The cost of deviating from the optimal monetary policy: a panel VAR analysis.

Departement of Economic and Buine Studie Univerity of Genova The cot of deviating from the optimal monetary policy: a panel VAR analyi. Chiara Guerello Working Paper 1/2014 ISSN

The cot of deviating from the optimal monetary policy: a panel VAR analyi. Chiara Guerello Abtract Exploiting the Panel VAR gmm etimator feature, macroeconomic country factor are combined with micro-economic bank data to tet for the rik taking channel in the Euro Area. According to prior expectation baed on an extended DSGE model, the analyi prove that the monetary policy feed the bank rik taking by increaing the bank leverage, but i not able to affect the level of credit rik. Wherea, deeper invetigation point out the Taylor gap to add to bank rik appetite in all it form, while movement in the interet rate, to the extent of reaction to the target variable, mooth the bank rik. Keyword: Rik taking channel; Taylor gap; Monetary policy; Credit rik: Panel VAR JEL: E51; E52; G21 E-mail: chiara.guerello86@vodafone.it Addre: Department of Economic (DIEC), Univerity of Genoa, via vivaldi 5, 16126 Genova, Italy 1

1 Introduction Thi eay enter the debate on whether the Central Bank ha to incorporate ome bank rik indicator in it reaction function (Borio and Zhu (2012)), or it i an excluive role of the bank upervior to control over the financial ector tability (Nier and Merrouche (2010)). It i invetigated to which extent the monetary policy affect, through the o called rik taking channel,the built up of rik in the financial ytem and might contribute to globally non-linear dynamic in the economy (boom but cycle). Specifically, it dientangle which component of the real interet rate (monetary policy rule or random hock) affect the financial intermediarie perception of rik, and hence their level of leveraging and credit rik, and with which relevance. A defined by Borio and Zhu (2012) the rik taking channel of monetary policy i the tranmiion mechanim of the movement in the interet rate that work through the conequently movement in rik perception and, hence, in the degree of rik in bank aet portfolio, on the pricing of aet, and on the price and non-price term of founding. In Europe, due to the central role played by bank in the founding proce, the rik taking channel i mainly explicated by quantitative and qualitative movement of the financial intermediarie balance heet. According to Angeloni et al. (2010) the tranmiion channel to the overall bank rik can be divided in two main effect affecting different balance heet component. A firt, more quantitative, regard the expanion/contraction of the balance heet after an expanive/retrictive monetary policy. A point out by Adrian and Shin (2010), the problem relie in the mimatch of the maturitie, becaue the bank prefer hort term founding when the hort term interet rate decreae and the yield curve become upward loping and teeper. A it regard the typologie of founding choen by the bank, it i uually called leverage effect, and i computed a ome relative meaure of hort run aet. Thi leverage effect, due to change in the cot of founding, make the balance heet rikier without any change in the aet mix. A econd channel, more qualitative, work through the aet mix: indeed, after of a prolonged looe monetary policy, bank how a tronger propenity to riky aet (ee Rajan (2006). The impact of interet rate on valuation, income and cah flow (for intance due to the reduction of the expected default probability of the borrower) affect the rik perception and the rik tolerance of the lender, a well a, relaxed the capital requirement. A it i related to the overall quality of the loan portfolio of the bank, it i uually named credit rik and meaured a the default probability of the loan portfolio. Even if the rik taking channel i an enhancing mechanim of the financial accelerator channel(bernanke et al. (1999)), there are ituation in which tabilizing force are le 2

effective (for example the cae of collective moral hazard propoed by Farhi and Tirole (2013)) and a looe monetary policy could cumulatively lead to an over-extenion of the balance heet, a built up of financial imbalance and a bank crii if it unwind. Many influential work ha empirically aeed the exitence of the rik taking channel for monetary policy. The reult are robut to different etimation method and credit rik proxie. However, the trength of the reult i arguable for Europe, for different reaon: 1. Angeloni et al. (2010) performed a VAR model with data at country level for US and European Monetary Union (EMU). The negative caual relation of monetary policy to credit rik and leverage i trongly ignificant and one way only for US. For the Euro Area thi caual relation move in the two direction, and the reult are le table. Even if the qualitative effect i peritent and ignificant for both area, the impact of the monetary policy on the overall balance heet vanihed in everal pecification. 2. Briimi and Deli (2010), accounting for the bank heterogeneity by a local GMM etimator, verified the preence of the rik taking channel only for mall bank. Healthier bank, with trong market power are not reponive to movement in the interet rate for what concern the aet rik. For ome unit the reult are overturned. The reult are ignificant for both US and Euro Area. 3. Deli et al. (2012) uing a factor augmented panel var to account for bank heterogeneity for US data, howed that, on average, the negative correlation between interet rate and credit rik hold only in the medium period, in the very hort period (firt lag) i the oppoite. 4. Ozuca and Akbotanci (2012), in oppoition with the actual literature, found a trong and robut negative relation between monetary policy and credit rik for a panel of Turkih bank. Learning from the finding lited above, in thi eay i ued a rich panel of European bank. A VAR pecification allow to overcome the endogeneity problem avoiding the ue of proxie for the interet rate, a well a, to conider other influencing economic factor a the financial market index or the GDP. Other bank pecific characteritic, quite table over the time, are accounted by a bank fixed effect. Even if the preliminary impule repone function analyi point out the bank indifference to the nominal interet rate while taking their aet rik portfolio deciion, many argued that keeping the nominal interet rate too low for too long (roughly between 2002 3

and 2005) tranformed the benign effect of the rik taking channel in a collective overtaking of rik and the boom-but cycle that led to the 2008 financial crii. But what doe it mean too low? Since 2002, the European central bank, and ECB in firt, were operating with a well etablihed paradigm, etting the price tability a a primary cope for the monetary policy. Thi practice i explicated by an interet rate rule, called Taylor rule, that linked the interet rate directly and excluively to the inflation rate. Thi hitorical rule i conidered a benchmark to define when an interet rate i too low. The deviation from the optimal rule, called Taylor gap, entered many invetigation but it effect on the rik taking channel are till unclear. On one ide Nier and Merrouche (2010) howed that the current account imbalance and the compreion of the pread between the hort rate and the long rate had driven the recent adding up in the bank leverage more than the difference in the monetary policy tance. In oppoition, Altunba et al. (2011) found a poitive relation between the overall portfolio quality and interet rate and negative one between the portfolio quality and different meaure of Taylor gap. Before arguing in favor or againt an extenion of the Taylor rule to include the leverage and/or the credit rik, it i worth to invetigate how thi two component (the precribed interet rate and the Taylor gap) interact with the rik variable. The main contribution of the paper i to analyze contemporaneouly the different rik variable and different monetary policy component to deeply dientangle how the rik taking channel effectively work. Many previu finding of the literature are put together to provide a complete and exhautive view of the phenomenon, and hence make the policy advice concluive and reliable. In order to provide a proper view of the phenomenon, we decide to bae our expectation on a extended DSGE model, rather that rely on a fragmented literature. The model, in addition to provide a benchmark to which compare the empirical reult, provide a quite realitic explanation of the dynamic and tranmiion channel behind the correlation empirically analyzed. The eay i divided a follow: ection 2 introduce a dynamic tochatic general equilibrium model that how the rik taking channel dynamic. Section 3 decribe the data and the econometric model. Section 4 diplay the main empirical reult and ection 5 conclude. 4

2 A DSGE model for an economy with rik taking channel 2.1 Model: feature and propertie A Dynamic Stochatic General Equilibrium model i ued to decribe the theoretical dynamic of the rik taking channel. It how the effect of three different hock, monetary, macroeconomic and financial, to an economy with a relevant financial ector. Specifically the model i an extended verion of the neoclaical growth model, a deigned in Schmitt-Grohe. and Uribe (2004), in which a financial ector how friction a in the Bernanke et al. (1999) and the banking ector i deigned a in Chritiano et al. (2010). Combining thee feature in a very imple way allow for an endogenouly determined level of bank leverage(hort term aet over depoit) and credit rik (the amount of entrepreneur that default). Therefore it i poible to conider the dynamic of the two component of the bank rik conidering only three hock. The model include four nominal rigiditie, price and wage tickine a la Calvo (Calvo (1983)) and Yun (1996) with indexation to pat inflation and money demand for both houehold and firm, and even real rigiditie (internal habit in conumption, monopolitic competition in factor and product market a well a in the liquidity ervice market (bank are aume to be perfectly competitive only in the loan market), invetment adjutment cot, variable cot of adjuting capacity utilization, and monitoring cot of entrepreneur productivity). The ytem i olved by an algorithm propoed by Schmitt-Grohe. and Uribe (2004). It allow for approximating the model in Taylor erie till the econd order. In thi way it i poible to conider the inefficiencie generate by the price and factor tickine, becaue two recurive variable repreenting the price and wage diperion arie in the linearized model. In addition it allow for a tochatic volatility repreentation of the hock and to properly deal with the non linearity of the model. Finally, it allow for a ditorted teady tate, in which the financial friction are till effective. In thi ection i decribed the exact nonlinear repreentation of the complete et of equilibrium condition. 5

2.1.1 Houehold The repreentative houehold i a family with a continuum of member, each of them i a worker. They maximize an utility function determined by the conumption growth, the labor upplied, and the liquidity ervice provided by holding money, depoit and hort term marketable ecuritie. The cot of conuming a certain amount c t of intermediate good i determinate by the total expenditure minimization:.t. min 1 0 p i,t c i,t di the firt order condition i then: c t = [ 1 0 c λ f i,t ] 1 λ f (1) λ f c i,t = ( P it ) ( λ f 1 ) c t where P t = [ P t 1 0 P 1 λ f i,t ] λ f (2) The labor deciion are taken collectively at union level, that give the labor upply ide a monopolitic power. The upply of labor follow: h t = h d t 1 0 ( w j t ) λw 1 λw d j (3) w t The houehold decide alo the wage level, but, due to the wage tickine, they can reoptimized it only with probability ξ w, otherwie they adjut it according to pat price inflation: w j t = w j t 1 π t 1 (4) Becaue the labor demand curve i identical acro all the labor market, the wage and employment can be aumed identical, and all the houehold face the contraint: 1 = [(1 ξ w ) 1 λw t + ξ w (w t 1 π t 1 ) 1 λw 1 ] 1 λ w (5) w 1 w t [ 1 ξ w( π t 1 π t ) 1 1 λw 1 λ w ] 1 λ w = φ L K wt F wt (6) K wt = h 1+σ L d,t + βξ w [ w t+1 w t 1 π t+1 ] λw 1 λw (1+σ L) K w,t+1 (7) F w,t = h d,t λ z,t λ w + βξ w ( w t w t+1 ) λw 1 λw ( w t+1 w t ) 1 1 λw 6 1 π t+1 F w,t+1 (8)

The houehold at the end of the period received an interet on the current depoit, jut inveted hort term marketable ecuritie and the lat period hort term aet. The houehold problem can be ummarized a:.t. max E 0 k=0 where w,t = (1 ξ w )( w t w t ) λw 1 λw ln(c t bc t 1 ) φ L ( w,t h t ) 1+σ L 1 + σ L + φ M (m t ) 1 σm 1 σ M + +φ D (d t ) 1 σ D 1 σ D + φ DM (dm t ) 1 σdm 1 σ DM + ξ w ( w t 1 ) λw 1 λw ( π t ) lambdaw λw 1 w,t 1 (9) w t π t 1 P t c t + m h,t m h,t 1 + (1 + rt d )d t d t + (1 + rt dm )dm t 1 dm t+1 + (1 + r a )a t 1 a t <= h d,t w j,t ( w j,t ) λw 1 λw d j w t The firt order condition for the Houehold are: 2.1.2 Firm 0 +(1 )(1 ι) n t+1 W e W e and 5 (10) ι λ z,t = 1 βb (11) c t bc t 1 c t+1 bc t λ w λ w 1 φ L w h σ L 1 1 1xi w βπ = λ z,t w (12) λ w 1 ξ w βπ 1 λ w λ z,t = β(1 + r A t )λ z,t+1 (13) φ M m σ M h,t r DM = λ z,t βλ z,t+1 π t (14) φ D d σ D t = rt D λ z,t (15) t+1 ra t+1 = φ DMdm σdm t+1 (16) λ z,t+1 A main feature ditinguihe the firm behavior from the tandard New Keyneian one: they face a cah-in-advance contraint for which they need holding money to elf-financing. 7

The problem faced by the repreentative firm i: max E t k=0 r t+ Q t+ P t+ ξ π [( P t P t ) 1 1 λπ k =1 ( π t++1 π t+ )y t+k r k k f i,t+k w t+k h i,t+k ( )] + mc i,t+ [(u t+k k 1 + r t+) f α (z t+ h d,t+ ) 1 α ψ t+ The firt order condition for the firm are: 2.1.3 Entrepreneur K π,t = y t p t λπ 1 λπ F π,t = mc t y t p t λπ 1 λπ ( P t P t ) λπ 1 λπ 1 = (1 ξ π ) p t 1 1 λπ λ z,t+1 + ξ π βe t ( π t ) 1 λ z,t π t+1 1 λπ ( λ z,t+1 + ξ π βe t ( π t ) 1 λ z,t π t+1 1 λπ ( k ( π t++1 )y t+k ] π t+ =1 + ξ π ( π t 1 π t ) 1 1 λπ (17) p t = 1 K π,t λ π F π,t (18) p t ) 1 λπ 1 K π,t+1 p t+1 (19) p t ) 1 λπ 1 F π,t+1 p t+1 (20) y t π = (u t+k k f t+) α (z t+ h d,t+ ) 1 α ψ (21) π = (1 ξ π ) p t λπ 1 λπ + ξ π ( π t π t 1 ) λπ λπ 1 π,t 1 (22) Conidering the entrepreneur among the actor allow to introduce a financial ector characterized a in Bernanke et al. (1999). The credit upply ide i characterized by perfect competition and aymmetric information deigned true a monitoring cot paid by the houehold/bank if the credit i not repaid. The demand ide i deigned uch that repreent a riky invetment. There are a large number of entrepreneur who combine net worth and loan to buy capital, that at the end of period i tranformed according to the productivity of the project (ω i a hock log-normal ditributed, with time varying variance, the rik premium hock σ t ). The entrepreneur repaid the debt only if the productivity i above a threhold that determine jointly with the interet rate, the loan contract. Allowing for the productivity hock, the financial friction are introduced in the form of aymmetric information due to the productivity be oberved by the bank at a monitoring 8

cot. After oberving the hock, the entrepreneur determine the price of capital utilization and rent out capital ervice to the firm. In order to avoid that entrepreneur et enough net worth that they are no longer credit contraint, each period they exit the market with probability ι and their net worth i in part conumed and in part given to the houehold. Each period enter the market a many new entrepreneur a the population remain contant. Therefore, defining repectively the probability of default G(ω t, σ t ) = ωt and the gro hare of profit accruing to the bank 0 ω t df(ω t, σ t ) Γ( ω t ) = ω t (1 F(ω t, σ t )) + G(ω t, σ t ) the entrepreneur maximization problem can be ummarized a follow: max E t k=0 [(u t+k r K t+k a(u t+k)) + (1 δ)q t+k ]ω t+k K t+k S.T. Firt Order Condition Γ( ω t+1 ) µg(ω t+1, σ t+1 ) = 1 + r t 1 + r k,t+1 (1 n t+1 Q t K t+1 ) (23) n t+1 = ι π t [r k,t r t 1 µg(ω t, σ t )]K t Q t 1 + ι π (1 + r t 1)n t (24) credit t+1 = Q t K t+1 n t+1 (25) 1 + r k,t+1 = (u t+1r K t+1 a(u t+1) + (1 δ)q t+1 )K t+1 π t+1 Q t (26) (1 Γ( ω t+1 )) 1 + r k,t+1 Γ ( ω t+1 ) + 1 + r t Γ ( ω t+1 ) µg (ω t+1, σ t+1 ) r K t = a (u t ) (27) ( 1 + r k,t+1 1 + r t (Γ( ω t+1 ) µg(ω t+1, σ t+1 ))) = 0 (28) 9

2.2 Bank A propoed by Chritiano et al. (2010), the repreentative bank combine feature of a commercial bank and invetment bank. It founding are repectively compoed by houehold depoit for the commercial part, and hort term aet and marketable ecuritie for the invetment part. Similarly it aet portfolio i compoed by riky aet and reerve. The liquidity ervice, provided by the bank for founding purpoe, origin another real friction in the model: aimetric information between houehold and bank. A credit crunch i due to either a contraction in entrepreneur demand for credit becaue the pread rie, or a contraction in the loan upply a the bank found themelve at an higher cot. On the output market the bank are competitive and the profit retained from the riky loan activity are equal to zero. A preented before the zero profit condition for the bank i: ωt+1 [(1 F( ω t+1 ) ω t+1 + (1 µ) ωdf(ω)](1 + r k,t+1 )Q t+1 K t+1 = (1 + r t+1 )B t+1 0 The part on the right of the equation i minimum gain the bank ha to retain from the next period loan. The left hand ide correpond to the found that the bank received at the end of the period: repectively the interet from the non-bankruptcy entrepreneur and the net entrepreneurial loan ubidiary from bankrupt entrepreneur. From the liabilitie ide of the balance heet, we can ee that the bank create three different liabilitie: bank depoit, hort term marketable ecuritie and hort term aet. The firt two clae provide liquidity ervice to the houehold, and are produce by the bank by tranforming the reerve and the capital, according to a Cobb- Dougla production function. d t + ςdm t = χ b (u t k t b) ϵ E (1 ϵ) The reerve are included in the production function to conider the precautionary aving of a bank concerned about the poibility of unexpected withdrawal and bank run. The depoit conit in the cah reerve for the bank, but only a part can be ued in the production. The other amount, τ b d t, i the minimum reerve required to hold againt depoit. E t = (1 τ b )d t The hort term marketable ecuritie and the hort term aet are iued at the end of the period to finance the new bank riky loan: credit t+1 = dm t+1 + a t 10

Given the balance heet decribed above and conidering that the bank face a monopolitic competitive market on the input market, the rapreentative bank optimizing problem can be ummarized a follow: S.T. F.O.C. max Π t = (1 + r t )credit t + d t + a t + dm t+1 credit t+1 (1 + r d t )d t (1 + rt dm )dm t (1 rt a )a t 1 rt K u t kt b d t + ςdm t = χ b (u t k b t ) ϵ ((1 τ b )d t ) ( 1 ϵ) (29) r t+1 = r a t+1 (30) r dm t+1 r t+1 = λ b,t+1 ς (31) kt b rt K u t = λ b,t (d t ςdm t ) (32) r d t = λ b,t (1 π t (1 dm t d t )) (33) 2.2.1 Capital producer A in Chritiano et al. (2010), the capital producer buy the capital from firm and bank (K t = k f t + kt b ), and generate new phyical capital, being contrained by a convex adjutment cot function for the invetment. S.T. max Firt Order Condition β k λ z,t+k (Q t+k K t+k Q t+k K t+k (1 δ) I t+k ) k=0 K t+1 = (1 δ)k t + S (I t, I t 1 )I t (34) 1 Q t S I t = 1 1 + r t+1 Q t+1 S I t+1 (35) 11

2.2.2 Market clearing condition and functional form The market clearing condition for product and factor repectively are: (u t+k k t+ ) α (z t+ h d,t+ ) 1 α ψ = π [c t + I t + a(u t )K t + µg(ω t+1, σ t+1 )(1 + r k,t )Q t 1 K t π t The main functional form are et a follow: + 1 ι (net t+1 W e )] ι (36) h = w h d,t (37) a(u t ) = γ 1 (u t 1) + γ 2 2 (u t 1) 2 (38) S It,I t 1 = 1 χ 2 (1 I t I t 1 ) 2 (39) F( ω t, σ t ) = 1 ln( ωt) µ σln t 2π 0 e ω2 t 2 dω (40) Finally to cloe the model, the monetary policy i the optimal one propoed by Guerello (2012), pecifically: 2.2.3 Shock r t = α R r t 1 + α π π t+i + α Y y t+i + α c credit t+i + ε (41) Three different hock are conidered in the model: a technology hock, a financial hock, and a monetary one. They are able to account for the mot of the volatility of real economy, financial ector, and interet rate. Specifically the technology hock affect directly the the production function of the repreentative firm. It can be define a an unobervable component that upward hift the production function, while maintaining the input contant. Defining the financial hock, the work of Chritiano et al. (2010) i followed. They propoed a financial hock that i related to the bank expectation about the productivity of the rik invetment. They proved that thi hock, i able to explain roughly 60% of macroeconomic variability. The rikine hock i deigned a the variance of the productivity ditribution and affect the external rik premium through the number of bankruptcy project. The monetary hock i included to account for the tochatic deviation from the optimal rule. The oberved interet rate in the lat decade ha diverged from the optimal one, howing that there are other factor that concur in determining the monetary policy. Since the 12

divergence from the optimal interet rule have been quite relevant (above the 10%), it i important to conider a tochatic component in the interet rule, of the form of an additive hock. All the hock dynamic are deigned a an autoregreive tochatic component, with conditional heterochedatic error. Specifically: 2.2.4 Calibration z t =ρ 0,z + ρ 1,z z t 1 + var(z t 1 )ϵ t 1 (42) σ t =ρ 0,σ + ρ 1,σ σ t 1 + var(σ t 1 )ϵ t 1 (43) ε t =ρ 0,ε + ρ 1,ε ε t 1 + var(ε t 1 )ϵ t 1 (44) The calibration follow the work of Schmitt-Grohe. and Uribe (2004), Chritiano et al. (2010) and Guerello (2012) a reported in the Table(1). In teady tate the labor input and the capacity utilization are et to unit, and the profit to zero. The target for inflation ha been choen uch that the optimal interet rate matche the long run interet data, a reported in Chritiano et al. (2010). The monetary policy parameter are taken from Guerello (2012). They are choen a to maximize the unconditional houehold utility, in a model with financial friction but without bank agent and liquidity ervice. The interet rule choen imply a quite mild repone to inflation, but a negative repone to both output gap and credit. The flexible monetary policy propoed allow the central bank to overcome the effect of both technology and financial hock. Additionally other impler rule are conidered. 2.2.5 Impule repone function In order to ae our prior expectation for the empirical part of thi paper, it i intereting to invetigate the dynamic of the model decribed in thi ection after movement in the interet rate. The model i perturbed by two different hock: an unexpected monetary hock in order to collected the movement of the interet rate that are not related to the movement in the interet rule target variable, and a technology hock that i the main hock in explaining the variability of GDP and inflation. A reported in table (6), after an unanticipated monetary hock the interet rate jump but jut after only a period experience drop larger then the initial jump due too the adjutment to the target variable. The driving force of the economy i the cot of capital. From eq(35) we can ee that a the interet rate increae the expected cot of capital increae too with ome degree of inertia due to the convexity of the invetment adjutment cot function. 13

The increaing movement of the Tobin Q lead to a built up in both the invetment and the capital, a well a credit and GDP. However, after a period, when the cot of the hort marketable ecuritie i determined to an higher level, the higher cot of founding puhe the bank to tight the upply of loan. Hence, a the upply of loan drop and the demand diminihe too a the net worth increae, the total loan drop uddenly after a period. The bank rik meaure how different reaction. The credit rik, proportional to the productivity minimum no default level, increae a the external premium increae, but a the interet rate drop in the econd period it revert to the teady tate value. Intead, the leverage rik, meaured a the ratio of aet to depoit, i influenced by movement in depoit and hort marketable ecuritie. The former increae uddenly, the latter with a lag of retard. Hence the leverage jump in the firt period and then after how a cyclical movement. Similar variable dynamic i oberved after a technology hock. However a technology hock both increae GDP and decreae inflation. hence it i a negative anticipated monetary hock. What come out from the analyi of the imulated dynamic of ome central bank i that unexpected or anticipated monetary hock have oppoite effect on the bank rik appetite. Since the damageful component of the monetary policy i the tochatic part, the only way to mooth it effect, with regard to the optimal operative interet rate rule, i to chooe a quite paive monetary policy (high degree of interet rate moothing). A reported in table(6) by the dotted line, if interet rate rule i deigned a paive and reacting to credit the effect of any interet rate hock, either unexpected or anticipated, on the leverage rik are moothed and the effect on the credit rik are not relevant. Hence, a long a the monetary policy i paive, the bank rik i not feed by the monetary policy a credit rik doe not repond to interet rate and leverage how a mall waving movement around the teady tate. Therefore any danger of boom buting cycle i contained. 3 Regreion pecification Rik meaure The firt purpoe of the eay i to invetigate the tranmiion channel of monetary policy to bank leverage and credit rik. In order to identify the different tranmiion channel, two different indicator of rik are adopted. The firt i a meaure of the leverage that approximate the rik due to the balance heet movement: when the balance heet become wider, and it i financed by hort term liabilitie, the probability of default increae becaue the bank i more expoed to mimatching problem. Specifically the ratio of aet to depoit i employed, a in Angeloni et al. 14

(2010), becaue i a direct meaure of the bank capitalization. It i preferred to pure meaure of the bank default probability, a EDF or Z-core, largely ued in literature 1 becaue they account for the health of the bank and the ytem, and do not reflect excluively the bank rik management. The econd meaure, approximating the credit rik, report the quality of the loan portfolio. Among the many meaure employed 2 the mot ued i the ratio of non-performing loan to total loan. It ummarize the current health of the bank loan portfolio, and indicate the percentage of loan that would probably default or generate lo for the bank in the hort term. Interet rate meaure Different meaure of interet rate have been employed, the average marginal refinancing rate and the overnight inter-banking rate, a well a, meaure of the monetary policy tance computed a the Taylor gap, the difference between the optimal interet rate rule precription and the current interet rate. Three optimal interet rate rule have been conidered, which coefficient are taken to maximize the conumer utility in a DSGE model with financial friction (for detail ee Guerello (2012)). Specifically: 3 i t ln 2.002 =1.5 ln in f lation t 2 i t (45) ln 2.002 =1.2956 ln in f lation t 0.0239 ln outputgap t (46) 2 i t ln 2.002 =0.9489 ln i t 1 2.002 + (1 0.9489)(3.6516 ln in f lation t 2 0.2192 ln outputgap t 2.5088 ln loangap t ) (47) 1 Gambacorta (2009) and citetagm10 ue EDF for Euro Area bank and Ozuca and Akbotanci (2012) employed either Z-core or EDF for Turkih bank. In addition etimated meaure of the bank default probability are employed. i.e. Karapetyan (2011) ue a logit model baed on balance heet data to generate a meaure of bank rik for Norwegian bank. 2 Angeloni et al. (2010) chooe the ratio of credit and mortgage loan on total loan, Deli and Koureta (2011) ue both riky aet to total aet and non performing loan to total loan, Briimi and Deli (2010) prefer a wider meaure ummarizing the problem loan to total loan. 3 Outputgap and loangap have been calculated a the detrended value of GDP and gro loan 15

Other factor The mot of the literature ue uni-variate model to tet the rik taking channel. It i preferred to etimate the effect of the monetary policy on bank rik alongide with the effect of bank rik on the buine cycle. A VAR model i etimated, incorporating macro and financial variable: a meaure of country real GDP and the main financial market index. Thi approach ha everal advantage: it overcome the problem of endogeneity of either the interet rate or Taylor gap, a well a, educe the repone of credit rik and leverage to movement in the interet rate. Converely many influential paper 4 underline the heterogeneity of the bank repone to the monetary hock. Of particular interet, the work of Briimi and Deli (2010), by a local GMM, invetigate the heterogeneity of bank to the extent of the credit rik repone to monetary policy. They found that large and healthy bank do not repond to interet rate for what concern their loan portfolio management trategie. Looking at the pat literature, in thi invetigation micro level data at bank level are combine with macro variable at country level. Data decription The databae i incluive of both micro data (credit rik, aet to depoit ratio and loan to aet ratio) of a et of almot 100 bank, taken from Capital IQ data-et, and macro data (GDP, market index and interet rate) of 10 European countrie 5 taken from the ECB Data-Warehoue. The ample cover a period of 9 year from 2003 to 2011, incluive of the recent financial crii. The ample frequency i annual. Even if while dealing with monetary policy data i better to ue quarterly data, Altunba et al. (2010), Altunba et al. (2011) and Briimi and Deli (2010) howed the reliability of analyi baed on annual data. The panel i unbalanced with a mean length of roughly 7 year. However the reult are not biaed becaue the problem i purely informative and the enter/exit of bank from the ample i not related to the variable examined. The variable ummarizing tatitic are reported in table(6). 4 Altunba et al. (2011), Deli and Koureta (2011) how how the degree of bank rik depend on the bank characteritic (liquidity, capitalization and dimenion); Altunba et al. (2010) found it true for loan too. Finally, Deli et al. (2012) and Briimi and Deli (2010) highlight that more pecifically the bank pecific characteritic hape the tranmiion channel. 5 The country are elected a all the countrie inide the Euro Area from 2002, le the Greece due to the recent economic turmoil. Specifically the countrie are: Autria, Belgium, France, Germany, Ireland, Italy, Luxemburg, Netherlander, Portugal, and Spain 16

Econometric pecification For the etimation of the model, a ix variable panel VAR i employed, in order: log of loan to aet, log of GDP, log of non-performing loan to total loan, log of aet to depoit, log of interet rate, log of market index. The panel VAR methodology allow for uing highly diaggregated data and account for ome degree of heterogeneity among the individual and countrie conidered, even if the repone i conidered the ame for all the ubject. The panel VAR methodology combine the traditional VAR approach, which treat all the variable of the ytem a endogenou, and the Arellano and Bond (1991) GMM etimator for panel data that allow for unoberved individual heterogeneity. Allowing a bank fixed effect to enter the equation, it i accounted for the bank pecific heterogeneity. Y i,t = A 0 + A(l)Y i,t 1 + α i + u i,t i = 1, 2,..., N t = 1, 2,..., T (48) The code and the trategie propoed by Love and Zicchino (2006) are followed. A uggeted by Canova and Ciccarelli (2013), if the data generating proce featured dynamic homogeneity, the bet etimator i the pooled etimator with fixed effect, becaue it capture idioyncratic but contant heterogeneity acro variable and unit, and thi i the cae. Since the bank fixed effect are correlated with the regreor 6 the mean differencing procedure produce biaed coefficient. It i better to ue a forward mean differencing procedure called Helmert procedure, that preerve the orthogonality between the tranformed variable and the regreor. Thi analyi i centered on the tudy of the orthogonalized impule repone function, and, due to the mall ample, the confidence band are etimated by a Monte Carlo imulation. In addition, the variance decompoition i advanced to determine the relevance of different monetary policy component for the volatility of the two rik variable. 4 Reult and dicuion 4.1 Some conideration on the baeline model In the firt place, we report what a ample of almot 100 European bank ay about the rik taking channel in Europe. The analyi i baed on the impule repone function over a period of 10 year. The impule repone function decribe the reaction of a variable to a unit innovation in another variable in the ytem, while holding the other hock equal to zero. 6 Lag of the dependent variable enter the equation a regreor and the dependency of the variable to the fixed effect hold at each point in time, hence alo for the lagged variable. 17

Since the actual variance-covariance matrix i unlikely to be diagonal, it i neceary to adopt a Choleky decompoition to obtain an orthogonal matrix. A the order of the variable et the identifying aumption, the variable that come earlier contemporaneouly affect the following variable and with a lag. In thi way the variable that come earlier are weakly exogenou becaue are affected by the later variable only with a lag. The order of the variable ha been choen following Angeloni et al. (2010). The underling aumption i that all the hock influence intantaneouly the tock market index and the interet rate, then balance heet variable adjut (it i irrelevant in which order) and the macro variable, in thi cae GDP, i the lat one. Since Deli et al. (2012) highlight that hock in bank rik appetite intantaneouly influence the amount of loan through the creation of new loan, we decide to place the growth rate of loan before the balance heet item, hence we aume that bank rik hock affect directly the amount of credit. The reulting order i GDP, leverage rik, credit rik, loan, interet rate, tock market index. Looking at table(6), the growth rate of total loan increae after a poitive hock in either GDP and real interet rate. For what concern the effective bank rik, we found clear evidence of the rik taking channel in the reult: indeed both leverage rik and credit rik increae after an expanionary monetary policy. In addition, table(6) point out that bank take over le rik after either an economic or financial expanion. It i confirmed the idea of Deli et al. (2012): indeed the bank profitability determine the individual level of rik, and a the profitability i pro-cyclical, the bank rik i countercyclical. Furthermore, a dicued by Borio and Zhu (2012), both a balance heet expanion and a built up in credit rik, if not exceive, i beneficial for the real economy. The problem arie when the force are no longer balanced and an overtaking in bank rik lead to a bank crii, a it recently happened. Thee reult are robut to different interet rate meaure (ee table(6)) and leverage meaure (Z-core).Many influential paper ued country or time fixed effect to take out any exogenou hock that affect the bank. Therefore, we conider, a in Love and Zicchino (2006), country-time fixed effect, taking out the country mean of each ingle year from the micro level variable (loan, leverage and credit rik). Table(6) how that the reult remain robut and, hence, we decide to relay on a model without country-time FE in the ret of the analyi. We believe that GDP and tock market index are able to capture the mot of the exogenou hock that affect the bank rik at an aggregate country level. 18

4.2 Capital regulation and rik taking channel Borio and Zhu (2012) identify two way in which the bank regulation, in particular the minimum capital tandard, affect the bank attitude toward rik. A thigh minimum threhold reduce the bank ability to extend their balance heet becaue enlarging the capital bae i more expenive than alternative founding reource. In the ame way it limit the ability of the bank to generate new loan, a it ha been prooved by Maddaloni and Peydró (2011). In 2006 a new bank regulatory act, called Bael II, ha changed the bank uperviion framework, and among the other, ha tightened the minimum capital requirement. In addition, it ha modified the way bank compute the rik of an aet, and, hopefully, ha changed the way they perceived the rik. Therefore it i intereting to invetigate how the rik taking tranmiion channel ha modified after 2006. A firt tep ha been run the analyi over two different ub-ample: before and after 2006. Table(6) report the finding. It i clear that after 2006 bank, to the extend of their rik management department, have become quite unreponive to monetary policy. There are not any evidence of the rik taking channel after the introduction of Bael II act. In order to invetigate the effect of Bael II act in a more detailed and robut way, we conider an interaction term in the analyi, following Maddaloni and Peydró (2011). A time dummy variable equal to 1 after 2006 i interacted with the real interet rate. Looking at table (7), we can ee that the effect of monetary policy i moothed with reference to leverage and loan. Thi upport the finding of Maddaloni and Peydró (2011), that the impact of monetary policy on the oftening of lending tandard i reduced by more tringent policy on either bank capital requirement. However, the effect of monetary tance on credit rik i enhanced by the new uperviory regime. A trong collaboration between monetary policy and bank uperviion, a advied by Agur and Demertzi (2012), eem to be able to...limit the amplification effect related to the rik-taking channel and increae the reilience of the banking ytem againt hock (Borio and Zhu (2012)). However i able to affect only the rik relative to the expanion of the balance heet, but it i unable to affect the quality of the credit undertaken by the bank. 4.3 Taylor Gap and rik taking channel Fata et al. (2009) how that worldwide the monetary policy wa over-looing before the recent financial crii. Indeed in the mot advanced economie, a US and Euro Area, the Taylor gap were negative, on average, one to three year before the but. It wa followed 19

by a harp tightening of monetary policy in 2007. Looking at table(6), bank rik i reponive to real interet rate but not to the nominal interet rate. Thi ugget that the nominal interet rate component (repone to inflation and Taylor gap) cancel out each other effect on bank rik. Even if it ha been proved that a looing monetary policy doe not affect bank rik, houe and aet price, few ha been aid on the effect of a prolonged negative Taylor gap on bank appetite for rik. In addition, Fata et al. (2009) argue the advantage of including ome macro prudential indicator in the imple Taylor rule. Many paper upport the uperiority of flexible credit augmented Taylor rule, over the imple one, in preventing the but. We decide to extend our analyi to different meaure of monetary policy. We conider both the effect of the precribed optimal rule and the Taylor gap, computed a the difference between the nominal interet rate and the precribed one.in addition to the claical benchmark, the Taylor rule eq(45), we conider two flexible optimized Taylor rule, a largely ued rule that conider both inflation and output-gap eq(46), and an augmented one, comprehenive of interet rate moothing, output-gap and credit eq(47). Looking at table(6), in which are reported the impule repone function for model including an alternative monetary policy variable at time, it i clear that a looe monetary policy, to the extend that it i the optimal repone to inflation and macroeconomic fluctuation, doe not built up bank rik. However, an over-accommodating monetary policy, incentive bank rik taking, both leverage and credit rik, and could lead to dangerou burn but cycle. Indeed the negative correlation between either the leverage rik or credit rik meaure and the Taylor gap prove that, ince the Taylor gap ha recently been negative, an increae in Taylor gap increae the bank propenity toward rik. Thi finding reinforce the reult of Altunba et al. (2011), becaue, in addition to prove the negative correlation between Taylor gap and bank rik, it ha been pecified that thi relation hold for both leverage and credit rik. Once conidered a credit augmented Taylor rule, the analyi of the impule repone function tell that the reult remain robut for what concern the credit rik. However, the optimal interet rate doe not affect the leverage anymore, and the leverage repond, till negatively, to the Taylor gap only in the long run (from 2 to 4 year). Therefore, once accounted for credit in the operational rule, the interet rate doe not affect the leverage, and an over-looing monetary policy affect credit rik only if prolonged. On the other ide, the Taylor gap i till able to generate boom but cycle through the built up of credit rik. While table(6) provide information on the dynamic repone of the bank rik variable 20

to different monetary hock, table(9),table(10) and table(11) jointly how the forecat error variance decompoition and allow for aeing the relative importance of each of the monetary hock. Specifically thee table report the percentage of variation in each variable that i explained by a hock in another variable, accumulated over time. We report the total effect accumulated over three different period: hort run (the two year horizon), medium run (the ix year horizon) and long run (the ten year horizon). Differently from what found by Buch et al. (2010), macroeconomic factor, a well a, monetary policy do not play an important role in explaining both credit rik and leverage fluctuation. Even if macroeconomic hock together explain an increaing percentage of leverage variance (from four percent in the hort run to twelve percent in the long run), the role of monetary policy i only marginal. In the medium-long horizon the two monetary policy hock together explain roughly five percent of the aet to depoit variance, but in the hort run thi percentage i le than one. Among the two hock, the random hock are able to explain more than eighty percent of monetary policy contribution, but looe relevance a the time horizon increae. The monetary policy role i even more marginal if we conider a augmented Taylor rule with total credit (eq(47)). In thi cae, while macroeconomic hock jointly explain from five to nine percent of leverage fluctuation, monetary policy hock doe not explain more than one percent of aet to depoit ratio in either the hort or long run. Buch et al. (2010) found that, for US bank, non performing loan ratio variance i explained ubtantially (27 percent) by macroeconomic hock. Therefore, it i urpriing that for Europe we found that macroeconomic factor are not able to explain more than five percent of the non performing ratio, in the medium and long run, and le than one percent in the hort run. Thi reult may be affected by the different, and le tight, claification of the non performing loan in Europe. Finally it i intereting to notice that credit rik and leverage hock repectively explain the other rik meaure variance in a greater percentage than the monetary policy and jut little maller than the aggregate macroeconomic hock. Thi how how the rik meaure uually move together, trongly increaing the bank overall rik and default probability. 4.4 Monetary union and Taylor Gap The countrie inide the Euro Area uually face a larger effect of Taylor gap, a the optimal monetary policy i decided at central level, looking at aggregate level variable. The peritent and large difference among the performance of the member economie, amplify the effect of the Taylor gap on the bank rik. Indeed bank of mall and medium ize look at the macroeconomic performance of the country in which they operate, rather 21

than at the aggregate performance of the European economy. Therefore it i intereting to evaluate how the monetary policy at central level can affect the bank appetite for rik, and which portion of bank rik variance i explained by the difference in the member macroeconomic performance. We contruct new variable for the optimal operative interet rule. In thi cae the benchmark interet rate i computed uing aggregate variable for CPI, GDP and total credit. Table(6) report the impule repone function for the three different Taylor rule conidered. Surpriing the reult reported for the ingle country optimal rule, are not longer upported. A the Taylor gap i negative, if the difference between the marginal rate and the benchmark increae, the bank rik decreae. However, a a more ophiticated benchmark i conidered (ee eq(47)), an over-looing monetary policy to the extend that doe not concern the financial tability objective, increae credit rik. Finally, a reported in table(13), table(14) and table(15), we perform a forecated error variance decompoition analyi. In addition to the optimal rule and the Taylor gap, we conider the difference between the agrregate optimal rule and the ingle country one. In thi way we ae the relevance of the different macroeconomic performance in explaining the credit rik and leverage. It i clear that a monetary union pay a cot in term of incentive to bank rik appetite. The larger Taylor gap i able to explain half of the monetary policy contribution to the credit and leverage rik variance. There withal, the Taylor gap computed at aggregate level explain le than one percent of the bank rik variance, in both it form (optimal operational rule and random hock). 5 Concluion The current literature provide ome evidence in favor of an active rik taking channel in Europe, that are not concluive. We tet both the preence of thi monetary tranmiion channel and the relevance of the different factor concurring in defining the monetary policy. Even if our reult provide evidence of an active rik taking channel, further conideration can be extrapolated from the reult over the mechanim of thi tranmiion channel. The analyi of the DSGE model propoed how that after a monetary hock it become more expenive for the bank to found themelve by hort-term aet, and hence it lead to a credit crunch. Actually the leverage jump after a negative monetary hock and drop after a poitive technology hock, hence a negative anticipated monetary hock. 22

In the ame way, conidering both the optimal rule and the Taylor gap in the regreion, the primary role played by the Taylor gap i highlighted, a well a, it ha hown a the bank rik increae after a adding up in the Taylor gap (in abolute term). However it ha hown that leverage and credit rik are moothed by a looe monetary policy, if it i precribed by the forecated value of inflation, GDP and credit. In addition, from forecated error variance decompoition analyi, it i clear that the monetary policy i not a relevant in explaining bank rik fluctuation a the other macroeconomic factor (GDP and market index). Wherea among the component of the monetary policy, the tochatic factor (Taylor gap) i ten time more powerful in explaining leverage volatility than the precribed interet rate. From the Central Bank point of view, the good new i that the monetary policy, to the extend of the optimal operational interet rule, i not affecting the bank rik appetite. The evidence of an active and damageful rik taking channel for the monetary policy, reflect the role played by the Taylor gap. Therefore, if the monetary policy i over-looing for a long time, the neutral role of the central bank in determining the level of bank rik could be overcome by feeding a damageful boom buting cycle. In concluion, a looe monetary policy i not damageful by itelf, but it could be riky if the interet rate i kept far away from what the real economy need. A the Taylor gap i mainly determined by policy objective miperception and the overconfidence on the weight the Central Bank give to the financial tability, uing different monetary policie, we hown that ha the objective of the financial tability i clearly tated, the Taylor gap effect i moothed, at leat for what concern leverage. Additional minor concluion are drawn from the reult. Firt it ha been proved that a joint work of Central Bank and Bank Regulator i beneficial for the economy. The work of Agur and Demertzi (2012) argue that a joint policy of Bank Regulator and Central Bank, whenever different intitution, allow to prevent and face any financial turmoil without huge cot for the economy in peace time. The reult of the paper upport and add to thi theory. Indeed they how that the rik taking channel ha been almot neutralized by the introduction of a more tightening capital requirement. It upport the improvement of the Bank regulation in the Euro Area by the Bank Union, a a unique intitution could coordinate in a better way the two policie. Finally omething ha been aid on the problem of the lack of convergence of the Euro Area fical policie. It ha been hown that a monetary union, a EMU, exacerbate the Taylor gap effect. The rik taking channel ha been amplified by the larger Taylor gap, a different countrie and economie need different level of interet rate. In addition, it ha been hown that the difference between the aggregate optimal operational rule and the 23

country one i able to explain roughly 90% of the monetary policy contribution to bank rik volatility, playing a central role. For the ECB, the rik taking channel i a factor to conider more accurately than for other Central Bank, becaue it could be damageful even if the monetary policy would be tighter that in the lat decade. 24

Reference T. Adrian and H. Shin. Financial intermediarie and monetary economic. FRB of New York Staff Report, (398), 2010. I. Agur and M. Demertzi. Exceive bank rik taking and monetary policy. ECB Working Paper Serie, 2012. Y. Altunba, L. Gambacorta, and D. Marque-Ibanez. Bank rik and monetary policy. Journal of Financial Stability, 6(3):121 129, 2010. Y. Altunba, L. Gambacorta, and D. Marque-Ibanez. Doe monetary policy affect bank rik taking? ECB Working Paper Serie, 2011. I. Angeloni, E. Faia, and M. LoDuca. Monetary policy and rik taking. ECB Working Paper Serie, 2010. M. Arellano and S. Bond. Some tet of pecification for panel data: Monte carlo evidence and an application to employment equation. The Review of Economic Studie, 58(2): 277 297, 1991. B. S. Bernanke, M. Gertler, and S. Gilchrit. The financial accelerator in a quantitative buine cycle framework. Handbook of macroeconomic, 1:1341 1393, 1999. C. Borio and H. Zhu. Capital regulation, rik-taking and monetary policy: a miing link in the tranmiion mechanim? Journal of Financial Stability, 8(4):236 251, 2012. S. Briimi and M. Deli. Bank heterogeneity and monetary policy tranmiion. ECB Working Paper Serie, 2010. C. Buch, S. Eickmeier, and E. Prieto. Macroeconomic factor and micro-level bank rik. CESifo Center for Economic Studie & Ifo Intitute for economic reearch, 2010. G. A. Calvo. Staggered price in a utility-maximizing framework. Journal of monetary Economic, 12(3):383 398, 1983. F. Canova and M. Ciccarelli. Panel Vector Autoregreive Model: A Survey. Centre for Economic Policy Reearch, 2013. L. Chritiano, R. Motto, and M. Rotagno. Financial factor in economic fluctuation. ECB Working Paper Serie, 2010. 25

M. Deli and G. Koureta. Interet rate and bank rik-taking. Journal of Banking & Finance, 35(4):840 855, 2011. M. Deli, I. Haan, and N. Mylonidi. The rik-taking channel of monetary policy in the ua: Evidence from micro-level data. SSRN working paper erie, 2012. E. Farhi and J. Tirole. Collective moral hazard, maturity mimatch and ytemic bailout. American Economic Review, 2013. A. Fata, K. Prakah, R. Pau, and S. Aladair. Leon for monetary policy from aet price fluctuation. In World Economic Outlook, chapter 3. International Monetary Found, October 2009. L. Gambacorta. Monetary policy and the rik-taking channel. BIS quarterly review, 2009. C. Guerello. Optimal operational monetary policy in a medium ize dge model with financial friction: the effect of financial rik variability. Univerity of Warwick - MSc Economic thei, 2012. A. Karapetyan. Credit, houe price, and rik taking by bank in norway. NB taff memo, 13, 2011. I. Love and L. Zicchino. Financial development and dynamic invetment behavior: Evidence from panel var. The Quarterly Review of Economic and Finance, 46(2):190 210, 2006. A. Maddaloni and J. Peydró. Bank rik-taking, ecuritization, uperviion, and low interet rate: Evidence from the euro-area and the u lending tandard. Review of Financial Studie, 24(6):2121 2165, 2011. E. Nier and Q. Merrouche. What caued the global financial crii? evidence on the driver of financial imbalance 1999-2007. IMF working paper erie, 2010. E. Ozuca and E. Akbotanci. An empirical analyi of the rik taking channel of monetary policy in turkey. Technical report, ERC-Economic Reearch Center, Middle Eat Technical Univerity, 2012. R. Rajan. Ha finance made the world rikier? European Financial Management, 12(4): 499 533, 2006. S. Schmitt-Grohe. and M. Uribe. Optimal operational monetary policy in the chritianoeichenbaum-evan model of the u buine cycle. NBER working paper erie, 2004. 26

T. Yun. Nominal price rigidity, money upply endogeneity, and buine cycle. Journal of Monetary Economic, 37(2):345 370, 1996. Acknowledgement I am very grateful to Prof. Marco Mazzoli for hi through comment at different tage of thi work. The ueful converation with Prof. Riccardo Lucchetti are alo gratefully acknowledge. Table and graph Thi ection contain all the table and graph decribing the feature of the theoretical model preented and the empirical reearch run Specifically table containing the model calibration and graph of the dynamic of the model after the main hock conidered. For what concern the empirical reult, in addition to the ome variable tatitic, all the impule repoe function graphe are provided along with the forecat error variance decompoition table, when neceary. 27

Table 1: Calibration Param. Value Decription Param. Value Decription β 0.999 Subjective dicount factor σd -1.2852 Curvature of utility of depoit b 0.56 Degree of habit peritence φdm 4068 Preference parameter S.T. marketable ecuritie θ 0.36 Share of Capital in value added φd 0.00010169 Preference parameter depoit ψ 0.5033 Fixed cot parameter W e 0.009 Lump-um tranfer to houehold to entrepreneur δ 0.02 Depreciation Rate F( ω 0.15 Percent of buine that go into bankruptcy λw 1.05 Steady State Mark-up upplier of labor ι 97.62 Percent of entrepreneur who urvive λπ 1.2 Steady State Mark-up intermediate good firm (1 µ) 0.77 Fraction of realized profit in bankruptcy ξw 0.712 Degree of wage tickine ς 0.889 Share of S.T. marketable ecuritie in depoit ervice technology ξπ 0.703 Degree of price tickine ϵ 0.94 Power of exce reerve in depoit ervice technology σl 1 Curvature of diutility of labor χb 102 Contant in front of depoit ervice technology φl 1.0772 Preference parameter labor τb 0.02 Bank reerve requirement φm 0.5393 Preference parameter money κ 2.92 Capital adjutment cot σm -2.6418 Curvature of utility of money γ1 0.042 Parameter of capacity utilization function σdm -3.43 Curvature of utility of hort term marketable ecuritie γ2 0.0017 Parameter of capacity utilization function Exogenouly determined teady tate value K HD 1 Labour demand b K 0.0093 Fraction of capital ued in bank π * 1.84 Inflation target QQ 1 Tobin Q U 1 Capital Utilization RK 0.042 Return on capital Shock calibration ρ 0 M 0 Contant in Monetary hock equation ρ M 0.9963 Autoregreive coefficient in Monetary hock equation ρ R 0 0 Contant in Rikne hock equation 1 M ρ R 1 0.6750 Autoregreive coefficient in Rikne hock equation ρ Z 0 0 Contant in Technology hock equation ρ Z 1 0.9020 Autoregreive coefficient in Technology hock equation The calibration value are taken from Schmitt-Grohe. and Uribe (2004), Chritiano et al. (2010) and Guerello (2012) 28

Table 2: Impule Repone Function after either technology or monetary hock IRF after a technology hock interet rate GDP Tobin Q Total loan Aet to Depoit NPL ratio Interet Rate GDP Tobin Q Total Loan Aet to Depoit NPL ratio 15 140 30 25 25 90 rule1 rule1 rule1 rule1 rule1 rule1 rule2 rule2 rule2 rule2 rule2 rule2 rule3 rule3 rule3 rule3 rule3 80 rule3 120 20 20 25 10 70 100 15 15 20 60 80 10 10 50 15 60 40 10 40 5 20 20 5 5 30 10 10 10 10 0 2 4 6 8 10 12 14 16 18 20 15 0 2 4 6 8 10 12 14 16 18 20 20 0 2 4 6 8 10 12 14 16 18 20 5 0 2 4 6 8 10 12 14 16 18 20 15 0 2 4 6 8 10 12 14 16 18 20 15 0 2 4 6 8 10 12 14 16 18 20 10 Interet Rate GDP Tobin Q Total Loan Aet to Depoit NPL ratio 15 140 35 25 25 90 rule1 rule1 rule1 rule1 rule1 rule1 rule2 rule2 rule2 rule2 rule2 rule2 rule3 rule3 rule3 20 rule3 20 rule3 80 rule3 120 30 10 15 70 15 100 25 10 60 10 80 20 50 40 60 15 5 30 5 5 10 20 40 10 10 20 10 15 10 15 20 0 2 4 6 8 10 12 14 16 18 20 15 0 2 4 6 8 10 12 14 16 18 20 0 rule1: imple Taylor rule - ln i t 2.002 = 1.5 ln in f lation t 2 0 2 4 6 8 10 12 14 16 18 20 0 0 2 4 6 8 10 12 14 16 18 20 20 rule2: flexible Taylor rule augmented with output - ln i t 2.002 = 1.2956 ln in f lation t 2 0.0239 ln outputgapt rule3: Smoothed flexible Taylor rule augmented with output and credit - ln i t +(1 0.9489)(3.6516 ln in f lation t 2 0.2192 ln outputgapt 2.5088 ln loangapt) 2.002 = 0.9489 ln i t 1 2.002 + 0 2 4 6 8 10 12 14 16 18 20 25 0 2 4 6 8 10 12 14 16 18 20 10 5 0 5 0 0 5 0 5 5 0 5 0 5 0 5 0 0 0 29

Table 3: Variable decription Variable Mean Std. Dev. Min Max Obervation lnltoa overall -.4949354.6240002-9.471066.1451363 N = 1322 between.5240806-4.89733 -.0447593 n = 178 within.3167093-7.105189 2.130934 T-bar = 7.42697 lngdp overall 11.50709 4.510904.3810007 14.71222 N = 1322 between 4.39558.732269 14.67755 n = 178 within.2049042 10.69178 12.33534 T-bar = 7.42697 lnatod overall.8268491.8732059.0426906 9.359208 N = 640 between.6971795.0435338 4.192399 n = 89 within.5552704 -.537189 8.547904 T-bar = 7.19101 NPLtoTL overall 324.4615 2637.215 0 47064 N = 1322 between 2065.524 0 24066.78 n = 178 within 1352.177-12274.32 23321.68 T-bar = 7.42697 lnmarket overall 7.973885 1.491046 4.901346 10.62183 N = 1322 between 1.503123 5.289353 10.60567 n = 178 within.2341733 7.153615 8.633486 T-bar = 7.42697 lneonia overall.3904973 1.069185-3.338345 1.719189 N = 1322 between.6745865-2.03785 1.512518 n = 178 within.8311182-2.189725 2.861853 T-bar = 7.42697 lnmrate overall 1.251893.7912259 -.6931472 2.509366 N = 1322 between.4884226 -.1269474 1.994174 n = 178 within.6195975 -.2494564 2.867512 T-bar = 7.42697 30

0.0283 0.0072 0.2471 0.5950 13.6614 0.1319 0.0291 0.6677 0.1802 0.1265 0.0241 0.0484 0.1269 1.0738 5.2752 0.0730 0.0276 0.3119 0.0944 0.0131 0.0037 0.6058 0.0795 2.7878 15.7735 0.0504 0.1823 0.3958 0.8755 0.0714 0.0195 0.0030 0.0077 0.5642 0.0001 1.8391 9.2584 0.0052 0.1669 0.1421 0.0855 0.0358 0.0472 0.0106 0.0017 0.0153 0.1255 34.6961 0.0725 0.0349 0.0571 0.2650 0.2031 0.0363 0.0330 0.0137 0.0018 0.1790 27.4324 0.1576 0.0467 0.0432 0.1496 0.0430 0.0477 0.0080 0.0010 0.0071 0.0583 0.0268 3.8106 0.9784 0.2612 0.0170 0.0768 0.2694 0.0010 0.0435 0.0019 0.0040 0.0336 0.0176 2.5379 0.8534 0.2616 0.0031 0.0293 0.0668 0.0043 0.0305 0.0125 0.0327 0.0811 0.3465 4.5235 0.0585 0.0103 2.0163 0.0198 0.0529 0.0028 0.0093 0.1300 0.0055 5.0707 1.8253 0.0367 0.0542 0.4128 0.1052 0.0492 0.0573 (p 5) lnmrater lnmrater (p 95) lnmrater (p 5) lnmrater lnmrater (p 95) lnmrater (p 5) lnmrater lnmrater (p 95) lnmrater (p 5) lnmrater lnmrater (p 95) lnmrater (p 5) lnmrater lnmrater (p 95) lnmrater (p 5) lnmrater lnmrater (p 95) lnmrater (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate 0.0160 0.0030 0.0295 0.2465 1.2255 14.1421 0.0922 0.0367 0.3516 0.3944 0.1306 0.0029 0.0057 0.0020 0.0256 0.1535 0.3698 6.1304 0.0574 0.0100 0.1071 0.0212 0.1171 0.0061 Table 4: Impule Repone Function of bae model Monetary policy variable: real marginal refinancig rate 1 lag VAR of lngdp lnatod NPLtoTL lnltoa lnmrater lnmarket repone of lngdp to lngdp hock repone of lngdp to lnatod hock repone of lngdp to NPLtoTL hock repone of lngdp to lnltoa hock repone of lngdp to lnmrater hock repone of lngdp to lnmarket hock repone of lnatod to lngdp hock repone of lnatod to lnatod hock repone of lnatod to NPLtoTL hock repone of lnatod to lnltoa hock repone of lnatod to lnmrater hock repone of lnatod to lnmarket hock repone of NPLtoTL to lngdp hock repone of NPLtoTL to lnatod hock repone of NPLtoTL to NPLtoTL hock repone of NPLtoTL to lnltoa hock repone of NPLtoTL to lnmrater hock repone of NPLtoTL to lnmarket hock repone of lnltoa to lngdp hock repone of lnltoa to lnatod hock repone of lnltoa to NPLtoTL hock repone of lnltoa to lnltoa hock repone of lnltoa to lnmrater hock repone of lnltoa to lnmarket hock repone of lnmrater to lngdp hock repone of lnmrater to lnatod hock repone of lnmrater to NPLtoTL hock repone of lnmrater to lnltoa hock repone of lnmrater to lnmrater hock repone of lnmrater to lnmarket hock repone of lnmarket to lngdp hock repone of lnmarket to lnatod hock repone of lnmarket to NPLtoTL hock repone of lnmarket to lnltoa hock repone of lnmarket to lnmrater hock Error are 5% on each ide generated by Monte Carlo with 500 rep repone of lnmarket to lnmarket hock Monetary policy variable: marginal refinancig rate 1 lag VAR of lngdp lnatod NPLtoTL lnltoa lnmrate lnmarket repone of lngdp to lngdp hock repone of lngdp to lnatod hock repone of lngdp to NPLtoTL hock repone of lngdp to lnltoa hock repone of lngdp to lnmrate hock repone of lngdp to lnmarket hock repone of lnatod to lngdp hock repone of lnatod to lnatod hock repone of lnatod to NPLtoTL hock repone of lnatod to lnltoa hock repone of lnatod to lnmrate hock repone of lnatod to lnmarket hock repone of NPLtoTL to lngdp hock repone of NPLtoTL to lnatod hock repone of NPLtoTL to NPLtoTL hock repone of NPLtoTL to lnltoa hock repone of NPLtoTL to lnmrate hock repone of NPLtoTL to lnmarket hock repone of lnltoa to lngdp hock repone of lnltoa to lnatod hock repone of lnltoa to NPLtoTL hock repone of lnltoa to lnltoa hock repone of lnltoa to lnmrate hock repone of lnltoa to lnmarket hock repone of lnmrate to lngdp hock repone of lnmrate to lnatod hock repone of lnmrate to NPLtoTL hock repone of lnmrate to lnltoa hock repone of lnmrate to lnmrate hock repone of lnmrate to lnmarket hock repone of lnmarket to lngdp hock repone of lnmarket to lnatod hock repone of lnmarket to NPLtoTL hock repone of lnmarket to lnltoa hock repone of lnmarket to lnmrate hock Error are 5% on each ide generated by Monte Carlo with 500 rep repone of lnmarket to lnmarket hock 31

0.0276 0.2323 0.6332 15.1666 0.1371 0.0296 0.7491 0.2406 0.1234 0.0286 0.0422 0.1309 5.9395 4.6234 0.0910 0.0440 0.7074 0.1205 0.0143 0.0039 0.6058 0.0915 2.9845 19.0866 0.0636 0.1836 0.3897 0.8362 0.0758 0.0201 0.0067 0.0021 0.5028 0.0264 1.1610 8.4305 0.0218 0.1437 0.1919 0.1640 0.0394 0.0104 (p 5) lnatod_ce lnatod_ce _CE (p 5) lnatod_ce lnatod_ce _CE (p 5) lnatod_ce lnatod_ce _CE (p 5) lnatod_ce lnatod_ce _CE (p 5) lnatod_ce lnatod_ce _CE (p 5) lnatod_ce lnatod_ce _CE 0.0104 0.0019 0.0207 0.1388 36.3113 0.0358 0.0554 0.3089 0.2430 0.0320 0.0350 0.0056 0.0061 0.0136 0.0958 42.0241 0.0090 0.0749 0.1463 0.1593 0.0266 0.0384 (p 5) NPLtoTL_CE NPLtoTL_CE _CE (p 5) NPLtoTL_CE NPLtoTL_CE _CE (p 5) NPLtoTL_CE NPLtoTL_CE _CE (p 5) NPLtoTL_CE NPLtoTL_CE _CE (p 5) NPLtoTL_CE NPLtoTL_CE _CE (p 5) NPLtoTL_CE NPLtoTL_CE _CE 0.0008 0.0063 0.0589 0.0254 4.3149 0.9037 0.2624 0.0172 0.0676 0.2698 0.0010 0.0389 0.0022 0.0048 0.0249 0.0179 1.4536 2.0828 0.2418 0.0062 0.1636 0.1326 0.0086 0.0292 (p 5) lnltoa_ce lnltoa_ce _CE (p 5) lnltoa_ce lnltoa_ce _CE (p 5) lnltoa_ce lnltoa_ce _CE (p 5) lnltoa_ce lnltoa_ce _CE (p 5) lnltoa_ce lnltoa_ce _CE (p 5) lnltoa_ce lnltoa_ce _CE 0.0145 0.0216 0.0922 0.3802 6.0289 0.0703 0.0132 2.1248 0.0625 0.0118 0.0281 0.0513 1.5831 1.6069 0.0375 0.0141 1.9802 0.0046 0.0490 (p 5) real_lneonia real_lneonia (p 95) real_lneonia (p 5) real_lneonia real_lneonia (p 95) real_lneonia (p 5) real_lneonia real_lneonia (p 95) real_lneonia (p 5) real_lneonia real_lneonia (p 95) real_lneonia (p 5) real_lneonia real_lneonia (p 95) real_lneonia (p 5) real_lneonia real_lneonia (p 95) real_lneonia (p 5) lnmrater lnmrater (p 95) lnmrater (p 5) lnmrater lnmrater (p 95) lnmrater (p 5) lnmrater lnmrater (p 95) lnmrater (p 5) lnmrater lnmrater (p 95) lnmrater (p 5) lnmrater lnmrater (p 95) lnmrater (p 5) lnmrater lnmrater (p 95) lnmrater 0.0167 0.0024 0.0256 0.2458 1.3028 16.3159 0.1011 0.0401 0.4409 0.4243 0.1301 0.0004 0.0142 0.0007 0.0272 0.1219 6.0430 2.8195 0.0729 0.0285 0.3655 0.0453 0.1311 Table 5: Impule Repone Function: robutne check Different monetary policy variable: overnight interbank rate 1 lag VAR of lngdp lnatod NPLtoTL lnltoa real_lneonia lnmarket repone of lngdp to lngdp hock repone of lngdp to lnatod hock repone of lngdp to NPLtoTL hock repone of lngdp to lnltoa hock repone of lngdp to real_lneonia hock repone of lngdp to lnmarket hock repone of lnatod to lngdp hock repone of lnatod to lnatod hock repone of lnatod to NPLtoTL hock repone of lnatod to lnltoa hock repone of lnatod to real_lneonia hock repone of lnatod to lnmarket hock repone of NPLtoTL to lngdp hock repone of NPLtoTL to lnatod hock repone of NPLtoTL to NPLtoTL hock repone of NPLtoTL to lnltoa hock repone of NPLtoTL to real_lneonia hock repone of NPLtoTL to lnmarket hock repone of lnltoa to lngdp hock repone of lnltoa to lnatod hock repone of lnltoa to NPLtoTL hock repone of lnltoa to lnltoa hock repone of lnltoa to real_lneonia hock repone of lnltoa to lnmarket hock repone of real_lneonia to lngdp hock repone of real_lneonia to lnatod hock repone of real_lneonia to NPLtoTL hock repone of real_lneonia to lnltoa hockrepone of real_lneonia to real_lneonia hock repone of real_lneonia to lnmarket hock repone of lnmarket to lngdp hock repone of lnmarket to lnatod hock repone of lnmarket to NPLtoTL hock repone of lnmarket to lnltoa hock repone of lnmarket to real_lneonia hockrepone of lnmarket to lnmarket hock Error are 5% on each ide generated by Monte Carlo with 500 rep Model with country-year FE 1 lag VAR of lngdp lnatod_ce NPLtoTL_CE lnltoa_ce lnmrater lnmarket repone of lngdp to lngdp hock repone of lngdp to lnatod_ce hock repone of lngdp to NPLtoTL_CE hock repone of lngdp to lnltoa_ce hock repone of lngdp to lnmrater hock repone of lngdp to lnmarket hock repone of lnatod_ce to lngdp hock repone of lnatod_ce to lnatod_ce hock repone of lnatod_ce to NPLtoTL_CE hock repone of lnatod_ce to lnltoa_ce hock repone of lnatod_ce to lnmrater hock repone of lnatod_ce to lnmarket hock repone of NPLtoTL_CE to lngdp hock repone of NPLtoTL_CE to lnatod_ce hock repone of NPLtoTL_CE to NPLtoTL_CE hock repone of NPLtoTL_CE to lnltoa_ce hockrepone of NPLtoTL_CE to lnmrater hockrepone of NPLtoTL_CE to lnmarket hock repone of lnltoa_ce to lngdp hock repone of lnltoa_ce to lnatod_ce hock repone of lnltoa_ce to NPLtoTL_CE hock repone of lnltoa_ce to lnltoa_ce hock repone of lnltoa_ce to lnmrater hock repone of lnltoa_ce to lnmarket hock repone of lnmrater to lngdp hock repone of lnmrater to lnatod_ce hock repone of lnmrater to NPLtoTL_CE hock repone of lnmrater to lnltoa_ce hock repone of lnmrater to lnmrater hock repone of lnmrater to lnmarket hock repone of lnmarket to lngdp hock repone of lnmarket to lnatod_ce hock repone of lnmarket to NPLtoTL_CE hock repone of lnmarket to lnltoa_ce hock repone of lnmarket to lnmrater hock Error are 5% on each ide generated by Monte Carlo with 500 rep repone of lnmarket to lnmarket hock 32

0.2121 0.1180 2.0683 22.8851 5.3628 0.8505 0.4904 0.3141 2.0752 0.4838 0.1264 0.0092 0.0007 0.0331 0.1141 4.1810 4.3925 0.0840 0.0413 0.2262 0.0929 0.0446 0.0179 0.0021 0.0701 0.7484 0.0189 2.5337 7.2095 0.1530 0.2598 0.7227 0.0163 0.0374 0.1492 0.0089 0.0091 0.6572 0.2095 26.3891 28.4201 0.4919 0.6005 0.9087 0.7830 0.0985 0.1607 0.0201 0.0386 0.4160 0.1513 10.6818 2.2104 0.0927 0.1671 0.3934 0.1655 0.0348 0.0843 0.0032 0.0035 0.1695 0.1264 28.0815 1.2417 0.1304 0.1893 0.2179 0.2342 0.0482 0.0406 0.0051 0.0243 0.1962 0.0329 0.6906 2.3863 0.2427 0.0619 0.2076 0.0444 0.0053 0.0470 0.0060 0.0052 0.1921 0.2998 16.7089 19.3195 0.3846 0.2782 0.4005 0.4478 0.0661 0.0864 0.1676 1.5957 4.6530 18.2507 0.4242 0.6468 1.6383 0.0748 0.0470 0.3775 0.0078 0.0057 0.2935 0.2295 21.6481 19.5595 0.2634 0.3051 0.4903 0.4088 0.0561 0.1079 (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate (p 5) lnmrate lnmrate (p 95) lnmrate 0.0683 0.0026 0.5962 6.8131 1.8627 0.2464 0.1511 0.0084 0.6903 0.1488 0.0067 0.0054 0.0037 0.1523 0.2390 9.3574 13.9842 0.2536 0.2801 0.5739 0.3515 0.1025 0.0582 Table 6: Impule Repone Function pre and pot BaelII (2006) Subample: BaelI regime - pre 2006 1 lag VAR of lngdp lnatod NPLtoTL lnltoa lnmrate lnmarket Sample : if year<=2006 repone of lngdp to lngdp hock repone of lngdp to lnatod hock repone of lngdp to NPLtoTL hock repone of lngdp to lnltoa hock repone of lngdp to lnmrate hock repone of lngdp to lnmarket hock repone of lnatod to lngdp hock repone of lnatod to lnatod hock repone of lnatod to NPLtoTL hock repone of lnatod to lnltoa hock repone of lnatod to lnmrate hock repone of lnatod to lnmarket hock repone of NPLtoTL to lngdp hock repone of NPLtoTL to lnatod hock repone of NPLtoTL to NPLtoTL hock repone of NPLtoTL to lnltoa hock repone of NPLtoTL to lnmrate hock repone of NPLtoTL to lnmarket hock repone of lnltoa to lngdp hock repone of lnltoa to lnatod hock repone of lnltoa to NPLtoTL hock repone of lnltoa to lnltoa hock repone of lnltoa to lnmrate hock repone of lnltoa to lnmarket hock repone of lnmrate to lngdp hock repone of lnmrate to lnatod hock repone of lnmrate to NPLtoTL hock repone of lnmrate to lnltoa hock repone of lnmrate to lnmrate hock repone of lnmrate to lnmarket hock repone of lnmarket to lngdp hock repone of lnmarket to lnatod hock repone of lnmarket to NPLtoTL hock repone of lnmarket to lnltoa hock repone of lnmarket to lnmrate hock Error are 5% on each ide generated by Monte Carlo with 500 rep repone of lnmarket to lnmarket hock Subample: BaelII regime - pot 2006 1 lag VAR of lngdp lnatod NPLtoTL lnltoa lnmrate lnmarket Sample : if year>2006 repone of lngdp to lngdp hock repone of lngdp to lnatod hock repone of lngdp to NPLtoTL hock repone of lngdp to lnltoa hock repone of lngdp to lnmrate hock repone of lngdp to lnmarket hock repone of lnatod to lngdp hock repone of lnatod to lnatod hock repone of lnatod to NPLtoTL hock repone of lnatod to lnltoa hock repone of lnatod to lnmrate hock repone of lnatod to lnmarket hock repone of NPLtoTL to lngdp hock repone of NPLtoTL to lnatod hock repone of NPLtoTL to NPLtoTL hock repone of NPLtoTL to lnltoa hock repone of NPLtoTL to lnmrate hock repone of NPLtoTL to lnmarket hock repone of lnltoa to lngdp hock repone of lnltoa to lnatod hock repone of lnltoa to NPLtoTL hock repone of lnltoa to lnltoa hock repone of lnltoa to lnmrate hock repone of lnltoa to lnmarket hock repone of lnmrate to lngdp hock repone of lnmrate to lnatod hock repone of lnmrate to NPLtoTL hock repone of lnmrate to lnltoa hock repone of lnmrate to lnmrate hock repone of lnmrate to lnmarket hock repone of lnmarket to lngdp hock repone of lnmarket to lnatod hock repone of lnmarket to NPLtoTL hock repone of lnmarket to lnltoa hock repone of lnmarket to lnmrate hock Error are 5% on each ide generated by Monte Carlo with 500 rep repone of lnmarket to lnmarket hock 33

Table 7: Effect of change in bank uperviion regime Reult of the Etimation by ytem GMM dep.var lnatod NPLtoTL lnltoa L.lnGDP -2.9264958-33.115574 1.2219396 (-.99598553) (-1.8648106) (1.1519832) L.lnAtoD.26821771-2.4017334.01744338 (2.1822716) (-.62690392) (.4195989) L.NPLtoTL -.00346042.82807571.00028113 (-1.7965061) (4.6370976) (.35646872) L.lnLtoA -.01555742-1.2193181.55701293 (-.28973949) (-.5164283) (4.7332525) L.lnMRATE.28263386-20.636701.05456813 (1.1924307) (-2.572168) (.39368463) L.baelII*lnMRATE -.07494514 23.833056 -.0971332 (-.34439052) (2.6372561) (-.79703425) L.lnMARKET -.77119116-14.01053.19905925 (-1.9262001) (-1.1351612) (1.0037422) t tatitic in parenthei 1. BaelII i a dummy variable with value 1 after 2006, year of introduction of BaelII regulatory act with more tight capital requirement. 2.The VAR model etimated alo the equation for the other variable, here not reported. 34

0.0234 repone of lngdp to lngdp hock 0.0090 0.2130 12.9902 repone of lnatod to lngdp hock 0.6296 repone of NPLtoTL to lngdp hock 0.1216 0.0296 repone of lnltoa to lngdp hock 0.5017 0.5447 0.1123 repone of rule1 to lngdp hock repone of lnmarket to lngdp hock 0.0250 repone of lngdp to lngdp hock 0.0030 0.2084 repone of lnatod to lngdp hock 0.7965 13.4159 repone of NPLtoTL to lngdp hock 0.1176 0.0280 repone of lnltoa to lngdp hock 0.4184 0.4834 0.1180 repone of rule2 to lngdp hock repone of lnmarket to lngdp hock 0.0248 repone of lngdp to lngdp hock 0.1087 0.1853 repone of lnatod to lngdp hock 2.9286 5.3677 repone of NPLtoTL to lngdp hock 0.0854 0.0558 repone of lnltoa to lngdp hock 0.2735 0.0167 0.1091 0.0047 repone of rule3 to lngdp hock repone of lnmarket to lngdp hock 0.0113 0.0057 repone of lngdp to lnatod hock 0.6201 0.0973 16.3899 repone of lnatod to lnatod hock 3.1049 repone of NPLtoTL to lnatod hock 0.0401 0.1862 repone of lnltoa to lnatod hock 1.1999 0.3165 0.0633 0.0253 repone of rule1 to lnatod hock repone of lnmarket to lnatod hock 0.0096 0.0055 repone of lngdp to lnatod hock 0.6214 0.0871 repone of lnatod to lnatod hock 3.0397 17.5400 repone of NPLtoTL to lnatod hock 0.0365 0.1857 repone of lnltoa to lnatod hock 1.0954 0.2541 0.0551 0.0243 repone of rule2 to lnatod hock repone of lnmarket to lnatod hock 0.0075 0.0120 repone of lngdp to lnatod hock 0.5354 0.0653 repone of lnatod to lnatod hock 4.3457 17.6782 repone of NPLtoTL to lnatod hock 0.0143 0.1980 repone of lnltoa to lnatod hock 0.2300 0.1421 0.0589 0.0719 repone of rule3 to lnatod hock repone of lnmarket to lnatod hock 0.0103 0.0014 repone of lngdp to NPLtoTL hock 0.0213 0.1544 repone of lnatod to NPLtoTL hock 37.0980 0.0099 repone of NPLtoTL to NPLtoTL hock 0.0442 0.0528 repone of lnltoa to NPLtoTL hock 0.3461 0.2830 repone of rule1 to NPLtoTL hock 0.0322 0.0414 repone of lnmarket to NPLtoTL hock 0.0103 0.0019 repone of lngdp to NPLtoTL hock 0.0306 0.1322 repone of lnatod to NPLtoTL hock 36.6693 repone of NPLtoTL to NPLtoTL hock 0.0390 0.0677 repone of lnltoa to NPLtoTL hock 0.2830 0.2335 repone of rule2 to NPLtoTL hock 0.0314 0.0398 repone of lnmarket to NPLtoTL hock 0.0063 0.0108 repone of lngdp to NPLtoTL hock 0.0695 0.0601 repone of lnatod to NPLtoTL hock 40.9903 0.2256 repone of NPLtoTL to NPLtoTL hock 0.0654 0.0838 repone of lnltoa to NPLtoTL hock 0.1518 0.0366 repone of rule3 to NPLtoTL hock 0.0099 0.0632 repone of lnmarket to NPLtoTL hock 0.0010 0.0065 repone of lngdp to lnltoa hock 0.0562 0.0259 repone of lnatod to lnltoa hock 3.5320 1.0136 repone of NPLtoTL to lnltoa hock 0.2606 0.0125 0.1405 repone of lnltoa to lnltoa hock 0.3450 0.0012 0.0436 repone of rule1 to lnltoa hock repone of lnmarket to lnltoa hock 0.0010 0.0058 repone of lngdp to lnltoa hock 0.0497 0.0276 repone of lnatod to lnltoa hock 4.0739 1.0629 repone of NPLtoTL to lnltoa hock 0.2607 0.0127 repone of lnltoa to lnltoa hock 0.3112 0.1282 0.0013 0.0413 repone of rule2 to lnltoa hock repone of lnmarket to lnltoa hock 0.0052 0.0100 repone of lngdp to lnltoa hock 0.0584 0.0468 repone of lnatod to lnltoa hock 6.3950 2.8579 repone of NPLtoTL to lnltoa hock 0.2491 0.0036 repone of lnltoa to lnltoa hock 0.0771 0.0905 0.0111 0.0620 repone of rule3 to lnltoa hock repone of lnmarket to lnltoa hock 0.0155 0.1095 0.0096 7.2960 0.7475 (p 5) rule1 rule1 (p 95) rule1 repone of lngdp to rule1 hock (p 5) rule1 rule1 (p 95) rule1 repone of lnatod to rule1 hock (p 5) rule1 rule1 (p 95) rule1 repone of NPLtoTL to rule1 hock (p 5) rule1 rule1 (p 95) rule1 0.0187 0.0746 2.9898 0.1981 0.0748 repone of lnltoa to rule1 hock (p 5) rule1 rule1 (p 95) rule1 repone of rule1 to rule1 hock (p 5) rule1 rule1 (p 95) rule1 repone of lnmarket to rule1 hock 0.0154 0.1052 0.0099 8.0443 0.6547 (p 5) rule2 rule2 (p 95) rule2 repone of lngdp to rule2 hock (p 5) rule2 rule2 (p 95) rule2 repone of lnatod to rule2 hock (p 5) rule2 rule2 (p 95) rule2 repone of NPLtoTL to rule2 hock (p 5) rule2 rule2 (p 95) rule2 0.0188 0.0725 2.5774 0.1872 0.0743 repone of lnltoa to rule2 hock (p 5) rule2 rule2 (p 95) rule2 repone of rule2 to rule2 hock (p 5) rule2 rule2 (p 95) rule2 repone of lnmarket to rule2 hock 0.0027 0.0274 0.1834 0.0496 20.8930 1.9214 (p 5) rule3 rule3 (p 95) rule3 repone of lngdp to rule3 hock (p 5) rule3 rule3 (p 95) rule3 repone of lnatod to rule3 hock (p 5) rule3 rule3 (p 95) rule3 repone of NPLtoTL to rule3 hock (p 5) rule3 rule3 (p 95) rule3 0.1257 0.1229 0.6510 0.1721 0.0012 0.1217 repone of lnltoa to rule3 hock (p 5) rule3 rule3 (p 95) rule3 repone of rule3 to rule3 hock (p 5) rule3 rule3 (p 95) rule3 repone of lnmarket to rule3 hock 0.0126 0.0031 repone of lngdp to lnmarket hock 0.0415 0.2620 repone of lnatod to lnmarket hock 1.0406 14.8126 repone of NPLtoTL to lnmarket hock 0.1027 0.0365 repone of lnltoa to lnmarket hock 0.5055 0.3017 repone of rule1 to lnmarket hock 0.1322 0.0026 repone of lnmarket to lnmarket hock 0.0144 0.0035 repone of lngdp to lnmarket hock 0.0318 0.2408 repone of lnatod to lnmarket hock 1.4081 15.9425 repone of NPLtoTL to lnmarket hock 0.0968 0.0376 repone of lnltoa to lnmarket hock 0.4703 0.2560 repone of rule2 to lnmarket hock 0.1333 0.0023 repone of lnmarket to lnmarket hock 0.0052 0.0138 repone of lngdp to lnmarket hock 0.0528 0.2764 repone of lnatod to lnmarket hock 2.8911 11.8216 repone of NPLtoTL to lnmarket hock 0.1193 0.0441 repone of lnltoa to lnmarket hock 0.1065 0.1861 repone of rule3 to lnmarket hock 0.1390 0.0323 repone of lnmarket to lnmarket hock 0.0275 repone of lngdp to lngdp hock 0.0033 0.2288 13.4227 repone of lnatod to lngdp hock 0.4115 repone of NPLtoTL to lngdp hock 0.1367 0.0249 repone of lnltoa to lngdp hock 0.8202 0.3236 0.1332 repone of diff1 to lngdp hock repone of lnmarket to lngdp hock 0.0277 repone of lngdp to lngdp hock 0.0070 0.2446 14.6855 repone of lnatod to lngdp hock 0.5769 repone of NPLtoTL to lngdp hock 0.1282 0.0179 repone of lnltoa to lngdp hock 0.7378 0.2683 0.1159 repone of diff2 to lngdp hock repone of lnmarket to lngdp hock 0.0420 repone of lngdp to lngdp hock 0.0410 0.1934 repone of lnatod to lngdp hock 1.6632 16.2062 repone of NPLtoTL to lngdp hock 0.1737 0.1372 repone of lnltoa to lngdp hock 0.4432 0.0134 0.1747 repone of diff3 to lngdp hock repone of lnmarket to lngdp hock 0.0118 0.0046 repone of lngdp to lnatod hock 0.6127 0.0841 17.6884 repone of lnatod to lnatod hock 2.9657 repone of NPLtoTL to lnatod hock 0.0515 0.1825 repone of lnltoa to lnatod hock 0.5239 1.3504 0.0682 0.0197 repone of diff1 to lnatod hock repone of lnmarket to lnatod hock 0.0126 0.0044 repone of lngdp to lnatod hock 0.6094 0.0843 17.5186 repone of lnatod to lnatod hock 3.1910 repone of NPLtoTL to lnatod hock 0.0539 0.1813 repone of lnltoa to lnatod hock 0.4732 1.1389 0.0621 0.0192 repone of diff2 to lnatod hock repone of lnmarket to lnatod hock 0.0271 0.0073 repone of lngdp to lnatod hock 0.5446 0.1068 repone of lnatod to lnatod hock 5.9104 19.2820 repone of NPLtoTL to lnatod hock 0.0571 0.1967 repone of lnltoa to lnatod hock 0.3209 0.2147 0.1246 0.0402 repone of diff3 to lnatod hock repone of lnmarket to lnatod hock 0.0106 0.0023 repone of lngdp to NPLtoTL hock 0.0190 0.1368 repone of lnatod to NPLtoTL hock 34.9562 repone of NPLtoTL to NPLtoTL hock 0.0386 0.0524 repone of lnltoa to NPLtoTL hock 0.3828 0.3067 repone of diff1 to NPLtoTL hock 0.0346 0.0358 repone of lnmarket to NPLtoTL hock 0.0107 0.0013 repone of lngdp to NPLtoTL hock 0.0151 0.1412 repone of lnatod to NPLtoTL hock 33.8992 repone of NPLtoTL to NPLtoTL hock 0.0396 0.0440 repone of lnltoa to NPLtoTL hock 0.3468 0.2430 repone of diff2 to NPLtoTL hock 0.0328 0.0323 repone of lnmarket to NPLtoTL hock 0.0069 0.0065 repone of lngdp to NPLtoTL hock 0.0292 0.1016 repone of lnatod to NPLtoTL hock 28.8160 repone of NPLtoTL to NPLtoTL hock 0.0458 0.0730 repone of lnltoa to NPLtoTL hock 0.0867 0.2022 repone of diff3 to NPLtoTL hock 0.0222 0.0343 repone of lnmarket to NPLtoTL hock 0.0010 0.0071 repone of lngdp to lnltoa hock 0.0566 0.0293 repone of lnatod to lnltoa hock 3.9847 1.0254 repone of NPLtoTL to lnltoa hock 0.2617 0.0160 0.3755 repone of lnltoa to lnltoa hock 0.1275 0.0018 0.0370 repone of diff1 to lnltoa hock repone of lnmarket to lnltoa hock 0.0011 0.0071 repone of lngdp to lnltoa hock 0.0549 0.0312 repone of lnatod to lnltoa hock 3.6447 1.1115 repone of NPLtoTL to lnltoa hock 0.2610 0.0156 0.3150 repone of lnltoa to lnltoa hock 0.1236 0.0018 0.0374 repone of diff2 to lnltoa hock repone of lnmarket to lnltoa hock 0.0034 0.0201 repone of lngdp to lnltoa hock 0.0952 0.0561 repone of lnatod to lnltoa hock 8.5057 2.4664 repone of NPLtoTL to lnltoa hock 0.2486 0.0190 repone of lnltoa to lnltoa hock 0.0650 0.2427 0.0070 0.1037 repone of diff3 to lnltoa hock repone of lnmarket to lnltoa hock 0.0132 0.0197 0.0948 0.4353 5.5248 (p 5) diff1 diff1 (p 95) diff1 repone of lngdp to diff1 hock (p 5) diff1 diff1 (p 95) diff1 repone of lnatod to diff1 hock (p 5) diff1 diff1 (p 95) diff1 repone of NPLtoTL to diff1 hock (p 5) diff1 diff1 (p 95) diff1 0.0655 0.0113 2.9889 0.1387 0.0617 repone of lnltoa to diff1 hock (p 5) diff1 diff1 (p 95) diff1 repone of diff1 to diff1 hock (p 5) diff1 diff1 (p 95) diff1 repone of lnmarket to diff1 hock 0.0134 0.0232 0.0900 0.3573 5.7876 (p 5) diff2 diff2 (p 95) diff2 repone of lngdp to diff2 hock (p 5) diff2 diff2 (p 95) diff2 repone of lnatod to diff2 hock (p 5) diff2 diff2 (p 95) diff2 repone of NPLtoTL to diff2 hock (p 5) diff2 diff2 (p 95) diff2 0.0626 0.0084 2.5866 0.0823 0.0570 repone of lnltoa to diff2 hock (p 5) diff2 diff2 (p 95) diff2 repone of diff2 to diff2 hock (p 5) diff2 diff2 (p 95) diff2 repone of lnmarket to diff2 hock 0.0198 0.0823 0.1048 0.4787 8.2495 (p 5) diff3 diff3 (p 95) diff3 repone of lngdp to diff3 hock (p 5) diff3 diff3 (p 95) diff3 repone of lnatod to diff3 hock (p 5) diff3 diff3 (p 95) diff3 repone of NPLtoTL to diff3 hock (p 5) diff3 diff3 (p 95) diff3 0.0898 0.0556 0.8108 0.0043 0.0846 repone of lnltoa to diff3 hock (p 5) diff3 diff3 (p 95) diff3 repone of diff3 to diff3 hock (p 5) diff3 diff3 (p 95) diff3 repone of lnmarket to diff3 hock 0.0152 0.0029 repone of lngdp to lnmarket hock 0.0300 0.2608 repone of lnatod to lnmarket hock 0.9136 14.5707 repone of NPLtoTL to lnmarket hock 0.0966 0.0338 repone of lnltoa to lnmarket hock 0.4307 0.5424 repone of diff1 to lnmarket hock 0.1308 0.0030 repone of lnmarket to lnmarket hock 0.0152 0.0031 repone of lngdp to lnmarket hock 0.0284 0.2497 repone of lnatod to lnmarket hock 1.1962 14.9240 repone of NPLtoTL to lnmarket hock 0.0981 0.0314 repone of lnltoa to lnmarket hock 0.3837 0.4996 repone of diff2 to lnmarket hock 0.1306 0.0023 repone of lnmarket to lnmarket hock 0.0236 0.0089 repone of lngdp to lnmarket hock 0.0732 0.3199 repone of lnatod to lnmarket hock 3.9519 17.0574 repone of NPLtoTL to lnmarket hock 0.1393 0.1018 repone of lnltoa to lnmarket hock 0.4499 0.0990 repone of diff3 to lnmarket hock 0.1238 0.0206 repone of lnmarket to lnmarket hock Table 8: Impule Repone Function of bae model Taylor rule 1 lag VAR of lngdp lnatod NPLtoTL lnltoa rule1 lnmarket 1 lag VAR of lngdp lnatod NPLtoTL lnltoa diff1 lnmarket Error are 5% on each ide generated by Monte Carlo with 500 rep Error are 5% on each ide generated by Monte Carlo with 500 rep 1 lag VAR of lngdp lnatod NPLtoTL lnltoa rule2 lnmarket Flexible Taylor rule with inflation and output-gap 1 lag VAR of lngdp lnatod NPLtoTL lnltoa diff2 lnmarket Error are 5% on each ide generated by Monte Carlo with 500 rep Error are 5% on each ide generated by Monte Carlo with 500 rep Flexible Taylor rule with interet moothing, output-gap, and total credit. 1 lag VAR of lngdp lnatod NPLtoTL lnltoa rule3 lnmarket 1 lag VAR of lngdp lnatod NPLtoTL lnltoa diff3 lnmarket Error are 5% on each ide generated by Monte Carlo with 500 rep Error are 5% on each ide generated by Monte Carlo with 500 rep 35

Table 9: FEVD - Taylor rule gap Main variable Explanatory variable Short Period: 2 year lngdp lnatod NPLtoTL lnltoa rule1 diff1 lnmarket lngdp.82337936.01397545.00436201.00193397.14699912.00368804.00566205 lnatod.01765963.94058433.01000554.00023047.00015296.00865889.02270818 NPLtoTL.00720607.01881193.96586244.0000838.00042106.00000000.00761402 lnltoa.00970677.22694878.00689998.7473811.00327146.00040337.00538855 rule1.01883386.04294761.00065444.00297678.93266242.00021353.00171135 diff1.01665646.05278616.00044083.00302022.9011924.02226493.00363901 lnmarket.25602856.03360954.00121722.02539057.0386464.0758069.56930081 Medium period: 6 year lngdp lnatod NPLtoTL lnltoa rule1 diff1 lnmarket lngdp.69828124.00985304.04890384.00413773.20704081.02582314.0059602 lnatod.03239471.8363428.05009213.00116608.00617436.04316652.03066341 NPLtoTL.01912843.0338818.92191833.00092063.00427106.00232271.01755704 lnltoa.0437882.21046567.00862486.69683578.01853724.0100404.01170786 rule1.0615591.04166172.01155245.00295313.87574229.00405881.00247249 diff1.03222254.05383913.00670592.00295924.87562025.02501062.00364229 lnmarket.26279593.04648841.02099021.0290228.05830463.14317577.43922225 Long Period: 10 year lngdp lnatod NPLtoTL lnltoa rule1 diff1 lnmarket lngdp.64273589.01354028.09917194.00414074.22027483.01511581.0050205 lnatod.03518645.81104474.07080057.00115559.01326771.03882909.02971585 NPLtoTL.02722577.03429169.90618139.00132185.00866425.00301454.01930052 lnltoa.05756456.20458608.0168486.6710842.02944928.00892358.0115437 rule1.0368983.05364512.01302351.00294118.88276616.00713569.00359005 diff1.07257697.04147294.02418935.00296531.83764221.01874092.00241229 lnmarket.26587409.04719323.0452132.02740882.08086269.12466121.40878676 * Variance-decompoition: percent of variation in the row variable explained by column variable ** rule1= Taylor rule. diff1=taylor gap, difference between nominal interet rate and Taylor rule. 36

Table 10: FEVD - Taylor rule gap Main variable Explanatory variable Short period: 2 year lngdp lnatod NPLtoTL lnltoa rule2 diff2 lnmarket lngdp.82319352.01398431.00435695.00193381.1508168.00005598.00565863 lnatod.01765722.94057977.01000244.00023088.0006389.00818777.02270301 NPLtoTL.007208.0188107.96586201.00008362.00041111.0000131.00761147 lnltoa.00971379.22696856.00689959.74735456.00353198.00014532.00538621 rule2.01662417.05270974.00044228.00301529.9190139.00456438.00363024 diff2.02124523.04128428.00069188.00295615.90901516.02335039.0014569 lnmarket.25587232.03348585.00120768.02538024.03971811.07522502.56911079 Medium period: 6 year lngdp lnatod NPLtoTL lnltoa rule2 diff2 lnmarket lngdp.69801114.00985323.0488684.00414009.22228932.01088358.00595424 lnatod.03239399.83633415.050088.0011677.0111178.0382474.03065096 NPLtoTL.0191245.03388388.9218983.00092081.00507733.00154672.01754846 lnltoa.04379423.21047146.00862344.69678953.02202393.00659503.01170237 rule2.03223293.05377005.00671652.00295436.89396224.00672991.00363398 diff2.06877378.03969277.01231854.00297269.85059322.02324947.00239955 lnmarket.26266809.04641563.02096647.02902266.07436543.12755783.43900389 Long Period: 10 year lngdp lnatod NPLtoTL lnltoa rule2 diff2 lnmarket lngdp.64251704.01353658.09912105.00414335.22055501.01511141.00501556 lnatod.03518834.8110356.07080085.00115719.0132761.03883824.02970368 NPLtoTL.02722315.03429319.90615229.00132221.00870225.00301626.01929065 lnltoa.05757402.20458957.01684718.67103433.02949063.00892606.01153821 rule2.03691993.0535763.0130454.00293641.88280467.00713554.00358176 diff2.08072434.03954396.02597968.00299197.82468656.02373444.00233905 lnmarket.26576826.04712354.04519119.02740875.08133348.12459494.40857982 * Variance-decompoition: percent of variation in the row variable explained by column variable ** rule2= flexible Taylor rule with different weight for inflation and output-gap. Taylor gap, difference between nominal interet rate and flexible Taylor rule. 37

Table 11: FEVD - Taylor rule gap Main variable Explanatory variable Short period: 2 year lngdp lnatod NPLtoTL lnltoa rule3 diff3 lnmarket lngdp.89708694.00067957.00009106.00192627.0744516.00194343.02382114 lnatod.02051063.94158609.00501573.00059019.00074613.00129296.03025826 NPLtoTL.00649281.01656703.96719489.00012828.00041889.00024939.00894871 lnltoa.00825798.29985481.00176283.67848326.00058112.00052951.01053048 rule3.13953279.06118966.00207337.00490691.55213598.23874734.00141395 diff3.05971519.03665782.00018589.00061696.48677896.40279815.01324703 lnmarket.40087955.02356005.000645.02610931.05647614.11220819.38012176 Medium period: 6 year lngdp lnatod NPLtoTL lnltoa rule3 diff3 lnmarket lngdp.69235294.02922272.02109655.00228646.11634459.09591812.04277863 lnatod.04560873.8871984.02023608.0005822.00112556.01025812.03499091 NPLtoTL.01248861.02178897.94440646.00015644.00506883.0056892.01040149 lnltoa.03773076.2914979.00454281.64519851.00312233.00406765.01384005 rule3.16427278.06619873.00867243.01777265.46497586.23626501.04184253 diff3.04921734.06810325.01453398.00633152.40926057.43704843.01550491 lnmarket.37326663.04088047.00930692.02817563.06357486.1794025.30539297 Long Period: 10 year lngdp lnatod NPLtoTL lnltoa rule3 diff3 lnmarket lngdp.67068086.03584195.04065772.00279187.11078876.09701535.04222349 lnatod.04842063.8777895.0258163.00058058.00138567.01114114.03486617 NPLtoTL.01578845.02137661.93920618.00023663.00560356.00747124.01031732 lnltoa.04711537.28975084.00981366.62870262.00394129.00719428.01348194 rule3.1673396.06663077.00999484.01768804.45998336.23678765.04157574 diff3.04959974.068021.01646544.00633024.40781586.43617903.0155887 lnmarket.37407124.04145122.01502723.0277214.06343518.17746219.30083155 * variance-decompoition: percent of variation in the row variable explained by column variable ** rule3= flexible Taylor rule with interet moothing, output gap and total credit. diff3= Taylor gap, difference between nominal interet rate and flexible Taylor rule. 38

0.0354 0.2347 repone of lngdp to lngdp hock 0.0066 12.8713 repone of lnatod to lngdp hock 0.9838 repone of NPLtoTL to lngdp hock 0.1373 0.0282 0.4572 0.0019 repone of lnltoa to lngdp hock repone of rule1_eu to lngdp hock 0.1548 repone of lnmarket to lngdp hock 0.0316 repone of lngdp to lngdp hock 0.0012 0.2469 repone of lnatod to lngdp hock 1.0233 13.7299 repone of NPLtoTL to lngdp hock 0.1389 0.0442 0.3874 repone of lnltoa to lngdp hock repone of rule2_gen to lngdp hock 0.1442 repone of lnmarket to lngdp hock 0.0277 repone of lngdp to lngdp hock 0.1582 0.3083 repone of lnatod to lngdp hock 2.5939 10.2386 repone of NPLtoTL to lngdp hock 0.1339 0.0834 0.2497 repone of lnltoa to lngdp hock repone of rule3_gen to lngdp hock 0.1486 0.0345 repone of lnmarket to lngdp hock 0.0098 0.0060 0.0867 repone of lngdp to lnatod hock 0.6285 12.4215 repone of lnatod to lnatod hock 3.1539 repone of NPLtoTL to lnatod hock 0.0252 0.1815 0.5613 0.2396 repone of lnltoa to lnatod hock repone of rule1_eu to lnatod hock 0.0480 0.0315 repone of lnmarket to lnatod hock 0.0090 0.0060 repone of lngdp to lnatod hock 0.6234 0.0880 repone of lnatod to lnatod hock 3.7310 13.8025 repone of NPLtoTL to lnatod hock 0.0223 0.1834 0.5201 0.2086 repone of lnltoa to lnatod hock repone of rule2_gen to lnatod hock 0.0486 0.0331 repone of lnmarket to lnatod hock 0.0001 0.0160 repone of lngdp to lnatod hock 0.6005 0.0955 repone of lnatod to lnatod hock 3.0442 8.7116 repone of NPLtoTL to lnatod hock 0.0525 0.1876 0.1533 0.1412 repone of lnltoa to lnatod hock repone of rule3_gen to lnatod hock 0.0594 0.1665 repone of lnmarket to lnatod hock 0.0136 repone of lngdp to NPLtoTL hock 0.0018 0.1705 repone of lnatod to NPLtoTL hock 29.1224 0.0565 repone of NPLtoTL to NPLtoTL hock 0.0542 0.0409 repone of lnltoa to NPLtoTL hock 0.1606 0.2413 repone of rule1_eu to NPLtoTL hock 0.0509 0.0234 repone of lnmarket to NPLtoTL hock 0.0138 repone of lngdp to NPLtoTL hock 0.0013 0.1806 repone of lnatod to NPLtoTL hock 29.2511 0.2516 repone of NPLtoTL to NPLtoTL hock 0.0628 0.0398 repone of lnltoa to NPLtoTL hock 0.1413 0.2055 repone of rule2_gen to NPLtoTL hock 0.0557 0.0224 repone of lnmarket to NPLtoTL hock 0.0083 0.0104 repone of lngdp to NPLtoTL hock 0.1140 0.1245 repone of lnatod to NPLtoTL hock 38.2990 0.0335 repone of NPLtoTL to NPLtoTL hock 0.0388 0.0698 repone of lnltoa to NPLtoTL hock 0.1180 0.0289 repone of rule3_gen to NPLtoTL hock 0.0217 0.0945 repone of lnmarket to NPLtoTL hock 0.0018 0.0066 0.0499 0.0273 3.1646 1.0489 repone of lngdp to lnltoa hock repone of lnatod to lnltoa hock repone of NPLtoTL to lnltoa hock 0.2600 0.0114 0.1088 0.0977 repone of lnltoa to lnltoa hock repone of rule1_eu to lnltoa hock 0.0041 0.0373 repone of lnmarket to lnltoa hock 0.0022 0.0066 0.0513 0.0261 3.5024 1.0703 repone of lngdp to lnltoa hock repone of lnatod to lnltoa hock repone of NPLtoTL to lnltoa hock 0.2602 0.0098 0.0973 0.0834 repone of lnltoa to lnltoa hock repone of rule2_gen to lnltoa hock 0.0047 0.0404 repone of lnmarket to lnltoa hock 0.0037 0.0039 0.0612 0.0288 2.7118 1.4868 repone of lngdp to lnltoa hock repone of lnatod to lnltoa hock repone of NPLtoTL to lnltoa hock 0.2649 0.0044 0.0301 0.0498 repone of lnltoa to lnltoa hock repone of rule3_gen to lnltoa hock 0.0188 0.0421 repone of lnmarket to lnltoa hock 0.0021 0.0022 (p 5) rule1_eu rule1_eu (p 95) rule1_eu repone of lngdp to rule1_eu hock (p 5) rule1_eu rule1_eu (p 95) rule1_eu 0.0741 0.0409 repone of lnatod to rule1_eu hock (p 5) rule1_eu rule1_eu (p 95) rule1_eu 0.8267 2.2218 repone of NPLtoTL to rule1_eu hock (p 5) rule1_eu rule1_eu (p 95) rule1_eu 0.0171 0.0532 repone of lnltoa to rule1_eu hock (p 5) rule1_eu rule1_eu (p 95) rule1_eu 1.2343 0.1301 repone of rule1_eu to rule1_eu hock (p 5) rule1_eu rule1_eu (p 95) rule1_eu 0.0527 0.0222 repone of lnmarket to rule1_eu hock 0.0021 0.0021 (p 5) rule2_gen rule2_gen (p 95) rule2_gen repone of lngdp to rule2_gen hock (p 5) rule2_gen rule2_gen (p 95) rule2_gen 0.0827 0.0415 repone of lnatod to rule2_gen hock (p 5) rule2_gen rule2_gen (p 95) rule2_gen 0.7926 2.2903 repone of NPLtoTL to rule2_gen hock (p 5) rule2_gen rule2_gen (p 95) rule2_gen 0.0187 0.0596 repone of lnltoa to rule2_gen hock (p 5) rule2_gen rule2_gen (p 95) rule2_gen 1.0642 0.1161 repone of rule2_gen to rule2_gen hock (p 5) rule2_gen rule2_gen (p 95) rule2_gen 0.0521 0.0220 repone of lnmarket to rule2_gen hock 0.0021 0.0315 repone of lngdp to rule3_gen hock (p 5) rule3_gen rule3_gen (p 95) rule3_gen 0.2534 0.0175 (p 5) rule3_gen rule3_gen (p 95) rule3_gen repone of lnatod to rule3_gen hock (p 5) rule3_gen rule3_gen (p 95) rule3_gen 18.6443 0.7117 repone of NPLtoTL to rule3_gen hock (p 5) rule3_gen rule3_gen (p 95) rule3_gen 0.0562 0.1398 repone of lnltoa to rule3_gen hock (p 5) rule3_gen rule3_gen (p 95) rule3_gen 0.3461 0.1887 repone of rule3_gen to rule3_gen hock (p 5) rule3_gen rule3_gen (p 95) rule3_gen 0.0162 0.1603 repone of lnmarket to rule3_gen hock 0.0148 0.0035 repone of lngdp to lnmarket hock 0.0291 0.2450 repone of lnatod to lnmarket hock 1.0868 9.9585 repone of NPLtoTL to lnmarket hock 0.0910 0.0262 repone of lnltoa to lnmarket hock 0.4640 0.0273 repone of rule1_eu to lnmarket hock 0.1167 0.0053 repone of lnmarket to lnmarket hock 0.0137 0.0034 repone of lngdp to lnmarket hock 0.0352 0.2446 repone of lnatod to lnmarket hock 1.1767 10.8604 repone of NPLtoTL to lnmarket hock 0.0888 0.0304 repone of lnltoa to lnmarket hock 0.3974 0.0257 repone of rule2_gen to lnmarket hock 0.1166 0.0065 repone of lnmarket to lnmarket hock 0.0007 0.0092 repone of lngdp to lnmarket hock 0.0815 0.2763 repone of lnatod to lnmarket hock 1.2716 5.2374 repone of NPLtoTL to lnmarket hock 0.1188 0.0327 repone of lnltoa to lnmarket hock 0.1084 0.0724 repone of rule3_gen to lnmarket hock 0.1590 0.0527 repone of lnmarket to lnmarket hock 0.0346 0.2394 repone of lngdp to lngdp hock 0.0132 14.3177 repone of lnatod to lngdp hock 0.9898 repone of NPLtoTL to lngdp hock 0.1555 0.0197 0.1646 0.2694 repone of lnltoa to lngdp hock repone of diff1_eu to lngdp hock 0.1398 repone of lnmarket to lngdp hock 0.0307 0.2443 repone of lngdp to lngdp hock 0.0224 13.4903 repone of lnatod to lngdp hock 1.3621 repone of NPLtoTL to lngdp hock 0.1423 0.0233 0.1556 0.1978 repone of lnltoa to lngdp hock repone of diff2_eu to lngdp hock 0.1356 repone of lnmarket to lngdp hock 0.0343 repone of lngdp to lngdp hock 0.0149 0.1970 repone of lnatod to lngdp hock 0.4211 12.6774 repone of NPLtoTL to lngdp hock 0.1363 0.0391 0.3820 0.0379 repone of lnltoa to lngdp hock repone of diff3_eu to lngdp hock 0.1492 repone of lnmarket to lngdp hock 0.0101 0.0063 0.0509 repone of lngdp to lnatod hock 0.6192 16.7902 repone of lnatod to lnatod hock 3.3545 repone of NPLtoTL to lnatod hock 0.0306 0.1906 0.4824 0.5857 repone of lnltoa to lnatod hock repone of diff1_eu to lnatod hock 0.0506 0.0326 repone of lnmarket to lnatod hock 0.0087 0.0063 0.0530 repone of lngdp to lnatod hock 0.6127 14.2156 repone of lnatod to lnatod hock 3.2401 repone of NPLtoTL to lnatod hock 0.0307 0.1911 0.4174 0.4517 repone of lnltoa to lnatod hock repone of diff2_eu to lnatod hock 0.0466 0.0353 repone of lnmarket to lnatod hock 0.0106 0.0065 repone of lngdp to lnatod hock 0.6179 0.1258 repone of lnatod to lnatod hock 4.1290 12.8281 repone of NPLtoTL to lnatod hock 0.0381 0.1798 0.2331 0.1108 repone of lnltoa to lnatod hock repone of diff3_eu to lnatod hock 0.0584 0.0392 repone of lnmarket to lnatod hock 0.0107 0.0028 repone of lngdp to NPLtoTL hock 0.0270 0.1354 repone of lnatod to NPLtoTL hock 35.5621 0.0033 repone of NPLtoTL to NPLtoTL hock 0.0527 0.0457 repone of lnltoa to NPLtoTL hock 0.2696 0.0889 repone of diff1_eu to NPLtoTL hock 0.0333 0.0385 repone of lnmarket to NPLtoTL hock 0.0105 0.0038 repone of lngdp to NPLtoTL hock 0.0325 0.1407 repone of lnatod to NPLtoTL hock 33.8446 repone of NPLtoTL to NPLtoTL hock 0.0484 0.0561 repone of lnltoa to NPLtoTL hock 0.2339 0.0557 repone of diff2_eu to NPLtoTL hock 0.0268 0.0424 repone of lnmarket to NPLtoTL hock 0.0105 0.0033 repone of lngdp to NPLtoTL hock 0.1509 repone of lnatod to NPLtoTL hock 32.3386 repone of NPLtoTL to NPLtoTL hock 0.0448 0.0411 repone of lnltoa to NPLtoTL hock 0.0499 0.0874 repone of diff3_eu to NPLtoTL hock 0.0299 0.0336 repone of lnmarket to NPLtoTL hock 0.0020 0.0047 0.0475 0.0283 3.7456 1.0939 repone of lngdp to lnltoa hock repone of lnatod to lnltoa hock repone of NPLtoTL to lnltoa hock 0.2614 0.0107 0.0782 0.1846 repone of lnltoa to lnltoa hock repone of diff1_eu to lnltoa hock 0.0035 0.0334 repone of lnmarket to lnltoa hock 0.0023 0.0051 0.0487 0.0281 3.1627 1.0757 repone of lngdp to lnltoa hock repone of lnatod to lnltoa hock repone of NPLtoTL to lnltoa hock 0.2609 0.0097 0.0621 0.1665 repone of lnltoa to lnltoa hock repone of diff2_eu to lnltoa hock 0.0059 0.0322 repone of lnmarket to lnltoa hock 0.0021 0.0072 0.0466 0.0303 3.3632 1.0344 repone of lngdp to lnltoa hock repone of lnatod to lnltoa hock repone of NPLtoTL to lnltoa hock 0.2592 0.0070 0.0380 0.0564 repone of lnltoa to lnltoa hock repone of diff3_eu to lnltoa hock 0.0053 0.0455 repone of lnmarket to lnltoa hock 0.0009 0.0116 (p 5) diff1_eu diff1_eu (p 95) diff1_eu repone of lngdp to diff1_eu hock (p 5) diff1_eu diff1_eu (p 95) diff1_eu 0.1587 0.0187 repone of lnatod to diff1_eu hock (p 5) diff1_eu diff1_eu (p 95) diff1_eu 8.2616 0.9390 repone of NPLtoTL to diff1_eu hock (p 5) diff1_eu diff1_eu (p 95) diff1_eu 0.0564 0.0494 repone of lnltoa to diff1_eu hock (p 5) diff1_eu diff1_eu (p 95) diff1_eu 1.2065 0.0576 repone of diff1_eu to diff1_eu hock (p 5) diff1_eu diff1_eu (p 95) diff1_eu 0.0637 repone of lnmarket to diff1_eu hock 0.0014 0.0132 (p 5) diff2_eu diff2_eu (p 95) diff2_eu repone of lngdp to diff2_eu hock (p 5) diff2_eu diff2_eu (p 95) diff2_eu 0.1899 0.0166 repone of lnatod to diff2_eu hock (p 5) diff2_eu diff2_eu (p 95) diff2_eu 10.3360 1.1343 repone of NPLtoTL to diff2_eu hock (p 5) diff2_eu diff2_eu (p 95) diff2_eu 0.0619 0.0589 repone of lnltoa to diff2_eu hock (p 5) diff2_eu diff2_eu (p 95) diff2_eu 1.0329 0.0565 repone of diff2_eu to diff2_eu hock (p 5) diff2_eu diff2_eu (p 95) diff2_eu 0.0660 repone of lnmarket to diff2_eu hock 0.0093 (p 5) diff3_eu diff3_eu (p 95) diff3_eu repone of lngdp to diff3_eu hock (p 5) diff3_eu diff3_eu (p 95) diff3_eu 0.1013 0.0581 repone of lnatod to diff3_eu hock (p 5) diff3_eu diff3_eu (p 95) diff3_eu 0.2503 6.3151 repone of NPLtoTL to diff3_eu hock (p 5) diff3_eu diff3_eu (p 95) diff3_eu 0.0372 0.0465 repone of lnltoa to diff3_eu hock (p 5) diff3_eu diff3_eu (p 95) diff3_eu 0.4217 0.0025 repone of diff3_eu to diff3_eu hock (p 5) diff3_eu diff3_eu (p 95) diff3_eu 0.0781 repone of lnmarket to diff3_eu hock 0.0155 0.0029 repone of lngdp to lnmarket hock 0.0281 0.2174 repone of lnatod to lnmarket hock 1.1034 12.6978 repone of NPLtoTL to lnmarket hock 0.0906 0.0272 repone of lnltoa to lnmarket hock 0.0485 0.4671 repone of diff1_eu to lnmarket hock 0.1205 0.0027 repone of lnmarket to lnmarket hock 0.0158 0.0032 repone of lngdp to lnmarket hock 0.0227 0.2234 repone of lnatod to lnmarket hock 1.2323 13.1787 repone of NPLtoTL to lnmarket hock 0.0857 0.0321 repone of lnltoa to lnmarket hock 0.0490 0.3582 repone of diff2_eu to lnmarket hock 0.1215 0.0027 repone of lnmarket to lnmarket hock 0.0138 0.0045 repone of lngdp to lnmarket hock 0.0291 0.2664 repone of lnatod to lnmarket hock 1.6113 11.5508 repone of NPLtoTL to lnmarket hock 0.0976 0.0344 repone of lnltoa to lnmarket hock 0.1858 0.0426 repone of diff3_eu to lnmarket hock 0.1110 0.0089 repone of lnmarket to lnmarket hock Table 12: Impule Repone Function for Euro Area Taylor rule 1 lag VAR of lngdp lnatod NPLtoTL lnltoa rule1_eu lnmarket Sample : if euro==1 1 lag VAR of lngdp lnatod NPLtoTL lnltoa diff1_eu lnmarket Sample : if euro==1 Error are 5% on each ide generated by Monte Carlo with 500 rep Error are 5% on each ide generated by Monte Carlo with 500 rep Flexible Taylor rule with inflation and output-gap 1 lag VAR of lngdp lnatod NPLtoTL lnltoa rule2_gen lnmarket Sample : if euro==1 1 lag VAR of lngdp lnatod NPLtoTL lnltoa diff2_eu lnmarket Sample : if euro==1 Error are 5% on each ide generated by Monte Carlo with 500 rep Error are 5% on each ide generated by Monte Carlo with 500 rep Flexible Taylor rule with interet moothing, output-gap, and total credit. 1 lag VAR of lngdp lnatod NPLtoTL lnltoa rule3_gen lnmarket Sample : if euro==1 1 lag VAR of lngdp lnatod NPLtoTL lnltoa diff3_eu lnmarket Sample : if euro==1 Error are 5% on each ide generated by Monte Carlo with 500 rep Error are 5% on each ide generated by Monte Carlo with 500 rep 39

Table 13: FEVD - Taylor rule gap for Euro Area Main variable Explanatory variable Short period: 2 year lngdp lnatod NPLtoTL lnltoa rule1eu diff1eu diff1country lnmarket lngdp.81915474.01722789.00553316.00238615.00005647.00010819.15100178.00453161 lnatod.0207118.94112043.00776024.00014124.00888213.00082079.00120506.01935831 NPLtoTL.00615932.01562048.97253702.00003947.00009223.00002969.00030463.00521716 lnltoa.01420991.2342252.0083505.72258812.00375453.0014691.00751198.00789067 rule1eu.05340699.01287179.0052467.00184875.90365241.00882465.0045596.0095891 diff1eu.01593421.00228152.00877045.00665618.82873504.12640249.00006487.01115524 diff1country.00930171.09282978.0000116.00313721.12737514.02408974.73944866.00380617 lnmarket.28288367.02499521.00080162.0287429.04957615.07178792.06327809.47793444 Medium period: 6 year lngdp lnatod NPLtoTL lnltoa rule1eu diff1eu diff1country lnmarket lngdp.69789359.01306642.04930184.00384323.00754408.01126802.21304707.00403575 lnatod.0440222.81593384.04835297.00165523.03833966.00356639.01767874.03045097 NPLtoTL.01686952.03134294.92925556.00079348.00238512.00016725.00516957.01401657 lnltoa.06401176.20893085.01038149.65900765.00653019.00206799.03367615.01539393 rule1eu.05478895.01533333.01979679.00201535.87530498.01173617.00994066.01108377 diff1eu.04043469.00642082.01423099.0062103.79898843.11734116.00531601.01105759 diff1country.01809125.09178274.0107932.00301632.14159429.028123.7023215.0042777 lnmarket.29993819.03154364.02114942.03254556.09136916.05771416.10351534.36222453 Long Period: 10 year lngdp lnatod NPLtoTL lnltoa rule1eu diff1eu diff1country lnmarket lngdp.64260795.01668342.09903825.00378299.01055689.01351981.21038127.00342941 lnatod.04942521.78636137.07046884.00172386.03781721.00403116.02088605.02928629 NPLtoTL.02433137.03210808.91401181.00119432.00364533.00043058.00862731.01565121 lnltoa.08229326.20078769.0203575.62754783.00802195.00293331.04293753.01512093 rule1eu.05573097.01579607.02667285.00200579.86642094.01173534.01064812.01098991 diff1eu.04613357.00639579.01424568.00625.79117004.11645402.0082638.01108709 diff1country.02119903.09122837.01789318.00298572.13987477.02795463.69463846.00422584 lnmarket.30436101.03237458.04717771.03063221.087443.05511286.11015279.33274585 * Variance-decompoition: percent of variation in the row variable explained by column variable * rule1= Taylor rule at aggregate level. diff1eu=taylor gap, difference between nominal interet rate and aggregate Taylor rule. diff1country= difference between the Taylor gap, computed at aggregate level, and the optimal rule for the ingle country. 40

Table 14: FEVD - Taylor rule gap Main variable Explanatory variable long period: 2 year lngdp lnatod NPLtoTL lnltoa rule2eu diff2eu diff2country lnmarket lngdp.81884776.01723723.00553017.00239804.00005953.00013033.15126048.00453646 lnatod.02072014.94112589.00778423.0001427.00882052.00082142.00124015.01934495 NPLtoTL.00616837.01563129.97250605.00003974.00009472.00002867.00031182.00521935 lnltoa.0142274.23408064.00834047.72276388.00374524.0014278.00751543.00789914 rule2eu.0534378.01276622.00531532.0018429.90351385.00870052.00481469.00960871 diff2eu.02689465.00287716.00907217.00752394.78016068.16249506.00045959.01051676 diff2country.00921988.09214706.00000894.00314227.12779656.02392554.73994284.00381691 lnmarket.28275034.02467085.0008236.02878721.04958638.07066974.06446351.47824836 long period: 6 year lngdp lnatod NPLtoTL lnltoa rule2eu diff2eu diff2country lnmarket lngdp.69744106.01307891.04934592.00385197.00732995.01134664.21360843.00399713 lnatod.04411014.81609832.04849422.00164306.03788386.00357918.01786366.03032755 NPLtoTL.01690661.03134492.92922312.00079212.00232945.00016886.00525018.01398474 lnltoa.0640446.20876542.01037887.65908122.00649062.00201939.03381603.01540385 rule2eu.05488501.01531279.01990445.00200233.87504647.01160912.0102055.01103434 diff2eu.06333509.00719808.01301215.00707239.74285992.1465448.00979316.01018441 diff2country.01811055.09124756.01085648.00302374.14160287.02796215.70294695.00424969 lnmarket.30024591.03142656.02126838.03255134.09030724.0569919.1049055.36230318 Medium Period: 10 year lngdp lnatod NPLtoTL lnltoa rule2eu diff2eu diff2country lnmarket lngdp.64214403.01672051.09917525.00378215.01026902.01360585.21091428.00338891 lnatod.04956897.78634289.07071742.0017096.0373527.00404857.02109962.02916024 NPLtoTL.02442424.0320971.91392991.00119059.0035601.00043569.00875663.01560574 lnltoa.08236973.20061547.02038371.62754495.0079447.00289522.04312533.01512087 rule2eu.05584569.01578066.02683253.00199237.86607541.01161035.01092266.01094032 diff2eu.07179425.00721658.0134765.0071091.73123194.14464352.01434538.01018273 diff2country.02126399.09069429.01803039.00299228.13985796.02779503.69516824.00419781 lnmarket.30471292.03228735.04747511.03061083.0863467.0544382.11151936.33260953 * variance-decompoition: percent of variation in the row variable explained by column variable ** rule2= aggregate flexible Taylor rule, with different weight for inflation and output-gap. diff2eu=taylor gap, difference between nominal interet rate and aggregate flexible Taylor rule. diff2country= difference between the Taylor gap computed at aggregate level, and the optimal rule for the ingle country. 41

Table 15: FEVD - Taylor rule gap for Euro Area Main variable Explanatory variable Short period: 2 year lngdp lnatod NPLtoTL lnltoa rule3eu diff3eu diff3country lnmarket lngdp.86066031.00693113 6.165e-06.00016834.0497036.00095827.06509513.01647706 lnatod.02261692.89571083.00815141.00007755.00528538.0007701.00351186.06387596 NPLtoTL.0083619.02721843.94964078.00001217.00104525.00107462.000753.01189383 lnltoa.01746473.3254209.00390715.5829625.0199543.00305563.00709273.04014205 rule3eu.35353848.07765037.00380707.02124681.44635265.0325197.01165905.05322587 diff3eu.20565059.23178071.0129611.00732925.37259782.09414074.00160094.07393886 diff3country.00277719.18855644.00139296.00456759.07828919.07077034.63653759.01710869 lnmarket.3065983.19318401.00847261.02172354.218992.04305025.00045725.20752204 Medium period: 6 year lngdp lnatod NPLtoTL lnltoa rule3eu diff3eu diff3country lnmarket lngdp.59051725.03004702.02831288.00127356.17535911.00516581.15191523.01740914 lnatod.04606555.84612694.02233753.00021824.00843833.00083891.00659549.06937902 NPLtoTL.01015198.02274938.93439841.00029234.01109574.00093522.00445655.01592039 lnltoa.04275195.31099169.00651395.55280636.02663449.00422394.01373567.04234195 rule3eu.26376289.06498789.01529974.01622123.41007908.02814521.06731731.13418665 diff3eu.20392791.16323362.01850703.01148536.38198923.07628016.04319599.10138069 diff3country.01784071.17587251.0102449.00753063.08116551.06643041.6047805.03613483 lnmarket.2669486.16060354.0160897.01926636.26257286.04194703.03575734.19681458 Long Period: 10 year lngdp lnatod NPLtoTL lnltoa rule3eu diff3eu diff3country lnmarket lngdp.56899056.03452158.05194065.0011911.17037769.00489553.15204794.01603495 lnatod.05029136.83351735.02887643.00026906.00954865.00088076.0082827.06833369 NPLtoTL.0143299.02237467.92695921.00028718.01266134.00092511.00593367.01652891 lnltoa.05386726.30488207.01359365.53574805.02760314.00410966.01800584.04219033 rule3eu.27202284.06436515.01833717.01578492.40350176.02747528.06736371.13114917 diff3eu.20262969.16172106.01849069.01148979.38164439.07534663.04371163.10496612 diff3country.02018038.17476862.0133492.00747953.08425463.0655598.59835403.03605381 lnmarket.26657237.15815962.018754.01915151.26339161.04118682.03886133.19392273 ** variance-decompoition: percent of variation in the row variable explained by column variable rule3= aggregate flexible Taylor rule with interet moothing, output gap and total credit. diff3=taylor gap, difference between nominal interet rate and aggregate flexible Taylor rule. diff3country= difference between the Taylor gap computed at aggregate level, and the optimal rule for the ingle country. 42