1、Journal of the Japanese and International Economies 16, 289316 (2002)doi:10.1006/jjie.2002.0507Measuring the Liquidity Effect: The Case of JapanIichiro Uesugi1Ministry of Economy, Trade and Industry, 1-3-1 Kasumigaseki, Chiyoda-ku, Tokyo 100-8901, JapanE-mail: uesugi-iichirometi.go.jpReceived Octobe
2、r 23, 2000; revised January 29, 2002Uesugi, IichiroMeasuring the Liquidity Effect: The Case of JapanIn this paper, we measure the liquidity effect in Japan, complementing the work doneby F. Hayashi (2001, Int. Econ. Rev. 42, 287316) and compare it to the liquidity effectin the United States. Since i
3、nstitutional features are similar across these two countries,we apply J. Hamiltons (1997, Amer. Econ. Rev. 87, 8097; 1998, Carnegie-RochesterConf. Ser. Public Pol. 49, 144) methodology to estimate the liquidity effect for each dayof the maintenance period. Detailed daily data supplied by the Bank of
4、 Japan enable usto obtain more accurate estimates for Japan. Our key findings, which are not found inF. Hayashi (2001, Int. Econ. Rev. 42, 287316) are as follows: (1) On the final day of areserve maintenance period, both countries show the strongest evidence of the liquidityeffect, and (2) in both c
5、ountries the liquidity effect tends to be larger and more statisticallysignificant toward the end of the period. J. Japan. Int. Econ., September 2002, 16(3),pp. 289316. Ministry of Economy, Trade and Industry, 1-3-1 Kasumigaseki, Chiyoda-ku,Tokyo 100-8901, Japan.c2002 Elsevier Science (USA)Journal o
6、f Economic Literature Classification Numbers: E43, E44, E52.Key Words: liquidity effect; market operations; central bank; instrumental variables;monetary policy.1. INTRODUCTIONWhat is the response of the short-term interest rate to a change in the reservessupplied by the central bank? This relations
7、hip, between short-term interest rates1This paper is part of my Ph.D. dissertation at the University of California, San Diego. I am indebtedto James Hamilton for his constant help and advice and to the referees and editors for their helpfulcomments and suggestions. I am also grateful to Wouter den H
8、aan, Marjorie Flavin, Fumio Hayashi,Takeo Hoshi, Kazushige Kamiyama, Guy Yamashiro, participants in the workshops at UCSD, UCSanta Cruz, NBER Japan project meeting (informal macro-session), Yokohama National University,and the University of Tokyo for helpful comments, and to Shigeru Fujita for provi
9、ding the useful datafrom the Bank of Japan.2890889-1583/02 $35.00c2002 Elsevier Science (USA)All rights reserved.290 IICHIRO UESUGIand reserves, has been widely investigated and is known as the liquidity effect.Most researchers believe that there exists a negative relationship between reservesand th
10、e interest rate.The theoretical foundation for this belief was established by Grossman andWeiss (1983), Rotemberg (1984), Lucas (1990), and Fuerst (1992) by imposingrestrictions on the ability of agents to engage in certain types of financial transac-tions. Empirical evidence on the existence of the
11、 liquidity effect was provided byCochrane (1989) for a certain part of his sample period and Strongin (1995) forall the subperiods he analyzed. Alternatively, Leeper and Gordon (1992) reportedthat the response of interest rates to a money-growth innovation is positive. Paganand Robertson (1995) and
12、Christiano (1995) documented that the liquidity effectvanished after 1982. Hence, the literature, which has relied mainly on structuralVARs, has not provided a decisive answer as to the response of the short-terminterest rate to changes in reserves.A number of authors have viewed the structural VAR
13、approach as unconvincing,partly because of the reliance on questionable identifying assumptions.2Hamilton(1997, 1998) suggested an alternative method based on the clever use of instru-mental variables to identify an exogenous supply shock to reserves. Using dailydata, he then provided concrete evide
14、nce for the existence of an instantaneousliquidity effect in the United States.3Using similar variables to identify exogenoussupply shocks to reserves and employing intraday fluctuations of the interest rate,Hayashi (2001) supplied evidence for the liquidity effect in the Japanese market.4However, H
15、ayashis approach has several limitations, such as his liquidity effectbeing too narrowly defined, his failure to properly measure the interest rate differ-ential in response to exogenous shocks, and his lack of data on the final days ofthe maintenance period.Due to the approach taken in this paper,
16、our estimates of the liquidity effectare not hindered by the limitations faced by Hayashi. Our approach is based onthe belief that the similarities and differences in the effect across the countriesdepend on several institutional factors including the way in which each countrysets required reserves
17、and the timing of market operations by the central bank. Forthis cross-country comparison, even though there are many institutional differencesbetween the reserve systems and money markets of the United States and Japan,theoretically the basic mechanisms behind the liquidity effect are expected to b
18、ethe same in both countries. Two important common features in the reserve systemof the United States and Japan are:(1) multiple days in each maintenance period, and(2) a reserve requirement that must be satisfied by the end of a period.2See, for example, Cecchetti (1995), Rudebusch (1998), and Thorn
19、ton (1998).3Thornton (2000) was not only unable to replicate these findings using earlier and later FederalFunds data than those used in Hamilton (1997, 1998) but was also critical of Hamiltons methodologyto measure the liquidity effect. We will evaluate his statements in Section 5 to see if they ap
20、ply to Japan.4Hayashis research was implemented independent of this study.MEASURING JAPANS LIQUIDITY EFFECT 291These common features allow us to apply Hamiltons methodology to theJapanese short-term money market in order to obtain coefficients estimating theliquidity effect. The basic empirical resu
21、lts of this paper can be summarized asfollows: Evidence of the liquidity effect is found in Japan, especially on the finalday of a period.5 Days later in a period tend to have larger liquidity effects.The plan of the paper is as follows. Section 2 describes how the liquidity effectis generated. Sect
22、ion 3 presents the institutional background of the Bank of Japans(BOJs) reserve management. In Section 4, we look for instruments that satisfycertain requirements. In Section 5, we discuss and estimate the reserve equationsand the interest rate change equations. Section 6 then compares the liquidity
23、 ef-fects across the United States and Japan and discusses their causes. Section 7concludes.2. THE RESERVE SYSTEM AND THE LIQUIDITY EFFECTIn this section, we analyze how the liquidity effect is generated based on thecurrent reserve systems in the United States and Japan. We compare the systemsin Tab
24、le I. The original purpose of the reserve system was to ensure the solvencyof each private bank in order to protect depositors. Basically, reserves must beheld in the central bank in the United States and Japan, although some countriesdo not assign such obligations. The amount of reserves each bank
25、must maintainis determined proportionally to its deposit amount. This institutional scheme to-gether with the precautionary motive to avoid overdrafts generates the demand forreserves.Despite many institutional differences in their reserve systems, there are twokey features common to both the United
26、 States and Japan:(1) multiple days in each maintenance period(2) A reserve requirement that must be satisfied by the end of a period.Both in the United States and Japan, reserve requirements are maintained within2 weeks (U.S.) or a month (Japan) and how these requirements are achieved is left tothe
27、 discretion of each private bank. Because of these common features, a commonframework based on simple demand and supply schedules can be used to analyzeeither country. (See Fig. 1).On the final day of a period, private banks must satisfy their reserve requirementsand as a result may bid up the inter
28、est rate in order to acquire the necessary fundsto reach their respective requirements. Therefore, the demand schedule is highlyinelastic with respect to the interest rate (Fig. 1b). In this case, only the central5This is consistent with the results for the United States in Hamilton (1997).292 IICHI
29、RO UESUGITABLE IThe Reserve SystemsUnited States JapanMaintenance (M) and (For transactional deposits) C is half a month prior to Mcomputation (C) C is 2 days prior to Maperiod lag (For vault cash) C is14 days prior to MbLength of one maintenance 2 weeks (from Thursday to 1 month (from 16th of month
30、period (for transactional Wednesday 2 weeks later) to 15th of the next month)deposits)Treatment of vault cash Included in required reserves Not included in requiredreservesCarryover Possible (Up to 4% of Not allowedrequired reserves)Reserve ratio 310% (for transaction accounts) 0.051.2% (for time de
31、posits)0.11.3% (for others)Penalty rate for the shortage Discount rate+2% Discount rate+3.75%of reservesaFrom July 30, 1998, C is 16 days prior to M. This means that there are no overlapping days betweenC and M now. There is a complete lagged reserve system for the transactional deposits.bFrom July
32、30, 1998, C is completely overlapped with M.bank has the funds to fulfill the private banks need for reserves. Hence, the centralbank is able to charge any interest rate, provided that the private banks accept it.On any day other than the final day of a period, the demand schedule is veryelastic wit
33、h respect to the interest rate. Private banks form expectations about thefinal day interest rate. If the present interest rate were significantly lower than thefuture expected rate, banks would try to borrow nearly all of their required reservesfor the period, while if the present interest rate were
34、 significantly higher than thefuture expected rate, banks would attempt to lend out almost all of their reserves.FIG. 1. Demand and supply schedules in the reserve market.MEASURING JAPANS LIQUIDITY EFFECT 293Hence, we have a demand schedule that is nearly horizontal at the level of theexpected inter
35、est rate on days other than the final day (Fig. 1a).The demand curve is not perfectly vertical on the final day for two reasons:(1) Demand for excess reserves.Reserve requirements must be satisfied and private banks are penalized if theydo not satisfy the requirement. As a result, private banks want
36、 to hold the exactamount of required reserves since they earn no interest for any amount abovethe level of required reserves. However, in reality banks hold additional reservesin order to avoid overdrafts during a day, which in turn generates the settlementdemand for reserves.6In addition, it is alm
37、ost impossible for banks to reach theexact level of required reserves since reserves fluctuate in response to exogenousfactors such as deposit withdrawals, tax collections, and Treasury expenditures.The end result is that the demand for excess reserves is elastic with respect to theinterest rate.(2)
38、 Demand for allowances for the carryover of required reserves or excessreserves to the next period.In the United States, banks are allowed to carry up to 4% of the total requiredreserves into the next reserve maintenance period.7Either required reserves orexcess reserves can be carried over to the n
39、ext period. The demand for this carry-over is elastic with respect to the interest rate. Given higher interest rates, bankswill not have much more reserves than the lower bound of required reserves minus4% allowances. At lower interest rates, banks will have reserves closer to the upperbound of requ
40、ired reserves plus 4% allowances.Note also that the demand curve is not perfectly horizontal before the final day.One possible reason is risk aversion on the part of private banks. When privatebanks are risk averse, they find it necessary to hold some reserves even if thepresent interest rate rises
41、above the expected future rate.As can be seen in Fig. 1, the size of the interest rate decrease correspondingto the same unit increase in reserves supply differs according to whether or notwe are looking at the final day.8The degree to which the above factors affectthe slope of the demand curve play
42、s an important role in the liquidity effect.Based on these factors, we might expect the liquidity effect to be different acrosscountries.6Although the settlement systems of the United States (RTGS) and Japan (Net Settlement System)differ, overdrafts at the end of the day are penalized in both countr
43、ies.7In Japan, this carryover is not allowed.8There is a healthy literature in this area. Hamilton (1996) and Furfine (2000), for example, expect tohave different sizes in the liquidity effect across days in a maintenance period. For example, Hamiltondemonstrated that the fluctuations of the FF rate
44、 in a maintenance period can be simulated by weekendaccounting conventions, transaction costs, and limits on the degree to which banks can lend funds toone another in the funds market. Furfine finds that the predictable patterns in the level of interbankpayments match those found in the daily FF rat
45、e.294 IICHIRO UESUGITABLE IIThe Balance Sheet for the BOJ, April 30, 1999 (100 million yen)AssetsGold 4,239Coins 2,374Bills purchased 24,237Government securities in custody 19,882Government bonds 584,436Commercial bills discounted 116Loans 9,054Loans to deposit insurance corporation 63,914Foreign as
46、sets 38,656Deposits with agencies 9,808Cash collateral in exchange for government securities borrowed from 21,046financial institutionsOther accounts 13,301LiabilitiesBanknotes issued 529,325Current deposits held by the institutions with reserve obligations (Reserves) 38,362Current deposits held by
47、the institutions without reserve obligations 4,373Other deposits held by foreign central banks and others 132Government deposits 11,616Bills sold 115,044Government securities borrowed from financial institutions 19,882Other accounts 24,400Allowances 26,691Capital 1Legal reserves to cover possible lo
48、sses 21,326Total assets/liabilities 791,1523. INSTITUTIONAL BACKGROUNDThis section exploits the institutional characteristics of the reserve market to se-lect appropriate instrumental variables for estimating the liquidity effect. A goodinstrument should have a high correlation with the endogenous e
49、xplanatory vari-ables, but must be exogenous.9Our first task is to develop analogous instrumentsfor Japan by looking closely at the data. We must then check if the requirementsfor a valid instrument are satisfied.Table II shows the balance sheet for the BOJ. The variable of primary interestis reserves. As discussed in Section 2, a change in the quantity of these reserveswill result in a change in short-term interest rates. Many transactions affect thelevel of reserves. For example, when banknotes are brought to a bank as a deposit,the bank deposits a certain portion of
