HE RANSMISSION MECHANISM O BARER by José Noguera, CERGE-EI Sepember 23 Conac Informaon: Jose Noguera, P.O. Box 882, Polckých vězňů 7, 2 Prague, Czech Republc, phone (42-2) 24.5.7, e-mal: Jose.Noguera@cerge.cun.cz
HE RANSMISSION MECHANISM O BARER by José Noguera, CERGE-EI Sepember 23 Absrac hs paper addresses he ssue of wha has caused barer among ndusral frms n Russa. Parcularly, proposes a heorecal framework o deermne wheher barer s he resul of cred raonng or frms opmal choce, and based on ha, conduc a mulvarae VAR analyss o nqure no he emprcs of he phenomenon. JEL: E, E6, P2, P2, P23, P26 Keywords: Barer, money, cred, neres rae, VAR, un roos, ransmsson mechansm.
HE RANSMISSION MECHANISM O BARER José Noguera, CERGE-EI I. INRODUCION he Russan moneary polcy of he 99s can be summarzed as follows: a loose moneary polcy causng he hgh nflaon epsode durng he early sage of he ranson, followed by a gh moneary polcy unl he ruble crss n 998. In fac, afer he hrd quarer of 993, money growh rae and real money balances decreased, reflecng he declne n segnorage. On he oher hand, from 994 on, governmen deb seadly ncreases unl 998. he segnorage for deb subsuon polcy s abandoned afer he 998 Russan crss. Fgure shows real money balances and governmen deb o banks hroughou hs perod o llusrae ha polcy. hs polcy was he scenaro for he rse and fall n barer ransacons n Russa durng he 99s. Accordng o he Russan Economc Baromeer, he number of barer ransacons n ndusral sales n Russa ncreased from 5-% n 992 o 5-6% n 998, and hen falls o -2% n 22. A number of heores have been wren o explan he phenomenon. Some of hem clam ha srucural facors cause barer n Russa. Blanchard and Kremer (997) argue ha s a consequence of dsorganzaon. Gaddy and Ickes (22) propose he vrual economy: choose o barer o evade he offcal economy. Marn and Schnzer (999) crcze he vrual economy. Prendergas and Sole (999, 2a, and 2b) sae ha barer allows marke segmenaon, whch ncreases frms prof. 2
Alhough srucural facors can explan a share of oal barer ransacons, he hgh correlaon wh a number of moneary varables suggess ha a moneary explanaon can beer explan he rapd changes n barer ndcaors. hs s emphaszed n a number of emprcal works lke Bevan e al (2), Brana and Maurel (2), Commander e al (2), Lnz and Krueger (998), and Makarov and Klener (2). One parcular open queson s wheher barerng s he resul of frms choce or s mposed o hem. In a gh cred marke, cred raonng may arse as lenders choose no o ncrease neres raes and cred supply n order o hedge hemselves agans ncreasng rsk. Raoned frms hus may respond by swchng o barer or oher alernave ransacon echnology. On he oher hand, f lenders respond by ncreasng neres raes, oupu decreases, and frms wh access o barer echnology may choose o barer f he cos of hs alernave ransacon echnology s lower han he cos of borrowng. Huang e al (22), and Noguera (22a, b) emphasze how barer can rse as an alernave o a cosly cred marke. hs paper conducs a VAR analyss o es wheher he rapd changes n he share of barer ransacons n Russa durng he 99s resul from frms opmal decson or cred raonng. Secon II provdes he necessary heorecal background. Secon III proposes he economerc model. Secon IV deals wh he daa and nqures abou un roos. Secon V shows and nerpres resuls. hs ncludes Granger causaly and mpulse response funcon analyses. Secon V concludes. II. SOME HEOREICAL BACKGROUND Consder a frm wh an ndvsble projec, whch yelds a reurn R s wh probably q f he projec succeeds, and a reurn R f f he projec fals. Suppose ha he frm borrows fnancal resources for a real value of L f o underake he projec, whch repays a a real 3
neres rae r f he projec succeeds, and he avalable reurn R f f fals. he frm s expeced reurn s l s f [ R ( r L ] E ( π ) q ). On he oher hand, f he frm barers, faces a barer cos per ransacon τ, and he expeced reurn becomes b s f E( π ) qr ( q ) R C( τ), C >, C >. Assume ha C(τ) > L f s f and R > C(τ) > R, for all. A frm borrows f ( l ) ( b E π > E π ). hs occurs f f C( τ) R () q < f f ( r) L R. Le r b denoe he cuoff neres rae ha makes he rgh-hand sde expresson equal o one, hen r C( τ) L b f If r < r b, nequaly () holds, all frms conduc cash ransacons and he cred demand s consan and equal o L f. Ye, f r > r b, a number of frms qu he cred marke, choose o barer and he demand for cred decreases. hus, he demand for cred can be wren as L d f L f r < rb, f µ ( r) L Oherwse where µ(r), and µ (r) <. On he oher hand, n he presence of rsk, s a wellknown resul ha he supply for loans funcon (L S ) s a concave curve 2 ha reaches a maxmum a some neres rae level r r *. Equlbrum s llusraed n Fgures wo. In Noguera (22b) formalze hs argumen. 2 See Sglz and Wess (98), and De Mezza and Webb (987). 4
Fgure 2a, he loan funcon reaches a maxmum a an neres rae r * ha s less han he cuoff neres rae r b, (r * r b ). he cred demand s consan and, f s low enough say L f, all frms oban a loan and here s no barer. Ye, f s oo hgh, greaer han L f, lenders do no respond o he ncreasng demand, here s cred raonng and a barer economy emerges. In hs case, here s no correlaon beween flucuaons n barer ransacons and he real neres rae, bu beween barer and he varable ha s causng he cred marke consran. Consder Fgure 2b, where r * > r b. If he neres rae s below r b, here s no barer. Ye, f he equlbrum neres rae a pon A where r b < r < r *, lenders respond an ncreasng demand for cred wh hgher neres raes, and barer resuls from frms choce. In hs case he ncreasng neres rae s wha drecly causes barerng and boh varables mus be correlaed. On he oher hand, f he cred demand ncreases o L f, some frms barer because of cred raonng, bu some ohers because hey choose o do ; hus, barer mus be correlaed wh he neres rae and also wh he varables causng he cred marke consran. III. HE MODEL From he dscusson n he prevous secon, we denfy hree possble cases n whch a gh moneary polcy resuls n an ncrease of he barer acvy. In each case, he ransmsson mechansm allows o denfy wheher barer resuls from frms choce, cred raonng or boh. For he purpose of llusrang he argumen, n every case we assume ha ncreasng governmen ndebedness s makng he cred marke gher. Frs, he neres rae ncreases, and frms responds by swchng o barer o lower ransacon coss (Pon A n Fgure 2b). Second, here s cred raonng; neres raes do 5
6 no respond o moneary polcy and frms have no oher choce bu o barer; nsead, barer should respond drecly o governmen deb (Fgure 2a). hrd, here s a mxed equlbrum. Ineres raes ncrease wh a gher cred marke makng some frms swch o barer, ye oher frms are smply subjec o cred raonng; barer s correlaed wh boh neres raes and governmen ndebedness (Pon B n Fgure 2b). hus, by denfyng he ransmsson mechansm from moneary polcy o barer, we can also deermne wha makes frms barer. For ha purpose, we conduc a VAR analyss wh fve endogenous varables: barer share (B), oupu (Q), nflaon rae (π), real neres rae (R), and he real deprecaon rae (e). We also nclude wo exogenous varables: he money growh rae (θ) and governmen deb (L). hen, he Granger causaly ess and mpulse response funcons allow denfyng he ransmsson mechansm and hus he causes of barer n Russa. We use he followng model: (2a) b b b b b b b m L e R Q B B ) ( ε φ ϕ β π α δ χ, (2b) m L e R Q B ) ( π π π π π π π ε φ ϕ β π α δ χ π, (2c) q q q q q q q m L e R Q B Q ) ( ε φ ϕ β π α δ χ, (2d) r r r r r r r m L e R Q B R ) ( ε φ ϕ β π α δ χ, (2e) e e e e e e e m L e R Q B e ) ( ε φ ϕ β π α δ χ, where s he number of lags. able summarzes how he ransmsson mechansm helps denfy he reasons for barerng.
IV. DAA AND UNI ROOS We use daa for Russa for he followng seven varables: barer share (B), nflaon rae (π), oupu (Q), he real neres rae (R), he real deprecaon rae (e), governmen deb (L) and money growh rae (θ). he ndcaor s chosen based manly on daa avalably. We use he annualzed percenage growh of he Consumer Prce Index as a measure of he nflaon rae π, and he percenage growh of M2 as a measure of he money growh rae θ. he daa source for boh varables was he Russan Economc rends unl December 993, and he Inernaonal Moneary Fund (Inernaonal Fnancal Sascs) from January 994 on, lne 92264.XXZF for CPI and he sum of money and quas-money (lnes 92234ZF and 92235ZF) for M2. We choose he real exchange rae agans he US dollar as a proxy varable for he real exchange rae e, and he ndusral ndex as a proxy for oupu Q. he source for boh varables s he Russan Economc rends. We use he oal governmen deb o banks as a proxy o nomnal governmen ndebedness. he source s Russan Economc rends. hen, hs varable s deflaced by he consumer prce ndex as measured above o oban a measure of he real governmen ndebedness, L. o measure barer acvy B, we use he barer share of oal ransacons for Russan ndusral frms repored by he Russan Economc Baromeer. As a proxy for he real neres rae R, we use he quoen beween he nomnal neres rae and he nflaon rae. As ndcaor of nflaon we use he ncrease n he consumer prce ndex (IFS/IMF), and for he nomnal neres rae, we use he money marke rae (IFS/IMF) from February 995 on, and before ha dae, we esmae he money marke rae usng a regresson beween he money marke rae and he curren and hree lags of he GKO rae. 7
We use ha daa from May 993 o March 22 o conduc a number of un roo ess. All varables are n measured n logs, excep for he nflaon and money growh raes, whch are n percenage. As a creron o deermne he exsence of un roos, we follow he algorhm suggesed by Ender (995, page 257). Frs compue he followng regresson: y a p a2 γy β y ε, 2 where p s he number of lags. hs regresson ncludes a rend or a drf plays. able 2A shows he coeffcens γ and he -sas of he Augmened Dckey-Fuller ess for un roos for he varables n levels. he crcal value for 95% confdence for he null hypohess of no un roo s 3.45. hus, we accep he alernave hypohess ha he nflaon, real neres and money growh raes are saonary or negraed of order one. he able also shows he F-sas for he null ha a 2 gven ha γ for he remanng regressons. In all cases, he null s rejeced, and so we conclude ha barer share, oupu, exchange rae and real governmen deb have a leas one un roo wh 5% sgnfcance. able 2B shows he resuls of he augmened Dckey-Fuller es for barer share, oupu, exchange rae and governmen deb n frs dfferences. In each case, he regresson compued has a drf and a rend; hen we repea he exercse. he null of un roo s rejeced n each case excep for oupu. he augmened Dckey-Fuller es for he second dfference of oupu rejecs he null of un roo. Resuls are shown n able 2C. herefore, we conclude ha nflaon, real neres and money growh rae are negraed of order zero, barer share, exchange rae and real governmen deb are negraed of order one, and oupu s negraed of order wo. 8
V. ESIMAIONS AND IMPULSE RESPONSE FUNCIONS o es for Granger causaly we run VAR regressons usng fve endogenous varables, barer share, nflaon, oupu, real neres and real exchange rae, and wo exogenous varables: real governmen deb and money growh. Based on he un roo analyss, we nclude nflaon, real neres and money growh raes n levels; barer share, real exchange rae and governmen deb n frs dfferences; and oupu n second dfferences. We also use an nercep n each equaon, and compue he F-sascs o es he null hypohess ha here s no Granger causaly [See Hamlon (994), page 35]. We use daa from July 993 o March 23. We fnd ha he lower he number of lags, he lower boh Akake nformaon and he Schwarz Bayesan crera, whch means ha he model has a very shor memory. We chose wo lags for each equaon. hs s he lowes number of lags ha do no show auocorrelaon n any of he regressons. able 3 shows he compued F sascs for he null ha he varable shown n he op row Granger causes he varable n column. he crcal value for rejecon of he null hypohess s 3.87. From he ess we make he followng conclusons: nflaon Granger causes barer share and neres rae; he real neres rae Granger causes barer share, oupu and real exchange rae; real governmen deb Granger causes barer share; and money growh rae Granger causes nflaon. I s noeworhy ha whle nnovaons o he neres rae affec barer acvy, only nflaon seems o Granger cause real neres rae. Ye, as we wll shorly see, nflaon affecs real neres negavely, and he effec of real neres on barer s posve; on he oher hand, nflaon n Russa was decreasng whle barer acvy was ncreasng durng he 99s. hus, a puzzle arses snce beng real neres an endogenous varable, wha deermnes real neres so as can affec barer acvy. o 9
solve he puzzle, we consder a srucural break around February 995. here are reasons o hnk abou ha. Frs, observe n Fgure ha governmen deb s neglgble before July 994, when he governmen sared o change segnorage for deb as source of governmen fnance. Also, average nflaon decreased from a roughly average of 5% before ha dae o an average of 5% aferward. If nflaon s as hgh as 5%, he adjusmen process s very quck, and can be even almos nsananeous; hus, changes n he money growh rae can barely affec real varables. On he oher hand, lower nflaon gves room o rgdes and a slower adjusmen process of he real neres and real exchange raes. hus, ncludng a moneary polcy varable whou consderng hs srucural break would cause a msspecfcaon ha weakens Granger causaly ess. We exend he model o nclude a dummy varable, whch s equal o zero before February 995, and equal o he money growh rae from ha dae on. Resuls on Granger causaly are shown n able 4. Resuls enhance all prevous Granger causaly fndngs shown n able 3, bu n addon show ha he money growh rae Granger causes he real neres and real exchange raes. Fgure 3 shows he mpulse response funcons for hose cases where Granger causaly s deeced. he Granger causaly analyss and he mpulse funcons ogeher allow mporan conclusons. Frs, nnovaon n nflaon causes an ncrease n barer acvy. hs resul s no new and s he usual explanaon o he barer acvy n oher epsodes lke he German and Hungaran hypernflaon of he XX cenury. However, hs does no seem o be he case of Russa durng he 99s, snce durng ha perod, nflaon n Russa decreased as barer acvy ncreased. We also fnd ha barer s affeced hrough he real neres rae and real governmen deb. In fac, from he Granger
causaly ess and he mpulse response funcons, nnovaon n eher varable drecly causes an ncrease n barer acvy. he neres rae channel means ha some frms have chosen o barer because of he hgh cred cos, and he real governmen deb channel means ha, for oher frms, barer has been a Hobson choce, ha s, he only choce hey has snce hey are subjec o cred raonng. On he oher hand, only nnovaons n he money growh rae make he real neres rae o ncrease. In summary, he moneary consran polcy made he real neres rae ncrease foserng frms o swch no barer; ye also caused cred raonng and some oher frms were forced o barer. hs he mxed equlbrum case llusraed by pon B n Fgure 2b. he VAR analyss conduced also allow oher conclusons ha ypcally appear n moneary models lke nnovaons n money growh make nflaon and real exchange rae emporary ncrease, nnovaons n he real neres rae make oupu and he real exchange rae emporary decrease, and nnovaon n nflaon makes he real neres rae emporary decrease. Fgure 4 llusrae he dynamc of causaly works n hs model. VI. CONCLUSION Despe he small sample avalable, hs VAR exercse provdes srong evdence of wha has caused barer n Russa durng he 99s. Cred raonng explans an mporan share of barer ransacons. he resul ha ncreasng governmen deb does affec neher neres raes nor oupu n he shor-run s a ypcal sympom of cred raonng. Ye, drecly affecs he share of barer ransacons among ndusral frms, whch means ha frms have chosen o barer because of he ncreasng raonng n he cred marke. On he oher hand, we also fnd ha many frms have chosen o barer because of he hgh cos of cred. he economerc exercse conduced shows clear evdence of he
neres rae channel. A resrcve moneary polcy made neres rae ncreases, and he hgher neres raes foser frms o barer. Barer can be seen as an alernave choce o bankrupcy, especally for hose frms ha were subjeced o cred raonng. I seems less clear wheher hs was he case of hose frms ha chose o barer. Ye, solvng hs puzzle s a maer of furher research. REFERENCES ) Bevan, A., S. Esrn, B. Kuzneov, M. Schaffer, M. Angelucc, J. Fennema and G. Mangaro (2) he Deermnans of Prvazed Enerprse Performance n Russa, Wllam Davson Workng Paper Number 452, June 2) Blanchard, O. and M. Kremer (997) Dsorganzaon, Quarerly Journal of Economcs, November, 2-4, 9-26. 3) Brana, S. and M. Maurel (2) Barer n Russa : Lqudy Shorage Versus Lack of Resrucurng, Workng Paper CEPII Research Cener 4) Commander, S., I. Dolnskaya and C. Mumssen (2) Deermnans of Barer n Russa - An Emprcal Analyss, Inernaonal Moneary Fund Workng Paper /55. Ocober 5) De Meza, Davd and Davd C. Webb (987). oo Much Invesmen: A problem of Asymmerc Informaon, Quarerly Journal of Economcs 2 (2), May: 28-292. 6) Enders, Waler (995). Appled Economerc me Seres, John Wley and Sons, Inc. 7) Gaddy, C. and B. Ickes (22) Russa's Vrual Economy, Washngon, D.C.: Brookngs Insuon Press 2
8) Hamlon, J. D. (994). me Seres Analyss, Prnceon Unversy Press, Prnceon, N.J. 9) Lnz, S. and G. Krueger (998) Enerprse Resrucurng n Russa's ranson Economy: Formal and Informal Mechansms Comparave Economc Sudes, 4(2), Summer 998, pages 5-52 ) Makarov, V. and G. Klener (2) Barer n Russa: An Insuonal Sage Problems of Economc ranson 42(), March, pages 5-79 ) Marn, Dala and Monka Schnzer (999) he Economc Insuon of Inernaonal Barer, Economc Journal Aprl 22; 2(479): 293-36. 2) Huang, Hazhou, Marn, Dala and Chenggang Xu (22). Fnancal Crses, Economc Recovery and Bankng Developmen n Former Sove Unon Economes, Cenre for Economc Polcy Research, DP Nr, 3734, London 3) Noguera, J (22). A heorecal Model of Barer n Russa, Cerge-EI Workng Paper Seres. 4) Noguera, J (23). Is Barer a Hobson s Choce? a heory of barer and cred raonng Cerge-EI Workng Paper Seres. 5) Prendergas, Cance and Lars Sole (999) Resrcng he Means of Exchange Whn Organzaons, European Economc Revew, Aprl, vol. 43, no. 4-6: 7-9. 6) Prendergas, Cance and Lars Sole (2a) he Non-Moneary Naure of Gfs, European Economc Revew, December, vol. 45, no. : 793-8. 7) Prendergas, Cance and Lars Sole (2b) Barer, Lqudy and Marke Segmenaon, unpublshed manuscrp, Unversy of Chcago. 3
8) Sglz, Joseph and Andrew Wess (98) Cred Raonng n Markes wh Imperfec Informaon. Amercan Economc Revew 7, 3 (June), 393-4. 4
ABLE BARER AND HE RANSMISSION MECHANISM Frms Choce: Increase n Governmen deb or Increase n expeced exchange or ncrease n R ncrease n B or decrease n θ Cred Raonng: Increase n Governmen deb or Increase n expeced exchange or ncrease n B or decrease n θ and ncrease n R no effec on B Frms Choce and Cred Raonng: Increase n Governmen deb or Increase n expeced exchange or ncrease n R ncrease n B or decrease n θ and Increase n Governmen deb or Increase n expeced exchange or ncrease n B or decrease n θ 5
ABLE 2A AUGMENED DICKEY-FULLER ESS FOR UNI ROOS Varables are n levels Un Coeff. -sa F-sa Lags Regresson Roo Barer share Yes.97 -.66.5. Drf and no rend Inflaon rae No.93-3.52. 2. Drf and rend Oupu Yes.92-2.25 3.76. Drf and rend Real Ineres Rae No.83-3.7.. Drf and rend Real Exchange Yes.96-2. 2.24. Drf and rend Governmen Deb Yes.98 -.66.69. Drf and rend Money Growh No.9-3.8.. Drf and rend ABLE 2B AUGMENED DICKEY-FULLER ESS FOR UNI ROOS Varables are n Frs Dfference Un Coeff. -sa F-sa Lags Regresson Roo Barer share No -.42-8.92.. Drf and rend Oupu Yes.87-2.74 5.4 2. Drf and rend Real Exchange No.39-6.2.. Drf and rend Governmen Deb No -.34-9.3.. Drf and rend ABLE 2C AUGMENED DICKEY-FULLER ESS FOR UNI ROOS Varables are n Second Dfference Un Coeff. -sa F-sa Lags Regresson Roo Oupu No -.5-3.96. 2. Drf and rend 6
ABLE 3 PAIRWISE GRANGER CAUSALIY ESS Compued F sascs for varable n op row Granger causes he varable n frs column Barer Share Inflaon rae Oupu Real neres Real exchange Gov. deb Money growh Barer Share -- 8.38. 7.87.4 7.29.6 Inflaon.76 --.82 2.8.2.26 5.7 Oupu.77.9 -- 3.62.3.3.43 Real neres.8 4.63.53 --.34.38.69 Real exchange.28.77.93 4.85 --.66.4 ABLE 4 GRANGER CAUSALIY ESS INCLUDING A DUMMY Compued F sascs for varable n op row Granger causes he varable n frs column Barer Share Inflaon rae Oupu Real neres Real exchange Gov. deb Money growh Money growh (Dummy) Barer Share -- 9.8.3 9.3.76 5.7.84 2.3 Inflaon.66 --.96.82.5.38 5.5.33 Oupu.77.7 -- 3.69.34.4.25.7 Real neres.6 5.9.46 --.8.26 2.6 4.79 Real exchange.4.85.58 3.9 --.44.32 3.7 7
FIGURE Real Money Balances and Governmen Deb o Banks n Russa 6 2 8 4 992-QI 993-QI 994-QI 995-QI 996-QI 997-QI 998-QI 999-QI 2-QI 2-QI 22-QI Real Money Balances Governmen Deb o Banks Source: IFS/IMF (real money balances) Russan Economc rends (Governmen Deb o Banks) 8
FIGURE 2a Barer When Frms Choose o Do I L f L f L S r * r b 9
FIGURE 2b Barer When here s Cred Raonng B A d L d L L s r b r * 2
2
FIGURE 3 Impulse Response Funcons Inflaon Granger causes barer share Inflaon Granger causes neres rae.3.25.2.5..5 -.5 -. -.5 -.2 2 3 4 5 6 7 8 9.5 -.5 -. -.5 -.2 -.25 -.3 -.35 -.4 -.45 3 5 7 9 3 5 7 9 2 23 25 27 29 Ineres Granger causes barer share Ineres Granger causes oupu.4.2.6.4.2..8.6.4.2 -.2 2 3 4 5 6 7 8 9 -.2 -.4 -.6 -.8 -. -.2 -.4 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 2 22 23 Ineres Granger causes real exchange Gov. deb Granger causes barer share,.2. -. -.2 -.3 -.4 -.5 -.6 -.7 -.8 -.9 3 5 7 9 3 5 7 9 2 23 25.35.3.25.2.5..5 -.5 -. 2 3 4 5 6 7 8 Money growh Granger causes nflaon Money growh Granger causes real neres.25.5.2..5.5. -.5 -. 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8.5 -.5 -.5 3 5 7 9 3 5 7 9 2 23 -.2 -.25 -.3 Money Growh Granger causes real exchange.8.7.6.5.4.3.2. -. -.2 2 3 4 5 6 7 8 9 22
FIGURE 4 Dynamc of Causaly Money Growh Rae Real Exchange Inflaon Barer Share Ineres Rae _ Governmen Deb Oupu 23