Why Commercial Banks Held Excess Reserves: The Japanese Experience of the Late 90s*

Transcription

1 December 2004 Revsed December 2005 Why Commercal Banks Held Excess Reserves: The Japanese Experence of the Late 90s* Kazuo Ogawa Insttute of Socal and Economc Research, Osaka Unversty * Ths s a substantal revson of the paper I submtted to the Fnancal Research Councl. I am grateful to Hrosh Fujk, Shn-ch Fukuda, Hdeo Hayakawa, Nobuo Inaba, Kazumasa Iwata, Toshk Jnush, Keme Kazuka, Hrosh Koyama, Ryuzo Myao, Hrosh Nakaso, Kuno Okna, Yoshnor Shmzu, Etsuro Shoj, Shgenor Shratsuka, Hsash Tanzak, Juro Teransh, Fukujyu Yamazak, Jyunj Yano, Yukhro Yasuda, the partcpants of the semnars at the Bank of Japan and Kobe Unversty and an anonymous referee for extremely valuable comments and suggestons on early verson of the paper. Ths research was partally supported by Grants-n-Ad for Scentfc Research of the Mnstry of Educaton.

2 Abstract We nvestgated, emprcally, why Japanese banks held excess reserves n the late 1990s. Specfcally, we pn down two factors explanng the demand for excess reserves: a low short-term nterest rate, or call rate, and the fragle fnancal health of banks. The vrtually zero call rate ncreased the demand for excess reserves substantally, and a hgh bad loans rato largely contrbuted to the ncrease n excess reserve holdngs. We found that the holdngs of excess reserves would fall by two-thrds f the call rate were to be rased to ts level pror to the adopton of the zero-nterest-rate polcy, and the bad loans rato were to fall by 50%. JEL Classfcaton Number: E42, E51, E52, and G21 Keywords: Excess Reserve, Bad loans, Zero-nterest-rate-polcy Correspondence to: Insttute of Socal and Economc Research, Osaka Unversty, 6-1 Mhogaoka, Ibarak, Osaka, JAPAN Tel: Fax: E-mal: ogawa@ser.osaka-u.ac.jp

3 1. Introducton Japanese banks have chroncally held excess reserves snce the late 90s. Fgure 1 llustrates the rato of actual reserves to requred reserves for commercal banks as a whole. The ncreasng trend n the excess reserve rato has been conspcuous snce the summer of 2001, and ths rato reached a hgh of 5.88 n October The excess reserve rato typcally parallels the supply of reserves. In fact, the reserve supply began to ncrease when the Bank of Japan announced that t would provde ample funds to push down the uncollateralzed overnght call rate, or short-term nter-bank money market rate, as low as possble, n February Ths acton s known as the zero-nterest-rate polcy. Furthermore, the reserve supply drastcally ncreased after the Bank of Japan (BOJ) swtched ts operatng target for money market operatons from the uncollateralzed overnght call rate to the outstandng balances of the current accounts held at the BOJ n March The target level of outstandng balances of the BOJ s current accounts was valued at around fve trllon yen n March 2001, but then rose to between 30 and 35 trllon yen by January However, t should be noted that the reserve supply does not necessarly automatcally create a demand for reserves. The am of ths study, therefore, was to shed lght on the demand sde of reserves and to conduct an emprcal nvestgaton nto why commercal banks hold excess reserves. In actual fact, banks have ther own motves for holdng excess reserves. We dentfed two factors that may have affected banks reserve demands snce the late 90s. The frst of these factors s the nearly zero level of the call rate: the uncollateralzed overnght call rate was pushed down to ts mnmum level after the adopton of the zero-nterest-rate polcy. Reserves and call loans are close substtutes; as a result, banks have more ncentve to hold reserves when the call rate s very low. The second relevant factor s the nstablty of the fnancal system. Ever snce the outbreak of fnancal nsttuton debacles n the late 90s, depostors have been very cautous about choosng whch banks to depost ther money n. Ths mples that, n general, once the balance sheet of a bank deterorates, depostors wll swtch deposts from one bank to another, more healthy one. Therefore, banks wth fragle balance sheets have an ncentve to hold reserves, for precautonary reasons. 2 Incorporatng these factors, 1

4 we constructed a smple theoretcal model of bank reserve demand. Usng a panel dataset for banks, we estmated these banks optmal demand functons for reserves, as derved from the theoretcal model. The quanttatve mportance of the aforementoned two factors was evaluated by means of a smulaton analyss, based on the parameter estmates n the banks reserve demand functons. 3 Our man fndngs may be summarzed as follows. Usng a theoretcal model, we were able to derve each bank s optmal demand for reserves, whch s a decreasng functon of the short-term nterest rate and an ncreasng functon of the bank s fnancal health, represented by ether ts bad loan rato or the rate of change n ts share prce. It turned out that both factors sgnfcantly mpact banks reserve demands, as suggested by the theoretcal model. Quanttatvely, the effect of the short-term nterest rate on reserve demand was qute large. We found that rasng the call rate to 0.25%, ts level just before the adopton of the zero-nterest-rate polcy, decreased the demand for excess reserves n the bankng sector as a whole by as much as 55-70%. The effect of a bank s fnancal health on ts excess reserve demand was also sgnfcant. Excess reserves were reduced by 13-27% when the bad loan rato was halved; ths consttutes the goal to be attaned by the end of FY 2004 n the Program for Fnancal Revval, under the Kozum Admnstraton. The paper s organzed as follows. The next secton provdes descrptve statstcs for the banks reserve holdngs, whch are based on the panel dataset for the Japanese commercal banks used n our study. Secton 3 constructs a smple theoretcal model of reserve demand and derves the optmal demand equaton for reserves, whch are to be estmated. Secton 4 presents the estmaton results, and Secton 5 conducts a smulaton analyss, based on the estmaton results from the prevous secton, n order to evaluate quanttatvely the effects of the short-term nterest rate and bank fnancal health on excess reserve demand. Secton 6 concludes the study. 2. To What Extent Dd Banks Hold Excess Reserves? We frst calculated the extent to whch commercal banks held excess reserves usng panel data for Japanese banks durng the perod FY 1991 to We compared the 2

5 actual reserve holdngs at the end of each fscal year wth the reserves requred at that tme. The requred reserves were calculated as follows. The reserve requrement ratos for deposts depend upon the type and amount of deposts, as s shown n Table 1. For example, when a bank holds three trllon yen as tme deposts, the bank should hold bllon yen (= ( ) ( ) ( ) (500-50) ) as requred reserves. The requred reserves at the end of a fscal year are computed for each bank n our panel dataset, and compared wth the actual reserve balance, to gauge the extent to whch banks hold excess reserves. Our panel data set conssts of 145 banks, of whch nne are cty banks, three are long-term credt banks, seven are trusts, and 126 are regonal banks. Merged and/or nsolvent banks were excluded from the sample. Therefore, our dataset consttutes an unbalanced panel. The sample perod covers the years from FY 1991 to FY Table 2 presents descrptve statstcs of the extent to whch banks held excess reserves. The medan of the rato of actual reserves to requred reserves was less than one untl 1996, whle t exceeded ths value after It should be noted that the year n whch the excess reserve rato began to exceed one was characterzed by a number of bankruptces of fnancal nsttutons. It s lkely that banks held excess reserves for precautonary purposes, to meet the demands of depostors swtchng from unhealthy banks to healthy ones. In 2001, the excess reserve rato agan exhbted a sharp ncrease. The medan of the excess reserve rato n 2001 was about twelve tmes as large as that n Ths sharp ncrease concded wth the BOJ s swtch of monetary polcy target: from the short-term nterest rate to the outstandng balances of the current accounts at the BOJ, consttutng the so-called quanttatve easng of monetary polcy. Under the new monetary polcy regme, the BOJ ncreased the monetary base substantally. The standard devaton of the excess reserve rato also ncreased after the mddle of the 90s, suggestng that there were dverse reserve holdng patterns across banks. Ths fact may also be nferred from Fgure 2, whch llustrates the frequency dstrbuton of the excess reserve ratos n 1994, 1999 and The frequency dstrbuton was more dspersed and less skewed to the rght n later years. 3

6 3. A Model of Bank Demand for Reserves Consder a bank, whch allocates a gven amount of deposts ( D ) to an nterest-bearng asset, and wth reserves at a central bank bearng no nterest, under the stochastc wthdrawal of deposts. 5 By allocatng all of ts deposts to the nterest-bearng asset, and keepng ts reserves to the mnmum requred level, the bank can maxmze ts ncome from nterest but runs the rsk of beng unable to meet large, unantcpated wthdrawals of deposts. Such a stuaton may be costly for the bank, as t gves rse to apprehenson among depostors regardng the fraglty of the bank s balance sheet, and depostors may thus be prompted to further wthdraw ther deposts. Therefore, the bank optmally allocates ts resources between the nterest-bearng asset and reserves, to maxmze ts expected nterest ncome, whle smultaneously takng nto account the cost ncurred n the case of reserve shortages. Formally, the objectve functon of the bank s stated as follows: r L ~ ( D R ) r E[ Max( 0, X R )] p, (1) where r L : nterest rate on nterest-bearng asset r, : penalty rate for the -th bank n case of reserve shortage ( r > r P R : reserve balance of the -th bank X ~ : stochastc depost wthdrawal at the end of the perod p, L ) The bank maxmzes eq.(1) wth respect to R subject to the followng requred reserve constrant: R δ D where δ :requred reserve rato for the -th bank The frst-order condton s gven by 6 4

7 r ~ [ X R ] = r λ p, Pr L (2) where Pr [ ] ndcates probablty λ : non-negatve Lagrangean multpler assocated wth the requred reserve constrant The LHS of eq.(2) represents the decrease n the lqudty shortage cost brought about by ncreasng reserves, or the margnal beneft of ncreasng reserves, whle the RHS provdes the margnal cost of ncreasng reserves. When λ > 0, the bank holds no more than the requred amount of reserves. On the other hand, when the optmal reserve * holdngs ( ) R exceed the requred level, then λ = 0. Fgure 3 llustrates the way n whch the optmal level of excess reserves s determned. When λ = 0, t s easy to see from the fgure that the demand for reserves depends upon the nterest rate ( r L ) and the penalty rate ( r, ). The demand for reserves ncreases as the nterest rate falls and the penalty rate P rses. Ths model, whle smple, can explan the surge n excess reserve holdngs by Japanese banks that took place n the late 90s. The call rate was vrtually zero under the zero-nterest-rate polcy, and the BOJ mantaned the call rate to a level as low as possble under the new monetary polcy regme. In the late 90s, banks wth massve amounts of bad loans became nsolvent, whch may have led to a rse n the penalty rate. For ths reason, depostors became very cautous about the balance sheet condtons of banks n an unstable fnancal system. If a bank s reserves fell short of depostors wthdrawals, then depostors may have perceved the balance sheet of the bank to be deteroratng. Ths percepton would have accelerated the wthdrawal of deposts, as well as the potental bankruptcy of the bank. To avod such a dsaster, banks would have set the penalty rate qute hgh and ncreased ther demand for excess reserves. 4. Estmaton of Bank Demand Equatons for Reserves Dervaton of Bank Demand Equatons for Reserves To derve a bank s optmal demand equaton for reserves, we must specfy the way n whch banks antcpate depost wthdrawals. We assume that the random varable X ~ has 5

8 a Pareto dstrbuton. Specfcally, the densty functon of θ ~ θx 0, ( X ) = ~ θ + 1 ~ f X 0, < X < (3) X where θ, : parameters of the dstrbuton θ X 0, 0, X, > 0 > 0 X ~ s wrtten as By specfyng the densty functon of depost wthdrawals, we may calculate the probablty that depost wthdrawals exceed reserves, as follows: ~ [ X R ] R Pr = (4) X 0, θ The substtuton of eq.(4) nto eq.(2) and the rearrangement of terms yelds the followng reserve equaton: log R 1 rl λ = log X 0, log θ rp, (5) It s qute lkely that the bank s subjectve dstrbuton of X ~ depends on the bank s attrbutes. When the balance sheet of a bank deterorates, the bank may antcpate a large amount of depost wthdrawals. In other words, the condton of a bank s balance sheet affects the dstrbuton of ts depost wthdrawals. We specfy a bank s balance sheet condton usng the bad loans rato ( BADLOAN ). That s, we express X 0, as a functon of deposts and the bad loans rato. η ε 0, αdi BADLOAN (6) X = where α > 0, η > 0, ε > 0 Deposts are ncluded n eq.(6) as a scale factor. Substtutng eq.(6) nto eq.(5) and arrangng terms, we obtan the followng optmal demand equaton for reserves: 6

9 log R 1 rl λ = logα + η log D + ε log( BADLOAN ) log (7) θ rp, For estmaton purposes, we transform the demand equaton for reserves nto the devaton from requred reserves. Frst, when λ > 0, banks do not hold any excess reserves, so the devaton of optmal reserves from the requred level s zero. In other words, the followng equaton s specfed n logarthmc form: log R logδ D = 0 (8) Second, when the optmal demand for reserves exceeds the requred level, the demand for excess reserves s expressed as 1 r log R logδ D = logα logδ + (9) ( ) L η 1 log D + ε log( BADLOAN ) log θ rp, Snce we can dentfy from the data whch banks held excess reserves, we may estmate the system of reserve demand equatons, (8) and (9), usng censored regresson models. Descrpton of the Data We brefly explan the varables used n estmaton. The varable ( R ) represents the reserve balance of an ndvdual bank at the BOJ. Deposts ( D ) consst of all types of deposts, ncludng certfcates of depost and debentures outstandng. The amount of requred reserves s calculated based on the depost balances and the requrement ratos provded n Table 1, as was dscussed n Secton 2. The requred reserve rato ( δ ) s obtaned by dvdng the requred reserves by deposts. The bad loans rato s calculated n two dfferent ways: (1) the rato of rsk management loans to total loans ( BADLOAN1 ) 7 and (2) the rato of non-performng loans (NPL) under the Fnancal Reconstructon Law ( BADLOAN 2). 8 Use of the rate of 7

10 change n the share prce has the mert that we can make full use of the whole sample perod from fscal year of 1991 n estmaton. The varable BADLOAN1 s avalable for the perod ncludng and followng 1997, whle the varable BADLOAN 2 s avalable for the perod ncludng and followng Accordngly, the estmaton perod n each case begns wth the year n whch the bad loans rato frst becomes avalable. We also use the rate of change n the share prce ( SHARE ) as a proxy for the market s percepton of the banks fnancal health. 9 Table 3 presents the descrptve statstcs for the bad loans rato. The varable BADLOAN1 exhbted an ncreasng trend untl 2001, n terms of both the mean and the medan, whle the same trend was observed for BADLOAN 2n terms of the medan. Fgure 4 presents the frequency dstrbuton for bad loans ratos n 1998 and 2002, both of whch are calculated from the rsk management loans. There are consderable dfferences between the shapes of the frequency dstrbutons n both years. In 2002, the peak of the frequency dstrbuton rose, and the dstrbuton was more skewed to the left. The correlaton coeffcent for these two measures s qute hgh (0.8982). However, the market s percepton of banks fnancal health does not necessarly closely reflect the bad loans rato. In fact, the correlaton coeffcent between BADLOAN1 and SHARE s , and that between BADLOAN 2 and SHARE s The penalty rate s lkewse treated n two dfferent ways. Frst, the penalty rate s treated as constant across banks and years. In ths case, the penalty rate s subsumed nto the constant term. Second, the rato of operatng profts to total equty s employed. The dea behnd ths s that the banks, the reserves of whch are short of depost wthdrawals, wll eventually go nsolvent and lose ther entre operatng profts. The nterest rate s the uncollateralzed overnght call rate. Econometrc Issues We estmated eqs.(8) and (9) by two dfferent methods for censored excess reserves. One s pooled Tobt. The followng procedure was used to classfy banks, accordng to whether they have excess reserves. When a bank s actual reserves exceed the requred reserves by at least 10%, then t s dentfed as holdng excess reserves. Ths conservatve 8

11 classfcaton wth respect to banks holdng excess reserves reflects the unavalablty of the daly depost data upon whch the reserve requrement should be based. 10 However, rgorously speakng, requred reserves correspondng to the average outstandng amount of deposts n March are requred to mantan over the perod from the March 16 to Aprl 15 on average. Thus, the requred reserve s not necessarly bndng on the sngle day at the end of March. Therefore t s lkely that excess reserves are measured wth errors and observatons are msclassfed. Tobt estmator s stll vald when the measurement error and the orgnal dsturbance have a jont normal dstrbuton snce the measurement error s ndstngushable from the dsturbance and the composte error term s normally dstrbuted. However, as was dscussed n calculaton of excess reserves, t s lkely that they are underestmated. Then normalty assumpton s no longer tenable and the Tobt estmator s nconsstent. To cope wth ths stuaton, we use censored least absolute devaton estmator (CLAD) developed by Powell(1984). Unlke the standard estmators of the censored regresson model such as Tobt or other maxmum lkelhood approaches, the CLAD estmator s robust to heteroscedastcty and s consstent for a wde class of error structures. Durng estmaton, we added dummy varables to account for the dfferent types of banks,.e., dummy varables for cty banks (DCITY), long-term credt banks (DCREDIT), and trusts (DTRUST). We also added the dummy varable (DPOLICY) to control for the monetary polcy regme change that occurred n March Thereafter, the monetary base ncreased substantally n order to satsfy the target set for the current accounts balance at the BOJ. Estmaton Results Table 4 presents the estmaton results for banks demands for reserves by pooled Tobt. The estmaton s conducted for sx specfcatons, where three varables are used to represent the banks fnancal health, and where two account for the penalty rate. The estmaton results are qute satsfactory, n the sense that the coeffcent estmates of the three key varables determnng the demand for reserves are statstcally sgnfcant, 9

12 rrespectve of the model specfcaton. Deposts exert a sgnfcantly postve effect on reserve holdngs. The call rate has a negatve effect on the demand for reserves, as s predcted from the theoretcal model n Secton The two bad loans ratos also affect the demand for reserves postvely, whch mples that banks wth hgher bad loans ratos ncreased ther reserves. Furthermore, ncreases n the rate of change n share prces decreased the demand for reserves sgnfcantly. Table 5 shows the estmaton results for banks demands for reserves by CLAD. The coeffcent estmates of the three key varables determnng the demand for reserves remans unaltered qualtatvely, although there are quanttatve dfferences. The effects of deposts on reserve holdngs are qute smlar to Tobt estmates when bad loan ratos represent the banks fnancal health, but the coeffcent estmates of depost are nsgnfcant when the rate of change n share prce s used. The call rate has a sgnfcantly negatve effect on the demand for reserves, rrespectve of the model specfcaton. The two bad loans ratos also exert a sgnfcantly postve effect on the demand for reserves, whle the effect of share prce change on the demand for reserves s slghtly weaker. To sum up, the basc results obtaned by Tobt estmaton are essentally unaltered even after takng measurement errors nto consderaton. Table 6 tabulates the elastctes of reserve demands wth respect to deposts, the call rate, and the bad loans rato, for the four types of banks holdng excess reserves. These elastctes are calculated on the bass of the results of the estmaton wth BADLOAN1 as the bad loans rato, and the rato of operatng profts to total equty as the penalty rate The depost elastcty s generally close to one, except for cty banks, n whch case t s much less than one (0.4675). The nterest-rate elastcty ranges from to , whle the elastcty wth respect to the bad loans rato ranges from to The magntude of ths elastcty s smaller for cty banks because, of these, a lower proporton of banks held excess reserves. 5. The Effects of the Short-term Interest Rate and Bank Fnancal Health on the Demand for Excess Reserves: A Quanttatve Evaluaton In the prevous secton, we found that both the short-term nterest rate and banks 10

13 fnancal health are sgnfcant explanatory varables of the demand for reserves. Based on the parameter estmates of the demand equaton for reserves, and usng smulaton analyss, we evaluated quanttatvely the extent to whch changes n the short-term nterest rate and/or banks bad loans ratos affected the demand for excess reserves. Specfcally, we took the followng steps. Frst, we calculated the theoretcal values of excess reserves by substtutng the actual exogenous varables for 2002 nto the estmated demand equaton for reserves, or eq.(9), and subtracted the requred reserves. Next, we calculated the predcted value of excess reserves under dfferent scenaros of the call rate and the bad loans rato. We assumed that the call rate was rased to ts level pror to the adopton of the zero-nterest-rate polcy. In other words, the call rate was set to ts level n March 1998, or to 0.25%. 15 Wth regard to the banks fnancal health, we consdered a case n whch the bad loans rato was halved. Ths scenaro s consstent wth the Kozum Structural Reform Plan, ntroduced n October 2002, whch specfed that non-performng loans should be reduced by 50% wthn one year. The predcted excess reserves thus calculated were summed up across banks, and the aggregated excess reserves were compared wth ther baselne value. Table 7 ndcates the extent to whch banks excess reserves are reduced under the dfferent scenaros for the call rate and the bad loans rato, and under four dfferent specfcatons of the reserve demand equaton estmated by pooled Tobt. When the call rate s rased, excess reserves are reduced by as much as 70% n the constant penalty case, and by about 60% n the varable penalty case. Although the magntude of the reducton n excess reserves s slghtly less for the case n whch the banks fnancal health s mproved, t s not neglgble n ths case. Excess reserves are reduced by 13% to 27%. When ncreases n both the call rate and banks fnancal health are mplemented smultaneously, the holdngs of excess reserves are reduced almost by two-thrds. 6. Concludng Remarks We emprcally nvestgated why Japanese banks held excess reserves n the late 90s. We were able to pn down two factors that explan the demand for excess reserves: a low 11

14 short-term nterest rate, or call rate, and banks fragle fnancal health. The nearly zero call rate substantally ncreased the demand for reserves, and the hgh bad loans rato also contrbuted to the observed ncrease n reserve holdngs. A quanttatve evaluaton of these factors was also conducted. It was found that excess reserve holdngs may be reduced by two-thrds by rasng the call rate to ts level pror to the zero-nterest-rate polcy, and by decreasng the bad loans rato by 50%. The Japanese fnancal system s ganng stablty as banks rapdly elmnate bad loans, and ths elmnaton should lead to a reducton n excess reserve holdngs. 12

15 Footnotes 1 There are two spkes n the fgures, each of whch corresponds to Year 2000 and Fscal Year 2002 problems. On those occasons, the polcy authorty provded ample lqudty to meet a surge n demand and secure stablty n the fnancal markets. 2 See Shrakawa (2002) for a detaled dscusson of the factors that have contrbuted to an ncrease n banks reserve demands snce the late 90s. 3 Hosono et al. (2001), Hayash (2001), and Uesug (2002) nvestgated the lqudty effect usng daly reserve data. The relatonshp between the equlbrum short-term nterest rate and the reserve supply may be obtaned by aggregatng banks reserve demands. Our focus n ths study was to examne what motvates ndvdual banks to hold excess reserves. 4 Accordng to our calculaton, actual reserves are always short of requred reserves untl 1996, whch hnts that the data on excess reserves mght be underestmated. To see ths possblty, we calculated the rato of the outstandng amount of reserves at the end of month to the average outstandng amount, usng the aggregate monthly reserve data. Ths rato was sgnfcantly larger than unty for March and September. Snce the average outstandng reserves are close to requred reserves pror to adopton of zero nterest rate polcy, t suggests that banks hold reserves n excess of requred reserves at the end of fscal year, whch cannot be captured n our excess reserve fgures. Underestmaton of excess reserves s taken nto consderaton n estmaton of reserve demand functon n subsequent secton. 5 The followng model of reserve demands comes from Chapter 8 n Frexas and Rochet (1997). 6 It can be shown that the objectve functon of the bank s a concave functon of reserves, so the soluton to the frst-order condton s guaranteed to be a maxmum of the objectve functon. 7 Rsk management loans nclude loans to borrowers n legal bankruptcy, past-due loans n arrears by sx months or more, loans n arrears by three months or more but less than sx months, and restructured loans. 8 The NPLs under the Fnancal Reconstructon Law are defned as the sum of the assets classfed as bankrupt/de facto bankrupt, doubtful, and specal attenton. 9 Some banks are not lsted on the stock exchange, so these are excluded from the sample n the estmaton of the demand equaton for reserves ncorporatng the share prce. 10 To check for robustness, we also used a more conservatve defnton of excess reserves. In ths case, banks wth actual reserves exceedng the requred reserve by 50% were dentfed as banks holdng excess reserves. Under ths defnton, the estmaton results remaned qualtatvely unchanged. 11 The effects of call rate on reserve demand mght be ntensfed for the followng reason. When the nter-bank nterest rate s vrtually zero, transactons n the call market have been declnng and lqudty-rsk sharng mechansm has been eroded substantally. In ths stuaton demand for reserves provded by the Bank of Japan wll be very senstve to change n the call rate. I thank the referee for ndcatng ths pont. 12 Loan and Wood (2002) also reported that the nterest rate has a sgnfcantly negatve 13

16 effect on the reserve demands of commercal banks n the U.S.A. 13 The elastctes are calculated by multplyng the adjustment factor by the orgnal parameter estmates. See Chapter 16 of Wooldrdge (2002) for a dscusson of adjustment factors. 14 The elastctes based on the other specfcatons are calculated n a smlar way. They are not reported here, snce they follow the same pattern as that reported n the text. These elastcty estmates are avalable upon request from the author. 15 It s assumed that the penalty rate does not change. 14

17 References [1] Frexas, X. and J.-C. Rochet(1997). Mcroeconomcs of Bankng, The MIT Press. [2] Hayash,F.(2001). Identfyng a Lqudty Effect n the Japanese Interbank Market, Internatonal Economc Revew 42, pp [3] Hosono,K., Sughara,S. and T.Mhra(2001) Knyu Sesaku no Yukose to Genka (Effectveness and Lmts of Monetary Polcy), Toyo Keza Shnposha, (n Japanese). [4] Lown,C.S. and J.H.Wood(2003). The Determnaton of Commercal Bank Reserve Requrements, Revew of Fnancal Economcs 12, pp [5] Powell,J.L.(1984). Least Absolute Devaton Estmaton for the Censored Regresson Model, Journal of Econometrcs 25, pp [6] Shrakawa,M.(2002). Ryotek Kanwa Sayogo 1 nenkan no Keken (Experence durng One Year after Adopton of Quanttatve Easng Polcy), n Komya,R. and Nppon Keza Center(eds.) Knyu Sesaku Rong no Soten: Nchgn Hhan to sono Hanron (Issues of Monetary Polcy Debates: Crtcsm of the Bank of Japan and Rebuttal), Nppon Keza Shnbunsha, pp (n Japanese) [7] Uesug,I.(2002). Measurng the Lqudty Effect: The Case of Japan, Journal of the Japanese and Internatonal Economes 16, pp [8] Wooldrdge,J.M.(2002). Econometrc Analyss of Cross Secton and Panel Data, The MIT Press. 13

18 Fgure 1 Excess Reserve Rato The Excess reserve rato s defned as the rato of actual reserve to requred reserve. Data Source: The Bank of Japan Homepage

19 Fgure 2 Frequency Dstrbuton of Excess Reserve Rato n Data Source: Nkke Fnancal Data Medan=0.91 Standard Devaton=1.56

20 Fgure 2 Frequency Dstrbuton of Excess Reserve Rato n Data Source: Nkke Fnancal Data Medan =4.00 Standard Devaton=39.38

21 Fgure 2 Frequency Dstrbuton of Excess Reserve Rato n Data Source: Nkke Fnancal Data Medan=15.18 Standard Devaton=39.25

22 Pr Fgure 3 Determnaton of Optmal Demand for Reserve ~ [ X ] R r r L p, 0 δ D * R R

23 Fgure 4 Frequency Dstrbuton of Bad Loans Rato n 1998 Based upon Rsk Management Loans Data Source: Nkke Fnancal Data

24 Fgure 4 Frequency Dstrbuton of Bad Loans Rato n 2002 Based upon Rsk Management Loans % Data Source: Nkke Fnacal Data

25 Table 1 Reserve Requrement Ratos (1) Reserve Ratos on Deposts Type of deposts Amount of deposts Reserve rato (%) Tme Deposts 2.5 trllon yen ncludng Certfcates of Depost trllon yen 500 bllon trllon yen bllon yen 0.05 Other deposts 2.5 trllon yen trllon yen 500 bllon trllon yen bllon yen 0.1 (2) Reserve Ratos on Debentures Outstandng Type of banks Reserve rato (%) Banks, long-term credt banks 0.1 and foregn exchange bank Notes: The reserve ratos n the table are effectve on October 16, Data Source: The Bank of Japan Homepage

26 Table 2 Descrptve Statstcs of Banks Excess Reserve Fscal year Rato of actual reserve to requred reserve Medan Standard devaton Data Source: Nkke Fnancal Data

27 Table 3 Descrptve Statstcs of Bad Loans Rato Fscal year rato of rsk management rato of non-performng loans loans to total loans under Fnancal Reconstructon ( BADLOAN1) Law to total loans ( BADLOAN 2) Mean Medan Mean Medan Notes: Rsk management loans nclude loans to borrowers n legal bankruptcy, past due loans n arrears by 6 months or more, loans n arrears by 3 months or more and less than 6 months, and restructured loans. The non-performng loans under the Fnancal Reconstructon Law are defned as the sum of the assets classfed as bankrupt/de facto bankrupt, doubtful, and specal attenton. Data Source: Nkke Fnancal Data (%)

28 Table 4 Estmaton Results of Optmal Demand Equaton for Reserve (Pooled Tobt) Choce of banks balance sheet varable and penalty rate BADLOAN1 Case 1 Constant penalty rate Case 2 Varable penalty rate BADLOAN2 Case 3 Constant penalty rate Case 4 Varable penalty rate Share prce change Case 5 Constant penalty rate Case 6 Varable penalty rate Constant Depost Call rate Bad loans rato Dummy varables or share prce Cty banks Long-term Trust Monetary σ NOBS change credt banks banks polcy change LOGL *** *** *** *** ** *** *** 757 (-15.10) (21.08) (-10.05) (4.29) (-2.08) (-1.18) (1.29) (3.65) (29.99) *** *** *** *** ** *** *** 735 (-13.74) (21.03) (-9.42) (4.72) (-2.25) (-0.82) (0.68) (4.71) (29.53) *** *** *** ** * * *** *** 548 (-13.49) (16.92) (-10.34) (2.09) (-1.75) (-0.53) (1.66) (2.66) (27.24) *** *** *** *** * *** *** 530 (-10.90) (16.47) (-8.28) (3.04) (-1.95) (-0.05) (0.88) (3.30) (26.80) *** *** *** *** * *** *** *** 1134 (-16.04) (21.77) (-11.40) (-3.31) (-1.67) (3.08) (5.65) (32.40) *** *** *** *** * *** *** *** 1121 (-15.26) (21.62) (-10.27) (-2.95) (-1.84) (2.64) (7.25) (32.12) Notes: σ : standard devaton of dsturbance LOGL: logarthm of lkelhood functon NOBS: number of observatons Values n parentheses are t-values. *, **, ***: sgnfcant at 10%, 5%, and 1% level, respectvely

29 Table 5 Estmaton Results of Optmal Demand Equaton for Reserve (Censored Least Absolute Devatons Estmator) Choce of banks balance sheet varable and penalty rate BADLOAN1 Case 1 Constant penalty rate Case 2 Varable penalty rate BADLOAN2 Case 3 Constant penalty rate Case 4 Varable penalty rate Share prce change Case 5 Constant penalty rate Case 6 Varable penalty rate Constant Depost Call rate Bad loans rato Dummy varables or share prce Cty banks Long-term Trust Monetary NOBS change credt banks banks polcy change *** *** *** *** *** *** 757 (-6.19) (7.99) (-11.14) (3.37) (-0.41) (-1.06) (3.95) (8.94) *** *** *** *** ** *** 735 (-3.57) (6.17) (-8.10) (4.25) (-0.54) (0.32) (1.30) (7.06) *** *** *** ** (-2.80) (4.40) (-8.01) (1.97) (-0.23) (0.01) (0.48) (0.12) *** *** *** *** * *** 530 (-2.57) (5.80) (-10.41) (5.66) (0.22) (1.78) (0.95) (3.93) *** * ** *** *** 1134 (0.63) (0.60) (-7.93) (-1.79) (2.39) (2.93) (3.20) *** * *** 1121 (0.29) (1.03) (-8.65) (-1.19) (1.13) (1.71) (3.68) Notes: The standard errors are estmated by bootstrap procedure. We performed bootstrap replcatons. NOBS: number of observatons Values n parentheses are t-values. *, **, ***: sgnfcant at 10%, 5%, and 1% level, respectvely

30 Table 6 Estmates of Reserve Demand Elastcty Types of banks Depost Call Bad loan rate rato Cty banks Long-term credt Banks Trust banks Regonal banks Notes: The elastcty s calculated for banks holdng excess reserve on the bass of estmaton results wth BADLOAN1 as the bad loan rato and the rato of operatng profts to total equty as penalty rate.

31 Table 7 Effects of Call Rate Increase and Reducton of Bad Loans Rato on Demand for Excess Reserve: Quanttatve Evaluaton by Smulaton Analyss Choce of banks balance sheet varable and penalty rate BADLOAN1 Case 1 Constant penalty rate (%) Change of excess reserve by ncrease of call rate reducton of to 0.25% bad loans rato total by 50% Case 2 Varable penalty rate BADLOAN2 Case 3 Constant penalty rate Case 4 Varable penalty rate