Measuring the Adjusted Monetary Base in an Era of ... · (a) Treasury cash holdings 0.271 (b)...

1 The concept of high-poweredmoney used in this section isslightly broader than the con-cept used by Friedman andSchwartz (1963).

2 This concept also is widelyreferred to as the source base;see for example Andersen andJordan (1968). Our usage con-forms to the earlier practice oflabeling it “the monetarybase.”

FEDERAL RESERVE BANK OF ST. LOUIS

3

NOVEMBER/DECEMBER 1996

Measuring the AdjustedMonetary Basein An Era ofFinancialChangeRichard G. Anderson andRobert H. Rasche

he adjusted monetary base is an indexthat measures the effects on a centralbank’s balance sheet of its open

market operations, discount windowlending, unsterilized foreign exchangemarket intervention, and changes in statu-tory reserve requirements. Such an indexis important because the long-run path ofa monetary economy’s price level isprimarily determined by the path of thecentral bank’s balance sheet, adjusted forthe effects of changes in statutory reserverequirements.

The St. Louis

adjusted monetary baseequals the sum of the monetary (or source)base and the reserve adjustment magnitude(RAM). This article presents a revisedmeasure of the monetary base and a new RAM. The revised measure of themonetary base differs from previous mea-sures by including all Federal ReserveBank deposits held by domestic depositoryinstitutions; previous measures haveexcluded the aggregate amount of deposi-tory institutions’ required clearing balancecontracts with Federal Reserve Banks. Thenew RAM recognizes that, since the Mone-tary Control Act of 1980, an increasingproportion of depository institutions havenot significantly changed their demand forbase money (vault cash and deposits at

Federal Reserve Banks) relative to transac-tions deposits following changes instatutory reserve requirements. PreviousRAM adjustments have assumed thatdepository institutions would matchchanges in their statutory requiredreserves about dollar-for-dollar withchanges in their holdings of base money,following a change in reserve-requirementratios. The new RAM, constructed from fifteen years of weekly data on more than10,000 individual depository institutions,measures more precisely the change in the amount of base money demanded bydepositories following changes in reserve-requirement ratios than did previous RAMadjustments based on aggregate data.

THE REVISED MONETARYBASE

The measure of the monetary base thatwas published by the Federal Reserve Bankof St. Louis through September 1996included most, but not all, deposits at Federal Reserve Banks held by domesticdepository institutions. The new measure,presented in this article and published bythe Bank since October 1996, includes allsuch deposits. The revision increases thelevel of the base by an amount that variesfrom zero in 1980 up to about $6 billion in1994 and 1996.

Sources and uses of high-poweredmoney for the U.S. economy in December1995 are shown in Table 1.1 Most of thehigh-powered money supplied by the Fed-eral Reserve and the Treasury is repre-sented by currency in circulation and thedeposits of domestic financial institutionsat Federal Reserve Banks; together, theseconstitute the monetary base.2 The oldmeasure of the St. Louis monetary base(line 6) equals the sum of currency in cir-culation outside the Treasury and FederalReserve (line 6a) plus the “reservebalances” of depository institutions (line

Richard G. Anderson is an assistant vice president and economist at the Federal Reserve Bank of St. Louis. Robert H. Rasche is a professor ofeconomics at Michigan State University and a visiting scholar at the Federal Reserve Bank of St. Louis. We thank Cindy Gleit and Daniel Steinerfor excellent research assistance. We also thank the staff of the Division of Monetary Affairs, Board of Governors of the Federal ReserveSystem, for providing the data used in this article.

T


4


6b). Reserve balances, measured bysubtracting the aggregate amount ofdepository institutions’ required clearingbalance contracts from their aggregateFederal Reserve deposits, is an accountingconcept intended to measure the aggregateamount of high-powered money availableto support deposit expansion. Uses of high-powered money other than as the monetarybase, including their use to satisfy requiredclearing balance contracts, were about $25billion in December 1995 (line 7).

The new measure of the monetarybase (line 8) equals the sum of currency incirculation (line 8a) plus all FederalReserve deposits held by domestic deposi-tory institutions (line 8b). The newmeasure recognizes the similarity betweenthe Federal Reserve deposits classified as

reserves balances (line 6b) and those classified as held to satisfy requiredclearing balance contracts (line 7d). Bothcategories of deposits are used by deposi-tory institutions to settle interbank pay-ments, and both are available to satisfylegal reserve requirements (albeit perhapsat the cost of failing to satisfy a requiredclearing balance contract). Including inthe monetary base those Federal Reservedeposits putatively held to satisfy requiredclearing balance contracts increases theamount of Federal Reserve deposits in thebase by about one-fourth.

Our new measure of the monetarybase is suggested by the definition ofBalbach and Burger (1976):

... (the monetary base) can thereforebe identified in any monetary system

Table 1

Current and Revised Measures of the Monetary Base, December 1995*

Factors Supplying Base Money

(1) Reserve Bank credit(a)Securities held by the Federal Reserve 387.132(b) Loans to depository institutions 0.209(c) Federal Reserve float 1.223(d) Other Federal Reserve assets 32.212

Total Reserve Bank credit 420.776

(2) Gold stock 11.050(3) SDR certificates 10.168(4) US Treasury currency and coin outstanding 23.969Total supply of base money other than Reserve Bank credit 45.187

(5) Total supply of base money 465.963

* $ billions, not seasonally adjusted.Components may not add to totals due to rounding.

Source: Board of Governors of the Federal Reserve System.

Factors Using Base Money: CurrentMeasure of the Monetary Base

(6) The Monetary Base: Current Measure

(a) Currency and coin in circulation 419.615(b) Reserve balances of depository institutions at Federal Reserve Banks 20.402

Total monetary base 440.016

(7) Uses of base money other than as the monetary base

(a) Treasury cash holdings 0.271(b) Deposits of other than domestic financial institutions at Federal Reserve Banks 7.349(c) Other Federal Reserve liabilities and capital 12.841(d) Deposits, other than reserve balances, of domestic financial institutions at Federal Reserve Banks, including contractual amount of required clearing balances 5.487

Total other factors using base money 25.947

Factors Using Base Money: RevisedMeasure of the Monetary Base

(8) The Monetary Base: Revised Measure

(a) Currency and coin in circulation 419.615(b) Deposits of financial institutions at Federal Reserve Banks (revised measure) 25.888

Total monetary base 445.503

(9) Uses other than as the monetary base

(a) Treasury cash 0.271(b) Deposits of other than domestic financial institutions at Federal Reserve Banks 7.349(c) Other Federal Reserve liabilities and capital 12.841

Total other factors using base money (revised measure) 20.460


5

3 The revised measure of thebase, like previous measures,excludes Federal Reservedeposits held by the U.S.Treasury and by foreign centralbanks, included in lines 7b and9b of Table 1. These depositsare not used to make interbankpayments nor to dischargedebts of nongovernmentalunits; see Advisory Committeeon Monetary Statistics (1976).

4 Nonmember banks held smallamounts of Federal Reservedeposits before 1980, appar-ently used for clearing housepayments. The measure of themonetary base proposed byBrunner (1961) excludedthese deposits. The measuresconstructed by Cagan (1965)and Friedman and Schwartz(1963, appendix A) include(estimates of) these deposits.


by ascertaining and summing the following:1. those assets which the consolidated

banking sector uses to settle interbank debt; and

2. those items, aside from bank liabilities, which are used as money;

and by the definition of the Advisory Com-mission on Monetary Statistics (1976, p. 8):

With respect to monetary aggregates,one basis for defining such a total is to regard money as corresponding toassets that are generally used to dis-charge obligations and that are not theexplicit liability of nongovernmentalentities in the society. Traditionally,such assets have corresponded tospecie. In the United States today they correspond primarily to the non-interest-bearing fiat issues of theultimate monetary authority. Theterms “high-powered money” and“monetary base” have been used torefer to this total. We shall refer to it as “the base.”For the United States today the baseincludes all currency outside the Fed-eral Reserve and the Treasury plus allbank deposits at Federal ReserveBanks.

Although broader than the oldmeasure it replaces, the new measure ofthe monetary base excludes an importantasset that these definitions suggest shouldbe included: the amount of intraday credit,in the form of Federal Reserve deposits,used by banks for interbank payments.During 1994, such intraday deposits aver-aged approximately $50 billion, or nearlytwice the close-of-business-day amount of Federal Reserve deposits included in the monetary base (see Richards, 1995, p. 1066). The major barrier to inclusion ofintraday deposits is the lack of timely pub-lished data: close-of-business depositlevels are published weekly on the Boardof Governors’ H.4.1 statistical release,while intraday credit is not published inany release.3

The need to revise the measure of themonetary base arises from changes in U.S.

financial markets since the Monetary Con-trol Act of 1980. The Act significantlychanged the demand for Federal Reservedeposits. Prior to the Act, almost alldeposits at Federal Reserve Banks wereheld by member banks of the FederalReserve System.4 Banks used thesebalances both to satisfy reserverequirements and to make payments onbehalf of customers. For most memberbanks, the latter came “free”: the amountof reserves that they were required to holdagainst deposits was more than sufficientto satisfy any demands arising from inter-bank payments (perhaps with someintraday Federal Reserve overdraft credit).Nonmember banks and thrifts, lackingaccess to the Federal Reserve’s books forfinal settlement of payments, madeinterbank payments and settled checksthrough correspondent accounts atmember banks.

The Monetary Control Act made non-member institutions subject to FederalReserve System reserve requirements and,at the same time, gave them direct accessto the payments system through depositsat the Federal Reserve Banks (see Gilbertand Summers, 1996). Because the newreserve requirements were phased in forthese institutions over an eight-yearperiod, many initially found their vaultcash more than sufficient to satisfy therequirements. Holding only small amountsof Federal Reserve deposits, some institu-tions found that overdrafts on FederalReserve accounts became a problem (seeFederal Reserve Bulletin, March 1981, pp. 247-49 and December 1982, p. 756).As a result, during the early 1980s theFederal Reserve required some depositoryinstitutions that used Federal Reserve pay-ments services to maintain so-calledrequired clearing balances, or levels of Fed-eral Reserve deposits above and beyondthe amounts necessary to satisfy the insti-tutions’ statutory reserve requirements. Tooffset the cost of holding these balancesand make the requirement more palatable,the Federal Reserve paid the institutions,at (approximately) the federal funds rate,“earnings credits” that could be used only


6

5 Contrary to the terminology inthe quotation, required clearingbalances are not (and neverhave been) included in pub-lished Federal Reserve data onreserve balances.

6 There is one exception to thisstatement. A nonmemberdepository institution may havea separate clearing balancedeposit account at a FederalReserve Bank if it satisfies itsrequired reserves via a pass-through contract with anothereligible depository institution.

7 See Feinman (1993) andHilton, Cohen and Koonmen(1993).


to defray charges for Federal Reservepriced services such as check clearing andwire transfers.

A clear statement of the rationale forthe exclusion of required clearing balancesfrom the current measure of the monetarybase is provided by Gilbert (1983), p. 23n:

Depository institutions maintainclearing balances at Federal ReserveBanks as a means of payment for thefees Federal Reserve Banks now chargefor services. Depository institutionsreceive implicit interest on theirclearing balances at the federal fundsrate, which may be used to pay thefees on services. Required clearing bal-ances are subtracted in computing the[monetary] source base becauseclearing balances are part of totalreserve balances held by depositoryinstitutions at Federal Reserve Banks,but are not related to the levels ofdeposit liabilities.5

Perhaps reasonable at the time, therequired clearing balance contract hasevolved into a flexible, voluntary tool ofdepository institution cash management.Although the contract obliges a depositoryinstitution to maintain a larger FederalReserve deposit than is necessary to satisfyits required reserves, the deposit is neithera distinct type nor separate category ofdeposit: all of the funds are available tosettle interbank payments and may be con-verted to vault cash if necessary.6

During the mid-1980s, and especiallyfollowing the February 1984 shift to con-temporaneous reserve accounting, anincreasing number of institutions realizedthat they could simplify their reserve man-agement by voluntarily agreeing tomaintain a required clearing balance.7

Maintaining a required clearing balancechanges the expected cost to thedepository of satisfying its statutoryreserve requirements because theadditional Federal Reserve depositsprovide an inexpensive cushion againstcostly shortfalls relative to statutoryreserve requirements. Deficiencies relativeto the contracted clearing balance imposelittle cost on the institution, while permit-

ting it to use all its Federal Reservedeposits to satisfy its statutoryrequirements. At the same time, theFederal Reserve deposits used to satisfythe clearing balance contract accumulateearnings credits at about the federal fundsrate.

The Federal Reserve deposits held tosatisfy a required clearing balance contractact as a buffer stock relative to the depositsneeded to satisfy statutory reserve require-ments because, under Federal Reserveaccounting rules, balances in a depository’sFederal Reserve account are applied first tosatisfy its statutory required reserves andonly thereafter to satisfy the clearingbalance requirement. Hence, when aninstitution’s Federal Reserve depositbalance falls below its expectation, theshortage is recorded in the FederalReserve’s accounting system as a deficiencyon a clearing balance requirement ratherthan as a deficiency on a statutory reserverequirement (provided the sum of vaultcash and Federal Reserve deposits exceedsthe institution’s required reserves).

No penalties are imposed for smalldeficiencies on voluntary clearing balancecontracts, and larger shortfalls arepenalized at only a 2 or 4 percent annualinterest rate (see Stevens, 1993). Deficien-cies relative to required reserves aresubject to significant penalties and“administrative counseling,” while compa-rable deficiencies relative to a clearingrequirement are subject to minimal penal-ties. An institution that sometimes hasbeen forced to borrow at either thediscount window or a penalty federalfunds rate to cover reserve deficienciesmay find the required clearing balanceaccount comforting.

By 1985, about 4500 institutions hadclearing balance contracts, totalingapproximately $1.2 billion (Figure 1).These numbers were about the same in thethird quarter of 1990, before the December1990 reduction in reserve requirements onnonpersonal time deposits and certainother liabilities. Two years later, during thethird quarter of 1992, the amount of con-tracted required clearing balances had


7

nearly tripled to approximately $4.5billion while the number of institutionshad increased to about 4700.8

In summary, proper measures of themonetary base should include depositsheld at Federal Reserve Banks to satisfyrequired clearing balance contracts, forseveral reasons:

• First, the size of a contract, and hencethe amount of additional FederalReserve deposits that must be held tosatisfy the contract, is determined by adepository institution’s asset andliability management strategy, not byany regulation. An institution maychange the size of its clearing balancecommitment when it desires,appropriate to its business needs.9

“Required” clearing balance contractsare voluntary commitments.

• Second, the Federal Reserve depositsheld to satisfy a clearing balance con-tract are available to settle interbankpayments in the same way as otherFederal Reserve deposits. Thesedeposits are maintained in the sameFederal Reserve account that containsdeposits used to satisfy statutoryreserve requirements.

• Finally, the Federal Reserve depositsused to satisfy required clearingbalance contracts are supplied by theFederal Reserve, through actions suchas open market operations, in the sameway as other high-powered money.They are not a distinct type of funds.

Including the amount of requiredclearing balance contracts in the monetarybase is not without objection. Some depos-itory institutions seem to adjust the size oftheir contract inversely to changes in thefederal funds rate, seeking perhaps to gen-erate only enough earnings credits to payfor their use of the Federal Reserve’s pricedservices. For these institutions, thedemand for Federal Reserve deposits maybe highly interest-elastic and largely unre-lated to either liquidity management orlending decisions. If so, argue someanalysts, required clearing balances shouldbe excluded from the monetary base.

Relatively simple macroeconomicanalysis shows that this argument has noimplications for definition or measurementof the adjusted monetary base. In previousarticles, we have explored the dependenceof most money multiplier components,such as k and e, on economic variablessuch as interest rates and income(Anderson and Rasche, 1982; Anderson,Johannes and Rasche, 1983). Although itseems likely that including requiredclearing balances in the monetary base willincrease the interest elasticity of the excessreserve ratio, e, this increase has no impli-cations for the importance of the adjustedbase as a policy indicator in models wherechanges in the adjusted monetary base aretransmitted to the economy solely throughchanges in a monetary aggregate, M. Inthese models, the role of the adjustedmonetary base as an indicator of the stanceof monetary policy is independent of thesize of the elasticity of multipliercomponents such as k (the nonbank pub-lic’s desired ratio of currency to checkabledeposits) and e (depository institutions’desired ratio of excess reserves tocheckable deposits) with respect to

8 The actual amount of FederalReserve deposits used to satisfydepositories’ clearing balancecontracts is not available. Thesedata are the nominal contractedamounts.

9 However, Federal Reserve oper-ating rules generally discouragechanges more frequently thanonce a month, or approximate-ly every third or fourth mainte-nance period.


Required Clearing Balance ContractAmounts

February 1994, Increase in federal funds rate target7

Jan 80 Jan 85 Jan 90 Jan 95 Jan 00

Billio

ns of

Doll

ars

0

1

2

3

4

5

6

April 1992, Reduction inreserve requirements

April 1995, Nationwide spreadof OCD sweep programs begins

December 1990, Reduction inreserve requirements

Figure 1


8

10 In this article, we make anadditional change to the Burgerand Rasche methodology. Theyproposed that a single RAMadjustment be used in creatingthe adjusted monetary base.We adopt the proposal ofTatom (1980), who observedthat the RAM adjustment wasappropriate only for changes inreserve-requirement ratios with-in a given overall structure ofrequirements.


variables such income and interest rates.For further discussion, see Anderson andRasche (1996b).

Comparison of U.S. and foreignpayments systems reinforces our decisionto include the aggregate amount ofrequired clearing balance contracts in themonetary base. In some countries, deposi-tory institutions’ need to settle interbankpayments in central bank deposits hasbeen cited as a justification for effective,binding statutory reserve requirements(Deutsche Bundesbank, 1994, pp. 70-80).In countries without statutory reserverequirements, all central bank deposits heldby depository institutions are clearing bal-ances; yet, such deposits play an importantrole in the central bank’s monetary policy(see Bank of Canada, 1987, 1989, 1991;Bank for International Settlements, 1993).

ADJUSTING THE MONETARYBASE FOR CHANGES INRESERVE REQUIREMENTS

The adjusted monetary base combinesin a single index three Federal Reserveactions that affect the supply of the mone-tary base—open market operations,discount window lending, and unsterilizedforeign exchange market intervention—with changes in statutory reserverequirements that affect depository institu-tions’ demand for the monetary base.Hence, measuring the adjusted monetarybase requires a mechanism that can trans-late changes in the demand for base moneydue to changes in statutory reserverequirements into equivalent changes inthe supply of base money due to openmarket operations or similar actions. Thereserve adjustment magnitude, or RAM,provides such a mechanism by measuringthe amount by which the aggregate quan-tity of base money demanded bydepository institutions during any givenperiod has been affected by changes inreserve requirements, relative to a selectedbase period.

Combining the effects of changes instatutory reserve requirements with thosefrom open market operations and similar

instruments depends, implicitly or explic-itly, on a model of depository institutions’demand for base money. Absent a fullyworked out model of bank liquidity man-agement and reserve demand, we usestatistical tests to identify the characteris-tics of depositories that likely have, andhave not, responded to changes in reserverequirements since 1980. Only the formerset of institutions is included in our newRAM adjustment.

The original St. Louis adjusted mone-tary base published by Andersen andJordan (1968) included an adjustment for“reserves released by changes in reserverequirements.” The adjustment,constructed as suggested by Brunner(1961), added to the monetary base ateach date the cumulative dollar amount bywhich past changes in reserverequirements had changed the level ofrequired reserves. Although each change inreserve requirements was viewed asabsorbing or liberating a certain dollaramount of required reserves, theseamounts depended only on the amount ofreservable deposits on the date of thereduction: They did not vary in laterperiods with changes in the levels ofreservable deposits.

In 1977, Burger and Rasche (1977)showed that Brunner’s adjustment for theeffects of reserve-requirement ratiochanges was inadequate because it did notconsider the amount by which pastchanges in reserve-requirement ratiosaffected banks’ current required reserves.They showed that a suitable adjustmentmust vary with current deposit levels if itis to remove the total effect of the changein reserve-requirement ratios from themonetary base multiplier (and no more).They proposed that the adjusted monetarybase be measured as the sum of the mone-tary (source) base and a time-varyingreserve adjustment magnitude (RAM), amethodology that has generally beenmaintained in subsequent revisions of theSt. Louis adjusted monetary base.10

Substantial changes in the structure ofreserve requirements since 1980 suggest areexamination of the Burger and Rasche


9


methodology. We formalize their analysisby considering a model with two classes of institutions. Those in the first classresemble the member commercial banksconsidered by Burger and Rasche: Theirrequired reserves exceed their vault cash,and they must hold deposits at FederalReserve Banks to satisfy the balance oftheir statutory requirements. As a result,statutory reserve requirements play animportant role in determining their demandfor base money. Institutions in the secondclass find legal reserve requirements muchless influential in their portfolio allocationdecisions: Their level of required reservesis less than their vault cash, and they neednot hold any deposits at Federal ReserveBanks to satisfy statutory requirements.The demand by these institutions fordeposits at Federal Reserve Banks dependslargely on their need to make interbankpayments in immediately available fundson the books of the Federal Reserve Banks,and perhaps on Federal Reserve restric-tions regarding daylight overdrafts.11

Monetary Base Multipliers before1980

We begin with a model that reflectsthe institutional environment before theMonetary Control Act of 1980. Since ourpurpose is to illustrate the dependence ofthe RAM adjustment on the distribution ofdeposits among different classes of deposi-tory institutions, we separate member andnonmember banks more explicitly thanhave previous authors.12 We assume (1) a central bank that issues twoliabilities, currency, Cu, and reservebalances (that is, deposits at the centralbank), RB, and (2) two types of depositoryinstitutions, indexed by superscripts Mand N (corresponding to member banks,and to non-member banks and thrifts,respectively), that issue demand D =DM

1DN and time T = TM1TN deposits. Thetwo types of depositories are dissimilar infour characteristics:

• Type M institutions are subject to cen-tral bank reserve requirements againstdeposits that may be satisfied by

holding either vault cash or deposits atthe central bank.

• Government deposits are only at typeM institutions.

• Type N institutions hold deposits attype M institutions but not vice versa.

• Type N institutions are not permittedto hold deposits at the central bank.

They are similar in two other ways:

• Both types of institutions hold vaultcash to satisfy reserve requirementsand/or to convert deposits intocurrency on demand.

• Both types of institutions issuedeposits that the nonbank publicregards as perfect substitutes.

We assume that transactions amongbanks, the government, and the nonbankpublic are settled in terms of currency, Cu,demand deposits held by the governmentat type M depositories, , demanddeposits held by the nonbank public attype M and N depositories, and ,respectively, and demand deposits held bytype N at type M depositories, .(Throughout, superscripts refer to theowner of the deposit and subscripts to theissuer of the deposit.) Define

and ,

and note that

.

The Federal Reserve imposes reserverequirements against demand DM and timeTM deposits at rates rD and rT, respectively,such that the required reserves of a type Minstitution are RR = rDDM1rTTM.13

The monetary base multiplier in thismodel is easily derived. Suppressing timesubscripts, the monetary base is by defini-tion

MB = RB1Cu

= rDDM1rTTM1k DP1VCN1

(VCM1RB2rDDM2rTTM)

= (rDdM1rTtM1k1vN1eM)DP,where

(rDDM1rTTM) are the required reserves of type M institutions,

D D D D DPM N

PMG

MN= + − −

D D D DM MP

MG

MN= + +D D DP

MP

NP= +

DMN

DNPDM

P

DMG

11 See for example Hancock andWilcox (1996).

12 Burger (1971) provides a simi-lar analysis without as explicit aseparation of different classesof institutions.

13 Historically, some nonmemberbanks and thrifts faced state-imposed reserve requirementsthat had to be satisfied withholdings of vault cash, depositsin other banks, U.S Treasurybills or certain other liquid secu-rities. See Gilbert (1978),Gambs and Rasche (1978),and Gilbert and Lovati (1978).


10

k DP is currency held by the nonbankpublic, VCM and VCN are vault cash held bytype M and N institutions, respectively,RB are the deposits held by type Minstitutions at the central bank,

and .

The term (VCM1RB2rDDM2rTTM)equals the amount of high-poweredmoney, vault cash plus reserve balances,held by type M institutions above andbeyond their required reserves.

For clarity, it may be useful to relatethese reserve constructs in our model tothose currently published. The FederalReserve Board’s reserve measures differfrom those of the Federal Reserve Bank ofSt. Louis by excluding surplus vault cashfrom the definition of excess reserves. Foran individual depository institution i,

• if VCi,M.(rDDi,M1rTTi,M), then thedifference (VCi,M2rDDi,M1 rTTi,M) isreferred to as surplus vault cash. IfVCi,M#(rDDi,M1 rTTi,M), then surplusvault cash is zero.

• If VCi,M.(rDDi,M1rTTi,M) and RBi,M.0,then RBi,M2RCBi,M is referred to asexcess reserves, where RBi,M representsthe total reserve balances held bydepository institution i at the FederalReserve and RCBi,M is the amount of itsrequired clearing balance contract (ifany). Note that RCBi,M may be zero,and RBi,M2RCBi,M may be negative. IfRBi,M2RCBi,M = 0, then excess reservesequal zero even though surplus vaultcash is greater than zero.

• If VCi,M,(rDDi,M1rTTi,M), then RBi,M2RCBi,M2(rDDi,M1rTTi,M2VCi,M) is excessreserves.

Measures of aggregate total and excessreserves published by the Board of Gover-nors of the Federal Reserve System omitsurplus vault cash and an amount of Fed-eral Reserve deposits equal to depositoryinstitutions’ required clearing balance con-

tracts.14 The old reserve measures pub-lished by the Federal Reserve Bank of St.Louis through September 1996 (discussedin the previous section of this paper)included surplus vault cash but omittedrequired clearing balance contracts.

The monetary base multiplier for M1is straightforward:

,

and for M2,

,

where

The reserve adjustment to the mone-tary base, RAM, maps the effects of changesin reserve-requirement ratios into equivalentchanges in the monetary base. The effect ofthis mapping is to define new adjusted-monetary-base multipliers for the monetaryaggregates M1 and M2 that are invariant tochanges in the reserve-requirement ratios rD

and rT, denoted as m1b(rD,rT,k,eM,t,tM,dM,vN)and m2b(rD,rT,k,eM,t,tM,dM,vN), respectively.At the same time, the adjustment shouldnot change the response of these multi-pliers to arguments other than the reserve-requirement ratios rD and rT, when comparedto the responses of m1b(rD,rT,k,eM,t,tM,dM,vN)and m2b(rD,rT,k,eM,t,tM,dM,vN).

Letting the adjusted monetary base bedefined as AMB = MB + RAM, we have:

M1 = m1b(MB1RAM) = m1bAMB

where

and mk

r d r t k v eb D

MT

M N M1

0 0

1=

+( )+ + + +

,

tM M

D

T T

DP

M N

P=

−( )=

+( )2 1.

M Cu D T k t D

k t

r d r t k v eMB

m r r k e t t d v MB

P P

DM

TM N M

D TM M M N

2 1

1

2

= + + = + +( )=

+ +( )+ + + +

= ( ), , , , , , ,

M Cu D k D

k

r d r t k v eMB

m r r k e t t d v MB

P P

DM

TM N M

D TM M M N

1 1

1

1

= + = +( )=

+( )+ + + +

= ( ), , , , , , ,

eVC RB r D r T

DM

MD

MT

MP

= + − −

dD

Dt

T

Dv

VC

DM

MP M

MP N

NP

= = =, , ,


14 See Table 1.20 in the FederalReserve Bulletin or the Board’sweekly statistical releaseAggregate Reserves ofDepository Institutions and theMonetary Base. (H.3)


11


and are the reserve requirementratios on transaction and time deposits in a chosen base period, respectively. Thereserve adjustment magnitude,

,

maps the change in required reserves dueto any change in reserve requirementratios since the specified base period intoan equivalent change in the monetarybase.

Similarly, for M2 we have

M2 = m2bAMB = m2b(MB1RAM).

Noting that

,

then

,

which is invariant to changes in the legalrequired reserve ratios rD and rT.

The above analysis may be extended tothe case where type M institutions (thosesubject to central bank reserve require-ments) issue i = (1,...,I) classes of trans-action deposits and j = (1,...,J) classes oftime deposits, each with possibly differentreserve requirement ratios.

At type M institutions, let

and

denote the ratios of the nonbank public’sholdings of demand and time deposits inthe ith and jth reserve classifications, respec-tively, to their total holdings of demanddeposits, DP. Then aggregate requiredreserves are15

,

and the monetary base multiplier is

Note the presence of base-period re-serve-requirement ratios for each class ofdeposit, and , in the denominator.The corresponding reserve adjustmentmagnitude is

Finally, as a caveat and extension toearlier remarks, note that the RAM adjust-ment does not make all money and creditmultipliers invariant to changes instatutory reserve-requirement ratios. Con-sider the monetary-base multiplier forbank credit (BC), or mBC , defined byBrunner and Meltzer (1968) as BC =mBCMB. In our notation,

[see Brunner and Meltzer (1968), equationA.8, p. 32]. Let the adjusted-monetary-basemultiplier for bank credit be defined as

BC = mBCbAMB = mBCb[MB1RAM].

Then:

,

so

which is not invariant to change in legalrequired reserve ratios rD and rT. Thus, inmodels in which intermediated (bank)credit provides a channel of monetarypolicy independent of that provided bymonetary aggregates, the adjustedmonetary base defined above is not an ade-

mt r d r t e v

r d r t k e vBCb

tD

m tT

m m N

Dm

Tm m N

=+( ) − + + +( )

+ + + +

1

0 0

,

1

0 0

+( ) − + + +( )

=

+ + +[ ] +

−( ) + −( )

t r d r t e v D

m r d r t e v D

m r r d r r t D

tD

m tT

m m NP

BCb tD

m tT

m m NP

BCbD

tD

mT

tT

mP

mt r d r t e v

r d r t k e vBC

tD

m tT

m m N

tD

m tT

m m N

=+( ) − + + +( )

+ + + +

1

RAM r r

r r D

MiD

MiD

Mi

i

I

MjT

MjT

Mj

j

JP

= −( ) +

−( )

=

=

∑

∑

0

1

0

1

δ

τ .

rMjT

0rMiD

0

mk

r r k v e

b

MiD

Mi MjT

Mj N M

j

J

i

I1

0 0

11

1= +

+ + + +

==∑∑ δ τ

.

r r DMiD

Mi MjT

Mj

j

J

i

IPδ τ+

==∑∑

11

τ MjMj

P

T

D=δMi

MiP

D

D=

mk t

r d r t k v eb D

mT

m N m2

0 0

1=

+ +( )+ + + +

1

2

2 0 0

+ +[ ] =

+ + + +[ ] +

−( ) + −( )

k t D

m r d r t k v e D

m r r d r r t D

P

b tD

m tT

m N mP

bD

tT

mT

tT

mP

RAM r r d r r t DDtD

MT

tT

MP= −( ) + −( )

0 0

rT0rD

0

15 Recall that only type M institu-tions are subject to statutoryreserve requirements in thismodel.


12


quate indicator variable for the stance ofmonetary policy.

Adjusting for Reserve RequirementChanges with “EconomicallyNonbound” Institutions

For periods prior to late 1980, deposi-tory institutions are easily separated intotwo groups based on their holdings of basemoney: Member banks held vault cash,were subject to Federal Reserve Systemreserve requirements, and generally helddeposits at Federal Reserve Banks;nonmember banks and thrifts held vaultcash and were not eligible to hold depositsat Federal Reserve Banks.

Studies of the adjusted monetary baseprior to 1980 generally assumed thatmember banks changed their holdings of base money about dollar-for-dollar fol-lowing changes in their required reservesdue to changes in reserve-requirementratios. During that period, member banksheld few excess reserves, and most bankslikely faced reserve-requirement ratios suf-ficiently high to constrain their portfolioallocation decisions.16 The use of FederalReserve deposits to settle interbankpayments was little discussed. The generalassumption among economists andbanking analysts seemed to be that eitherbanks’ deposits at Federal Reserve Bankswere more than adequate to absorb debitsand originate payments, or that the Fed-eral Reserve would supply adequateintraday credit. Banks also had verylimited ability to alter their demand forbase money by inducing customers to shiftfunds from transaction to nontransactiondeposits. Time deposits, with the exceptionof large negotiable certificates of deposit,were subject to effective Regulation Qinterest rate ceilings. Overall, both banksand their customers likely were sufficientlyconstrained that other multiplier compo-nents (such as the ratio of time and savingsdeposits to transaction deposits, or t) wereunaffected by changes in legal reserve-requirement ratios on deposits.

Under this regime, the total amount ofbase money demanded by depositories

equaled the sum of member banks’required and excess reserves, plus thevault cash held by other depositories.Because member banks applied essentiallyall their vault cash to satisfy reserverequirements, and because requiredclearing balances were approximately zero,excess reserves at member banks equaledthe difference between their deposits atFederal Reserve Banks (denoted as RB)and the portion of their required reservesnot satisfied by vault cash:

ERM = RBM2(rDDM1rTTM2VCM).

(Note that RB denotes both the reservebalances and total Federal Reserve Bankdeposits held by member banks; withoutrequired clearing balance contracts, theseare exactly the same.) Excess reserves forthe banking system as a whole equaled thesum of excess reserves at member banks plusvault cash at nonmember banks, VCN. Theaverage aggregate excess reserve ratio was

Today’s environment is considerablydifferent. The Monetary Control Actextended reserve requirements to alldepository institutions, reduced to zerorequired reserves on savings and personaltime deposits, and significantly reducedother reserve requirements on memberbanks. During December 1990 andJanuary 1991, reserve-requirement ratioson nonpersonal time deposits and Euro-dollar liabilities were reduced to zero forall depository institutions. In April 1992,reserve-requirement ratios on transactiondeposits were reduced to 10 percent from12 percent.17 Depository institutions alsogained greater freedom to adjust their mixof reservable and nonreservable depositsduring the 1980s, following the end ofRegulation Q ceilings on deposit offeringrates.

Following implementation of theMonetary Control Act, many depositoryinstitutions found that their vault cash,although largely held for retail businessreasons, also satisfied their reserve require-ments.18 In the Federal Reserve System,depository institutions that fully satisfy

e e vER VC

DM N

M NP

= + = +.

16 In 1977, required reserveratios at member banks rangedfrom a minimum of 7 percenton the first two million of netdemand deposits to 16.25 per-cent on net demand deposits inexcess of 400 million dollars.The required reserve ratio onsavings deposits was 3 percentand the reserve requirementson time deposits maturing inless than 180 days were 3 per-cent on the first $5 million and6 percent on time deposits inexcess of $5 million. (FederalReserve Bulletin, December,1977, p. A9)

17 In 1995, the reserve require-ment ratio on the first $3.8million of net transactiondeposits was zero (the reserveexemption amount), and only3 percent on the next $51 mil-lion (the low reserve tranche).The cutoff for the low reservetranche is changed annually.

18 Reserve requirements wereincreased from zero on all non-member depository institutions.The full imposition of reserverequirements on these institu-tions was phased in over theperiod between 1981 and1987.


13


their required reserves with vault cash areknown as “nonbound” institutions; otherinstitutions are known as “bound” institu-tions. In this article, we refer to theseinstitutions as L-Nonbound and L-Bound,respectively. Table 2 shows the percentagedistribution of L-Bound and L-Nonbounddepository institutions among depositoriesreporting data to the Federal Reserve forselected years from 1981-95. (The rowslabeled E-Bound and E-Nonbound areexplained later.) Part A of the tableincludes only institutions that reporteddata weekly, while Part B includes institu-tions that reported quarterly andannually.19 In mid-1983, after the initialphase-in of the Monetary Control Act,about 40 percent of the total depositsreported by weekly-reporting institutionswas at L-Nonbound institutions; for allreporting institutions, shown in Part B,about 43 percent of deposits was at L-Non-bound institutions. By mid-1989, theseproportions had fallen to approximately 18percent (in Part A) and 25 percent (in PartB). The 1990-91 reduction in reserve-requirement ratios increased the propor-tion of total deposits at L-Nonboundweekly reporting institutions to about 27percent in 1991.

We regard L-Nonbound institutions asfacing no effective reserve requirement–related constraint because they seemunlikely to change the asset mix of theirportfolios following a change in statutoryreserve requirements. L-Nonbound institu-tions satisfy their required reserves withvault cash, and we assume that their hold-ings of vault cash are determined almostentirely by business needs, rather than bystatutory reserve requirements. As a result,these institutions seem unlikely to changetheir portfolio mix of assets in response toa change in reserve-requirement ratios.

Vault cash held by L-Nonbounddepository institutions in excess of theirrequired reserves is known as “surplusvault cash.” Surplus vault cash is “surplus”only in the sense that some part of thebank’s vault cash is not used to satisfystatutory reserve requirements. Nostatutory requirement determines a deposi-

tory’s vault cash; these amounts are volun-tarily chosen by the institutions’ managers.As such, they presumably reflect anticipatedbusiness of the institution, and hence arenot surplus in the economic sense of indi-cating a portfolio disequilibrium. Surplusvault cash is included in both the old andnew St. Louis’ measures of the adjustedmonetary base, and in the Board of Gover-nors’ measure of the monetary base, notadjusted for reserve requirement changesand not seasonally adjusted. It is excluded,however, from the Board of Governors’measure of the monetary base, adjusted forchanges in reserve requirements.

Historical data on surplus vault cashare shown in Figure 2 (see page ••).Before 1959, vault cash could not be usedto satisfy reserve requirements, so all vaultcash was surplus. Surplus vault cashdecreased sharply during 1959-60 whenFederal Reserve member banks were grad-ually allowed to apply vault cash towardsatisfying required reserves. (The fractionof vault cash eligible to satisfy requiredreserves increased linearly at the rate ofone-twelfth per month, reaching 100 per-cent of vault cash in December 1960.)From 1961 to 1981, surplus vault cashequaled the vault cash held by nonmemberbanks and thrift institutions, sincemember banks applied virtually all theirvault cash to satisfy reserve requirements.Surplus vault cash grew rapidly during the1970s as the fraction of banks that weremembers of the Federal Reserve Systemdeclined. Although the Monetary ControlAct extended reserve requirements to alldepository institutions, the requirementswere phased in during 1980-1987. Duringthese years, surplus vault cash generallydecreased. During the later 1980s, surplusvault cash remained relatively constant butexhibited substantial seasonal variation.

Although the vault cash holdings of L-Bound institutions are less than theirrequired reserves, some of these institu-tions also might not be constrained bylegal reserve requirements. For some ofthese institutions, especially smaller ones,statutory reserve requirements might notbe an important factor in their portfolio

19 For weekly-reporting institu-tions, the data are the firstreserve maintenance periodthat begins and ends in July ofthe specified year. June dataare used for quarterly andannual reporters.



14

Statistics on Legally and Economically Bound and Nonbound DepositoryInstitutions, Selected Years

A. Weekly Reporting Depository Institutions (Federal Reserve FR2900 report)

1983 1985 1987 1989 1991 1993 1995

Distribution of Reporting Depository Institutions, by reserve status (percent of reporting institutions)L-Bound 33.7 33.6 45.4 47.5 27.4 34.4 30.2L-Nonbound 66.3 66.4 54.6 52.5 72.6 65.6 69.8E-Bound 33.7 33.6 45.4 47.5 5.9 6.0 5.5E-Nonbound 66.3 66.4 54.6 52.5 94.1 94.0 94.5

Distribution of Total Deposits, by reserve status (percent of total deposits of weekly reporters)L-Bound 60.4 70.3 79.3 81.6 72.9 78.2 75.1L-Nonbound 39.6 29.7 20.7 18.4 27.1 21.8 24.9E-Bound 60.4 70.3 79.3 81.6 54.8 57.7 56.3E-Nonbound 39.6 29.7 20.7 18.4 45.2 42.3 43.7

Distribution of Net Transactions Deposits, by reserve status (percent of aggregate net transactions deposits of weekly reporters)L-Bound 74.6 78.9 85.1 86.1 80.0 85.3 82.1L-Nonbound 25.4 21.1 14.9 13.9 20.0 14.7 17.9E-Bound 74.6 78.9 85.1 86.1 64.5 67.7 66.6E-Nonbound 25.4 21.1 14.9 13.9 35.5 32.3 33.4

Distribution of Required Reserves, by reserve status (percent of aggregate required reserves of weekly reporters)L-Bound 92.0 93.0 94.9 95.2 93.1 94.8 91.8L-Nonbound 8.0 7.0 5.1 4.8 6.9 5.2 8.2E-Bound 92.0 93.0 94.9 95.2 82.8 83.1 81.2E-Nonbound 8.0 7.0 5.1 4.8 17.2 16.9 18.8

B. Statistics on the Sum of Weekly, Quarterly and Annual Reporting Depository Institutions

1983 1985 1987 1989 1991 1993 1995

Distribution of Reporting Depository Institutions, by reserve category (percent of reporting institutions)L-Bound 19.1 19.8 22.1 19.8 11.1 14.4 11.7L-Nonbound 80.9 80.2 77.9 80.2 88.9 85.6 88.3E-Bound 19.1 19.8 22.1 19.8 2.4 2.5 2.1E-Nonbound 80.9 80.2 77.9 80.2 97.6 97.5 97.9

Distribution of Total Deposits, by reserve category (percent of reported total deposits)L-Bound 56.9 66.8 73.9 75.5 67.1 72.1 69.1L-Nonbound 43.1 33.2 26.1 24.5 32.9 27.9 30.9E-Bound 56.9 66.8 73.9 75.5 50.4 53.2 51.8E-Nonbound 43.1 33.2 26.1 24.5 49.6 46.8 48.2

Notation: L-Bound denotes legally bound, L-Nonbound denotes legally nonbound (applied vault cash exceeds required reserves), E-Bound denotes economicallybound (as defined in this article), E-Nonbound denotes economically nonbound. All quarterly and annual reporting institutions are considered as both legally andeconomically nonbound in the construction of this table.

Source: tabulations by the authors from unpublished Federal Reserve data.

Table 2

decisions. In this article, we denote sucheconomically-nonbound institutions as E-Nonbound, and other institutions—forwhich reserve requirements are binding inthe traditional sense of constraining theirasset portfolio choices—as E-Bound.

How might the portfolio reactions ofE-Bound and E-Nonbound institutions tochanges in reserve-requirement ratiosdiffer? The depository institutions’ordinary business somewhat restricts theirresponses. Generally, an institution mustmaintain adequate stocks of vault cash toconvert customer deposits into currencyon request, and of Federal Reserve Bankdeposits to originate and absorb interbankpayments. However, both constraints aresomewhat flexible. There is an intradaymarket in vault cash, at least within largercities, suggesting that a bank might ask acustomer who is seeking a large amount ofcash to wait until later in the day, whenadequate currency can be obtained fromthe Federal Reserve or from a correspon-dent. Some banks require customers whoare planning to withdraw a significantamount of currency to provide at least onebusiness day’s notice. It also is notuncommon for ATM machines to run outof currency. For Federal Reserve Bankdeposits, there is a national secondarymarket, the federal funds market. Forinterbank payments, the Federal Reservemay delay an interbank payment if itexceeds applicable daylight or overnightoverdraft limitations. Because a failure toconvert a deposit into currency or to makea requested interbank payment maydamage a customer relationship, a deposi-tory cannot be indifferent to its mix ofvault cash and Federal Reserve Bankdeposits.

The economic distinction between E-Bound and E-Nonbound institutions isillustrated by the response of their excess-reserve ratio to changes in reserve-require-ment ratios. For E-Bound institutions,changes in reserve-requirement ratios,within the range where the requirementremains a constraint on the institutions’portfolios, will induce the institutions tomatch changes in required reserves with

dollar-for-dollar changes in base money. Asa result, excess reserves (base money heldin excess of statutory requirements) will beapproximately unchanged. If all depositoryinstitutions are E-Bound, then a change inreserve-requirement ratios will leave theaggregate excess reserve ratio, e, almostunaffected. The best-known historicalexample of this type of portfolio adjust-ment is member banks’ reaction to the1936-37 increase in reserve-requirementratios. Contrary to expectations of FederalReserve officials, member banks’ excessreserve ratios did not fall sharply followingthe increases.20 Surplus deposits at FederalReserve Banks, excess in the sense thatthey were not applied to satisfy statutoryrequired reserves, were in fact an optimalportfolio choice by member banks; thedeposits were not excess in any economicsense. The reserve-requirement ratios of1935 were effective constraints on thebanking system, with almost all memberbanks E-Bound.

In contrast, consider the portfolioresponse of E-Nonbound depository insti-tutions to a change in reserve-requirementratios that leaves the institutions E-Nonbound after the change. The institu-

20 See Friedman and Schwartz(1963), pp. 521-34, for a dis-cussion of the changes inreserve requirements and docu-mentation that the Fed antici-pated that the increases inreserve-requirement ratioswould substantially reduce theexcess reserves of the bankingsystem. Note that only depositsat Federal Reserve Banks wereeligible to satisfy reserverequirements (vault cash didn’tbecome eligible until 1959).



15

Surplus Vault Cash

Jan 55 Jan 60 Jan 70 Jan 80 Jan 90 Jan 00

Billio

ns of

Doll

ars

0

1

2

3

4

5

6

7

8

Phase-in of applied vault cashat member banks

November 1980, Monetary Control Act

Figure 2

tions’ business needs, not legal reserveratios, are the primary determinant of theirbase money holdings. The excess reservesof E-Nonbound institutions will varyapproximately dollar-for-dollar but in theopposite direction to the change in requiredreserves, leaving total base moneyholdings largely unaffected.

The behaviors of surplus vault cashand required clearing balance contractsafter the 1990-91 and 1992 changes inreserve requirement ratios suggest that asubstantial proportion of depository insti-tutions are E-Nonbound. Surplus vaultcash, shown in Figure 2, increased sharplyin 1991 after reserve-requirement ratioswere reduced to zero from 3 percent onnonpersonal time deposits and Euro-dollar liabilities. This increase suggeststhat at least some depository institutionswith surplus vault cash were E-Non-bound during 1990. More dramaticperhaps was the sharp increase in requiredclearing balance contracts, shown inFigure 1.

Although almost all E-Bound insti-tutions reduced their holdings of FederalReserve Bank deposits after the reductionin reserve-requirement ratios, some found that attempts to match the reduc-tion in their required reserves with adollar-for-dollar decrease in their FederalReserve deposits caused an unacceptableincrease in the probability that they mightexperience either an overnight overdraftor exceed their regulatory cap on day-light overdrafts.21 This increase also waslikely due, at least in part, to depositoryinstitutions’ recognition that FederalReserve Bank deposits no longer neededto satisfy statutory reserve requirementscould be used to satisfy required clearingbalance contracts. Note that the 1991surge in required clearing balancecontracts occurred after several years of stability in the amount of suchbalances.

The aggregate data on requiredclearing balance contracts also are consis-tent, at least in part, with the alternativehypothesis that some part of requiredclearing balances is held primarily to

defray the cost of Federal Reserve pricedservices. Clearing balances surged during1991 and 1992 as growth of the monetarybase accelerated and the federal funds ratefell, and they decreased sharply during1994 as growth of the base slowed and thefederal funds rate rose. Although someinstitutions likely adjust the size of theirclearing balance contracts inversely withrespect to changes in the federal fundsrate, the changes in 1991 seem too large tobe primarily a reaction to a lower federalfunds rate.

If E-Nonbound institutions represent asignificant share of the monetary base heldby depository institutions, it is importantto separate them from E-Bound institu-tions when measuring RAM. To make theanalysis more precise, consider an econ-omy with two distinct groups of deposi-tory institutions, both subject to FederalReserve System reserve requirements.Define economic excess reserves as

ERi = RBi2(rDDi1rTTi2VCi),i = (EB,EN) ,

where RB denotes total Federal ReserveBank deposits held by all depository insti-tutions (with no deduction for the amountof required clearing balance contracts),and let the subscripts EB and EN denotegroups of E-Bound and E-Nonbound insti-tutions, respectively. E-Bound institutionsare assumed to change the amount of basemoney they demand (relative to reservabledeposits) approximately dollar-for-dollarfollowing a change in statutory requiredreserve ratios. For this group, changes inreserve requirement ratios leave theirexcess reserve ratio,

,

approximately unchanged. E-Nonboundinstitutions do not change their holdingsof base money (relative to reservabledeposits) following a change in reserve-requirement ratios. Their excess reserveratio,

,eER

DEN

ENP

=

eER

DEB

EBP

=



16

21 See Richards (1995) orHancock and Wilcox (1996).

changes in equal absolute amount, but inthe opposite direction, to the reserve-requirement ratio.

In the previous section, we derived theRAM adjustment proposed by Burger andRasche (prior to passage of the MonetaryControl Act) from a decomposition of themonetary base into the amounts of basemoney held by member banks, by otherdepository institutions, and by thenonbank public:

MB = RB1Cu

= rDDM1rTTM1k DP1VCN1

(VCM1RB2rDDM2rTTM)

= [rDdM1rTtM1k1vN1eM]DP .

The analysis of this section suggests that asimilar decomposition between E-Boundand E-Nonbound institutions would beuseful for the period since implementationof the MCA. In obvious notation, the mon-etary base may be written as

.

The appropriate RAM for inclusion in theadjusted monetary base is then

,

where and denote the statutoryreserve-requirement ratios on demand andtime deposits, respectively, during the baseperiod of the RAM adjustment.22 Note thatthis RAM includes only deposits at E-Boundinstitutions. In all essential aspects, the treat-ment of E-Nonbound institutions in thisRAM is analogous to the treatment of non-member banks in Burger and Rasche (1977).

The adjusted monetary base may bewritten as

AMB = MB1RAM

The adjusted monetary base multiplier is

By assumption, eEB does not change whenor changes because E-Bound

institutions match reductions in theirrequired reserves due to changes in statu-tory reserve-requirement ratios aboutdollar-for-dollar with reductions in theirholdings of base money. In contrast, eEN

is assumed to change when or changes so as to leave the second term inthe denominator ( )unchanged.

To measure accurately this post-Mone-tary Control Act RAM, it is necessary todetermine the time-varying fractions of transaction deposits, dEB, and timedeposits, tEB, at E-Bound depository insti-tutions. These fractions cannot be identifiedin aggregate data; they must be obtainedfrom data on individual financial institu-tions. Later in this article we present astatistical analysis of individual bank datathat allows us to develop criteria for sepa-rating E-Bound and E-Nonbound institu-tions. Some of the results of that analysis areshown in Figures 3 and 4, and in Table 2.The estimated number of E-Bound deposi-tory institutions, shown in Figure 3, fell by 80 percent, to about 500 institutions,following the 1990-91 reduction in reserverequirements.

The proportions of transaction andnontransaction deposits at E-Bound insti-tutions, shown in Figure 4, fell from peaksin 1990 to approximately 65 and 54 per-cent, respectively, following the 1990-91reduction in reserve requirements. (Recallthat the reserve-requirement ratio on timeand savings deposits was reduced to zeroin December 1990.) The rows labeled E-Nonbound and E-Bound in Table 2 showadditional data. In 1995, we estimate thatonly about 2 percent of U.S. depositoryinstitutions were E-Bound—that is, foundstatutory reserve requirements to be animportant determinant of their businessdecisions. Only deposits at these E-Bound

r r eEND

EN ENT

EN ENδ τ+ +

rENTrEN

D

rEBT

,0rEBD

,0

M

AMB

k

r r e r r eEBD

EB EBT

EB EB END

EN ENT

EN EN

1

1

0 0

=

+

+ +( )+ + +( ), ,

.δ τ δ τ

= + + +[+ + + ]

r r r

r k e e D

EBD

EB EBT

EB END

EN

ENT

EN EB ENP

, ,

.

0 0δ τ δ

τ

rEBT

,0rEBD

,0

RAM r r r r DEBD

EBD

EB EBT

EBT

EBP= −( ) + −( )

, ,0 0δ τ

MB r r r

r k e e D

EBD

EB EBT

EB END

EN

ENT

EN EB ENP

= + + +[+ + + ]

δ τ δ

τ

22 The new St. Louis adjustedmonetary base incorporates fourRAM adjustments, each builtusing a different base period.



17

institutions are included in the new RAMadjustment for the St. Louis adjusted mone-tary base (see Anderson and Rasche, 1996a).

Time Deposit Ratios and ReserveRequirement Changes

Our discussion of RAM has focused tothis point on the direct impact of changesin reserve-requirement ratios on the mone-tary base multiplier, assuming that theother ratios in the adjusted monetary basemultipliers for M1 and M2 are unaffectedby changes in the reserve requirementratios, or that

In the previous section, we showed how,for E-Nonbound institutions, the excessreserve ratio, e, might be a function of r. In this subsection, we explore whether the time deposit ratio, t, might be a func-tion of r.

The end of Regulation Q ceilings ondeposit offering rates during the early1980s gave depository institutions a tool,changes in deposit offering rates, thatcould in principle be used to adjust theirreserve positions by inducing customers toshift among categories of deposits with dif-ferent reserve-requirement ratios. In acompetitive market without legal interestrate ceilings, it seems reasonable to expectthat changes in reserve-requirement ratioswill affect the rates offered by E-Boundinstitutions on different types of deposits.Further, it seems reasonable that economicagents will base decisions about theproportion of their wealth to hold in theform of time deposits, in part, on the ratesof return offered on time deposits. If so,following implementation of the MonetaryControl Act, competitive pressures mayhave caused increases in offering rates onsavings and time deposits, relative to thoseon transaction deposits. Similarly, theDecember 1990–January 1991 reduction to zero of reserve requirements on nonper-sonal time deposits may have increasedoffering rates on large negotiable CDs rela-tive to other instruments. Both events mayhave increased the ratio of time deposits totransaction deposits at E-Bound institu-tions, tEB, that enters the adjusted mone-tary base multiplier.23

dm

dr

m

r

m

r e k= = −

+ +

∂∂

.



18

Number of Institutions Included in RAM


Numb

er of

Insti

tution

s

0

1000

2000

3000

4000

5000

6000

First maintenance period in July of each year

Figure 3

Proportions of Deposits in Economically

First maintenance period in July of each year

Net TransactionDeposits (Delta)

December 1990

Time and Savings Deposits (Tau)


Ratio

0.46

0.52

0.58

0.64

0.70

0.76

0.82

0.88

Figure 4

Testing for a shift in the time depositratio circa 1980 is difficult, due to Regula-tion Q controls. Data from the latter period(1990-91) suggest, however, that banks donot alter their offering rates in response tosubstantial changes in reserve requirements.The spreads between rates on largenegotiable CDs and on Treasury bills andcommercial paper are shown in Figure 5.Although there is considerable variabilityfrom week to week, there is no discernibletrend. The spread of CD rates over three-and six-month Treasury bill rates fluc-tuates around 50 basis points; the spreadof CD rates over commercial paper ratesfluctuates around zero. Neither has anydiscernible spikes or shifts at the beginning

of 1991 when the reserve-requirementratio was reduced to zero on nonpersonaltime deposits. Hence, we do not include inRAM any adjustment for potential indirecteffects of reserve-requirement ratio changeson the multiplier via changes in the time-deposit ratio at E-Bound institutions.

MICROECONOMIC EVIDENCE: ARE BANKSBOUND BY RESERVEREQUIREMENTS?

In this section, the reactions ofcommercial banks to the December1990–January 1991 and April 1992 reduc-tions in reserve requirement ratios are

23 The multiplier discussed here isfor transaction money, M1.Multipliers for the broader mea-sures of money such as zero-maturity money (MzM), M2,and M3 include additionalterms in their numeratorswhich, in a more detailed anal-ysis, would be shown as com-ponents of the time depositratio, t. For examples, seeRasche and Johannes (1985).



19

Rate Spreads on 3-Month Instruments

75 79 8377 81

Negotiable CD Rate – Coupon Equivalent T-Bill Rate

Negotiable CD Rate – Commercial Paper Rate

85

Monthly data, percent

1.5

2.5

0.5

1.0

2.0

0.0

20.5

87 89 91 93 95 97

Figure 5

Rate Spreads on 6-Month Instruments

Negotiable CD Rate – Coupon Equivalent T-Bill Rate

Negotiable CD Rate – Commercial Paper Rate

1.5

2.5

0.5

1.0

2.0

0.0

20.5

75 79 8377 81 85 87 89 91 93 95 97

examined in an analysis of variance(ANOVA) framework.24 The analysis seeksto quantify the different reactions of L-Bound and L-Nonbound banks, of varioussizes, to changes in statutory reserve-requirement ratios. The goal of theanalysis is to develop criteria thatdistinguish E-Bound from E-Nonboundinstitutions, consistent with the construc-tion of RAM outlined above. Our datasample consists of commercial banks thatreported weekly deposit and reserve levelsto the Federal Reserve from mid-1990through the end of 1992.

To this point in this article, the termsbound (L-Bound) and nonbound (L-Nonbound) have described the reserveposition of an individual bank during asingle reserve maintenance period. (Recallthat a bank is said to be L-Nonbound if itfully satisfies its required reserves withvault cash.) In this section, we use theseterms more broadly. For tractability in sta-tistical analysis, we classify individualbanks as being either of type L-Bound ortype L-Nonbound according to a specifiedcriterion applied to the bank during anumber of reserve maintenance periods.This classification system would be trivial ifall institutions, during all reserve main-tenance periods, were either L-Bound or L-Nonbound and did not change status. In fact, small and medium size institutionsoften change categories, being L-Nonboundin periods when retail cash demands areheavy and L-Bound in others; larger bankstend to remain consistently in one category.

In our analysis, we examined threealternative criteria for classifying a bank astype L-Bound: (1) the bank was L-Boundduring maintenance periods in the secondhalf of 1992 (1992 H2), regardless of itsstatus in other years; (2) the bank was L-Bound during maintenance periods in1990 H2, regardless of its status in otheryears; and, (3) the bank was L-Boundduring all maintenance periods in 1990H2, 1991 H2 and 1992 H2. We find thatstatistical inferences regarding the reactionof banks to changes in statutory reserverequirements are robust to reasonablealternative criteria. Parts A and B of Tables

3, 4, 5, and 6 show comparative resultsbased on cases (1) and (3), respectively.

Our size groupings are broadly consis-tent with categories used in other bankingstudies. Banks classified as L-Bound aresubsequently separated into four sizeclasses—small, medium, regional and large—based on net transaction deposits, whileL-Nonbound banks are separated into onlytwo: small and medium. There are too fewlarger L-Nonbound banks for analysis. TheMonetary Control Act of 1980 establisheda tiered system of reserve requirementswherein the first $25 million of net trans-action deposits, the “low reserve tranche,”was subject to a 3 percent requirement,while larger amounts were subject to a 12percent requirement. Initially set at $25million, the low reserve tranche is indexedannually to the change in the aggregateamount of net transaction deposits. Weclassify banks in our sample data as smallif their holdings of net transaction depositsdid not exceed the low reserve trancheduring any reserve maintenance period inthe second half of 1990 (the low reservetranche was $41.1 million in 1990).25 Weclassify banks as medium-sized if theiraverage level of net transaction depositsduring the second half of 1990 was greaterthan the low reserve tranche but less than$135 million, as regional-sized if their nettransaction deposits averaged more than$135 million but less than $500 million,and as large if their net transactiondeposits averaged more than $500 million.

Statistical inferences regarding banks’reactions to reserve-requirement changesare robust to reasonable alternative size-classification schemes. Use of the lowreserve tranche for delineating small banksprovides an important control in our anal-ysis, because the April 1992 change inreserve-requirement ratios affected onlybanks with transaction deposits above thetranche. Results for the medium andregional groups are not sensitive to theprecise cut-off selected to separate thegroups because there are relatively fewbanks with net transaction depositsbetween about $100 to $150 million. The$500 million cut-off places about 150

24 Because of the unsettled stateof the thrift industry during thisperiod, we exclude thrifts fromthe analysis.

25 The Garn-St. Germain Act of1982 created the reserveexemption amount, which issubject to a zero reserverequirement. Originally $2.1million of deposits, it also isindexed. See Anderson andKavajecz (1994) or theFederal Reserve System’sRegulation D, ReserveRequirements, for details.



20

banks in our large category, similar to thegroup of large weekly reporting banks onthe list published by the Federal Reserve.26

The estimated models shown below alsoare not sensitive to inclusion or exclusionof banks that acquired other institutions.

Summarizing our results, we find that:• L-Nonbound banks did not change their

holdings of base money (vault cash plusFederal Reserve Bank deposits), relativeto transaction deposits, when reserverequirements changed in 1990-91 or1992.

These banks chose to hold enough vaultcash to satisfy their reserve requirements.If vault cash is held primarily to convertdeposits into currency on request, thentheir holdings of vault cash are likely to beinsensitive to changes in statutory reserverequirements. The quantity of FederalReserve Bank deposits held by these banks,if any, is unlikely to be affected by changesin statutory reserve requirements becausethe deposits are not being held to satisfyreserve requirements.• Small and medium-sized L-Bound banks

changed their holdings of base moneysomewhat in response to the 1990-91reduction of the reserve-requirementratio on time and savings deposits tozero from 3 percent, but they did notrespond to the 1992 reduction of thereserve requirement ratio on net transac-tion deposits to 10 percent from 12percent.

Small L-Bound institutions primarily differfrom small L-Nonbound institutions bychoosing to hold less vault cash; the twogroups of banks have similar mixtures ofdeposits. In the absence of a fully workedout model of bank reserve management, itis difficult to identify factors that mightaccount for the differing responses of smallL-Nonbound and small L-Bound to the1990-91 changes in reserve-requirementratios. In contrast, the 1992 reduction inthe reserve-requirement ratio on net trans-action deposits affected (algebraically)only larger institutions with nettransaction deposits above approximately$42 million. The reduction had no effecton banks subject to only a 3 percent

marginal reserve requirement, and only aweak effect on medium-sized banks thatfaced a 12 percent requirement on only apart of their transaction deposits.• Larger L-Bound banks responded

strongly to the 1990-91 reduction andsomewhat less strongly to the 1992reduction.

These banks chose to satisfy their requiredreserves by holding relatively largeamounts of Federal Reserve Bank deposits.In general, these amounts are more thanwhat is necessary for the banks’ payments-related activities, such as check clearingand wire transfer. If statutory reserverequirements are binding for any group ofbanks, it must be for these.

Our models seek to estimate theresponse of banks’ base money holdings tochanges in reserve requirements.Measuring the amount of base money heldby some nonmember institutions is prob-lematic, however. While all banks in ourdata report their daily holdings of vaultcash to the Federal Reserve, somenonmember banks do not hold FederalReserve deposits in their own name;rather, they hold them indirectly via apassthrough contract with a correspondentbank.27 In addition, some nonmemberbanks hold Federal Reserve deposits bothindirectly (through a correspondent) anddirectly (in their own account). Weincreased the Federal Reserve depositsreported by banks with passthroughreserve arrangements by an amount equalto the difference between the bank’srequired reserves and its applied vaultcash. At the same time, we reduced eachcorrespondent bank’s reported FederalReserve deposits by the amount of itsrespondent banks’ required reservescharged against the correspondent’sFederal Reserve account. Given the datareported by banks to the Federal Reserve,this is the only feasible method formeasuring the amount of Federal Reservedeposits held (indirectly) by banks withpassthrough reserve contracts.

Summary statistics for our sample ofbanks are shown in Table 3. The sampleconsists of commercial banks that reported



21

26 See Table 1.27 in the FederalReserve Bulletin.

27 Nonmember depository institu-tions may contract with theFederal Reserve to satisfy theirrequired reserves (beyond vaultcash) by having the amount ofthe requirement chargedagainst the Federal Reservedeposit account of an eligiblecorrespondent institution (a so-called “passthrough” reservecontract). Nonmember institu-tions that satisfy requiredreserves via passthrough con-tract may open an additionalFederal Reserve account in theirown name. Federal Reservedeposits in this second accountmay be used to satisfy arequired clearing balance con-tract but may not be used tosatisfy required reserves. (Ofcourse, the funds could beloaned via the federal fundsmarket to the correspondent.)

data weekly to the Federal Reserve frommid-1990 through the end of 1992.28

Banks without data for all included reservemaintenance periods are omitted, as arebanks involved in mergers or acquisi-tions.29 Part A and Part B of the table showbanks classified as L-Bound and L-Non-bound via two alternative schemes. In PartA, banks are classified as L-Bound if theywere L-Bound in 1992 H2, after both the1990-91 and 1992 reductions in reserverequirements; banks not classified as L-Bound are classified as L-Nonbound, evenif they were bound during some reservemaintenance periods in 1990 and 1991. InPart B, banks are classified as L-Bound ifand only if they were L-Bound in everyreserve maintenance period during 1990H2, 1991 H2 and 1992 H2; otherwise, theyare classified as L-Nonbound. Applicationof the latter criteria reduces the number ofL-Bound banks from 1822 in part A to 710in part B, primarily by pushing banks thatare close to being L-Nonbound (or inother words, they hold enough vault cashto fully satisfy their required reserves insome maintenance periods) from the L-Bound group in part A into the L-Non-bound group in part B.30

Part A of Table 3 shows that the 1990-91 reduction in reserve requirements onnonpersonal time and savings depositsreduced required reserves at small, medium,regional and large banks by about 44 percent, 19 percent, 17 percent and 18 percent, respectively, after allowing forincreases in their net transactions deposits.On average, these banks satisfied aboutone-half of their required reserves withvault cash (column 4).31 Contractedclearing balances increased sharply from1990 to 1992, approximately doubling forthe smallest banks and increasing by almostan order of magnitude for large banks.

A Traditional Fixed Effects ANOVAModel

In Table 4 (see page ••), we presenttraditional ANOVA fixed-effects regressionestimates for the effects of changes inreserve requirements in 1990-91 and 1992on the ratio

during the last 13 reserve maintenanceperiods in 1990, 1991 and 1992, a total of39 observations on each bank. The modelis

,

whereyijt = the ratio of base money (vault

cash plus deposits at Federal ReserveBanks) to net transaction deposits held bybank i in maintenance period j in year t, (i = 1,..., N), (j = 1,...,13), (t = 90,...,92).

= 1 for bank i, and 0 otherwise,

= 1 in year t, and 0 otherwise,

= 1 in maintenance period j, and 0 otherwise,

= 1 if bank i had a clearingbalance contract in maintenance period jof year t, and εi

jt is an assumed i.i.d. distur-bance.

The dummy variables index therelative position of reserve maintenanceperiod j within the year, with the firstperiod in July of each year numbered “1”and the last period of the year numbered“13”. As such, they absorb seasonal fluctu-ations that may differ in strength acrossbanks. We interpret the as repre-senting the effects of changes in reserverequirements between 1990, 1991, and1992, although as dummy variables theymay also pick up other year-specificeffects. Estimates presented in Parts A andB of Table 4 correspond to the statisticsshown in Parts A and B of Table 3. Becausethe estimates in Parts A and B are similar,we discuss only the estimates shown inPart A. Standard errors reported in thetable are Huber-White robust estimates ofthe regression covariance matrix.

For all groups of banks, the nullhypothesis of no significant year effects in

DijtY

DijtP

DijtR

DijtP

DijtY

DijtB

y D

D D

D

ijt i N ijtB

i

N

t ijtY

t

j ijtP

j

ijtR

ijt

= + −( ) +

−( ) + −( ) +

+

=

−

= =

∑

∑ ∑

φ α α

λ λ γ γ

ξ ε

1

1

92

90

91

13

1

12

vault cash +Federal Reserve deposits

net transactions deposits

28 Our sample may underrepre-sent small depository institu-tions that are not required toreport data weekly to theFederal Reserve. We assumethat virtually all of these institu-tions would be classified as eco-nomically nonbound andexcluded from the calculationof RAM. For a discussion ofFederal Reserve data reportingrequirements, see Andersonand Kavajecz (1994).

29 Banks that acquire other depos-itory institutions are permittedunder the Federal Reserve’sRegulation D to phase out, dur-ing the following eight quar-ters, the benefit of the acquiredinstitution’s low reservetranche. All banks involved insuch acquisitions are excludedfrom our statistical analysis.Such banks are included in thecalculation of RAM, where weallow for this effect by adjust-ing the size of the tranche lossadjustment to reflect the size ofthe tranche in the base period,January 7, 1991; seeAnderson and Rasche (1996a).

30 The overall sample size also issmaller in Part B because 13regional and large L-Boundbanks in Part A are reclassifiedas L-Nonbound in Part B, anddropped from the analysis.

31 In December 1995, for exam-ple, required reserves of alldepositories were $56.6 bil-lion, of which $37.5 billionwas satisfied with vault cash.See for example AggregateReserves of DepositoryInstitutions and the MonetaryBase, Board of Governors’weekly statistical release (H.3),April 25, 1996, Table 2.



22



23

1990Small 1.16 0.62 0.11 0.68 24.3 1.37 1139Medium 2.67 2.04 0.14 1.37 33.0 2.89 512Regional 3.41 2.81 0.12 1.37 27.4 3.57 122Large 8.68 7.12 0.10 3.64 61.1 8.86 49



L-Nonbound

1990Small 1.45 0.85 0.21 1.33 35.78 2.33 2209Medium 0.14 0.10 0.02 0.13 2.27 0.19 45

1991Small 0.93 0.93 0.25 0.93 38.37 2.33 2209Medium 0.11 0.11 0.02 0.10 2.29 0.18 45

1992Small 1.08 1.08 0.30 1.08 43.35 2.48 2209Medium 0.11 0.11 0.02 0.11 2.46 0.20 45

1 Classifications are based on banks’ legal reserve status in 1992 H2 and on their size in 1990 H2. Banks are classified as L-Bound if they were legally bound in oneor more reserve maintenance periods during 1992 H2; if not, they are classified as L-Nonbound. Banks are classified as small if their net transaction deposits didnot exceed the low reserve tranche ($40.4 million) in any reserve maintenance period during 1990 H2, and are classified as medium or regional if their averagelevel of net transaction deposits during 1990 H2 did not exceed $125 million or $500 million, respectively. Banks with net transaction deposits averaging more than$500 million during 1990 H2 are classified as large.

Table 3

Summary Statistics for Commercial Banks Used in ANOVA(by reserve status and size, billions of dollars except number of institutions)

A. Classified by Legal Reserve Status in 1992 and by Size in 19901

L-Bound

RequiredReserves,

Total

RequiredReserves

Against NetTransaction

Deposits

RequiredClearingBalances

(contractedamount)

Applied VaultCash

AggregateNet

TransactionDeposits

Amount ofBase Money

Held (Vault Cash+ FederalReserve

BankDeposits)

Number ofBanks



24

Table 3, continued

Summary Statistics for Commercial Banks Used in ANOVA(by reserve status and size, billions of dollars except number of institutions)

B. Classified by Legal Reserve Status in 1990, 1991 and 1992, and by Size in 19902

L-Bound

RequiredReserves,

Total

RequiredReserves

Against NetTransaction

Deposits

RequiredClearingBalances

(contractedamount)

Applied VaultCash

AggregateNet

TransactionDeposits

Amount ofBase Money

Held (Vault Cash+ FederalReserve

BankDeposits)

Number ofBanks




L-Nonbound

1990Small 2.28 1.32 0.30 1.92 54.6 3.34 3112Medium 0.86 0.63 0.75 0.67 12.8 1.01 241

1991Small 1.47 1.47 0.38 1.44 59.0 3.18 3112Medium 0.72 0.72 0.11 0.63 13.6 0.96 241

1992Small 1.77 1.77 0.47 1.67 67.7 3.46 3112Medium 0.81 0.81 0.13 0.69 15.4 1.04 241

2 Banks are classified as L-Bound if they were legally bound in all reserve maintenance periods in 1990 H2, 1991 H2 and 1992 H2; if not, they are classified as L-Nonbound. Banks are classified as small if their net transaction deposits did not exceed the low reserve tranche ($40.4 million) in any reserve maintenance periodduring 1990 H2, and are classified as medium or regional if their average level of net transaction deposits during 1990 H2 did not exceed $125 million or $500million, respectively. Banks with net transaction deposits averaging more than $500 million during 1990 H2 are classified as large.



25

Table 4

ANOVA Model Estimates for Weekly-Reporting Banks

L-Bound Banks L-Nonbound Banks

Statistic Small Medium Regional Large Small Medium

A. Classified on Legal Reserve Status in 1992 and on Size in 19901

F Statistics (numerator degrees of freedom)

for bank fixed effects 25.9 (1138) 101.3 (511) 72.4 (121) 69.0 (48) 119.7 (2208) 144.7 (44)p-value (Pr > F) 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001

for year fixed effect 430.2 (2) 3917.2 (2) 346.4 (2) 1152.5 (2) 2781.9 (2) 10.4 (2)p-value (Pr > F) 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001

Coefficient Estimates (t-statistic)Year effects

1990 versus 1992 0.021 (25.9) 0.017 (84.9) 0.032 (26.5) 0.036 (47.9) 0.009 (75.1) 0.003 (4.0)p-value (Pr > | t |) 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001


Required Clearing Balance 0.004 (2.0) 0.006 (9.4) 0.022 (6.8) 0.015 (11.3) 0.009 (26.9) 0.001 (0.4)p-value (Pr > | t |) 0.0465 0.0001 0.0001 0.0001 0.0001 0.6756

Summary StatisticsModel degrees of freedom 1153 526 136 63 2223 59Error degrees of freedom 43267 19441 4621 1847 83927 1695R-Squared 0.41 0.75 0.67 0.75 0.76 0.79

B. Classified on Legal Reserve Status in 1990, 1991 and 1992, and on Size in 19902




Coefficient Estimates (t-statistic)Year effects


1991 versus 1992

20.002 (1.9) 0.004 (14.4) 0.013 (10.0) 0.012 (15.5) 0.003 (9.7) 0.002 (8.4)p-value (Pr > | t |) 0.0612 0.0001 0.0001 0.0001 0.0001 0.0001

Required Clearing Balance 20.004 (1.1) 0.006 (6.9) 0.030 (7.4) 0.015 (11.3) 0.007 (8.8) 0.004 (4.7)p-value (Pr > | t |) 0.2631 0.0001 0.0001 0.0001 0.0001 0.0001


1 Banks are classified as L-Bound if they were legally bound in one or more reserve maintenance periods during 1992 H2; if not, they are classified as L-Nonbound.Banks are classified as small if their net transaction deposits did not exceed the low reserve tranche ($40.4 million) in any reserve maintenance period during1990 H2, and are classified as medium or regional if their average level of net transaction deposits during 1990 H2 did not exceed $125 million or $500 million,respectively. Banks with net transaction deposits averaging more than $500 million during 1990 H2 are classified as large.

2 Banks are classified as L-Bound if they were legally bound in all 39 reserve maintenance periods in 1990 H2, 1991 H2 and 1992 H2; if not, they are classified asL-Nonbound. Size criteria are the same as in the previous footnote.

the behavior of yijt is rejected. Consider,then, the regression results for each groupof banks:

Small L-Bound banks. The model esti-mates suggest a strong response by smallL-Bound banks to the 1990-91 reductionin reserve requirements. The reasonable-ness of the estimated ANOVA regressioncoefficients can be judged by comparingthem to estimates calculated from thesummary statistics shown in Table 3,under the null hypothesis that the banksare E-Bound—that is, that the banks’reductions in their holdings of base moneywill match, dollar-for-dollar, the decreasein their required reserves. Approximatelyhalf of the required reserves of these banksin 1990 was due to net transactiondeposits ($620 million), and about halfwas due to nontransaction deposits. Eachcategory of deposits was subject to a 3 per-cent marginal requirement. Ignoring thezero reserve requirement on the reserveexemption amount, the reduction to zeroof the reserve-requirement ratio onnonpersonal time and savings depositsmight reduce the banks’ average ratio ofbase money to net transaction deposits byabout one-half of its 1990 value of 0.056,or 0.028. The estimated ANOVA effectequals the estimated coefficient on the1990 year dummy variable less the coeffi-cient on the 1991 year dummy variable, or

(λ̂902 λ̂91) = 0.02120.001 = 0.02.

Both λ̂90 and λ̂91 are estimated relative to1992 because is omitted from theregression. The size of ANOVA effect isclose to, but slightly smaller than, the effectcalculated from the summary statistics.

The regression coefficient for 1991,λ̂91, measures small L-Bound banks’reaction to the April 1992 reduction in themarginal reserve requirement on net trans-action deposits to10 percent from 12percent. Because these banks’ net transac-tion deposits were below the low reservetranche, their required reserves were unaf-fected by the change. The estimatedANOVA effect, 0.0006, is about zero, asexpected.

Medium L-Bound Banks. The results formedium-sized L-Bound banks are similarto those for small L-Bound banks. Asabove, we can judge the reasonableness ofthe estimated ANOVA effect by calculatinga preliminary estimate from the summarystatistics in Table 3, under the nullhypothesis that the banks are E-Boundboth before and after the change in thelegal requirements. About three-fourths ofthe required reserves of these banks weredue to net transaction deposits in 1990,and their total required reserves averagedabout 8.1 percent of their net transactiondeposits. Thus, reducing to zero thereserve requirement on nonpersonal timeand savings deposits seems likely to reduceby about one-fourth their overall ratio ofrequired reserves and base money holdingsrelative to net transaction deposits, adecrease of about 0.02.

The estimated ANOVA effect of the1990 reduction equals the coefficient onthe 1990 year dummy less the coefficienton the 1991 year dummy variable, or

(λ̂902 λ̂91) = 0.016720.0034 = 0.013.

The effect is economically significant,although smaller in size than our prelimi-nary estimate which assumed that thereduction in base money holdings wouldmatch, dollar-for-dollar, the decrease inrequired reserves.

The estimated effect of the 1992reduction is only 0.0034. While statis-tically significant with such a large sample,it is less than one-sixth of the 0.02 changein the marginal statutory reserve require-ment ratio. This estimate suggests thatbanks in this size range reduced theirholdings of Federal Reserve deposits little,if at all, following the decrease in thereserve-requirement ratio. There is someevidence of a smaller response by banksthat had required clearing balancecontracts. This difference seems consistentwith our conjecture that required clearingbalances are, for some banks at least, alow-cost type of excess reserves. Further,banks with required clearing balance con-tracts likely are purchasing payments-related services from the Federal Reserve

DijY,92



26

and need sufficient Federal Reservedeposits to avoid overdrafts.

Overall, for small and medium L-Bound banks, the estimated responses tothe 1990-91 reduction are economicallysignificant, while responses to the 1992reduction are not. The insignificance of the latter coefficient for small banks isexpected, since their marginal reserve ratioremained unchanged at 3 percent. While the1992 year effect for the medium L-Boundbanks is statistically significantly greaterthan zero, its small size makes it difficult toattribute the effect to changes in the mar-ginal reserve requirement ratio. In part, theestimated effect may reflect banks’ reactionto the lower federal funds rate that prevailedduring 1992, relative to 1991. Below, wecompare these responses to those of similarL-Nonbound banks.

As a result of this analysis, small andmedium-size L-Bound depository institu-tions are assumed to have been E-Boundprior to January 1991 and are included inRAM through December 1990, but areassumed to have become E-Nonbound inJanuary 1991 and are thereafter excludedfrom RAM. With the exclusion of thesedepository institutions, only about 5-1/2percent of weekly reporting institutions, and2 percent of all institutions (see Table 2), areincluded in the revised RAM adjustment forthe St. Louis adjusted monetary base.

Regional and Large L-Bound Banks. Onbalance, regional and large L-Bound banksare estimated to have respondedsignificantly to both the 1990-91 and 1992reductions; in other words, they were E-Bound. Preliminary estimates of thereductions’ effects may be calculated, asabove, from Table 3. Required reservesheld against net transaction deposits wereabout 83 percent of these banks’ totalrequired reserves in 1990, and their ratiosof total required reserves to net transactiondeposits were about 12.2 and 13.9 percent,respectively. If the 1990-91 reductions inrequired reserves were reflected fully inreduced holdings of base money, we wouldproject an effect of more than 0.02. The esti-mated ANOVA effect for regional banks is

(λ̂902 λ̂91) = 0.032120.0128 = 0.0193,

and for large banks is

(λ̂902 λ̂91) = 0.036420.0121 = 0.026.

Both are almost precisely what wouldoccur if these banks matched thereduction in their required reserves withdollar-for-dollar reductions in theirholdings of base money.

In contrast, these banks seem to havereduced their holdings of base money lessthan proportionately, following the April1992 reduction in reserve requirements.The estimated coefficients, λ̂92, for regionaland large banks—0.0128 and 0.0128,respectively—are significantly greater thanzero statistically and less than the 0.02reduction in the marginal statutoryrequirement. Again, banks with requiredclearing balance contracts are estimated tohave reduced their holdings of base moneyless than other banks. The lower federalfunds rate during 1992 likely attenuatedthe reduction in Federal Reserve depositsthat otherwise would have followed thereserve-requirement reduction. Inaddition, the smaller estimated coefficient(relative to the statutory change of 0.02)likely reflects some large banks becomingE-Nonbound after the 1990-91 and 1992reductions.32

Small and Medium-Size L-NonboundBanks. Estimates for L-Nonbound bankssuggest economically insignificantresponses to changes in reserverequirements since 1990. Table 3’ssummary statistics show that requiredreserves held against nontransactiondeposits were about 40 percent and 30 per-cent, respectively, of the total requiredreserves held by these banks in 1990. Theaverage aggregate ratios of requiredreserves to net transaction deposits atthese banks were 4.1 percent and 6.2 per-cent, respectively, suggesting that thesebanks’ ratios of base money to net transac-tion deposits might decrease by as much as1.8 percent. The estimated ANOVA effectsof the reduction—0.005 for small banksand 0.003 for medium bank—are fairlysimilar and less than one-third of the pre-



27

32 See Feinman (1993) andHilton, Cohen and Koonmen(1993).

liminary estimate based on Table 3, underthe assumption that the banks are E-Bound. Considering the generally lowerfederal funds rate that prevailed duringthis period, we have difficulty attributingthe change in these banks’ base moneyholdings, relative to net transactiondeposits, to changes in legal reserve ratios.

Like medium-size L-Bound banks,medium-size L-Nonbound banks did notrespond to the 1992 reduction in thereserve-requirement ratio on nettransaction deposits. The estimatedANOVA effect, 0.0035, is not economicallydifferent from zero.

An ANOVA Model withIdiosyncratic Bank-Year Interactions

The ANOVA model shown aboveincludes a single fixed effect for each bank,ai, and assumes that the response of allbanks to the changes in statutory reserverequirements is the same, measured by λ̂90 and λ̂91. Because it seems unlikely thatall banks responded in the same way, weestimated a second ANOVA model thatpermits idiosyncratic responses by eachbank to the year effects:

,

where yijt is the same as above. This modelbecomes the same as the previous ANOVAif the bank-year effects are constrained tobe equal for all banks during each year, orin other words,

In the ANOVA, the effect of the 1990-91reduction in reserve requirements is measured by

β̂i,902 β̂ i,91 = ( ̂βi,902 β̂ i,92)2( β̂i,912 β̂ i,92),

and the effect of the 1992 reduction by

β̂i,912 β̂ i,92.

Distributions (histograms) of theseindividual bank-year interaction effects areshown in Figures 6-8. Summary statisticsand hypothesis tests for this model areshown in Table 5.33 Although Parts A andB of Table 5 show estimates under thesame alternative L-Bound classification cri-teria used in Tables 3 and 4, we limit ourdiscussion to Part A. Figures 6-8 are basedon the regressions summarized in Part A ofTable 5.

The null hypothesis that there was nochange in the behavior of yijt across 1990,1991, and 1992 is easily rejected by the F-statistics reported in Table 5. Theestimated responses of individual small,medium, regional and large L-Boundbanks to the 1990-91 reduction in reserve-requirement ratios are shown, respectively,in panels A and C of Figures 6 and 7(pages •• and ••). The number ofestimated coefficients plotted in each panelis shown in the panel title. On balance, L-Bound banks responded by significantlyreducing their holdings of base money:Most of the shaded area in eachdistribution is well to the right of zero(marked by a vertical line). Substantialvariation in the responses of individualbanks is evident in the figures, in partbecause different banks held different pro-portions of transaction and nontransactiondeposits (recall that the dependent variableis the ratio of the bank’s base money hold-ings to its transaction deposits).

In contrast, the estimated response byL-Bound banks to the April 1992reduction in reserve-requirement ratios,shown in panels B and D of Figures 6 and7, is more varied. As expected, thedistribution for small L-Bound banks(panel B of Figure 6) is tightly centeredabout zero. Medium-size L-Bound banks(panel D of Figure 6) also respondedweakly to the change, most commonlyreducing their ratios of base money totransaction deposits by about half of whatwould be implied if they had matched thedecrease in their required reserves dollar-for-dollar.34 For regional-size banks (panel

β β λ11

90 91, , , , .t N t tNt= = = =L

y D

D D

D D

ijt i N ijtB

i

N

i

N

it i ijtB

t

ijtY

j ijtP

j

ijtR

ijt

= + −( ) +

−( ) +

−( ) + +

=

−

= =

=

∑

∑ ∑

∑

φ α α

β β

γ γ ξ ε

1

1

1

92

90

91

13

1

12

,

33 The ANOVA models are estimat-ed with the GLM and REG pro-cedure in SAS, version 6.11,on an HP Unix workstation.

34 Medium-size L-Bound banksaveraged about $80 million innet transaction deposits (seeTable 3), the first $3.6 millionsubject to a zero reserverequirement ratio, the next$38.6 million to a 3 percentratio and, before the April1992 reduction, the balance toa 12 percent ratio. Their ratioof base money to net transac-tion deposits would havedecreased by about 1 percent-age point if the banks hadmatched the reduction in theirrequired reserves with a dollar-for-dollar reduction in their hold-ings of base money.



28

B of Figure 7), the ratio fell about 1-1/2percentage points, close to what would beprojected from their average nettransaction deposits of about $280 million(see Table 3). Large L-Bound banks (panelD of Figure 7) most commonly reduced

their base money holdings by about thefull 2 percentage points.

The ANOVA effects for L-Nonboundbanks are shown in Figure 8 (see page ••).Because these banks satisfied theirstatutory reserve requirement with vault



29

Table 5

ANOVA Model Estimates for Weekly-Reporting Banks, with Bank-YearInteraction Effects






for bank-year fixed effect 14.0 (2278) 52.7 (1024) 20.3 (244) 100.9 (98) 27.6 (4418) 21.3 (90)p-value (Pr > F) 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001

Coefficient Estimates (t-statistic)Required Clearing Balance 0.005 (1.3) 0.010 (12.0) 0.022 (4.2) 0.018 (9.2) 0.006 (10.9) 0.014 (4.1)

p-value (Pr > | t |) 0.20 0.0001 0.0001 0.0001 0.0001 0.0001




for bank-year fixed effects 76.8 (235) 236.9 (315) 127.7 (113) 252.0 (43) 54.6 (3111) 310.1 (240)p-value (Pr > F) 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001


Coefficient Estimates (t-statistic)Required Clearing Balance 0.007 (1.09) 0.008 (8.11) 0.020 (3.1) 0.0195 (10.4) 0.006 (3.6) 0.014 (10.6)

p-value (Pr > | t |) 0.27 0.0001 0.0018 0.0001 0.0003 0.0001




cash, we anticipated little reaction to thereductions in reserve-requirement ratios.Although there is some variety in effectsfor individual banks, the distributions forL-Nonbound banks generally aresymmetric about zero, for both the 1990-91 changes (panels A and C) and the 1992change (panels B and D).

Repeated-Measures Analysis ofVariance

The panel, or longitudinal, structureof our data requires attention to the implied covariance structure of the data-

generating process. Each bank in oursample is observed for 13 reserve mainte-nance periods in each of three years, 1990,1991, and 1992. As such, it seems unlikelythat the disturbances in our ANOVAmodels, εijt, are independent andidentically distributed. If they are not, thecoefficient estimates are unbiased and inef-ficient, while the estimated covariancematrix is biased and inconsistent. Anappropriate covariance structure likelywould be block-diagonal, with a separateblock for each bank. A test for theresponses of depository institutions which



30

a. Small Bound Banks, N=11391990 vs. 1991Frequency

200

100

300

0

20.0

602

0.050

20.0

402

0.030

20.0

202

0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

b. Small Bound Banks, N=11391991 vs. 1992Frequency

300

400

500

200

100

02

0.060

20.0

502

0.040

2 0.0

302

0.020

20.0

100.0

000.0

100.0

200.0

300.0

400.0

500.0

60

Figure 6

Frequency

d. Medium Bound Banks, N=5121991 vs. 1992

170180

150

130

160

140

120

100110

8090

60

40

70

50

30

100

20

20.0

602

0.050

20.0

402

0.030

20.0

202

0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

c. Medium Bound Banks, N=5121990 vs. 1991Frequency

110

120

100

90

80

60

70

50

40

30

20

10

0

20.0

602

0.050

20.0

402

0.030

20.0

202

0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

incorporates this block-diagonal covar-iance structure may be constructed byviewing the banks as if they were clinicalsubjects engaged in a laboratory ex-periment. It is commonplace in clinicalstudies to measure certain characteristicsof subjects both pre- and post-treatment,asking whether the change in the measure-ment for each subject, when averagedacross all subjects, is statistically sig-nificant. Since there are multiple obser-vations on each subject, the models arewidely referred to as repeated measuresmodels.35 In these models, the repeated

observations for each subject are treated asmultiple time series, and the disturbance isassumed to be multivariate normal.

In our data, we observe the ratio ofbase money to net transaction deposits foreach bank during thirteen reserve mainte-nance periods in each of three years: 1990,1991 and 1992. In the repeated measuresANOVA, the observations for each year aretreated as thirteen realizations of a singletime series process; pooled across the threeyears, the observations are regarded as amultiple time series process composed ofthree univariate processes. The data for

35 See Crowder and Hand(1990), Diggle, Liang andZeger (1994), or Davidian andGiltinan (1995). An earlier ref-erence is Hsiao (1986).



31

a. Regional Bound Banks, N=1221990 vs. 1991Frequency40

30

20

10

0

20.0

602

0.050

20.0

402

0.030

20.0

202

0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

b. Regional Bound Banks, N=1221991 vs. 1992Frequency

30

40

50

60

70

20

10

02

0.060

20.0

502

0.040

20.0

302

0.020

20.0

100.0

000.0

100.0

200.0

300.0

400.0

500.0

60

Figure 7

c. Large Bound Banks, N=49 d. Large Bound Banks, N=491990 vs. 1991 1991 vs. 1992Frequency Frequency

8

6

7

5

3

1

4

2

0

30

20

0

10

20.0

602

0.050

20.0

402

0.030

20.0

202

0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

20.0

602

0.050

20.0

402

0.030

20.0

202

0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

1990-91 form a pre- and post-treatmentcontrast for the 1990-91 reserve require-ment reduction, and the data for 1991-92form a similar contrast for the April 1992reduction. Inferences regarding the re-sponse of various groups of banks to thereserve requirement changes are made bytesting for the existence of significantinteraction effects across years between(and among) each individual year’s reservemaintenance-period effects.

Repeated measures ANOVA results areshown in Table 6 and in Figures 9 and 10.Two tests are shown in each of Parts A and B in the table. The first is based on an esti-

mated multivariate ANOVA, or MANOVA,model wherein the dependent variable isthe vector [yij,90, yij,91, yij,92]′, the explanatoryvariables are dummy variablesrepresenting each reserve maintenanceperiod and the presence of a requiredclearing balance contract, and thedisturbance vector for each bank isassumed multivariate normal without anyrestrictions on its covariance matrix. Thevalue of Wilks’ lambda, a multivariateanalog of more familiar F-tests, suggestsrejection of the hypothesis that coefficientson the period dummy variables are thesame in all three equations. The second



32

a. Small Nonbound Banks, N=2209 b. Small Nonbound Banks, N=22091990 vs. 1991 1991 vs. 1992Frequency Frequency600

500

400

300

200

100

0 0

600

500

400

300

200

100

20.0

602

0.050

20.0

402

0.030

20.0

202

0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

20.0

602

0.050

20.0

402

0.030

20.0

202

0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

Figure 82

0.060

20.0

502

0.040

20.0

302

0.020

20.0

100.0

000.0

100.0

200.0

300.0

400.0

500.0

60

20.0

602

0.050

20.0

402

0.030

20.0

202

0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

c. Medium Nonbound Banks, N=451990 vs. 1991

d. Medium Nonbound Banks, N=451991 vs. 1992

Frequency Frequency121110

6

89

7

5

3

1

4

2

0

10

6

8

9

7

5

3

1

4

2

0

test shown in the table is based on an esti-mated single-equation regression thatincludes interaction effects between thereserve-maintenance period dummy vari-ables and the year dummy variables (for1990 and 1991, relative to 1992). The nullhypothesis of no year effects is againstrongly rejected. The test statistics shownin Parts A and B of Table 6 reinforce theinferences obtained from the fixed-effectsANOVAs: L-Nonbound banks also were E-Nonbound at the time of the 1990-91reduction, and larger L-Bound institutions

responded more strongly to the reductionsthan did smaller banks.

Our final repeated measures test isgraphical, shown in Figures 9 and 10. The test is based on the differences Dij,91 = yij,912yij,90 and Dij,92 = yij,922yij,91, respectively, where yijt is the ratio of basemoney to net transaction deposits held bydepository institution i in reserve mainte-nance period j, j = 1,...13, during year t, t = 90, 91, 92. Letting D•jt denote the meanof the Dijt for maintenance period j, (t = 91,92) then under suitable regularity condi-



33

Table 6

Test Statistics for Repeated Measures ANOVA ModelsNull Hypothesis: No Year-Period Interaction Effect for 1990, 1991 and 1992




MANOVA Model Wilks’ Lambda

Value of statistic 0.996 0.909 0.957 0.775 0.982 0.924F Value 2.02 25.1 2.67 6.52 20.1 1.77(degrees of freedom) (24, 27310) (24, 12262) (24, 2902) (24, 1150) (24, 52990) (24, 1054)p-value 0.0023 0.0001 0.0001 0.0001 0.0001 0.013

Univariate ANOVAF Value 1.98 27.8 1.51 5.92 20.45 1.75(degrees of freedom) (24, 27312) (24, 12264) (24, 2904) (24, 1152) (24, 52992) (24, 1056)p-value 0.0029 0.0001 0.0543 0.0001 0.0001 0.014


MANOVA Model Wilks’ Lambda

Value of statistic 0.982 0.891 0.952 0.768 0.997 0.942F Value 2.13 18.8 2.84 6.05 4.99 7.32(degrees of freedom) (24, 5638) (24, 7558) (24, 2710) (24, 1030) (24, 74662) (24, 5758)p-value 0.0011 0.0001 0.0001 0.0001 0.0001 0.0001

Univariate ANOVAF Value 2.67 20.3 1.52 6.10 3.62 7.57(degrees of freedom) (24, 5640) (24, 7560) (24, 2712) (24, 1032) (24, 74664) (24, 5760)p-value 0.0001 0.0001 0.0522 0.0001 0.0001 0.0001



tions a (1002

g) percent confidence inter-val for the null hypothesis that D•jt = 0 is

,

where t(11

g)/2 is the [(11

g)/2]th quantile ofthe t distribution with n-1 degrees offreedom.36 These confidence intervals areplotted as horizontal line segments in Fig-ures 9 and 10. The length of each linesegment shows the width of theconfidence interval, the numbers on thevertical axis index the reserve maintenanceperiod j, and the means D•jt are indicatedby the large dots on each line. We show

the confidence intervals graphically (ratherthan reporting significance levels or p-values for rejection/acceptance of the nullhypothesis of no response to a change inthe reserve requirement ratio) becauseinferences drawn from a graphical presen-tation likely are more robust to deviationsfrom the regularity conditions that justifyuse of the (asymptotic) t-distribution inconstruction of the intervals. A graphicalpresentation also is somewhat easier tointerpret than classical test statistics whenbetween-bank, within-group variances aresmall, as reflected in the short length ofthe confidence intervals.

Figure 9 shows banks’ responses to the1990-91 reduction in reserve-requirementratios. The results are clear: L-Boundbanks, shown in the second row of thefigure, reduced their holdings of basemoney relative to net transaction depositsby about 20.02, close to the estimatesobtained in the ANOVA models, while L-

D t

D D

n nD

t

D D

n n

j

ij j

ij

ij j

i

• /

•

•

/

•

,−−( )−( ) +

−( )−( )

+( )

+( )

∑

∑

1 2

2

1 2

2

1

1

γ

γ

36 See for example Mood, Graybilland Boes (1974), p. 387.



34

Confidence IntervalsFor the Difference between 1990 and 1991 Monetary Base-to-Net Transaction Deposit Ratios.

Small Medium Regional Large

Nonbound

Maint

enan

ce Pe

riod

Maint

enan

ce Pe

riod

20.05 0.00 0.05 20.05 0.00 0.05

123

7

54

6

8

10

12

9

11

13

123

7

54

6

8

10

12

9

11

13

Figure 9

Bound

Maint

enan

ce Pe

riod

Maint

enan

ce Pe

riod

20.05 0.00 0.05 20.05 0.00 0.05

1234

6

8

10

12

9

11

13

5

7

1

3

7

5

9

11

13

2

4

6

8

10

12

Maint

enan

ce Pe

riod

Maint

enan

ce Pe

riod

20.05 0.00 0.05 20.05 0.00 0.05

1

3

7

5

9

11

13

2

4

6

8

10

12

1

3

7

5

9

11

13

2

4

6

8

10

12

Nonbound institutions did not respond.Of special interest is the confidence inter-val for the change between the thirteenthreserve maintenance periods in 1990 andin 1991. The reserve-requirement reduc-tion was phased in, with only one-half ofthe reduction in force during the finalreserve maintenance period of 1990. E-Bound banks would be expected to respondto the phased reduction by displaying asmaller response for this thirteenth periodthan for the other twelve periods. Such aresponse is distinct for L-Bound banks andabsent for L-Nonbound banks. We concludethat: (1) L-Nonbound depository institu-tions likely were E-Nonbound before the1990-91 reduction, and hence should beexcluded from RAM; and (2) L-Boundinstitutions generally were E-Bound, andresponded as expected to the reduction inrequirements.

Figure 10 shows similar intervals forthe April 1992 reduction in reserve require-

ments. Small and medium-size L-Non-bound and L-Bound banks did notrespond: their confidence intervals eitherinclude, or are very close to, the origin.Although larger L-Bound banks reducedtheir holdings of base money, on averagetheir response was less than the 0.02reduction in the statutory requirement.The less-than-proportionate response per-haps reflects the falling federal funds rate;it might also be signalling, however, thatthe banks were becoming, or had become,E-Nonbound. The latter hypothesis is sup-ported by the observation that, by the endof 1992, about half of these banks hadagreed, through required clearing balancecontracts, to maintain Federal Reservedeposits in excess of amounts necessary tosatisfy their statutory reserve requirements.We conclude that by the end of 1992 onlya small number of U.S. depository institu-tions found statutory reserve requirementsgoverning their demand for base money.



35

Confidence IntervalsFor the Difference between 1991 and 1992 Monetary Base-to-Net Transaction Deposit Ratios.

Small Medium Regional Large

Nonbound

Maint

enan

ce Pe

riod

Maint

enan

ce Pe

riod

1

3

7

5

9

11

13

2

4

6

8

10

12

20.05 0.00 0.05 20.05 0.00 0.05

1

3

7

5

9

11

13

2

4

6

8

10

12

Figure 10

20.05 0.00 0.05 20.05 0.00 0.05

Maint

enan

ce Pe

riod

Maint

enan

ce Pe

riod

Bound

1

3

7

5

9

11

13

1

3

7

5

9

11

13

2

4

6

8

10

12

2

4

6

8

10

12

Maint

enan

ce Pe

riod

Maint

enan

ce Pe

riod

20.05 0.00 0.05 20.05 0.00 0.05

1

3

7

5

9

11

13

2

4

6

8

10

12

1

3

7

5

9

11

13

2

4

6

8

10

12

CONCLUSIONSThe evidence presented in this article

suggests that depository institutions’demand for the monetary base haschanged sharply since implementation ofthe Monetary Control Act in 1980.Increasingly, depository institutions aremaintaining deposits at Federal ReserveBanks to settle interbank payments, not tosatisfy statutory required reserves.Measures of the monetary base must beexpanded to include all the deposits heldby domestic depository institutions at Fed-eral Reserve Banks. The exclusion of thenominal amount of required clearing bal-ance contracts from the old St. Louismeasure and from the Board of Governors’measure, albeit perhaps justified in theearly 1980s, is today incorrect.

The importance of statutory reserverequirements as a determinant ofdepository institutions’ demand for basemoney also has diminished since 1980. Ifthe adjusted monetary base is to remaininterpretable as an index of quantitativemonetary policy actions, care must beexercised when combining the effects ofchanges in reserve requirements withchanges in the monetary base. Inparticular, the RAM adjustment includedin the St. Louis adjusted monetary basemust be redefined so as to include onlythose depository institutions that respondto changes in statutory reserverequirements by changing their holdingsof base money, or in other words, are E-Bound. Both the apparently less thandollar-for-dollar adjustment of many largebanks to the April 1992 reduction inreserve requirements and the continuingspread of OCD-based sweep programs sug-gest that few, if any, depository institutionswill be E-Bound in the future.

[Editor’s Note: In September 1996, theFederal Reserve Bank of St. Louis beganpublishing the new adjusted monetarybase measure that is presented in thispaper.]

REFERENCESAdvisory Commission on Monetary Statistics. Improving the Monetary

Aggregates, Board of Governors of the Federal Reserve System,1976.

Andersen, Leonall C., and Jerry L. Jordan. “The Monetary Base - Explan-ation and Analytical Use,” this Review (August 1968), pp. 7-11.

Anderson, Richard G., and Kenneth A. Kavajecz. “A Historical Perspectiveon the Federal Reserve’s Monetary Aggregates: Definition,Construction and Targeting,” this Review (March/April 1994), pp. 1-31.

_______ and Robert H. Rasche. “A Revised Measure of the St. LouisAdjusted Monetary Base,” this Review (March/April 1996)(1996a),pp. 3-13.

_______ and _______. “Redefining the Adjusted Monetary Basein an Era of Financial Change,” Research Working Paper 96-012,Federal Reserve Bank of St. Louis, October 1996. (1996b)

_______ and _______. “What Do Money Market Models Tell UsAbout How to Implement Monetary Policy?” Journal of Money, Creditand Banking (November 1982), pp. 796-828.

_______, James Johannes, and Robert H. Rasche. “A New Look atthe Relationship Between Time Series and Structural Models,” Journalof Econometrics (October 1983), pp. 234-51.

Balbach, Anatol B., and Albert E. Burger. “Derivation of the MonetaryBase,” this Review (November 1976), pp. 2-8.

Bank for International Settlements. Payments Systems in the Group ofTen Countries, Basle, Switzerland, December 1993.

Bank of Canada. “Implications of Monetary Policy in the Absence ofReserve Requirements,” Discussion Paper No. 1, September 1987.

_______. “The Implementation of Monetary Policy in a System withZero Reserve Requirements,” Discussion Paper No. 2, February 1989.

_______. “The Implementation of Monetary Policy in a System withZero Reserve Requirements,” Discussion Paper No. 3, May 1991,revised September 1991.

Brunner, Karl. “A Schema for the Supply Theory of Money,” InternationalEconomic Review (January 1961), pp. 79-109.

_______ and Allan Meltzer . “Liquidity Traps for Money, Bank Credit,and Interest Rates,” Journal of Political Economy, (Jan.-Feb. 1968),pp. 1-37. Reprinted in Monetary Economics, Karl Brunner and Allan H.Meltzer, eds., Oxford: Basil Blackwell, 1989.

Burger, Albert E. The Money Supply Process, Wadsworth PublishingCompany, 1971.

_______ and Robert H. Rasche. “Revision of the Monetary Base,”this Review (July 1977), pp. 13-27.

Cagan, Phillip. Determinants and Effects of Changes in the Money Stock,1875-1960, National Bureau of Economic Research, 1965.

Crowder, M.J., and D.J. Hand. Analysis of Repeated Measures, Chapmanand Hall, 1990.



36

Davidian, Marie, and David M. Giltinan. Nonlinear Models for RepeatedMeasurement Data, Chapman and Hall, 1995.

Deutsche Bundesbank. The Monetary Policy of the Bundesbank,Frankfurt: Deutsche Bundesbank, 1994.

Diggle, Peter J., Kung-Yee Liang, and Scott L. Zeger. Analysis ofLongitudinal Data, Oxford: Clarendon Press, 1994.

Federal Reserve Bulletin, various issues.

Feinman, Joshua. “Reserve Requirements: History, Current Practice and Potential Reform,” Federal Reserve Bulletin (June 1993), pp.569-89.

Friedman, Milton, and Anna J. Schwartz. A Monetary History of theUnited States, 1867-1960, Princeton University Press, 1963.

Gambs, Carl M, and Robert H. Rasche. “Costs of Reserves and theRelative Size of Member and Nonmember Bank Demand Deposits,”Journal of Monetary Economics (November 1978), pp. 715-23.

Gilbert, R. Alton. “Two Measures of Reserves: Why Are They Different?”this Review (June/July 1983), pp. 16-25.

_______. “Effectiveness of State Reserve Requirements,” thisReview (September 1978), pp. 16-28.

_______ and Jean M. Lovati. “Bank Reserve Requirements and TheirEnforcement: A Comparison Across States,” this Review (March1978), pp. 22-31.

_______ and Bruce J. Summers. “Clearing and Settlement of U.S.Dollar Payments: Back to the Future?” this Review (Sept./Oct.1996), pp. 3-27.

Hancock, Diana, and James A. Wilcox. “Intraday Management of BankReserves: The Effects of Caps and Fees on Daylight Overdrafts.”Journal of Money, Credit and Banking, (November 1996, Part 2), pp.870-908.

Hilton, Spence, Ari Cohen, and Ellen Koonmen. “Expanding ClearingBalances,” in Reduced Reserve Requirements: Alternatives for theConduct of Monetary Policy and Reserve Management, Ann-MarieMuelendyke, ed., Federal Reserve Bank of New York, 1993.

Hsiao, Cheng. Analysis of Panel Data, Cambridge University Press,1986.

Kohn, Donald L. “Comments on Anderson and Rasche,” this Review(19), pp. ••.

Mood, Alexander M., Franklin A. Graybill, and Duane C. Boes.Introduction to the Theory of Statistics, McGraw Hill, 1974.

Rasche, Robert H., and James M. Johannes. Controlling the Growth ofMonetary Aggregates, Kluwer Academic Publishers, 1985.

Richards, Heidi Willman. “Daylight Overdraft Fees and the FederalReserve’s Payment System Risk Policy,” Federal Reserve Bulletin(December 1995), pp. 1065-77.

Stevens, E.J. “Required Clearing Balances,” Economic Review, FederalReserve Bank of Cleveland (vol. 29, no. 4, 1993), pp. 2-14.

Tatom, John A. “Issues in Measuring An Adjusted Monetary Base,” thisReview (December 1980), pp.11-29.



37

Date post:	23-Jul-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

Measuring the Adjusted Monetary Base in an Era of ... · (a) Treasury cash holdings 0.271 (b)...

Documents