Full text of Review (Federal Reserve Bank of St. Louis) : February 1984

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

Review
Vol. 66, No. 2




February 1984

5 The Dairy Price Support Program :
A Study o f M isdirected E conom ic
Incentives
15 Does Higher Inflation Lead to M ore
Uncertain Inflation?
27 Calculating the Adjusted M onetary
Base under Contem poraneous
Reserve Requirem ents

The Review is published 10 times per year by the Research and Public Information Department o f
the Federal Reserve Bank o f St. Louis. Single-copy subscriptions are available to the public free o f
charge. Mail requests f o r subscriptions, back issues, o r address changes to: Research and Public
Information Department, Federal Reserve Bank o f St. Louis, P.O. Box 442, St. Louis, Missouri
63166.
The views expressed are those o f the individual authors and do not necessarily reflect official
positions o f the Federal Reserve Bank o f St. Louis o r the Federal Reserve System. Articles herein may
be reprinted provided the source is credited. Please provide the Bank’s Research and Public Informa­
tion Department with a copy o f reprinted material.




Federal Reserve Bank of St. Louis
Review
February 1984

In This Issue .. .




In the first article of this Review, "The Daily Price Support Program: A Study of
Misdirected Economic Incentives,” Michael T. Belongia finds that the large and
rapidly expanding surplus of dairy products in the United States can be traced
directly to the production incentives offered by the dairy price support program.
He shows, by applying elementary economic principles, why effective price sup­
ports will create surplus production and, moreover, why the surplus will grow over
time if limits are not placed on production.
Belongia then applies the same economic principles to recent milk price and
production data to evaluate the likely impact of changes in the dairy price support
program approved by Congress in 1983. He finds that these changes are likely to
have little, if any, effect on current surpluses and will offer no solution to the
problem of larger surpluses in the future.
In the second article in this issue, “Does Higher Inflation Lead to More Uncertain
Inflation?” A. Steven Holland investigates the relationship between the rate of
inflation and the level of inflation uncertainty. He cites evidence from previous
studies of this relationship and discusses the theoretical arguments concerning
the likely consequences of greater inflation uncertainty on the economy.
Holland demonstrates that the results from empirical tests of the hypothesis
that higher inflation leads to more uncertainty about future inflation are sensitive
both to the measure of inflation uncertainty used and to the inclusion of energy
shocks into the analysis. He finds that, if individuals incorporate the impact of
energy shocks in their predictions of future inflation, there is no significant link
between inflation and inflation uncertainty. On the other hand, using measures of
inflation uncertainty derived from the Livingston survey of inflation expectations,
he finds a positive relationship between the rate of inflation and inflation uncer­
tainty even if the positive impact of energy shocks on inflation uncertainty are
included in the analysis.
Holland argues that the latter result is more likely to reflect the actual rela­
tionship between the rate of inflation and inflation uncertainty. Thus, he con­
cludes, the reduction of inflation uncertainty is an important, albeit often over­
looked, benefit of anti-inflation policies.
In the third article, “Calculating the Adjusted Monetary Base under Contempo­
raneous Reserve Requirements,” R. Alton Gilbert describes how the method of
calculating this Bank’s adjusted monetary base is modified to reflect the timing of
reserve accounting under contemporaneous reserve requirements. One important
result of the change is that the reserve adjustment magnitude (RAM), the compo­
nent of the adjusted monetary base that measures the effects of changes in reserve
requirements, must be estimated for the most recent weekly observations. The
adjusted monetary base for the most recent one or two weeks, therefore, will be
preliminary. Past data indicate that errors in estimating RAM will be small.
The adjusted reserves series will no longer be published on a weekly basis,
because of a change in the timing of data on currency in the hands of the public.
This Bank, however, will continue to publish adjusted reserves on a monthly
average basis.
3




The Dairy Price Support
Program: A Study of Misdirected
Economic Incentives
Michael T. Belongia

T

JL HE dairy price support program is likely to be the
focal point of agricultural policy in 1984. Dramatic in­
creases in the program’s cost have made it a visible
target for politicians concerned about federal budget
deficits. Consumer groups, who favor lower dairy
prices, also are opposed to the current program. Live­
stock producers, however, opposed recent changes in
the dairy program, fearing that such efforts to reduce
surplus daily production will promote a significant
slaughter of dairy cows that will keep beef prices at low
levels. These groups and their opposition to the dairy
program were confronted, as usual, by a politically
powerful dairy lobby.1
This article first reviews the history and mechanics
of the dairy price support program. Elementary eco­
nomic principles show why the dairy program has
generated an increasing volume of dairy surplus by
effectively maintaining milk prices above the competi­
tive market level. The program’s guarantee to purchase
all surplus product at the support price is shown to
produce an inefficient allocation of resources and a
transfer of wealth to dairy producers and suppliers of
production inputs. The analysis also demonstrates

Michael T. Belongia is an economist with the Federal Reserve Bank
of St. Louis. Robert W. Hess provided research assistance.
’ Since January 1, 1981, the dairy lobby has contributed over $1.3
million to 293 members of the House of Representatives. Two-thirds
of these officials voted against reductions in dairy price supports.
Moreover, much of legislative support for the dairy program — which
raises the cost of dairy products to consumers — comes from con­
gressmen who, for all practical purposes, have no dairy farmers in
their districts. For example, in the 1982 election, the dairy lobby
contributed to 117 congressmen from districts with less than 1 per­
cent of their populations engaged in farming. Seventy-two of these
congressmen voted against reductions in price supports. See Jack­
son and Birnbaum (1983).




why a lower support price would reduce both surplus
production and the prices of daiiy products without
large increases in the program’s cost. The article's final
section evaluates the likely effects of the compromise
dairy legislation, passed by Congress and signed by the
President last November, that provides for direct pay­
ments to farmers for reducing output.2

THE DAIRY PROGRAM:
A BRIEF HISTORY
For many years, milk prices have been supported,
both directly and indirectly, by a variety of government
initiatives.3In 1922, the Capper-Volstead Act effectively
exempted dairy cooperatives from antitrust actions,
thereby allowing producer organizations to restrict
output, charge higher prices for milk and earn
monopoly profits. Later, in 1935, amendments to the
Agricultural Adjustment Act established marketing
orders that set minimum prices for milk; the USDA was
charged with enforcing the payment of established
minimum prices to farmers.
Since 1949, the federal government has supported
the price of milk directly by guaranteeing to purchase
all milk that cannot be sold in the market at the federal­
ly established support price ($12.60 per hundred­
weight (cwt.) currently; $13.10 prior to last November).
The price of milk is maintained by Commodity Credit
Corporation (CCC) purchases of manufactured dairy
products from dairy processors. The CCC actually

2For general details on provisions of the legislation, see King (1983a).
3The dairy program is discussed in detail in Manchester (1983).
Donahue (1983) and Malcolm (1983) provide brief surveys of the
dairy program and its history.

5

FEBRUARY 1984

FEDERAL RESERVE BANK OF ST. LOUIS

C h a rt 1

C C C P u rch ase of D a iry Products: Butter, Ch e ese a n d Dry M ilk 11
Billions of

1970

pounds

1971

Billions of dollars

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

[X D a ta f o r f is c a l y e a r 1983 a re p r e lim in a r y .

processed milk products: butter, cheese and non­
fat dry milk. Prices also are supported indirectly by
food stamps, import restrictions on foreign daily prod­
ucts, and government purchases of milk for use in the
National School Lunch Program. To support milk
prices at the pre-November level, the CCC has pur­
chased more than 10 percent of all milk marketed by
farmers in recent years.
At the previous support price of $13.10 per cwt., the
volume of CCC purchases of surplus daiiy product
grew rapidly since 1979.4In the fiscal years 1977-79, net
CCC expenditures on the daiiy program averaged less
than $500 million annually.5 As the data in chart 1

“ Previous legislation had mandated an adjustment in the support level
each April and October such that the new support price represented
80 percent of parity. The last such increase, which raised the support
level to $13.10/cwt., occurred in April 1981.
5This average expenditure, however, includes only direct government
outlays for the purchase and storage of surplus product. Several


6


indicate, this expenditure more than tripled to an aver­
age cost of more than $1.8 billion for fiscal years 198082. In the fiscal year ending September 30, 1983, CCC
outlays for the purchase and storage of surplus dairy
product exceeded $2.7 billion.6 These rapid increases

studies have attempted to measure the additional social loss associ­
ated with the higher prices paid by consumers for the smaller volume
of dairy products they consume. Although these studies are some­
what dated, the indirect social cost of the dairy price support program
in the early 1970s was estimated to average nearly an additional
$100 million per year. In other words, the direct cost of the dairy
program represented, on average, about half of its full social cost.
Since large increases in the volume of surplus product purchased in
recent years implies that the gap between the support price and
competitive market price has widened substantially, the indirect so­
cial cost of the program may now be considerably larger. See Heien
(1977), Ippolito and Masson (1978), and Dahlgran (1980) for details
on the derivation of the social cost estimates.
6The United States is not alone in trying to curb large increases in the
costs of price support programs. See Tangermann (1983), for exam­
ple, on the structure and cost of the European Economic Commu­
nity’s dairy price support program.

FEDERAL RESERVE BANK OF ST. LOUIS

FEBRUARY 1984

C h a rt 2

Trends in M ilk Production

occurred, in part, because the dairy program — unlike
the price support programs of other commodities —
places virtually no restrictions on the volume of milk
that a dairy farmer can market at the support price.
The cost of the program increased, of course, be­
cause the program’s incentives caused output to grow
at a faster rate than the demand for dairy products. As



the data plotted in chart 2 indicate, milk production
increased at a 1.3 percent annual rate since 1970. Aside
from the incentive effects of the dairy program, this
steady increase in output also is partially attributable
to the 1.9 annual increase in average output per dairy
cow. Most important to the present analysis, however,
is the much larger 2.7 percent average rate of increase
in milk production since 1978.
7

FEDERAL RESERVE BANK OF ST. LOUIS

F ig u r e 1

How

P r i c e S u p p o r ts

FEBRUARY 1984

F ig u r e 2

Produce

Surpluses

That Grow

Larger O ve r lim e

The most recent version of the dairy program —
prior to the adoption of the paid diversion plan last
November — supported the price of milk at $13.10 per
cwt., but imposed two 50-cent-per-cwt. fees. The first
50-cent fee was collected on all milk marketed, effec­
tively lowering the 1983 support price to about $12.60
per cwt. The second 50-cent fee was to be collected
only if CCC purchases of surplus product were ex­
pected to exceed 7.5 billion pounds (milk equivalent);
this second fee was to be refunded, however, to pro­
ducers who reduced their production to specified
levels. Since this program was not in effect long enough
for the second assessment to be binding, the analyses
that follow deal in terms of only one 50-cent deduction.

THE SIMPLE ECONOMICS OF
DAIRY PRICE SUPPORTS
To illustrate how the existence of a price support
program affects the daiiy market, consider the model
of the daiiy market shown in figure 1. Without a sup­
port price, the long-run market equilibrium for milk
would be at point A where, at the competitive price PA,
the quantity of milk supplied to the market by milk
producers (Qa ) is equal to the quantity demanded by
consumers (Qa). At PA, where the market supply and
demand schedules intersect, there is neither a surplus
nor a shortage of milk. Since the quantity of milk
Digitized for
8 FRASER


How

Price S u p p o rts

Encou rage

M ore

Production

and M ore

Producers

brought to the market by producers at that price exact­
ly satisfies consumer demand, there are no incentives
for either producers or consumers to change their
rates of production or consumption.
If the support price is above PA, both consumers and
producers will modify their behavior in predictable
fashion. A price support like that represented by Ps
guarantees producers a higher return for their output
than they would otherwise have obtained. In the short
run, this higher return will provide an incentive for
dairy farmers to increase output to point B. At the
higher market prices, however, consumers will reduce
their milk consumption until they reach point C. The
net result of the increased production and reduced
consumption produces a short-run milk surplus, in
this instance equal to the difference between the quan­
tities supplied and demanded at Ps ( Q — Q c ) .
b

SHORT-RUN AND LONG-RUN
CONSEQUENCES OF MILK
PRICE SUPPORTS
Figure 2 shows why these changes in the dairy mar­
ket occur by focusing on the revenue and cost curves of
a representative dairy farmer. In the absence of a sup­
port price for milk, the representative producer (Farm
I) would produce qaI units of milk at the competitive
price of PA. At this level of output, both his short-run

FEBRUARY 1984

FEDERAL RESERVE BANK OF ST. LOUIS

and long-run marginal costs (SMC and LMC) of in­
creasing output by one unit are exactly equal to the
price he would receive from selling one more unit. He
has no incentive to increase milk production beyond
qaI, however, because the marginal cost of doing so
exceeds PA, the price he would receive for the milk.
Increasing output beyond qaI, then, would produce a
loss equal to the vertical distance between average cost
(SAC) and price times the total amount produced.
The stability o f the representative produ cer’s
equilibrium at a, is reflected in the market equilibrium
at point A in figure 1, where Qa units would be pro­
duced and consumed at the price PA- Output levels Qa
(for the industry) and qaI for daily Farm I depict a
long-run equilibrium for several reasons. First, because
the price received by each producer for his milk is
equal to both his marginal and average costs of produc­
tion, the representative producer is earning only a
normal rate of return on his production.7The absence
of short-run losses or rents indicates that there are no
incentives to attract new producers to the industry
(e.g., Farm II) or induce existing producers to change
their rates of output.
The figure also indicates that the total output pro­
duced by all dairy farmers is sold in the market for the
price at which the market supply and demand sched­
ules intersect. Until some exogenous factor changes
either the price or the position of the curves depicted
in figure 1, the representative producer’s output and
the market price and quantity will remain at their
respective long-run equilibrium points.
The introduction of a price support increases the
price received by producers and upsets the long-run
equilibrium at point A in figure 1. In the short run,
farmers engaged in milk production respond by in­
creasing their rates of output until their SMC is equal to
the new higher marginal revenue (which is equal to the.
support price, Ps). Thus, Farm I increases its rate of
output to units, which, when added to the increased
production of all other current daily farmers, increases
the quantity supplied in the market to Qb (figure 1). At
the higher support price, however, consumers pur­
chase only Qc units of milk. In the short run, the price
support generates a market surplus equal to the differ­
ence between quantities supplied and demanded at
the support price (Qb — Qc).

in clud ed in these costs is the capitalized value of the net earnings of
the dairy farm; this represents the sale value of the farm and, hence,
is a “ cost” of staying in the dairy business to the current farmer. See,
for example, Stigler (1966).




This initial surplus, however, understates the longrun impact of the price support program. The surplus
will continue to increase because the maintenance of a
support price above average cost introduces short-run
economic rent equal to the difference between the
support price and average cost times the higher level of
output produced. This rent gives an incentive to new
producers to enter the industry (e.g., Farm II) and to
existing producers to increase permanently the size of
their capital stock (larger herds, larger bams, etc.).
As new farms begin and existing farms expand pro­
duction, there will be increased demand for the scarce
resources needed to produce milk: daiiy cows, feed,
equipment, land and the specific skills necessary to be
a successful dairy farmer. Increased demand for these
inputs eventually will raise the marginal and average
costs of producing milk to LMCi and LAC'i in figure 2
where, at the new long-run equilibrium position (qdI
and qdn), economic profits for each producer in the
industry equal zero, just as they did originally. Notice
in figure 1, however, that the higher support price at Ps
eventually produces a long-run surplus of dairy prod­
ucts equal to the difference Qd — Qe. The surplus is
larger in the long-run because supply and demand
schedules become more elastic over time.8

BENEFICIARIES OF THE
DAIRY PROGRAM
Who, then, benefits from the dairy program and who
lobbies for its continuation? First, farmers who own
daiiy operations have benefited from price supports
because the values of the specific inputs (including
their own specific knowledge) used to produce milk
have increased. Without a support program, dairy
farming would be less profitable and, consequently,
the land, equipment and dairy cows used in milk pro­
duction would be less valuable.
The suppliers of inputs used in the dairy industry
also have benefited from the price support program.
The increased demand for their inputs by both old and
new dairy farmers tends to raise input prices and per­
mits suppliers of these inputs to earn greater profits

8The absolute value of the slope of a supply or demand schedule is
smaller as it becomes more elastic. Demand is more elastic in the
long run because substitutes can be found for higher priced dairy
products. The long-run supply cun/e is more elastic than the shortrun supply schedule because of the entry of new producers to the
industry. It is easy to see, for any level of price supports, that flatter
supply and demand curves will increase quantity supplied, reduce
quantity demanded and increase the size of the market surplus.

9

FEDERAL RESERVE BANK OF ST. LOUIS

than they would in the absence of a price support
program. Thus, not surprisingly, both input suppliers
and daiiy farmers oppose large reductions in the pro­
gram’s benefits because such reductions would re­
duce their wealth.

IMPACT OF THE 50-CENT
DEDUCTIONS IMPOSED IN 1983
An attempt was made in 1983 to reduce the growing
volume of surplus production — caused by behavioral
relationships like those in figures 1 and 2 — with the
imposition of a 50-cent fee on all milk produced.
Although adopted early in 1983, court rulings delayed
the actual collection of fees until late in the year. Essen­
tially, the fee amounted to a reduction in the support
price.
In terms of figure 3, the 50-cent deduction can be
treated as a parallel downward shift in the support
price line from $13.10 to $12.60 per cwt. From basic
economics, we know that a decrease in output price
will, ceteris paribus, lead to a decrease in output; pro­
ducers move down and to the left along their upwardsloping marginal cost curves. Starting from an
assumed long-run equilibrium at point Ap, the 50-cent
deduction would be expected to move producers to­
ward point Bp as they attempt to equate marginal cost
with the new, lower level of marginal revenue (price).
The net effect of all producers decreasing output in
this manner would be a reduction in quantity supplied
to the market, similar to the movement from point B to
point A in figure 1. In short, an effective decrease in
prices will cause individual producers and, hence, the
industry to scale back production to the point where
the new support price is equal to marginal cost.
A second effect of a lower support price — through
its negative effect on production — would be a reduc­
tion in the demand for inputs used in the dairy industiy. The reduced input demand would tend to reduce
input prices and exert pressure on some inputs to
leave dairy production. This market reaction would
lower costs and ultimately shift LAC to LAC' as figure 3
shows. In fact, producers would continue to exit from
the dairy industry until LAC' = SMC' = $12.60 for all
existing farmers and a new long-run equilibrium exists
at point Bp.
Thus, the 50-cent deduction should have promoted
a decrease in milk production. The absence of produc­
tion controls, however, led to the speculation among
some analysts that farmers would compensate for the

10


FEBRUARY 1984

50-cent deductions by producing more milk even if it
meant producing at a loss.9 The argument supporting
this conclusion is that farmers need to generate a mini­
mum level of revenue — or cash-flow — to meet oper­
ating expenses. Therefore, if prices are reduced, suf­
ficient cash-flow can be generated only by increasing
output. Thus, the argument goes, the deduction plan
would cause an increase in the dairy surplus rather
than a curtailment in its growth.
This reasoning is specious, however, as can be seen
from figure 3. If the representative producer’s SMC
curve is upward sloping, a lower price will be associ­
ated with a reduced volume of output as producers
move down and to the left along the SMC curve. There­
fore, unless the costs facing dairy producers behave
contrary to usual relationships, a lower support price
should cause reductions in output.10 The gaps be­
tween the predictions of economic theory and produc­
ers' actual response to the 50-cent deductions suggest
an alternative explanation for the increase in milk pro­
duction in 1983.

An Alternative Explanation fo r
Increased Output Under
Fee Assessment
A more conventional explanation of the increased
dairy production in 1983 can be based on a different
view of the cost structure facing milk producers. In­
stead of showing farmers producing at a short-run loss
after the 50-cent deduction as in figure 3, available data
indicate that the effective support price, even with the
deduction, still was greater than LAC. Under these
conditions, new and existing producers could con­
tinue to earn short-run economic rents by increasing
output as they did last year.
The best evidence that a $12.60/cwt. support level
remained above average cost can be found in data on
the size of dairy herds. Despite the 50-cent deduction

9ln recent congressional debate over changes in dairy legislation,
comments like the following were common: “ Instead of helping to
lower the milk supply, the $1 assessment program . . . has forced too
many dairy farmers to increase production in order to keep their cash
flow from declining to stay in business.” See Albosta (1983).
10Aside from the downward-sloping marginal cost argument, one
other explanation could explain increased production in response to
a lower support price. If the labor of the farm owner is treated
explicitly as an input and the owner faces a tradeoff between extra
revenue and leisure in his use of free time, it is possible to construct
a theoretical preference mapping that would allocate marginal time
to the production of extra revenue from increased milk output.

FEDERAL RESERVE BANK OF ST. LOUIS

F ig u re 4

F ig u r e 3

Eff ec ts o f th e F i f t y - C e n t D e d u c t i o n

Plan

and higher feed costs, farmers increased the size of
their herds during 1983. Moreover, with a record pro­
portion of replacement heifers, farmers are likely to
expand herds even further in 1984. These data suggest
that the LAC facing the typical producer still was less
than the support price of $12.60; consequently, farmers
were finding it profitable to expand production. The
data also imply that short-run rents could be earned by
dairy producers until the entry of new firms and in­
creased input demand increased the LAC to $12.60.
The production decisions of dairy farmers in recent
years appear to be consistent with the relationships
shown in figure 4. Historical data suggest that produc­
ers responded to higher support prices in 1980 by
increasing production from, for example, qa to qc in
figure 4. Furthermore, good weather and the incentives
of the grain price support programs produced large
stocks of relatively cheap feed grain. Lower feed prices
would shift SMC and SAC downward to positions like
SMC' and SAC' in figure 4. If these cost shifts have, in
fact, occurred, a lower support price of $12.60 still
produces short-run economic rents for all dairy farm­
ers operating at point Dp. This explanation suggests
that reductions in the support price beyond those
achieved by the 50-cent fee are necessary to reduce
surplus production. The relationships in figure 4 also
suggest that if the support price is not reduced further,
surplus production will continue to grow until com­
petition for inputs increases SAC' to a level that passes
through point Dp.



FEBRUARY 1984

C o s t and Production Relationships S uggested

b y M a r k e t Da ta

POSSIBLE EFFECTS OF THE
PAID DIVERSION PLAN
Efforts to reduce the support price further to $11.60
per cwt. have been defeated in Congress. Instead, com­
promise legislation that combines the 50-cent fee col­
lection and a lower $12.60 per cwt. support price
($12.10 effective support level) with a plan to pay farm­
ers directly for reducing output has been passed; the
payment is $10 per cwt. to farmers who reduce output
by up to 30 percent of historical levels. The program is
scheduled to be in effect for 15 months beginning
January 1, 1984. The bill also contains provisions for
further adjustments in the support level — either up or
down — in 1985 if the Secretary of Agriculture expects
CCC purchases to be less or greater than 5 billion
pounds milk equivalent. The analysis that follows con­
siders whether this combination of a lower support
price and paid diversion is likely to achieve the desired
reductions in surplus dairy production and program
costs.11
The key elements in plans to pay farmers directly to
reduce output are the 1981-82 base used in determin­
ing their historical production levels and the net bene­
fits to reducing output. Several existing pieces of data
are especially pertinent to this analysis. First, a USDA
study has determined that, since this 1981-82 base
period, 58 percent of dairy farmers have increased

11 Much of this analysis is based on Tipton (1983).

11

FEDERAL RESERVE BANK O F ST. LOUIS

output, while 36 percent decreased output.12 These
data imply that individual dairy farms have very differ­
ent cost structures and that their responses to a diver­
sion payment will vary substantially.
The main consideration is the calculation of net
benefits to reducing output. A first approximation of
this value is simplified below (all entries are in $/cwt.):
+ $10.00 diversion payment
+ $ 7.00 variable cost saved by reducing output
—$12.60 income lost from milk not sold
~

4.40 net benefit of reducing output.

This calculation actually is made more complicated,
however, by the unknown costs and benefits of taking
resources out of production, then, in 15 months when
the diversion program expires, adding them back into
production. In other words, a correct assessment of
net benefits must be based on a present value calcula­
tion that includes the full costs and benefits of partici­
pating in the diversion program. These other costs and
benefits include — among many others — net rev­
enues from cows slaughtered now, future replacement
costs of new cows and the increased long-run efficien­
cy of the herd if older cows are replaced by younger
animals. It is not clear, a priori, that the net benefits of
the diversion program — crudely estimated above at
$4.40/cwt. — still are positive when the present values
of expected costs and benefits are taken into account.13

Likely Response to Current
Diversion Incentives
The response of farmers to the paid diversion might
be best analyzed by considering how different groups
of dairy farmers have altered their rates of output in
recent years. Consider first dairy farmers who have
decreased production since the 1982 base year. These
farmers voluntarily have reduced output in response
to their relatively higher operating costs. By participat­
ing in the diversion plan, they will receive payments for
output reductions already achieved in 1983 or planned
for 1984. Therefore, while these producers have a
strong incentive to participate in the diversion plan, it
is not clear whether these payments will reduce their
12The remaining 6 percent of farms were formed after 1982 and would
not qualify for the diversion program.
13Just prior to publication of this article, the USDA announced that
only 12 percent of U.S. dairy farmers agreed to participate in the
diversion program. This low rate of participation is expected to
reduce dairy production by no more than 6 percent, about one-half
the production cutback sought by the legislation. A significant cost
factor cited by farmers who elected not to participate in the program
was the deterioration of their breeding stock’s genetic pool that
would result from selling some cows to reduce production for just 15
months. See King (1984) and Shipp (1984).


12


FEBRUARY 1984

production beyond the levels already achieved volun­
tarily in response to higher operating costs.
In contrast, consider those dairy producers with
relatively lower operating costs who have increased
output since 1982. To qualify for the diversion pay­
ments, these producers would have to reduce output
not only below the level planned in 1984, but also
below the increased level of 1983. Moreover, because
these producers can produce at lower cost, the diver­
sion payments may not be sufficient to offset the lost
revenues from output reductions. Therefore, the ex­
tent to which dairy farmers who increased production
since 1982 will participate in the diversion plan is un­
known. Chart 3 shows the break-even points for partic­
ipating in the diversion plan for farmers who have
expanded by different amounts since 1981-82. These
data indicate that participation will be less profitable if
output has been increased substantially in recent
years.
The diversion plan, then, appears to focus on the
following issues: relatively efficient producers — near­
ly 60 percent of the total — are less likely to participate
in efforts to reduce output; 6 percent more are not
eligible to participate in the program. Less efficient
producers — about 35 percent of the total — already
have reduced output below the 1982 levels to be used
as the historical basis for payments. Therefore, these
producers will be paid for output reductions already
achieved. Finally, it is unclear that the program’s bene­
fits will offset the full costs of adjusting production for a
plan scheduled to last only 15 months. The question
remaining is whether the incentives to reduce output
are sufficient to generate further reductions beyond
those realized since 1982.

One Study’s Results
One analyst has investigated and produced esti­
mates of the diversion program’s likely effects in 198414
Under certain assumptions about participation in the
program by different classes of producer, the plan
would show an intended reduction o f 15 billion
pounds of milk at a taxpayer cost of $1.5 billion (15
billion pounds X $10 per cwt.). Because 5.5 billion
pounds of this reduction are likely to have already
occurred, however, the diversion plan will pay $1.5
billion for the 9.5 (15.0 — 5.5) billion pound reduction
in output attributable to the program itself.
This reduction in output would be offset somewhat,
however, by the continuing increase in production by
the more efficient producers. After estimating further

14Tipton.

FEBRUARY 1984

FEDERAL RESERVE BANK OF ST. LOUIS

C h a rt 3

Returns from Participating at Different Levels in the
D a iry Diversion Program 11
Thousands of dollars
D ollar advantage to participation in dairy program

Percent reduction eligib le for payments
S o u r c e : F e l t e s , e t a l . (1983)

L l Returns are ca lcu la te d for a 5 0 -co w h erd with a 14,000-pound p ro d uction a v e r a g e .

effects from projected increases in dairy product de­
mand and revenues collected from the 50-cent fee
assessment, the diversion plan was projected to have
the following impacts next year:

o Seven billion pounds of surplus daily products

o $650 million would be collected from the 50-cent
fee assessments.15
Under these assumptions, the net cost of the dairy
program in 1984 (in billions of dollars) is estimated to
be:
$1.20 + $1.50 - $.65 ~ $2.00

would be produced;

o The surplus product would cost $1.2 billion to
purchase and store;

o The diversion payments would cost $1.5 billion;
and



15The proposed legislation also intends to reduce the surplus by
increasing domestic demand for dairy products. This is to be
achieved through increased advertising expenditures paid for with
fees assessed on dairy producers. Planned expenditures of $200
million per year for dairy advertising would be a 250 percent in­
crease over 1982’s advertising expenditures.

13

FEDERAL RESERVE BANK OF ST. LOUIS

FEBRUARY 1984

This figure is about double the Office of Management
and Budget estimates of the cost of a daiiy program
based solely on a reduction in the support price to
$11.60.16The OMB also estimates that the program will
increase the cost of dairy products to consumers by
$1.8 billion.17Furthermore, the diversion payments are
expected to have little effect on the long-run surplus
problem because the oldest and least productive cows
will be slaughtered.18 This will leave the dairy herd
younger and more productive when the program and
its payments end early in 1985.

port price further. Congress already has defeated a
proposal to reduce the daiiy price support level and
has passed instead a plan to pay farmers directly for
reducing output. Under these conditions, consumers
can expect to pay higher prices for daiiy products,
while taxpayers can expect further increases in the
costs of this program.

CONCLUSIONS

Dahlgran, Roger A. “ Welfare Costs and Interregional IncomeTransfers Due to Regulation of Dairy Markets,” American Journal of
Agricultural Economics (May 1980), pp. 288-96.

The foregoing analysis suggests several important
conclusions about the daiiy price support program.
First, a price support program without production
controls will generate increasing surpluses and pro­
gram costs. Second, the dairy price support — at least
since 1980 — has been kept substantially above what
would have otherwise been a competitive market price.
This has caused an inefficient allocation of resources
(too many resources allocated to daiiy production)
and transferred wealth from consumers and taxpayers,
in general, to daiiy producers and suppliers of inputs
to the dairy industry. Third, efforts to reduce surplus
production by paying farmers not to produce are likely
to have little impact on surplus production, particu­
larly in the long run, but will keep program costs near
their current levels. Finally, the only effective way to
reduce surplus daiiy production is to reduce the sup­

REFERENCES
Albosta, Paul. U.S. Representative, Congressional Record,
November 9, 1983, p. H9523.

Donahue, John D. “ The Political Economy of Milk," Atlantic Monthly
(October 1983), pp. 58-68.
Feltes, Linda, et al. "The Dairy Compromise Program: Who Should
Participate?” FM 522, University of Minnesota Agricultural Exten­
sion Service, December 1, 1983.
Heien, Dale. “The Cost of the U.S. Dairy Price Support Program:
1949-74,” Review of Economics and Statistics (February 1977),
pp. 1- 8.
Ippolito, Richard A. and Robert T. Masson. “ The Social Cost of
Government Regulation of Milk," Journal of Law and Economics
(April 1978), pp. 33-65.
Jackson, Brooks and Jeffrey H. Birnbaum. “ Dairy Lobby Obtains
U.S. Subsidies With Help From Urban Legislators,” Wall Street
Journal, November 18, 1983.
King, Seth S. “ Dairy Bill Poses Perils For A Presidential Veto,” New
York Times, November 19, 1983a.
________ “ How the Dairy Lobby Put the Squeeze on the White
House," New York Times, December 4, 1983b.
Malcolm, Andrew. “ Dairy Output is Still Rising Despite Laws,” New
York Times, November 11,1983.

16Jackson and Birnbaum.

17King (1983b).
18More to the point, any reduction in output achieved through smaller
numbers of dairy cows will be short-lived because output increases
have come primarily from greater productivity per animal. This point
is highlighted by comparing the 1.6 percent increase in the number
of dairy cows between 1980 and 1983 to the 5.1 percent increase in
average output per cow (from 11,889 lbs. to about 12,400 lbs. per
year) over the same period.

Digitized for
14FRASER


Manchester, Alden. The Public Role in the Dairy Economy (Westview Press, 1983).
Stigler, George J. The Theory of Price, 3rd ed. (The Macmillan
Company, 1966).
Tangermann, Stefan. “ Europe and the Road to Serfdom,” The
Financial Times, November 2, 1983.
Tipton, E. Linwood. “ Dairy Outlook ’84” (paper presented atthe 60th
Annual USDA Outlook Conference, Washington, D.C., November
1, 1983).

FEDERAL RESERVE BANK OF ST. LOUIS

FEBRUARY 1984

Does Higher Inflation Lead to
More Uncertain Inflation?
A. Steven Holland

I n recent years, many countries have experienced
"stagflation,” a period of high and rising inflation and
unemployment. Over this time, higher inflation in­
creasingly has come to be blamed for higher unem­
ployment and reduced growth of real output. This
contrasts sharply with previously held notions that
there was either a long-run tradeoff between inflation
and unemployment or a “natural rate of unemploy­
ment” regardless of the inflation rate.

Thus, if reducing inflation produces sufficiently larger
output growth and lower unemployment in the long
run, it is a worthwhile venture, even if doing so would
produce a large short-term loss of output and rise in
unemployment.3 While Friedman’s discussion pri­
marily concerns the variability of inflation — not
necessarily identical to the notion of inflation uncer­
tainty — it is clear that he considers them to be closely
related.

One reason why many people have changed their
minds about inflation’s impact on the economy is the
presumed impact of “inflation uncertainty.” Many now
argue that there is greater uncertainty about future
prices during periods of higher inflation.1 This in­
creased uncertainty leads to a less efficient allocation
of resources.

This argument can be split into three separate
hypotheses: (1) higher inflation leads to greater
variability of inflation; (2) greater inflation variability
implies greater uncertainty about future inflation; and
(3) greater inflation uncertainly has a detrimental effect
on economic activity. For policymakers to be con­
cerned about the relevance of hypothesis 3, they must
believe that they can influence the level of inflation
uncertainty. Hypotheses 1 and 2 state that they can do
this by controlling the rate of inflation. If exogenous
factors, such as energy shocks, are primarily responsi­
ble for greater inflation uncertainty, then policymakers
can do little to affect it.

The best-known statement of this view came from
Milton Friedman in his Nobel Lecture. Briefly stated,
Friedman argued that greater inflation uncertainty
shortens the average duration of contracts and re­
duces the efficiency of the price system. These two
forces combine to lower the growth rate of real output
and potentially increase the rate of unemployment.2

A. Steven Holland is an economist at the Federal Reserve Bank of St.
Louis. Jude L. Naes, Jr., provided research assistance.

This article focuses on the validity of the first two
hypotheses, which together imply that higher inflation
leads to greater inflation uncertainty. Besides analyz­
ing the causes of inflation uncertainty, an assessment
of its potential effects is presented as well.

1Some have suggested that uncertainty begins to increase once the
rate of inflation rises above some threshold. For example, see Logue
and Willett (1976) and Hafer and Heyne-Hafer (1981).
2Friedman (1977). He suggests that the natural rate hypothesis holds
for the very long run (a period of decades), because the economy’s
institutional structure for dealing with inflation eventually will adjust to
eliminate the real effects of inflation.




3To determine whether the long-term benefits of anti-inflation policies
would offset the short-term costs, one must consider the timing of the
output effects and the rate at which future output gains are dis­
counted. See Meyer and Rasche (1980).

15

FEDERAL RESERVE BANK OF ST. LOUIS

Since energy shocks have been the single most im­
portant factor accounting for temporary price level
changes, this article also investigates the impact of
changes in the relative price of energy on both the rate
of inflation and the level of inflation uncertainty.4Ener­
gy shocks and inflation uncertainty should be positive­
ly associated, because the magnitude and timing of the
effects of an energy shock on the rate of inflation are
bound to be viewed with uncertainty.

WHAT IS INFLATION UNCERTAINTY?
Inflation uncertainty arises from a lack of complete
knowledge about how future price levels are deter­
mined. Of course, an individual typically will have
enough information to make some forecast of future
inflation rates. A given estimate of next period’s infla­
tion can be thought of as the mean of some underlying
probability distribution.
The forecaster’s inflation uncertainty may be esti­
mated by looking at the size of some specified confi­
dence interval for his forecast. For example, a person
may have predicted at the end of 1982 that 1983 infla­
tion had a 90 percent probability of being between 3
percent and 5 percent. If the same individual’s 90 per­
cent confidence interval for 1984 inflation (forecast at
the end of 1983) is wider, say 4 percent to 7 percent,
then his uncertainty about 1984 inflation is greater
than it was for 1983 inflation.
The analysis presented here deals with inflation un­
certainty for a representative individual. Though the
level of an individual's uncertainty about inflation is
not directly observable, ways of estimating it have been
suggested in the literature. One of these is to use the
variance or standard deviation of the errors made in
forecasting inflation. A forecaster is trying to predict
the outcome of a process that has both systematic and
random components. With an unbiased forecast of the
inflation rate, the variance of the forecast errors indi­
cates the importance of the random component and
can be considered an estimate of the level of inflation
uncertainty.5 An implicit assumption in this type of
analysis is that the variance need not be constant but
may vaiy over time.

4See Tatom (1981).
5lt is the ex ante, not the ex post, variance of forecast errors that is
relevant. Estimates of the latter, however, are commonly used as
proxies for inflation uncertainty. See, for example, Klein (1978),
Engle (1983), and Pagan, Hall and Trivedi (1983).

Digitized for
16FRASER


FEBRUARY 1984

WHY DOES INFLATION UNCERTAINTY
MATTER?
The real effects of inflation uncertainty arise in part
because inflation expectations enter into the contract­
ing process. Any contract that provides for payment in
nominal rather than real terms incorporates an ex­
pectation of future inflation. If actual inflation ends up
higher than was expected when the contract was
made, a redistribution of wealth occurs: those making
the contracted nominal payments gain and those re­
ceiving them lose. If actual inflation is lower than was
expected, the opposite wealth redistribution occurs.
When there is greater inflation uncertainty, riskaverse individuals will attempt to shorten the duration
of contracts to reduce the risk of loss caused by devia­
tions of actual from expected inflation. More frequent
negotiation of contracts will divert economic resources
to the contracting process from other, previously more
efficient uses.6
As the accompanying insert demonstrates, greater
inflation uncertainty increases the risk associated with
both saving and investing, since both require a contract
of some kind. Individuals faced with greater inflation
uncertainty may choose to reduce both their planned
savings and investment. The result is likely to be lower
long-term real economic growth.
Another potential recil effect of inflation uncertainty
is reduced efficiency of the price system in allocating
resources. The basic idea is this: the more uncertain is
inflation, “the harder it becomes to extract the signal
about relative prices from the absolute prices.”7 Be­
cause individual prices adjust to unexpected inflation
at different rates due to the presence of long-term
contracts and the costs of adjusting prices, relative
prices may be temporarily distorted.8They also may be
incorrectly perceived, because information does not
flow smoothly across markets. As a result, economic
efficiency is reduced, producing lower output growth

indexation of contracts can reduce (though not totally eliminate) the
risk associated with contracting, and one would expect indexation to
increase as inflation uncertainty increases. For a theoretical analysis
of indexation in this context, see Gray (1978). Klein finds evidence
that an increase in “ long-term price uncertainty” leads to a reduction
in the average term to maturity of outstanding corporate debt.
7Friedman, p. 467. Again, Friedman’s discussion is in terms of infla­
tion variability; if this variability were anticipated, however, adjust­
ments could be made that reduce or eliminate this effect. His discus­
sion of this effect is based on the work of Hayek (1945) and Lucas
(1973) among others.
8See Bordo (1980) and Sheshinski and Weiss (1977).

FEBRUARY 1984

FEDERAL RESERVE BANK OF ST. LOUIS

How Inflation Uncertainty Creates Greater Risk for
Savers and Investors
To see how greater inflation uncertainty affects
savings and investment, consider the consequences
of an unexpected price level increase for a saver. The
expected real rate of return (r*) on savings can be
written:
(1) r* = i - p*,
where i is the nominal rate of return (assumed to be
constant) and p* is the expected rate of inflation.
There is risk to the saver, because the realized real
rate of return (r) will only equal the expected real
rate if the actual inflation rate (p) equals the ex­
pected inflation rate. The difference between the
realized and expected real rates can be written as:
(2)

r - r* = - (p — p*).

Using the variance of the difference between the
actual and expected real return as a measure of risk
and the variance of inflation forecast errors as an
estimate of inflation uncertainty, inflation uncer­
tainty and risk are equated:1
'A similar analysis can be carried out for other types of contracts.
For exam ple, if one considers w ag e contracts, the risk m easure is
the variance of th e actual less th e expected real w age.

and possibly higher unemployment than if all relative
prices were correctly perceived.9
The notion that greater inflation uncertainty leads to
reduced economic growth and higher unemployment
has been supported by empirical research. Mullineaux
finds some measures of inflation uncertainty to have a
negative effect on the growth of industrial production
and a positive effect on unemployment; Levi and Makin
get similar results for employment growth. Further­
more, Blejer and Liederman find that increased disper-

9Carlton (1981) discusses in detail the impact of inflation uncertainty
on the organization of markets. He concludes that (p. 19):
. . . in response to inflationary uncertainty, we expect to see fewer con­
tracts with fixed prices for long time-periods, fewer customized goods,
greater use of standardized goods sold in a liquid market, a move from
outside contracting of customized goods to internal production through
vertical integration, and a move from vertical integration to reliance on
standard quality goods sold in a liquid market where “the market" price is
easy to observe. All of these changes are undesirable from an efficiency
standpoint.




(3) var (r — r*) = var (p — p*).

The effect of greater risk on the flow of savings is
not clear a priori. The greater risk could reduce the
savings of risk-averse individuals and, as a conse­
quence, reduce the actual amount of investment as
well. If a person’s goal in saving is to assure a given
level of recil wealth in the future, however, greater
risk could actually lead to increased savings.
For investment in physical capital, the analysis is
not as straightforward, because the nominal rate of
return (i) varies to some degree with the rate of
inflation. A deviation of actual from expected infla­
tion does not necessarily indicate that the realized
real rate is different from the expected real rate of
return. Therefore, the effect of inflation uncertainty
on investment risk depends on how the nominal
rate of return is expected to respond to a change in
the rate of inflation. This response may also be
viewed with uncertainty. If investors are risk averse,
then risk-increasing inflation uncertainty will re­
duce investment and lower output growth in the
long term.

sion of relative price changes leads to a significant
reduction in real GNP and increased unemployment.10

WHY SHOULD HIGHER INFLATION
LEAD TO GREATER INFLATION
UNCERTAINTY?
The relationship between higher rates of inflation
and inflation uncertainly is based more on empirical
regularities than on theoretical rationale. Beginning
with Okun in 1971, several researchers have found that
there are significant positive correlations between
rates of inflation and the variability of inflation across
countries and across time for a given countiy. Others

10See Mullineaux (1980), Levi and Makin (1980), and Blejer and
Leiderman (1980). Evans (1983) finds an unstable price level to
have a negative effect on real GNP growth, and Able (1980) finds a
negative impact of inflation variability on the rate of investment.

17

FEBRUARY 1984

FEDERAL RESERVE BANK OF ST. LOUIS

have found a positive relationship between inflation
variability (or inflation itself) and proxies for inflation
uncertainty, the latter including the dispersion of infla­
tion expectations across survey respondents and the
variance of estimated inflation forecast errors. The in­
sert on pages 20 and 21 provides a summary of findings
from previous studies.
The theoretical rationale centers on the hypothesis
that a more inflationary economy produces greater
uncertainty about the future direction of government
policy, causing greater uncertainly about future infla­
tion. Okun states that the application of fiscal and
monetary policies is apt to be less consistent (i.e., pre­
dictable) during inflationary times because of the diffi­
culty in reducing inflation without causing unacceptably high rates of unemployment and interest.11 In a
similar vein, Friedman states that:
A burst of inflation produces strong pressure to coun­
ter it. Policy goes from one direction to the other, en­
couraging wide variation in the actual and anticipated
rate of inflation. And, of course, in such an environment,
no one has single-valued anticipations. Everyone recog­
nizes that there is great uncertainty about what actual
inflation will turn out to be over any specific future
interval.12
One can argue that an inflationary economy creates
an environment in which major policy changes be­
come more likely and the effects o f such policy
changes become more uncertain. To support this
argument, one need only look at some of the policy
measures taken or proposed in recent years at least
partially in response to an inflationary economy: de­
regulation of financial institutions, wage and price
controls, indexation of income taxes and changes in
methods for implementing monetary policy.

INFLATION FORECASTS AND THE
VARIANCE OF ERRORS
The discussion above suggests that the variance of
errors in forecasting inflation could be used as one
measure of inflation uncertainty. If the variance of the
forecast errors remains constant over time, so does the
level of inflation uncertainty. One way to determine
whether inflation uncertainty has changed over time is
to test for non-constant variance (i.e., heteroscedasticity) in the residuals from a model of inflation expec­
tations.13

"S e e Okun (1971).
12Friedman, p. 466.
13This is the approach suggested by Engle (1982) and Pagan, Hall
and Trivedi.


http://fraser.stlouisfed.org/
18
Federal Reserve Bank of St. Louis

Table 1
Two Models of Inflation Expectations:
11/ 1954- 111/1983
( 1)

(2)

Intercept

0.108
(0.40)

0.376
(1.46)

P i-i

0.285
(3.35*)

0.210
(2.59*)

Pi —2

0.099
(1.17)

0.100
(1.26)

0.309
(3.92*)

0.261
(3.56*)

M.- 1

0.109
(2.47*)

0.125
(3.05*)

M.-2

0.023
(0.48)

0.037
(0.84)

0.022
(0.44)

0.026
(0.56)

0.102
(2.08*)

0.107
(2.31*)

pf-1

—

0.017
(1.43)

Pf-2

—

0.039
(3.10*)

P t-3

M .- 3

M,

4

D1,

-1 .0 1 7
(-1 .8 5 *)

-0 .9 3 6
(-1 .8 5 *)

D2,

2.022
(3.98*)

1.189
(2.35*)

Rz

0.770

0.804

SE

1.33

1.23

h

1.46

0.50

t-statistics in parentheses.
'Significant at the 5 percent level (one-tailed test).

First, we need an inflation expectations model that
provides unbiased forecasts over both lower and high­
er inflation periods; we can then test whether the error
variance is larger for the higher inflation period. A
model obtained by regressing the quarterly growth rate
of the GNP deflator (p) on its own lagged values, lagged
values of the growth rate of M l (M), and dummy vari­
ables for periods ofwage-price controls and their aftermath (D1 and D2) is given by equation 1 in table I.14The
equation was estimated using data from 11/1954—111/

14AII growth rates are expressed in annualized log differences. D1 has
a unity value during the control period of 111/1971—1/1973 and zero
otherwise. D2 represents the period during which controls were
being phased out, taking a unity value for the period 1/1973-1/1975
and zero otherwise.

FEDERAL RESERVE BANK OF ST. LOUIS

Table 2
Tests for Inflation Uncertainty Using
Regression Models of Inflation
Expectations_____________________
(1)

ef = 0.732 - 0.035 p ,_ , + 0.163 p t- 2
(1.80*) (-0 .2 5 9 )
(1.17)
- 0.144 p ,_3 + 0.225 p,_4
(-1 .0 3 )
(1.67*)
R2 = 0.082

(2)

SE = 2.24

DW = 2.13

e f = 0.881 - 0.055 p ,_ , + 0.127 p t_2
(2.25*) (-0 .4 2 )
(0.942)
- 0.140 p ,_3 + 0.180 p , _4
(-1 .0 4 )
(1.39)
R2 = 0.040

SE =2.17

DW = 2.00

t-statistics in parentheses.
'Significant at the 5 percent level (one-tailed test).

1983, and the number of lags was chosen using stan­
dard t and F tests. When we divide the sample into a
lower inflation period, 11/1954—IV/1967, and a higher
inflation period, I/1968-III/1983, we can reject the
hypothesis that the error variance is the same in both
periods.15 As expected, the variance is higher in the
period of higher inflation.16

Another test of the constancy of the variance of the
forecast errors over time is obtained by regressing the
squared value of the inflation forecast error for period t
(ef) estimated from equation 1 on the variables thought
to cause changes in the variance. When four lagged
values of the inflation rate are used, the estimated
equation yields the results shown in equation 1 in table
2. The results indicate, once again, that inflation affects
the variance of forecast errors using this model of ex-

15The average quarterly rate of growth of the GNP deflator between
11/1954 and IV/1967 was 2.18 with a maximum of 4.57 and a mini­
mum of -0 .8 7 ; for 1/1968-111/1983, the average growth rate of the
GNP deflator was 6.31 with a maximum of 11.41 and a minimum of
2.83. The value of the calculated F-statistic (F ^ 47 = 1.77) from the
Goldfeld-Quandt test is statistically significant at the 5 percent level.
For an explanation of this test for heteroscedasticity, see Goldfeld
and Quandt (1965).
16A Chow test does not indicate that the structure of the model is
different for the two periods. The Chow test statistic is F10 98 =
0.705, far below the level required for statistical significance at the 5
percent level.




FEBRUARY 1984

pected inflation17 The effect over four quarters is both
positive (as indicated by the sum of the coefficients of
the lagged values of the inflation rate [0.209]) and statis­
tically significant at the 5 percent level.18

Relative Energy Price Changes and
Expected Inflation
The above result seems to suggest rather strongly
that a higher inflation rate is associated with more
inflation uncertainty. This conclusion must be careful­
ly viewed, however; the results are quite sensitive to the
way in which the model of inflation expectations is
specified. In particular, if one considers the possibility
that individuals anticipate some impact of a higher
relative price of energy on the rate of inflation, then
inflation does not affect the variance of the errors. An
estimated inflation expectations model that incorpo­
rates two lagged values of the change in the relative
price of energy is presented in equation 2 of table l.19
When the sample was divided into the same lower and
higher inflation periods as before (and the impact of
energy prices is taken into account), the hypothesis
that the error variance is the same in both periods
cannot be rejected at the 5 percent level of sig­
nificance.20
Furthermore, as equation 2 in table 2 shows, lagged
values of the inflation rate do not affect the squared
inflation forecast error estimated from equation 2 in
table l.21 Therefore, when this inflation expectations
model is used, there is no indication that higher infla­
tion is associated with greater inflation uncertainty.

17The test statistic TR2 (where T is the number of observations) has a
X 2 distribution with degrees of freedom equal to T minus the number
of regressors. This statistic is used to test for heteroscedasticity. In
equation 1 in table 2, TR2 = 9.62, which is statistically significant at
the 5 percent level. This test is suggested by Engle (1982).
18The t-statistic for the sum of the coefficients is 2.59. Additional
lagged values of p up to a total of 12 had no effect. Lagged values of
the rate of inflation are used instead of the current rate, because the
rate for period t is not known at the time the forecast is made. This
procedure of regressing squared residuals on a set of variables as a
test for heteroscedasticity is suggested by Breusch and Pagan
(1979).
19The relative price of energy is defined as the ratio of the “fuels and
related products and power” component of the producer price index
(PPI) to the business sector deflator. See Tatom for a slightly
different model of the inflation rate itself (rather than expected
inflation).
2aThe Goldfeld-Quandt F-statistic is Fsn 4S = 1.47.
21 Neither the value of TR2 (4.72) nor the sum of the coefficients of
lagged inflation (0.112, t = 1.44) are statistically significant at the 5
percent level.

19

FEDERAL RESERVE BANK OF ST. LOUIS

FEBRUARY 1984

Previous Research on the Relationships among
Inflation Rates, Inflation Variability and
“Inflation Uncertainty”
Article

Countries Studied

Time Periods

M ajor Findings

Okun (1971)

17 industrialized
OECD countries

1951-1968

High correlation between the average
annual percentage increase in the GNP de­
flator and the standard deviation of annual
inflation rates.

Gordon (1971)

same as Okun

1960-1968

Smaller correlation than in Okun. Also, if
five relatively small countries are omitted,
correlation disappears.

Logue and Willett
(1976)

41 industrialized
and nonindustrialized
countries

1949-1970

Inflation rates of no more than 2-4 percent
cause no problem of increased variability of
inflation.

Jaffee and Kleiman
(1977)

same as Okun
United States
(survey)

1951-1971
1955-1971
(survey)

(a) Positive correlation across countries
between inflation and its variance,
though correlation is weak during
1960s.
(b) Positive relationship between mean
and standard deviation of inflation
rates expected in the SRC survey.

Foster (1978)

40 countries

1954-1975

Large correlations between inflation and
the average absolute change in inflation.

Cukierman and
Wachtel (1979)

United States

1948-1975
1955-1976

Variance of expected inflation across indi­
viduals is positively related to variance of
actual inflation using both the Livingston
and SRC surveys.

Taylor (1981)

7 large industrialized
countries

1954-1979

Strong correlation (except during 1960s)
between average inflation and its standard
deviation, which is at least partially due to
association between average inflation and
inflationary shocks.

Fischer (1981)

United States

1806-1979
1954-1976
(survey)
1950-1980
(survey)
1948—1980

(a) Positive relationship between inflation
and its variability.
(b) Variance of expected inflation across
individuals is positively associated with
actual, expected and unanticipated in­
flation using the Livingston and SRC
surveys.
(c) No significant association between the
rate of inflation and the variance of re­
siduals from a forecasting equation of
the inflation rate.


20


FEBRUARY 1984

FEDERAL RESERVE BANK OF ST. LOUIS

Frohman, Laney and
Willett (1981)

United States

1954-1976

Positive relationship between both actual
and expected inflation and the variance of
expected inflation across individuals using
the Livingston survey.

Hafer and
Heyne-Hafer (1981)

same as Logue and
Willett, excluding
Chile

1970-1979

Inflation and its variability are positively
related; may require rates as high as 9 per­
cent.

Pagan, Hall and
Trivedi (1983)

Australia

1973-1981
1968-1982

Inflation affects variance of errors in fore­
casting, as measured by squared estimated
forecast errors.

Engle (1983)

United States

1947-1979

Inflation does not affect squared value of
estimated forecast errors. Past values of
squared forecast errors do.

INDIVIDUAL INFLATION
UNCERTAINTY AND THE VARIABILITY
OF INFLATION EXPECTATIONS
AMONG INDIVIDUALS
The preceding tests illustrate one of the major prob­
lems involved in trying to estimate an individual’s
uncertainty about future inflation: estimates can be
sensitive to one’s assumptions about the nature of the
information used to forecast inflation. In this section,
we use a different approach to estimating inflation
uncertainty based on very different assumptions about
the way individuals form inflation expectations.
In contrast to the models of inflation expectations
estimated previously, individuals may use consider­
ably more information to forecast inflation rates than
the past growth rates of such aggregates as the price
level and money supply. For example, each forecaster
may have personal information regarding the histori­
cal relationship between the price of a specific product
and the general price level. This specialized informa­
tion is likely to be too costly for all individuals to obtain.
If there is greater heterogeneity in the inflation signals
that forecasters receive from this type of marketspecific information, then greater dispersion of indi­
vidual inflation forecasts can result. An individual who
observes a wider variety of predictions of next period’s
inflation rate (through published sources, for example)
may become more uncertain about the accuracy of his
own forecast, especially if he believes that different
forecasts are based on information he does not have.22
22This kind of partial information framework is used by Cukierman and
Wachtel (1982). There is, however, an alternative explanation for




In the analysis to follow, it is assumed that greater
dispersion of inflation forecasts among individuals
leads to increased inflation uncertainty. Therefore, we
use measures based on the variability of responses to
the Livingston survey of inflation expectations to fur­
ther investigate the relationship between inflation and
inflation uncertainly23
The standard deviation of the individual inflation
forecasts from the Livingston survey is the first proxy
for inflation uncertainty. Chart 1 shows the actual infla­
tion rate over the forecast period and the mean and
standard deviation of six-month inflation forecasts
from the first half of 1954 to the first half of 1983. The
shaded areas of the chart represent periods of energy
shocks.24 The chart indicates that both energy shocks
increased variability of individual inflation forecasts: forecasters
may all use the same information but in different ways. This would
not necessarily imply greater inflation uncertainty for a particular
individual since each forecaster could be just as certain as he ever
was about the accuracy of his forecast.
23Joseph Livingston of The Philadelphia Inquirer conducts a survey
each spring and fall, requesting respondents to indicate their predic­
tions about a number of economic indicators including the consumer
price index (CPI). Because the survey results published, for exam­
ple, in June contain predictions for the following December, Living­
ston refers to them as six-month-ahead forecasts as does this
article. (The survey also includes 12-month forecasts, which are not
used here.) Because the respondents to the June survey are
thought to know only the April CPI, however, they are actually
predicting an eight-month rate of change. For a detailed discussion
of the Livingston expectations data, see Carlson (1977). This article
uses the data in Carlson’s revised form updated to the present.
24The periods of energy shocks are the first half of 1973 to the second
half of 1974, and the first half of 1979 to the first half of 1981. The
quarterly deflator for fuels and related products and power divided
by the business sector deflator grew at an annual rate of 22.9
percent from IV/1972 to IV/1974 and 23.4 percent from 1/1979 to
11/1981.

21

FEDERAL RESERVE BANK OF ST. LOUIS

FEBRUARY 1984

C h a rt 1

A ctu al Inflation, Expected Inflation and Standard D eviation
of Six-M onth Inflation Forecasts

S h a d e d a re a s re p re s e n t p e r io d s o f e n e rg y p ric e sh ocks.
L a te s t d a t a p lo tte d : F irst h a lf 1983

and inflation may have a positive impact on inflation
uncertainty. All three series rose substantially during
periods of energy shocks, and there are significant
positive correlations between the uncertainty measure
and the other two series in other periods.25
The root-mean-squared error (RMSE) of the indi­
vidual forecasts of inflation from the survey serves as
the second proxy for inflation uncertainty. An ex­
amination of chart 1 indicates that the survey mean
inflation expectation is biased; it consistently underpredicts the inflation rate over most of the sample
period. The RMSE of the inflation forecasts incorpo­

25The correlation coefficient between the standard deviation and the
expected inflation rate is 0.787 for the entire period and 0.667 for the
period omitting the two energy shock periods. Between the standard
deviation and the actual inflation rate, the correlations are 0.724 and
0.597, respectively. These figures are all statistically significant at
the 5 percent level.


22


rates these errors. The squared value of this variable is
the sum of the variance of inflation expectations across
survey respondents (the standard deviation squared)
and the squared forecast error using the survey mean
as the expected inflation rate.26The use of this variable

26The mean-squared error of the forecasts can be written:

1 2n

MSE, = £

(pS - p,)2

i= 1

= (Pt*“ Pt)2 + fj

2

(pH - p?)2,

i= 1

where n is the number of forecasters, p* is the expected rate of
inflation for the ith forecaster, and p* is the mean expected inflation
rate among the forecasters. The first term on the right-hand side of
the equation is the squared forecast error, and the second is the
variance of individual inflation expectations. We use the square root
of this variable and the standard deviation of expectations because
regressions using the mean-squared error and the variance exhib­
ited heteroscedasticity.

FEBRUARY 1984

FEDERAL RESERVE BANK OF ST. LOUIS

A ctu al Inflation, Expected Inflation and R o o t-M ean -Sq u ared Error
of Six-M onth Inflation Forecasts

S h a d e d a re a s re p re s e n t p e r io d s o f e n e rg y p ric e sh ocks.
L a te s t d a t a p lo tte d : F irst h a lf 1983

as a measure of inflation uncertainty assumes that
there is greater inflation uncertainty, holding constant
the variance of inflation expectations, when a large
mean forecast error occurs than when a small mean
forecast error is observed.27
Chart 2 plots the RMSE along with the actual infla­
tion rate and the mean expected inflation rate from the
survey. Again there is a positive association between
the uncertainty measure and the other two series, with
the largest increases in RMSE occurring during periods
of energy shocks.28 As chart 2 shows, the RMSE is

27The standard deviation of the forecasts has one advantage over
RMSE as a proxy for inflation uncertainty: it does not contain any ex
post information. RMSEt+1 includes the actual inflation rate from
period t + 1 , pt+1.
28The correlation coefficient between RMSE and the expected infla­
tion rate is 0.559 for the entire sample period and 0.433 for the
period exclusive of the periods of energy shocks. Between RMSE
and the actual rate of inflation, the correlations are 0.826 and 0.658,
respectively.




considerably more variable than the standard devia­
tion over the sample period. The most interesting dif­
ference in the two series, however, is their behavior
during the energy shock periods: the standard devia­
tion remains higher than normal throughout each of
the energy shock periods and does not decline until
the period is over; the RMSE peaks, then declines sub­
stantially while relative energy prices are still rising.
Therefore, these two measures imply different re­
sponses of inflation uncertainty to energy shocks.

INFLATION AND THE VARIABILITY
OF INFLATION FORECASTS
This section provides more detailed evidence on the
effects of inflation and energy shocks on the two mea­
sures of inflation uncertainty discussed above. Table 3
presents results from regressions based on six-month
inflation forecasts. The data used are from the same
sample period shown in the charts.
23

FEDERAL RESERVE BANK OF ST. LOUIS

FEBRUARY 1984

Table 3
Tests for Inflation Uncertainty Using
Livingston Survey Data
(1)
SDt+ 1ll,

(2)
RMSEt + 1

Intercept

0.802
(8.32*)

1.407
(4.32*)

Pt

0.103
(5.40*)

0.267
(3.66*)

P t-i

—

Pt —2

—

0.180
(2.64*)

Pt —3

—

-0 .2 2 3
(-3 .0 6 *)

Sum

—

0.186
(2.69*)

P?+i

—

0.023
(1.84*)

Pf

0.006
(1.42)

0.000
(0.02)

Pta-1

0.003
(0.73)

-0 .0 2 4
(-1 .7 5 *)

Pf 2

0.014
(3.54*)

Sum

0.023
(2.55*)

P

0.389
(3.25*)

0.448
(3.85*)

R2

0.769

0.663

SE

0.30

0.84

DW

2.05

1.91

-0 .0 3 8
(-0 .5 7 )

—
- 0.001
( - 0.01)

t-statistics in parentheses.
‘ Significant at the 5 percent level (one-tailed test).

Inflation's Effect on the Standard
Deviation o f Forecasts
In equation 1, the dependent variable is SDt + ilI„
which is the standard deviation of inflation expecta­
tions for period t +1 as calculated from responses to
the Livingston survey at period t.29 The most recent
six-month rate of inflation known to the forecasters, pt,
has a positive and strongly significant effect on the
standard deviation of the forecasts. Lagged values of
this variable had no significant effect. The value of the

29The variable is written SDt+ 1ll, to indicate that it is based on fore­
casts of period t + 1 inflation given an information set from period t, I,.


24


intercept implies that, in the absence of inflation or
changes in the relative price of energy, the standard
deviation of inflation forecasts would be about a 0.8
annualized percentage rate. The coefficient for p, in­
dicates that for every 1 percentage-point increase in
the annual rate of inflation, the standard deviation
increases by about 0.1 percentage point. Therefore,
with 8 percent inflation, the standard deviation would
be twice as high as with zero inflation.
The three most recent values of the annualized sixmonth change in the relative price of energy (pe) also
have a significant positive impact on this measure of
inflation uncertainty.30 A 1 percent increase in this
variable leads to an increase in the standard deviation
o f 0.023 percentage points over three six-month
periods. In other words, an energy shock affects this
measure of inflation uncertainty for up to 18 months. A
20 percentage-point increase in the relative price of
energy — not uncommon in the last decade — causes
the standard deviation of inflation expectations to in­
crease by about 0.45 percentage points.31

Inflation’s Effect on the
Root-Mean-Squared Error o f Forecasts
Equation 2 presents results using the RMSE of infla­
tion forecasts for period t +1 (RMSEt + i) as the depen­
dent variable.32 The conclusion that inflation exerts a
positive influence on inflation uncertainty is the same
as in equation 1, although the impact occurs over four
six-month periods. The sum of the coefficients of cur­
rent and lagged inflation is positive and significant.
Over 24 months, a 1 percentage-point increase in infla­
tion leads to an increase in RMSE of about 0.19 percent­
age points. Although this is about twice the impact that

30The series for the inflation rate and changes in the relative price of
energy are constructed to include the most recent numbers known
by the forecaster, so monthly data are used. The spring forecaster is
assumed to know the April levels of the CPI and the relative price of
energy, so the six-month rate of change is calculated between
October and April. For the fall forecast, the rate is calculated be­
tween April and October. The denominator in the relative energy
price variable for monthly data is the finished goods component of
the PPI.
31The regressions also were run with a somewhat different dependent
variable, the standard deviation across individuals of the expected
level of the CPI divided by the mean expected level. This is the
coefficient of variation of the CPI level forecasts. The results were
similar to those for the standard deviation of the inflation rate fore­
casts. The coefficient of variation of the inflation rate forecasts is
clearly an inappropriate variable to use, since, as the expected
in flatio n rate approaches zero, the coefficient of variation
approaches infinity_________________________
32RMSE, +1 = y j (SDt+ 1llt)2 + (Pf+ill, - pl+1)2.

FEBRUARY 1984

FEDERAL RESERVE BANK OF ST. LOUIS

inflation had on the standard deviation, the constant
term is nearly twice as high in this equation; thus, the
impacts actually are quite similar. The initial impact of
inflation on RMSE is much greater than it is on the
standard deviation, but this effect is partially offset
after 24 months have passed.
The impact of relative energy price changes is quite
different in this regression than it was in equation 1.
The initial impact on the uncertainty measure is posi­
tive, but the effect is totally offset 12 months later.33
Consequently, if the relative price of energy were to
increase by the same amount each period, it would
cease to have any effect on the RMSE after 12 months.
In contrast, for the standard deviation of expectations
to stabilize, the level rather than the growth rate of the
relative price of energy must stabilize.34
In both equations, the effect of higher inflation on
the measure of inflation uncertainty is positive and
permanent. There is no indication that, over time, fore­
casters come to be just as certain about higher rates of
inflation as they were about lower rates. This evidence
supports the hypothesis that higher inflation leads to
more uncertain inflation.

different inflation expectations model — one incorpo­
rating the effects of changes in the relative price of
energy on expected inflation — led to the opposite
conclusion. On the other hand, there are positive rela­
tionships between the rate of inflation and the stan­
dard deviation and root-mean-squared error of infla­
tion forecasts taken from the Livingston survey. Energy
shocks also affect these two measures of inflation un­
certainty, but in quite different ways.
Because the results of empirical tests based on infla­
tion forecasting models are sensitive to the specifica­
tion of the model, the usefulness of these results is
questionable. Therefore, uncertainty measures based
on the variability of “observed" inflation forecasts or
forecast errors should be given more attention. In this
article, these measures indicate that inflation uncer­
tainty can be reduced if the rate of inflation is reduced.
In light of recent evidence that greater inflation un­
certainty has a detrimental effect on the levels of eco­
nomic activity and unemployment, the reduction of
inflation uncertainty is an important potential benefit
of anti-inflation policies.

CONCLUSION
Researchers have compiled considerable evidence
suggesting that the rate and variability of inflation are
positively related and a lesser amount of evidence link­
ing these variables to inflation uncertainty. This article
has explored the relationship between the rate of infla­
tion and the level of inflation uncertainty in greater
detail, looking also at the impact of energy shocks on
inflation uncertainty.
The empirical results presented here are somewhat
mixed and are sensitive to the method chosen for
measuring inflation uncertainty. On the one hand, a
model of inflation expectations was introduced and
estimated for which the variance o f the estimated infla­
tion forecast errors is related to the rate of inflation. A

REFERENCES
Able, Stephen L. “ Inflation Uncertainty, Investment Spending, and
Fiscal Policy,” Federal Reserve Bank of Kansas City Economic
Review (February 1980), pp. 3-13.
Blejer, Mario I., and Leonardo Liederman. “ On the Real Effects of
Inflation and Relative-Price Variability: Some Empirical Evidence,”
Review of Economics and Statistics (November 1980), pp. 53944.
Bordo, Michael David. "The Effects of Monetary Change on Relative
Commodity Prices and the Role of Long-Term Contracts,” Journal
of Political Economy (December 1980), pp. 1088-1109.
Breusch, T. S., and A. R. Pagan. “A Simple Test for Heteroscedasticity and Random Coefficient Variation,” Econometrica (Septem­
ber 1979), pp. 1287-94.
Carlson, John A. “A Study of Price Forecasts,” Annals of Economic
and Social Measurement (Winter 1977), pp. 27-56.
Carlton, Dennis W. “ The Disruptive Effect of Inflation on the Orga­
nization of Markets," National Bureau of Economic Research, Con­
ference Paper No. 104 (1981).

“ The inclusion of the variable pf+1 in the regression is not meant to
imply that forecasters know the value of this variable, only that it
affects RMSEt+1.
34The regressions in table 3 also were run with several other indepen­
dent variables, none of which was statistically significant at the 5
percent level. These included current and lagged values of the
absolute value of unanticipated inflation (based on the survey mean
expectation), a dummy variable for the period of wage and price
controls, and a time trend. In regressions excluding the relative price
of energy, the estimated effects of inflation on the uncertainty mea­
sures were somewhat larger.




Cukierman, Alex, and Paul Wachtel. “ Differential Inflationary Ex­
pectations and the Variability of the Rate of Inflation: Theory and
Evidence,” American Economic Review (September 1979), pp.
595-609.
_________ “ Relative Price Variability and Nonuniform Inflationary
Expectations," Journal of Political Economy (February 1982), pp.
146-57.
Engle, Robert F. “Autoregressive Conditional Heteroscedasticity
With Estimates of the Variance of United Kingdom Inflation,” Econ­
ometrica (July 1982), pp. 987-1007.

25

FEDERAL RESERVE BANK OF ST. LOUIS

FEBRUARY 1984

_________ "Estimates of the Variance of U.S. Inflation Based upon
the ARCH Model,” Journal o f Money, Credit and Banking (August
1983), pp. 286-301.

Klein, Benjamin. “The Measurement of Long- and Short-Term Price
Uncertainty: A Moving Regression Time Series Analysis,” Eco­
nomic Inquiry (July 1978), pp. 438-52.

Evans, Paul. “ Price-Level Instability and Output in the U.S.," Eco­
nomic Inquiry (April 1983), pp. 172-87.

Levi, Maurice D., and John H. Makin. “ Inflation Uncertainty and the
Phillips Curve: Some Empirical Evidence,” American Economic
Review (December 1980), pp. 1022-27.

Fischer, Stanley. “Towards an Understanding of the Costs of Infla­
tion: II,” in Karl Brunner and Allan H. Meltzer, eds., The Costs and
Consequences o f Inflation, Carnegie-Rochester Conference
Series on Public Policy (Autumn 1981), pp. 5-41.
Foster, Edward. "The Variability of Inflation,” Review of Economics
and Statistics (August 1978), pp. 346-50.
Friedman, Milton. “ Nobel Lecture: Inflation and Unemployment,”
Journal o f Political Economy (June 1977), pp. 451-72.
Frohman, Deborah A., Leroy O. Laney and Thomas D. Willett. "Un­
certainty Costs of High Inflation,” Voice of the Federal Resen/e
Bank of Dallas (July 1981), pp. 1-9.
Goldfeld, Stephen M., and Richard E. Quandt. “ Some Tests for
Homoscedasticity,” Journal of the American Statistical Associa­
tion (June 1965), pp. 539-47.
Gordon, Robert J. “ Steady Anticipated Inflation: Mirage or Oasis?”
Brookings Papers on Economic Activity (2:1971), pp. 499-510.

Logue, Dennis E., and Thomas D. Willett. “ A Note on the Relation
between the Rate and Variability of Inflation,” Economica (May
1976), pp. 151-58.
Lucas, Robert E., Jr. “ Some International Evidence on OutputInflation Tradeoffs,” American Economic Review (June 1973), pp.
326-34.
Meyer, Laurence H., and Robert H. Rasche. “ On the Costs and
Benefits of Anti-Inflation Policies,” this Review, (February 1980),
pp. 3-14.
Mullineaux, Donald J. “ Unemployment, Industrial Production, and
Inflation Uncertainty in the United States,” Review o f Economics
and Statistics (May 1980), pp. 163-69.
Okun, Arthur M. “The Mirage of Steady Inflation,” Brookings Papers
on Economic Activity (2:1971), pp. 485-98.

Gray, Jo Anna. “ On Indexation and Contract Length,” Journal of
Political Economy (February 1978), pp. 1-18.

Pagan, A. R., A. D. Hall and P. K. Trivedi. “ Assessing the Variability
of Inflation,” Review o f Economic Studies (October 1983), pp.
585-96.

Hafer, R. W., and Gail Heyne-Hafer. “ The Relationship between
Inflation and Its Variability: International Evidence from the 1970s,”
Journal o f Macroeconomics (Fall 1981), pp. 571-77.

Sheshinski, Eytan and Yoram Weiss. “ Inflation and Costs of Price
Adjustment,” Review o f Economic Studies (June 1977), pp. 287303.

Hayek, F. A. “The Use of Knowledge in Society,” American Eco­
nomic Review (September 1945), pp. 519-30.

Tatom, John A. “ Energy Prices and Short-Run Economic Perfor­
mance,” this Review (January 1981), pp. 3-17.

Jaffee, Dwight, and Ephraim Kleiman. “The Welfare Implications of
Uneven Inflation,” in Erik Lundberg, ed., Inflation Theory and Anti­
inflation Policy (Westview Press, 1977), pp. 285-307.

Taylor, John B. “ On the Relation Between the Variability of Inflation
and the Average Inflation Rate,” in The Costs and Consequences
of Inflation, pp. 57-85. See Fischer.


26


Calculating the Adjusted
Monetary Base under
Contemporaneous Reserve
Requirements
R. Alton Gilbert

T

M . HE adjusted monetary base is designed to be a
single measure of all Federal Reserve actions, including
changes in reserve requirements, that influence the
money stock. It is equal to the source base plus a
reserve adjustment magnitude (RAM) that accounts for
changes in reserve requirements by the Federal
Reserve.1
The adoption of contemporaneous reserve require­
ments (CRR), which became effective in February of
this year, did not affect the reserve requirement ratios
applicable to any group of deposit liabilities. It did,
however, alter the relationships between deposit lia­
bilities and the periods over which depository institu­
tions are required to hold reserves against them.2 This
article describes how the adoption of CRR has mod­
ified the calculation of RAM and, hence, the adjusted
monetary base.

Ft. Alton Gilbert is a research officer at the Federal Reserve Bank of
St. Louis. John G. Schulte provided research assistance.
1The following articles describe and explain the adjusted monetary
base in greater detail: Gilbert (1980), Tatom (1980) and Gilbert
(1983).
2For a general description of the new system of contemporaneous
reserve requirements, see Gilbert and Trebing (1982).




THE CALCULATION OF RAM AND
THE ADJUSTED MONETARY BASE
RAM is calculated as the difference between the
reserves that would have been required (given current
deposit liabilities) if the base period’s reserve require­
ments were in effect and the reserves that are actually
required (given current requirements). Adding RAM to
the source base produces an adjusted monetary base
series that shows what the source base would have had
to be, given the deposit liabilities for each period, if the
reserve requirement ratios had always been those of
the base period.3Thus, this procedure converts reserve
requirement changes into equivalent changes in the
source base, holding the reserve requirements con­
stant.
The base period for calculating RAM is January 1976
through August 1980. Base period reserve require­
ments are the average reserve requirements over that
period for two categories o f deposit liabilities: check-

3The source base equals the reserve balances of depository institu­
tions with Federal Reserve Banks, which excludes their required
clearing balances, plus total currency in circulation, whether held by
depository institutions or the public. It is derived from the combined
balance sheets of the Federal Reserve Banks and the U.S. Treasury.

27

FEDERAL RESERVE BANK OF ST. LOUIS

able deposits and total time and savings deposits. For
member banks, the average reserve requirement was
12.664 percent on checkable deposits and 3.1964 per­
cent on total time and savings deposits.4 For nonmember institutions, base period reserve requirements
were zero, since they were not subject to reserve re­
quirements of the Federal Reserve in the base period.
Thus, RAM is calculated as the current checkable de­
posits of member banks multiplied by 0.12664, plus the
current total time and savings deposits of member
banks multiplied by 0.031964, minus the current re­
quired reserves of all depository institutions.

CALCULATION OF RAM UNDER THE
PRIOR SYSTEM OF LAGGED
RESERVE REQUIREMENTS
The specific data on deposit liabilities and required
reserves used to calculate RAM necessarily reflect the
system of reserve accounting in effect. Under lagged
reserve requirements (LRR), the average reserves of a
depository institution over the seven-day reserve
maintenance period ending each Wednesday must
equal or exceed its required reserves. The required
reserves for a depository institution in each mainte­
nance week were based on its average deposit liabili­
ties over the seven-day period ending 14 days before
the end of the current maintenance week. Thus, in
calculating RAM under LRR, data on the required re­
serves of depository institutions for each maintenance
week were matched with the deposit liabilities of mem­
ber banks of two weeks earlier.

THE CALCULATION OF RAM
UNDER CRR
Under CRR, the reserve maintenance periods, during
which average reserves must equal or exceed required
reserves, have been lengthened to two-week periods
that end every other Wednesday. Required reserves on
checkable deposits for the current two-week mainte­
nance period are based on daily average checkable
deposits for the 14-day period ending two days before
the end of the current maintenance period. Required
reserves on time and savings deposits are based on
daily average deposits over a 14-day period ending 30
days before the end o f the current maintenance
period.5Table 1 presents the timing of reserve account­
ing for maintenance periods in 1984.
4Gilbert (1980).
5The timing of resen/e accounting for vault cash is also altered under
CRR. Under the previous LRR system, the average vault cash held


28


FEBRUARY 1984

Calculating RAM: An Example
The calculation of RAM must be adjusted to the new
timing of reserve accounting under CRR. The new pro­
cedure for calculating RAM is illustrated for the first
maintenance period under CRR: February 2 through
February 15,1984. Required reserves for that period are
based on daily average checkable deposits over the
period January 31 through February 13, and daily aver­
age time and savings deposits from January 3 through
January 16. For this maintenance period, RAM is calcu­
lated as the average of total time and savings deposits
of member banks over January 3-16 multiplied by
0.031964, plus average checkable deposits of member
banks over January 31-February 13 multiplied by
0.12664, minus the required reserves of all depository
institutions for the period February 2-15. In calculating
RAM for the next maintenance period, February 16
through February 29, observations for each category of
deposit liabilities and required reserves apply to
periods brought forward 14 days.

Calculating a Weekly Adjusted
Monetary Base with a Biweekly RAM
Given the timing of reserve accounting under CRR,
RAM is calculated for two-week periods ending every
other Wednesday. This Bank’s adjusted monetary base
will still be published on a weekly basis for seven days
ending each Wednesday. The measurement of the
source base is not affected by the change to CRR, since
it is derived from the balance sheets of Federal Reserve
Banks and the U.S. Treasury. The adjusted monetary
base is calculated for each week by adding the value of
RAM that covers that week to the value of the source
base. Because of the timing of reserve accounting
under CRR, each value for RAM is used in calculating
two weekly observations for the adjusted monetary
base.

THE MOST RECENT VALUES OF RAM
MUST BE ESTIMATED
The change from LRR to CRR changes the availability
of data necessary to calculate RAM. Previously, under
LRR, data on required reserves and the lagged values of

by a depository institution over the seven-day period ending 14 days
before the end of the current maintenance period counted as part of
reserves in the current maintenance period. Under CRR, the average
vault cash held over a 14-day period ending 30 days before the end of
the current maintenance period counts as reserves in the current
maintenance period. The timing of reserve accounting for vault cash
is not discussed in the text because it does not influence calculation
of the adjusted monetary base.

Table 1
1984 Reserve Periods
Computation and Maintenance Dates Under Contemporaneous Reserve
Requirements (weekly reporters)
Reserve
period
number

Computation period
Maintena nee period

Non transaction accounts
and vault cash

Transaction accounts

1

21 2/84 -

2/15/84

1/ 3/84 - 1/ 9/84
1/10/84 — 1/16/84

1/31/84 — 2/ 6/84
2/ 7 /8 4 - 2/13/84

2

2/16/84 -

2/29/84

1/17/84 — 1/23/84
1 /2 4 /8 4 - 1/30/84

2 /1 4 /8 4 2 /2 1 /8 4 -

2/20/84
2/27/84

3

3/ 1/84 -

3/14/84

1/31/84 — 21 6/84
2/ 7/84 - 2/13/84

2/28/84 3/ 6/84 -

3/ 5/84
3/12/84

4

3/15/84 -

3/28/84

2 /1 4 /8 4 2 /2 1 /8 4 -

2/20/84
2/27/84

3 /1 3 /8 4 3/20/84 -

3/19/84
3/26/84

5

3/29/84 -

4/11/84

2/28/84 3/ 6/84 -

3/ 5/84
3/12/84

3/27/84 4/ 3/84 -

4/ 2/84
4/ 9/84

6

4/12/84 -

4/25/84

3 /1 3 /8 4 3/20/84 -

3/19/84
3/26/84

4 /1 0 /8 4 4 /1 7 /8 4 -

4/16/84
4/23/84

7

4/26/84 -

5/ 9/84

3/27/84 4/ 3/84 -

4/ 2/84
4/ 9/84

4/24/84 - 4/30/84
5/ 1/84 — 5/ 7/84

8

5/10/84 -

5/23/84

4 /1 0 /8 4 4 /1 7 /8 4 -

4/16/84
4/23/84

5/ 8/84 5 /1 5 /8 4 -

5/14/84
5/21/84

9

5/24/84 -

6/ 6/84

4/24/84 5/ 1 /8 4 -

4/30/84
5/ 7/84

5/22/84 5/29/84 -

5/28/84
6/ 4/84

10

6/ 7/84 -

6/20/84

5/ 8/84 5 /1 5 /8 4 -

5/14/84
5/21/84

6/ 5/84 6 /1 2 /8 4 -

6/11/84
6/18/84

11

6/21/84 -

71 4/84

5/22/84 5/29/84 -

5/28/84
6/ 4/84

6 /1 9 /8 4 6/26/84 -

6/25/84
7/ 2/84

12

7/ 5/84 -

7/18/84

6/ 5/84 - 6/11/84
6 /1 2 /8 4 - 6/18/84

7/ 3/84 7 /1 0 /8 4 -

7/ 9/84
7/16/84

13

7/19/84 -

8/ 1/84

6 /1 9 /8 4 6/26/84 -

6/25/84
7/ 2/84

7 /1 7 /8 4 7/24/84 -

7/23/84
7/30/84

14

8/ 2/84 -

8/15/84

7/ 3/84 7 /1 0 /8 4 -

7/ 9/84
7/16/84

7 /3 1 /8 4 8/ 7 /8 4 -

8/ 6/84
8/13/84

15

8/16/84 -

8/29/84

7/17/84 7/24/84 -

7/23/84
7/30/84

8/14/84 8 /2 1 /8 4 -

8/20/84
8/27/84

16

8/30/84 -

9/12/84

7/31/84 8/ 7/84 -

8/ 6/84
8/13/84

8/28/84 9/ 4/84 -

9/ 3/84
9/10/84

17

9/13/84 -

9/26/84

8 /1 4 /8 4 8 /2 1 /8 4 -

8/20/84
8/27/84

9 /1 1 /8 4 9 /1 8 /8 4 -

9/17/84
9/24/84

18

9/27/84 - 10/10/84

8/28/84 9/ 4/84 -

9/ 3/84
9/10/84

9/25/84 - 10/ 1/84
10/ 2 /8 4 - 10/ 8/84

19

10/11/84 - 10/24/84

9 /1 1 /8 4 9/18/84 -

9/17/84
9/24/84

10/ 9 /8 4 - 10/15/84
10/16/84 —10/22/84

20

10/25/84 - 11/ 7/84

9/25/84 - 10/ 1/84
10/ 2/84 - 10/ 8/84

10 /23/84 - 10/29/84
10 /3 0 /8 4 - 11/ 5/84

21

11/ 8/84 -1 1 /2 1 /8 4

10/ 9/84 - 10/15/84
10/16/84 —10/22/84

1 1 / 6 /8 4 - 11/12/84
11/13/84 —11/19/84

22

11/22/84 - 12/ 5/84

10/23/84 - 10/29/84
10/30/84 - 11/ 5/84

11 /2 0 /8 4 - 11/26/84
11 /2 7 /8 4 - 12/ 3/84

23

12/ 6/84 - 12/19/84

11/ 6/84 - 11/12/84
11 /13/84 - 11/19/84

12/ 4/84 - 12/10/84
12 /1 1 /8 4 - 12/17/84

24

12/20/84 - 1/ 2/85

11 /20/84 - 11/26/84
11 /2 7 /8 4 - 12/ 3/84

12 /18/84 - 12/24/84
12 /2 5 /8 4 - 12/31/84




FEDERAL RESERVE BANK OF ST. LOUIS

deposit liabilities used in calculating RAM were avail­
able before the end of a maintenance period. Thus,
RAM could be calculated before data were available on
the source base. Under CRR, however, it will not be
possible to calculate RAM until more than a week after
the end of each maintenance period.
The difference in the timing of the data necessaiy to
calculate RAM under LRR and CRR, and the problem
CRR creates for calculating a weekly adjusted mone­
tary base, are illustrated for the last maintenance
period under LRR and the first two maintenance
periods under CRR. Required reserves for the week
ending February 1 were based on average deposit lia­
bilities for the week ending January 18. The data on
deposit liabilities and required reserves necessaiy for
calculating RAM for the week ending February 1 were
available by February 1. Because the source base for the
seven days ending February 1 was available by Febru­
ary 3, the exact adjusted monetary base for the week
ending February 1 was published in this Bank’s U.S.
Financial Data on February 3, 1984.
The data necessary for calculating the value of RAM
for the next maintenance period, the 14 days ending
February 15, were available on February 23. If publica­
tion of the adjusted monetary base was delayed until
all data necessary for calculating RAM were available,
the adjusted monetary base for the weeks ending
February 8 and 15 would not be published until Febru­
ary 23. Such delays can be avoided only by estimating
RAM for the most recent maintenance period. This is
done for the adjusted monetary base series published
in this Bank’s U.S. Financial Data release, with pre­
liminary data published for the most recent one or two
weeks (see table 2).
If there is no change in reserve requirements, the
estimate of RAM used to obtain the preliminary weekly
adjusted monetary base is the value of RAM for the
most recent maintenance period. If a change in reserve
requirements becomes effective during the current
maintenance period, however, the estimated RAM for
this period equals its lagged value plus an estimate of
the effect of the change in reserve requirements on
required reserves.
A change in reserve requirements became effective
in the maintenance period covering the two weeks
ending February 15, the last phased reduction in re­
serve requirements of member banks specified in the
Monetary Control Act of 1980. The prior phased reduc­
tions in member bank reserve requirements reduced
total required reserves by about $2 billion (see figure 1).
Digitized for 30
FRASER


FEBRUARY 1984

Table 2
Weekly Schedule for Publishing
Preliminary Data on the Adjusted
Monetary Base in U.S. Financial Data
Preliminary data
Publication date

for w eek(s) ending

February 10, 1984
February 16
February 23
March 1
March 8
March 15
March 22
March 29
April 5

February 8
February 8, 15
February 22
February 22, 29
March 7
March 7, 14
March 21
March 21, 28
April 4

Thus, the estimated value of RAM used in calculating
preliminary values of the adjusted monetary base for
the weeks ending February 8 and February 15 was the
value of RAM calculated for the week ending February 1
plus $2 billion. A preliminary value of the adjusted
monetary base for the week ending February 8, pub­
lished on February 10, was calculated as the source
base for the week ending February 8 plus that esti­
mated value of RAM.
Note that the deposit data for the week ending Janu­
ary 18 were used to derive this value of RAM. This value
is subject to further revisions due to revisions of both
the deposit data for the week ending January 18 and
the required reserves for the week ending February 1.
The preliminary number for the adjusted monetary
base for the following week, the seven days ending
February 15, which was published on February 16,
equals the source base for the week ending February 15
plus the estimated value of RAM.
By February 23, the data on checkable deposits and
required reserves were available to calculate RAM for
the two weeks ending February 15. At that time, the
preliminary adjusted monetary base data for the weeks
ending February 8 and 15 were revised to incorporate
the new value for RAM. Moreover, this latest value for
RAM was used in calculating the preliminary value of
the adjusted monetary base for the week ending Febru­
ary 22 (published on February 23), and the preliminary
value for the week ending February 29 (published on
March 1). When using this approach to calculate the
weekly adjusted monetary base series, either one or
two of the most recent weekly observations are pre­
liminary. RAM for the current and prior weeks remains

FEDERAL RESERVE BANK OF ST. LOUIS

FEBRUARY 1984

C h art 1

T w o -W e e k A v e r a g e s of R A M

1980

1981

1982

1983

1984

E n d i n g d a t e of 2 - w e e k p e r i o d
N O T E : N u m b e r s i n d i c a t e e f f e c t iv e d a t e s

o f m a j o r c h a n g e s in r e s e r v e r e q u ir e m e n t s .

1. The 8 percentage point marginal reserve requirement was raised to 10 percent. In addition, the base upon which the marginal reserve
requirement is calculated was reduced. This action increased required reserves about $1.7 billion.
2. The marginal reserve requirement was reduced from 10 to 5 percentage points and the base upon which the marginal reserve
requirement is calculated was raised. This action reduced required reserves about $980 million.
3. The 5 percent marginal reserve requirement on managed liabilities and the 2 percent supplementary reserve requirement against large
time deposits were removed. These actions reduced required reserves about $3.2 billion.
4. Required reserves of member banks and Edge Act corporations were reduced about $4.3 billion and required reserves of other
depository institutions were increased about $1.4 billion due to the implementation of the Monetary Control Act of 1980.
5. In conjunction with the transitional phase-in program underthe Monetary Control Act, required reserves of member banks were reduced
$2.0 billion, and required reserves of other depository institutions were increased $0.9 billion.
6. In conjunction with the transitional phase-in program under the Monetary Control Act, required reserves of member banks decreased by
$2.0 billion.
7. In conjunction with the transitional phase-in program under the Monetary Control Act, required reserves of member banks were reduced
$2.1 billion, and required reserves of other depository institutions were increased $0.9 billion.
8. In accordance with provisions of the Depository Institutions Act of 1982 that exempted the first $2.1 million of reservable liabilities at all
depository institutions from reserve requirements, required reserves were reduced by an estimated $800 million.
9. In conjunction with the transitional phase-in program underthe Monetary Control Act, required reserves of member banks were reduced
by approximately $1.9 billion. Also, the reserves released by the growth of money market deposit accounts (available after midDecember 1982) produced an upward drift in RAM during 1983, especially during the first half of the year.
10. In conjunction with the transitional phase-in program under the Monetary Control Act, required reserves of member banks were reduced
$2.0 billion, and required reserves of other depository institutions were increased $0.9 billion.




31

FEDERAL RESERVE BANK OF ST. LOUIS

FEBRUARY 1984

subject to revisions due to revisions in the data on
deposit liabilities and required reserves.

quirements, covering the week ending February 22,
would be published on February 23.

ERRORS IN ESTIMATING RAM ARE
LIKELY TO BE SMALL

THE PUBLICATION OF ADJUSTED
RESERVES UNDER CRR

Errors in estimating RAM with its lagged value gener­
ally will be small relative to the size of the adjusted
monetary base. The size of the errors using this
approach are simulated for the period from 1980
through early 1984, using the average RAM calculated
for each two-week period over these four years. For
periods when no changes in reserve requirements oc­
curred, the error in using the value for the prior twoweek period to estimate RAM for the current period
was less than $100 million in half of the periods, and
less than or equal to $200 million in about 84 percent of
the periods.

Adjusted reserves will not be published weekly
under CRR. Adjusted reserves are calculated by sub­
tracting seasonally adjusted currency in the hands of
the public from the adjusted monetary base, seasonal­
ly adjusted. Until February 1984, weekly observations
for currency in the hands of the public covered sevenday periods ending each Wednesday, the same periods
that applied to weekly observations of the adjusted
monetary base. Weekly currency data now cover the
seven days ending each Monday, matching the timing
of weekly average checkable deposits under CRR. With
this change in timing, it would be inappropriate to
subtract weekly average currency from weekly average
adjusted monetary base to obtain a weekly adjusted
reserves series; the periods for currency and the ad­
justed monetary base do not match up. The change
described above in the timing of data on currency in
the hands of the public probably has little effect on
data for monthly average currency. This Bank, there­
fore, will continue to publish adjusted reserves on a
monthly average basis.

As chart 1 indicates, large changes in RAM typically
have occurred only when there have been major
changes in reserve requirements. Changes in RAM,
other than those resulting from the 10 major changes
in reserve requirements identified in chart 1, have been
relatively small.
Errors in estimating the effects of changes in reserve
requirements should not generally be large. The Feder­
al Reserve is generally able to estimate the effects of a
change in reserve requirements on required reserves
very accurately. Furthermore, most changes in reserve
requirements have applied to time and savings
deposits.6 Since under CRR, the time and savings de­
posits subject to reserve requirements in a reserve
maintenance period are lagged four weeks, data
should be available to indicate the effects of those
changes on required reserves when the preliminary
data on the adjusted monetary base are published.
To illustrate the timing, suppose that reserve re­
quirement ratios on time and savings deposits were
raised, effective in the maintenance period covering
the two weeks ending February 29. Required reserves
for that maintenance period are based on average time
and savings deposits over the two weeks ending Janu­
ary 30. The first weekly observation of the adjusted
monetary base affected by that change in reserve re­

CONCLUSIONS
The timing of data used in calculating the adjusted
monetary base has been changed to reflect the timing
of reserve accounting under the new system of con­
temporaneous reserve requirements. Observations for
the adjusted monetary base for the most recent one or
two weeks will be preliminary, because the most recent
values of the reserve adjustment magnitude will be
estimated. The adjusted reserves series will no longer
be published on a weekly basis, due to a change in the
days covered by weekly average data on currency in the
hands of the public. This Bank will continue to publish
the adjusted reserves series on a monthly average
basis.

REFERENCES
Gilbert, R. Alton. “ Revision of the St. Louis Federal Reserve Ad­
justed Monetary Base,” this Review (December 1980), pp. 3-10.
_________ “Two Measures of Reserves: Why Are They Different?”
this Review (June/July 1983), pp. 16-25.

6Changes in reserve requirements since November 1980 are dictated
by the Monetary Control Act of 1980 and the Depository Institutions
Act of 1982; they apply to both checkable deposits and to time and
savings deposits. From 1960 until November 1980, the Board of
Governors changed reserve requirements 35 times. Only 11 of those
changes involved demand deposits.


32


Gilbert, R. Alton, and Michael E. Trebing. “The New System of
Contemporaneous Reserve Requirements,” this Review (Decem­
ber 1982), pp. 3-7.
Tatom, John A. “ Issues in Measuring an Adjusted Monetary Base,”
this Review (December 1980), pp. 11-29.

Agriculture

—

An Eighth District Perspective

Banking & Finance —An Eighth District Perspective
Business

—

An Eighth District Perspective

The Federal Reserve Bank of St. Louis is publishing a new package of
publications that analyze the effect of current economic trends on the
Eighth Federal Reserve District. Single subscriptions to the new re­
gional package — which includes quarterly reports on agriculture,
banking and finance, and business — are offered to the public free of
charge. To subscribe, please write: Research and Public Information,
Federal Reserve Bank of St. Louis, P.O. Box 442, St. Louis, MO 63166.