Full text of Economic Review (Federal Reserve Bank of Cleveland) : 1995 Quarter 1, Vol. 31, No. 1

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

Vol. 31, No. 1

ECONOMIC REVIEW
1995 Quarter 1
Restoring Generational
Balance in U.S. Fiscal Policy:
What Will It Take?

v /2

by Alan J. Auerbach,
Jagadeesh Gokhale,
and Laurence J. Kotlikoff

Vagueness, Credibility, and
Government Policy

13

by Joseph G. Haubrich

Federal Funds Futures as an
Indicator of Future Monetary
Policy: A Primer
by John B. Carlson,
Jean M. Mclntire,
and James B. Thomson




FEDERAL RESERVE BANK
OF CLEVELAND

20

■
E C O N O M I C

R E V I E W

1995 Quarter 1
Vol. 31, No. 1

Restoring Generational
Balance in U.S. Fiscal Policy:
What Will It Take?

2

by Alan J. Auerbach, Jagadeesh Gokhale,
and Laurence J. Kotlikoff
What are the magnitudes of tax increases, transfer cuts, or reductions in
government purchases required to restore a generationally balanced U.S.
fiscal policy? Under the authors' conservative baseline of updated genera­
tional accounts, income taxes would have to be raised permanently by 43
percent, federal transfers cut by 33 percent, or government purchases low­
ered by 32 percent beginning in 1996. The required policy changes will be
larger if their implementation is postponed. The authors also find that the
outlay reductions in nondefense and non-Social Security spending that
Congress recently considered would still leave an unsustainably large im­
balance in the generational stance of U.S. fiscal policy.

Vagueness, Credibility, and
Government Policy

13

Economic Review is published
quarterly by the Research Depart­
ment of the Federal Reserve Bank of
Cleveland. Copies of the Review are
available through our Corporate
Communications & Community
Affairs Department. Call 1-800-5433489 (OH, PA, WV) or 216-5792001, then immediately key in 1-5-3
on your touch-tone phone to reach
the publication request option. If you
prefer to fax your order, the number
is 216-579-2477.

Coordinating Economist:
Jagadeesh Gokhale
Advisory Board:
Charles T. Carlstrom
Joseph G. Haubrich
Peter Rupert

by Joseph G. Haubrich
This article examines the economic reasons why it may be in a government
agency’s — and society's— best interest to be vague about policy objectives.
The author uses the recently developed concept of "cheap talk" to explain that
when an agency faces a trade-off between precise and credible announce­
ments, its best move may be to provide truthful but limited information.

Federal Funds Futures as an
Indicator of Future Monetary
Policy: A Primer

20

by John B. Carlson, Jean M. Mclntire,
and James B. Thomson
Unlike most futures contracts, which are drawn on commodities or financial
instruments whose price or yield is determined in competitive markets, the
federal funds futures rate is essentially determined by a deliberative deci­
sion of the Federal Open Market Committee (FOMC). As such, the fed funds
futures market is a place where one can place a bet as to what future mone­
tary policy will be. The FOMC can thus assess in fairly precise terms what
markets expect it to do. In this paper, the authors examine the predictive
accuracy of the fed funds futures market and consider some policy implica­
tions. They find that accuracy clearly improves in the two-month period
leading up to the contract's expiration and that the largest prediction errors
occur around policy turning points.




Editors: Tess Ferg
Robin Ratliff
Design: Michael Galka
Typography: Liz Hanna

Opinions stated in Economic Re­
view are those of the authors and
not necessarily those of the Fed­
eral Reserve Bank of Cleveland or
of the Board of Governors of the
Federal Reserve System.

Material may be reprinted pro­
vided that the source is credited.
Please send copies of reprinted
material to the editors.

ISSN 0013-0281

2

Restoring Generational Balance
in U.S. Fiscal Policy:
What Will It Take?
by Alan J. Auerbach,
Jagadeesh Gokhale,
and Laurence J. Kotlikoff

Introduction
Generational accounting is a relatively new
method o f reorganizing the governm ent’s
budget data to understand how the burden o f
paying for governm ent spending on goods and
services is distributed among living and future
generations .1 To study this distribution, genera­
tional accounting estimates lifetime net tax
rates facing different generations under current
policies .2 For a given generation, the lifetime
net tax rate is its per capita lifetime net tax bur­
den as a share o f the present value o f its per
capita lifetime labor income.
The lifetime net tax burden, in turn, is the
present value o f per capita taxes net o f trans­
fers that members o f a generation pay over
their lifetimes, evaluated as o f their year of
birth. For generations currently alive, the life­
time net tax burden includes net taxes they

■ 1 The technique of generational accounting was developed in
Auerbach, Gokhale, and Kotlikoff (1991) and in Kotlikoff (1992). See also
Auerbach, Gokhale, and Kotlikoff (1994). Unless stated otherwise, spend­
ing in this paper refers to government purchases of goods and services.

■ 2 A generation is defined as individuals of a particular sex born in

the same year.


Alan J. Auerbach is a profes­
sor of economics at the
University of California,
Berkeley, and an associate of
the National Bureau of Eco­
nomic Research; Jagadeesh
Gokhale is an economic ad­
visor at the Federal Reserve
Bank of Cleveland; and
Laurence J. Kotlikoff is a pro­
fessor of economics at Bos­
ton University and an
associate of the National

Bureau of Economic Re­
search. The authors thank
the Office of Management
and Budget for providing
critical data on FY1996
budget projections and the
Social Security Administra­
tion for providing popula­
tion projections. They also
thank Robert Anderson,
Darrel Cohen, Robert
Kilpatrick, and Patrick Locke
for helpful comments.

have paid in the past and those they may e x ­
pect to pay in the future. Similar remarks apply
to the calculation o f the present value of a g e n ­
eration’s per capita lifetime labor incom e.
In contrast to the three previous years, a gen­
erational accounting analysis of U.S. fiscal policy
was not included in the Budget of the United
States for fiscal year 1996.3 This paper presents
such an analysis. It reports updated lifetime net
tax rates using the latest long-range tax and ex­
penditure projections made by the Office o f M an­
agement and Budget (OM B ).4
Earlier presentations o f lifetime net tax rates
indicated that current U.S. fiscal policy contains
a large generational im balance — a result that
this update confirms. If the current fiscal treat­
ment o f living (including new born) gen eration s
continues throughout their lifetimes, the life­
time net tax rate on those born in 1993 would
be about 34 percent, while future generations
■ 3 The last generational accounting presentation in the U.S. Budget
appeared in Office of Management and Budget (1994), chapter 3.

■ 4 These projections are an extension of the OMB's 1994 Mid-Session
Review baseline projection and incorporate, among other things, long­
term demographic and fiscal projections of the Social Security Adminis­
tration and the Health Care Financing Administration

3

would face an average rate o f 84 percent."’
That is, under current policies, future generations
would bear a fiscal burden two-and-a-half times
as large, on average, than that on the current
newborn generation. Further, a sizable fiscal im­
balance remains despite incorporating optimistic
assumptions about the path o f future federal pur­
chases and health care outlays in the calculations.
Such large projected fiscal burdens on future gen­
erations imply that current fiscal policy is "unsus­
tainable"— a conclusion that is robust to alterna­
tive assumptions about future productivity growth
and interest rates.
This method of calculating the imbalance in
U.S. fiscal policy has been criticized on several
grounds. O ne objection focuses on the assump­
tion that living generations will continue to be
treated per current fiscal policy throughout their
lifetimes, while the tax treatment o f those bom in
the future w ill differ. To some, this assumption
seems to imply that the incidence of future pol­
icy changes to correct the imbalance would fall
exclusively on future generations. They suggest
that the calculations be altered to include the im­
pact o f future policy changes on the lifetime net
tax rates of living generations, since this will nor­
mally be the case. Then, they contend, lifetime
net tax rates on future generations would decline
from the high levels suggested in earlier genera­
tional accounting presentations to more plausible
and acceptable levels, and most o f the dramatic
conclusions drawn by generational accounting
would disappear.6
The assumption o f unchanged tax treatment
o f living generations was only a heuristic and
was not intended to suggest that future policy
changes will apply only to future generations.
Nevertheless, this paper responds to the criti­
cism directly by posing a question: What are
the magnitudes o f tax increases, transfer cuts,
or spending reductions necessary to equalize
the lifetime net tax rates o f current newborn
and future generations — that is, to restore a
generationally balanced fiscal policy?
The experiments assume that policy changes,
w hen introduced, will apply to all generations
alive then and in every year thereafter. Hence,
the new policies will affect the lifetime net tax
rates o f most generations alive in 1993, our base
year. The tax, transfer, and spending policy e x ­
periments are conducted for a set of baseline
projections o f future revenues and outlays as
well as for alternative assumptions about the
growth paths o f federal purchases and health
care outlays. In each case, we report the changes
in taxes, transfers, or purchases needed to equal­
ize lifetime net tax rates of future and current




newborn generations. We also present the val­
ues o f the equalized lifetime net tax rates.
The calculated tax hikes, transfer reductions,
or spending cuts required for achieving a genera­
tionally balanced fiscal policy are immense —
much larger than those recently considered by
Congress as part o f the debate to balance the
budget by the year 2002. Thus, achieving a bal­
anced budget by that date would not place U.S.
fiscal policy on a sustainable path unless budget
balance were preserved thereafter. The reason is
that under current projections, growth in outlays
after 2002 will far outstrip growth in revenues,
and maintaining a balanced budget beyond 2002
is likely to require cuts in addition to those
needed just to balance the budget by that year.
The policy changes required to equalize life­
time net tax rates o f newborn and future gen­
erations can be viewed as alternative measures
o f the im balance in current U.S. fiscal policy.
Unlike the critics’ conjecture, these measures
also suggest that a substantial im balance is em ­
bedded in current U.S. fiscal policy.

I. How Are
Generational
Accounts and
Lifetime Net Tax
Rates Computed?7
Generational accounts refer to the present val­
ue of taxes net o f transfers that a m em ber o f
each generation may expect to pay on average
now and in the future. Thus, generational ac­
counts reveal the prospective net tax burdens
on different generations. In contrast, lifetime
generational accounts include net taxes paid in
the past and refer to the present value o f net
taxes as o f the generation’s year o f birth.
■ 5 The estimates presented in Office of Management and Budget
(1994), chapter 3, were 36.3 percent on current (1992) newborns and 82
percent on future generations. The differences in the estimates reported
here stem from technical improvements incorporated in the calculations
as well as from the use of previously unavailable long-range budgetary
projections provided by the 0MB. The lifetime net tax rates reported are
averaged across male and female generations.

■

6 For examples of such criticism, see Eisner (1994) and Haveman
(1994). Another criticism, not dealt with here, stems from the Ricardian
equivalence proposition, which states that current generations, perceiving
the tax increases on future generations implicit in the deficit financing of cur­
rent government spending, will respond by increasing their saving and be­
quests. However, formal tests fail to detect the altruistic behavior required for
Ricardian equivalence. See Altonji, Hayashi, and Kotlikoff (1992).

■ 7 This section presents a brief discussion of the method of genera­
tional accounting. For more detailed treatments, see Auerbach, Gokhale, and
Kotlikoff (1991) and Kotlikoff (1992). See also Office of Management and
Budget (1994), chapter 3.

A. Living
Generations
Lifetime generational accounts are used here to
com pute the lifetime net tax rate facing each
generation bom betw een 1900 and 1993- The
calculations use National Incom e and Product
Account data on federal, state, and local taxes,
transfers, and spending for each year up to
1993, as well as OMB projections o f these ag­
gregates up to 2030.8
In the computational procedure, total taxes
and expenditures are classified into several cate­
gories for each year between 1900 and 2030. We
include taxes on incomes from labor and capital,
payroll taxes, and indirect taxes. Expenditures re­
fer to transfers such as Social Security, Medicare,
Medicaid, and other welfare payments, plus gov­
ernment purchases. The amount in each tax and
transfer category is distributed among generations
alive in a certain year— cohorts by single year of
age and sex ranging from newborn to 100 years
old. For years prior to and including 1993, we
use actual population data to perform this distri­
bution; for future years, we use population pro­
jections from the Social Security Administration.9
The amounts o f per capita taxes or transfers
distributed to members o f each generation are
determined according to relative profiles of tax
payments and transfer receipts obtained from mi­
croeconomic surveys.10 Current and past taxes
and transfers are distributed among different gen­
erations using available information on age- and
sex-specific payments and receipts for those
years. For some categories, such as Social Secu­
rity transfers, relative profiles are available for
each year between I960 and 1992. For others,
profiles are available for only a few o f the years.
For each payment and receipt category, the earli­
est available profile is used for distributing pay­
ments and receipts in prior years. Similarly, the
latest available profile is used to distribute the
amounts in later (including future) years.
For years beyond 2030, we project the per
capita amounts o f taxes and transfers by apply­
ing a growth factor to the values for the year

■

8 All outlays and receipts are measured in 1993 dollars.

■

9 We use the intermediate population projections through 2066
made by the Social Security Administration. We then extend these projec­
tions through 2200 using the mortality, fertility, and immigration assump­
tions applicable in 2066.

■ 10 These surveys include the Survey of Consumer Expenditures by
the Bureau of Labor Statistics, the Survey of Income and Program Participa­
tion by the Bureau of the Census, the Current Population Survey by the Bu­
reau of the Census, the Annual Abstracts of the Social Security Bulletin by
the Social Security Administration, and the Survey of Consumer Expendi­

tures
http://fraser.stlouisfed.org/ by the Federal Reserve System.

Federal Reserve Bank of St. Louis

2030. The prospective generational account for
each current (1993) generation is com puted by
subtracting total transfer receipts from total tax
payments in each future year that the genera­
tion will be alive, actuarially discounting the re ­
sulting net tax payments back to 1993 using an
assumed rate of interest, r, and summing over
the remaining years o f life for that generation.
The computation o f the lifetime generational
account for a given generation alive in 1993 uses
the same type of calculation, except that net
taxes paid in the past are also included. More­
over, the annual net taxes are actuarially dis­
counted back to the generation’s year o f birth. In
the case of the generation aged 43 in 1993 (those
bom in 1950), for example, per capita net taxes
paid up to 1993 and projected net taxes paid up
to 2050 (age 100) are capitalized to yield a gen­
erational account as o f 1950.
The present value of lifetime labor incom e
is used as a base to calculate the lifetime net
tax rate for each generation. As mentioned ea r­
lier, the lifetime net tax rate is the lifetime g e n ­
erational account as a percent o f the present
value of lifetime labor income. For each gen­
eration, the stream o f per capita labor incom e
earned in each year up to 1993 and projected
incom e for future years is capitalized to pro­
duce the present value o f lifetime labor in­
com e. W e derive the estimates o f per capita
labor income in a manner similar to that for
deriving per capita taxes and transfers: In e a c h
year, labor’s share o f net national incom e is
distributed by relative profiles of labor incom e
These profiles are based on individual w age
and salary data from the Census Bureau’s Cur­
rent Population Survey and are constructed fo r
the years 1963 through 1992.

The implications o f current fiscal policy for th e
lifetime net tax rates on future generations
(those born after 1993) can be derived by u sin g
the accounts o f generations currently alive. T h is
computation requires a consideration o f the
governm ent’s intertemporal budget constraint
which can be specified as
(1)

PVSPEND, = GWt + PVCt + PVFt.

Equation ( 1 ) states that the present value o f the
government’s current and projected purchases,
PVSPEND', must equal the government’s current
net worth, GWt , plus the present value of pro­
spective net tax payments o f all generations

5

currently alive, PVC,, plus the present value of
net tax payments o f all future generations,
PVFt . The sum o f prospective generational ac­
counts over all individuals currently alive pro­
vides an estimate o f PVFt .
We estimate the value o f PVSPENDt by com ­
puting the present value o f current and projected
government spending on goods and services.
Projections o f purchases through 2030 assume
that government purchases will keep pace with
population growth and with increases in labor
productivity. Spending projections beyond 2030
are made by applying a growth factor to per cap­
ita spending in 2030. Under the assumption that
the 2030 spending per capita will be maintained
thereafter (except for an adjustment for growth),
aggregating the per capita amounts across the
(projected) population for years beyond 2030
yields total spending for these years.
The per capita amounts o f purchases in
2030 are obtained by dividing the 2030 value
o f total purchases into one general and three
age-specific categories and distributing these
equally across the relevant (projected) popula­
tion segments for the year 2030. Finally, we es­
timate GWt by cumulating annual government
deficits over tim e .11 For the United States, the
value o f GWt is negative because government
budgets have been in deficit for most years
during the last several decades.
Knowing three o f the four terms in equation
( 1 ) enables us to derive the remaining item,
PVFt , as a residual. Thus, PVFt is the amount
o f the present value o f government purchases
not covered by current government net worth
plus the present value o f current and future
net tax payments by living generations. This re­
sidual must b e paid for by net tax payments to
be levied on generations as yet unborn.
Although the manner in w hich the residual
burden will be distributed across unborn gen­
erations is unknown today, we can illustrate its
magnitude by distributing it according to som e
predetermined rule. Here, w e adopt the crite­
rion that the distribution should equalize the
lifetime net tax rates of all future generations.
This requires that the residual burden be dis­
tributed equally across all future generations
except for an adjustment for growth .12 Thus,
generations born in year t pay net tax burdens
1 + g times the net tax burdens o f generations
born in year t - 1 , w here g is the annual rate
o f growth o f labor productivity .13 Because fu­
ture labor incom e is assumed to grow at rate
g, this adjustment im poses equal lifetime net
tax rates on all future generations.



A com parison o f the lifetime net tax rate on
future generations with that on new born gen­
erations is one way to estimate the degree o f
generational im balance em bedded in current
fiscal policy. The lifetime net tax rate on new ­
born generations is derived by finding the ratio
o f the present value o f their net tax payments
under current policy projections to the present
value o f their lifetime labor incomes. If a growthadjusted distribution o f the residual burden
among future generations produces a lifetime
net tax rate significantly larger than that on cur­
rent new borns, fiscal policy can be viewed as
being biased against future generations. If the
lifetime net tax rate on future generations is
judged as being prohibitively high, current fis­
cal policy may be deem ed unsustainable.

II. Generational
Accounts and
Lifetime Net
Tax Rates for the
United States
A. Prospective
Generational
Accounts
Baseline prospective generational accounts for
selected generations alive in 1993 are show n in
tables 1 and 2 . The calculations include all fed­
eral, state, and local government taxes, trans­
fers, and spending on goods and services and
assume that governm ent spending on goods
and services will keep pace with population
and productivity growth. They also incorporate
conservative estimates o f growth in governm ent

■

11 This method does not include the value of government physical
assets in GWt . However, if it did, one would have to include the present
value of imputed rent on these assets in PVSPEND,, representing the
government's purchase of the service flow from these assets for public
consumption. Because these two items would be equal in present value,
constraint (1) would be unaffected.
■ 12 Equal Bbsolute distribution of the residual burden would suc­
cessively reduce the lifetime net tax rates on generations born later be­
cause continued productivity growth will cause their labor income to
exceed that of generations born earlier. A growth-adjusted distribution of
the residual burden would result in the imposition of equal lifetime net
tax rates on all future generations. For a further discussion of these is­
sues, see Kotlikoff and Gokhale (1994).

■ 13 We assume that the ratio of per capita net tax burdens on future
male and female generations is the same as that on newborns.

6

T A B L E

1

The Composition of Male Generational
Accounts ( r * 0.06, g ■ 0.012)
(present values in thousands of 1993 dollars)
Taxes Paid
Generation's
Age in 1993

0
5

10
15

20
25
30
35
40
»5
50
55
60
65
70
75
H
O
85
90
Future
g e n era tio n s’1

Net Tax
Payment

87.2
107.0
130.3
159.6
188.7
199.9
195.7
182.7
158.6
119.7

68.0
7.1
-5 7 .0
-105.1
-1 0 8 .3
- 100.8
-8 6 .3
-7 6 .2
-5 8 .9
215.5

Labor
Income
Taxes

Capital
Income
Taxes

39.9
49.1

12.1

60.0
73-4

86.6
92.2
90.8

86.1
77.9
65.7
50.5
33.9
18.0
7.2
3.1

1.6
0.9

0.6
0.5
—

9.6
15.1
19.1
24.1
28.5
33.7
39.9
44.9
47.6
48.0
46.0
42.3
37.2
29.4
19.7
9.9

0.0
0.0
—

Transfers Received

Payroll
Taxes

Excise
Taxes

Social
Security

38.3
47.6
59.0
73.4

34.4
40.0
46.0
52.5
57.0
57.2
56.0
54.6
53.3
50.1
45.7
40.2
34.0
28.2

8.8
10.8
12.8

88.1
94.5
93.0

88.0
79.7
67.6
52.4
35.4
18.9
7.2
3.2

1.6
1.0
0.7
0.5
—

22.6
17.1

12.0
8.0
6.4
—

Health

22.4
26.2
30.8

Welfai

3.9
4.9
6.3

14.7

36.1

8.0

16.6

40.6
42.4
44.2
47.8
53.2
59.8
67.4
74.7
80.8
86.3
76.6
65.6
52.8
41.4
31.4

9.7
10.3
9.9
9.3

19.8
23.6
28.8
35.5
43.5
53.9
67.0
83.6
93.5
85.5
71.5
54.5
42.3
33.5
—

8.6
7.9
7.3

6.6
5.9
5.1
4.5
3.7
2.7

1.8

—

1.4
—

—

—

Percentage Difference in Net Payments

—
147.1
generations
and age zero____________________

Future

—

—

—

—

a. Generations bom in 1994 and thereafter.
SOURCE: Authors calculations.

health care outlays. The growth o f Medicare
and Medicaid expenditures averaged 7.4 and
15.5 percent, respectively, over the last five
years. The baseline incorporates a rapid growth
in these outlays until 2005, with somewhat
slower growth thereafter .14
The prospective net tax burdens shown in ta­
bles 1 and 2 exhibit a pronounced life-cycle pat­
tern. Working-age generations, who are in their
high earning and taxpaying years, have positive
net tax burdens: The present values of their in­
come, payroll, and indirect taxes are large, but
values of receipts from Scxial Security and health
care transfers are small. The opposite result holds
true for older generations.
In 1993, newborn males may expect to pay
$87,200, and newborn females $53,2(X), on net,
under baseline policies during their remaining




lifetimes. In contrast, average lifetime net tax
burdens amount to $215,500 for future males
and $131,500 for future females if the fiscal
treatment o f living generations continues un­
der baseline policies.
As mentioned earlier, prospective genera­
tional accounts can be com bined with past net
tax payments to calculate lifetime net tax burdens
for all living generations. Taken as fractions o f life­
time labor incomes, they yield lifetime net tax rates
Table 3 shows baseline lifetime gross and net tax
rates and gross transfer rates for generations

■ OMB'sPost-2005 growth rates for Medicare and Medicaid outlays are
14
the
best estimates. The growth rates used in all calculations are
available from the authors upon request.

D
T A B L E

2

The Composition of Female Generational
Accounts ( r = 0.06, 0 = 0 .0 1 2 )
(present values in thousands of 1993 dollars)
Taxes Paid

Generation’s
Age in 1993

0
5

10
15

20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
Future
generations 1
1

Net Tax
Payment
53.2
64.3
77.2
92.9
109.2
114.7
109.2
97.3
76.1
42.6
- 0 .3
- 4 9 .9

Capital
Income
Taxes

Labor
Income
Taxes

10.2

23.0
28.2
34.5
42.1
49.3
51.0
48.3

12.7
16.0

20.2

4 4 .3

25.4
30.6
35.7
41.0

39.0

44.4

31-8

45.2
43.9
41.6
38.0

23.4
14.7
7.3

- 101.0
-139.1
-1 4 0 .4
-1 3 1 .3
-1 1 1 .7
- 88.8
- 6 4 .8

2.6
1.0

32.0

Transfers Received

Payroll
Taxes

Excise
Taxes

23.3
28.9
35.8
44.5
53.1
55.6
53.0
48.9
43.5
35.7
16.8
8.4
3.0

33.1
38.3
43.7
49.2
53.0
53.7
53.3
52.8
51.4
48.5
44.2
39.3
33.6
28.0

26.6

22.8

0.5
0.3

17.3

0.1
0.1

0.0
0.0

0.1
0.1

—

131.5

1.2

0.5
0.3

22.5
12.4
4.7

—

—

9.5
7.2
—

12.6

Social
Security
8.3

10.2
12.1
13.8
15.5
18.6
22.3
27.3
33.6
41.6
52.0
65.6
82.8
91.9
84.8
72.0
57.2
43.1
32.6
—

Health
18.3
21.4
25.4
30.1
34.3
38.3
43.0
49.9
58.8
69.6
80.7
92.0

101.6
109.3

100.0
87.3
70.3
53.8
38.4
—

Welfare
9.8
12.3
15.4
19.1
21.7
19.4
15.8

12.6
9.7
7.3
5.7
4.6
3.9
3.5
3.1
2.7

2.2
1.7
1.3
—

a. Generations lx>m in 1994 and thereafter.
SOURCE: Authors' calculations.

T A B L E

3

Lifetime Net Tax Rates
Living and Future Generations
under Baseline Assumptions
Generations by
Year of Birth

Net
Tax Rate

1900
1910
1920
1930
1940
1950
1960
1970
1980
1990
1993
Future
generations 3

23.6
27.0
29.1
30.4
31.4
32.6
33.5
34.1
34.2
34.2
34.2
84.4

Ciross
Tax Rate

Gross
Transfer Rate

27.2

3.6
5.7
7.1
8.4
9.7

32.8
36.1
38.7
41.0
44.3
46.7
49.0
50.3
51.3
51.4
—

a. Generations tx>m in 1994 and thereafter.
NOTE: Calculations incorporate OMB projections.
SOURCE: Authors’ calculations.




11.6
13.2
15.0

16.1

b om in the year 1900 and in every tenth year
thereafter. It also presents these rates for 1993
newborns and future generations.
The lifetime net tax rates are populationweighted averages over male and fem ale gen ­
erations born in the same year. Table 3 show s
that lifetime net tax rates have risen from
nearly 24 percent on generations born in 1900
to m ore than 34 percent on those b om in
1993.15 For newborns in 1993, the net tax rate is
the difference between a gross tax rate o f 51 per­
cent and a gross transfer rate o f 17 percent. The
gross tax rate includes taxes on lalx>r and capital
income, payroll taxes, and indirect and other
taxes. The gross transfer rate encompasses re­
ceipts from Social Security, Medicare, Medicaid,

17.0
17.3
—
■ 15 More precisely, this rise occurred between 1900 and 1970. Life­
time net tax rates on generations born after 1970 will be maintained at
34.2 percent if generations currently alive continue to be treated per base­
line fiscal policies.

8

1 T A B L E 4
Lifetime Net Tax Rates for Living and
Future Generations under Alternative
Health Care and Federal Spending Paths

Generations
by Year
of Birth

1900
1910
1920
1930
1940
1950

Baseline

23.6
27.0
29.1
30.4
31.4
32.6
33.5
34.1
34.2
34.2
34.2

1960
1970

1980
1990
1993
Future
generations0

Slower
Health
Care
Growthb

Slower
Spending
Growth3

23.6
27.0
29.1
30.4
31.4
32.6
33-5
34.1
34.2
34.2
34.2

84.4

73.1

Slower
Health Care
and Spending
Growth

23.6
27.1
29.2
30.7
31.9
33.4
34.4
35.3
35.7

23.6
27.0
29.2
30.7
31.9
33.4
34.4
35.3
35.7

36.0
36.0

36.0
36.0

70.4

59.1

a. Federal spending is held constant in real terms after the year 2000.
b. Health care spending grows at a 2 percent slower rate than the baseline
through 2005, followed by baseline growth.
c. Generations bom in 1994 and thereafter.
NOTE: Calculations incorporate OMB projections.
SOURCE: Authors' calculations.

T A B L E

5

Percentage Difference under Alternative
Interest and Growth Rates: Baseline

r=
0.03
0.06

0.007

0.012

0.017

120

8 =

119
147

122

158
280

0.09

261

137
243

SOURCE: Authors’ calculations.

T A B L E

6

The baseline and other policies discussed so
far use a 6 percent rate of discount (r = 0.06)
and a 1.2 percent rate o f average productivity
growth (g = 0.012) to project taxes, transfers, and

Percentage Difference under Alternative
Interest and Growth Rates: Slower Health Care
Growth and Constant Real Federal Purchases
8 =
r=
0.03
0.06
0.09

and other welfare transfers. The lifetime net
tax rate on future generations is a staggering
84 percent, w hich is almost tw o-and-a-half
times as large as that on new borns in 1993.16
Table 4 reports lifetime net tax rates under
alternative future paths for outlays on health
care and federal purchases. Specifically, co l­
umn 1 o f table 4 repeats the baseline lifetime
net tax rates o f table 3. Column 2 shows the e f­
fect o f freezing real federal spending on goods
and services permanently beginning in 2000.
Lifetime net tax rates o f all living generations
are unchanged, since neither future tax nor
transfer payments are affected by this policy.
However, because reducing federal purchases
lessens the residual burden on future genera­
tions, their lifetime net tax rate is lowered to
73 percent. This result suggests that freezing
federal purchases permanently is not sufficient
to put the U.S. fiscal house in order from a g e n ­
erational accounting perspective.
Column 3 of table 4 reports the effect of as­
suming a 2 percent slower growth in health care
outlays until 2005, with baseline growth thereaf­
ter. Slower growth in health care spending raises
the lifetime net tax rates of young and middleaged living generations — those who will receive
lower health care transfers as a result. It also re­
duces the lifetime net tax rate on future genera­
tions by 14 percentage points. Thus, although
slower growth in government health care expen ­
ditures over the next decade will reduce the g en ­
erational imbalance in U.S. fiscal policy, a sizable
imbalance may still remain.
Column 4 of table 4 shows the effect o f com ­
bining the policies o f columns 2 and 3 — an opti­
mistic scenario. This reduces the lifetime net tax
rate on future generations from 84 percent to 59
percent. Thus, even if federal purchases are not
increased beyond current levels and growth in
health care outlays is 2 percentage points low er
than the baseline over the next 10 years, future
generations will incur lifetime net tax rates that
are 64 percent larger, on average, than those fac­
ing current newborns.

0.007

0.012

0.017

49
72
149

43
64
137

38
57
125

SOURCE: Authors’ calculations.




16 Note that future generations'lifetime net tax rate is derived by
distributing the residual of the present value of government spending af­
ter government net worth and the net contribution of living generations
have been deducted. Hence, it cannot be subdivided into gross tax and
transfer rates.

■

9

T A B L E

7

Permanent Tax Increases, Transfer Cuts,
or Spending Cuts Needed to Achieve a
Generationally Balanced Fiscal Policy
(percent)

Baseline

Slower
Spending
G row th3

Slower
Slower
Health
Health
Care and
Care
Spending
Growth*5 Growth

A. Policy Change in 1996

Tax Increases
Incom e taxc
Income tax (fed. only)
Payroll tax
Indirect taxes
All taxes

42.6
51.9
64.5

69.8
18.6

Transfer Cuts
Social Security
Health
All transfers

95.0
59.2

Spending Cuts
Entire government
Federal
Federal nondefense

97.4

32.8
31.6
_ d
_

32.9
40.1
49.9
54.0
14.4

29.1
35.5
44.1
47.7
12.7

19.6
23.9
29.7
32.1
8.5

73.5
45.8
25.3

65.0
49.0
24.8

43.7
33.0
16.7

26.3
93.7

21.7
67.9

60.2

_ d
_

_ d
_

15.8
_ d
_

B. Policy Change in 2001

Tax Increases
Incom e taxc
Incom e tax (fed. only)i
Payroll tax
Indirect taxes
All taxes
Transfer Cuts
Social Security
Health
All transfers
Spending Cuts
Entire government
Federal
Federal nondefense

39.9
48.5
61.3
67.7
17.6

35.2
42.8
54.1
59.8
15.6

23.7
28.8
36.4
40.3
10.5

66.4
37.8

87.2
51.4
29.3

77.0
55.8
28.9

51.8
37.5
19.5

38.8

32.9

26.7
84.9

19.7
80.2

51.5

62.6
79.2
87.5

22.8
_ d
_

_ d
_
_ d
_

_ d
_
_ d
_

_ d
_

_ d
_

purchases beyond 2030.17 Table 5 presents the
percentage difference between the lifetime net
tax rates on future and 1993 newborn genera­
tions under alternative interest and productivity
growth rates for the baseline .18 Table 6 depicts
the same calculation for the optimistic scenario
o f slower health care outlay growth and constant
real federal spending.
Using a higher discount rate while keeping
the productivity growth rate constant can have
an ambiguous effect on the percentage differ­
ential. In present-value calculations, a higher
rate o f discount reduces the relative weight on
net payments that are further into the future.
Hence, if the profile o f aggregate net tax pay­
ments by living generations is rising through
time while that o f government purchases is fall­
ing, a higher discount rate will tend to increase
the residual burden on future generations. If
the slopes o f the time profiles o f spending and
net payments are reversed, a higher discount
rate may reduce the residual burden. Similar re­
marks apply for varying the rate o f productiv­
ity growth while keeping the discount rate
fixed. Despite the ambiguity, however, it is use­
ful to exam ine w hether the conclusion o f an
imbalanced U.S. fiscal policy is sustained over
a reasonable range o f interest and growth rates.
Table 5 shows that for many such rates, the
lifetime net tax rate of future generations is more
than twice as large as that o f 1993 newborns. Un­
der optimistic projections (table 6 ), the percent­
age differentials range from 38 percent to 149
percent. Thus, the conclusion that current U.S.
fiscal policy is severely imbalanced remains true
under a wide range o f interest and growth rates,
despite using optimistic assumptions about future
federal purchases and health care outlay paths.

C. Policy Change in 2016

Tax Increases
Incom e taxc
Incom e tax (fed. only)i
Payroll tax
Indirect taxes
All taxes
Transfer Cuts
Social Security
Health
All transfers
Spending Cuts
Entire government
Federal
Federal nondefense

97.7
118.2
156.4
189.2
45.2

75.6
91.5

121.0
146.4
35.0

66.8
80.8
106.9
129.4
30.9

45.0
54.4
72.0
87.1

20.8

_ d
_
_ d
_

_ d
_

_ d
_

87.4

83.2

61.0

63.4

49.1

90.7
48.9

73.0
d

65.3

50.5

39.3
__d

d

d

32.9

a. Federal purchases are held constant in real terms alter the year 2000.
b. Health care spending grows at a 2 percent slower rate than the baseline
through 2005, followed by baseline growth.
c. Federal, state, and local income taxes.
d. Requires a reduction of more than 100 percent.
SOURCE: Authors’ calculations.




B. Fiscal Policies
Required to
Eliminate the
Imbalance
Next, to address the methodological criticism
discussed earlier, we compute the tax increases,
transfer cuts, or spending reductions necessary

■ 17 Earlier presentations of generational accounting assumed a
0.75 percent rate of productivity growth. The OMB’s latest budget projec­
tions through 2030 incorporated the assumption of a 1.2 percent rate of
productivity growth (defined in terms of GDP per worker). This dictated
the use of the same rate for years beyond 2030.

■ 18 The percentage difference is calculated as ((f/C l-1 ) x 100,
where F is the lifetime net tax rate on future generations and C is the
same rate on 1993 newborns.

10

T AB L E

8

Equalized Lifetime Net Tax Rates for
Newborn and Future Generations
Resulting from Table 7 Policies
(percent)

Baseline

Slower
Spending
Growth*

Slower
Health
Slower
Care and
Health
Spending
Care
G rowthb Growth

A. Policy Change in 1996

T ax Increases
Incom e taxc
Incom e tax (fed. only),
Payroll tax
Indirect taxes
All taxes

42.7
42.8
43.9
44.8
43.6

40.8
40.9
41.7
42.4
41.5

41.9
41.9
42.6
43.3
42.4

39.9
40.0
40.5
40.9
40.3

Transfer Cuts
Social Security
Health
All transfers

38.1
39.9
39.7

37.2

38.7
39.8
39.8

37.8

Spending Cuts
Entire government
Federal
Federal nondefense

34.2
34.2
__d

34.2
34.2
__d

36.0
36.0

36.0
36.0

38.6
38.5

__ d

38.6
38.5

__d

B. Policy Change in 2001

Tax Increases
Incom e taxc
Income tax (fed. only)
Payroll tax
Indirect taxes
All taxes

44.5
44.6
46.0
46.4
45.4

42.2
42.2
43.4
43.6
42.9

43.1
43.1
44.1
44.4
43.7

40.8
40.8
41.5
41.6
41.2

Transfer Cuts
Social Security
Health
All transfers

__ d

37.6
39.0
39.0

39.1
40.2
40.3

38.1

40.4
40.4
34.2
__ d
__ d

34.2
__ d
__ d

36.0
36.0

36.0
36.0

Spending Cuts
Entire government
Federal
Federal nondefense

__ d

38.8
38.9

__ d

C. Policy Change in 2016

T ax Increases
Incom e taxc
Incom e tax (fed. onl )
Payroll tax
Indirect taxes
All taxes
Transfer Cuts
Social Security
Health
All transfers
Spending Cuts
Entire government
Federal
Federal nondefense

48.6
48.6
50.7
48.0
49.0

44.5
44.5
45.9
44.1
44.8

__ d

__ d

38.8

43.0

40.8
41.0

41.8
42.0

39.9
40.0

34.2
__d
__d

34.2
__ d
__d

36.0

36.0

52.5
52.6
55.6
51.7
53.2

48.4
48.4
50.8
47.8
48.9

__ d
__d

__ d
__ d

__ d
__ d

a. Federal purchases are held constant in real terms after the year 2000.
b. Health care spending grows at a 2 percent slower rate than the baseline
through 2005, followed by baseline growth.
c. Federal, state, and local income taxes.
d. Requires a reduction o f more than 100 percent.
SOURCE: Authors’ calculations.




to eliminate the generational im balance in U.S.
fiscal policy. Various combinations o f all three
policies are introduced beginning in 1996,
2001, and 2016. Because the new policies are
applicable to all generations alive w hen they
are introduced, they will affect the lifetime net
tax rates o f most living generations. In each
case, w e calculate the permanent percentage
increase (or reduction) required in taxes, trans­
fers, or purchases in order to equalize the life­
time net tax rates o f 1993 new born and future
generations.
Panel A o f table 7 presents the percentage
by which various taxes, transfers, and spend­
ing will have to change beginning in 1996 to
eliminate the generational imbalance. The re­
quired percentage increases are shown for the
baseline and for the alternative federal spending
and health care outlay growth paths analyzed
in table 4. Under baseline projections, incom e
tax revenues would have to increase perm a­
nently by almost 43 percent beginning in 1996
to equalize the lifetime net tax rates o f new ­
born and future generations. This implies that
the average income tax rate would have to rise
from 15.7 percent currently to 22.3 percent im ­
mediately and permanently.
Under the fortuitous case o f slow growth in
health care outlays and zero growth in federal
purchases, income taxes would have to increase
by about 20 percent. If only federal income taxes
are considered, the required increases in annual
revenues range between 24 and 52 percent; those
necessary under payroll or indirect tax hike poli­
cies are even larger. If all taxes are considered,
eliminating the imbalance in U.S. fiscal policy
would require tax hikes of about 19 percent un­
der baseline projections and 8.5 percent under
the optimistic scenario.
Cuts in transfers to establish equal lifetime net
tax rates on newborn and future generations
would also be severe. Under the baseline projec­
tion, a 33 percent permanent and across-theboard reduction in transfers beginning in 19%
would be necessary to restore a generationally
balanced policy. Alternatively, restoring balance
would require permanently reducing the size o f
combined federal, state, and local government
purchases by 32 percent beginning in 1996.
Table 8 shows the value at which the lifetime
net tax rates on 1993 newborns and future gen­
erations would be equalized under the corre­
sponding policies shown in table 7. Under
baseline projections, for example, increasing all
taxes permanently by 19 percent beginning in
1996 would raise the lifetime net taxes of 1993
newborns from 34 percent to 44 percent and

TABLE

9

100 percent. For exam ple, eliminating health
care transfers entirely beginning in 2016 would
not be sufficient to restore a generationally bal­
anced policy.

Impact of the Balanced Budget Proposal
by the Year 2002 on Lifetime Net Tax
Rates of Living and Future Generations

Baseline

Balanced
Budget
Proposal

1900

23.6

23.6

1910
1920
1930
1940
1950

27.0
29.1
30.4
31.4
32.6
33.5
34.1
34.2
34.2
34.2
84.4

27.1
29.2
30.6
31.7
33.1
34.0
34.8
35.2
35.2
35.1
72.5

Generations by
Year of Birth

1960
1970
1980
1990
1993
Future generations 1

Difference1
5

0.0
0.1
0.1
0.2
0.3
0.5
0.5
0.7

1.0
1.0
0.9
-1 1 .9

a. Present value of lifetime net taxes as a ratio o f the present value o f life­
time labor income.
b. Percentage-point increase in the net tax rate if the balanced budget pro­
posal is adopted.
c. Generations born in 1994 and thereafter.
SOURCE: Authors’ calculations.

reduce that on future generations from 84 percent
to 44 percent. That is, increasing all taxes perma­
nently by 19 percent is equivalent to increasing
lifetime net tax rates of 1993 newborns by almost
30 percent. Note that the equalized lifetime net tax
rates on newborn and future generations are differ­
ent for different policies. If an across-the-board
transfer cut were adopted instead of an across-theboard tax hike, lifetime net tax rates on newborn
and future generations would be equalized at 40
percent instead o f 44 percent.
Delaying policy changes to restore a generationally balanced fiscal policy is likely to prove
costly. This can be seen from panels B and C
in tables 7 and 8 . Raising incom e taxes begin­
ning in 2001 instead o f in 1996 will necessitate
an increase o f 52 percent instead o f 43 percent.
Similarly, initiating cuts in government pur­
chases in 2001 instead o f in 1996 will deepen
the cuts to 39 percent from 32 percent. Intro­
ducing these policies in 2016 will push the re­
quired income-tax hike to 98 percent and will
increase the cuts required in government pur­
chases to 73 percent.
The same is true for all other tax increases
and transfer or spending cuts. Indeed, some
spending and transfer cuts that will restore gen­
erational balance if implemented in 1996 are
no longer feasible if implemented in 2001 or
2016 because the required cuts would exceed




The required hikes in taxes or cuts in trans­
fers and spending to restore generational e q ­
uity are quite considerable. The main m essage
o f this section is that no matter how one
chooses to calculate it, the mammoth size o f
the im balance in U.S. fiscal policy cannot be
made to disappear. Moreover, policy changes
to correct the im balance need to be introduced
sooner rather than later: Procrastination will
only make the m edicine more bitter.

III. The Balanced
Budget Amendment
This section contrasts the policies required for
restoring generational balance in fiscal policy
with those being considered by policymakers
today. While debating the adoption of a balanced
budget amendment to the U.S. Constitution,
Congress recently considered proposals to cut
all outlays except for defense and Social Secu­
rity. Here, we consider the impact o f similar
cuts on the generational stance o f U.S. fiscal
policy. The outlay reductions involve cuts in
nondefense discretionary spending ranging
from 1 percent in 1996 to 4 percent in 2002
from our baseline values. For Medicare and
Medicaid, the reductions range from 3 percent
in 1996 to 14 percent in 2002. Finally, cuts in
other mandatory spending categories range
from 4 percent in 1996 to 16 percent in 2002.
For each category, the percentage cut for 2002
is preserved in later years .19
Table 9 shows the impact o f this proposal
on the lifetime net tax rates o f living and future
generations. The rates are higher for living,
especially younger, generations. The rate for
generations born in 1950, for exam ple, in­
creases by 0.5 percent, while that for 1993 new ­
borns is almost 1 percentage point higher.
The proposal would imply a reduction in the
lifetime net tax rate o f future generations from
84 to 73 percent.
The outlay cuts analyzed here redress the
imbalance to som e extent, but still leave an unsustainably large lifetime net tax rate on future
generations. Thus, under what w e consider to

■ 19 These cuts balance the federal budget by the year 2002 from a
“current law” baseline in which federal discretionary spending is frozen
in nominal terms. Under our conservative baseline, however, the budget
remains in deficit in all future years.

be conservative but reasonable budget projec­
tions, future Congresses may need to rein in
outlays or increase revenues further to restore
generational balance to U.S. fiscal policy.
Given the results o f the previous section, leav­
ing such large adjustments for future considera­
tion is likely to prove costly.

IV. Conclusion
The generational stance o f current U.S. fiscal
policy is badly out o f balance. It is impossible
to avoid this conclusion no matter which of
many alternative measures one uses to analyze
the generational distribution o f net tax burdens.
Although tax cuts seem to have widespread
political appeal today, the analysis presented
here suggests that enacting them may be the
wrong thing to do.
In fact, the early adoption o f fiscal measures
to reduce the projected heavy net tax burdens
on future generations is imperative. This re­
quires either increasing taxes or reducing gov­
ernment outlays today. Redressing the current
U.S. fiscal im balance is important because such
heavy burdens will prove econom ically infeasi­
ble to impose on future generations in view of
the fact that gross tax rates would have to be
higher than net tax rates. Moreover, imposing
high lifetime net tax burdens on future genera­
tions may depress their incentives to work,
save, and invest, thereby hurting future Ameri­
cans’ living standards. Finally, the analysis
show s that postponing the adoption o f correc­
tive measures will only w orsen the choices
available to policymakers in the future.




References
Altonji, Joseph, Fumio Hayashi, and Laurence J.
Kotlikoff. “Is the Extended Family Altruisti­
cally Linked? Direct Tests Using Micro
Data,” American Economic Review, vol. 82,
no. 5 (D ecem ber 1992), pp. 117 7 - 98.
Auerbach, Alan J., Jagadeesh Gokhale, and
Laurence J. Kotlikoff. “Generational Ac­
counts: A Meaningful Alternative to Deficit
Accounting,” in David Bradford, ed., Tax
Policy a n d the Economy , vol. 5. Cambridge,
Mass.: MIT Press and the National Bureau o f
Econom ic Research, 1991, pp. 5 5 - 1 1 0 .
------------ »--------------» an d _________ . “G enera­
tional Accounts and Lifetime Tax R a tes__
1900-1991,” Federal Reserve Bank o f C leve­
land, Economic Review, vol. 29, no. 1
(Quarter 1 1993), pp. 2 - 1 3 .
------------ >------------- »and__________ “
Generation­
al Accounting: A Meaningful Way to Evaluate
Fiscal Policy,”Journal o f Economic Perspec­
tives, vol. 8, no. 1 (Winter 1994), pp. 7 3 - 9 4 .
Eisner, Robert. “The Grandkids Can Relax,”
The Wall Street Journal, November 9, 1994.
Haveman, Robert. “Should Generational Ac­
counts Replace Public Budgets and Deficits?”
Journal o f Economic Perspectives, vol. 8, no. 1
(Winter 1994), pp. 9 5 -1 1 1 .
Kotlikoff, Laurence J. Generational Accounting:

Knowing Who Pays, a n d When, f o r What
We Spend. New York: The Free Press, 1992.
------------ , and Jagadeesh Gokhale. “Passing the
Generational Buck,” The Public Interest, no.
114 (Winter 1994), pp. 7 3 -8 1 .
Office o f Management and Budget. Analytical

Perspectives, Budget o f the United States Gov­
ernment, Fiscal Year 1995. Washington,
D C.: U.S. Government Printing Office, 1994

13

Vagueness, Credibility,
and Government Policy
by Joseph G. Haubrich

Joseph G. Haubrich is an econo­
mist and consultant at the Federal
Reserve Bank of Cleveland. The
author thanks Loretta Mester for
helpful comments.

Introduction
Have more than thou showest,
Speak less than thou knowest,
Lend less than thou owest.
— William Shakespeare,

King Lear
(Act I, sc. iv, line 132)
Should the Federal Reserve — or any other
government agency — make precise statements
about its policy objectives? Determining the
proper amount o f secrecy in government gen­
erates controversy whether the agency involved
undertakes espionage, banking, or monetary
policy. Betw een the broad areas of agreement
(classifying military strategies, publishing legis­
lation) lie equally broad areas of contention.
This article explores the econom ic reasons
why a government agency may find it in its
own — and society’s — interest to be vague
about policy objectives. Circumstances arise in
which it is optimal for agencies to release only
partial information about their decisions. For
that reason, vagueness, and the secrecy n eces­
sary to preserve it, represent an accom m oda­
tion with an imperfect world rather than a

conspiracy o f silence.


Unlike complaints about the Central Intelli­
gence Agency or the National Security Agency,
the objections against banking and monetary
authorities center not around a total lack o f pub­
lic announcements, but around the vagueness of
their policy statements. This results from three re­
lated but separable policies: closed meetings,
delayed release o f decisions and minutes, and
uninformative releases. Immediate release o f a
videotaped meeting may matter little if the poli­
cies agreed upon remain vague and imprecise,
while a blacked-out, highly secret meeting
could in principle result in detailed, precise
statements o f policy.
In the area o f banking regulation, Irvine
Sprague, a former director of the Federal Deposit
Insurance Corporation (FDIC), described his am­
biguity about announcing which banks were too
big to fail: “Comptroller Todd Conover hinted
that the eleven largest banks in the nation were
immune from failure. In my Boston speech, I
identified the top two as being absolutely safe.
The right number is elusive .” 1

■

1 See Sprague (1986), p. 259.

14

Closure policy is not the only area where
banking rules seem vague, nor do regulators
have a m onopoly on ambiguity. Regulatory en ­
forcem ent o f commercial lending standards —
a serious concern during the last recession— has
also been criticized for imprecision (McLemore
[1991]). In the realm o f monetary policy, Con­
gressman Henry B. Gonzalez, former chairman
o f the House Banking Committee, has called
for videotaping Federal O pen Market Commit­
tee (FOMC) meetings and for the immediate re­
lease o f monetary policy objectives. Outside
the government, credit-rating agencies do not
always announce precise standards for each
rating (Hansell [19931). More recently, both
types o f ambiguity have surfaced in the area of
derivatives. There is apparently still some un­
certainty about how regulators will treat bank
investment in derivatives (Karr and. Gaylord
[1994]) and about what banks will tell their cus­
tomers (Tomasula [1994]).
In this article, I explore the concept techni­
cally know n as “cheap talk” as a simple eco ­
nomic reason for secrecy and vagueness. Cheap
talk illustrates an incompatibility betw een preci­
sion and credibility in policy announcem ents
and provides an econom ic explanation of why
such announcem ents provide a limited, but
still real, amount o f information. The cheaptalk explanation for secrecy em phasizes the
cooperative nature o f the problem. In that re­
spect, it differs greatly from the vagueness and
secrecy o f a lazy worker hiding from his boss
or o f a junta trying to keep its human rights
violations from the press. Cheap talk presents
an agency that wants to communicate, but that
for reasons detailed below', cannot do so with
perfect precision.
This article presents a simple exam ple of
points first raised by Stein (1989), along with an
intuitive introduction to the econom ic theory
o f cheap talk. It then uses some recent advances
to look at why Stein’s arguments for secrecy
may fail and why precise announcem ents
would be useful .2

■ 2 Other authors have suggested different reasons for vagueness
and secrecy. See Goodfriend (1986) and Kane (1980) for a more detailed
examination of this issue.

■ 3 Signaling works, then, when its benefits outweigh its costs— but
things don't always happen that way. Economists thus distinguish be­
tween “separating" equilibria, where different types split out, and “pool­
ing" equilibria, where everyone acts the same. See Spence (1973).




I. Cheap Talk and
Communication
"Then you should say what you m ean , ”
the March Hare went on. “ do, "Alice hastily
I
replied; “ least— at least I mean u 'hat I sa \— ”
at
— Lew is Carroll,

Alice 's Adventures in W onderland
Secrecy and vagueness describe aspects of
communication. Consequently, any econom ic
theory o f secrecy and vagueness must address
the econom ics o f communication. The facet
that appears most useful, and that I therefore
concentrate on, is technically called cheap talk.
Cheap talk refers to unverifiable m essages that
are costless to send and receive. This stands in
contrast to “signaling,” a better-known eco ­
nomic theory o f comm unication that refers to
messages which are both costly and verifiable.
Signaling builds on the intuition o f "put
your money where your mouth is.” The e c o ­
nomics o f signaling, for instance, explain why
a company will erect a costly headquarters to
demonstrate its intent to stay around, or w hy
skilled workers undertake the expense o f a c o l­
lege education to distinguish them selves from
less skilled workers. In each case — construc­
tion or education — the costly action serves
notice of something important, such as d ep en d ­
ability or quality. Every firm wishes to appear
reliable, and every worker w ishes to appear
highly skilled. Those w'ith a true advantage dif­
ferentiate themselves by bearing the cost o f sig­
naling, which acts as a device to screen out
less desirable types .3
Cheap talk, in contrast, arises when different
types do not wish to appear the same and w hen
there is no costly investment option. An example
here would be the classified ads. Nothing pre­
vents me from listing a piano for sale, but it
serves no purpose if I really wish to sell my
comic hxx)k collection. Likewise, a SBF (single
black female) would most likely not list herself as
a DJM (divorced Jewish male), though in princi­
ple she could.
More abstractly, the communication envi­
sioned by cheap-talk theory involves a sender
and a receiver. The sender has private inform a­
tion that matters to the receiver, w ho must
choose an action. The outcom e depends on
both the sender’s type (that is, the private infor­
mation the sender has) and the action taken
by the receiver. Thus, a receiver’s action might
be to visit my house writh the intent to buy m y
comic book collection.

15

T A B L E

1

Coordination Game
R eceiver
Sender
Type
Type

A ction A

A ction C

2,3

0,0

1,2

0,0

a
b

A ction B

2,3

1.2

SOURCE: Adapted from Matthews, OkunoFujiwara, and Postlewaite (1991).

FI GURE

1

Utility Functions
m+b

The classified ad exam ple pinpoints one big
advantage o f cheap talk: coordination. It wastes
everyone’s time if aspiring pianists, rather than
X-men aficionados, com e to my house. Like­
wise, agreeing on a place to m eet if one gets
separated from a group o f friends at the mall
gives another simple exam ple of the advan­
tages o f cheap talk as coordination.
Table 1 describes the coordination role of
cheap talk in the formalism of game theory. The
sender may Ix . type a or type b, while the re­
1
ceiver may take action A, B, or C. The first num­
ber of each pair denotes the payoff to the sender;
the second is the payoff to the receiver. If the
sender does not send a message about his type,
the receiver takes action C, because the certain
payoff o f 2 beats the average of 1.5 from choos­
ing A or B in ignorance. The sender, however,
has an incentive to send a message — and to
send the taith — because delivering the wrong
message hurts the sender as well as the receiver.
If a type a sender announces “I’m type b." then
both the sender and receiver get zero .1




This sort o f com m unication or coordination
game has been justified here with rather hom ey
exam ples o f pianos, com ic books, and malls,
but it has a direct bearing on policy announce­
ments. Consider a central bank that, for whatever
reason (internal politics, the latest econom ic re­
search), has a particular position on how much
banks should rely on discount-window borrow ­
ing for short-term liquidity. An easy central
bank would let banks borrow substantial
amounts at short notice. Banks, if they knew
this, would want to structure their loan portfo­
lios to exploit this possibility. A tough central
bank would discourage lending, and if banks
w ere aware o f that, they would not want to be
caught short. In this case, it benefits the central
bank to com m unicate its position to the banks
— that is, to declare w hether it is type a (easy)
or type b (tough) in the game o f figure 1 .
To take another exam ple, a regulator may
look at low-capitalized financial institutions,
such as savings and loans, and decide how it
wants to deal with their risky investments.
O ne type o f regulator may prefer to prosecute
management vigorously for undertaking what it
deems to be inappropriate risks, while another
type may view denying those investments as
an unfair hardship on a w ell-m n organization.
Clearly, it matters to the thrift ow ners — and to
their investment strategy — which position the
regulator takes. Just as clearly, the regulator is
much m ore likely to get its way by talking
cheaply and revealing its type to the industry.

II. Secrecy and
Vagueness: The
Partition Equilibrium
Men use ... speech only to conceal
their thoughts.
— Voltaire, D ialogue 14.

Le Chapon et la Poularde
In the previous section, cheap talk served a coor­
dinating role, being both credible and precise.
Vagueness and secrecy had no place. This
section describes a m ore subtle effect in which
■ 4 Even in this simple example, things are not as straightforward as
they seem. For example, another cheap-talk equilibrium exists in which the
receiver ignores all messages, and hence the sender can report any arbitrary
message. Game theorists accurately describe this as the babbling equilib­
rium, which points out another difficulty with cheap-talk games: They often
have several equilibria, only one of which may have the desired properties.
The example also leaves unspecified the language of the messages, whether
verbal, code, or the number of lamps left in the tower of Boston’s Old North
Church. Readers interested in a deeper treatment of these issues should con­
sult Matthews, Okuno-Fujiwara, and Postlewaite (1991).

16

precision and credibility conflict with each
other, leading to secrecy and vague policy pro­
nouncements.
The increased subtlety o f this result also re­
quires a more formal approach. Let the sender
be the bank regulator and the receiver b e a
bank or the banking system. The regulator has
a preferred risk level for banks that strikes
som e balance betw een safety and profitability
and that takes into account the cost o f a bail­
out. This preferred risk level, denoted m and
distributed uniformly betw een 0 and 1 , deter­
mines the sender’s type, but is unknown to the
bank. T he bank, perhaps because it does not
internalize the cost o f the safety net provided
by the regulator (or perhaps because it under­
stands the risks better), prefers to undertake
more risk. The regulators know the extent o f
this bias, denoted b. The bank must put to­
gether a loan portfolio with risk level y, also
falling somew here betw een 0 and 1 .
The regulator’s utility is
(1)

UR= - ( y ~ m )2.

The bank’s utility is

(2)

UB = - { y - [m + b ] ) 2

Figure 1 illustrates these functions. Reflecting the
difference in preferred risk levels, equation ( 1)
has a maximum at y = m, while equation ( 2) has
a maximum at y = tn+ b. The bank and the regu­
lator know each other’s utility function.
Equations (1) and (2) embody several impor­
tant assumptions. First, the interests o f the regu­
lator and the bank are not perfectly aligned.
Nonetheless, the bank does care about what the
regulator chooses, since a bank far from the regu­
lator’s preferred risk level may face increasingly
intrusive regulation. In the terminology o f Buser,
Chen, and Kane (1981), the regulatory tax be­
comes more and more burdensome as the bank’s
risk deviates further from the regulator’s preferred
level. For example, although increasing risk may
boost the bank’s income, the higher regulatory
taxes could mean that profits will drop.
Items falling under the regulator’s discretion
include the handling o f branch and merger
proposals, the extent and thoroughness o f ex ­
aminations, and, in extrem e cases o f failure,
lawsuits or overly stringent regulation. Such
procedures may mean the difference betw een
current managers remaining in place during a
reorganization, a new management team being
brought in, or even prosecution for malfeasance.
Making this problem nontrivial is the private




nature o f m. Only the government agency o b ­
serves m, which reflects either the regulator’s
exact feelings, som e bureaucratic/political out­
com e, or econom ic analysis based on confiden­
tial inputs, such as BOPEC or CAMEL ratings.^
It is possible that this value changes over time,
with new administrations and new appoint­
ments. Formally speaking, in the model pre­
sented here, the level o f m is given to the
government by such a process, rather than b e­
ing freely chosen.
Equally important, the regulator wishes to
com m unicate its m type — it d oesn’t just want
to m ake all banks think that it is tough. For ex ­
ample, a regulator with a low m views banks
investing a large share o f deposits in safe T-bills
as prudent. A regulator with a high m views
such banks as lending too little. As Stein (1989)
puts it, “Not all types want to create the same
expectations” (p. 36). Hence, regulators want
to let banks know the level o f m 6
Now w e are in a position to discuss secrecy
and vagueness. W e must proceed, however, in
a way that may seem backwards. That is, we
start with the answer and then show that it
works. Specifically, a particular type o f vague­
ness, announcing a range o f m rather than a
specific value, solves the credibility problem.
In gam e-theoretic terminology, we conjecture
an equilibrium and show our conjecture to be
correct. Though econom ically and logically pre­
cise, this approach is unsatisfying — a bit like
knowing that 17 X 17 is 289 without knowing
how to extract square roots.
With these preliminaries out o f the way, we
can understand how vagueness and secrecy
play a role. Suppose, as in the earlier examples,
that the regulator notices the coordination as­
pect o f the problem and announces m. The
bank, however, believes that a slightly higher
risk level is appropriate and, knowing m ,
chooses a risk o f y = m + b. The regulator
doesn’t like this, so instead o f announcing m,
it announces m - b, figuring that when the
bank increases its risk above the announced
m, it will return to the risk level most preferred
by the regulator. But the bank isn't stupid. It
knows that the regulator wants to understate

■ 5 BOPEC ratings apply to bank holding companies, while CAMEL rat­
ings apply to banks. Both are confidential assessments of these institutions’
health filed by their regulators. See Spong (1990) for additional details

■6

In Stein's model of monetary policy, some distortion (caused
either by the government or by a market imperfection) means that the
monetary authority wishes to fool people and drive down the unemploy­
ment rate. The imperfect correlation of interests thus takes a slightly dif­
ferent form than in this paper.

D
m, so it overstates y even more. Understanding
this, the regulator wants to understate m further
yet, meaning that the bank adjusts risk y up
even more, meaning that the regulator .... Obvi­
ously, credibly communicating m proves impos­
sible. Because the regulator has an incentive to
manipulate banks' expectations, it cannot credibly
and precisely announce its preferred risk level.
Divergent interests make this impossible .7
Banks and regulators have similar, but not
identical, interests. This makes communication
desirable, but precise announcements useless. On
the other hand, it makes imprecise — or vague
— announcements useful. Suppose that instead
o f announcing that the preferred risk for banks is
m = 0.57721, the regulator simply announces
whether its preferred risk is high, medium, or
low. Because interests are not identical, the regu­
lator wants to manipulate banks’ expectations.
However, because interests are similar, a regula­
tor with a high preferred risk (large m) will not
manipulate expectations too far. It will not want
to tell banks that its preferred risk is in the low
category, since the difference is just too large.
With only three choices, the coordination side o f
communication becomes more important than
the manipulation side. The regulator in effect
commits itself to not telling little white lies —
only big lies are possible. And while the regulator
wishes that its hard-charging loan machine would
take a little less risk, it really doesn't want the
bank to becom e a conservative bond investor.
More formally, consider the regulator an­
nouncing a “partition” o f three intervals [0, « ,],
[ a v a 21, and [a2, 1]. (For completeness, I define
the first and last terms as a 0= 0 and « 3= 1 .)
W henever m falls betw een 0 and a v the regu­
lator announces that it favors low risk, or that
m is in the interval [0, a t).
For any such announcement, the bank, know­
ing m has a uniform distribution, makes a best
dj + cii + j
guess o f it a s ------2------an<J consequently
chooses its risk level as

(3)

y-

a i + a >+1 , u 8

-------- --------- + o.

The bank pushes up its risk level by b from its
best guess o f the regulator’s true m. For exam ­
ple, w henever m falls betw een 0 and a x, the
bank sets

each region. It must be true that if m falls in
the interval [ap a i+ 1], the regulator prefers to
announce that particular interval rather than
any other.
At the boundaries, an arbitrage condition
holds: The regulator, with a target risk level o f
t n - a i , must be indifferent betw een announc­
ing interval [ai _ v a i ] or [a t , a i+1 ]. From equa­
tions ( 1 ) and ( 3), this condition becom es

(4)

a . + a.^,
- (^ - J m + b _ a y
_

a i-i + a i

2

------ 2------+ b - a.) .

Equation (4) reduces to a difference equation
having the form a i+1 = 2a i - a i_ 1- 4 b , subject
to a 0 = 0 and a } = 1.
Standard methods exist to solve such differ­
ence equations (see Goldberg [1958]), and us­
ing them delivers the results

a l = —+ 4b and
2 /,
a 2 = - + 4b.
If we set b =
then the three intervals (or
partitions) become low = [ 0, \ ], medium = [ ± J-],
and high = [
1 ], Notice the asymmetry in this
partition equilibrium. The intervals are not all
the same size, meaning that the regulator can be
more precise when its preferred risk level ex ­
ceeds the mean (that is, when m > -|). Because
the bank tends to set risk above what the regu­
lator prefers, the regulator can use the natural
endpoint, m = 1 , to create a more precise an­
nouncement. The result is that announcem ents
will be vaguer and secrecy will be higher when
the regulator’s risk is relatively low.
These numbers make the exam ple particu­
larly simple, but the main points carry through
in general. The number and size o f the parti­
tions may vary as the exact trade-off between
coordination and manipulation changes. Thus,
partitions remain, as does the asymmetry b e­
tween them.
To summarize, the regulator wishes to com ­
municate its preferred risk level to the bank.
The gaming caused by the bank desiring more
■ 7 This scenario assumes that the interaction is a one-shot game.

y = T +b
In order to show that this vagueness tactic
 actually works, we need to be more specific
http://fraser.stlouisfed.org/
and calculate the a- s, or the boundaries for
Federal Reserve Bank of St. Louis

Considering repeated interactions between the bank and the regulator may
lead to different results, but only, as Stein (1989) notes, under very strong
assumptions.

■8

This analysis closely follows Crawford and Sobel (1982). Banks
choose y to maximize their expected utility, given by equation (2).

18

risk than does the regulator means that any
precise announcem ent will not be credible. The
partition equilibrium, on the other hand, deliv­
ers a credible announcement that is not precise.

III. Small Lies
and Small Banks
Striving to better , oft we m ar what 's well.
— William Shakespeare,

King Lear
(Act I, sc. iv, line 371)
The partition equilibrium provides an intuitive
justification for secrecy and vagueness. It repre­
sents a way to communicate credibly when inter­
ests are similar but not identical. A closer look at
the reasoning involved, however, casts some
doubt on the general applicability o f the results.
Because an exacting analysis of the criticisms
would involve some highly technical aspects in­
appropriate for an Economic Review, this section
concentrates on econom ic intuition instead.
The first problem concerns how the regulator
(sender) tries to influence the receiver. In the par­
tition example, if the regulator announces that it
prefers medium risk, the bank guesses that

m = ~ (because ^

1 ] = | ) an<^ chooses a

risk level of y = § + ^ = § •This response may
tempt the regulator into announcing a "revised”
m essage o f "m is in the interval (y-j’ ^)- If the
bank reasons as before, this will lead to a risk

announces that m is in the interval

bank may believe, "Things are totally fouled
up. W e’d better assume that m =




Such a b e­

lief will once again allow the partition equilib­
rium to exist. That is, the regulator realizes that
any deviation from the standard announcement
could lead to an undesirably large change in
bank expectations. In this case, because the
bank becom es too conservative, it would be
better for the regulator to stay with its original
three announcements.
Another critical assumption is that the regula­
tor faces only one bank, or a completely homo­
geneous banking system that acts like one bank.
If, instead, many banks each have different pre­
ferred risk levels (bj s), problems can once again
arise. In this case, if the regulator makes an unex­
pected announcement, the average of the poten­
tially different responses may lead to a sm(X)th
response. Any big shifts get averaged out. and
the equilibrium again unravels.10
Put another way, with a large audience, the
sender has an incentive to "fine tune” the aver­
age audience reaction. This leads receivers to
attempt to offset the anticipated fine tuning,
and communication breaks down.

IV. Conclusion
He was a pow er politically fe r years, but
b e never got prom inent enough t ’ have bis
speeches garbled.
— Abe Martin.

level o f y =
The bank may not reason as before, how ­
ever. The original partition equilibrium defined
the ranges, but w hat if the sender changes the
announced range? What does the bank believe
when the regulator does something unexpected?
This puts the econom ist in the uncomfortable
position o f playing psychologist. It also makes
the ultimate result somewhat uncertain. For ex ­
ample, if the bank recognizes what the regula­
tor is doing with the revised announcement, it
will shade its choice o f y somewhat higher,
the regulator will shade the interval lower, and
the partition equilibrium will break down. As
the originator o f this critique explains, "The
cheap-talk equilibrium breaks down entirely if
small differences in government announce­
ments can cause only small differences in pub­
lic expectations” (Conlon [19941, p. 420).
An unexpected announcement can have
various consequences.9 When the regulator

| ), the

M2 6

Abe Martin s Sayings an d Sketches
How much detail a government should com ­
municate to its citizens remains controversial,
especially in the areas o f m oney and banking.
On many issues, the gov ernment com m unicates
to foster coordination with the public. There
are simply som e things it is useful for citizens
to know, and the government tells them. In
other cases where interests may not align ex ­
actly, communication cannot always b e both
precise and credible. Vagueness and secrecy
present o n e way around the problem by allow ­
ing partial communication.
The conflict between credibility and precision
suggests that pressuring an agency to release
information may not always be productive. Re­
leasing bank regulators’ m eeting notes or

■ 9 This is the problem of multiple equilibria, mentioned in footnote 4
■

10

See Conlon (1992) The detailed argument is quite complex

19

videotaping FOMC deliberations will most
likely result in reports and videotapes display­
ing the lamented vagueness o f current official
releases. The partition equilibrium remains the
optimal solution to the problem facing the
government and the public; videotaping will
not change the trade-off betw een vagueness
and credibility.
Pressure may result in truthful, precise an­
nouncements if it leads to an appropriate
change in institutional structure. The change
must som ehow further align the interests o f
the two parties or introduce a credible commit­
ment mechanism. Less drastic changes, per­
haps occurring as agencies com e to grips with
the trade-offs involved, may alter the amount
o f information released. The FOMC's recent
policy announcem ents are a case in point .11
These conclusions should be treated with a
healthy skepticism, however. As w e have seen,
further examination o f the econom ic issues re­
veals that the benefits o f vagueness may be
sensitive to particular modeling assumptions.
Cheap talk represents an intriguing, but not
entirely compelling, justification for imprecise
policy announcements.

References
Buser, Stephen A., Andrew H. Chen, and Edward
J. Kane. “Federal Deposit Insurance. Regula­
tory Policy, and Optimal Bank Capital," Jo u r­
nal o f Finance, vol. 35. no. 1 (March 1981),
pp. 51 - 60.
Conlon, John R. “Robustness o f Cheap Talk
with a Large Audience,” University o f Missis­
sippi. Department o f Economics and Finance,
Working Paper, Ju n e 1992.
_________. "Can the Government Talk Cheap?
Communication, Announcements, and
Cheap Talk." Southern Economic Journal,
vol. 60, no. 2 (O ctober 1994), pp. 4 1 8 - 2 9 .
Crawford, Vincent P., and Jo el Sobel. “Strategic
Information Transmission." Econometrica, vol.
50. no. 6 (November 1982), pp. 1 4 3 1 -5 1 .

Federal Reserve Bank o f Cleveland. Circular Let­
ter 9 5 -3 3 , March 10, 1995.
Goldberg, Samuel. Introduction to Difference
Equations. New York: Joh n Wiley & Sons,
1958.
Goodfriend, Marvin. "Monetary Mystique: Secrecy
and Central B a n k in g Journal o f Monetary
Economics, vol. 17, no. 1 (January 1986),
pp. 63 - 92.
Hansell, Saul. “Big Bank Goals: Higher Ratings,"
New York Times, Ju n e 8. 1993.
Kane, Edward J. “Politics and Fed Policymak­
ing: The More Things Change the More
They Remain the Sam e,” Jo u rn al o f Mone­
tary Economics, vol. 6, no. 2 (April 1980),
pp. 1 9 9 - 2 1 1 .
Karr, Albert R., and Becky Gaylord. “New Guide­
lines to Toughen Monitoring o f Derivatives
Transactions by Banks," Wall Street Journal.
O ctober 24, 1994.
Matthews, Steven A., Masahiro Okuno-Fujiwara,
and Andrew Postlewaite. “Refining CheapTalk Equilibria,” Jou rn al o f Economic Theory,
vol. 55, no. 2 (D ecem ber 1991), pp. 2 4 7 - 7 3 McLemore, Joel. "What Bank Examiners Are
Guilty o f — and Aren't," Wall Street Journal,
D ecem ber 5, 1991.
Spence, Michael. “ b Market Signaling," Quar­
Jo
terly Jou rn al o f Economics, vol. 87, no. 3
(August 1973)'. pp. 3 5 5 - 7 4 .
Spong, Kenneth. Banking Regulation: Its Pur­
poses, Implementation, a n d Effects, 3d ed.
Federal Reserve Bank o f Kansas City, 1990.
Sprague, Irvine H. Bailout: An Insider's Ac­
count o f B an k Failures an d Rescues. New
York: Basic Books, 1986.
Stein, Jerem y C. “Cheap Talk and the Fed: A
Theory o f Imprecise Policy Announcements,”
American Economic Review, vol. 79, no. 1
(March 1989), pp. 3 2 - 4 2 .
Tomasula, Dean. “BT Is Sued for S130M by
P&G in Swaps Deal," American Banker,
O ctober 28, 1994.

■

11 In the first quarter of 1995. the Federal Reserve adopted a pol­
icy of announcing changes in the stance of monetary policy the day they
are made For details, see Federal Reserve Bank of Cleveland (1995).




20

Federal Funds Futures
as an Indicator of Future
Monetary Policy: A Primer
by John B. Carlson, Jean M. Mclntire, and James B. Thomson

John b . Carlson is an economist,
Jean M. Mclntire is a senior re­
search assistant, and James B.
Thomson is an assistant vice
president and economist at the
Federal Reserve Bank of Cleve­
land. The authors thank Charles
Carlstrom, Jagadeesh Gokhale,
Joseph Haubrich, Spence Hilton,
Peter Rupert, and E.J. Stevens for
helpful comments.

Introduction
Monetary policy attracted considerable media
attention in 1994. The focus was largely con ­
centrated on the six increases in the federal
funds rate objective during the year. The fed
funds rate is the interest rate banks pay w hen
they borrow Federal Reserve deposits from
other banks, usually overnight. It is closely
w atched in financial markets because the level
o f the funds rate can be immediately and pur­
posefully affected by Federal Reserve open
market operations.
The Federal O pen Market Committee (FOMC),
the main policymaking arm o f the Federal Re­
serve System, communicates an objective for the
fed funds rate in a directive to the Trading Desk
(hereafter Desk) at the Federal Reserve Bank of
New York. Actions taken to change an intended
level o f the fed funds rate are motivated by a de­
sire to accomplish ultimate policy objectives, es­
pecially price stability. Permanent changes in the
fed funds rate level are thus the consequence of
deliberative policy decisions .1
Although the Desk does not achieve the in­
tended funds rate on a daily basis, it effectively
does so on average. Figure 1 illustrates the ef­




fective fed funds rate daily over the past six
years relative to the intended rate .2 The an­
nualized effective yield varies substantially on
a daily basis, but the monthly average rate is
generally close to the rate specified by the
FOMC. Since O ctober 1988, the mean absolute
deviation o f the monthly average o f the fed
funds rate from the intended level has been
less than six basis points (six one-hundredths
o f a percent).
Because the average monthly fed funds rate
remains close to the intended level (and hence
is independent o f permanent market influences),
it is unique among other short-term rates. Thus,
predicting what the average monthly rate will
b e in the future is tantamount to predicting
what the fed funds rate objective will be over
the course o f the month.
In 1988, the Chicago Board o f Trade began
trading an interest-rate futures contract based
on average monthly fed funds rates. This co n ­
tract, known as the 30-day fed funds futures
■ 1 Indeed, over most of the post-World War II period, the fed funds
rate or its equivalent has been the Fed's policy instrument

■ 2 The daily effective rate is based on a survey of the transactions
arranged through five fed funds brokers.

21

F I G U R E

1

Daily Fed Funds Rate
and Intended Level
Percent

SOURCE: Chicago Board of Trade.

contract, may be written for any calendar month
up to 24 months ahead. The market price o f fed
funds futures essentially embodies a prediction
o f the monthly average o f the daily fed funds
rate. Because markets understand that deviations
of the overnight funds rate from its desired level
tend to average out over the span o f a month,
the implied rate is essentially the market’s expec­
tation of the intended rate. Thus, the FOMC can
assess in fairly precise terms what the markets—
at least the fed funds futures market— believe its
actions will be.
This paper exam ines the predictive content
o f the fed funds futures contract and considers
som e policy implications. The next section de­
scribes the fed funds market and how the funds
rate is determined. W e exam ine how closely
the average monthly rate matches the monthly
average o f the intended rate. In sectron II, w e
describe the fed funds futures instrument and
market. In section III, w e exam ine the predic­
tive accuracy o f the implied fed funds futures
rates and com pare it with alternative forecasts.
W e offer policy implications and som e co n ­
cluding remarks in sections IV and V.

I. The Fed
Funds Market
Participants in any futures market have every in­
centive to understand the fundamental determi­
nants o f the price of the commodity or financial
instrument on which the futures contract is
drawn. Perhaps the most striking example o f this
 is illustrated by Roll (1984), who examines the


market for frozen orange juice futures. The supply
of frozen orange juice is highly “concentrated” in
the sense that 80 percent o f the oranges typically
used come from Orange County, Florida. Because
frost can destroy a large share of the market, fro­
zen orange juice futures prices are clearly highly
sensitive to changes in weather. Indeed, Roll
shows that these futures prices can be used to
provide weather forecasts for Orange County that
are marginally superior to the forecasts of the
National Weather Service .3
Exogenous factors, such as bad weather,
can also affect the daily average funds rate by
creating payment delays and hence float, but
such effects are transitory and tend to average
out on a monthly basis. Moreover, the Desk
monitors float closely and stands ready to en ­
ter the market to offset any anticipated effects.
Nevertheless, unanticipated float and other
daily factors can influence monthly average
rates and hence lead to marginal deviations
from the monthly average funds rate specified
by the objective.
To appreciate better the unique forces at
play in the fed funds market, it is useful to re­
view som e critical characteristics o f fed funds
and the determinants o f the fed funds rate .4
Goodfriend and W helpley (1993) identify three
features that, taken together, distinguish fed
funds from other money market instruments.
First, they are borrowings o f immediately avail­
able m oney — funds that can b e transferred b e ­
tw een depository institutions in a single day.
(About three-quarters o f fed funds in 1991
w ere overnight borrow ings.) Second, fed funds
can b e borrow ed only by those depository in­
stitutions that are required to hold reserves
with Federal Reserve Banks. Third, fed funds
borrowings are exem pt from reserve require­
ments and interest-rate ceilings .5
The fed funds market serves as an effective
device to redistribute reserves in the banking
system. For exam ple, som e banks, typically
large ones with w ide access to financial mar­
kets, find themselves persistently in need o f
reserves. O ther banks, typically small ones
with m ore limited investment opportunities,

■ 3 Price is a slightly better predictor of the error of the National
Weather Service Forecast at 5:00 a m than of the forecast made the pre­
vious night or that same night (Roll [1984], p. 871).

■ 4 For a more complete description of the fed funds instrument and
market, see Goodfriend and Whelpley (1993).

■ 5 Reserves refer to bank assets held in the form of vault cash and de­
posits at Federal Reserve Banks. Reserve requirements, on the other hand,
are the amount of assets that must be held as reserves against a liability.

22

F I G U R E

2

Timing of Contemporaneous
Reserve Accounting System

Reserve
accounting for
liabilities
A cco u n tin g
for reserv es

SOURCE: Authors, adapted from Meulendyke (1989).

have a persistent surplus o f reserves. Although
banks may lend reserves directly to each other
through their correspondent relationships,
about 40 percent o f total fed funds transactions
in 1991 w ere arranged through brokers, with
the remainder purchased directly from counter­
parties .6 Moreover, as payments flow through
the banking system, individual banks face wide
swings both in their reserve balances and in
their reservable deposits. The fed funds market
thus also provides a convenient outlet in which
banks can buy or sell reserves to offset the an­
ticipated and unanticipated impact o f payments
on their reserve positions.
While the actions o f individual banks within
the fed funds market may effectively redistrib­
ute reserves to w here they are most needed in
the banking system, they do not affect the ag­
gregate supply o f reserves, which is determined
by Desk actions and market factors outside the
control o f individual banks and the Desk. The
demand for reserves in the aggregate is criti­
cally dependent on the nature o f reserve re­
quirements, especially the reserve accounting
system, on the public’s demand for reservable
deposits, and on bank funding decisions.
Required reserves are computed as fractions
o f daily average deposit levels, as specified in
Regulation D. (Since D ecem ber 1990, only
transactions deposits have been reservable.)
Under the current reserve accounting system,
daily average deposit levels are based on a
two-week computation period beginning every'
other Tuesday (see figure 2 ).
Although banks may ultimately affect the de­
mand for their transactions accounts (and hence
required reserves) by raising or lowering the in­

terest rate paid, depositors typically respond


with a lag. In fact, within the span o f the re­
serve computation period, the effect on depos­
its demanded is negligible; hence, the level o f
required reserves is largely predetermined.
The time interval over which daily average
reserves must equal or exceed computed re­
quired reserves — called the reserve mainte­
nance period — is specified as a tw o-week
period beginning two days after the start o f the
reserve computation period. Total reserves con ­
sist o f depository institutions’ deposits at Fed­
eral Reserve Banks net o f contractual clearing
balances and applied vault cash .8 It is within
the reserve maintenance period, then, in
which demand must equal supply (that is,
w hen the market must clear).
The ultimate supplier of reserves is o f course
the Federal Reserve, which provides reserves
through either open market operations or dis­
count window lending. Since the demand for
reserves is essentially predetermined over the re­
serve computation period, the operating prob­
lem faced by the Federal Reserv e is how much
reserves it will supply through open market op­
erations .9 This decision essentially determines
the equilibrium level o f the fed funds rate.
The operating procedure is com plicated by
the fact that the Desk does not know precisely
what the levels o f required reserves will be nor
the demand for reserves in excess o f required
holdings. It must estimate them daily as new in­
formation becom es available. Moreover, b e­
cause discount window borrowing occurs at
the volition o f banks, the Desk does not know
what the level o f borrowing will be. The level
o f discount window borrowing, however, is re­
lated to the spread betw een the fed funds rate

■6

A correspondent relationship is one in which one bank (corre­
spondent) holds the deposits of another (respondent). Large banks often
act as correspondent banks tor smaller banks because they may have ac­
cess to a variety of services not directly available to the smaller banks.
For example, small banks may choose to hold deposits with the large
bank, which in turn provides payment services. Because respondent de­
posits are reservable, large banks typically find themselves in need of re­
serves, while small banks typically hold a surplus. Thus, respondent
banks may lend their excess reserves directly to their correspondent, but
also sell them in the fed funds market

■ 7 See, for example, Federal Reserve Bulletin, vol. 81, no. 1 (Janu­
ary 1995), table 1.15, p.A9.

■

8 Applied vault cash equals average vault cash over a two-week pe­
riod beginning 30 days before the end of the reserve maintenance period
Thus, applied vault cash is determined before required reserves are known
■ 9 Total reserve demand equals required reserves over the computa­
tion period plus the demand for reserves in excess of required reserves
(which are also largely predetermined)

23

T A B L E

1

Deviation of Monthly Average Fe
Funds Rate from Intended Level
(percent)
Mean
Deviation

M ean Absolute
Deviation

1988-1994

0.04

0.06

1992-1994

0.03

0.05

SOURCES: Chicago Board of Trade; and authors' calculations.

B 0 X 1

Fed Funds Futures
Market Terminology
Open interest

Total number of contracts outstanding on a
given day.

Volume

Daily volume in number o f contracts traded.

Settlement price

Official price set by the exchange at the end
of the day to determine daily gains and losses
and margin requirements.

Derivative

Security whose value depends on the value
of underlying simpler securities.

Futures contract

Agreement between two parties to buy or sell
an asset at a future date at a specified price.

Fed funds
market

Collective interbank borrowing and lending
activities designed to maintain required re­
serve ratios.

Fed funds
effective rate

Average daily rate on overnight fed funds as
reported by the Federal Reserve Bank of New
York.

Trading unit

S5 million overnight fed funds held for a
minimum of 30 days.

Price bias

Settlement price calculated as 100 minus the
monthly average overnight fed funds rate.

Hedging

Taking a position that is equal and opposite
to the risk exposure relative to a market
position in an attempt to offset any losses
incurred by the underlying position by gains
in the future position.

and the discount rate, so initial estimates are
obtained for the desired spread.
At the beginning o f a maintenance period,
the Desk projects reserve needs based on esti­
mates of required reserves, excess reserves, and
discount window borrowing. It formulates a pro­
gram to add or absorb reserves smoothly over
the course o f the two-week period. It also esti­
 mates the effect o f market factors on the level of


nonborrowed reserves. As the period unfolds,
the Desk continually monitors the appropriate­
ness of its estimates and revises its program for
reserves provision accordingly. Over the course
of the maintenance period, it is also guided by
the behavior o f the fed funds rate. For exam­
ple, if the rate is persistently above its desired
level, the Desk may choose to supply more re­
serves than the program calls for.
Although the fed funds rate may swing
widely from day to day, the Desk’s actions are
generally successful in achieving its objective on
average. Table 1 presents the monthly average
and the mean absolute deviation o f the daily fed
funds rate from its intended level since October
1988. The funds rate over this period tended to
be only three basis points above, and the mean
absolute deviation only about six basis points
above, its target level. Thus, the Desk achieves
its objective rather closely on a monthly average
basis. Over the same period, the daily funds rate
ranged between 10 percent and 3 percent. The
key determining factor in this movement is the
deliberative policy choice o f the FOMC.

II. The Fed Funds
Futures Contract
The fed funds contract, also known as 30-day
fed funds futures, calls for delivery o f interest
paid on a principal amount of $5 million in over­
night fed funds (see box l ).10 In practice, the
total interest is not really paid, but is cash-settled
daily. This means that payments are made when­
ever the futures contract settlement price changes.
The futures settlement price is calculated as 100
minus the monthly arithmetic average o f the daily
effective fed funds rate that the Desk reports for
each day of the contract month.
To illustrate, consider the situation in which
a bank sold 10 December contracts at 94.42 just
before the market’s close on O ctober 4, 1994.
This was the contract’s price around market
closing as reported in The Wall Street Jou rn al
the following day (see the fourth [“settle”] col­
umn in table 2). It em beds a market expecta­
tion o f a D ecem ber fed funds rate o f 5.58
percent (that is, [100 - 5.58] = 94 .4 2 ). For
deferred-m onth contracts, such as the D ecem ­
b er contract, each basis-point ( 0.01 p ercent)
chan ge cau ses the price o f the contract to
m ove by o n e tick, or $41.67 (that is, 0.01
percent tim es [30 d ays/360 days] tim es $5

10 See Chicago Board of Trade (1992).

24

T A B L E

2

Interest Ratea

Open

10/94
11/94
12/94
1/95

94.98
94.78
94.44
94.28

High

94.99
94.78
94.44
94.28

Low

94.96
94.78
94.41
94.27"

Lifetime
Open
Settle Change High
Low Interest

94.96
94.76
94.42
94.27

0.02
0.02

95.63 94.63
95.52 94.50

4,392
3,779

0-03

96,00 94.41

1,082

0.02

94.66 94.24

162

a. 30-day federal funds (Chicago Board of Trade) - $5 million; pts. of 100
percent.
SOURCE. Wall StreetJournal, October 5, 1994.

m illion ).11 Thus, if the D ecem ber settle price
rises to 94.45 on O ctober 5, the seller o f the
contract ow es the contract holder $1,250.10
($41.67 times three ticks times 10 contracts).
Payments are made through margin accounts
that sellers and holders have with their b ro­
kers. At the end o f the trading day, sellers’
and holders' accounts are debited or credited
to facilitate payments.
Fed funds futures are a convenient tool for
hedging against future interest-rate changes.
To illustrate, consider a regional bank that con ­
sistently buys $100 million in fed funds. Sup­
pose the bank’s analysts believe that econom ic
data to be released in the upcoming w eek will
induce the FOMC to increase the objective o f
the fed funds rate by 50 basis points at its next
meeting. If the contract settle price (for the
meeting m onth) implies no change from the
current rate, the bank may choose to lock in
its current cost by selling 20 contracts (or tak­
ing a short position) and holding the position
to expiration. Conversely, suppose that a net
lender o f funds expects a policy action to
low er the fed funds rate. It can protect its re­
turn by buying futures contracts (or taking a
long position).

its ability to achieve its objective, the conse­
quences for the funds rate may be predictable.
Speculators w ho anticipate such effects may
find it profitable to buy or sell current contracts.
Figure 3 illustrates the monthly average of
both the number o f outstanding contracts
(open-interest) and the volume for each o f the
six contracts studied. Although it reveals that
the market has grown appreciably in a rela­
tively short time, this growth has not been
shared equally among contracts o f various du­
rations. For exam ple, open interest has trended
upward for contracts o f less than four-months’
duration, while it peaked in late 1992 and then
receded for the four- and five-month contracts.
Current-month and one-month contracts are
most heavily traded throughout the period .12
Two-month and three-month contracts have
also enjoyed active trading; however, when
the length o f the contract extends beyond this
point, trading activity diminishes. Indeed, the
monthly average volume in the five-month
market has rarely exceeded 100 contracts. The
market for four- and five-month-ahead con ­
tracts peaked in 1993 after the fed funds rate
had plateaued at its cyclical low. Contracts
over five months long do exist, but their ap­
pearance is sporadic.

III. Predictive
Accuracy
Figure 4 illustrates monthly average futures
rates and the corresponding forecast errors
since O ctober 1988 (w hen contracts w ere first
traded) for each o f the contract horizons. Not
surprisingly, predictive accuracy diminishes as
the contract horizon is extended. Also, errors
tend to be relatively large when the funds rate
changes direction or when it changes rapidly
over a short period. Neither the 1989 peak in

Participants in the fed funds futures market
need not be banks that borrow in the fed
funds markets. Anyone w ho can satisfy margin
requirements may participate. Thus, traders
w ho make their living as “Fedwatchers” may
speculate with fed funds futures. This would
suggest that to the extent Fed policy is predict­
able, speculators would drive futures prices to
em body expectations o f future policy actions.
Since the level o f the fed funds rate is essen­
tially determined by deliberative policy deci­
sions, the fed funds futures rate should have
predictive value for the size and timing o f fu­
ture policy actions. Moreover, given that the
Desk may face systematic problems that hinder




B

11

Although December has 31 days, a 30-day-month standard is

used to define a constant tick size. Also, the structure of current-month
pricing is different from deferred-month pricing in that the price of the
current contract reflects a day-weighted average of the rate experience to
date and the implied term rate to the end of the month. Contracts are
listed on the Chicago Board of Trade exchange for the current month and
for each of the 24 months that follow.
■ 12 However, on a daily basis, current-month volume often drops
below one-month volume given the dramatic decline in the number of
contracts generally associated with trading during the final days of the
month. At the same time, there is an opportunity for arbitrage as trading
forces the convergence of the futures price with the spot price as the con­
tract approaches maturity. As the closing price becomes a virtual cer­
tainty, the incentive to place a bet on the settlement price declines as
speculative profits are reduced to zero.

25

F I G U R E

3

Size of Market
Number of contracts

Number of contracts

Number of contracts

Number of contracts

Number of contracts

Number of contracts

Number of contracts

Number of contracts

Number of contracts

Number of contracts

SOURCE: Chicago Board of Trade.

the funds rate nor the policy turnaround in
February 1994 was anticipated at any contract
horizon. Nor did the market adequately fore­
see the sequence o f funds-rate reductions initi­
ated in m id-1990 and again in 1991.
That the fed funds futures market failed to
anticipate these episodes may not be all that
damning. Because such decisions are often
based on information that surprises both fore­
casters and policymakers alike, there may be
no way to predict the timing o f such events.
Moreover, the market may b e dominated by
hedgers, w ho seek to reduce risk rather than
 to speculate on a projected change. The uncer­


tainty surrounding the response o f policy may
be too great for som e speculators to act on the
projection. That is, the expected rate o f return
may not be sufficient to com pensate for the
level o f risk to which the position is exposed.
One might expect that the current month’s
futures rate would be a good predictor o f the
month’s fed funds rate. Alter all, by the middle of
the month, the market already knows half of the
daily rates used in the monthly average calculation.
Moreover, as time moves on, more information
relevant to policy decisions becomes available,
which in turn should enhance the predictive per­
formance of a given contract. For example, one

26

F I G U R E

4

Fed Funds Futures Rate as a
Predictor of the Effective
Fed Funds Rate

Percent

Percent

Percent

Percent

Percent

Percent

a. Line breaks reflect periods during which no contracts were traded.
SOURCES: Chicago Board o f Trade; and Board o f Governors o f the Federal Reserve System.

would e x p e a the probability o f an unantici­
pated shift in Fed policy to diminish as the ex­
piration date of a contract approaches.
T o exam ine how readily the futures market
incorporates available information into its pric­
ing decisions, w e estimate the mean absolute
deviation betw een the daily rate and the co n ­
tract standard for each o f the trading days lead­
ing up to the expiration date. In principle, if
the market is efficient, the mean absolute de­
viation should diminish. Figure 5 illustrates that
the deviation declines steadily as the expiration




date approaches. Indeed, the mean absolute
deviation is virtually zero by the last trading
day. Moreover, the mean absolute deviation
averaged over the month is less than six basis
points, about the same as the mean absolute
deviation o f the fed funds rate from its monthly
average intended rate. This suggests that all
systematic variation in the fed funds rate is an­
ticipated by the market and incorporated into
the future’s price. If the fed funds futures mar­
ket w ere not incorporating all the information
about future fed funds yields, one might expect

27

F I G U R E

5

Mean Absolute Deviation of Future
from Effective Funds Rate

Basis points

Trading days remaining in calendar month
SOURCES: Chicago Board of Trade; and authors' calculations.

B 0 X 2

Alternative Forecasting Models
Naive Model (random walk)

where r t is the effective fed funds rate and £, is a random
disturbance.

Univariate Model (estimated)
A rt = 0.011 + 0.367 A r(. , + e,
(0.046) (7.912)
w here A rt is the first difference o f the effective fed funds rate
and e, is an independent, identically distributed (i.i.d.) ran­
dom disturbance. The equation is estimated from Septem ber
1954 to Septem ber 1988.

that the mean absolute deviation o f the futures
rate would materially exceed that o f the fed
funds rate from its intended level .13
Market participants are clearly able to im­
prove their estimates o f the current month’s
average as the month progresses. What's more
impressive is that the predictive accuracy o f
the one-m onth-ahead futures rate also im­
proves over the period leading up to the end
o f the prior month. The mean absolute error

on the
http://fraser.stlouisfed.org/ last day o f the previous month is about
Federal Reserve Bank of St. Louis

one-third lower than the mean absolute error
20 days earlier. The only exception to this
trend occurs for a few days in the middle o f a
calendar month. Nevertheless, the predictive
performance is not significantly diminished.
To the extent that the fed funds futures market
is efficient, contract rates should predict fed funds
rates at least as well as alternative forecasting
models. As a preliminary investigation o f market
efficiency, we compare the prediction errors of
fed funds futures with those o f a naive model
and an estimated univariate model (first-order
autoregressive model of the change in the fed
funds rate).14 The naive model simply assumes
that the best forecast of the future fed funds rate
is the current rate (see box 2). This model is
sometimes called a random walk because it im­
plicitly assumes that changes to the fed funds
rate are random and permanent. The univariate
model also assumes that changes to the level of
the fed funds rate are permanent, but it allows
for some persistence o f the change. That is, if a
change occurs in one period, it can occur again
(at least partially) in the subsequent period.
Table 3 presents the mean prediction error
(MPE) and the mean square error (MSE) for each
of six forecast horizons and for each of the alter­
native forecasting approaches .15 The prediction
error is defined as the forecast less the actual
(monthly average effective fed funds rate). All
three approaches tended to overpredict over the
whole period. The bias was uniformly larger for
predictions based on fed funds futures rates, the
only exception being for the five-month-ahead
horizon. This suggests that fed funds futures pric­
ing may be dominated by consistent borrowers
of overnight funds who are willing to pay a pre­
mium to hedge against the risk o f interest-rate in­
creases .16 Given the limited sample, however, it
may be too early to draw such a conclusion .17
■

13 Both measures of variability are small.

■

14 This model, an ARIMA (1,1,0), was identified using a method
proposed by Box and Jenkins (1970).

■ 15 The /'-month-ahead prediction errors for the futures contract are
simply the difference between the futures rate on the /-month-ahead con­
tract and the average of the fed funds rate for the same month. All con­
tract rates are averaged over the month that they are recorded.

■ 16 One might ask why this premium exists. It is possible that trans­
actions costs may preclude any profitable strategy to exploit the premium.

■ 17 Indeed, Spence Hilton at the Trading Desk of the Federal Reserve
Bank of New York believes that the prediction bias may be a quirk of the sam­
ple period. He notes that over most of the sample period, the market (as well
as the F0MC) was surprised by the lack of strength in the economy. The
F0MC often responded to evidence of economic weakness by lowering the
fed funds rate immediately. He believes that this experience could dominate
the average prediction error given the limited sample.

28

T A BL E

3

Relative Predictive Accuracy
of Fed Funds Futures
Panel A. W h ole F o re ca st Period (O c to b e r 1 9 8 8 -D e c e m b e r 1 9 9 4 )
Federal Funds Fu tu res
F o recast H orizon

MPEa

MSEb

Current
One month ahead
Two months ahead
Three months ahead
Four months ahead
Five months ahead

0.01
0.06
0.10
0.17
0.25
0.26

0.00
0.03
0.09
0.20
0.36
1.62

Naive M odel
MPE

Univariate Model
MSE

MSE

—
0.05
0.17
0.33
0.54
0.77

—
0.04
0.08
0.14
0.20
0.27

MPE

—
0.03
0.08
0.15
0.22
0.30

—
0.04
0.13
0.27
0.48
0.72

Panel B. S econd Half o f F o recast Period (after 1 9 9 1 )
Federal Funds Fu tu res
F o recast H orizon

Current
One month ahead
Two months ahead
Three months ahead
Four months ahead
Five months ahead

MPEa

0.01
0.06
0.10
0.13
0.18
0.23

Naive M odel

Univariate Model

MSEb

MPE

MSE

MPE

MSE

0.01
0.03
0.06
0.11
0.18
0.29

—

—
0.04
0.12
0.22
0.38
0.75

—
-0 .0 1
-0 .0 2
- 0 .0 0
0.03
0.08

0.03
0.09
0.16
0.31
0.49

-0 .0 4
- 0 .0 7
- 0 .0 6
- 0 .0 5
-0 .2 2

—

a. Mean prediction error.
b. Mean square error.
SOURCE: Authors' calculations.

Although alternative models may provide
less-biased predictions than the fed funds fu­
tures, investment strategies based on the mod­
els would be more risky. This is evident when
comparing the MSEs o f alternative forecasts.
The MSE provides a measure o f the dispersion
o f forecast errors and hence o f the uncertainty
associated with the prediction. In all but one
case, the MSE o f the fed funds futures predic­
tion is less than the alternatives .18 Thus, al­
though the average gain could be greater for
alternative predictions, the potential for losses
is also higher.
Because the fed funds futures market is
young and the volume o f trades is small relative
to some other comparable instruments (for exam­
ple, Eurodollar futures), one might question
whether the market is “deep” enough to accom­
modate large trades. If the market is deep, large
trades should not appreciably affect market rates
unless they reflect the incorporation o f new infor­
mation in futures prices. To assess the potential
relevance of this issue, w e exam ine whether
the increased volume o f the market has led to
better predictions. The second panel in table 3
presents the MPE and MSE statistics for the pe­
riod since 1991. These results reveal that the




dispersion o f forecast errors declines sharply
for horizons o f three months or more. How­
ever, the improved predictive performance
over the latter period may reflect the fact that
the fed funds rate was relatively more stable
over this period.
In sum, the preliminary' evidence presented
above suggests that fed funds futures are use­
ful for predicting future fed funds rate changes
(and hence policy moves), especially over the
shorter forecast horizons. Prediction error is
shown to diminish almost daily leading up to a
contract’s expiration date. The fact that the
MSEs o f fed funds futures predictions are rela­
tively small provides som e evidence that fed
funds futures markets efficiently incorporate in­
formation into pricing decisions.

■ 18 The only exception is lor the five-month-ahead futures contract,
which was not actively traded over the first three years of the market

29

IV. Some Policy
Implications
The fed funds futures rate, by virtue of being a
market-determined expectation about future de­
liberative actions, provides potentially useful infonnation for Fed policymakers. For example,
the FOMC may find the futures rate helpful in
assessing the credibility o f alternative policy
choices. To illustrate, consider a situation in
which financial markets clearly perceive increas­
ing inflationary pressures and e x p e a the FOMC
to counter with a fed funds rate increase.
A key market concern may be that the FOMC
must demonstrate sufficient resolve to ensure
that short-term objectives — such as interest-rate
smoothing — do not interfere with the achieve­
ment o f longer-tenn price stability. Under these
circumstances, the absence o f an anticipated ac­
tion could induce expectations o f rising inflation
and in turn becom e embedded in longer-term in­
terest rates as increased inflation premia. Thus, if
the market expects an anti-inflationary move, the
FOMC may feel compelled to act even if it be­
lieves inflationary pressure will ebb so as to pre­
vent a flare-up of inflationary expectations.
To what extent should the FOMC react to fed
funds futures as a signal o f expectations regard­
ing future changes? In principle, participants in
the fed funds futures market will base their trad­
ing decisions on expectations o f the fed funds
rate path they believe the FOMC will choose
over time. If the FOMC were to base its decision
solely on the market’s expectation, it is not clear
what would ultimately determine the fed funds
rate path. That is, the equilibrium outcome of
such a policy may be indetenninate. This prob­
lem is described by Keynes (1936, p. 156) in an
analogy with newspaper competitions:
... the competitors have to pick out the six prettiest
faces from a hundred photographs, the prize being
awarded to the competitor w hose choice most
nearly corresponds to the average preferences of
the competitors as a whole; so that each competi­
tor has to pick, not those faces which he himself
finds prettiest, but those which he thinks likeliest to
catch the fancy of other competitors, all of whom
are lcx)king at the problem from the same point of
view . It is not the case of choosing those which, to
the best of one’s judgment, are really the prettiest,
nor even those which average opinion genuinely
thinks is the prettiest. We have reached the third
degree where we devote our intelligences to antici­
pating what average opinion expects average opin­
ion to be. And there are soriK*. I believe, who
practise the fourth, fifth and higher degrees.



O ne cannot know at which degree participants
choose to make their decision; hence the inde­
terminacy.
The FOMC, o f course, does not base its deci­
sion solely on what the market expects it to do, as
clearly evidenced by the failure of the fed funds fu­
tures market to anticipate turning points in fed
funds rates. Rather, the FOMC looks at many
things, and bases its decision on the majority’s
assessment o f the fed funds rate level needed to
accomplish ultimate objectives. In this context,
however, the Committee may find knowledge
of market expectations useful in assessing the
financial-market consequences o f alternative ac­
tions. For example, the estimated impact o f any
given action may differ depending on whether
the policy change is anticipated by the market.
Thus, fed funds futures rates are helpful as part
o f an array of indicators considered by the
FOMC in its policy deliberations.

V. Concluding
Remarks
Futures contracts are typically drawn on com ­
modities or financial instruments w hose price
or yield is determined in competitive markets.
In the case o f fed funds, however, the rate is
essentially determined by a deliberative deci­
sion o f the FOMC, the main policymaking arm
o f the Federal Reserve System. Hence, the fed
funds futures market must anticipate actions
taken by the FOMC. In short, through the fed
funds futures market, one can place a bet on
what future monetary policy will be. The Com­
mittee then can get a clear reading o f what
these market participants expect them to do,
which may at times be helpful for FOMC mem­
bers who place great weight on knowing if a
policy choice would surprise the market.
If they are to be instructive for policymakers,
fed funds futures rates should have some predic­
tive content. The predictive accuracy of futures
rates clearly improves over the two-month pe­
riod leading up to the contract’s expiration, pro­
viding some evidence that the market is efficient
in incorporating new information into its pricing.
The largest prediction errors occur around policy
turning points. Nevertheless, the evidence above
suggests that the fed funds futures markets are ef­
ficient processors of information concerning the
future path o f the fed funds rate.

30

References
Chicago Board of Trade. 30-Day Fed Funds Fu­

tures-. Flexible Futures f o r Managing Risk.
Chicago: Board o f Trade o f the City o f
Chicago, 1992.
Box, George E.P., and Gwilym M. Jenkins. Time
Series Analysis: Forecasting a n d Control. San
Francisco: H olden-Day, 1970.
Goodfriend, Marvin, and William Whelpley.
“Federal Funds,” in Timothy Q. Cook and
Robert K. LaRoche, eds., Instruments o f the
Money Market. Richmond: Federal Reserve
Bank o f Richmond, 1993, pp. 8 - 2 1 .
Keynes, John Maynard. The General Theory o f
Employment, Interest, a n d Money. New
York: Harcourt, Brace, and World, 1936.
Meulendyke, Ann-Marie. U.S. Monetary Policy
a n d Financial Markets. New York: Federal
Reserve Bank o f New York, 1989.
Roll, Richard. “Orange Ju ice and W eather,”
American Economic Review, vol. 74, no. 5
(D ecem ber 1984), pp. 8 6 1 -8 0 .




31

First Quarter
Working Papers
Current Working Papers of the
Cleveland Federal Reserve Bank
are listed in each quarterly issue
of the Economic Review. Copies
of specific papers may be re­
quested by completing and
mailing the attached form below.

Single copies of individual pa­
pers will be sent free of charge
to those who request them. A
mailing list service for personal
subscribers, however, is not
available.

Institutional subscribers, such
as libraries and other organiza­
tions, will be placed on a mail­
ing list upon request and will
automatically receive Working
Papers as they are published.

■

9417
Bank Diversification:
Laws and Fallacies of
Large Numbers

■

9419
Posted Rates as Signals
in Mortgage Lending
Markets

■

by Joseph G. Haubrich

by Robert B. Avery,
Patricia E. Beeson, and
Mark S. Sniderman

by Robert B. Avery,
Patricia E. Beeson, and
Mark S. Sniderman

■

■

■ 9418
The Effects of Inflation
on Wage Adjustments
in Firm-Level Data:
Grease or Sand?
by Erica L. Groshen and
Mark E. Schweitzer

9420
The Computational
Experiment:
An Econometric Tool
by Finn E. Kydland and
Edward C. Prescott

9421
Underserved Mortgage
Markets: Evidence from
HMDA Data

9501
More on the Differences
between Reported and
Actual U.S. Central
Bank Foreign Exchange
Intervention
by William P. Osterberg and
Rebecca Wetmore Humes

Please complete and detach the form below and mail to:
Research Department
Federal Reserve Bank of Cleveland
P.O. Box 6387
Cleveland, Ohio 44101




Check item(s)
requested

Please send the following Working Paper(s):

□ 9417
□ 9418

□ 9419
□ 9420

□ 9421
□ 9501

Send to:
Please print
Name

Address

City

State

Zip

32

The Federal Reserve Bank of Cleveland
and the Journal of Money, Credit and Banking
Announce:

Derivatives and Intermediation
November 2-3,1995
Cleveland, Ohio

The Federal Reserve Bank of Cleveland and
the Journal of Money, Credit and Banking
are jointly sponsoring a conference on
Derivatives and Intermediation:
Theory and Evidence.




The growing derivatives market poses several chal­
lenges for policymakers. The first is to understand
the sources of financial innovation resulting in the
proliferation of these products. What economic
forces make derivatives viable instruments? What
gains do rational participants obtain from these con­
tracts, and why do they dominate transactions in
the cash securities markets? To what extent does
regulatory policy— bankruptcy rules, capital require­
ments, accounting rules, and deposit insurance—
affect the market? In other words, does derivativerelated financial innovation stem from changes in
the marketplace, or from changes in the regulatory
environment? The answer is crucial to understand­
ing both derivatives and intermediation. The second
challenge for policymakers is to understand how
derivatives impact regulatory concerns in the areas
of bank risk, payments system reform, and
intermediary powers.

Call for Papers
The conference proceedings will be published in the
Journal of Money, Credit and Banking, and authors
will receive an honorarium. Prospective contribu­
tors are invited to send a completed paper or de­
tailed abstract by May 30,1995 to:
Joseph G. Haubrich
Research Department
Federal Reserve Bank of Cleveland
P.O. Box 6387
Cleveland, OH 44101-1387