Discussion of Large Banks, Loan Rate Markup, and Monetary Policy

Transcription

1 Discussion of Large Banks, Loan Rae Markup, and Moneary Policy Dean Corae Universiy of Wisconsin Madison and NBER 1. Inroducion Cuciniello and Signorei (his issue, hereafer CS) provide a New Keynesian dynamic sochasic general equilirium (NKDSGE) model wih imperfec compeiion in he anking indusry and collaeral-consrained orrowers o address some imporan quesions. How much does anking indusry marke srucure amplify usiness cycles? How does sraegic ineracion eween ig anks and he cenral ank affec aggregae oucomes? The idea ha ank marke srucure and real economic oucomes are conneced is imporan, and I hink he auhors make a nice conriuion o he lieraure. CS esalish some ineresing resuls. Firs, sraegic ineracion among a small numer (n) of anks generaes counercyclical loan markups which depend on (i) he elasiciy of susiuion eween loan ypes (ɛ ), firm leverage, and how he cenral ank responds o inflaion (φ π ) in is policy rule. Second, counercyclical markups amplify he ransmission of moneary and echnology shocks on he real economy, which is asen in he perfecly compeiive case (i.e., consan markups when n ). Third, he level of he spread is posiively relaed o firm leverage (using he accouning measure of capial asses o ne worh). Fourh, amplificaion of financial (or echnology) shocks is weaker (sronger) he more aggressive he cenral ank s response o inflaion. There is susanial heerogeneiy in he anking secor. Firs, ank marke srucure varies susanially across counries. For insance, in 2011, he asse marke share of he op hree anks in I am graeful o Akio Ino for compuaional assisance on his discussion. 179

2 180 Inernaional Journal of Cenral Banking June 2015 Tale 1. Measures of Compeiion in Banking Value Sd. Error Corr. wih Momen (%) (%) GDP Ne Ineres Margin Lerner Index Markup Rosse-Panzar H Source: FDIC, Call Repors, and Thrif Financial Repors. Values correspond o he ime-series average of he loan-weighed cross-secional averages. Consisen daa availale from 1984 o Noes: Daa correspond o commercial anks in he Unied Saes. he following counries varied eween Porugal a 89 percen, Germany a 78 percen, he Unied Kingdom a 58 percen, Japan a 44 percen, and he Unied Saes a 35 percen. While his heerogeneiy provides a poenially imporan source of variaion for idenificaion, i is no necessarily exogenous and can vary, for insance, wih governmen policy. Second, even wihin counries here is susanial heerogeneiy among anks. If all anks wihin a counry are symmeric, hen asse marke share is 1/n. Symmery would hen imply, for example, ha U.S. op hree marke share should e roughly 3/7000, no 35 percen. Oviously, symmery is a srong simplifying assumpion. Again, his heerogeneiy is no necessarily exogenous and can vary, for insance, wih governmen policy and region-specific shocks. CS cie numerous papers which documen counercyclical ank markups. For insance, ale 1 (aken from Corae and D Erasmo 2013) uses Call Repor daa on all U.S. commercial anks from 1984 o Several of hese papers provide evidence for imperfec compeiion in he anking secor including high margins and imperfec pass-hrough of coss (as measured y he Rosse-Panzar H-saisic). Counercyclical markups in he anking secor provide a new amplificaion mechanism. For insance, if markups on loans are counercyclical, hen loan raes rise in a downurn, choking off invesmen and furher amplifying he downurn. This is apparen in impulse response funcions (shown elow) for a model wih an imperfecly compeiive anking secor versus he perfecly compeiive case.

3 Vol. 11 No. 3 Discussion: Corae 181 This discussion will e organized in he following secions. Secion 2 will riefly descrie essenial pars of CS s model, while secion 3 will presen some resuls. Finally, secion 4 will discuss quesions concerning rousness of he resuls as well as fuure direcions. 2. Model 2.1 Environmen There are equal masses of wo ypes of agens: (i) households (P ) who work, deposi resources in ank, and consume; and (ii) enrepreneurs (E) who produce inermediae goods, accumulae capial, and ake loans from anks which mus e collaeralized. Households discoun he fuure a a lower rae han enrepreneurs (i.e., β P >β E ), wih β E low enough ha he enrepreneurs collaeral consrain always inds. There are hree ypes of non-financial firms: (i) capial goods producers who operae in a perfecly compeiive marke a real price q k sujec o adjusmen coss; (ii) enrepreneurs who produce inermediae goods via a consan-reurns-o-scale echnology wih whie-noise aggregae produciviy shock A E ; and (iii) reailers who differeniae inermediae goods, sell a a markup, and face quadraic price adjusmen coss (which delivers a New Keynesian Phillips curve). There are a finie numer n 2 of idenical anks which collec deposis from households and issue loans o enrepreneurs. The deposi marke is assumed o e perfecly compeiive, while he loan marke is modeled along he lines of Gerali e al. (2010), where loans of differen ypes are aggregaed using a Dixi-Sigliz syle aggregaor wih each ank having a share of oal loans equal o 1/n. Finally, he cenral ank ses he inerank rae R i o lean agains inflaion wind (wih sensiiviy φ π ) according o he following rule: R i = R i π φ π exp ) (ε Ri,

4 182 Inernaional Journal of Cenral Banking June 2015 where he inflaion rae is given y π = log(p ) log(p 1 ) and ε Ri is a whie-noise policy innovaion. 2.2 Equilirium Mos of he model environmen is familiar from earlier NKDSGE models. The wo key differences, which I will highligh here, are he enrepreneur s collaeral consrain and, more imporanly, he finie numer of anks choosing loan raes. Enrepreneur i chooses consumpion, capial, and loans (c E (i),k E (i), E (i)) o maximize he expeced presen discouned value of log uiliy sujec o a sandard udge consrain and an ex ane collaeral consrain 1 E (i) E [ m E q+1k k E (i)(1 δ k ) ] R, (1) where m E is he loan-o-value raio. Using sufficien condiions such ha he collaeral consrain (1) is always inding, he firs-order condiion of he enrepreneur yields he loan demand decision rule: E = β E χ q (R χ ) nwe, (2) where χ (1 δ k )m E E [ q k +1 ] 1 Consrain (1) is used in several DSGE models (e.g., Iacoviello 2005). Is microfoundaions are, however, differen from Kiyoaki and Moore (1997). In Kiyoaki-Moore, here is no aggregae uncerainy, so ex ane is equal o ex pos and he consrain makes defaul su-opimal. Bu he ex pos condiion is wha is relevan for he commimen prolem. In paricular, in period +1, if enrepreneur i does no pay ack, he gains R E (i) u loses his collaeral q k +1k E (i)(1 δ k ). For sufficienly low realizaions of q k +1, he enrepreneur migh defaul ex pos in he curren model, which is no accouned for in he ex ane analysis.

5 Vol. 11 No. 3 Discussion: Corae 183 and ne worh is defined as nw E r k k E 1 R E 1 + q k (1 δ k )k E 1. (3) Afer susiuing necessary condiions, leverage is given y LV = ( ), 1 E q kke ( ) which yields counercyclical leverage (since cor Y, E < 0). q kke An alernaive inerpreaion of CS is ha loans of size E (i) mus e syndicaed according o a loan producion funcion wherey enrepreneur i needs loans from anks j =1, 2,..., n, which solves he following cos-minimizaion prolem: min { E (i,j)} j s.. β E n R(j) E (i, j) (4) j=1 ( 1 n ) 1 ɛ n j=1 E (i, j) ɛ 1 ɛ The soluion o his prolem is given y E (i, j) = ( )( 1 R (j) n R ) ɛ ɛ ɛ 1 E (i). (5) E (i), (6) where 1 R 1 1 ɛ n R n (j) 1 ɛ. (7) j=1 CS assume ha each ank j solves a saic profi-maximizaion prolem, choosing loan rae R(j) o maximize profis: [R max (j) R i ] R (j) E (i, j)di, (8)

6 184 Inernaional Journal of Cenral Banking June 2015 where he demand for syndicaed loan E (i, j) is given y (6). Given a finie numer of anks, a ank s choice of ineres rae acually affecs ne worh of enrepreneurs in he fuure, which is no eing aken ino accoun owing o he saic profi-maximizaion assumpion. The firs-order condiion which solves (8) weighs he coss and enefis o ank j from raising is ineres rae: (A) Direc posiive effec of raising own rae on profis: ( ) R ɛ (j) E 1 R (B) Indirec negaive effec of compeiion from oher anks (noe ha he elasiciy of susiuion eween loans ɛ maers direcly): ɛ [ R (j) R i ] E ( R (j) R ) (ɛ +1) (C) Indirec negaive effec on overall loan demand: + [ R (j) R i ] ( R (j) R ) ɛ (+) R R(j) R ( ) R 2 ( ) (+) E R R R(j) (D) Indirec posiive effec on cenral ank policy choice: ( ) R i R (+) R R (j) ( R (j) R ) ɛ E R (j) Imporanly, in a symmeric Nash equilirium, he soluion o he aove firs-order condiion yields he loan markup over he ineres i pays deposiors (also he cenral ank policy rae): B C D {}}{{}}{{}}{ ɛ (n 1) + Ξ R, + Ξ R i = ɛ R i (n 1) +Ξ }{{}, n }{{}}{{} B C A M R i, (9)

7 Vol. 11 No. 3 Discussion: Corae 185 where Ξ, is he asolue value of he elasiciy of aggregae loan demand o loan raes, and Ξ R i is he asolue value of he elasiciy of cenral ank policy raes o loan raes (i.e., an increase in ank j s rae R (j) increases he overall loan rae R, which lowers aggregae demand and inflaion, which ulimaely leads o a decrease in he cenral ank policy rae R i ). In a compeiive environmen where n, he markup M in (9) simply depends on he elasiciy of susiuion eween loans from differen anks given y lim n M = ɛ so ha a higher degree of susiuion leads o lower markups. The compeiive case also implies ha markups do no vary wih he cycle. Wih a finie numer of anks, he cyclical properies of markups depend on many facors. Using he definiion of markups in (9), one can compue dm dy o deermine he imporan facors in he cyclical properies of he markup. In paricular, markups are counercyclical (i.e., dm dy < 0) if [ ] n +Ξ R i dξ, dy + [ ɛ (n 1)+Ξ, n ] dξ R i [ ɛ +(ɛ 1) Ξ R i Ξ, ] dy dn dy < 0, (10) dξ Ri dy where dξ, dy < 0 since leverage is counercyclical and < 0y an argumen similar o ha in ulle 2 following (9). Noe ha in (10) I have also considered how markups would vary in heir model if he numer of anks also varied wih he cycle. Since dn dy =0y assumpion in CS s model, markups are counercyclical if he commercial versus cenral ank sraegic impacs (i.e., he second erm in (10)) ouweigh he leverage effecs across he cycle (i.e., he firs

8 186 Inernaional Journal of Cenral Banking June 2015 erm in (10)), provided ɛ is sufficienly large (which he caliraion akes elow). 3. Resuls There are hree key (new) parameers in CS s model: (i) he elasiciy of susiuion across loan varieies ɛ ; (ii) he sensiiviy of he cenral ank policy ineres rae o inflaion in he policy rule φ π ; and (iii) he numer of anks n or marke share 1/n. Tale1inCS provides he parameer values. In paricular, loan elasiciy (which affecs he seady-sae loan spread) is se a ɛ = 161. The parameer is caliraed so ha, in he case of aomisic anks (i.e., n = ), he seady-sae gross markup M equals This value, in urn, corresponds o a ne spread eween he loan rae and he policy rae of around 2.5 percenage poins in annual erms. The cenral ank policy rule akes φ π =1.5. The auhors vary marke srucure (i.e., n) eween 3 and infiniy (he aomisic case). Figure 1 elow reproduces he impulse response funcions under hese parameer values for he more sandard case where here is an AR1 echnology shock process wih persisence equal o 0.9 insead of i.i.d. shocks as in CS s figure 4 under oh he aomisic (n = ) case and he non-aomisic (n = 3) case. Noe ha in order o make comparisons eween he aomisic and non-aomisic cases, CS use a susidy which fully offses he disorion associaed wih monopolisic compeiion in he anking secor in he seady sae (see secion 5 of heir paper). 2 Figure 1 wih persisen echnology shocks is qualiaively similar o CS s figure 4 wih i.i.d. shocks. Figure 1 makes clear ha he 2 In paricular, hey se Υ so ha M ss =1, (11) 1+Υ which means ha he afer-ax loan markup is 1. For he aomisic ank case, since M ss = ɛ 1, hey se Υ =. For he non-aomisic case, here is no a ɛ 1 ɛ 1 closed-form soluion for Υ, as M ss depends on Υ. So one mus find a fixed poin of Υ under which he afer-ax markup is 1. CS se Υ = M ss(υ) 1.

9 Vol. 11 No. 3 Discussion: Corae 187 Figure 1. Impulse Response Funcions in CS Model aomisic ank case (n = where loan markups do no vary wih he cycle) yields less amplificaion in key variales han he nonaomisic case (n = 3 where loan markups are counercyclical). Here, however, figure 1 shows ha he CS model predics ha profis wih a finie numer of anks are counercyclical in he shor run. Empirical papers end o find procyclical ank profis (see, for example, Alerazzi and Gamacora 2009). Procyclical profis and counercyclical markups are consisen if loan originaions are sufficienly procyclical (a propery consisen wih daa).

10 188 Inernaional Journal of Cenral Banking June Rousness 4.1 Exogenous Numer of Banks Oviously, in he very shor run, one may wan o ake he numer of anks n as a parameer, u surely in he long-run marke share is no exogenous nor invarian o policy experimens as assumed y CS. To see he implicaions of exogenous marke srucure, one can use he same parameer values as in ale 1 of CS s paper wih persisen echnology shocks u add a free enry condiion. As a resul of he ax policy assumed in CS, seady-sae ank profis are zero. Hence, o consider enry ino he anking secor, we need o assume enry coss are zero. In ha case, he zero-profi condiion in a model wih ank enry is ( M 1+Υ 1 ) R i E =0. (12) Since E > 0, his implies ha in he equilirium wih zero enry cos, he afer-ax markup should always e 1. Thus he equilirium allocaion is exacly he same as in he case of he aomisic ank case, u (12) pins down he numer of anks endogenously (i.e., he aomisic and endogenous n lines lie on op of one anoher). As one can see from our figure 1 (compared wih CS s figure 4) here are some key differences eween he case wih free enry (i.e., endogenous n ) and he aseline model ( non-aomisic ) in CS. Free enry implies ha he numer of anks mus acually fall in order o mainain he zero-profi condiion, a counerfacual predicion (counercyclical enry in he model while here is procyclical enry in he daa). This is relaed o he counercycical profis in CS s aseline non-aomisic case. Ceeris parius, negaive ank profis requires exi in order o e consisen wih he zero-profi condiion. Counercyclical enry (i.e., dn dy < 0) in condiion (10) is why here is less amplificaion in he endogenous n impulse response funcions in figure 1. Thus, he channel eween he financial acceleraion markup and propagaion of shocks is no necessarily rous. There are oher ways o generae he link eween he financial acceleraion markup and propagaion of shocks. I is possile o wrie DSGE models wih counercyclical markups due o he compeiive effecs of procyclical enry and counercyclical exi in he

11 Vol. 11 No. 3 Discussion: Corae 189 Figure 2. Impulse Response Funcions in CI Model case of a finie numer of symmeric firms. For insance, Jaimovich and Floeoo (2008) do his in he conex of non-financial firms, and i is possile o apply heir ideas o he anking indusry. Figure 2 graphs he impulse response funcions for a case wih endogenous ank enry in Corae and Ino (2015) (hereafer CI) wihou New Keynesian fricions u wih posiive enry coss. The figure shows ha procyclical ank enry in response o a favorale produciviy shock can lead o counercyclical markups even wih perfecly flexile prices. 3 Of course, hese DSGE models canno speak o he vas heerogeneiy in he ank size disriuion due o heir simplifying assumpion on symmery. 3 For his paricular parameerizaion, however, he effecs are no large in figure 2.

12 190 Inernaional Journal of Cenral Banking June Saic Bank Opimizaion CS assume ha anks are myopic and choose ineres raes no aking ino accoun ha heir choice of ineres rae may in fac aler fuure orrowing y enrepreneurs (and hence fuure profis) since each non-aomisic ank in heir model has marke power. I is possile, however, o incorporae forward-looking ehavior and srucure layers of non-compeiive and compeiive forward-looking firms (e hey financial or non-financial). See, for insance, Gerali e al. (2010) and he appendix in Jaimovich and Floeoo (2008). This will e imporan if his framework is o e used o undersand ank capial and macroprudenial policies. 4.3 Symmery Ever since Kashyap and Sein s (2000) influenial empirical paper, we know he moneary ransmission mechanism works differenly for ig and small anks, u here has een lile srucural work on ha opic. In order o simply he analysis, CS assume ha all anks are symmeric, which is oviously inconsisen wih he daa. I is possile o inroduce ank heerogeneiy ino srucural models wih imperfec compeiion and dynamic ank opimizaion using enry/exi (like ha in Jaimovich and Floeoo) o generae counercyclical markups. In a series of papers, Corae and D Erasmo (2013, 2014a, 2014) also generae a financial acceleraion markup wih heerogeneous anks a he expense of compuaional complexiy. 4.4 Binding Collaeral Consrains No all non-financial firms are collaeral consrained in he economy. I may prove useful o apply mehods from occasionally inding consrains o his framework (e.g., Guerrieri and Iacoviello 2015). This is also relaed o he fac ha he seady-sae de-o-oupu level is 10 in CS s model u a mos 3 in he daa. If one ses m E =0.375 and n =3, he model produces a seady-sae de level of 3 as opposed o m E =0.8 and n = 3 in CS s enchmark which yields a seady-sae de level of 10. The former case, however, yields a smaller response o echnology shocks. In paricular, oupu and

13 Vol. 11 No. 3 Discussion: Corae 191 invesmen are nearly half as large in response o a echnology shock when he de-o-oupu level is a 3 raher han Conclusion CS ake an imporan sep o ring ank marke srucure ino he DSGE framework. Their work provides a simple framework o ask how policy ineracs wih an exogenous marke srucure o affec aggregae oucomes and welfare. A paricularly ineresing and novel par of heir analysis is how he sraegic ineracion eween commercial anks wih marke power and he cenral ank can amplify shocks in he economy (i.e., erm D in (9)). CS s resuls are complemenary o oher papers. For insance, Corae and D Erasmo (2013, 2014a, 2014) also generae a financial acceleraion markup wih an endogenous size disriuion of anks hrough enry and exi affecing compeiive marke forces. Endogenizing marke srucure means ha policy counerfacuals can affec marke srucure (i.e., here are no issues wih he Lucas criique) as well as marke srucure affecing policy. I elieve his o e a very fruiful avenue of fuure research. References Alerazzi, U., and L. Gamacora Bank Profiailiy and he Business Cycle. Journal of Financial Sailiy 5 (4): Corae, D., and P. D Erasmo A Quaniaive Model of Banking Indusry Dynamics. Mimeo a. Capial Requiremens in a Quaniaive Model of Banking Indusry Dynamics. Mimeo Foreign Compeiion and Banking Indusry Dynamics: An Applicaion o Mexico. Mimeo. Corae, D., and A. Ino A Simple DSGE Model of Banking Indusry Dynamics. Mimeo. Gerali, A., S. Neri, L. Sessa, and F. Signorei Credi and Banking in a DSGE Model of he Euro Area. Journal of Money, Credi and Banking 42 (s1):

14 192 Inernaional Journal of Cenral Banking June 2015 Guerrieri, L., and M. Iacoviello OccBin: A Toolki for Solving Dynamic Models wih Occasionally Binding Consrains Easily. Journal of Moneary Economics 70: Iacoviello, M House Prices, Borrowing Consrains, and Moneary Policy in he Business Cycle. American Economic Review 95 (3): Jaimovich, N., and M. Floeoo Firm Dynamics, Markup Variaions, and he Business Cycle. Journal of Moneary Economics 55 (7): Kashyap, A., and J. Sein Wha Do a Million Oservaions on Banks Say aou he Transmission of Moneary Policy? American Economic Review 90 (3): Kiyoaki, N., and J. Moore Credi Cycles. Journal of Poliical Economy 105 (2):