View original document

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

CO

o
GO
O
c

03

February 1980
Vol. 62, No. 2

QQ
CD

>

CD
CO
CD

Od
"cU
a5

TD
CD



3 On the Costs and Benefits of
Anti-Inflation Policies
15 Access to the Discount Window for All
Commercial Banks: Is It Important for
Monetary Policy?
25 The New M onetary Aggregates

The R e v i e w is published monthly by the Research Department of the Federal Reserve Bank of
St. Louis. Single-copy subscriptions are available to the public free of charge. Mail requests for
subscriptions, back issues, or address changes to: Research Department, Federal Reserve Bank of
St. Louis, P.O. Box 442, St. Louis, Missouri 63166.
Articles herein may be reprinted provided the source is credited. Please provide the Bank’s Re­
search Department with a copy of reprinted material.




On the Costs and Benefits
of Anti-Inflation Policies
LAURENCE H. MEYER and ROBERT H. RASCHE

A

l \ prominent policy issue of the 1970s and one
that seems certain to dominate the early 1980s is the
appropriate response to a prevailing high rate of in­
flation. The view that there is a long-run trade-off
between inflation and unemployment, widely accepted
at the end of the sixties, is now held by only a small
minority of economists. It is still widely believed,
however, that there is a short-run trade-off between
inflation and unemployment, which implies that re­
strictive macroeconomic policies designed to reduce
inflation would cause a temporary rise in the unem­
ployment rate. Therefore, both the time pattern of the
response of inflation and unemployment to demand
management policies and the relative cost of inflation
and unemployment remain dominant issues in the
design of macroeconomic policy.
There is relatively little consensus on either the costs
or benefits of reducing inflation. Both income-expenditure and monetarist macroeconometric models indicate
that a deceleration in monetary growth would gradu­
ally eradicate inflation but at a sizable cost in terms
of foregone output. In contrast, recent theoretical anal­
ysis based on “rational expectations” suggests that the
cost of reducing inflation could be small. Moreover,
the literature contains still less information on the
cost of inflation, which makes it difficult to obtain a
careful balancing of the costs and benefits of policies
intended to reduce or eliminate inflation.
This paper develops three views of the dynamics of
inflation and unemployment: the expectations-augmented Phillips Curve model, a monetarist model of
the relation of monetary change to both inflation and
unemployment, and a rational expectations model.
Based on each of these models, estimates of the cost
of reducing inflation are presented. Finally, the size
of the permanent per period gains associated with



eradicating inflation that would justify incurring these
temporary costs are estimated using both the Phillips
Curve and monetarist models.

THREE VIEWS OF THE RELATION
HETWEEN INFLATION AND
UNEMPLOYMENT
The Phillips Curve approach, as employed in
virtually all large-scale, income-expenditure macro­
econometric models, relates inflation to the unemploy­
ment rate and inflation expectations and almost uni­
formly specifies inflation expectations as dependent
exclusively on past actual inflation rates. Some mone­
tarists, however, prefer to relate both inflation and
unemployment directly to monetary change and reject
the regularity between inflation and unemployment
embodied in the Phillips Curve. A third view accepts
the Phillips Curve but introduces an alternative hy­
pothesis about expectations formation. This rational
expectations approach yields conclusions quite differ­
ent from the first two approaches.

The Phillips Curve
The Phillips Curve relates the rate of change in
nominal wages to both the unemployment rate and
the rate at which wages and/or prices are expected to
rise. This is usually combined with a mark-up model
of price determination in which prices are related to
wages. A combination of the two hypotheses yields a
relation between inflation (p ), unemployment (u),
and expected inflation (pe):
(l ) p = a + P u + p*, | < 0.
3

There are two basic sources of inflation identified
in equation 1: a demand factor and an expectations
factor. The unemployment rate is a proxy for the
3

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

balance between supply and demand in the labor
market. The lower the unemployment rate, the greater
the demand relative to the supply of labor. When
there is excess demand for labor, wages are bid up
at a rate proportional to the degree of excess demand.
Workers and firms bargain directly about nominal
wages, but labor supply and demand depend on the
real wage rate.1 Hence, the bargaining for nominal
wage increases over any period will also depend on
the rate of inflation expected over that period.
According to the Phillips Curve perspective, de­
mand management policies affect inflation by affect­
ing aggregate demand and, hence, unemployment.
While in principle economic policy could affect in­
flation expectations directly, the specification of the
Phillips Curve in macroeconometric models generally
assumes that inflation expectations are formed adap­
tively, that is, they depend exclusively on past infla­
tion rates. A simple form for such an equation is:
( 1 ') p t = a + p ut + pt-i,

where pt is the rate of inflation over some period, ut
is the average unemployment rate over the period,
and pt_! is last period’s inflation rate and, hence, this
period’s expected rate.
The role of lagged inflation in the Phillips Curve
may also reflect the direct influence of past wage
and price changes on current inflation due to catch-up
effects and long-term contracts. Contracts that fix
wages over some period, typically from one to three
years, permit current wage settlements to influence
wages and prices over the duration of the contract,
building an element of persistence or inertia into the
inflation process.2
In the specification given by equation 1', there exists
a critical unemployment rate consistent with either
price stability or constant inflation.3 Setting pt = pt-i,
the critical value of u is u* = -a/p. This is often
referred to as the natural rate of unemployment or
the nonaccelerating inflation rate of unemployment
JIf escalator clauses were both universal and complete, bar­
gaining would be in terms of real wages. The existence of
partial escalator clauses for some workers speeds the response
of wage to price change, but doesn’t alter the fact that nom­
inal wage bargains must directly reflect inflation expectations
over the duration of the contract.
2See John B. Taylor, “Staggered W age Setting in a Macro
Model,” American E conom ic Review, Papers and Proceedings
(M ay 1 9 7 9 ), pp. 108-13, for a model that incorporates both
forward-looking and backward-looking elements in the wage
setting decisions.
3The existence of an equilibrium or natural rate of unemploy­
ment independent of the rate of inflation depends on the co­
efficient of inflation expectations in equation 1 or lagged in­
flation in 1' being equal to unity.

4



FEBRUARY

1980

Figure 1

(NAIRU). Anti-inflation policy operates by raising u
above u°; as long as u is greater than u°, inflation de­
celerates and ultimately is eradicated.
Figure 1 depicts this cycling down process. First,
policy reduces aggregate demand. This raises u above
u° and induces a deceleration in inflation (link 1).
The decline in the actual inflation rate, in turn, re­
duces inflation expectations via the p i term (link 2),
which further reduces actual inflation in the next
period (link 3). As long as u remains above u°, this
cycling down continues. Ultimately, u returns to u*
when inflation has been fully eradicated. Thus, eradi­
cating inflation requires a temporary rise in the un­
employment rate during the transition to price
stability.

Monetarist Reduced-Form Equations
Stein has developed a “monetarist” framework for
assessing the relation between inflation and unemploy­
ment.4 Stein’s basic inflation equation can be expressed
as:
( 2 ) p t - pt-i = a ( m t - pt - i ) ,

where mt is the rate of monetary growth in period t.
According to this specification, inflation accelerates
when monetary growth exceeds the previous period’s
rate of inflation.
The distinctive feature of the monetarist equation
is not that it identifies monetary growth as the key
factor driving inflation while the Phillips Curve ig4Jerome L . Stein, “Inflation, Employment and Stagflation,”
Journal o f Monetary Econom ics (April 1 9 7 8 ), pp. 193-228.
Similar approaches have been presented by Carlson and
Tatom: Keith M. Carlson, “Inflation, Unemployment, and
Money: Comparing the Evidence from Two Simple Models,”
this R eview (September 1 9 7 8 ), pp. 2-6; and John A. Tatom,
“Does the Stage of the Business Cycle Affect the Inflation
Rate?” this R eview (September 1 9 7 8 ), pp. 7-15.

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

nores the influence of monetary change on inflation.
The Phillips Curve itself is consistent with monetary
change as a dominant influence on the inflation rate.
However, it is a specification of the structure of the
inflation process, that is, how monetary change affects
inflation. According to the Phillips Curve interpreta­
tion, a decline in monetary growth moderates inflation
by temporarily raising unemployment. The monetarist
equation, in contrast, directly relates monetary growth
to inflation and is essentially a reduced-fomi equation
relating inflation to policy instruments. In contrast to
the Phillips Curve, however, equation 2 allows onlij
monetary change to affect inflation. It also appears to
make the acceleration of inflation independent of de­
mand conditions in the economy. In fact, however,
the unemployment rate itself is also affected by mone­
tary growth in the Stein model.5 Hence, in both the
Phillips Curve and monetarist frameworks, a decline
in monetary growth both increases unemployment and
reduces inflation. Both approaches therefore permit us
to calculate the temporary rise in unemployment asso­
ciated with anti-inflation policy.6

Equilibrium Models and
Rational Expectations
Recently there has been renewed interest in equi­
librium models in which monetary change results in an
immediate proportional change in the price level.7
A useful point of departure here is a simple inflation
reduced-form equation in which the inflation rate
equals the rate of monetary growth:
( 3 ) p = m.

This differs from traditional monetarist models in
allowing the full effect of monetary change on prices
to occur immediately. The Phillips Curve then deter­
mines the unemployment rate. It is convenient to
rearrange equation 1 as:
( 4 ) u = u ° + 1/|3 (p - pe),

which demonstrates that unemployment deviates from
5Stein’s unemployment rate equation is presented later when
his model is used to derive the response of inflation and un­
employment to monetary change.
6Note, however, that inflation is expected to fall as long as
u > u* according to the Phillips Curve; in the monetarist
model, on the other hand, an acceleration in monetary growth
induces an acceleration in inflation even if the unemployment
rate initially is above its critical u“ level. Despite this dif­
ference, the two approaches yield similar implications for the
time path of unemployment and inflation in response to a
deceleration in monetary growth.
7Fo r example, see Thomas J. Sargent and Neil Wallace, “Ra­
tional Expectations and the Theory of Economic Policy,”
Journal o f Monetary Econom ics (April 1 9 7 6 ), pp. 169-85. The
model developed below is similar to the one developed by
Sargent and Wallace.




FEBRUARY

1980

its natural rate only in response to unanticipated in­
flation.
Monetary change determines the actual inflation
rate (via equation 3). Inflation expectations, accord­
ing to the rational expectations view, are formed on
the basis of the relevant economic theory — in this
case on the relevant model of the inflation process —
and are conditional on all relevant available informa­
tion. Taking the expected value of inflation from
equation 3,
( 5 ) p* = E ( p ) = E ( m ) ,

where E(m ) is the expected rate of monetary growth
conditional on information available prior to the
period over which the expectations apply, equation 4
can be rewritten as:
( 4 ') u = u* + 1 /p (m - E ( m ) ).

This implies that unemployment is affected only by
unanticipated monetary change.
In the previous two models, a deceleration in mone­
tary growth reduces the growth in nominal demand,
but inflation unwinds gradually; the decline in the
growth of nominal demand, therefore, initially falls
heavily on real demand and, hence, initially reduces
output and employment. In this model, in contrast,
monetary deceleration increases unemployment only if
the monetary deceleration is unexpected. If it is an­
nounced in advance and reflected in expectations,
a rapid deceleration of inflation results with no teno
porary rise in unemployment and, hence, no cumu­
lative output loss.
However, a question remains: How rapidly would
the expected rate of monetary growth decelerate even
if the Fed announced a policy of phased deceleration?
Barro’s attempt to implement this model empirically
holds that economic agents base their expecta­
tions of monetary growth in part on past monetary
growth.8 In a sense, such a specification substitutes
past actual rates of monetary growth for past actual
inflation rates and, therefore, does not entirely re­
move the inflation inertia embedded in the traditional
Phillips Curve. Meltzer also has recently emphasized
the gradual process whereby expectations of future
monetary growth respond to current observations of
monetary policy actions.9 The overall framework,
8Robert J. Barro, “Unanticipated Money Growth and Unem­
ployment in the United States,” American Econom ic Review
(M arch 1 9 7 7 ), pp. 101-15.
8Allan H. Meltzer;> “The Case for Gradualism in Policies to
Reduce Inflation, ’ Stabilization Policies: Lessons from the
1970s and Im plications for the 1980s (St. Louis: Washington
University, Center for the Study of American Business),
forthcoming.

5

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

FEBRUARY

1980

C hart ]

M oderation in Inflation: Perry Equation

R ate o f
W age

Change

Rate of
W age

Percent

Change
Pe rce n t

11.0

11.0

10.0
9.0

8.0
7.0

6.0
5.0
4.0
3.0

2.0
1.0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

1

Year

however, suggests that the Fed can minimize the
cumulative output loss by carrying out its anti-infla­
tion policy in a manner that makes it easy for the
public to discern its intent; this means developing
a reputation for meeting its monetary aggregate tar­
gets and reducing the volatility of monetary growth
so it is easy to recognize changes in the target rate
when they occur.

THE RESPONSE OF INFLATION AND
OUTPUT TO ANTI-INFLATION POLICY
Results Based on Estimated Phillips Curves
When developing an estimate of the cost of anti­
inflation policy using a Phillips Curve, anti-inflation
“policy” is identified with an increase in the unem­
ployment rate above the critical rate. Underlying the
change in the unemployment rate, but implicit in the
analysis, are changes in monetary and fiscal policy
instruments. As long as the unemployment rate is
maintained above the critical rate, inflation will de­
celerate. Based on the assumed initial inflation rate
and on the estimated parameters in the Phillips Curve,
it is simple to calculate how long it will take to eradi­
cate inflation.
The final step in estimating the cost of an anti-inflation policy is to convert the increased unemployment
into a measure of the cumulative output loss. This

6


is done via Okun’s Law: Each 1 percentage point
increase in the unemployment rate reduces real out­
put by 3.2 percent.10 Thus, at the 1978 value for
potential output, for example, a 1 percentage-point
rise in unemployment translates into a 45.6 billion
dollar loss in output. The cumulative but undis­
counted loss can be found by assuming that potential
output will rise at a 3.3 percent rate in line with pro­
jections by the CEA. The cumulative loss (L ) in this
case is:
(6)L =

Z
t

=

y ( u - u * ) (1 + p )',
0

where,
Y = the Okun coefficient,
u = the level of unemployment brought on by policy,
u#= the critical unemployment rate,
p = the rate of growth of potential output, and
n = the number of years required to eradicate
inflation.
10Arthur M. Okun, “Potential GNP: Its Measurement and Sig­
nificance,” from Proceedings of the Business and Economics
Statistics Section of the American Statistical Association
(1 9 6 2 ), pp. 98-104. More recent estimates of Okun’s Law
suggest that the output loss might be only 2.5 percent for
each 1 percent increase in unemployment. See, for example,
the estimates in the St. Louis model in Leonall C. Andersen
and Keith M. Carlson, “A Monetarist Model for Economic
Stabilization,” this R eview (April 1 9 7 0 ), pp. 7-25, and
Tatom’s discussion of Okun’s Law in “Does the Stage of
the Business Cycle?”

F E D E R A L . R E S E R V E B A N K O F S T . L O U IS

FEBRUARY

1980

Chart 2

C u m u la tiv e O utput Loss:
O utput

Perry Equation
O u tp u t Loss

Loss

lio n s o f 197 2 D o lla r s

B illio n s o f 1972 D o lla rs

2,200

2,200

2,000

2,000

1 ,800

1 ,800

1 ,600

1,600

1 ,400

1,400

1,200

1,200

1,000

1,000

800
600
400

200

1

12

13

14

15

1

Year

The discounted output loss is simply the product of
the initial year’s loss and the number of years required
to complete the program.11
The estimate of the cumulative output loss based
on a Phillips Curve equation presented in this section
is derived from the results of a study by Perry.12
Perry’s results are based on a wage change equation,
using the inverse of his weighted unemployment rate
and lagged wage change, estimated using annual ob­
servations over the 1954-77 period:
(7 ) A in W = -1 .8 8 + 7.44 (1 /U w ) + 0.79 A l n W - +
( - 2 .2 )
( 3 .5 )
(4 .6 )
0.21 A in W -, + 1.07 DNIX
SE = 0.70,
( 1 .1 )
(2 .9 )

where W is adjusted hourly earnings in the private
nonfarm sector and DNIX is a dummy for the con­
trols equal to -1 in 1972 and 1973 and -f-1 in 1974
and 1975. His preferred equation yielded a NAIRU of
4.0 in terms of his weighted unemployment rate
(which corresponds to about 5.5 percent in the of­
ficial unemployment rate in 1977). Hence, any un­
n This is the same as discounting future years’ losses by a 3.3
percent real interest rate.
12George L. Perry, “Slowing the W age Price Spiral: The Macroeconomic View,” Brookings Papers on Econom ic Activity
( 2 : 1 9 7 8 ), pp. 259-91. Similar estimates have been presented
by Okun ancf Gramlich. Arthur M. Okun, “Efficient Disinfla­
tionary Policies,” American Econom ic Review, Papers and
Proceedings (M ay 1 9 7 8 ), pp. 348-52; and Edward M. Gram­
lich, “Macro Policy Responses to Price Shocks,” Brookings
Papers on Econom ic Activity ( 1 : 1 9 7 9 ), pp. 125-66.




employment rate above 5.5 percent, if maintained
long enough, would eradicate inflation.
In the following simulations, inflation is assumed
initially to be 7.5 percent, and the economy is as­
sumed initially to be at the critical unemployment rate
(NAIRU). The time it would take to eliminate infla­
tion if unemployment were raised by either 1 or 3
percentage points is then calculated.
The response of inflation to a rise in the unemploy­
ment rate and the accompanying cumulative output
loss are depicted in charts 1 and 2: Perry 1 cor­
responds to a 1 percentage point rise in the unem­
ployment rate and Perry 2 to a 3 percentage-point rise.
Beginning with Ain W (approximately the percentage
change in the wage rate) equal to 10 percent, the
unemployment rate is raised above NAIRU and held
there until the rate of wage change declines to 2.5
percent, the rate presumed consistent with the trend
growth in labor productivity and, hence, with price
stability. When unemployment is raised 1 percentage
point, the rate of change in the wage rate falls from
10 percent to 9.6 percent in the first year and declines
about 0.3 percentage points per year thereafter, taking
23 years to reach the 2.5 percent rate consistent with
zero price inflation. The undiscounted cumulative out­
put loss is $1.5 trillion, and the discounted cumulative
output loss is $1 trillion.
7

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

FEBRUARY

1980

C hart 3

M o d e ra tio n in Inflatio n :

St. Louis a n d Stein M o d e ls
I n f l a t i o n Ra te

I n f l a t i o n Rate

Percent

Percent

8.0

8.0

5

6

7

8

9

10

11

12

0

Year

If unemployment is raised by 3 percentage points,
inflation is eliminated after 11 years. The cumulative
output loss, however, is greater in this case: $1.8 tril­
lion in the undiscounted case and $1.5 trillion in the
discounted case.13

Results Based on a Monetarist Model
According to Stein’s monetarist model, monetary
change affects both inflation and unemployment.
Stein’s two-equation model is:
( 8 ) Aut — 3.0 - 0.6 Ut-i + 0.4 p t-i - 0.4 m t~
i
( 9 ) Apt = - 0.4 p t-i + 0.4 m u ,

If monetary growth remains constant, inflation con­
verges to the rate of monetary growth, and unemploy­
ment converges to a constant rate, equal to 5 percent
in Stein’s model. Hence, the equilibrium rate of in­
flation equals the rate of monetary growth, and the
critical unemployment rate is 5 percent. If monetary
growth declines below the rate of inflation, inflation
13While most reported Phillips Curves yield high estimates
of cumulative output loss in line with Perry’s, there are
some that imply much lower estimates. Fo r example, see
the Phillips Curve presented by Phillip Cagan in “The Re­
duction of Inflation by Slack Demand,” in William Fellner,
Project Director, Contemporary Econom ic Problems in 1978
(Washington, D .C .: American Enterprise Institute for Public
Policy Research, 1 9 7 8 ), pp. 13-45. The cumulative output
loss based on Cagan’s equation is only about one quarter of
that implied by Perry’s equation.




decelerates and unemployment temporarily rises above
its equilibrium rate.
The simulation used to derive an estimate of the
cumulative output loss from Stein’s model differs from
that used in the Phillips Curve approach. Since inflation
and unemployment are both linked directly to mone­
tary change in the monetarist model, the rate of mone­
tary growth can be used as the policy instrument. As­
sume that the rates of monetary growth and inflation
are both 7.5 percent initially and that the economy is
at the equilibrium unemployment rate. Anti-inflation
policy is identified with a deceleration in the rate of
monetary expansion. Now, consider two scenarios: a
phased deceleration of monetary growth by 1 percent­
age point per year until the rate of monetary growth
declines to a rate which, if maintained, would be con­
sistent with price stability (zero in this model) and an
immediate deceleration to the rate consistent with
long-run price stability. Imposing these alternative
paths of monetary change on the model generates the
associated implied paths of inflation and unemploy­
ment; the rise in unemployment above 5 percent is
then translated into a measure of the cumulative out­
put loss.
The effects of each policy on inflation and output
loss are depicted in charts 3 and 4; the Stein 1 lines
correspond to the gradual deceleration in monetary

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

FEBRUARY

C ha rt 4

C u m u l a t i v e O u t p u t Loss:
O u t p u t Lo ss

St. L o u i s a n d S t e in M o d e l s
O u t p u t Loss

B illio n s o f 1972 D o lla rs

B illio n s o f 1972 D o lla rs

growth, the Stein 2 lines represent the more aggres­
sive policy. Under the gradual policy, unemployment
begins to rise in year 2, peaks in year 8 at 6.6 per­
cent, and returns to almost 5 percent by year 16. The
inflation rate, on the other hand, begins to decelerate
in year 2, initially at a 0.4 percentage point a year
rate, ultimately reaching 1.0 point per year by year
7. The inflation rate is down to 2 percent by year 8
and thereafter declines gradually to about zero by
year 16. The undiscounted cumulative output loss is
$687.5 billion. Interestingly, the more aggressive pol­
icy incurs a smaller undiscounted output loss, $613
billion.
Note that, qualitatively, the results are similar to
those based on the Phillips Curve: Restrictive demand
management policy temporarily will raise unemploy­
ment as it induces deceleration in the rate of inflation.
The size of the cumulative output loss in the Stein
model, however, is dramatically lower than that based
on Perry’s equation.

1980

key equations in the St. Louis model are
a reduced-form equation for the rate of
growth in nominal income based on the
Andersen-Jordan equation and an expectations-augmented Phillips Curve.15 The rate
of monetary change is the principal de­
terminant of the rate of change in nominal
income, although the rate of change in
high-employment government expenditures
also has a small, transitory effect. Thus, a
decline in the rate of monetary growth is
quickly translated into a decline in the rate
of increase in nominal income. The distri­
bution of the latter decline between prices
and output depends on the Phillips Curve;
the slower the deceleration of inflation as
nominal income falls, the greater the im­
pact of monetary change on output and the
greater the resulting cumulative output loss
of anti-inflation policy.

To begin, a base run in which the rate
of monetary growth is held steady at 7.5
percent from III/1968 through IV/1978
was generated. This builds in inflation in­
ertia and provides the base against which to evaluate
the effects of gradual monetary deceleration. Begin­
ning in 1/1973, monetaiy growth was gradually decel­
erated by 1 percentage point in the first quarter of
each year. The results from this policy run were then
compared with those from the base run and the cumu­
lative output loss associated with this policy was
derived by comparing the output solution assuming
monetary growth remains at 7.5 percent per year with
that assuming a phased monetary deceleration.16
The first set of simulations was generated using the
St. Louis model estimated over the sample period
I/1953-IV/1978. The results, labeled StLl, are re­
ported in charts 3 and 4. The inflation rate begins to
decline slowly; indeed, it takes 2 years to reduce it
by 1 percentage point. Thereafter, the deceleration
speeds up; after 5V2 years, inflation has declined by
7.5 percentage points. The unemployment rate initially
rises slowly, and the maximum increase is only 1.8

Results Using the St. Louis Model
The St. Louis model is in some sense a compromise
between the Phillips Curve and the monetarist reduced-form approaches developed above.14 The two

15Leonall C. Andersen and Jerry L. Jordan, “Monetary and Fis­
cal Actions: A Test of Their Relative Importance in E co ­
nomic Stabilization,” this Review (November 1 9 6 8 ), pp.
11-24.

14The St. Louis model is described in Andersen and Carlson,
“A Monetarist Model.”

16Because the model produces direct solutions for the response
of output to monetary deceleration, the use of Okun’s Law
is not required.




9

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

percentage points. The undiscounted cumulative out­
put loss is only about $200 billion.17
The estimate of the cumulative output loss in this
case is dramatically lower than for either the Stein
model or the Perry equation. However, the small size
of the cumulative output loss reflects, in part, the
suspiciously large coefficient on the demand slack
variable in the model’s Phillips Curve — almost three
times the size of the same coefficient estimated
through 11/1971 or 1/1975, for example. This rise in
the demand slack coefficient is neither readily ex­
plained nor mirrored in other estimates of Phillips
Curves. Consequently, two additional runs with modi­
fied versions of the St. Louis model were made.
First, a simulation of the model estimated through
III/1971 in which the coefficient on the demand slack
variable is substantially smaller was run. The results
are depicted by the StL2 lines in charts 3 and 4. In­
flation decelerates much more gradually; after six
years, it has declined by only 4 percentage points.
The cumulative output loss, already at $350 billion,
is escalating rapidly.
In the second modified version of the St. Louis
model, the Phillips Curve was replaced with a mone­
tarist reduced form for the inflation rate in which in­
flation depends on a 20-period distributed lag on the
rate of change in the money supply.18 The lines la­
beled StL3 in charts 3 and 4 present the implications
of gradual monetary deceleration on inflation and out­
put in this case. The results are remarkably similar to
those generated by the first modified version of the
St. Louis model (StL2 lines in charts 3 and 4). The
inflation rate declines somewhat more rapidly and
the output loss is a bit smaller, but both the time
pattern and magnitude of inflation deceleration and
output loss are very close.
The StL2 and StL3 simulations were not run long
enough to eradicate inflation and, therefore, are not
directly comparable with the Phillips Curve and mone­
tarist reduced-form results. Nonetheless, the results
17In the Perry and Stein results, the initial level of potential
output was that for 1978. The $200 billion cost estimate for
the St. Louis model is based on an initial level of income in
1973. To make the St. Louis result comparable with the
Perry and Stein results, it would be appropriate to multiply
it by a factor equal to the ratio of potential output in 1978
to that in 1973.
18The inflation reduced form was provided by Tatom and is
similar to the one he presented in “Does the Stage of the
Business Cycle?” In addition to the distributed lag on mone­
tary change, it also includes a four-quarter distributed lag on
the differential in the rates of change in producer prices for
. energy and the price index for the nonfarm business sector,
as well as two dummies, one for the effects of the freeze
and Phase II and one for the subsequent catch-up effect.




FEBRUARY

1980

at the end of six years were qualitatively similar
to the Perry and Stein results: Anti-inflation policies
impose a sizable cost in the form of lost output dur­
ing the transition to lower inflation rates.19

The Credibility Effect and
Rational Expectations
The Phillips Curve-based results reported above re­
lated inflation to a distributed lag of past inflation
rates, which implies a gradual unwinding of inflation
in response to anti-inflation policies. In this specifica­
tion, inflation expectations are formed exclusively on
the basis of past actual inflation. This ignores the pos­
sibility that the public will adjust their inflation ex­
pectations to both recent policy actions and expecta­
tions about future policy. A well-defined, credible
anti-inflation policy might induce a more rapid decel­
eration of inflation expectations than is suggested by
the conventional equations. Fellner, for example,
maintains that “. . . the standard model coefficients
. . . would change significantly for the better — in the
direction of a much more rapid rate of reduction of
inflation for any given slack — if a demand manage­
ment policy . . . changed to a credible policy of con­
sistent demand disinflation.”20
But, by how much do standard econometric ap­
proaches overestimate inflation inertia and the associ­
ated cumulative output loss? Unfortunately, reliable
quantitative estimates of the extent to which policy­
makers can speed the deceleration of inflation by
clearly defining their anti-inflation policies and con­
vincing the public that they intend to follow through
do not exist. Nevertheless, there is widespread agree­
ment that anti-inflation policies ought to be set out
clearly and supported by both the Federal Beserve
and the Treasury to maximize credibility.
There are, however, two empirical applications of
rational expectations macro models that provide some
insight into the predictions of that approach for the
response to a phased monetary deceleration. Paul A.
1!,Dewald recently presented simulations of the response of in­
flation, output, and unemployment to monetary deceleration
based on a modified version of the St. Louis model. William
G. Dewald, “Fast vs. Gradual Policies for Controlling Infla­
tion,” Federal Beserve Bank of Kansas City Econom ic R e­
view (January 1 9 8 0 ), pp. 16-25. He estimates the Phillips
Curve in the rate of change as opposed to the first difference
form used in the St. Louis model. This procedure does not
yield a coefficient on the demand slack variable as high as
in the St. Louis specification. Hence, Dewald also finds that
monetary deceleration yields a large cumulative output loss.
20William Fellner, “The Credibility Effect and Bational E x­
pectations: Implications of the Gramlich Study,” Brookings
Papers on Econom ic Activity ( 1 : 1 9 7 9 ), pp. 167-78.

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

Anderson modifies the St. Louis model by respecifying
its Phillips Curve to be consistent with rational expec­
tations.21 He begins with a Phillips Curve of the fol­
lowing form:
( 1 0 ) p = a + p x + ep* + e,

where x is a measure of demand slack in the economy,
and e is a random disturbance term with mean zero.
Instead of specifying pe as a distributed lag on past
actual inflation rates as in the equation in the St. Louis
model, Anderson imposes rational expectations by set­
ting p equal to the expected value of inflation based
on equation 10. Setting pe = E ( p), he solves for the
expected value of inflation:
(10-)E(P) = T^

+TL »

Anderson uses this equation to determine the infla­
tion rate in the St. Louis model, based on the estimates
of a , (3, and e from the St. Louis Phillips Curve. In
particular he sets e = .86. This procedure, in effect,
dramatically raises the response of inflation to changes
in demand slack. He runs simulations of the response
to an acceleration in the rate of monetary growth. In
the original St. Louis version, inflation increases grad­
ually and unemployment declines; in the rational ex­
pectations version, inflation increases more rapidly
and the effect on unemployment virtually disappears.
If e is viewed as the coefficient on expected infla­
tion, however, it seems inappropriate to employ its
value of .86 as estimated in the St. Louis model in
the rational expectations version of the St. Louis model
because it was estimated originally under the assump­
tion that expectations are formed adaptively. Taking
e = 1, as seems essential to the rational expectations
framework, equation 10' is no longer a meaningful
equation for p. Instead, setting pe = E(p ) and solv­
ing for E (p ), we obtain:
( 10" ) 0 = c c + P x ,

which indicates that there is a unique value of the
demand slack variable (x ° = -a/ p), corresponding,
of course, to the natural rate of unemployment. Only
random disturbances (with zero mean) can cause x
to differ from x*. In this case, the impact of monetary
deceleration on the rate of growth of nominal income
is transformed immediately and fully into a decline
in inflation without any cumulative output loss.
An alternative empirical application of a rational
expectations macro model is presented by Barro.22 He
-’■Paul A. Anderson, “Rational Expectations Forecasts from
Nonrational Models,” Journal o f Monetary Econom ics (Jan ­
uary 1 9 7 9 ), pp. 67-80.
22Barro, “Unanticipated Money Growth.”




FEBRUARY

1980

relates deviations in unemployment from its natural
rate to unanticipated monetary change ( as in equation
4' above) and, in addition, provides a model describ­
ing how economic agents form expectations about the
rate of monetary growth. On the surface at least,
Barro’s model seems well suited to provide an estimate
of unemployment’s response to a policy of phased
deceleration in monetary growth. The crucial issue
here is how rapidly economic agents learn that the
policy rule has in fact changed. They may learn this
from an announcement by the Fed. Given some doubt
about the Fed’s commitment to follow through on any
announced deceleration, however, economic agents
may insist on learning the new policy rule by observ­
ing the new pattern of monetary growth rates. This
involves reestimating the policy rule and incorporat­
ing new observations each period. Eventually, eco­
nomic agents will learn that the Fed intends to
decelerate monetary growth and then stabilize it at
a noninflationary rate. But this learning process may
take some time; meanwhile, monetary change will be
less than expected and unemployment will exceed
the natural rate. Hence, the Barro model also allows
for the existence of a cumulative output loss during
the transition to price level stability.23
The survey above provides the following cost esti­
mates: the extremely large estimate of the cumulative
output loss based on Perry’s Phillips Curve, the smaller
but still sizable loss based on Stein’s monetarist
model, the evidence from simulations with the St.
Louis model which, on balance, also suggest a large
output loss, and the rational expectations results as
modeled by Anderson, which suggest virtually no out­
put loss if monetary deceleration is perfectly antici­
pated. Unfortunately, in addition to the uncertainty
surrounding the actual cumulative output loss likely
to be associated with anti-inflation policy, there is
also uncertainty about the benefits to be derived from
eradicating inflation.

BALANCING THE GAINS
AGAINST THE COSTS
The cumulative output loss is a measure of the costs
of anti-inflation policies. To evaluate the desirability
of such policies, an assessment of the gains from re­
ducing inflation is required. Unfortunately, the costs of
23This is an application of the learning mechanism emphasized
by John B. Taylor, “Monetary Policy During a Transition to
Rational Expectations,” Journal o f Political Economy (O c­
tober 1 9 7 5 ), pp. 1009-21; and by Benjamin M. Friedman,
“Optimal Expectations and the Extreme Information Assump­
tions of ‘Rational Expectations’ Macromodels,” Journal of
Monetary Econom ics (January 1 9 7 9 ), pp. 23-41.

11

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

inflation (and hence the benefits of reducing inflation)
are not as clear cut or easily quantifiable as the costs
of unemployment. Currently, no studies provide esti­
mates of the benefits that would accrue from reduc­
ing or eliminating inflation, which could in turn be
compared directly to the cumulative output loss re­
quired to eradicate inflation. What can be computed,
however, is the minimum size of the permanent gain
in output per year due to the eradication of inflation
that would justify incurring the cumulative output
loss associated with the transition to price stability.

The Costs of Inflation
There are at least three dimensions to the costs of
inflation.24 First, there are the costs associated with
anticipated inflation that would be incurred even in
a fully indexed economy where institutions have com­
pletely adapted to an inflationary environment. Second,
there are the costs of anticipated inflation that arise
from a set of institutions that have only partially
adapted to the inflationary environment. Third, there
are the costs associated with unanticipated inflation
and uncertainty about the rate of inflation.
A fully indexed economy is one in which all nominal
payments and receipts (including wages, coupon pay­
ments, taxes, transfers, etc.) are tied (indexed) to the
inflation rate. All debt instruments except for currency
are indexed also. Currency is not indexed because in­
dexing it is assumed to be impractical.
The major costs of anticipated inflation in a fully
indexed economy can be labeled “menu” and “shoe
leather” costs. Menu costs refer to the resource costs
imposed by the necessity of frequent adjustments to
published price lists in an inflationary economy. Shoe
leather costs describe the costs incurred by more fre­
quent trips to the bank (or to the market) as a result
of the incentive to economize on currency holdings.25
Recently there has been a growing emphasis on the
costs imposed by inflation that reflect the existence
of institutions that are not fully adapted to an infla­
tionary environment. The major source of these costs
is the tax system, and the major effect is on saving
and investment incentives and, therefore, on capital
accumulation and the growth of output. The taxation
24The discussion of the costs of inflation in this section draws
upon the recent survey by Stanley Fischer and Franco
Modigliani, “Towards an Understanding of the Real Effects
and Costs of Inflation,” W eltw irtschaftliches Archiv ( 4 :
1 9 7 8 ), pp. 810-33.
25For a discussion of the welfare cost of anticipated inflation,
see John A. Tatom, “The Welfare Cost of Inflation,” this
R eview (November 1 9 7 6 ), pp. 9-22.


http://fraser.stlouisfed.org/
12
Federal Reserve Bank of St. Louis

FEBRUARY

1980

of personal interest income, for example, may induce
a decline in the after-tax real rate of return to savers
as inflation increases.26 Furthermore, the tying of the
depreciation deduction for tax purposes to the histor­
ical rather than the replacement cost of capital goods
tends to raise the cost of using capital goods in an
inflationary environment. While inflation has many
other effects on saving and investment, there is grow­
ing concern that its net effect is to discourage both
saving and investment.
Unanticipated inflation imposes costs by inducing
redistributions of income and wealth. These “transfer
effects” arise because contracts have been written in
nominal terms embodying expectations about future
inflation which turn out to be incorrect. The social
cost of such redistributions is difficult to assess because
there is a gainer for every loser. However, many con­
sider the “transfer” costs associated with unanticipated
inflation the most serious cost associated with
inflation.27
Uncertainty about the inflation rate may impose ad­
ditional costs by increasing the uncertainty associated
with the outcome of economic decisions. Above, a cost
was ascribed to the actual redistributions that follow
from unanticipated inflation. There may also be utility
losses associated with the increased likelihood of such
arbitrary transfers when there is considerable uncer­
tainty about expected inflation. A number of studies
have suggested that inflation uncertainty tends to be
systematically related to the level of inflation. If this
is the case, reducing the level of inflation will also re­
duce inflation uncertainty.28
Fischer and Modigliani do not provide estimates for
the various effects of inflation since “any measures
26For a discussion of the effect of taxation of interest income
on the response of nominal interest rates to a change in
expected inflation, see Martin Feldstein, “Inflation, Income
Taxes, and the Rate of Interest: A Theoretical Analysis,”
American Econom ic Review (Decem ber 1 9 7 6 ), pp. 809-20.
27Franco Modigliani and Lucas Papademos, “Optimal De­
mand Policies Against Stagflation,” W eltw irtschaftliches
Archiv ( 4 : 1 9 7 8 ), pp. 736-82.
28See, for example, Arthur M. Okun, “The Mirage of Steady
Inflation,” Brookings Papers on E conom ic Activity (2 :1 9 7 1 ) ,
pp. 485-98; and Dennis E . Logue and Thomas D. Willett,
“A Note on the Relation Between the Rate and Variability
of Inflation,” Econom ica (M ay 1 9 7 6 ), pp. 151-58. These
studies generally associate an increase in inflation uncertainty
directly with an increase in the welfare cost of inflation. This
follows only if increased uncertainty about inflation increases
uncertainty about real income, real wealth, etc. For a dis­
cussion of the relation of inflation uncertainty to the welfare
cost of inflation, see Lionel Kalish III, Laurence H. Meyer,
and David H. Resler, “Inflation Uncertainty and the Welfare
Cost of Inflation,” mimeographed (Federal Reserve Bank of
St. Louis, 19 8 0 ).

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

FEBRUARY

1980

F ig u r e 2
Output

Table 1
The Minimum Value of the
Per Year Gain (g) That Justifies
Eradicating Inflation (billions of
1972 dollars)
Equation/
model

Value of g

Perry 1

$73.0

Perry 2

Evaluating the Minimum Gain Per Year
Required to Justify Policies to
Eradicate Inflation
The solid line (X ) in figure 2 is the rate of growth
of potential output if inflation remains indefinitely at
7.5 percent. If anti-inflation policies are pursued, out­
put is assumed to follow the dashed line (X '). The
transitional costs of eradicating inflation occur be­
tween t0 and tn as unemployment rises above the rate
associated with potential output.
The cost of inflation may involve decreases in po­
tential output due to disincentives to saving and/or
investment and/or welfare losses due to anticipated
or unanticipated inflation. The benefit of eradicating
inflation is shown in figure 2 as an increase in output
above the level that would have prevailed had infla­
tion continued to average 7.5 percent; hence, the
29Fischer and Modigliani, “Towards an Understanding of the
Real Effects and Costs of Inflation,” p. 813.
30This approach was suggested to us by Jerry Jordan and
Allan Meltzer.




31.0

Stein 2

would be totally speculative at this stage.”29 In the
absence of a measure of these costs, however, it is
possible to compute the minimum total costs associ­
ated with continued inflation necessary to justify in­
curring the previously calculated costs of eradicating
inflation.30

70.9

Stein 1

25.4

dashed X' line rises above the solid X line after tn.31
This analysis emphasizes the necessity of comparing
the transitional cost incurred over the period during
which inflation is eradicated with the permanent bene­
fit attributable to the eradication of inflation.
G is the present value of the permanent per period
output gain, evaluated from period n to oo:

To simplify further, gt is assumed to be constant for
all t > n. The value of g, which equates the dis­
counted cost of unemployment and the gain from
eradicating inflation is then calculated. This is the
minimum value of the permanent per period gain from
eradicating inflation that would justify incurring the
transitional costs. The values of g, based on transi­
tional costs estimated from the Perry equation and
Stein model and calculated under the assumption of
a 3.3 percent discount rate, are presented in table 1.
The minimum value of g varies from about $25 billion
per year in the Stein model to $73 billion based on
Perry’s Phillips Curve under a moderate policy.
This analysis provides an alternative perspective
on the case for gradualism. A gradual policy will im­
pose a lower cost of eradicating inflation if the Phil­
lips Curve is nonlinear. For Perry’s nonlinear Phillips
Curve, for example, the discounted cost was $1.0 bil­
lion for the gradual policy and $1.5 billion for the
31The gains of reducing inflation should begin being regis­
tered during the transition period. To simplify the calcula­
tion, the benefits are assumed to begin only at t„, when infla­
tion is eradicated. This assumption biases the calculation of
the present value of benefits downward.

13

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

more aggressive policy. A more gradual policy, how­
ever, also postpones the benefits from eliminating in­
flation. The size of the permanent per period gain
required to justify the anti-inflation policy may there­
fore be smaller under the more aggressive policy. In­
deed, this is the case for the Perry results. Although
the cumulative output loss is smaller under the grad­
ual policy (Perry 1), the size of the per period gain
required to justify eradicating inflation is smaller
under the more aggressive policy (Perry 2). The more
aggressive policy also vields a smaller minimum per
period gain using the Stein model, although this was
expected since the cost turned out to be lower in the
radical case using Stein’s model.
The calculations reported above presumed that the
gains from reducing inflation could be meaningfully
represented as a fixed real sum per period. Suppose,
instead, that the gains are more meaningfully specified
as a real sum which grows at the same rate as poten­
tial output. For example, the cost of a fully anticipated
increase in inflation is generally measured by the re­
duction in the area under the demand curve for real
money balances as wealthowners reduce their demand
for money in response to the associated rise in nomi­
nal interest rates. The decline in demand for real
money due to a rise in the interest rate is generally
considered proportional to the overall scale of money
holdings which, in turn, is determined by the level
of transactions (e.g., real income). The cost of a given
rate of inflation and, hence, the benefits of eliminating
inflation may, therefore, grow at the rate of increase
of potential output. In this case:


http://fraser.stlouisfed.org/
14
Federal Reserve Bank of St. Louis

FEBRUARY

1980

where g is the value of the gain in period n (the first
period in which a gain is registered). For p > r, G
oo. This corresponds to the result recently derived
by Feldstein: If the cost of inflation grows at a rate
equal to or greater than the discount rate, any positive
initial gain (any g > 0) is sufficient to justify incurring
any finite transitional cost!32

CONCLUSION
The existence of large transitional costs of eradicat­
ing inflation is not a sufficient reason to reject anti­
inflation policies. The potential existence of large
benefits associated with reducing inflation requires
a careful assessment of the net effects of anti-inflation
policies. Unfortunately, the range of the estimates of
the cumulative output loss, the uncertainty about the
adjustment in those results required to allow for the
credibility effect, and the lack of quantitative esti­
mates of the cost of inflation make it extremely diffi­
cult to obtain a meaningful comparison of the costs
and benefits of an anti-inflation policy. Narrowing the
range of estimates of output loss and developing a
measure of the benefits associated with anti-inflation
policies should be high on the priorities for macroeconomic research in the 1980s.

:i-Martin S. Feldstein, “The Welfare Cost of Permanent Infla­
tion and Optimal Short-Run Economic Policy,” Journal of
Political Econom y (August 1 9 7 9 ), pp. 749-67.

Access to the Discount Window for All
Commercial Banks: Is It Important
for Monetary Policy?
R. ALTON GILBERT

ROPONENTS of legislation to induce or require
more banks to be members of the Federal Reserve
System often argue that the existence of nonmember
commercial banks creates problems for the conduct of
monetary policy. Two of the most frequently men­
tioned of these problems are: First, the ratio of net
demand deposits to bank reserves may become more
variable as nonmember banks have a larger share of
demand deposit liabilities.1 Second, as banks with­
draw from membership, the average ratio of net de­
mand deposits to reserves rises. With a lower average
reserve base, fluctuations in reserves due to such
factors as changes in float and currency holdings of the
public make net demand deposits more variable.2
These potential problems have been subjected to ex­
tensive theoretical analysis.
Recently, a third possible problem with declining
membership has been suggested — the fact that non­
member banks do not have access to the discount
window on a day-to-day basis. Lack of direct access
to the discount window for all commercial banks, it
has been alleged, may preclude the adoption of appro­
priate monetary policy because of the Federal Re­
serve’s concern about the differential impacts of these
policies on member and nonmember banks. During
periods of tight monetary policy, for example, declin1Dennis R. Starleaf, “Nonmember Banks and Monetary Con­
trol,” Journal o f Finance (September 1 9 7 5 ), pp. 955-75; and
Kenneth J. Kopecky, “Nonmember Banks and Empirical
Measures of the Variability of Reserves and Money: A Theo­
retical Analysis,” Journal o f Finance (M arch 1 9 7 8 ), pp. 31118.
2See, for example, J. A. Cacy, “Reserve Requirements and
Monetary Control,” Federal Reserve Bank of Kansas City
Monthly R eview (M ay 1 9 7 6 ), pp. 3-13.




ing Federal Reserve membership might increase the
liquidity risk for the entire banking system, since
fewer banks would be able to use the discount win­
dow to provide a temporary offset to unexpected
reserve outflows.3
The validity of this argument hinges on whether
credit from the discount window significantly would
help nonmember banks adjust to deposit withdrawals.
Under existing legislation, the Federal Reserve has the
authority to lend to nonmember banks in unusual
emergencies in which these banks would fail without
additional reserves.4 Increasing the number of banks
that are members, therefore, would not increase the
ability of the Federal Reserve to respond to such
emergency situations.
3G. William Miller, “Statement,” Monetary Control and the
M em bership Problem , U.S. Congress, House, Committee on
Banking, Finance, and Urban Affairs, 95th Congress, 2nd Ses­
sion, July 27, 1978, pp. 60-62; and statements by Sen. William
Proxmire and Paul Volcker, Chairman of the Federal Reserve
Board, before the U.S. Senate Committee on Banking, Hous­
ing, and Urban Affairs, February 4, 1980.
4Credit from Federal Reserve Banks is classified as reserve ad­
justment credit, seasonal credit, and emergency credit. Adjust­
ment credit is available for member banks to meet unexpected
temporary credit demands caused by sudden deposit with­
drawals or unanticipated loan demand. Seasonal credit is avail­
able to relatively small member banks that have seasonal
patterns in their deposits and loans. Emergency credit may be
made available to member or nonmember banks with severe
financial difficulties. For additional information on the condi­
tions under which the Federal Reserve makes credit available
to banks, see R. Alton Gilbert, “Benefits of Borrowing from
the Federal Reserve when the Discount Rate is Below Market
Interest Rates,” this R eview (M arch 1 9 7 9 ), pp. 25-32. This
paper analyzes use of adjustment credit by member banks.
For an analysis of seasonal borrowing, see Stanley L . Graham,
“Is the Fed’s Seasonal Borrowing Privilege Justified?” Federal
Reserve Bank of Minneapolis Quarterly Review (F all 1 9 7 9 ),
pp. 9-14.

15

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

The type of monetary policy that the Federal Re­
serve may be precluded from adopting out of concern
for liquidity pressures on nonmember banks is pre­
sumably not that designed to create liquidity emer­
gencies for the banking system. Concern that declin­
ing membership would increase liquidity risk to the
banking system suggests, rather, that nonmember
banks have greater difficulty than member banks in
adjusting to unanticipated deposit withdrawals or
changes in demands for credit in more normal cir­
cumstances than those emergencies in which the
Federal Reserve would make credit available to non­
member banks. Since monetary policymakers are jus­
tifiably concerned about possible influences of Federal
Reserve membership on the conduct of monetary
policy, it is important to clarify whether lack of direct
access to the discount window for nonmember banks
does, indeed, pose a serious problem for monetary
policy.
This article takes an indirect approach to determin­
ing whether nonmember banks have greater reserve
management difficulties due to lack of access to the
discount window. If borrowing from the discount
window is a useful way to adjust to unanticipated
reserve outflows, member banks would borrow from
the discount window on at least a few occasions each
year, taking advantage of their regular, reliable, dayto-day access to the discount window to meet unex­
pected withdrawals or to cushion themselves against
temporary liquidity pressures. If, however, most mem­
ber banks manage their reserve positions without
borrowing at the discount window, it is doubtful that
nonmember banks have significantly greater difficulty
than members in responding to similar reserve out­
flows. In this case, lack of access to the discount win­
dow for nonmember banks is simply irrelevant to the
membership issue.

DO MEMBER BANKS BORROW
FREQUENTLY FROM THE
DISCOUNT WINDOW?
Most member banks do not borrow from the dis­
count window; those few that do so generally bor­
row infrequently. From 1974 to 1977, the proportion
of the 430 member banks in the Eighth District that
borrowed in any one year was as high as 25 percent
only in 1974, when the discount rate was substantially
below alternative short-term interest rates.5 Of the 115
member banks that borrowed during 1974, only 21
5Fo r further discussion of member bank borrowing, see Gilbert,
“Benefits of Borrowing from the Federal Reserve.”


16


FEBRUARY

1980

did so on more than 10 occasions, and only 55 bor­
rowed five or more times (table l ) . 6 During 1975,
only nine banks borrowed five times or more, and in
1976 only 10 banks borrowed that frequently.
The infrequent borrowing of member banks from
the Federal Reserve indicates that, in most circum­
stances, they adjust to reserve losses without resort
to the discount window. This is sufficient evidence for
rejecting the view that nonmember banks necessarily
have greater problems than member banks in coping
with reserve outflows because they lack direct access
to the discount window. If borrowing from the dis­
count window were an important means for banks to
adjust their reserve positions to reserve outflows, most
member banks would borrow from the discount win­
dow.7 Yet, they do not do so.

HOW DO MEMBER BANKS ADJUST
TO RESERVE DRAINS?
Most member banks manage their reserve positions
by means other than borrowing at the discount win­
dow. This has implications for the significance of the
discount window for the banking system’s operation.

Effects of Deposit Fluctuations on the
Reserve Positions of Member Banks
Cash management by member banks is investigated
by considering the factors that determine their de­
sired cash holdings, the types of events that cause
their cash holdings to be different from desired levels,
^Although Reserve Banks classify borrowing by member banks
as reserve adjustment or seasonal credit, it is often difficult
to distinguish the purpose of borrowing in actual practice.
Member banks that have the privilege of receiving seasonal
credit often change the amount of seasonal credit they borrow
daily or weekly. Use of the discount window for reserve
adjustment is gauged in this article by the number of occa­
sions on which a bank borrows from the discount window
dining a year, measured as the number of times when
borrowing is positive followed by periods when borrowing
is zero. Member banks that use the discount window for
reserve adjustment on a routine basis tend to borrow on
several occasions each year.
"This conclusion is reinforced by noting that Fed reserve re­
quirements are binding for most member banks, in the sense
that they hold larger cash balances than they would in the
absence of Fed reserve requirements, whereas state reserve
requirements are not binding for most nonmember banks. In
most states, required cash reserves of nonmember banks are
substantially smaller than the cash reserves nonmember banks
actually hold. See R. Alton Gilbert, “Effectiveness of State Re­
serve Requirements,” this R eview (September 1 9 7 8 ), pp. 1628. If most member banks do not borrow at the discount win­
dow to offset reserve drains in order to meet binding reserve
requirements, nonmember banks would be even less likely to
borrow at the discount window, if made available to them,
unless they were made subject to member bank reserve
requirements.

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

FEBRUARY

1980

Table 1
Frequency of Borrowing from the Federal Reserve by Member Banks
Bank size
(Total deposits in
millions of dollars)

Number of Eighth District member banks that borrowed
on the following number of occasions each year
2-4
5-10
11-15
16-20
Over 20

1

Tot®
1
ofT a n k s
borrowing

1974
$ 0 - $ 10

7

8

7

8

0

0

30

10-

25

14

21

8

0

1

0

44

25-

18

50

3

2

11

2

0

0

5 0 - 100

2

1

4

1

0

1

9

Over 100

1

1

4

2

3

3

14

1975
$ 0 - $ 10

2

2

1

0

1

0

6

10-

25

2

8

1

1

0

0

12

25-

50

6

4

1

1

0

0

12

50- 100

3

2

0

0

1

0

6

Over 100

4

2

0

2

0

0

8

$ 0 - $ 10

2

2

1

0

0

6

1976
1

10-

25

4

0

1

1

0

0

6

25-

50

2

3

5

0

0

0

10

5 0 - 100

1

1

0

0

1

0

3

Over 100

4

4

0

0

0

0

8

$ 0 - $ 10

1

5

1977
1

0

0

1

8

10-

25

3

4

4

0

1

0

12

25-

17

50

5

6

3

3

0

0

5 0 - 100

2

4

2

0

0

0

8

Over 100

2

5

7

3

1

0

18

and their response to deviations of actual from de­
sired cash balances. Cash balances of member banks
can be classified into three components: vault cash,
reserve balances at Federal Reserve Banks, and
demand balances due from correspondents.
The amount of vault cash banks desire to hold is
based upon their expectations of depositors’ demand
for currency. Reserve balances held by most member
banks at the Federal Reserve are determined by
their required reserve balances, which are based
upon their deposit liabilities and vault cash held two
weeks earlier. Required reserve balances change each
week, and member banks must meet their required
reserves on a weekly average basis. Finally, demand
balances of member banks due from correspondents



are determined by both the volume of transactions
through those accounts and the balances their corres­
pondents require as compensation for services pro­
vided. Correspondent banks generally do not charge
respondents explicit fees for services; they require,
instead, that respondents hold certain average de­
mand balances with them.8
The primary cause of deviation between a bank’s
actual and desired cash balances is unanticipated fluc­
tuation in demand deposit liabilities. Time depos­
its mature on specific dates, and therefore bank man­
agement can anticipate when it must be ready to
8Robert E . Knight, “Correspondent Banking Part III: Account
Analysis,” Federal Reserve Bank of Kansas City Monthly
Review (December 1 9 7 1 ), pp. 3-17.

17

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

make payments to holders of time deposits. Demand
deposit liabilities, however, fluctuate more from day
to day and are, therefore, more difficult to estimate.
The effects of unanticipated loan demand on
banks’ reserve positions are reflected in changes in
their demand deposit liabilities. When banks make
loans, they increase the demand deposit balances of
borrowers. If borrowers did not withdraw those de­
posits to make payments, there would be no dis­
turbance in the current week to the reserve positions
of banks making loans. Increases in loans cause banks
to lose reserves only when borrowers withdraw
deposits.
Bank customers withdraw deposits from their de­
mand accounts either by demanding currency or by
writing checks. When depositors demand currency, a
bank’s assets (vault cash) and liabilities (demand
deposits) decrease simultaneously. Member banks
lose reserves when depositors demand currency, but
due to lagged reserve accounting, this decline in
vault cash does not affect reserves available to meet
reserve requirements in the current week. Likewise,
the decline in demand deposit liabilities that results
when depositors demand currency does not affect
required reserves in the current settlement week, but
two weeks later instead.9 Therefore, member banks
may choose to make no initial response to a loss of
reserves in the form of vault cash, unless total vault
cash falls below some critical level at which the risks
of temporary currency shortages become sufficiently
great.
When customers withdraw demand deposits by
writing checks, the effect on a member bank’s cash
assets depends upon how the bank clears checks and
how it pays for checks drawn on its customers’ ac­
counts. Banks receive checks as their customers make
deposits and, therefore, need a mechanism for collect­
ing checks drawn on other banks. Some member
banks receive payment for checks by having their
reserve balance at their Reserve Bank credited, while
paying for checks drawn on their depositors’ accounts
by having their reserve balances debited. Frequent
debits and credits to member banks’ reserve accounts
cause these balances to change on most business days.
9Required reserves of a member bank in the current settlement
week are based upon its deposit liabilities two weeks earlier.
The reserves that count toward meeting a member bank’s
required reserves in the current settlement week are average
daily vault cash held two weeks earlier and average daily
reserve balances at the Fed in the current week. Fo r a com­
parison of how contemporaneous and lagged reserve account­
ing affects the reserve management of member banks, see R.
Alton Gilbert, “The Effects of Lagged Reserve Requirements
on the Reserve Adjustment Pressure on Banks,” Financial
Analysts Journal ( September-October 1 9 7 3 ), pp. 34-43.

Digitized for 18
FRASER


FEBRUARY

1980

If a member bank clears checks through its reserve
account, net withdrawals of deposit liabilities cause
its reserve balance to decline. Required reserves for
the current week, however, are not affected by de­
posit withdrawals in the current week due to lagged
reserve requirements. Thus, if its initial reserve bal­
ance just equaled its required balance for the cur­
rent week, deposit withdrawals during the week
would cause a member bank’s reserve balance to be
deficient. Borrowing through the discount window is
one means of increasing reserves on short notice.
Other member banks clear checks through their bal­
ances at correspondents. They deposit checks with
their correspondents and pay for checks drawn on
depositors’ accounts by having their demand bal­
ances due from correspondents debited, including
checks presented for payment by the Federal Reserve.
If these banks do not use other Reserve Bank services
that cause their reserve balances to change, these
balances would change only when banks adjusted
them to equal their required reserve balances.10 Thus,
the reserve balances of member banks that use serv­
ices of correspondents tend to change less frequently
than those of member banks that make greater use
of Reserve Bank services.
For a member bank that clears checks through a
correspondent bank, net withdrawals of deposit liabil­
ities would not disrupt its balance between actual and
required reserves for the current week. The decline in
demand deposit liabilities instead would affect the
bank’s required reserves in coming weeks. Since
checks are cleared through the bank’s correspondent
account, its reserve balance is unaffected by net
deposit withdrawals.
The response of these member banks to deposit
withdrawals depends upon the constraints placed on
them by their correspondents. If respAident banks
were required to hold certain minimum amounts of
demand balances at correspondents on a weekly av­
erage basis, they occasionally would have to obtain
additional cash balances when experiencing net with­
drawals of deposit liabilities. Studies of correspondent
banking, however, indicate that respondent banks may
average their balances over longer periods of time,
such as a quarter or even a year, in meeting the bal­
ance requirements of correspondents.11 Respondent
10Use of the following Reserve Bank services may involve
debits and credits to reserve balances: wire transfers, coin
and currency shipments, and collection of coupons on secur­
ities held in safekeeping and matured securities.
n Knight, “Correspondent Banking.”

FEBRUARY

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

banks may have enough flexibility in managing their
cash positions to simply let their demand balances due
from correspondents fluctuate as their deposit liabil­
ities fluctuate with no additional response to deposit
outflows.
One method of determining whether respondent
banks may temporarily reduce their demand balances
due from correspondents by amounts equal to short­
term deposit outflows is to compare the dollar magni­
tude of changes in demand deposit liabilities to
changes in their demand balances due from corre­
spondents. When large decreases in demand balances
due from correspondents were compared to large de­
creases in gross demand deposit liabilities for 95 mem­
ber banks that obtain most of their services through
correspondents, the ratio of the former to the latter
averaged 1.04.12 Thus, large weekly decreases in de­
mand balances due from correspondents are of ap­
proximately the same dollar magnitude as large
weekly decreases in gross demand deposits. These re­
sults indicate that respondents have enough short­
term flexibility in managing their cash positions that
they can cope with relatively large decreases in their
gross demand deposit liabilities by letting their de­
mand balances due from correspondents decline
temporarily.

Are Demand Deposit Withdrawals Large
Enough to Induce Banks to Borrow?
Only those member banks that clear checks through
their reserve accounts have reductions in their re­
12The 9 5 member banks, located in the Eighth District portions
of Illinois and Missouri, had total deposits of less than $50
million in 1976. None of these banks cleared checks through
the Reserve Bank regularly. These banks had so little ac­
tivity in their reserve balances at the Fed that the dollar
amounts in their reserve balances remained unchanged for
150 days or more during 1976. Relatively large declines in
the demand deposit liabilities at individual banks are meas­
ured by calculating the changes in gross demand deposits
from each reserve settlement week to the next and deter­
mining the fifth largest decline. That amount is compared
to the fifth largest weekly decline in demand balances due
from correspondents, a measure of relatively large weekly
declines in balances due from correspondents. An alternative
comparison would be of the greatest weekly decline in de­
mand deposit liabilities to the greatest weekly decline in de­
mand balances due from correspondents. Such comparisons
were not reported because, for many banks, the greatest
weekly declines were substantially different from declines in
other weeks and, therefore, not representative of the rela­
tively large weekly declines in these series. Examination of
weekly declines in both series indicated that the fifth greatest
weekly declines tend to be fairly representative of large de­
clines. Fo r instance, the third greatest weekly decline in
demand balances due from correspondents divided by the
third greatest decline in gross demand deposits averages
1.06.




1980

serve balances when experiencing net demand de­
posit outflows. These are the banks, therefore, that
tend to borrow frequently from the discount window.
Even for those banks that use their reserve balances
for clearing checks, however, fluctuations in demand
deposit liabilities may not be large enough to induce
them to borrow to avoid reserve deficiencies. If fluc­
tuations in their reserve balances are smaller than
their normal excess reserves, no response to declines
in demand deposit liabilities is necessary.
This issue is investigated by comparing average ex­
cess reserves to large decreases in demand deposit
liabilities for a group of 102 member banks that make
extensive use of Reserve Bank services.13 The fifth
largest weekly decrease in gross demand deposits
was larger than their average excess reserves in 1976
for all but one of these banks. Those relatively large
decreases in deposits were, on the average, about
60 times larger than average excess reserves.14 There­
fore, excess reserves could not fluctuate by as much
as demand deposits during at least several weeks each
year. When demand deposit liabilities decline by
relatively large amounts, member banks that clear
checks through their reserve accounts must obtain
additional reserves to avoid reserve deficiencies.

Do Frequent Borrowers Clear Checks
Through Their Reserve Accounts?
The next step is to determine whether member
banks that borrow from the discount window most
frequently are, in fact, primarily among those banks
that clear checks through their reserve accounts. A
reliable indicator of whether a member bank uses its
reserve balance for settlement in check collection is
the frequency of changes in its reserve balance from
day to day. Check collection is the only Reserve Bank

l:iThe reserve balances of these member banks were unchanged
on no more than five business days during 1976.
1*The ratio of the fifth largest weekly decline in gross demand
deposits to average excess reserves was higher for larger
banks, as indicated in the following display:
Size group
(Total deposits,
in millions
of dollars)

$

Number
of
banks

Sum of the fifth largest
weekly decrease in gross
demand deposits divided by the
sum of average excess reserves

10

7

4.46

25

31

10.02

25-

50

33

84.88

50-

100

18

83.16

13

92.50

0 -$
10-

100 and over

19

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

FEBRUARY

1980

Table 2
Comparison of the Frequency of Changes in Reserve Balances
of Frequent Borrowers and Other Member Banks1

Size cateaorv
(Annual average

Distribution of banks by the measure of frequency
of change in reserve balances at the Fed (Number
°* cla^s ^ur‘n9 a year when a member bank’s reserve
balances at the Fed were the same as on the previous day)
Number
121- 131- 151- 176- 201251- 301
i ie of banks
120
130
ISO
200 250
300 & Over
175

Up to
115

Infrequent borrowers

3

2

4

3

5

3

1

2

1975

2

1976

$0-$10

Type of bank

1974

total deposits,
millions of dollars)

1

5

18

8

1

3

1

7

2

53

Frequent borrowers in:

1977
$10-$25

Infrequent borrowers

2
2

1

2

1

1
17

15

9

8

10

1

7

1

5

14

10

2

82

Frequent borrowers in:
9

1974

2

2

1976

2

2

1977
$25 - $50

7

1975

5

5

Infrequent borrowers

15

5

5

4

5

3

1

38

Frequent borrowers in:
1974

2

1976

4

1977
$5-$100

12

1975

6

Infrequent borrowers

11

1

13
2

1

5
6

2

1

14

Frequent borrowers in:
1974

6

1

1

1976

1

1

1977

2

2

0

0

1974

12

12

1975

2

2

1976

0

0

1977

Over $100

6

1975

11

11

Infrequent borrowers
Frequent borrowers in:

1Banks designated as frequent borrowers are all Eighth District member banks that borrowed at the discount window on
five or more occasions during a year. Those designated as infrequent borrowers are member banks in the Eighth District
portions of Illinois and Missouri that did not borrow at the discount window in 1976 and were not frequent borrowers
during 1974, 1975, or 1977.


20


F E D E R A L R E S E R V E B A N K O F S T . L O U IS

FEBRUARY

1980

service that is likely to cause a member bank’s re­
serve balance to change each business day.15

row from the discount window frequently and others
never do?

The distribution of nonborrowers and frequent bor­
rowers by activity in their reserve balances is pre­
sented in table 2. Almost all of the Eighth District
member banks that borrowed on five or more occa­
sions per year in 1975-77 had changes in their reserve
balances on each business day. Incentive to borrow
from the discount window was relatively great during
1974 due to the large spread between market interest
rates and the discount rate. Several banks that bor­
rowed frequently in 1974 had relatively idle reserve
balances. More of the frequent borrowers during 1974,
however, had relatively active reserve balances than
nonborrowers of comparable size.18 Thus, evidence in
table 2 supports the hypothesis that banks which bor­
row frequently are primarily among those that have
relatively active reserve balances.

One possible answer is that, among member banks
with relatively active reserve balances, the frequent
borrowers have more highly variable deposit liabil­
ities, and thus are induced to borrow from the dis­
count window more frequently to replace reserves
lost due to deposit withdrawals. Table 3 indicates that
frequent borrowers in the years 1975-77 tend to have
more highly variable demand deposit liabilities than
other banks of comparable size with similar activity
in their reserve balances.17

Why Are Only Some of the Member Banks
with Active Reserve Balances Frequent
Borrowers?
Fluctuations in deposit liabilities at member banks
that use correspondent services tend not to induce
frequent, short-term borrowings from the discount
window since those banks can adjust to decreases in
deposit liabilities by letting their balances at corres­
pondents decline temporarily. Thus, banks that use
services of correspondents may not be induced to
borrow from any source in response to relatively
large withdrawals of deposit liabilities. But among
member banks that clear checks through their re­
serve accounts, and consequently have frequent
changes in their reserve balances, why do some bor15The relation between use of Fed services and frequency of
changes in reserve balances can be illustrated for member
banks in the Eighth District portions of Illinois and Missouri,
based on a survey of utilization of Fed services during
1976-77. Frequency of change in reserve balances is meas­
ured over 126 calendar days from mid-September 1976 to
mid-January 1977. The minimum number of days that a
member bank’s reserve balance would be unchanged during
that period is 39. For the 68 banks that deposited more
than five checks to their reserve balances during January
1977, reserve balances were unchanged an average of 39.9
days. A contrasting group is composed of 69 banks that
cleared no checks through the Fed and used the reserve
balances of their correspondents in remitting for the Fed’s
cash letters and handling transactions with the Fed for coin
and currency. Their reserve balances remained unchanged
for an average of 75.6 days.
16Among member banks with total deposits up to $50 million,
about 65 percent of the banks that borrowed frequently in
1974 had reserve balances that remained unchanged for
115 days or less, whereas only 20 percent of nonborrowers
had such active reserve balances.




Another factor that might account for the difference
in frequency of borrowing is the difference in ratios
of loan to deposits. Banks with higher ratios of loan
to deposits tend to have smaller amounts of liquid as­
sets, which they can sell quickly to replace reserve
drains. Frequent borrowers had significantly higher
loan-to-deposit ratios in 1976; the difference was
positive, but not statistically significant at the 5 per­
cent level in 1977. Thus, there is some evidence that,
among the relatively small member banks with active
reserve balances, those that borrow frequently from
the discount window have higher ratios of loans to
deposits.18
Differences in federal funds positions might also
explain the varied use of the discount window by
banks with active reserve balances. Reserve balances
can be increased either by borrowing from the dis­
count window or by reducing federal funds sold to
correspondents. Loans in the federal funds market
17The characteristics of member banks that borrowed from
the discount window on five or more occasions in 1974, but
were not frequent borrowers in the following three years,
were also compared to those of member banks with rela­
tively active reserve balances that did not borrow fre­
quently in any of the years 1974-1977. The banks that
borrowed frequently only during 1974 did not have greater
variability in their demand deposit liabilities than the other
banks. They did have higher ratios of loans to deposits than
nonborrowers. However, the banks that borrowed fre­
quently only during 1974 had such large proportions of their
assets invested in federal funds sold on average, and positive
net federal funds sold so often, that they were not induced
to borrow frequently from the discount window during the
following years. Analysis in table 3 is limited to banks with
total deposits up to $50 million because there were only
two frequent borrowers in 1975-77 with total deposits be­
tween $50 million and $100 million (average for 1 9 7 6 );
including those two banks in the analysis, and making com­
parisons to nonborrowers with total deposits up to $100
million might bias some of the comparisons in table 3.
18A study of Tenth District member banks found that banks
which borrowed from the discount window had higher ratios
of loans to assets than nonborrowers of comparable size.
See J. A. Cacy, “Determinants of Member Bank Borrow­
ing,” Federal Reserve Bank of Kansas City Monthly Review
(February 1 9 7 1 ), pp. 11-20.

21

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

FEBRUARY

1980

Table 3
Comparison of Frequent Borrowers and
Nonborrowers with Relatively Active Reserve Balances (1975-77)1
Measure

Frequent
borrowers

Nonborrowers2

Difference

T-Statistic

Variability of demand deposits3

6.60%

4.82%

1.78%

2.0260

Percentage of average loans to average
total deposits
1976
1977

61.81%
63.71

54.18%
58.52

7.63%
5.19

2.2991
1.5895

Number of weeks in which net federal funds
sold was zero or negative
1976
1977

20.44
20.44

7.98
10.64

Average percentage of net federal funds
sold to total deposits4
1976
1977
Number of banks

2.20%
2.32
18

5.87%
5.41

12.46
9.80

-3.67%
-3.09

3.5466
2.7838

-2.2940
-2.1797

47

1Analysis is limited to banks with average total deposits in 1976 of less than $50 million dollars.
2The reserve balances of these banks were unchanged on no more than 120 days during 1976.
3Variability of demand deposits is measured in the following w ay: A bank’s demand deposit liabilities as of each business day
are compared with average demand deposit liabilities in the five previous business days, and the percentage deviation of de­
posits of each day from the five-day moving average is calculated. The standard deviation of those percentage differences
(positive and negative signs retained) is calculated for each year as a measure of the day-to-day variability of a bank’s
demand deposit liabilities. For interpretation of this measure, suppose a bank has a standard deviation of 5 percent. For
that bank the percentage deviations of demand deposit balances from the five-day moving average would lie within a range
5 percentage points below and above the mean on about two-thirds of the days during the year. Measuring deposit varia­
bility with this standard deviation weights the measure toward the more extreme percentage deviations from the five-day
moving averages, which is appropriate for a measure that might be correlated with frequency of borrowing by banks to
avoid reserve deficiencies.
4Net federal funds sold are counted as zero for weeks in which banks purchased more federal funds than they sold.

generally have maturities of one day, and, therefore,
banks that sell federal funds can increase their
reserve balances quickly by reducing the amount of
federal funds they sell. Federal Reserve regulations
do not permit member banks to receive reserve
adjustment credit through the discount window while
selling federal funds. Infrequent borrowers with
active reserve balances may sell federal funds so often
that they are seldom permitted to borrow from the
discount window.
Two aspects of federal funds positions are analyzed
in table 3. One is the number of weeks during each
year in which a bank’s net federal funds sold was
zero or negative. A bank with positive net federal
funds sold during most weeks tends to have few
occasions on which it would be induced to borrow.
The number of weeks during which net federal funds

http://fraser.stlouisfed.org/
22
Federal Reserve Bank of St. Louis

sold were zero or negative is significantly greater for
frequent borrowers in both 1976 and 1977.
The other measure of federal funds position is aver­
age federal funds sold as a percentage of average
total deposits, a measure of the cushion of liquidity
a bank maintains in the form of federal funds sold.
This percentage is significantly lower for frequent
borrowers in both 1976 and 1977.
Analysis of the characteristics of banks that bor­
row frequently from the discount window indicates
that use of the discount window for reserve adjust­
ment is a valuable service for frequent borrowers.
Banks that borrow frequently are among those
that make extensive use of Reserve Bank services.
They tend to have greater short-term variability in
their demand deposit liabilities and have higher

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

ratios of loans to deposits. In addition, net federal
funds sold, which are smaller on average and zero
more often than for other banks, indicate that they
substitute use of the discount window for holding
liquid assets.
Most member banks, however, prefer to adjust their
reserve positions to deposit fluctuations by methods
other than borrowing at the discount window. Among
relatively small member banks (those with total de­
posits less than $100 million), most clear checks
through correspondents. These banks can cope with
deposit fluctuations by allowing their demand bal­
ances due from correspondents to fluctuate. Of the
relatively small member banks that clear checks
through their reserve balances, most prefer to adjust
their reserve positions to deposit outflows by draw­
ing system that resulted from, or were compounded
Only a small minority of member banks borrow
from the discount window as a method of reserve
adjustment.

FEBRUARY

Table 4
Variability of Demand Balances Due
to Commercial Banks and Private
Nonbank Depositors1
Annual average
of demand balances
due to respondents
(millions of dollars)

Measure of variability of demand
balances due to2
Nonbank depositors
Commercial
other than the
banks
U.S. government

(10 largest correspondent banks)
$ 162

5.5%

5.4%

141

7.3

5.7

141

7.8

6.7

91

6.1

3.7

60

14.1

3.0

59

10.6

3.8

One of the reasons for creating the Federal Reserve
System was concern over liquidity crises of the bank­
ing system that resulted from, or were compounded
by, simultaneous demands for cash by respondent
banks from their correspondents. This concern might
still be relevant, given the continuing reliance of most
banks on correspondents for liquid balances.
Percentage changes from week to week in demand
balances of individual banks due from correspondents
are often quite large. Does it follow that demand
balances of correspondent banks due to respondents
are also highly variable from week to week? Are
correspondent banks vulnerable to liquidity problems
as a result of large fluctuations in their demand bal­
ances due to respondent banks?
The variability of total demand balances due to
respondents depends upon the size of the bank’s total
balances due to respondents (see table 4). For the
three largest correspondent banks, total demand
balances due to respondents are about as variable
as total demand deposit liabilities due to private non­
bank depositors. For smaller correspondent banks
those balances are more variable than their other
demand deposit liabilities. This contrast is greatest
for the 21 smallest correspondent banks: The meas­
ures of variability in their demand deposit liabilities
due to private nonbank depositors were about the
same as for the larger correspondent banks, but bal


53

8.0

4.0

35

STABILITY OF THE CORRESPONDENT
BANKING SYSTEM

1980

9.8

7.1

27

11.7

7.5

8

12.0

3.5

(Averages for 21 smaller correspondent banks)
$ 1.06

30.3%

5.1%

1These banks were selected from among all Eighth District
member banks that borrowed from the discount window
during 1976 and other Eighth District member banks
in Illinois and Missouri. A bank is assumed to be offering
correspondent services to other banks if its demand bal­
ances due to respondents are positive each day and change
each business day.
V ariability of these balances is measured by calculating
average balances for each reserve settlement week in 1976,
percentage changes in balances from each week to the
next, and then taking the standard deviation of the per­
centage changes.

ances due to respondents were substantially more
variable. Their demand balances due to respondents
were about as variable from week to week as de­
mand balances due from correspondents of individual
banks that clear checks through correspondents. For
95 member banks that obtain most of their services
through correspondents, the average measure of
weekly variability in their demand balances due from
correspondents is 29.5, which is approximately equal
to the average measure of variability in demand
balances due to respondents of 30.3 for the 21
smallest correspondent banks.19
19VariabiHty of demand balances of correspondents due to re­
spondents is measured as the standard deviation of percent­
age changes in those balances from week to week. Variability
of demand balances due from correspondents is measured
the same way. The 95 member banks which use correspond­
ent services are described in footnote 12.

23

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

These comparisons indicate that the relatively large
correspondent banks have an advantage over smaller
correspondents in coping with the variability in de­
mand balances due to individual respondent banks.
For correspondents that serve the largest number of
respondent banks, fluctuations in balances due to indi­
vidual respondent banks tend to cancel each other
out, causing their total demand balances due to other
banks to be no more variable from week to week
than the sum of other demand deposit liabilities. Cor­
respondents that serve fewer respondents do not
benefit as much from such cancellation. Thus, the
largest correspondent banks are able to offer respond­
ents the service of reserve adjustment mechanisms
with no more vulnerability to occasional liquidity
problems resulting from fluctuations in their balances
due to respondents than that from fluctuations in bal­
ances due to nonbank depositors.

SUMMARY AND CONCLUSIONS
It recently has been suggested that one adverse
effect of declining Federal Reserve membership is
that the Federal Reserve may become more reluctant
to pursue restrictive monetary policy because of its
uneven impact on the banking system. Nonmember
banks, without access to the discount window, might
not have the means to cope with liquidity pressures
resulting from restrictive monetary policy. Therefore,
extending access to the discount window to more
banks, by inducing or requiring more banks to be
members, would facilitate the implementation of
monetary policy.
The validity of this argument is investigated by
examining the frequency with which member banks
borrow at the discount window. If access to the dis­
count window helps banks adjust to occasional liquid­
ity pressures, most member banks would borrow


24


FEBRUARY

1980

frequently from the discount window as a means of
coping with such pressures. Only a small minority of
member banks, however, actually borrows frequently
from the Federal Reserve. Therefore, while there are
valid reasons for suggesting that reductions in mem­
bership may interfere with monetary control, analysis
of reserve management by banks does not indicate
that direct access to the discount window for non­
member banks is important for the conduct of
monetary policy.
Many member banks adjust to declines in
their cash balances by letting their demand balances
due from correspondents decline temporarily. This
reaction is automatic for banks that clear checks
through correspondent balances. For banks that clear
checks through accounts at their Reserve Bank, net
withdrawals of deposit liabilities cause reductions in
their reserve balances, and may occasionally cause
reserve deficiencies unless offset. Banks that borrow
frequently at the discount window are among these
banks that clear checks through their reserve ac­
counts. Most member banks that clear checks through
reserve accounts, however, do not borrow frequently
from the discount window, but instead rely upon
federal funds sold to correspondents as a cushion of
liquidity, reducing federal funds sold when expe­
riencing reserve outflows.
The relatively large correspondent banks benefit
from offsetting fluctuations in the deposits due to in­
dividual respondent banks. Thus, large correspondent
banks can offer respondents means of adjusting their
cash positions to deposit fluctuations without incur­
ring any more variability in their own liabilities than
results from fluctuations in demand balances due to
nonbank customers. This indicates that correspondent
banks offer member and nonmember banks adequate
means of adjusting their cash position to fluctuations
in deposit liabilities.

The New Monetary Aggregates
R. W. HAFER

A
Jnumber of major financial innovations over the
last decade have changed the composition of assets
used by the public to make payments. Examples in­
clude the introduction of negotiable orders of with­
drawal (NOW accounts), the implementation of auto­
matic transfer systems ( ATS accounts) whereby funds
from savings accounts can be automatically trans­
ferred to checkable deposits, the growing use of
money market mutual funds as substitutes for conven­
tional savings accounts, and the dramatic growth in
repurchase agreements (RPs).1 Because of these and
other developments, it increasingly was argued that
the existing monetary aggregates did not measure the
true financial position of the public and, therefore,
were inadequate tools of monetary policy.
In response to these developments, the Federal Re­
serve Board recently announced redefinitions of the
monetary aggregates.2 This article describes the new
aggregates, compares them to the old measures, and
discusses some technical issues involved in their
measurement.
1For a discussion of NOW and ATS accounts and their effect
on the old monetary aggregates, see Steven M. Roberts, “De­
veloping Money Substitutes: Current Trends and Their
Implications for Redefining the Monetary Aggregates,” Im prov­
ing the Monetary Aggregates: Staff Papers, Board of Gover­
nors of the Federal Reserve System, Washington, D.C., 1978,
pp. 147-70. Hereafter, this publication will be referred to as
Staff Papers. See also John A. Tatom and Richard W . Lang,
“Automatic Transfers and the Money Supply Process,” this
R eview (February 1 9 7 9 ), pp. 2-10. An introduction to re­
purchase agreements is found in Norman N. Bowsher, “Re­
purchase Agreements,” this R eview (September 1 9 7 9 ), pp.
17-22.
2“Announcement,” Board of Governors of the Federal Reserve
System, Washington, D.C., February 7, 1980.




Definitions
Tables 1 and 2 compare the old and new monetary
aggregates. As shown in table 1, the new basic trans­
actions measure — MIA — is essentially the same as
the old M l measure, except for the deletion of de­
mand deposits due to foreign commercial banks and

Table 1
Comparison of Old and New
Transactions-Type Monetary
Aggregates
Component

Old
M1

New
M1A

New
M1B

Currency in circulation

X

X

X

At commercial banks:
Demand deposits inclusive
of deposits due to foreign
commercial banks and
official institutions
Demand deposits exclusive of
deposits due to foreign
commercial banks and official
institutions

X

X

X

NOW accounts

X

ATS accounts

X

At thrift institutions:
Demand deposits
NOW accounts

X

ATS accounts

X

Credit union share draft balances

X

X

25

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

official institutions. This change, based on a recom­
mendation of the Advisory Committee on Monetary
Statistics, was made because such balances consist
primarily of compensating balances held by foreign
commercial banlcs at U.S. commercial banks for serv­
ices performed.3 Since the total of such balances is a
small percentage of the old Ml, this deletion does not
produce large discrepancies between the old Ml and
new MIA measures.4
Until the early 1970s, a clear distinction between
interest-bearing deposits and non-interest-earning de­
posits held for transactions purposes existed. Since
then, however, a series of financial innovations and
regulatory changes have blurred this distinction.5 The
significant changes have taken the form of NOW ac­
counts, which were established in several New Eng­
land states in the early 1970s, ATS accounts, and the
rapid growth of credit union share drafts as an alter­
native payments mechanism. The new M1B aggregate
combines those financial items that have the dual
characteristic of being held both for check-writing
purposes and as savings accounts. Nearly all of these
items are interest-bearing checkable deposits.8 The
new MIA and M1B aggregates focus on those mone­
tary components that are employed primarily as a
means of payment.
As shown in table 2, the old M2 aggregate was de­
fined as the sum of currency, demand deposits, sav­
ings deposits, time deposits, and time certificates of
deposit (CDs) other than those issued in denomina­
tions of $100,000 or more by large weekly reporting
banks. The new M2 measure is much broader in
scope. It is calculated by adding savings deposits and
small time deposits (those issued in denominations
of less than $100,000) at all depositary institutions,
overnight RPs issued by commercial banks, overnight
Eurodollars (issued by Caribbean branches of mem­
3See Improving the Monetary Aggregates: Report o f the Ad­
visory Com m ittee on Monetary Statistics, Board of Governors
of the Federal Reserve System, Washington, D.C., 1976, pp.
15-20. Hereafter, this publication will be referred to as

Report.
4Fo r example, during 1978 the amount of demand deposits
due to foreign-related banking offices averaged less than 2
percent of total demand deposits.
5For an excellent description of this, see Roberts, “Developing
Money Substitutes.”
6This distinction is necessary because not all of the new check­
able deposits are interest-bearing at present. For example,
some depositary institutions currently offer non-interest-bear­
ing NOW accounts (N IN O W s), and demand deposits at mu­
tual savings banks do not currently pay interest. Also, some
non-interest-eaming demand deposits that are held at thrift
institutions and cannot be separated from interest-bearing
checkable deposits are included in M1B. At present, the
amount of such deposits is small.


26


FEBRUARY

1980

ber banks) held by U.S. nonbank residents, money
market mutual fund shares, and a consolidation com­
ponent to the new M1B measure. This definition in­
cludes such a broad array of monetary components
and institutions (e.g., commercial banks, U.S. agencies
and branches of foreign banks, Edge Act corporations,
foreign investment companies, mutual savings banks,
savings and loan associations, and credit unions), that
it is more directly comparable to old M3 than to the
previous M2 definition (see table 2).
There is, however, one major difference between
the new M2 measure and the old M3 definition: The
new M2 includes overnight RPs and Eurodollars, and
money market mutual funds shares which were not
included in the old M3 measure.
The introduction of these items stems from their
increasing substitutability for other non-transactionstype financial holdings already included in the broader
monetary measures. For instance, money market mu­
tual funds shares are viewed as substitutes for other
non-transactions-tvpe financial assets, despite the fact
that owners of these shares are offered check-writing
privileges. The fairly large minimum denomination re­
quirement (usually $500 or more) for checks written
on these accounts and the fact that these balances
typically exhibit relatively slow turnover rates suggest
that these accounts are used primarily as savings
rather than transactions accounts.7
The transactions and investment characteristics of
overnight RPs have been subjected to considerable
investigation in recent years. For example, studies by
Garcia and Pak, Wenninger and Sivesind, and Tinsley,
Garrett, and Friar have viewed these RPs as close sub­
stitutes for existing demand deposits. Consequently,
they have explained a large part of the decline in
the public’s demand for transactions balances which
occurred in the mid-1970s by including RPs in the
definition of a transactions-type money (i.e., old M l).
In contrast, others have regarded RPs as short-term,
highly liquid investment items that are significantly
different from demand deposits.8
"Thomas D. Simpson, “The Redefined Monetary Aggregates,”
F ederal Reserve Bulletin (February 1 9 8 0 ), p. 100.
8See Gillian Garcia and Simon Pak, “Some Clues in the Case
of the Missing Money,” American Econom ic Review ; Papers
and Proceedings (M ay 1 9 7 9 ), pp. 330-34; John Wenninger
and Charles Sivesind, “Defining Money for a Changing F i­
nancial System,” Federal Reserve Bank of New York Quar­
terly R eview ( Spring 1979) pp. 1-8; Peter A. Tinsley, Bonnie
Garrett, and Monica Friar, “The Measurement of Money
Demand,” Staff Study # 1 3 3 (Board of Governors of the Fed­
eral Reserve System, 1 9 7 8 ); and Thomas D. Simpson, “The
Market for Federal Funds and Repurchase Agreements,” Staff
Study # 1 6 6 (Board of Governors of the Federal Reserve
System, 19 7 9 ).

FEBRUARY

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

1980

Table 2
Comparison of Old and New Non-Transactions-Type Monetary Aggregates
Component
Currency

Old M2

New M2

Old M3

New M3

Old M4

Old M5

L

X

X

X

X

X

X

X

X

X

At commercial banks:
Demand deposits inclusive of deposits
due to foreign commercial banks and
official institutions

X

X

Demand deposits exclusive of deposits
due to foreign commercial banks and
official institutions

X

X

X

NOW accounts1

X

X

X

ATS accounts1

X

X

X

X

X

X

Overnight RPs
Savings deposits

X

X

X

X

X

X

X

Small time deposits (< $100,000)

X

X

X

X

X

X

X

Large time deposits
Other than large negotiable CDs
Including large negotiable CDs

X

X

X
X

X
X

X
X

X
X

X

Term RPs

X

At thrift institutions:
Demand deposits

X

X

X
X

NOW accounts1

X

X

ATS accounts1

X

X

Credit union share draft balances

X

X

X

X

X

Savings deposits (Mutual savings banks
and savings and loan associations)

X

X

X

X

X

Small time deposits (< $100,000)

X

X

X

X

X

X

X

X

X

Large time deposits (> $100,000)
Term RPs (Commercial banks and savings
and loan associations)

X

X

X

Other:
Overnight Eurodollar deposits of U.S.
nonbank residents2

X

X

X

Money market mutual funds shares

X

X

X

Term Eurodollars held by U.S. nonbank
residents

X

Bankers acceptances

X

Commercial paper

X

U.S. savings bonds

X

Liquid Treasury securities
M2 consolidation component3

X
X

X

X

1These accounts were included previously in the savings deposit component of the definitions.
2Overnight Eurodollars issued by Caribbean branches of member banks.
3See text, p. 30, for a discussion of this component.




27

FEBRUARY

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

There presently is no consensus as to whether over­
night RPs and Eurodollars, and money market mutual
funds shares primarily constitute transactions- or in­
vestment-type assets. These items are included in the
new M2 definition. Data on each of these series will
be published separately, however, so a direct com­
parison of these components with the new transactions
aggregates MIA and M1B will be possible.
The new M3 series is defined as new M2 plus large
time deposits (those issued in denominations of
$100,000 or more) at all depositary institutions and
term RPs issued by commercial banks and savings
and loan associations. The new M3 aggregate is simi­
lar to the old M5 definition primarily because of the
large-denomination time deposits component. The
combination of the large-denomination time deposits
and term RPs in this aggregate is based on the belief
that these items are relatively close substitutes in
many financial portfolios.9
The broadest of the new monetary aggregate defi­
nitions is the “L” series. This aggregate, which meas­
ures total liquid assets, adds to the new M3 series such
financial items as other Eurodollar holdings by non­
bank U.S. residents, bankers acceptances, commercial
paper, U.S. savings bonds, and liquid Treasury obliga­
tions.10 This measure closely approximates the credit
expansion generated through the commercial banking
sector and other financial channels.

1980

Table 3
Rates of Growth for Transactions-Type
Monetary Aggregates
Annual Rate of Grow th 1

Year

Old M1

New M1A

New M1B

1970

4.8%

4.8%

4.8%

1971

6.6

6.6

6.6

1972

8.4

8.5

8.5

1973

6.2

5.7

5.8

1974

5.1

4.7

4.7

1975

4.6

4.7

4.9

1976

5.8

5.5

6.0

1977

7.9

7.7

8.1

1978

7.2

7.4

8.2

1979

5.5

5.5

8.0

Q uarterly Rate of Grow th 2

Quarter

Old M1

New M1A

New M1B

A comparison of the growth rates of the old and new
monetary aggregates provides a useful way to assess
the differences resulting from the redefinitions. As
shown in table 3, there is relatively little quantitative
difference between the annual growth rates of the Ml
aggregate and the new MIA and M1B measures over
the 1970-1979 period. For example, the average differ­
ence in annual growth rates between Ml and MIA
over this period is only 0.18 percentage points, the
largest divergence occurring in 1973 when Ml grew
0.5 percent faster than MIA.

2.0%

2.6%

2.9%

5.8

5.9

5.9

111/1975

7.2

7.0

7.3

IV/1975

3.0

2.9

3.2

1/1976

4.6

5.4

5.7

11/1976

6.4

5.8

6.3

111/1976

4.1

3.4

3.9

IV/1976

COMPARISON OF GROWTH RATES

1/1975
11/1975

7.4

7.0

7.6

1/1977

7.4

8.8

9.3

11/1977

7.4

6.7

6.9

111/1977

8.6

6.0

6.5

IV/1977

7.4

8.4

8.7

1/1978

6.6

7.6

7.9

11/1978

9.2

8.7

9.1

111/1978

7.9

7.1

7.3

IV/1978

4.3

5.6

7.4

-1.3

0.2

4.8

1/1979
11/1979

8.1

7.8

10.7

111/1979

9.7

8.8

10.1

IV/1979

5.0

4.7

5.3

8Simpson, “The Redefined Monetary Aggregates,” p. 102.
10The Eurodollar holdings included in this measure incorpo­
rate those that are not captured in overnight Eurodollars
issued by Caribbean branches of member banks. Liquid
Treasury obligations consist of those issues with 18 months
or less remaining to maturity. See Simpson, “The Redefined
Monetary Aggregates,” p. 98.
It should also be noted that the new M2, M3, and L ag­
gregates exclude the amounts held by depositary institutions,
money market mutual funds, the federal government, the
Federal Reserve, and foreign commercial banks and official
institutions. See Simpson, “The Redefined Monetary Aggre­
gates,” pp. 98, 108.

Digitized for 28
FRASER


1Fourth-quarter-to-fourth-quarter growth rates.
-Annualized growth rates based on seasonally adjusted data.
SOURCE: “The Redefined Monetary Aggregates,” table A l.

Although the average difference between the M l
and M1B measures is somewhat larger (0.51 percent),
M1B has demonstrated a faster rate of growth over
recent years relative to Ml. This faster growth—1.0
percent in 1978 and 2.5 percent in 1979 — results from

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

FEBRUARY

Table 4
Rates of Growth for Old M2, Old M3,
and New M2 Monetary Aggregates1
Annual Rate of Growth

Year
1970

Old M2

Old M3

7.2%

7.2%

New M2
5.8%

1971

11.3

13.5

13.5

1972

11.2

13.3

12.9

1973

8.8

9.0

7.3

1974

7.7

7.1

6.0

1975

8.4

11.1

12.3

1976

10.9

12.7

13.7

1977

9.8

11.7

11.5

1978

8.7

9.5

8.4

1979

8.3

8.1

8.8

Quarterly Rate of Growth

Quarter

Old M2

Old M3

New M2

1/1975

6.4%

11/1975

9.5

12.4

14.9

111/1975

10.0

12.8

14.6

IV/1975

6.8

9.4

9.9

1/1976

10.5

12.0

13.0

11/1976

10.0

11.9

12.7

8.2%

7.8%

111/1976

8.9

11.0

11.3

IV/1976

12.6

13.8

15.2

1/1977

10.9

12.4

13.7

11/1977

9.0

10.5

11.2

111/1977

10.1

11.8

9.6

IV/1977

7.9

10.1

9.7
7.5

1/1978

7.0

8.1

11/1978

8.4

8.4

7.5

111/1978

9.8

10.3

8.2

IV/1978

8.5

9.8

9.5

1/1979

2.8

5.3

6.3

11/1979

8.8

7.9

10.2

111/1979

11.9

10.5

10.3

IV/1979

8.9

7.8

1980

Over the last five years, the average difference be­
tween the M l and MIA quarterly growth rates was
0.80 percent; the average difference between Ml and
M1B growth was 1.16 percent. The impact of NOW
and ATS accounts again is demonstrated — during
1979, M1B grew 2.5 percent faster than either M l or
MIA. As these figures suggest, the extension of NOW
accounts nationwide may temporarily produce wider
divergencies between the MIA (and old M l) and
M1B growth rates.
Annual and quarterly growth rates for the old M2,
old M3, and new M2 measures are presented in
table 4. As these figures show, growth rates of new
M2 tend to be closer to those of the old M3 definition
than to old M2. For instance, the average annual
growth rate of old M2 was 9.2 percent over the last
decade while the averages for old M3 and new M2
were 10.3 percent and 10.0 percent, respectively. An
examination of the quarterly data reveals a similar
relationship: From 1/1975 to IV/1979, old M2 grew
at an 8.9 percent average rate while the average
growth rates for old M3 and new M2 were 10.2 and
10.5 percent, respectively. In addition, the proportion
of new M2 that consists of money market certificates
and money market mutual funds has increased sharply
since 1978.1
1
Table 5 presents the annual and quarterly growth
rates for the old M4 and M5 aggregates together with
the new M3 and L definitions. As noted earlier, the
new M3 aggregate is relatively closer in construction
to the old M5 measure than to old M4. The difference
between the average annual rate of growth of old M4
and new M3 is 1.2 percentage points; that between
old M5 and new M3 is only 0.5 of a percentage point.
On a quarter-to-quarter basis, movements in new M3
and old M5 are even more similar. For example, over
the period I/1975-IV/1979, the average quarterly rate
of growth of old M4 was 8.1 percent while that of
old M5 and new M3 was 9.6 percent and 10.4 per­
cent, respectively.

7.2

JSee footnotes accompanying table 3.
SOURCE: “The Redefined Monetary Aggregates,” table A2.

the increased use of NOW and ATS accounts as de­
mand deposit and other balances are shifted into
these interest-earning checkable deposits.
Growth rates of the old and new M l measures ex­
hibit greater divergence on a quarter-to-quarter basis.



The growth rates of L — total liquid assets — have
been closer to new M3 than to the other monetary ag­
gregates. While the average quarterly growth rates
of L and new M3 have been roughly similar over the
past five years (11.1 percent and 10.4 percent, respec­
tively), there has been a growing divergence between
these measures in more recent years. This is explained
by the rapid growth of liquid assets issued by non­
n Simpson, “The Redefined Monetary Aggregates,” p. 105.

29

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

depositary institutions which form the distinction be­
tween new M3 and L.12

MEASURING THE NEW AGGREGATES:
TECHNICAL CONSIDERATIONS
Several technical problems arise in the measure­
ment of the new monetary aggregates: Certain de­
posits held by depositary institutions must be consoli­
dated to avoid double counting, the series used in
calculating the new aggregates must be seasonally ad­
justed, and the data needed to construct the new
aggregates must be gathered.
In calculating the old Ml aggregate, the problem
of double counting deposits was resolved by netting
out cash items in the process of collection, interbank
deposits, and Federal Reserve float from total com­
mercial bank demand deposits.1 A similar proce­
3
dure is followed in measuring the new MIA. At
the M1B and M2 levels, however, it is assumed
that thrift institutions hold demand deposits at com­
mercial banks to service their checkable deposits and
ordinary savings deposits. Thus, in calculating M1B,
the estimated proportion of demand deposits owned
by thrift institutions used to service their checkable
deposits will be removed;14 for new M2, total demand
deposits owned by thrift institutions are currently
netted out.
At the new M2 and M3 levels, further consolidation
measures are employed. For instance, in the calcula­
tion of new M2, savings and time deposits owned by
all depositary institutions are netted out, and money
market mutual funds’ holdings of RPs are deducted
from the public’s holdings of overnight RPs. In addi­
tion, CDs held by these funds are also netted out of
large time deposits in calculating new M3. Both of
these latter items are netted out in the derivation of
the total liquid assets aggregate (L ).
12Fo r example, the percentage increase in dollar amounts be­
tween January 1978 and December 1979 for these items,
using seasonally adjusted data, are:
Bankers accep tan ces............................ + 1 2 0 %
Commercial paper ............................... + 48
Short-term Treasury securities ...... + 39
U.S. savings bonds .............................. +
4
and, based on seasonally unadjusted data, + 1 3 6 % for term
Eurodollars.
13Fo r a discussion of this problem, see Report, pp. 12-14;
Darwin Beck, “Sources of Data and Methods of Construc­
tion of the Monetary Aggregates,” Staff Papers, pp. 117-33;
and Simpson, “The Redefined Monetary Aggregates,” pp.
107-10.
14At the present time, the amount of such holdings is negli­
gible and, therefore, is not omitted from M1B.

30



FEBRUARY

1980

Table 5
Rates of Growth for Old M4, Old M5,
New M3, and L Monetary Aggregates1
Annual Rate of Growth

Year

Old M4

1970

10.2%

1971

12.8

14.3

14.8

10.4

1972

12.3

13.9

14.0

12.9

1973

12.0

11.0

11.7

12.3

1974

10.7

9.0

8.7

9.6

1975

6.6

9.7

9.4

9.8

1976

7.1

10.2

11.4

11.0
12.6

Old M5

New M3

9.2%

8.9%

t
6.5%

1977

10.1

11.7

12.6

1978

10.6

10.6

11.3

12.3

1979

7.5

7.6

9.5

N.A.

Q uarterly Rate of Growth

Quarter

Old M4

Old M5

New M3

L

1/1975

7.6%

8.9%

7.2%

7.1%

11/1975

5.5

9.5

9.4

9.5

111/1975

6.2

10.1

10.7

10.5

IV/1975

6.2

8.8

9.1

10.7

1/1976

6.0

9.0

9.9

10.1

11/1976

6.0

9.4

11.3

11.1

111/1976

6.3

9.2

10.3

10.0

IV/1976

9.5

11.8

12.1

10.8

10.1

11.8

12.4

11.5

1/1977
11/1977

8.3

10.0

11.4

11.8

111/1977

10.0

11.7

11.7

12.2

IV/1977

10.4

11.5

12.5

12.8

1/1978

10.2

10.0

10.5

11.2

11/1978

10.6

9.8

11.1

12.4

111/1978

9.9

10.4

10.3

11.3

IV/1978

10.1

10.7

11.5

12.2

1/1979

5.4

6.8

7.9

10.4

11/1979

3.7

4.9

8.8

13.1

111/1979

9.2

8.9

10.3

11.7

IV/1979

11.0

9.1

9.8

N.A.

!See footnotes accompanying table 3.
SOURCE: “The Redefined Monetary Aggregates,” table A3.

Derivation of seasonally adjusted aggregates fol­
lows past procedures wherein the individual compo­
nents of the series are seasonally adjusted first, then
aggregated to the desired level.15 At this time, how15See Simpson, “The Redefined Monetary Aggregates,” pp. 11011. Fo r a general treatment of the seasonal adjustment

F E D E R A L R E S E R V E B A N K O F S T . L O U IS

FEBRUARY

1980

Table 6
New and Proposed Data Sources
Institution
Member banks

Component Collected

Coverage / Freq uency
125 large member banks/weekly
125 large member banks/weekly

NOW and ATS accounts

all member banks/weekly

Overnight Eurodollars at Caribbean branches
Nonmember banks

Term RPs
Overnight RPs

approximately all/weekly

Demand deposits

sample/weekly

NOW and ATS accounts

sample/weekly

Savings and small-denomination time deposits

Mutual savings banks

sample/weekly

Large-denomination time deposits

sample/weekly
sample/weekly (Wednesday)

Savings and small-denomination time deposits

sample/weekly (Wednesday)

Large-denomination time deposits
Savings and loan associations

NOW accounts and demand deposits

sample/weekly (Wednesday)
sample/thrice-monthly
sample/thrice-monthly

Large-denomination time deposits
Credit unions1

NOW accounts
Savings and small-denomination time deposits

sample/thrice-monthly

Share drafts

sample/weekly (Wednesday)

Savings and small-denomination time deposits

sample/weekly (Wednesday)

1The weekly sample (scheduled to begin in March 1980) will consist of 70 of the nation’s largest credit unions, plus a sampie of smaller credit unions to be collected once a month.
SOURCE: “The Redefined Monetary Aggregates.”

ever, several of the components used to calculate
some of the new aggregates are not seasonally ad­
justed because of data insufficiencies or technical diffi­
culties. The individual series that have not been
seasonally adjusted include NOW accounts, ATS ac­
counts, credit union share drafts, demand deposits
at thrift institutions, overnight RPs and Eurodollars,
money market mutual fund shares, term RPs at com­
mercial banks and savings and loan associations, and
term Eurodollars held by U.S. nonbank residents.
A much wider diversity of financial institutions
now participates in the data reporting and collection
process (see table 6). Financial institutions that have
not been active participants in the previous deriva­
tion of the monetary aggregates will play an impor­
tant role. For example, the Federal Home Loan Bank
Board now collects data on NOW accounts held at
savings and loan associations; beginning in the spring
of 1980, a sample of large credit unions will provide
data on credit union share drafts and related items;
problem, see Report, pp. 37-40 and, for a technical discus­
sion, David A. Pierce, Neva Van Peski, and Edward R. Fry,
“Seasonal Adjustment of the Monetary Aggregates,” Staff
Papers, pp. 71-90.




the Investment Company Institute provides a weekly
survey of money market mutual fund shares; and a
daily survey of 125 large member banks forms the
basis for the RP series. As this incomplete listing sug­
gests, the comprehensiveness of the new monetary ag­
gregates is greater than the previous measures.

CONCLUSION
The Federal Reserve Board recently has redefined
the monetary aggregates to provide better measures
of financial assets held by the public. The new basic
transaction measure, called MIA, is equal to the
former M l minus demand deposits held at commer­
cial banks due to foreign commercial banks and offi­
cial institutions. Large discrepancies between the
growth rates of Ml and MIA are not anticipated.
In addition, a broader transactions measure —
M1B — has been introduced. This aggregate com­
bines those deposits that are held, for the most part,
both for check-writing purposes and as savings ac­
counts. M1B, therefore, equals MIA plus NOW ac­
counts, savings accounts subject to automatic trans­
fer (ATS accounts), credit union share drafts, and
31

demand deposits at mutual savings banks (non-inter­
est-bearing). The growth of NOW and ATS accounts
has contributed to faster growth of M1B relative to
Ml or MIA. Consequently, if NOW accounts are
legalized nationwide, more rapid growth in M1B rela­
tive to MIA is expected.
The Board also has redefined M2 and M3 and has
introduced a new aggregate, L, which is intended
to measure total liquid assets held by the public.
These redefinitions represent a consolidation of the

32


former M2, M3, M4, and M5 measures, For instance,
the new M2 is similar in definition to the old M3; the
new M3 is similar to the old M5. The new L aggre­
gate, unlike any previous measure, includes such
items as term Eurodollars held by U.S. nonbank resi­
dents, bankers acceptances, commercial paper, U.S.
savings bonds, and other liquid Treasury obligations.
This broad measure of liquid financial assets is be­
lieved to provide a useful measure of credit in the
economy which arises either through the banking sec­
tor or through other financial channels.