The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Federal Reserve Staff Study- Volume II
New Monetary Control Procedures
Board of Governors of the Federal Reserve System
February 1981
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Federal Reserve Staff Study
New Monetary Control Procedures
VOLUME II
David Lmdsey and Others
Monetary Control Experience Under the New Operatmg Procedures
David Pierce
Trend and Noise m the Monetary Aggregates
Lawrence Shfman and Edward McKelvey
The New Operatmg Procedures and Ecpnom1c Act1V1ty since October 1979
J>eter Tmsley and Others
Money Market Impacts of Altern~uve Operating Proc~dures
Edwm M Truman and Others
The New Federal Reserve Operatmg Procedure An ixterl).al P~rspective.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
~MONETARY CONTROL EXPERIENCE UNDER
THE NEW OPERATING PROCEDURES
February 1981
Paper Written for a Federal Reserve
Staff Review of Monetary Control
Procedures
by
David Lindsey and Others
February 1981
MONETARY CONTROL EXPERIENCE UNDER
THE NEW OPERATING PROCEDURES
CONTENTS
I.
II•
SUMMARY OF PRINCIPAL FINDINGS•••••••••••••••••••••••••••••••
A.
B.
C.
D.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1
OBSERVED VARIABILITY OF MONETARY AND RESERVE AGGREGATES
A.ND UUTED KEA.SURES ••••••••••••••••••••••••••• o • • • • • • • • • •
III.
PAGE
7
Monetary Aggregates•····•·•······•••••••••••••••••·•••
Reserve Aggregates•••••••••••••••••••••••••·•••••••·••
Reserve Multipliers •••••••••••• o••••••••••••••••••••••
Relationship of M-lB, M-2, Reserve Aggregates and
Multipliers.........................................
12
COMPARATIVE ACCURACY OF JUDGMENTAL VERSUS ECONOMETRIC
PROJECTION PROCEDURES AND SELECTION OF THE RESERVE
INSTR.mlENT. • • • . . • • • . . • . . • . • . • . . • • . . • . • . . • • . . . • • • . . . . • • . • . •
IS
A.
B.
c.
D.
7
10
11
The Nature of the Multiplier Prediction Tests.........
An Analysis of the Multiplier Prediction Resultso•••••
1. M-lB results •••••..•.•..••....•...••......... o••••
2. M-lA results ■ •••••••••••••••••••••••••••••••••••••
3. M-2 results.......................................
19
32
32
35
35
4.
Conclusions ••••••••• o•••••••••••••••••••••••••••••
35
The Nature of the Money Stock Prediction Tests........
An Analyais of the Money Stock Prediction Results.....
1. Money stock target misses.........................
2. Money stock target misses versus econometric
prediction errors •• @••••••••••·••·••••··••••••••
3. Comparison of multiplier and money stock errors
for econometric models ••• o•••••••••••••••••••••o
4. Results across the econometric models.............
5. The choice of a particular reserves instrument....
6. A federal funds rate versus a nonborrowecl
38
46
47
reserves operating target•••••••••••••••••••••••
59
7.
The feasibility of close short-run monetary
control.........................................
47
49
50
51
61
CONTENTS (cont.)
PAGE
IV.
VARIABILITY IN INTEREST RATES AND MONEY DEMAND ANALYSIS.......
64
CHA.RTS AND TABLES ..•.•••••..•.•.••.•...••..•.• •"• •. :................
75
APPENDIX:
ERROR EXPRESSIONS IN TABLES 4-6 .........................
100
'RE"FE'RENCES ••••••••..•...•••.•••••••••••••••.••••.•••••••••• IP!\ • • • • • •
101
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
February 1981
MONETARY COt-lTROL EXPERIENCE UNDER
THE NEW OPERATING PROCEDURES*
1.
Summary of Principal Findings
This study examines the record regarding monetary control and re-
lated issues for the first year after the adoption of the reserves-oriented
operating procedures.
1.
A summary of our principal findings follows:
Observed variability of monetary and reserve aggregates and
their multipliers.
a.
Using currently available seasonal factors, monthly and
quarterly growth rates of the monetary aggregates are significantly more variable over the period of the new operating
procedures than over the last decade.
However, by smoothing
pre-1980 data more than 1980 data, current seasonal factors
substantially exaggerate this increase in variability.
b.
When the seasonal factors in use at the time are applied to
monthly data over each of the last 10 years, the increase
in variability of monthly money growth since last October is
still statistically significant.
(In contrast, the changes
in variability of reserve measures and their multipliers in
the last year--when constructed with original seasonals--are not
* David Lindsey of the Division of Research and Statistics of the Federal
Reserve Board staff was reponsible for the preparation of this study
with the collaboration of other staff of the Federal Reserve System.
Major contributions were made by Helen Farr, Gary Gillum, Kenneth Kopecky,
Eileen Mauskopf, Edward Offenbacher, and Richard Porter of the Division
of Research and Statistics of the Federal Reserve Board staff; John Judd
and John Scadding of the San Francisco Bank staff; and Albert Burger of
the St. Louis Bank staff. Assistance also was provided by Wayne Smith
and Fran Weaver of the Board staff. Thanks also are due to James Johannes
and Robert Rasche of Michigan State University for providing simulation
results of their model.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 2 statistically significant.)
Similar quarterly measures of
the narrow monetary aggregates also display significantly
higher variability in growth rates than over th~ previous
decade, but quarterly measures of M-2 do not.
c.
It may be noted that, over the last decade or so, the variability of quarterly rates of growth of the monetary aggregates in the U.S. has been well within the range observed
in other major industrial countries.
2.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Comparative accuracy of judgmental versus econometric projection
procedures and selection of the reserve instrument.
a.
Judgmental projections of the nonborrowed and total reserve
multipliers made at the beginning of intermeeting periods
were significantly more accurate than the forecasts of the
Johannes-Rasche and San Francisco Bank models.
They also
were superior to the Board's monthly model forecasts of the
nonborrowed reserves multiplier, but they were no better
than the Board model forecasts of the total reserves multiplier.
The 1udgmental multiplier errors, after incorporation
of intermeeting adjustments to the narrow reserves targets,
were in all cases smaller than the econometric forecast
errors.
In predicting the two base multipliers, the results
were more even, with the Board model showing a slight edge.
b.
The multiplier-prediction experiment was intentionally designed so that all the multiplier errors would include misses
- 3 in reserves as well as in money.
This feature provided com-
parability between the judgmental and model results.
However,
this experiment is incapable of evaluating which model could
best predict money given a fixed level of reserves or which
reserve aggregate could provide the closest monetary control
if chosen as an exogenous operating target over the intermeeting period.
This is because induced movements in
actual reserves distort the true relation going from reserves
to money and bias the obs;rved multiplier errors.
c.
Two alternative procedures for predicting monetary aggregates,
rather than their multipliers, were designed for the Board
and San Francisco models to circumvent this problem of reserve endogeneity.!./
lJ
Unfortunately, neither procedure could be
Both procedures determine exogenous levels of reserve aggregates in
an initial step. The first procedure assumes that actual nonborrowed
reserves have been the instrument controlled exogenously by the Desk since
October 1979. Hence, this procedure derives exogenous levels of the three
broader reserve measures from simulations of the ~oard and San Francisco
models given the actual level of nonborrowed reserves. Then, each reserve
measure, in turn, is held at this level and, in a second step, the model
is simulated in the prese~ce of the observed errors in all of the structural
equations. The model's prediction of money in this simulation is compared
to its prediction in the absence of these errors, and the difference interpreted as the prediction error that would have occurred if that particular
reserve aggregate had been exogenously fixed over the month.
The second, alternative procedure attempts to refine the first by
correcting for the induced movements of actual nonborrowed reserves within
the month that reflect Desk adjustments to the nonborrowed reserves target
in response to unanticipated developments as revealed by incoming data. In
this procedure, exogenous levels of all four reserve aggregates are determined in the first step by simulating the models given the staff's expected
federal funds rate for the month. The last step in this procedure is identical to that of the first procedure described above.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 4 applied to the Johannes-Rasche model, which is incapable of
addressing the problem of reserve endogeneity.
Nor could
these procedures be applied to the judgmental method, although the actual experience of misses of monetary aggregates from their intermeeting targets may be used as a
benchmark in evaluating the other two models' money forecast
errors.
d.
In both procedures, the model errors for the narrow monetary
aggregates with nonborrowed reserves or the nonborrowed base
taken as exogenous ranged from slightly to somewhat smaller
than the actual misses of money from intermeeting targets.
However, in most of these model simul3tions, even for the nonborrowed measures of reserves, the federal funds rate occasionally moved outside the FOMC's limits.
The results with total
reserves and the total base exogenous ranged from about the
same as actual misses of money from intermeeting targets to
dramatically worse in the case of the Board model forecasts
given total reserves.
The deterioration of total relative to
nonborrowed measures in both models largely stems from the enlarged
importance of misforecasts of average required reserve ratios-either implicit in the San Francisco model or explicit in the
Board model.
Without having the discount window operate to
alter levels of total reserves and the total base when required
reserves change randomly, supply-side errors in the models
substantially destabilize money.
- 5 -
e.
By contrast, when the Board model is simulated under the assumption of contemporaneous, uniform, and universal reserve requirements on demand deposits and zero reserve requirements on other
bank liabilites, its money errors--given total reserves and, to
a lesser extent, the total base--drop dramatically.
The results
given nonborrowed reserves and the nonborrowed base, however, are
scarcely affected by this assumption.
On the other hand, the
model's money errors for the two nonborrowed reserve aggregates
do show some improvement when the error in the discount borrowing
equation is suppressed.
These results suggest that refonns to
the structure of reserve requirements--some of which are in
train as requirements under the Monetary Control Act become
phased in--are a prerequisite to giving more emphasis to total
reserves or the total base in short-run open market operations.
They also suggest that consideration might be given to a restructuring of the discount window, if nonborrowed reserves are retained
as the main operating target, which seems warranted under an institutional structure similar to the present one.
f.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
All of, the model results indicate that close monetary control is
impossible in the short run of a month or so under the present
,i,..,.
....
institutional structure.
....
-~
!I
I.
.....,. ... .,
When simulated with a federal funds
rate instrument, the two models yield money errors similar to
those with the nonborrowed reserve measures as the instrument.
Under either instrument, the money errors, particularly in
the Board model, tend to fall over longer periods, reflecting
the partial averaging out of monthly errors over periods even
as short as a quarter.
- 6 3.
Variability in interest rates and money demand ~ehavior.
a.
In none of the seven money demand funct_ions examine9 is the
effect of interest rate movements, even whe~ combined with
income and price-movements, sufficient to e~plaiµ alJ
of the large variability ~n money growth since last 0~tobe~.
b.
The quarterly model w:ith the largest interest rate i~pact-and with the best annual forecasting ,recor_d in rec,ent years-is one proposed by Porter an~ ~impson of the ~pard staff that
incorporates a ~hort-term interest rate and a i.ong-term interest
rate variable representing ~he profitability of investments
in cash managem~nt.
This eqqation indicates that interest
rate movements depressed M-:1.A demand by 7 percent at an annual
rate in the fir~t quarter of 1980 and raised it by 2~1/2
percent in ·the th:f.-rd quarter ,,of this year.
c.
However, even w_ith depressing effects on M-lA growth f,rom
interest rates ,of about 4 percent and from real income,oj 11b9ut
3 p~rcent in the ,second quarter of -1980, ·this equation's over,prediction in,that quarter was about 6 per~~nt, suggesting
that other factors-were at -Wo~k.
d.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Some preliminary
evidence of,a link from a change in bank .,._loans
- "" ~
to the level of money suggests -that the large short,fall in M-lA
growth in the second quarter of 1980 r~lative ~o m9st model
predictions was related to the i~pQsi~iQn Qf credit controls
in March.
However, other ~vidence, over a variety of sample
p_eriods prior to the tnid-1970s,, casts some doubt on •the change
in total bank loans as a reliable variable~~ ~oney,demand
- 7 -
II.
Observed Variability of Monetary and Reserve Aggregates and Related
Measures*
The first issue examined is the observed variability of various mone-
tary and reserve aggregates, and of their multipliers, over the October 1979 to
September 1980 period. Weekly, monthly and quarterly variability over this 12month period--measured by standard deviations of annualized growth rates--is
compared to that over other October to September "fiscal years," beginning with
October 1970.
Results for M-lA are not reported because they are very similar to
those for M-lB.
A.
Monetary Aggregates
The widespread perception of increased variability in growth rates of
the monetary aggregates since last October appears to be borne out in Charts 1
and 2, which indicate the variability of seasonally adjusted growth rates in
M-lB and M-2 for each of the last 10 October-to-Septembe~ fiscal years. These
monetary aggregates have been seasonally adjusted using the current series of
seasonal factors.
The summary statistics plotted are standard deviations of
growth rates over weekly (for M-lB only), monthly, and quarterly intervals. With
these current seasonal factors, the data for M-lB in Chart 1 show an uptrend in
variability since the 1977 (for quarterly data, 1978) fiscal year that has continued in 1980.
The variability of M-2, shown in Chart 2, has risen since fis-
cal 1978, but from a rather low level.
* Contributors to this section:
Fran Weaver of the Board staff.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
While the standard deviation for monthly
Helen Farr, Gary Gillum, David Lindsey, and
-8-
M-2 growth in fiscal 1980 reached the peak level shown for this series--as
is also true of the weekly, monthly, and quarterly measures for M-lB--the
standard deviation for quarterly M-2 growth in fiscal 1980 only matched its
previous peak level recorded in fiscal 1975.
It might be noted that substantial variability in money growth is
not unique to the United States. Indeed, in comparison with the variability
of quarterly money growt~ in other major industrial countries in the period
1973 to the present, the variability of U.S. money growth has been fairly
low, as shown in Table 1.
However, one should be careful in comparing the
measures of variability shown in the table, for several reasons.
First, the
measures are biased in favor of relatively low U.S. variability because of
the fact that
n.s.
data are averages of daily observations while in other
countries the money data typically are based on only one observation per
month ..!./ Second, the extent to which central banks attempt to control
money growth has varied across countries and over time. Third, the institutional setting, which may affect, for example, the availability of money
substitutes and the interest sensitivity of the demand for money, differs
across countries.
Some of the observed increase in variability in U.S. monetary
growth since October 1979 can be expected to arise from the standard seasonal
adjustment techniques used for the aggregates.
These techniques tend to
produce smoother seasonally adjusted data for earlier years than for recent
1/ The series on Canadian M-1 and M-2 and German Central Bank money also are
averages of daily data.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 9 -
years because, as each year recedes into the past, the change in its seasonal
pattern relative to earlier years becomes better captured by its estimated seasonal factors.
In order to correct for this bias in judging the relative vari-
ability of the aggregates in fiscal 1980, the monetary aggregates in earlier
years were seasonally adjusted using original seasonal factors--that is,_those
available at the time •.!}
The effects of employing the seasonal factors in use at the time are
shown in Charts 3 and 4. In Chart 3, the variability of weekly growth rates
for M-lB in fiscal 1980 is no larger than in 1979, and only slightly larger
than the average for the 1971 to 1979 period. This result contrasts sharply
with the increased weekly variability for data with current seasonal factors,
shown in Chart 1.
For monthly and quarterly data, however, the variability of
M-lB growth rates still increases after fiscal 1978, although the increase is
much less pronounced.
The effects on M-2 of replacing current with original
seasonal factors, shown in Chart 4 for monthly and quarterly data, are similar
to those for M-lB.
One important outcome is that the increase in monthly and
quarterly variability since 1978 is smaller in Chart 4 than in Chart 2.
The bias implicit in the use of current seasonal factors also affects-the comparison of individual monthly growth rates.
Growth rates in
fiscal 1980 that are large in absolute value are exaggerated in comparison to
those of earlier years.
For example, the 21.6 percent growth rate of M-lB in
1/ To the extent possible, the currently available data for the not seasonally
adJusted money stock in any given year were adjusted, component by component,
with seasonal factors originally used during that year. Changes in definitions
of the money stock and its components made some ad hoc adjustments necessary.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 10 -
August 1980, which appears to be a record high, is exceeded ·by'a 24.0 percent
growth rate in April 1979 and nearly matched by a 21.4 percent growth rate in
March 1977 when the original seasonal factors are used for those earlier years.
As a check on the procedure using original seasonal factors, we employed an analogous technique for constructing seasonal .factors for earlier
0
•
,
years. In this procedure, seasonal factors were generated by the X-11 seasonal
adjustment program for each fiscal year using only aata for earlier years that
are currently available in the not seasonally adjusted series.
The results
from this technique confirmed the conclusions reached using original seasonal
factors.
B.
lJ
Reserve Aggregates
We applied a similar analysis to the variability of nonborrowed and
total reserves.
'!J Standard deviations of weekly, monthly, and quarterly growth
rates are shown in Charts 5 and 6. They are based upon data that were seasonally adjusted with the implied original seasonal factors, for the same reasons
1/ An analysis of not seasonally adjusted growth rates uncovered the surprising result that not seasonally adjusted M-lB and M-2 monthly growth rates have
varied over a significantly smaller range since October 1979 than in most earlier
years. The lower standard deviations for not seasonally adjusted monthly data
in fiscal 1980 than in several recent fiscal years may simply be a historical
accident. The seasonal factor for M-lB, for example, called for an enormous
not seasonally adJusted increase in April 1980, which did not materialize.
In addition, in August, the not seasonally adjusted growth rate of M-lB was
near the average for the year, rather than showing a drop in the ievel ~alled
for by the seasonal factor. Of course, the adoption of the new procedures
may have affected the seasonal pattern, and the System's attempt to reinforce
the existing seasonal factors may be imparting a new policy seasonal to the
data.
2/ Table 2, in 13'.lbsection D, contains results for the nonborrowed and total
monetary base.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 11 -
discussed above with regard to the monetary aggregates.
The implied sea-
sonal factors were derived by dividing the not seasonally adjusted reserve
measure by the seasonally adjusted reserve measure, both as originally published during that year.
The variability of monthly growth rates for non-
borrowed reserves in fiscal 1980 shown in Chart 5 is near the high levels recorded in the fiscal years 1973 to 1975. However, the jump of this monthly
variability from fiscal 1979 to 1980 is only half as large as would be evident if current seasonal factors were used instead.
In contrast, the weekly
and quarterly growth rate data for nonborrowed reserves decline in variability from fiscal 1979 to 1980.
The variability in the growth rate of total
reserves, presented in Chart 6, declines from fiscal 1979 to 1980 with all
three data frequencies and is not as high as the variability in several earlier
years.
c.
Reserve Multipliers
Multipliers defined as the ratio of M-lB to either nonborrowed or
total reserves have been constructed from data that have been seasonally adjusted with the original factors ..!/ As shown in Chart 7, month-to-month variability of the nonborrowed reserves multiplier, as with nonborrowed reserves
themselves, increased in~£Lsca1~~98Q to.near~the highs' attained~in-thecm~d,~L
1970s. In ·contrast, weekly and quarterly variability of the nonborrowed reserves multiplier declined in 1980, to a level below the average variability
for the past decade.
1/ Table 2, in subsection D, contains results for the nonborrowed and total
monetary base multipliers.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 12 -
The monthly and quarterly variability of the total reserves multiplier, shown in Chart 8, evinces only slight increases in ,fiscal 1980. Monthly variability in 1980 was well below past peaks, but quarterly variability was.at a
relatively ,high level.
In contrast, weekly variability of this multiplier de-
clined sha,rply to its lowest level of the decade.
D.
Relationships of M-lB, M-2, Reserve Aggregates, and Multipliers
The relationships among the results for M-lB and M-2, the reserve ~ggre-
gates, and their multipliers are presented in Table 2.
'
,
Suunnary stattstics of
monthly and quarterly growth rates for fiscal 1980 are compared both to those
.
for fiscal 1979 and to those of the average of the fiscal years 1971-79.
In
.
addition, tests of statistical significance of the change in variability in 1980
from 1979 and from the average of the 1970s are reported.
The variation in monthly and quarterly M-lB and M-2 growth rates over the
1970s is smaller than the growth-rate variation both for the narrow reserve aggregates and for their associat~d multipliers.because of substantial negative correlation between reserve anq m~ltiplier growth:rate variability.
In other words,
there was a strong tendency for narrow reserve aggregates to move in the opposite
direction-to changes in their multipliers. For the nonborrowed and total monetary
.-
~
~
'".
l
1971-79 period for both monthly and quarterly data, but it is not large enough to
make M~lB and M-2 growth vari~bility smaller than that for these reserve measures
themselves.
Although a comparable pattern generally emerges over monthly intervals
in fiscal 1980, the attenuation of the negative correlation between the monetary
. '
base and its multiplier makes the variability of its multiplier changes smaller than
money-growth variability.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
For quarterly data in fiscal 1980, both base measures
- 13 -
and their multipliers were positively correlated, causing the variation in M-lB
and M-2 growth to be larger than the variation in growth of either the multiplier
or the base measure.
Tests have been made of the statistical significance of the changes
in variability in fiscal 1980 from earlier periods.
Changes in variability that
are significant at the 10 percent level (for a two-tailed test) are indicated
by three special symbols in Table 2.
Only M-2 monthly growth rates show a
statistically significant increase in variability in fiscal 1980 when compared
both with fiscal 1979 and with the fiscal 1971-79 period as a whole. However,
variability in quarterly average growth rates for M-2 in fiscal 1980 is not
significantly higher than in either earlier period.
For M-lB, variability in
both monthly and quarterly average growth rates for 1980 is significantly higher than for the 1971-79 average, but not significantly higher than for 1979 taken
alone. Statistically significant decreases were recorded for nonborrowed reserves
and its multipliers with M-lB and M-2, but only for quarterly average rates of
change.
Despite these indications of significantly higher variability for the
monetary aggregates and significantly lower quarterly variability for nonborrowed reserves and its multipliers in fiscal 1980, some years in the 1970s also
registered statistically significant changes in variability of these measures.
Lagged reserve accounting appears to be a major factor explaining the
high variability of the reserve multipliers reported in Table 2.
To estimate ap-
proximately the impact of lagged reserve accounting, multipliers were constructed
in two ways f-rom four-week averages of not seasonally adjusted levels of M-lB
and the various reserve measures.
-
In the first, nonborrowed and total reserves
were measured over the same weeks as the monetary aggregates, representing
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 14 -
the present lagged accounting system.
In the second, nonborrowed and total
reserves were measu~ed over a four-week period ending two weeks after the corresponding period for the aggregates, representing a system of cont~mporaneous
reserve accounting.
For :both the nonborrowed and total monetary base, the
currency and nonmember bank vault cash components were not shifted ~orward,
although reserve components were.
Table 3 reports the measured reduction in multiplier variability due
to the two-week forward shift of reserves.
rhe variabil~ty of the total re-
serve multiplier is de~reased going from column (1) to column (2) by.an amount
ranging from 12 to 14 percentage points (depending on the period), roughly half
of the measured variability under lagged reserve accounting.
The other reserve
measures, especially the nonborrowed and total monetary base, show less improvement.
Also, the adjustment ~or contemporaneous reserve accounting produces much less of
an improvement in the nonborrowed reserve multi,plier in fiscal 1980 than in
earlier years, particularly when viewed as a fraction of the variability of the
multiplier in column (1).
This oddity of 1980 should be kept in mind in
interpreting certain results in the next section.
The procedure embodied in Table 3 probably overestimates the reduction
in multiplier variability that would have obtained had contemporaneous reserve
accounting been in existence during these years.
Under contemppranous accounting
the variability of excess reserves might have risen and, in any event, the outcome of interest ,rates and the quantity of ,money demanded would have differed from
that actually ob.served.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 15 -
III. Comparative Accuracy of Judgmental Versus Econometric Projection
Procedures and Selection of the Reserve Instrument*
The departures of the monetary aggregates from longer-run and
interim targets as well as wide swings in short-term interest rates since
last October have given rise to criticisms of various aspects of the operating procedures.
In one way or another~ all the criticisms involve the tech-
niques used in selecting and adjusting target paths for the reserve aggregates.
Under the new procedures, lnitial intermeeting levels for a family of reserve measures are derived, largely judgmentally, from intermeeting
targets for the money aggregates and from associated projections of separate
components of the aggregates, other liabilities subject to reserve requirements, excess reserves, vault cash, and discount window borrowings.
The in-
termeeting money stock target reflects the FOMC's desired speed of return to
the longer-run objective following observed deviations.
This target repre-
sents the FOMC's chosen average growth rate for the entire interim period of
several months, adjusted for lagged effects of policy actions and special factors) both of which give rise to expected temporary variations of money demand
around that average growth rate.
After the intermeeting average money stock
targets and related projections are converted to weekly paths over the intermeeting period, the associated weekly levels of total reserves and the total
* Contributors to this section: Helen Farr» Gary Gillum, Kenneth Kopecky,
David Lindsey, Richard Porter, and Wayne Smith of the Board staff; John Judd
and John Scadding of the San Francisco staff; and Albert Burger of the St.
Louis Bank staff.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
base are derived.
The initial intermeeting targets for these reserve aggre-
gates are simply the averages of these weekly levels.
The FOMC's stated as-
sumption for discount window borrowings is then used to derive the nonborrowed reserves target and associated nonborrowed base level.!/
The implicit money/reserve multipliers built into the initial intermeeting targets for reserves can, of course, be derived by dividing the
targeted levels of the monetary aggregates by the targeted levels of reserves.
During intermeeting periods, the staff typically adjusts the re-
serve targets in light of incoming information about unexpected changes in
the multipliers.
These adjustments are made cautiously, either to avoid
overreaction to transitory, self-correcting changes in the multiplier or
because the multiplier variation is recognized too late in the intermeeting
period for a change in reserve targets to have a perceptible effect on the
intermeeting average level of the monetary aggregates.
While many observers, including market participants, have
objected to the enlarged variability of interest rates since October 1979,
others, particularly those in the monetarist camp, have complained that the
operating procedures still embody excessive emphasis on smoothing movements
in short-term interest rates. Some monetarists argue that the Federal Reserve
should make adjustments to the reserve targets more aggressively, both within
1/ For a detailed discussion of the establishment of reserve target paths,
see Appendix B of the Federal Reserve's Monetary Policy Report to the Congress, February 1980. Also see Stephen Axilrod and David Lindsey, "Federal
Reserve System Implementation of Monetary Policy: Analytical Foundations
of the New Approach," Federal Reserve Board, processed; presented at the
Denver Meeting of the American Economic Association, September 6, 1980;
forthcoming in American Economic Review, Papers and Proceedings, May 1981.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 17 -
interm~eting perio<ls and from one intermeeting period to the next, regardless
of interest rate consequences. Other monetarists suggest instead that the Desk
should simply maintain a predetermined growth rate of some reserve measure
from one intermeeting period to the next, or even over a considerably longer
time, and accept whatever interest rate movements result.
This section first addresses a somewhat narrower, more technical
criticism that originates with those who have recommended econometric, rather
than judgmental, techniques for forecasting the various multipliers.
The
Shadow Open Market Committee, besides proposing the monetary base as either
an operating or intermediate target, also suggests replacing judgmental projections of the multiplier with a statistical time-series method devised by Professors James M. Johannes and Robert H. Rasche of Michigan State University.!/
This committee advocates maintaining over the control period a level of the reserve instrument equal to the long-run money stock target divided by this multiplier estimate.
(Professor James Pierce of the University of California at
~erkeley makes a somewhat different criticism.
He argues that modern methods
of statistical "filtering" and. "optimal" forecasting should supplement judgmental procedures.)Y
1/ For the most recent description of this technique, see James M. Johannes
and Robert H. Rasche, "Can the Reseryes Approach to Monetary Control Really
Work?" April 1980.
2/ James L. Pierce, "Making Reserves Targets Work," in Controlling the Monetary Aggregates III, (Federal Reserve Bank of Boston, forthcoming). It may be
noted that both the money demand side of the Board staff's monthly money market model and Banking Section time series models of the monetary aggregates
have been considered in preparing the initial intermeeting money targets. In
addition, the Banking Section is engaged in a long-term project to integrate
time-series models and filtering methods into the judgmental projections made
between FOMC meetings.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 18 In order to assess the gains, if any, that would have resulted
if econometric models had replaced judgmental methods in October 1979, we
conducted two experiments drawing on the evidence accumulated since that time.
iirst, we compared the accuracy of judgmental projections of multipliers for
nonborrowed reserves, total reserves, the nonborrowed monetary base, and the
total monetary base with the accuracy of one-month-ahead postsample forecasts
of three monthly econometric models.!/
The purpose of this test was to see
whether the various multipliers could have been better predicted by econometric techniques.
Second, we contrasted the.misses of the monetary aggregates
from their intermeeting and monthly targets with the prediction errors of the
models when each of the four reserve measures was treated in turn as the exogenous control instrument.
Because the second test corrects for induced move-
ments in the reserve aggregates, we believe it is a more reliable indicator
of the potential value of the econometric models in helping to derive reserve
targets than the first procedure.
Another purpose of this test was to see
whether money could have been better controlled by hewing to ·an operating
target other than nonborrowed reserves.
Thus, these results also suggest the
-
degree of monetary cont~ol that would have been attained if each of the four
reserve measures had been used as the primary operating target.
1/ The advent of the new reserve-oriented operating regime no doubt has altered the coefficients of the variqus models' equations. Practical ways of
mitigating this drawback of econometric procedures, other than judgmental adjustments to the equations, have not been advanced.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
.;. 19 -
A;
The Nature of the Multiolier-Prediction T~sts
The multiplier predictions of the three models were compared with
three conceptually distinct judgmental projections 6f the multiplier. The
first is the initial projection of the average multiplier over the intermeeting period.
It is simply the targeted average level of the relevant
monetary aggregate divided by the initially targeted average level of the
relevant reserve measure.
These projections were made at the beginning of
the intermeeting period just after the FOMC meeting; in cases of a long interval between FOMC meetings, the intermeeting period was divided into two
subperiods for reserve targeting purposes.
The error in the multiplier is
calculated as the difference between the realized multiplier and its predicted value.Y
It equals the percent miss of money from its intermeeting tar-
get less the percent miss of reserves from the initial intermeeting target,
both at annual rates.~
1/ In all the calculations of forecast errors for all the procedures, the
actual and predicted multiplier or money stock values were based on data
that were either not seasonally adjusted or were converted to not seasonally
adjusted levels using the Board's current seasonal adjustment factors.
Then, the natural logs of the levels were calculated and percent errors
derived as the difference beween the log of the actual and the log of the
predicted level times 100. Finally, the percent errors were converted to
represent annual rates of change by multiplying by 12.
2/ The multiplier errors to be reported in Table 4 are calculated as
where Mis the monetary aggregate; R is the reserve measure; and Act, IniPred,
Target, and IniTar represent actual, initially predicted, targeted, and ini- tially targeted values respectively. Rearranged, this expression becomes
1200• (ln mAct _ ln miniPred)
= 1200• (ln MAct - ln MTarget)
- 1200• (ln RAct - ln RiniTar).
Thus, the multiplier error is composed of a monetary aggregate error and a
reserve aggregate error. The appendix shows the expressions for all the errors
whose summary statistics are shown in Tables 4-6 of this section.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 20 -
It will be important in the later analysis to keep in mind that
this judgmental multiplier error incorporates misses of both money and reserves~from the levels'built into the multiplier projection.
That is, in-
termeeting deviations of the monetary aggregate from its,target and of the
reserve measure from its initial target both contribute to a multiplierprojection error.
This characteristic of incorporating both money and
reserve errors will be preserved in designing comparable experiments for
the econometric models.
The second judgmental multiplier projection, examined is for the
adjusted intermeeting-period multiplier.
It is simpl~ the same intermeeting
money target divided by t~e "final" adjusted intermeeting reserve target.
The reserve target determined at the beginning of the last statement week
_of the intermeeting period was considered the final one.
It should be noted
that reserve path adjustments do occur in the last week; in fact, about twothirds of the misses of the final adjusted reserve path for nonborrowed reserves were intentional.!/ The error in this adjusted multiplier projection
represents not only the percent miss of money from target less,the percent
miss of reserves from their adjusted target, but also the extent to which
intermeeting adjustments-to targeted reserves prior to the final week did
not compensate for actual errors in the initial multiplier projection.
This latter relationship is shown formally in tpe appendix.
Initial
"-intermeeting-period-multiplier percent_errors,.plus t;l!e_perc~nt reserve path
'
1/ See Fred J. Levin and Paul Meek, "Implementing the New Operating Procedures: The View from the Trading Desk," in this compendium.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
.,
.
- 21 -
adjustments-made as compensation for expected errors in the initial multiplier projection--equal the adjusted intermeeting-period-multiplier percent
errors.
Since the adjusted multiplier error equals the initial multiplier
error plus the reserve target adjustment, the adjusted multiplier error
will be lower than the initial multiplier error to the extent that intermeeting
reserve adjustments are in the opposite direction to initial multiplier
errors.
In fact, adjusted multiplier errors for both nonborrowed and total
reserves have been smaller, on average and ignoring sign, than the initial
intermeeting period multiplier errors because reserve path adjustments have
partially offset the initial multiplier errors, owing to a negative correlation
between the two.
It should not be expected that reserve path adjustments would fully
offset initial multiplier errors.
For one thing, unexpected multiplier errors
occur in the last week of intermeeting periods, after "final" reserve path adjustments have been made. For another thing, lagged reserve accounting creates
certain problems for total reserves.
Total reserves are predominantly deter-
mined by the amount banks need to satisfy required reserves based on deposits
two weeks earlier.
This permits certain recognized multiplier disturbances
originating on the money-supply side that affect the intermeeting average
total and nonborrowed reserves multiplier values--such as changes in the
mix of demand deposits that alter required reserves given the same amount of
demand deposits in total--to be fairly readily offset through adjustments to
both the total and nonborrowed reserve targets. Other disturbances on the
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 22 supply-of-money side, such as unexpected changes in demands for excess
reserves, and even more importantly money-demand-side disturbances,
pose more of a problem.
For example, a permanent surge in the demand for
transactions balances in the second week of a four-week intermeeting period
will not raise required reserves, and hence demanded total reserves, until
the last week.
Thus, due to the delay in this increase of total reserves,
the average total reserves multiplier over the intermeeting period would move
above the initially projected level.
But, in this case, total reserves also
would necessarily overshoot its initially targeted average level, given the
now higher demand for reserves in the last week.
Nonborrowed reserves, by contrast, are susceptible to fairly close
week-to-week control, and near-term path adjustments in response to money demand shocks to its multiplier are more practicable for this reserve aggregate.
Even so, adjustments to the nonborrowed reserve target can completely offset
such multiplier disturbances only if the monetary aggregate is fully returned
to its targeted intermeeting average level following recognized divergences.
But such divergences from target, particularly those recognized late in the
intermeeting period, are difficult to completely eliminate.
An attempt to
adjust the nonborrowed reserve path late in the period in order to compensate
fully for a demand-side disturbance to its multiplier would only slightly
affect the intermeeting average of money, and thus would be frustrated by
a further, nearly proportional offsetting change in the observed value of
the multiplier in the opposite direction. In addition, efforts made late
in the intermeeting period to compensate fully for perceived multiplier
disturbances through adjustments in the nonborrowed reserve target run the
risk of violating the funds rate limits set by the FOMC.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 23 The third judgmental multiplier forecast analyzed is the multiplier
projection for the current month.
each FOMC meeting.
midmonth.
These projections also were made just after
Since October 1979, FOMC meetings took place on average at
Accordingly, this forecast is reported in order to indicate how
much the receipt of incoming data improves the quality of the judgmental projection.
This projection is constructed as the average of those past weekly esti-
mates and future weekly targets for the money aggregate that are encompassed
by the current month divided by the average of those past weekly estimates and
future weekly targets for the reserve aggregate that also are included in the
current month.
Thus, the judgmental projector of current-month multipliers typi-
cally had access to one week of first published data and one week of preliminary
data for the monetary aggregates at the time the projection was made. Also, the
projector typically had knowledge of interest rate developments over the first
half of the month, as well as estimates of required reserves for two more weeks
based on the first published and preliminary deposit data.
These considerations
thus mean that the judgmental projector of current-month multipliers enjoyed
advantages in the tests unavailable to the monthly econometric models.
The postsample predictions of three econometric models with
quite different structures also were examined: the Johannes-Rasche model;
·-~the~Board _staff's money market model; and the San Francisco Federal Reserve
!
Bank staff's money market model.
..,._
'C-
..... it.
-
-
~i:,.,~~-.i
.::.~~
_.__J
.;..J__
~
,._
-
The basic characteristics of each of these
models will be summarized here briefly and citations given for more detailed
descriptions.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 24 -
The Johannes-Rasche model originally was devised to predict the monetary-base multiplier but recently has been adapted by the authors to forecast
the other reserve multipliers as well.!./ Their model predicts separately six
component ratios of the various multiplier expressions: the ratio of reserves
to demand deposits, the ratio of currency to demand deposits, and other relevant ratios comprising the multiplier.! Time-series equations--univariate
Box-Jenkins ARIMA models--that capture the influence of history, seasonal movement9, and previous errors are used to forecast these ratios.l/ Forecasts of the
various multipliers are constructed from these predictions of the component
ratios.
In the present study, the simulations were conducted by Johannes and
Rasche with the assistance of the staff at the Federal Reserve Rank of St. Louis.
The Board's monthly money market model, by contrast, develops explicit
equations for the demand for and supply of demand deposits. Interest rates
equilibrate the two when nonborrowed or total reserves are taken as exogenous.
The demands for demand deposits and currency depend inversely on current
and lagged levels of interest rates, and directly on current and lagged
levels of exogenous real personal income and exogenous prices. The supply
of demand deposits depends on the predicted amount of reserves available
1-/ See- Johannes, and Rasche, '!Can the Reserves -Approach Real:l..y WorkJ •: ,~a!ld ,,_ r
Johannes and Rasche, "Predicting the Money Multi plier," Journal of Monetary
Economics, vol. 5 (July 1979), pp. 301-25.
2/ As the reserve concept, the model used the Board's reserve measures adjusted
for changes in Regulations D and M rather than the St. Louis Bank's variants.
This approach assumes perfect knowledge of required reserves on nondeposit ltems
and of changes in the marginal and supplemental reserve requirements. In the
simulations of all the econometric models, special borrowings were treated as if
they were known with certainty.
3/ Each successive month's multiplier was forecast after incorporating the error
made in the previous month; thus, the final not seasonally adjusted values of
the monetary aggregates and their components in the previous month were assumed
to be known with certainty.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 25 to support demand deposits divided by the predicted average required reserve
ratio on demand deposits.
With nonborrowed reserves taken as the exogen-
ous control instrument, for example, predicted reserves available to support
demand deposits equal actual nonborrowed reserves plus predicted discount window borrowings less predicted excess reserves less predicted required reserves
on savings and time deposits less actual required reserves on nondeposit items.
The supply of demand deposits, given nonborrowed reserves, depends directly on
market interest rates because of the positive interest-rate-demand elasticity
of borrowings and the negative interest-rate-demand elasticity, on balance, of
savings and small time deposits.
Having determined predicted equilibrium levels of demand deposits
and interest rates, the predicted levels of the monetary aggregates are derived by adding the predicted quantities demanded of currency and the other
components of the aggregates to predicted demand deposits.!./ This model was
originally developed by Board staff in the early 1970s and has since gone
through many respecifications.!/
1/ All the Board model simulations used Board staff data and projections
for real personal income and prices in current and earlier months. Thus,
for example, no model simulation was based on the revisions in personal
income for July-September 1980 published November 18, 1980. However, perfect knowledge of the currently published (mid-November 1980) monetary
aggregates for the previous month was assumed. The actual average discount
rate in the current month also was used. Except for the equations predicting
required reserve ratios, the sample period for all the model's equations
ended in September 1979. The equation for the required reserve ratio on
demand deposits was refit after each postsample projection to include the
latest month.
2/ For the original specification see Thomas D. Thomson, James L. Pierce and
Robert T. Parry, "A Monthly Money Market Model," Journal of Money, Credit and
Banking, vol. 7 (November 1975), pp. 411-31. For the most recent description
of the model, see Helen T. Farr, "The Monthly Money Market Model," working
paper (Board of Governors of the Federal Reserve System, July 1980).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 26 The last monthly model examined wa~ recentlY, estimat~d by the
staff of the Federal Reserve Bank of San Francisco.
The model incorporates
an equation for the demand for nominal demand deposits that depends on current and lagged values of nominal personal income and the commercial paper
rate.
This equation also includes as an explanatory variable the change in
total loans at-commercial banks, which is designed to capture the effects
on the money stock of temporary shocks arising from net loan extensions
or repayments that make actual money depart from the long-run demanded
quantity.
In addition, the model has behavioral equations for bank demands
for total reserves and for discount-window borrowings, together yielding
the bank demand for nonborrowed reserves.
The the~retical underpinnings
of the model focus upon bank management of federal funds and RPs as substitutes for other managed liabilities, including large CDs and discount-window
borrow\ngs, and upon the effects of funds rate movements on the bank supply
of demand deposits.
The banking system supports more managed liabilities
and fewer demand deposits at a higher federal funds rate for a given volume
of assets to be financed.
An increase in the volume of such assets causes
increases in both managed liabilities and demand deposits, thus providing a
direct supply-side link between bank loans and the quantity of demand deposits supplied.
The model's implicit demand deposit supply function depends
directly on the commercial paper rate and inversely on the federal funds rate.
1/ For additional discussion, see section IV of this paper and John Scadding and John Judd, "The Disequilibrium Demand for Money," forthcoming.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 27 The interaction of this demand deposit supply function with the demand
deposit demand function, together with the interaction of the actual supply
of nonborrowed reserves with the bank demand for nonborrowed reserves, which
1s
a function of the commercial paper rate and the funds rate, determines the
equilibrium levels of interest rates as well as the model's predicted quantities . ..!./
Before explaining how the predictions of the multiplier in the
Board and the San Francisco models were obtained, a digression on the "pure
theory" of deriving reserve targets is warranted.
The desired average
growth rate for the monetary aggregates over a horizon of several months is
first established by the FOMC.
iod growth rates is constructed.
Then a pattern of targeted intermeeting-perIn doing so, consideration is given to any
1/ The model is actually estimated for lunar month--that is, four-week-blocks of data, with required reserves based on deposits lagged two weeks to
account for lagged reserve accounting. Although actual levels of time and
savings deposits for the four weeks ending two weeks earlier are used in computing required reserves, predicted levels of demand deposits and managed liabilities shifted back two weeks also enter into the computation. Lunar-month
predictions were interpolated to obtain calendar-month predictions. The model's equations were fit over a mid-1976 through September 1979 sample period.
In the simulations of the model, all exogenous variables for the current month other than the reserve instrument and the discount rate were projected using time-series models. Perfect knowledge of seasonally adjusted deposits, bank loans,- personal income, and so on was assumed for periods prior to
- the month being forecast_._ ,TbeJ Bo~r_di~ ,~eserve measur:es a9jt1;sted ,for chan_geEl, ~
in Regulations D and M were used in all simulations. _This approach assumes
perfect·knowledge of required reserves on nondeposit items and of changes in
the marginal and supplemental reserve requirements. For the four months from
May through August, when the federal funds rate fell below the discount rate
and adjustment borrowings dropped to a frictional minimum, the model was simulated with the actual federal funds rate treated as the exogenous control instrument.
A more detailed description of the model and projection technique
appears in John Judd and John Scadding, "Contribution to the Study of the
Monetary Control Experience Under the New Operating Procedures," forthcoming.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 28 expected temporary variation in money demand around the specified average
growth rate--due to the operation of lags in the impact of policy actions and
to known special factors, say, tax rebates.
Simultaneously, an expected
level of short-term interest rates is implied that is consistent with anticipated money demand relationships at the quantity of money given by the
upcoming intermeeting target.
Of course, the existence of such an expecta-
tion does not imply that the Federal Reserve uses the funds rate as an operating target.
Unanticipated developments over the intermeeting period cause
the actual interest rate outcome to differ from initial expectations. Indeed,
the primary virtue of reserve targeting is that unexpected changes in money
demand cause interest rates to react automatically and offset some of the
miss of money from the target level.
(These induced interest rate movements
also will be in the appropriate countercyclical direction to the extent that
the unanticipated change in.money demand is related to an unwanted strengthening or weakening of aggregate spending.!/)
The next stage of the process involves selecting an initial expectation for discount-window borrowings.
This selection is conditioned by
the initial expectation of money market interest rates that are thought con1/ To the extent that changes in money demand represent shifts in the demand
fun~tiQn itself, the induced interest rate movements with an !~variant rese'rve target··wn1~not 'be·itr=-,the-appropriafe countercyclicai dire-ction ... ,To_.
the extent errors in the money supply function occur, the induced interest
rate movements will neither contribute to hitting the money target nor be in
the appropriate countercyclical direction, other things equal.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
...
- 29 sistent with the public's money demand at the intermeeting target for the
money stock.
In light of this ~xpectation of short-term interest rates and
the level of the discount rate, an initial average level of discount-window
borrowings is assumed for the interme~tlng period. Then 1ntermeetlng targets
for total reserves and the total base are determined from the money targets
and projections of reservable items based on expected interest rate levels.
Finally, subtracting the expected level of borrowing~ from the target& for
total reserves and the base yields the targets for nonborrowed reserves and
the nonborrowed base.
The manner in which the models generated multiplier forecasts
may now be•described.
In contrast to the Judgmental multiplier projections,
which were based upon money and reserve targets, for Lhe models predictions
of both aggregates were used. Initially, howe~er, we tried an alternative
approach.
For the multiplier prediction with the Board model, we first
attempted to derive reserve predictions consistent with the monthly target
for M-lA.
However, these targets reflected the judgmental projector's
estimates about money demand, and at times implicitly incorporated information
about the model's errors unavailable to the model.
of money demand typically were different.
The model's predictions
Even if these differences were
relatively small (and even if the model's predictions of money demand were
more accurate), large errors in the model forecasts of interest rates and
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 30 of nonborrowed reserves would be implied.
In fact, the model simulations
yielded very large multiplier forecast errors.Y
Therefore, an alternative experiment was conducted that permitted the Board and San Francisco models to predict levels of the money stock,
as well as reserves, that are consi9tent with the detenninistic structure-of
each model.
Specifically, the two models predicted money and reserves given
the same staff expectatio~ of the federal funds rate that the judgmental projector used.
That is, the predictions of money and reserves in the BoarJ
and San Francisco model simulations were hased on the judgmental expectation
of the average federal funds rate in the current month, made around the time
of the FOMC meeting.
This procedure thus allowed for misses in predictions
of both money and reserves.
These multiplier predictions were then subtract~d
from the realized multiplier values, and the errors compared to the Jo~annesRasche monthly errors and the judgmental intermeeting and current-month fo~ecast errors.
While these experiments were designed to be as even-handed as possible in comparing the different multiplier forecasting techniques, we believe on balance that the initi~l judgmental inter~eeting multiplier projections are most handicapped, and the current-month judgmental multiplier fore1/ The comparable siwulation with the San Francisco model was aborted when
the results for the Board model became known. This problem'of unrealistic
interest rate variability does not arise for the Johannes-Rasche procedure
because the multiplier forecasts are invariant to changes in iPterest rates,
which do not appear in the model.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 31 casts most advantaged . .!./
1/ Several factors put the initial intermeeting judgmental projections of the
multiplier at a disadvantage in the experiments compared with the model forecasts.
Since intermeeting periods (or subperiods) are, on average, shorter than calendar
months, there is less automatic smoothing of the actual figures owing to reversals of transitory noise with the passage of time. As a minor compensation for
this effect, the judgmental forecast errors for intermeeting periods expressed
as percentages were converted to annualized rates-of-change errors by a factor
of 12, rather than a factor of around 13. Moreover, the judgmental projector
typically goes into intermeeting periods at midmonth having received only first
published monetary aggregate data for the previous month, in contrast to the
perfect knowledge of final data afforded the models. (Since these
forecasts were compared with final data, the judgmental forecasts were adjusted,
as necessary, for benchmark revisions, but not for other data revisions.) Furthermore, the projector does not know the final values of the previous month's
measures of economic activity, such as personal income and prices--unlike the
San Francisco experiment-but, like the Board model runs, only has the benefit
of ~taff estimates. The projector also must predict reserve requirements
against nondeposit items and impacts of changes in Regulations D and M, which
the models were allowed to know with certainty. The funds rate expectation
in the current month that was used to obtain the Board and San Francisco models'
forecasts on average was based on two weeks of observed data, giving the Board
and San Francisco models an advantage relative to the initial intermeeting
judgmental projection. Finally, unlike these two models, the judgmental projector would not know of upcoming discount rate changes. Owing to knowledge
of additional data, the adjusted intermeeting projections and current-month
judgmental projections are, of course, much less disadvantaged than either
the initial intermeeting projections or the three econometric approaches.
There is a difficulty with the concept of a current-month multiplier
forecast. When averaged over the current month, the actual reserve measures,
even nonborrowed reserves, frequently differed from the average of the weekly
levels falling in the current month that comprise the initial intermeeting
target path., The Desk.was instructed to aim at .,a ~onbor~~w~4~r~~~~v,es. f~rg~;,
defined as an intermeeting ~verage. Thus, the Desk did not always attempt to
follow week-by-week the individual weekly compon~nts of the intermeeting
average, even aside from reserve ta~~et adju~tments. This reason partly
explains why the actual outcome for nonborrQwed reserves in the current
month deviated from the current-month average of weekly target path levels.
Given the strong negative correlation between nonborrowed reserves and
their multipliers, particqlarly in the v~ry short run, this effect enlarges
the current-month judgmental e~rors of the no~borrowed reserves ~ul~iplier
projection.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 32 B.
An Analysis of the Multiplier-Prediction Results
Table 4 presents summary statistics for the multiplier-prediction
errors of the four techniques ..!/ The findings may be summarized briefly:
1.
M-lB results.
Results for M-lB are available for all the pro-
cedures.
a.
Judgmental results.
(1)
The error dispersion statistics--mean absolute and root
mean squared errors--for the adjusted judgmental intermeeting projections of nonborrowed and total reserve
multipliers are consistently lower than the figures for
the initial projection.
This improvement, noted earlier,
results from reserve path adjustments that partially
compensate for recognized multiplier disturbances.
(2)
In contrast, reserve path adjustments yielded no average
improvement for the nonborrowed or total base multiplier
forecasts.
However, because the Trading Desk was instruct-
ed to focus on nonborrowed and total reserves as operating
targets, formal reserve path adjustments were made only
to these two reserve targets. Thus, the only adjustments
to the two base paths were these formal
adjustments
for
..,J.
~
..
nonborrowed and total reserves.
-
.J
...;JI<-
,_
..
J- -
t..,,,
...
';:::
-
In other ~ords, the
.
1/ The appendix presents the expressions for the errors.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 33 table <loes not include any intermeeting adjust~ents
to the currency or vault cash components of the base,
even though proJections of these items also were
altered as the intermeeting period progressed.
(3)
The current-month Judgmental proJections of all the H-lB/
reserves multipliers are more accurate than either of the
intermeeting pro]ections in terms of error dispersion statistics, reflecting the gains from additional information.
(4)
Over the period examined the judgmental mean error statistic was negative for the total reserves measure, implying
an overestimate of the total reserves multiplier, but it
was positive for nonborrowed re9erves, meaning the nonborrowed reserves multiplier typically was underestimated.
The source of this reversal was an initial underestimate,
on average, of the ratio of discount-window borrowings to
deposits ..!/ Part of this average underestimate typically
kept the nonborrowed reserves multiplier prediction from
being too low, unlike the average total reserves multiplier
prediction.
But the borrowing error also was, on average,
sufficiently large to make the nonborrowed reserve multi-
1/ The only difference in the two multiplier expressions is the presence of a
negative ratio of discount-window borrowings to transactions balances in the
denominator of the nonborrowed reserves multiplier.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 34 plier-prediction higher than the observed multiplier.
The tendency for discount borrowings to be higher than
the FOMC's initial assumption was amplified, on average,
by a reduction of nonborrowed reserves below the initial
target, partly undertaken in order to offset the effects
on money of higher borrowings than expected given the
funds rate, discount rate, and required reserves.
b.
Judgmental versus model results.
The initial intermeeting judg-
mental projections consistently outperform the Johannes-Rasche
and San Francisco model predictions for all four reserve
multipliers.
The edge is fairly small for the nonborrowed and
total monetary base multipliers.
The initial intermeeting judg-
mental projections and Board monthly model forecasts are more
evenly matched.
The measures of error dispersion of nonborrowed
r
reserve multiplier forecasts are substantially lower for the
judgmental procedure than for the Board model.
For total
reserves multipliers, however, the Board model has a slight
edge.
It may be noted that the intermeeting projection incor-
porating adjustments to the reserve targets reverses the relative performance of the two procedures for total reserves.
For the nonborrowed and total base multipliers, error dispersion statistics are somewhat lower for the Board model than
for the initial judgmental projections.
The judgmental mean
errors, 'though, are lower in all cases than the Board model
with the exception of total reserves.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 35 2.
M-lA results.
M-lA.
The Johannes-Rasche technique was not applied to
The error dispersion statistics for M-lA tend to be a bit
larger than for M-lB (except for the San Francisco model) but
otherwise are quite similar.
1.
M-2 results.
The San Francisco model does not predict M-2.
Also,
since no weekly data for redefined M-2 exist, there are no intermeeting judgmental results for this aggregate.
However, all of
the judgmental error dispersion statistics for the current-month
projection are lower than those for the narrower monetary aggregates, as is the case for most of the Johannes-Rasche statistics.
The reverse is true of all the comparable Board model statistics,
except for the nonborrowed base.
4.
Conclusions.
a.
Value of alternative procedures.
(1)
One conclusion emerging from Table 4 is that the initial
intermeeting judgmental projections of multipliers for
the narrow reserve measures were either superior to or
about the same as econometric forecasts derived without
the benefit of judgmental "add factors."
Inteno.eeting
adjustments to the Judgmental proJections improve their
performance.
Given that under the new procedures the
effective operating targets have been nonborrowed and
total re&erves, these results do not indicate that replacing judgment with econometric multiplier forecasts in
- 36 -
October 1979 would have provided consistently better multiplier projections.
(2)
The various model forecasts obviously contain information,
however, and supplementing judgmental forecasts with model
forecasts would provide some gain in precision.
In principle,
a weighted-average "consensus" forecast could be constructed
from judgmental and econometric predictions, with the predictions of the historically more accurate procedure
weighted more heavily.,Y
b.
Multiplier-projection techniques versus selection of money
targets.
(1)
It might be argued that if another procedure for predicting multipliers, say the Johannes-Rasche technique, had
been used to derive reserve paths each month that were
consistent with the midpoint of the longer-run ranges for
the monetary aggregates, the divergences from the midpoints of the longer-run ranges since October 1979 would
have been reduced.
1/ As noted earlier, the Board's monthly model already is considered in determining the particular pattern of intermeeting money targets consistent
with the FOMC's average interim path.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 37 -
(2)
This position really boils down to a criticism of the
selection of intenneeting money targets, not of how
multiplier forecasts are made.
Even with intermeet-
ing money targets always determined by the midpoint of
the longer-run range, the results in Table 4 suggest
that a better approach would be to derive reserve paths
using judgmental multiplier projections and then make
intermeeting adjustments in response to new information.!./
There are, of course, good reasons why the FOMC does not
attempt an abrupt return to the long-run target following
observed discrepanc'ies, but this is a separate issue alto'gether .'!:./
c.
Multiplier-projection techniques versus choice of reserve
aggregate.
(1)
It might also be argued that the money stock could have
been kept under closer control if another reserve aggregate, say the nonborrowed or total monetary base, had been
used as the operating target.
Indeed, at first glance, the
results in Table 4 would appear to provide strong evidence
in support of this view.
The error dispersion statistics
for all the econometric procedures consistently decline as
the reserve aggregate considered is successively broadened.
1/ The extent to which the results in Table 4 bear on the question of the
best technique to use in deriving reserve paths is questionable, as the next
subsection will make clear.
2/ For an analysis of this issue, see Peter Tinsley, Peter von zur Muehlen,
Gehard Fries, and Warren Trepeta, "Money Market Impacts of Alternative Operating
Procedures," in this compendium.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 38 (2)
This position is essentially a criticism of the choice
of the reserve measure for use as a target, rather
than of judgmental projection methods.
Table 4 shows
that judgmental predictions of the multipliers for the
two base measures were similar to those of all the econometric forecasts except the Board's monthly model, which
were somewhat better.
(3) More fundamentally, however, Table 4 is incapable of providing reliable evidence on the question of the best
reserve concept for use as an operating target or on the
question of the best econometric method available for use
in deriving the appropriate level of the reserve target.
The reason simply is that the reported multiplier-prediction errors contain endogenous movements of reserves
away frOJn their predicted values.
These error statistics
are not instructive regarding the closeness of monetary
control in a regime in which a model's predicted level
of reserves was taken as an invariant operating target.
The next subsection will address this issue in depth,
and conduct alternative empirical tests.
c.
The Nature of the Money Stock Prediction Tests
The error statistics of multiplie~ predictions shown in Table 4 are
quite misleading regarding the reserve measure that could provide the closest
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 39 ,,
monetary control, because they include reserve misses as well as money stock
forecast errors. The design of the multiplier tests discussed in the last two
sub~ections underscores the fact that none of the reserve measures has been
'
)
truly exogenous sine~ October 1979.
In the case of nonborrowed reserves,
there were intermeeting adjustments in response to recognized multiplier disturbances, as well as misses of the final adjusted path owing to noncontrolled
l,_
I
)
,-
-
--."'
':> - ,
~
-
factors affecting reserves, like float, or to other considerations.
....J ✓
j
\
"
!. J.
,
Because
-
adjustments 'to ~oiiborrowed reserves tended to be in the opposite 'direction to
',
deviations of the monetary aggregates from target, the prediction errors of
judgmental nonborrowed r~serve multipliers suffer from an upward 'bias .
.., .:.
I
-
\
~
-:J
1
~
.-;
--,. _
~ !.ti",
l- ...,.. _
1
For example, assume that, even th~ugh nonborrowed reserves are main~
1
n.
_1
tained at the i~itial target, the money stock unexpectedly jumps in the first
half of the lntermeeting period to a level that, if maintained, would imply
J ,t.
~
::._
•
a
~
!
;:
,
~~
"-J
f
!
an annual growth rate for the entire intermeeting period 10 percent faster than
I
~
,_... t:.
~
:
1 \
~
.,.);
_., I ,
\
1
;::
targeted.
::1: .._ _
-i I
:]
'r
Now suppose that the Desk in response adjusts the nonborrowed'reserves
.,.
-
target d~~wa;d so that'for the entire period its annual rate of change is 10
percent less than initially targeted.
Assume that, as a result, growth of the
money stock in ~he intermeeting
period is reduced 2 percentage points to 8 perr ._
.,.
J
'
-
.:::,
cent ~bove the targeted gr~wth rate.
Although the intermeeting adjustment to
nonborrowed reserves brings the rate of growth of the money stock 2 percentage
-
>
'
_point-s .clos_er, to_ :target', ~-t also produces an 18 percent error in the initial
-.,.
~
(.
l,
J
., ,pr~pict_ion,.of ,the ann~al, rate of change of the multiplier, rather than the 10
_. .:_
.1.
l
.,
,-
f
-
~-
percent error that would have occurred in the absence of a reserve path adjust-
ment.
.,
1
)
Thus, if nonborrowed reserves literally had been held at their initial
,,
~ , -ta:i;g~t level _throug'hout the intermeeting period, their multiplier error staI
•
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
J
...
,
'
'
J
- 40 tistics shown in Table 4 would have been lower, although the misses of the monetary
aggregates from their intermeeting targets of course would have been larger.
As a result of this inverse relationship between deviations of nonborrowed reserves from initial intermeeting targets and the prediction errors of
the associated multiplier-since October 1979 the correlation coefficient between
the two was -.74--the judgmental multiplier-prediction errors were considerably
larger than the monetary-aggregate intermeeting target misses.
For example,
the mean and root mean squared errors of the initial judgmental M-lB/nonborrowed
reserves multiplier projection were 2.7 percent and 14.9 percent at an annual
rate respectively.
But, as shown in the top row of Table 5, the comparable
error statistics for the miss of M-lB from intermeeting targets built into
this multiplier projection were -0.9 percent and 9.8 percent respectively.
The error statistics in Table 4 for judgmental predictions of multipliers for the other reserve aggregates and for model predictions of multipliers for all the reserve measures also are potentially quite misleading, although
the direction of bias is less clear and more dependent on the particular
judgmental or model technique being employed.!/
To be sure, the results of
1/ The total reserve and nonborrowed base judgmental M-lB multiplier error statistics for the initial projection are close to those for M-lB deviations from
its intermeeting target, while the comparable error statistics for the total
monetary base multiplier are below those for M-lB deviations. For the econometric models, the particular equations subject to the largest errors and, in
the Board and San Francisco models, the various equations' interest elasticities
all play a role. To illustrate how multiplier error statistics like those in
Table 4 could be biased downward, consider a hypothetical example involving total
reserves and abstracting from lagged reserve accounting. If the Federal Reserve
held nonborrowed reserves constant over the intermeeting period but exerted
little administrative pressure on banks borrowing at the window, then shortterm interest rates would react little when the public unexpectedly increased
its desired holdings of demand deposits. Although required reserves would rise,
discount-window borrowings would increase to fill the gap between nonborrowed
and required reserves, raising total reserves. The assumed muting of interest
rate movements would affect other deposits only
(continued on page 41)
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 41 Table 4 are intended to provide a fair test of the relative ability of the
various procedures to predict the different multipliers under these circU1nstances of endogenous movements in various reserve aggregates. Nevertheless,
these statistics are unreliable as a test of the relative attractiveness of
the various reserve aggregates as potential operating targets because the
multiplier errors include endogenous misses of the reserve prediction and
forecast errors of money.
These results also are unreliable as a test of
the relative improvement i~ monetary control that could be obtained by
relying on each particular model for determining the appropriate value of
the reserve target.
We have developed two procedures for circumventing'the problem of
the endogeneity of reserve aggregates.!/
The first makes the working assumption
1/ (continued from page 40). little, even if Regulation Q ceilings made savings
depositors quite sensitive to movements in short-tem interest rates. Thus, total
t'eserves would go up by about the same proportion as demand deposits and the M-1/
total reserves multiplier would remain relatively stable.
This case can be compared with one in which the Federal Reserve maintained
total reserves at a predetermined level. Now, as the assumed surge in the demand
for demand deposits increased required reserves, market•interest rates would rise,
as discount borrowings were offset by open market sales. The rise in interest rates
would, by assumption, induce large outflows of savings deposits, making more reserves
available to support demand deposits. Hence, some of the increase in the demand for
demand deposits would be accommodated automatically. In this latter example, holding total reserves exogenous causes the observed M-1/total reserves multiplier to
increase noticeably, in contrast to the case with a nonborrowed reserves target.
Examples of supply-side shocks can be constructed that give the same result.
1/ These procedures involve stochastic model simulations in which the chosen instrument is held constant at an exogenous level determined in a prior step. Selection
of alternative policy instruments influences the ultimate allocation but not the
total impact of random disturbances on the financial system. This conclusion is
well established in control theory and has also been the subject of a number of
early inquiries in macroeconomics, such as Martin J. Bailey, National Income and
the Price Level: A Study in'Macrotheory (McGraw-Rill, 1962); William Poole,
"Optimal Choice of Monetary Policy Instruments in a Simple Stochastic Macro Model,"
Quarterly Journal of Economics, vol. 84 (May 1970), pp. 197-216; John H. Kareken,
Thomas Muench, and Neil Wallace, "Optimal Open Market Strategy: The Use of Information Variables," American Economic Review, vol., 63 (March 1973 ), pp. 156-72.
A more general discussion of this phenomenon with empirical illustrations and
further references may be found in P. Tinsley and P. von zur Muehlen, "A Maximum
Probability Approach to Short-Run Policy," Journal of Econometrics, vol. 15
(January 1981), PP• 31-48.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 42 that the actual level of nonborrowed reserves may be treated as exogenous,
since it, in fact, has been the instrument under Federal Reserve control.!/
The Board and San Francisco models were then simulated to predict money,
given actual nonborrowed reserves.
Obviously, no disparity between actual
and predicted nonborrowed reserves is allowed to occur.
Hence, the two models
were permitted· to generate money stock forecasts given the actual level of
nonborrowed reserves provided by the Trading Desk.
These predictions were
compared with realized money stock levels.
If nonborrowed reserves are assumed to be determined-exogenously, then
the total reserve and the two base aggregates would be endogenously related to
'
the money stock.
This endogeneity would present a problem for evaluating money
stock forecasts generated using actual levels of total reserves, the,nonborrowed base, or the total base as if they were exogenous.!/
Hence, rather than
-
following this approach, an alternative set of simulations of the Board and
San Francisco models was cond~cted that explicitly treated total reserves, the
nonborrowed base, or the total base as exogenous.
Rather than simulating the
models with the actual level of the three broader reserve measures treated as
exogenous, the model was run using as the exogenous policy variable the prediction of these rese~ve measures derived from the simulation with the actual level of nonborrowed reserves treated as exogenous.
In other words, predicted
1/ Intermeeting adjustments to the nonborrowed reserve path in response to observed money stock deviations from target represent a feedback from currently
evolving errors and violate the assumption of exogeneity. ,The second procedure
described below attempts to correct for this correlation.
2/ From the vantage point of the two monthly models and assuming nonborrowed
reserves are exogenous, actual total reserves, for example, would implicitly contain information about all the structural errors in the model, except for the
error in the demand for currency. If the models then were simulated by treating actual total reserves as the exogenous control variable, the resulting~
ante model prediction of money would in fact be-based on~ post deposit errors
implicitly captured in the level of actual total reserves. Such a procedure
would violate the purpose of the comparison between an~ ante prediction and
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
y
- 43 -
levels of borrowed reserves and/or currency were added to actual nonborrowed
reserves in order to determine in a first step the predicted levels of total
reserves, the nonborrowed base, and the total base.
Thus, from the viewpoint
of both models, levels of these reserve measures were constructed to be internally
conslstent with the assumed exogenous level of nonborrowed reserves.
On the
assumption that the Federal Reserve maintained nonborrowed reserves at the
observed level in an effort to achieve a particular monetary aggregate objective, it follows that if another reserve measure instead had been used as the
operating target, the Federal Reserve would have chosen a setting of that
measure consistent with the same monetary objective.
In computing monetary aggregate errors for this set of simulations,
the predicted money levels were thus the same as those derived from the run
with actual nonborrowed reserves exogenous in the absence of the model errors.
However, the "actual" levels of the monetary aggregates used to evaluate the
forecasts were not based on the actual money levels observed in the data,
which reflect deviations of actual from predicted levels of total reserves
and the two base measures.
Instead, the respective model simulations held
these reserve aggregates constant at their predicted levels, and imposed on
the model the observed errors in all the structural equations in a second
step.
Accordingly, the predicted values of the monetary aggregates in each
of these second-stage simulations represented the levels that the models
suggest actually would have resulted had the reserve measure been held constant
at the assumed exogenous level.
In sum, the money stock error in each simulation for the three broader reserve aggregates was defined as the difference between the model solution
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 44 for the monetary aggregates given the assumed exogenous value for each reserve measure--after imposing the~ post errors in each equation--and the
prediction of the monetary aggregate derived from the nonborrowed reserves
exogenous simulation in the absence of these errors.!/
The basic assumption underlying this procedure-that actual nonborrowed reserves are an exogenous variable--obviously abstracts from deliberate
intermeeting adjustments to the nonborrowed reserve target.
An alternative
procedure to correct for this effect was used with the Board and San Francisco
models.
It involved defining the exogenous level of all the reserve measures
in the first step as the predictions of these reserve measures obtained from
the simulations used to generate the multiplier predictions underlying Table
4.
In that experiment the models were solved for money and reserves given
the judgmental expectation of the federal funds rate in the current month.
Hence, this procedure takes that_predicted level of reserves as the exogenous
level. The remaining step is then carried out just as in the first procedure.
This second procedure is more likely than the first to provide
settings for all the reserve measures that are truly exogenous, since it
eliminates the residual bias in the forecast errors for money based on actual
nonborrowed reserves in the first ~rocedure.!/
Be that as it may, we believe
1/ It should be emphasized that these monetary aggregate errors are not related to any observed levels of the aggregates. They are the errors that
would have emerged according to the models if the specific reserve measure
had in fact been held exogenous at the assumed level in the presence of the
the same errors as actually occurred in all the equations.
2/ Our in1tial expectation was that the first procedure involved a residual
upward bias in the errors using actual nonborrowed reserves relative
to the errors using the,other reserve measures. For example, a positive shock
to money during the control period would at times have induced an intentional
reduction in actual nonborrowed reserves, which would move the predicted level of money further below the realized level. The results of the two procedures
confirmed this expectation for the San Francisco model but not for the Board
model.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 45 that both procedures--reported below in Tables 5 and 6--provide better tests
than Table 4 of the relative merits of various reserve measures as potential
operating targets.
We also believe that either approach better indicates the
usefulness of the Board and San Francisco models for deriving target reserve
paths associated with near-term money stock objectives. The model prediction
errors will indicate how closely the models would have come to predicting levels
of the monetary aggregates if the reserve aggregate had been held at the assumed
exogenous level during the course of a month.
Unfortunately, the Johannes-Rasche technique is incapable of addressing this problem of reserve endogeneity.
The conclusion that the total mone-
tary base would be the best operating target rests on results like those reported in Table 4.!/
But this conclusion must be viewed with considerable skepticism,
given the endogeneity of this reserve measure, even over the period of the new
reserves-oriented operating procedures.
Johannes-Rasche predictions of mon-
ey necessarily are based on actual, rather than exogenized, levels of the
broader reserve aggregates.
The results therefore are not in accordance with the
spirit of the experiments reported in Tables 5 and 6, unless one makes the restri~tive assumption that these multipliers would not have changed if the re-
serve measure had, been held exogenous at a level different from that observed.Y
1
Nor could this counterfactual experiment be applied directly to the
-'l..lt"""Lt
.t:-'--'
.,..,,,..,;::..
'~
:i,J:~.s...-. ......
1...-•r
->-----•-]'-.-,?"'
the judgmental projection procedure.
"'-""
-
_,,
:1--.1"""f<.t
r
1-.
fl
./
t
er~"-
ua.--~.-.::..,,
r'"
Instead, the actual experience of misses
1/ See Johannes and Rasche, "Can the Reserves Approach Really Work?"
2/ One interpretation of this assumption is that the multiplier has a zero interest rate elasticity, and that the full adjustment of the monetary aggregate
to a change in the reserve aggregate occurs in the current month so as to restore the multiplier to its predicted value.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
...r
1
'i.
- 46 of each monetary aggregate from its target, both for intermeeting periods and
for current months, is reported for comparison with the mo~el errors.
At the
risk of redundancy, a review of the selection of these targets follows.
As
discussed earlier, targets for the monetary aggregates are established shortly
after FOMC meetings.
T~ey are based on the average growth rates selected by
the FOMC over the entire interim period, which incorporate the Committee's
desired speed of the attempted !eturn to longer-run objectives following recognized deviations.
As also previously noted, h9wever, near-term growth rate
targets occasionally differ from the specified average growth rate over the
interim ~eriod.
Regarding the monthly targets, half of the current month
typically has elapsed by the time the target is established.
The potential
for significantly affecting the monthly average growth rate by influencing
the behavior of the monetary aggregates in the last two weeks of the month is
limited.
The intermeeting-period targets often reflect some adjustment
for anticipated temporary vari~tions in future money growth around the
FOMC's average path that is associated with lagged effects of past or current
actions or with known special factors.
D.
An Analysis of the Money Stock Prediction Results
Tables 5 and 6 present summary statistics both of the misses in the
growth rate of- eaclf monetary•aggregate ~from~ it!s, -1-ntermeeting~,o.r,.monthly _t..!J-rget;,__
expressed at an annual rate, and of model forecast errors of annualized money
growth.
It should be kept in mind that these statistics represent errors over
-
a particular period-encompassing the 15 intermeeting periods or 13 months
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 47 following the adoption of the new operating procedures.
Generalizations from
these results about the future should be made cautiously.!./ The findings in
both tables may be summarized briefly.
1.
Money stock target misses.
a.
Intermeeting periods. As noted earli~r, the root mean squared
miss of M-lA is 11 perc~nt and of_M-lB is_nea~ 10 percent,
both at annual rates,.
b.
Current months.
The misses from the current-month targets,
which contain estimates for two weeks of realized data, are
only about half as much.
c.
Various monetary aggregates.
(1)
The size of the available error dispersion statistics,
expressed as growth rates, decreases as the monetary
aggregate concept broadens.
(2)
The mean errors are negative for the narrow aggregates,
reflecting large shortfalls in three of the four intermeeting periods from late March through early July. However, they were positive for current-month M-2 misses.
~
2. _Money4stock target misses versus econometric prediction errors.
a.
Conclusions.
A comparison of the statistics in Tables 5
and 6 that summarize model money errors with those that
summarize observed money target misses reveals that model
predictions based on either of the nonborrowed reserve
measures ranged from slightly to somewhat lower than
1/ For a study that attempts to get around this problem by looking at sets of
errors in the Board model that are typical of the experience of the 1970s, see
Tinsley and others, "Money Market Impacts."
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 48 misses of M-lA and M-lB from intermeeting targets.
b.
Caveats.
These results are misleading, however, unless
properly interpreted.
(1)
As indicated earlier (seep. 31), the models enjoyed
certain advantages in the tests relative to actual
experience;
In addition, since the models were simulated
without a constraint on the federal funds rate, these
results ignore the FOMC's upper and lower funds rate
bounds.!}
These bounds were violated for some of the
predictions. Violations of the funds rate limits in
the planning stage, in which the exogenous level of
the reserve target level is selected, mean that these
levels of the reserve measures 'used in the simulations
would not in fact have been chosen if the models alone
had been used to set the operating target. Violations
of the funds rate limits in the execution stage, in
which the models' errors are imposed with reserves held
fixed at the exogenous level, mean that the reserve
aggregate actually would not have been maintained at
the exogenous target level over the control period,
but would have been altered to keep the funds rate
within desired limits.
However, it may be noted that
the San Francisco model violated the federal funds
range in Table 6 by only a minimal amount for the two
nonborrowed measures.
1/ As noted in the previous subsection, the San Francisco model errors summarized
in Table 5 represent an exception for the four months when the actual federal
funds rate, rather than actual nonborrowed reserves, was used in the first
stage of the simulations.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 49 (2)
Even considering these caveats, the value of the
Board and San Francisco models for use in predicting
money given reserves, and hence for deriving reserve
target paths, appears much greater in Tables 5 and
6 than in Table 4.
We believe the results in
Tables 5 and 6 are more relevant to this issue.
3.
of multi lier and money stock errors for econometric
The differences between the models' typical multiplier
errors, shown in Table 4, and typical money stock errors, shown
in Tables 5 and 6, are instructiye.
a.
Nonborrowed reserves.
(1)
Using actual nonborrowed reserves in the Board and San
Francisco models (Table 5) or predicted nonborrowed reserves (Table 6) to forecast the monetary aggregates,
rather than using the expected funds rate to predict
the money/nonborrowed reserves multiplier (Table 4),
resulted in a striking decline in the error statistics
for both the Board and San Francisco models.
1/ The error statistics for the Johannes-Rasche approach of course would be
identical in Tables 5 and 6 to those in Table 4 because this method would have
to use actual levels of all the reserve aggregates, despite their evident endogeneity.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 50 -
(2)
This result indicates that the multiplier statistics
in Table 4 suffer from a marked upward bias owing to
intermeeting adjustments or other factors that involve
a negative correlation between the departure of nonborrowed reserves from its path and the forecast error in
the multiplier.
b.
Total reserves.
(1)
The Board model shows a marked deterioration in the money stock forecasts of Tables 5 and 6, which exogenize
total reserves, compared with the multiplier forecasts of
Table 4, which treat total reserves as endogenous.
(2)
The San Francisco model statistics, in contrast, show a
clear improvement in the latter two tables.
c.
Nonborrowed base.
(1)
The Board model improved somewhat going from Table 4 to
Table 5, but changed little going from Table 4 to Table 6.
(2)
The San Francisco model improved dramatically in the latter two tables, particularly Table 5.
d.
Total base.
(a)
The Board model worsened noticeably in Tables 5 and 6.
(b)
The San Francisco model was about unchanged in the latter
tables.
4.
Results across the econometric models.
for both M-lA and M-lB.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
The statistics are similar
The San Francisco model's errors stack
- 51 up favorably in all comparisons, but the Board model also does
quite well for nonborrowed reserves; except perhaps for M-lA in
Table 6.
5.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
The choice of a particular reserves instrument.
Concerning the
important ques'tion of which' reserve measure would have afforded
the closest monetary control since October 1979 if held at a predetermined 1,evel over each control period, it is convenient to
consider the Board and San Francisco models in turn.
,a.
The Board model results.
(1)
In both Tables 5 and 6 this model shows the best results
'
for ,nonborrowed reserves an~ the nonborrowed base, with
the la'tter slightly better for the n.arrow aggregates
and the former slight!y bett~r or about even for M-2.
(2)
In contrast, the ,error dispersion statistlcs for to'tal
reserves were two to three times as high as for nonbor'
rowed reserves.
Similarly, these total base statistics
were about ~ouble those of their nonborrowed base courtterparts in Table 5 and aiso higher, although less dramatically
so, in Table 6.
~he deter±oration in the,predictlons of_the model Wl\Etn,
the total reserves and base measures are treated as exogenous arises principally ,from the enlarged importance
of demand depos~~ supply-rerated errors.
With a given
.level of total reserves or ·the total base, ,the discoun't
,window is' not permitteit to play its role as a sarety
- 52 valve in muting the impact of supply-related shocks on
interest rates.
With total reserves given exogenously,
for example, the demand deposit supply curve becomes
quite interest inelastic in the model.
But prediction
errors of the average required reserve ratio on demand
deposits and of o~her reserv~ble items cause., shifts i~
this curve, inducing interest rate fluctuations large
enough to produce changes in the quantity of demand
deposits demanded of a similar size.
This effect accounts
for the large prediction errors for the money stock
given exogenous total reserves or total base.
(4)
The system of lagged reserve accounting makes the monthly
average required reserve ratio by type of deposit quite
unstable, as was suggested by the standard deviations
of actual multipliers shown in Table 3, and quite
unpredictable as well.
It is inherently difficult for
a monthly model to capture adequately the effects of
the two-week lag in required reserve accounting •.!/
1/ In the model simulations reported in these tables, the average required
reserve ratio against demand deposits is forecast as an inverse function of
predicted demand deposits. Alternatively, the model was simulated using timeseries models of weekly reservable demand deposit data. This method reduced
the root mean squared forecast errors of the monthly average required reserve
ratio against demand deposits by about 30 percent, and reduced the annualized
percent root mean squared errors of M-lA and M-lB shown in Table 5 in the runs
with total reserves exogenous by about 2 and 4 percentage points respectively.
The errors in runs with actual nonborrowed reserves, however, were little
changed.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 53 In addition, the system of graduated reserve requirementf that prevailed over the period added to the
difficulties of forecasting the average required
reserve ratio on demand deposits because deposit
distribution among banks affected the average ratio
of required reserves to deposits.
Finally, forecast
misses of other reservable deposit and nondeposit items
also made the position of the demand-deposit supply function under any reserve aggregate control variable
more difficult to predict.
This general problem is exac-
erbated with a total reserves or base instrument. Accompanying these reserve instruments is a demand deposit supply
function with a very low interest rate sensitivity.
(5)
'
The first set of memo items in Tables 5 and 6 shows the
money stock errors with the various reserve measures
treated as exogenous at the'same time 'that the average
required reserve ratio on demand deposits and the level
of required reserves against small time and saving deposits and large time deposits are asswned to be known withcertainty.
The resulting error statistics, compared to
those in the body of Tables 5 and&, provide an upper
limit to the improvement in monetary control since
October 1979 that would have arisen if legislated reserve
requirements on demand deposits had been contemporaneous,
uniform, and universal, and if there had been no reserve
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 54 requirements on other deposit categories.
This experi-
ment is relevant because full phase-in of the Monetary
Control Act of 1980 makes reserve requirements on transactions balances more nearly uniform and universal and removes reserve requirements on all other deposit categories except nonpersonal time deposits.
Moreover, the
requirements on these latter deposits could be eliminated at the Board's discretion.
The error statistics
with total reserves and the total base exogenous -reported
in these memo items show a marked improvement over the
comparable errors without such certain knowledge in the
body of the tables.
Indeed, in both tables total reserves
evince quite small prediction errors for M-lA, relative both
to other reserve measures and to M-lB, as might be expected
with required reserves effectively applying only to demand
deposits in the simulation.
The M-lA error statistics
for the total monetary base, however, show less improvement
than those for total reserves.
(6)
I~ CQntra~t, th~ improveme~t in the errors for both of the
nonborrowed measures shown in the first set of memo items is
,trivial in Table 6 and nonexistent in Table 5.
In part, this
result again reflects the success of the discount window in
muting the effects of supply-side-shocks on the money stock
with a nonborrowed reserve measure as the operating target.
However, these results must be interpreted with care, as
~
55 -
they only reflect Jhe particular history of errors
since October 1979.
As noted in the last section's
discussion of Table 3, the portion of variability of
the M-lB/nonborrowed reserves multiplier due to
lagged res~rve accounting for some reason declined
dramatically in fiscal 1980 relative to the experience
of earlier years.
Consequently, the apparently trivial
impact of these institutional changes on the model's
money predictions, given nonborrowed reserves or the
nonborrowed base in Tables 5 and 6, may well be specific
to the unusual pattern of equation errors experienced
over the past year.
(7) These results suggest great caution in putting more dayto-day emphasis'on a total reserves operating target
until the average required reserve ratio on transactions
balances and required reserves against other items
become more predictable.
As just noted, the Monetary
Control Act, particularly after full phase-in, will
certainly help in this regard .
.Y
However, the reinsti-
tution of contemporaneous reserve accounting would seem
1/ The gain in monetary control would be larger still if, after the phase-in of
the new reserve requirement structure under the MCA has proceeded for a few
years, the Board were to impose the supplemental reserve requirement. This a~tion
would raise average reserve requirement ratios on transactions balances and bring
more depository institutions under binding reserve requirements. Without the
supplemental, a sizable fraction of transactions deposits will be at institutions that can meet their reserve requirements with vault cash held for day-today operations.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 56 -
('
l
I''
to be a prerequisite for strictly maintaining a total
'
,
I
reserves or total base operating target, judging from
I
the results in Tables 5 and 6 for
the Board model.
I
In
I'
I
addition, close short-run control over these reserve
I
measures would become more feasible .
(8)
.Y
The second set of memo items in Tables 5 and 6 adds certain knowledge of the function determining the ratio of
I
discount borrowings to deposits to the certain knowledge
I
concerning required reserves embodied in the first
I
I
set of memo items. The resulting error statistics, compared with those in the first s,et of memo items, provide
an upper limit to the improvem~nt in monetary control
I
since October 1979 that would have arisen if the discount!\
window equation had not been allowed to generate any supplyside multiplier prediction errolrs.Y That is, these statistics show how closely M-lA (or M-lB) could have been
controlled if the only relevant,, errors in the model's
11
1/ For discussion of the controllability of various reserve aggregates under
lagged and contemporaneous reserve accounting, see Axilrod and Lindsey,~- cit.
2/ These results could be interpreted' as applying to the situation prevailing
without administrative pressure or arbitrage restrictions but with a graduated
marginal discount rate that rises with increases in borrowings as a percent of
deposits. In this case, the borrowing ratio as a func tion of the spread of the
funds rate over' the discount rate could become quite predictable. For a discussion of variants of such a proposal, see Perry D. Quic~, "Federal Reserve Discount
Window Reforms: Policies Without Administrative Presspre," Board paper, July 1980 ..
For a discussion of this and other proposals to make discount borrowings more
predictable, see Peter Keir, "Impact of Discount Polic~ Procedures on the Effectiveness of 'Reserve Targeting," in this compendium.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1
- 57 equations were for excess reserves and the demands for
currency and demand deposits (or transactions balances).
The M-2 statistics show the precision of control over this
aggregate if, in addition to these errors, random disturbances also affected the remaining components of M-2.
The results with the total reserves or total base
measures held exogenous are, of course, identical in
the first and second set of memo items in Tables 5 and
6.
The money error is the same regardless of whether
or not unexpected movements in discount borrowings occur, since any such movements would have to be fully
offset by open market operations to keep total reserves
or the total base on target.
(9)
For the two nonborrowed measures, a slight improvement
in monetary control owing to certainty about the borrowings function is evident in Table S, but a little larger
improvement appears in Table 6.
These results suggest
that some consideration might be given to a restructuring of the discount window.
These monetary control ad-
vantages would have to be balanced against the disadvantages of any such institutional change . .!/
b.
The San Francisco model results.
The same two procedures
for exogenizing reserve aggregates are shown in the body of
Tables 5 and 6 for this model. These results yield rather similar
conclusions, although it may be reiterated that the error
l/
Ibid.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 58 ,statistics for all reserve measures were lower than those
reported for the Board model.
(1)
As in the Board model, nonborrowed reserves and the nonborrowed base have the lowest error statistics in the
two tables. The nonborrowed base recorded slightly smaller errors than nonborrowed reserves in Table 5, but the
two measures ran a dead heat in Table 6.
(2)
The ,error dispersion statistics for total reserves and
the total base were.about two to three times higher than
for their nonborrowed counterparts inlboth tables.
_( 3)
The money prediction errors for total reserves were much
smaller for this model than for the ;Board model.
In
marked contrast to the Board model, these error statistics actually improved going from Table 4 to Tables 5
and 6.l/
The explanat~on for the better performance of
,total reserves in this model ,appears ,to involve ,a.combination of -three ,factors.
First, the demand deposit .supply
"function in the San Francisco,model is much more interest
elastic than the Board's, which reduces the effects1of
supply-side disturbances on the model's money stock error
,given ,total ,reserves in the last itwo tables.
Second,,
the San Fr,ancisco model us,es staggered ,four-week periods
for r~serves and deposits, affording a more successful
1/ The bulk of the San Fran~isco model's multiplier forecast errors ;given ,the
projected funds rate in Table 4 stems from forecast errors of ,reserves rather
than forecast errors of money. (The last set of memo i:tems·in Tables 5 and 6
indicate that the model's forecasts of money given the projected funds rate
are ,fairly accurate.) These ,reserve.prediction e~rors ,are-el:iminated-going
from Table 4 to Tables 5 and 6.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 59 tr~atment 9f l~gged ,reserve ~c~ounti~g.
Thir~, the model
do~~ not make ~plicit pr~!;littiO]lS of ,r.equ:t,,re~ reserve
rati_,o~ on de~and ~~posits a~d 11tanaged 1i~bilities.
The
1!19d~J. 's :f:.!Jlpli_cit ,forecJt~t.!:1 Qf 1t}:te_s~ r~~:l,.os a_~pear to yield
,b~~t~r results th~n ~tte~~ts ,to
c.
Re~~r;ves "l(ersus ~?S~ ~ellsures.
m~~~l
the ~atios explJcttly.
Jhe ~Qtal ~a~e ,dpes quite~
,bit better ,,than. to,to;tl ,resj:!ry_e~ it_1 ,,t;_he Board !!lodel ~md just a
b_it ~b~j:ter :I.~ ,th_e
~~n
_F~~i;_ic,!~~9 JP9~~1..
.!!?"!J!Y~r_, U!l~,r t1!,e in-
sti tutio~al ~tr,ucture .tJ1a,t ._has ,o],tained since Oc_tober 1979, their
nonborrowed count~rpar,ts were cl~Jtrly ,st1;perior. In both mQ9.els,
the nonborFowed ba~e does a 'l,it pet~e~ than ~onborrowed re~erves
for M-L\ and M-lB ~n Tab~e '5, ~ut this slight advan~age ~t~tually
disapp~ars in the perhaps more ~eliable Table 6.
Althoµgh
these results only apply •to a 13-month period, they do not
suggest that a.change in day-to-day emphasis from nonborrowed
reserves to the non~orrowed base would afford much, if any,
improvement,in monetary controi.
6.
A federal funds rate versus~ nonborrowed reserves operating tariet.
The third set of memo ~tems ~n Tab~es 5 and 6 ,in~icat~s th~ ~:l,.ze
of money errors for ~he Board and San Francisco ~odels takin~ as
given either the actual federal funds rate in the current month
or the judgmental federal funds ra~e prediction ma4e as of midmonth on average.
The similarity of the funds rate and nonborrowed
reserves error statistics for the narrow aggregates with the Board
and the San,Francisco model simulations in both Tables 5 aq_d 6 is
remarkable; the only real divergence appears for the Bo~rd model
forecast of M-lA in Table 6.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 60 a.
This general result is broadly consistent with earlier work,
done both internally and outside the System ..!./
b.
It may be noted that this result can only be interpreted to
indicate that a fixed funds rate operating target would have
provided as close monetary control as a fixed nonborrowed
reserves target (in the absence of funds rat~ constraints)
on the assumption that, at the beginning-of each,control
period, the funds rate target would have been fully adjusted
to the level thought consistent with the average money
target over the period.Y
c.
This result also only holds for the institutional structure
prevailing since October 1979.
In Table 5, the errors
for the funds rate instrument with the Board model are a
little larger than the errors for nonborrowed reserves in the
second set of m~o items, which assume institutional changes to
make reserve requirements and discount borrowings more predictable.
However, in Table 6, the errors for the funds rate are
somewhat below those for nonborrowed reserves in the second
memo item.Y
1/ This literature was initiated by James Pierce and Thomas Thomson, "Some
Issues in Controlling the Stock of Money," in Controlling the Monetary Aggregates II: The Implementation, Federal Reserve Bank of Boston Conference Series
9 (September 1972). For the most recently published extension, see Charles
Sivesind and Kevin Hurley, "Choosing an Operating Target for Monetary Policy,"
Quarterly Journal of Economics, vol. 94 (February 1980), pp. 199-203.
2/ See Axilrod and Lindsey, "Federal Reserve System Implementation," for a
discussion of the realism of such an assumption.
3/ Recall the discussion (on p. 55) of the perhaps atypically large multiplier
errors for nonborrowed reserves over this period, even with these institutional
changes.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 61 -
7.
The feasibility of close short-run monetary control.
a.
The evidence from Tables 4-6.
Close short-run monetary con-
trol over periods as brief as one month is not possible with
either a funds rate or a reserve aggregate operating target
under the current or any reasonably similar institutional
structure. Even the lowest error statistics in Tables 4-6
support this conclusion.
(1)
The San Francisco model's root mean squared prediction
error of M-lB monthly growth rates of 3.2 percent at an
annual rate for the nonborrowed monetary base in Table 5
implies that, over the long pull, in one month out of
twenty the annualized growth rate of M-lA would move outside a band of 12.5 percentage points centered on the
monthly target.
(2)
The lowest root mean squared error for monthly M-lB
growth in the J~hannes-Rasche model, 9.0 percent, implies that, on the average, in one month out of twenty
M-lB growth will vary outside a range of 36 percentage
points centered on the monthly targeted growth rate.!/
,:
The averaging out of monthly errors over quarterly periods.
Y1
<J11..:",4'l::"Cdle' L..... "t1'QtJrc..,{...,.(t....flf\,'.:"~...... ,t._r_'\n.,.
u ..... _ .,..,.. ..,,J
"•~11
.i
.1.
"'
.....
h ~
l~--•
.J;
... ?
..,.
In light of the sizable monthly errors, the degree to which
~uch errors average out over a longer time horizon is of interest.
1/ As emphasized earlier, translating multiplier errors into money errors when
the reserve measure is endogenous is potentially misleading.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
r 1 .,.,, 1
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 62 -
(1)
Table 7 displays summary statistics of quarterly prediction errors of the monetary aggregates expressed
at an annual rate, based on the money errors reported
in Table 5. The intermeeting-period "quarterly" misses
are derived from the nonannualized intermeeting errors
underlying Table 5 by grouping the errors into five
'
sets of three adjoining intermeeting-period errors,
averaging each set of errors,'and then annualizing by a
factor of 4.
The quarterly errors derived from monthly
observations are simple averages of monthly errors over
calendar quarters, expressed at annual rates.
A sizable
reduction both in the bias and in the measures of dispersion is evident for the quarterly average misses of
money from intermeeting targets.
However, this largely
reflects the reduction of the annualizing factor from
12 to 4, although some averaging out of the individual
monthly misses is evident.
The proportional declines of
the San Francisco model's root mean squared errors for
predictions of th~ narrow aggregates from monthly to
However, the Board model's quarterly statistics for M-lA
and M-lB show a more sizable improvement, reflecting essentially no systematic tendency for errors to run on
the same side from one month to th~ next.
If, the monthly
errors were serially uncorrelated, the quarterly root mean
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
squared error on average would be reduced t6 19 percent
of the monthly root mean squared error--a result that
is achieved exactly for the Board's monthly model predictions of M-lB given actual nonborrowed reserves.
(2)
Table 8 presents the comparable quarterly errors for
the models derived from the second procedure's monthly
errors reported in Table 6.
The reductions for the San
Francisco quarterly errors are•similar to the changes
from Table 5 to 7.
The reductions for the Board's
quarterly errors, however, are larger than the changes
from Table 5 to 7, reflecting a negative serial correlation of monthly errors.
)
- 64 IV.
Variability in Interest Rates and Money Demand Analysis*
Many observers have noted that interest rates as well as money
growth have registered enlarged variability since the new operating procedures were established.
Some even have suggested that more variable
interest rates have accentuated the movements of money; others have
argued that the causation has worked in the opposite direction.
This section addresses these issues from the perspective of the
demand for money.
Several alternative money demand equations are examined.
In each, movements in money demand are decomposed by source into the separate
effects on the quantity of money demanded of each of the variables appearing in the equations.
These individual sources include interest rates,
real income, and prices, as well as other variables in several alternative equations, both for quarterly and monthly data.
The size of the
residual errors in these equations also is examined, and an attempt is
made to explain why they occurred.
The accuracy of these models' money
growth predictions over various time spans--monthly, quarterly, and
annually--also is asssessed.
These results serve as background for other papers in the overall
project that address the issue of whether or not attempts to control
money over the past year have produced either cycles or greater volatility
in short-tenn interest rates, in real economic activity, and, through feedback effects, in the monetary aggregates themselves.
One study uses the
Board's monthly model to examine the variability of interest rates and
*Contributors to this section: Helen Farr, David Lindsey, Eileen Mauskopf,
Edward Offenbacher, and Richard Porter, of the Board staff; John Judd and
John Scadding, of the San Francisco Bank staff.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 65 -
money under alternative targeted speeds of return to the longer-run
money target following deviations and under alternative instruments . .!/
This study holds the real sector and price level exogenous.
The assump-
tion of no feedback effects from real economic activity or prices to
money is dropped in another,study, where macroeconomic models allowing
for, various interactions-between the monetary and real sectors are-con-,
sidered. 2 /
Our paper examines predicted growth rates for M-lA derived from
five quarterly and two monthly money demand equations.
The quarterly
money demand equations consist of the Board (MPS) equation, -the Wharton
and DRI equations, one proposed by Michael Hamburger of New York University and one recently developed by Richard'Porter and Thomas Simpson of
the Board staff.
The monthly equations are taken from the Board's and
the San Francisco Bank's money market models.
In their original form, the quarterly equations were estimated
over somewhat different time periods and explain somewhat different monetary concepts.
The properties and predictive performance of the original
vers~ons of the MPS, Hamburger, DRI, and Wharton equations have,been,dis-
:~ussed"elsewhere.lf
:l
The Porter-Simpson equation, which is not discussed
in~thiJ earlier, paper, is 'simila.r_to,the MPS equation but incorporates
a five-year bond rate ratchet variable--with·an increasing elasticity as
J
1/
Tinsley and others, "Money Market ~mpacts."
2/ Jared Enzler and Lewis Johnson, "Cycles Resulting from Money Stock
Targeting," January 1981.
3/ See Jared Enzler, Eileen Mauskopf and Edward Offenbacher, "Other Money
Demand Equations , " October 1980; and Michael J. Hamburger. "Behavior
of the Money Stock: Is There a Puzzle?" Journal of Monetary Economics,
vol. 3 (July 1977), pp. 265-88.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 66 the rate ri$es.
This variable is intended to serve as a proxy for the
opportunity cost of not making cash management in~est~ents.
Thus, the
total interest rate impact in this equation is captured by the sum
of the standard interest rate-term and this cash management variable . .!/
For purposes of comparison, the equations were reestimated over
a common s4mple.pe~iod, 1960:4-1974:2. _ The results_are reported
~~r
a
connnon monetary concept, M-lA, by adding the predictions of a standard
currency equation to those of the MPS, DRI, and Wharton equations, which
were estimated for demand deposits only.
The estimated elasticities of
the various quarterly equations are given in Table 10.
Predicted values of M-lA and their decomposition in te~s of
explanatory variables are,pre~ented in Tables 10-15.
Table 10 summarizes
the results from all these equations for fiscal years 1979 and 1980.
Tables 11-15 present quarter-by-quarter predictions and decompositions.
Predicted M-lA growtp rates are obtained by dynamically simulating the
demand equations beginning in 1974:3, as shown in the "predicted M-l_A"
cplumn in each table.
The column labeled
"pre-1977:4 values" is obtained
by fixing the values of all explanatory variables for 1977:4-1980:3 at
their 1977:3 values.
The predicted growth of ~-lA in this column represents
;: the ~eff~cts,:,of ~ovell\_eqts in~ valu~~- oj all the explanatory variables only.,_ '"'"'
I
up through 1977:3.
..,_
t'
..-
,-
..,
~ j:.
.....
,,.,/
., "'
\ l _ ..,..
.,,_-..
fr<
I'-
In each subsequent column, a single group of explanatory
variables, as labeled, is permitted to take on actual historical values
for 1977:4-1980:3 rather than fixed 1977:3 values.
The figures in
1/ See Thomas Simpson and Richard Porter, "Some Issues ,Involving the
Definition and Interpretation of the Monetary Aggregates," in Controlling
the Monetary Aggregates III (Federal Reserve Bank of Boston, forthcoming).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
,_.
_
,.._
- 67 these columns represent the growth in predicted M-lA due to the actual
movements in only that group of explanatory variables.
In addition,
as a basis for judging predictive perfonnance, each table presents actual
M-lA growth rates as well as an "adjusted M-lA" measure designed to
indicate what M-lA growth would have been if other checkable deposits
and related deposit substitutes had never been introduced.
This adjustment
is obtained by adding to M-lA two-thirds of the increase in other checkable
deposits plus approximately one-fourth of business savings deposits and
one-fifth of state and local government savings deposits.
Prediction
errors are calculated relative to growth in both actual and adjusted
M-lA.
As shown in Table 10, all of the equations except Hamburger's
overpredict actual M-lA growth during each of the two most recent fiscal
years.
Hamburger's equation makes no prediction error for actual M-lA
growth during fiscal 1980, but it underestimates M-lA growth by 1.5 percentage points over fiscal 1979.
Relative to the growth in adjusted M-lA, the
Porter-Simpson equation has the smallest annual prediction errors--overpredicting growth rates during each of these two fiscal years by less
than one percentage point.
However, the quarterly results reported in
Tables 11-15 indicate that, on average, even this equation does considerably worse during shorter periods, particularly the second quarter of
1980; indeed, no equation predicted that quarter's actual decline in M-lAo
The decompositions of predicted growth rates indicate that the
increase of prices is by far the most important factor contributing to
the high growth of predicted nominal money demand.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
This reflects the
- 68 public's attempt to, attain their _desired level of real balances.
In
addition, differences in the, equations' estimated response to price changes
are an important factor in accounting for, their di~ferences in pr,edictive
performance over the past two fiscal years.
The increase in prices of 9.2
percent (logarithmic change) during fiscal 1980, for example, leads to
an increase in predicted M-lA of about equal magnitude in the MPS, Wharton,
and Porter-Simpson equations, other things equal.
By contrast, the DRI
price component amounts to 7.6 percent in fiscal 1980.
While this equa-
tion suggests that most of the adjustment of money holdings ,to price level
changes is completed within a year, nom~nal balances will never grow by the
full 9.2 percent rate of inflation because the long-run price elast~city is
less than unity.
The Hamburger equation implies an increase in M-lA in
fiscal 1980 due to price increases since 1977:3 of 2.6 percent, owing to
its glacially slow speed of adjustment.
In this equation, M-lA takes 16
years to ~om.plete 90 percent of its ultimate change in r~sponse to a
change in prices.
Multiplying its implied coefficients on lagged inflation
by the associated actual inflation rates over the last 16 years yields a
price component of moaey growth over fiscal 1980 of 6.2 percent.
The 3
,percentage point shortfa~l ,from the 9.2 percent actual inflation rate
arises in part because recent inflation rates have been well above the
average rate over the past 16 years.
Hamburger's equation will reflect
a given rate of inflation to the same extent as the other equations only
if that rate is sustained for several decades.
The superior forecasting results of the Hamburger equation for
unadjusted M-lA also depend on the inclusion of the dividend-price ratio
as an explanatory variable.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
The presence of the dividend-price ratio can
- 69 be criticized, on a number-of grounds.
Theoretically, the dividend-price
ratio (or the earnings-price ratio) behaves more like a real rate of
interes~ than a nominal rate but it is the nominal rate of return on an
.,
,.
~
l
, ~-aJ.:ternatdcve, asset, t:J1at should properly be included in the equation.
Moreo~er, ,is ince the-,al,,ternative to holding money balances is an investJ
_
ment to ;be qeld, fpp-p,a ,short, period of time, the appropriate yield is an
-
~
e~r~ings•ra~e,adjustedJ:p~ anticipated nominal capital gains or losses
over~th~~-p~riod.- ~Or!, pra~tical grounds, the difficulty in for~casting
the dividend-price ratio would complicate the policymaking process if the
,ftquatio,n~we~~ µsed fq~ ;his purpose.
In view of these deficiencies, further
•, :- •dts~u~sion C?J· this -,e,quation will be limited.
, , , -· - The -tables ~.lso sho~ that changes in other factors exert a con- siderably ~ore,nodest influence than changes in prices on the average
-gro;~tb_of,,pr~d.ict;e4 M-:1~,' as opposed, to its variability.
The continued,
, ,,- :.but :decl:1,ning, 1r~a\ 1e']fpans!o,n of the economy through early 1980 is generally
reflect~d in ~-pos!ti~e.-b~t declining, contribution to predicted money
,growth. 1 .-l'he .,actua! 1dec~ine ,in real income later in 1980 is reflected in
-a negative c9~tribution~to predicted money in all equations but Hamburger's.
·simflarly,:the ,impact on,the average growth of M-lA stemming from changes
in interest-·rates ,is, in general, considerably smaller than the estimated
impactJarising from-changes in the price level.
This outcome reflects rela-
tively low estimated interest elasticities as well as offsetting movements
.,
·in interest-rates,themselves, especially in 1980.
'r
"
--
On the other-hand, except in the Hamburger equation, the vari-
ability of both the real income and interest rate components have contributed to relatively sizable quarter-to-quarter movements in predicted money
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 70 growth, as indicated by Tables 11-14.
Movements in real 1 income alone,
caused predicted quarterly M-lA growth at an'annual rate to vary over a
3. 1-percentage-point range in the DRI model and up 'to 'a 9 .•2...:-percentagepoint range in the Wharton model.
Changes in interest rates alone caused
predicted M-lA growth to vary over a 3.2-percentage~point range in the
DRI model and a 6.9-percentage-point range in'the •Wharton model.
Interest
rate movements cause predicted M-lA growth to var-y over,~a-10.9-percentagepoint range in the Porter-Simpson equation·,' when the cash management
variable is included in the calculation. 1 /
Regarding the interest rate component, an' 1ep1-sodic- review ~als'o
is warranted.
The rapid rise in interest rates throughithe first quarter
of 1980 induced a ,deceleration in predicted M-lA growth-from 1979:3 to
1980:1 in a range from 0.3 percentage points for the~DRI equation to 5.0
percentage points in the Porter-Simpson equation:
The:decline •in rates
starting in the second quarter is sufficient to'offset the lagged'.interest
rate effects in the MPS equation in that quarter, turning the overall
interest rate impact positive.
In the Wharton, 'DRI~-and Porter-Simpson
equations, the absolute value of the negative interest~rate effect is reduced.
In none of these equations does predicted,M-lA growth in 1980:2
become negative.
Thus, the 3.9 percent decline in actual M-lA in the
second quarter of last year cannot be attributed solely to the effects of
1/ The interest rate impact on money growth in the Porter-Simpson model
is captured by the sum of the last two columns in Table 12. With this
interpretation, the pattern of interest rate effects is-similar to the
MPS, Wharton, and DRI models, although it is amplified.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 7-1 -
current and lagged interest rate movements on money demand in the various
equations.!/
Table 16 ,shows the effects of a dynamic simulation of r the Board's
monthly model.
Movements in interest rates over this whole period are es-
timated to have caused considerable variation in the growth rate of money
demanded.
For example, ~he depress~ng effect of interest rate movements
-on March,money·growth of 7.0 percent has, by June, become.a stimulating
force of 18.9 percent.
Even the quarterly averag~ figures for this model,
shown in the bottom panel of the table, involve considerably more variation in• the interest rate ~ffects than the quarterly models examined
earlier.
However, the implied response of money does not correspond very
well with actual money-growth.
The estimated- impact ·of·interest-.rates
on: M-lA growth by' the secdnd q1:1arter1,has become quite •expansionary,
rather than depressive; an~ the model thus makes a verY,·large error in
that quarter.
For the·year as a whol~f interest•rate effects average out,
ind the model substantially ov~rpredicts M-lA growth,
1/ The issue of whether or not a change occurred in the variability of
the qtta~terly interest rate component in fiscal 1980 can be,examin~d
formally. For each equation, "F" statistics were calculated to test the
null hypothesis of equality between the variance of this COJ!!.ponent_for
fiscal 1979 and fiscal 1980. Against the alternative hypothesis of a
change in the variance in either direction--that is, a two-tailed testthe nult hypothesis of no change in variance could not be, rejected at
the 5 percent significance level in three of the five models--MPS, .
Porter-Simpson, and Wharton. However, in a one-tailed test of the alternative hypothesis of an increase in the variance, all but the PorterSimpson equation rejected the null hypothesis of no change at the ,
5 percent level. Thus, while the evidence is mixed, there _is a hint of
an increase in money variability associated with heightened variability
in interest rates. It should be noted, that the extremely small sample
' size in these tests precludes definitive conclusions on this point.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
•
- 72 -
The large short-fall of actual relative to predicted M-lA
growth in the second quarter of 1980 in all the models discussed so far
may in part reflect the imposition of credit controls in March.
Evidence
on the role of credit controls has recently been provided by John Judd
and John Scadding of the San F~ancisco Bank staff.
As noted in Section
III, they have estimated a monthly demand equation for demand deposits
in which the amount of deposits the public ho~ds reflects not only current
and lagged values of interest rates and of nominal personal income but
also disequilibrium caused by shifts in the supply function of demand
deposits.
Such shifts act like shocks to the demand for demand deposits,
causing the public temporarily to hold more or less than the amount of
demand deposits desired on the basis of longer term considerations.
Judd and Scadding argue that changes in the volume of commercial
bank loans constitute an important source of these demand deposit shocks.
In the San Francisco model,.the banking system is viewed as responding to
I
exogenous changes in the demand for bank loans by changing either demand
deposits or managed liabilities.
The deposits that are created in the
process of new loan extensions essentially are byproducts, held only temporarily until they are spent.
Hence, Judd and Scadding include in their
demand deposits equation the net change in total bank loans to proxy for
~
-
..-,.
....
deposit shocks that leave the public temporarily off its demand function.V
1/ This specification involving the change in bank loans differs from a
epecif~cation involving a level of loans variable. Earlier work by Board
staff found that the level of nonfinancial business demand deposits depended
significantly on the level of business loans. More recent work by the staff
suggests that over some periods the level of total_ bank loans is a significant deteminant of the aggregate demand for demand deposits.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Their equation'was'estimated over the sample period mid-1976 to
September 1979. - Table 16 dis-plays the M-lA "errors from this equation
monthly, quarterly, and annually over the last fiscal year.
)
The equation
-
comes closer to predicting the decline in money in 1980:2 than all the
other equations, lending credence to the view that the imposition of
credit controls helped to reduce money in that quarter.
This conclusion, of course, is conditional on the validity of,
'
the equation
itself.
'
Although
the equation predicted adjusted M-lA growth
quite accurately over fiscal 1980 as a whole, it registered relatively
large errors in the first and final quarters of that period.
Other evi-
dence drawn from earlier periods raises questions about the robustness
of this specific variant of the Judd-Scadding hypothesis.
Attempts to
explain demand deposit movements using this specification in other sample
periods have not met with uniform success.
Using the San Francisco
specification--nominal money'holdings on nominal income--the approach
works only in the 1970s; it perfoms poorly in the 1960s as well as for
the period from 1960 to mid-1974.
Furthermore, a specification that
imposes the property of homogeni ty of d'egree one in prices gives reasonable
results only in the latter half of the 1970s, from mid-1976 to 1980.
On
balance, the explanatory power of the loan shock variable seems limited to
recent years .Y
1/ The San Francisco model 'assumes that errors in the demand for money are
uncorrelated over time, whereas an alternative specification currently
being examined by the Board staff allows these errors to be correlated. In
all other respects, this alternative specification is essentially the same
as the San Francisco model. Preliminary results from this research suggest
that the San Francisco approach may substantially overestimate--even over
the late 1970s--both the size of the initial impact of bank loans on
the money stock and the degree of money stock disequilibrium over time
caused by such loan changes.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
'
In su~~ary, the overall record of the various money demand equations for fiscal,years 1979 and 1980 is somewhat mixed.
For the most
recent two-year period as a whole, the forecasting record is not too bad
by recent historical standards, particularly with respect to adjusted
~-lA.
The Port~r-Simpson equation has the best record, with annual growth
rate errors of less than 1 percent in both years, while the Wharton equation averages less than 1 percent for both years together.
Also as noted,
the San Francisco equation shows an error of less than 1 percent during
fiscal 1980 as a whole.
On the other hand, no single sati,factory explan-
ation for the spectacular overprediction in the second quarter of i980
has emerged.
Some evidence suggests that these developments may well have
reflected in part the effects of the imposition of credit controls on
bank _lending and, in turn, on money.
Thus, while it may be noted that
interest rate movements induced in, the various equations slightly
more
,
quarterly variability in the predicted quantity of money demanded in
fiscal year 1980 than_in the previous fiscal year, our analysis as a
whole indicate~ that other ~nfluences not captured in the standard
equations significantly suppl~ented these effects.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
~ Z E D UPDWifl RATES OF M-la
S'l'ANI:MD DE.VIATlOJS
CUrrent .seasonal Factot'S
PERCan'
60.0
40.0
30.0
20.0
10.0
.a.o
4.0
3.0
2.0
'
1.0
.8
I
1972
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
.6
t {:,
1974 ,_,
1978
, r~ FIISCAL YEARSI
1980
ANNUALIZED GIOWIH:RATES OF M-2
STANUZ\RD DEVIATIONS
CUrrent Seasonal Factors
PERCENT
20.0
10.0
8.0
6.0
4.0
3.0
2.0
1.0
QUARTERLY
.8
.6
.4
.3
.2
1972
'19'76
1974
''
FISCALJI,. YEARS
~
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
'.'\J
1978
1980
CllART 3
.,,. :
-
'
r
ANNUALIZED GROWm RATES OF M-18
STJ\NDMD DEVIATIOOS
'{
I
Inphed Original Seasonals
PERCEN'.f
100
80
60
-40
30
20
10
8
6
4
3
QUARTERLY
2
1u l 1
1972
1974, \
p,·,
t
'
l
,.
,ll / 11,976
1), i
1
(
';- '_, ', ,.,1
FISCAL YEARS
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
l-4.i.1'
llf
,t
1978
1980,
rnART 4
ANNUALIZED GROWIH' RATES OF M-2
srANIWID DEVIATICNS
Implied Orig1.nal Seasonals
PERCENI'
40.0
30.0
20.0
10.0
8.0
6.0
4.0
3.0
2.0
LO
.8
.6
.4
1972
1974
, ,1976
1980
1978
FISCAL
YEARS
l
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
,
.
ANNUALIZED GRafflI RATES OF ~RROWED RESERVES
sr.ANIWID DEVIATIONS
Implied Original Seasonals
PERCENI'
400.0
200.0
WEEKLY
100.0
60.0
40.0
20.0
10.0
6.0
4.0
QUARTERLY
2.0
1.0
.6
1972
1974
1976
FISCAL YF1IRS
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1978
1980
ANNUALIZED GROWill RATES OF 'OC/l'AL RESERVES
STANDARD DE.VIATIONS
Implied Original Seasonals
PERCEN'I'
400.0
200.0
WEEKLY
100.0
60.0
40.0
20.0
M:NIHLY
10.0
6.0
4.0
QUAR'I'ffiLY
2.0
.6
.4
1972
197.1
1976
~
,
FISCAL ¥EARS
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1978
1980
QIART7
M-lB
~
RESERVE MJLTIPLimsl/
S1'ANDARD PEVJATIOOS
Jnplied Original Seasonals
PERCENI'
600.0
400.0
200.0
wm<LY
100.0
60.0
40.0
20.0
tum-n.Y
10.0
'-........_
6.0
4.0
2.0
1.0
.6
1972
.l/ Armualized rates of
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1974
change.
1976
fISCAL '¥EARS
'
1978
1980
awua
I
'M-1B rorAL RESERVE ·MULTIPL1msll
srANDARO DEVIATIOOS
Inplied Original Seasonals
PERCENI',
400.0
200.0
100.0
60.0
40.0
l
' '
20.0
KNIHLY
10.0
6.0
4.0
QUARTERLY
2.0
1.0
.6 ,1
.4
1972
1/ AnnUdl.lzed rateb of chanye.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1974 '
ln6
197b
1980
'
TABLE 1
VARIABILITY OF QUARTERLY MONETARY GROWTH RATES IN MAJOR INDUSTRIAL COUNTRIES
(seasonally adjusted annual rates)
Standard deviation
Mean
Standard devlation/Mean
Sample
Narrow Money
Canada
:ZMl
France
Ml
Germany
Ml
Japan
Ml
Switzerland: Ml
U.K.
Ml
U.S.
MlA
1.763
1.349
1.426
1.976
2.735
2.228
0.690
2.212
2.545
1.907
2.483
1.012
2.866
1.369
0.797
0.530
0.748
0.800
2.703
o. 777
0.504
1973:Ql-1980:Q3
1973:Ql-1980:Q3
1973:Ql-1980:Q3
1973:Ql-1980:Q3
1973: Ql-1980: Q3
1973:Ql-1980:Q3
1973:Ql-1980:Q3
0.931
0.882
o. 714
0.888
1.330
1.879
o. 777
3.501
3.301
1.952
3.013
1.979
3.151
2.316
0.266
0.267
0.366
0.295
0,. 6 72
0.596
0.335
1973: Ql-1980: Q3
1973:Ql-1980:Q3
1973: Ql-1980: Q3
1973:Ql-1980:QJ
1975:Q4-1980:Q2
1973:Ql-1980:Q3
1973:Ql-1980:Q3
Broad Money
Canada
France
Germany
Japan
Switzerland:
U .K.
U.S.
M2
M2
CBM
M2
M2
LM3
M2
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 2
STANDARD DEVIATIONS OF M-lB, RESERVE MEASURES, 'AND ASSOCIATED MULTIPLIERS
MONTHLY AND QUARTERLY AVERAGE GROWTH RATES
SEASONALLY ADJUSTED USING IMPLIED ORIGINAL SEASONAL FACTORS
(in annualized percent changes)
Nonborrowed
Total
Nonborrowed
Total
monetary base
reserves
reserves
monetary base
lcorrelcorrelcorrelcorreI
I
I
I
M-lB I
llation
llation
llation
llation
I
I
I
or I
IMultilcoeffi-1
IMultilcoeffi-l
IMultilcoeffi-l
IMultilcoeffiM-2 IReserveslplierlcient!./IReserveslplie~lcient!./IReserveslplierlcient!..JIReserveslplierlcient.![
I
I
Monthly Growth Rates
M-lB
1971-79 period
1979
1980
6.1
7.6
9. 611
16 .1
16.0
20.2
17 .4
14.1
22.8
--0. 94
-0.88
-0.91
11.6
10.1
9.0
13 .1
11.0
11.2
-0.92
-o. 74
--0. 59
5.8
5.8
6.3
8.0
7.6
10.4
-0.65
-0.38
--0 .47
4.7
4.7
3.4
6~9
8.1
8.1
--0. 50
-0.39
-0.17
4.2
3.5
6.0fl*
16 .1
16.0
20.2
17 .4
15.0
20.3
--0. 97
-0.98
-0.96
11.6
10.1
9.0
12. 7
9.3
11.3
--0. 94
-0.94
--0 .85
5.8
5.8
6.3
7.4
5.5
6.8
-0.82
-0.81
-0.58
4.7
4.7
3.4
6.1
4.8
5.4
-0.73
-0.73
--0 .13 -
--0. 96
-0.91
-0.15
4.2
5.4
4.6
5.4
5.8
6.2
--0. 83
-0.77
-0.33
3.0
3.5
2.1
4.7
5.1
4.6
-0.78
-0.67
0.80
1.9
2.8
2.3
3.2
4.5
4.7
-0.40
-0.54
0.63
--0. 97
-0.95
--0. 01
4.2
5.4
4.6
5.6
4.6
5.7
-0.84
-0.89
--0. 65
3.0
3.5
2.1
4.9 -0.80
3.5 -o. 74
2.41/ 0.91
1.9
2.8
2.3
3.5
2.8
3.3
-0.23
-0.30
0.21
M-2
1971-79 period
1979
1980
Quarterly Average
Growth Rates
M-lB
1971-79 period
1979
1980
3.0
3.8
6.411
8.5
8.3
3. 511
10 .2
9.0
3.1
2.5
4.4
8.5
8.3
3. 5//
10. 5
5.911
M-2
1971-79 period
1979
1980
7.6
2. 711
1/ Correlation between reserve measure and associated multiplier.
7T The increased (or decreased) standard deviation for 1980 compared with that of the 1971-79 period is statistically significant at the 10 percent level.
* The increased (or decreased) standard deviation for 1980 compared with that of 1979 is statistically significant at the
10 percent level:.
..
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 3
STANDARD DEVIATIONS OF RATES OF CHANGE OF FOUR-WEEK AVERAGES OF
M-lB DIVIDED BY FOUR-WEEK AVERAGES OF RESERVE MEASURES, ALL NOT SEASONALLY ADJUSTED
·(in annualized percent changes)
Reserve measure
not shifted
forward
(1)
Reserve measure
shifted forward
two weeks
(2)
Higher variability
associated with lagged
reserve accounting
( 1) - (2)
1971-79
26.9
16 .1
10.8
1979
29.2
22. 5
6.7
1980
34.8
30.7
4.1
1971-79
24.4
12. 7
11.7
1979
28.0
16. 2
11.8
1980
22.4
8.4
14 .o
1971-79
13.6
11.4
2.2
1979
13. 5
11.3
2.2
1980
14 .1
12.9
1.2
13. 5
11.3
2.2
1979
14.0
11.5
2.5
1980
13 .1
11.8
1.3
Reserve measure
Nonborrowed Reserves, NSA
Total Reserves, NSA
Nonborrowed Monetary Base, NSA
Total Monetary Base, NSA
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
,
TABLE 4
MONTHLY AVERAGE ERROR STATISTICS FOR MONTHLY RATE OF CHANGE OF MONEY MULTIPLIERS, NSA
JUDGMENTAL AND ECONOMETRIC FORECASTING TECHNIQUES
October 1979-0ctober 1980
(in annualized percent)
M-lA
Mean
error
FORECASTING TECHNIQUES BY
RESERVE MEASURE
Nonborrowed Reserves
Board Judgmental
Initial Intermeeting Period !J
Adjusted Intermeeting Period l/ !/
Current Month 'lJ
2,6
2.8
1,2
Mean
absolute
error
error
Mean
error
RMS
M-lB
Mean
absolute
error
RMS
error
13 7
10. 7
8.9
16 2
13 5
2 7
3 0
12 7
14.9
9 9
11,7
1 8
8 6
12 0
11 0
Mean
error
n.a.
n a
M-2
Mean
absolute
error
RMS
error
3 9
n a
n a
7 4
n a
n.a.
9.0
Johannes-Rasche
n.a.
n,a
n,a
-4 8
20.7
26 2
-3.4
19.4
23.4
Board Monthly Model
-9 2
19.9
25 9
-9 0
18 9
24 7
-8 5
22 6
26 3
San Francisco Model
13 8
31 0
35 8
13.6
31.4
36 0
n a
n.a.
na
-4.4
-3 5
9.0
6 1
5.0
10.7
8 5
-4 3
-3 4
8 9
10 3
8 2
n a
n a
n a.
n a
n a
6 5
6,1
-1 9
5 0
6 1
0,8
3 5
n a
4 1
n.a
-3 0
14 6
16 6
-2.1
15 5
17 3
Total Reserves
Board Judgmental
Initial Intermeeting Period !J
Adjusted Intermeeting Period l/ !/
Current Month lJ
-1,9
n a
Johannes-Rasche
n a
Board Monchly Model
-3,1
8 7
10 3
-2.8
8 2
9 5
-2.4
11.0
13,3
San Francisco Model
-5.4
15 6
20.0
-5 6
16 0
20.4
n a.
n a
n a.
8.4
10 4
n a
8 1
4 5
10 5
5 6
n a
2.4
n a
n a.
3 1
n a
3 8
-o
2
5 8
7.0
-3 1
8 0
9 0
Nonborroved Monetary Base
'
Bosrd Judgmental
Initial Intermeeting Period !J
Ad~usted Intermeeting Period l/
Current Monlh 'lJ
Y
0.1
9,4
11.5
0 9
9 0
11 6
-0 3
5.1
6.5
0.2
1.0
0.3
n a.
-0 6
9 9
11.7
n a
Johannes-Rasche
n.a.
Board Monthly Model
-3 7
6 9
9 2
-3 5
6.1
8 2
San Francisco Model
2,6
9 0
11 0
2 4
9.4
11 2
n a
n a
n a,
-1,4
7 1
7 5
4.7
8 3
-0,4
-1 1
n a
n a.
1,6
n a
n a
3.0
n a,
n a
Johannes-Rasche
n.a.
Board Monthly Model
-2 1
Total Monetary Base
Board Judgmental
Initial Intermeeting Period l/
Adjusted Intermeeting Period l/ !/
Current Month 'lJ
n a
9 6
-o
-1 3
3
6 1
6 7
5 6
-0 6
4.1
7 2
8 5
4 7
n a.
n a,
-0.5
8,0
9 0
0.8
5 4
6 4
4 0
5 4
-1 9
4.0
s.o
•-1.5
5 0
6 3
3 6.
San Francisco Model
-2,3
6,3
8,3
-2 4
6,6
8,6
n a,
n a
n a
1/ From October 10, 1979 to February 6, 1980, projection errors of old M-1 are reported for M-lA and ~-lB All the percent
errors are annualized by 12 and include the October 29-November 19, 1980 period
2/ Error of initial multiplier forecast,adjusted for intermeeting changes in targeted reserve path
3/ From October 1979 to January 1980, projection errors of old M-1 are reported for M-lA and M-18 and projection errors of old
M-2 are reported for M-2.
n a.--not available
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Ir
•
.
TABLE 5
ERROR STATISTICS FOR THE
ACTUAL VERSUS TARGFTED
Mean
error
FOMC INTERIM MONEY STOCK TARGETS
Intermeeting Period Path
Current Month Path '!:.J
Y
FIRST PROCEDURE
MONTHLY RATE OF GROWTH OF THE MONETARY AGGREGATES, NSA
AND ACTUAL VERSUS PREDICTED FROM ECONOMETRIC MODELS
October 1979-0ctober 1980
(in annualized percent)
M-lA
M-1B
Mean
Mean
Mean
absolute
Mean
RMS
RMS
absolute
error
error
error
error
et"ror
error
M-2
Mean
absolute
error
RMS
error
-0.9
-1.5
8 6
4.9
11 0
6 0
-0.9
-1 0
7 6
4.2
9.8
5 0
n a.
1.2
n a
3 1
n.a
3 4
-2.4
4 6
7.0
-2.1
4.8
6.5
-1.9
5 3
6.4
0.4
4 2
5.4
0.1
4 4
5,6
n.a.
n,a
Total Reserves
Board Monthly Model
Given Exogenous Total Reserves!/
San Francisco Model
Given Exogenous Total Reserves!/
-5.6
17.3
23.1
-4.6
13.2
18.6
-6.4
20.1
-0 8
8 1
10 6
0.5
8.3
10 6
n a
n.a
Nonborrowed Monetary Base
Board Monthly Model
Given Exogenous Nonborrowed Base!/
San Francisco Model
Given exogenous Nonbot"rowed Base!/
-2.0
4 0
6 1
-1 7
3.9
5.6
-4.5
6 8
8.4
1.0
3 0
2 8
3 2
n a
n.a
n.a.
ECONOMETRIC PROCEDURES BY RESERVE
MEASURE
Nonborrowed Reserves
Board Monthly Model
Given Actual ~onborrowed Reserves
San Francisco Model
Given Actual Nonborrowed Reserves
0.9
,
24.4
Total Monetary Base
Board Monthly Model
10 0
12.3
-3.7
7.9
-3 1
9 9
-5.1
13 4
15 5
Given Fxogenoue Total Base!/
San Francisco Model
Given Exogenous Total Base!/
0.4
5 5
8 5
0,4
5 3
8,1
n,a.
n a
n a.
Memo
'-----;------------~;--------------;------------'-.=c..
Board Monthly Model
Required Reserve Ratio on Demand Deposits and Required Reserves Against Savings and Time Deposits Known
With Certainty
Given Actual Nonborrowed Reserves
-2.4
4.7
7 1
-2,2
4,9
6.6
-2.0
4.8
5 9
6,1
-0.8
2.2
2,9
-1 8
5 5
7.0
-3.8
Given Exogenous Total Reserves!/
8 0
Given Exogenous Nonborrowed Base!/
6 8
-2 0
3 8
5 9
-1,7
3,8
5,5
-4 6
8 2
8.5
10 0;..._
Given Exogenous Total Base !/=----_____.,__0_•..,6______5....,...6_ _-=...,7,..,.,...3_...,,.-_o__,7,---...,...,,_-4_,_4_,__-,-_5_9_..,__-2__.2__________
Board Monthly Model Borrowing Ratio to Deposits Function ae Well ae Required Reserve Ratio on Demand Deposits and Required
Reserves Against Savings and Time Deposits Known With Certainty
Given Actual Nonborrowed Reserves
I -2,0
4,4
6,3 I -1,8
4,7
6.1
-1.5
3.7
4.9
8,0
Given Exogenous Total Reserves!/
I
-0.8
2 2
2.9 I -1,8
5.5
7 0
-3.8
6 1
5 3
Given Exogenous Nonborrowed Base!/
I -1.9
3.6
5 5 I -1,6
3 7
-4 3
6.0
7.6
5.9
-2.2
8 5
10.0
Given Exogenous Total Base!/
I O6
5.6
7 3 I -0.7
4 4
Federal Funds Rate
Board Monthly Model
-2.0
5 0
-1 5
4 8
Given Actual Federal Funds Rate
-2.2
6.5
3.4
4 8
6 8
6.6
Given Judgmental Federal Funds Rate
-2 1
5 1
-1.6
3 6
5 0
-2 3
4 9
6 9
San Francisco Model
Given Actual Federal Funds Rate
-0.5
4.2
5.0
-0 6
4 3
5 2
n.a.
n a
n.a
n,a
n a,
Given Jud ental Federal Funds Rate
-0.5
4.0
5.2
-0 7
4.6
5 4
.!. From October 10, 1979 to February 6, 1980, errors for old M-1 are reported for M-lA and M-1B. All the percent errors are
annualized by 12 and include the October 29-November 19, 1980 intermeeting period.
2/ From October 1979 to January 1980, projection errors for old M-1 are reported for M-lA and M-1B and errors for old M-2 are
reported for new M-2,
1J The exogenous level is equal to the model's prediction of this reserve aggregate given the actual level of nonborrowed
reserves.
n a.-not available
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 6
SECOND PROCEDURE
ERROR STATISTICS FOR THE MONTHLY RATE OF GROWTH OF THE MONETARY AGGREGATES, NSA
ACTUAL VERSUS TARGETED AND ACTUAL VFRSUS PREDICTED FROM ECONOMETRIC MODELS
October 1979-0ctober 1980
(in annualized percent)
M-lA
M-lB
Mean
Mean
Mean
absolute
Mean
absolute
RMS
RMS
Mean
error
error
error
error
error
error
error
FOMC INTERIM MONEY STOCK TARGETS
-0.9
Intermeeting Period Path !J
8.6
11 0
7.6
9.8
-o 9
n a.
-1.0
4.9
6.0
4.2
Current Month Path!/
-1.5
5.0
1.2
ECONOMETRIC PROCEDURES BY RESERVE
MEASURE fl
Noiiii'iirrowed Reserves
Board Monthly Model
Given Exogenous Nonborrowed Reserves
San Francisco Model
Given Exogenous Nonborrowed Reserves
6 7
9.5
-2 1
0.1
4 2
5.2
0 0
Total Reserves
Board Monthly Model
Given Exogenous Total Reserves
San Francisco Model
Given Exogenous Total Reserves
-5.5
17.3
23.1
0 8
Nonborrowed Monetary Base
Board Monthly Model
Given Exogenous Nonborrowed Base
San Francisco Model
Given Exogenous Nonborrowed Base
-2.5
6.0
M-2
Mean
absolute
error
n.a.
3,1
RMS
error
n a.
3 4
8 2
-4.9
4.0
4.9
n,a
n.a.
n,a
-4 6
13 2
18,6
-6.0
19 6
24.0
7 2
9,4
0,8
6 8
9 0
n a,
n a.
n a
-0.3
6,9
9 8
-0,4
5.5
7 5
-2.8
8 6
10 5
0 1
4.2
51
0,1
4.0
4.9
n,a
8.4
n.a
10.9
n.a.
Total Monetary Base
Board Monthly Model
11.8
-2.5
10,4 11 -4 2
9.3
8 4
Given Exogenous Total Base
-3.1
14.5
17.3
San Francisco Model
I
9.4
-0.3
6.7
6 4
8 9 I n a
-0.3
Given Exogenous Total Bsse
n a
n.a.
Memo
Board Monthly Model
Required Reserve Ratio on Demand Deposits and Required Reserves Against Savings and Time Deposits
Known With Certainty '}.J
I
'I
7.9 I -4.7
Given Exogenous Nonborrowed Reserves
I -2.2
6,6
9.2
I -1,9
5 8
8 6
10 7
Given Exogenous Total Reserves
I -0.8
2.1
2 9
I -0 7
3.7
4.9 I -3 3
10 2
12.9
7 2 I -3.0
8,3
Given Exogenous Nonborrowed Base
I -0.4
6 8
9 6
I -0.5
5.4
10 2
7.4 I -2.1
Given Exogenous Total Base
I 0.2
5 0
5.8
I -0,1
5,4
9 8
12 .4
Board Monthly Model
Borrowing Ratio to Deposits Function as well as Required Reserve Ratio on Demand Deposits and Required
Reserves Against Savings and Time Deposits Known With Certainty '}.J
Given Exogenous Nonborrowed Reserves
I -0.6
5 5
7 8
I -0 6
5,1
7 0 I -3 4
7.5
9.7
I -0 8
2 1
2,9
I -0,7
3,7
4 9 I -3 3
Given Exogenous Total Reserves
10.2
12.9
Given Exogenous Nonborrowed Base
I -1 4
3,7
5 5
I -1,2
4 0
5.4 I -3.7
5.7
7 4
9,8
Given Exogenous Total Base
I 0,2
5,0
5,8
I -0 1
5.4
7.4 I -2,1
12,4
Federal Funds Rate
I
I
I
Board Monthly Model
I
I
I
Given Actual Federal Funds Rate
I -2 2
4 8
6.8
I -2 0
5.0
6 5 I -1,5
3 4
4.8
Given Judgmental Federal Funds Rate
I -2 3
4 9
6.9
I -2,1
5.1
6 6 I -1 6
3 6
5 0
San Francisco Model
I
I
I
Given Actual Federal Funds Rate
I -0.5
4 2
5.0
I -0.6
4 3
5.2 I n a,
n.a.
n a,
Given Judgmental Federal Funds Rate
I -0 5
4 0
5 2
I -0.7
4,6
5.4 I n a
n a.
n,a,
1/ From October 10, 1979 to February 6, 1980, errors for old M-1 are reported for M-lA and M-18, All the percent errors are
1
annualized by 12 and include the October 29-November 19, 1980 intermeeting period.
From October 1979 to January 1980, projection errors for old M-1 are reported for M-lA and M-lB•and errors for old M-2 are
reported for new M-2
3/ The exogenous l~vel of each reserve measure is equal to the model's prediction of this reserve aggregate given the judgmental prediction of the federal funds rate
n 1,--not available,
Y
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 7
FIRST PROCEDURE
ERROR STATISTICS FOR THE QUARTERLY RATE OF GROWTH OF THE MONETARY AGGREGATES, NSA
ACTUAi VERSUS TARGETED AND ACTUAL VERSUS PREDICTED FROM ECONOMETRIC MODELS
October 1979-5eptember 1980
(in annuall~ed percent}
I
I Hean
I
FOMC INTERIM MONEY STOCK TARGETS
Intermeeting Period Path!/!/
Current Month Path~'!/
I
I
ECONOMETRIC PROCEDURES BY RESERVE
MEASURE 2J
Nonborrowed Reserves
Board Monthly Model
Given Actual Nonborrowed Reserves
San Francisco Model
Given Actual Nonborrowed Reserves
I
Total Reserves
Board Monthly Model
Given Exogenous Total Reserves!/
San Francisco Model
Given Exogenous Total Reserve~
Nonborrowed Monetary Base
Board Monthly Model
Given Exogenous Nonborrowed Base!/
San Francisco Model
Given Exogenous Nonborrowed Base!/
I
I
error
-o,3
-o.5
1
I
I
1 -0,9
I
I 0.2
I
I
I
I -2.0
I
I 0.1
I
I
M-lA
Mean
absolute
error
2 2
1,1
RMS I Mean
error I error
I
1
I
M-lB
Mean
absolute
error
RMS
error
Mean
error
M-2
Mean
absolute
error
RMS
error
1.8
0,9
2,4
-0,4
1.1
n.a.
0.4
n.a.
0.7
1.1 I -0.0
0 8
1,2
-0,6
1,0
1,1
1.2
1,6
n.a
n a.
n.a.
3 1
4.4
-2,8
5.6
6.7
2,2
2 8
n.a.
n.a.
n.a.
7
o.7
1.0
-1,5
2,3
2.5
0.2
0,8
1 1
n.a.
n.a.
n.a.
2.a
1,4
-0,3
n.a
0.8
I
I
I
I
0,9
I
2,0
I
1 5 I
0.1
I
I
I
5,1 I -2 3
I
2.1 l
o.6
-0.1
0 7
0,9
0.1
0 7
1.0
1,1
3 1
I
l
I
-o
Total Monetary Base
Board Monthly Model
-0,8
2 2
-1.2
1,5
1.8
1.2
-1.5
Given Exogenous Total Base!/
2 6
3 1
San Francisco Model
Given Exogenous Total Base!/
0 5
1,7
2,1
0 4
1,6
2,1
n a,
n,a,
n a
Federal Funds Rate 2/
=------+---;..._----='---------'--t------.;;;..;...;.__ _;;;....;.;=---+---'::....;:'-'--....=.:=---=::....;:c....
Board Monthly Model
1,3
-0,8
-0.8
1.0
1,4
0.9
0,6
Given Actual Federal Funds Rate
-0 6
0.7
-0,8
1 2
0,9
-0.9
1.0
-0,6
1.3
0,6
Given Judgmental Federal Funds Rate
0.8
San Francisco Model0,0
1 4
-0.1
1,4
1.1
1.1
Given Actual Federal Funds Rate
n.a.
n.a.
1.5
o.o
o.o
1 2
1.0
n,a
Given Judgmental Federal Funds Rate
1.4
n.a.
1/
Quarterly" errors calculated as averages of three adjoining intenneeting periods, annualized by a factor of 4,
(Averages
of four adjoining internieeting periods give very similar results,)
2/ From October 1979 to January 1980, projection errors for old M-1 are reported for M-lA and M-lB and errors for old M-2
are reported for new M-2,
3/ Quarterly errors calculated as three-month averages of monthly percent errors in each calendar quarter annualized by a
factor of 4,
!!J The exogenous level is equal to the model's prediction of this reserve aggregate given the actual level of nonborrowed
reserves.
n,a,--not available
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 8
SECOND PROCEDURE
ERROR STATISTICS FOR THE QUARTERLY RATE OF GROWTH OF THE MONETARY AGGREGATES, NSA
ACTUAL VERSUS TARGETED AND ACTUAL VERSUS PREDICTED FROM ECONOMETRIC MODELS
October 1979-September 1980
(in annualized percent)
Mean
error
FOMC INTERIM MONEY STOCK TARGETS
Intermeeting Period Pathf/Y
Current Month Path!;J1/
-0.3
-0.5
ECONOMETRIC PROCEDURES BY RESERVE
MEASURE!J
Nonborrowed Reserves
Board Monthly Model
Given Exogenous Nonborrowed Reserves I -o.9
San Francisco Model
I
Given Exogenous Nonborrowed Reserves -0.1
M-lA
Mean
absolute
error
RMS
error
Mean
error
M-lB
Mean
absolute
error
2.2
1.1
2.8
l 4
-0.3
-0.4
0.9
1.1
1.1
M-2
Mean
absolute
error
RMS
error
Mean
error
2.4
1.1
n.a.
0.4
n.a.
n.a.
0.9
0.7
0.8
-0.8
0 8
1.0
-1.7
2.6
2.1
1.3
-o 1
1.0
1.3
n a.
n.a.
n a.
Lil
RMS
error
Total Reserves
Board Monthly Model
Given Exogenous Total Reserves
San Francisco Model
Given Exogenous Total Reserves
-2.8
3.1
5.1
-2 3
3.1
4.4
-2.1
5.4
6.5
0.7
1.8
2.s
0.7
1.8
2.4
n.a.
n.a.
n.a.
Nonborrowed Monetarl Base
Board Monthly Model
Given Exogenous Nonborrowed Base
San Francisco Model
Given Exogenous Nonborrowed Base
-0.1
0.8
1.0
-0.2
0.6
0.6
-0.9
2.6
2.8
-0.1
1.1
l 3
-0.1
1 0
1.3
n.a.
n.a.
n.a
Total Monetary Base
Board Monthly Model
-0.7
1 7
-0.8
1 5
1.9
2.0
-1 2
3 9
4.3
Given Exogenous Total Base
San Francisco Model
o.o
2.2
o.o
1.7
1.6
2.3
n.a.
n.a.
Given Exo enous Total Base
Federal Funds Rate
Board Monthly Model
-0.8
1.0
1.3
0.9
1.4
-0.6
0.7
Given Actual Federal Funds Rate
-0.8
0.6
-0.8
0.9
1.3
-0 6
Given Judgmental Federal Funds Rate -0.9
1.0
1.2
0.6
0 8
San Francisco Model
1.1
1.4
Given Actual Federal Funds Rate
0.0
1.1
1.4
-0.1
n.a.
1.0
1.4
n.a
n.a.
n a.
Given Jud mental Federal Funds Rate
OO
1.2
1.5
0.0
1 'Quarterly" errors calculated as averages of three adjoining intermeeting periods, annualized by a factor of 4.
(Averages of four adjoining intermeeting periods give very similar results.)
From October 1979 to January 1980, projection errors for old M-1 are reported for M-lA and M-lB and errors for old
M-2 are reported for new M-2.
3/ Quarterly errors calculated as three-month averages ~f monthly percent errors in each calendar quarter, annualized
by a factor of 4.
4/ The exogenous level ie equal to the model's prediction of this reserve aggregate given the judgmental prediction
of the federal funds rate.
n.a.-not available.
y
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 9
'
SUMMARY OF EQUATION ELASTICITIES FOR SELECTED
VARIABLES
Equation
Prices
Current Past
Real income
Current Past
MPS
1.000
0
.317
.642
Porter-Simpson
1.000
0
.331
.205
Wharton
1.000
0
.469
-.047
Hamburger
.036
.964
.036
.964
DRI
.211
.584
.128
.355
1/
2/
3/
4/
5/
T-bill, federal funds,
commercial
paper rate
Current Past
Passbook rate
Current Past
-.o5o 2 / -.045 3 / -.017
.002
-.040
-.037
Dividend:erice ratio
Current Past
Bond rate
Current Past
-.112
0
4/
5/
4/
5/
-.030- -.197- -.030- -.197-
-.007
-.019
Other..![_
Current Past
.077
.183
-.039
-1.015
.003
.007
-.060
-.165
-.003
-.053
.026
-.025
.025
-.016
-1.56
-.078
The other variables are as follows: MPS, time trend= -1.52 percent per year; Porter-Simpson, opportunity cost of
cash management proxy with the table entry being the elasticity evaluated at the 1979:4-1980:3 mean of this proxy,
53.23; Wharton, elasticity with respect to the discount rate; DRI, sum of elasticities with respect to lagged stocks
of nonfinancial corporate holdings of Treasury and agency bonds [-.012, -.033 current and past] and with respect
to lagged stocks of other deposits (mainly large CDs, flow-of-funds concept) [-.004, -.123 current and past].
Sum of T-bill rate elasticity (-.041) and federal funds rate elasticity (-.009).
Sum of lagged T-bill rate elasticities. Lagged funds rates not in equation.
Evaluated at 0.10 for commercial paper rate.
Evaluated at 0.05 for passbook rate.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 10
SUMMARY OF SIMULATION DECOMPOSITIONS, 1978:3-1979:3 AND 1979:3-1980:3
(percent rate of change from preceding period based on seasonally adjusted data)!./
Equation
Actual
M-lA
Adjusted
M-lA '!:.J
Predicted
M-lA
Error
Error
(adjusted
Predicted M-lA due to movement in
I
(actual
M-lA
post-77:3 post-77:3
I
M-lA minus
minus
Ipre-77: 4l/ post-77:3 real
interest
predicted) predicted) !values
prices
income
rates
other~
I
1978:3-1979:3
A.
MPS
5.1
7.0
7.8
-2. 7
-0.8
Porter-Simpson
5.1
7.0
7.9
-2.8
Wharton
5.1
7.0
7.5
Hamburger
5.1
7.0
DRI
5.1
MPS
I
I
0.2
8.5
2.7
-2.2
-1.:::
-0.9
o.o
8.9
1.1
-1.5
-0.4
-2.4
-0.5
-0.5
8.5
2.2
-2.6
3.6
1.5
3.4
3.2
1.7
0.7
-1.9
7.0
9.3
-4.2
-2.3
0.7
6.4
1.7
-0.8
2.2
4.0
5.0
7.4
-3.4
-2.4
0.1
9.4
o.o
-o. 7
-1.2
Porter-Simpson
4.0
5.0
5.7
-1.7
-0.7
o.o
9.3
-0.6
-1.2
-1.6
Wharton ,
4.0
5.0
6.2
-2.2
-1.2
-0.5
9.5
-0.6
2.0
Hamburger
4.0
5.0
4.0
o.o
1.0
2.7
2.6
0.6
-1.9
DRI
4.0
5.0
8.9
-4.9
-3.9
0.2
7.6
0.3
-0.3
1979:3-1980:3
B.
1/
1.2
The percent changes are changes in natural logarithms multiplied by 100.
2/ The adjustments are based on the assumption that the introduction of ATS accounts nationwide, NOW accounts in the North-
'1J
4/
east, and savings accounts for businesses and for state and local governments has had a depressing effect on M-lA
growth. An adjusted M-lA series is constructed as an estimate of what M-lA would have been if these new deposit categories had not been created. The series added to M-lA essentially consists of two-thirds of other checkable deposits,
one-fourth of business savings deposits, and one-fifth of state and local savings deposits. Since the latter two series
tend to fluctuate with interest rates, the actual adjustment is made by assuming that these series grow at half the rate
of increase of nominal income after the initial introductory phase for each.
That is, for a given row (equation), this column represents the effects of pre-1977:4 movements in all the variables
for which there are entries in subsequent columns.
"Other" variables differ by equation and are as follows: MPS, time trend; Simpson-Porter, interest rate ratchet
variable as proxy for the opportunity cost of cash management services; DRI, prior-period stocks of nonfinancial
corporate holding~ of government bonds and time deposits (primarily large CDs).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 11
BOARD'S QUARTERLY ECONOMETRIC MODEL EQUATION (MPS).!./
(percent rate of chang~ from preceding period, ~nnualized_, based on ,seas~nally adjusted data)~
Error
'(adjusted, ,1 _____P_r_e_d_i_c.;;..te_d_..;;M.;;..-_l;.;;.A;;....;;d;.;;;u-=-e....,;;.to,;....;m;;;.;o;..;v:..:e:;;:m:;:e;,;;:n~t--=:;in~---M-lA
I
post-77:3 post-77~3
minus
lpre-77:4
post-77:3
real
interest
time
predicted) values~~--P~r_i_c_e_s_ _.;;..in~c.;;..o_m~e;;.__ _~r~a~te;;;.;s;;......_~t~r~e~nd=--
Actual
M-lA
Adjusted
M-lA
Predicted
M-lA
Error
(actual
M-lA minus
predicted)
1978:4
5.5
6.8
8.8
-3 .3
-2.0
0.4
7.7
5.4
-3.3
-1.3
1979: 1
0.2
3.3
7.3
-7.1
-4.0
0.3
8.5
2.6
-2.8
-1.3
7. 2
9.2
7.6
-0.4
1.6
0.2
8.6
1.6
-1.5
-1.3
1979:3
7.8
8.9
6.7
1.1
2.2
0.1
8.2
1.0
-1.4
-1.2
1979:4
4.~
4.7
4.3
0.2_
0.4
0.1
8.2
0.9
-3.6
-1.2
Date
!979:2
'
''
1980:1
4.8
5.5
6.6
-1.8
-1.1
0.1
9.2
2.2
-3 .5
-1.2
1980: 2
-3.9
-2.8
8.7
-12.6
-11.5
10.1
-2.0
1.8
-1.2
1980:3
11.0
12.4
9.0
2.0
3.4
o.o
o.o
2.4
-1.2
1978:3-79:3
5.1
7.0
7.8
-2.7
-0.8
0.2
8.5
2.7
1979:3-80:3
4.0
5.0
7.4
-3.4
-2.4
0.1
9.4
o.o
-
8.9
-0.7
---------------------------------'-----------------------1/ The MPS equation is for demand deposits. Predicted values from a separate currency equation were added to demand
deposit predicted values to obtain M-lA predicted values.
/I
2/
The percent changes are changes in natural logarithms multiplied by 100.
3/ That is, pre-1977:4 movements in all variables that appear in subsequent columns.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 12
PORTER-SIMPSON EQUATION!/
(percent rate of change from preceding period,
annualized, based on seasonally adjusted data)!/
Error
(actual
M-lA minus
redicted)
Error
(adjusted I
Predicted M-lA due to movement in
post-77:3 post-77:3 post-77:3
M-lA
I
real
interest
cash
minus
lpre-77:41/ post-77:3
income
predicted) !values
prices
rates
mana ement!!/
Actual
M-lA
Adjusted
M-lA
Predicted
M-lA
1978:4
5.5
6.8
9.2
-3. 7
-2.4
o.o
8.3
2.6
-o. 7
-1.2
1979:1
0.2
3.3
8.6
-8.4
-5.3
8.9
1.4
-1.2
-0.6
1979:2
7.2
9.2
6.2
1.0
3.0
8.9
-0.4
-2.1
-0.2
1979:3
7.8
8.9
7.0
-0.8
1.9
o.o
o.o
o.o
8.2
0.6
-2.3
0.4
1979:4
4.5
4.7
6.4
-1.9
-1. 7
o.o
8.0
1.1
-0.7
-2.1
1980:1
4.8
5.5
3.0
1.8
2.5
o.o
9.1
0.8
-0.6
-6.3
-3.9
-2.8
3.4
-7 .3
-6.2
o.o
10.2
-3.1
-1.9
-1.9
11.0
12.4
9.8
1.2
2.6
o.o
8.7
-1.4
-1.4
4.0
1978:3-79:3 5.1
7.0
7.9
-2.8
-0.9
8.9
1.1
-1.5
-0.4
1979:3-80:3 4.0
5.0
5.1
-1.7
-0. 7
o.o
o.o
9.3
-0.6
-1.2
-1.6
Date
1980:2
19~0:3
1/
2/
3/
4/
'
The Porter-Simpson equation is for M-lA.
The percent changes are changes in natural logarithms multiplied by 100.
That is, pre-1977:4 movements in all variables that appear in subsequent columns.
This variable is a proxy for the opportunity cost of investment in cash management services.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 13
WHARTON MODEL EQUATION.!./
(percent rate of change from preceding period,
annualized, pased on ~eason~lly adjusted data),?_/
Date
1973:4
1979~1
Actual
M-lA
Adjusted
M-lA
Predicted
5.5
6.8
7,6
0.1
3.3
M-lA
7.7
Error
( actuc.l
M-lA minus
predicted)
-2.1
-7 .5
1979:2
7.2
6.9
0.3
1979:3
7.8
7.0
0.8
Error
(adjusted I- - - - - - - - Predicted
M-lA due to movement in
-------,,,..,,,---,,-------,e--=--M-lA
post-77: 3
post-77:3
1
minus
lpre-77:4l/
post"77:3
real
interest
predicted) I values
prices
income
rates
-0.8
I
I -0.1
7.6
4.0
-3.9
8.3
2.1
-2.3
8.9
0.2
-1.4
8.1
2.4
-2.7
7.5
2.4
-s.o
9.0
1.3
-4.3
10.6
-5.2
-0.5
9.4
-1.0
1.8
8.5
2.2
-2. 6
9.5
-0.6
-2 .o
I
-4.4
1 -o .5
1.9
I
I -o.9
I
I -o.7
'
I
1979:4
4.5
4.7
4.4
0.1
0.3
1980:l
4.8
5.5
5.5
-0.7
o.o
1980:2
-3.9
-2.8
4.4
-8.3
-7 .2
1980:3
11.0
12.4
9.8
1.2
2.6
1978:3-79:3 5.1
7.0
7.5
-2.4
-0.5
079:3-80:3 4.0
5.0
6.2
-2.2
-1.2
I -o.6
I
I -o.6
I
I -o. 5
I
l -o.4
I
I
I -o.5
I
I -o.5
I
1/
2/
3/
The Wharton equation ~s for demand deposits. Predicted values from a separate currency equat1on were added
to de'l!and deposit predicted values to obtain M-lA predicted values.
The percent changes are changes in natural logarithms multiplied by 100.
That is, pre-1977:4 movements in all variable~ that appear in subsequent columns.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 14
HAMBURGER EQUATION!./
(percent rate of change from preceding periodA annualized,
based on seasonally adjusted data)!/
Error
Predicted M-lA due to movement in
(adjusted I
post-77:3
post-77:3
M-lA
I
post-77:3
real
interest
minus
lpre-77: 4l/
income
rates
predicted) I values
Erices
Actual
M-lA
Adjusted
M-lA
Predicted
M-lA
Error
(actual
M-lA minus
predicted)
1978:4
5.5
6.'8
3'.4
2.1
3.4
3.3
1.3
0.7
-1.8
1979: 1
0.2
3.3
3.5
-3.3
-0'.2
3.2
1.5
0.7
-1.9
1979:2
7.2
9.2
3.3
3.9
5.9
3.1
1.8
0.6
-2.2
1979: 3
7.8
'
8.9
4.0
3.8
4.9
3.0
2.0
0.7
-1. 7
1979:4
4.5
4.7
3.4
1.1
1.3
2.8
'2. 2
0.7
-2.4
1980:1
4.8
5.5
3.8
1.0
1.7
2.7
2.5
0.8
-2.2
1980:2
-3.9
-2.8
3.1
-7 .o
-5.9
2.6
2.8
0.4
-2. 7
1980:3
11.0
12.4
5.4
5.6
7.0
2.6
3.0
0.4
-0.5
1978:3-79:3 5.1
7.0
3.6
1.5
3.4
3.2
1.7
0.7
-1.9
2.7
2.6
0.6
-1.9
Date
I
,
I
1979:3-80:3 4.0
1/
2/
3/
5.0
4.0
o.o
1.0
I
-
,I
I
I
'
The Hamburger equation is for M-lA.
The percent changes are changes in natu~al logarithms multiplied by 100.
That is, pre-1977:4 movements in all variab'les that appear in subsequent_c?lumns.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
,,
TABLE 15
DRI EQUATION.!/
(percent rate of change from preceding period,
annualized, based on seasonally adjusted data)Y
Error
(actual
M-lA minus
predicted)
Error
(adjusted 1____....;P=-r=-e=-d=-i=-c::..:t:;.:e:..:d~M=--..;:l;;.:A:.....:::d..::u..::e_t=-o=--m=-o=-v:....:e:..;:m=.:e:..;:n:..:t:.....:::i;.:;:n:___ __
M-lA
I
post-77:3 post-77:3
minus
lpre-77:4
post-77:3
real
interest
4/
predicted) lvalues;u~--p~r=-i~c~e=-s=---~i=-n=-c~o;.:;:m:..:e_ _ _r=-a~t~e=-s=---~o~t:..:h~e~r-
Actual
M-lA
Adjusted
M-lA
Predicted
M-lA
1978:4
5.5
6.8
8.3
-2.8
-1.5
1979:1
0.2
3.3
9.0
-8.8
-5. 7
1979: 2
7.2
9.2
9.7
-2.5
-0.5
1979:3
7.8
,8. 9
9.0
-1.2
--0 .1
1979:4
4.5
4.7
10.1
-5.6
-5.4
Date
1
1980: 1
4.8
5.5
9.2
-4 .4
-3. 7
1980:2
-3.9
-2 .8
6.0
-9.9
-8.8
1980:3
11.0
12 .4
9.4
1.6
3.0
1978:3-79:3
5.1
7.0
9.3
-4 .2
-2.3
1979:3-80:3
4.0
5.0
8.9
-4.9
-3.9
I 1.0
I
l 0.1
I
I o.5
I
I o.4
I
I , o.3
I
I 0.2
I
1 0.1
I
I 0.1
I
5.2
2.2
-1.3
1.3
6.0
1.7
-1.2
1.8
6.4
0.8
-1.3
3.2
6.7
1.2
-0.6
' 1.4
6.8
1.3
-1.4
3.1
7.3
1.0
-0.9
1.5
7.7
-0.9
-0.9
-0.1
1.8
0.1
7.6
I
I 0.1
I
I 0.2
6.2
1.5
-1.1
2.0
7.6
0.3
-0,4
1.2
I
1/
2/
3/
4/
The DRI equation is for demand deposits. Predicted values from a separate currency equation were added to demand
deposit predicted values to obtain M-lA pred~cted values.
The percent changes are changes in natural logarithms multiplied by 100.
That is, pre-1977:4 movements in all variables that appear in subsequent columns.
Includes prior-period stocks of nonfinancial corporate holdings of government bonds and time deposits (primarily large
CDs).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 16
BOARD'S MONTHLY MODEL
(percent change from preceding period, annualized).~/
Actual
M-lA
A.
Error
(adjusted
M-lA
minus
predicted)
Monthly results
1979:10
1979: 11
1979:12
1980:01
1980:02
1980:03
1980:04
1980:05
1980:06
1980:07
1980:08
1980:09
B.
Adiusted
M-lA '!:.J
Error
(actual
M-lA
Predicted
minus
M-lA
predicted)
I
I
Predicted M-lA growth due
to movement in
post-77:09
I
post-77:09
lpre-77:10 post-77:09 real personal
income
interest rat
I values~ prices
-=--------------------
1
2.2
4.6
5.6
3.6
9.3
-1.8
-17.8
0.6
11.5
7.7
19.4
12.5
2.2
4.4
6.3
4.7
9.6
-0.6
-15.1
-0.6
13.4
9.9
20.8
14.5
0.4
2.1
5.4
7.3
6.6
3.5
2.9
20.4
26.7
20.2
18.7
11.2
1.8
2.5
0.2
-3.7
2.7
-5.3
-20.7
-19.8
-15.2
-12.5
0.7
1.8
2.3
0.9
-2.6
3.0
1.3
-18 .. 0
-21.0
-13.3
-10 .. 3
2.1
3.3
4.5
4.8
-3.9
11.0
4.8
5.5
-2.7
12.5
2.6
5.8
16.2
16.6
1.9
-1.0
-20.1
-5.6
2.2
-0.3
-18.9
-4.1
4.0
5.3
10.1
-6.1
-4.8
-4.1
Quarterly results'!../
I
I
1
I
I
I
I
I
I
1
I
I
I
I
I
I
1979:4
1980:l
1980:2
1980:3
c.
Annual results
1979:3-80:3
I
I
I
I
I
I
I
I
I
o.o
o.o
o.o
o.o
o.o
o.o
o.o
o.o
o.o
o.. o
o.o
o.o
12.6
13.4
13.4
13.1
14.4
13.4
14.9
15.0
9.8
13.2
10.9
o.o
o.o
o.o
o.o
12.6
13.7
14.4
11.3
o.o
13.0
11.8
-0.8
-9.6
-9.4
-7.3
-1.5
-4. 7
-2.6
-3.9
-5.8
-7 .o
0.4
-3.9
-7.0
-4.7
12.s
18.9
14.1
6.9
-0.1
-1.8
-1.0
-7 .2
-3. 7
-L3
-1.2
-8.8
-2. 7
-5.2
-6.7
-1.5
8.4
6.8
-3 .1
o.o
1/ The percent changes in the first five columns are standard percent changes. However, in subsequent columns percentage
changes are measured as a percent of the predicted M-lA (third column) level for the previous period.
2/ M-lA adju~ted essentially equals M-lA plus two-thirds of other checkable deposits, one-fourth of business savings
deposits, and one-fifth of state and local savings deposits. Since the latter two series tend to fluctuate with interest
rates, the actual adjustment is made by assuming that each of these series grow at half the rate of increase of nominal
income after the initial introductory phase for each.
3/ That is, pre-1979:10 movements in all variables that appear in subsequent columns.
4/ Growth rates computed from quarterly averages of levels.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 17
SAN FRANCISCO MONTHLY MODEL
(percent change from preceding period, annualized)!/
Actual
M-lA
A.
2.3
4.6
s.s
3.6
9.4
-1.9
-17.7
0.7
11.4
7.8
19.3
12.3
2.2
4.4
6.3
4.7
9.6
-0.6
-15.1
-0.6
13.4
9.9
20.8
14.5
-2.3
-5.5
4.6
10.1
8.7
9.3
-3.2
-5 .o
-5.1
0.1
1.3
-12. 7
1. 7
4.9
-6.6
0.3
-3.7
4.8
o.s
-1.0
6.5
14.4
19.0
16.0
s.o
4.5
9.9
5.8
-4.0
0.3
2 .-6
-10.1
0.4
6.9
-4.5
1.8
-1.5
Quarterly results 4/
1979:4
1980:1
1980:2
1980:3
c.
Error
(adjusted
M-lA
minus
predicted)
Monthly results
1979:10
1979: 11
1979:12
1980:01
1980:02
1"')80: 03
1980:04
1980: OS
1980:06
1980:07
1980:08
l<l80:09
B.
Adjusted
M-lA 2/
Error
(actual
M-lA
minus
Predicted
M-lA
predicted)
4.5
4.8
-3.9
11.0
-2. 7
12.S
-2.4
4.9
0.2
16.5
4.0
5.3
4.6
s.s
7.2
0.6
-2 .9
-4 .o
Annual results
1979:3-80:3
!J
-0.6
0.7
I___P_r_e_d_i_c_t_e_d_M_-_,,1,,,,,A~g'-=-r_o_w.;..th_d_u_e-=-et~o'----m...:o_v...:e_m:..;:.e..:.cn.;..t___in=--1
lpre-77: 10
I values 3/
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
o.o
o.o
o.o
o.o
o.o
post-77:09
nom. pars.
income
post-77:09
interest
rates
post-77:09
changes in
bank loans
o~o
o.o
7.5
8.2
8.4
8. li
7.6
7.3
6.1
5.4
5.3
6.6
6.6
6.9
-4.5
-4.6
-3. 7
-2.9
-2.9
-5.2
-3.1
5.9
8.1
7.5
4.9
2.1
-5.1
-8.9
-4.2
3.2
4.7
-5.3
-8.0
-12.4
-6.7
0.3
7.5
7.0
o.o
o.o
o.o
o.o
8.0
7.8
5.6
6.7
-4 .3
-3 .6
'3.7
4.9
-6.1
0.9
-9.1
4.9
o.o
7.0
o.o
-2.5
0.0
o.o
o.o
o.o
0.0
The percent changes in the first five columns are standard percent changes. However, in subsequent columns percentage
changes are measured as a percent of the predicted M-lA (third column) level for the prevlous period.
2/' M-lA adjusted essentially equals M-lA plus two-thirds of other checkable deposits, one-fourth of business savings
deposits, and one-fifth of state and local savings depoelts. Since the latter two series tend to fluctuate with inter
est rates, the actual adjustment is made by assuming that each of these series grow at half the rate of increase of
nominal income after the initial introductory phase for each.
3/ That is, pre-1979:10 movements ln all variables that appear in subsequent columns.
4/ Growth rates computed from quarterly averages of levels.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
APPENDIX
ERROR EXPREs'sroNS IN TABLES 4 -
f)./
Table 4
Multiplier Errors
Forecasting Technique
Board Judgmental
Initial Intermeeting Period
ln mAct _ ln miniPred
=
AdJusted Intermeeting Period
Act
adJPred
1nm
- 1nm
=
Current Month
Johannes-Rasche
Board Model and
San Francisco Model
Act
IniPred
ln mcm - ln m
cm
+
=
(ln -J,-ct - ln MTarget)
(ln RAct _ ln RadJTar)
=
(ln -J,-ct _ ln MTarget)
cm
cm
IniTar)
(ln RAct
cm - lnR cm
Act
Pred
ln m
m
cm - ln cm
Act
Pred
ln mcm - ln m
cm
I-'
0
0
=
(ln ~ct
cm
Tables 5 and 6
Money Stock Errors
FOMC Interim Money Stock Targets
Intermeeting Period Path
ln }('-ct _ ln MTarget
Current Month Path
ln -J,-ct _ ln MTarget
cm
cm
Board Model and
San Francisco Model
"Act" ln MPred
ln M
cm
cm -
1,/ All level errors annualized by a factor of 1200.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
ln MPred)
cm
RPred)
(ln RAct
cm - ln cm
- 101 REFERENCES
1. Stephen ,i\xilrod and David Lindsey. "Federal Reserve System Implementation
of Monetary Policy: Analytical Foun4ations of the New Approach," Federal
Reserve Board, processed; presented at the Denver Meeting of the American
Economic Association, September 6, 1980; forthcoming in American Economic Review,
Papers and Proceedings, May 1981.
2.
Martin J. Bailey. National Income and the Price Level, New York, 1962.
3. Jared Enzler and Lewis Johnson. "Cycles Resulting from Money Stock Targeting," January 1981.
4. Jared Enzler, Eileen Mauskopf and Edward Offenbacher. "Other Money Demand
Equations," October 1980.
5.
Helen T. Farr. "The Monthly Money Market Model," FRB working paper, July
1980.
t1ichael J. Hamburger. "Behavior of the Money Stock: Is There a Puzzle?"
Journal of Monetary Economics, vol. 3 (July 1977), pp~ 265-88.
6.
7. James M. Johannes and Robert H. Rasche. "Predicting the Money Multiplier,"
Journal of Monetary Economics, vol. 5 {July 1979), pp. 301,25.
8. ______. "Can the Reserves Approach to Monetary Control Really Work?"
April 1980.
9. John Judd and John Scadding. "Contribution to the Study of the Monetary
Control Experience under the New Operating Procedures,'' forthcoming.
10. John H. Kareken, Thomas Muench, and Neil Wallace. "Open Market Strategy: The
Use of Information Variables," American Economic Review, vol. 63 {March 1973),
PP• 156-72.
11. Peter Keir, and others. "Impact of Discount Policy Procedures on the Effectiveness of Reserve Targeting," January 1981.
12. Fred J. Levin and Paul Meek. "Implementing the New Operating Procedures:
The View from the Trading Desk,'' February 1981.
13. James L. Pierce. "Making Reserves Targets Work," in Controlling the Monetary Aggregates III, Federal Reserve Bank of Boston, forthcoming.
14. James Pierce and Thomas Thomson. ''Some Issues in Controlling the Stock of
Money," in Controlling the Monetary Aggregates II: The Implementation, Federal
Reserve Bank of Boston Conference Series 9, September 1972.
15. William Poole. "Optimal Choice of Monetary Policy Instruments in a Simple
Stochastic Macro Model," Quarterly Journal of Economics, vol. 84 {May 1970),
pp. 197-216.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 102 -
16. Perry u. Quick. "Federal Reserve Discount Window Rf:'forms:
Admi;:1istrutive Pte~sures" Board paper, July 1980.
Policies Without
17. Jqhn Scadding and John Judd. "The Disequil;l.brium Demand for Money," forthcoming.
1!=3. Thomas Simpson and Richard Porter. "Some Issues Involving the Definition
and Interpretetion of th_e Uonetary Aggregates," in Controlling the Monetary
Aggregates :rr, Fen~ral R~serve Bank ◊f Boston, forthcoming.
19. Charles Sivesind and Kevin Hurley. "Choosing an Operating Target for Monetary
Policy,'' Quarterly Journal of Economics, vol. 94 (February 1980), pp. 1'19-203.
20. Pet-er Tinsley and Peter von zur Muehlen. "A Maximum Probability Approach
to Short-Run Policy," Journal of Econometrics, vol. 15 (January 1981), pp. 31-48.
?.l. Peter Tinsley, Peter von zur Muehlen, Gehard Fries, and Warren Trepeta.
"Money Mark.et Impacts of Alternative Operating Procedures," January 1981.
22. Thomas o. Thomson, J;:imes L. Pierce, and Robert T. Parry. "A Monthly Money
Market Model," Journal of Money, Credit and Banking, vol.'7 (NovembeJ;" 1975),
PP• 411-31.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-
T
TREND AND NOISE IN THE
MONETARY AGGREGATES
David A. Pierce
Special Studies Section
Division of Research and Statistics
December 1980
Any views expressed or errors contained herein are solely the responsibility
of the author, who is grateful to William P. Cleveland, Darrel w. Parke and
several other staff members for helpful comments, to Gerhard Fries for computational assistance, and to Valerie Watkins for typing.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
SUMMARY
This paper reports results of an investigation attempting to m~asure
the le~th of run over which short-run fluctuations in aggregates may, with
reasonable probability, be said to reflect change in trend.
The results are
enumerated after the next paragraph and are illustrated by Figure 1, which
shows th~ relationships between the sizes of the fluctuations in M-lA (annualized grQwth rates) and the number of months ~ecessary before that degree of
fluctuation reflects, with 70% probability and (in parenthesea on vertical
axis) 95% probability, a change in trend.
(Values for M-2 may be obtained
by replacing numbers on the vertical axis by entries about 75% as large.)
As ~n eJ!:am.ple of interpreting this figure, if trend, "growth rate in M-lA (or
M-lB) ha~ been 5% (seasonally adjusted annual rate) and a current month's
figure is 8%, we could not say with even 70% probability that a change in
trend had occurred since that would requ~re a 4.5% deviation from the current
trehd, contrasted to our observed 3% deviation.
Examining Figure 1 sho~s
that it would require two months of growth averaging 8% to say with 70%
probability, and four months of (average) 8% growth to say with 95% probability, that the trend was now different from 5%.
The paper does not develop a specific procedure for the estimation
or specifica,tion of trend.
Insteaq, the m~asu"tes oJ noise d~veloped are used
to assess the plausibility of de~artures from a desired or hypothesized value
of the trend.
For example, ~he presumed 5% trend in the previous paragraph's
illustration of Figure 1 could be the desired or targeted M-lA growth path
over the current period.
The incoming M~lA figures would then be examined
relative to the noise they are likely to contain (as in Figure 1) to see if a
statistically significant departure from 5% growth had occurred.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
i
Deviation from
trend growth rate
5%(10%)
4%(8%)
3%(6%)
2%(4%)
1%(2%)
Length of run
(in months)
1
Figure 1,
2
4
5
6
Required deviation of observed ~rowth rate from trend to
say with 70% (95%) probability that a change in trend has occurred,
M,... tA and M-lB
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
m½P
nrincipal findings of this qtunv are as follows.
1.
Annualized growth rates determined from the first published monthly
monetary aggregates have estimated standard deviations of 4-1/2% for M-lA and
M-lB [or Sl.5 billion, assuming a level of $400 billion for these aggregates]
'
and 3-1/2% for M-2, attributable to noise (error, uncertainty, randomness)
I
in these series.
Transitory variation and seasonal factor uncertainty are
the principal contributors to this noise.
2.
Growth rates over longer periods than one month have markedly
decreasing noise levels.
The estimated standard deviation of noise in an
annualized three-month growth rate is about 1.7% for M-lA and M-lB and 1.3%
for M-2.
For six-month growth rates the analogous measures are 0.8% for M-lA
and M-lB and 0.6% for M-2.
3.
Noise in levels of monthly data has an estimated standard
deviation (not annualized) of 0.29%for M-lA and M-lB and 0.23%for M-2 or
about± $1 billion and+ $4 billion for current levels of these series.
These figures steadily decrease to 0.13% and 0.10% (± $.5 billion and
l
± $1.6 billion) for six-month averages of the aggregates.
4.
For changes in weekly data (M-lA and M-lB) the estimated noise
standard error is± $3.3 billion.
-
Transitory variation and seasonal factor
uncertainty ~re again the main contributors.
5.
There is substantial negative correlation between transitory
variations and revisions in seasonal factors, providing further evidence that
factor revisions tend to smooth the series.
Moreover, considerable informa-
tion on the seasonal revision is available at the time of initial publication
(from a concurrent adJustment), so that it is evidently possible to construct
improved first-published monetary aggregate data with decreased noise.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
ii
~he m~thod of this study was to construct and use models for the
monetary aggregates (particularly M-lA, M-iB and X-2) to separate the firstpublished versions of these serie$ into trend and noisa components.
The
noise is assumed to be composed pf errors in seasonal adjusttnent; irregular/
transitory variat~ons; and, to a lesser e~tent, sampling errors.
These com-
ponents are analyzed separately and jointly to derive an overall measure, the
standard deviation, of tpe total of th~se sources of uncertainty on weekly
and monthly bases and over several months.
Then, regarding trend as the series net of these sources of error/
uncertainty, there are three quantities such that one can in principle be
determined from the other two:
(a)
Length of time period of the fluctqations.
(b)
Size(s) of the fluctuations.
(c)
Probability that a change in trend has occurred.
In particul~r the probability that a change in trend has occurred is the
probability that the fluctuations,in the aggregate cannot be accounted for by
noise alone, when the noise is measured relativ~ to its standard deviation.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
iii
1•
INTRODUCTION
That the monetary aggregates are subject to considerable uncertainty,
irregularity, noise, error, etc. is known to virtually all observers of these
series.
Such movements obscure changes in the underlying more permanent
aspects of these series, such as changes in trend.
The problem addressed in
this paper ig the extent to which this obfuscation necessarily takes place,
versus the extent to which a given sequence of movements in the observed
aggregates can impart information concerning trend, that is, the assessment
of experience about length of run over which short-run fluctuations may,
wtth reasonable probability, be said to reflect changes in trend.
There are three matn dimensions to this problem:
(1) the size(s) of
the fluctuations, (2) the length of run (number of weeks or months) over which
the fluctuations or movements occur, and (3) the probability that a change in
trend has occurred.
In principle, given suitable definitions of the trend and
non-trend components of the aggregates, any of these three elements is determined from the other two; for example, if the trend growth rqte in M-lA has
been 5% and in the last three months the observed growth rate averages 10%,
what is the probability that the trend is now in excess of 5%?
Or, more in
line with the phrasing of the a1,ove quotation, how many months of an observed
10% growth rate are required to ensure, with 70% probability (or 95% probability), that a change in trend (from 5% growth) has occurred?
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
To answer such questio~s we require the following:
(a)
A notion of what we mean by trend.
(b)
An enumeration and measurell).ent of the major ways in which
an observed money supply figure can depart from this trend.
- 2 -
Concerning (a), we regard trend simply as the "underlying" or "true"
series (or series growth) that would be observed except for the presence of
error or noise.
The trend is thus unobservable; however, we can still make
jnferences about possible values or ranges for the trend, as indicated in the
third paragraph of the summary.
This approach has several features, vis~ vis
the alternative of constructing an explicit model for trend based on prior
knowledge and assumptions relating the aggregates and other variables.
There
is substantial uncertainty and disagreement regarding the appropriate specification of such relationships, for example, whether monetarist or Keynesian or
whether and how shifts in money demand have occurred.
Indeed a major cause
of this uncertainty is errors in variables of the type that produce the
random fluctuations in the aggregates under investigation here.
Any estimate
of trend is sensitive to these assumptions and is itself subject to error.
Furthermore, many sources of error or randomness in the monetary aggregates
are already accorded a non-structural treatment.
For example, a major source
of uncertainty is due to the seasonal adjustment process; and, whether appropriately or not, published seasonal factors for (say) demand deposits are
determined from other demand'deposit values and not from relationships to
time deposits, interest rates, etc.
Finally, the problem as posed is in a
sense statistical, to address whether a change in trend has occurred rather
than why.
This problem is most directly addressed by regarding trend as the
series net of various sources of noise (error, uncertainty, randomness),
which are now described,, and to measure the extent and impact of this noise.
Concerning (b) above, there are several reasons why an observed
monetary aggregate series, even after seasonal adjustment, will depart from
its underlying trend; we will distinguish the following:
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1.
~ranqitory variation.
Irregular, evanescent fluctuation in a
data series, due to causes extraneous to those related to our concept of the
series.
This ppenome?on was examined in some detail by the Committee on
Monetary Statistics [l] a~d, by Porter et al. ([6]).
2.
Sampling and reporting error.
The true series is a population
total (for example bank deposits) and only a sample from this population is
av~ilable, from which an estimate of the population figure is constructed.
An important example for monetary statistics has been the presence of nonmember
banks that report their deposits only one week each quarter.
Further~ore,
member and nonme~ber banks alike may commit reporting errors, which may or
may not be discovered at a later date.
3.
Seasonal adjustment error.
This is partly conceptual insofar
as we do not know very well what we want to remove from a series, the seasonal
adjustment techniqu.e may be faulty, and even the best method generally provides
only an estimate of any "true" seasonal factor that we are able to specify.
And the first· published data on the aggregates have a further source of error
because of subsequent revisions of the preliminary seasonal factors.
A second classification of noise in the monetary aggregates is
according to whether (a) it exists only in preliminary or first-published
data and is eliminated in a subsequent version of the series, or (b) it is
imbedded in the final data, as well as in any preliminary versions of the
series.
Revisions are errors that are discovered and removed from preliminary
data serieq when further information qubsequently becomes available, an
example being revisions due to improved estimates of the qeasonal factors.
Remaining (unobservable) sources of noise include transitory variation and
parts of seasonal adjustment and sampling errors.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-
!,_
-
~he orohlem of detecting a change in tren~ i~•~n_q~g~i~g one and
is in general based on an analysis of current and recent data~
It is thus
necessary to measure the extent of uncertainty or randomness in the preliminary monetary aggregate data, including' both revisions anci' final-data error'.
The basic framework of this study is as~follows. _Let
mi
denote
the first published, seasonally adjust~d monetary aggregate (usually in
logged form).
Write this as the sum
(1.1)
of the trend mt and all sources of noise, randomness, uncertainty, error,
irregularity, etc., nt•
distribution
~
A major task is then to' estimate the -probability
or the standard deviation, assuming normality~- of the
noise nt and of successive averages of this term.
Given this, we could
then state how large the noise term would need to be in order to have (say)
a 70 (or 95)% probability of a change in trend:
this would occur if the
observed value of this term were larger than its standard error' (or twice- '
its standard error).
Results of this type (see Figure 1) show the· tradeoff,
between the size of a fluctuation and the length of time over which that
size of departure would need to persist, in order to signal, at~ g~ven
probability level, a trend change.
Even a single'week's number- could
strongly indicate such a change if deviant enough (for example,' the $9
billion increase on August 6, 1980), whereas a more modest change in trend
or level would show itself only after several weeks or months.
The estimation of the standard deviation of the noise is accomplished in Section 3, after investigating the component noise sources
(seasonal, transitory, sampling) in Section 2.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Also in Section 3 are
the basic results (summarized in Figure 1 and Tahle 2) relating the size
of fluctuations, length of run, and probability of a change in trend.
Section 4 presents some results for weekly data.-
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 6 -
2.
MEASUREMENT OF NOISE COMPONENTS
We enumerated several ways in which an observed monetary aggregate
series, especially as first published, can depart from its trend, namely,
transitory variations, seasonal factor errors, and sampling and reporting
errors.
The composite of these is the overall noise term in the representa-
tion of the first published aggregate as the sum
(2 .1)
of trend and noise.
To estimate the standard deviation of nt, it is necessary to evaluate
the variances and covariances of the sources of error and randomness comprising
it.
We therefore write nt as the sum
Y
nt = -rt + 0 t + F' t + e: t
of the following components:
(a)
revisions rt, due mainly to seasonal factor revisions but
also to such things as more complete reporting, correction
of reporting errors, and heretofore, benchmarking; 1/
(b)
historical seasonal factor errors ot;
(c)
transitory variations Ft; and
(d)
sampling errors e:t (as with benchmark revisions, much
less important after implementation of the Monetary Control
Act).
1/ The minus sign in front of rt in equation 2.2 reflects the fact that
the revision rt itself is added to the preliminary data to obtain the
revised data, so that the revision error present in the preliminary data
is the negative of rt•
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
(2 .2)
7 -
~he ~etermination of the variation and covariation of these noise
components is based on autoregressive-integrated moving average (ARIMA)
models fitted to the monetary aggregates.
The series analyzed were M-lA,
M-lB, and M-2, using monthly data from 1973 through 1979 inclusive.
The
models were fit on the changes in logarithms (approximately the rates of
gro~th) of these series.
The main feature of these models is that they
are all similar to a model given by Cleveland [2], which accurately characterizes the X-11 seasonal adjustment procedure.
This means that inferences
for such quantities as the seasonal adjustment error and the transitory
variance can be based on the known characteristics of the Cleveland model,
as X-11 is the primary means of seasonal adjustment for these series.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 8 -
Seasonal Adjustment Error
Le~ Xt denote a ~ seasonally adjusted monetary aggregate series,
assumed to be written in logged form as
(2 .3)
where mt is the trend, st the unknown true seasonal ~omponent, and the
"irregular" component et =
sources,
~
t + et represents noise from other-than-seasonal
The first published seasonally adjusted series, as in equation
2. 1, !a then
(2 .4)
and the final seasonally adjusted ser~es (after seasonal factor revisions)
is
(2. 5)
wheres~ ands{ are preliroinary and ftnal seasonal factors.
Writing equa-
tion 2.4 as
m~ = xt - si =mt+ (st - s{) + (si - si) + et
(2 .6)
=mt+ ~t - rt+ rt+ €t
shows explicitly the contribution of seasonal revisions and final seasonal
adjustment error to the overall ~oise nt in equation 2.2.
Thus the uncertainty
in monetary aggregates stemming from sea~onal adJustment may be broken down
into seasonal factor revisions (rt) and error in the final seasonal factors
(&t)•
Applying results in Pierce [3] to the new aggregates it was found
that
where cr* is the standard deviation of the year-over-year difference in tqe
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
monthly change uxt
= Xt
- Xt-1 of Xt, that is, the standard deviation or
ARIMA model for the aggre-
~Xt - ~Xt-52, a quantity that appears in the
gate series.
These standard deviations were found to be!/
a*
for
=- ( .0052, .0052, .0043)
(M-lA, M-lB, M-2 );
thus
crr = (.17%, .17"%, .13%)
for
(M-lA, M-lB, M-2) .
(2. 7)
The interpretation of this result is that if, for example, the first published
M-lA figure were $400 billion then a 95% confidence interval for the final
revised figure (due only to seasonal revisions and ignoring benchmarking and
other effects) would be $400 billion [1 ± 2crrl or $400 billion± $1.4 billion.
Similarly, for the error ot in the final seasonally adjusted data
(which is also present in the preliminary data as in equation 2.2) it was
found that
= (.10%,
.10%, ,08%)
for
(M-lA, M-lB, M-2).
Note that the~e error figures and resulting confidence limits are
expressed as perce~tages of the levels of the series, as they are computed
from the series' logarithms.
However, the quantities of greatest interest
are usually the growth rates of the aggregates, which,are essentially the
changes in the logs of these series.
For these rates of change,it is thus
the standard devtation of
-
1/ The size of the seasonal factor revisions can also be measured empirically
if enough first-published and revised data are available. This is not possible
for the new aggregates, however. Thus it is noteworthy that in Pierce [3]
the model-based and empirical revision standard errors for old M-1 were found
to be in ciose agreement with each other.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- ,n -
that is of interest, anrl sirni1arlv of
The basic result relating standard deviations for a series Ut and its series
of changes ~ut is that
(2 .8)
where Pu(k) is the lag-k autocorrelation of u, or the correlation coefficient
between ut and Ut-k (here k=l).
For the seasonal adjustment errors, autocor-
relations of rt are given by Pierce [3]' and those for Ot were kindly
supplied by W.P. Cleveland.
In particular Pr(l)
= 0.52
and p 0 (1)
= -0.27,
whence
crvr = (.17, .17, .13)
for
(M-lA, M-lB, M-2)
crv 0 = (.16, .16, .13)
for
(M-lA
and
all in percentage terms.
'
M-lB
'
M-2)
,
Note that the positive serial correlation coefficient
for rt tends to hold down the value for crvr (and in fact to make it essentially
the same as or), whereas the negative value of p 0 (1) tends to increase crv 0 ,
both relative to the values obtaining if these seasonal 'adjustment errors
were not autocorrelated.
AJso note that all growth-rate results are not
annualized.
An interpretation analogous to that following equation 2.7 is
that if a first published seasonally adjusted monthly growth rate for M-lA
were 0.5% (6% annualized), a 95% confidence interval for the final rate (again
ignoring benchmarking and other revisions and errors) would be 0.5% ± .34% or
(1.9% to 10.1%) annualized.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 11 -
Sampling and Reporting Brror
Sampling error arises because data are available for most of the
8,700 nonmember banks only,on the call reports once each quarter, except for
a sample of 600 nonmember banks that report their deposit balances each
week.
The sampling error for monthly data has been estimated at $320 million,
or less than 0.10% of M-lA or M-lB.
,In addition, there are benchmark revisions
as new data from the quarterly call reports are used to update the nonmember
bank deposit estimates.>'·,
We shall not deal further with sampling error in this study.
It
is relatively small, and since successive sampling/benchmark errors are very
highly autocorrelated (some statistics on this were kindly supplied by Darrel
Parke), the effect or these errors on growth rates would be even smaller.
(On the other hand, we caution that occasional reporting errors, which can
occur in member or.nonmember bank data, would behave much as transitory
variations discussed earlier.)
Moreoyer, the Monetary Control Act greatly
decreases'the importance of sampling and of call report data.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 12 -
Transitory Variation
Over time the monetary aggregates are subject to very short run
variations that bear little or no relation to the economy in -general!.
'These -
kinds of variations are called transitory bPcause they are 1fl~eting •iu- nature
and provide no information about underlying economic processes.
Such vari-
ations were studied by the Comm!ttee on Monetary Statistics [l] and in
greater detail by Porter et al. [6].
Widely ranging1 estim:ates,of transitory
standard deviation were found, depet'lding on the frequency- of, data employed,
in the model and on the model employed for the systematic part of the series.
A single precise definitio~ of transitory variation does not exist.
In the present study we are interested in separatin:~f all 'short ...run-' irregular'
variation from the longer-run, more slowly varying part of tbe- series (the
trend-cycle) and from the seasonal part of the series.
It thus seems reason-
able to label whatever part of the series that is purely random, or-serially
uncorrelated, as transitory.
values of the aggregate.
Such a component is unrelated to past or· future
This concept of transitory variation has the· furthe'r
feature that the Cleveland model for X-11 incorporates an irregular component
that has this property, that it is the random or serially uncorrelated component of the aggregate with maximum. variance (Tiao and Hillmer [8]).
In
this sense the transitory component is similar to the irregular component
estimated by the X-11 procedure.
The qualification needed in adopting this
approach is that~ s~rially uncorrelated component may still be related to
other series, such as interest rates, so that the identification of such a
relationslu.p could alter what is labeled as irregular or as trend, as discussed
in Section 1.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
.,.. J.3 -
As before we denote the transitory component by ~t•
From calcu-
lations with the Cleveland model,
and thus the trans+tory standard deviation for the three monetary aggtegate
series undeh study is
ar, = (.21%, .21%, .17%)
for
(M-lA~ M-lB, M~2).
As the trans~tory cpmponent is by definition serially uncorrelated,
it follows that Pf(k) = O, and the~efore from equation 2.8 the transitory
standard ~eviation for the growth rate of each series is ✓ 2 times a~,
or
O'iJF,
= (,30%, .30%, .24%)
for
(M-lA, M-lB, M-2)
This is the large~t stngle source of uncertainty iq the monetary aggregates,
co~pared with seasonal aqj~stment and sampling error.
For example, a reported
M.,..lA growth rate of 8% could, within~ 1 standard-error limit, be as low as
4.4% or as.high as 11.6% due to irregular or transitory variation.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 14 -
'3.
OVER-ALL J',W,ASTTRR~ Oli' NflTR'li'.
~easures of randomness or noise in the first published seasonally
adjusted aggregates can now be obtained, based in part on the standard
deviations derived in Section 2.
The additional information needed is the
contemporaneous covariances and correlations between the various sources
of uncertainty, for ass~s~ing noise in aggregates at a single month, and
autocorrelations and lagged cross-correlations o~ these components, for
aggregates measures over several months.
Noting as in the previous section
'
that the inclusion of sampling/reporting error would have relatively little
effect on the result and that much of the role of sampling'and call report
procedures is being eliminated, we concentrate on transitory variations ~t
and the two seasonal adjustment errors rt and ot•
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-
, 'i -
Noise in Single-Month Monetary Aggregate nata
It is shown in Pierce [3, 4] that seasonal revisions are uncorrelated with error in the final seasonal factors, and that rt and ot are
negatively correlated with the transitory component St•
-
In particular, for
'
!kl < 12 it is shown in Pierce [4] that
where vk is..-:the k~h term:..iii. the moving average of the X-11 _seasonal adjustment procedure, that is, the coefficient of xt-~ when the X-11 seasonal factor
is written as
In particul:ar, --v 0 == 0 .181: for k
=0
1
and
(-.80, -.80, -.52) x 10-6
have·
we
for
(M-lA, M-lB, M-2).
Now the variance of nt is the sum of the variances of ot, rt, and St plus
twice the nonzero covariances, an~ since rt enters with a negative sign in
equation 2.2, these covariances cancel each other.
Thus the variance of nt
is the sum of the three component variances, and taking square roots,
crn = (.29%, .29%, .23%)
for
(M-lA, M-lB, M-2).
For growth rates the calculations are analogous except that the first order
serial covariances also need to be taken into account.
We have, as in equa-
tion 2.8,
=
(.377%, .377%, .297%)
for
(M-lA, M-lB, M-2),
as Pn(l) = 0.15 for M-lA, Band 0.14 for M-2.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
(3. 2)
~nnualized, these grqwth rate standard deviations are 4:~2% for
M-lA and M-lB and 3.58% fo+ M-2, which we round to 4-1/2 and 3-1/2% respectively.
As an example, if a first-published seasonally adjusted monthly
M-iA growth ri;tte i.s 8%, a 70% confidence interval for the "true" M-lA, or
the trend, would be
8% ± a or
(3.5% to 12.5%),
and a 95% confidence intervals for the trend rate of growth would be8% :!: 20
or (-1% to P%)
Alternatively, if the previous trend were anywhere from 3-1/2 to 12-1/2%, we
aoulQ not say even with 70% probablity that a change in trend had- occurreq,
on the basis of a one-month observation of 8% growth.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
rable 1 summarizes the results of Section 2 and this discussion.
_
TABLE 1. Stanoard deviations of noise
in monthly mon~tary aggregates
.
M-lA and M-lB
Levels
M-2
Growth rates
Growth rates
Levels
se
lrce
Percent
:nsitory
iati.ons
.21
.8
.30
3.6
.17
2.9
.24
2.9
sonal
isions
.17
.7
.17
2.0
.13
2.2
.13
1.6
or in final
sona.1
(:ors
.10
.4
.16
1.9
.oa
1.4
.13
1.6
.29
1.2
.38
4.5
.23
3.9
.30'
3.6
al
Billions 1/
of dollars-
Percent
Based on a level of S400 billion.
Based on a level of $1.7 tril+ion.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Annual rate
(percent)
Percent
Billions 2/
of doll'ars-
Percent
Annual rat
(percent)
- 17 -
Noise in Data Spanning Several Months ,
The example concluding the previous subsection shows clearly that
uncertainty, noise, error, irregularity, etc. in the monetary aggregates are
so great that very l~ttle can be said about trend or underlying movements in
these series on the basis of one month's movement in the current data.
We
are now able to address _th7 question motiv~ting this paper, of how long it
does take before these fluctuations in the observed-data begin to signal '
possible_changes.in trend.
We shall determine the standard deviation of k-month averages and
k-month growth rates in M-lA, M-lB, and M-2, as a function of k.
To do this
we note that the transitory component ~tis serially independent and that,
while the seasonal adjustment errors Ot and rt were autocorrelated at lag
k=l (Section 2), their autocorrelations were small at lags k
> 1,
at least
up to the annual lag of 12; thus this is also true for nt·
If nt is the total noise term for a one-month average of the
,aggregate at month t, then (1/2)(nt + nt+l) is the noise for a two-month
average of the aggregate, and in general the- total deviation from trend for
a k-month average is
(3.3)
It is straightforward to determine the standard deviation of nik) given the
standard deviation and the lag-1 autocorrelation ~f nt (below equation 3.2)
and the fact that other autocorrelations of nt can be neglected.
Table 2
shows the resulting standard deviations, along with those of the k-month
growth rates.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Table 2.
Standard deviations of noise ink-month averages
and growth rates of monetary aggregates
Y
In pe,:-cent
k-month average
M-2'
M-lA and
M-lB
.-29
.23
.38 (4.5)
.30 (3 .6)
.22
.17
.21 (2 .s)
.16 (1.9)
3
.18
.14
.14 (1. 7)
.11 (1.3)
,4'
- .16 -
.12
.10 (1.2)
.08 (1.0)
5
.14
.11
.08 (1.0)
.06 (0. 7)
6
.13
.10
.07 (0.8)
.os
M-lA and
k
' - M-l'B
1
a>
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
k-month growth rate
2
1/
-
'
M-2
Annualized growth rates are shown in parentheses.
(0.6)
- 18 -
The analogous calculation for growth rates is simpler since the
noise in an average k-month raie is just
(3 .4)
multiplied by 12 to annualize.
The standard deviation of equation 3.4 is
thus Gn lf/k, or 0.41/k for M-lA and H-lB and 0.32/k, for ,M-2, e-,ccept for k=l
where for k=l where the nonzero tag-1 serial correlation~of {nt} also plays
a role.
This is a more rapid decrease, ask increa~e&, ~han for the average
levels, an etfect that is apparent in Table 2, where :for k=l the growth-rate
standai;-d deviations are higher but where for larger k they drop well below
those fork-month averages.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Figure 1 is a plot, against k, of the third column of Table 2.
- 19 -
4.
NOISE IN WEEKLY DATA
'
It was seen how the noise in the ' monetary
aggregate data decreased
as we went from a single-month figure to averages over two and more months.
Since monthly data themselves can be regarded as (suitably prorated) averages
of weekly data, clearly the opposite effect holds (that is, a sharp increase
in all forms of randomness or noise) as one goes from a monthly to a weekly
frequency of observation.
-
However, in terms of their serial correlation
patterns and in other ways, weekly data are more difficult to analyze.
In
particular there is no known model (such as the Cleveland X-11 model for
monthly data) to characterize the Board's weekly seasonally adjustment procedure or the extent of irregular/transitory variation.
Thus we will necessarily
be less precise and more ad hoc in our assessment of noise in weekly data,
though we can still make some reasonable approximations.
Since a month is an average of slightly over four weeks, it would
follow that any noise component in weekly data that is serially uncorrelated
from week to week would have a standard deviation approximately double that
in monthly data.
-
Consider first the transitory error.
Assuming that the
arguments for serial independence of transitory variations in monthly data
are also valid for weekly data, we would have a transitory standard deviation,
from Table 1, of
aE = 2(.21%) = .42%
for weekly M-lA or M-lB.
This figure is in line with results reported in
Porter et al. [6J based on a signal extraction me;hod.
Concerning weekly seasonal adjustment errors, we would expect some
negative within-month serial correlation in both preliminary-factor revisions
and final-factor errors, since weekly seasonal factors need to be consistent
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 20 -
with those for monthly data.
On the other hand, a given revision in a monthly
factor would be on average applied to factors for each week in that month,
inducing positive autocorrelation.
If these effects approximately offset
each other, our "doubling_" rule would again apply, giving
ar
= 2(.17%) =
.34%, a 8
= 2(.10%) =
.20%
respectively for seasonal revision and final-factor errors.
To combine these standard error estimates into an overall measure
of noise in weekly M-lA and M-lB it is necessary to take account of the
correlation between the preliminary seasonal, final seasonal, and transitory
disturbances.
~or monthly data, in was found in Section 3 that transitory
variations were negatively correlated with both revisions and final-factor
errors, the latter two being independent of each other.
Since revisions are
adjustments.!:£_ preliminary data and are therefore the negative of the revision
errors in that data, those errors are positively correlated with transitory
variationc;.
In particular in Section 3 it was seen that the correlation
between these two disturbances exactly offsets that between final-factor
errors and transitory variations,_as both are related in the same way to
the "central" moving average weights in X-ll.
Assuming that such an offset
also occurs with weekly data, the standard deviation of the total noise in
weekly M-lA and M-lB is
For changes in weekly data, assuming serial independence (see below),
the standard deviation of the noise is
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 21 -
and similarly the standard deviations of preliminary seasonal, final seasonal
and transitory noises for weekly changes are respectively 0.48%, 0.28%, and 0.59%
These results are displayed in Table 3, together with their dollar effects
C°\~ .) ("I
i
-
<-., f[1-,, ,.
,.
assuming a level of M-lA 6r M-lB of $400 billion •
. ~r-7~;
f
It is also possible to separate the measurement of preliminary
seasonal factor error, which will be corrected when the series is revised,
and fpe;;~~~J-1f,~mene~of-tl04.se, which remains in the final factors.
--b ..
,J_,.;:
io
To do this
-!.11.S~J"'f!.lq
it is -n~cessary-to-·aceount-•for the correlation between the revisions and
transitory variations.
From equation 3.1 it can be shown that for monthly
data the correlation between these was -0.24 (+o.24 for the xevision errors
present iil~nitial data).
Assuming this result also holds for weekly M-lA,
M-lB, then ~t can be
that the,standard deviation for noise in final
i.. I!!
t..
Jl,~Own
.,.
i..
.I!'
(revised) data is 0. 38% ($1.5 billion) for levels and O. 54% ($2 .2 billion)
for changes.
These are substantially less than the standard deviations that
would ha✓e· '-c:>btained (0:47% and O. 66%) without the negative serial correlation,
a result that confirms the widely held view that not only does revision of
seasonal factors remove the preliminary-data seasonal adjustment error~, but
such revision algo partially'smooths the transitory variations.
As noted earlier in this section, there is more uncertainty surrounding these estimates of noise than in the case of monthly data, largely because
of the possible effects of serial and contemporaneous correlation among the
noise components.
As a simple example of the effects of serial correlation,
and hence of the sensitivity of this paper's results, suppose the noise in
weekly data, instead of being serially independent as assumed, had correlation
coefficient p = 0.3 between adjacent weeks within a month.
[Possible argu-
ments for positive serial correlation between weeks were given above for
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
TABLE 3.
Estimated standard deviati~ns of noise
in weekly aggregates (M-lA and ,M-lB)
'
"
Source of
noise
Levels
,Billions 1 /
of dollarsPelicent
.Changes
'
Percent
l3illions l /
of dollars-
1. Transitory
.42
l.17
.59
2.4
2. Seasonal
i;-evisions
.34
;). .4
.48
1.9
3. Error in f:lnal
seasonal factors
.20
.8
.48
1.1
4, Total
.58
2.3
. 82
3.3
5. NOise in
final data 2/
.38
1.5
, .54
2.2
variations
1/
Assuming aggregate level of $400 billion.
2/ ,Lin~ 1 combined with line 3.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
c'
r
I
- 22 -
were given above for seasonal adiustment errors and can al&o be given for
transitory variations.}
Then (1) instead of doubling, the noise standard
deviation would increase by a factor of 1.8 in going from levels of monthly
to levels of weekly data, and (2) the growth rate standard deviation would
increase by a factor of ✓2(1 - p) = 1.2 rather than ✓2 = 1.4.
The combined
effect of these is that the noise standard deviation for changes in weekly
data would be
1.8
2
X
1.2
1.4
x
$3.3 billion=
$2. 5 billion
rather than $3.3 billion as previously calculated.
Conversely, the presence of negative serial correlation in noise
between weeks within a month (for example as a result of consistency constraints
between weekly and monthly data) would increase the weekly noise stan<lard
deviation relative to the monthly noise standard deviation.
In general, these results are further evidence that very little can
be inferred from any but the most atypical movements in weekly data.
However,
the converse should also be noted -- a significant move in the "true" data,
that is, a pronounced change in underlying trend or level, will likely go
undetected in one or a few weeks' time.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 23 -
REFERENCES
1.
8ach, G.L., Phillip D. Cagan, Milton Friedman, Clifford G. Hildreth,
Franco Modigliani, and Arthur Okun. Improving the Monetary
Aggregates: Report of the Advisory Committee on Monetary Statistics,
Washington, D.C.: Board of Governors of the Federal Reserve System,
1976.
2.
Cleveland, William P. Analysis and Forecasting of Seasonal Time Series.
Ph.D. dissertation, University of Wisconsin, Department of Statistics,
1972.
3.
Pierce, David A. "Data Revisions in Moving Average Seasonal AdJustment
Procedures," Journal of Econometrics, vol. 14 (Special Federal
Reserve Issue, 1980), pp. 95-114.
4.
_____
5.
_ _ _ _ _, Darrel
6.
Porter, Richard D., Agustin Maravall, Darrel w. Parke, and David A. Pierce.
"Transitory Variations in the Monetary Aggregates," in Improving
the Monetary Aggregates: Staff Papers. Washington, D.C.: Soard
of Governors of the Federal Reserve System, 1978.
7.
Shiskin, Julius, Allan H. Young, and John c. Musgrave. "The X-11 Variant
of the Census Method-II Seasonal Adjustment Program." Technical
Paper 15. U.S. Bureau of the Census, February 1967.
8.
Tiao, George c., and Steven r,. Hillmer. "Some Consideration of
Decomposition of a Time Series," Biometrika, vol. 65 (December
1978), pp. 497-502.
"Sources of Error in Economic Time Series." Washington,
D.C.: Board of Governors of the Federal Reserve, Special Studies
Section, 1980.
w. Parke, William 'P. Cleveland, and Agustin Maravall.
"Uncertainty in the Monetary Aggregates: Sources, Measurement and
Policy Effects," Journal of Finance, forthcoming (Papers and Proceedings of Denver Meeting, September 1980).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
The New Operating Procedures and Economic
Activity since October 1979.
Paper Written for a
Federal Reserve
Staff Review of Monetary
Control Procedures
by
Lawrence Slifman and Edward McKelvey
CONTENTS
Section
I.
II.
III.
IV.
V.
VI.
VII.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
SUMMARY OF PRINCIPAL FINDINGS
1
PRELIMINARY CONSIDERATIONS
The new procedures
The monetary control problem
The transmission mechanism
Longer-run effects
The economic enyironment
10
10
RESIDENTIAL STRUCTURES
A review of the experience since October 1979
Analytical considerations
14
14
17
THE CONSUMER SECTOR
The experience since October 1979
Summary
23
6
6
7
8
22
26
BUSINESS FIXED INVESTMENT
Theoretical considerations
The experience since October 1979
Summary
27
27
30
INVENTORY INVESTMENT
The determinants of inventory investment
Interest rate-variability and inventory investment
The experience since October 1979
34
34
INFLATION EXPECTATIONS
42
APPENDIX I: THE EFFECTS OF THE 1979 OIL PRICE SHOCK
Analytical considerations
Review of energy studies
Summary
Simmulation results
APPENDIX II: INFLATION EXPECTATIONS
Survey data
Evidence from financial markets
Theoretical models
Concluding remarks
REFERENCES
32
35
36
AI-1
AI-1
AI-4
AI-7
AI-8
AII-1
AII-2
AII-5
AII-8
AII-9
R-1
THE NRW OPERATING PROCEDURES AND ECONOMIC
ACTIVITY SINCE OCTOBER, 1979~I.
SUMMARY OF PRINCIPAL FINDINGS
On October 6,, 19i9, the Federal Reserve announced a fun~amental
change in oper~ti9g,procedures designed to assure-greater control over growth
in the mone~a~y agg~egates as pa~t of a more compreh~nsive pa~kage of policy
changes. 1
In-the~mon½hs i~ediately preceding_this.action economic activity
had been surprisingly robµs~, and_ inflation-and--infl~Fionary expectations had
in~~nsifie~.
The shift:in pro~edures was intended to,provide more effective
restr~int on ex~ftssive aggregat~ de~~P~ pres~ures an~ ultimately to reduce
inflation and the inflationafyrp~ychology that had been developing.
This
pap~r revi~ws the course of_economi~ activity since October 1979 and assesses
In analyzing_the, infJu~nce,of the new proc~dures on economic activity
it is important to reme~er that the objectives of monetary policy were not
affected; the only alteration wa~ to the shorter-run procedures used to help
achieve. the_ longer-run objectiv~s.
Thus, while interest rates were allowed
to fluctua~e more widely,ov~r the short run in response to market ~orces, the
t~rget ranges for th~. gr~wth rates,~of_the monetary.aggregates were unchanged.
Hen~e, in asse~sing the events since October 1979, the fo~us primarily is on
the implications for real s~c~o~ activity arising from the increased frequency
and amplitude of interest rate movements th~t haye occurred over the past 16
months.
1/
*
The analysis is made more.difficult because of the large_ number of
The other actions were al percentage point increase in the discount rate
from 11 to 12 percent and the establishment of an 8 percent marginal
reserve requirement on managed liabilities.
This study was coordinated by Lawrence Slifman and Edward McKelvey, major
•
•
contributions
were made by Susan Burch, Carolk Corrado, James Freund,
James Glassman, David Gree~, Owen Irvine, Charles Steindel, and Che Tsao.
The manuscript drafts were typed by Karen Pashkevtch, Sharon ~herbert, a~d
Debbie Vorce. Research assistance was provided by Ron Se~e and ~artha
Waldheger.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
atypical events that have buffeted the economy since October 1979, inclu<lin~
an oil price shock, credit controls, intensif1ed 1nflation expectations and
an attendant commodity speculation, and heightened hud~et uncertainties.
One central purpose of the new operating procedures
1s
to prov1de
greater assurance that the target growth rates for the monetary aggregates will
be met.
Over time, success in achieving this obiective should lower expectations
about prices, and, in turn, facilitate ~he planning of long-term savinP, and
investment commitments by businesses and households.
Thus, while wider
cyclical movements in interest rates may affect the short-run (for exaMple,
quarterly) pattern of changes in GNP, over the longer run the new operating
procedures should have a favorable influence on the growth path of the economy.
The principal findings of this paper are:
(1)
Economic activity, as measured by real GNP, almost certainly
would have contracted in ·any event <luring 1980, as a result of fundamental
forces already at work before October 1979.
The near doubling of
011
prices
during 1979, which generated a sizable transfer of income 'to foreign oil
producers, combined with a decline in labor productivity, had led to slow
growth in real disposable income and a deteriorat1on in household balance
sheet positions.
'
At the same time, an acceleration in the overall inflation
rate--in part a result of the oil price shock--and a concomitant rise in
inflation expectations had generated additional imbalances and overextensions
by both consumers and businesses, which left the economy vulnerable to a
slowdown in activity.
(2)
It is difficult to assess the effects of the new operating
procedures per se, as compared to the constraining effects of the monetary
targets themselves and other forces operating on the economy, on the timing
and composition of output changes during 1980.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
To the extent interest rates
-3-
redcted more promptly to fluctuations in money and credit demands, the
contraction of activity in 1980 may have been hastened, however, the subsequent
rebound also developed more quickly.
The decline in output during the second
quarter wa& intensified by the imposition of credit controls in connection
with the administration's anti-inflation program announced March 14, 1980.
In particular, the reduced availability of credit and the unwillingness of
consumers to borrow appears to have been an important factor in the sharp
drop in household spending between March and June.
As the program was phased
out and credit conditions eased early in the summer, spending for housing and
consumer goods rebounded sharply.
(3)
In the housing sector, it is especially difficult to separate
the impact of the new procedures from the effects of other factors.
The
depth and speed of the decline in housing activity between October 1979 and
May 1980 probably was magnified by the unprecedented movement of mortgage rates
to historically high levels; and the subsequent sharp rebound in residential
construction reflected the swift decline in mortgage rates that occurred
during the spring and summer.
Some of the rate variability in mortgage markets
likely reflected the switch in procedures; however, developments in real
estate markets also were heavily influenced by other factors.
In particular,
credit cont~ols apparently had an adverse, though largely unintended, effect
on the availability of real estate financing at some institutions.
In addition,
various institutional and regulatory changes affecting the thrift industry
and tne mortgage market probably contributed s1gnif1cantly to the pattern of
housing activity throughout the year.
(4)
Surveys of consumer attitudes taken immediately after October 6,
1979, indicated a significant deterioration in attitudes towards the purchase
of debt-financed items such as cars and large household durable~.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
The heightened
-4-
pessimism primarily reflected fears that the new monetary package would be
associated with reduced credit availability, and the drop in consumer spending, which began early in 1980, probably was influenced somewhat by these
uncertainties.
But, the more dominant factors were sluggish growth of real
income, a relatively low saving rate, high debt burdens and financing costs,
and households' concern about the acceleration of inflation.
These fundamental
forces were exacerbated by the psychological reaction of consumers to the
credit control program.
With the dismantling of Lhe controls program in
early July, however, consumer markets began to recover.
(5)
Because of the lags in the capital spending process, the behavior
of business investment in 1980 was, to a sizable degree, dependent on commitments made before O~tober 1979.
Nevertheless, the slump
1a
real outlays that
did occur probably was influenced by the une~pectedly sharp decline in aggregate
demand during the first half of 1980--a decline that was exacerbated by the
unusually sharp cyclical rise in interest rates and by credit controls.
The
subsequent easing of financial conditions during the summer and the rebound
in aggregate demand helped to arrest the contraction of real capital spending
in the last half of the year.
(6)
The new procedures apparently affected the pattern of inventory
movements in 1980, but the impact is hard to disentangle from other influences.
The excess accumulation and subsequent liquidation of stocks that occurred
in 1980 reflected sharp swings in final sales that were associated with the
1 cyclical movements in income, unusually wide fluctuations in interest rates,
and the credit control program.
The steep rise in credit costs during the
first part of the year, coupled with the rapid stock buildup, also caused
serious financing problems for many firms--most notably auto de~lers--and
probably provided additional stimulus for liquidation.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-5-
(7)
On balance, the shift in procedures has not, as yet, had a
clear effect on inflation expectations, although there was some improvement
for about half a year following the imposition of credit controls.
The
evidence does suggest, however, that expectations have not worsened since
October ~79, despite the persistence of rapid price increases.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
II.
PRELIMINARY CONSIDERATIONS
The new procedures.
Economic activity in the third quarter of
1979 was surprisingly robust in the face of deteriorating fundamental forces
Coupled with rapidly rising prices, this unexpected resilience in aggregate
demand produced strong deman4s for money and credit.
1n this turbulent
environment, the FOMC announced on October 6, 1979, a fundamental change
in its operating procedures.
Since then less emphasis has been placed on
containing day-to-day fluctuations in the federal funds rate, and more
attention has been focused on controlling reserves.
At the time the new procedures were instituted it was recognized
that they would entail greater freedom for the funds rate to change over the
short run in response to market forces.
Thus, it was expected that interest
rates would exhibit greater short-run variability (on a day-to-day or week-toweek baiis) as well as more rapid and possibly larger adJustments to cyclical
variations in aggregate demand.
As discussed in the paper by Dana Johnson,
"Interest Rate Variability under the New Operating Procedures and the Initial
Response in Financial Markets," in this compendium, there is evidence that
interest rates indeed have become more variable since October 6 on both a
short-run and cycl:Lcal basis.
Of course, it is impossible to-know how events would have unfolded
, during 1980
in
the absence of the change in procedures and therefore difficult
to draw firm conclusions about the nature and extent of their effects on the
economy.
The task is made harder by the fact that the past year was in 'many
other ways quite different from most other ·,years in the postwar period.
The
economy was still absorbing ,the effects •of 'the 1979-80 oil price shock, rapid
inflation was distorting traditional patterns of behavior, and infl~tion
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-7-
expectations were quite high
Indeed, by the spring of 1980,the situation
seemed to require special measures, and on March 14 a governmentwide antiinflation program wag announced, which included selective credit controls.
This actton complicates further any attempt to separate the effects of the
new procedures from the influence of other events and circumstances.
In spite of these difficulties, this paper attempts to draw inferences
_ about the effects of the new procedures _from a r,eview ot, ~hat, did h~pRen . .,,., In
the sections that follow, developments in seve~al key credit-related sectors
are discussed.
In the cours~ of this study, three fundamental issu~s are
raised:
-.
(1) Did the increased cyclical amplitude and frequency of interest
rate movements lead to a shorter and sharper recession in the
first part of 1980 and hasten the subsequent recovery?
(2) Did the increased short-run variability of interest rates, by
increasing ~ncertainty, permanently redu~e the level of aggregate
demand and hence output?
(3) How have the new procedures affected expectations?
In order to address these issues, it is first necessary to outline the_
analytical framework that describes the relationship between monetary control
procedures and economic activity.
The monetary control problem.
In discussing the effects of adopt-
run ob1ectives of monetary policy and the short-run method by which the FOMC
attempts to meet th~se objectives.
Over the past decade, the Connnittee has
followed a dual strategy in planning and executing monetary policy.
In the
first stag~, the Committee has,established longer-run (one-year) target
r~nges for,the monetary aggregates that are thought to be consistent with
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-9-
The evidence reviewed below indicates that the cyclical contractions in
housing and in consumer expenditures for durable goods experienced during the
spring of 1980 were unusually sharp.
Perhaps surpri~ingly, the resumption of
activity in these sectors following the rapid declines in interest rates that
occurred during the late spring and early summer also appears to have been more
prompt than might have been expected under such unprecedented circumstances.
But, as is suggested below, these altered cyclical patterns more likely
reflected, at least to a significant degree, the effects of the credit control
program that was operating between March and July.
The second credit-related mechanism relies on a presumption that
average yields on longer-term securities would increase as a result of the
higher short-run volatility in interest rates that appears to have been
associated with the October 6 action.
is not clear cut, however.
The evidence supporting this presumption
On the one hand, the results_ reported by Johnson
in "Interest Rate Variability under the New ... Procedures" in this compendium
suggest that the liquidity premium on Treasury coupon issues was little
changed after October 6.
On the other hand, the Board's quarterly econometric
model has shown a statistically significant link between the variability of
short-term rates and the average level of yields on long-term corporate bonds. 1
Economic theory suggests at least two reasons why an increase in rate volatility
may boost the level of long-term yields.
One possibility is that the removal
of policy constraints on interest rate movements has had asymmetric effects
that are biased upward, this would have raised average rates across the
maturity spectrum during 1980.
Alternatively, it is conceivable that efforts
by borrowers and lenders to minimize risks by lengthening liabilities and
1/
The relevant equation in the current version of the Board's model was
estimated over a period that does not include the experience since
October 1979.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
ECONOMIC INDICATORS
Real GNP
Billions of 1972 dollars
1600
(Quarterly)
1500
1400
1300
1200
1-----.J.:....----L----.J-----'-----'------L-----.L.-----11100·
Gross Business Product Fixed-Weighted Price Tndex
Percent change, Q4 to Q4
rs
12
9
6
3
.
3-Month Treasury Bill Rate
Percent
20
(Monthly)
16
12
8
4
0
1973
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1974
1975
1976
1977
1978
1979
1980
-10-
shortening dssets could have increased liquidity premiums and thus made the
y1eln curve steeper.
Under either circumstance, a permanently lower level
of output would result as activity in the interest-sensitive sectors was
discour~ged, however, 1t must be emphasized that the evidence on this mechanism
is far from conclusive.
Longer-run effects.
One principal purpose of the change in procedures
was to reduce inflation expectations by providing greater assurance that the
System would meet its basic money growth obiectives.
The reduction in the
level and in the variability of inflation that might be achieved through a
smoother growth path for the aggregates could ultimately exert a moderating
influence on expectations.
Moreover, once the new procedures show signs of
success, consumers and business firms might change their basic outlook about
prices.
Lower and more stable expectations about prices would facilitate the
planning of long-term spending commitments and thereby spur capital formation.
The evidence on price expectations so far is inconclusive, as
discussed further in section VII and appendix II; however, 1t may yet be too
early to rule out a significant effect.
At best, one can say that expectations
did not deteriorate further immediately following the October 6 announcement,
even though prices continued to accelerate.
There was an upward movement in
expectations around the beginning of the year, but 1n the spring they abated
somewhat--posstbly in response to the credit controls as well as the sharp
drop 1n output-and they generally held at this lower level until near the
end of the year.
The economic environment.
The unexpected resilience of aggregate
demand and the rapid rate of 1nflat1on that prompted the October 6 actions
continued throughout the fourth quarter of 1979.
Real final sales rose at
about a 3 percent annual rate, boosted by a substantial advance in consumption.
At the same time, consumer prices were rising at a 13 percent annual rate.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-11-
Despite the strength of aggregate demand, the economy at the end of
1979 appeared on the brink of a contraction in activity;
In addition to the
Federal Reserve's efforts to restrain the growth of the monetary aggregates,
the nearly 80 percent increase in the price of imported crude
011
that occurred
during 1979 transferred some $30 billion of income to foreign oil producers
and added perhaps as much as 2-1/4 percentage points to the overall inflation
rate.l
As a result of the sharp change in the relative cost of this price-
inelastic good, consumers increased their total nominal spending in order to
maintain lifestyles.
The higher consumer outlays were financed out of reduced
saving and increased borrowing, driving the personal saving rate to a relatively
low level and keeping the debt-income ratio near its record high of the third
quarter.
Inflation-induced credit demands by businesses and a rising federal
deficit also pushed interest rates up further.
Reflecting these pressures, economic activity began to turn down
early in 1980.
Initially the contraction was concentrated in household
sector demand, with residential construction, autos and other durable goods
most severely affected.
Despite the emerging weakness, inflation and inflation
expectations continued to intensify--fanning the flames of speculation in
many commodity markets and pushing interest rates even higher.
growth in money and credit surged in February.
Nonetheless,
Thus, on March 14 the President
invoked the Credit Control Act of 1969, and under the provisions of this
legislation the Federal Reserve announced a program of credit controls.
These measures hastened the reductions in credit availability that
were already in train at many lenders.
In addition, some lenders reportedly
imposed tighter nonprice credit terms, including stricter approval standards,
lower maximum borrowing limits, and higher minimum monthly payment requirements.
1/ Estimates of the effects of recent oil market developments on output and
prices differ substantially.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
These matters are analyzed in appendix I.
-12-
The announcement of the pr~gram appa!ently induced consumers to curtail their
., t ..
~
'
. . ..
~
..
Retail stores in particular reported a steep decline
us~ of cr~dit as well.
,,in crsedit ,use,,and
...
I
-
_,..
.,-
-,.
Judden drop in application~ for new accounts.
Banks also
nQticed sharp ;educ~ions ,in credit ca~d use.
These adjustmepts in the supply and demand for credit reinforced
'
'
f~
fundamental factqrs--~~~h ~s t~e rise
o~l ~rices, slugg~s? growth of real
~income, illiquid b~lanc: sheets, and an accelerating price level--to produce
the sharpest single-quarter_contraction in output recorded for the postwar
I
period,
, quarter.
,l,
..,
)
,I
Real- GNP fell nearly 10 _percent a~. an annual rat;_e in the second
Over the.first h~Jf of the ye~r, the index of industrial production
dropped a cumul~tive 7~.3 pe~cent {not at an annual rate), and employment
'
declined by 1-1/4 million.
J
The outp~t redu~tions were largest in the motor
vehicle and construction-related s~ctors, alth?ugh cr~dit controls were not
intende,d to restrLct.lending
~~
these areas.
Nevertheless, outside the auto
'
~
and,hou~ing sectors industrial production dropped a total off 7 percent
between
-
,January and July.
l
~
}-
At the same time, firms responded to the high interest
J
-
-
j
-
rates prevailipg during the first part of 1980 py cutting their stocks and
r
J
'.
,.:-
'\
l _:
reducing their orders.
Credit cond1t1ons eased abruptly in the spring in response to the
developing slack in the economy.
Consequently, the rise in most interest
rates came to a halt in, late March ~nd early April, and yields began to drop
at--.,,a recor~ pac~.
.
Most . private~short-term rates
fell 7 to 9 percentage points
'
in less than four ~onths,, to thejr lowest levels since the spring of 1978.
,
~
As lQan ,de)lland fell in mos; sectors~ the credit restraint guidelines were
phased out beginning in the late spring, and the program was completely
di~ma~tled ~n ea~ly July .
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
•
'
-13-
With the easing of credit market conditions, economic activity began
to revive in the mid-summer.
Those sectors hardest hit during the first half
of the year--autos and housing--led the rebound.
Consumer outlays for household
durable goods such as furniture and appliances, which often are credit financed,
also improved during the summer.
In addition, the rapid drop in rates relieved
the inventory financing pressures that had constrained many businesses earlier
1n the year.
Interest rates began climbing again later in the summer, and by the
end of 1980 most rates were at or above their previous peaks.
However, the
immediate response to this run-up in rates was less dramattc than had occurred
earlier in the year.
Housing starts and retail sales continued to rise
through the autumn, and economic activity as a whole maintained considerable
momentum through the end of the year.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-14-
III.
RESIDENTIAL STRUCTURES
A dramatic contraction in the housing industry took place in 1980.
For the year as a whole, real residential investment expenditures were nearly
20 percent below their 1979 level.
Private housing starts averaged 1.3 million
units in 1980--a rate of production lower than every year in the past decade
except 1975.
Housing activity declined with unprecedented speed early in
the year, and the recovery that developed during summer and early fall was
unusually strong.
These rapid movements in construction mirrored developments
in mortgage markets, where costs of credit climbed to record levels and then
fell quickly to their pre-1980 levels.
By the end of 1980, housing activity
had rebounded substantially, even though credit conditions once again were
unusually tight.
A review of the experience since October 1979.
Real residential
construction began to slow dramatically in the autumn of 1979 at about the
time the new operating procedures were announced.
Activity had been unexpectedly
strong in preceding months, with total housing starts hovering in the 1.7 to
1.9 million unit range throughout the spring and summer.
The decline in new
residential construction that ensued was both deep and rapid; by February
starts had dropped to a 1.3 million unit rate.
Adjustments to the higher financing costs that followed in the
wake of the October 6th action were particularly visible in the single-family
sector.
Mortgage rates rose quickly to an unprecedented 13 percent level and
remained there through the end of 1979; between September and December total
home sales dropped 15 percent, and the rate of increase in home prices slowed
markedly for both new and existing units.
Indeed, the average sales price
of new homes actually fell during the fourth quarter of the year, reflecting
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-15-
both a slowing in price increases for units of a given quality and downgrading
of units purchased.
In the same period, prices of existing units rose by
only half the rate of the preceding year.
Despite the softening in real
estate markets, builders managed to trim stocks of unsold houses, thus limiting
increases in the inventory/sales ratio.
In the spring of 1980 financial conditions became more disorderly-with mortgage rates climbing rapidly to above the 16 percent level--and production adjustments in the housing sector became even more acute.
Between
March and May builders sliced another 400,000 units from the pace of new
activity, bringing housing starts near their postwar low.
forces also were sharp during the spring.
Cutbacks in work
By April new home sales had hit
a nadir of 345,000, and while reduced production kept the stock of unsold
units declining, the inventory/sales ratio reached a record high of 12.6
months' supply.
Taken together, the retrenchment in housing activity in
late 1979 and early 1980 was unusually rapid, with starts declining by
900,000 units at an annual rate from 1.84 million in September to 940,000
in May.
In contrast, it took nearly a year and a half for starts to fall by
a similar amount during the downturn in 1973 and 1974.
The role of credit controls in the 1979-80 housing contraction is
somewhat problematical.
The program applied mainly to commercial banks and
placed no specific limits on mortgage credit.
Indeed, banks were encouraged
to treat such lending normally in light of general market conditions, and
they were specifically urged to maintain the availability of funds to small
businesses, farmers, and homebuyers.
Nevertheless, mortgage rates rose
sharply in the first few weeks after the program was announced, and the flow
of funds in this market slowed to a trickle.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-16-
Initially, those banks specializing in real estate lending mav have
curtailed these activities in order to satisfy the Boar~•s overall guideline
of 6 to 9 percent growth in total loans.
In a subsequent statement, the
Board clarified its original intent to avert even such an indirect limitation
on extensions of mortgage credit.
Of course, it is impossible to estimate the
degree to which institutions--both banks and thrifts--may have cut back such
lending in spite of the provisions of the program, but the sharp decline that
did occur suggests that the effect probably was significant.
Mortgage rates declined rapidly in the late spring, in conjunction
with a widespread easing in credit conditions.
The substantial reductions in
financing costs prompted a resurgence in housing starts in the summer and
fall that was far more swift and robust than had been experienced in previous
postwar housing cycles.
Underlying demand for owner-occupied housing had
reportedly remained quite strong throughout the preceding contraction, and
thus fueled the surge in activity when financial constraints eased.
Moreover,
thrift institutions--through the use of new deposit instruments--had avoided
the liquidity squeeze characteristic of previous cycles, and were thus in a
better position to resume lending when the demand resurfaced.
Between April
and July new-home sales rebounded almost 90 percent, and the average price
of these units reaccelerated.
Builders apparently were quite aggressive in
restarting production and reassembling work crews, possibly anticipating that
the upturn in activity would be prolonged.
The renewed strength in housing production continued well into
autumn, despite the resumption of increases in mortgage interest rates that
began in early August.
Part of this strength may have reflected transactions
postponed from the second quarter; some buyers also may have purchased homes
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-17-
in anticipation of further rate increases.
In addition, builders seemed
reluctant to disband recently reassembled work forces.
At the end of 1980
construction activity continued to be surprisingly strong in spite of
worsening financial conditions, with total housing starts remaining stable
at about 1-1/2 million units.
Analytical considerations.
Residential construction is one of the
sectors of the economy most sensitive to changes in financial conditions;
thus, it is in this area that the change in operating procedures would have
been expected to have its largest impact.
Purchases of single-family homes
usually entail substantial mortgage financing and therefore depend heavily
on both the cost and the-availability of credit.
Multifamily structures--
whether owned as condominium units or built for rental use--also are highly
leveraged in most cases.
In addition, for all types of residential construc-
tion the profits realized by developers hinge on financing costs.
By no means can all of the ups and downs in residential construction
over the past five quarters be attributed to the new procedures.
Activity in
this sector appeared to be on the verge of a downturn when the new procedures
were announced in October 1979, and it seems highly likely that a substantial
contraction would have taken place in 1980 in any event.
Moreover, as already
indicated, the credit restraint program appears to have had a significant,
though largely unintended effect.
Finally, after the previous housing cycle
in 1974-75, there had been other important changes in the financial landscape-most notably in the ability of thrift institutions to attract funds in periods
of high interest rates.
The enhanced competitive position of thrift institutions and other
regulatory changes in the mortgage market created a new regime in which
price rationing rather than the availability of credit determined the level
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-18-
of housing activity as interest rates rose to cyclical highs.
The most
important element of this new regime was the ability of thrift institutions
to issue deposit liabilities with yields tied to open-market rates of interest.
The new instruments--money market certificates and later the small saver
certificates--insulated these institutions somewhat from the sharp drop-offs
in deposit flows that had been characteristic of previous cycles.
In addition,
advances from the Federal Home Loan Bank System were more readily available
than in earlier periods of credit stringency, thus augmenting further the
resources available to the thrift institutions.
Under these circumstances,
thrifts were able to maintain a steadier flow of funds to the mortgage market
as interest rates rose.
Moreover, even when rates eventually peaked and
deposit flows dropped o£f, the liquid asset positions of these institutions
remained relatively comfortable, and they were therefore better prepared to
reenter the mortgage market when rates began to plummet.
Finally, state
usury ceilings also were rendered ineffective during most of the 1979-80
housing cycle by a federal statute that took effect in January 1980, and
borrowers also enjoyed unprecedented access to funds through secondary
markets, which functioned well throughout the period.
It is especially difficult to disentangle the impact of the new
procedures'from the effects of improvements in the ability of thrifts to
compete for funds, since the latter may well have contributed to the amplitude of interest rate movements over the last year or two.
In previous
cycles, cutbacks in credit flows to the housing sector, induced by increases
in market rates above deposit rate ceilings, had been a key element by which
monetary restraint had limited real activity.
The ability of thrifts after
May 1978 to offer deposit instruments paying market yields relaxed the constraints on the availability of funds and thus may have created a situation
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-19-
in which higher interest rates were required to achieve a given degree of
restraint.
In this regard it is noteworthy that mortgage flows cont-inued to
be relatively strong throughout 1978 and most of 1979, even-though yields on
substitutes for deposits were rising during most of that period to levels
that were then considered high.
Indeed, even when market rates peaked in the
spring of 1980, there were reports indicating that, at least in some areas,
credit was available "at a price."
Nevertheless, adoption of the new monetary control procedures-perhaps by contributing to a speedier response of interest rates to underlying changes in supply and demand--probably had some effects on real estate
transactions.
Because the financial arrangements associated with these
transactions generally require several months to consummate, the more rapid
changes in interest rates that occurred in 1980 increased the risk exposure
for those attempting to purchase new or existing homes.
During the spring
of 1980, for example, potential homebuyers found that mortgage rates were
increasing 50 to 100 basis points over periods as short as three or four
weeks, and that these higher financing costs either would disqualify them
from meeting lenders' standards for mortgages or simply impinge too heavily
on their own budgets.
Under these circumstances they were understandably
reluctant to incur the risk of contracting for financing at rates to be
determined at the time of settlement; yet for the same reasons lenders were
hesitant to make fixed-rate commitments for loans to be made several months
later.
In principle, the emergence of renegotiable and variable-rate
mortgages should have helped alleviate the situation, at least from the
lenders' perspective.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
However, these instruments still were very new, and
-20-
borrowers were understandably skeptical of their advantages.
Moreover, the
variable-rate mortgage was not designed with sufficient interest rate flexibility to neutralize the lenders' risks of rapid, short-run rate fluctuations.
Consequently, the standard fixed-rate mortgage remained the dominant instrument
in this market in 1980.
The risks associated with more rapid movements in interest rates
also affected builders, both directly and indirectly.
To the extent that
borrowers and lenders could not make contracts, builders were faced with
big~ carrying costs for unexpected inventories of unsold houses.
Also, the
frequency and magnitude of interest rate movements probably increased the
number of contracts that ultimately were broken.
These risks were largest
in the case of custom-built homes, for which the lags between commitment
and settlement dates typically are longest.
Even though builders typically
would receive cancellation penalties in instances where contracts were broken,
they still were faced with the problem of financing and reselling these
units during a period when prospects for finding buyers were dim.
Builders also were subject to greater risks arising out of their
own construction financing.
Speculative building apparently was curtailed,
as builders sought to avoid the costs of carrying unsold inventory.
Moreover,
those working on units with sales contracts found that their profits were
more uncertain, since the sales price would already be fixed while the construction loan often would stipulate a floating rate tied to the prevailing
prime rate.
Although this risk theoretically could have been symmetric, the
tendency for purchasers to sign contracts during periods of low interest
1
rates skewed the builders' risk to the upside.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-21-
In summary, several factors combined in 1980 to produce the sharpest
ho~sirig'cycle -of-th~-postwar period.
t
L
I
:
i
•
Adverse developments in the fundamental
.J1:.
determinants of housing--such as sluggish real income growth and the effects
-
I •
" •
-
-
r
''of ris1ng mortgage' rates on monthly financing costs--laid the groundwork for
some retrenchme~t in'~~nstruction activity even before changes in operating
procedures were implemented.
I
I
To the extent that the new procedures made
-r
,-:
interest rates more responsive to changes in underlying credit conditions,
they ma"y have contributed- to the speed of both the downturn in real estate
activity in the first half of the year and the subsequent rebound 1n the
third quarter.
'However, 'other changes of a more institutional nature, most
no'tably the great'er fiexib1hty of thrift institutions to attract funds, also
-
-
were important in this regard.
Finally, although the credit controls program
was not intended to restrict mortgage flows, the pattern of developments 1n
the second and third quarters strongly suggest that the controls did influence
events
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
10
a signif~cant way.
-22-
IV.
THE CONSUMER SECTOR
The October 1979 monetary policy pack~ge, including the new proce~
\ '"1
~
'1
'
dures, had an immediate impact on consumer attitudes toward
purchasing durable
r_:;
{
J
;I
goods and on opinions about the economic outlook~
l
i
I
C
}
,#
-
The University of Michigan
J
t.
;. l
-
J
)
-..,I
r
\
Survey Research Center (SRC) made a special tabulation
of responses
to their
'
I
r
J
i_
,,
usual monthly survey questions on the basis of whether
the
questions were
,1
~
answered before or after October 6.
1
In this tabulation theJ SRC found that
~
I
;_,
r
...
1-
I,.
I
initially households did not think the new package would increase, interest
J
-,
;
\
I
I
f
._
rates or--at least over the next 12 months--lower the inflation ra~e.1
For
over a year prior to the policy change, the Center had reported that consumers
'
\
thought that interest rates were at record levels; in October respondents
1
-
;
'
indicated no further increase in expected interest rates despite an increase
'
c
in the number of those who said that "credit was in short supply."
In contrast,
as shown in table 1, the rate of inflation expected for the succeeding 12
l f
t
~
i
\
f ...
months, which had dipped during the sunnner of 1979, resumed its upward movement
r
in late 1979--after the October 6 announcement.
_,.
t _..
f
(•
l
.)
A literal interpretation of
the SRC data--that is, higher reported expected inflation rates, but unchanged
expected interest rates--implies that respondents thought the real rate of
interest would decline.
Since economic theory suggests that lower real
interest rates provide an incentive to boost spending on long-lived goods,
the implied decrease in the real rate of interest should have been associated
with a desire for higher outlays on consumer durables and houses.
Yet the
SRC found a significant deterioration in attitudes towards the purchase of
It is somewhat puzzling that households did not correctly predict a
further rise in interest rates. This may have been because consumer
credit rates already were so high by historical terms--11.9 percent for
a new auto loan at a commercial bank and 20.4 percent for a personal
loan at a finance company--that most survey respondents thought further
increases impossible.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Table 1
SELECTED INDICATORS OF CONSUMER ATTITUDES
(Percentage of households reporting)
Question
Dec.
Jan.
Feb.
Mar.
Apr.
1980
May
June
July
Aug.
Sept.
20
14
7
9
22
24
16
9
5
5
8
Large household goods
5
11
9
13
17
12
11
19
28
31
24
19
16
13
Autos
7
7
8
18
18
13
14
17
24
29
22
11
11
10
Bad time to buy--credit
is tight, interest rates
are high:
31
n.a.
n.a.
68
n.a.
n.a
56
n.a.
76
75
66
53
41
so
Households expecting interest
rates to rise during next
12 months
61
70
70
62
40
45
47
71
56
26
21
21
39
53
Expected change in prices
next 12 months (mean percent)
10.4
9.7
1/
11.2
10 8
13.1
10.7
12.0
11.1
8 2
8.1
9.0
8 0
8.5
Perceived change in prices
over last 12 months (mean
percent)
n.a.
n.a.
n.a.
16.0
16.4
16.6
16.1
16.6
17 4
15.9
14.4
14.5
14 4
14.5
Houses
1/
Nov.
9
2.
5.
1979
Oct.
6
Credit is tight
4.
Sept.
5
1.
3.
Aug.
Responses to the October survey were tabu1ated separately, depending on whether the questions were answered before
October 6 or October 6 and later
the earlier figure was 8.6 percent, the later responses were 9.5 percent. The
average for the month was 8.9 percent.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
ALLOCATION OF DISPOSABLE PERSONAL INCOME
Consumer Durables
s a share of Disposable Income
(Quarterly)
Percent
\.
Food and Energy Outlays
, As a share of Disposable Tncorne
(Quarterly)
!Percent
Personal Saving'Rate
·(Quarterly)
~Percent
T960
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
196.5
;1980
-23-
cars~arid~i~rge.household' durables after the announcement, and opinions about
f ).,.,,..
(.
~
\
I
f"1 --,
},
r
J
'1
r-
-
whether it was·a good'time to buy a'house became more pessimistic.
~ ,-_ -,r,Th~ 'cbnsumer reaction 'evident in the' SRC data apparently reflected
{ \
._ i°
-
-
) .-.
F
a perception by householas that tliere would be 1ess credit available over the
coming ~ohths than· tliey had'anticipated and hence that they had little choice
but to' scale' h1ck .plans'for· durable goods outlays and rebuild liquid asset
) ,
\ l
positions.
:;;
-
... ,
".?-
r..,
~
-r
~
'
l,
- )
r
l.i.....
SRC' inaication of
f
r
,,.._ t ,
\~ I
t
•
•
•
Moreover; the·rise in'inflation expectations, coupled with the
I,_
.....,.
.,.~
a greater
r •
I
- t
() I
,-
-
-
.,
des'ire to cut back on dl'.scretionary durable goods
I
'
\
•
•
purchases, suggests that households--for precautionary reasons--wished to
accumul.":at1 -'fi~anbia( r~th'e'r than tangible assets.
Consistent with these
survey··re-suli:s:;' househo'tds· began·' to shift their portfolios towards shorter-term
as";ets 1iii 'd{E!' '·auiumn ro'f 1979 Tn order to reduce the capital risk associated
with increa;ed 'interest: rate.variability.
The experience since October 1979.
were"--in
a we'11-1fst'abli's;hed'
downtrend.
At the time of the change in·
This reflected households I concern
a~out·sluggisli 'real'income growth over the past year, ·the acceleration of
{nflation, •~nd~h1gh a~d 'rising debt' burdens.
Thus, faced with unfavorable
credit market and incdme' 'developments, consumers had seriously, begun to cut
back·o~)Jiscretionlry outlays'prior to October 6.1
)
f,
.._..
-....
,.
.-
t)
'-<
f
~
_
.,
'tr.
.-
I I
J.
e
In particular, despite
j
vigorous promotional 'efforts unit auto sales had declined to a sales rate of
; only ~iti:•s··mifllon 'un1ts"in tlie ·third quarter of 1979 from a 12.1 million
unit'peJi in thetsec6nd'quartet of'l978.
ll
1
Contemporaneous economic'analysis during 1979 and early 1980, however,
was somewhat obscured by the severe 1978-79 winter and gasoline shortages
of'the- spring; which 'artificially shifted substantial consumer demand to
the second half of 1979.
:..!',~
l.'
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
I.I.._
fl....,_~
( 1
-i(""
I
I~
\
1•
-24-
At the same time that consumers were r~d~_cing-t_J:!~ir d,iscretionat;Y-,
outlays in 1979, the share of family budgets <!,ev~t~~ .to .. it~ms ~ften ,-c?13s~~~,red
essential was rising dramatically.
Thi~ reallocation of, household income
...
'"
j
,,
t
...
-
~
-
...
toward essentials reflected the sharp jump in the re~~t!V~rprices~~f 1 f~~4
and energy--goods that have relatively low_ short-run ptJ<::e, elastic;i~Jes ., ,, .)
Thus, with real income growth sluggish and relativecprices shifting in an
l-'
..-
~~
.,f \
t_ \
f
1
I 1
-,_.
adverse direction, households in the second half-of 1979 chose to reduce
~
-
\
-
J
J
t"J
-
This behavior, however, could not be,sustained for.long,
and
in
.2
..!
~
-
.h
;
1
-1 -
_l
J,.
/
February 1980 consumer outlays began to slide as l;ious~holds) ~~te~p~~~ ,~o_, ,:
- build savings and reduce borrowing in the face of rap,i4ly, rising in~ere~,t _
rates.
The bulk of the subsequent drop in consumption was, fo~ ~iscretion~ry
J
._.
-
,,_
l
•
items such as autos, furniture, and appliances, whjch are,t?e most credit~~
sensitive consumer purchases.
The factors that acted to retard consumption early. in,~980_we:e
reinforced by the credit control program.
The major eleme~tc.of, tp~p~ogra)ll
as it affected consumers was to limit the growth, of- open-end- credi.t,
- .,. . , - _such, as
,.,
~
,._
~
credit-card debt, and those forms of closed-end credit that were either
r
'-
.,
-
.._
--
'
_ _ ..,
unsecured or secured by collateral not being purchase9 ~ith t~e PFO~~eds,
the credit.
o!
Extensions of automobile and mortgage... credit were
- exempted
- from
~
.;
specific limitations because consumer demand in these, sec;oi;~. alr_e,ady was .~eak.
In the first few weeks after controls, were annou~c~d, -~~y counne~Fial
banks and some retailers took steps to restrict-the availability
of consumer
~
(
f
- ---
J...,
credit, usually by adopting more stringent credie-approval _stan~a1j,4,s;.,--'Man~
.,
= {
-
f
banks instituted user fees on credit cards, lowe,;ed b;rro~ing li~its; ~r
,
>
,., ;
-
I
'
I
-.,
stopped issuing cards, perhaps taking advantage of the program'to.do what
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-25-
they had wanted to d~ anyway but feared might anger established customers.
Retailers most common,ly tightened credit terms through higher lending standards
and by raising minimum monthly payment requirements.
However, there were many
reports by retailers that consumers had cut back volunta!ily on credit card
use after the controls program was invoked, and that applications for new
accounts were off dramatically.
All forms of consumer installment credit
dropp~d at an 8 percent annual rate in April, the first full month under
cre~it controls, compared with increases of 5 percent in March and 7 percent
for the first quarter as a whole.
Foll~l<nng the imposition of credit controls, spending at re~ail
stores continued d~clining through the spring, and by May r~al outlays were
dow~ more than,9 percent from their January peak.
Despite the exemption of
closed-end auto loans from the credit control program, unit auto sales declined
in each successive month until May, when they reached a 5-year low of 7.3
million units.
The peak-to~tr9ugh decline in retail sales was the most
precipitous drop in consumer spending in the postwar period--about 25 percent
deeper than in the 1974-75 cycle.
Inflation expectations eased s~gnificantly in the spring--apparently
.J
the combined result of the sharp cycle in interest rates during the winter, and
spring, the anti-inflation program announced March 14, 1980, and the dramatic
cutbacks in consumer spending.
r~
:.
As indicated in line 4 of table 1, the expected
,
inflation rate declined from the double digit level of late 1979 and early 1980
to an 8 to 9 percent range beginning in May.
Although there was concurrently
some reduction in the actual inflation rate as measured by the consumer price
'
.
index,
SRC data' on the perceived,- rate of inflation did not,change_as much as
.
the expectational data.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
'
-26-
1
With the dismantling of credit cont;ols in early July, consumer
markets began to recover.
'
Unit auto sales picked up sharply in July and
remained in an 8-1/2 to 9 million ~nit r~te range through the autumn.
Spend-
ing for general merchandise, apparel, ' and furniture and appliances was up a
strong 4 percent in the third quarter, and consumer outlays generally were
well maintained through the end of the year.
Sunnnary.
Iheoretically; consumers adjust expenditures to expecta-
tions of longer-term earnings and to developments that affect returns from
>
accumulated savings or wealth.
'
The second fa~tor suggests a principal role
'
for interest rates as a determinant of consumption.
As a practical matter,
'however, it appears that apart from·investm~nt in housing, most households
in thE1 :past typically were concerned more with the avail~bility than the
'
'
~
cost of credit.l ~Increases in interest rates on credit-fina~ced purchases--
.
,
such as automobiles or major durable ,goods--used to play a secondary role,
since the movement in rates was relatively small and therefore -added little
to the contracted monthly.payment stream. 'While ' the SRC data suggest that
1
credit availability continued to be the primary concern of house holds .in
'
1980, reports from retailers indicated that higher financing costs gained
new importance.
Thus, in terms of th~ir effects on households; the ,new
operat:mg procedures fi.rs1t re,inforced fears '-that credit ~ight not be .available
to support additional expenditures or to meet emergency needs, and 'later
induced
J:/
a m~re' pronounced
respo~s~ to changes· in inter~st rates.
...
'
Higher interest rates can depress the value of household holdings ,of
corporate equities and credit-market debt; but these negat,ive ,wealth
effects on consumption are thought' to be si:nall--especiall~ ''in ,the short
run.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-27-
V.
BUSINESS FIXED INVESTMENT
The determinants of business capital spending can be divided into
those that determine the level of the capital stock, and those that regulate
the speed at which the actual amount of capital is adjusted to the desired
level.
Somewhat more problematically, the adjustment process may be divided
into two lags, the first being the lag between the recognition of deficiencies
in the capital stock and the decision to invest, and the second being the
lag between the decision to invest and the actual installation of new capital.'
Because of the lags in the capital spending proc~ss, it is unlikely that
wider ~r more frequent ~yclic~l swings in inte~est 'rates--such as those in
1980--would have a very significant effect on capital spending.
However, to
the extent the average level of (long-term) interest rates was higher in 1980
than it might have been under alternative operating procedures or policy
objectives, the long-run capital stock ultimately will be lower.
Theoretical considerations.
The neoclassical theory of investment,
developed by Dale Jorgenson and his associates,l starts by deriving the
stock demand for capital.
In this theory the optimal level of capital is an
increasing function of the expected, long-run level of real output ("accelerator"
effects) and a decreasing function of the real "service cost of capital."
Th~ service cost of capital is the value of the after-tax cash flow produced
by a unit of new capital over a period, and in equilibrium it will be
equal to the cost of raising the funds to hold the new capital for the
period.
1.
Although tax and portfolio considerations complicate the issue,
See, for example, Hall and Jorgenson (1967).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-28-
in general the service cosl of capital is an increasing function of market
interest rates and a decreasing function of inflation rates. 1
Within the framework of the neoclassical theory of investment,
unusually sharp cyclical fluctuations in interest rates could be expected to
affect capital spending directly through changes in the cost of capital and
indirectly through induced swings in output.
Neoclassical models of business
investment typically employ long distributed lags on the output and cost of
capital variables.
This reflects the belief that business spending decisions
are not based on the interest rate or sales demand prevailing at a particular
moment, since these data may contain a good deal of "noise."
Rather, 1t is
believed that firms use a longer planning horizon and base spending decisions
on expected "permanent" output and capital costs, which are represented
empirically as distributed lags.
Thus, given the lags in the capital-stock
adjustment process and firms' concerns about permanent rather than actual
output and capital costs, the effects ~f cyclical interest rate movements on
business fixed investment are likely to be small.
While the overall effect of sharp interest rate cycles on capital
spendin& should theoretically be small, there could be some effect on the
short-run timing and composition of investment.
During periods when interest
rates are cyclically high, especially if cash flow is deficient, firms might
postpone orders of items-with short lead times, defer purchases of goods
bought "off the shelf," or attempt tc;, stretch out delivery dates for previously
An important issue is whether the cost of capital depends on short- or
long-term interest rates. Most empirical work has been done with long-term
rates. Strictly spe,king the use of long rates is an implication of the
"putty-clay" hypothesi11; that is, that the ratio of capital to labor inherent
in -the existing capital stock cannot be modified to reflect the optimal
proportions called for by current interest and wage rates. This hypothesis
is controversial. For a discussion of the pros and cons of the putty-clay
hypothesis, see Hall (1977) and the comment by Modigliani (1977).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-29-
ordered items.
Although there is some flexibility for firms to engage in
these types of short-run timing adjustments, there is little evidence to
suggest an unusual amount of this behavior in 1980.
As was noted above, the demand for capital is believed to be a
decreasing function of long-term interest rates.
It has been argued by some
observers of financial markets that the average level of long-term rates in
1980 was higher th~n would have been the case under alternative operating
procedures or if the economy had not been subjected to such atypical events as
the imposition and removal of credit controls, an inflation-induced commodity
speculation, and unusually robust credit demands by households and businesses.
To the extent the average level of long-term rates was unusually high last
year, the desired stock of capital may have been lowered, and investment
might well be reduced until such time as the actual stock adjusts to the new
desired level.
To a certain extent firms could counteract the higher cost
of capital resulting from increased long-term rates by reducing th~ proportion
of investment financed by debt.
Issuing new shares of stocks probably would
not be a less costly means of raising capital, given the tendency of stock
and bond yields to move together, but increasing the share of investment
financed internally might prove advantageous.
In the long run, the new operating procedures could promote greater
cyclical stability in capital spending.
By allowing-interest rates to respond
more promptly to shifts in credit demands and supplies, the new procedures
might eventually contribute to a shortening and damping of business cycles.
If so, future cyclical changes 1n output would be less likely to be viewed as
"permanent," and firms would be more confident of a quicker return to normal
demand following a decline in sales.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Thus, to the extent that demand for
-30-
capital depends upon permanent rather than actual output, the new operating
procedures ultimately could reduce the procyclical variation in capital
spending.
The experience since October 1979.
The fundamental determinants of
capital spending generally were not supportive prior to October 1979.
Growth
in real GNP, although somewhat erratic on a quarterly basis, slowed from
nearly a 5-1/2 percent annual rate during 1978 to about a 2 percent rate over
the first three quarters of 1979.
Reflecting the sharp deceleration in final
demands, constant-dollar orders arid contracts for new fixed investment were
relatively flat during the first nine months of 1979.
On the whole, these
movements suggest that a slide in real investment spending probably would
have occurred during 1980 even if the operating procedures had not been
changed.
The investment environment became even less hospitable after October
1979.
Although a downturn in overall activity had long been anticipated,
final demands began to contract in early 1980 at a rate that surprised most
observers, and probably was not expected by most businesses.
Moreover, the
credit control program restricted business access to most sources of shortterm financing at a time when cash flow was dropping rapidly.
Under these
conditions, real business fixed investment (BFI) declined at a 20 percent
annual rate in the second quarter of 1980.
It is interesting to note that despite the rapid rise in nominal
interest rates that occurred in late 1979 and early 1980, the real long-term
interest rate apparently moved up little, on average, if at all.
During the
autumn and winter, the corporate bond rate rose about 4 percentage points
before peaking in April.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Based on evidence from the McGraw-Hill plant-and-
Billions of 1972
dollars, Ratio Scale
Real Business Fixed Investment
7
1so
!
1160
~140
~120
100
80
Billions of 19 72
dollars, Ratio Scale
Business Outlays for Motor Vehicles
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
60
40
30
20
10
5
1968
1970
1972
1974
1976
1978
1980
-31-
equipment spending survey, the expected inflation rate for business product
prices also rose 4 percentage points over the period, suggesting little change
in the real cost of capital.l
Similarly, the real cost of capital variables
in the Board's econometric model, which are based on equity prices, only rose
about 1/2 to 1-1/2 percentage points during the run-up in interest rates that
occurred around the turn of the year.
Similarly, an ~x post calculation of
real corporate bond rates showed an increase of only about 1-1/4 percentage
points between September 1~79 and March 1980.2
The evidence of little
movement in real rates suggests that the cyclical swing in nominal interest
rates probably had only limited effect on real BFI during 1980-Q2.
The bulk of the second-quarter decline in business capital outlays
was concentrated in reduced spending for trucks and autos, items whose
acquisition is easily postponed when demand or financial conditions deteriorate.
Indeed, a sharp decline in motor-vehicle outlays often occurs in the early
stages of a cyclical contraction.
Excluding these two items, real business
fixed investment fell at a 13-1/4 percent annual rate in the second quarter.
Business purchases of motor vehicles rebounded sharply in the third quarter,
but this was more than offset by widespread investment cutbacks elsewhere-especially for structures--and total BFI in constant dollars slipped another
1-1/2 percent (annual rate).
~owever, in the fourth quarter of 1980 real
capital spending edged up, with the increase concentrated in nonresidential
structures.
1/
2/
In the McGraw-Hill survey taken during late September and early October of
1979, firms expected to raise their product prices 8 percent over the next
year; in the March-April 1980 survey, they expected,, a 12 percent rise
in their product prices. The putty-clay hypothesis of investment indicates that product prices rather than capital goods prices should be
used to measure real interest rates.
The ex post calculations are based on the AAA corporate bond rate and
a measure of expected product price increases. The price expectations
variable is calculated as an exponentially declining weighted 3-year
moving average of the producer price index (PPI) ror finished goods
excluding food.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-32-
Summary.
Business fixed investment generally is considered to be
coincident with overall economic activity at cyclical peaks.
(One exception
is the 1973-75 recession, when the significant reductions in real BFI did not
begin until 1974-Q3.)
Thus, the timing of the decline in real BFI during
1980-Q2 was not out of line with the usual experience in previous business
cycles.
The intensity of the decline, however, was a bit different from
other cycles (table 2).
Often, as a capital spending downturn begins the
biggest loss (in percentage terms) occurs during the second quarter of the
contraction.
In contrast, during 1980 the biggest loss was in the first
quarter of the BFI cycle (1980-Q2), and it was considerably larger than most
first-quarter cyclical losses.
Averaging over the first two quarters
of previous BFI cyclical contractions, however, the magnitude of the 1980
- decline was well within-the range of previous cyclical experience.
The
composition of the reductions in real BFI during 1980 also resembled earlier
investment ~ycles.
In particular, the 58 percent (annual rate) drop in real
outlays for business trucks and cars in 1980-Q2 was similar to the reductions
evident in earlier cycles.
The behavior of real BFI during 1980--particularly in the construction area--was to a sizable degree dependent on commitments made before
October 1979.
Thus, the pattern of capital spending during 1980 was deter-
mined primarily by movements 1n final demand that already had been observed
or were expected prior to the announcement of the new operating procedures.
To the extent that the new procedures influenced business investment during
1980, the effects mainly occurred indirectly through changes in overall
activity (accelerator effects) that might have been induced by the new procedures.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
The data do suggest, however, that the unexpectedly sharp decline
Table 2
CYCLICAL CHANGES IN REAL BUSINESS FIXED INVESTMENTl
(Percent change, compound annual rate)
Beginning quarter
of BFI
contraction
Change during ...
First quarter
Second quarter
of contracti'on
of contraction
Average for
first two
quarters
1949-Ql
-18.2
-18.9
-18.7
1953-Q4
-2.6
-9.3
-5.9
1957-Q4
-9.2
-22.9
-16.5
1960-QJ
-8.8
0.4
-4.4
1969-Q4
-3.8
-6.7
-5.2
1974-Q3 2
-8.3
-17.1
-12.7
1980-Q2
-19.9
-1.5
-11.2
1/
2/
Cycles are based on the contraction in business fixed investment; these
cycles may vary in timing from the NBER-designate~ contractions in
overall activity.
There was a fractional decline in real BFI during 1974-Q2, but the
significant contraction did not begin until 1974-Q3.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-33-
in output during the winter and spring of 1980--which was exacerbated by the
unusually rapid cyclical rise in interest rates and by credit controls--may
have intensified the 1980-Q2 drop in real BFI.
But, the easing of financial
market conditions during the summer and the rebound in economic activity
probably ameliorated the contraction of real BFI in 1980-Q3.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-34-
VI.
INVENTORY INVESTMENT
The determinants of inventory investment.
Observed changes in
inventory levels, that is, inventory investment, can be decomposed conceptually
into planned and unplanned changes., Unplanned changes result from unanticipated
events such as production disruptions or differences between actual and forecasted sales.
Planned changes, on the other hand, result from firms adjusting
their inventories toward desired or target inventory levels.
Since rapid
adjustments in stocks to their desired levels are costly, these changes
generally are spread over several months or quarters.
Reflecting the fact that inventories serve as a buffer, the target
level of inventories depends on the expected level of future sales and production.
In addition, target inventory levels depend inversely on per unit
inventory carrying costs, which consist of good-specific carrying costs (such
as maintenance costs and depreciation) and financial carrying costs.
financial inventory-carrying cos~s consist of two components.
These
First there
is the opportunity cost (or direct cost if external financing is used) of
the funds invested in a unit of inventory.
The second component is the
reduction in carrying costs that comes about from increases in the price of
the good while it is held in inventory.
Hence, financial inventory-carrying
costs are properly measured by a real interest rate, that is, the nominal
interest rate used to measure the opportunity cost of the funds invested
minus the expected rate of price inflation (for specific goods) over the
inventory holding period.
Multiplying this real interest rate (assuming it
is expressed at an annual rate) by the good's price gives the number of
dollars per year it costs to hold a unit of the good in inventory.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-35-
Unfortunately, over any given period of time, changes in actual
inventory levels do not necessarily track very well changes in target inventory levels.
This is because:
(1) in the presence of adJustment costs, it
is optimal for firms to plan to spread the adjustment over several periods,
and (2) actual inventory levels can be heavily influenced by unplanned inventory investment.
Indeed, when unplanned inventory investment is large,
actual inventory levels and target inventory levels can even move in opposite
directions.
Estimates of the speed at which inventory levels are adjusted
to their target values range from three months to several years and are the
subject of much dispute in the economic literature.
The slowest speeds of
adiustment have been estimated for manufacturing inventories, the fastest
for retail trade.
Interest rate variability and inventory investment.
Traditionally,
retail firms have depended heavily on short-term bank loans to finance part
of their inventory.
For items like appliances and automobiles, specific
arrangements to finance "floor plans" are common.
Some automobile dealers
also obtain financing from their manufacturers' credit corporation.
Less is
known about the extent to which manufacturers finance their inventories
through external borrowing arrangements.
Greater variability in short-term interest rates increases the firm's
risk of encountering the cash-flow problems associated with refinancing at
high levels of ~nterest rates.
To minimize these risks, firms are likely to
seek alternative sources of financing for their inventories.
One possibility
is to tap longer-term sources, an option that generally will be more expensive
if pursued for any length of time.
In addition, firms may also depend more
on internal financing, which can make inventory investment more sensitive
to cash-flow fluctuations.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-36-
In either case, assuming firms attempt
to minimize costs,, any
,
shifts in sources of financing induced by greater variability in short-term
ratea presumably raises financial carrying costs and in turn leads to lower
target inventory levels.
In the trade sector, this probably will ~ause more
frequent shortages, less selection in the stores, and more special ordering
of goods.
Also, since inventory financing costs ~re an operating cost!
upward pressure will be placed on prices.
In the manufacturing sector,
smaller target inventory levels will likely lead to more freq~ent shortages,
more back orders, and perhaps larger fluctuations in output and employment
as firms attempt to gear production more closely to sales.
In summary, over the short run the effects of greater interest rate
variability are likely to be higher inventory carrying costs and smaller
target inventory l~vels.
The fewer inventories that are available to serve
as a buffer stock, the more other variables (such as output and employment)
will have to fluctuate in response to shocks hitting the economy.
On the
other hand, to the extent the new operating procedures improve the ability
of the Federal Reserve to pursue a policy that ultimately reduces both the
level and volatility of the aggregate inflation rate, variations in spending
and production could diminish over the longer run.
In this event, the economy
,
'
will need smaller buffer stocks and the short-run effects on production and
output might be mitigated or offset altogether.
The experience since October 1979.
Changes in inventory levels
during 1980 were heavily influenced by the unplanned inventory investment
that accompanied the business-cycle ~urning point in January 1980.
While
there were reports t?at businesses were cautious about ~uilding large stocks
toward the end of 1979, it 1s likely that few anticipated the steepness of
{
the decline in final sales that actually occurred in the first half of 1980.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-37-
Hence, much of the inventory buildup in early 1980 probably was due to errors
in forecasting sales.
In table 3 the inventory levels held by each maJor sector are listed;
changes over the past year are given in columns (4) and (5).
While there
was substantial liquidation by retail-sector firms, stocks accumulated in
the manufacturing sector.
This pattern is typical at the early stage of a
cyclical contraction in activity:
in response to declining sales, retail
firms cut orders and begin to liquidate excess inventories rather quickly.
In turn, manufacturers accumulate unplanned inventories until the rate of
production can be reduced to a level below their unexpectedly low sales
rate.
Often these excess manufacturing inventories are not liquidated until
retail firms have nearly completed selling their excess inventories and have
resumed ordering at a pace consistent with the level of final sales.
The observed changes in inventory levels reflected, in part, planned
responses toward altered inventory target levels.
The change in the maior
determinant of each sector's target level--its sales or shipments--over the
first year of the new procedures is reported in columns (8) and (9) of table 3.
In the retail sector, by far the largest liquidation occurred in the automotive
group, which experienced the largest sales decline.
.
The inventories of other
retail durable goods also were drawn down, presumably reflecting the sharp
drop-off in sales.
The absence of liquidation in nondurable retail inventories
~
is consistent with the fact that sales for firms in this sector declined only
slightly.
On the other hand, manufacturing sales fell 3.8 percent, and yet
there was a small buildup of stocks over the year.
This accumulation at
manufacturing firms reflected unplanned inventory accumulation over the year,
as well as the divergence of trends affecting various specific industries--in
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Table 3
INVENTORIES, ~ALES AND INFLATION EXPECTATIONS
Avg inventory
1960-79 as a
percent of
!,IA level
' Inventory
end of
Sept 1979
(1972
dollars)
Inventory
end of
Sept 1980
(1972
dollars)
Secto:-
(1)
(2)
(3)
(4)
(5)
'!.;,tail sector
19 03
65 3
62 6
-2 7
-4 13
Inventory investment
Change in sales
over the ):'.ear
Sector sales level
over the ):'.ear
1972
Percent of
Percent of
Sept 1979 Sept 1980
(1972
initial
dollars
initial level
dollars)
level
(8)
(6)
(7)
47 7
44 7
-3 0
(9)
Inflation exoectat1on measures
Actual inflation observed
Actual inflation rate
over previous year as of over next six months as ol
Sept
Sept
Sept
Apr
Aor
1979
1980
1980
1979
1980
(10)
(11)
(12)
(13)
-6 29
12 02
13 66
11 49
13 81
8 65
-14 43
6 73
7 09
12 15
8 25
5 14
(14)
Auto g:-oup
4 13
16 3
14 ,o
-2 3
-14 11
9 7
8 3
-1 4
1:0,..i.ura!J le
' 10 61
35 5
36 l
6
l 69
30 0
29 3
- 7
-2 33
13 36
16 16
12 70
15 75
7 39
4 29
13 5
12 5
-1 0
-7 41
8 0
7 1
- 9
-11 25
10 43
10 23
9 87
11 70
10 56
ve-c~a,t vholesalers
12 50
49 9
50 1
2
40
38 1
39 6
1 5
3 94
9 70
11 75
10 82
16 10
6 03
"'a"lJ'fact..1r1ng sector
43 00
142 5
141 4
9
63
76 2
73 3
-2 9
-3 81
11 15
14 04
14 61
15.76
9 21
74 53
257 7
256 1
-1 6
- 62
161 7
157 6
-4 1
-2 54
-:-a-m, se-ctor
15 20
43 0
43 4
4
93
Other sector
10 28
43
0
42 8
- 2
- 47
100 00
343 7
342 3
-1 4
- 41
goods
Ot~er durable goods
S.1?>total
'!a'l.ifactun'lg and trade
:'otal ",IA inventory
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-38-
particular the l~rge accumulation of stocks at norf;uto transportation
manufacturers (such as aircraft pl~nts and shipbuilding yards).
Without this
buildup, manufacturing inventories would have declined.
Presumably, in late 1979 and in 1980, target inventory levels also
were inftuenced by the large swings in nominal interest rates and inflation~
'
'
rates, as'well as by unexpected changes in sales.
Between September 1979
'
and April 1980 the prime rate rose from around 13 percent to 20 percent, and
the coDIIIlercial paper rate rose from about 12 percent to 16 percent.
This
large i~crease in nominal· rates does not appear to hav~ been accompanied by
a concomitant change in short-run inflation expectations of the same magnitude.
Lacking data on movements in the expected rate of change of each subsector's
prices, the right-hand colunms of table 3 report two proxy inflation forecasts:
(i) ~ naive forecast that assumes the inflation rate over the relevant inventory
holding period (taken here to'be six months) would have been equal to that
observed over the preceding 12-month interval, and (2) the perfect foresight
forecast that prices would increase at the rate that actually materialized
over the following six months.1
It seems reasonable to assume that the
actual inflation expectations held by most firms lay somewhere between these
two proxy forecasts.
The values of these proxy forecasts are reported in
table 3 for September 1979 and for April 1980 (the month that nominal interest
rates peaked).
The value of the naive forecast also is reported for September
1980.
It is important to observe two differences between the inflation expectations measures shown in table 3 and those reported on page 31: (1) the
measures in table 3 attempt to reflect the inflation rate expected over
the near t~rm (since they are to be compared with short-term interest
rates), while the evidence on page 31 pertains to longer-run expectations;
and (2) these measures are industry specific while those on page 31 are for
for all product prices. Thus, it is possible·for real short-term rates to
rise in a specific industry, while real long-term rates for all businesses
on average are little changed.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-39-
The expected inflation proxies for the retail auto subsector suggest
'
that the expected rate of inflation for automotive products probably declined
somewhat between September 1979 and April 1980.
Henc~, the increase of 4 to
6 percentage points in short-term nominal interest rates was associated with
a rise in the real interest rate faced by automobile dealers of somewhere
between 4 and 9 percentage points.
These sharply higher carrying costs
partly explain why automobile inventories declined substantially'more than·
-
'
did aQtomotive sales; econometric estimates suggest that around ~1 billion
of the $2.3 billion liquidation in automotive stocks can be attributed to
the rise in the real interest rate.
After April, both the decl~ne in nominal
rates and the pickup in the rate of auto price inflation worked to reduce
these financial carrying costs.
_The proxies reported in table 3 suggest that the expected inflation
rate for other retail durable goods did not increase between September 1979
and April 1980.
Hence, the rise in nominal i~terest rates was translated
into higher real inventory-carrying costs for these goods and thus was,a
factor behind the substant~al liquidation observed in this sector.
During
this period the naive forecast of nondurable-goods price inflation rose by
j
~
I
\
".
nearly 3 percentage points, whereas perfect foresight would have yielded,a
cpnsiderable drop in the e~pected nondurable-goods inflation_rate.
it is not clear which way actual inflation expectations changed.
Hence,'
In any
case, between September and April the increase in nominal rates probably
helps account for the small liquidation observed here in late 1979 and early
1980.
The rise in real interest costs between September and April probably
also helped stimulate the small liquidation of merchant wholesaler stocks
that occurred in late 1979 and early
1980.
I
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-40-
The rapid cyclical run-up in interest rates during the first part of
1980 probably also affected inventory investment indirectly by depressing sales
of automobiles, houses, furniture and appliances, and other interest-sensitive
items.
The height of the cyclical peak in nominal rates and the impact of
credit controls undoubtedly surprised most firms, causing_actual sales to
fall short of their sales forecasts, and thus led initially to at least some
unplanned inventory investment.
Also the effects of the higher rates and
credit controls on sales during the spring may have led to lower target
inventory levels.
Hence the indirect effects·on sales of·an-unusually sharp·
cyclical rise in interest rates, coupled with credit controls, probably
accentuated recent inventory movements; by causing a larger accumulation
initially and possibly a reduction in desired stocks, more liquidation was
needed_ to bring inventory levels back down to their target values.
In those sectors characterized by relatively fast speeds of adjustment (such as retail trade establishments), the unusually high cyclical peak
in real rates-during the spring may have caused a higher observed rate of
liquidation.
However, the prompt and precipitous decline in rates after the
peak probably moderated the amount of this extra liquidation.
For the more
slowly adjusting manufacturing sector the additional liquidation induced by
the extremely sharp cyclical run-up in rates was smaller; by the ~ime mos~
manufacturers were in a position to liquidate their inventories (the third
quarter of 1980) short-term interest rates were well below their April 1980
cyclical peaks.
Hence it is reasonable to conclude that most of the extra liquidation was confined mainly to the rapidly adjusting trade inventories.
Moreover,
to the extent that the new operating procedures provided a steadier availability
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-41-
of bank loans over the high interest rate period, this too moderated the
observed liquidation somewhat by decreasing the number of forced sales of
excess inventory.
Because most of the extra liquidation probably took place
in the trade sector, the overall inventory investment cycle that accompanied
the 1980 recession will appear t~ have occurred somewhat earlier in the
business cycle.
In summary, the inventory cycle that accompanied the 1980 recession
was relatively mild in comparison with previous business cycles.
This
primarily reflected the cautious inventory behavior practices followed by
firms after the 1973-75 recession.
However, the rapid rise in financing
costs and tlie unexpectedly sharp drop in sales in the first part of 1980
caused serious difficulties for many firms--most notably auto dealers.
The
ensuing liquidation of stocks was exacerbated by the effect of high interest
rates on final sales and the unexpectedly strong impact of credit controls on
consumer demand.
Despite the easing of credit markets during the summer and
the pick up in final sales in the last half of 1980, firms continued to
maintain tight control over inventories through the end of the year.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-42-
VII.
INFLATION EXPECTATIONS
One obJective of the shift to a reserve operating guide for open
market operations was to help reduce inflation expectations by providing
greater assurance that targeted growth rates for the monetary aggregates
would be realized.
In this section data from several surveys and inferential
information from the behavior of long-term interest rates since October 1979
are examined in an effort to assess the effects of the change in procedures
on expectations.
Appendix II provides a more detailed analysis of the data,
as well as a discussion of related theoretical and empirical problems.
On
balance, the evidence does not suggest that the October 6 action led to an
immediate improvement in inflation expectations; however, it may be too early
to rule out such an effect, since many survey respondents and market participants
presumably would have awaited some signs of success for the new procedures
before revising their expectations.
Measures of inflation expectations from three surveys are summarized
in table 4.
These data show no reduction in tbe expected rate of inflation
in the.;months immediately following the October 6 action; indeed, all three
series indicated some deterioration by the end of the year.
Data from the
University of Michigan Survey Research Center (SRC) show that consumers'
expectations of inflation over the ensuing 12-month period rose quite sharply
in November 1979 following a steady decline since the spring of that year. 1
Thereafter, the average expected rate of inflation peaked at 13 percent in
1'/
The SRC asksseveral questions every month that are designed to measure
consumers' expectations about inflation. The data on anticipations are
constructed from the following questions: "During the next 12 months do
you think that prices in general will go up, or go down, or stay where
they are now?" and "By about what percent do you expect prices to go up,
on average? during the next 12 months?" The questions refer to the CPI.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Table 4
MEASURES OF INFLATION EXPECTATIONS 1
(Percent)
Period
University of Michigan
(SRC) survey2
1977-Q4
6.5
1978-Ql
Q2
Q3
Q4
7.4
7.8
8.3
8.1
1979-Ql
Q2
Q3
9.1
11.1
10.3
1979-September
October
November
December
9.7
8.9
11.2
10.8
1980-January
February
March
April
May
June
13.1
10.7
12.0
11.1
8.2
8.1
July
August
September
October
November
December
~981-Janua.ry
9.0
8.0
8.5
9.2
8.7
10.1
7.8
Livingston
survey 3
Blue Chip
Economic
Indicators4
6.3
n.a.
6.7
n.a.
n.a.
n.a.
n.a.
7.1
8.5
n.a.
n.a.
8.3
8.5
8.5
8.8
9.6
~-9
10.1
8.2
8.6
8.9
9.1
9.0
8.9
8.9 •
8.7
8.9
9.1
8.9
10.3
9.5
n.a.
All expectations are for the CPI, except that the GNP deflator is
shown for the Blue Chip Economic Indicators.
2/ Mean increase of responses to the question: "By aqout what percent
do you expect prices to go up, on the average, during the next 12
months?" The question refers to the CPI.
3/ Expected increase constructed by the Federal Reserve Bank of
Philadelphia from disaggregated Livingston data; data are for the
last month 9f the quarter indicated.
4/ Consensus forecast; series begins in May 1979.
1/
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-43-
January 1980 and remained high through April.
From May to November expecta-
tions were more moderate, possibly as a result of the credit controls, with
the average expected rate of inflation generally fluctuating in "the 8 to 9
percent range--the same as in late 1978 and early 1979.
The other' two surveys generally confirm the pattern of expectations
.shown in the SRC data.
The Livingston biannual survey of "informed" business
.
economists indicated a clear increase in the inflation expectations of these
respondents between June and December of 1979, and again in the subsequent
half-year ~eriod.1
The Blue Chip Economic Indicators survey data failed
to show any immediate impact of the revised open market procedures on the
anticipations of-private economic forecasters.
In an analytical sense, the relevance of the survey data could be
'
discounted because there is no evidence that transactions actually are based
'
on these expectations.
The behavior of long-term interest rates, on the
other hand, conveys information about changes in inflation expectations that
1s directly the outcome of financial transactions.
In the economic literature,
the expectations hypothesis of the term structure of interest rates--a widely
accepted view of interest rate relationships--holds that long-term rates are
weighted averages of current and anticipated short-term rates, adjusted
.appropriately for liquidity and risk premiums.
Because expected future rates
presumably incorporate expected rates of inflation, movements and shifts in
the yield curve should embody some inf~rmation about changes in the expecta)
tions of market participants.
1/
since 1947 Joseph A. Livingston has collected biannually the anticipations
of economic variables from businessmen, economists, and professional forecasters. Livingston mails questionnaires in early December and May, and
asks for 6-, 12-, and (in May) 18-month ahead forecasts of the CPI and
the PPI for finished goods. Results of these surveys are published
regularly in Livingston's column in either the Philadelphia Inquirer or
the (Philadelphia) Bulletin.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-44-
Unfortunately, the post-October 1979 data are difficult to interpret;
'
at best one can say that they too provide no evidence of a significant Jmprovement in expectations following the change in procedures.
Analysis is complicated
by the possibility that several other factors can also cause long-term rates to
change.
These include changes in liquidity and risk premiums, as already
s_ug~ested, 4:s well as movement"'s i~ t~e real rat_e of interest .,an_d differ!;!nc~s
that may arise from the segmentation of securities markets from one another.
These difficulties can be minimized by confining the investigation
to yields on Treasury securities.
These yields, whose average level and
volatility increased dramatically after October 1979, have been studied
extensively in the paper by Johnson, "Interest Rate Variability Under the
New . . . Procedures, 11 in this compendium.
That study produced little, if any,
evidence to support the hypothesis that liquidity premiums have risen following
the change in procedures, a result which increases the possibility that
changes in inflation expectations were responsible for the behavior of
longer-term Treasury yields.
If so, the data would indicate that inflation
expectations rose sharply in the fall of 1979 and have remained both high
and highly volatile since then.
However, such a conclusion must be highly
tentative, given the difficulties in measuring liqurdity premiums and
uncertainties about the movement of the real rate of interest over this
period.
On balance, the available information on expectations does not
indicate any clear improvement in expectations following the Octob~r 6 action.
They generally suggest a worsening in expectations over the subsequent six
months or so, followed by an improvement later in the spring of 1980.
It
should be emphasized, however, that little is known about how households and
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
ONE-YEAR FORWARD RATES IMPLIED
IN THE YIELD CURVE FOR TREASURY SECURITIES!/
Percent
15
14
13
12
11
10
9
8
7
6
5
1976
1978
1980
!/ According to the expectations hypothesis, the series plotted represents the
yield on one-year Treasury obligations expected by market participants to
materialize one year ahead of the dates indicated (the "one•year forward
rate"). This rate is calculated using the one- and two-year constant
maturity Treasury yields, and is arrived at by the following formula:
wo 1/ [
r f, t = 100 [ ( 1 + r 2 t]
where
~
r 1 t) - 1 ,
1 + -fffff
rf,t is the one-year forward rate, in percent,
rl,t is the one-year constant maturity yield, in percent,
and r2 t is the two-year constant maturity yield, in percent.
,
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-45-
businesses form expectations, and particularly about how quickly they react to
changing events; moreover, measurements of inflation are subJect to potentially
large error.
Consequently, it may be too early to conclude that the change in
procedures had no effect in reducing expected rates of inflation.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
AI-1
Appendix I
THE EFFECTS OF THE 1979 OIL PRICE SHOCK
In October 1979, when the Federal Reserve System was implementing
~
its new operating procedure, two important events were affecting the domestic
and international energy markets.
First, the political situation in Iran
during early 1979 and strong worldwide demand were leading to sharp increases
in international oil prices.
Second, the initial phase of the domestic oil
decontrol program had Just begun and already was proceding at a rapid pace.
These events continued to put upward pressure on domestic energy prices and,
by worsening the general inflationary outlook, played a key role in shaping
the economic environment over the past two years.
This appendix addresses
the question of the impact of the 1979 energy price shock on both real activity
and the general level of prices.
Analytical considerations.
The transmission of energy price shocks
to overall economic activity can be viewed using a number of increasingly
broader and more realistic analytical frameworks.
Consider first the impact
of the price shock within the production sector of the economy.
As energy
prices go up, the initial effect is an increase in production costs, which
drives up output prices and leads to a first-round decline in final demand
and some unintended inventory accumulation.
This first-round market response
will then trigger secondary reductions in interindustry flows, higher wages
and higher prices of intermediate goods, and ultimately further declines in
intermediate and final demands.
The initial impact plus the subsequent
interindustry adJustments will lower the levels of real income and employment
and raise prices.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
AI-2
The above scenario captures the essential response of producers in
the very short run.
Over a longer period of time, other considerations
become important and tend to imply further depressing effects on real activity
and increasing pressure on prices.
1.
They are briefly discussed below.
Changes in the aggregate input structure:
As energy prices rise
and energy resources become more costly relative to other inputs, a producer's
optimal input structure will change.
take place.
Inevitably some input substitution will
Between energy and labor inputs the empirical evidence seems to
suggest that higher energy costs induce producers to substitute labor for
energy.
What happens between energy and capital is less clear.l
In the
short run, especially if there is some flexibility in capacity utilization,
the weight of evidence seems to suggest complementarity between energy and
capital services.
In this case, higher energy costs will lead to increased
use of labor and reduced capacity utilization in most industries.
In the
aggregate, this type of input ad3ustment tends to slow down capital formation
and labor productivity growth.
This line of reasoning is consistent with
the post-1973/74 experience of a high employment, low investment recovery
path in the U.S. economy.
In this context, energy price hikes further depress
real output.
2.
Monetary and financial repercussions:
Energy price increases will
also have important effects outside the production sector, as interest rates
respond and induce still further changes in the economy.
Regarding the
I
energy price effect transmitted through interest rates, the actual course of
1
the economy obviously depends on the res;ponse of the monetary authority.
I
I
1/
For an interesting discussion see B~rndt and Wood (1979).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
I
AI-3
Assuming a neutral, nonaccommodative policy stance, we can focus our attention on repercussions that are endogenous to the system.
At the 1n1t1al
stage, real liquidity deMand will tend to fall after an oil price shock
because of the reduced level of real activity.
Whether nominal transaction
demands will rise or fall depends on how rapidly the general price level
rises at this time.
It is generally impossible to tell beforehand whether
the falling real output or the rising price level will have the predominant
effect on m9ney demand.
It is quite possible that the output effect and the
price effect largely will offset each other.
If so, a non~ccommodative
policy would, at least at the initial stage, leave short-term interest rates
little changed.
On the long end of the market, however, results will depend
on market participants' inflationary expectations and risk assessments.
It
is almost inevitable that rapidly rising energy costs would lead to renewed
inflationary expectations, and would have the tendency to drive up long-term
bond yields and lower bond prices.
Such a development would tend to depress
domestic demand for consumer durables and capital goods and retard capital
formation.
This is yet another possible source of depressing forces on real
output.
3.
Balance of payments and trade effects·
Increases in 011 prices act
like an excise tax on the economy--reducing demand, raising unemployment, and
generating more inflation.
If the price increase originates from abroad, as
is the case with OPEC price hikes, the adverse effect on domestic real output
will be made more serious by the outflow of dollars, causing deterioration in
the nation's trade balance and payment positions.
Further, the adverse
impact of increased 011 payments overseas is not limited to current accounts
alone.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Excess oil payments to foreign producers cause concern abroad about
AI-4
I
the effects of US. energy policy on the dollar's future and tend to weaken
the dollar.
In such an environment, private capital outflows are likely tb
rise, leaving deficits in U.S. capital: accounts as well.
There are other compll cations in the international market that tend
to make foreign 011 price increases h~ve a more severe impact than comparable
I
domestic price increases.
I
For instance, OPEC price shocks affect all oil-
consuming nations at the same time, arr,d bring higher_ 'price levels and slower
I
growth to most U.S. trading partners.' The interdependence of major 011consuming nations on the world market makes the fi~al impact of OPEC price
I
shocks worse for the United States than would be the case if the United
II
States were the only country affected.:
Review of energy studies.
''I
In order to sort out these effects, we
i
have reviewed a number of energy studies that use simulation techniques to
examine the impact of the 1978-79 oil price shocks.
Results from four studies
I
are briefly summarized to provide a sense
of the magnitude of the impact.
I
I
The four are selected because each one deals with at least one of the aspects
discussed above.
The results are summarized in table AI-1.
I
Thurman and Berner (1979) used the MPS econometric model of the
U.S. economy in their simulations.
I
Their basic price senario started from
l
the June 1979 OPEC price schedule, and \involved raising the average price of
I
imported oil in the United States by 62 percent by the end of 1979, with an
additional increase of 9.5 percent in 11980.
I
Using this price assumption,
they found that the level of real outpu[t would have been reduced O.4
percent by th: end of 1979 and 0.9 perclnt at the end of 1980.
The inflation
I
rate for domestic consumption prices wohld have been only fractionally higher
in 1979, and 0.7 percentage points highlr in 1980.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Table AI-1
SUMMARY OF RESULTS AND PRICE ASSUMPTIONS
Price increase scenario
Author(s)
for
imported crude oill
Impact on the level
of real GNP
(percent deviation
from control)
Impact on rate
of 1nflat1.on
(percenta~e
points)
'End of
ThurmanBerner
RascheTatom
Tatom
MorkHall
1/
2/
Scenario I:
62 percent in 1979,
9.5 percent in 1980
Scenario II:
68 percent in 1979,
28 percent 1n 1980
Actual through 1979
(roughly 28 percent rise in
the relative price of energy
from 1978-Q4 to 1979-Q4
1980
1979
1980
-.4
-.9
.1
.7
-.5
-1.5
.1
1.3
-6.0
.7
2.4
1.8
1.3
-3.1
Actual through 1980-Q3
relative price of energy
assumed constant 1n 1980-Q4
72 percent increase 1n 1979
and additional 25 percent
in 1980
During
1979
-1.1
In Thurman and Berner's study, the "control" scenario assumes crude import prices
increase 26.5 percent 1n 1979 and 2.7 percent 1n 1980. In the other studies, the
price scenario described was compared to a hypothetical case in which real 011
prices remained roughly constant.
For Thurman-Berner and for Tatom this is the rate of change of the GNP deflator,
for Mork and Hall this 1s the rate of change 1n the domestic price level.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
AI-5
In November 1979, when their paper was written, Thurman and Berner
could not foresee the price changes that actually did occur.
scenario, however, they assumed that the price of imported
additional 26 percent.
In an alternative
011
increases an
Under this more inflationary assumption their results
suggest that real output would have been reduced by 0.5 percent at the end of
1979 and would have been 1.5 percent lower at the end of 1980, the inflation
rate would have been boosted 1.3 percentage points in 1980.
We should note
that even these more drastic price assumptions underestimated the nominal
I'
price changes that have actually occurred.
Eoth the basic scenario and the
higher-priced one suggest that increases in
011
prices of the magnitudes
assume<l by Thurman and Berner lowered the level of nominal GNP from what it
would have been in the absence of an energy price shock.
This is because
higher petroleum prices induce increases in the overall price level that are
more than fully offset by decreases 1n real output.
In part this seems to
be caused by an emphasis on r1s1ng import prices as the "driving" variable,
and the underestimate of the 1979 increases in domestic crude prices,
This
overestimates the income transfer abroad (see Thurman and Berner, page 21)
I
and underestimates the increase 1n domestic refiner costs and subsequent
I
petroleum-product price increases.
•
the impact of the increases 1n crude
Thus, the model tends to underestimate
,I
01'1
'
prices on the domestic price level.
I
Mork and Hall (1979) used a macroeconomic model in which energy,
1
labor, and capital demands are derived '.from an implied aggregate production
I
function.
Their model allows for energy substitution when 1 relat1ve input
I
prices change,
I
Their price scenario involved a crude-oil price increase
I
from $12.50 per barrel in 1978 to $21 5? per barrel 1n the second quarter
I
of 197g,
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
At the beginning of 1980, oil'prices were assumed to rise another
AI-6
14 percent, making the 1980 average about 25 percent higher than that of
1979.
(Again this underestimates the actual price changes.
By June, the
average price of imported 011 had risen 53 percent over the average of 1979. 1 )
The model suggests that the 011 price shock caused the level of real GNP to
be 1.1 percent lower by the end of 1979 and 3.9 percent lower by the end of
1980 than it would have been otherwise.
In addition, the domestic inflation
rate was 1.8 percentage points higher in 1979 and 1.3 percentage points
greater in 1980.
Since the underlying 011 price assumptions were too low,
the conclusions of their model also should be taken as underestimates.
Rasche and Tatom (1981) used the argument that a rise 1n the
-
relative price of energy reduces the economic capacity of producers, causes
more inflation, and reduces the full-employment level of output.
Given
time, the energy price hikes also will reduce business investment in plant
and equipment, and lower the desired capital-labor ratio.
Their empirical
results were obtained from production function estimates,
Increases in the
relative energy price were calculated to be 28 percent from 1978-Q4 to 1979-Q4,
based on the actual crude-011 price changes from $12.93 per barrel in December
1978 to $23.63 in December 1979.
This relative price increase was estimated
to have slowed output growth by 3.1 percent over the four-quarter period. 2
They did not provide explicit estimates of the impact of increased energy
prices upon the general rate of price inflation.
1/
2/
Department of Energy, Monthly Energy Review, September 1980, p. 74.
These results are, of course, sensitive to the specific price assumptions.
In an alternative exercise, the authors examine the impact of a relative
energy price change equivalent to the difference between the average
price for 1979 and that for 1978. In this exercise the output growth is
est}mated_ to have ,been slowed by 1.6 percent.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
AI-,7
'
In a follow-up studv, Tatom (1980) µsed a price equation and a
'
variant of the Anderson-Jordan equation;from the Federal R~serve Bank of
St. Louis Model to,assess the impact of 1energy price increases on GNP and
'
inflation.!
I
Both equations were estimated
using quarterly data for the
I
period 1955-Ql to 1978-Q3.
The energy price scenario in Tatom's simulations
I
I
was based on actual ,price developments up to 1980-Q3; the relativ~ price of
I
i
energy wa~ then assumed to remain unchaqged after th~t.
The impact of energy
I
price changes was measured ,b,y implicitly assuming an alternat,ive price scenario
in which relati~e energy prices remainedI con~tant.
Tatom's equption e~timates indicate that -the 1979-80 oil price
I
increases caused nominal GNP _growth to be 1.1 percent~ge points lower than
it otherwise would have been in 1979, arid 2.0 percentage points lower in
I
1980.
I
In 1981, his model predicts, that jnominal GNP will increase as a result
I
of rising prices.
The 1979-80 energy price changes are also esti~ate~ to
'
have ,added O. 7 perce!}tage points to the :measured inflation rate ,in 1979 and
2.4 percentage points in 1980.
I
Combining these two sets of results, Tatom
foun~ tpe impact on real GNP to be a 1.7 percent lower level of output at
the end of 1979 and 6.0 percent lower a~ the end of 1980.
SullDllary.
I
The diverse model structures and feedback mechanisms .in
I
these simulations make it difficult to dompare the results directly.
H~wever,
I
if their oil-price scenarios were unifo~ly adjusted to approximate what
I
actually took place from mid-1979 on, ttteir simulations suggest that the
level of real GNP at the end Of 1980, coulld have been as much ,as 6 percent
I
I
lower than what would have occurred in the absence of the oil price run-up.
I
1/
Tatom's price equation b~sically re~ates the rate of increase in the implicit
GNP deflator to a distributed lag of rates of growth in money stock.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
II
I
I
AI-B
In terms of the impact on the general level of prices, the studies
reviewed suggest that the oil price hikes boosted the inflation rate between
0.1 and 1.8 percentage points during 1979, and between 0.7 and 2.4 percentage
points in 1980.
Given the wide variation in these results and the concern
that these studies might underestimate inflation impacts, we have performed
a simulation exercise of our own.
Simulation results.
The impact of the actual increase in crude-
oil prices since the beginning of 1979 can be estimated by calculating the
increase in the national "oil bill" due to higher import prices.1
First,
total payments made for crude oil--both domestic and imported--during 1979
and 1980 are calculated.
This is contrasted with an estimate of the size
of these payments under the assumption that 011 prices had continued to
increase at the rate established in 1978 (the base-line in figure AI-1.)
We have assumed as a matter of simplicity that the levels of imports
and of domestic production are constant and equal for both price scenarios.
By this we do not mean to ignore such changes as the decreased level of 011
imports seen over the last year.
Rather these assumptions were made in order
to concentrate on the short-term effects of oil-price changes.
The changes
that have occurred in both production and consumption are largely the result
of long-run adaptations and we ignore them in our calculations. 2
1/
2/
The actual calculations utilized data and programming supplied by John
Rosine, of the Board's Wages, Prices, and Productivity Section. He is,
of course, not responsible for any errors in interpretation or assumption
that have been made.
Referring to Professor L.D. Taylor's excellent review of the small amount
of literature on the demand for petroleum products, we note that, for
example, the estimates of the short-run price elasticity of demand for
gasoline vary from -0.07 to -0.80. (See Taylor (1977), page 32, table 1.7.)
The estimates of short-run income elasticity for gasoline range from 0.30
to 0.74. These are widely varying results and reflect both different data
sets and models; they make it difficult to suggest any one picture concerning
the level of imports and consumption under the alternative price scenario
In short, there seems no simple alternative to the assumptions we have
chosen. These assumptions, however, probably bias upward our estimate
of the inflationary impact of the recent international oil-price increases
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Fi~ure AI-1:
REFINER ACQUISITION COSTS FOR IMPORTED CRUDE OIL
Dollars per barrel
38
34
30
26
22
18
Base line projection
14
--,1g9-=J771_ __.~--=1i19r=i71Qs----'--~,:;;0~~;;;9:----1---:--:-~ ~-__JlO
1 98 0
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
AI-9
While the quantities of oil are assumed equal in both our scenarios,
prices, of course, -are not.
In particular, the prices of domestic crude are
assumed to follow a "decontrol path" designed to bring their prices to the
level, in each case, of i~ported oil by the end of 1981. 1
These assumptions allow us to form an estimate of the national
crude-oil bill under our two different price paths.
We assume that the
increased oil bill following the early 1979 price explosion was fully passed
on, partly in terms of higher petroleum production prices, and partly in
terms of higher prices for other goods and services.
The ratio of the change
in the total oil bill to the total value of final goods and services provides
an estimate of the change in the CPI due to higher oil prices.
Table AI-2 presents the results of the simulation.
Column 1 shows
the impact on domestic prices of the actual change in import 'prices.
Column 2
shows the hypothesized direct impact of a continuous increase in import oil
prices of 3 percent per year.
these two price scenarios.
The final column shows the difference in
Thus, adding up the quarterly impacts, the rise
in imported oil prices over the past two years added about 2.2 percentage
points to the inflation rate in 1979 and 2.3 percentage points in 1980.
To
the extent the price increases have not been passed on immediately, these
changes have tended to come in 1980 (and probably in 1981) rather than in
1979 and 1980.2
1/ See the program discussed in Carson and Harnish (1979).
-
2/
This exercise
was performed before domestic crude oil was ordered decontrolled on
January 28, 1981; however, any adjustments to reflect this action would,
of course, be confined to 1981 and later.
Increases in oil prices have noticeable effects on the prices of such
substitute energy forms as coal and natural gas. Therefore, we should
not look at the impact of an increased oil bill, but at the impact of an
increased energy bill. Very rough calculations suggest that these
additional price increases might aggravate the inflationary impact shown
in table 3 by as much as 20 percent. Concentration only on the effects
of increased petroleum expenditures offsets the upward bias contained in
some of the assumptions underlying these calculations.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
'
Table AI-2
1
IMPACT OF OIL PRICE INCRE~SES ON THE INFLATION RATE
(Percentag~ points)
I
I
Estimated Contribut1on to Increase in CPI
Period
Actual crude oil
price changes
Hypothetical
price changes
Differencel
1978-Q4
.13
.04
.09
1979-Ql
Q2
Q3
Q4
.45
. 85
.52
.89
.09
.09
.20
.12
.36
.76
.32
1980-Ql
Q2
Q3
Q4
.68
. 70
.43
.56
.01
. -03
.02
.01
.67
.66
.41
.54
1/
. 77
Differences may not agree with entr~es in the first two columns because
of rounding errors.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
AII-1
Appendix II
'
INFLATION EXPECTATIONS
Inflation expectations play a critical role in many im~ortant
behavioral- macroeconomic relationships,
yet very little is known about them.
'
;
One problem is.measurement, since expectations are inherently unobservable .
.,
""
..
J.
:
"'!
However, infere~ces abou; them c~n be drawn from several surveys as well as
from observed economic behavior. :rwo sources of injormation are, commonly
used and are discussed in the body of this paper.
First, information abou~
the inflation expectations of various types of economic agents is collected_,
directly in several privat~ survexs.
Second, indirect evidence on inflation
expectat~ons can be gleaned from ~he financial securities markets.
While
such data ~ay indicate roughly how price ,expectations have changed over time,
they suffer from ~1gnif1cant uncertai~ties and difficulties in measurement
and, by themselves, can convey no insight into how expecta~ions might shift
in a changing environment.
In principle, models of inflation expectations, derived from economic
I
theory, can provide such insights in a framework that is consistent with the
ax1Qms of rational economic behavior.
These models, however, frequently are
not conducive to emp~ric~l estimation of crucial parameters.
Moreover, their
-
basic structure may change when a new regime in monetary policy is introduced,
thus making assessments of the change in operating procedures quite difficult
1£ not impossible.
In ~his appendix both the conceptual and the empirical
'
difficulties in examining changes in inflation expectations are explored
'
;
'
furth~r~ :The di~pussion is divided into a r~view of probl~s inherent in
.,.
"'
...
('
...
survey data, the' evidence
from financial markets, and theoretical models.
'
-
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
AII-2
Sur~ey data.
Private surveys offer the most direct measures of
inflation expectations, but the samples
I
ften are statistically deficient.
01
Three surveys are discussed in.the main text.
The University of Michigan
Survey Research Center· (SRC) asks several, questions every month that are
designed to measure directly consumer inflation expectations.1
The Livingston
biannual survey of "informed" business economists provides a less timely
measure of expectations. 2
Finally, the Blue Chip Economic Indicators news- '
letter provides monthly results from a survey of several private economic
forecasters.
As noted in the text of the paper, all three~surveys indicated
roughly the same pattern of changes in expectations following the policy ''action
of October 6, 1979.
In the months that followed,
the average expected rate
I
of inflation increased noticeably, in some instances dramat'ically, and it
remained high in the first few months of 1980.
After the introduction of
1/ The SRC anticipations data are constructed from the following questions:
"During the next 12 months, do you think that prices in general will go
up, or go down, or stay where they are now?" and."By about what percent
do you expect prices to go up, on the average, during the next 12 months?"
The questions refer to the CPI.
2/
Since 1947, Joseph A. Livingston has collected biannually the anticipations
of economic variables from economists, businessmen, and professional forecasters. Livingston mails questionnaires to respondents in early De~ember
and May and asks for a 6-, 12-, and (in May) an 18-month ahead forecast
of the level of the CPI and the PPI for finished goods. Results of these
surveys are published regularly in Liyingston's column in either the
Philadelph1a Inqu1rer or the (Philadelphia) Bulletin.
Early research used the reported average survey response (Turnovsky,
1970; Turnovsky and Wachter, 1972; Pekando, 1975), but these figures are
often arbitrarily adjusted by Livingston. More recently, analysts have
obtained individual forecaster responkes to calculate a better measure' of'
the expected inflation rate (Carlson,' 1977; Wachtel, 1977; Figlewski and
Wachtel, 1978; Hafer and Resler, 1980~. Moreover, it fs not' clear whether
the responses represent predictions of the price level using May and
December as the base period or April ~nd October· (months for which' index
numbers are supplied to the responden~s by Livingston). Recent research
(Hafer and Resler, 1980) has suggeste~ that rationality tests of the
Livingston survey are quite sensitive: to the interpretation of the length
of the forecast horizon.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
I
1
Table AII-1
PERCEIVED AND ANTICIPATED INFLATION
Perceived 1
Anticipated 2
1979-0ctober
November
December
n a.
8.9
11.2
1980-January
February
March
April
May
June
16.6
16.1
16.6
17.4
15.9
14.4
Period
J,uly
August
l?eptember
October
November
December
16.0
16.4
14.5
14.4
14.5
15.0
14.7
t5.2
10.8
13.1
10. 7
12.0
11.1
8.2
8.1
9.0
8.0,
8.5
9.2
8.7
10.1
1/ Mean incre 9 se obtained from "By }mat percent do you think prices
2/
have gone up, on the average, during the last 12 months?"
Mean increase obtained from "By about what percent do you expect
prices to go up, on the average,-during the next 12 months?"
1
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
AII-3
I
I
credit controls in March 1980 and a recor~ decline in the money supply in
April, expectations eased significantly. : They generally remained at these
lower levels until the end of 1980.
I
Analysis of survey data suggestsI a note of caution in the use of
these indicators of inflation expectations.
exhibits several deficiencies.
The Michigan survey, for example,
Changes i~ this measure tend to be fairly
sensitive to movements in the prices of food
and gasoline--items that are~
I
I
I
observed by consumers more frequently tha~ most other categories.
Another
I
characteristic of this survey is the persistent contrast between expectations
'
I
and perceptions of inflation--perceived inflation
is usually much higher than
I
expected inflation. 1
I
The magnitude of the difference between perceived and
I
anticipated inflation rates is quite startling, as is the failure of subsequent
I
I
expectations to respond to changes in per9ept1.ons.
Livingston's surveys of price e~pectat1ons have been the subj~ct of
I
r
considerable analysis.
Most stl,!dies have lbeen interested in the rati~nality
l
of surveyed expectations.
The test of rat\ionality has usually taken the
form of:
(AII-1)
t and Et-1 { Pt } is the previous
I
Ratipnality requires that' the expectations
where Pt is the inflation rate for period
period's expected rate of inflation.
I
be unbiased--in equation AII-1 this is equivalent to the joint hypothesis
I
that ao=O and a1=l, where the errors, Ut, ~re assumed to be serially independent.
I
1/
This information is obtained from the &uestion: "During the last 12 months,
have prices of the things you buy remained unchanged, or have they gone uo,
or have they gone down?" and "By aboutlwhat percent do you think prices
have gone up, on the average, during the last 12 months?"
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
I
I
I
AII-4
With these parameter restrictions, the actual inflation rate deviates from
the expected r~te only in an unanticipated way.
Early tests for efficiency (Pesando, 1975; and Carlson, 1977) employed
the following structure:
(AII-2)
(AII-3)
If expectations are efficient and prices can be ~haracterized by the timecseries process in equation AII-2, then ai=bi for all lags i.
This ,test is
inappropriate, however, if the errors are not homogeneous and independent.
An alternative test regresses forecasting errors on past inflation rates:
(AII-4)
If the forecasts are efficient, they should be uncorrelated with any past
information. 1
This form of the efficiency hypothesis requires all coefficients
to be zero.
Early studies of the Livingston data (Turnovsky and Wachter, 1972;
Pyle, 1972; Gibson, 1972; deMenil and Bhalla, 1975) employed the average
reported mean of inflation expectations.
A general consensus emerged from
the studies that expectations could be described by an adaptive or extrapolative scheme.
However, these studies were marred by the quality of reported
expectations, which often were arbitrarily adjusted by Livingston.
More recent tests of rationality expressed in the form of equation
Ail-1 have presented mixed results.
Pesando found the data to be consistent
with the rational expectations hypothesis; however, Figlewski and Wachtel
]:_/ Note that this statement of rationality is a weak form of the hypothesis
since "past info~tion" includes only the past history of prices.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
ail-5
(1978) were able to reJect the hypothesis using a pooled time-series, cross-
section sample that included more recent data.
More recently, Hafer and
Resler (1980) found characteristics o~ bias in the Livingston forecasts,
I
regardless of the forecast horizon and the sample period used.
Some tests for efficiency (Carlson, 1977; and Pesando, 1975), ,which
are based on equation AII-2, are flawed by the presence of nonhomogeneous
'
residuals--under those conditions the,F-statistic does not take on the
properties usually assumed.
An alternative test using equation AII-4 was
proposed by Mullineaux and also was used by Hafer and Resler (1980).
Mullineaux's test reJected eff1ciency 1 for the sample period 1959-1969.
However, Hafer and Resler have shown that the efficiency tests are not robust
and depend on the sample selected and the forecast horizon of the survey
respondents.I
Nevertheless, their evidence suggests that forecasters are
more efficient in predicting longer-term inflation than short-term price
developments.
In summary, most evidence suggests that the Livingston surveys have
serious limtations.
Because they fail to conform with rationality criteria,
I
'
either (1) the survey does not accurately measure inflation expectations, or
(2) expectations are slow to absorb new information.
From a theoretica~
I
point of view the second implication is difficult to accept.
I
Evidence from financial market~.
In an analytical sense, surveys
I'
of inflation expectations may not be ~egarded as a useful.source of informatior
since there is no evidence that economic actions are directly based on these
I
I
expectations.
-1/
However, there are sev~ral sources of data, directly linked to
Confusion in the interpretation of the appropriate time horizon in the
Livingston data has led to the use of several different forecasting
horizon lengths: (1) Mullineaux kssumes the horizon to be from April to
October, (2) Jones-Jacobs employep a May-December horizon, and (3) others
assume it to be from April to Dec~ber.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
I
AII-6
economic acti~ities of one sort or another, from which inferences about
expectations potentially can be extracted.
These include the financial
futures markets, the implied forward rates in long-term securities, the
commodity futures markets, and the wage bargaining process--particularly
for the unionized sector, where coverage by formal escalator clauses is
concentrated.
Because an extensive literature is available on the movement
of interest rates, we examine most closely the information available in
financial market transactions.
The expectations hypothesis of the term structure of interest
rates is a widely accepted explanation of interest rate relationships.
This
hypothesis holds that long-term rates are a weighted average of expected
short-term current and forward rates; because forward rates should reflect
the expected rate of inflation, observations on levels of and shifts in the
yield curve should embody some jnformation about market expectations of
inflation.
However, alternative hypotheses suggest other sources of change
in the forward rates, including:
(1) a nonconstant real rate,1 (2) risk
premiums, (3) liquidity premiums, or (A) differences arising from the possibility that securities may be traded in segmented markets.
1/
It may be possible
Earlier studies have shown that long-term rates reflect market expectations of inflation and are efficient forecasts of future prices (Granger
and Rees, 1968; Bierwag and Grove, 1971; Laffer and Zecher, 1975, Phillips
and Pippenger, 1976; Sargent, 1976, 1979; Mishkin, 1978; Pesando, 1978;
Fama, 1975; Barro, 1978; Lucas, 1973, Sargent and Wallace, 1975; Phelps
and Taylor, 1977; Fischer, 1977). For example, Fama (1975) finds that
market rates use all relevant information about price developments and
that the real rate is conqtant. However, Shiller (1979) suggests that
the relatively constant real rate in Fama 1 s sample may be attributable
to Federal Reserve behavior, not the inability of the Federal Reserve
to induce unanticipated surprises on the market. He identifies and
analyzes tests of three nested hypotheses about expectations that are
common in the literature and concludes that· "none of the hypotheses is
likely to be so strictly true as to rule out completely a predictable
effect of systematic monetary policy on expected real interest rates"
(p. 65).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
AII-7
to abstract from considerations regarding risk and differences due to trading
1n segmented markets by concentrating on Treasury securities, which have
relatively uniform risk character1st~cs.
Since late 1979 yields 1n 1ntermed1ate- and longer-term Treasury
markets have risen noticeably and displayed greater volatility.
Analysis by
the staff, presented by Johnson 1n "Lnterest Rate Var1ab1l1Ly'Under the
'
New ... Procedures" 1n this compendium, found little evidence that l1qu1d1ty
premiums rose during periods of 1nter,est rate volatility.
At best, this
factor may acGount for only a very small fraction of the rise 1n interest
rates.
However, at the same time rates on securities with maturities of
about one year were much more volat 1l'e than would have been implied by an ex
post rational long-term rate--that is:, a hypothetical series which would
have resulted from a perfect forecast of short-term rates.
This result
raises the question whether the volat'1l1ty 1n 1ntermed1ate-term rates
discredits the expectations hypothesis or whether there is information 1n
the forward rates about market expect,ations of inflation.
Shiller (1979), among many other analysts, has addressed this issue
and argues that conventional tests of rationality may be weak if long~term
I
interest rates are too volatile.
Thej high relative volatility of long rates
I
compared with that of short rates violates some of the assumptions which lead
I
to the traditional characterization o~ long rates as a weighted average of
I
I
expected short rates.
His tests generally reject the expectations model 1n
!
favor of a model that allows for long, rates to be influenced-by transient
I
effects unrelated to expectations.
influence by expectations.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
I
Hpwever, he makes some allowance for
AII-8
On balance, the literature on this subJect suggests that discretion
should be used in making inferences about inflation expectations based on the
implied forward rates in the term structure of interest rates.
If inflation
expectations were the predominant influence on the expected forward rates
embedded in the term structure over the past two years (chart AII-1), two
observations are suggested:
(1) inflation expectations became more pessimistic
last fall and remained high even at the trough of long-term rates in June
1980 and (2) the relationship among forward rates for various time horizons
~
was so erratic in 1980 that it is difficult to explain how relevant incoming
information 'could have been systematically utilized.
Theoretical models.
Models of inflation expectations are useful
because they provide a ,characterization that is consistent with the principles
of rational economic behavior.
A wide variety of models for inflation expecta-
tions have been used in past research.
Most efforts to model inflation
expectations are special cases of the more general autoregressive expectations
mechanism:
Et { Pt+l}
=
n
Ea Pt-i·
i=O i
They range from naive expectations models, where
Et { Pt+l }
= Pt,
to extrapolative models, where
A< 1,
to adaptive expectations, where
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Et { Pt+l}
= Et-1
{ Pt } + 8 (Pt-1 - Et-1 {Pt} )
8 < 1.
AII-9
These ~odels are thought to be consistent with a limited concept of rationality.
However, Feige and Pearce (1976) suggested a framework in which autoregressive
1
models might be ec'onomically rational given the costs of obtaining'\ additi~nal
information.
In fact, they demons trat ed that an autoregressive expectations
1
I
,I
mechanism will provide an efficient forecast of inflation; the residual of
~
'
J
'
an appropriate ARIMA model for inflation was uncorrelated with lagged innovations
j
in monetary or fiscal policy.
This question is an empirical issue and could
be tested in a more general class of m6dels that allows for other expla~atory
variables, in addition to information ~vailable in the.past history of prices.
The modeling approach to inflation expectations,suggests that an
appropriate model of the inflation process be selected and then used to forecast inflation.
This procedure is subJect to several criticisms.
First, the
structure of the model should change with every transition to a new regime of
monetary policy.
In addition, to the extent that the Federal Reserve does
not innnediately affect the actual inflation rate, an extrapolative model of
inflation will ignore essentially all information about the new operating
procedures.
However, these models may be useful in analyzing the behavior
of real interest rates.
I
Concluding remarks.
This appendix
has explored several sources and
I
methods that frequently have been used in the past to make inferences about
inflation expectations.
On balance, moaeling techniques are of limited
usefulness because they require more data than are available.
Inferences
about inflation expectations that are bksed on financial market transactions
are highly tenuous because of the implitations of high interest rate volatility
I
I
for the expectations hypothesis of the term structure.
Finally, although
survey data provide a direct estimate of inflation expectations, they exhibit
a number of peculiar characteristics.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-Rl-
REFERENCES
Barro, Robert. "Unanticipated Money, Output and the Price Level in the
United States," Journal of Political Economy, August 1978, 86(4),
pp. 549-80.
Berndt, Ernst R., and David O. Wood. "Engineering and Econometric Interpretations of Energy-Capital Complementarity," American Economic Review,
June 1979, 69(3), pp. 342-3.
Bierwag, G. O., and M.A. Grove. "A Model of the Structure of Prices of
Marketable U.S. Treasury Securities," Journal of Money, Credit and
Banking, August 1971, 3, pp. 605-29.
Carlson, John. "A Study of Price Forecasts," Annals of Economic and Social
Measurement, Winter 1977, 6, pp. 27-56.
Carson, William, and Douglas Harnish. "Final Regulatory Analysis of Rulemaking on Phased Decontrol of Upper Tier Crude Oil," U.S. Department
of Energy, Economic Regulatory Administration, November 1979.
deMenil, George, and SurJit Bhalla. "Direct Measurement of Popular Price
Price Expectations," American Economic Review, March 1975, 65(1),
pp. 169-80.
Fama, Eugene F. "Efficient Capital Markets· A Review of Theory and Empirical
Work," Journal of Finance, May 1970, 25, pp. 383-417.
"Short-Term Interest Rates as Predictors of Inflation,"
American Economic Review, June 1975, 65(3), pp. 269-82.
"Forward Rates as Predictors of Future Spot Rates," Journal
of Financial Economics, October 1976, 3, pp. 361-77.
Fe1.ge, Edgar L., and Douglas K. Pearce. "Economically Rational Expectations.
Are Innovations in the Rate of Inflation Independent of Innovations in
Measures of Monetary and Fi.seal Policy?" Journal of Political Economy,
June 1976, 84(3), pp. 499-522.
Figlewski, Stephen, and Paul Wachtel. "The Formation of Inflationary Expectations," New York University, Graduate School of Business Administration,
November 1978.
Fischer, Stanley. "Long-Term Contracts, Rational Expectations and the
Optimal Money Supply Rule," Journal of Political Economy, February 1977,
85(1), pp. 191-205.
Gibson, William E. ''Interest Rates and Inflationary Expectations: New
Evidence," American Economic Review, December 1972, 62(5), pp. 854-65.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-R3-
Phil lips, Llad, and John Pippenger. "Preferred Habitat vs Efficient
Market . A Test of Alternative Hypotheses," Federal Reserve Bank of
St. Lou~s Review, May 1976, 58(5), pp 11-19.
Pyle, David H. "Observed Price Expectations and Interest Rates," Review of
Economics and Statistics, August 1972, 54, pp. 275-80.
Rasche, Robert, and Tatom, John
Energy Price Shocks, Aggregate Supply
and Monetary Policy· The Theory and International Evidence. CarnegieRochester Conference Series on Public Policy. Vol. 14
1981.
Sargent, Thomas J. "Rational Expectations and the Term Structure of Interest
Rates," Journal of Money, Credit and Banking, February 1972, 4, pp 74-97.
"A Classical Macroeconometric Model of the United States,"
Journal of Political Economy, April 1g76, 84(2), pp. 204-37.
"A Note on Maximum Likelihood EstiP1ation of the Rational
Expectations Model of the Term Structure," Journal of Monetary EconoP1ics,
January 1979, 5, pp. 133-43.
Wal lace. 11 'Rational I Expectations, the Optimal
and the Optimal Money Supply Rule," Journal of
Political Economy, April 1975, 83(2), pp. 241-54.
, and Neil
-------Monetary Instrument,
Shiller, Robert J. "Can the Fed Control Real Interest Rates?" Research Paper
39. University of Chicago, April 1979.
"The Volatility of Long-Term Interest Rates and Expectations
Models of the Term Structure," Journal of Political Economy, December
1979, 87(6), pp. 1190-1219.
"Alternative Tests of Rational Expectations Models· The
Case of the Term Structure." Working Paper 563. National Bureau of
Economic Research, October 1980.
Tatom, John A. "Energy Prices and Economic Performance." Federal Reserve
Bank of St. Louis, November 1980 (mimeo).
.,, , ,,, ,
Taylor, Lester D. "The Demand for Energy· A Survey of Price and Income
·ElasticLties," i,n,William,D. Nordhaus,, ed;, International Studies-~of
the Demand for Energy. Amsterdam· North-Holland, 1977, pp. 3-43.
Thurman, Stephan, and Richard Berner. Analysis of Oil Price Shocks in the
MPS Model, CEP~ Conference on Energy Prices, Inflation, and Economic
Activity, Cambridge, Massachusetts, November 1979.
Turnovsky, Stephen J. "Some Empirical Evidence on the Formation of Price
Expectations," Journal of the American Statistical Association, December
1970, 65, pp. 1441-54.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-R4-
, and Michael L. Wachter. "A
-------Hypothesis Using Directly Observed Wage
Test of the Expectations
and Price Expectations,"
Review of Economics and Statistics, February 1972, 54, pp. 47-54.
U.S. Department of Energy, Monthly Energy Review, September 1980.
Wachtel, Paul. "Survey Measures of Expected Inflation and Their Potential
Usefulness," in Joel Popkin, ed., Analysis of Inflation: 1965-1974,
Ballinger for the National Bureau of Economic Research, 1977.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Money Market Impacts of
Alternative Operating Procedures
January 1981
Paper Written for a Federal Reserve
Staff Review of Monetary Control Procedures
by
Peter A. Tinsley, Peter von zur Muehlen,
Warren Trepeta, and Gerhard Fries
MONEY MARKET IMPACTS OF ALTERNATIVE OPERATING PROCEDURES*/
In October 1979, reserves-oriented operating procedures were adopted
for the execution of short-run monetary policy.
The historical record of
money market volatility since that date has not been encouraging.
As shown
in figure 1, the standard deviations of both the monthly growth rate of M-lA
and the monthly change in the federal funds rate increased markedly during the
12 months subsequent to the alteration of procedures in contrast to the
standard deviations for the preceding 12 months.
This paper explores the short-run volatility consequences of money
stock targeting under current and alternative operating procedures.
The
focus is narrowly drawn on the feasibility of money stock targeting, an issue
that may be considered independently of the desirability of intermediate
targeting on monetary aggregates.
o
Two principal issues are considered:
Was the money market buffeted by atypical events in
...)
1980, or is there an inherent flaw in current operating
procedures that tends to induce volatility in money
markets?
o
Does there exist a well-behaved trade-off between the
volatility of deviations of M-lA from long-run targets
and the volatility of short-term interest rates under
current and alternative operating procedures that may
be exploited by short-run monetary policy?
*/ Sections of this paper were prepared by P.A. Tinsley, P. von zur Muehlen,
and G. Fries (sections 1-4 and appendix A) and w. Trepeta (section 5 and
appendixes Band C) with special assistance from H. Farr, B. Garrett, J. Lovin,
V. Watkins, and c. Wilson.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 2 -
Figure 1:
1.a
Historical Variability in the Monthly Growth Rates
of Monetary and Reserve Aggregates and the
Monthly Change in the Federal Funds Rate
M-lA
1978.10 - 1980.10
Gt(M-lA)
·-
0
--,--------------------------,
s;j""
0
CJ
0
0
0
('J---~.,..-,--,--,--,--,--,--,---,--,--,--..--,--,--,--,--,--,--,---,--,---,--..........,.--,.---.--1
I
78.10
80.10
79.10
- annualized growth rate of M-lA in month t
- standard deviation of growth rates 78.10 - 79.09
- standard deviation of growth rates 79.11 - 80.10
1.b
RFF
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
EiRFF
0
0
+- cr l (t:.RFF t) = . 27---+j ~ o (LlRFF) = 2. 4 ~
2
C0-4-----,-~-..-.--,----,---,--,--1......,.--,--,--,--..--,--,----,---,--,--..----t-
l
78.10
79.10
- change of federal funds rate in month t
t:.RFF
cr l (t:.RFF t) - standard deviation of monthly change 78.10 - 79.09
standard deviation of monthly change 79.11 - 80.10
cr 2 (t:.RFFt)
- 3 Figure 1, continued
1.c
NBR
Gt (NBR)
0
o~-----------------------,
78.10
l.d
TR
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
79.10
80.10
Gt(NBR) - annualized growth rate of nonborrowed reserves (seasonally
adJusted) in month t
standard deviation of growth rates 78.10 - 79.09
crl(Gt)
standard
deviation of growth rates 79.11 - 80.10
cr2 (Gt)
Gt(TR)
Q,~----------------------,
l'--..
0
IV)
0
22.2~ I+-- cr 2 (G ) = 23. 7 -+
I
t
0
l.{)....J---+-.,........,.........----.---,--~.--.--.."""T""--r--r-r.----r-r~-r,r-r"""T"--r-.---,-...,.--r-,
I
78.10
79.10
80.10
Gt (TR) - annualized growth rates of total reserves (seasonally
adJusted) in month t
cr 1 (Gt) - standard deviation of growth rates 78.10 - 79.09
cr 2 (Gt) - standard deviation of growth rates 79.11 - 80.10
- 4 The results of this study, obtained by both deterministic and
stochastic simulations of a monthly money market model used by the staff,
are the following:
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
o
The odds are at least two to one that a portion
of the increased money market volatility observed
in 1980 should not be solely ascribed to current
operating procedures.
o
There exists a well-behaved trade-off between the
volatility of money stock targeting performance and
the volatility of short-term interest rates in the
sense that an improvement in the performance of one
objective can be exchanged for a bounded deterioration in the performance of the other.
Other results of this study include the following:
o
Both interest rate and reserves policies are more
successful in attaining year-over-year money stock
targets than in maintaining close adherence to a
money stock target path within the year.
o
Examination of short-run M-lA objectives in 1980'
suggests that the FOMC attempted to make up about
30% of the perceived gap between the proj'ected
money stock and the annual target path in the following month.
o
Tight restrictions on tne target range of admissible
monthly variations in the federal funds rate will
dominate variations fn the desired1 speed of reent~y
to the annual money stock target path.
- 5 -
o
If the target range on the federal funds rate is
sufficiently relaxed, the estimated speed of reentry
to the annual money stock target path estimated for
historical procedures is approximately effici~nt in
the sense that faster speeds of reentry would yield
much larger fluctuations in the federal funds rate
with only small improvements in the volatility performance of the money stock.
o
There is some evidence that approximately the same
money stock targeting performance can be achieved by
a federal funds rate policy as by a nonborrowed reserves
policy at a lower cost in interest rate volatility if
the planned settings of the federal funds rate are
sufficiently aggressive.
o
No evidence is provided in this study of unstable interest
rate cycles induced by money stock targeting.
Thus, at
least in the context of this study, it is unlikely that
interest rate instability is a significant constraint on
the design of short-run ~onetary policy.
I
Discussion in the paper is organized along the following lines:
First, the concept of short-run stochastic volatility is introduced briefly
in section 1.
One consequence of the design of short-run operating procedures
is the allocation of random disturbances within money markets.
The selection
'
of a reserves-oriented or an interest-rate-oriented
policy affects the allocation of transient disturbances between the money stock and short-term interest
rates.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 6 -
Second,
in
order to examine the volatility allocations of alterna-
tive operating procedures, it is necessary to have some method of generating
the expected consequences of alternative policies.
Section 2 describes a
procedure for stochastic simulations of the econometric model of monthly
activity in money markets used by the staff.
The design and execution of
monthly monetary policy is characterized by simulations of policy planning
and execution stages.
In the policy planning stage, the policy authority selects a shortrun money stock objective (for M-lA) and a policy instrument setting (either
nonborrowed reserves, total reserves, or a federal funds rate setting) that
will achieve the short-run money stock objective in the next month, at leab.
in the absence of forecast errors.
Generally, if the money stock has been
displaced from the annual money stock target path in recent history, the
short-run money stock objective does not represent a plan for an immediate
return to the annual target path within the next month because month-to-month
departures from the annual target path are viewed as partially the result of
transient disturbances that will tend to "wash out" over time.
In the policy execution stage, the ex ante money stock targeting
plans of the policy authority may be partially frustrated by the impacts of
unforeseen random disturbances.
These impacts are represented by stochastic
simulations of the monthly money market model, given the ex ante policy
instrument setting selected by the policy authority.
simulations are used in this study:
Two types of stochastic
(1) in pseudo-history simulations, the
forecast errors of the monthly model in 1980 are added to estimate the
performance of monetary policies under cpnditions that existed in 1980;
(2) in average-history simulations, random disturbances similar in size and
pattern to the forecast errors of the monthly model from the nine-year sample
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 7 -
1971-79 are added to gauge the robustness of money stock targeting performance when policies are exposed to a full spectrum of plausible money market
shocks.
Section 3 contrasts the performance of current operating procedures
under pseudo-history and average-history stochastic simulations in 1980.
The
purpose of this comparison is ~o determine if the random disturbances observed
in 1980 were unusually severe in contrast to shocks in the 1970s, or if
reserves-oriented operating procedures are largely responsible for the
increased volatility of money markets in 1980.
Simulation experiments described in section 4 contrast the allocation
of money market volatility associated with alternative operating procedures.
The results suggest that money stock targeting performance is sensitive to
the range of variation permitted for the federal funds rate.
Another important
determinant of money stock targeting performance is the planned monthly speed
of reentry to the annual money stock target path represented by the selection
of the short-run money stock objective.
The results suggest that the current
rate of reentry to the annual money stock target path implied by historical
short-run money stock objectives in 1980 is approximately efficient in the
sense that attempts to close the gap between the money stock and the annual
target path more rapidly would produce large increases in the expected volatility of the federal funds rate in exchange for small improvements in money
stock targeting performance.
Section 5 explores the possibility that close control of the money
stock may induce undamped cycles in short-term interest rates.
Results of an
experiment with the staff monthly model suggest that interest rate instability
is not an effective constraint on the design of money stock targeting procedures.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 8 -
Finally, three appendixes provide more explicit descriptions of
(1) tte e~sential economic structure of the staff monthly money market model,
the ~im~lation characteriz~tion of monthly policy used in this paper, ~nd
the methodology underlying stochastic simulation~; (2) an approximation of
the FOM9 selection of short-run objectives for the narrow money stock (M-\A)
in 1980; ~tnd (3), an examin~tion of the potential for the staff monthly
money m~rket model to ?X~ibit interest rate instability.
1.
The Concept of Short-Run Stochastic Volatility
High-frequ~ncy oscillations in the in9icators of ~onetary policy
may be viewed w~th dismay by ~oney mark~t particip~nts, in P,~rt, b~~ause t~ey
obscure the underlyin~ intentioqs of the policy authority.
Ho~e~er, npt a+-
kinds of meas~r:ed in~rea~es, in the variability of money market instr~me~ts
imply less information about policy intentions.
In the case of money stock
targeting, the gross varia~ility of the monthly growth rate of the money
stock may be an inappropriate me?sure of policy per+ormance.
If the money
stock is forced off a target path by an unanticipated disturbance, the growth
rate of the money stock must be ag,gressive~y alt:ered in subsequent months to
recover the targeted p~th.
In this ca~e, a more suitabl~ measure of undesir-
able volatility may be the standard deviation of monthly departures from the
money stock target path.
Sµnilarly, the dispersion of unexpected changes in
short-term tnt~rest rates ~ay be a mor~ r~levant measure of undesirable
interest rate volatility than the fluctuations of total changes in interest
rates.
Thus, in this p?per, undesirable volatility is differentiated from
gross variability when volatility is a short-run stochastic concept referring to the dispersion of outcomes around planned objectives or expectations.
Stochastic volatility is unavoidable in an economic environment
that is subjected to unpredictable and sizable disturbances.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Where that
- 9 -
volatility is allocated within money markets depends importantly on the
design of monetary policy.
Attempts to eradicate transitory variations in
target variables, for example, may increase the short-run volatility of
other money market variables in exchange for little improvement in the longrun performance of the target variables.
The extent of this unavoidable
short-run stochastic volatility and the nature of the trade-off allocations
available to money stock targeting procedures are examined in the next three
sections of this paper.
2.
An Econometric Portrayal of
u.s.
Money Markets and Alternative
Operating Procedures
An econometric model of monthly financial behavior
Estimates of short-run stochastic volatility in money stock targeting
procedures have been obtained from stochastic simulations of an econometric
model used by the staff to generate monthly econometric forecasts of money
market behavior.
The stochastic simulation approach was adopted to circumvent
the lack of an extensive historical track record with the new operating procedures.
Stochastic simulations permit 1980 to be "rerun" under alternative
random disturbances or under competing policy procedures.
In the current version of the model, 20 estimated equations plus
several accounting identities project reserve aggregates, the components of
M-3, and selected short-term interest rates, given judgmental projections of
the monthly paths of personal income and the consumer price index (CPI), and
an assumed path for the policy instrument -- nonborrowed reserves (NBR), total
reserves (TR), or the federal funds rate (RFF).!/
Real economic activity is
1/ See appendix A of this paper for a discussion of the economic structure
of the monthly model. For a complete description of the current monthly econometric model, see H.T. Farr, "The Monthly Money Market Model," working paper
(Board of Governors of the Federal Reserve System, July 1980, revised November
1980 ).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 10 -
exogertous in this model and not affected by short-term alterations in the
conduct of monetary policy.
Essentially, the model provides a characteriza-
tion of the interactions of money demand and supply based on projections of
nonbank demand for bank liabilities and the consequent short-term portfolio
adjustments by banks.
Three equations were added to the existing monthly model:
Judg-
mental projections of both personal income and the CPI were replaced by
econometric time series projections since a complete history of judgmental
forecast errors was not available.
Also, an estimate of a historical
reaction ruie for the Federal Reserve discount rate was added to provide
a description of the probable adjustment of the discount rate in hypothetical simulation experiments.
in counterfactual simulations, such as those
described in section 4, the federal funds rate may move far off its historical path, and it is unlikely tnat the policy authority would permit
a
large
spread between the discount rate and the federal funds rate to persist for
an indefinite period.
As described in appendix A, the estimated historical
reaction rule adjusts the discount rate toward the federal funds rate with
a mean lag of about three and one-half months.
Given the econometric model description of monthly financial activity, a characterization of the design and impacts of monthly monetary policy
is represented by the following three steps.
1.
Setting_ the__iIJ.tedm _moge,1 _st_oc}c__target_.
At each FOMC meeting, the policy authority selects an interim M-lA
target for the month ahead.
This procedure is illustrated in figure 2.
In
monthly terms, the 1980 target path for M-lA (denoted as MT) is represented
by a 4.75% growth path from 1979.11 to 1980.11, approximating the midpoint
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 11 -
Figure 2:
Illustration of Selection of Interim M-lA Target
ln M
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
••
•
•
I
/
/
/
/
/.
ln M1 '
/
,,,,, ,,,,,,
,_,.,,,, ln MI (A.=O)
2
,,,,.
~
• • • •
1
MT
2
--------t
12
- M-lA target path (4.75% annual growth)
M1 - P~oJected M-lA in month 1
MI 2 - InterimM-lA target fer menth 2
(a) A=O; no closure of relative target gap (base drift),
(b) A=l; full closure of relative target gap in one month
A
- Monthly rate of reentry to annual target path (MT)
ln
- natural logarithm
- 12 -
of the fourth-quarter-to-fourth-quarter target range selected for M-lA by
the FOMC.
At the time of the FOMC meeting, which typically occurs near the
end of the month (month 1 in figure 2), the policy authority is faced with
an unplanned deviation of the projected money stock of the current month,
M1, from the target path, MT.
The FOMC then selects an interim money stock
target for the next month (month 2).
The fraction of the gap between the
money stock and the annual target path that the committee plans to eliminate
in the next month is termed in this paper the "monthly rate of reentry" to
the annual target path, denoted by A.
If A= 1, the policy authority plans
to return to the annual target path, MT, in one month.
On the other hand,
if A= O, the authority plans to achieve an annualized monthly growth in
M-lA equal to the annualized growth rate of the target (4.75%) but starting
from the current money stock, Mt, rather than the target value for month 1.
Thus, in the case of A= O, the policy authority does not intend to reduce
the relative money stock target gap in month 2, a choice that leads to planned
base drift.
In this characterization of interim target selection, the selection
of the reentry rate, A, determines the persistence of past random disturbances in the current money stock target gap, MTtlMt•
If A= 1, the effect of
a random disturbance in period 1 that causes the money stock, Mt, to deviate
from targeted money, MT1, is eliminated in one month.
Since the reentry rate
is fixed for all months in the planning horizon, a zero reentry rate (A= O)
implies that the effective duration of the impact of a random disturbance is
infinite since there will be no planned offset.
A reentry rate, A, can be
converted to an implied monthly age, A, of random disturbances in the money
stock target gap, as shown in table 1.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
The typical monthly reentry rate of
- 13 -
Table 1.
Translation of Monthly Reentry Rate (A) to
Average Age (A) of Money Stock Target Misses
ReentrI rate (A)
Average monthly
age (A) Jj
1
.333
.292
1
3
3.4
J:..I
.167
.111
6
9
0
00
1/ Average age (in months) of random disturbances in the money stock
target gap (ln MTt - ln Mt). More explicitly,
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
00
A= AL i(l - A)i-l,
i=l
= 1/A.
]:_/ Estimated reentry rate of historical interim targeting procedures
80.02 - 80.11; see appendix B.
- 14 -
current targeting procedures in 1980, estimated by
w.
Trepeta, is 0.292.~./
This implies an average age of random disturbances in the annual money stock
target gap of about three and one-half months.
2o Setting the policy instrument.
A planned setting for the intermeeting interval (represented by the
following month in this discussion) is then selected for one of three possible
\
policy instruments -- nonborrowed reserves, NBR, total reserves, TR, or the
federal funds rate, RFF.
The selection of the policy instrument setting is approximated in
model simulation exercises by the following procedure:
It is assumed tha,t
the projection of the money stock in the current month of the meeting is sufficently accurate so that any remaining forecast error may be neglected.
A
forecast of ,money market behavior in the following month is then simulated
by the staff monthly model as if the interim money stock target, MI, for that
month is the effective policy instrument.
This forecast provides settings
for nonborrowed reserves, NBR, total reserves, TR, and the federal funds
rate, RFF, that are consistent with achieving the interim money stock target,
MI, at least in the absence of forecast errors.
In some of the cases analyzed, the FOMC is assumed also to place a
target range on the federal funds rate so it cannot move by more than 300
basis points from the current month to the following month.~
In these
2/ Estimation of the historical reentry intentions of current procedures
is discussed in appendix B.
3/ One motivation f or a ,tar,get range on an auxiliary variable, such as the
federal funds rate in a nonborrowed reserves policy, is to provide a rough
check for operational breakdowns of the planning model. If actual events
move the auxiliary variable outside the auxiliary target range, actual events
may not be statistically compatible with the ex ante forecast. When this
occurs, the planning model may be missing some ingredient in the structure
of the actual economy, and the policy authority may wish to reconsider planned
policy. Using this interpretation, the target range for the auxiliary variable should bear some resemblance to a confidence interval of the ex ante
projection of the auxiliary variable.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1
- 15 cases, the funds rate constraint must be satisfied in both the ex ante policy
planning stage and in the ex post policy execution stage (when random disturbances are encountered, as explained shortly).!!./
3.
Simulation of subsequent "history."
After selecting a policy instrument setting (either nonborrowed
reserves, total reserves, or the federal funds rate) that will achieve the
interim money stock target, Mlt, in the absence of random disturbances, the
model is then resimulated in the second month with nonzero random disturbances.
All variables except the policy instrument are affected by the random
disturbances.
There are two types of stochastic simulations:
(1) In pseudo-history
stochastic simulations of a month in 1980 (1980.01-1980.10), the historical
forecast errors for the model are included in the simulation.
Thus, if the
policy instrument is set at its historical path, actual monthly history would
be simulated for all variables in the pseudo-history simulations.
(2) In
average-history stochastic simulations, random disturbances similar in pattern
and size to those encountered during the nine-year sample period, 1971.01
through 1979.12, are incorporated. 5 /
The purpose of average-history simula-
!!J In all policy simulations, the federal funds rate was subject to a floor
of two percentage points and a ceiling of forty percentage points to prevent
simulation of events that are far removed from the sample experience. In
simulations in which a reserve aggregate is the policy instrument, if the
federal funds rate hit a target range boundary on the monthly change or a
floor-ceiling boundary on the level, the federal funds rate became the
effective policy instrument for the planning stage and/or the execution
stage of that month.
5/ The random disturbances of the stochastic simulations reproduce the cross
correlations of the historical monthly forecast errors of the model, both
over time and across equations in a given month. The standard deviation of
the monthly forecast error of demand deposits was increased by about 8% to
account for ex post information on recent shifts in the money demand function
that is incorporated in the current version of the model. This information
is introduced into the model by shift parameters, which include rough approximations for the impact of repurchase agreements, the appearance of ATS accounts,
and so on.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 16 -
tions is to examine the robustness of the response of alternative operating
procedures to a full spectrum of plausible random disturbance patterns.
Thus, the planning stage of a monthly policy operating procedure
is characterized by two components:
(1) selection of the interim target
for the money stock, Mlt, as determined by the rate of monthly reentry, A,
to the annual target path, MTt; and (2) selection of the policy instrument
nonborrowed reserves (NBR), total reserves (TR), or the federal funds rate
(RFF) that will be held invariant to incoming random disturbances throughout
the subsequent month.
The execution stage of monthly policy is represented
by a stochastic simulation in which the effect of the planned policy is eval
uated by a monthly model simulation having nonzero random disturbances.
The policy cycle consisting of the following:
1.
selection of the monthly interim money stock target,
2.
selection of the monthly policy instrument setting, and
3.
execution of monthly policy under random disturbances
is repeated in each month of the effective policy horizon, 1980.01 to 1980.10.
3.
A Comparison of Pseudo-History with Average-History Performance
of Current Operating Procedures
The annual performance of an operating policy procedure depends to
a great extent on the type of random disturbances encountered.
As noted in
appendix A, some policies are more vulnerable to shocks to the demand for
money while others are more affected by supply side shocks.
It is of inter-
est to determine whether the intrayear deviations of the actual money stock,
M, from the annual target path, MT, were due to some inherent flaw in the
current operating procedure or whether the random disturbances encountered
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 17 -
in the first 10 months of 1980 were atypical (incorporating the impacts of
unusual events such as the imP,osition of special credit restraints in midMarch).
Three policy "histories" for M-lA are presented in figure 3.
first is actual M-lA, denoted by M.
The
This can be obtained by simulating the
monthly model over the first 10 months of 1980 with historical disturbances
and with monthly nonborrowed reserves, NBR, maintained on its historical path.
The two remaining "pseudo-historical" paths, MSa and MSb, are also obtained
by simulations with historical disturbances but the nonborrowed reserves
paths, NBRa and NBRb, of these simulations are obtained by average approximations of current operating procedures.
In both cases, the constant monthly
rate of reentry, A, to the target money stock path was set at the typical
value estimated for historical planned policy in 1980, A= 0.292.
Also, both
policies were subject to the restriction that the monthly change in the
federal funds rate could not exceed 300 basis points.
Since both the monthly
rate of reentry, A, and the federal funds rate target range are only approximate characterizations of historical policy procedures, the simulated results
will recover only approximations of the consequences of actual policy.
In the first approximation of historical procedures, labeled "NBR
policy (restrict L\RFF)," the simulated policy authority selects that level
of nonborrowed reserves, NBR, that will attain the interim money target, MI,
in the absence of random disturbances but subject to a monthly target range
of six percentage points on the federal funds rate, RFF.
Under historical
1980 disturbances, this approximation of policy, denoted NBRa, produces the
money stock MSa•
As shown in figure 3, this policy closely mimics movements
in the historical money stock, M, in the first six months of 1980, although
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 18 Figure 3.
Historical and Expected Performance
of Current Procedures
M-lA
r-.---,--,--,---,--,-r---.----,-~--r--~
BILLIONS
HIGH
395
390
M
s
·~
b/
//
I
/__
~
385
-MT
_.,,.-~LOW
,..,,,,.,,.,,,-- ✓•
,,/✓ ;;
380
-
~:,,.,-- - -
~
370
80.03
MONTH
-----
8001
8002
8003
8004
8005
8006
8007
8008
8009
8010
80.06
LOW
M
370.759
371.955
373.113
374.615
376.332
377.625
379.268
.380.604
382.500
373.684
373.118
307.585*
307.782*
371.292*
373.670*
379.713
383.668
:;:86. 658
--------------369.793
370.810
80.09
MSA
MSB
369.912
372.42'5
371.992
367. 844 +369.364*
372. 137*
371.382*
374.865*
377.604*
382.138*
370.569
373.~81
374.283
371.094*
373.689*
377.,093
377.587*
381.944
384.530
386.341
LOW (+-+) - lower boundary of 70% confidence interval
M
MS
(-) (----) -
a
MSb (---) -
MT
E--) -
HIGH (+-+) -
*-
MT
HIGH
-------- -------- -------- -------370.837
372. 274
373.716
375.164
376.618
378.077
379.542
381.013
382.489
383.971
372.845
375.573
377.457
379.495
380.474
382.530
384.291
386.418
388.645
390.738
for NBR policy (restrict AR.FE)
historical M-lA
simulated "history" - NBR policy (restrict &RFF) with historical disturbances
simulated "history" - NBR policy (BOR = BOR_ ) with historical disturbances
1
target M-lA (4.75% annual growth 79 11-80 11)
upper boundary of 70% confidence interval for NBR policy (restrict AR.FF)
outside 70% confidence interval
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
A
- 19 -
it returns more slowly to the annual target path, MT, in the remaining four
months. 6 /
The second approximation to current operating procedures is similar
to NBRa, but the simulated policy authority first selects a total reserves
estimate, TRt, consistent with the interim money stock target.
In this policy,
the nonborrowed reserves path, NBRb, is obtained by subtracting the current
"pseudo-history" of borrowed reserves, BOR_1 .
?J
In other words, the assump-
tion of a continuation of current borrowings by the policy authority leads to
the selection of the nonborrowed instrument setting, NBRb
=
TR - BOR-1•
This
policy approximation, NBRb, is identified as "NBR policy (BOR = BOR-1 )" and
is also subject to the restriction that the monthly target range for RFF
cannot exceed six percentage points.
The money stock attained by this policy
under 1980 disturbances is labeled MSb•
As noted in figure 3, the 10-month
growth of the money stock of this policy approximation is quite close to
actual history except during April 1980 when the decrease is not so pronounced.~
The results in figure 3 suggest that the two nonborrowed reserves
policies are reasonable approximations of current operating procedures since
6/ The return to target path is inhibited by the restriction that the monthly
RFF change must not exceed three percentage points, whereas in actual history
the federal funds rate dropped by 6.63 percentage points in May 1980.
7/ "The amount of nonborrowed reserves -- that is total reserves less member
bank borrowing -- is obtained by initially assuming a level of borrowing near
that prevailing in the most recent period." p. 82, "The New Federal Reserve
Technical Procedures for Controlling Money," attachment to Chairman Volcker's
testimony before the Joint Economic Committee on The 1980 Economic Report of
the President, February 1, 1980. See also related discussion in M.
Hadjimichalakis, "Precision of Monetary Control and Volatility of Rates: A
Comparative Analysis of the Reserves and the Federal Funds Operating Targets,"
working paper (Board of Governors of the Federal Reserve System, December 1980).
8/ This policy tends to produce more modest changes in the federal funds
rate since the simulated policy authority does not implicitly recognize the
projected offset in borrowed reserves when selecting the planned change in
the supply of nonborrowed reserves.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 20 -
the resultant money stocks, MSa and MSb, bracket the result of historical
operating procedures, M.
To estimate the range of money stock outcomes that might have
occurred under current operating procedures had 1980 been an "average" ye_~r,
a 70% confidence interval was generated for the first approximation of current NBR policy (restrict ~RFF).
'!.J'
The 70% confidence interval is
obtained from 100 stochastic simulations of the first 10 months of 1980.
Random disturbances for each simulation differ but ~re selected to replicate
the historical pattern of the forecast errors encountered by the monthly
model over the nine-year sample, 1971.01 - 1979.12.
After 100 money ~tock
paths are generated by the average approximation of current operating procedures, the upper 15 and lower 15 of the simulated money stock paths a~e
removed to define the boundaries of the 70% confidence interval shown in
figure 3.
As indicated, both the actual money stock, M, and the "pseudo-
historical" money stocks, MS 8 and MSb, obtained using 1980 historical
disturbances fall below the 70% confidence interval in at least 3 of the
first 10 months of 1980.
Under the assumption that the relative accuracy of the model
description of money market behavior is not substantially affected by the
shift in operating procedures, this result suggests that the odds are at
9/ Relative to the annual target path, MT, there is a discernible "upside
risk" implied by the effective midline of the 70% confidence interval.
This is due, in part, to the logarithmic formulation of money demand in the
monthly model. To illustrate, if the logarithmic forecast error is normally
distributed
lnM - inM
= E,
E ~
N(0, cr2),
the mean of the simulated forecasts, E(M), will exceed the certainty-equivalent
(zero residual) forecast, M.
2
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
A
E (M) = M ecr /2 •
- 21 -
least two to one that a portion of the gyrations of the money stock observed
during 1980 can be ascribed to unusually severe disturbances encountered in
1gso and not to instability generated by current operating procedures.
4.
Expected Trade-offs in the Volatility of Monthly Money Stock Target
Gaps and of Changes in the Federal Funds Rate Under Alternative
Operating Procedures
This section examines the volatility implications of varying the
monthly rate of reentry, A, to the money stock target path, MT.
One mea-
sure of the volatility of the federal funds rate is the standard deviation
of the monthly change, cr (~RFF).
Under general conditions, this statistic
can be interpreted as one-half the width of the 70% confidence interval for
month-to-month variations in RFF.
One measure of the volatility of money stock performance is the
standard deviation of monthly deviations of the annualized cumulative growth
rate of the money stock from the annualized cumulative growth rate of the
money stock target, where the latter is 4.75% for every month in the policy
horizon.
The precise measure used is the square root of squared deviations
from 4.75% or the root mean squared error, RMSE.
The RMSE also penalizes
persistent "biases" in performance when the money stock consistently grows
below or above the target path as in the case of base drift.
In all cases
reported below, the mean bias is negligible (since the average random disturbance is zero) so the RMSE corresponds closely to the standard deviation
and is approximately equal to one-half the range of the 70% confidence
interval for monthly departures from 4.75% growth.
It can be demonstrated that this volatility measure of monthly
cumulative growth rates around 4.75% is equivalent to the RMSE of the
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 22 logarithm of the annualized monthly target gap, RMSE (GAP), wheu the annual-ized target gap for month tis
X
100.
Thus, in the results discussed below, if the root mean squared error of the
money stock target gap is two, RMSE(GAP) = 2.0, this indicates that the
annualized cumulative growth rate of the money stock in a given month will
fall between 6.75% (4.75 + 2.0) and 2.75% (4.75 - 2.0) with approximately
70% probability. 10/
The results that follow tabulate the expected trade-off between
target-gap volatility of the money stock and the volatility of the federal
funds rate.
Points on the volatility trade-off "frontier" are generated by
altering the monthly rate of reentry, A, to the money stock target path.
As the monthly rate of reentry moves from zero (base drift) to unity (full
gap closure), it is of interest to determine if the frontier is "unstable"
(positively sloped) or well-behaved (negatively sloped).
In the case of the
former, the volatility of both the money stock target gap and the federal
funds rate would increase with the speed of monthly reentry, suggesting that
a viable trade•Qff does not exist.
In the latter case, an increase in the
volatility of the federal funds rate can be exchanged for a reduction in the
volatility of the monthly target gap.
The volatility frontier is estimated for a particular policy by
average-history
simulations of the monthly econometric model.
As discussed
10/ The simulation results do not include an estimate of "noise" introduced
by preliminary seasonal adjustment. An examination of recent work by
D. Pierce suggests that estimates of the target-gap volatility presented
later underestimate total money stock volatility by about 1 percentage point;
see D. Pierce, "Data Revisions with Moving Average Seasonal Adjustment Procedures," Journal of Econometrics, vol. 14 (September 1980), pp. 95-114.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 23 -
earlier, the pattern of ra~dom disturbances of average-history stochastic
simulations resembles the fattern of historical forecast errors generated by
the monthly model over the sample span, 1971.01 - 1979.12.
The following four operating procedures are examined:
1.
NBR.
Nonborrowed reserves are selected as the policy
instrument that is held invariant to random disturbances
during the policy execution stage of each month.
2.
NBR (restrict RFF).
The nonborrowed reserves policy
is subject to a federal funds target range of six percentage points.
That is, the monthly change in the
federal funds rate cannot exceed 300 basis points in
either the policy planning stage, the policy execution
stage, or both.
3.
RFF.
The federal funds rate is selected as the policy
instrument that is held invariant to random disturbances
in the policy execution stage of each month.
Thus, all
monthly changes in the federal funds rate under this policy
are planned changes selected in the planning stage to
return the money stock to its interim target level.
4.
TR.
Total reserves is selected as the policy instrument
that is held invariant to the impact of random disturbances
in the policy execution stage of each month.
Trade-offs in the volatility of the federal funds rate and tpe
volatility of the money stock target gaps under the four alternative operating procedures are displayed in figures 4-6.
The horizontal axis indicates
interest rate volatility as represented by J:he standard deviation of ~onthly
.. \J:.,
changes in the funds rate, a(ARFF), measurtf--in percentage points.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
~~
;I•
·The
- 24 -
Figure 4.
Reentry Trade-off Schedules under
Nonborrowed Reserves Policies
RMSE(GAP~)
,3
NBR Policy (restrict ARFF)
NBR Policy
2
1
a(MU'Ft)
0
3
1
9
6
/
>..=.292 ~
>..=0 --+-I + A=l
/
/
-
~+
--- t
>..=l
>..=.292
I
t
>..=0
2
NBR Policy (restrict ARFF)
NBR Policy
3
RMSE(GAP BO. 10)
a(~RFFt) - Standard deviation of monthly change in RFF
RMSE
- Root mean square error
- Log of annualized relative gap between target money and actual money
in month t
GAP 80 _10 - Log of annualized gap in last month of horizon (80.10)
A
- Monthly rate of reentry to the money stock annual target path
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 25 vertical axis provides two mea~ures of money stock target gap volatility.
The upper panel (going up from the horizontal axis) indicates the monthly
volatility in money stock target gaps for all months in a sample of 10-month
policy horizons.
Thus, the upper panel measures month-to-month variability
of the money stock target gap within a 10-month "year."
The lower panel
(going down from the horizontal axis) measures the volatility of the terminal
gap at the end of a sample of 10-month "years."
It is assumed that the
policy authority wishes to reduce all measures of volatility --
money stock
target gap volatility within a policy year, RMSE (GAPt); terminal target gap
volatility, RMSE(GAPso.10); and funds rate volatility, cr(~RFFt)•
However,
the results indicate that this is not possible.
The unrestricted nonborrowed reserves policy, NBR policy, is displayed on the right side of figure 4. 11 /
For the case of base drift (A= O),
the upper panel indicates that a target gap volatility of about 2.26% is
obtained by the NBR policy at the cost of a monthly funds rate volatility of
about 5.9 percentage points.
As the monthly reentry rate,A, moves toward
unity, the volatility of the monthly money stock target gap is reduced at
the cost of an increase in monthly funds rate volatility.
Thus, the trade-off
moves in a southeasterly direction as the reentry rate, A, increases.
At
A= 1, the unrestricted NBR policy obtains a 60% reduction in monthly money
stock target gap volatility at the cost of a 75% increase in the monthly
volatility of the federal funds rate, RFF. g/
11/ As noted earlier, this policy is not subject to a target range restriction on monthly changes in the federal funds rate.
12/ This may be a relatively optimistic trade-off since the short-run and longrun interest rate elasticities of the demand for money in the staff monthly
model were the highest among models examined by the authors. A procedure for
selecting the monthly rate of reentry that minimizes undesirable consequences
of money market volatility is explored in P. von zur Muehlen and P. Tinsley,
"A Measure of the Cost of Money Market Volatility Associated with Money Stock
Targeting," working paper (Board of Governors of the Federal Reserve System,
December 1980).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 26 -
As the reentry rate, 11., goes to unity, the monthly money stock
target gap volatility does not go to zero.
This is because the r,eentry r,ate
is establi•shed in, the planning stage without benefit of ~erfect f,oresight of
the random disturbances that will, be encountered duri.)lg the subsequent month,.
Thus, ~.3% is the lowest monthly gap in volatility that can be achieved qy
the unrestricted, NBR policy,, at the cost of a monthly funds ra~e volat~Lity
of 10.2 P,ercentage pointso
Note aJ.,so tha,t not much, :ls gained, in, terms, o~ a,
reduction in monthly money stock volati1ity by moving from, 11.
estimated historical rate of reentry) to 11.,
= 0.292
(the
= 1 (plann,ed comP,lete closur~ of
the money stock target gap in one month).
The bottom, panel is roughly a mirror image of the toP. panel ~xcept
that the terminal gap volatility measures for correspond~ng rates of reentry,
'
11., are uniformly lower.
'
This property was found for a:U P,Olicie!!, exam,:f.:n,ed
and indicates that all policies will be more successful in attaining year-overyear targets than in maintaining close adherence to the target path within
the year.
The left side of the panels fn figure 4 indicate the expected volatility trade-off for a nonborrowed reserves policy that fs subj,ect; to a range
restriction of 6 percentage points on monthly changes in the federal funds
rate, RFF.
This policy is a closer approximation to current operati~g proce-
dures than the unrestricted NBR policy.
The results in figure 4 suggest that
variation in the reentry rate under current prodedures is largely futile since
performance is dominated by the imposition of the target range restriction on
monthly variation of the federal funds rate.
Figure 5 indicates the expected trade~offs for an unrestricted
interest rate policy when the federal funds rate, RFF, is held constant
during the month rather than nonborrowed reserves, NBR.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Of course, the
- 27 Figure 5.
Reentry Trade-off Schedules under
a Federal Funds Rate Policy
RMSE(GAP t)
J
11.=0
+
2
1
0
cr(liR.FF t)
T
6
3
t
I
1
11.=.292
9
t
11.=l
t
11.=0
2
-3
~SE(GAP BO. lO)
cr(~RFFt) - Standard deviation of monthly change in RFF
RMSE
- Root mean square error
- Log of annualized relative gap between target money and actual money
in month t
GAP 80.10 - Log of annualized gap in last month of horizon (80.10)
A
- Monthly rate of reentry to the money stock annual target path
GAPt
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 28 -
federal funds rate is reset at the beginning of each month to obtain
the interim money stock obj ective.
1
One characteristic of the unrestricted interest rate policy, RFF,
is that comparable reductions in both monthly and terminal gap volatility
are obtained at lower levels of interest rate volatility.
That is, at the
historical rate of planned reentry, A= 0.292, the monthly money stock gap
volatility under an unrestricted' nonborrowed reserves policy, NBR, is 1.6,
a result that is close to the 1.5 obtained for the unrestricted interest rate
policy, RFF.
Slmilarly, the terminal gap volatility at X,
the unrestricted NBR policy and 0'.58 for the RFF policy.
=
0'.292 is 0.56 for,
However~ the corr,
ponding monthly REF volatility measure at~= 0.292 is 8.0 percentage points
for the NBR policy and only 4.8 percentage points for the RFF policy. l3/
Simulation experiments with total reserves, TR, as a policy instrument were not encouraging.
As indicated in figure 6, both monthly and terminal
money stock target gap volatility remain large and the trade-off that exists is,
associated with extremely large measures of monthly volatility in the federal
funds rate, RFF.
As indicated in appendix A, this may occur if the projections
of required and excess reserves are inaccurate.
Two concluqions may be drawn
from this result regarding the use of total reserves as a policy instrument.
If one believes that projections of the total reserves money stock multiplier
provided by the monthly econometric model are inferior to those that can be
obtained by other models (judgmental or econometric), one may reject the
results in figure 6.
Alternatively, if the monthly model projections of the
total reserves money stock multiplier are representative of projections under
13/ Since, as noted in appendix A, planned settings of the funds rate are
identical under all policies, the results in figure 5 could be obtained
approximately by a nonborrowed reserves policy with a relatively wide target
range on planned changes in the funds rate and tight restrictions on unplanned
changes in the funds rate.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 29 -
Figure 6.
Reentry Trade-off Sohedules under
ar Total Reserves Policy
RMSE(GAP t)
(J)
25
26
11.=0
o{6RFFt) - Standard deviation of monthly change in RFF.
RMSE
- Root mean square error.
GAPt
- Log of annualized relative gap between target money and actual money in month t.
GAP 80 _10 - Log of annualized gap in last month of horizon (80.10).
'
A
- Monthly rate of reentry ~o
the money stock annual target path.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
current institutional arrangements, additional modifications in policy procedures (such as the design of more predictable reserve requirements on current
deposits) might be considered befor~ evaluating procedures involving the use
of total reserves as a policy instrument.
5.
The Issue of Interest Rate Instability
In contrast to the case of short-run volatility examined in preced-
ing sections by which money markets may be subJected to frequent and sizable
transient disturbances, the possibility exists that an attempt to exert
close control over the money stock may induce undamped cycles in short-term
interest rates.
This condition, termed interest rate instability, is not
related to the pattern of unforeseen disturbances but, rather, to the nature
of lagged interest rate effects on the demand for money.
Suppose, for example, that lagged impacts of the federal funds rate
on the demand for money are more powerful than the contemporaneous impact.
In
this case, ever-larger changes in the funds rate might be required to offset
the current impacts of previous settings of the federal funds rate.
Examination of the staff monthly money market model suggests that
even very tight month-to-month control of the money stock, M-lA, would not
produce interest rate instability.
(Details of this exploration of the
dynamic structure of the monthly econometric model are discussed in appendix
C.)
However, this conclusion must be tempered With several qualifications.
First, alternative specifications of the distributed-lag impacts of interest
rates on money demand may yield different stability conclusions. 14 /
Second,
appendix C examines only the direct impacts of interest rates on money demand
14/ Indeed, J. Ciccolo found evidence of
estimated prior to the mid-1970s ~hift in
"Is Short-Run Monetary Control Feasible?"
Policy (Federal Reserve Bank of New York,
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
instability in several models
money demand. See J. Ciccolo,
Monetary Aggregates and Monetary
1974), pp. 82-91.
- 31 -
and does not explore the impact of indirect transmission channels such as the
interest rate elasticity of capital investment. 15/
Third, the dynamic
response structure of money delnand may alter as the public reacts to perceived changes in the operating procedures of monetary policy.
Since no evidence of interest rate instability was uncovered in
this study, it would appear that interest rate instability is not a maJor
barrier to money stock targetipg procedures.
A number of studies in the
early 1970s explored the general problem of policy instrument instability. 16 /
One result of this literature suggests that if instability seems to exist,
the difficulty may lie in the ~esign of the policy strategy.
That is, for
any model with an arbitrary lag structure, it is possible to concoct a policy
rule that may yield (unstable) cycles in the target variable, the policy
instrument, or both.
This does not imply that instrument instability will
exist for a policy that recognizes the dynamic response structure of the
model in question.
In many instances, the difficulty may lie in the selec-
tion of an inappropriate indicator of policy performance or the adoption of
an inflexible policy rule that disregards available measurements.
Two additional factors may also reduce the likelihood of interest
rate instability.
First, there is mounting evidence that the structure of
the economy may be more adequately represented by stochastic coefficient
15/ For discussion of model simulation experiments incorporating both direct
and indirect impacts of interest rates on money demand, see J. Enzler and
L. Johnson, "Cycles Resulting from Money Stock Targeting," working paper
(Board of Governors of the Federal Reserve System, December 1980).
16/ A partial list includes: R.S. Holbrook, "Optimal Economic Policy and
the Problem of Instrument Instability," American Economic Review, vol. 62
(March 1972), pp. 57-65; G.C. Chow, "Problems of Economic Policy from the
Viewpoint of Optimal Control," American Economic Review, vol. 63 (December
1973), pp. 825-37; M. Aoki, Optimal Control and System Theory in Dynamic
Economic Analysis (Amsterdam: North-Holland, 1976); and especially S.J.
Turnovsky, "The Stability Properties of Optimal Economic Policies,"
American Economic Review, vol. 64 (March 1974), pp. 136-48.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 32 -
models than by fixed coefficient models,
0
a result that may rationalize
cautious market interventions by both public and private agents. l8/ - Second,
it is likely that normal arbitrage in money markets would tend to dampen or
eliminate predictable cycles in interest rates.
6.
A Note of Caution
All policy analysis is model specific and the results of this paper
are not exempt from that dictum.
Two limitations of the existing monthly
econometric model may be noted.
1.
As indicated in appendix A, the economic structure and
scope of the current monthly econometric model are limited.
It would be desirable to incorporate a full spectrum of
portfolio adjustments by bank and nonbank sectors as well
as interactions between real and financial economic activity. 1 9/
Efforts in these directions are impeded by the limited scope
of available monthly data.
2.
Stochastic simulations provide a more robust method of
analysis than deterministi,c simulations since they account
17 / See recen,t empiriLcal evidence ,on annual and ,quart.erly models with
stochastic structures in iP.A.V.13. Swamy and P.A.. Tinsley., "Linear Prediction
and "Estimation Methods for Regression Models with Stationary Stochastic
Coefficients," Journal ,of Econometrics, vol. 12 (February 1980 ), pp. 103-42;
and P. Tinsl ey., J. "Berry,, G. Fries, B. 'Garre,tt, A. Norman, P.A.V.B. Swamy,
and P. von zur Muehl•en, "The Impact of U.nc,eit'tainty on the Feasibility of
Humphrey-Hawkins Objectives.,"' Journal of Finance (1980 pr,oceedings of the
American Financial Association, forthcoming).
1
18/ That is, aggressive interven'tion in a stoch'.a,stic coefficients model may
increase the unpredictabili,ty of ,the res:ponse ,to the intervention.
19/ The r~chness of analysis that can be obtained by examination of a complete •capi•tal account -model is demonstrated in the theoretical analysis of
M.G. Hadjim:khail..akis, ",Precision of Monetary Control and Volatility of Rates:
A Comparative Analysis of the Reserves and the Federal Funds Operating Targets," working paper '(Board ,of Governors ,of the ,Federal Reserve System,
December 1980).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
- 33 -
for the histori~al forecast error record of the model
employed.
Although the boundary limits of uncertainty are
delineated by tpis technique, all uncertainty is allocated
to additive external "surprises."
Evidence is accumulating
that the essential structure of the economy is better
described by allocations of forecast uncertainty over all
model coefficients, in contrast to the conventional assumption that the model structure is fixed over time. 20/
Although progress in this area of inquiry is slow and tedious,
it is strongly suspected that the existence of stochastic
policy multipliers requires prudent policy interventions if
the aim of policy is to reduce, rather than increase, volatility indices of performance.
20/
See references cited in note 17.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
APPENDIX A:
Planned and Unplanned Changes in the Money Stock_ (M) ape! the
Federal Funds Rate (RFF) Under Alternative Operating_Procedures */
Table A.l presents the essential structure of the staff monthly
econometric model used in stochastic simulations described in this paper. JI
The structure is sufficiently simple that most deviations in the patterns
of planned and unplanned changes in the money stock and short-term interest
rates due to alterations in operating procedures can be interpreted by direct
inspection of the model as shown below.
The model structure
This model is a characterization of short-run behavior.
Changes
in variables (denoted by~) refer to changes induced by altered settings
of the policy instrument and the impacts of random disturbances.
The predic-
tions of all excluded variables (such as GNP and price inflation) are presumed
invariant to short-run changes in the policy instruments.
Prediction errors
of excluded variables are contained in the relevant random disturbances.
For
example, if GNP is overpredicted, the money demand disturbance (ao) will
include a negative component.
As shown in table A.l, the skeletal model consists of seven equations.
The first equation indicates that the demand for money is inversely
related to the federal funds rate.
effective supply of money.
The next five equations comprise the
The second equation defines required reserves.
This equation contains a random disturbance (bo) representing errors in
projecting the change in required reserves that is associated with a given
change in the money stock.
*/
This error term includes errors in predicting
Prepared by P. Tinsley.
1/ The complete structure of the FRB staff monthly econometric model is presented
in H.T. Farr, "The Monthly Money Market Model," working paper (Board of Governors
of the Federal Reserve System, revised November 1980).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
A-2
Table A.1
A Skeletal Money Market Model
Equation
= a0
Description
money demand (stochastic)
1.
6M
a 8RFF.
1
2.
8RR = b
3.
8EXR =
4.
8TR
5.
8BOR
= d0 + dl (8RFF - 8RDIS).
borrowed reserves (stochastic)
6.
6NBR
= 8TR
nonborrowed reserves (identity)
-
0
+ b 1liM.
required reserves (stochastic)
c .
excess reserves (stochastic)
0
= 8RR + 8EXR.
total reserves (identity)
- 8BOR.
discount rate (policy rule)
Variable Definitions
1.
M
- money stock
2.
RFF
- Federal funds rate
3.
RR
- required reserves
4.
EXR
- excess reserves
5.
TR
- total reserves
6.
BOR
- borrowed reserves
7.
NBR
- nonborrowed reserves
8.
RDIS - FR discount rate
Coefficient Properties
(i)
slope coefficients
a , b , and d are all positive
1
1
1
(ii)
discount rate reaction rule
e
(iii)
lies between O and 1
1
random disturbances (intercept coefficients)
a0
,
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
b0 , c0
,
and d
0
have zero means and constant variances
A-3
the distribution of the money stock over different types of deposits and
among banks of different sizeso
Since the own rate on excess reserves is
zero, the third equation suggests that net changes in excess reserve holdings
are unplanned.
The fourth equation defines total reserves, and the fifth
indicates that borrowings are positively related to the spread between the
federal funds rate and the cost of borrowing.
Finally, the sixth equation
defines nonborrowed reserves.
The last equation is a characterization of discount rate policy.
The discount rate is pegged at a given level if e1 is zero; alternatively,
the spread between the discount rate and the federal funds rate is maintained
if e1 is unity.
Historical policy lies between these two extremes.
The
historical reaction rule for the discount rate (RDIS) incorporated in the
monthly model suggests that e 1 is about 0.25. -2/
Planning and execution stages
The policy authority may choose one of three variables as its
policy instrument -- total reserves, TR; nonborrowed reserves, NBR; or the
federal funds rate, RFF.
p
denoted by b. •
In table A.2, planned settings of variables are
To illustrate, under a funds rate procedure, the planned
change in RFF is determined by the planned target objective for the money
stock
b.
p
RFF =
p
b. M
-- '
where -a1 is the interest rate coefficient in the demand for money (equation
1 in table A.l).
2/ This is the coefficient for the first-month reaction. As fitted by
von zur Muehlen, the historical reaction function suggests a mean lag in
adjustment of about three and one-half months and full adJustment to an
RFF change in about nine months.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
A-4
-
Table A.2:
Plannea and Unplanned-Consequences of
Alternative Operating Procedures!./
Change in federal funds rate
Change in money stock
planned
planned
Policy
all policies
unplanned
---------------------'
unplanned
-------------------------
RFF policy
TR policy
NBR policy
/J. uRFFnbr
= 6/J. URFF tr
/J.uMnbr
= 6/J.~tr
+ (1-6
, , , ~)tJ.UM rff
d0 6
+r __
bL,
d 6
0
a1b1
where 6
=
albl
a1b1 + d1(l - e1)
1/ Subscripts (rff, tr, nbr) denote policy selection. For example, /J.uMy.ff
is the unplanned change in the money stock under an RFF operating procedure.
The following superscripts denote planned and unplanned changes:'
p
/J. M - planned change in M (before random disturbances)
u
tJ.
M - unplanned change (forecast error)
t:.M
-
total observed change (/J.M =~PM+ /J.uM).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
'f16te tha't~, in "the plarinin~ s'tage, 'the pl-anned cliahge ih _;-tne mortey
,.:,",stci'ck '(8i>M,) so~glii
po'1.lcy 'in's'trtlment.
!ie'h1n:g
all
~y 'th~
'pol'icy,authori-t'y can b~ viewed a~ tbe effective
In 'this stage, 'the model is e,as't in-to 'a forech~'t moae 'by
rand'om., di'stiirbances equai to zero '(ao == bo
co
:£::
.
zero
ls
the ;,b'es't'· 1forec1ist
•of
each pr'edic'tlon ·ertor.
= do :!:: ,o) since
,Glven pianned 'money
,s\
change
t8P,M), thi 'c;even-·equaftons
of 'tffe mo'del ar~ then so,lved to 1give the
p1anned settrngs of the remaining s~ven varlab1es.
Since ~here1fuust b~ only
·one 'solution or 'the ,l'ine'af model 'for a given moftey sS'tock 1target,, 't'he •pYanned
"cnangi!s 'of ali vaf:la'bles mu'st 'be iden'ticai •uhd'et any 'operatfilg ~procedure·.
'(Thi<'s i>roperty i's
-
-expll:c11tly 'in~ri'cate'd on1.y for ·piannecl :RFF se•t'tlngs 'f,n
1
Given the 1ex 8llte ,p'l:anned ~ett:f!'ngs~
7
cbe :et~cutdon fStage
'o"f .;policy
'is <ieffihect 'by aad'lhg ·rtonzero values of =the ran'dbm -ai.•i3'turba:nce·s 'tao, 'ho, co~
Dlst'incefons ·ainohg operat:hig ,p'roc'eclures ·are 'tte'termin'ed 1by 'the 1selectaon
'
of one ,;ari'ible ;(des'fgnat:e<i the j,o'licy 'fnstrum'ent:) 't'ha't ,i-s hela constant or
·ao ).
'invariant 'to 'th·e 'Tand'oin c:li"sturl>ahces ·ehcounterea 'dur':irig the ;pol"icy ·execution
1s"'tage
::· \ '
(liowever shbr-t in "aura"t:lon).
HoTa:Hig 1on"e evar4tab1.e 'c'ons'taht "'fo-tces ,the
lmpac't. of ,"th'e 'random d:f.sturbance's 'onto the i-em~rn'tn'g ,seven ·varlables 'i(tha't
h
now 'include :tne unplanne& change "ih ~the money ·s·tock ~ M).
'pa'ttern of uhp1-ann~'d ·changes (denotecl ,by
!).u
Thus, the expec,ted
-in -table -A:2-) -is ,ent:tr~ly de'te·r-
m'ined 'b~y (1·) the se·l:e~tlon of 'the 'poficy 'instrument, ·atfd {2) the dis•t-rl.'butions
or typical historical -patterns cff :fhe ·rani:lom ·dist•urbances.
'JI
The~methodology of the·stochastic simulatibns and that of under'tying classic'al econometrtcs an·d control theory in general
probabifities of 'the random di'stu'rbances •of "the model
i's that ,the
structural Ji-elations
ar·e ·1nvariant to variatio·ns in the selected :po-~icy ih's~rument's.
ln ·other
3/ Tlie purpose of ·stochastic ·simulations ls to i~o'Tate -and ·quant'ify ·the rol:e
of the selection of tne .:pol'icy instrument' ,by :.using 'r"ahdom dl1st'ur.bance
patterns simi~ar to those observed 'in recent 'bfstory.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
A-6
words, in the case of n variables subject tom structural "laws," the distributions of the random disturbances are invariant to the motion of the
n-m "instruments."
This does not imply necessarily that the instruments
are statistically independent of the realizations of the disturbances as
would not be the case for feedpack policies.
For models with additive random disturbances, it may be argued that
the stochastic volatility is merely allocated by policy since the impact of a
random disturbance may be partially or fully absorbed by an instrument without
diminishing or magnifying the additive disturbance impact.!±../
This would not
be true for models with stochastic coefficient structures where stochastic
disturbances interact multiplicatively with the instrument settings.
Alternative selections of policy instruments that are held invariant between policy intervention dateq influence the ultimate destinations
of random disturbances.
This alteration in the allocation of volatility is
a principal reason that apparent correlations between target variables and
potential instruments (caused by the impacts of common disturbances) seem to
break down when the potential instruments are, in fact, employed as policy
instruments.
This phenomenon, well-known in control theory, may be inter-
preted as the raison d'etre of Goodhart's law:
"Any statistical regularity
will tend to collapse once pressure is placed upon it for control purposes." 51
Some unplanned consequences of alternative policy procedures
It is useful to sketch some of the maJor differences in unplanned
consequences for the money stock and the federal funds rate under alternative
policies.
4/ The allocation of uncertainty by alternative feedback strategies and
the dramatic alterations in projected confidence regions that may result
are discussed and illustrated in P. Tinsley and P. von zur Muehlen, "A
Maximum Probability Approach to Short-Run Policy," Journal of Econometrics,
vol. 15 (January 1981), pp. 31-48.
5/ As cited in Albert M. Wojnilower, "The Central Role of Credit Crunches in
Recent Financial History," Brookings Papers on Economic Activity, 1980:2, p. 324.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
A-7
RFF.
1.
as poJicy instrument.
There is no unplanned change in RFF (tRFF = /J.PRFF) and unplanned
u
changes in the mbhey stock (/J. M) are determinea only by the riarrdofu disturbance of the money demand schedule (ao).
2.
TR_as policy instrument.
In counterpoint to the RFF policy, the unplann~d change in the
money stock under a total reserves policy (TR) is wholly detefminea by two
"supply side" shocks -.:. bo, the forecas't error of requif'e'd t-es'etve's; arrd 'c'cr,
the forecast error of excess reserves.
The results of stochastic sifuulatibns
preseh'tea in the paper suggest that 'the performance of 'a tot'al r'es'erve's pbl'i'cy
l's apparently sensitive to the accuracy of the required reserves foreca'st'••
As indic~i'ted
ih table
2, the unplanned change in RFF is a function of b'oth
demand ( 8.Q) and supply ( bo, co) shocks and inversely tel:ated to the interest
rate coefficient of money demand ( a 1 ) and re·serve requirements (bi).
Uii.p"ianned
changes ih neither money nor RFF are affected by forecast errors of borrowings
(BOR) or the discount ·ra'te policy (e1).
In severai respects,, a nonborrowed reserves policy,, NBR, may 'be
interpreted as
a
hybrid pollcy, mixing elements of both
RFF
and
TR
pollcres.
The unplahned c'hange 'ln the tnoney s'tock,, for example, 'is •a wei•gbt~ed av·era:ge
•o'f 'the 'unplanned changes that wduld 'be bbs~ryed under the c·ompeting poli'c:1.es.
That is, the we'ights on unplanned money stoc'k changes under an RFF policy,,
u
1
,
_,
.
'b,. 'H'rff, and unplannea money stock changes under a TR policy,
u
/J. Mfr, 'sum 'to
unl'ty and are fractional 'for fractl:ona!L e:i_, the dl.scoun't rate reaction ·coef-
1
ffcient.
Thus, 1under a nonborrowed reservifs pol:1.cy,, d:lsc'oun't rate policy is
an :l.mportant deterininan't of the relative impacts of demand and supply sid'e
shocks.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
If larger demand shocks are expected judgmentally in the near term,
A-8
e1 might be raised closer to unity; conversely, if difficulties are expected
in forecasting required reserves, the response of the discount rate might be
muted (moving e1 toward zero).
NBR is the only policy under which unplanned changes in both the
money stock and the funds rate, RFF, may be induced by forecast errors in
borrowings (do)•
Indeed, the presence of the borrowing projection error
(do) and a discount reaction (e1) of less than unity are the only elements
that provide a distinction between NBR and TR policies.
That is, if do= 0
and e1 = 1, NBR policy is identical to TR policy since ~BOR, in this
case, woul~ always be zero (see equations 5 and 6 of table A.1).
The results in table A.2 also indicate that unplanned changes in
the funds rate will tend to be smaller under NBR policy than under TR policy
due to the positive slope of the effective total reserve supply schedule
under the NBR policy (e1 less than unity).
Thus, under current assumptions,
dispersion of total changes in the federal funds rate will tend to be smallest
under a funds rate policy, RFF, and largest under a total reserves policy,
TR.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
APPENDIX B:
Constructfon• of the Desired Speed of Reentr~ GA),
to the I'..ong-Run M-1:A 'llarget Path in 1980 *//
r.
Characterization of FOMC In~entions
At each meeting, the FOMC selects short-run targets for severa~
monetary, agg,regates ,, expressed as seasonally adjusted aMerage gi::owth, ra te~4
over a horizon of two or more months.
Staff then trans1ate the EOMG '·s,
desfre~ short-run growth rates for the aggregates into monthly, target le~e]a
of the aggregates1.
At times, this translation reqµ:l!res var:tabl:e month:-t;q.-
month growth rates in o,rder to accommodate anticipated transitory, var:tat1lons,
in money d'emand.
This dd1scuss..ton focuses on the F0MC'·s1 sho,rt-i;-un obj1ect1'.r<>q
for M-1A alone, and! bases1 fts estimates o,f. the typkal intended M-11\ reen,trx,
speed on monthly translations of the FOMC's short-run obJectives.
Specific.ally, the FOMC short-run obj,ect:iive for M-1:A ±s rel?resented,
as a plan to reduce the gap between the long-run annual target path for M-~~
and the level of M-lA in the month following the FOMC meeUng to a g,iv~11i
fraction of the current gap in the month of the FOMC meeting,, as project:_ed
by the staff at the time of the meeting. lp
An algebraic formulation of
this linkag,e between short-run and long-run M-lA objectives of the FOMC is
*/' Prepared by w. Trepeta with research assistance from H. Hayssen,
M. McLaughlin, and A. Reilly.
1/ This characterization does not incorporate FOMC intentions to influence,
subsequent to its meeting, the level of M-lA that is projected fo~ the current month of a meeting. This abstraction seems permissible, given that the
FOMC met on average around mid-month and often later in the ~onth, when little
could be done to alter the average level of M-lA projected for that month.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
where the subscript t denotes the month of the F0MC meeting; t+l the
month following the meeting; ln indicates a natural logarithm; MT is the
long-run annual target path for M-lA; MI is the F0MC's short-run objective
for M-lA as iudgmentally translated by staff; Mt denotes the staff's
judgmental projection of the level of M-lA in the month of the F0MC meeting;
and A denotes the desired speep of reentry to the long-run target path
implied by equation B.1.
The left-hand side of equation B.1 (multiplied by 100) is the
desired percentage gap in the month following the F0MC meeting between the
long-run target path for M-lA (MTt+l) and the level of M-lA (Mt+l)•
Simi-
larly, the second term on the right-hand side (multiplied by 100) is the
percentage gap projected for the month of the meeting.
The desired ratio
of the target gap in month, t+l, relative to the current proJected gap in the
current month, t, is 1 - A•
That is, A, the desired speed of reentry to
the annual target path, is that fraction of the current projected gap that
the F0MC desires to close over the coming month.
If A= 0, no closure of the
target gap is planned; this is equivalent to planned base drift, when it is
desired that M-lA grow in the next month at an annualized rate equal to the
annual target rate of growth.
Alternatively, if A= 1, the intention is to
eliminate fully over the next month the projected gap between M-lA and its
long-run target path.
Note that the desired speed of reentry, A, is only
an ex ante intention that may be frustrated subsequently by forecast errors.
2.
Empirical Estimates
Estimates of the desired speed of reentry, A, were based on F0MC
decisions at nine meetings from February 1980 through November 1980.
Y
From
February through May, it is assumed that the long-run target path for M-lA
2/
The F0MC did not meet in June.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
B-3
corresponded to the midpoint of the announced FOMC target range of 3-I/2 to
6 percent growth for M-lA from 1979 Q4 through 1980 Q4.
This long-run path
of 4-3/4 percent growth is anchored to a base, centered on November 1979, of
$369.7 billion, which was the estimate until June 1980 of the average lev,el
of M-1A in the fourth quarter of 1979.
From July onward, the long-run path
was anchored to a base of $368.1 billion, which, following June benchmarking,,
was the revised estimate of average M-lA: in the fourth quarter of 19,79,.
The estimates assume also that, from July onward, the long-run
target path for M-lA was lowered to 4-1/4 percent growth.
This one-half
percent decrease in the long-run target reflects the fact that, in July, staff
increased by 1/2 percent its estimate of the depressing effect of ATS deposi
growth on M-lA expansion.
This revised estimate of ATS growth implies that
a 1/2 percent downward revision of the long-run target range for growth of
M-lA would be consistent with the original FOMC intentions embodied in the
target range announced in February.
Indeed, from July onward, the FOMC's
short-run target paths for M-lA consistently pointed toward year-end levels
below the midpoint of the target range selected in February.
Given these assumptions, the ordinary least-squares estimate of
the desired speed of reentry to the effective long-run target path of M-lA
is A. = 0.292. ~
This estimate implies that, in 1980, the FOMC did not
3/ Specifically, an ordinary least-squares regression of the dependent variable (ln MTt+l - ln Mlt+l) on the independent variable (ln MTt - ln MP)
yields the following results: a coefficient on the independent variafile of
0.708, having a standard error of 0.072 and at-statistic of 9.821; R2 = 0.9234;
and the standard error of regression= 0.0022.
Alternatively, if it is believed that ordinary least squares place disproportionate emphasis on large target gaps, an arithmetic mean estimate of the
desired speed of reentry is 0.392, slightly higher than the ordinary leastsquares estimate of 0.292. The arithmetic mean estimate suggests an average
"age" of random disturbances in the desired M-lA target gap of about two and
one-half months.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
B-4
wish the average age of random disturbances to M-lA to exceed three and onehalf months. 4/
3.
Concluding Remark
The average desired speed of reentry,A, is a simple charac-
terization of the short-run objectives of the FOMC.
This representation
treats M-lA as the sole intermediate target of policy and specifies that the
desired reentry speed is independent of observed and forecast values of GNP,
the inflation rate, and other variables of potential concern to the FOMC.
Nevertheless, this approximation of FOMC intentions is superior to a number
of more complicated specifications 5/ and may serve as a useful benchmark
for policy discussion.
4/ In months, the average "age" of random disturbances in the long-run target
gap implied by equation B.1 is 1/A. Due to the nature of exponential decay,
complete elimination of the influence of a given disturbance to M-lA implied
by equation B.1 can be a lengthy process. For A= 0.292, seven months are
required to eliminate 90% of a given disturbance. (One month is required for
A= 1 and twenty months for A= 0.111.)
5/ Several tests were conducted to explore the possibility that the planned
reentry speed,A, was systematically related to selected explanatory variables, such as
(1) time remaining in the policy horizon (80.01 - 80.12),
(2) the absolute value of the money stock target gap projected for the current
month, or (3) the signed value of the target gap projected for the current
month (to allow for asymmetric responses to positive and negative deviations
from the target path). Ordinary least-squares regressions of A on a constant
and these potential explanatory variations, both singly and in various combinations, indicate that A was not related to these variables at a 90 percent
level of confidence.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
APPENDiX C':
An Examination of the Interest Rate Instaibil:tty of the
Staff Monthly Money Market Model ~
The :f!ssue of interest rate instability was examined by analyzing
the• .tmp1ic.it stability of the distributed lag impacts of the federal fundts,
rate• on the currency and demand deposit equations of the staff monthll:y money,
Elasticities of the demand for M-lA with respect to, current and
l!agged ffed'eral funds rates were approximated by volume-weighted averag,es o,f
the component ellasti!cit:f!es of M-1:A wfth we,:f:ghts, of 01.3 and O•. 7 appH!edi
currency and d'emand deposit components respectively.
t(!)J
the
Combining the :finflu-
enc•e• o,f a].11 other var:i!ab]:es affecting money demand together w:lfth random d:ll.s,turbances in an error term,, Vt, yielded' the foll'.ow.tng function d'escuibi:fLng;
lllQntfu,]y g17owth :f!n, the demand for M-lA:
Y'
7
l!n Mt
= constant
r
+
a1 ll.lnRFFt-i + Vt ,
i=0
where•
a0, =
a] =
az, =·
a 3, =
-.02'54467',
- . 0:2 53'940,
-.0,240>"131'. ,,
- .. 021:3'02'4,
814 = -.0•17262 7,
a5 = -.O>l'.25832,
8'6, = -.0068288',,,
a = -9·. 842'1' x r o;- 9•,
7
and!
wft~
estfmated to equal 0.69.
*/' Prepare~ by
w.
Trepeta.
]/' The !eft-han~ variable, ~ln Mt, times 1200• percent equals the annualized
monthly growth rate of dem•and1 for M-lA.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
C-2
In equation C.1, M apd RFF denote the levels of M-lA and the federal
funds rate respectively; the subscript t indicates month t;
monthly change; and ln the natural logarithm.
~
indicates a
Equation C.2 indicates that
a disturbance to the growth of money demand in month t, Vt, tends to equal
69 percent of the disturbance in the previous month plus a random component,
Et, with a mean of zero.
The policy authority is assumed to manipulate the funds rate in
order to minimize the expected value of the current month's squared deviations of the monthly growth rate of M-lA from a fixed monthly growth rate
target given knowledge of p, the coefficient of serial correlation between
values of Vin successive months. 2 /
This strategy involves setting
[~ *ln M - (1 - p) constant
7
-
E (ai - pai-l)~ln RFFt-i + pa7~ln RFFt-8
i=l
(C.3)
where ~*In M denotes the fixed target for monthly growth of M-lA.
Intuitively, this equation represents a federal funds rate reaction
function involving a monthly setting designed to offset fully all predictable
deviations of the monthly growth rate of money from target.
With the federal
funds rate setting thus specified, algebraic analysis parallel to Ciccolo's
2/ Alternatively, the analysis could assume that the policy authority varies
the monthly growth rate target in order to return M-lA to a given long-run
target path whenever random disturbances have driven M-lA off this path. In
this case, the monthly growth rate target,~ *ln M, would contain a random
component, and the variance of the federal funds rate would be somewhat
different from that discussed above. However, pursuit of a fixed long-run
target, by itself, would not induce interest rate instability, if instability
does not arise in the case of a fixed monthly growth rate target.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
C-3
reveals that, after a disturbance to money demand, the federal funds rate
converges to a stable value rather than exhibiting ever-larger cycles.
'if
However, for the very tight monetary control procedure assumed, the margin
between stability and instability is extremely small, especially in light
of the standard errors of the coefficient estimates.
This margin is greater,
though, for less rigid control procedures involving an expected return of
the monthly growth rate of money to target in more than one month.
3/ Ciccolo, "Is Short-Run Monetary Control Feasible?" This algebraic
analysis involves examining the Schur determinants of the difference equation
7
a 0 ~ln RFFt +
E (ai - Pai-1) ln~RFFt-i - pa7~ln RFFt = constant,
i=l
to see if all exceed zero, a necessary and sufficient condition for stability.
Use of the Schur theorem in stability analysis is discussed in A. C. Chiang,
Fundamental Methods of Mathematical Economics, 2d ed. (~cGraw-Hill, New York,
1974), pp. 599-600.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
The New Federal Reserve Operating Procedure:
An External Perspective
February 1981
Paper Written for a Federal Reserve
Staff Review of Monetary Control
Procedures
by
Edwin M. Truman and others
February 1981
The New Federal Reserve Operating Procedure:
An External Perspective
CONTENTS
Page
I.
Introduction and Summary
l
II.
Background and Framework
7
III.
Exchange Market Developments
A.
Var1abil1ty of Exchange Rates
20
B.
Responses of Exchange Rates
32
1. Tests for Structural Shift
33
2.
C.
IV.
Variability of Exchange Rates
due to Interest Rates
Exchange Market Intervention
The Foreign Experience under the New Federal
Reserve Operating Procedure
A.
V.
20
39
41
49
49
Introduction
B. U.S. and Foreign Interest Rates
52
C.
Reactions in Major Industrial Countries
57
D.
Reactions in Deyeloping Countries
67
U.S. International Capital Flows
70
':
VI.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Concluding Topics
A.
79
Consequences of Exchange Rate Variability
79
l.
Impact on Trade Flows
79
2.
Impact on Foreign Direct Investment
and Domestic Fixed Investment
81
3.
14.
B.
Impact,on Domestic Rr1£es:
Ratchet,Effects
Impact 10n .Qf-f\i01a1l ,and ,Pr.wate
•Do:rn ar Ho1l dings
The Exchange Rate as ilnforma't<1on 'Variab'le
,and ,Ro)li1q1 Hnstnumen;t
l.
The tExchange ,Ra·te ras '.Tn'formati•on Wa'r~iab'l e
2.
The 1Exchange 'Ra,te as 'Policy Instrumen't
R,eferences
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
'83
91
96
February 1981
The New Federal Reserve Operating Procedure:
An External Perspective*/
Section I -- Introduction and Summary
When the Federal Reserve adopted ,ts new operating procedure
with its greater emphasis on the supply of bank reserves, its decision
was motivated ,n part by the pronounced weakness of the dollar 1n
September 1979.
The adoption of the new procedure followed by 6½ years
the structural shift that occurred in exchange rate arrangements among
maJor currencies in March 1973; the change in procedure has many elements
of s1m1lar1ty with that earlier shift to managed floating exchange rates,
and our analysis concentrates on a comparison of experience since October
1979 with experience between March 1973 and that date.
Section II presents a brief overview of developments since
October 1979 and lays out the framework for our analysis. That analytical framework is based upon the fact that, holding other factors
constant,l/ the link between the new procedure and the variability of
spot and forward exchange rates (which 1s analyzed in Section III)
depends on whether the new procedure has been associated with greater
variability in nominal dollar interest rates (a topic that is investigated
*/ This paper was prepared by the staffs of the Federal Reserve
Bank of New York and the Division of International Finance of the Board of
Governors of the Federal Reserve System and was coordinated by Edwin M. Truman.
The paper benefited from comments on an earlier draft by the staff of the
Federal Reserve Bank of San Francisco.
1/ For the period under consideration this is a particularly
, strong assumption that is, nevertheless, useful.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-2in another paper in this study),l/ on the extent to which interest rates
on assets denominated in other currencies move in line with dollar interest rates (as a consequence, for example, of the policies of other countries), and on whether the new procedure influences the variability of
expectations of U.S. inflation.
It is desirable to distinguish, to the extent possible, between
the effects of the new operating procedure~~ and the possible consequences for the effective stance of monetary policy of the adoption of
the new procedure, e.g., the possibility of a tighter policy on average.
This distinction is especially relevant to the assessment of reactions of
foreign countries in Section IV.
Changes in the pattern of capital flows may
be induced by the new procedure; experience in this area is examined in
Section Vin the context of the balance of payments identity.
Finally, to
the extent that exchange rates have become more variable since October 1979,
it is useful to consider the consequences of such increased variability, the
use of the exchange rate as an information variable under the new procedure,
and the possible scope under the new procedure for using the exchange rate
as a policy instrument.
These issues are discussed in Section VI.
Our principal findings are as follows.
1. The foreign exchange value of the dollar appreciated
immediately following the Federal Reserve's adoption of its new operating
procedure.
It rose sharply in the spring of 1980, fell back i~ the
l/ That paper, "Interest Rate Variability under the New
Operating Procedures and the Initial Response in Financial Markets,"
concludes that there has been an increase in the variability of U.S.
interest rates since October 1979.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-3summer, and rose again more recently.
By the end of November 1980, the
dollar was 5½ percent above its level at the end of September 1979
(Section II).
A portion of this appreciation may be attributable to U.S.
monetary policy that may have been tighter on average over the period than
it otherwise would have been as an indirect consequence of the new procedure.
However, in the absence of an accepted set of counter-factual
assumptions about the performance of the U.S. and world economy and about
U.S. economic policies during the past year, the size of this portion cannot be quantified.
2.
Since October 1979, the variability of international
interest rate differentials over daily, weekly, and monthly intervals has
increased significantly because of the increase 1n the variability of
dollar interest rates, which, in turn, is attributable at least in part
to the new operating procedure (Section II.A).
3.
The increase in the variability of interest rate differ-
entials since October 1979 has contr1buted to a significant increase,
compared with earlier periods, in the day-to-day variability of spot
dollar exchange rates (Section III.A).
4.
The evidence of an increase 1n the month-to-month variability
of spot dollar exchange rates although clear is somewhat less conclusive
The month-to-month variability of interest rate differentials, however, has
increased significantly, and the responsiveness of exchange rates to changes
in such differentials appears not to have changed after October 1979. These
results suggest a decline 1n the variability of determinants of exchange
rates other than interest rate differentials.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Exchange rate variability
-4of course is a function not only of developments in the U.S. economy but
also of economic developments abroad.
This Joint determination is illus-
trated most dramatically in the case of the rise in the variability of
the yen-dollar exchange rate since October 6, 1979
(Sections III.A and
III.B).
5. One-year forward dollar exchange rates for individual
currencies, especially over monthly intervals> have exhibited some cases
of reduced variability since October 1979. This phenomenon has not been
observed for five-year forward rates.
This evidence provides limited
support for the hypothesis that the new procedure could lead to a
reduction of the variability of forward exchange rates (Section Ill.A).
6. We found little evidence of a fundamental change 1n exchange market intervention behavior since October 1979. Our analys1s
did identify a possible shift back toward the pattern of less active
intervention prevailing prior to November 1, 1978 (Section III.C).
This finding suggests that patterns of exchange rate movements have not
been contaminated by changes in intervention behavior; it also suggests
that we cannot read into intervention behavior any evidence of foreign
countries' being sufficiently unhappy with the new Federal Reserve operating procedure to alter such behavior.
/. We found l1ttle·evidence of a significant increase in the
month-to-month variability of foreign interest rates related to the increase
in the variability of dollar interest rates or to the new operating procedure.
Canada is an important exception (Section IV.B).
8. The available anecdotal evidence supports the view that,
at least until recently and aside from Canada, the Federal Reserve 1 s new
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-5-
operating procedure has not resulted in significant deviations from what
policies in the major industrial countries otherwise would have been.
Recently, German policies have been constrained by high dollar interest
rates, which some observers have attributed to the new operating procedure (Section IV.C).
The uncertainty surrounding the wide swings in
dollar interest rates have caused technical policy problems, espectally
for some developing countries (Section IV.D).
However, some of these
apparent problems may reflect unfamiliarity with the implications of the
new procedure during its initial use over the past year.
9. Although gross U.S. international capital flows have been
quite variable during the past year, we have not been able to identify
any significant developments,that can be associated with the new operating
procedure per se. Such flows were influenced importantly by other developments during the past year, e.g., the credit restraint and managed liabilities programs (Section V).
10. Our review of the available literature revealed little
empirical evidence that an increase in exchange rate variability, such as
has occurred since October 1979, has adverse economic and financial effects.
In particular, we found no direct or indirect evidence of a link between
the variability of dollar exchange rates and the level of domestic prices.
In other words, the so-called ratchet effect, which hypothesizes that
fluctuations in exchange rates raise the average inflation rate, does not
appear to be a feature of the U.S. economy (Section IV.A).
11. The adoption of the new operating procedure neither reduced
nor enhanced the role of the exchange rate as one of several financial
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-6variables useful as information variables in carry~ng out monetary policy.
However, it may well "be neither feasible ,nor desirable to adopt the spot
exchange rate as a policy instrument under the new operating procedure.
While attempts to stabilize
spot 'exchange rates through sterilized inter,
vention may be-successful, the variability of forward exchange rates could
well be increased -- with uncertain economic consequences (Section VI.B).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-7-
Section II -- Background and ,F,rameworkl/
An extended sljde of the dollar was arrested-toward the end of
1978 following the November l package,'and the dollar moved higher over
the first half of 1979. Weakness reemerged in early summer and again in
September 1979.
In both episodes, the decline in the value of the dollar
generated heavy net exchange market purchases of dollars by both U.S. and
foreign authorities, averaging almost $1 billion equivalent per week.
In
early October 1979, rumors of a new policy package, followed by tne announcements from the Federal Reserve on October 6, provided substantial
support to what had been, in late September, a ver_y weak dollar. The dollar
rebounded sharply during October, and the rebound was accompanied by substantial net official sales,of dollars at about the same rate as the previous purchases, that is, about $1 billion per week.
The strength in the dollar was not long lived.
Following the
taking of the U.S. hostages and the subsequent freezing of Iran-ian official
assets, as well as the round of substantial increases in the price of oil
in late 1979, the weighted-average exchange value of the dollar declined in
November and December and ended the year below its September trough.
An upward movement of the dollar in early 1980 was fueled by
increases in U.S. interest rates that apparently outweighed the effects
of a deterioration in the outlook for U.S. inflation. A stronger dollar
was also encouraged through the first part of 1980 by increasingly optimistic assessments of the likely U.S. current-account position compared
with the expected positions of other major industrial countries, especially
l; The principal contributors to this section were Peter Isard
and Karen-H. Johnson.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-8-
Japan and Germany. This upward momentum accelerated dramatically in
March, in the wake of the_introduction of the credit :estra1nt program
-
and the further sharp increases in U.S. interest rates relative to those
abroad.
From late January through ear]y April, the dollar's weighted-
average exchange value rose by more than 10 percent.
, When U.S. interest rate$ began their steep decline in April, falling even more rapidly and substantially than they had earlier risen at
a time when foreign interest rates were declining only moderately,,the spot
exchange value of the dollar plunged -- falling by about 9 percent from early
April to the end of May.
While offJcial dollar sales had been very heavy
during the runup of the dollar, the net purchases were relatively light
as the dollar declined in April and May.
The dollar's foreign exchange value continued to decline gradually
from the end of May through mid-July and, subsequently, fluctuated in a
narrow range through mid-October.
Meanwhile, interest rates abroad declined
somewhat ,n response to evidence of slower real economic growth, dollar
interest rates began to rise again, and the United States moved int~ currentaccount surplus.
Dollar interest rates rose significantly after mid-October,
and the dollar's weighted-average foreign exchange value also increased
significantly to a level at the end of November 1980 about 7 percent above
,ts July 1980 low and 5½·percent-abov~ ,ts level"at the end of September 1979.l/ Net intervention sales of dollars, especially by U.S. author1ties, also increased dramattcally.
1/ The dollar rose somewhat further through mid-December as dollar
interest rates continued to rise relative to rates on foreign-currency,denominated assets. Note that the analyses and material cited in
this paper use the end of November 1980 as a common cutoff date. V
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-9-
Chart l provides a perspective on the foreign exchange value of
the,dollar not only during the past two years but also since the beginning of January 1973. The relatively sharp fluctuations on a weekly
aver~ge basis during the past 18 months are quite evident in the Bhart,
but they are not unprecedented since the w1despread adoption of floating
exchange rates in March 1973. Chart 2 presents the same data in the form
of three-month moving averages.
Casual inspection of these two charts
suggests that while the dollar's average value has exhibited quite marked
short-term fluctuations since October 1979, the fluctuations over somewhat
longer periods have been less pronounced.
Finally, by way of introduction,
Chart 3 presents weekly average observations on spot, one-year forward,
and five-year forward bilateral OM-dollar exchange rates during the past
three years. The chart suggests somewhat less variability in the one-year
forward rate than 1n the spot rate dur1ng 1980, although the five-year
forward rate appears to have been no less variable. We will return to
the data presented 1n Chartsl-3 in Section III.
Changes 1n the degree of exchange rate variability since October
1979, of course, may reflect more than the shift in the Federal Reserve's
operating procedure.
Interpretation of the charts and the statistical
data, therefore, will be factlitated by a brief examination of how exchange rates·may be influenced by the Federal Reserve 1 s new operating
procedure.
Hol'ding other factors constant,- the link bewteen the new procedure
and the variability of exchange rates depends on several considerations:
(1) whether the new operat1ng,procedure produces greater variability in
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-10-
Chart l
Weighted Average Foreign Exchange Value of the
U.S. Dollar Against Ten Major Foreign Currencies:
Weekly Average
120
1 10
100
90
1973
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1974
1975
1976
1977
1978
1979,
1980
,
,,_
_
Chart 2
Weighted Average Foreign Exchange Value of the, 1
U.S. Dollar Against Ten MaJor Foreign Currencies:
lhree-Month Moving Average
120
1 10
100
90
80
1973
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1974
1975
1976
1977
1978
1979
1980
.,,.12Chart 3
Oiyl-Oollar Exchange Rates
(Weekly Average)
OM per dollar
-=--,
I
~200
I
~180
I
J60
I
I
I
f
I
L_
0
L..L.J,_ LLl _L_ ~- .LLL...-.......-,___._______..._...____,______
, ...J.----' --L---L..l-J.--
L~ l
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1978
1979
J
.
.
1980
1 40
-13nominal dollar interest rates (which was expected to occur in the short
run especially for short-term rates but need not continue in the long
run especially for longer-term rates); (2) the extent to which interest
rates on assets denominated in foreign currencies move in line with dollar
interest rates; and (3) whether the conduct of monetary policy influences
the variability of U.S. inflation expectations. To the extent that the
new operating procedure produces greater variability in dollar interest
rates, especially longer-term rates, that is not offset by similar movements in foreign interest rates (as a result, for example, of a policy
response by foreign authorities), it will tend to increase the variability
of spot dollar exchange rates.
But if the new operating procedure has
caused, or eventually causes, market participants to expect less relative
variability of U.S. price inflation than under the earlier procedure, the
new procedure will have contributed to greater stability of long-term
forward dollar exchange rates, other things remaining equal, and, thereby,
might actually reduce the variability in spot dollar exchange rates as well.
The analytical structure outlined in the paragraph above rests on
two premises. The first is the presumption that the influence of monetary
policy on exchange rates is transmitted primarily through interest rates
and inflation rates, both actual and expected. The second is the approximation that the difference between the spot and forward exchange rates
the forward discount -- can be equated with the nominal interest rate
differential .l/ Neither premise denies the fact that exchange rates (spot
1/ This is the covered interest rate parity condition, which holds
exactly, Tn the absence of current or prospective capital controls, for
interest rates on securities that are comparable (in terms of maturity, tax
treatment, liquidity~ and default risk), reflecting the fact that competitive
exchange markets bid away any profit that might be earned by coveriDg spot
foreign exchange transactions with equal and opposite forward transactions.
These cond1t1ons are closely met in the Eurocurrency markets.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-14and forward) are also influenced (directly or 1nd1rectly) by factors that
have little to do with the Federal Reserve 1 s operations: for example, by
shifts 1n countr1es 1 relative current-account pos1t1ons as the result of
real phenomena such as 011 discoveries: changes in productivity, or other
changes in competitiveness, by the imposition and removal of credit controls, or by exogenous variations in expectations of high inflation rates.
Under the first premise, monetary policy influences exchange rate
movements over the long run primarily by influencing expected inflation
rates.
Accordingly, since the level of the forward exchange rate reflects,
inter alia, expectat1ons about the future level of the spot rate, revisions
in expectations about inflation rate differentials are the primary monetary
factor contr1but1ng to changes in the forward rate.
When U.S. inflation
expectations change without an accompanying change in nominal dollar interest
rates, spot and forward exchange rates will move by the same amount, other
things equal.
In such cases a change 1n the expected inflation differential
results in an equal change in the real interest rate differential, where the
real interest rate is the difference between the nominal interest rate and
the expected 1nflat1on rate.
At the other extreme, when a revision in U.S.
inflation expectations is accompanied by an equal change in nominal dollar
interest rates, such that no change occurs 1n the differential between U.S.
and foreign real interest rates, the forward rate will still respond to the
change in inflation expectations but the spot rate will remain unchanged.l/
1/ Strictly speaking, interest rates and inflation expectations
have time-horizons or term structures, and the changes in spot and forward
exchange rates that accompany a given change in short-term interest rates
or 1nflat1on expectations depend on how these term structures have shifted.
Even a very large increase 1n the.variability of the overnight federal funds
rate, in parti~ular, would have a small impact on exchange rate variability
if 1t were not accompanied by substantial increases in the variability of
interest rates on monthly or annual maturities, other things equal. Thus,
an assessment of the maturity-dimension of interest rate var1abil1ty is
important 1n cons1der1ng the degree to which the new operating
procedure has influenced the variab1l1ty of exchange rates.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-15Thus, stab1l1ty of the spot exchange rate requires stability of
differentials between real interest rates, other things remaining equal.
Consequently, a monetary policy operating procedure that tends to reduce
the variability in U.S. inflation expectations (and thus the variability
of the forward rate, assuming expectations of inflation abroad are not
negatively correlated with expectations of U.S. 1nflat1on) should contribute to stability of the spot rate -- because the expected inflation
differential is a component of the real interest rate differential.
In
addition, policy adjustments involving sharp movements in real interest
rates may be less likely when inflation expectations are more settled.
Whether or not the new Federal Reserve operating procedure has to date
reduced the relative variabi1ity of expectations of U.S. inflation is
addressed in other contributions to this study.
Without persuasive evi-
dence that expectations have become significantly less variable as a consequence of the
new operating procedure, any change in the variability of
interest rate differentials relative to that of exchange rates during the
past year reflects changes in the variability of the non-monetary factors
that contribute to exchange rate determination.l/ Empirical findings concerning the relative variability of interest rates and exchange rates
since October 1979 are discussed in Sections III.A and III.B.
The volatility of interest rates and exchange ra~es has sometimes
been associated with capital flows.
(See Section V.) The potential for
actual capital flows should be considered in the context of the balance of
• 7/ It should be recognized, however, that real interest rates
can also vary becau~e of fluctuations in money and credit demand ~hat are
not accorrrrnodated by the monetary authority.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-16payments accounting identity, which constrains net private capital flows
to equal, with the sign reversed and including the statistical discrepancy,
the sum of net current-account flows and net official intervention flows.
With no exchange-market intervention and a given current-account position,
a strengthening of net private demand for assets denominated in one currency matched by a weakening of demand for assets denominated in another
currency cannot generate any net change in private capital flows, but
instead results in a change in the exchange rate.
In particular, exchange
rates will adJust to offset the influence of interest rate movements on net
private demands for currencies.
On the other hand, official intervention can resist changes in
exchange rates by purchasing currencies for which net private demand has
weakened and selling currencies for which net private demand has strengthened,
thus permitting net private capital flows to occur.
In this context it 1s
important to distinguish between intervention that is "sterilized" through
open market operations or other procedures that prevent the intervention
from leading to changes in bank reserves 1n either country whose currency
is being bought or sold and intervention that is "unsterilized." Unsterilized intervention is generally more effective than steril1zed intervention
-
in resisting changes in exchange rates following a shift in net private
demands for assets denominated in different currencies, since the changes
in bank reserves and, hence, in money supplies will operate through interest
rate adJustments to moderate or offset the effects of the initial shift in
currency demands.
Sterilized intervention -- particularly when it is visible
to market participants -- may provide a sense of policy commitment that also
influences interest rates or expectations about inflation and may thereby
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-17succeed rn ,moderating ex-change rate van,a1b:i ~ J ty wi"thot1t ,an ,as•soc,,ated
adj1ustment ,of ,moneta,ry growth rates~l/
.Whet.he,r ,or not the ,moderatrnn 0f ,exchange irate ,v,ar;iab1J 1ty should
1
be a po] j(cy ob3ec;t~ve ,depends •bot'h or,i whetlhe,r ex,change rate varrnbfl ;i ty
has 1ut;1des~ra'ble •consequences f.or ithe tlJ.S. q•r,i:flatrnn :rate and ·neal ,activity
(as 1ul,tiimate tanget Mar:i,ab~,es) .and ,on the extent
t0
•whq ch 1exchange
rate •v,a1r1J,ab<1 ij ;j t,.y ma1kes 1t more ·d1ffikt1lt for ;for.ei,gn COJ!.mtra•es to ,sta1b:i ~ dze
-their econom1 es.
In this context, the evfi,dence ·suggests that greater shor:t-
1run exchange ,rate variabi~;ity ,does ·not genera:l]y 'ha,ve signifka·nt •undes1ra'bJce
consequences for 1U.S,. or fo,re1gn infqation nate•s or r.eal a,ct;jv;ity '1.eve1s~ 'U
has 1seer.i ,hyp0thes1zed, lin part:icuJar., t•hai.t increases in iimport ,prices ratchet
up 1domesit1c .price teveijs wh:iqe declines ;i.n import prices do not have a symmetric ,downward ,eft.ect, •whiich \W(i)Ul,d imp]y a •net rnifl at:wnary impact -of
1
greater ,e,xc'bange rate var:;i,abdli.ty.
,Little empiir:ical evi.dence of such ratohet
ef;fiects has been mound~ especia~ l;Y for the
.u .-5~
economy. <(See Sect fora 'VL/LJ
'Ne,v,er-theless,, to the ,exteflt that tighter U.S. monetary rpo~ 11cy as
a cor,isequence of the new Ol'!)erati•ng iprooedur.e ihas l,ed to str.onger dol ~-a,r
exchange ,rates ·0r to ,po;l ii cy adJ ustments a•broad, the s·h1 ft in operatrn_g
0
,procedures may rndeed have ,had ;impp.r.tan:t impacts on fore, gn pr, ces and
1/ Ster~lized 1ntervent~on may a~so 1nfluer.ice exchange rates through
a second channel. In an unj:ertam·world, ,assets denominated in d1f,ferent
currencies that offer the same ~xpected yte~cls w1~l n(i)~ necessar1ly ~eregarded as perfect substitut,es by ri•s'k-aNerse .investors. Accordingly,
J;o,n,,ard exchange r.ates may dJffer firorn expected future si;>ot ra,tes .by ±he
11
11
' n s'k premrnm
that pr, vate 1 nvestors ,r.equ1 re to ,match the currency com:pos 1trnn of their aggregate ,portfolio w;ith tt:ie re~ati.ve stocks •of pub~,c
,debts that authorit1'es have t'hrus't ,upon them. ~Jhether changing the currency
1compos:it;i on o'f pr, vate portfo9 jos through ster, 11zed rntervent,,on has ,a
,quanti,ta'tively s1gn,f,icant impact •on nsk premiums and thereby on exchange
rates, 1however~ ,remains a•n open emp,1rical queshon.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1
-18-
activity levels, with additional feedback effects' on U.S. price and
activity variables. The principal types of impacts on U.S. and foreign
prices and activity can be describea by considering two basic cases.
When policies in foreign countries do not respond and the
dollar appreciates (in both nominal and real terms), foreign currencies
wi 11 face higher l oca 1-currency prices for their import's and consequent
upward pressure on their domestic price indexes.
The United States will
face lower dollar prices for imports and less domestic inflationary pressure.
Furthermore, the appreciation of real dollar exchange rates, although
not permanent, will have lagged effects on trade flows for several years.
Foreign countries will be led to substitute away from U.S. exports toward
competing products, while U.S. consumers will also shift away from U.S.
products and increase their imports. Thus, in this case, the tighter U.S.
policy will have a depressing influence on U.S. activity that is reinforced •
~
by international substitution effects away from U.S. output~ whereas foreign
activity will be promoted by the substitution effects but held down
tiy
lower
U.S. import volumes associated with lower U.S. activity.
If, as an alternative case, foreign countries respond to tighter
U.S. monetary-policies by letting their interest rates rise in order to
stabilize exchange rates, the tighter U.S. and foreign-country policies will
put downward pressures on both U.S. and foreign real activity variables (with
feedback or international multiplier effects operating through lower import
volumes).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
,,
While in this case both U.S. and foreign import prices will remain
relatively statile in the absence of any exchange rate change, the downward
,.J
•
-19pressures on activity are transmitted to downward pressures on prices in
all economies.
Because of ,ts impacts on prices and activity, the tighter U.S.
monetary policy if sustained -- whether it leads to tighter policies
;
abroad or to stronger dollar exchange rates -- may also induce significant
changes ,n trade and current-account balances.
Trade volumes will change
to reflect both the income effects of changes ,n activity levels in the
United States and abroad and the substitution effects of changes ,n real
exchange rates.
Trade balances, measured ,n value terms, will be further
influenced by changes in the prices of tradable goods and may adJust over
time to exh1b1t the familiar J-curve pattern.
In addition to its effects on the variability of exchange rates
and the actual tightness of U.S. monetary policy, the new operating procedure may have been associated with greater uncertainty among foreign
policy authorities with regard to their perceptions about U.S. monetary,
policy.
Such uncertainty was particularly notable when U.S. interest
rates rose 1n March and, again, in November of
countries may not have desired to follow.
1980
to levels that foreign
(See Section IV.) At such times
foreign authorities might not have great concerns if they were confident of their expectations that the extreme movements ,n U.S. interest
rates and exchange rates would be short lived -- that is, confident of
their underlying perceptions about the general stance of U.S. monetary
policy and the general performance of the U.S. economy.
It can be argued
that the increased uncertainty about U.S. policy over the past year or so
may reflect an unfam1l1arity with the 1mpl1cat1ons of the new operating
procedure and may not be inherent ,n the use of the procedure 1n the future.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-20Section III -- Exchange Market Developments
In this section we present the results of our ~nvesttgations
of exchange market developments since October :6., ~979.
'First, ·we
analyze the statistical behavior of international rnte,rest rate differentials as a necessary introduction to our ana~ysis ,of the 1behavior
of dollar exchange rates (spot and forward); next., we Jook for possible
changes ~n the responses -0f exchange rates ~o changes dn determ,~ants of
exchange ,rates, ,particuJarly interest ,rates; frnal1y, we a,ook for poss,ible
systemat1 c changes in the intervention :be'havior ,of monetary authori t:ies.
,A.
Variab11 i ty of Exchange Rates!_!
As 1s re.ported :in anothe,r ,paper in this study., s~(nce Octobe,r 6,
1
1979, the variabi~ilty of the federa1 funds rate and of interest rates on
Treasury securities across the maturity spectrum has increased .V
Other
things be1•ng ,equal , such an increase fo van abri l 1 ty ,might 'be expected to
ihave ,been a·ssocrnted wit·h an ,nc•rease
;in
the ,va,riabi~ ity of rnternationaJ
rnterest ,rate ,<:Jifferentials and, ,n turn, wHh an increase in the vari'.In this sub-
abiq 1rty of exchange ,rates -- a·t 1 east spot exchange rates.
section, we nep0rt on ·our empfr•iicaJ examination o_f these two related
qtlest,,o,r;is ,us,mg a comm0n method010,gi1ca~ framewo,rk.
We fi-rst ,had to define what we meant ,by the term
11
varrnb1 l ity. 11
It may refer to the changes ·(absol~te or al1ebraic, fro~ one observation
to the next or to the ~~spersion -0f such ,changes ilietween success)ve ~bservat ions.
It may al so refer to the extent of error :in predkt:wn.
ifhe
1/ The principal contributor to th~s Section was Ralph M. 'Smith.
11
Jnterest ;Rate Varna'M1 ;i,ty 1under
the New Operating Procedures and the Initial Response ;in 'F;inanci:a~- 'Markets. 11
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
2/ See Dana Johnson and others,
-21time interval over which var1ab1l1ty 1s measured -- daily, weekly,
monthly, etc. -- must also be specified.
In the case of exchange rates,
one might look at forward exchange rates for various time periods as
well as at spot exchange rates.
In the case of interest rates, one might
look at rates on assets of various maturities.
In the case of the differential between the interest rate on
dollar-denominated assets and the interest rate on foreign-currencydenominated assets, we examined three-month interest rates (the Eurodollar rate minus a representative three-month rate in the relevant
domestic market).
We calculated algebraic changes in these series over
intervals of l day, 1 week (5 days) and l month (21 days)l/ and calculated
the standard deviations of the changes as the basic measure of variability.
Each series ·was divided into three periods:
(1) the period from March 1973
(or somewhat later, as determined by data availability) when floating exchange rates began to October 5, 1979; (2) the period of the new operating
procedure from October 8:, 1979, to the end of tlovember 1980; and ( 3) the period
from November l, 1978,to October 5, 1979. lhe third period was selected
to begin with the date of t'he ' d0llar 'defense package, which some observers
11
viewed as signifying a change
,n
11
itJ.S. exchange rate policy towa,rd providi,ng
rrore intervention and •poli·cy support for the dollar to assure less varia'bifity
~/ The 5-day interval corresponded to 7 calendar clays {l week), and
the 21-day interval corresponded to the average number of market days in a
calendar month. We constructed the weekly and monthly series as changes
from a single rday to a single day to avoid the downward bias to the measure
of variability that would have been imparted by averaging daily observations.
Usi:ng this procedut'le we 9enerated 5 ",weekly" series and ,21 "monthly" series
from the 5th and 21st differences in t·he 'basic series.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-221n the dollar's exchange value.
Comparisons were made between vari-
ability in the post-October 6 period and variability 1n each of the
other two periods.
Table 1 pre~ents the results of the analys1s of the var1ability
of three-month interest rate differentials. The results show a statis"
tically s1gnJf1cant increase 1n the variability of all the calculated
differentials after October 6, 1979, regardless of the country, the time
T
interval (daily, weekly, monthly), or the time period used in the comparison.11 As is discussed in more detail in Section IV.B below, 1n a few
countries the variability of three-month interest rates appears to have
increased somewhat after October 1979, but the increases are much smaller
than those for the three-month Eurodollar interest rate.
Thus, the results
presented in Table l reflect primarily the large increase in the variability
of three-month dollar 1nterest rates that is reported in another paper in
this study.
Given the general results showing an i~crease in the variability
of interest rates on Treasury securities across the maturity spectrum, it is
not surprising that we found similar results for Eurocurrency interest rate
differentials for longer maturities -- one-year rates and five-year rates. 2/
'
1/ The formal statistical test is of the hypothesis that the '
variances-of the series in two periods are equal -- that their ratio
equals 1 .00. If the test showed tnat the probability of obtaining the
calculated value of the ratio (when the ''trueuvalue was 1.00) was less
than 5 percent, then the equal-variance hypothesis was rejected. The
fact that for some of the tests reported below the estimated variances
increased or decreased is not useless information, even 1n the case
when the change was not statistically sign1ficant. We can ~t least
say that it is more probable that the variance increased (decreased)
than that it decreased (increased).
2/ For the series examined, the only exception to the pattern
was the series for the one-year Eurodollar-Eurosterling interest rate
differential since October 1979 compared with the entire 6½-year preceding
period. Mere either there was little change or the increase was not significant in a large number of cases.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-23-
Tab 1e 1
Var1abili'ty of Three-Month Interest Rate Differentials:
Eurodo~lar Minus Foreign Rate
(Standard deviations of changes)
3/73 - 9/,79
l'0/79 - 11 tao
11,178 - 9/79
.252
.236
.220
.284
.376
.407
.394
.412
A83
.271
.250
.231
.256
.314
.446
.373
.408
.473
.876
. 781
.938
. 754.966
.360
.388
.338
.337
.517
2.601'
2.371
2.766
1.902
2.608
.661
.761
.672'
.643
1.092
DAILY
Germany
Switzerland
Japan
Canada
United Kingdom
AlO
WEEKLY
Germany
Switzerland
Japan
Canada
United Kingdom
.678,
MONTHLY
Germany
Switzerland
Japan
Canada
United Kingdom
.965
.751
.904
.761
1.212
Note. Standard deviations of weekly and monthly changes are means of standard
deviations of 5 series of -5-day• changes, and-,2-h ser•ies of.· 21-day, changes res11ectively . .,,
All series showed a statistically significant increase in variability in the postOctober 6 period compared with the two earlier periods.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-24_ In the case of exchange rate variability we used as our basic
measure of variability the standard deviation of algebraic changes in
percent.l/ Changes, rather than levels, were chosen in order to eliminate the influence of strong time trends in some of the series.
Percent-
age changes were used because they seemed more appropriate, that is, consistent with the formulation of most models of exchange rate qetermination.
The rest of the procedures was similar to those summarized above for,the
interest rate differentials.
Table 2 presents this measure of the daily variab1l1ty 1n spot
and one-year forward exchange rates for the weighted average dollar,a~d,
for five bilateral dollar exchange rates and two five-year forwara b1lateral
exchange rates.
The spot 1O-currency weighted average dollar, shown
in the top
line of the top panel of Table 2, increased in variability after October 6,
1979, compared with either the preceding 11-month per1od or the entire 6½-year
period.I/ The results for the bilateral exchange rates are similar except
that the increase in the variability of the rate with the Swiss franc is
not significant when the period since October 6, 1979, is compared with the
longer preceding pen od .. _
~
...
~
y
'
..# .... ~ ............ "'I;..-
The results for the one-year forward weighted-average dollar,
shown in the top line of the middle panel, show
a significant
increase
1/ We also computed, but do not report here, mean absolute changes,
which yieTded the same general pattern of results.
2/ As shown by the asterisks in the first and third columns the
hypothesis that the variances are equal can be rejected at the 5 percent
level of significance.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
::i._ ~"
,,
-25Table 2
Daily Exchange Rate Variability
(Standard deviations of percentage changes)
3/1/73 - 10/5/79
10/9/79 - 11/28/80
SPOT
Weighted-average dollar
German mark
Swiss franc
Japanese yen
Canadian dollar
Sterling
.373*
.573*
.738
.488*
.195*
.462*
.428
.707
.770
l. 337
.244
.770
.337*
.427*
.596*
.590*
. 211 *
.512*
1-YEAR FORWARD
Weighted-average dollar
German Mark
Swiss franc
Japanese yen
Canadian dollar
Sterling
.408*
.627*
.799
.528*
.255*
.609
.489
.578
.788
.783
,318
.548
.627*
.541
.719
.681*
.284*
.610*
.759*
.910*
1.070
1.136
.769*
.897*
'
11/1/78 - 10/5/79
5-YEAR FORWARD
German mark
Swiss franc
*Significantly Q1fferent from 10/9/79 - 11/28/80 period at .05 level of significance.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-26-
1n var1abil1ty after October 6, 1979, compared with the entire 6½year period, but a significant decrease in variability compared with
the 11-month period from November 1978 to October 1979.
However, the
results for the weighted-average dollar are somewhat contaminated by the
averaging process, and ,tis appropriate to look again at the bilateral
exchange rates.
Here a somewhat more-mixed pattern emerges than was
the case with the spot exchange rates.
When the comparison of the recent
experience is made with the shorter preceding period (columns 2 and 3),
two of the bilateral one-year forward rates show significant increases in variability (the yen and Canadian dollar}, two show increases in variaoi1ity
that are not significant (the mark and Swiss franc),' and one shows a
significant decrease in variability (sterling}.
'
'
When tne comparison is
made with the longer pre.ceding period (columns 1 and 2}, the same two
exchange rates show significant increases in variaoility, but the other
three rates show reductions in variability -- a significant reduction in
the case of the. mark. As shown in the last panel, the five-year forward
rates for the mark and the SWiss franc show significant increases in
variability since Octooer 1979 compared with either earlier period.
One other aspect of the results reported 1n Table 2 1s interesting.
In the post-October 6 period, three of the five one-year forward
bilateral exchange rates (mark, yen,and sterling) exhioit less variability than do the spot rates for the same currencies.
middle and top panels in the second column.l
(Compare the
In contrast, in the two
earlier periods (first and third columns), the variability of the oneyear forward exchange rates is greater than the variability of the spot
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-27rates -for a~'l five purr_!=nc:ies ~n .:the ta·b~e
_,,T.his findfog 1s ,co,1i1S1stent
wi-th the •hypot·hes1s, deveij,o,ped :an Seotion -:n, that th_e increased v,ariabrlity of re 9,l ,dollar interest •raite s1f,fere.nt1al,s should .affect the
varia·bil'lty of spot ,mor:~ .rt:h~F,1 ·f@nward •exchan9e rate-s.
two :Five-year forw_ard rates shown
-,fl
t~
~
Howe.ver, tbe
as·t pane'l of Tab'l,e 2 ,do not
,exhib:i-t less var1ab-, ..Hty COITJpared .w-'lt•h the spot ,rates §"wee October 1~979.
Jh1 s res_1..1~l,t sugges-ts .the ,~y,pothes~s .t'ha·t the -yar1ab1 h:t;y ,oif the f11-ve-year
interest rate d1 fferent1a·1 s ,n the recent peri_gid :refl,ected, to a ,g,re-a.:ter
,degree than has b.e~n the •C9-se f.or some p.f the one-year ,imte,rest rate
@1fferentaals, N~ri9t4ons in ,nomina~ rather than real interest rat~s.
Tables 3 ar,ig 4 ,present meas_ur:es of exchange irate varta6.f1l Hy
',oyer ,week0:Y ar;ip ,mor;ithily j)~:terval s.
,As for th,e ,cal.cu~at10ns 1:1srng ,the
jr.it,er~.?t r,fl,te,diiff._erentiaJs pre_sented i:n Jab]e ~. ;f~,ve •non-ove,rJapping
•s.~r~,es 0f w.eekiy int.erw.a~s aod twenty-o-ne ,n.on~overla,P,ping series ,,of
,monthly i,n,t~r:v,al-s 1w~re ,co,nstr:.ucted.
The ,resuj;ts in Ta-b]es 3 ana 4 ar~
the ,m~ams oif stand~.rd t<ilr~j.a;t.10ns from :tbe re.s;p,e.ctw,e s,eries } l
,As the •1 ength _pf the inte,rva 1 o,ver which the se,r;f es ,of ,exchange
•r,a;te chijn,g_e,s 1,r:ic,neased fr.om ..dai~y t~ wee:kll,Y to m0r,,th~¥, tf:ie ,pro~ortion
,of the ser;ie,s showim_g 1c;ln tncre,ase 1 ir.i .v,ar,iabjl ity ,decl in~Q ar.id ,1the pr,.o,por:- _
ti9n ,s,howrng reduced ,v,_an,abflity rnc,reas,e..d.
Tt.i.e month]y r._esults J,n Table
4 in comRarison with th~ ~aily results in Table 2 illustrate this pattern.
Al J of th.e spo;t ,exs:,hange rate ,cQtn,pari,s,ons showr,i in the top pane1l, exc~pt
that ;for st~rl jng ,co111pa.red wjth tbe ppst,...,Nov.ember ~·978 ,p,e.rj,od, ,s;h!))w a
rjse j_n ~be ,mean st~ndard .deyi,at~0A..
HoweYer, for the •weighted-average
,dollar pnJy ,aboµt half pf th~ 2~ s~ri~s show a statistjcalJy significant
1/ S1gnif1cance tests were performed for each non-overlapping
series, and the results are summar1zed in Appendix Tables l and 2.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-28Table 3
Weekly Exchange Rate Variability
(Means of standard deviations of five series of
five-day percentage changesl/)
3/73 - 9/79
10/79 - 11/80
11/78 - 9/79
SPOT
Weighted-average dollar
German mark
Swiss franc
Japanese yen
Canadian dollar
Sterling
.870
1.290
l .630
l. 128
.469
1.069
l. 021
1.459
1.705
2.260
.565
l.393
.789
.977
l .471
l .316
. 511
1.263
1-YEAR FORWARD
Weighted-average dollar
German mark
Swiss franc
Japanese yen
Canadian dollar
Sterl mg
.888
l.381
1.736
l. 210
.546
l .345
l.011
l . 241
1.555
1 . 712
.638
l. 120
l . 319
1.200
l. 707
l.491
.616
l.476
1 .633
l.893
l. 983
2. 165
l .891
2.259
5-YEAR FORWARD
German mark
Swiss franc
1/ Significance tests of the individual weekly series are summarized in Appendix
Table 1.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
•-
~
-29-
Table 4
Monthly Exchange Rate Variab1l1ty
(Means of standard dev,at,ons of 21 series of
21-day percentage changesl/)
3/73 - 9/79
10/79 - 11/80
SPOT
"We1ghted.:.giverage' dollar
German mark
Swiss franc
Japanese yen
Canadian dollar
Sterling
2. 04'1
3.046
3.430
2.609
l. 158
2.450
2.777
3.537
3.946
4.529
1.325
2.617
1.748
2. 197
2.886
2 .150
1.309
2.830
1-YEAR FORWARD
Weighted-average dollar
German mark
Swiss franc
Japanese yen
Canad, an do 11 ar
Sterling
1.915
3. 120
-3.727
2.852
1.289
2.975
2.334
2.819
3 .181
3.834
l.289
1. 958
2.369
2.797
3.622
2.676
1.509
3.113
5-YEAR FORWARD
German mark
Swiss franc
3.472
4.032
4.527
4.224
4.436
4.396
-
11/78 - 9/79
-
1/ Sign1f1cance tests of the individual monthly series are summarized ,n Appendix
- Table 2.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-30-
increase in variability after October 6 compared with either preceding
period.
For the bilateral spot exchange rates, the yen (in the compari-
son with preceding 11-month period) shows a significant increase in
variability after October 6 in more than half of the 21 series, and the
mark (in the same comparison) shows a significant increase in about half
- ,.: ,
oJ the 21 _s.eries.
Both currencies show less" evidence of a--signif1cant-
increase in variability after October 6 when the comparison is made
with the longer preceding period.
For the 1-year forward exchange rates (middle panel), the
comparison with the November 1978 to September 1979 period shows a
slight reduction in the mean standard deviation in the monthly variability of the weighted-average dollar (none of the 21 series showed a
significant change} as 'well as reductions in the monthly varia6ility 'fo'r
the Swiss franc, the Canadian dollar, and sterling.
(Only for the last
'
currency did any of the 21 series show a significant decline in varia,bility.)
A small increase was recorded in the mean standard deviation in the monthly
variability of the 1-year forward mark (none of the 21 series showed a
significant change) and e larger increase for the yen (onlyl of the
21 series showed a significant increase).
The results of the comparison
with the longer preceding period are similar -- a bit more evidence of
an increase in the monthly variability of the weighted-average dollar,
the yen, and the Canadian dollar and a bit more evidence of a reduction
in the monthly variability of sterling, the Swiss franc, and the mark. For
the 5-year forward exchange rates (bottom panel), only the results for
the mark yielded substantial evidence of any change in monthly variability.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-3'1-
f_or the mark., there ,was some 1,ncrease in .month l:Y var, ab1·l 1ty in the postOctober 6 period compared with the longer ·pneced1ng period.
Again comparing the variab1l1,ty,of the spot and forward rates,
one can ~ee 1n Table 4 that tn the post-October 6 ,period the mean
standar,d ·deviations of the month~y series for ,each of the ,fwe one-~ear1
f.orward b;ilateral exchange r.ates was lower than for the correspond;f•ng spot
exchange ,r::a.te ii n contr.as,t with the patter.n in the two preced,,ng 1perii ods.
·However, the mean standard dev~at,on ,of the monthly series for the two
fiive-year-forward rates .again shows ,an increase relaNve t0 the spot
r.ates in all the periods.
Cross-country analys~s of the results presented in Tables 2-4 on
daily, .weekly, and monthly exchange rate variability (and, indeed, the
,results for interest rate d1fferent1als 1n Table las well) does not reveal many striking patterns.
Such analysis does focus attention on the
fact, as was noted in Section II, that the variability of bilateral
exchange rates depends not only on developments Jn the U.S. economy but
a~s0 on developments in individual economies abroad and common developments, e.g., foternational •oil price increases, tha,t may have different
effects_ on different economies., One exampl'e from Tables 2-4 may help to
illustr.ate this point.
The variability 0f the :Yen-dol~ar ·exchange rate
(spot and one-year forward) in the period after October 6 incrseased more
in comparison with the two precedang~periods for all three time intervals
(daifly, weekly, ,and monthly) than .did the variability for any-'of the
other four 'bi~ateral ,rates shown :in Tab]es 2-4.l/
Jn€1eed, in s•ome cases
1/ The ~easure used was the r,tio of the standard ~eviation (or
the mean-standard deviation for the ~eekly ana monthly series) in the
,post-October 6 period to the standard deviat~on an each of the two pre'Ceding periods.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-32-
for the other currencies, the variability declined. This result prqbably
reflects not so much effects of the Federal Reserve's operating procedure
but rather the strong cyclical movement in the yen over this period -induced by Japan's inflation and current-account performance -- as well
as the somewhat more relaxed attitude of the Japanese authorities toward
fluctuations in the yen's exchange rate noted in Section IV.C below.
The results of all of these calculations point to a definite
increase of variab1lity of spot exchange rates measured over daily,
weekly, and monthly intervals. The evidence for forward rates is not
conclusive, though there are certainly cases of decreased variaoility
especially for the longer intervals. This latter evidence offers limited support to the hypothesis that the new operating procedure could
lead to a reduction of the variability of forward exchange rates.
Another approach to measuring variability would be to measure
the variability of prediction errors from a "structural" economic model
of exchange rate determination.
An examination of residual variances
in the model reported on in the following section indicated no significant change in prediction error of monthly average exchange rates
~
after October 6.
B.
Responses-Of Exchange Rates.!!
Ha'ving established that ·th.e variaf>iltty of spot exchange rates
increased after October 1979, we next examined the causal factors under-
J/ The principal contributors to this section were Peter
Hooper and John Morton.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
lying the increased variaotlity-with a focus on the ~esponse of exchang~ rates to the 1ncreased ~ariab1lity of 1nterest rates -- more precisely
to the increased variab1l1ty of interest rate differentials.
We sought,
first, to determine if the responsiveness of exchange rates to changes in
U.S. interest rat-es increased or decreased following the October 1979
measures.
No statistically significant evidence_of a shift in this
causal relationship--attributable to the new operating procedures was
found.
We then estimated the change in the variability of interest rates
since October 1979, assuming that no shift in the causal relationshipbetween interest rates and exchange rates had taken place. This analysis
suggested that, b_y one measure ~t least, the month-to-month variability
of exchange rates attributable to interest rate ,changes increased threefold after October 1979, compared with the average experience during the
previous six year~.
However~ the total variability of exchange rates
increased much less between these two periods, as fluctuations in other
,-
= ~'
'factors that affect-exchange rate declined. ~- -• ,._,
1.
~
'
Ch
r
Tests for Strnctural Shift
To test for a structual shift in exchange rate relationships,
~•e estimated a model of exchange rate determination that express
dollar 1 s weighted-av~rage foreign exchange value as a function of the
differential between U.S. and foreign short-term interest rates, relative
U.S. and foreign prices, and a variable indicating the degree of imbalance
in the U.S. trade position.l/ The lvst two factors are included to explain,
respectively, changes in the underlying nominal and real exchange rate
lf To be consistent with the other series used in the model,
monthly averages of the exchange rate and interesi rate series were used
rather than the pure series constructed for the analysis in Section III.A.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
• leve1s that are not direttly associ~ted wfth thang~s iti interest rate
different1als.ll This equation, estimated over the period August 1973 to
October 1980, 1s shown in column 1 of Table 5.
In order to test for structural shifts, the period December
1973 to October 1980 was partitioned into subperiods, divided before and
after October 1979. Chow tests were employed to test for structural
stability of the whole equation. To test for structural stabil{ty of
the interest rate coefficient in particular, dummy variables and t-tests
were used.
(These resu1ts are shown in line 4 of Table 5; the estimated
coefficients indicate the addHional responsiveness of exchange rate's to
interest rates during the period when the possible shift occurred.l
The results suggest that the structure- of exchange rate determination was not the same after 0Ctober 1979 (based o,n a Chow test for
the results shown in columns 5 and-6 of Table 5). Also, the responsiveness of the exchange rate to interest rate changes appears to have increas·ed, as indicated by the· sigmfica-nce oLthe, las.t coeffjcient _in ~olumn 2.1
-
-
~
While these resul·ts clPP'ear to support the hyp()thesis that th.e change 1n
I
operating procedure in October 1979 was associated ~ith a shift in the
responsiveness of exchange rates to interest rates, the evidence is not
conclusive.Any of a number of events during the floating exchange' rate
period could have precipitated a shift in the exchange rate determination
11 The variables are described in more detail in Table 5, It
should be noted that interest rates are assumed to be exogenous. In
some circumstances, this assumption may be questionable. For a discussion
of the theoretical Justification for modeling the process of exchange rate;
determination 1n this way, see Hooper and Morton (1980).
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-35Table 5
Monthly Exchange Rate Equations for the Weighted-Average Dollarl/£1
(t-ratios in parentheses)
"
Determinant
(1) Relative pr,cesll
c
'
''
l
I (
(2) Cumulative trade balance
(3) Nominal short-\ rm interest
differential_
7
(4) Nominal short-term interest
di fferenti a1 5/
Oct. 79 - Oct. 80
-1
.49
0
86)
. 18
(5.69)
~
.55
(2.88)
\
,
Full sample period
8/73 - 10/80
2
-3
-
.62
( 1. 32)
. 17
(6.46)
.09
(.34)
.67
(2 .16)
(5) Nominal short-term interest
differential§./
Nov. 78 - Oct. 80
-2
Sum of Squared residua1s
R
.9545
.0211
. 77
(1.37)
.17
(5.44)
- . 16
(-.51)
.9563
.0201
. 78
(1. 57}
.16
(5.98)
-.33
'(-.95)
Sample Split
Samp 1e s,p 1it
at October 1979
at November 1978
8173-9179
10/79-10/80 8/73-10/7f 11/78-10/80
-5
-6
-7
-8 - 9
.86
2.78
1.28
.02
.47
(2.44)
(. 36)
(3.44)
(.01) (. 16)
.22
. 31
.00
-.06
-.00
(6 .37)
(5.82)
(.
13)
(-.39)
(-.14)
.69
.99
.73
.40
. 70
(. 18)
(.28)
(3.09)
(3.71) (1.00)
.36
(1.02)
.34
(. 90)
-4
.86
(2.73)
(2.08)
.9578
.0194
.9578
.0192
. 72
.9433
.0167
.5461
.0018
.9067
.0137
.5817
.0031
.5772
.0030
1/ Index of weighted-average exchange value of U.S. dollar against currencies of other G-10 countries plus Switzerland.
- Weights are 1972-76 total trade of each of the 10 countries.
y Equations corrected for autocorrelation with Cochrane-Orcutt technique.
3/ Weighted-average (10-country) foreign CPI, divided by U.S. CPI with both series expressed as 6-month weighted moving
averages.
y U.S. 3-month CD rate minus 10-country weighted average of foreign 3-month rates.
§_/ Interest differential times 0-1 dummy which takes value 1.0 beginning in October 1979.
§j Interest differential times 0-1 dummy which takes value 1.0 beginning 1n November 1978.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-36-
process.
.
-
To test for uniqueness of the apparent October 1979 shift, the
same tests were run spl1tting the sample at November 1978, coinctd1ng
with the adoption of the dollar defense program, the shift in U.S. exchange market,_intervention behavior, and U.S. monetary policy actions
aimed at strengthening the dollar. The results indicate an even stronger
reJection of the hypothesis that the exchange rate determination process
did not change (columns 7 and 8 of Table 5} and a more significant shift
>
in the exchange rate-interest rate relat1onsh5p (column 3) after November
1978 than after October 1979. Column 4 in Tahle 5 reports the results of
an equation in which hotb sb1fts were tested simultaneously with dummy
variables, and column 9 reports those for an equation estimated from
November 1978 to October 1980, in which the October 1979'shift alone was
tested.
These results indicate that, after allowing for the shift in
November 1978, the add1tional shift in October 1979 was not sfgnificant.
The tests reported in Table 5 ~mployed monthly data and the
ten-currency weighted-average dollar index.
Similar tests were run
using monthly equations for bilateral (dollar} exchange rates a9a1nst
the German mark, Japanese yen, Canadian dollar,and British pound, as well
addition, these tests and those reported above were repeated with the
real interest differential substituted for the nominal ·interest differential and, in the case of the quarterly equations, using a more complex
mopel of exchange rate determination.!/ The results of these tests, while
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
(1980}.
l/ The ~uarterly-model is described in Hooper and Morton
' ''
I
-37d1ffer1ng in detail,-,n general supported the conclusion that the October
-
'
1979 change ,n operating procedure by itself was not associated with a
significant shift in the structure of exchange rate determination, after
,_alJqwJng for tf)e_ possibility of a shift following the Novemoer 1978
-;;,. '
measures.
Finally, we al,so compared the 1980 interest rate and exchange ,5
.:i
rate cycles with those of 1974-75, a time period roughly similar in many
respects including the apparent importance of interest rate developments
for exchange rates.,
Table 6 shows the net movement of exchange rates and three-month
'
interest rate differentials from peak to trough and from trough to peak.
Troughs and peaks were dated by months in the case of the weighted-average
,
"
exchange rate and by weeks in the case of the DM-<lollar exchange rate. The' '
weighted-average and bilateral comparisons show similar results:
the
swings in interest rate differential~ in 1980 were larger than~in 1974-75
,
(although at a higher level of rates and at a more rapid pace}, whereas
,
the swings in spot exchange rates were smaller at least through November.
While the ratios reported in the last column in Table 6 suggest that the
movement, of exchange rates relative to interest rates was_Jower in1980 than in the 1974-75, it should be recognized that the ratios do·not ,-
..
t
~
-
t\
J
•
!
....
'-
'
'I-
•
take into account factors other than interest rates that affected exchange,
rates during these periods.
In any event, the data at least provide
additional, though weak, evidence that the responsiveness of exchange
rates to interest rates following the October 1979 measures did not
increase significantly relative to comparable historical experience.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-38-
Table 6
Interest Rate and Exchange Rate Cycles.!/
1974-75 and 1980
Percentage change in
Interest rat~ _
Exchange rate
d1fferentialYL
Weighted-average rates
, :Rat10J/
Period:
1974-75
Peak (s,pt-) to.trough (Mar.)
Trough (Mar.) to peak (Sept.)
- .. 8. 7
+ 9.7
-
1980
Trough (Jan.) to peak (Apr.)
Peak (Apr.) to trough {July)
Trough (July) to November
-
-- -
2 r7 ,·• -
-i;
+ 3.1, - 8.6
+ 8.3
+ 6.5
- 7.0
+ 5.5
OM-dollar rates
,3 .:e2· -~'::
3.3
:P
+ 2.9
.
, 2 .1
<
0.8
-
0.7
-'
I,,
Period:
1974-75
,
Peak (Sept. 11) trough {Mar. 5)
Trough (Mar. 5l to peak {Sept. 24)
-14.4
+16. l
- 4.3
+ 4.1
+13.8
-10.6
+10. l
+ 5.1
-10.4
+ 9.2
'
:• ,,l
3.3
-" 3,. 9,
1980
J
Trough (Jan. 9) to peak (Apr. 9)
Peak ~(Apr. 9) to trough (July 23)
Trough (July 23) to Nov. 26
2.7
,._ 0
1. l
' '
'
1
l/ Peaks and troughs are dated by exchange rates. Peaks and troughs in
interest rate differentials, reported here, generally prec~decl~~change r9te
peaks and troughs by one or two periods.
2/ U.S. three-month CD rate minus foreign three~month interest rate,in percentage
points.
.
'
\
"
3/ Exchange rate change divided by the change in the interest rate differential.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
"\
-
-392.
Variability of Exchange Rates due to Interest Rates
Based on the conclusion tnat the responsiveness of exchange rates
to interest rates was not significantly altered by the change in operating
procedure, Table 7 provides an indication of the impact on exchange rate
variability of the increased variability of interest rates since October
1979.
Line 2 of the table indicates that since October 1979 the average
absolute monthly variation in the differential between nominal U.S. and
foreign short-term interest rates has more than tripled to 1.56 percentage points, compared with previous experience beginning in either March
1973 or November 1978 ..l/ A sta6le relationship between exchange rates
and interest rate changes would suggest a similar tripling of the monthly
variation of exchange rates associated with interest rate changes.
Based
on the significant interest rate coefficient reported in column 8 of
Table 5 (.70), the average ,11-.solute percentage dtat1ge in the monthly
average weighted-average value of the dollar due to changes in the
interest ~ate differential rose from .32 percent during November 1978to
September 1979 to l .09 percent during October 197g to October 1980.I/
l/ Note from lines l and 2 that on this measure the increased vari~
ability of the interest rate differential is more than accounted for by
the increased var1ab1lity of the U.S. interest rate.
21
The interest rate coefficient for the period August 1973 to November
1978 was about the same magnitude (compare columns 7 and 8 of Ta61e
5), though not statistically significant. Based on this coefficient,
the variability of exchange rates attributable to interest rate changes
also about tripled after October 1979 when compared with the average
experience during the previous six years, as shown by the figure in
parentheses in Table 7.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-40Table 7
Variability of Interest Rates and Exchange Rates
(Average absolute month-to-month changes)
March 73-Sept. 79
Oct. 79-0ct. 80
Nov. 78-Sept. 79
U.S. 3-month CD rate
{percentage points)
,.42
1.56
.38
U.S.-foreign 3-month
interest differential
(percentage points)
.45
1.56
.46
Exchange rate changes
due to changes in
interest differential
(percentage changes)}/
(. 31 )
1.09
.32
Exchange rate change
{percent)
1.37
1.68
1.10
1/Based on estimated interest rate coefficient of .7 for the period
- November 1978-0ctober 1980 (reported in column 8 of Table 5). ,
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-41-
A comparison of the bottom two lines of Table 7 shows that
whfle interest rate var1ab1lity alone would have caused monthly exchange
rate variability to increase after October 1S79, the total variability of
exchange rates due to all. causes increased by much less, on this measure.!/
This suggests that a decline in the variabili,ty of exchange rate determinants other than interest rates -- such as actual and expected relative
price movements or trade balance changes -- partially offset the impact
of increased interest rate variation.
,
2
C. Exchange Market Intervention_/
As Section III.A of this paper reported, fluctuations in short,
'
term interest rate differentials have increased since October 1979 and
'
day-to-day exchange rate changes also have increased. Somewhat greater
variability in spot exchange rates over weekly and monthly intervals has
also been experienced.
Has this increased variability led to more force-
ful intervention action by monetary authorities to resist spot exchange
'
rate changes, or has it perhaps occurred because the authorities have
'
been less willing to commit intervention resources now than they might
~
have_been jn past years? This question is.relevant to two aspects of,_
11 The post-October 6 increase' in the measure of month-to-month variability of the spot weighted-average dollar shown in the last line of
Table 7 (about 50 percent compared with the period from November 1978
to September 1979 and about 25 percent compared with the period from
March 1973 to September 1979} is roughly equivalent to the increase
shown by the measure in the top line in the top panel of Table 4 (about
60 percent compared with the shorter period and 35 percent t"Ompared with
the longer period).
21 The principal contributor to this section was John P. Wilson.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
the overall analysis.
F1rst, if intervention behavior has cnanged,
the observed pattern of exchange rate movements, in turn, may have been
influenced.l/ Second, if intervention behavior has cnanged; sucb._a
change might be interpreted as~evidence that the Federal Reserve's new
operating procedure has caused·difficulties for other countrfos .:_,a
topic that fs discussed in more detail in Section IV.
Table 8 provides sultmlary information on annual changes;in tbe
weighted-average value of the dollar and on U.S. and foreign net tntervention from March 1973 through November 1980.l/ The annual amount of
net intervention rose sharply in 1977 in -comparison with earlier years
and has remained high since then.
However, one cannot Judge on the basis
'
"
of the amount of intervention alone whether monetary authorities have
changed their intervention beha,vior.
The question of a possible cflange in intervention behavior
'
can be explored by statistical methods that search for changes in
I
amounts of intervention per unit of exchange rate change.
W
..
M
\
However,
--
October 1979 was not the only recent landmark which might be associated
with a basic change in intervention behavior. An important earlier,
possiole benchmark was tne Novemoer 1978 announcement of a massive
cQoper.ative program of support for the dollar.
-..
...,,,
J'I>
.
~
In the interval, from
I.,
,.
'
"'(
March ,1973 to November, 1980, tb.er~fore, structur~l ch~nges in intervention 'behavio~ may have 'occu;red at least twice.
'
l/ , Thi~ statement assumes that -e:change market interven.tion affects
exchange rates at least in the short run.
2/ Net intervention is presented because the table presents the net
change in the dollar's exchange value for the relevant period.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-43-
Table 8
Exchang~ V~~ue o~ the Qollar ,and Int~rvention
by ~aJor Countries: 1973-80
Change in weightedaverage 1yalu~ of •the
,dollar'"-' (per_cent)
Period
N~t
~nt~rventJon ~b1llions o~ do1~ars)
~.S.
Foreign2/
1otal
J•L2
- .l
-.14.8
.-'l.4.9
197.~
..~4)8
~- 1
.-11 .c4
--11._3
197:5
,6. 2
-.i
_5.0
-4. 7
1.97_6
1_.
.o
-- .4
-""~3.)
,4.. ,1
J::.977
,. 7,. 8
-- .4
,36.,l
"35.. 6
'l Q7.j3
-10.,3
5:7
27_.8
33 ..4
19]_9
- .. 8
- ...8
-J,7 .. 2
-as.a
1980 ({j)_a•n. -Noy. )
_5 .,]
-7 .1
,-14,. 7
.e,;.21.,8
1973 (Mar.-Dec.,)
-
-
-
-
J/ ·End 1of year ·(or mQnth,) ifrom the (end of th!:;! prec~d,tng Jear (9r
=-- ,month).
.
'1:_/ •G-J_Q countries p-lu_s ·SwHzer'land, _Denmark, Ireland, and Norway.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-44-
Our rnvestigation of th.ese possi.ble shifts involved linear
.
regression analysis. The absolute value of U.S., foreign,and total
(U.S. plus foreign} monthly net intervention, deflated by the U.S.
CPI, was related to the absolute percentage change 'in the dollar's value
in that month. The possibility of as.Y11111etries in intervention behavior
when the dollar is appreciating and when it is depreciating was examined
by partitioning the independent variable accordingly. The tests for
structural shifts were performed by including, for the two subperiods
of interest, dummy variables in the equation along with the basic
1/
explanatory variables.-- It should be noted that there are a number of
potential statistical problems with this procedure,, including the possibility that the exchange rate change is endogenous.
From this perspective
as wel~ as for several other reasons the results should be regarded as
illustrative.
Table 9 presents the results for the basic equation. The
results for U.S. intervention, shown in column l, indicate that there
was a significant shift in U.S. intervention behavior in the direction
l/ The basic equation was the following:
I= aORl + a,D1R1 + a2D2R1 + bo~ + b1D1R2 + b2D2R2
where
I= absolute value of monthly net intervention, deflated by the U.S. CPI;
R (r. ) = absolute value of monthly percentage change (end of month} in
1 2
the spot dollar I s weighted-ave.rage va 1ue (1 C-currency index}
when the dollar is appreciating (depreciatingl; other values O;
D1 = dummy variable, O prior to November 1978 and l thereafter; and
D2 = dummy variaole, 0 prior to October 1979 and l thereafter.
A better "deflator" of intervention activity might be some measure of
exchange market volume, but this is not available for the 1ong sample or
on a month~y ~as,~. Since the CPI rose far less over the sample than
benchmark 1nd1cat1ons of exchange market volume, the price index can be
regarded as a conservative deflator.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
~
-45Table 9
Estimates of Central Banks' Intervention Response, Percentage
Change in Weighted-Average Dollarl!
{March 1973 - November 1980)
United StatesY
l
ForeigJ/
2
U.S.+ Foreign
3
44.2
( l. 5)
490.5
{4.5)
497.4
(4.0)
Post 10/78 shift
164.5
(3.5)
-195.5
{-1.2)
-30.3
(-.2}
Post 9/79 shift
37.3
{. 7)
271.9
(1. 5)
337.6
(1 . 6)
62.8
{2.4}
579.5
{6.2)
605.6
{5.7)
377.4
(5.7}
128.6
(. 5)
513.4
(l. 9)
-340.9
(-4.4)
-467.1
(-1.7}
-883.5
{-2.8}
.59
.49
.53
Dollar appreciating
Whole sample
Dollar depreciating
Whole sample
Post 10/78 shift
Post 9/79 shift
-2
R
Note. See footnote l, p.f3, for details of equation specification.
l/ t - ratios in par~ntheses.
y Includes Desk operations for both System and Treasury accounts.
1f Japan, Canada, United Kil)gdom, Germany, France, Switzerland, and Italy.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-46...'.
of offering greater resistance to exchange rate changes (appreciation as
we11 as depreciation} following the November 1, 1978, announcement.
In
the period since October 6, 1979; there apparently has been a significant reduction in U.S. resistance to th.e dollar's depreciation which has
approximately offset the increase after November 1, 1'978'. Thus, on,
balance, the only significant net ch.ange in U.S. intervention beh.avfor
since November l, 1978, has involved heavier purchases ot foreign currencies when thE: dollar was appreciating. This apparent shift in behavior,
in turn, may merely reflect th.e fact, that a signi,ficant amount of U'.S.
swap debt was outstanding on November. 1, 1978, which the U.S. monetary
authorities sought to cover as promptly as possible.
The results for the com6ined group of foreign countries shown
-
in column 2 suggest no significant shifts in intervention behavior
either after November
1,,
1978, or after October 6, 1-979.
Combining the U.S. and foreign net intervention yields the
results shown in the last column of the table.
Here there is weak evi-
-
dence of a somewhat greater response· to the dollar's depreciation after
November 1, 1978, and considerably stronger evidence of reduced response
(from this higher rate} after October 6, 1979.
Some exploration was also conducted of the influence of time _
trends (as proxies for omitted 'influences} and of measures of intramonth
exchange rate variability as additional explanatory'factors in intervention
behavior. Orre or the other factor, taken alone; often improves the equation
fit, but when entered together the trend effect tends to-dominate·, leav~
ing the variability coefficient insignificant. This is suggestive of
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-47some 11 underdeflat1on 11 of th.e data, but may a1so result from other influences.
In any case, enter1ng such. terms did not affect the basic pattern of resu1ts.
Overall, the empirical findings lead toward the conclusion that
it was U.S., rather than foreign, intervention behavior that has
changed the most following the 1978 dollar-defense measures and the
1979 change in operating procedure. The U.S. results suggest somewhat
greater tolerance of depreciation of the dollar since October 1979,
and this apparent tolerance, when combined with weaker evidence for the
foreign countries as a group, carries through in the overall equation
shown in the last column 1n the table.
However~ this effect in part is an
offset to the shift in the opposite direction following the November l
package and may only refl~ct the fact that the dollar has not experienced
a period of sustained decline since October 1979.
It should be emphasized that these resu1ts are quite sensitive to the specificiation of the equations. We have not presented
here a full model of intervention behavior, but we feel comfortable
in conc1uding th.at we have not, to date, observed a dramatic change in
intervention behavior by monetary authorities as a group since October
1979. Rather we appear to have found a possible shift back toward the
basic pattern that prevailed prior to November l, 1978, and weak evidence of somewhat greater tolerance of depreciation of the dollar.
Subject to these qualifications, the results, in turn, suggest
two conclusions.
First, it is unlikely that the observed patterns of
exchange rate movements have been contaminated by changes in intervention
behavior~ Any bias is likely to be small and would be in the direction
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-48of observing increased exchange rate variation in the period since
October 1979 compared with the prev1ous year but not the previous
6½ years.
Second, the results do not provide any evidence in support
of the hypothesis that the new operating procedure has caused difficulties for other countries.
However, we examine other evidence relat-
ing to this hypothesis in the next section.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-49Section IV -- The Foreign Experience under the New Federal Reserve
Operating Procedure
A.
Introduction
As discussed in Section II above, the new Federal Reserve
operating procedure, through its effect on the level or variab1l1ty of
U.s.- interest rates or of the exchange value of the dollar, may influence foreign output, prices, and current-account balances, and thereby
also have feedback effects on the United States.
However, the impact
of greater short-run (i.e., day-to-day, week-to-week, or even month-tomonth) exchange rate var1ab1lity per se is likely to be small.
The
impacts are likely to be confined to increased uncertainty as it
affects private and public decis1on-mak1ng.
Empirical studies, while
not denying the theoretical possibility of such effects, have generally
not been able to isolate them.
(See Section VI.A.)
Thus, the maJor impact on foreign countries of the new Federal
Reserve operating procedure has been through its possible effect on the
average level of dollar interest rates and exchange rates.
However, ,n
the absence of discretionary policy reactions abroad, the net effect of
tighter U.S. monetary policy on foreign economic variables is ambiguous.
Lower demand 1n the United States will tend to reduce foreign output and
prices and reduce foreign current-account balances.
On the other hand,
the tendency for the foreign currencies to depreciate against the dollar
will tend to raise foreign price levels somewhat and, with a longer lag
to divert demand from U.S. to foreign goods, thereby raising foreign
output and increasing current-account balances.11
1/ These qualitative statements are consistent with the
results of-s1mulat1ons with the Multi-Country Model (MCM) developed
in the Division of International Finance at the Board of Governors of
the Federal Reserve System.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-50,-
However, if fore:ign officials act, through exchange market
intervention and other monetary policy actions, to peg their exchange
rates, the impact of tighter U.S. policy becomes unambiguous with
respect to output and prices (but not current-account balances).
Indeed,
the negative impact on foreign output and prices of weaker U.S. demand
would be reinforced by the restrictive policy abroad and, by assumpti'On,
no competitive gain from a currency depreciation would offset this impact.
It is this aspect of an "interest rate war" -- raising the specter
of a synchronized and mutually reinforcing global recession -- that was
discussed in the press and in international fora early in 1980.
Changes in dollar interest rates and exchange rates have
induced policy reactions abroad, especially in Canada, in continental
Europe, and in some developing countries with currencies pegged to the
dollar.
The level and, especially, the timing of movements of foreign
interest rates were altered.
However, until recently, the available
evidence supports the view that the new Federal Reserve operating
procedure did not result in significant deviations from what policies in
the major industrial countries would otherwise have been.
Fundamentally,
domestic economic conditions abroad (notably the high level of inflation
and increased oil bills) were sufficiently similar to those in the United
States that essentially similar policy stances would have been called for
and would have been adopted in any,case.
In contrast to the situation prevailing for most of th~ period
since October 1979, a case,can be made that domestic conditions abroad,
especially in the key German,economy, in recent months have called ~ar
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-51~
policy actions that are different from those appropriate to developments
in the United States. Such a conflict between domestic and, external
needs is by no means the result of the new operating procedure per 2!L
but may be exacerbated by it.
Given that macroeconomic developments in
the United States and abroad are not synchronized {and policies are not
hannoni~ed), it is inevitable that conflicts between domestic and external
needs may ari~e from time to time. The new Federal Reserve operating
procedure may accentuate such conflicts to the extent that the new technique is designed to ensure a prompter and more automatic response of
interest rates to changes in money demand.
Even 1f monetary policies in the foreign industrial countries
have not deviated significantly
over the past year from what they other,
~
wise would have been, the large swings in U.S. interest rates have caused
problems for foreign officials. Several governors of foreign central
banks have said that fluctuations in U.S. interest rates, which were produced by excessive concentration on week-to-week fluctuations in the
money supply, were too large and imparted large fluctuations to exchange
rates with serious consequences for other countries.
To some extent foreign concerns may simply be a matter of a
general dislike of variability. Two specific aspects can also
be identified. One is the increased uncertainty about what the
level of U.S. interest rates will be, which makes it difficult for
monetary authorities to be responsive to domestic conditions and at the
same time to achieve short-run exchange rate objectives and also makes it
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-52difficult to predict or control budgeted borrowing costs. Another
aspect is the increased interest r.ate variability, which has caused
problems especially in some developing countries where domestic interest
rate structures are relatively rigid.
B.
U.S. and Foreign Interest Rates 1/
'• In order to examine-the relationship, if any, between the
variability of interest rates on dollar-denominated and foreign-currency
denominated ~ssets, several exercises were performed.
Standard deviations of changes in interest rates over daily,
'Aeekly,and monthly intervals were calculated for three-month interest
rates in five foreign countries and for the three-month Eurodollar
rate __y_ The results' are shown in Tabl~' 10. 31 All the series show an
increase in variabil'ity of interest rates in the post-October 6 period
compared with the preceding 11-month period for all three intervals; i~~
a number of cases, however, the increase was not significant.
When com-
'
pared with the entire 6½-year preceding period, German and British interest
'
rates show a reduction in variability; the reduction was sometimes
l/ The principal contributor to this s,ect10n was Ralph W. Smith.
2/ See Section III.A for a fuller description of the methodology employed. A full daily series for the U.S. three-month CD rate
was not available. The results for the three-month Eurodollar rate
appear to show an increase in variability of about the same order of
magnitude as that shown for three-month u.s.-Treasury bills in Table 5
of "Interest Rate Variability under the New Operating Procedures and
the Initial Response in Financial Markets. 11
3/ Generally similar results, not reported here, were found
for three-month domestic interest rates in France, the Netherlands, and
Belgium, for selected one-year Eurocurrency rates and for five-year
Eurocurrency rates.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-53Table 10
Variability of Foreign and Eurodollar Three-Month IRterest Rates
(Standard deviations of ~hanges)
3/73 - 9/79
10/79 - 11/80
11/78 - 9/79
Germany
Switzerland
Japan
Canada
United Kingdom
.156*
.178
.099:
.213
.321
.130
.190
.153
.304
.313
. 119
.110*
.076*
.141*
, .206*
Eurodollar
.194*
.389
.223*
Germany
Switzerland
Japan
Canada
United Kingdom
.297
.348
.224
.357
.603
.242,
.424
.433
.593
.508
.207
.261
.176
. 158
.419
Eurodollar
.355
.886
.301
.647
.674
.499
.618
1.073
.499
-.862
1.291
1.392
1.005
.391
.519
.408
.250
.849
.781
2.647
.612
DAILY
WEEKLYl/
MONTHLYll
Germany
Switzerland
Japan
Canada
United Kingdom
Eurodollar
* Significantly different from 10/79 - 11/80 period at .05 level of significance.
Significance tests of the individual weekly and monthly series are summarized in
Appendix Table 3.
1/ Standard deviations of weekly and monthly changes are means of standard
deviations of 5 series of 5-day and 21 series of 21-day changes respectively.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-54significant in the German case ..1' The ~ariability of foreign interest
rates, whether over daily, weekly, or monthly intervals, was substantially
less, however, than the variability of Eurodollar rates.
Four conclusions can be drawn from the results in Table 10.
First, the increased variability of dollar interest rates during the past '
year, which presumably at least in part is attributable to the new Federal
Reserve operating procedure, has led to an increase in the variability of
I
>
Canadian interest rates, as the Canadian authorities responded to the •increased variability of dollar interest rates.
Second, the observed in~·
crease in the variability in Japanese rates probably reflects a secular
trend toward greater flexibility in Japanese rates that has little to do'
with the new procedure or the variability of dollar interest rates.
Third,
it is of particular interest that the evidence concerning the variability
German 3-month interest rates is inconclusive.
The variability of
German 3-month interest rates increased somewhat compared with the
preceding 11-month period and declined somewhat compared with the entire
,6½-year preceding period.I/ Fourth, the results for the United Kingdom
,
and Switzerland are mixed but, on balance, lend little support to the
hypothesis that the rise in the variability of dollar interest rates
induced a rise in the variability of foreign rates.
1/ Statistically significant changes are marked by an asterisk
in Table 10 for daily rates; for the individual weekly and monthly series,
tests of signficance are presented in Appendix Table 3.
2/ One-year and five-year Euro-OM rates showed a pattern of increases compared with both preceding periods, which may have reflected the
greater variability of underlying economic and financial conditions. It
is also of some interest that the series on three-month interest rates in
France, the Netherlards,and Belgium in all but one period showed the German
pattern -- more variability compared with the shorter preceding period and
less compared with the longer period.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-55-
The relative stab1l1ty of foreign interest rates might be
thought to res.ult from act1,ons by foreign, authoriti ~s that would stab,1ize those interest rates but at the same time increase the variability
of the respective money supplies.
growth
An examination of monthly rates of
monetary aggregates for the maJor fore, gn countr, es shows 'that
,n
for most countries the standard deviation of the monthly growth rate of
M-1 relative to the mean growth rate was somewhat higher since October 1979
than during the whole per1od since January 1973 ...!./ However, these results
most likely reflect the var1abil1ty of the underlying economic situation
rather than the 1nd1rect influence of the Federal Reserve 1 s new procedure.
We further examined the relationship between U.S. and foreign
' '
interest rates by regression techniques.
Specifically, we regressed
changes in (monthly average) foreign 3-month interest rates on contemporaneous changes 1n the (monthly average) U.S. CD rate for the same three
time periods.I/ Table 11 shows the coefficient on the U.S. interest rate
2
(b) and the coefficient of determination (R) for each equation.
The
-
overall explanatory power,of the equations is very low except in the case
of the Canadian interest rate, and the coefficient on thi U.S. interest
rate is significant in only two cases aside from the three Canadian
equations.
The evidence presented in Table 11 is consistent with that
in Table 10 in that variations in U.S. interest rates apparently have
strongly influenced rates abroad, since October 6, 1979, only in the case
of Canada.
1/ Us mg Just the standard deviation to measure var, ab111 ty
yields fewer cases of increased variability, and no general pattern of
increased variability was found for the broader aggregates.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
_g_; An exploration of lagged responses did not alter the results.
-56-
Table 11
Relationship Between U.S. and Foreign Interest Ratesl/
WEIGHTED-AVERAGE FOREIGN
March 1973-September 1979
October 1979-November 1980
No~ember 1978-~eptemb~r 197~
2
R
b
.339*
.092
.220
,-
.214
.138
.094
GERMANY
March 1973-September 1979
October 1979-November 1980
November 1978-September 1979
. 181
.045
.225
.038
.040
.288
.159
.144
.044
.214
.048
.071
SWITZERLAND
January 1975-September 1979
October 1979-November 1980 •
November 1978-September 1979
JAPAN
' '
March 1973-September 1979
October 1979-November 1980
November 1978-September 1979
.053
.042
.275
.005
.008
.156
.550*
.40:fC
.349*
.431
.559
.448
.381 *
.097
.521
.055
.096
.134
CANADA
March 1973-September 1979
October 1979-November 1980
- October 1978-September 1979
UNITED KINGDOM
March 1973-September 1979
October 1979-November 1980
November 1978-September 1979
* Significant at .05 level.
JJ
Equation: change in foreign interest rate= b (change in U.S. CD rate)
Note. Monthly averages of three-mon~h interest rates were used.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-57C.
Reactions in Major Industrial Countriesl/
The announcement of ·the new Federal Reserve operating procedure
in October 1979 was welcomed by officials in other countries. Given the
relatively high U.S. inflation rate observed in the first three quarters
of 1979 and the weakness of the dollar in September, a policy initiative
that promised a more stable U.S. policy (and, therefore, helped to stabilize the dollar) and that was perceived to promise a somewhat tighter U.S.
policy was deemed appropriate. At a press conference on October 25, 1979,
shortly after the new U.S. operating procedure was announced, President
Leutwiler of the Swiss National Bank stated that Swiss authorities welcomed the new U.S. package and that the Bank would not do anything to
enQangerits success. More ,recently, the governor of a foreign central
bank remarked that, despite some problems for his country caused by dollar
interest rate volatility, U.S. authorities are doing what others had urged
them to do with respect to monetary policy.
Similarly, the increase in dollar interest rates and the rise
in the exchange Value of the dollar that followed the announcement were
not viewed abroad as a problem.
Inflation rates abroad also had been
rising and real growth was unexpectedly strong, so that some upward presure on interest rates worldwide was consistent with most countries'
domestic economic objectives. To be sure, some concern was expressed that
interest rates had risen too far.
In late 1979, representatives of some
1/ The principal authors of this and the following section were
Karen H. Johnson and Larry J. Promisel with the assistance of their colleagues at the Board of Governors and the Federal Reserve Bank of New York.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-57-
c.
Reactions
,n
Major Industrial Countriesl/
The announcement of the new Federal Reserve operating procedure
in October 1979 was welcomed by officials in other countries. Given the
relatively high U.S. inflation rate observed in the fi'rst three quarters
of 1979 and the weakness of the dollar in September, a policy initiative
that promised a more stable U.S. policy (and, therefore, helped to stabilize the dollar) and that was perceived to promise a somewhat tighter U.S.
policy was deemed appropriate.• At a press conference on October 25, 1979,
shortly after the new U.S. operating procedure was announced, President
'
Leutwiler of the Swiss National Bank stated that Swiss authorities welcomed the new U.S. package and that the Bank would not do anything to
engangerits success. More recently, the governor of a foreign central
'
bank remarked that, despite some problems for his country caused by dollar
interest rate volatility, U.S. authorities are doing what others had urged
them to do with respect to monetary policy.
Similarly, the increase in dollar interest rates and the rise
in the exchange value of the dollar that followed the announcement were
not viewed abroad as a problem.
Inflation rates abroad also had been
rising and real growth was unexpectedly strong, so that some upward presure on interest rates worldwide was consistent with most countries•
domestic economic obJectives. To be sure, some concern was expressed that
interest rates had risen too far.
In late 1979, representatives of some
1/ The principal authors of this and the following section were
Karen H. Johnson and Larry J. Promisel with the assistance of their colleagues at the Board of Governors and the Federal Reserve Bank of New York.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
of the smaller European countries expressed the v1ew that the worldwide
interest rate structure was. becoming too high.
However, the focus of
their attent1on was not the United States, but Germany, where short-term
1nterest rates had -risen from around 4,percent 1n early 1979 to 9½ percent
in November-December.
See Table 12.l/
The per1od February-May 1980, when 1nterest rates on dollar
assets first rose and then declined very sharply, posed more significant problems for other countries, especially those of continental
Europe.Y
The upward pressure on dollar exchange rates that resulted from
the rise in dollar interest rates in February-March was resisted partly
by a rise in foreign interest rates and partly by heavy intervention
in the foreign exchange markets.
The weighted average foreign interest
rate rose about 150 basis points from January to March-April while dollar
rates rose almost 450 basis points from January to March.
Total net
foreign intervention sales of dollars were substantial in March.
When dollar interest rates subsequently fell 750 basis points
by May, foreign interest rates on average declined only slightly (on
the order of 50 basis points).
Foreign monetary authorities sold dollars,
net, in April, but in May there were net intervention purchases.
1/ All interest rates cited in this section are monthly averages of
daily quotations on three-month rates.
Y The behavior of U.S. interest rates during this period, especially
after March, reflected to an important extent the influence of the March
credit restraint program and not the new operating procedure. We ignore
this aspect in what follows.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-59Table 12
Three-Month Interest Rates
(Interbank loan or nearest equivalent, average of daily rates)
Period
NetherSwitzer- United
Kingdom
lands Sweden land
10
7
8
9
Weighted
average
foreign
11
'
U.S.
CD s
12
Eurodollars
13
Belgium
l
Canada
2
France
3
Germany
4
Italy
5
Japan
6
1979-Jan.
Feb.
March
April
May
June
July
Aug.
Sept.
Oct.
Nov.
Dec.
8.93
8.22
7.63
7.63
8.16
9.09
11. 18
11. 42
11.88
12.99
14. 17
14.49
l 0.87
10.94
11.08
11. 18
11.26
11. 17
11. 29
11.78
11. 89
13.34
14. 19
14.02
6.55
6.83
7.05
6.96
7.63
8.63
9.90
10.85
11. 67
12. 14
12. 72
12.55
3.85
4. 13
4.42
5.50
5.89
6.40
6. 77
7.03
7.82
8.84
9.57
9.54
11.12
11.38
11.45
11. 52
11.37
11. 27
11 .46
11.50
11. 51
12. 71
13. 13
16. 01
4.52
4.50
4.55
5. 13
5.25
5.46
6.26
7.00
7.00
7.01
8. 13
8.42
8.69
7.42
7.35
7.23
7.82
8.55
9.53
9.51
9.82
10.09
11.86
14.56
5.84
5.84
5.84
5.84
5.83
5.83
6. 14
6.60
6.60
7.06
9.03
9.74
0.05
0.13
0.93
0.93
1.54
1.51
1.19
1.66
1.94
2.57
3.97
5.67
12. 61
13.28
11.98
11.64
11.76
13.02
13. 87
14.06
14. 11
14. 12
16.09
16.74
5.98
6.03
6.08
6.86
7.29
7.76
8.37
8.86
9.26
9.94
11.12
11.70
10.51
10.19
10.13
10.06
10.16
9.95
10.11
l 0. 71
11.89
13.66
13.90
13.43
11. 16
10. 79
10.64
10.60
10. 75
10.52
l 0.87
11.53
12.64
14. 59
15. 00
14. 51
1980-Jan.
Feb.
March
April
May June
July
Aug.
Sept.
Oct.
Nov.
14.38
14.45
16.23
17 .10
16.31
14.69
13.30
12. 52
12.35
12.24
12.40
13. 93
13.96
14.72
16. 31
13.23
11.73
10.91
10.47
10.73
11 . 71
12.96
12. 31
12.63
13.94
12.84
12.62
12. 37
11. 87
11. 20
11. 81
11. 69
11. 26
8.79
8.94
9.51
l 0. 12
10.18
10.00
9.56
8.93
8.90
8.99
9.37
17.00
17.88
18. 12
16.92
17.20
17.25
17.49
17.30
17. 50
18.16
17. 51
8.44
9.10
12.37
13. 51
13.63
13. 51
12 .89
12.04
11.46
10.98
9.74
11. 85
11.99
11.48
10.76
11. 18
10. 72
10.06
9.97
10.31
9.63
9.59
10.79
10.79
10.79
10. 78
12.89
12.89
12.89
12.89
12.84
12.84
12.90
5.45
_ 5. 19
6.57
6,87
5.85
5.64
5.29
5.52
5.57
5.40
5.53
17.30
17. 72
18.07
l 7. 70
16.97
16.68
15.82
16.45
15. 89
15.87
15.84
11.44
11. 77
12.86
13. 05
12.72
12.40
11. 81
11. 42
11 .43
11 . 41
11. 35
13.39
14.30
17.57
l 6.14
9.79
8.49
8.65
9.91
11. 29
12.92
15.68
14.33
15.33
18. 72
17. 81
11. 20
9.41
9.33
10.82
12.07
13.55
16.46
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1
I
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-60In June and July, dollar interest rates declined an
additional 200 basis points, and foreign interest rates on average
declined
100
basis points.
Although dollar interest rates began to
'
rise again in August, foreign interest rates on average declined somewhat further.
From August through November foreign interest rates
remained essentially unchanged on average -- although OM rates edged
up after late October, and yen and sterling rates eased -- while dollar
interest rates again rose sharply.
In the four months ending in
November, the dollar appreciated by almost as much as it did earlier
in the year, but net foreign intervention sales of dollars were smaller.
I
Given Germany's dominant role in Europe, the German policy
reaction is a central element in the general reaction in continental
European countries to U.S. economic developments in general and the
new,Federal Reserve operating procedure in particular. As noted above,
I
I
interest rates ih Germany had risen significantly during 1979, and in
I
the first quarter of
1980
the Bundesbank saw no reason to relax its
I
relatively restr1ctive stance.
Economic activity remained (surpris-
ingly) strong, with no evident signs of the generally forecast slowdown.
Inflation rates were high and the rate of increase of producer
prices, a leading indicator of consumer prices, was not letting up.
Central bank money growth was near the upper limit of the target
range announced in December 1979.
The German current account was
recognized as on its way to a record deficit.
Thus, when dollar
I
-
interest rates rose, Gennan'•'fnterest ,ratesf s'eemed'=to 1ris'e -in"response,
but the rise in Gennan int~~est~rates~seemed 2to be 'tiased on domestic
considerations, as was noted ~by the Bun'desbank at the· 'time.
This 1atter
view is reinforced by the fact that Gennan interest rates' did-not decline
significantly from April to :Ju'ne 'despite the 'pllmge'"in dollar rates.
In late 1980, in th~~5bs~ht~ of-rising tloll~~'interest •
1
rates and the consequent exch ange 'rate impl'ica'tions, Gennan autho'rities
appeared to have wanted to pe'rmi1: interest rates to aecrine further.II
This would have been clearly 'in 'line with domestic economic conaitfons·;
economic activity had been un~xpectedly weak -- industrial output,fell 1
substantially during the second and'·'th'ird 'quarters:,_ inflation rates·
declined, and central bank money 'growth was below the lower end of the
5-8 percent target range for 1980.
The pattern of Gennan in'terest rates seems to have influenced other continental European coun'tries more than did u.s:, rates.
That·
1
is a natural result of the dominant position of Gennany_ wi_thin-Europe'.-,
To the extent that the European Mone'tary System·comriiits'its~merfibers to·
greater exchange rate stability, th~'EMS constrains·independent mone-tary po 1icy further and thus enh'~fices' Germany I s-domi'nanc:e.
a speech at Pforzheim on Octbb~r' 20'~1 ok Scn~~singer•, Vice' Presi :.dent of the Bundesbank, said "Th~~faeirthai1 th~~C~nlra1'B~n~ Cbancil 1
at its last meeting in Berl in didr n8t 1 detide-tol'low'erf'the"Bank''
s'
1
interest rates further was due firstCan8 fore\nost~to'·the-n'arrowJlimitss
imposed by the present external ecbnom1crcons'te'llation,on' the"'scope~or
our interest rate policy."
l/ In
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
0
-62The Bank of Italy introduced its new bank credit enforcement
procedures in March.' Interest rates rose at that time and remained
high during the spring; in June the Bank of Italy announced additional
credit tightening. The March action had been under consideration for
some time, as a means to fight inflation and discourage stockpiling,
especially of imports; there was widespread recognition that the credit ceilings in force were being exceeded. The timing of the Italian
package with the U.S. credit restraint program apparently was
coincidental. The June action came in response to pressure on the
lira within the EMS.
Similarly, it appears that the behavior of Dutch and Belgian
interest rates in 1980 -- to the extent that they deviated from
what would have been desired purely on domestic grounds -- were
determined primarily by German rates.
Only to the extent that German
rates, in turn, reflected dollar rates can one attribute changes inDutch and Belgian rates 'to the new Federal Reserve-operating procedure~
The same can be said of French interest rates.
However, tn
France more than in most other countries~ the.volatility,of interes~
0
rates (as distinct from the level) has been an important issue, as_
well.
Since the French banking system and financial structure are based
on large amounts of refinancing at the central bank, and since French
industry is heavily reliant on fairly short-term bank financing; ,the
economy is seen to be very sensitive to changes in inter~st rates.,~,,
The inertia that many economies have due to the fixed nature of a
large share of capital costs is not present in France unless interest
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-63-
rates are stable.
Thus, despite pronounced changes 1n world interest
rates, French interest rates were held fairly steady.
the whole year
1980
Indeed, over
the range for the Bank of France's money market
1ntervent1on rate (the key rate 1n the French system) was less than 2½
percent.
The continued strength of the French franc within the EMS
made it possible for the Bank of France to pursue the above policy despite
whatever impact it may have had on the franc-dollar rate.
However, the
downward adJustment in early November in the Bank of France's intervention rate was precipitated by the weakness of the German mark within
the EMS arrangement, which, in turn, was attributed to a rise in dollar
interest rates.
In sum, authorities in continental European countries were
affected by the new operating procedures; they were affected by both
the higher level and, to a much lesser extent, the volatility of U.S.
interest rates.
However, the problems caused were not great, given
that internal and external obJectives were broadly consistent.
Any
problems stemmed primarily from German policy actions and conflicts
and were thus at most only indirectly related to the Federal Reserve's
new operating procedure.
Three other countries also merit discussion.
Canada because
it was the country most clearly influenced by U.S. policies, the
United Kingdom because it was least influenced, and Japan because it 1s
so big.
The Bank of Canada announces targets for the growth rate of
M-1. These targets, to which the Bank of Canada tries to adhere quite
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-64-
systematically, determine in principle the path of Canadian interest
rates.
Nevertheless, as noted above, Canadian interest rates are
correlated significantly with U.S. rates.
In a press release accompany-
ing the increase in the Bank of Canada's Bank Rate on October 9, 1979,
I
Governor Bouey cited the riew Federal Reserve opetating procedure as
an important factor behind the Bank Rate action.
Governor Bouey argued
that a depreciation of the Canadian dollar would have added to inflation without benefiting the real economy, since the key export sectors
l
(excluding automobiles) were at full capacity.
The perceived need to
follow U.S. interest rates persisted in 1980.
In addition, the sharp swings in U.S. interest rates in the
spring caused problems for the Bank of Canada.
Rather than suffer
the announcement effects that would have accompanied frequent changes
in the Bank Rate (which must change with market rates to keep it a
penalty rate), the Bank of Canada gave up its discretionary setting of
''
the Bank Rate and on M~rch 13 tied it to market rates.
Subsequently,
when U.S. interest rates fell the Bank of Canada tried to moderate
the Canadian interest rate decline; .!!the Bank felt that the U.S. decline would be reversed (at least partially) and wanted to avoid an
excessive swing in Canadian rates, which Canadian authorities viewed
as undesirable.
Similarly, when U.S. interest rates increased again
during tne fall, Canadian rates rose but less rapidly and by a smaller
amount. Analysis of Canadian exehange market interventi~n behavior
conducted 1n connection with the investigation summarized in Section III.C
yielded no evidence of a change 1n behavior since October 1979.
l!From April to June, 3-month interest rates declined about 450 basis
po1nts 1n Canada, while U.S. 3-month CD rates decl1ned 750 basis points.
Again from July to November, 3-month interest rates rose about 200 basis
po1nts 1n Canada, while U.S. 3-month CD rates rose 700 basis points.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-65-
The new Federal Reserve operating procedure almost certainly
did not significantly affect British policy.
British macroeconomic
policy throughout the past year has been concerned with achieving a
monetary growth strategy upon which the government had embarked
before the Federal Reserve's October announcement.
However, the Fed-
eral Reserve's new operating procedure has become involved in the
internal U.K. debate on techniques of monetary control.
In Japan, various monetary policy actions were adopted last
November and again in February-March.
These actions were linked both
to domestic inflation and to strong selling pressure on the yen that
developed when dollar interest rates rose.
However, one must be care-
ful about drawing any strong conclusions about the effect on Japan or
Japanese policymaking of the System's new procedure on the basis of
these two episodes.
It 1s certainly the case that the yen's value
has been quite variable over the past year, but much of that
variability appears to reflect the behavior of Japan's external accounts
and other-events not directly related to U.S. polic1es ..l/ Nevertheless,
some tendencies in the Japanese policy response have developed that may
bear watching in future episodes.
The Japanese authorities are inclined to reserve their conventional monetary measures -- i.e., discount rate changes, adjustment
in money stock growth, the use of credit controls -- for domestic
.l/The results reported in Section III.A for the yen exchange rate (spot and
forward) against the dollar point most consistently in the direction of an
increase in variability for that currency.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-66-
objectives, such as reducing inflation.
The important implications
for monetary policy of external events -- including capital flows and
exchange market pressures generated by international interest rate
differentials -- are, of course, acknowledged.
However, the preferred
response to such external influences seems to be to rely on official
exchange market intervention and capital controls -- in the latter case,
by relaxation first of controls currently in place to encourage capital
flows in the (officially) desired direction and in extremis by placement of new controls to limit capital movements.
This principle of policy assignment appears to have been
maintained roughly intact during the recent episodes.
Although the
successive tightening of monetary policy between October 1979 and March
appears to have been directed largely at control of inflation, it is
noteworthy that the effect on the yen has been given a more-than-usual
prominence in official characterizations of the measures.
Whether this
development constitutes anything more than a minor innovation remains
to be seen, however, as the tightening was consistent with both internal
and external objectives at the time.
It is interesting to note that
the Bank of Japan allowed interest rates to decline significantly
from August to November although dollar interest rates were rising.
The relative stability of the yen against the dollar during this period
may have contributed to this relaxed attitude.
Nevertheless, the Jap-
anese also appear to be willing to tolerate fairly wide swings in
interest differentials and in the yen's exchange rate before resorting
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-67to extraordinary measures, such as capital controls
There is some
evidence, however, that the Japanese authorities have felt that their
economic policy decisions have been complicated by the greater unpredictability of dollar interest rates which they associate with the
Federal Reserve 1 s new operating procedure.
D.
Reactions in Developing Countries
The impact of the new Federal Reserve operat~ng procedure on
the problems confronting policymakers in the developing countries
during the past year was somewhat different from that discussed above
for the developed countries.
As in the industrial world, inflation in-
creased sharply during 1979 and 1980 in most developing countries.
At
the same time, the high level and the variability of dollar interest
rates on world capital markets created diff1cult1es for those developing countries that are maJor commercial borrowers internationally.
In
addition, local institutional problems arose 1n some of the developing
countries in Latin America and elsewhere whose exchange rates are
pegged to the dollar.
The generally high level of world interest rates during the
first months of 1980 raised the cost to developing coun~r,es of funds
obtained abroad and added to their balance of payments problems.
Since
the need for tighter monetary conditions and higher interest rates at
that time was felt in many of the developed countries, it is not clear
to what extent the new operating procedures~ se added to the burdens
of the developing countries.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
At least one country, Brazil, ceased new
-68-
borrowing during early 1980 1n part because the authorities viewed
interest rates at that time as excessively high.
It should be noted,
however, that the interest rates on most such borrowings float, so
Brazil would not have been obli_ged to continue high interest payments
once world rates came down. Moreover, the 1nternation~l ·reserve~
Brazil expended in lieu of additional borrowing would have earned the
current market rate had they remained invested.
The increased variabilitJ of interest rates in world capital
markets creates additional uncertainty for borrowing countries 'as ·theyattempt to plan government budgets and manage their balance of payments flows.
More experience with the new procedure and its impacts
may enable them to 'forecast better their expected borrowing costs over,
for example, a year or more.
Since many developing countries have fixed or crawling-peg
exchange rate regimes, the increased variability of,world interest
rates has the potential for creating large, unwanted, short-term capital flows and disintermediation:at local financial institutions.
Some countries (Venezuela and Mexico) prevented such flows during ~980
by allowing greater flexibility of domestic interest rates. The induced
movements in domestic rates were welcomed by author,it1es in Mexico,
.but less so in Venezuela, because of differences in the underlying
domestic economic situations at the time. Many other developing
countries, however, have fixed (or at least somewhat inflex:ible-)
domestic interest .rates rand were forced by the rise in world interest
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-69-
rates to take actions they would most likely not otherwise have taken.
Some, such as the Dominican Republic and Guatemala, itensified exchange controls.
Brazil slowed the rate of crawl of its exchange
rate -- even though its inflation rate rose -- and introduced a tax
on domestic borrowings.
Thailand, which had flexible domestic interest
rates but legislated ceilings on them, changed some financial market
regulations and taxes and adjusted the ceilings in order to prevent
large capital outflows.
Many authorities of developing countries re-
gard stable interest rates as necessary for the health of domestic
financial institutions and to promote sustained levels of domestic investment.
They view as troublesome a change toward greater volatility
of interest rates on world markets.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-70,
I
Section V -- U.S. International Capital Flows-'
The Federal Reserve 1 s adoption of its new operating procedure
could have 1nfluenced the structure of U.S. international capital flows
by changing pr1vate investors• expectations concerning, 1n particular,
interest rates, exchange rates, or inflation rates.
Changes in private
investors• expectations would be manifested in revised portfolio preferences Hhic~,
rn turn, would be observed as net private capital flows
(including changes in the statistical discrepancy) to the extent that
central banks chose to accommodate such revised preferences through intervention in the foreign exchange market rather than allowing exchange
rates to absorb the changes.
During the four quarters commencing in October 1979 and ending
in September 1980, the G-10 countries reduced their net reserve asset
holdings in the United States by $15.2 billion.
See Table 13.
This
official net capital outflow was accounted for by four countries
Germany, Switzerland, Japan, and Italy. In addition, the reserve assets
of the United States increased (an outflow) by $4.5 billion during this
period.Y With the exception of Italy, most of this activity took
place during the fourth quarter of 1979 and the first quarter of 1980,
an interval of time over which the weighted-average exchange value of
the dollar appreciated on balance about 10 percent.
I/ The principal contributor to this section was Patrick Par~1nson
Z/ About $2.2 billion of this total consisted of Jn SCR al1ocation and-Carter notes issued in the first quarter of 1980.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Table 13
U.S. International Capital Flows
(Billions of dollars, outflow=(-), not seasonally adJusted)
1979Q4
l 980Ql
1980Q2
1980Q3V
Change in net foreign posif/ions
of banking offices in U.S.-
-5.0
9.0
-23. l
-12. l
Net private securities transactions
- 1. l
4.9
-2.0
Change in foreign official
reserve assets
- .3
-7.4
G-10 countries and Switzerland
OPEC
All other countries
-7.2
6.0
.9
-10. 7
,n U. S,
... 6
-3.3
.5
-7.9
- l 1. l
-5.8
All other transactions
6.0
.8
Stat1st1cal discrepancy
8.9
.5
Change
reserve ajisets
Trade balance
Memo:
Current-Account Balance
1978Q4-1979Q3
-31. 2
6.4
-0.8
1.0
2.8
7.0
7.7
7.0
3.5
1.3
4.3
1.4
1.4
3.9
2.4
-15.2
17.5
4.7
1.7
1.4
.4
-4.5
- .3
-5.8
-30.6
-27 .3
4.7
3.8
15.3
.3
7. l
18. 7
8.3
43.0
14.6
-2.5
- .7
.5
-2.2
- l .5
3.3
*
*/ Less than 50 m1ll1on.
1/ Excluding liabilities to foreign official inst1tut1ons.
£1 Prel1mrnary.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1979Q4-1980Q3
,.~
-
I
-72An examinatfon ,of the intervention behaYior of monetary
authorities 1 rep0r,ted in Section :III.C of this study, found little
evidence that the offici•al capital outflows starting in the fourth
quarter of 1979 resulted from significant changes in historical
patterns of i'ntervention.
If no change occurred in interv~ntion
behavior f.o~1owing the adoption of the new operating procedure by the
Federal Reserve, and' there was . no change in the average level ot' the
exchange ,rate as a consequence of th~ new procedure, the new procedur~
,
could not have had a large impact on net private capital flows.
The '
sum of private and official capital flows (includi'ng the statistical
.
'
.
'
discrepancy) is by definition the mirror image of the balance on current
transactions. The current-account balance is largely predeterm1ned in '
the sHort run by past developments in real exchange rates and the levels
of U.S. and foreign economic_activity.
Hence, the sum of net private
and official capital1 transactions is largely predetermined.
To the extent that the ,adoption of the new procedure did not
affect the average level of the exchange rate after October 1979, but
the new procedure did affect the month-to-month variab1lity of the
exchange rate, one would expect to observe periods of substantJal
-:,
off1c1al capital 1nflows and outflows even if intervention behavior were
-
unchanged.
For example, from the end of December 1979 to the end of
March 1980, the dollar appreciated by about 8 percent. The examinat1on
of intervention behavior 1n Section 111.C suggests that a 1 percent
appreciation of the,dollar results in between $0.9 b1llion and $1.5
b1llion in intervention sales of dollars by the ~aJo~ G-10 countries
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-73(1nclud1ng the United States). This 1s consistent with their recorded
volume of net intervention and net official capital outflow 1n the first
quarter of 1980. See Table 13. Of course, given this net official
outflow and an essentially unchanged current-account position,l/ net
private capital flows had to compensate, as they did.
To the extent that the new operating procedure resulteq 1n
slower growth in U;,S. real" economic activity, the U.S. current-account
deficit, assuming an unchanged exchange rate, was smaller and the sum
of net private and official capital inflows was smaller. 2/ To the
extent that the dollar appreciated, this would have reduced the current
account deficit somewhat further in the short run (J-curve effect}.
The net effects on U.S. international transactions would depend
on the size and timing of the effect on the current account, on the
size of any tendency for the dollar to appreciate, and on the vigor
of any intervention response to such an apprec1at1on.
It 1s likely
that for tne U.S. economy over a period as short as one year the
increase 1n the official capi'tal outflow associated with the stronger
dollar would more than compensate for the lower U.S. current-account
deficit~
''
1/·In fact, the current account did move into deficit 1n
the first-quarter reflecting sharply higher prices for imported 011.
2/ It should·'be ·emphais1zed that the discussion in this
paragraph-concerns monetary policy results that might have been
brought about by the new procedure rather than the o,peration of
the procedure itself.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-74Even in the absence of any exchange market intervention induced
directly or indirectly by the change in operating procedure, the change
could have affected the structure of private capital flows.
As ,shown.in
Table 13, in the four quarters following adoption of the new operating
procedure, the notable elements in the private component of the capital
account were a reported net outflow of $31.2 billion from banking offices
-in
the United States, reversing a trend- of net j-nf-lows that- began in-_----
early- 1979, and a $17.5 billion increase (inflow) in the reserve assets
held in the United States by tne OPEC countries.
In addition, the
statistical discrepancy totaled $43 billion during this period. The
most likely source of the discrepancy is net unrecorded private capital
inflows.
These developments probably were caused by factors other than
but coincident with the change in operating procedure.
One factor that
clearly influenced the structure of private capital flows was the
imposition, at the time of the change in operating procedure, of a
marginal reserve requirement on Eurodollar borrowing and other managed
liabilities of member banks and U.S. agencies and branches of foreign
banks.
The introduction of this marginal reserve requirement had a
pronounced effect on the observed capital-account transactions of banks.
In particular~ U.S. agencies and branches of foreign banking offices
,..
'
found it profitable to reduce their liabilities to foreign banking
r
'1 (~ "fl
t'("
I'
r
,
;
1
.-
I
-
_, ...,.
I
..,
.. .,
Ir
\
,/''1
1
("
""""l:"'I If'
offices and, in turn, shift the booking of loans to nonresidents to their
offshore offices. The outflow from U.S. banking offices totaled $5.0
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
'<
?..r.t-,_,._..., -. ~.t"'
billion in the fourth quarter of 1979. Jn early 1980 the banks were
large1y abl~ to avoid the 'program's impact. The subsequent tighteni'ng
of the program in March 1980 partially account~ for the $23.l billion
outflow nrom U.S. banking offices in the second quart~r of 1980.
)
A second factor that posstbly influe~ced the structure of
pr1vate capital flows was the freezing of Iranian assets in the United
States and in foreign ,branches of U.S. banks by •President Carter in
November 1979. This action may have discouraged nonresidents from holijing financial assets at domestic ana foreign offices of U.S. banks or
otherwise in the United States where their owner-ship was identifiable.
Such a response by nonresidents would have reduc~d ban'k-reported capital
inflows and also might hav~ increased the volume of unrecorded capital
.inflows.
A third factor that possibly influenced the structure of
private capital flows was the anticipati·on and implementation of the
U.S. credit restraint program. This program may have induced partially
unrecorded, roundtrip capital flows as U.S. borrowers went abroad to
borrow funds, through unrecorded channels, including funds that,had been
• deposited abroad -by U.S. residents.
In the second quarter of 1980, the
'
period when this program had its impact, the statistical discrepancy
indicated unrecorded inflows total mg $18.7 billion.
Finally, as illustrated in Chart 4, since October 1979 there
has been a sharp increase in the variability of the differential between
the interest rate on overnight Eurodollar deposits and the federal funds
rate -- weekly average in both cases. Again most of this increase
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-76CHART 4
OVERNIGHT EURO-DOLLAR AND FEDERAL FUNDS RATES
Weekly series
Percent per annum
20
I
)
l\
l
l
l
ll '
1•
I
16
OVERNIGHT EURO-DOLLAR DEPOSITS
12
8
4
2
DIFFERENTIAL
+
--------------~~P-¥----------t-tt---t---t--:-H--..,..-Hr---ttt---0
2
4
1977
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
1978
1979
1980
-77reflects the imposition (October 1979), avoidance (end 1979), tightening
(March 1980), and elimination (July 1980) of the marginal reserve program.
However, a residual amount may have reflected the influence of the new
operating procedure~ se through its effect on the variability of the
federal funds rate.11 The increase in the week-to-week variability of
the federal funds rate, in turn, appears to have discouraged weekend
Eurodollar reserve avoidance activity somewhat. A bank does not realize
its reserve reduction until two weeks after it has paid its accomplice
a share of the expected gain from avoiding reserves.
The increased
variability of the federal funds rate increased the variance of that
)
expected gain.
In summary, no significant change in the intervention behavior
of monetary authorities has been observed since the adoptiQn of the new
operating procedure.
Given that the current-account balance is largely
predetermined in the short run, this implies that the new procedure has
had no direct effect on net private capital flows on average since October
1979.
Changes in the structure of private capital transactions and in the
statistical discrepancy have been observed, but these changes can be largely
accounted for by other factors.
However, to the extent that the new procedure
may have indirectly facilitated a tighter Federal Reserve policy on average
over the period, 'an appreciation of the dol,lar and a reduction
current-account deficit may have resuJted.
1n
the U.S.
Under such circumstances, one
would expect an increase in the net official capital outflow to G-10 countries
1/ This, phenomenon is another form of roundtrip flows, i.e.,, would
not affect-net flows.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-78and a smaller increase in the net private capital inflow (recorded and
unrecorded).
However, over periods longer than a year the s1ze and
direction of these influences depend on many other factors as was discussed
1n Section II above.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-79Section VI -- Concluding Topics
A.
Consequences of Exchange Rate Variability
The evidence presented in Section III of this study indicates
that the change -in operating procedure probably led to some increase in
the short-run variability of dollar exchange rates.
On the assumpt,on
that this phenomenon persists, the question arises as to whether the
additional variability attributable to the new procedure can be expected
to have any perceptible adverse economic and financial effects on the U.S.
economy.
To provide an answer to this question, we examined the possible
impact of increased variab1lity of exchange rates on trade flows, on
foreign direct investment and domestic fixed investment, on domestic
prices, and on the attractiveness of dollar-denominated assets for private
and official holders.
Our review of the existing literature, as well as
our own examination of some of the data, did not uncover any evidence suggesting that these effects are likely to be strong and significant.
l.
Impact on Trade Flows l /
A number of contributions have been made to the theoretical and
{
empirical literature on the effects of exchange rate variability on export
and import volumes and prices.
The theoretical work in this area, e.g.,
Clark (1973) and Hooper and Kohlhagen (1978), indicates that an increase
in the variance of nominal exchange rates per se will reduce the volume
- of international trade if firms are risk averse.
Empirical verification of this effect to date has been only
partially successful.
Four studies {Clark and Haulk {1972), Makin (1976),
Hooper and Kohlhagen (1978), and Kenen {1979)) found no significant
negative effect of exchange rate variability in equations explaining
1/ The principal contributor to this and the following section
was Peter B. Clark.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-80-
trade volume, although the next-to-last study did find a significant
impact on prices of traded goods. ,A study Pl.l,blished by GATT (Blackhurst
and Tumlir (1980)) found no sign1ficant change 1n the ratio of the growth in
world exports to the growth in world output
- over
. the past 25 years.
One
recent study (Abrams (1979)), however, has found a signif1cant negative
effect of exchange rate variability on trade volume us,ing annual crosssection and t1me-ser1es data.
In addition, another study (Cushma,n, (1980))
using a methodology similar to that of. Hooper and Kohlhagen did_find some
evidence of a negative impact on trade volume.
Nevertheless, given the
difficulties in interpreting the results of these studies it seems reasonable to conclude that there is no firm evidence relating adverse effects on
trade flows to exchange rate variability.
One reason these studi~s may,not
have uncovered much evidence is that they may have underestimated th~ time
period over which an increase in exchange rate variability must be recprded
in order to affect trade flows.
Based on our review of the literature, we would conclude that
the adoptipn of the new operating procedure in October 1979 has not had
a significant negative impact on the volume of U.~. trade flows to date.
First, the results presented in Section III.A do not indicate an un,
ambiguous increase
,n
'
exchange rate volat,il ity beyond ,the short run
'>
(measured in weeks) in the post-October 1979 period. There is also no
strong empirical evidence that increased exchange ~ate variability has
a significant negative impact on trade flows, and the large exchange
rate movements 1n 1980 were concentrated in a few months, which was
'
probably too short a time period to affect the longer-run c~nsiderations
that presumably influence the extent to which'a firm will engage 1n
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-81~
1nternat1onal trade.
An additional piece of information is that, since
the adoption of the Federal Reserve's new operating procedure, the
equations used by the Board's staff to help to forecast the volume of
non-011 imports and non-agricultural exports have underpredicted the
level of real trade flows, with the exception of the prediction of the
volume of non-agricultural exports in the fourth quarter of 1979.
These
in-sample errors can be viewed as weak evidence that exchange rate
variability (a variable that does not appear in the equations) has not
had a negative impact on U.S. trade.
2.
Impact on Foreign Direct Investment and Domestic Fixed Investment
Very little theoretical or empirical work exists on the pos-
sible effects of exchange rate volatility on foreign direct investment
or domestic fixed investment.
Therefore, firm conclusions on the basis
of existing work are not possible at this point.
On the theoretical side, one might conJecture that an increase
in exchange rate variability would reduce the level of foreign direct
investment to the extent that the exchange risk cannot be directly hedged
or is not offset by variations in other prices.
However, the only direct
examination of this question, that by Cushman (1980), indicates that the
impact of increased exchange rate variability on foreign direct investment is ambiguous when a firm can substitute foreign investment for
exporting in order to exploit a foreign market.
In other words, 1t is
possible that a firm might engage in foreign direct investment and produce
abroad rather than export from the home market in the face of greater
exchange rate volatility.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
In his empirical work Cushman does in fact
-82-
find some weak empirical evidence indicating that foreign investment
may be positively affected by fluctuations in exchange rates.
The evidence is also very scanty regarding the effects of
exchange rate volatility on domestic fixed investment.
From a theoret-
ical point of view one might expect that such volatility could reduce
domestic investment for at least two reasons: (1) larger exchange rate
fluctuations could increase the variance of both domestic and/or foreign
sales, thereby increasing the risk associated with the profits arising
from a given level of fixed investment, and (2)greater volatility in
exchange rates could increase the variance of both input and output
prices, and to the extent that these price movements are not offsetting,
increase the variance of the profit stream associated with domestic investment. These considerations were among the reasons for the formation
of the European Monetary System, which has as one of its objectives a
reduction in exchange rate uncertainty among the major European currencies.
The empirical evidence linking exchange rate variability
directly to the level of domestic investment is limited to one study by
Kenen (1979).
He finds some weak evidence of a negative impact of such
variability (the average of the absolute monthly change fo real and
nominal exchange rates over a 36-month interval) on a country's real
gross capital formation.
As Kenen himself admits, however, these
results are far from definitive since in the crp$s-country regressipn
for 16 advanced countries the only explanatory variable~ he uses are
measures of the trend and variability in exchange r,ates.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-83-
3.
Impact on Domestic Prices: Ratchet Effects 1 /
The ratchet hypothes1s states that domestic prices rise when
a currency depreciates, but do not fall (or do not fall proportionately
as much) when the currency appreciates, resulting in net inflationary
pressure when the currency fluctuates.
The results of our empirical
tests of this hypothesis for the-United States were mixed.
Based on
quarterly data over the floating rate per1od to date (1973 Q3-1980 Q2),
weak evidence of a ratchet effect in the impact of the exchange rate on
U.S. non-oil import pr1ces was found.
However, no significant direct or
i~direct l1nk between this ratchet effect and domest1c prices was evident.
Recent published work on ratchet effects 1s limited to a study
by Morris Goldstein (1977),.
Goldstein tested for ratchet effects in the
impact of fluctuations in aggregate import prices on domestic prices in
five industrial countries using annual data over the period 1958-73.
In
a model relating changes in U.S. domest1c prices (GDP deflater) to changes
in wages (or unemployment), productivity,and import prices he obtained
mixed results -- under some specifications of the model ratchet effects
were found, and under others they were not.
However, Goldste1n 1 s work is not directly relevant to the
question of ratchet effects with respect to exchange rate movements since
his model does not test directly the relat1onships between exchange rates
and either 1mport prices or domestic prices. Moreover, his emp1rical
analys1s covered a per1od of relative stability in exchange rates.
In our analysis we tested for the existence of ratchet effects
at three different levels. The first test (I} used an import price model
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
]._/ The principal contributor to this section was Peter Hooper.
-84-
relating changes in nonoil import prices to changes in (1) a weighted
average of foreign consumer prices, (2) an index of world coffee and
sugar prices, and (3) the weighted-average value of the dollar. The
second test (II) involved a domestic price model that relates changes
in (alternatively) the absorption deflater and the CPI to changes in
(1) domestic unit labor cost, as measured by the domestic wage rate
divided by a 5-quarter moving-average index of productivity, (2) the oil
import price, and (3) non-oil import prices. The third test (III)
employs the same domestic price model but with changes in' foreign prices
and the exchange rate substituted for non-oil import prices.11
The tests for ratchet effects were performed including in the
models an additional exchange rate variable times a 1, 0 dummy variable
which took the value l when the dollar depreciated and O when it appre'
ciated.Y Given that the exchange rate is expressed in terms of foreign
currency units per dollar, the expected sign of its coefficient in the
price equations is negative. The existence of a ratchet effect would be
indicated by a significantly negative coefficient on the additional
exchange rate variable.
The results for the import price model (I) are summarized in
Table 14. Equation IA, which excludes the test for a ratchet effect,
shows significant current and lagged exchange rate coefficients w~th the
expected sign.
In equation IB, the last two coefficients indicate the
l/ In each test it is assumed that macroeconomic policy is unaffected, or is affected symmetrically, by exchange rate depreciation or
appreciation. For example, if higher inflation associated with depreciation
leads to faster money growth and lower inflation associated with appreciation
does not lead to slower growth, the tests are -biased in favor of finding a
ratchet effect.
2/ In the second model the 1, 0 dummy variable was applied to the
non-oil import price variable, taking the value 1 when these import prices
rose.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
/
-85Table 14
Tests for Ratchet Effects 1n the Impact of
Exchange Rate Changes on U.S. Non-oil Impor,t Pr1cesll
(Estimated coefficients; t-ratios in parentheses)
Explanatory variables
IA
-2.3
(-1.83)
-2.4
(-2.04)
%8Fore1gn consumer prices
l. 90 (3.67)
1.71
(3.38)
%8World coffee-sugar Price
.09
(3.41)
.08
(3.09)
%8Exchange rate
-.35
(-2.82)
-.48
(-1. 97)
%8Exchange rate
(t-1)
'
-.46
(-3.63)
-.02
(-.07)
Constant
%8Exchange rate{depreciation
only)
IB
.26
(. 71)
,
%8Exchange rate (t-1)
(depreciation only)
-.73
(-1.97)
.5495
.5967
DW
2.00
1.97
Rho
0.00
-.05
.!/ Dependent variable is quarterly percentage changes in U.S. nonfuel
import,,unit value; equations estimated over ,1973 Q2-l980 Q2, corrected- for lst-oraer autocorrelation. Exchange rate and foreign price data
are 10-country weighted averages, using multilateral trade weights.
Exchange rate 1s expressed 1n terms of foreign currency units per dollar.
r
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-86-
additional impact on U.S. non-oil import prices of the exchange rate change
when a depreciation takes place. The results suggest the presence of a weakly significant ratchet effect in the 1-quarter lagged impact of exchange
rate changes, but no significant ratchet effect with respect to contemporaneous exchange rate changes.
However, this empirical result may
reflect the fact that there were very few episodes of sustained dollar
appreciation during the sample period . . Moreover, while the coefficient_
on the lagged exchange rate changes was weakly significant, the combined
current and lagged effects were not statistically significant.
Table 15presents the results of attempts to relate ratchet
effects of exchange rate changes directly to domestic prices. This connection was not supported by the data.
First, as shown in equation II,
'
nonoil import prices have only a marginally significant impact on domestic
prices ...!/ No evidence was found of ratchet effects in the impact of nonoil import prices on domestic prices, perhaps because during 1973 Q2-1980
Q2 those prices actually fell quarter-to-quarter only twice. Second,
the last coefficient in equation IIIB indicates the absence'of significant ratchet effects in the direct impact of exchange rate changes on
domestic prices.
Theh~sults for'e'quations IIIA and' IIIB 'presentea ih Table·15 ~-"'"_"__ _
,
are problematical in that the foreign price variable has a marginally
significant coefficient with the wrong sign. This result may reflect
1/ The results reported use the domestic absorption deflator
as the dependent variable. Very similar results were obtained when the
CPI was employed. The equation numbers in Table15 correspond to the
second and third levels of analysis outlined above.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-87Table 15
Tests for Rachet Effects in the Impact of E~change
Rate Changes on U.S. Domestic Prices!/
(Estimated coefficients; t-ratios in parentheses)
Explanatory
variables
II
Equation
IIIA
1118
Constant
0.8
(3.23)
1.3
(4.98)
1.3
(4.70}
%8U.S. unit labor costb/
.44
(2.70)
.47
{4.29)
.46
(3.99)
.03
(4.22) _
.03
(2.98}
%8U.S. oil import price£!
.02
(2.57)
%8Non-oil import pric~/
.06
{1.36)
%6Foreign consumer pricesc/
-.23
(-1.82}
- .22
(-1.60}
%6Exchange ratrf:.I
-.04
{-1. 76}
-.06
(-1.05)
%6Exchange ratefl-1
(depreciation
only)
.03
{.49}
.6947
Rho
.8535
2.00
2.26
2.17
-.06
- .41
-.27
a/ Dependent variable is quarterly percent change in U.S. absorption
deflater; see footnote lJ to Table 14.
b/ Four-quarter distributed lag on U.S. wage rate divided by "normal"
productivity index.
c/ Three-quarter distributed lag.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
.8317
-88-
the effects of collinearity with either the oil import price or the unit
labor cost variables.
In an effort to correct for these possible sources
of bias in the estimates, equation IIIB in Table 15 was reestimated, first,
using domestic prices excluding energy as the dependent variable and
dropping the 011 import-price, and second, splitting the normal unit
labor cost variable into wages and normal productivity variables and
substituting' the unemployment rate for wages in a reduced-form specification. These adJustments (not reported here) yielded positive (though
not significant) coefficients on the foreign price variable, but did not
provide any further evidence of ratchet effects.
4.
Impact on Official and Private Dollar Holdings
The new operating procedure has apparently led to an increase'
in fluctuations in dollar interest rates, and this has caused greater
fluctuat1ons in international interest rate differentials and increased
the short-run volatility of dollar exchange rates.
It is difficult to
judge the likely effects of these developments on the incentives of
official and private dollar holders to diversify their portfolios.
It
is possible that increased exchange rate variability could lead to diversificat1on away from dollar assets.
However, this is by no means neces-
sarily the case.
Theoretical work, e.g., Dooley (1975), shows that the effect
of an increase in the variance of exchange rates on the optimal shares
of assets in a portfolio is ambiguous; the effect depends on the initial
conditions as well as the character of the asset holder's utility function.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-89Hence, it is not clear on theoretical grounds whether any increase in
exchange rate volatility generated by the new operating procedure would
lead to diversification away from dollar assets.
There is little empirical evidence on the impact of exchange
rate variability on private asset demands.
One study by Akhtar and
Putnam (1980) did find some evidence that exchange rate variability (the
standard deviation of daily dollar-OM spot rates) had a negative effect
on the demand for money in Germany.
-
Yet even if this diversification
effect were widespread, the impact on the demand for dollar assets is unclear, since presumably increased variability in dollar exchange rates
causes some diversification out of non-dollar assets into dollar assets,
as well as diversification out of dollar assets into other currencies.
Furthermore, fluctuations in interest rates must also be taken
into account in assessing the impact of exchange rate variability on
portfolio demands.
It is not clear that the real earnings on dollar-
denominated assets have become more uncertain than the earnings on, say,
mark-denominated assets.
The nominal earnings on dollar-denominated
assets have become more uncertain because of ~he increase in interest
rate variability, but
SO'
have the·nomrnal earnings· (expressed in dollars)·
on mark-denominated assets because of the exchange rate volatility.
However, if these nominal earnings are expressed in real terms by
deflating by the rate of change of an index of, say, U.S. and German
prices, then part of the variability in the real rate of return on markdenominated assets is offset to the extent that German goods have a
weight in the deflator.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
The variability in the real rate of return on
-90the dollar-asset earnings depends on the correlation between the movements in U.S. interest rates and the exchange rate.
Consequently, the
effect of the new procedure on the attractiveness of the dollar denominated-assets is ambiguous.
Other factors may also affect central banks' incentives to
diversify away from the dollar.
For example, some OPEC investors still
may be nervous about the precedent set by the freezing of Iranian assets,
and the way in which this situation is resolved is likely to affect
their attitude toward dollar assets in the future.
Germany, Japan, and
Switzerland are now faced with current-account deficits, and this seems
to have caused them to reconsider their position against the use of their
currencies as reserve assets.
Germany and Japan are reported to have had
some direct dealings with Saudi Arabia and other OPEC investors, and both
countries have been taking steps to make assets denominated in their
currencies more attractive or available to countries with current-account
surpluses.
Although reliable data on global reserve diversification trends
are not available past the first quarter of 1980, available evidence does
not support the concl,us~on that the System's change in _operating procedure has had much effect on reserve preferences of foreign central banks.
For the five maJor foreign reserve centers, data through
October 1980 show that their foreign exchange reserves continue to be
overwhelmingly held, in dollar-denominated assets, with little sign of a
change in proportions since the end of September 1979. Data collected by the
IMF for non-reserve-center countries indicate their reserves were about
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-91~
58 percent in dollars at the end of the third quarter of 1979, with
German marks accounting for around 14 percent, and other currencies much
smaller sha~es.
Figures for most of the non-reserve centers through the
first quarter of 1980 suggest that the dollar's share rose to about 61
percent of the total.l/ Developments since the first quarter are more
impressionistic, but aside from periodic reports about purchases of markand yen-denominated securities by some OPEC investors, there is little
evidence of increased diversification. The amount of these OPEC transactjons may total several billions of dollars equivalent, but this
probably will not have a major effect on the share of dollar assets in
global central bank portfolios.
B.
2
The Exchange Rate as Information Variable and Policy Instrument /
1. The Exchange Rate as Information Variable
Data on financial variables become available before data on the
variables that are the ultimate targets of monetary policy.
Financial
data contain information about the disturbances that are affecting the
economy and, therefore, about the likely values for ultimate target
variables.
For many years the investigation of how best to extract the
information contained in financial data focused on the search for a
single indicator of the stance of monetary policy.
More recently it
1/ This estimate may, overstate the rise in the dollar's share
because it-does not take account of valuation effects. The dollar appreciated from the end of September 1979· to the end of March 1980.
2/ The principal contributor to this section was Dale W.
Henderson.-
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-92-
has been recognized that more information can'be obtained if movements
in a number of financial variables are analyzed simultaneously. According to that more recent approach, the authorities select desired values
for their ultimate target variables; the actual values of these ultimate
target variables are unobservable in the current period. The authorities then choose some financial variables as policy instruments.
Another group of financial variables is regarded as information variables.
Values for the policy instruments consistent with desired values for
the ultimate target variables are selected and forecasts of the information variables are made.
Unanticipated movements in the information
variables are used to make inferences about the disturbances that'are
affecting the economy and, therefore, about the values of the unobservable ultimate target variables that are likely to emerge if monetary
policy remains unchanged.
On the basis of these inferences the values
of the policy instruments are changed to increase the likelihood that
/
the desired values of the ultimate targets will be attained.
Under the Federal Reserve's old operating procedure increases
in the demand for output and increases in the demand for money would
have caused little or no change in the value of the dollar in the short
run because they would have been accommodated at an unchanged nominal
interest rate.
However, shifts in desired asset holdings away from
the dollar and increases in expected inflation would have led to dollar
depreciation. Thus the second pair of disturbances could have been
distinguished from the first pair on the basis of exchange rate movements.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
Under the new operating procedure the first pair of disturbances
-93-
causes dollar appreciation while the second pair of disturbances leads
to dollar depreciation ..!/ Thus, the adoption of the ~w operating
procedure neither reduced nor enhanced the role of the lchange rate
as an information variable.
First consider an increase in.the demand for U.S. output. 2/
As a result of this disturbance output tends to rise.
Under the old
operating procedure with the interest rate held constant the money stock
would have risen but there would have been little or no change 1n the
value of the dollar in the short run. 3/ Under the new operating procedure with nonborrowed reserves held constant, the money stock increases,
and the interest rate rises.
Dollar-denominated securities become
more attractive, so the dollar would appreciate in the short run. 41
The money stock increases because the rise in output causes private
agents to raise their demand for transactions balances at the expense
of other reservable deposits even though the interest rate rises .
.!/Asimilar point has been made Frenkel and Mussa (1980).
2/ It is assumed that this and the next two disturbances considered leave the expected future value of the dollar un~~anged. For
example, the expected future spot exchange rate would be unaffected if
market participants regarded the disturbances as temporary.
3/ The dollar would have appreciated in the short run if the
rise in the transactions demand for money had come at the expense of the
demand for foreign securities to any significant extent. Over time as
current-account developments became more important the dollar would have
tended to appreciate or depreciate depending on ~hether the demand for
U.S. output increased because of a decrease in the demand for foreign
output or a drop in U.S. savings.
4/ Over time the dollar would tend to appreciate further or
depreciate-depending on the reason for the shift up in the demand for
U.S. output as explained in the preceding footnote.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-94-
Now consider an increase in the demand for money.
Previously
/~•
i wou,ld have been accommodated, so the money stock would
1
have been increased with no change in the interest rate or in the value
this disturban
of the dollar. There would have been no change in output.
new procedure the money stock and the interest rate rise.
Under the
Dollar-denom-
inated securities become more attractive, so the dollar appreciates.
However, output tends to fall.
This comparison of the implications of
an increase in money demand with those for an incr-ease in the demand for
U.S. output indicates that under each operating procedure the effect of
the two disturbances on the value of the dollar is the same.
Next consider a shift in asset preferences away from dollardenominated securities and toward foreign-currency-denominated securities.
Under the old procedure this disturbance would have led to a
depreciation of the dollar. This depreciation would have caused an increase in output and, therefore, a rise in the money stock.
Under the
new procedure the dollar still depreciates, but the interest rate tends
to be pushed up.
It seems likely that the depreciation of the dollar
would be large relative to the rise in the interest rate, at least
initially.
Both of these adJustments would work to equilibrate the
market for dollar securities, but, as long as output remains constant,
the interest rate can only rise to the extent that the depreciation of
the dollar increases the demand for nonborrowed reserves. The demand
for nonborrowed reserves 1s probably not very sensitive to changes 1n
the value of the doll~r, so the interest rate woul~ probably not rise
very much initially. Thereafter, output would tend to rise, the
interest rate would then rise, and the money stock would probably rise.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-95-
Finally, consider an increase in expected inNfttion in the
United States relative to inflation abroad.
This disturbance would,
of course, lead to an increase in the expected future price level and
a depreciation of the expected future exchange rate.
At the initial
nominal interest rate, price level~ and exchange rate there would be
an increase in aggregate demand because of the drop in the real interest
rate anq a decrease in the demand for dollar-denominated securities
since foreign-currency-denominated securities would be relatively more
attractive.l/ Under the old procedure the increases in output and,
perhaps, the price level would have raised the money stock.
The rise
in output would have further decreased the demand for dollar-denominated
securities, so the dollar would have depreciated.
procedure the nominal interest rate would rise.
Under the current
This increase would
partially, but probably not completely, offset the drop in demand for
dollar-denominated securities, so the dollar would probably depreciate.
This comparison of the implications of a shift in asset preferences
away from the dollar with those for an increase in expected inflation
indicates that under each operating procedure the effect of the two
disturbances on the value of the dollar is the same.
Under the old operating procedure all four disturbances increase the money supply and the demand for nonborrowed reserves.
How-
ever, the first pair of disturbances (the increase in the demand for
U.S. output and the increase in money demand) leave the value of the
1/ The demand for money would be decreased if the expected
rate of inflation were a separate argument in the demand for money. In
that case the final result could be a lower or higher money stock depending on whether this effect or the net impact of those mentioned in the
next sentence was more important.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-96-
dollar unchanged while the second pair of disturbances ~he shift in
asset preferences away from the dollar and the increase in expected
inflation) lead to dollar depreciation.
Under the new operating '
'
procedure all four disturbances increase the money supply and the ,.
interest rate.
However, the first pair leadsto dol~ar appr~tiation
while the second pair leads to dollar depreciation. Thus, ~he information contained in exchange rate movements makes it possible to
distinguish between the two pairs of disturbances under both operat~ng
procedures.
2.
The Exchange Rate as Policy Instrument
It has been reported above that the dollar's spot exchange
rates, forward exchange rates, and the differentials between U.S. and
foreign interest rates have been more variable in the period since the
'
adoption of the Federal Reserve's new operating procedure than they
were in previous periods.
This increase 1n variability has led some
to suggest that the author1t1es should undertake through intervention
to reduce or eliminate variation in the spot exchange rate at least
over short intervals such as a month or a quarter. That is, according
to some there is a good case for adopting the spot exchange rate as a
policy instrument.
It 1s not at all obvious that 1t is feasible or
desirable to follow this course of action especially under the new
operating procedure.
It is ·1,kely that larger variations in the foreign exchange
reserves of the United States or of other countries would have been
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-97requ1red to reduce or eliminate the larger spot exchange rate variation
exper1enced under the new procedur~.
If spot exchange rate variation
rema1ns larger than otherwise would be the case, substantial sw1ngs in
foreign exchange reserves could be required to stabilize spot rates in
the future.
In addition, doubt remains regarding the efficacy of
exchange market intervention that leaves bank reserves unchanged
so-
called sterilized intervention -- in affecting the spot exchange rate.
Even quite substantial variations in foreign exchange reserve might
not be sufficient to reduce significantly variations in spot exchange
rates.
Even if it were possible to reduce spot exchange rate variation
through sterilized interv~ntion, it might not be desirable to do so.
If
this strategy were adopted, less of the variation in d1fferentials
between the U.S. and foreign interest rates would be reflected in spot
exchange rates and more would be reflected in forward exchange rates.
For the major currencies so-called covered interest parity holds fairly
exactly.
That is, the difference between the U.S. interest rate and
'
a foreign interest rate is approximately equal to the forward discount
on the foreign currency.
Sterilized intervention probably has little
or no effect on interest rate differentials. Thus, if sterilized
intervention .is employed to stabilize spot rates in the face of
substantial variations in interest rate differentials, forward rates
will become more variable.
Whether private agents would be better off if spot exchange
rates were less variable and forward exchange rates were more variable
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-98-
is unclear.
If spot exchange rates were less variable, there would
be less incentive for private agents whose transactions involve the
payment or receipt of foreign currencies to hedge against exchange
risk.
However, some hedging would continue to occur either because
the authorities did not attempt completely to fix spot exchange rates
or because private agents,would doubt that the authorities could be
successful in keeping spot exchange rates fixed even though they
indicated their intention to do so. Those who chose to cover would
have more variable forward exchange rates.
If forward contracts for
all, including quite long, maturities were readily available at low
cost and forward exchange needs could be very accurately anticipated,
variable forward rates would constitute no problem.
On the day that
transactions were undertaken forward rates for all available maturities
would be known, and all anticipated transactions could be covered.
However, neither of these conditions 1s met.
Forward markets for
maturities beyond one year are thin or non-existent and needs for
forward exchange are no easier to forecast than other variables
relevant to business decisions.
Thus an agent making a decision involv-
ing substantial fixed costs at a given time would face the prospect
of having to choose at a later date either to hold an open position in
foreign currency or to cover that position at a forward rate that is
-
unknown at the time of the original decision.
In Section VI.A above the limited evidence that is available
on the effects of the exchange rate uncertainty on U.S. international
transactions was discussed. The conclusion was that there is little
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-99-
conclusive evidence that an increase in exchange rate variability has
important negative effects on the types of transactions that have been
studied.
Here it has been argued that for a given amount of variation
in interest rate differentials, the stabilization of spot exchange rates
implies the destabilization of forward rates. There are no studies of
the effects of this kind of redistribution of exchange rate uncertainty,
but it is by no means self-evident that it would be beneficial.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-100Appendix Table l
Weekly Exchange Rate Var1ability
(Number of 5-day series showing increases {+} or decreases(-} in
variability in the 10/79-11/80 period compared with the previous
periods. The number of series showing statistically significant
(.05 level) changes is in parentheses.}
10/79 - ,11/80
compared with
3/73 - 9/79
+
10/79 - 11/80
compared with
11/78 - 9/79
+
SPOT
Weighted-average d'ollar
German mark
Swiss franc
Japanese yen
Canadian dollar
Sterling
4(2)
5(2)
3(0)
5(5)
5(3)
5(3}
1-YEAR FORWARD
Weighted-average dollar
German mark
Swiss franc
Japanese yen
Canadian dollar
Sterling
5(1)
2(0)
1(0}
5(5)
4(3)
0(0)
0(0)
3(2)
4(2)
0(0)
5(0)
5(1)
4(0)
0(0)
5(3)
0(0)
4(0)
l (0)
0(0)
2(0)
0(0)
0(0)
0(0)
l(0)
5(2}
5(5)
5(1)
5(5)
5(0)
3(1)
0(0)
0(0)
0(0)
0(0)
0(0)
2(0)
0(0)
4(0)
5(2)
1(0)
4(0)
0(0)
l{0)
5-YEAR FORWARD
German mark
Swiss franc
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
5(1)
0(0)
l(0)
l(O)
5(3)
-l'OlAppendix-Table 2
Monthly Exchange Rate Variability
(Number of 21-day series showing increases(+) or decreases(-)
1n variability in the 10/79-11/80 period compared with the
previous periods. The number of series showing statistically signf1cant
(.05 leve13 changes is in parentheses.)
SPOT
Weighted-average do11ar
German mark
Swiss franc
Japanese yen
Canadian dollar
Sterling
10/79 - 11/80
compared with
3/73 - 9/79
+
10/79 - 11/80
compared with
11/78 - 9/79
+
20(11)
18(1)
19(2)
21(2)
15(4)
10(4)
3(0)
2(0)
0(0)
6(0)
11(0)
21(8)
21(9)
19(4)
21(17)
11 (0)
4(0)
0(0)
0(0)
2(0)
0(0)
10(0)
17(0)
'18(3)
7(0)
3(0)
21(8)
9(5)
0(0)
3(0)
14(0)
18(1)
0(0)
12(6)
21 (11)
11(0)
12(0~
6(0
21(1)
5(0)
0(0)
10(0)
9(0)
15(0)
0(0)
16(0)
21(8)
l(O)
1-YEAR FORWARD
Weighted-average dollar
German mark
Swiss franc
Japanese yen
Canadian dollar
Sterling
'5-YEAR FORWARD
German mark
Swiss franc
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
4(0)
11(0) 10(0.)
17(10)
11(0) 10(0)
TO(0)
11 (0)
(
-102Appendix ,Table 3
Variability of Foreign and Eurodollar Three-Month Interest Rates
(Number of 5-day and 21-day series showing increases (+) or
decreases (-)invariability in the 10/79-11/80 period compared
with the previous periods. The number of series showing statistically
significant (.05 level) changes is in parentheses.)
10/79 - 11 /80
compared with
3/73 - 9/79+
l 0/79 - 11 /80
compared with
11 /78 - 9/79
+
WEEKLY
Germany
Switzerland
J'apan
Canada
U. K.
0(0)
5(41
5(5)
5(5)
Eurodollar
5 (5)
l(O}
4(0)
5(1)
5(_5}"'
5 (5)
5(5)
3 (2)
0(0')
2(0) '
O{O)
5(5)
0(0)
5 (5)
0(0)
0(0)
0(0)
O'(O)
0(0)
0(0)
MONTHLY
Germany
Switzerland
Japan
Canada
U.K.
0(01 21 (6}
20(4)
l (0)
21 (21 } 0(0)
21 (21 l 0(0)
7(0) 14(0)
Eurodollar
21 (211
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
'
0(02
4 (0 ),
17 (3}
21 (10) - 0(0)
21 (21) 0(0)
21 (21 ) -() ( 0 )_
4(0}
17 (l l
21 (21} • ·0(01
I
:.
-103REFERENCES
Abrams, R1chard K., 11 Actual and Potent1al International Trade
Flows and Flex1ble Exchange Rates. 11 unpublished paper,
Federal Reserve Bank of Kansas City, August 1979.
Akhtar, M.A., and Bluford Putnam, 11 Money Demand and Fore1!Jn
Exchange Risk: The German Case, 1972-1976, 11 The
Journal of F1nance, Vol. 25, June 1980, pp. 787-794.
Blackhurst, R1chard, and Jan Tuml ir, 11 Trade Rel at10ns Under
Flex1ble Exchange Rates, 11 GATT Stud1es 1n International
Trade, number 8, Geneva, September 1980.
Clark,·Peter B., 11 Uncerta1nty, Exchange Risk, and the L:evel of
Internat10nal Trade, 11 Western Economic Journal, Vol. 11
(September 1973), pp. 302-13.
and Charles J. Haulk, "Flexible Exchange Rates and the
- - - -, Level
of Trade: A Preliminary Analysis of the Canadian
Experience," unpublished paper, Federal Reserve Board,
1972.
Cushman, David b., 11 The Effects of Exchange Rate Risk on International Trade and Direct Investment," unpubl1shed
Ph.D dissertation, Vanderb11t University, 1980.
Frenkel, Jacob A., and Michael L. Mussa, 11 Monetary and Fiscal
Policies in an Open Economy," National Bureau of Economtc Research Working Paper No. 575, National Bureau of
Economic Research, October 1980.
Dooley, M1chael, 11 Exchange Rate Expectat1ons, Portfol10 Balance
and the Reflow Hypothesis," Proceedings of the Southwestern Finance Association, Bureau of Business
Research, University of Texas at Austin, March 1975.
Goldstein, Morris, 11 Downward Price Inflex1bil ity, Ratchet Effects
and the Inflationary Impact of Import Price Changes:
Some Emp1rical Tests, 11 IMF Staff Papers, November -1977, pp. 569-612.
Hooper, Peter, and Steven Kohlhagen, 11 The Effects of Exchange
Rate Uncerta1nty on the Prices and Volume of International Trade, 11 Journal of International Economics,
Vol. 8, November 1978, pp. 483-511.
Hooper, Peter, and John Morton, 11 Fluctuations m the Do 11 ar:
A Model of Nominal and Real Exchange Rate
Determination, 11 International Finance Discuss10n Paper,
No. 168, November 1980.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
-104-
Kenen, Peter B., "Exchange-Rate Variability -- Measurement and
Implications," memorandum for Consultative Group on
International Economic and Monetary Affairs, International Finance Section, Department of Economics,
Princeton University, June 20, 1979.
Makin, John, Eurocurrencies and the Evolution of the
International Monetary System, 11 Eurocurrencies
and the International Monetary System, edited
_ by Carl Stem, et al., Washington, D.C.: American
Enterprise Institute, 1976, pp. 17-52.
https://fraser.stlouisfed.org
Federal Reserve Bank of St. Louis
11
@FedFRASER