The Impact of Discriminatory Credit Constraint on Business Cycle: A DSGE Model with Endogenous Loan-to-Value

Transcription

1 The Impac of Discriminaory Credi Consrain on Business Cycle: A DSGE Model wih Endogenous Loan-o-Value Yuchao Peng Lili Yan February 12, 215 Absrac This paper builds a banking DSGE model wih endogenous loan-o-value raios, aking he differen relaionship beween differen ypes of firms and banks ino accoun. Due o he poliical connecions beween he bank and enerprises, loan o value raio for sae-owned enerprises is endogenously higher han ha for privae enerprises, which is referred o discriminaory credi consrain in his paper. Compared o non-discriminaory credi consrain, we find ha discriminaory credi consrain can furher amplify he impac of negaive echnology shocks on oupu, and reduce he effeciveness of expansionary moneary policy. Empirical evidence from China indusrial firms daa suppors our conclusion. JEL Classificaion: E32, E34 Key Words: Discriminaory Credi Consrain, Endogenous Loan-o-Value, Business Cycle, Poliical Connecions, DSGE model. We are indebed o Dr. Zhichao Zhang a Durham Universiy Business School for coninuous suppor and supervision. We are graeful o Professor Andrea Gerali in Banca d Ialia for help on model diagnosics. This research is funded by Cenral Universiy of Finance and Economics Key Topic Suppor Program in 214. All remaining errors and omissions are ours. Deparmen of Finance, Business School, Durham Universiy, Durham, UK. School of Finance, Cenral Universiy of Finance and Economics, Beijing, China. yuchaopeng@homail.com. School of Finance, Cenral Universiy of Finance and Economics, Beijing, llyanll@163.com 1

2 The longer erm implicaions of focussing lending on sae firms and sarving he high growh privae secor could reurn o haun he banks wih bad debs in he fuure. Le e al. 214) 1 Inrodcuion The US subprime morgage crisis and he following world financial crisis sugges ha financial secor plays an increasingly imporan role in modern economy s running, and i is also he key o undersand how economy flucuaes. Over he las several decades, here has been an increasing lieraure in his area Kiyoaki & Moore, 1997; Bernanke e al., 1999). However, mos of researches pay lile aenion o an exising phenomenon in financial aciviies: discriminaory credi consrain, which may influence he economic flucuaion and he effeciveness of fiscal and moneary policies. For insance, China suffered a sharp drop in oal impors and expors, and a fall in GDP growh rae in he fourh quarer of 28 because of he sluggish global economy. In order o simulae he economy, he Chinese governmen decisively launched he four rillion rescue plan, and simulaneous counercyclical expansionary moneary policy. However, mos of he liquidiy exraced by People Bank of China flew o sae-owned enerprises, causing privae enerprises suffering more binding credi consrain see Figure 1). Even hough he firms provided he collaeral wih he same value, he banks were willing o lend more money o sae-owned firms, which are backed up by Chinese governmen. Banking discriminaion of ownership leads o he misallocaion of financial resources, hus hindering he economy s recovery: China s economic growh experienced a shor recovery in 29 and 21, hen following by a coninuous decline. This paper presens a dynamic sochasic general equilibrium model where he loan-ovalue raios are endogenously by firms and banks. Our main objecive is o undersand weaher discriminaory credi consrain can affec economic flucuaion or no, and o find he reason why counercyclical moneary policy has failed o save he China s economy. [FIGURE 1 ABOUT HERE] This paper fis exising researches on financial fricions and business cycle. Since Kiyoaki and Moore 1997) and Bernanke e al. 1999)henceforh, BGG), numerous scholars have researched on he impac of financial fricions on business cycle 1. On he demand side, he lieraure is mosly based on wo alernaive approaches 2 : one 1 See he survey of Brunnermeier e al. 212) o know he lieraure on financial fricions in macro economics. 2 See Brzoza-Brzezina e al. 213) abou he comparison of wo approaches. 2

3 is abou exernal finance premium financial acceleraor), e.g., Carlsrom and Fuers 1997), Chrisensen and Dib 28), Le e al. 214) and so on; he oher is abou credi consrain e.g., Iacoviello 25, 215). They find financial fricions can sharpen economic flucuaion by amplifying primary shocks. On he supply side, many DSGE models are consruced revolving around he banking secor, e.g. Gerali e al. 21) henceforh, GNSS), Gerler and Karadi 211), Angeloni and Faia 213) and so on. They propose ha bank capial regulaion and he deerioraion of banks balance shee are he essenial facors driving he business cycle. These researches have dramaically conribued o he heoreical developmen of financial fricions. However, when researching on credi consrain, he majoriy of scholars regard he loan-o-value raio as consan or exogenous shock, merely caring for he banking discriminaion on borrowers asse scale and asse price, nay, ignore ha differen ypes of firms may confron differen loan-o-value raios, which may furher influence he economic flucuaion in anoher way. Despie recen researches abou financial crisis and financial regulaion propose ha he loan-o-value raio is an imporan insrumen of macro prudenial policies Mendicino & Punzi, 214; Rubio & Carrasco-Gallego, 214), by assuming ha he loan-o-value raios couner-cyclically respond o oal loan or oupu, however, his assumpion does no fi he realiy, because i is he firms and he banks who decide he raios in order o cope wih he informaion asymmery and o resis loan defaul risks. As a consequence, hese models wih non-discriminaory credi consrain and exogenous loan-o-value raios canno help o solve he problem of China s economy. Our research is also relaed o he lieraure abou poliical connecions, banking discriminaion on he ownership and financial misallocaions. Despie he markeizaion reforms ha have had a major impac on all aspecs of he Chinese economy, sae-owned enerprises and privae enerprises are sill subjec o differen reamen Dollar and Wei, 27)). Especially, sae-owned enerprises have more access o finance Brand & Li, 23; Ge & Qiu, 27; Li e al., 28; Cull e al., 29; Song e al., 212). Because of incenives for promoion, local governmen officials may force he banks o provide more loans o sae-owned enerprises, which may bring fiscal revenue Gordon & Li, 212). Recen paper of Cull e al. 213) conribue o his opic by empirically esing he relaionships among invesmen behaviour of saeowned enerprises, poliical connecions and financial consrain, proving ha financial misallocaion in China indeed resuls from poliical connecions. Numerous researchers implemen pleniful research on banking discriminaion and is influence on financial misallocaion and TFP losses Midrigan & Xu, 21; Song e al., 212; Gilchris e al., 213; Carvalho, 214; Moll, 214), however, hey rarely ake he influence on business cycles ino accoun. Recen work of Le e al. 214) conribues o he issue by consrucing a DSGE model of China, and proposing ha he longer erm implicaions of focussing lending on sae firms could be harmful o China s economic growh. Bu 3

4 hey do no characerize he wo ypes of firms ino he model, and do no explain he implicaions of discriminaory credi consrain on economy s flucuaion. In his paper, based on GNSS s work, we build a dynamic sophisic general equilibrium model, where he banks can decide he loan-o-value raio, aking boh saeowned enerprises henceforh, SOEs) and privae enerprises henceforh, PEs) ino accoun, o analyze he impac of discriminaory credi consrain on business cycles 3. The mos specific par of our model is o allow he banks and sae-owned enerprises o decide loan-o-value raios when solving he opimizaion problems. We disinguish performing loans and non-performing loans, which are no known unil he loan becomes maure. Because loan-o-value raios are depended on banks behavior of resising risk, we inroduce a Cosly Sae Verificaion henceforh, CSV) cos, following BGG, o evolve he mechanism of he loan o value raio endogenously. This model describes he poliical connecions among he governmen, sae-owned enerprises and he banks by inroducing a subsidy ino he banks profi funcion. Thus, we can derive he differen credi consrains ha differen ypes of enerprises confron, under he assumpion ha loan ineres raio is he same for boh sae-owned enerprises and privae enerprises according o non-arbirage condiions. We parameerize he model according o exising lieraure and he daa of China s economy. Subsequenly, hrough several experimens designed o illusrae how he model behaves, we compare he ransmission mechanism of a negaive echnology shock under discriminaory credi consrain and non-discriminaory credi consrain respecively. We find he oal oupu and oal loan o enerprises have experienced a more significan decline due o discriminaory credi consrain. Moreover, according o he impulse-responses of an expansionary moneary shock, we find ha he oupu and oal loan increase less in he model of discriminaory credi consrain han hose in he model of non-discriminaory credi consrain. These resuls reveal ha discriminaory credi consrain could amplify he echnology shock and reduce he effeciveness of moneary policy. For robusness, we furher empirically examine he impac of discriminaory credi consrain on he relaionship beween moneary policies, echnology improvemen and economic growh. The empirical evidence from China s indusrial level daa has suppored our conclusion from he model. This paper conribues o he lieraure by exending DSGE models on credi consrain and business cycle. We combine he financial fricion of boh demand side and supply side, wih credi consrain depending on boh he firms financial posiion and he banks siuaion. The model can capure no only he effec of real economy on business cycles, bu also he impac of financial facors on economy s flucuaion. Moreover, his model can characerize risk-resising behavior of banks wih he loan-o-value raios henceforh, LTV raios) endogenously evolving. As o he realiy, his paper 3 To be clear, he model proposed by his paper may also be used o analysis he poliical connecions and bank discriminaion in oher counries. 4

5 proposes a new explanaion of he recen decline of China s economic growh. Our conclusion suppors he furher deep reforms of markeizaion in China, he amelioraion of financial sysem and he marke-orienaion ransformaion of financial archiecure. Finally, his paper indirecly inroduces poliical facors ino DSGE model, filling he gap of exising lieraure o some exen. The remainder of his paper is organized as follows: Secion 2 inroduces he full DSGE model, while Secion 3 describes he calibraion of he model. Secions 4 and 5 presen he impulse response analysis and empirical analysis respecively. Secion 6 comprises concluding remarks. 2 Model Seup In his secion, we exend he GNSS model by disinguishing wo ypes of enerprise: sae-owned enerprises and privae enerprises. We aemp o describe he discriminaory credi consrain confroned by sae-owned firms and privae firms, and simmulae is impacs on he economy when exogenous shocks occur. The whole economy incorporaes paien and impaien households, wholesale banks, bank lending and bank deposi branches, SOEs and PEs, reailers, capial goods producers and a cenral bank. All he LTV raios are decidable for wholesale banks, who also underake he defaul cos CSV cos). Due o srong relaionship wih governmen, SOEs may parly decide LTV oo. As o LTV raio for impaien households, we assume i consan for simplificaion, while ha for enerprises is ime-varied. Considering ha he banks in China will provide less sric loan condiions o SOEs because of heir srong governmen background, we assume ha he governmen will pay subsidies o wholesale banks, which are direcly conneced o he loan amoun supporing SOEs. 2.1 Households Paien Households Following GNSS, he represenaive of paien households maximizes is whole-life expeced uiliies: E = β P 1 a P ) log c P a P c P 1) + log h P n P,A ) 1+φ A 1 + φ A n P,B ) 1+φ B 1 + φ B 1) where, c P is he consumpion of each paien household, c P is he observable aggregae consumpion of paien household in las period, a P is he parameer o measure he consumpion habi, h P is housing service, n P,A and n P,B are he hours worked in he sae-owned firms Firm A) and privae firms Firm B) respecively, which have inverse 5

6 Fischer elasiciy φ A and φ B. The represenaive paien household earns a wage wih a real wage rae w P,A of sae-owned firms and w P,B of privae firms o suppor heir consumpion and accumulaion of housing h P wih housing price q h. The budge consrain is: c P + d P + q h h P w P,A n P,A ) + w P,B n P,B + dp r d 1 + P 2) π where d P is he deposi, r d 1 is he ne deposi ineres rae of 1 period, and P is lump-sum axes and dividends Impaien Households Similarly, he impaien households also choose he labour supply, consumpion and housing invesmen o maximize heir expeced uiliies as: E = β I and subjec o: 1 a I) log c I a I c I 1) + log h I ) c I + bi r bh 1 π n I,A + q h h I w I,A n I,A ) 1+φ A 1 + φ A n I,B ) 1+φ B 1 + φ B 3) + w I,B n I,B + b I 4) where b I is he money borrowed from banks wih a ne loan ineres r bh. The households should provide heir housing as collaeral, so hey confron a borrowing consrain se by banks b I 1 + r bh ) m H E [q h +1h I π +1 ] Tha means he oal amoun of loan and ineres should be less han he expeced value of housing muliplied by loan o value raio m H. 5) 2.2 Enerprises In his model, we incorporae wo ypes of enerprise, sae-owned enerprises and privae enerprises, indexed as s = A, B respecively. We inend o describe he differen efficiency of sae-owned and privae enerprises, and o analyse he differen financial fricion hey confron. In common wih mos sandard se-ups, he represenaive enerprise selecs he labour, capial, consumpion and loans o maximize is whole-life 6

7 uiliy of consumpion c s. E = βs 1 a s ) log c s a I c s ) 1 6) where, β s is he subjecive discoun facor of enerprise s, c s is he observable aggregae consumpion in las period, and a s is he consumpion habi parameer Sae-owned Enerprises The producion funcion of sae-owned enerprises is like Cobb-Douglas echnology y A = z u A k A 1) αa n A ) 1 αa 7) where, k A 1 is he capial, z is echnology shocks, and u is ime-varying uilizaion rae of capial See Schmi-Grohé & Uribe, 26), which leads o a cos: ψ u A ) k A 1 = k 1 A κ u 1 2 u A ū A ) 2 + κ u 2 u A ū A ) ) where κ u 1 and κu 2 are posiive parameers. Uilizaion rae of capial reflecs he capaciy. If he uilizaion rae decreases, he firm can be regarded as having excess capaciy. The labour n A is aggregaed from paien labour and impaien labour by C-D echnology wih parameer µ n : n A = n I,A ) µ n n P,A ) 1 µ n The sae-owned enerprises earn revenue by selling heir producs. 8) Their main expendiure consiss of purchasing capial goods o inves, paying wages and heir own consumpion. They can borrow money from banks o help hem sar he business and smooh he consumpion. Then we can wrie he budge consrain as follows: y A x A + b A + q k 1 δ) k A 1 = q k k A + ψ u A ) k A 1 + w P,A n P,A +c A + ba 1 + w I,A n I,A 1 + r be 1 ) π 9) where, δ is he depreciaion rae of capial, q k and 1/x A are he relaive price of saeowned enerprise-made producs and capial goods compared o consumpion goods price, b A w P,A is he amoun of borrowed money, r be is he ne loan ineres for enerprises, and w I,A are real wage raes for paien households and impaien households respecively. In order o inroduce he endogenous law of evoluion for loan o value raio, we consruc a scenario o show he loan defaul and informaion asymmery. We assume 7

8 he projec has probabiliy o succeed of 1 η, and hen he enerprise will reurn he oal loan payable o banks including he ineres. However, here is η probabiliy ha he projec may fail. If he projec fails, boh he enerprise and he banks will incur a cos. The enerprises are no willing o reurn he oal deb, so loan defaul occurs. The bank will pay a cos for Cosly Sae Verificaion CSV, menioned in BGG, 1999), which is assumed o be posiively relaed o loan o value raio and he oal loan and ineres payable m A 1 /2κ w muliplied by oal deb). For enerprises, he loan payable is diminished, so hey only reurn 1 m A 1 /2κ w o he bank. This is a hreshold value: if he reurn rae is less han he hreshold, he bank will pay CSV cos o hire a person o recoup he loan payable. However, he enerprise should also incur he oal cos of projec failure. We assume he projec failure cos is κ f m A 1 qk k 1 A, which is posiively relaed o he loan o value raio and capial. This is because when he LTV raio is relaively higher, he enerprise will have higher leverage and will be moivaed o inves in more risky projecs, which will lead o more loss when he projec fails. Then we can aggregae he budge consrain of enerprises under wo differen condiions by weigh of heir probabiliies, 1 η and η respecively: y A x A + b A + q k 1 δ) k A 1 = q k k A + ψ u A ) k A 1 + w P,A n P,A ) ) + ba r be 1 1 η ma 1 π 2κ w + w I,A n I,A + c A + κ f m A 1q k k A 1 1) The firm also confrons borrowing consrain se by banks b A 1 + r be ) [ ] m A E q+1k k A π +1 1 δ) 11) In our model, we assume he loan o value raio is endogenous, which means he loan o value raio depends on he negoiaion of banks and enerprises. However, sae-owned enerprises have more power in he loan marke, so we assume ha only sae-owned enerprises and banks can make a decision abou loan o value raio Privae Enerprises Following Song e al. 211) we assume he privae enerprises can hire managers wih remuneraion w m o improve heir labor efficiency by χ. Then he producion funcion of privae firm is: y B = z 1 ψ) u B k 1) B αb ) χn B 1 αb 12) Meanwhile, o preven he manager divering funds from he company o heir own benefi, he remuneraion saisfies an incenive consrain w m > ψy B, where ψ is he 8

9 parameer saisfying an assumpion χ > 1 ψ) 1 1 α B proved by Song e al. 211) o keep he firm willing o hire a manager. Similarly o sae-owned enerprises, we can wrie he budge consrain of privae enerprises as y B x B + b A + q k 1 δ) k B 1 = q k k B + ψ u A ) k B 1 + w P,B n P,B ) ) + bb r be 1 1 η mb 1 π 2κ w + w I,B n I,B + c B + κ f m B 1q k k B Banking Secor Following Gerali e al. 21), he bank secor in our model is also spli ino hree pars: a wholesale bank and wo reailer branches. Unlike Gerali e al. 21), however, we assume he banks know he collaeral provided by enerprises, and will choose LTV raio o ensure ha he borrowing consrain of enerprises is always binding Wholesale Branch Following GNSS, each wholesale branch operaes under perfec compeiion. On he liabiliies side hey will combine heir own bank capial k w wih deposi d ransferred from deposi reail branches on he liabiliies side. On he asses side, hey will issue loans b I o impaien households, and loans b A and b B o sae-own and privae enerprises respecively. The law of evoluion for bank capial is π k w = 1 δ b) k 1 w +jbank 1, where δ b measures resources used o manage he bank capial, j 1 bank is profi of oal banks. According o he analysis on he scenario of loan defaul and informaion asymmery, banks will incur a CSV cos. We assume ha all of his cos is underaken by wholesale branches, and hen hey will choose deposi and loans o maximize heir expeced profi as follows: max r we m s b A + b B,bs,bI ) + r wh b I r d κ kb 2 η 2κ w m A + m B ) 1 + r be k w b ) 2 ν b ) + T b A where, r we, r wh are wholesale loan rae for enerprises and impaien households respecively, k w and b are bank capial and oal loan, ν b is he capial acquiremen raio se by he regulaor, κ kb and kappa w are parameers, T b A ) is he subsidy from governmen 4 for loans o sae-owned firms, T b A ) κ = T 1 2 b A b A) 2 / ba +κ T 2 ba, where κ T 1 and κt 2 are posiive parameers. This subsidy is o reflec he igh relaions beween banks and sae-owned enerprises. As we know, he majoriy of banks in China are 4 When closing he model, he cos of his subsidy is finally borne by paien households. ) 9

10 really conrolled by governmen, which includes boh cenral governmen and local governmen. Then he leaders and shareholders of sae-owned firms may influence he banks o reduce he loan condiions and requiremens. Moreover, moivaed by he pursui of promoion, officials of local governmen coninuously improve heir poliical performance by exering pressure on banks and sae-owned firms for GDP growh. Furhermore, when he loans of sae-owned firms became non-performing, he governmen helps he banks o srip away bad asses. These phenomena reflec he srong relaions beween banks and sae-owned enerprises, so we use he subsidy o reveal his relaionship, which is also he main driver of discriminaory credi consrain beween sae-owned and privae firms. Following GNSS, we assume he deposi rae faced by he wholesale banks is equal o Taylor Rule rae r, which is paid o he deposi reailer. The balance shee consrain of wholesale branches is b A + b B + b I = b = k w + d. In addiion, hey choose he loan o value raio under he condiion m s q+1 k ks π δ) = 1 + r be ) b s. Afer some algebra wih regard o he firs order condiions, we can derive he ineres spread beween wholesale loan raes and Taylor Rule rae: r wh = r κ kb k w b ) ) k ν b w 2 13) b Similar o GNSS, he ineres spread comes from he cos of capial acquiremen. Furhermore, we can ge he ineres spread beween wholesale loan raes for impaien households and privae enerprises: r we = r wh + ηm B 1 + r be κ w 14) This spread reflecs he cos of probabiliy of loan defaul or CSV cos. Meanwhile, we also can derive a similar equaion from he firs order condiion for loan of sae-owned enerprises. Combining hem, we can derive he relaionship of loan o value raios for wo ypes of firms: ) κw m A m B = T b A η ) 1 + r be 15) I is obvious ha he exen of discriminaory credi consrain is posiively relaed o he marginal subsidy rae o banks for SOE loans, and negaively relaed o he loan ineres rae of enerprises Deposi Reailer Branch The deposi reailer branches operae under monopoly compeiion. They collec deposis from paien households wih deposi raes r d i) and ransfer he deposis o 1

11 he wholesale branch wih ineres rae r, equal o he Taylor Rule rae. According o Beneš and Lees 27), he oal deposi marke each reailer confrons is d = [ 1 d i) 1/ε d] εd, where εd is he elasiciy of subsiuion of deposi. Afer cos minimizaion, we can derive he deposi demand of each reailer as d i) = [ r d i) /r d ] εd d. We assume he deposi reailers have a quadric adjusmen cos when adjusing he deposi rae; hen we can wrie he profi funcion of deposi reailer branch as: max r d i) E = Λ P, r d i) r d i) d i) κ d 2 ) 2 r d i) r 1 d i) 1 r d d 16) where, Λ P, is random discoun facor of paien households, and κ d is parameer of sicky exen. equilibrium, is: 1 + ε d ε d r r d κ d The firs order condiion for r d i) yield, afer imposing symmeric r d r d 1 1 ) r d r d 1 [ ) λ P + β P E +1 r d λ P κ +1 r d d r d 1 +1 d +1 r d d In he seady sae he deposi ineres is r d = ε d ε d 1 r. Because ε d <, he deposi rae is marked down o Taylor rule rae. ] = 17) Loan Reailer Branch Similar o deposi reailer branches, we assume he loan reailer branches operae under monopoly compeiion. The loan demands of each loan reailer [ branch ] for enerprises and impaien households indexed by j = E, H ) are b j i) = r bj i) /r bj εbj b j, where ε bj is he elasiciy of subsiuion of loan demand. Loan reailer branches selec loan ineres rae o maximize heir profi as follows: max r bj i) E = Λ P, r be +r bh i) b E i) r we b E i) κ be 2 i) b H i) r wh b E i) κ bh 2 i) r 1 be i) 1 r bh i) r 1 bh i) 1 r be ) 2 r be ) 2 r bh where, Λ P, is random discoun facor of paien households, κ be and κ bh are parameer of sicky exen. The firs order condiions for r be i) and r bh b H b E 18) i) yield, afer imposing 11

12 symmeric equilibrium, 1 ε bj + ε bj r wj r bj κ d r bj r bj 1 1 ) r bj r bj 1 + β P E [ λ P +1 λ P r bj +1 κ bj r bj 1 ) r bj In he seady sae he loan ineres is r bj = ε bj ε bj 1 rwj. Because ε bj >, he loan ineres rae is marked up o wholesale loan ineres rae. Then, we can wrie he oal profi of banking secor as +1 r bj b j +1 b j ] = 19) j bank = r be b A + b B ) + r bh b I r d d κ kb 2 η 2κ w m A + m B ) 1 + r be k w b ) 2 ν b ) + T b A ) 2) 2.4 The Res of he Economy Capial Goods Producer Following Iacoviello 25) and GNSS, we assume he capial goods producer produces capial goods wih a quadric adjusmen cos, and hen maximizes profi as follows: max i i) E = Λ P, [ q k [ 1 κ I 2 ) ] ] 2 i 1 i i i 1 where, i is invesmen, Λ P, is random discoun facor of paien households, κ I is he adjusmen cos parameer. In our model, boh sae-owned and privae enerprises use homogenous capial, so he capial can be freely raded, which means all capial goods have a unique capial goods price q k. The firs order condiion is: 1 q k ) i i = 1 κ I 1 κ ) [ 2 I i λ P ) ] 1 + β P κ I E +1 i+1 i+1 q+1 k i 1 i 1 2 i 1 λ P 1 i i q k And he aggregaed capial is k A +k B 1 δ) [ k 1 A + ) ) ] 2 kb 1 = 1 κ I i 2 i 1 1 i Reailer In our model, he reailers combine he differen producs of wo ypes of enerprises, by C-D echnology y = y A ) µ y ) y B 1 µ y where, µ y is he weigh of sae-owned enerprises made producs in oal final goods. By maximizaion of profi under cos consrain, we can derive 21) µ y 1 µ y = ya y B x B x A 22) 12

13 We inroduce he sicky price as Calvo, 1983). Only γ of reailer may change he price in each period, and hen we can derive he New Keynes Philips Curve, log π π = β P log π +1 π 1 γ) 1 βγ) [ ] µ y log xa γ x A + 1 µy ) log xb x B where, x is seady sae of he relaive price of he final goods o inermediae goods ha is equal o he mark-up rae. 23) Marke clear condiion and Cenral bank To close he model, we give he marke clear condiion for final goods as y = q k [ 1 κ I 2 ) ] 2 i 1 i + [ c A + c B + c H + c P ] δ b k + 1 w + Adj 24) i 1 π As o housing marke, we assume he housing held by all households is equal o an exogenous consan h, ha is, h I + h P Taylor rule, = h. The cenral bank follows he sandard log 1 + r r = ρ log 1 + r 1 r [ + 1 ρ) φ y log y y 1 + φ π log π ] + ε MP 25) π where, ρ measures he coninuiy of moneary policy, φ y and φ y are he weighs assigned o inflaion and oupu sabilizaion respecively, and r is seady sae value of ineres rae. 3 Calibraion In he previous secion, we buil he full DSGE model wih discriminaory credi consrain. In his secion, we calibrae he srucural parameers according o relaed lieraure and China s daa, in order o analyze he impulse-response figures of he main economic variables under echnical shock and moneary policy shock. 3.1 Srucure Parameers Affecing Seady Sae We parameerize he model by season, according o China s quarerly daa covering he period from 24Q1 o 214Q1. Table 1 shows he calibraed parameers which influence he seady sae of he model. We selec average Chinese average deposi benchmark ineres rae as seady sae value of deposi ineres rae. According o he seady sae deposi ineres rae, we calibrae he objecive discoun facor of paien households as.9926, which is close o he value calibraed by Chen e al. 212). 13

14 For he discoun facors of impaien households and of sae-owned enerprises, we follow GNSS o calibrae hem as.975. Considering ha privae enerprises are more moivaed o enlarge producion and borrow money, we calibrae heir discoun facor as.97, which means hey are more impaien han sae-owned enerprises. [TABLE 1 ABOUT HERE] We calibrae he seady sae of LTV according o Chinese indusrial firm daa from Naional Bureau of Saisics of China. We use he oal deb minus amoun payable o calculae he long-erm deb, and hen use he long-erm deb divided by oal fixed asses o calibrae LTV raio. According o he daa, we calibrae he LTV raio of sae-owned enerprises as.5, while he LTV raio of privae enerprises is only.46. As he daa of household loan is no available, we calibrae he LTV raio of impaien households as.7, he same as GNSS, which is in line wih Chinese policy ha buyers should pay a leas 3% of housing price as a haircu. According o he Commercial Bank Repor of he People s Bank of China, he non-performing loan rae is around 1%, so we calibrae he seady sae loan defaul rae as equal o.1. Then we can calculae he parameer of defaul cos for firm κ f as We use Chinese average loan benchmark ineres rae as seady sae value of loan ineres rae o firms, and choose 7-day Shanghai iner-bank offered rae SHIBOR) as he proxy of seady sae value of Taylor Rule rae. There is an imporan assumpion ha he seady sae value of banks leverage raio is equal o he requiremen of.8 se by he cenral bank similar o GNSS). According o hese ineres raes, we can calibrae he parameer of CSV cos κ w as 1, and we calibrae he subsiuion elasiciy of deposi marke and loan marke for enerprises in order o mach he seady sae LTV raio. Wih regard o he subsiuion elasiciy of he loan marke for households, we follow GNSS o calibrae i as The depreciaion rae of capial is calibraed as normal,.25 Guo e al., 215), which means 1% per year. Values for he capial share in producion funcion vary in he lieraure, wih Chinese scholars calibraing i in a range beween.4 and.5 Guo e al., 215; Chen e al., 212). To describe he characerisic of invesmenled growh of SOEs, we calibrae he capial share of SOEs and PEs as.5 and.33 respecively. Following Gerler and Karadi 211) and GNSS, we calibrae he seady sae of uilizaion of capial as 1, and calibrae he parameer of adjusmen cos for capial uilizaion as κ u 1 =.478 and κu 2 =.478. Following he majoriy of sudies, we calibrae he seady sae mark-up rae in he goods marke as 1.2, which maches he subsiuion elasiciy of 6. Considering ha he privae economy is growing rapidly and already occupies more han 6% of he GDP Szamosszegi e al., 211), we calibrae he weigh of SOE-produced goods as.4. Following GNSS, he share of impaien households is calibraed as.2. 14

15 We calculae he managemen cos of bank capial as.865. The subsidy o banks for SOE loans is hard o calibrae. The seady sae of marginal subsidy rae κ T 2 can be calculaed as.4. As o he second order marginal subsidy rae κ T 1, we calibrae i as.2, and will es is range in he furher analysis. 3.2 Oher Srucure Parameers Table 2 repors he calibraion values of oher parameers, which do no affec seady sae. We calibrae he parameer of sicky price γ as.75, which means he enerprise has one chance o change heir price a year. The parameers of adjusmen cos are calibraed according o GNSS poserior mean values. This is for wo reasons. Firs, Chinese financial archiecure is bank-oriened, as is he euro area, so heir banking sysems are similar o some exen. Secondly, our model is based on he GNSS model, so he parameers may be more suiable o show he mechanism of financial fricions. Wih regard o he coefficien of Taylor rule, we calibrae he coefficien on inflaion rae φ π as 1.5, while he coefficien on oupu is.125 Chen e al., 212; Rannenberg, 213). Following GNSS, we calibrae he coninuiy parameer of moneary policy as.75. Abou consumpion habi, we calibrae i as.5, according o he esimaion resul by Ma 214); Ma and Li 215). [TABLE 2 ABOUT HERE] 4 Simulaion and Discussion In his secion, we analyse he role of discriminaory credi consrain based on he impulse responses of echnology shocks and moneary policy shocks. By comparing resuls from he model wih discriminaory credi consrain wih hose from a model where he wo ypes of enerprise have he same LTV raio, we find he impac of discriminaory credi consrain on he economy. We also explain he mechanism whereby i amplifies he echnology shocks and hinders he effeciveness of moneary policy. Finally, we focus on he source of discriminaory credi consrain. When enlarging he parameer value of marginal subsidy rae for SOE loans, we find ha he impac of discriminaory credi consrain decreases. 4.1 Technology Shock Figure 2 describes he response flucuaions of he main macro-economic variables afer one uni of negaive echnology shock. As a resul of echnology decrease, boh he producion cos and he produc price increase, leading o a decline in consumpion. While being aware ha he echnology shock is emporary, represenaive paien 15

16 households may increase heir curren consumpion and diminish heir deposi, causing he deposi supply o decrease. For impaien households and enerprises, heir loan demand increases. Impaien households would like o borrow more money o mainain heir consumpion, while enerprises would like o borrow money o smooh he influence of negaive echnology shock. However, limied by equilibrium, he oal loan decreases as deposi supply decreases, herefore causing he ineres rae o increase. These resuls are in line wih exising lieraure. [FIGURE 2 ABOUT HERE] In our model, endogenous LTV raios and differen reamen of SOEs and PEs are described clearly. From Figure 2 we can see ha under he siuaion of discriminaory credi consrain, loan o SOE grows slighly, while loan o PE falls significanly. This may be inerpreed according o he differen LTV raios. Boh he LTV of SOEs and ha of PEs is decreasing, which reveals ha he LTV of SOEs is couner-cyclical, while ha of PEs is pro-cyclical. SOEs have more power o influence he bank because of heir srong background, so he credi consrain hey confron is less binding. However, PEs can only accep he loan condiion proposed by banks. As a consequence, more loan resources flow o SOEs. Due o China?s special characerisics he financial marke is no perfec, leading o banks unfair reamen owards differen ypes of enerprise. In China, cenral and local governmen force banks o suppor SOEs wih low requiremens for collaeral, and may even vouch direcly for he SOEs so ha hey receive loans. These facors ensure ha he loan direcion is biased oward SOEs. Under he condiions of negaive echnology shock, discriminaory credi consrain can enlarge he flucuaion of oupu. By changing he parameers o make he credi consrain of boh ypes of enerprise he same, eliminaing he subsidy o banks for SOE loans, we can ge anoher series of impulse responses wih no discriminaion. From Figure 2, we can see ha he growh of oupu wih no discriminaory credi consrain is slighly higher. This means ha he financial acceleraor effec KM ype) is enlarged when he credi consrain is discriminaory. When he economy deerioraes, banks end o lower LTV raio o conrol he CSV cos. Due o he subsidy for SOE loans, banks only lower he LTV raio of PEs, bu mainain or increase he LTV raio of SOEs. The relaively lower LTV raio will induce a sharp decrease in capial demand. Because SOEs and PEs use he same capial for invesmen, he capial prices are influenced by capial demand of boh ypes of enerprise. As he figure shows, compared o he model wih no discriminaion, he decrease of LTV for PEs has a greaer influence on he capial price, making i decrease furher. This is due o resource misallocaion. A he beginning, despie of he differen LTV raios, PEs and SOEs have almos he same produciviy and here is no misallocaion. However, he borrowing consrains wih differen binding exen lead o differen operaion siuaion of wo ypes of firms. Wih more binding borrowing consrain, PEs 16

17 may only decrease loan and invesmen, or even fire sale he capial, which induce o supply s furher decline. Correspondingly, he supply of SOEs is less decreasing. Due o subsiuion elasiciy of demand beween wo ypes of firms, he demand srucure canno change inermediaely. Thus, SOEs profiabiliy and capial reurn become lower han PE s. Coniguous enlargemen of he loan gap beween SOE and PE leads o more severe misallocaion, herefore causing aggregae capial price s furher decline. Furher decline of capial price makes boh he borrowing consrains of PEs and SOEs more binding, and causing aggregae oupu s decline amplified. As a consequence, discriminaory credi consrain se by banks sharpens he decrease of capial price, and hen he oal collaeral is less valuable, while he oal loan of enerprises declines furher, hus leading o more decline of oal oupu. Furhermore, discriminaory credi consrain may worsen he economic srucure and sharpen he excess capaciy of SOEs. From Figure 2, we can see ha he capial of SOEs increases, while ha of PEs decreases, and he capial uilizaion of SOEs decreases faser han ha of PEs. This reflecs he change of economic srucure and greaer excess capaciy of SOEs. As we know, SOEs in China are less efficien, so heir oupu growh relies on invesmen. Coninuous invesmen improves SOEs? capaciy of producion, bu he demand canno saisfy he supply. A he beginning of he 21s cenury, benefiing from he rend of globalizaion, China s expors increased rapidly, which suppored he excess capaciy of manufacuring. However, afer he global crisis of 28, exernal demand decreased sharply, and he problem of excess capaciy became more imporan. Alhough our model canno describe he resuls of economic srucure imbalance and excess capaciy, we know ha hey are disincenives o economic developmen. 4.2 Moneary Policy Shock Figure 3 repors he impulse responses of he main macro-economic variables under a sandard deviaion of expansionary moneary policy shock. As a resul of expansionary moneary policy, he deposi ineres rae decreases, and hen households decrease deposis o increase curren consumpion. The growh of aggregae demand simulaed by expansionary moneary policy causes oupu o grow. For banks, decreasing policy ineres rae makes hem lower loan ineres raes, which leads o growh of oal loan. These resuls are similar o hose in he relaed lieraure. [FIGURE 3 ABOUT HERE] We pay furher aenion o he loans o differen ypes of enerprise. Compared wih he model wihou discriminaion, LTV raio of SOEs decreases slighly, bu ha of PE decreases more significanly. More loans flow o SOEs raher han PEs. Due o he subsidy for SOEs, banks se relaively higher LTV for SOEs, bu lower LTV 17

18 raio for PEs, in order o lend more money o SOEs and improve heir profi. Wih a similar mechanism o he siuaion under negaive echnology shocks, decreasing PEs LTV has a dramaic influence on he capial price, as he figure shows. Toal loan o enerprises decreases more sharply, herefore leading o lower effeciveness of expansionary moneary policy when simulaing oupu growh. Since expansionary moneary policy canno promoe economic growh effecively, growing liquidiy canno flow o he real economy, and hen i flows o he housing marke. Compared o he LTV raios for enerprises, LTV raio for impaien households is higher and more fixed, as household loan is less risky. When he real economy canno absorb he liquidiy released by cenral banks, i is reasonable ha he liquidiy flows o households. As a consequence, housing and real esae become more valuable due o heir collaeral value, leading o growing housing demand and a prosperous housing marke. This is in line wih China s housing marke boom since 29. Furhermore, in he model wih discriminaory credi consrain SOEs confron he more serious problem of excess producion capaciy. This has he same mechanism as ha under negaive echnology shocks. Banks unfair reamen o differen ypes of enerprise leads o differen capial invesmen decisions of SOEs and PEs. Having obained financial suppor, SOEs implemen more capial asses invesmen. However, unsusainable growh of aggregae demand canno suppor fas growing supply, hus causing excess producion capaciy in SOEs. This is also in line wih he acual siuaion since Furher Analysis on Amplificaion Effec of Discriminaory Credi Consrain According o he above analysis on impulse response figures, we find ha discriminaory credi consrain has an amplificaion effec on he ransmission of negaive echnology shocks. In order o make clear he source of amplificaion effec and how i changes according o he exen of discriminaion, we furher examine he impulse response figures under differen parameers of subsidy. As he firs order marginal subsidy rae κ T 2 is depends on he seady sae of LTV raios of SOEs and PEs, we can adjus only he second order marginal subsidy rae κ T 1 o change he exen of discriminaion. As we can see from Figure 8, higher subsidy rae leads o relaively smaller amplificaion effec. This is due o he double effec of subsidy. In fac, he subsidy for SOE loan has boh income effec and subsiuion effec on credi consrain and oal loans o enerprises. On one hand, when he governmen improves he marginal subsidy rae, he banks may earn more income, and hen hey will properly lower he ineres rae under negaive echnology shocks) and increase loan supply o maximize heir profi. This is income effec. On he oher hand, when he governmen improves he marginal subsidy rae, he banks are willing o lend more 18

19 o SOEs raher han PEs, and hen hey will se higher LTV for SOEs and lower LTV for PEs. Resource misallocaion makes he capial price decrease more sharply, leading o decline of oal loan o enerprises. This is subsiuion effec. Wih regard o he oal loan o enerprises, hese wo effecs play opposie roles. Generally, he subsiuion effec is more significan han he income effec, so we find he amplificaion effec of discriminaory credi consrain. However, when we furher improve he marginal subsidy rae, he subsiuion effec changes less, bu income effec increases o some exen; herefore, he amplificaion effec is slighly weakened. We can also find his mechanism hrough he equaion of he gap beween LTV raios. Under he assumpion of he same loan defaul rae for all ypes of enerprises, from Eq.15) we can have he expression of gap of LTV raios: m A = m B + κ w κ T 1 b A / b A 1 ) + κ T ) 2 η ) 1 + r be 26) Where b A is SOE loan, κ T 1 and κt 2 are parameers of subsidy for SOE loan, rbe is he loan rae o enerprises. I is obvious ha he gap of LTV raio is posiively relaed wih he wo parameers of subsidy for SOE loan, and negaively relaed wih loan ineres. When we increase he κ T 1, under he condiion of increasing ba, he numeraor of he fracion a he righ hand side of Eq.26) also increases. However, he income effec of increasing subsidy lowers he loan ineres r b E, decreasing denominaor of he fracion, so we canno judge he value of fracion increases or decreases. Through he figure, we know generally he gap of LTV raios is posiive relaed o he subsidy parameer κ T 1, bu he marginal effec is negaive relaed o he parameer. [FIGURE 4 ABOUT HERE] 5 Empirical Evidence from China Finally, we analyse in deail he impac of discriminaory credi consrain on ransmission of echnology shocks and moneary shocks. In his secion, we empirically es he robusness of he model resul using daa of Chinese indusrial firms. We focus our esing on wo conclusions from he model analysis: firs, ha he gap beween LTV raios of SOEs and PEs will be enlarged by expansionary moneary policy; secondly, ha loans flow more o SOEs han o PEs under expansionary moneary policy shocks, due o discriminaory credi consrain. 19

20 5.1 Regression model Firs, we build a benchmark regression model o analyse he facors influencing he economic growh: IAV i, = β + β 1 T F P i, + β 2 In + β 3 DCC i, + β 4 N i, + µ i + ε i, 27) where, IAV i, is logarihm of indusrial added value of indusry i in he year, as he proxy of economic growh, T F P i, is logarihms of TFP of indusry i in he year, measuring he echnology level improvemen, In is real ineres rae of year, measuring he moneary policy, is he raios of average SOE LTV raio o average PE LTV raio in indusry i in he year, measuring he discriminaion exen of credi consrain, N i, is logarihm of employee numbers of indusry i in he year, measure he labor 5, µ i and ε i, are fix effec and residual error respecively. We esimae TFP using Solow Residual Mehod Barro, 1999; Felipe, 1999). We firs esimae he capial share of producion funcion, under he assumpion of consan reurns o scales by 28), and hen calculae TFP by equaion 29). The wo equaions are as follows: IAV on i, = β + αkon i, + ε i, 28) T F P i, = IAV i, ˆαK i, 1 ˆα) N i, 29) where IAV on i, is he raio of logarihm of indusrial added value o logarihm of employee numbers, KoN i, is he raio of logarihm of capial asses o logarihm of employee numbers, α is he capial share coefficien, ˆα is is esimaed value. According o he conclusion above on discriminaory credi consrain drawn by impulse response figures, following Fisman and Love 23) we can build he regression model as follows o examine he moderae effec of discriminaory credi consrain: IAV i, = β + β 1 T F P i, + β 2 In + β 3 DCC i, +β 4 DCC i, T F P i, ) β 5 N i, + µ i + ε i, 3) where, DCC i, T F P i, ) represens produc iems of TFP and he discriminaion exen of credi consrain. Because he improvemen of TFP may promoe economic growh, so β 3 should be significanly posiive. If coefficien β 4 is also significanly posiive, i means he higher discriminaion exen may amplify he influence of TFP on indusrial added value growh. Similarly, in order o es moderae effec of discriminaion exen on he effecive- 5 We use capial and labour o esimaed TFP by Solow Residual Mehod. Due o he co-lineariy problem, we selec only wo of he hree variables included in he regression model. 2

21 ness of moneary policy, we build anoher regression model wih a produc erm as follows: IAV i, = β + β 1 T F P i, + β 2 In + β 3 DCC i, +β 4 DCC i, In ) β 5 N i, + µ i + ε i, 31) where, DCC i, In ) represens produc iems of In and he discriminaion exen of credi consrain. Because he indusrial added value is negaively relaed o ineres rae, if coefficien β 4 is significanly posiive, i means he higher discriminaion exen may hinder he influence of moneary policy on indusrial added value growh. 5.2 Daa Source and Descripion The daa se we use come from he China Annual Survey of Indusrial Firms from 1999 o 28. Firs, we sor he firms according o wheher hey belong o SOE or PE, and eliminae hose firms wih fewer han 8 employees in order o ensure he firms exisence and scale. Secondly, o mainain he effeciveness of LTV raio, we eliminae he firms wih long-erm deb and negaive capial asses, because we use he raio of long-erm deb o capial asses as he proxy of LTV raio. Moreover, we exclude he firms whose LTV raio is bigger han 3, o ensure a reasonable and suiable scale of firms. Following his procedure, we have a sample wih 397,69 observaions. Thirdly, for he purpose of calculaing he discriminaion exen of credi consrain, we aggregae he firms daa o indusrial-level daa. In he China Annual Survey of Indusrial Firms, all firms are classified ino 39 indusries, such as Coal Mining, Oil and Gas, and Texiles. Taking all he facors ino accoun, we obain an indusrial-level sample of 335 observaions of 39 indusries from 1999 o Finally, we esimae he TFP of each indusry in each year. As shown in he firs column in Table 5, he capial share is significan a 5% level, wih a value of.79. I is marginally bigger han oher scholars esimaed resuls, because for our model we selec indusrial firms, which have a relaively higher capial share han he agriculure and public services indusries. Then we calculae he TFP hrough Eq.29). The measuremen and descripive saisic of variables are repored in Table 3 and Table 4 respecively. The correlaion analysis is no repored, and all he correlaion coefficiens are smaller han he co-lineariy hreshold of.7 Mason e al., 199). [TABLE 3 ABOUT HERE] [TABLE 4 ABOUT HERE] 6 Due o he daa availabiliy, we canno obain he daa covering he years from 28 o 214. However, since discriminaory credi consrain is long-sanding in China, he daa from 1999 o 28 can be used o examine is impac. 21

22 5.3 Resuls and Analysis Discriminaory Credi Consrain, TFP and Economic Growh Table 5 repors he regression resuls. As shown in Table 5, Reg.5) esimaes he model3) using he Fix Effecs mehod 7. The coefficien of produc iem DCC i, T F P i, is.257 and significan a 5% level. This reveals ha, in indusries where he exen of discriminaion of credi consrain beween SOE and PE is higher, when TFP decreases, he growh rae of indusrial added value will decrease more dramaically. This is in line wih our conclusion from he model analysis. Discriminaory credi consrain can amplify he impac of echnology shock on oal oupu. The coefficien of TFP is significanly posiive, and he coefficien of ineres rae is significanly negaive, also in line wih exising heory. Reg.3) is he benchmark esimaion, which is he reference o regression models wih produc iems. All esimaed coefficiens are significan, and in line wih exising lieraure. The coefficien of DCC i, is significanly negaive, which shows ha discriminaory credi consrain is damaging o economic growh. [TABLE 5 ABOUT HERE] Moneary Policy and Loan Flow Direcion Table 5 also repors he regression resuls of Eq.31). As shown in Reg.5), he coefficien of produc iem DCC i, In is and significan a 5% level; he coefficien of real ineres rae is -27.7, significan a 1% level. According o our calculaion of coefficiens, we find ha if he exen of discriminaion of credi consrain increases by 1 uni, when real ineres rae decreases by.1 percenage poins, he increase of indusrial added value growh rae will decrease by.5238 percenage poins. Tha is, in indusries where discriminaory credi consrain beween SOEs and PEs is greaer, expansionary moneary policy will be less effecive in promoing economic growh. Therefore, discriminaory credi consrain may hinder he effeciveness of expansionary moneary policy. In Reg.5), he esimaed values of oher coefficiens are also significan, in line wih he benchmark model. For robusness, we also add he wo produc iems in regression a he same ime. All he resuls are significan and consisen wih Reg.4) and Reg.5). 7 The Hausman es has been made for all panel regressions. All he resuls suppor he fixed effec model. 22

23 6 Conclusion Remarks This paper proposes a new explanaion for he recen decline of China s economic growh. We formulae a DSGE model ha incorporaes endogenous LTV raios. In our model, poliical connecions resul in he long-sanding discriminaory credi consrain which plays an imporan role on amplifying economic flucuaion and reducing he effeciveness of expansionary moneary policy. When he economy suffers a negaive shock, he inequiy in credi marke can induce o he financial misallocaion, furher leading o deerioraion of oal loan and oupu. The empirical evidence from China s indusrial daa finds ha he discriminaion of credi consrain beween SOE and PE will amplify he impac of echnology shock on oal oupu and hinder he effeciveness of expansionary moneary policy. These finding are consisen wih he model s predicions. Therefore, we suppor o he marke-orienaion ransformaion of financial archiecure, he amelioraion of financial sysem, and he improvemen of equal righs on he access o finance, which is in line wih Le e al. 214) and he curren policy of China. In fac, poliical connecions widely exis in many counries, especially in emerging counries. From he micro sigh, poliical connecions may help he firms o obain more resources and beer performance. From he macro sigh, however, poliical connecions may induce o he inequiy in resource allocaion, even resource misallocaion, leading o TFP loss and sharp flucuaion of capial price ha are harmful o boh economic growh and flucuaion. Exising lieraure on business cycle pays more aenion o real economy and financial facors, bu neglecs he poliical and insiuional facors. This paper indirecly inroduces poliical facors ino DSGE model, filling he gap o some exen. References Angeloni, I., & Faia, E. 213, April). Capial regulaion and moneary policy wih fragile banks. Journal of Moneary Economics, 6 3), Barro, R. J. 1999). Noes on growh accouning. Journal of Economic Growh, 4 2), Beneš, J., & Lees, K. 27). Monopolisic Banks and Fixed Rae Conracs: Implicaions for Open Economy Inflaion Targeing. Bernanke, B. S., Gerler, M., & Gilchris, S. 1999). The financial acceleraor in a quaniaive business cycle framework. Handbook of macroeconomics, 1, Brand, L., & Li, H. 23). Bank discriminaion in ransiion economies: ideology, informaion, or incenives? Journal of comparaive economics, 31 3),