View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.


Prepared for the Steering Committee for the Fundamental Reappraisal of the
Discount Mechanism Appointed by
the Board of Governors of the Federal Reserve System

The following paper is one of a series prepared by the research staffs of the Board of Governors
of the Federal Reserve System and of the Federal Reserve Banks and by academic economists
in connection with the Fundamental Reappraisal of the Discount Mechanism.
The analyses and conclusions set forth are those of the author and do not necessarily indicate
concurrence by other members of the research staffs, by the Board of Governors, or by the Federal
Reserve Banks.


April 1968


Some Proposals for a Reform
of the Discount Window

Franco Modigliani
Massachusetts Institute of Technology




Goals to be Achieved


Proposed Devices to Improve Federal Reserve
Control Over the Money Supply While Permitting
Unrestricted Use of the Window


Sources of Slippage Between Nonborrowed Reserves
and the Supply of Demand Deposits, and How to
Reduce Them


Outline of Proposed Reform—Basic Features


Elaboration of the Proposal


Behavior of Free Reserves and Supply of Demand
Deposits Under the Proposed System
Choice of the Penalty Rate—Case for a
Sliding Differential


A Possible Minor Improvement:
Interest on Excess Reserves



Payment of

Proposal for Special Borrowing Facilities Aimed at
Improving the Spatial Allocation of Bank Credit


Proposal Outlined


How the Term Window Could Contribute to Allocative


Operational Aspects


How Large a Target Volume of Borrowing?


Some Minor Complementary Suggestions


- 1 I.



The purpose of this paper is to outline several proposals for
a reform of the Federal Reserve discount window.

These proposals

are aimed at achieving the following major goals:

To eliminate the discretionary and sometimes capricious

elements that characterize the present administration of the window
by permitting unrestricted use of such borrowing facilities by all
creditworthy borrowers willing to pay the discount rate.

To reduce the slippage that exists between nonborrowed

(bank) reserves and the supply of demand deposits by reducing swings
in free reserves, especially those of a procyclical character, and
thus improve control of the Federal Reserve over the money supply
and interest rates,

To make available to smaller banks facilities analogous

to those provided by the markets for Federal funds and certificates
of deposit, from which these banks are now partially or totally
excluded because of the small size of their operations.

To contribute through item (iii) above and other devices

to an improved spatial allocation of bank credit.
It is believed that the proposed reform would also make possible
the achievement of two other goals:

To eliminate the announcement effects that result from

sporadic and hence sizable changes in the discount rate, and

- 2 vi)

To provide a stronger inducement than now exists for banks

to become members of the Federal Reserve System, which would contribute to goal (ii)«

- 3 II.

This section discusses six devices that it is believed would

improve Federal Reserve control over the money supply while permitting unrestricted use of the discount window.

Sources of Slippage BetweenNcnborrowed Reserves and the Supply
of Demand Deposits, and How to Reduce Them
As is well known, the slippage that exists between the volume

of nonborrowed reserves, which the Federal Reserve controls largely
through open market operations, and the supply of demand deposits
can be traced primarily to variations in free reserves.

Our primary

concern in this section is with methods of reducing such variations
while at the same time keeping the discount window open to all
creditworthy borrowers willing to pay the price.
Other major sources of slippage—such as variations in the
reserve ratio as a result of shifts of deposits between banks with
different reserve requirements and between member and nonmember
banks, changes in time deposits, and currency drain—would not be
affected by the proposed reforms except indirectly through goal
(vi) above.
The operation of the discount window obviously affects free
reserves through borrowing.
volume of bank borrowing

It is reasonable to suppose that the

is influenced in part by the profitability

of borrowing, as measured by the spread between the discount rate

- 4 and short-term market yields; this supposition is supported by the
empirical evidence.

It is also generally agreed that a rise in

aggregate demand and economic activity tends initially to be
accompanied by a rise in short-term market yields unless it is
accommodated by a commensurate expansion of the money supply.—
Under these conditions, as long as the discount rate is kept unchanged,
a rise in demand tends to increase the profitability of borrowing;
and since the rise in market rates also tends to reduce the demand
for excess reserves, the result is likely to be a reduction of free
reserves and thus a procyclical movement in the money supply, relative to nonborrowed reserves.
Under the present system this tendency is, of course, moderated
by various limitations on the use of the discount window through
regulations, frowns, and suasion.

Such administrative limitations

in turn seem unavoidable so long as the discount rate is changed
only infrequently, permitting sizable fluctuations in the spread
between it and short-term market yields and corresponding variations
in the incentive to borrow.

Furthermore, since changes in the dis-

count rate have tended to occur infrequently, and often only after
some debate within the Federal Reserve System, they have come to
acquire a symbolic meaning (even if often a rather obscure one) apt



Section 1.4 below.

- 5 to generate wide repercussions.

And this very feature in turn has

contributed to the practice of avoiding frequent changes in the
Clearly the source of slippage between nonborrowed reserves
and the money supply described above would be reduced if the discount
window could be redesigned so as to minimize fluctuations in the
incentive to borrow.

The simplest way to achieve this result, of

course, would be to shut the window altogether.

But this solution

is clearly inconsistent with preserving the essential role of the
central bank as a "lender of last resort."

Individual banks must

have an outlet to which they can turn in case of "justified need"
and, similarly, some methods must be provided by which the banking
system as a whole can manage to satisfy reserve requirements in a
way that is not unreasonably painful.
However, there is no reason in principle why borrowing from
the discount window should not involve some significant penalty.
Accordingly, one possible device to limit significantly the use of
the window would be to set the discount rate at some level substantially above short-term market rates. There would then be an
incentive for banks to avoid the risk of having to borrow and to
repay promptly any borrowing that might have been incurred due to
miscalculations of or unanticipated contractions in nonborrowed
reserves. Yet banks could be allowed unrestricted use of the window

- 6 subject only to normal and prudent standards of creditworthiness,
for use of the window would be limited by the cost of borrowing,
without any need for fiat or frowns.
However, this approach has two major drawbacks:



the "penalty11 would depend on the relation between the discount
rate and market rates, it would still be necessary, in order to
keep the penalty reasonably uniform over time, to change the discount rate from time to time.
ment effects.


And that would perpetuate the announce-

The method would in effect discriminate against

small banks, which cannot make effective use of the Federal funds
market as a source of funds.

Indeed, if the banking system as a

whole were out of debt, which presumably would be the normal circumstance under a penalty borrowing rate, the Federal funds rate
would tend to hover below the rediscount rate and around short-term
market yields, say the rate on 3-month Treasury bills (hereinafter
referred to as 3-month bills; or in some instances, bills).


individual banks having access to that market could make up their
deficiencies at that cost.

Yet the smaller banks would have to pay

the significantly higher penalty rate.
It is suggested that these shortcomings could be eliminated,
while retaining the basic idea of a wide-open window at a penalty
rate, by reorganizing the operation of the window along the lines
set forth in sections 2 to 6 of P a r t H .

- 7 2.

Outline of Proposed Reform—Basic Features

The window would be open to all borrowers willing to pay

the discount rate as long as they met some appropriate tests of

To avoid uncertainties, each bank would be

informed about the maximum amount of accommodation that it could
expect to receive.

The ceiling would be reviewed at stated inter-

vals, except under special circumstances requiring a reappraisal
of the bank f s credit standing.

The borrowing rate would be tied to a short-term market

rate, say for the moment, the 3-month bill rate.

This device would

eliminate sizable, discontinuous changes in the discount rate and
associated announcement effects.

To maintain the penalty character of the window, the

borrowing rate would be fixed at, say, last week's bill rate plus
a fixed number of basis points, or plus a fixed percentage.


siderations relevant in setting the size of the penalty are set
forth later,

Borrowing at the window would be for very short terms--

usually for a single day—though automatically renewable at the
option of the borrower.

To avoid discrimination against smaller banks, the Federal

Reserve would provide, for such banks, accommodations similar to
those obtainable through the Federal funds (hereinafter abbreviated

- 8 FF) market.

Specifically, those entitled to the special accommoda-

tion would be allowed to borrow at the window at a daily rate equal
to that day's average FF rate plus a commission, consisting of a
fixed but moderate number of basis points or a moderate percentage
charge, as noted above. This facility could be provided for banks
not exceeding a certain size or at particular locations, or perhaps
more equitably, for loans not exceeding a stated modest size.


the latter device were adopted, one might expect that this facility
would, in fact, be used only by the smaller banks with inadequate
access to the FF market.
3. Elaboration of the Proposal
It should be noted that the reform outlined in section 2 is
only part of a broader plan.

Indeed, what has been proposed so

far would be of no help in achieving goal (iv)--improved spatial
allocation of bank credit—of Part I. To that end there is a
separate proposal, described in Part III, to provide facilities for
longer-term borrowing.

Accordingly, the rest of Part IE is concerned

only with the operation of the


1-day window."

The first questions that need to be considered are: To what
rate should the discount rate be tied?
premium be?

And how large should the

These two questions are closely interrelated.


it would be desirable to anchor the discount rate to the yield of
some market instrument of major importance--one that has a broad,

- 9 well-organized market.

This would insure that the chosen rate would

be "representative11 and relatively free of erratic movements.


this point of view, the 3-month bill rate would seem to be an obvious
choice, at least under present arrangements.
There are, however, two related problems to be considered.
First, any specific instrument may, at times, reflect special influences.

Second, there are some delicate issues involved in tieing

a 1-day rate to a 3-month rate, if at the same time borrowing is

In particular, "term structure effects11 (for example,

expectations of a forthcoming fall in the 3-month rate) might make
it profitable to borrow short at a rate negligibly higher than the
3-month rate.
To avoid these problems, it would seem desirable to peg the
discount rate substantially above the 3-month bill rate.


relevant guide in deciding on the size of the premium is provided
by the consideration that, with a truly open discount window, the
discount rate, by and large, would set the ceiling for the FF rate.
In other words, as the FF rate approached the discount rate, the
demand for funds would become highly elastic as would-be borrowers
turned to the window.

This consideration suggests that the premium

should be sufficiently large to allow the FF rate to deviate from
the bill rate as much as might be justified by term structure and
other special circumstances affecting the bill rate, without making
it "profitable11 to borrow at the window.

- 10 While it is impossible to set an absolute limit, the above
considerations suggest a premium on the order of 100 basis points,
a margin somewhat larger than the largest amount by which the
(weekly average) FF rate has exceeded the 3-month bill rate in
recent years.

(Unfortunately this experience is a very limited one

since, as is well known, until early 1965 habits and convention
prevented the FF rate from being bid above the discount rate; and
this, of course, also tended to distort the relation of the FF
rate to other rates.)
With such a differential, one could accommodate substantial
variability in the proper relation between the FF rate and the
short-term rate to which the discount rate was tied, without creating incentives to borrow at the window and hence without causing
undesirable flurries in the volume of borrowing.

An alternative,

and probably more effective, device to guard against this source
of difficulty is discussed below. This would be to rely on a floating differential.
But first, it is well to examine more closely both the longrun and short-run behavior of the proposed system--assuming the
premium over the bill rate to be substantial but of fixed size.
We propose to show that, under this system, free reserves would
tend to fluctuate rather narrowly around a substantially constant
"equilibrium11 level.

In other words, while fluctuations of free

- 11 reserves would not be (and should not be) altogether eliminated,
deviations of free reserves from the constant equilibrium level
would set up strong forces tending to move these reserves back toward

Behavior of Free Reserves and Supply of Demand Deposits Under
the Proposed System


The analysis that follows is based largely on some very definite

views as to the major forces that shape banks' portfolio management
and their use of the discount window, as well as the behavior of
short-term market yields. These views in turn appear to receive
strong support from the empirical analysis of recent experience
undertaken in the course of the MIT-FRB econometric research on the
working of stabilization tools.—
This evidence supports the view that the volume of free
reserves outstanding at any given time reflects two basic sets of
(1) An "equilibrium11 component, to wit, the desired or
equilibrium level of free reserves. This equilibrium level itself
2) Some of these results have already been published in de Leeuw
and Gramlich, "The Federal Reserve-MIT Econometric Model,11 Federal
Reserve Bulletin, Jan. 1968, especially pp. 13-16 and equations (1),
(2), (3), p. 31; and Rasche and Shapiro, "The FRB-MIT Econometric
Model: Its Special Features," American Economic Review, May 1968,
especially Section IV C. Others are contained in yet unpublished
memoranda of the project; it is expected that these results will be
published in the near future. (See in particular Modigliani and
Rasche, "Central Bank Policy and Money Supply," multilith, presented
at the 1967 meetings of the Midwestern Economic Association.)

- 12 is the difference between (i) desired excess reserves, which depend
on short-term market yields and tend to decrease as these yields
increase, and (ii) the optimum volume of borrowing at the window,
which is basically controlled by the spread between short-term
market yields--such as the FF rate or the 3-month bill rate--and
the discount rate.

(Note, however, that the FF rate is an adequate

measure of short-term yields only for the very recent period, in
which that rate was not conventionally kept at, or below, the discount rate.)

A disequilibrium component reflecting the inability and/or

undesirability for banks to adjust instantaneously to unforeseen
(or transient) changes in their deposits or in the demand for commercial loans.

The unforeseen changes in demand deposits in turn

reflect (i) unforeseen changes in nonborrowed bank reserves due to
Federal Reserve operations and changes in currency holdings (and
time deposits), and (ii) the unforeseen effect on demand deposits
of expansion and contraction in bank credit itself.

Component (ii)

implies that, even in the absence of the changes under (i), the
adjustment of free reserves to their equilibrium level tends to
occur gradually over time--somewhat along the lines of the textbook
description of the process of expansion of deposits in response to
an initial disequilibrium.

In addition, the process of adjustment

gets disturbed by the changes under (i). Thus free reserves tend

- 13 to be high when there is an unforeseen increase in nonborrowed reserves
or an unforeseen slackening in the demand for commercial loans, and
to be low when the unforeseen changes are in the opposite direction.
The evidence referred to above also supports the view that
changes in short-term market yields (say, the 3-month bill rate or
the commercial paper rate) are accounted for largely by the interaction of the supply of demand deposits, controlled by the forces
outlined above, and the demand for demand deposits, which is basically
controlled by current and past levels of income and short-term market

It further suggests that in the "short run11 (say, a quarter

or less) the level of income is largely unaffected by variations
in the money supply or short-term interest rates (at least as long
as these remain within realistic limits).

It therefore follows

that, in the short run, the level of short-term market yields is
controlled, in the last analysis, by the behavior of the outstanding
stock of demand deposits.

In the light of the above interpretation of bank behavior, we

can examine first what would happen to the "equilibrium11 level of
free reserves under the proposed system.

To this end, it is con-

venient to ignore initially operations at the special discount
window provided for small borrowings.
It should be evident that by floating the discount rate sufficiently above short-term market yields the desired or equilibrium

- 14 level of borrowing can be brought essentially to zero, and this
would be true regardless of the level of the short-term rate.


contrast, under*the present system a rise in market yields increases
the equilibrium level of borrowingsunless, and until, counteracted
by a rise in the discount rate.)

It then follows that the equili-

brium level of free reserves would itself tend to be constant except
for the effect of market yields on excess reserves. However, this
effect appears to be fairly moderate, except possibly for extremely
low levels of market yields where "liquidity-trap phenomena11 could
become significant.—

Furthermore even this effect could be elimi-

nated by the device of paying interest on excess reserves, as is
discussed in section 6 beginning on page 26 • But even without this
device we are led to the conclusion that the equilibrium level of
free reserves would tend to be stable (and prevailingly positive)
under normal conditions, except for some tendency to decline mildly
with the prevailing level of market yields*
Finally, we may note that if the system were in a position of
equilibrium, with borrowings near zero and excess reserves in
equilibrium, then one could expect the FF rate to hover around the
2/ Analysis of recent years suggests that an increase of 100
basis points in short-term yields—in, say, the 3-month bill rate—
tends to reduce excess reserves by somewhat less than $50 million
within 1 month and by somewhat less than $100 million within one to
two quarters.

- 15 bill rate. More precisely, in the absence of term-structure effects
(that is, if short-term rates were anticipated to stay unchanged in
the near future), it should tend to be quite close to the bill rate,
though term-structure effects could cause it to deviate, within
limits, on either side of the bill rate.

This proposition seems

fairly obvious and can be supported by a more rigorous analysis,
which we need not spell out here.—

The above considerations have

the following implication, which is important for an understanding
of the workings of the proposed reform:

provided the discount rate

were set significantly above the 3-month bill rate, and if the system
were in a position of equilibrium, the FF rate could be expected to
be significantly below the discount rate.
We can now reintroduce the special "Federal funds window'1 for
small operators and show that this does not significantly change the
above conclusion.

We need observe only that those who are eligible

to use the special window and who find it economical to do so would
be borrowing at a rate differing only by a small commission from the
{±1 The above conclusion rests on the assumption that bills will
continue to represent an important component of secondary reserves
and source of short-term liquidity. Should this premise lose its
validity and bills cease to be held by banks in significant quantity-except for the purpose of satisfying collateral requirements—then
the bill rate would no longer provide a reliable yardstick of shortterm yields and hence would no longer be a suitable rate on which to
anchor the discount rate.

- 16 rate available to any would-be borrower, namely, the FF rate.
Since the relation between the FF rate and short-term yields was
just shown to be such as to induce the banking system as a whole to
hold positive free reserves, regardless of the level of short-term
yields, we can infer that this relationship would provide the same
incentive to special borrowers as a whole.

Hence, they would also

tend to hold a positive and relatively stable amount of free
reserves; these reserves would probably exhibit some tendency to
move inversely, but moderately, with the level of market yields.
Note, however, that the amount of borrowing at the special, in
contrast to the regular, discount window would not be zero since
some operators would on the average be borrowing there, just as
many other banks would, on the average, be borrowing from the FF

All we are asserting is that the net free reserve position

of the group as a whole would tend to be stable.-

5/ At a more refined level of analysis one should recognize
that the incentive structure would be a little different for those
operating at the special window. In the first place, if they were
short of reserves, they would be paying somewhat more than the
larger operator using the FF market. In addition, they would
probably earn less if they were long on reserves. In fact, small
operators would probably tend to hold any excess funds in the
form of excess reserves yielding nothing, instead of lending them
in the FF market where they would yield the FF rate less transactions costs. (However, these qualifications do not require modifying our conclusions that their "equilibrium11 level of net free
reserves would tend to be stable though probably somewhat higher
than for the larger operators, and probably also somewhat less
responsive to variations in market yields.)

- 17 4.3

We can now examine the short-run, dynamic behavior of the model

in response to developments pushing it out of equilibrium.


analytical purposes we can distinguish between disturbances originating
in the economy and those originating from Federal Reserve actions,
though of course in general both types of disturbances may occur
simultaneously and reinforce or offset each other.

Consider first the response of the banking system to a

situation in which the Federal Reserve wished to hold down the
money supply and raise short-term market rates.

To this end the

Federal Reserve would force an (unanticipated) contraction in nonborrowed reserves (relative to the normal seasonal and secular

As a result, free reserves would initially fall short of

the planned level.

This implies an increase in the demand for,

and/or a decrease in the supply of, funds in the FF market, which
would immediately raise the FF rate.

If the Federal Reserve action

were sufficiently strong, the shortage of reserves would be such as
to push the FF rate to the ceiling provided by the discount rate.
At this point some banks would be induced to borrow at the window,
thus acquiring the additional reserves needed to satisfy reserve
requirements (plus the demand for excess reserves, probably somewhat reduced).

But now the higher cost of borrowing (whether at

the window or in the FF market) relative to other short-term market
yields would put pressure on banks to reduce their asset portfolios,

- 18 thereby shrinking the supply of demand deposits and required reserves,
until the borrowing had been eliminated and free reserves had moved
back to equilibrium.

In the process short-term market yields would,

of course, tend to move up, which is presumably what the Federal
Reserve intended.

But note that this rise would not per se reduce

the pressure for the banking system to get out of debt.


with the discount rate floating above the bill rate, a rise in the
latter would not reduce the incentive for individual banks to avoid
a net borrowed position.

Suppose instead that the Federal Reserve wished to expand

the money supply and lower short-term market rates, and accordingly
brought about an unanticipated expansion of nonborrowed reserves.
Here initially free reserves would exceed the planned amount, causing an increase in supply and a fall in demand in the FF market.
This would lower the FF rate relative to the bill rate, encouraging
an expansion of banks1 portfolios and the money supply and leading
to an increase in required reserves.

The incentive to expansion

would persist until free reserves had moved back to equilibrium,
and thus the FF rate had reestablished its equilibrium relation to
the bill rate.

Here again the bill rate would presumably fall in

response to bank expansion, as intended.

But this fall would not

reduce the incentive to expand the money supply until the additional
reserves had been absorbed by higher required reserves; for as long

« 19 as free reserves remained above equilibrium, the FF rate would tend
to remain below the rate on bills and other short-term instruments.

By relying on the reasoning developed earlier, one can

also readily establish that the pressures and responses described
in (i) and (ii) apply equally to the subset of "small11 operators
having access to the special "Federal fund window."
In summary, an (unanticipated) expansion or contraction of nonborrowed reserves would, initially, be reflected largely in opposite
movements of free reserves and of the FF rate, relative to the bill

But this would be only a temporary and (appropriate) cushion-

ing reaction.

For the movement of the FF rate in turn would generate

incentives to actions that would tend to bring free reserves back
to the initial equilibrium (except for the small effect of the
change in short-term market yields on excess reserves).

With free

reserves moving back to the original position, the supply of demand
deposits would tend to move commensurately with the change in
nonborrowed reserves.

The final change in short-term market yields

(that is, the bill rate) would then depend on the size of the change
in nonborrowed reserves and the (short-run) elasticity of demand
deposits with respect to short-term yields.

Consider next the effect of an increase in the demand

for money (an upward shift in the demand schedule relating money
demand to short-term yields).

This would tend to raise short-term

- 20 market yields, unless the money supply increased.

But there could

be no significant increase in the money supply so long as the Federal
Reserve kept the level of nonborrowed reserves unchanged.


under these conditions, the money supply could expand significantly
only through an increase in borrowing.

But since the rise in the

bill rate would be accompanied by a commensurate rise in the discount
rate, there could be no incentive for banks to expand their borrowings. With borrowing unchanged, free reserves would also be unchanged,
except again for a moderate decrease in response to the higher
market yields, and hence the money supply would be basically
unchanged, as stated above. Needless to say, if the Federal Reserve
wished to prevent the bill rate from rising, it could do so by
supplying additional reserves, in amount sufficient to increase
the supply of demand deposits pari passu with the increased demand.
The reasoning can be repeated mutatis mutandis, in the presence of
a decrease in the demand for money and falling interest rates.
(v) A different and Very important type of disturbance originating from the economy would be an (unanticipated) surge of demand
for commercial loans. Banks, as we suggested earlier, would
initially tend to accommodate the increase without a commensurate
reduction in the rest of their portfolio.

Hence, the supply of

demand deposits and required reserves would in the first instance

But with nonborrowed reserves unchanged, the system would

- 21 be short of reserves, and hence the FF rate would tend to be pushed
to the discount rate ceiling, opening up the window to an amount
of borrowing needed to satisfy reserve requirements.

But again the

increase in the cost of borrowing (whether at the window or in the
FF market) relative to short-term market yields would generate an
inducement for banks to reduce their portfolios and the supply of
demand deposits, until the borrowing has been eliminated and free
reserves had moved back to equilibrium.

Of course, the reshuffling

of bank portfolios would likely involve some net liquidation of
bills and other market instruments to accommodate the expansion of
loans, which would result in some increase in short-term market

But once more this would not modify the incentive for the

banking system to get out of debt to the Federal Reserve since the
discount rate would be moving pari passu with the bill rate.
Similar conclusions hold mutatis mutandis, if an unanticipated
decline occurred in the demand for loans.
The above analysis has one implication that is worth noting.
It should be apparent that under the proposed reform a level of
free reserves in excess of the constant equilibrium level would
tend to be accompanied by an expansion of the money supply.


more, the larger the excess, the larger the rate of expansion of the
money supply would tend to be.

Conversely, free reserves below that

constant level would tend to be accompanied by a contraction of

- 22 the money supply at a rate commensurate with the negative gap.


proposed reform would thus tend to validate a view of long standing
that there is a direct, reliable association between the volume of
free reserves and the rate of change of bank credit and the money

Yet, paradoxically, this view is not warranted under the

existing set-up in which the equilibrium level of free reserves is
not stable over time because of variations in the spread between
the discount rate and market yields.

It is not inconceivable that

reliance on that unwarranted view may have been responsible for
certain past failures in monetary management.

Choice of the Penalty Rate—Case for a Sliding Differential
We are now in a position to set forth the main considerations

that would seem relevant in setting the size of the differential
between the discount rate and the bill rate, or other short-term
rate, to which it was tied.
It follows from the analysis of section 4 that the essential
implication of a large differential is that banks would tend to
find it undesirable to stay substantially in debt for extended

But this means, in turn, that the volume of demand deposits

could be kept under close control by the Federal Reserve through
its control over nonborrowed reserves (and reserve requirements).
At the same time, since a temporary shortage of reserves could
push the FF rate as high as the discount rate, a large differential

- 23 would imply the possibility of substantial short-run variability
of the FF rate and related very short-term market rates.

By the

same token, a small differential would imply more limited variability of the FF rate but at the cost of tolerating a larger and
longer-lasting departure of the money supply from the level determined by nonborrowed reserves—that is, in essence, a looser coupling
between nonborrowed reserves and the money supply.
The above considerations suggest that the choice of the
differential would be dependent in large part on one's view concerning the nature of the monetary mechanism.

Those holding that the

cutting edge of monetary policy rests on the effects of such policy
on interest rates and related financial yields would presumably be
led to favor a set-up that minimized unintended movements of interest
rates and hence to prefer a relatively small differential.

On the

other hand, those leaning toward the view that the money supply
affects economic activity directly might well be led to favor a
system that minimized unintended movements in the money supply,
even if at the cost of larger short-run fluctuations in interest
In my view, however, a reasonable choice of differential does
not really require settling the thorny issue about the nature of
monetary linkages.

For whatever one's view on that issue, presum-

ably it is generally agreed that departures from the intended

- 24 course, whether of the money supply or of very short-term market
rates, can have a noticeable effect on the economy if they persist-but not if they are ephemeral.

And this is particularly true once

the rules of the game are well understood and stable and the participants have had a chance to adjust to them.
Hence, insofar as purely transient disturbances are concerned,
the choice of the differential is unlikely to be of real consequence.
On the other hand, in the case of marked and/or persistent departures,
the Federal Reserve would soon have to reach a decision as to
whether the most suitable response involved modifying the interest
rate target or the money supply target, or some combination thereof.
In such circumstances, the choice of the differential would therefore control only the character of the short-run, semiautomatic
response of the model while the Federal Reserve made up its mind
as to the appropriate eventual response.
In any event, the dilemma of choosing between a high or a low
differential could be avoided by adopting a "compromise" system,
which should prove largely agreeable to both points of view.


compromise would consist of tieing the discount rate to the bill
rate with a variable peft.

Under this scheme the differential would

remain fixed at some base level as long as aggregate borrowing at
the discount window remained below some stated amount.

But if bor-

rowing were to exceed this amount, then the differential would rise

- 25 with the volume of aggregate borrowing, according to a pre-established
By making the base level of the differential relatively modest,
moderate and transient variations in the demand for money could be
absorbed by an elastic money supply, with minor effects on market

This feature appears especially desirable in light of the

difficulty in determining reliably the demand-for-money schedule,
and hence the supply of deposits appropriate to a certain level of
short-term rates.

Yet, larger and more persistent disturbances,

while still initially accommodated at the window, would be accompanied by a larger increase in the differential cost of borrowing,
which would put pressure on banks to eliminate rapidly at least a
portion of their borrowing.

Such disturbances would thus tend to

be communicated promptly to interest rates, unless of course the
central bank decided to accommodate the larger demand by increasing
the supply of reserves, thus reducing the effect on interest rates.
Finally it should be noted that while the schedule of penalty
rates would influence the response to a tight money situation,
such a schedule would have little influence in shaping the response
to a loose situation, characterized by a rise in free reserves.
That response would be controlled by the fall in the FF rate below
market yields and by the speed with which banks would respond to
this situation by expanding their portfolios of earning assets.

- 26 6.

A Possible Minor Improvement:

Payment of Interest on Excess

I have noted, in setting forth the anticipated behavior of the
banking system under the proposed reform, that some variation in
free reserves would continue to be present because of the negative
association between equilibrium excess reserves and short-term
market yields.

Even this source of variation in equilibrium free

reserves could be largely, if not totally, eliminated by the simple
device of paying interest on excess reserves at a rate tied to
short-term market yields.
It is suggested that the most effective arrangement would be
to peg the interest on excess reserves a certain number of basis
points (or a fixed percentage) below the FF rate.

(The differential

could be thought of as something in the nature of a commission paid
to the Federal Reserve Bank for investing the free reserves of the
banks owning them.)
This arrangement could be expected to have the following major


If the differential were made sufficiently large--say

on the order of 50 basis points, or even somewhat larger—there
would still be an incentive, at least for larger banks, to invest
unneeded reserves directly in the FF market to avoid the differential.
Thus one would largely preserve a well-working FF market*


Smaller banks not having ready access to the FF market would be able

- 27 to derive an income from their reserve surpluses commensurate with
that earned by the larger banks, except for a reasonable commission,

Because those now relying on the FF market as an outlet for

their surplus funds would gain less from this activity than under
the old system—to be precise, they would earn the differential
instead of the full FF rate—one would expect that the equilibrium
demand for excess reserves would increase,


But under the

new system the amount gained by investing surplus funds in the FF
market instead of keeping them as excess reserves would become a
constant—the differential—that would be independent of the level
of the FF rate.

Thus, while the equilibrium level of excess reserves

would presumably tend to be generally larger than under the present
system, it would become independent of fluctuations in short-term
market yields.
Even greater stability in the level of excess reserves could
be achieved by use of a sliding differential similar to that proposed above for the discount rate.

That is, the differential would

be kept constant as long as excess reserves were below some stated
amount; but if excess reserves grew larger, the differential could
be increased--thereby encouraging investment in the FF market,
which would lead to a lower FF rate, and thus finally, through
portfolio expansion, to a reduction in excess reserves.
One important caveat must be entered at this point.

The sta-

bilization of excess reserves results from making the opportunity

- 28 cost of holding excess reserves independent of the level of market

However, a difficulty would arise if the FF rate became so

depressed as to be lower than the posted differential.

Under such

circumstances, because the interest paid on excess reserves cannot be
less than zero, the differential itself would have to be reduced-which would lead to an increase in the desired level of excess reserves.
What this means, of course, is that the payment of interest on excess
reserves is an effective stabilizer of excess reserves only so long
as the

banking system does not encounter liquidity traps; but it

affords no protection against a liquidity trap.—
j>/ It would be possible, in principle, to design the proposal so
that it could afford some protection even against situations of very low
returns from investments and associated very low market yields. But
this would require the radical step of applying the differential even
when the FF rate were so low as to imply a negative interest--or in
other words, the levying of a penalty—on excess reserves. The general
effect of such a penalty, of course, would tend to be that of making it
possible for market yields to become extremely low--indeed in principle
even negative. This, in turn, would clearly be a useful stabilizing
mechanism. I must hasten to add, however, that it is hard to say just
how effective this mechanism would in fact prove to be. For one thing,
faced with a penalty on excess reserves, banks not finding any
adequately yielding market instrument might endeavor to turn depositors
away. It is unlikely that they would refuse deposits outright-because of long-run considerations. They might instead have recourse
to service charges aiming at the same results. But this would still
be a stabilizing influence for it would imply a negative return for
holding money--a storage charge—which again would facilitate bringing
market rates to very low or even negative levels and would encourage
investments in real assets.
A great danger is the possibility that banks might artificially
increase their deposits, in order to absorb excess reserves, by making
ficticious loans to customers. It is impossible to predict just how
widespread such a practice might become, what its effects might be,
and how it could be prevented or limited. However, it hardly seems
worthwhile to dwell on the issue of a penalty on excess reserves because
for the moment at least, the likelihood of short-term market yields being
so low as to create real problems seems rather remote.

- 29 III.


This section describes in brief fashion a proposal for creation of special borrowing facilities, the purpose of which would
be to improve the allocation of bank credit--that is, goal (iv)

Proposal Outlined
The reform sketched in Part IE hopefully would go a long way

toward achieving the first three goals set forth in the Introduction.

But it would do little toward the fourth—improvement of

the spatial allocation of bank credit—except possibly insofar as
it would tend to make more uniform, across the banking system, the
cost of very short-term borrowing and the rate of return from very
short-term lending.
In order to achieve goal (iv) and as a further contribution
to goal (iii), I would like to advance a second proposal, the
adoption of which, incidentally, would be largely independent of
the implementation of the reform set forth in PartH.

The essence

of the proposal is to set up a second discount window (hereinafter
referred to as the "term window11)--one that would grant credit for
an essentially fixed term, say 3 months, not repayable until maturity
(except under special circumstances and/or with some appropriate

The term window too would be open to any bank willing

to pay the price, up to some limit determined by a creditworthiness

- 30 standard.

At the same time, lending conditions would again be

structured so as to stabilize the amount of borrowing, making it
independent of the level of short-term market yields or of other
indicators of monetary stringency.

But in contrast to the first

window, which would be designed to have a minimum of use, the
term window would be designed to function as a substitute for an
interbank loan market; and to perform this function on an adequate
scale, the volume of borrowing outstanding might have to be substantial.
A market for interbank lending seems to have developed only
to a very limited extent, except for overnight lending in FF market
and through the correspondent banking system--a rather surprising
phenomenon considering the very large number of banks that make
up the U.S.

banking system.

To be sure, a satisfactory spatial

allocation of funds could be achieved even in the absence of interbank lending, if there were adequate devices by which banks could
attract funds from "surplus11 areas to "short" areas.

But until

rather recently such a possibility has been very much limited by


of ceiling rates on time deposits and the prohibition

of interest payments on demand deposits.

More recently increases in

ceilings on time deposit rates and the development of a market for
CD's have presumably led to some improvements.

But there is reason

to believe that even these developments fall short of adequacy

- 31 since the CD market is, in practice, accessible only to large,
prime banks.

The proposed term window could also be regarded as a

device to extend to smaller banks facilities that are analogous to
those provided by the CD market.

How the Term Window Could Contribute to Allocative Efficiency
Before inquiring how the terms of borrowing could be set so

as to reconcile the goals of an open and extensively used window with
that of a stable volume of loans outstanding, it would be well to
ascertain in what sense the existence of the term window could be
expected to improve the spatial allocation of bank credit.
Basically, the answer lies in the consideration that a window
open to all on the same terms would tend to equalize the opportunity
cost of funds among banks, and hence presumably also the terms on
which credit would be available to would-be borrowers.

It might be

argued that this uniformity already tends to prevail, in the sense
that under the present system all banks have the opportunity to
invest in a common set of market instruments, and—what is more
important--they all do invest, by and large, in certain instruments
such as Treasury bills.

It would therefore appear that the Treasury

bill rate represents the common opportunity cost for all banks.
But this is, in fact, not a valid inference since such bills are
held not merely for their cash income but also, in part at least,
to satisfy liquidity requirements (as well as certain other

- 32 requirements).

One rather striking piece of evidence in support

of this proposition is provided by the observation that many banks
holding some bills in their portfolios have been willing to issue
CDfs at a significant premium over the rate on bills.
We must conclude then that, even though the cash yield is the
same for all holders, the "total11 yield, including the "liquidity"
component, need not be the same.

It follows that the opportunity

cost of funds invested in other assets need not be the same for all
banks, even if they all hold bills.

In particular one would expect

that when banks are compared on the basis of the relation of their
supply of funds for lending relative to their lending opportunities,
the opportunity cost would be higher for banks with lower supplyto-demand ratios than for banks with higher ratios.
Under these conditions we might expect that if banks that are
relatively short of funds were enabled to borrow from banks with
excess funds at a rate somewhat above the current bill rate they
would, within some limits, tend to take advantage of this opportunity.
The borrowing banks would then use the funds for expanding their loan
portfolios (and possibly even their portfolios of short-term market

For the lending bank the investment in the loan

would presumably displace other assets, including some loans. And
the redistribution of loans would presumably increase allocative

- 33 The proposed term window would accomplish the same general
result, though by a somewhat different route.

Suppose the rate at

the term window had been set somehow and that at this rate banks
would borrow a certain volume of funds with which to expand their

In order to accommodate this demand, while keeping total

reserves unchanged, the Federal Reserve would have to liquidate
some of its portfolio of market instruments.

This would raise

market yields, thereby encouraging some banks--presumably those
better supplied with funds— to acquire market instruments at the
expense of their other investments, including loans.

Thus the

final effect would be a redistribution of loans from more amply
provided to less well provided areas through a somewhat circuitous

In other words, the surplus bank would choose as a substi-

tute for direct loans to its regular customers not loans to the
less well supplied customers of other banks, but rather market
instruments such as Treasury bills; such purchases by the bank with
surplus funds would enable the Federal Reserve to exchange securities
for cash, which it would lend to the "short11 bank, which in turn
would use those funds to expand its loans.
It might be noted from the above that one implication of the
proposed reform might well be an increase in the yield on market
instruments, especially short-term ones such as bills.

As is well

known, this is an effect that typically tends to accompany any

- 34 restructuring of the financial system whose result is more reliance
on pure price rationing and less on other forms of rationing.


also follows from this analysis that the improvement in allocative
efficiency one might expect from the proposed reform would depend
on the views one had about the effectiveness of present arrangements
for the spatial allocation of funds.
3. Operational Aspects
Having thus laid out the basic argument in favor of a term
window, we can take up the problem of how to achieve simultaneously
an open window and a substantial and yet relatively stable volume
of borrowing.
Abstracting for a moment from "practicality,11 one could readily
suggest a device that would accomplish the desired aim.


one could auction off on a regular schedule, say every week, a block of
funds equal to the volume of loans that would come due in that week,
somewhat along the lines of the present bills auction.
This approach probably deserves consideration in the light of
the experience gained with the bills auction. Major drawbacks
might be (i) administrative complexity and (ii) the fact that the
auction might again give an edge to the larger banks, which are
better equipped to participate in it.

It is hard to say without

further careful study how serious these shortcomings might be.

- 35 As an alternative, it may be possible to "simulate11 closely
an auction by a device similar to that suggested for the 1-day

reset the borrowing rate at frequent intervals — say, once

a week—and tie that rate to a short-term market rate—say, the
3-month bill rate or the CD rate—with a flexible differential,
one that increases as the volume of borrowing increases.
With this arrangement one could not altogether avoid some variations in the volume of borrowing, but the variations could be kept
within moderate bounds.

One important feature that would tend to

insure this result is that the shrinkage in the volume outstanding
in any given week could not exceed the amount reaching maturity.
Assuming a 3-month maturity, this amount would be approximately
1/13 of the outstanding volume.

Furthermore, since this window

would not be designed as a device for meeting short-run reserve
requirements (which would be handled through the 1-day window), it
would be quite appropriate to require that applications for loans
to be taken down in a given week be filed some time in advance.
Under these conditions the Federal Reserve would know in advance
how the volume of borrowing at the term window would vary from
week to week and could, if it wished, offset such variations by
open market operations.
One could readily conceive of slightly more complex designs.
For instance, the window could announce, say, 2 weeks in advance two

- 36 or more possible rates and ask for preliminary applications at each
of the indicated rates.

On the basis of this information it could

set a final rate 1 week in advance and could accept as final all
applications received at that rate.

In short, it should not prove

too difficult to design a system that would minimize fluctuations
in the volume of borrowing outstanding and that furthermore could
offset any remaining fluctuations through open market policy.


should be noted in this connection that there is little reason to
be concerned with the danger that, in slack periods, the volume of
borrowings would shrink beyond control.

In fact, with a variable

differential one could always go so far as to push the rate to a
level below the bill rate, at which point the volume of borrowings
would obviously become highly elastic because any bank could make
a "hedged11 profit by borrowing and using the proceeds to buy bills.
(It is an open question whether under such conditions it would be
preferable to let the borrowing shrink and to let the central bank
purchase bills through open market operations*)

How Large a Target Volume of Borrowing?
It should be apparent from the discussion of section 3 that

the smaller the target volume of borrowing, the easier the task of
minimizing fluctuations in reserves caused by fluctuations in borrowings at the proposed term window.

But it should be equally

apparent that keeping the volume of borrowing low would reduce the

- 37 effectiveness of the proposal in achieving a better spatial allocation of bank credit.
To see how far these goals might be reconciled we might first
ask this question:

If one neglects the problem of stabilizing

borrowing, how large a volume of borrowing at the term window might,
in "the long run,11 be optimal?

An answer to this question might be

obtained by pursuing the idea that for smaller banks the term
window should provide an alternative to the CD market.
suggests the following answer:

This criterion

The volume of borrowing should be

such that the rate necessary to induce it would be somewhat above
prevailing CD rates for a maturity comparable to that offered at
the window.
To understand the rationale for the suggested criterion, we
may first note that a term-window rate close to the CD rate would
tend to equalize roughly the opportunity cost of funds for all
banks that were issuing CD's and/or using the window.

To be sure,

for banks not using either device the opportunity cost could be
lower, presumably as low as the bill rate.

However, since CD rates

have tended, at least so far, to stay reasonably close to the bill
rate, the difference in opportunity costs would remain within modest

At the same time it should be recognized that a lower

rate at the term window would hardly be feasible, unless the window
were somehow closed to banks issuing CDfs--which would seem quite

- 38 undesirable and even inconsistent with the spirit of the proposal.
The reason is that, with a completely open window, the borrowing
rate would tend to set a ceiling on the CD rate.

Or to look at

this from a different angle, the demand for borrowing at the term
window might be expected to become very elastic as the rate approached
the CD rate.
These considerations suggest an operational and pragmatic
approach toward the development of the term window.

Suppose the

Federal Reserve started out with a fairly modest target, say around
$2 billion to $3 billion, which would imply a weekly turnover of
$150 million to $200 million.

If it then turned out that the rate

needed to clear that volume of borrowing were on the average substantially above the CD rate, one could make two inferences:


that the window seemed to be contributing significantly to an
improved allocation, filling a function not performed by present

(this inference could of course be further tested

by examining the distribution of borrowing among banks and the
apparent use made of the marginal funds acquired); and (ii) that
there was a prima facie case for moving in the direction of increasing the target.

It would then be possible to plan to have such an

expansion occur gradually over time as experience was gained with
operation of the window and with problems that might conceivably

- 39 If on the other hand, even with a modest target, one should
find that the rate tended to hover close to the CD rate and that
the window was being used by banks that could have issued CD's,
then one could infer that the target should be reduced, or even
that the reform was contributing so little to the improvement of
the system as to justify abandoning it.

Some Minor Complementary Suggestions
(i) It would seem appropriate to make the term window available

only to banks that are members of the Federal Reserve System.


limited-availability feature, when coupled with the payment of
interest on excess reserves (and a graduated system of reserve
requirements), could go some distance toward providing an incentive
for nonmember banks to become members, contributing to a better
control of the money supply and short-term market rates.
(ii) One could relax the requirement that all borrowing at
the window be for a single fixed term and allow some choice of
terms, say between 2 months and 6 months. The rate for such loans
could be tied to rates for the corresponding maturities on the
chosen market instruments, be it CD's or bills, through a single
differential applied to all maturities.

However, this greater

flexibility would complicate the task of stabilizing the volume of
loans maturing in any given week.

It is not clear that this refine-

ment is worth the cost since banks could presumably manage, through

- 40 other transactions in short-term markets, to reconcile a fixed-term
borrowing with their requirements.

If the volume of borrowing at

the window were sufficiently large, one might, as an alternative,
conceive of developing some sort of secondary market, with or without the participation of the Federal Reserve.

Federal Reserve Bank of St. Louis, One Federal Reserve Bank Plaza, St. Louis, MO 63102