% of annual GP). N [Cie empirical evidence.] The preference parameers, are se in a manner ha delivers (1 ) and K / Tha calibraion implies ha, in seady sae, he 1 Namely, I se and N GP =2.469/Y. N GP depends on seady sae GP ( Y ). For a N N given value of he model has a unique seady sae. affecs he scale of hours worked, oupu, consumpion, capial, invesmen, deposis and loans. The raios beween hese variables and ineres raes N N are no affeced by. Hence, he choice of (or equivalenly he choice of seady sae GP) does no affec he cyclical properies of ineres raes, deposis, loans and real aciviy. ae GP equals he sum of he hree agen s consumpion plus gross invesmen. GP corresponds also o final good oupu minus he bank s cos. 9

11 consumpions of he banker and of he enrepreneur represen 0.17% and 6.83% of GP, respecively, and ha deposis represen 54% of annual household consumpion. The simulaions below are based on a linearizaion of he model around a deerminisic seady sae. I hus have o pick a value for he second derivaive of he cos of excess bank capial (evaluaed a he seady sae). The baseline calibraion assumes GP "(0) 2/ Y. This implies ha a reducion in excess bank capial by 1% of quarerly GP seady sae GP ( Y ) raises he (quarerly) loan/deposi ineres rae spread by 20 basis poins (= ), as can be seen from (15). quivalenly, a rise in excess bank capial by 1% of annual GP raises he ineres rae spread by 3.2% per annum ( ). I assume ha TFP follows an AR(1) process: ln ln 1,, where, is whie noise. As is common in he RBC lieraure (eg King and Rebelo (1999)), I se 0.95, 2 2 (, ) (0.007). The defaul rae likewise follows an AR(1) process: (1 ). I assume 0.95, 1, 2 2 (, ) (0.01). As poined ou above, only unanicipaed shocks o he defaul rae maer for real aciviy. Hence, he variance of real aciviy induced by credi losses only depends on 2 (, ) (he persisence of defaul only maers for he behavior of he conracual loan rae R, bu i is irrelevan for he behavior of he expeced effecive loan rae R 1 R 1 (1 1) and for real aciviy). 5. Quaniaive resuls 5.1. Impulse responses Table 1 repors dynamic % responses o 1% TFP and credi defaul innovaions (he responses of excess bank reserves (x) and of deposis and loans are normalized by seady sae GP; he responses of he wage rae, consumpion, dividends, invesmen, hours and GP are normalized by seady sae values; ineres rae responses are expressed in % per annum erms). 10

12 Resuls for he baseline model Panel (a) of he Table shows responses under he baseline calibraion. As in sandard neoclassical models, a posiive TFP shock raises oupu, consumpion, invesmen and employmen. As TFP decays gradually afer he shock, he household saves more, by holding more deposis, and he bank makes more loans. The simulaions confirm he analyical resul (see above) ha, on impac, a posiive TFP shock lowers he bank s excess capial (x). In fac, he simulaion shows ha he fall in excess bank capial is persisen. Hence, he loans/deposi ineres rae spread rises persisenly. On impac, a 1% TFP shock raises he loan rae by 19 basis poins (bp), while he deposi rae increases by 14 bp. Panel (b) shows ha a 1% posiive innovaion o he loan defaul rae has a sizable, bu ransien, effec on GP. On impac GP falls by 1.30%; GP 4 quarers afer he shock rises by 0.10%. Wihin he firs year, annual GP falls by 0.68%. A 1% credi loss corresponds o 0.43% of annual GP. Bu he effec on GP is no very persisen: in he second year, annual GP falls by merely 0.02% (and here afer GP is slighly above is level wihou he shock). According o he IMF s April 2010 Global Financial Sabiliy Repor, credi losses of US banks during he curren financial crisis amoun o 6% of US GP, while credi losses of uro Area banks amoun o 5.3% of A GP. The model here predics ha a credi loss of his size generaes a fall in annual GP of abou 5%, in he firs year. On impac, he supply of loans fall sharply in response o he shock, by 3.05% of seady sae (quarerly) GP, which explains he sizable reducion in physical invesmen ha drives he fall in GP. eposis fall noiceably less, by -1.31% of seady sae GP. As a resul, he bank s excess capial falls (-1.43% of GP). Ineresingly, he expeced effecive loan rae falls in response o he credi loss, -8% bp p.a., bu he deposi rae falls more srongly, -128 bp. The loan/ineres rae spread increases hus by 120 bp. (The loan rae ha is no correced for expeced defaul rises by 281 bp). An economy wihou bank capial requiremen Panel (b) of Table 1 repors impulse responses for a model varian in which he bank does no face a capial requiremen. Specifically, I now assume ha he cos funcion of excess 11

13 bank capial x is linear in x (i.e. '' 0), which implies ha he loan/deposi ineres rae spread is independen of he socks of deposis and loans. Under his specificaion, an unanicipaed credi loss riggers a permanen (consan) rise in he enrepreneur s consumpion and a permanen fall in he bank s dividend. The bank cus lending, in order o dampen he effec of he defaul shock on her consumpion. The credi loss now has no firs-order effec on GP, invesmen, household consumpion, deposi and ineres raes. The responses o TFP shocks are qualiaively similar o hose in he baseline srucure. However, he shor run responses of deposis, loans, invesmen and oupu are somewha sronger. For example, GP rises by 1.87% on impac (compared o 1.63% in he baseline model). Inuiively, his is due o he fac ha he ineres rae spread is consan under he alernaive specificaion (in he baseline model, a posiive TFP shock raises he ineres rae spread, which dampens he increase in real aciviy). An economy wihou bank Panel (c) of Table 1 considers a model varian in which here is no bank. The household now lends direcly o he enrepreneur. (In ha model varian, I se he weigh of deposis in he household s uiliy funcion o zero, 0, as oherwise no seady sae exiss, given he assumpion ha he household and he enrepreneur have he same subjecive discoun facor.) In he No Bank case, he effecs of a TFP on real aciviy are noiceably sronger, in he shor run, compared o he baseline srucure (e.g. GP now rises by 2.03% in response o a 1% TFP shock). A 1% credi loss shock has a very small posiive effec on GP (+0.02%), which is due o he fac ha he shock lowers he household consumpion, which lowers he wage rae (see he household s firs-order condiion (4)), and raises labor demand Sochasic simulaions Table 2 repors prediced momens generaed by he model (sandard deviaions of HP filered variables, and heir correlaion wih GP). The prediced momens confirm he analysis above: he presence of a bank wih a capial consrain dampens he flucuaions 12

14 of real aciviy under TFP shocks, bu i generaes wider flucuaions in real aciviy in response o defaul shocks. The prediced sandard deviaion of GP [invesmen] under simulaneous TFP and defaul shocks is 2.03% [9.67%] under he baseline calibraion, compared o 1.69% [5.91%] in he model varian in which here is no binding bank capial consrain ( '' 0). This suggess ha he bank capial requiremen has a nonnegligible effec on business cycle behavior. Meh and Moran (2010) provide empirical evidence on empirical behavior of he raio of bank capial divided by bank asses, in he US. A a quarerly frequency ( ), ha raio has a relaive sandard deviaion of 0.43 (compared o he sandard of GP); is correlaion wih GP and bank loans are and -0.70, respecively. (All saisics discussed here and below are based on HP filered series.) In oher erms, US bank capial, normalized by asses, is couner-cyclical. As repored by Roeger (2009), in US quarerly daa ( ) he credi spread is negaively correlaed wih GP, eposis and bank loans o he privae nonfinancial business secor are posiively correlaed wih GP (0.08, 0.45). The baseline model here, wih TFP and credi loss shocks, maches he volailiy of he bank capial/bank asse raio, bu he model predics ha ha raio is pro-cylical (prediced sandard deviaion: 0.43; prediced correlaion wih GP: 0.64). When here are jus TFP shocks, he bank capial/asses raio is no volaile enough, bu counercyclical. The baseline model predics a couner-cyclical credi spread (-0.55), and procyclical deposis (0.20) and loans (0.37). Baseline model All Jus TFP Jus defaul shocks shock shock US ATA Bank capial/asses Relaive sandard dev Correl. wih GP Correlaions wih GP: Credi spread eposis oans

15 6. Two-counry version of he model I now assume a world wih wo counries. Boh counries produce and consume an idenical final good ha can coslessly be raded inernaionally. As before, I assume ha each counry is inhabied by a household and by an enrepreneur. There is one global bank (ha receives deposis in boh counries, and channels hem o enrepreneurs in boh counries). The bank acs compeiively, and hus he deposi rae and he expeced effecive loan raes are idenical across counries. 2 The only difference compared o he baseline model is ha I now assume ha he enrepreneur bears a quadraic invesmen adjusmen cos. The adjusmen cos is calibraed in such a fashion ha he model generaes a realisic relaive volailiy of invesmen--in he absence of an invesmen adjusmen cos, invesmen is exremely volaile, when here are counry-specific echnology shocks. (A small adjusmen cos is sufficien for ha purpose.) Table 3 repors impulse responses o 1% innovaions o counry 1 TFP and o he counry 1 loan defaul rae. The counry 1 TFP shock raises counry 1 GP and invesmen (by 1.31% and 3.53%, respecively, on impac), bu has basically no effec on counry 2 GP and invesmen. By conras, he 1% counry 1 defaul shock riggers falls in oupu and invesmen in boh counries; he reducions are very similar across counries; e.g., on impac GP and invesmen drop by abou 0.36% and 1.99%, respecively, in boh counries. A credi loss lowers he bank s excess capial, which raises he credi spread in boh counries; deposis and he deposi rae fall, while consumpion rises, in boh counries. This is accompanied by a rise in he wage rae, and a fall in employmen and oupu, in boh counries. The effec on (world) GP is weaker han in he closed economy. As menioned above, a 1% counry 1 credi loss corresponds o 0.43% of he counry s annual GP; his riggers a fall or domesic and foreign GP by -0.19%, during he firs year afer he shock. Thus, a credi loss of abou 5% of annual domesic GP in one counry (as observed in he US, during he curren crisis), is prediced o rigger a reducion of annual GP by 2.6%, in boh counries, during he firs year afer 2 All oher preference and echnology parameers are se a he same values as in he baseline closed economy model (he second derivaive is se a '' 2/( world GP). 14

16 he shock. In he second year afer he shock, annual world GP says below is preshock level by Hence, he effec on GP is non-negligible, bu shor-lived. Table 4 repors seleced prediced momens generaed by he wo-counry model. In a model varian wih jus credi loss shocks, oupu and invesmen are (almos) perfecly correlaed across counries. Wih jus TFP shocks, he cross-counry correlaions of oupu and invesmen are close o zero. Wih simulaneous defaul rae and FTP shocks, he cross counry oupu correlaion are 0.23 and 0.49, respecively. 7. Conclusion This paper has presened a SG model wih a bank. An unanicipaed credi loss was shown o generae a sizable, bu relaively shor-lived, recession. Wih a global bank, a loan defaul shock in one counry riggers a fall in boh domesic and foreign oupu. 15

17 Table 1. Closed economy model: % impulse responses ( periods afer shock) x r r r C d B d I N W GP (a) BASIN CAIBRATION WITH BANK 1% TFP shock % credi loss shock (b) BANK WITHOUT BINING CAPITA RQUIRMNT ( '' 0) 1% TFP shock % credi loss shock (c) NO BANK (IRCT HOUSHO NING TO NTRPRNUR) 1% TFP shock % credi loss shock

18 Noes: The Table shows % responses o 1% TFP and credi loss shocks (afer =0,1,4,8,40 quarers). Responses of excess bank capial (x), deposis () and loans () are normalized by seady sae GP (responses of deposis and loans perain o end-of-period socks). Responses of household consumpion (C), B enrepreneur s dividend ( d ), bank dividend ( d ), invesmen (I), hours worked (N), he wager rae (W) and GP are normalized by seady sae values. eposi rae ( r ), expeced effecive loan rae ne of defaul ( r ), and loan rae before defaul ( r ) are expressed in % per annum erms. 17

19 Table 2. Closed economy model: prediced momens (HP filered) TFP & defaul shock Jus TFP shock Jus defaul shock % Sd. CorrY % Sd. CorrY % Sd. CorrY (a) BASIN CAIBRATION WITH BANK x r r r C d B d I GP (b) BANK WITHOUT BINING CAPITA RQUIRMNT ( '' 0) x r r r C d B d I GP (c) NO BANK (IRCT HOUSHO NING TO NTRPRNUR) x r r r C d B d I GP

20 Noes: The Table shows prediced model saisics. % Sd: sandard deviaion in %. CorrY: correlaion wih GP. Saisics for bank capial (x), deposis () and loans () perain o series ha were normalized by seady sae GP (responses of deposis and loans perain o end-of-period socks). Saisics for household B consumpion (C), enrepreneur s dividend ( d ), bank dividend ( d ), invesmen (I), hours worked (N), he wager rae (W) and GP perain o series ha were expressed as relaive deviaions from seady sae values. eposi rae ( r ), expeced effecive loan rae ne of defaul ( r ), and loan rae before defaul ( r ) are expressed in % per annum erms. % Sd: sandard deviaion (in %); CorrY: correlaion wih GP. 19

21 Table 3. Two-counry model: % impulse responses ( periods afer shock) x r r 1 1 C1 I1 GP1 2 2 C2 I2 GP2 (a) BASIN CAIBRATION 1% shock o counry 1 TFP % credi loss shock in counry Noe: i, i, Ci, Ii, GPi: deposis, loans, household consumpion, invesmen and GP in counry I (i=1,2) Table 4. Two counry model: prediced momens (HP filered) TFP & defaul shock Jus TFP shock Jus defaul shock % Sd. Corr1&2 % Sd. Corr1&2 % Sd. Corr1&2 (a) BASIN CAIBRATION WITH BANK I GP Noe: %Sd: sandard deviaion (in %), Corr1&2: cross-counry correlaion 20

22 References evereux, M., J. Yeman, everage Consrains and he Inernaional Transmission of Shocks, Working Paper, Universiy of Briish Columbia and BIS. de Walque, G., O. Pierrard, A. Rouabah, Financial (In)Sabiliy, Supervision, and iquidiy Injecions: a ynamic General quilibrium Approach, Working Paper, Naional Bank of Belgium, forhcoming in: conomic Journal. Gerali, A., S. Neri,. Sessa, S. Signorei, Credi and Banking in a SG Model of he uro Area, Working Paper, Bank of Ialy. Goodfriend, M. and B.T. McCallum, Banking and Ineres Raes in Moneary Policy Analysis: a Quaniaive xploraion, Journal of Moneary conomics, Vol. 54, pp Meh, C. and K. Moran, The Role of Bank Capial in he Propagaion of Shocks, Journal of conomic ynamics and Conrol, Vol. 34, pp Roeger, W., The Financial Crisis 2008 in he QUST Model: Impac on urope, Working Paper, U Commission. Van den Heuvel, S., The Welfare Cos of Bank Capial Requiremens, Journal of Moneary conomics, Vol. 55,