BIS Working Papers. Macroprudential Policies in a Commodity Exporting Economy. No 506. Monetary and Economic Department

Transcription

1 BIS Working Papers o 56 Macroprudenial Policies in a Commodiy Exporing Economy by Andrés González, Rodríguez Franz Hamann and Diego Moneary and Economic Deparmen July 215 Paper produced as par of he BIS Consulaive Council for he Americas Research ework projec Incorporaing financial sabiliy consideraions ino cenral bank policy models JEL classificaion: E3, E58. Keywords: credi, leverage, financial acceleraor, business cycle, moneary policy, macro-prudenial policies, Colombia

2 BIS Working Papers are wrien by members of he Moneary and Economic Deparmen of he Bank for Inernaional Selemens, and from ime o ime by oher economiss, and are published by he Bank. he papers are on subjecs of opical ineres and are echnical in characer. he views expressed in hem are hose of heir auhors and no necessarily he views of he BIS. his publicaion is available on he BIS websie Bank for Inernaional Selemens 215. All righs reserved. Brief excerps may be reproduced or ranslaed provided he source is saed. ISS prin ISS online

3 Macroprudenial Policies in a Commodiy Exporing Economy Andrés González Franz Hamann Diego Rodríguez Absrac Colombia is a small open and commodiy exporer economy, sensiive o inernaional commodiy price flucuaions. During he surge in commodiy prices, as income from he resource secor increases oal credi expands, boosing demand for radable and nonradable goods, appreciaing he currency and shifing resources from he radable secor o he nonradable. Alhough his adjusmen is efficien, he presence of financial fricions in he economy exacerbaes he resource allocaion process hrough credi. In his phase, as oal credi expands, he appreciaion erodes he ne worh of he radable secor and booss he nonradable one, and hus credi ges concenraed in ha secor. A sudden reversal of commodiy prices causes a rapid adjusmen of resources in he opposie direcion. However, he abiliy of he radable secor o absorb he freed resources is limied by is financial capaciy. In his scenario, macroprudenial policies may help o resrain aggregae credi dynamics and hus preven or ac prudenly in anicipaion o he effecs of large oil price shock reversals. In his work we wrie a model ha accouns for hese facs and quanify he role of hree policy insrumens: shor erm ineres rae, FX inervenion and financial regulaion. We explore his issues in a DSGE model esimaed for he Colombian economy and find ha boh FX inervenion and regulaion policies complemen he shor-erm ineres raes in smoohing he business cycle by resraining credi, raising marke ineres raes and smoohing economic aciviy. However, hese addiional insrumens have undesirable secoral implicaions. In paricular, he use of hese policies implies ha credi o he radable secor dries and becomes more expensive, weakening is financial posiion, which in urn implies a sharper fall of his secor during he price reversal and a longer recovery. hese effecs, noneheless, appear o be quaniaively small according o he esimaed model. Keywords: credi, leverage, financial acceleraor, business cycle, moneary policy, macro-prudenial policies, Colombia JEL Classificaion Codes: E3, E58. he views expressed in his documen are hose of he auhors and no necessarily hose of he Banco de la República or he IMF. We are deeply graeful o Alessandro Rebucci and Enrique Mendoza, who commened his paper a he meeing of he BIS CCA Research ework in 214 and o Gianluca Benigno for sharing his insighs wih us during his visi o he Banco de la República. Also, o Paula Belrán and Joao Hernández, who assised us a differen sages of his projec. Of course, any misake in his paper is our full responsibiliy. Inernaional Moneary Fund, IMF insiue. agonzalez@imf.org Banco de la República, Deparameno de Modelos Macroeconómicos. fhamansa@banrep.gov.co Banco de la República, Deparameno de Modelos Macroeconómicos. drodrigu@banrep.gov.co 1

4 1 Inroducion Colombia, like many Emerging Marke economies, is a small open and commodiy exporer economy. he predominan ype of he commodiy expored has varied, from coffee in he pas cenury o oil in he curren one. oneheless he imporance of he commodiy exporing secors for he Colombian business cycle is prominen. In he las 4 years, empirical evidence shows a srong and posiive associaion beween he cyclical componen of he real price of Colombian expors and he cyclical componen of real GDP. Periods of high expor prices coincide wih economic booms, while periods of lower han usual prices are associaed wih recessions. In addiion, besides conribuing significanly o he GDP cycle volailiy, he effecs of hese shocks are widespread as hey affec real variables, like consumpion and invesmen, as well as oher financial variables, like credi. Figure 1 is clear ha abou he growing imporance of oil for he Colombian economy. Colombia became an imporan ne oil exporer by he mid 8 s, bu in he las decade, oil producion increased from 6% of GDP o 1%; he share of oil expor in GDP jumped from 4% in 22 o 8% in 212. In urn, fiscal revenues from oil as a share from oal public revenues increased from under 1% in 22 o close o 2% in 211. Paerns like hese were observed in he 8 s and he 9 s, bu a significanly lower inernaional real oil prices. While by he end of 1998 he nominal oil price was close o US$12 per barrel, by mid 28 he price had muliplied by a facor of en. Few quarers laer he price dropped abruply o US$3. Since hen, he oil price has recover seadily o levels close o a US$1 per barrel. Because of heir poenially significan macroeconomic consequences, commodiy price shocks are relevan for moneary policy. here has been growing ineres in undersanding he consequences of commodiy price shocks on small open economies, like Colombia. Gómez and Julio 214 idenify episodes of buoyancy and rerenchmen in global risk and find ha hey were ransmied o Colombia s counry risk premium being imporan drivers of Colombia s oupu and unemploymen gaps. Furhermore, aking ino accoun he ineracion of foreign shocks and using a gap model of moneary policy for Colombia, hey find ha aggregae demand-relaed shocks are no imporan as drivers of Colombian non-core inflaion. oneheless, hey find ha global risk was relevan a explaining Colombia s oupu gap, paricularly during he global financial crisis. Recenly, Arango, Chavarro and González 214 using a small-scale moneary policy model sudied he impac of commodiy price shocks on inflaion, finding evidence of a posiive albei small impac from oil and food inernaional prices shocks o inflaion. he auhors aribue he lack of effec from oil and energy price shocks o a subsanial decrease in he pass-hrough from oil prices o headline inflaion and o a moneary auhoriy has faced correcly hese shocks. In heir view, inflaion expecaions emerge as he main deerminan of inflaion during he inflaion argeing regime and commodiy price movemens are already par of he informaion se used by agens o form expecaions. hese resuls conras wih he empirical sudy of Arango, Chavarro and González 213, which finds a posiive and significan pass-hrough from food and oil inernaional prices o he domesic prices of some seleced, ye imporan, iems of he CPI and PPI baskes. Moreover, Jalil and amayo 211, using also an empirical model, esimaed firs and second round effecs of food inernaional prices on inflaion of Brazil, Chile, Colombia, México and Peru, finding ha, for Colombia, he effecs of commodiy price shocks imply an elasiciy of.27 of domesic prices o he inernaional prices, bu hese effecs vanish afer four monhs. Despie he increased ineres ha has recenly emerged among academics and policy makers abou he impac of commodiy prices and heir ineracion wih moneary policy, here is sill room o learn abou how do hese shocks ransmi o he Colombian economy. Empirical models ofen fail o capure he ineracion of hese shocks o real, nominal and financial variables as well as he economic policy reacion funcions and heir feedback o he economy. Small scale models, on he oher hand, are very useful o illusrae key mechanisms and ideas of he ransmission of hese shocks, bu also fall shor or being considered a sufficien descripion of a ample se of ransmission channels, which may be relevan for policy makers. In his sudy, we proceed in wo sages. In he firs one, we perform an oil price shock idenificaion 2

5 Figure 1: Oil shock 3

6 analysis in which we analyze how a key se of macroeconomic variables behave around hese evens. In paricular, we are ineresed in sudying large increases in inernaional oil prices. Once we idenify he shocks, we observe how counry risk, GDP, privae consumpion, domesic credi, rade balance and he real exchange rae evolve during he commodiy price surge as well as during is collapse. Our main findings are ha before he peak of a large and seady oil price hikes, counry risk falls, GDP rises, privae consumpion increases, domesic credi booms, rade balance improves and he real exchange rae appreciaes. In general, afer he sudden oil price reversal all hese paerns shif back in he opposie direcion. hese facs are consisen wih he inuiion shared by mos economiss who sudy small open economies in which resource secors are imporan. As he oil price grows, income from he resource secor increases and risk premium falls wih he improved overall crediworhiness, creaing a surge in demand for radable and nonradable goods, inducing a real exchange rae appreciaion and a shif of economic resources from he radable secor o he nonradable secor. Credi expands, especially in hose favored secors by he real appreciaion. Overall economic aciviy and demand booms, in andem wih asse prices. However, sharp oil price reversals runcae his process and a rapid reallocaion of resources happen ogeher wih a collapse in asse prices and he currency. In he second sage, we wrie a large-scale dynamic sochasic general equilibrium model o accoun for hese facs. he model capures some prominen feaures of he Colombian economy. We consider a hree secor economy resource, radable and nonradable secors populaed by households, enrepreneurs, monopolisically compeiive reailers, capial producers, banks and he public secor governmen and he cenral bank. Households enjoy he revenues from he windfall secor, consume and supply labor o firms. hey can also save in he form of bank deposis. Moneary policy has real effecs because here are nominal fricions in he nonradable secor. Enrepreneurs, boh in he radable and nonradable secors face financial fricions and heir exernal financing cos is decreasing in ne worh, as in Bernanke Gerler and Gilchris BGG, Since we are ineresed in sudying moneary and macroprudenial policies, we consider a cenral bank wih hree policy insrumens: he nominal ineres rae, which adjus in response o inflaion deviaions, he sales/purchases of inernaional reserves, which adjuss in response o real exchange rae misalignmen, and finally, a financial regulaion insrumen, which enlarge or compress exernal financing depending on aggregae credi dynamics. Banks are key in he ransmission of shocks o he res of he economy. hey give commercial loans o enrepreneurs, purchase serilizaion bonds and finance heir operaions by deposis from households and borrowing abroad in he form of foreign deb. he exernal ineres rae is increasing in he raio of aggregae exernal deb o he ousanding value of he sock of oil. his las mechanism capures he spiri of Kiyoaki s role of collaeral: he sock oil can be hough as an asse which reduces he risk premium of he exernal ineres rae of he economy. We esimae he model o use i for moneary policy analysis. Our aim is o assess he benefis and he coss of convenional and unconvenional insrumens in he face of a sudden reversal of high inernaional oil prices. he model shows ha, even hough he adjusmen of he economy is efficien, i can be also be a source of financial and macroeconomic insabiliy. he source of hese insabiliies can be associaed wih he fac ha when commodiy shocks are persisen bu end up abruply, he presence of financial fricions exacerbaes he resource reallocaion process hrough he credi process. In paricular, wih a financial fricion as in BGG, a persisen appreciaion erodes he ne worh of he radable secor and booss he nonradable one, redirecing no only labor and capial o he nonradable secor bu also credi. he collapse of he commodiy price afer a prolonged period of high prices, riggers a rapid adjusmen of resources beween secors. However, he abiliy of he radable secor o absorb he freed resources is limied by is financial capaciy. However, despie ha macroprudenial policies are able o resrain aggregae credi dynamics and hus preven or ac prudenly in anicipaion o he effecs of large oil price shock reversals, hey inensify he credi allocaion from he affeced radable secors owards hose who are he beneficiaries of he appreciaion he nonradable secor. 4

7 Figure 2: Price of oil and he oil shocks 2 Macroeconomic impac of large oil shocks: an even-sudy analysis. In his secion we perform an even-sudy o analyze how key macroeconomic variables behave around oil price shocks. In paricular, we are ineresed in sudying he effec of large and persisen increases in inernaional oil prices. We observe how domesic oupu, consumpion, invesmen, rade balance, he counry risk premium, credi and he real exchange rae evolve during he commodiy price surge as well as during is collapse. Our sample runs from 1988 o 212. he even analysis is carried ou a quarerly frequency. However no all variables are available for he full sample neiher are observed a same frequency. In paricular, we ake monhly daa for he oil price and he counry risk measured by EMBI-Colombia. Our measure of oil prices is he Europe Bren Spo Price FOB Dollars per Barrel 1988m1 o 212m12 adjused using he Unied Saes CPI. he remaining variables are quarerly and aken from he naional accouns and he balance of paymens saisics. he observed sample for hese las variables is 1999Q2 o 212Q4. We follow Hamilon 23 o find he quarers during which here were oil price shocks. Hamilon defines an oil shock as a large increase in he oil price. Specifically, an oil shock is he maximum value of he oil price during he las 36 monhs. An oil shock even occurs when he oil shock is larger han wo sandard deviaions 1. A he quarerly frequency here is an even if a leas one monhly shock even occurs. Figure 2 shows he oil price and is shocks for he Europe Bren Spo Price. As can be seeing from he figure, oil shock have become larger and more frequen in he las 1 years. Figure 3 repors he average evoluion of he cyclical componen of each variable six quarers before and afer he oil shock. 2 he value of zero in he horizonal axis represens he dae of he oil price 1 his leaves us wih a se of special or singular evens, no jus noise. We believe ha given he special naure of hese evens and given he daes a which hese evens were idenified, i is unlikely ha hey will coincide wih oher special or prominen evens oher shocks In Hamilon s sense ha may have hi Colombia a he same ime. Addiionally, using average in he macroeconomic series idenifies shocks since he average responses are inegraed over oher shocks presen. 2 We ake he average across evens. We do no compue he cyclical componen for he rade balance. 5

8 Quarers Quarers Quarers Figure 3: Macroeconomic effecs of oil shock an even-sudy analysis Real GDP Privae Consumpion rade Balance Real Exchange Rae EMBI index rade Balance excl. mining Manufacuring Credi Consrucion Credi oal Credi shock even. Our main findings are 3 : 1. Counry risk premium falls and says below normal levels while he oil price increase and jumps righ afer he fall in he price Real GDP and privae consumpion increase above rend during oil price hikes and fall back afer he price reversal. 3. oal real credi o he privae secor expands above normal as oil prices increase, and scales back down abou wo quarers afer he price shock vanishes. Credi o radable secor conracs and credi o nonradable secor expands during he booming phase. his credi allocaion process reverses during he collapse. 4. he peso appreciaes seadily in real erms agains he US dollar during he oil price boom and depreciaes sharply hereafer. 5. he oal rade balance improves during hree quarers before oil price peak, and deerioraes afer he price collapse. he non-oil rade balance deerioraes during he boom and also afer he price collapse. In he nex secion we wrie a model o accoun for hese facs. 3 he Model Our baseline model is a hree secor economy resource, radable and nonradable secors populaed by households, enrepreneurs, reailers, capial producers, privae banks, he governmen and he cenral bank. Households receive he revenues from he resource secor, supply labor o firms and consume final goods and save in he form of bank deposis. Oupu is produced in several sages, 3 he naure of he oil shock supply or demand may have differen effecs on macroeconomic variables, neverheless, our objecive in he paper is no o idenify he naure of he shocks, insead we focus on a small open economy being hi by inernaional oil price spikes unusual price movemens, which he economy akes as given, regardless of he naure of hese price movemens. 6

9 including a monopolisic compeiive nonradable secor wih nominal rigidiies. Enrepreneurs, boh in he radable and nonradable secor face financial fricions and heir exernal financing cos is decreasing in heir ne worh, as in BGG. In he baseline specificaion of he model, he cenral bank ses he nominal ineres rae using a moneary policy rule. We also enhanced furher he model o consider exchange rae and credi policies. We model he firs as he sales/purchases of inernaional reserves, which adjuss in response o real exchange rae misalignmen, and he second as any financial regulaion insrumen, which respond o aggregae credi dynamics by enlarging or compressing he exernal financing premium in he economy. We repor he complee se of equaions in Appendix Households he economy is populaed by households who discoun he fuure a he facor β, 1 and choose consumpion, c, labor supply, h, and deposis, d, o maximize expeced lifeime uiliy ] E β [z u c 1 σ 1 σ χ h1+η 1 + η = where z u is an exogenous preference shock, which evolves according o z u = ρ z u z u ρ z u log z u + ε zu. 1 he represenaive household budge consrain is c + d = 1 ζ q oil + τ + ξ + w h h r 1 d 1 2 where w h is he real wage, r denoe he real deposi ineres rae, q is he real exchange rae, ξ are he profis of he producers of nonradable goods, τ are lump-sum ransfers from he governmen o households and 1 ζ oil is he fracion of oil revenues earned by households. Households also allocae labor beween radable and nonradable secor, hus he ime consrain is h = h h + h h. he consumpion bundle for household is defined as c = γ 1 ω c ω 1 ω + 1 γ 1 ω ω c ω 1 ω 1 ω 3 where c and c is he consumpion of radable and nonradables, ω is a parameer ha deermines he elasiciy of subsiuion beween radable and nonradable goods and γ deermines he household s imporance of nonradable goods. Under hese assumpions, he opimal household choices of consumpion, labor supply and deposis are: z u c σ = λ 4 χh η = w h λ 5 λ = βe [λ r ] 6 c = γ p ω c 7 c = 1 γ p ω c 8 where λ is he lagrange muliplier of he budge consrain and p and p are he relaive prices of radable and nonradable goods. 7

10 he relaionship beween he real and nominal raes in he economy is deermined by he Fisher equaion, 1 + r = 1+i 1+, and by 1+π+1 C 1+π C = q q 1 in he case of he exchange raes, where is he nominal depreciaion rae. Also, as here are wo goods heir inflaion raes are: C = p p 1 9 and = p C p Capial Good Producers In boh radable and nonradable secors, here is a represenaive capial good producer acing in a perfecly compeiive environmen. Every period boh producers buy x and x of final goods and old capial ne of depreciaion, 1 δ k 1 and 1 δ k 1, and ransform hese ino new capial a a quadraic cos. hus, he echnology o produce each ype of capial is k = z x x + 1 δ k 1 ψ x 2 k 1 δ 2 k 1 11 k = z x x + 1 δ k 1 ψ x 2 2 k 1 δ k Boh z x and z x are exogenous invesmen effciency shocks which evolve according o z x = ρ z x z x ρ z x log z x + ε zx 13 z x = ρ z x z x ρ z x log z x + ε zx 14 Under hese assumpions he radable and nonradable capial prices are: p k p k x z x ψ k 1 δ = p 15 x z x ψ k 1 δ = p Enrepreneurs Enrepreneurs produce an homogeneous good in boh radable and nonradable secors. In he producion process, he enrepreneurs buy capial from he capial producer firm and finance hese paymens by heir own funds and aking loans from banks. Addiionally, hey work for he firm and hire labor from households. Once he producion is made, enrepreneurs sell back he depreciaed capial o he capial producer firm. During he producion process, each enrepreneur is subjec o an idiosyncraic shock ha affecs he produciviy of is capial. While he radable good is sold a inernaional given prices, he non radable homogeneous oupu is sold o a reail firm ha differeniaes he produc and sells i o households. hus, during he period, his process can be characerized by he following echnologies 8

11 y y = z = z α α 1 ω 1 ω h ω 1 ω + 1 α 1 ω h ω 1 ω + 1 α 1 ω ω k ω 1 ω 1 ω 1 ω k ω 1 ω 1 ω where h = h h Ω h e 1 Ω and h = h h Ω h e 1 Ω. Boh z and z are exogenous echnology shocks which evolve according o z = ρ z z ρ z log z + ε z 19 z = ρ z z ρ z log z + ε z 2 he opimal allocaion of labor services implies ha in he nonradable secor and in he radable secor w h = p W Ω z w e = p W 1 Ω z w e E r+1 k = E w h = p Ω z α α y z h α y z h α 1 ω y z h 1 ω y h h h 1 ω h h h 1 ω 21 h h e h h e. 24 = p 1 Ω z z h he opimal allocaion of capial depends on he expeced reurn of one uni of capial. p W +1z+1 y 1 α 1 ω δ p k +1 E r+1 k = E p +1z +1 z+1 k p k 1 1 α y ω +1 z+1 k p k + 1 δ p k Given ha enrepreneurs do no have enough resources o finance heir oal capial expenses, hen heir borrowing, b and b, is given by: b = p k k n 27 b = p k k n. 28 which in urn depends on he ne worh of he firm, n and n, which evolves in ime according o: n = φ v + w e h e 29 n = φ v + w e h e. 3 9

12 where v v = r k p k 1k 1 [ ] E 1 r k b 1 31 = r k p k 1k 1 [ ] E 1 r k b 1, 32 are he proceeds per uni of capial acquired, ne of he financing cos. Finally, enrepreneurs consume p c e = 1 φ v and p c e = 1 φ v. 3.4 Reailers Reailers operae in a monopolisic compeiion environmen, buy he homogeneous nonradable goods from he nonradable enrepreneurs a a wholesale price, p W, differeniae i a no cos and sell i o households and o he capial producer firms a he reail price, p. here are nominal price rigidiies in he nonradable secor, as each firm maximizes profis under cosly price changes as in Roemberg hus, he opimal price se is = 1 θ p y + θ p W y p κ +E [ β λ +1 p λ +1κ π 1 ι 1 ι 1 + π 1 ι ι π 1 ι where θ is an exogenous markup shock, which evolves according o he reailers profis are 1 ι 1 + π 1 ι 1 ι 1 ] 33 θ = ρ θ θ ρ θ log θ + ε θ. 34 ξ = p y p W y p 2 κ 2 ι π 1 ι where κ = ɛ 1 ɛ1 ɛβ θ 1 is a parameer ha deermines he slope of he Phillips curve in he nonradable secor. 3.5 Banks he banking secor operaes under perfec compeiion and each bank is owned by households. Banks can make commercial loans o enrepreneurs by aking deposis from households and borrowing b from inernaional financial markes a he ineres rae, r. his financial inermediaion process is subjec o fricions, in paricular a cosly sae verificaion problem on he side of he asse side of he banks, which shows up in loan ineres raes in he form of spreads. In addiion, banks can also purchase serilizaion bonds, b, from he cenral bank. he rae of reurn of hese bonds is r. hus, he balance shee of he banks is b + b + b = d + q b. herefore, ineres raes are relaed by he following condiions [ ] q r = E 1 + r 36 q [ ] E r k n +1 = p k k [ ] E r k n +1 = p k k 1 ν ν 1 + r rp r rp. 38

13 Boh ν and ν are exogenous spread shocks which evolve according o ν = ρ ν ν ρ ν log ν + ε ν 39 ν = ρ ν ν ρ ν log ν + ε ν 4 Equaions 37 and 38 sae ha real ineres raes of commercial loans are a decreasing funcion of he ne worh relaive o capial in each of he secors. his funcion is equivalen o a more deailed descripion of he BGG financial acceleraor wih one excepion, he erm rp, which inroduces a regulaion premium. his premium capures he essence of he role of regulaion on ineres rae spreads. I enlarges he ineres rae spreads in response o rapid aggregae credi growh, for insance. he precise mechanism by which his happens is no explici in our paper, bu we believe his erm could represen any regulaion measure, like counercyclical buffers, capial requiremens, reserve requiremens, or any mechanism which makes privae credi more cosly. 3.6 Governmen and Cenral Bank We characerize he governmen and he cenral bank as a se of policy rules: a moneary policy rule, ha responds o deviaions of inflaion relaive o he arge π, i i = i ρi 1 π π ϕπ exp ε µ 41 a FX inervenion rule, which responds o real exchange rae deviaions from is seady sae value, q ri = ri Ψ q q q 1 42 a regulaion premium rule, which responds o oal credi deviaions from is seady sae value, credi rp = exp µ rp credi 1 43 and a se of accouning equaions ha have o be saisfied every period: q ri = b 44 τ = b 1 + q 1 + r 1 ri r 1 b 1 + q ri he firs equaion is he balance shee of he cenral bank: on he lef hand side appears he value of he inernaional reserves in real erms of local currency and on he righ hand side he bonds issued o perform serilized operaions. he second equaion saes ha he proceeds of he cenral bank operaions are he resul of bond sales and FX purchases boh ne of ineres paymens. hese proceeds are ransferred back o households in a lump-sum way. 3.7 Oil secor One key aspec of he model is he role of oil. he value of oil aciviies in unis of he final composie goods in he economy evolves exogenously according a process 4 : oil = ρ z oiloil ρ z oil log oil + ε oil We assume ha oil shocks are exógenous o he small open economy and we do no diferenciae he naure of he oil shok, his involves modeling a global block in which he dynamics of he oil price is endogenous and depends on he overall aciviy. 11

14 he revenues from oil aciviies are ransferred o households. hus changes in hese revenues affec he households budge consrain. In addiion o his effec, we model he role of oil as affecing he ineres rae premium ha he economy faces when borrowing or lending abroad. qb exp ν b 1 + i = 1 + r z i GDP b exp 47 ν oil oil oil where GDP = q 1 ζ oil + p y + p y and z i = ρ z i z 1 i + 1 ρ z i log z i + ε zi π = ρ π π ρ π log π + ε π are exogenous exernal shocks. his equaion saes ha he ineres rae a which he economy borrows abroad grows as he sock of exernal deb increases bu falls wih he value of he sock of oil in he economy. Besides helping o close he small open economy model, his device is a simple way o capure he idea ha an economy which discovers new oil or enjoys higher oil prices, no only relaxes he households budge consrain bu also makes he economy more credi-worhy and reduces is risk premium. Alhough his is an ad-hoc device, we believe i capures he idea ha a real asse, like he sock of oil available for exracion, influences he exernal borrowing premium. For insance, a counry like Venezuela wih a large sockpile of oil, faces a lower exernal premium when compared o oher counries wih fewer oil reserves. 4 Esimaion here are wo ypes of parameers in he model. One ype of parameers are calibraed and he oher ype are esimaed. he se of calibraed parameers are hose ha affeced he seady sae of he model and herefore are chosen in order o mach he long run relaions observed in he daa. able 1 shows he values of he calibraed parameers. We begin by discussing he parameers ha affec he uiliy funcion. he discoun facor β is se in order o obain a real ineres rae of 2.5% in he long run. he risk aversion coefficien σ is se o one, implying a logarihmic uiliy funcion in consumpion. he parameer η is also se o one, herefore he labor supply elasiciy is uniary. And finally, as is sandard in he lieraure, he parameer χ is chosen in order o obain an average supply of hours of 1/3. he exernal real ineres rae, r, is also se o 2.5% and he domesic and exernal inflaion arges, π and π, are se o 3%. his implies a long run real depreciaion in he model equal o zero. he consumpion bundle in he model is composed by nonradable and radable goods, we are assuming an almos uniary elasiciy of subsiuion, ω, beween hese wo goods and since γ affecs he share of nonradable consumpion in he bundle, he parameer is chosen o mach he long run share of nonradable consumpion. he nonradable echnology in he model includes labor and capial, he elasiciy of subsiuion beween hem, ω, is se near o one and he share of labor in he producion funcion, α, implies a labor inensive echnology. he oal amoun of labor used in he producion combines labor supplied by he household and supplied inelasically by he enrepreneurs h e. he share of household labor relaive o oal labor, Ω, is se o.95, in line wih he share of business owners in he oal labor workforce as repored in Mejia 29. Anoher parameers ha affec producion are hose relaed wih he capial demand from he enrepreneurs. In his case, he exernal finance premium depends on ν, depends also on he parameer 12

15 able 1: Parameer Calibraion Parameer Descripion value β Discoun Facor.9938 σ Risk aversion 1. η Inverse of labor supply 1. χ Scale parameer r Exernal real ineres rae 1.62 π Inflaion arge 1.74 π Exernal inflaion arge 1.74 ω Elasiciy of subsiuion b/w and.9 γ Share in consumpion bundle b/w and.7 ω Elasiciy of subsiuion b/w Labor and Capial.9 α Share of labor in he producion.6 Ω Share of household labor in oal labor.9 ν Exernal finance premium parameer.1 φ Share of ne worh consumed by enrepreneur.9 δ Depreciaion rae.25 h e Enrepreneur labor supply.1 θ Elasiciy of subsiuion b/w inermediae goods 5. ω Elasiciy of subsiuion b/w Labor and Capial.9 α Share of labor in he producion.6 Ω Share of household labor in oal labor.9 ν Exernal finance premium parameer.1 φ Share of ne worh consumed by enrepreneur.9 δ Depreciaion rae.25 h e Enrepreneur labor supply.1 b Sock of ne foreign asses 1.2 ri Sock of inernaional reserves.1 oil Sock of oil reserves ζ Exracion rae.95 z Mean exogenous echnological process 1. z Mean exogenous echnological process 1. z i Mean exogenous exernal ineres rae process 1. z u Mean exogenous preferences process 1. z x Mean exogenous invesmen process 1. z x Mean exogenous invesmen process 1. 13

16 able 2: Long run relaions Relaions Model Daa labor 32.21% 28.58% Real ineres rae 2.47% 2.5% Inflaion rae 2.96% 3.% ominal ineres rae 5.43% 6.7% ominal nonradable loan rae 1.5% 13.86% ominal radable loan rae 9.95% 1.24% Consumpion / GDP 73.67% 82.34% Consumpion nonradable / Consumpion 62.99% 87.32% Consumpion radable / Consumpion 38.56% 12.82% Invesmen / GDP 11.37% 21.22% Invesmen nonradable / Invesmen 6.95% 59.95% Invesmen radable / Invesmen 41.59% 4.2% nonradable oupu/ GDP 64.77% 59.3% radable oupu/ GDP 24.22% 33.76% Oil producion/ GDP 11.% 7.21% Sock of oil / Annual GDP 55.2% 64.63% e foreign asses / Annual GDP 3.% 28.29% Inernaional reserves/ Annual GDP 1.% 13.13% Credi / Annual GDP 96.48% 28.2% φ which defines he consumpion for he enrepreneurs and depends on he depreciaion rae, δ. All hese parameers are chosen in order o fix observed long run value of he ineres rae spread. In addiion, he model assumes imperfec compeiion in he nonradable secor. he elasiciy of subsiuion beween he differeniaed nonradable goods θ, implies a mark up of 25%. his value is aken from previous sudies for Colombia Resrepo 25 and Perez 25. he same argumens are used o pick he parameers for he radable secor, he only difference arises because in his secor here is perfec compeiion and herefore here is no mark up. he parameer b deermines raio of ne foreign asses o GDP, ri deermines he raio of inernaional reserves o GDP, oil he sock of oil reserves relaive o GDP and ζ he exracion rae of he oil reserves. Finally, he mean for all exogenous process are se equal o one. he able 2 shows he long run relaions obained from he calibraed parameers and heir empirical counerpars. he second se of parameers are esimaed using Bayesian echniques. hese parameers mainly affec he dynamic behavior of he model and are relaed wih invesmen adjusmen coss, nominal rigidiies, ineres rae spreads, he policy rule coefficiens and he parameers for he exogenous process. During he esimaion he parameers ha deermine he FX inervenion Ψ q and he macroprudencial regulaion µ rp are se equal o zero. he parameer able 3 shows heir prior-poserior disribuions. he daa used in he esimaion is expressed in quarerly growh raes and is assumed o have a measuremen error. We included oupu in radable, nonradable and mining secors, oal consumpion public and privae, oal invesmen, commercial loans for radable and nonradable and saving deposis. Addiionally we included radable and non radable inflaion, nominal devaluaion, inerbank 14

17 able 3: Esimaion resuls Prior Mean Poserior Mean disribuion Mean Var Mean sd parameers ρ i bea ϕ π gamma ɛ bea ι bea ρ z bea ρ z bea ρ i bea ρ π bea ρ z oil 1 bea ρ θ bea ρ z u bea ρ z x bea ρ z x bea ρ ν bea ρ ν bea ψ gamma ψ gamma ν b gamma ν oil gamma variances ε z inv_gamma.125 Inf ε z inv_gamma.125 Inf.55.8 ε zi inv_gamma.125 Inf.32.5 ε oil inv_gamma.125 Inf ε π inv_gamma.125 Inf.45.8 ε µ inv_gamma.125 Inf.28.3 ε θ inv_gamma.125 Inf ε zu inv_gamma.125 Inf ε zx inv_gamma.125 Inf ε zx inv_gamma.125 Inf ε ν inv_gamma.125 Inf ε ν inv_gamma.125 Inf

18 lending rae and a measure of exernal ineres rae augmen wih risk premium. In appendix B here is deailed descripion of he daabase used for he esimaion of he model. 5 Macro Policy in Response o Oil Shocks We use he esimaed model o perform several quaniaive exercises. Firs, we compue he shock decomposiion analysis. We quanify he hisorical imporance of exernal and domesic shocks in he colombian daa. Alhough our focus is commodiy expor shocks, we also sudy he relaive imporance of he remaining ones. Second, we compue he Bayesian impulse-response funcions. he objecive is o visualize in more deail he macroeconomic impac of one-off commodiy shocks and he policy response of he cenral bank. he hird exercise is a counerfacual experimen o assess he effecs of an unexpeced reversal of a commodiy shock. here, agens ake decisions based on he idea ha he value of commodiy producion will increase persisenly for several quarers. However, hey do no anicipae he possibiliy of a sudden reversal. Our ineres is o use he model o analyze he role of convenional moneary policy and macro-prudenial policies when he sudden commodiy collapse akes agens by surprise. 5.1 Shock decomposiion Our firs quaniaive experimen is a shock decomposiion analysis. he model considers 12 shocks. We label hree of hem as exernal and he res as domesic. he exernal shocks are: one o he exernal ineres rae spread ha he economy faces when borrowing abroad, anoher is a foreign inflaion shock and he hird is a shock o commodiy expors. I can be eiher commodiy prices or quaniies. he domesic ones are shocks o produciviy, invesmen and ineres rae spreads all o boh secors radable and nonradable, markup as well as a moneary policy shock. Our aim is o gauge he esimaed conribuion of he considered shocks on he observed movemens of radable and nonradable oupu and credi, aggregae consumpion, inflaion, he real exchange rae and nominal ineres raes. Figure 4 repors he evoluion of hese variables solid black line from he second quarer of 2 o he firs quarer of 213. he colored bars represen he shock conribuions. he shock decomposiion of hese ime series confirms a convenional finding in many of he models used in he inernaional macro lieraure. he role of foreign ineres rae spread shocks appear o be small. Also he role of foreign inflaion shocks is negligible. Mos of he imporance of foreign shocks in he Colombian macro series sems from commodiy expor movemens, as we suspeced by he documened evidence in he firs par of he paper. he mechanism hrough which his shock ransmi o he economy is no apparen a firs sigh. he decomposiion of oupu reveals ha boh radable and nonradable produciviy as well as preference shocks explain mos of he observed oupu flucuaions in Colombia. he conribuion of commodiy expor shocks o he aggregae oupu cycle is small. his is no surprising as he share of hose expors on GDP is small. However, a deeper inspecion by secors reveals a clearer picure of he role of commodiies in he economy: he larges conribuion of such shocks is o radable and nonradable oupu, he real exchange rae, inflaion and policy raes. In he las en years commodiy expor shocks have been an imporan drag on radable oupu and a booser o nonradable producion. his conribuion is also eviden on relaive prices as commodiy expors have been a major source of real exchange rae appreciaion. his fac implies ha radable inflaion is also affeced by commodiy shocks. From 2 up o he firs half of 28 hese shocks were an imporan source of inflaionary pressures. his condiions have reversed since hen. ow commodiy shocks are helping o ease inflaion. As he cenral bank ses is policy rae focusing mainly on inflaion, ineres rae movemens reveal essenially he same shocks which impac inflaion. Despie hese large real effecs on secoral oupu and relaive prices, commodiy expor shocks have a smaller conribuion o radable and nonradable credi and consumpion.credi flucuaions 16

19 Figure 4: Hisorical Shock decomposiion 213 onradable Oupu onradable Credi oal inflaion radable oupu Real exchange rae Policy rae GDP radable produciviy onradable produciviy Exernal ineres rae spread Foreign inflaion.1 radable Credi.5.5 Commodiy revenues radable spread onradable spread Iniial values Consumpion.5.1 Moneary policy Preferences onradable invesmen radable invesmen

20 are mosly dominaed by invesmen specific shocks, specially in he case of nonradable credi. he iner-secoral effecs of specific shocks go from nonradable credi shocks o radable credi, bu no he oher way around. More precisely, radable invesmen shocks have no impac on nonradable credi, while nonradable invesmen shocks spillover o radable credi. his ype of credi is much more responsive o ineres rae credi spread shocks han nonradable credi. hus, based on hese findings, we are inclined o conclude ha he imporance of commodiy expor shocks appears o lie more on is secoral effecs han on is impac on aggregae aciviy. GDP flucuaions may mask he reallocaion effecs ha commodiy expor shocks enail. Preference shocks appear o be an imporan source of macroeconomic flucuaions, as hey affec GDP hrough nonradable oupu, consumpion, inflaion and he policy rae. However, hey do no affec boh ypes of credi. heir conribuion o real exchange rae and radable oupu flucuaions has also been small. Pu ogeher, hese resuls sugges ha credi cycles in Colombia does no appear o be driven by aggregae domesic shocks, nor by foreign shocks bu mosly by secoral specific shocks and heir ineracion. 5.2 Impulse response he model can be used o explain he dynamics of he economy afer a commodiy expors shock. here are several reasons ha induced us o hink ha undersanding he effecs of hese ypes of shocks is imporan. Firs, here is he documened evidence in he firs par of his paper. Second, even if we disregard i, here is large body of lieraure in he inernaional macroeconomics field ha sudies secoral reallocaion afer such shocks. We have se up a model in which we assume ha he nonradable secor in aggregae operaes in domesically shielded final oupu markes and faces lile direc compeiion from abroad. Also, nonradable secors impor a significan amoun of heir inpus. hese wo facs have been documened in Echavarría, González and Mahadeva 213. he mechanism ha we have in mind o explain he response of he economy o a commodiy boom, illusraed in Figure 5, is as follows: assume ha many reail producs are originally produced abroad. Given he relaive sheler from foreign compeiion and he imporance of impored inpus for nonradable secors, he exchange rae appreciaion lowers he cos of nonradable secors inpus wihou lowering heir domesic oupu price. his booss he nonradable secors value-added incomes and permis an expansion in he scale of producion. Finally, anoher favorable impulse o he nonradable secor arises hrough he reacion of he consumer o he lower exchange rae. In he Duch Disease lieraure, his was called he spending effec. In Colombia, an imporan share of impored goods goes ino household consumpion. hus, he real exchange rae appreciaion raises he real value of household income. In aggregae, his offses he loss of income from hose consumers who work in he radable secors. he rise in consumers income leads hem o demand more. As mos of domesic demand is supplied by he nonradable secor, his represens anoher channel hrough which he oil price rise benefis he nonradable secors. Circumsancial evidence suggess ha he combinaion of all hese effecs can be powerful, bu especially when a boom in bank credi is also riggered by he appreciaion. In our model, when here is a posiive commodiy expors shock, households experience an increase in heir income and wealh and herefore raise heir spending. his implies a greaer demand for boh radable and nonradable goods. he fall in he exernal risk premium associaed wih he improved ne foreign deb o real asses posiion generaes an appreciaion of he real exchange rae as well as he increased demand of nonradable consumpion. he increased demand for radable goods is me by a greaer share of impored goods han previously. he demand surge afer he oil price shock also drives enrepreneurs o increase heir demand for loans. his ype of adjusmen is convenional in a radable-nonradable model. However, in our model a key mechanism works hrough he exernal ineres rae risk-premium. his premium has an endogenous componen, which depends no only on ne exernal deb bu also on he sock of he real commodiy resource. A commodiy price shock raises he value of his real asse lowering he risk premium ha he economy faces in inernaional financial markes. hus, he income 18

21 2 4 6 Figure 5: Impulse response o a commodiy expors shock nominal ineres real exchange rae inflaion consumpion 5 radable oupu nonradable oupu credi radable credi nonradable credi effec on households budge consrain may be small bu he wealh effec, specially since overall he economy has a negaive ne foreign asse posiion, may be large. Wihou his mechanism, he effecs of a commodiy boom on he real exchange rae in a hree secor model would be smaller. here is an addiional channel in our model induced by he presence of he secoral financial acceleraor. he appreciaion of he exchange rae also leads o a fall in he value of he asses of he radable secor, lowering he value of is collaeral and consequenly, rising he exernal financing premium ha radable firms pay o commercial banks. his increase in financing coss coupled wih he lower demand of domesically produced radable goods drives furher down employmen in his secor. In conras, he nonradable secors benefi from an exchange rae appreciaion. his channel is presen in he model bu i is quaniaively small in he Colombian daa, as he shock decomposiion has already revealed. 5.3 Unexpeced commodiy expors reversal he adjusmen of he economy described by he impulse response funcions reflecs he bes reacions of economic acors, given he financial resricions hey face, and, also crucially, under he assumpion ha hey expec ha he commodiy boom will no be reversed. However, in pracice, agens may fail o appreciae ha commodiy prices are nooriously volaile. During he boom, he economy s aggregae deb and privae credi are higher and also he "size" of he radable secor is lower. hus, his secors abiliy o absorb nonradable secor employees is limied by is capial sock. his creaes a form of vulnerabiliy and a pruden policy maker may consider he use of alernaive policy insrumens, like regulaion policies aimed o conain aggregae credi. Our quesion is o deermine quaniaively if alernaive policies have differen effecs a he macro level and/or a he secoral level. he firs and convenional opion is o adjus he cenral bank reference rae in response o oal inflaion. In his case, he bank conrols boh he expansion of aggregae demand and credi. A second policy opion is o combine he policy rae adjusmen wih macro-prudenial policies, which respond o credi expansion. For insance, economic auhoriies have implemened policies o increase ineres raes on commercial loans. o capure he idea of a macro-prudenial policy in our model, we added 19

22 Figure 6: Effecs of macroprudencial policy 2 nominal ineres 2 real exchange rae 2 inflaion consumpion 5 radable oupu 3 nonradable oupu credi 2 radable credi 2 nonradable credi Black: here is a unexpeced commodiy price reversal and moneary policy is conduced in a convenional way. Blue: here is a unexpeced commodiy price reversal and convenional moneary policy is complemened wih a macroprudenial policy on credi. Red: here is no commodiy price reversal and moneary policy is convenional. a regulaion premium o he lending raes in boh secors. his premium increases when domesic credi exceeds is seady sae value and i can be se independenly from he policy rae. For furher deails on his ype of policy rule, see Unsal 213. ow, o mimic he dynamic paern of a ypical commodiy shock, we simulae i as an AR2 process. So, we replace he AR1 process used for esimaion, wih he AR2. he res of he parameers of he model remain inac and hose of he AR2 process are configured in such a way ha afer an iniial increase in he price of oil agens can expec addiional increases in he fuure racing a hump shape. his ype of paern canno be reproduced using an AR1. We are aware ha by changing he commodiy shock process we would obain possibly differen policy rules and parameer esimaes. However, we do so jus o perform a counerfacual experimen o capure he idea of a sudden reversal in commodiy prices. hus, an unexpeced negaive shock is added six quarers afer he iniial commodiy price hike. Figure 6 depics he response of he economy afer a commodiy shock under hree scenarios. he red line shows he case in which here is no sudden unexpeced commodiy price reversal and moneary policy is conduced as he esimaed aylor rule. he black and blue lines show wha happens when here is a reversal. he difference beween he black and he blue lines is ha, in he laer, convenional moneary policy is complemened wih a macro-prudenial policy. We sar considering he red line. I describes he scenario of a long lasing rise in he commodiy expor value. Iniially here is a shock of 25 bps which does no reverse immediaely bu rises o 6 bps six quarers ahead. hus, according o he esimaed parameer values, he rise in commodiy expors lead o an immediae significan appreciaion of he exchange rae of 5 bps, which is unwound slowly hereafer. 5 he RER appreciaion lowers radable secor oupu iniially by 2 bps. radable oupu gradually reurns o is former level, as he exchange depreciaes back slowly. Moneary policy 5 he real exchange rae is shown as dollars per peso so ha a fall is an appreciaion. 2

23 is adjused o keep inflaion under conrol. As here is no macroprudenial objecive or insrumen, privae credi is allowed o increase, and, indeed, is level rises o reach 1bps a is peak. hus, if commodiy expors rise and are no expeced o be reversed, moneary policy would have o be ighened o deal wih non radable inflaion no shown. In realiy, here is no much cerainy ha any commodiy boom will be susained. For insance, commodiy prices are affeced by supply and demand shocks and he possibiliy of sorage implies ha he price is sensiive o fuure prospecs. Agens would do well o incorporae he sudden reversal risk in heir acions. In our exercise, hey fail o do so. However, o analyze he role of macro-prudenial policies, we consider he case in which he cenral bank for some unexplained reason implemens a regulaion o ackle he commercial credi expansion. he necessary adjusmen of he economy afer a sudden commodiy boom reversal can be seen by comparing he red and black lines. As before, only moneary policy is acive. Figure 6 reveals ha a he ime of he reversal, here is a sharp real depreciaion ogeher wih a fall in aggregae consumpion. onradable secor deb also conracs. his reflecs he sudden realizaion ha pas invesmens in he secor will no be jusified. Moreover, nonradable inflaion no shown declines srongly a he ime of he reversal. Inflaion falls because of he labor marke adjusmen. In paricular, real wages fall due o several facors. Firs, he real exchange rae depreciaion riggers a flow of employmen from he nonradable o he radable secor causing an excess supply in he radable secor. Second, labor demand in he radable secor is lowered because here is shorage of domesic demand for hese goods. hird, he radable secor s abiliy o "absorb" he greaer labor supply a any given wage is limied by is capial sock which had decreased in he booming phase. oe, however, ha invesmen in he radable secor increases upon he commodiy reversal and he consequen depreciaion. o some exen, credi reurns o his secor and oupu recovers. hus, he secoral asymmeries linked o he exchange rae movemen are also presen in he case of a reversal. he use of macro-prudenial policies has imporan macroeconomic implicaions. Firs, commercial credi is sabilized as he policies are designed o do. oe ha his happens even wih a looser moneary policy. Exchange rae pass-hrough lowers radable inflaion and dominaes he nonradable inflaion pick-up. However, his macro-prudenial policy has secoral consequences. Credi o radable secor falls even furher during he boom, and credi o nonradable secor expands more rapidly. As a resul, here is a sharper real exchange rae appreciaion. his exacerbaes he ype of vulnerabiliy ha we have sressed. oe also ha he secoral effecs are so srong ha he fall in radable secor credi overshadows he nonradable one, leading o a fall in commercial credi. Consequenly, once he commodiy boom reverses, oal commercial credi reacs less, nonradable credi conracs less and radable credi reacs more slowly. hese resuls illusrae ha macro-prudenial policy can be an effecive complemenary ool in sabilizing credi. However, in he presence of secoral financial acceleraors i can exacerbae he reallocaion process. he radable secor faces a higher ineres rae because of macro-prudenial policy and, addiionally, suffers from he devaluaion of is collaeral; a reflecion of he real exchange rae appreciaion, which leads o higher exernal finance premium for hose companies. As a resul, invesmen in he radable secor falls more when macro-prudenial policy is employed han wihou i. herefore, he adjusmen of radable secor afer commodiy price reversal akes longer. While he macroprudenial policy limis he buildup of deb, our simulaions illusrae ha i also is likely o boos furher he nonradable secor and hamper he radable one. 6 Final Remarks In he las four decades empirical evidence shows a srong and posiive associaion beween he cyclical componen of he real price of Colombian expors and he cyclical componen of real GDP. Periods of high expor prices coincide wih economic booms, while periods of lower han usual prices are associaed wih recessions. In addiion, besides conribuing significanly o he GDP cycle volailiy, he effecs of hese shocks are widespread as hey affec real and nominal variables, like consumpion, 21