(IX@), 7, 23-35 zyxwvutsrqponmlkjihgfedcbaZYXWVUTS
Journai 01‘International .tlonr), and Fimncr
Economic
News, Exchange
Rates and Interest
Rates
GILDS
Department
of Economics,
A.
HARDOUVELIS*
Barnard Cd&,
iYE’
10027,
zyxwvutsrqponmlkjihgfedcbaZYXWVU
Cohnha
Universit_v, NW
l’ork,
USA
and
Rrsmrch
Department,
Ftdrrai
&wrw
Bank of ~Vrw York,
1~1’
100- /j,
US.4
The paper esamines the post-October
1979 response of exchange rates
and interest
rates to the new information
contained
in the first
announcement
of fifteen US macroeconomic
series. illarkets respond
primarily to monetary
news, but also to news about the trade deficit,
domestic inflation, and variables that reflect the state of the business
cycle. For all fifteen macroeconomic
variables, an increase (decrease) in
interest rates is accompanied
by an appreciation
(depreciation)
of the
dollar, which is consistent
with models that stress price rigidity and
absence of purchasing
power parity.
The
effect
the
economic
expectations
information
literature,
of economic
news
literature
on asset
because
hypothesis
about future
prices
an
has received
outcome
of
the
increasing
efficient
attention
is that flesible
asset prices
change
the moment
fundamentals
arrives in the market.
In the exchange
the role of news
has gained
importance
because
in
markets/rational
structural
models
new
rate
have
failed to predict most of the variation
in exchange
rates during the 1970s (see iMeese
and Rogoff,
1983). Researchers
were naturally
led to investigate
the types of news
that make exchange
rates so volatile,
and to utilize news in order to test which
model
of exchange
experience.
rate determination
Notable
studies
are
is consistent
those
of
with
the floating
Dornbusch
(1980),
Edwards
(1982a,b,
1983), Copeland
(1984), and Rose (1984).
In this paper, I also examine
the effects of macroeconomic
rates,
but I adopt
a different,
more
direct
strategy,
eschange
Frenkel
news
one that complements
rate
(1981),
on eschange
previous
work, avoids
some of its shortcomings,
and promises
interesting
conclusions.
I
examine
the instantaneous
response
of exchange
rates the moment
a piece of
economic
news hits the market.
If one can isolate the exact time during a business
day when news arrives,
it is straightforward
to examine
the response
of exchange
rates because
financial
data are continuously
available.
The advantages
of the
present
methodology
are many: first, it is possible
to identify
the exact types of
economic
news and how they affect exchange
rates; second,
the simultaneity
bias,
which plagues
the previous
literature
here. The independent
variables
that
on exchange
rates and news, is not present
represent
news can be interpreted
as causal
* The paper was presented at the 1985 meeting ofthe American Economic Association.
The author
wishes to thank Sebastian
Edwards,
R. W. Hafer, Nicholas
Karamouzis,
Andrew
Rose, and two
anonymous
referees for helpful comments;
Kim Rupert of the Money Market Services for providing
the survey data; Soo Kounne for able research assistance; and Barnard College for research support.
0261-5606jSS/Ol;OO23-l3SO3.00
0 1988 Butterworth
& Co (Publishers)
Ltd
21
Economic
Sews,
Exchange
Rafts
and Intcrtvt
Rafrs
variables because they are predetermined.
There is a precise analogy to input and
output variables of a controlled
experiment.
Third, it is possible to examine the
simultaneous
reaction of prices in other asset markets and, from the direction of the
various
reactions,
gain a better understanding
of how markets
interpret
the
information
they receive, and which exchange rate model they perceive as the
correct one. Thus, in addition to eschange rates, I study the reaction of short- and
long-term
interest rates.
I examine the exchange rate and interest rate responses to news contained
in the
first announcements
of a wide variety of US macroeconomic
series: four monetary
series (Ml, bank reserves, and the Fed discount and surcharge rates), two inflation
series (the consumer and producer price indices), the trade deficit, and eight other
monthly macroeconomic
series which provide information
about the state of the
business cycle and are closely watched by economic forecasters (the unemployment
rate, the industrial production
index, personal income, orders of durable goods, the
index zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
of leading indicators,
retail sales, consumer credit, and housing starts). Some
of the above macroeconomic
series have been partly analyzed by previous authors.’
But most of the empirical
evidence
presented
here is new. In addition,
the
estimation
of the market
responses
is done simultaneously
for all fifteen
independent
variables. This provides a unified framework
and avoids possible bias
when two or more announcements
occur within the same time interval.
The rest of the paper is organized as follows: Section I contains a discussion on
the model underlying
the estimation
results, and an esposition
of the different
hypotheses.
Section II presents the empirical evidence. Section III summarizes
the
main conclusions.
An appendis
contains a detailed description
of the data.
I. Market
The equation
which
I estimate
Responses
to
Economic
has the following
News
form:
/=,
where DP, represents
the percentage
change in a foreign currency
price or the
change in an interest rate in business day t; $1. is the unanticipated
component
of
economic series x, announced
at t; and U, is a random error term uncorrelated
with
information
prior to the announcement
at t. Each vector s:’ contains zeros for the
business days the series x, is not announced.
I estimate equation (1) separately for
each bilateral currency and each interest rate using ordinary least squares. Although
the daily changes of the different exchange rates and interest rates are correlated,
the OLS estimates are as efficient as the estimates from a seemingly
unrelated
regressions procedure because the set of independent
variables is identical in ever!
equation. The OLS estimates are also unbiased. There is no apriori reason why the
escluded information
contained
in the error term uI, which is composed
of other
types of daily news that hit the market, may be correlated with the predetermined
right-hand-side
variables.
&Iy discussion
will focus on the interpretation
of the estimated parameters
a,.
Equation
(1) is a reduced form equation and, therefore, the sign and size of the
estimated a,s depend on many underlying
factors, such as the policy rules of the
monetary
and fiscal authorities
and the autocorrelation
properties
of the
GILAS A. HARDOLVELIS
25
macroeconomic
series. Note also that the surprise about a variable, x”, retlects, by
definition,
a surprise about both its demand and its supply side; and typically, it is
the relative persistence
of shocks to demand
versus shocks to supply which
determines
the algebraic sign of the market reactions.*
story which market
My aim will be to identify
the ‘model’ or economic
participants
have in mind when they respond
to the announcement
of a
macroeconomic
variable. This may not be always possible, given the plethora of
plausible scenarios.
However,
it is possible to discriminate
between two sets of
interesting
hypotheses.
The first set of hypotheses is consistent
with models that
assume price stickiness and absence of purchasing
power parity and explain the
reactions of exchange rates and interest rates as a result of a change in the expected
future risk-free real rate of interest. An appreciation
(depreciation)
of the dollar
concurs with an increase (decrease) in nominal interest rates. The second set of
hypotheses is consistent
with models that assume price flexibility and purchasing
power parity and explains the reactions of exchange rates and interest rates as a
result of a change in the espected
future rate of inflation.
An appreciation
(depreciation)
of the dollar concurs with a decrease (increase) in nominal interest
rates.
II. Empirical
Evidence
I will first make some general remarks on the empirical evidence of Tables 1,2, and
3, and then analyze the market responses to each announced
series separately. The
independent
variables in the regressions
are the unanticipated
components
of the
announced
series and were constructed
using survey forecasts. The unemployment
rate (UN), consumer credit (CC), housing starts (HS), the trade deficit (TD), and
the discount and surcharge rates (DISC, SUR) represent changes in levels; all other
independent
variables represent percentage
changes. Table 1 presents the results
for the federal funds rate, the three-month
Treasury bill rate and the twenty-year
Treasury bond rate. Table 2 presents the results for seven major currencies.
The
results for an additional
ten currencies
are reported in Hardouvelis
(1985a).
My sample period runs from October 11, 1979 through August 16, 1984 and
Table 3 presents tests of structural
change by partitioning
the sample into two
subperiods with the break point occurring on October 15, 1982.3 During the period
October 1979 to October 1982 the Fed followed strict Ml targets and allowed wide
fluctuations
in interest rates. After October 1982, it began paying more attention to
and
abandoned
non-borrowed
reserves
as its
interest-rate
smoothening
intermediate
target. We, therefore,
have reason to suspect structural
instability,
especially
with regard
to monetary
announcements,
and more
precisely,
announcements
about bank reserves which are more closely related to the Fed’s
intermediate
targets. The hypothesis of parameter stability is consistently
rejected
in the case of bank reserves, RES. A joint test for all variables rejects the hypothesis
only in the case of the three-month
Treasury bill rate. Because there is apparent
instability in the Treasury bill responses, in Table 1 I present the interest rate results
for both the whole sample period and each subperiod
separately. Notice also that
interest rates (but not exchange rates) show considerable
heteroskedasticity
across
the two subperiods.
Thus the interest rate results for the entire sample come from a
weighted
least squares regression
with weights equal to the standard
error of
estimate of each subperiod.
26
Economic Sews,
TABLE
Exchange Rater and Interest Rates
Interest
1.
ocrober 19-9tO.iugusr1981
F-funds
0.004
n.no4
(o.lno4)
(0.003)
October
October
19-O to October
1982
T-bond
7.bill
-n.m
-I).OW
O.lOl*
0.2-j.
n.22i*
0.069
(0.015)
(O.lOl)
(0.039)
(0.105)
DISC
-0.011
--0.114*
-n.lw
(‘l.‘r)
(O.OI I)
(0.030)
n.061
0.221
0.3’1
(K212)
(0.0:3)
(0.042)
(0.256)
(ln.099)
0.41 I *
(0.152)
0.129‘
(0.050)
II.1723
(10.029)
II 178.
0.390’
(0 154)
--0.X0
n.lv
$1.059)
0.113
0.244*
0 233
0.139
(10.339)
(0.1 IS)
CPI
n.152
ll,IlW
((1. lx-)
(0.055)
-11.2-2.
PPI
-n.lsJ
10.119 I )
I,.166
0.1128
0.01x,
(ONl5)
-cl.n53
_
0 093
0.6-6
0.205*
(11.21 I)
(0.521)
(0.124)
0.1 18’
(0.050)
0.255
0.159
(0.209)
- I).,-II
-ll.S86*
(Ci.601)
0.01’
i).Ztrl
(n.093)
(0.390)
(0. lb-)
-n.zzl
(DN5)
0.041
0.2IlO
(lW68)
O.DoJ
--0.259
l
zyxwvutsrqpo
(O.oGI)
(0.431)
ll.l59*
[O.W2)
--O 1 iG*
(0.306)
IP
(Cl.545)
(n.n-5)
T-bond
0.1 19.
0.181’
(0.015)
T-b,11
(O.004)
(0.023)
-0.031’
I984
O.(W)4
0.197’
;O.OJ-)
to .\ugust
(Ml6)
(0.012)
(0.030)
I982
F-fundc
(0.0’2)
-0.li20
L’s
news.
F-funds
O.coI
hII
SLR
T-bond
(0.014)
C
RES
7.bill
rate responses to economic
(0 233)
l).lljO
-0.183’
(n.355)
!0.085)
11.151
-0.1~21
(Il.llbR)
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
(I, 1132)
(‘1.1.36)
(0.1133)
fl~.lK!3)
0). I-6)
(‘1. I108)
I).%3
PI
--11.1109
--1).,,,2*
(ll.1i68)
~\l.IW)
(0.21 I)
II 016
DG
(lUlO6)
--I, ,,,,o
0.,,,,-
LI
(0.039)
- 11.118I
RS
cc
(0 Ill j)
1).lMl2
I).ll3l
11.,,26*
l
(“.“l1L)
1,,11,,.;
(0.0.39)
(‘1 01 I)
(1U1119)
0.037
0 IN!6
(“.I IO)
(“V loo)
--0.110Kl
--rl.014
(O.lM2)
(lU~l2)
I .OJ’
11.332.x
2.23
4 II*
22.24’
first
(c)
change
personal
-.68*
I.31
I .iu
5.42’
I .96’
I .6-l
2.(,n*
2.1”
l.li4
1.26
2.-3’
2.19’
7.61’
5,Sl*
2.32’
T.3O*
errors
are inslde
adjusted
the parentheses.
for degrees
IO MI,
Asterisk
of freedom.
RES.
the
second
subperiod,
variables
(PI). orders
raw (LX).
installmax
ofdurable
housing
crcdjt
and
(RES):
starts
m bilhons
rnfc (DISC.
SUR).
(d)
rxres are one business
statistical
is rhe
regrrssion
and
SCR:
the enure
the
goods
xn millmns
oidollars
day changes
sample
of units
ret&s
weekly
co maturq.
is the
the constant
and trade
pcrccnrxge
in the
of leading
lndlcarors
(HS).
percent
D-W
10 CPI and PPI:
except
change
and the rctul
at the 95
error.
level.
R2 is rhe coefficient
Durban-\X’ltson
H3 IC F(8. ,I) and rcftrs
term,
of
sfatist~c.
C: n= Y38.446,
to US,
I I79 for the
respecrivcly.
ofrhe
percenrnge
index
in yields
standard
period
componeots
(CC);
s~gn~ticancc
to all the vauabler
monthly
(DC),
n.86
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQP
3.28.
H2 IS F(2. n) and
n) and refers
are rhc unuxiciparcd
rescues
dcnores
SEE
DISC.
1.1, RS, CC, and HS: H4 1s F(l5,
income
Intewsr
I).44
0 j-3
in non-borrowed
tn consumer
20.72.
9.94’
independent
unemployment
4.99’
2.115
subperlod,
The
13.r*
I1.j.l
(I,) HI 1s I’(_). n) and refers
IP. PI. DG.
(IUil,)
--o.IXlJ
I .1x1
(a) Standard
n.Illl?
ll.lU3
ii I.Cl48)
O.lll5
decermlnation
(n.011~)
(1)W)
(&032)
(l~.O83)
n.ln,l
-0.-41
--o,li92’
-0.11)l
(I’.lN’“)
O.l)l)j
(I~.Oll~)
(Il.OJI)
(11.4-l)
(1.22)
(KOl2)
(0.05 I)
1)02l
(O.cr5 I)
-0.333
1).O~)l
(n.uln)
1).~~28’
--o.M?
(0.0.X)
--o.o,,i
(IN 133)
0.l113’
(0.111ffi)
0.015
il.ci58
(0.i I”-)
-I).lN4
(M25)
I,.O3ll
(I W2)
(I).O.+O)
(n.059)
(11.153)
-,,.I
3’)
(I’ 3t”‘)
11.1119
-(l.\N-
--11.14-
(1l.1113)
11.11111
(13.385)
0.11ns
ID
ll.lij6
(O.II3Jj
(ItlMS)
i).OlJ
-0.6<9-
HS
O.lilK,
(CUGi)
(l’.Oll9)
(Il.lrl I)
0 2x4
22ll
(0 118)
l~.ll??.
(I.014
(‘I.020)
-0
Il.364
~(I.31 14)
(LI).
deticlr
prospective
,A coeftic~enr
CPI.
change
PPI,
in XII.
indusrrtal
and retail sales (RS):
in billions
changes
ofdollars
the weekly
index
the monthly
level m rhr
(TD):
in the Fed discount
of. say, 0. I99 denotes
pcrcenrzg:c
production
rhc monthly
(IP).
change
race and surcharge
an lncreasc
oftwenty
basts
po ints.
The explanatory
power of the fifteen variables is low, varying from an 8’ of
0.006 for the Canadian dollar to an Rz of 0.076 for the three-month
Treasury bill.
This is not surprising.
The dependent
variables are not levels but changes in the
levels. Within a given business day a variety of news hits the market and our fifteen
series represent a very small subset of such news. Furthermore,
there are many days
during which none of the variables that we examine is announced.
Notice that the
Durbin-Watson
statistic is close to 2.0 for most markets indicating
no apparent
zyxwvutsrqp
GIKAS
TABLE 2. Exchange
A.
HARDOUVELIS
rate responses
October
SWiSS
franc
G e rm a n
mark
27
to economic
news.
1979 to August 1984 zyxwvutsrqponmlkjihgfedcbaZYXW
Britrsh
pound
French
franc
Canadian
dollar
Italian
lira
-0.039’
-0.033
-0.005
-0.0381
-0.064*
-0.062’
-0.009
(0.022)
(0.020)
(0.017)
(0.018)
(0.020)
(0.019)
(0.008)
-0.22”
-0.2.M’
-0.224*
-0.197*
-0.203’
-0.178’
-0.370
Ml
(0.083)
(0.0’6)
(0.0’7)
(0.325)
(0.067)
(0.0-l)
(0.0’3)
0.174’
0.179*
0.129*
0.173’
0.250*
O.Oj4*
0.132”
RES
(0.061)
(O.Oj6)
(0.054)
(0.050)
(0.024)
(0.053)
(0.057)
-0.435*
-0.340
-0.119
-0.3’9
-0.028
-0.131
-0.404*
DISC
(0.223)
(0.203)
(0.180)
(0.087)
(0.196)
(0.191)
(0.205)
- 0.2-9’
-0.245
-0.006
-0.317*
-0.114*
-0.080
-0.254*
SUR
(O.Oj4)
(0.112)
(0.119)
(0.138)
(0.126)
(0.122)
(0.128)
- 0.005
-0.013)
0.210
0.314
-0.766
-0.167
0.050
CPI
(0.398)
(0.435)
(0.170)
(0.353)
(0.374)
(0.383)
(0.402)
-0.105
-0.os1
0.041
-0.045
0.021
WI zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
0.136
0.038
(0.295)
(0.269)
(0.259)
(0.239)
(0.272)
(0.1 lj)
(0.253)
0.-9-I*
0.932’
-0.028
0.528
0.375
0.137
0.652
UN
(0.44j)
(0.407)
(0.392)
(0.361)
(0.41 1)
(0.1;4)
(0.382)
-0.060
-0.108
0.113
-0.108
-0.124
-0.123
-0.031
IP
(0.140)
(0.128)
(0.123)
(0.114)
(0.055)
(0.120)
(0.130)
0.306
0.174
0.028
0.080
0.4j2*
PI
0.186
0.415
(0.251)
(0.215)
(0.229)
(0.220)
(0.232)
(0.203)
(0.098)
_0,rJ77*
-0.085’
-0.063’
-0.063’
- 0.034
-0.062”
-O.Oll
DG
(0.028)
(0.026)
(0.025)
pot 1)
(0.023)
(0.021)
(0.026)
0.057
0.047
0.01’
0.002
-0.008
LI
0.043
0.034
(0.075)
(0.068)
(0.066)
(0.064)
(0.029)
(0.060)
(0.069)
-0.193”
-0.139’
-O.ljZ*
-0.089
-0.185*
-0.172*
-0.047
RS
(0.067)
(0.063)
(0.060)
(0.056)
(0.063)
(0.027)
(O.Oj9)
0.037
- 0.064
0.053
o.oi5
0.055
-0.015
cc
0.021
(0.072)
(0.066)
(0.067)
(0.059)
(0.028)
(0.062)
(0.063)
-0.j28
-0.811
0.158
-0.794
-0.042
-0.683
HS
-0.305
(0.640)
(0.632)
(0.562)
(0.270)
(0.609)
(O.jOj)
(0.692)
0.062
0.112
0.083
TD
0.087
0.0-6
0.076
0.017
(O.Oj?)
(0.050)
(0.061)
(0.056)
(0.054)
(0.053)
(0.024)
Constant
R2
SEE
D-W
0.039
0.650
1.99
0.023
0.675
1.97
0.032
0.739
1.99
0.016
0.635
1.95
0.041
0.683
2.04
0.006
0.288
2.29
0.039
0.600
2.09
j.Tj*
9.59*
10.42’
6.50*
4.03’
3.8’*
9.56’
1.OO
0.02
1.86
0.32
0.30
0.02
0.02
2.87*
1.77
2.86’
2.97,
2.89’
1.88
0.76
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
2.92’
3.67’
4.31’
4.45*
2.31*
1.j2
4.31*
Hl : F(4, n)
H2: F(2, n)
H3: F(8, II)
H4: F(15, n)
No/es: See notes (a), (b), and (c) of Table 1. The exchange rates represent one business day percentage
changes in the prices of foreign currencies in US dollars. A coefficient of, say, -0.241 denotes
appreciation of the dollar by 0.24 per cent.
Also the F-statistics
at the bottom of Tables 1 and 2 show that the
four monetary
variables
carry most of the explanatory
power and are jointly
significant
in all interest rate and all currency markets (hypothesis
Hl); the two
inflation variables are jointly significant
only in the twenty-year
Treasury
bond
market (hypothesis H2); and the eight ‘cyclical’ variables are jointly significant
in
four of the currency
markets and in the three-month
Treasury
bill market
(hypothesis
H3).
misspecification.
Economic Sews, Exchange Rates afui Interest Rates zyxwvutsrqponmlkjihgfedcbaZ
28 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
TABLE 3.
October
TD
t-stat
Federal
funds
Treasury
bill
Treasury
bond
X:. German
mark
Japanese
ven
S&s
franc
British
pound
French
franc
Canadian
dollar
Italian
lira
(a) Marginal
(b) The test
the value of
appear in the
PI, DG, LI,
three interest
least squares
1.52
- .2.62*
1.17
0.20
0.18
Tests
of structural
1979 to October
hll
t-stat
RES
t-stat
-1.40
1.50
-1.33
3.36*
0.76
1.49
1982 vs. October
Monetary
F(3, n)
1.60
-2.21*
-0.38
0.82
- 2.52*
0.52
-0.51
-2.39*
0.58
- 1.48
-2.80*
0.43
-1.90
-0.99
0.22
-0.02
-1.43
0.14
- 0.98
-1.18
change.
(0.05)
2.42
(0.06)
3.9-(0.01)
1.60
1982 to ;\ugust
1984
Inflation
F(2, n)
Cyclical
F(8, n)
1.21
(0.30)
0.61
(0.54)
1.42
(0.19)
1.03
(0.36)
0.20
(0.82)
0.60
(0.55)
1.81
(0.16)
1.05
(0.35)
3.21*
(0.04)
0.96
(0.38)
0.38
(0.93)
1.83
(0.07)
1.51
(0.13)
0.88
(0.53)
1.41
(0.19)
1.10
(0.36)
0.64
(i:zj)
-_
1 .oo
(0.45)
2.72*
(0.01)
1.44
(0.13)
1.04
(0.41)
1.38
(0.16)
1.25
(0.23)
1.54
(0.09)
0.82
(;:;;)
(y.;;)
(0.63)
1.34
(0.22)
(0.38)
All
V4, n)
(‘:$)
significance
levels are in the parentheses.
n= 1185.
statistics were generated
by including individual
coefficient dummy variables which took
zero in the first subsample.
‘Monetary
refers IO Ml, RES, and DISC (SCR does not
second part of the sample). ‘Inflation’
refers to CPI and PPI. ‘Cyclical’ refers to UN, IP,
RS, CC, and HS. ‘All’ refers to Monetary,
Intlanon,
Cyclical, and TD. The tests for the
rates correct for the heteroscedasticity
across the subsamples;
we performed
weighted
with weights equal to the SEE of each subperiod.
II.
A. Monetary
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
A nnoortrurments
Let us begin by examining
the market reactions to the announcement
of XII, a topic
that has been analyzed extensively
in the past. Table 1 shows a positive reaction of
all interest rates. A 1 per cent unanticipated
weekly increase in hll (which has an
average sample size of approximately
five billion dollars) increases the federal funds
rate by twenty basis points, the three-month
T-bill rate by eighteen basis points,
and the twenty-year
T-bond rate by ten basis points .A The exchange rate reactions
confirm
the evidence
of previous
authors
that the dollar appreciates
after an
unanticipated
increase in Ml and show that the result is robust and independent
of
the foreign currency examined. For example, after an unanticipated
increase in Ml
by 1 per cent, the dollar appreciates against the West German mark by 0.21 per cent.
Previous
authors have emphasized
that the eschange
rate results show that
markets expect the risk-free real rate of interest to change, and that the Fed has
credibility in the market. A positive surprise about Ml signals a persistent increase
in money demand, but (perhaps) only a temporary
increase (because the Fed is
espected
to stick to its pre-announced
hll targets)
in money
supply.
The
GIKAS A. HARDOCVELIS
29
expectation of a future escess demand for money leads to an upward revision in the
expected market clearing (in the face of price rigidity)
real interest rate. This
explanation
cannot justify, however, the strong long-term
interest rate response
because a liquidity effect is short-lived.
And there is no consensus in the literature
about what exactly causes long-term
interest rates to respond so strongly.
In
Hardouvelis
(1985b), I attempt to rationalize
this response as a simultaneous
expected inflation effect. I develop a theoretical model in which both the liquidity
and inflation effect are present simultaneously,
but the liquidity effect dominates
the reactions of short-term
interest rates and exchange rates, while the inflation
effect dominates
the reactions of long-term
interest rates. The two effects can
coexist when market participants
attach some positive probability
weight on the
possibility
that the Ml targets will be abandoned.
Thus the strong positive
response of long-term
interest rates could be interpreted
as evidence that the Fed
lacked fi& credibility
in its fight against inflation.5
An unanticipated
increase in free reserves by 1 per cent of non-borrowed
reserves (this equals approximately
400 million dollars in the sample) decreases the
three-month
T-bill rate by eleven basis points during the first subperiod but has an
insignificant
effect on the twenty-year
T-bond
rate in both subperiods.
In
Hardouvelis
(1987), I present a model of the reserves (and money) market which
explains this reaction as an expected liquidity effect. 6 The exchange rate reactions to
free reserves announcements
represent entirely new evidence. The dollar shows a
significant
depreciation
against all currencies.
The depreciation
of the dollar
confirms that the accompanying
decrease in interest rates represents a decrease in
expected real interest rates. Both the interest rate and the eschange rate responses
show the importance
market participants
attached to the Fed’s inter-week
behavior
in the reserves market during the October 1979 to October 1982 period. However,
unlike the money announcement
responses, we cannot interpret the responses to
reserves announcements
as evidence of Fed credibility,
because there is no one-toone relationship
between the Fed’s inter-week
reserves target and the Ml target.
Announcements
of the Fed discount and surcharge rates have a positive effect on
interest rates and a negative effect on exchange rates, implying that markets expect
a change in the real rate of interest. A prospective
increase in the discount
or
surcharge rate is interpreted
by markets as a signal of a future tightening
by the Fed,
which increases the expected future real interest rate due to an expected liquidity
effect.’ Notice that the reactions to announced
surcharge rate changes are slightly
stronger for the case of the federal funds rate and the exchange rates, and slightly
weaker for the other interest rates than the corresponding
reactions to announced
discount rate changes. Evidently,
markets interpreted
changes in the surcharge rate
as only temporary attempts by the Fed to either tighten or loosen credit, which is
quite reasonable
given the nature of surcharge
rates.8 zyxwvutsrqponmlkjihgfedcbaZYXWV
U.B.
Inflation Announcements
The short-term
interest rate reactions to the CPI and PPI announcements
were
previously
analyzed by Roley and Troll, and Urich and \Vachtel.s They find a
refined
positive but insignificant
response. Urich and Wachtel, after examining
time intervals, are able to find a statistically significant delayed response to the WI
announcement
during
the afternoon
of the day of the announcement
(the
announcement
takes place in the morning).
These authors have attributed
the
30
Economic Stws, zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHG
Exchange Rater and Intcrcst Rater
response of short-term
interest rates to revisions of inflationary
expectations.
But
the response could also be due to changes in the risk-free real rate of interest as the
Fed may be expected to counteract
past increases in the rate of inflation.
The
twenty-year
T-bond rate and the exchange rates of the present paper can provide
evidence that allows us to discriminate
between these two competing
hypotheses.
The espected
inflation
hypothesis
predicts a depreciation
of the dollar and an
increase in long-term
interest rates that is larger the more permanent
the increase
in inflation
is expected
to be. The expected
liquidity
hypothesis
predicts an
appreciation
of the dollar and an increase in long-term
interest rates that is much
smaller than the observed increase in short-term
interest rates because the liquidity
effect disappears in the long-run.
Contrary to the short-term
interest rate reactions, the reactions of the twentyyear T-bond rate are statistically significant.
An unanticipated
monthly increase in
the PPI by 1 per cent increases the twenty-year
T-bond rate by sixteen basis points
(with a t-statistic of 3.8). An unanticipated
monthly increase in the CPI by 1 per
cent increases the twenty-year
T-bond
rate by eighteen
basis points (with a tstatistic of 2.4). The exchange rate reactions are insignificantly
different from zero,
and the algebraic signs show a delayed appreciation
(see Hardouvelis,
1985a). We
conclude
that neither hvpothesis
is able to explain all market reactions.
The
expected inflation hypothesis
cannot explain the absence of a dollar depreciation,
and the expected liquidity hypothesis cannot explain the strength of the reaction of
long-term
interest rates relative to short-term
interest rates.10
1l.C.
Cyclical
_;lnnorincmtMs
Let us now consider variables that provide
information
about the state of the
domestic
macro-economy.
Tables 1 and 2 show that significant
responses
are
present only after the announcement
of the unemployment
rate (UN), personal
income (PI), manufacturers’
orders of durable goods (DG), and retail sales (RS).
Evidently,
the announcements
of the industrial production
index (II?), the index of
leading indicators
(LI), consumer installment
credit (CC), and housing starts (HS)
do not provide significant
new information
about future economic developments
or provide mixed signals.
The market responses to cyclical news (as well as monetary news) are consistent
with the textbook aggregate demand (IS-LM) -aggregate
supply model in an open
framework,
in which
prices
are sluggish,
asset markets
adjust
economy
instantaneously
(the interest parity condition
holds), purchasing
power parity does
not hold in the short and intermediate
run, and the foreign interest rate is allowed
to change. Obstfeld (1985) has recently surveyed this model (which is an extension
of the hlundell-Fleming
model with perfect capital mobility) and used it to explain
the post-1971 floating eschange rate experience. According
to the model, a positive
shift in the IS (LM) curve causes the real rate of interest to increase (decrease) and
the dollar to appreciate (depreciate).
A positive shift in aggregate supply causes the
real rate of interest to decrease and the dollar to depreciate.
The market responses show that unanticipated
changes in the unemployment
rate, manufacturers’
orders of durable goods, and retail sales were interpreted
as
evidence of a persistent change in aggregate demand originating
in the real sector (a
shift in the IS curve) rather than aggregate supply. This is reasonable because the
presence of inventories
drives a wedge between production
and demand, and items
GIKAS
:
A.
HARDOCVELIS
31
such as retail sales or orders of durable goods are more directly related to demand
movements.
An unanticipated
increase in ,the level of the unemployment
rate by
one percentage point signals a future decrease in aggregate demand and causes the
T-bill and T-bond
rates to decrease by twenty-seven
and sixteen basis points
respectively,
and the dollar to depreciate. Unanticipated
increases in retail sales or
durable goods signal a future increase in aggregate demand and cause an increase in
interest
rates and an appreciation
of the dollar. Notice that the interest
rate
responses to retail sales and durable goods are significant
in the second subperiod
unanticipated
changes
only.
Finally,
in personal
income
were apparently
interpreted
as evidence of a persistent
change in aggregate
supply rather than
This is also reasonable,
since income
to the factors of
aggregate
demand.
production
is more directly
related to production
rather than demand., An
unanticipated
increase in personal income causes a significant decrease in long-term
interest rates and a depreciation
of the dollar.
The announcement
of the trade deficit has a statistically
significant
effect on the
three-month
T-bill rate during the first subperiod,
but an insignificant
effect on the
other interest rates and the eschange rates. From October 1979 to October 1982, an
unanticipated
increase in the monthly trade deficit by 1 billion dollars decreases the
three-month
T-bill rate by nine basis points. In general, an unanticipated
increase
in the trade deficit decreases interest rates and depreciates the dollar. When markets
learn that the trade deficit of last month was larger than they had anticipated,
they
apparently
expect a further future increase in the trade deficit. This is consistent
with the positive autocorrelation
of the monthly
trade deficit series shown in
Hardouvelis
(1985a). Markets probably
attribute
their surprise about the trade
deficit mostly to unanticipated
exogenous positively autocorrelated
changes in the
foreign demand for domestic products. A future contraction
in the foreign demand
for domestic products is expected to lead to a large deficit, temporarily
lower real
interest rates, and a depreciation
of the dollar. This scenario is consistent with the
model that we described earlier.
III. Conclusions
The post-October
1979 exchange rate and interest rate reactions to news about
different macroeconomic
variables are consistent
with an extended version of the
in which
open interest
parity
holds
traditional
Mundell-Fleming
model,
instantaneously,
purchasing
power
parity
does not hold in the short or
intermediate
run, the foreign interest rate is allowed to change, and expectations
play a crucial role, as in Obstfeld (1985). A basic characteristic
of this model is that
shifts of the textbook IS curve, the LM curve, or the aggregate supply curve affect
the real rate of interest and, subsequently,
the change in the real rate of interest
affects the exchange rate through
the interest rate parity condition.
Our main
empirical
regularity
is consistent
with this characteristic.
We discovered
that
an appreciation
(depreciation)
of the dollar is accompanied
by an increase
(decrease) in nominal
interest rates. This is true for all fifteen macroeconomic
series, although the responses to some of the series are not statistically significant.
Evidently,
during our sample period, exchange rate movements
were primarily
32
Economic
A’ews, Exchange
Rates and Interest Rates
driven by expectations
of future changes in real interest rates rather than the
expected rate of inflation. This empirical regularity is inconsistent
with the simple
monetarist
model of exchange rate determination
(see, for example, Frenkel, 1976),
which assumes price flexibility and purchasing
power parity.”
Monetary news carries most of the explanatory
power in our regressions.
The
four monetary variables are jointly significant
in all interest rate and exchange rate
markets
that we examine.
The response
to bank reserves announcements
is
particularly
strong in the period until October
1982, when the Federal Reserve
followed
non-borrowed
reserves
as intermediate
targets.
The responses
to
monetary
news are caused by expected
developments
in the money market,
developments
which are expected to affect the real rate of interest. Some nonmonetary variables also show significant
effects. These variables reflect news about
the unemployment
rate, manufacturers’
orders of durable goods, retail sales and
personal income. The first three signal future changes in aggregate
demand,
while
personal
income
the trade deficit
signals future changes
in aggregate
decreases
short-term
interest
rates
value of the dollar is insignificant.
Finally, inflation
on short-term
interest
rates, a strong positive
effect
no effect on exchange
rates.12
supply.
but the
Adverse
adverse
news about
effect on the
news has a weak positive
effect
on long-term
interest
rates, and
Appendix
The Data Srt
XII interest
rates are annualized
yields to maturity, expressed in percentages.
They were
T-bill and T-bond
rates are collected
at
provided
by Data Resources
Incorporated.
3:30 p.m. Eastern
Time, but the federal funds rate is a daily average.
A regression
coefficient of, say, 0.5 implies a change in interest rates of half a percentage point or 50 basis
points. Exchange rates, defined as the prices of foreign currencies in terms of US dollars, are
noon bid rates from the New York market as reported by the Board of Governors
of the
Federal Reserve System. In the regressions,
the dependent
variables are in the following
form: DP,, = lOO[log(s,,) -log(s,,_,)],
where J,, denotes the spot price of foreign currency;
in terms of US dollars during business day (. The factor 100 transforms
the units of the
dependent
variables to percentages.
Thus the magnitudes
of the regression coefficients are
comparable
to those in the interest rate regressions.
;i coefficient
of, say, 1.2 implies a
change of 1.2 per cent or 120 basis points.
The survey data were provided
by Money Market Services Incorporated
of Belmont,
California.
They represent
median forecasts of about forty market specialists.
Douglas
efficiency,
Pearce and V. Vance Roley (1985) among others have analyzed the unbiasedness,
and forecast performance
of Ml, the CPI, the WI, the unemployment
rate, and the
industrial
production
index. I have performed
a similar analysis (1985a) using the latest
sample period and the additional
eight survey variables (there are no survey data for
changes in the discount
and surcharge
rates). The overall conclusion
is that the survey
forecasts, although not always unbiased or efficient, have a smaller root mean square error
than forecasts which are based on autoregressive
models. In the same paper, I also present
the results on possible delayed responses to announcements,
as well as the responses to the
anticipated
components
of the series.
The data on Ml begin in October 1979 and end in February 1984, at the end of lagged
reserve accounting.
The data on reserves begin in late April 1980, when the first survey data
on free reserves become available, and also end in February 1984. The data on discount and
surcharge rate changes, unemployment,
industrial production
index, CPI, and PPI begin in
GIKAS A. HARDOL.VELIS
33
October 1979 and end in zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFED
Aug ust
1984. The remaining series begin in February 1980, when
the survey data become available, and end in August 1984.
The unanticipated
component
of Ml is defined
as lOO( -1 +[Ml(t--2)-FDM]/
Ml(t-3)),
where Ml (t -2) is the announced during fiscal week t level of Ml of fiscal week zyxwvutsrqpo
t-2,
Ml(t-3)
was announced
at t- 1, and FDM is the survey forecast of the weekly
change in Ml from t-3
to t -2. RES is defined as lOO(FR(t -1) -FRR)/NBR(t-2),
where FR(t - 1) is the announced
at zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJ
t level of free reserves of fiscal week t - 1, FFR is the
corresponding
survey forecast, and NBR(t-2)
is the level of non-borrowed
reserves of
fiscal week t -2. DISC and SUR are the announced prospective
changes in the discount and
surcharge
rates. WI, CPI, IP, PI, RS, DG, and LI refer to the unanticipated
monthly
percentage change in the producer price index, consumer price index, industrial production
index, personal income, retail sales, manufacturers’
orders of durable goods, and the index
of leading indicators.
UN is the unanticipated
monthly level in the unemployment
rate. CC
is the unanticipated
monthly change in consumer installment
credit (S billions). HS is the
unanticipated
monthly level in housing starts (millions of units). TD is the unanticipated
monthly level in the trade deficit (S billions). All independent
variables are seasonally
adjusted except RES. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDC
Notes
1. Cornell (1982), Engel and Frankel (1984). Frankel and Hardouvelis
(1985), Hardouvelis
(1984).
Husted and Kitchen (1985), and Gavin and Karamouzis
(1984) examined
hll announcements.
Hardouvelis
(1987) examined
the reaction
of the term structure
of interest
rates to the
the reaction of the three-month
announcement
of free reserves. Roley and Troll (1983) examined
.
T-bill rate to the announcement
of the Fed discount and surcharge
rate, the unemployment
rate,
the industrial
production
index (the latter two are also analyzed by Wachtel and Aleleer, 1984),
and the CPI and PPI announcements.
Urich and Wachtel (1984) analyzed the reaction of shortterm interest rates up to approximately
one year to the CPI and PPI announcements.
Batten and
Thornton
(1984) examined the exchange rate reactions to the announcement
ofthe discount rate.
2. See, for example, the models of Nichols et A. (1983), or Hardouvelis
(1985b).
3. Some of the series do not cover the whole sample. See the appendix
for details.
4. Note that the federal funds rate does not respond to the hI1 announcement
after October
1982.
Before October
1982, non-borrowed
reserves were the instrument
of monetary
control.
An
announced
unanticipated
change in hi1 provided
information
about the current week’s demand
for reserves (recall that until February
1984, required reserves of the week of the announcement
were a function of deposits two weeks earlier, and that the announced
level of AI1 referred to two
weeks earlier). This affected the market expectation
of the demand for borrowed
reserves and the
market clearing federal funds rate. After October
1982 the Fed adopted borrowed
reserves as its
instrument
for monetary
control, and thus accommodated
directly a change in the demand for
reserves which made the federal funds rate less likely to respond in order to clear the reserves
market.
5. Another
hypothesis
that can explain the data in combination
with the expected
liquidity
hypothesis,
which also presumes
lack of full credibility,
is the inf(attion-risk hypothesis.
This
hypothesis
was first advanced by Cornell (1983) and predicrs that long-term
interest rates respond
positively because inflation risk increases (decreases) after an unanticipated
increase (decrease) in
Ml. (See Taylor,
1981, for empirical
evidence on the positive relationship
between inflation
variability
and the average rate of inflation.) It also predicts a slightly positive response for shortrerm interest rates, and no response for exchange
rates, because the risk premium
is already
incorporated
in domestic real interest rates and, therefore,
no incipient capital movement
could
occur either into or out of the USA that would affect exchange
rates.
6. The basic argument
can be summarized
as follows:
Market participants
have a fairly accurate
estimate of total reserves of week t - 1 before the announcement
at t, because total reserves are a
at t - 1 (our data on reserves
function of demand deposits of week t - 3, which were announced
and Ml end on February
1984, at.the end of the lagged reserve accounting
period). Thus the
announcement
at t of the monetary base of week t - 1 with its components
provides information
about the composition of total reserves.
An overestimate
of non-borrowed
reserves
due to
34
.!% no m ic
L\ c wr,
Exchange
Rates and Interert Ratrj
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
7.
8.
9.
10.
11.
12.
unanticipated
expansion
of reserves by the New York desk implies an equal underestimate
of
borrowed
reserves which, at the observed
level of the federal funds rate and the discount
rate, is
due to an unanticipated
increase in reserve pressure at the discount window. Thus, assuming that
the surprise about excess reserves is zero (empirically
the assumption
does not make a difference),
RES represents
a proxy for an unanticipated
increase (decrease) in non-borrowed
(borrowed)
reserves. After an announced
positive RES, markets will expect real rates to go down in the future
if they perceive that actions taken by the New York desk are more persistent than actions taken at
the discount
window.
This is reasonable,
particularly
in the period between October
1979 and
October
1982 when the Fed followed
non-borrowed
reserves targets.
Batten and Thornton
(1984) argue that it is difficult to distinguish
between an expected liquidity
and an expected inflation effect using the exchange
rate reactions,
but eventually
interpret
the
appreciation
of the dollar after an (unanticipated)
increase in the discount
rate as evidence of an
expected inflation effect. They consider the expected future Fed tightening
as permanent,
i.e., as a
permanent
future decrease in the growth
rate of the money supply that leads to a decrease in
inflationary
expectations.
(See the Batten-Thornton
Table 3, p. 285, which leads the authors to
favor the expected inflation hypothesis.)
Batten and Thornton
treat the exchange rate reactions in
isolation,
without
examining
the simultaneous
interest rate reactions.
Their interpretation
is
rejected by the podhe and significant
response
of, particularly
long-term,
interest rates.
I treat the actual announced
changes in the discount
and surcharge
rates as unanticipated,
but
some of these changes may be widely anticipated
due to, say, rumors or official pronouncements.
Aly coefficient
estimates may, therefore,
be biased slightly towards zero. Batten and Thornton
(1984) present a detailed study in which they separate discount rate changes into technical (mostly
anticipated)
and discretionary
(mostly unanticipated).
They claim that during our sample period
discount
rate changes were mostly discretionary
rather than technical
and, therefore,
largely
unanticipated.
Smirlock and Yawitz (1985) also separate discount
rate changes into anticipated
and unanticipated
components.
Contemporaneously
with the initial draft of this article,
Hakkio
and Pearce
(1985) have
independently
examined the exchange
rate reactions
to the announcement
of \lI, the 01,
the
PPI, the industrial
production
index, and the unemployment
rate, and have examined
more
refined time intervals.
Their exchange
rate results are similar to mine.
The inflation-risk
hypothesis
of note 5 is, however,
consistent
with a/l the market reactions.
It
predicts a strong positive reaction for long-term
interest rates, a weak positive reaction for shonterm interest rates, and no reaction for exchange rates, which are exactly the reactions we observe.
h similar conclusion
was recently reached by Cornell and Shapiro (1985) who esamined
daily
correlations
between exchange rates and interest rates. However,
as the authors recognize,
their
approach
suffers from simultaneity
bias because they regress one endogenous
variable,
the
exchange rate, on another endogenous
variable, the interest rate. (See Engel, 1986, for a detailed
criticism.)
In contrast,
the present paper does not suffer from a similar simultaneity
problem
because I explicitly treat exchange
rates and interest rates as endogenous
variables.
As I mentioned
in notes 5 and 10, the strong positive response of long-term
interest rates to the
surprise components
of hll, the CPI, and the WI could be the result of changes in inflation risk.
This would imply that during our sample period, the Fed was not able to establishfullcredibility
among market participants
about its fight against inflation. zyxwvutsrqponmlkjihgfedcbaZYXWV
References
B~rres,
D.S.,
AND D.L.
THORNTON, ‘Discount
Rate Changes and the Foreign
Exchange
Market,’
1984, 3: 279-302.
Dollar Exchange
Rate and the News,’ Economics Letters,
Jorrrnal of International M oney and Finance, December
COPELAND, L.S., ‘The Pound Sterling/US
1984, 15: 109-113.
CORNELL, B., ‘Money Supply Announcements,
Interest
Rates, and Foreign
Exchange,’
Journal of
International M oney and Finance, August
1982, 1: 201-208.
CORSELL, B., ‘The Money Supply Announcements
Puzzle: Review and Interpretation,’
American
Economic Review, September
1983, 73: 644-657.
CORNELL, B., AND A.C. SHAPIRO, ‘Interest Rates and Exchange
Rates: Some New Empirical
Results,’
Journal of Infermtionai M oney and Finance, December
1985, 4: 431-442.
DOR.UBUSCH,,,R., ‘Exchange
Rate Economics:
Where do we Stand. 3,’ Brooking Papers on Economic
Activity , Spring
1980, 1: 14>185.
GIKAS
A. &RDO~VELIS
35 zyxwvutsrqp
S., ‘Exchange
Rates and News: A Multi-Currency
Approach,’
journalof Infemnational~lo~y
amf Finance, December
1982, 1: 21 l-224 (1982a).
EDW.I RDS.
S., ‘Exchange
Rate Market Efficiency and New Information,’
Economic1 Lstterr, 1982, 9:
377-382 (1982b).
EDWARDS, S., ‘Floating
Exchange
Rates, Expectations
and New Information,’
Journal of Monetary
Economics, May 1983, 11: 321-336.
ESGEL, C.M., ‘On the Correlation
of Exchange
Rates and Interest Rates,‘Journul of International Money
and Fim~~e, March 1986, 5: 125-128.
ENGEL, C.M., AND J.A. FRANKEL, ‘Why Money Announcements
Move Interest Rates: An Answer
from the Exchange
Rate Market,
Journa/ of Moneta~ Economics, January
1984, 13: 31-39.
FRANKEL, J.A., and G.A. HARDO~~ELIS, ‘Commodity
Prices, Money Surprises, and Fed Credibility,’
Journai of Money, Credit and Bunking, November
1985 (part I), 17: 425-438.
FRENKEL, J.A., ‘A Monetary
Approach
to the Exchange
Rate: Doctrinal
Aspects and Empirical
Evidence,’
Scandinavian Jawno/ of Economics, LMay 1976, 78: 200-224.
FRESKEL, J.A., ‘Flexible Exchange
Rates, Prices, and the Role of News: Lessons from the 197Os,’
Journal of Political Economy, August 1981, 89: 665-705.
G~vrs, W.T., AN D N.V. KARA~IOUZIS, ‘Monetary Policy and Real Interest Rates: New Evidence from
the Money Stock Announcements,’
working
paper,
Federal
Reserve
Bank of Cleveland,
December
1984.
HAI;X:IO, C.S., AND D.K. PEARCE, ‘The Reaction of Exchange
Rates to Economic
News,’ working
paper, Federal Reserve Bank of Kansas City, January
1985.
HARDOUVELIS, G.A., ‘Market Perceptions
of Federal Reserve Policy and the Weekly hlonetary
Announcements,’
Journal of Monetav Economics, September
1984, 14: 225-240.
HARDOUVELIS, G.A., ‘Economic
News, Exchange Rates and Interest Rates,’ working paper, Barnard
College, Columbia
University,
February 1985, revised as First Boston Working Paper Series No.
86-12 (1985a).
HARDOUVELIS, G.A.,
‘Exchange
Rates, Interest
Rates, and hIoney
Stock Announcements:
A
Theoretical
Exposition,’
Journai of International hlone_y and Finance, December
1985: 4: 443-454
(1985b).
HARDO~VELIS, G.A., ‘Reserves Announcements
and Interest Rates: Does hlonetary
Policy 1Iatter?,’
Journa/ of Finance, June 1987, 42: 407-422
HUSTED, S., AN D J. KITCHEN, ‘Some Evidence on the International
Transmission
of Money Supply
Announcement
Effects,‘Journa/
of Money, Credit and Banking, November
1985 (part 1). 17: 456466.
MEESE, R.A., AN D K. ROGOFF, ‘Empirical Exchange Rate Models of the Seventies: Do They Fit Our
Sample?,’ Journal of International Economics, February
1983, 14: 3-24.
N I CH OLS,
D.A., D.H. SM ALL, AN D C.E. WEBSTER, ‘Why Interest Rates Rise When an Unexpected
Large Money Stock is Announced,’
American Economic Review, June 1983, 73: 383-388.
OBSTFELD, M., ‘Floating
Exchange
Rates: Experience
and Prospects,’
Brookingr Paperr on Economic
.-lctivig, Fall 1985, 2: X9%464.
PEARCE, D.K., AN D V.V. ROLEY, ‘Stock Prices and Economic
News,‘Journalof
Burinerr, January 1985,
58: 49-67.
ROLEY, V.V., .~ND R. TROLL, ‘The Impact of New Economic
Information
on the Volatility of ShortTerm Interest Rates,’ Economic Review, Federal Reserve Bank of Kansas City, February
1983,315.
ROSE, A.K., ‘Testing for News in Foreign Exchange
Markets,’ Economirs Letters, 1984, 14: 369-376.
SM I RLOCK ,
M .,
ASD J. YAWITZ, ‘Asset Returns,
Discount
Rate Changes,
and Market Efficiency,’
Journal of Finanrc, September
1985, 40: 1141-l 158.
TAYLOR, J.B., ‘On the Relationship
between the Variability
of Inflation
and the Average Inflation
Rate,’ The Cortz and Consequenw of Infhtion, Amsterdam:
North
Holland,
1981, CarnegieRochester
Conference
Series on Public Policy, 15: 57-85.
URICH, T., AND P. WACHTEL, ‘The Effects of Inflation
and Money Supply Announcements
on
Interest Rates,’ Journal of Finance, September
1984, 39: 1177-1188.
WACHTEL, P., AN D I M . MELZER, ‘Interest
Rates and Announcements
of Information
on Real
Economic
Activity,’
working
paper, NYU Business School, July 1984.
EDW.~RDS,